A survey of antiprion compounds reveals the prevalence of non-PrP molecular targets

Prion diseases are fatal neurodegenerative diseases caused by the accumulation of the misfolded isoform (PrP(Sc)) of the prion protein (PrP(C)). Cell-based screens have identified several compounds that induce a reduction in PrP(Sc) levels in infected cultured cells. However, the molecular targets of most antiprion compounds remain unknown. We undertook a large-scale, unbiased, cell-based screen for antiprion compounds and then investigated whether a representative subset of the active molecules had measurable affinity for PrP, increased the susceptibility of PrP(Sc) to proteolysis, or altered the cellular localization or expression level of PrP(C). None of the antiprion compounds showed in vitro affinity for PrP or had the ability to disaggregate PrP(Sc) in infected brain homogenates. These observations suggest that most antiprion compounds identified in cell-based screens deploy their activity via non-PrP targets in the cell. Our findings indicate that in comparison to PrP conformers themselves, proteins that play auxiliary roles in prion propagation may be more effective targets for future drug discovery efforts.


Misfolding
and aggregation of endogenously expressed proteins cause several neurodegenerative disorders (1)(2)(3). Prion diseases belong to this class of proteinopathies and result from the misfolding of the -helixrich, cellular prion protein (PrP C ) into a -sheetrich, disease-associated, infectious isoform termed PrP Sc (3)(4)(5)(6) . Unlike other neurodegenerative disorders, prion diseases are readily transmissible to laboratory animals and cultured cells. The availability of laboratory models harboring the infectious aggregate has enabled the development of an empirical drug discovery strategy against prion diseases. Typically, prion-infected, neuronally derived cell lines that accumulate and stably propagate PrP Sc (7,8) are used as a primary screen for the identification of compounds that reduce prion levels in culture. Subsequently, the in-vivo efficacy of putative antiprion compounds is assessed by analyzing their ability to prolong disease incubation periods in prion-infected rodents. Using this approach, numerous antiprion compounds have been identified, including pentosan polysulfate, dextran sulfate, HPA-23, Congo red, suramin, dendritic polyamines, 2-aminothiazoles and quinacrine (9). However, none of these compounds have been shown to be effective against a range of human prion strains in animal models when administered at a late, post-symptomatic stage, and none have been shown to have significant disease-modifying properties in human clinical studies.
While measuring PrP Sc levels in infected cultured cells can be used to assess antiprion activity, this method does not elucidate the molecular targets of active compounds. As a result, the mechanisms of action of most antiprion compounds remain unknown. In principle, a compound can reduce the prion load in a cell by interacting with a number of molecular targets. The most direct mechanism is through direct binding to PrP C and stabilization of its native conformation (10,11). Alternatively, a drug may directly interact with PrP Sc , leading to its disaggregation (12), or may target auxiliary factors or proteins that play a role in PrP C expression, localization or conversion to PrP Sc (13).
To investigate whether antiprion compounds identified in prion-infected neuronal cell lines have a tendency to interact with PrP C , PrP Sc , or other targets, we screened a library of 2,160 known drugs and natural products and identified 206 compounds that cleared PrP Sc in neuroblastoma (N2a) cell lines at a concentration of less than 1 M. Of these initial hits, we validated the activity of 16 compounds and assessed their ability to bind to recombinant PrP, directly disaggregate PrP Sc , reduce the expression level of PrP C , and alter the localization of PrP C . Taken together, the results suggest that the antiprion activity of these compounds is mainly mediated by non-PrP targets.

EXPERIMENTAL PROCEDURES
Chemical library (MicroSource Discovery Systems, Inc.) The chemical library of 2,160 compounds screened in both cell-based and direct-binding assays was obtained from the MicroSource Discovery System (MSDI, Gaylordsville, CT, USA; http://www.msdiscovery.com/), and includes known drugs, bioactives, and natural products. Compounds were solubilized at 10 mM in DMSO and stored in a 96-well format by the Small Molecule Discovery Center at the University of California San Francisco (http://smdc.ucsf.edu/).

Cell-based antiprion activity and toxicity
A mouse neuroblastoma (N2a) cell line was infected with the Rocky Mountain Laboratory (RML) strain of scrapie prions to produce ScN2a cells (14). Screening the chemical library for antiprion activity was performed in a high-throughput ELISA assay. Briefly, 4 10 4 ScN2a cells were treated with compound of interest for 5 days at 1 M final concentration. Untreated ScN2a cells were used as negative controls; ScN2a cells treated with quinacrine (1 M) were used as positive controls (15,16). A toxicity screen was conducted in parallel at the same compound concentration and time of exposure in a 96-well format using an acetomethoxy derivate of calcein (calcein-AM) assay. Untreated ScN2a cells were used as negative controls. Both of these methods have been described previously (17,18).

Antiprion activity by immunoblotting
ScN2a cells (5 10 5 ) were propagated in a 10cm plate and treated for 5 days with the compound of interest at 50, 20, 10, or 1 M, depending on cellular toxicity. Negative controls were performed by treating cells with the highest percentage of DMSO applied to the cells. As a positive control, cells were treated with 1 M quinacrine. Cells were lysed with lysis buffer (0.5 % NP-40; 0.5 % deoxycholate; 10 mM Tris-HCl, pH 8; 100 mM NaCl) and protein concentration was normalized to 1 mg/ml using the BCA assay. Samples were incubated with 20 l/ml of proteinase K for 1 h at 37 °C. Digestions were stopped with 2 mM phenylmethylsulfonyl fluoride (PMSF), and samples were centrifuged at 100,000g for 1 h at 4 °C. Supernatants were discarded and pellets were resuspended in reducing SDS sample buffer for SDS-PAGE. Western blotting was performed according to standard procedures. PrP was detected by using D13 antibody Fab fragment conjugated (19) with a horseradish peroxidase (HRP) probe (Rockland Immunochemicals Inc.).
Isothermal Titration Calorimetry ITC measurements were performed at 37 °C using the ITC 200 Microcalorimeter (GE Healthcare). To counteract the backlash effect observed during the first injection (22), ITC titrations were carried out with one 0.2 l injection of ligand (compound), followed by 10 consecutive injections of 3.8 l of the compound with injection durations of 4-s and a 150-s interval between injections. The amount of energy released was measured following each injection. Prior to each measurement, DMSO was added to the protein sample to reach a final concentration of 1%. The sample chamber was filled with 15 M of recombinant MoPrP(89-230), which had been dialyzed previously in the screening buffer (see protein preparation). Solution stocks of compounds were prepared at 20 mM in 100% DMSO. Solutions of titrants were freshly prepared by diluting them with the dialyzed buffer to reach a 200 M final concentration (1% final DMSO). Titrant solutions were centrifuged for 5 min at 14,000 rpm, and supernatants were loaded into the syringe. The final molar ratio of ligand:protein exceeded 2.5. Isotherm data were analyzed with Origin7 software (MicroCal) supplied by the manufacturer.
Circular Dichroism -CD spectra of both truncated and full-length recombinant MoPrP were recorded at 10 M, with a 0.1-cm cuvette using a Jasco J-715 spectrometer with a Jasco PTC-348WI Peltier-effect temperature control device, where the temperature reported reflects that of the heating block. Scans were acquired at 50 nm/min, with a bandwidth of 2 nm and data spacing of 0.1 nm. Thermal unfolding curves were measured at 10 μM of recombinant MoPrP(89-230) and MoPrP . All compound solutions were made fresh from 100 mM DMSO stock solutions. Prior to each run, samples were equilibrated with the compound of interest for 15 min at 20 °C, then a temperature ramp rate of 2 °C/min was applied from 20 to 90 °C. To attenuate DMSO noise background, a wavelength of 230 nm was used to monitor the -helical signal, rather than the more typical 221-223 nm. The melting point was determined from the thermal unfolding curve fit by the two-state-folding EXAM program, which has been widely used for this purpose (23,24). A u C p value of 4.76 kJ mol 1 K 1 was set in the fitting of the van't Hoff equation to determine the melting temperature (T m ) and the H of the transition. A significant thermal upshift was defined as T m exceeding the standard deviation of the technique ( =0.4 °C) by a factor of three. The standard deviation was calculated based on three T m measurements of native PrP.
Thermal shift monitoring by Differential Scanning Fluorimetry (DSF) Protein stability was assessed in a 96-well format using an MxPro3005P qRT-PCR Detection System (Stratagene, Agilent Technologies, Cedar Creek, TX). Sypro-Orange dye (Invitrogen) was used to monitor the fluorescence by applying ROX filter for the fluorescence emission (610 nm) and FAM filter for the fluorescence excitation (492 nm). Optimal conditions to perform the assay were determined by varying protein concentrations, Sypro-Orange dilutions, buffers and thermal ramp parameters.
To experimentally screen compounds, recombinant MoPrP(89-230) and Sypro-Orange dye were plated manually. Stock solutions of compounds at 10 mM (100% DMSO) were freshly diluted in the screening buffer (1:10 ratio), then added to the protein by using a 96/384 pipettors robot (Apricot Designs) to achieve a final compound concentration of 1 mM (1% final DMSO).
DSF spectra of 10 M recombinant MoPrP(89-230) with a 1:2000 Sypro-Orange dilution in the screening buffer were recorded. Samples (150 L final) were heated at 2 °C/min, from 40 °C to 90 °C, and the fluorescence values were recorded after every 1-°C increase. Approximations of the melting temperature (T m ) have been assessed by using the maximum value of the first derivative generated by the QPCR software (MxPro QPCR software, Stratagene, Agilent Technologies, Cedar Creek, TX). To evaluate changes in T m , for each run, first derivatives of the melting curves were generated from the MxPro software, exported as a text file and imported in Mathematica software. The first derivative of the curves was fitted with a polynomial function. T m was approximated with the maximum value of the first derivative.
PrP C expression -To observe the effects of the compounds at their maximum of efficacy, uninfected N2a cells were incubated with the selected molecules for 5 days.
Compounds were tested at the same concentration used during the screening on ScN2a cells. Negative control cells were incubated with DMSO (0.5% final concentration). After treatment and cell lysis, normalized crude extracts were analyzed by Western immunoblotting using conjugated D13-HRP antibody to detect PrP. Normalization of the total quantity of protein present in the crude extract was checked by actin staining (Fig. 6A).
To assess PrP C expression in the presence of an increased dose of amcinonide, N2a cells were treated for 3 days and quantified by Western immunoblotting as described above.

Western-blot quantification
Westernblots were quantified using ImageJ software (http://rsb.info.nih.gov/ij/index.html). PrP Sc and PrP C levels in compound-treated cells were normalized against the levels in untreated control cells.

In vitro susceptibility of PrP Sc to protease
Brain homogenates (10% wt/vol) were prepared from ill CD-1 mice infected with RML prions, then diluted 10-fold using an acidic sodium acetate buffer supplemented by detergents (5 mM sodium acetate, pH 3.5; 1% NP-40). Ninety l of 1% brain homogenates (0.6 mg/ml) were incubated with 10 l of compounds or PAMAM generation 4.0 and 4.5 to reach a final concentration of 100 M or 100 g/ml, respectively. Samples were incubated for 2 h with constant shaking at 37 °C. After neutralization with 100 l of a freshly prepared buffer containing 0.2 M HEPES pH 7.5; 0.3 M NaCl and 4% Sarkosyl, samples were subjected to PK digestion (20 l/ml, 1 h of incubation at 37 °C). Proteolytic digestions were stopped by the addition of PMSF (2 mM final concentration), and samples were analyzed by Western immunoblotting using conjugated D13-HRP antibody to detect PrP.

Lipid raft isolation
For lipid raft isolation, N2a cells were treated with the compounds for 3 days in 6-well plates. Plates were placed on ice, cells were rinsed twice with chilled PBS and incubated for 20 min with icecold Triton X-100 lysis buffer made with a Mesbuffered saline (25 mM Mes; 150 mM NaCl; pH 6.5) containing 1% (v/v) Triton X-100. The lysates were then homogenized by passing through a Luer 22-gauge needle and centrifuged at 500g for 5 min at 4 °C to pellet cell debris. Cold supernatants were harvested, normalized to 0.5 mg/ml, and mixed with an equal volume of 80% (v/v) sucrose to obtain a 40% (v/v) sucrose solution. A 1-ml aliquot of the sample was then transferred to the bottom of a SW-60 centrifuge tube. Two ml of 30% sucrose was then added to the top, followed by the addition of 1 ml of 5% sucrose to create a discontinuous sucrose gradient. Tubes were centrifuged at 4 °C for 18 h at 140,000g in a SW-60 rotor (Beckman Instruments, Fullerton, CA). Fractions (1-8; 500 μL) were collected from the top and analyzed by Western immunoblotting using D13-HRP antibody (see immunoblotting section). Flotillin-1 (Sigma) and transferrin receptor (Santa Cruz Biotechnology) were probed by immunoblot and used as markers for detergent-resistant membranes and detergent-soluble fractions, respectively.

Immunocytochemistry
N2a cells (2.5 10 5 ) were plated on a coverslip (Fisher Scientific, Circles No. 1.5) placed in a 24-well plate format, and treated for 3 days with the selected compounds. Cells were washed with warm (37 °C) PBS media and fixed with 4% paraformaldehyde solution for 20 min at room temperature. Cells were rinsed 3 times for 5 min with room-temperature PBS, permeabilized with 0.3% Triton X-100 in PBS for 5 min, and then rinsed 3 times with PBS buffer. Cells were blocked with 10% normal goat serum (NGS) in PBS for 30 min at room temperature, then incubated with primary D18 antibody (5 μg/ml) in 10% NGS overnight at 4 °C. Cells were washed successively 3 times with PBS for 10 min, and incubated in the dark at room temperature for 2 h with a FITC-conjugated goat anti-human IgG (H&L) polyclonal antibody (5 μg/ml diluted in 10% NGS, Jackson Immunoresearch). Samples were rinsed 3 times with PBS for 10 min in the dark. Coverslips were rinsed briefly with water, then mounted on Superfrost Plus microscope slides (Electron Microscopy Sciences, Hatfield, PA) with Vectashield with the counterstain DAPI (Vector Laboratories), and sealed with Cytoseal. Slides were analyzed at the QB3-UCSF Nikon Imaging Center (http://nic.ucsf.edu/) on a Nikon Eclipse Ti-E Motorized Inverted Microscope. Images represent individual Z-slices taken from the middle of the cell.

Total RNA Purification and Prnp
Quantitative PCR Uninfected N2a cells were incubated with the amcinonide at 50 μM or 0.5% DMSO (untreated control cells) for 6, 24, 48, or 72 h. Total RNA was isolated with TRIzol reagent. cDNA was synthesized from 1 g of total RNA using the SuperScript II First-Strand Synthesis System for RT-PCR at 42 °C for 60 min in the presence of random primers (Invitrogen), and then diluted 10-fold in water. Two microliters were used in duplicate for quantitative PCR amplification of Prnp, and actin as an internal control, using the MxPro3005P qRT-PCR apparatus (Stratagene, Agilent Technologies, Cedar Creek, TX). The following program was used: denaturation step at 95 °C for 10 min, 40 cycles of PCR (denaturation at 95 °C for 10 s, annealing at 5 5°C for 8 s, elongation at 72 °C for 15 s). Primers were as follows: Actin: forward, gatcattgctcctcctgagc 5'; reverse, ctcatcgtactcctgcttgc 3'; Prnp: forward, cgagaccgatgtgaagatga 5'; reverse, atcccacgatcaggaagatg 3'. The curves of amplification were read with MxPro 3005p software using the comparative cycle threshold method. Relative quantifications of the target mRNAs were calculated after normalization of cycle thresholds with respect to actin levels. Values are expressed as -fold change compared to untreated cells (0.5% DMSO).

PrP C degradation kinetics
N2a cells were preincubated with the amcinonide at 20 μM or 0.5% DMSO (untreated control cells) for 3 days. Subsequently, 30 g/mL cycloheximide (Sigma) was added to the culture to inhibit protein synthesis, and cells were incubated at 37 °C for various durations. Cells were then lysed and residual PrP C levels were evaluated by Western immunoblotting.

Identification of compounds with antiprion activity.
We screened a chemical library of 2,160 compounds containing known drugs and natural products (Microsource) for antiprion activity in ScN2a cells. The cell-based assay was carried out in a multi-well format and PrP Sc levels were measured using an ELISAbased assay (18).
Cells treated with DMSO (carrier) and quinacrine (a known antiprion drug (15)) were used as negative and positive controls, respectively. These measurements were used as reference to calculate the normalized percentage change in PrP Sc level following treatment with experimental compounds. ScN2a cells were treated with compounds for 5 days at a concentration of 1 M. In parallel, we analyzed the cytotoxicity of all compounds at this concentration by employing the Calcein AM assay for membrane integrity (25). The level of cytotoxicity was normalized with respect to untreated cells.
Based on the distribution of activity and cytotoxicity measurements ( Fig. 1A and B), we defined a "hit" as a compound that reduced PrP Sc load by 40% with less than 20% cytotoxicity as measured by the Calcein AM assay (Fig. 1C). Using these criteria, we identified 206 nontoxic compounds with antiprion activity. A previous screen of the MicroSource compound library also identified active compounds (26), most of which were included in our hit set as well (Fig.  1C, red dots).
We randomly selected 40 hits for secondary validation. Initially, compounds were added to ScN2a cells at a concentration of 50 M. For compounds that proved toxic at this concentration, the experiments were repeated at 20, 10 and 1 M to assess the antiprion activity of the compound at the highest possible nontoxic concentration. Changes in PK-resistant PrP Sc levels were analyzed by Western blots. We were able to confirm the antiprion activity of 16 of 40 compounds (Fig. 1D). Some of these 16 compounds belonged to chemical classes previously known to have antiprion properties (statins, flavones, resveratrol, chalcone, quercetin, phenothiazine and corticosteroid (26,27)). Several were, as far as we are aware, novel: these include dehydrovariabilin, 3-deoxy-3 -hydroangolensic acid methyl ester and glycosides (Table 1). Experiments conducted on prion-infected N2a cells in the presence of the 16 compounds revealed a proportional decrease of all three glycoforms (unglycosylated, monoglycosylated and diglycosylated) of digested PrP Sc (Fig. 1E), suggesting that the compounds neither alter PrP glycosylation nor are dependent on PrP glycosylation for their antiprion activity.
Identification of compounds that interact with PrP: Isothermal Titration Calorimetry Next, we assessed the ability of these 16 antiprion compounds to interact directly with recombinant MoPrP(89-230) containing the structured domain of PrP found in the proteinase-resistant core of PrP Sc (28). Isothermal titration calorimetry (ITC) was used to measure direct binding between the compounds and recombinant PrP. We first sought to establish the ITC protocol with a control compound that is known to interact with PrP. We used suramin as a positive control as it has been shown to nonspecifically interact with PrP, inducing its aggregation (29,30). We used ITC to analyze the interaction between suramin and PrP at concentrations of 200 M and 15 M, respectively. In the presence of suramin, we observed a substantial release of energy for the first injections (approximately -6.0 Kcal/Mol of injectant), demonstrating an interaction between the partners ( Fig. 2A). Using the same experimental parameters, we performed ITC with the 16 confirmed antiprion compounds in order to detect any significant release of energy upon injection. No substantial release of energy was detected for any of the 16 compounds (Fig.  2B), suggesting that none interacted with recombinant MoPrP(89-230).
Identification of compounds that interact with PrP: Thermal-denaturation upshift assay We next sought to measure the binding of compounds to PrP by a thermal-denaturation upshift assay. Ligand binding and protein folding are thermodynamically linked, and the binding of a ligand stabilizes proteins against denaturation (31). We therefore used thermal denaturation to quantify potential binding of antiprion compounds by detecting induced upshifts in the melting temperature of PrP. Having validated the assay with known stabilizing and destabilizing agents, we proceeded to analyze the potential binding interactions of the 16 cell-active antiprion compounds with the structured domain of PrP. The compounds (100 M) were added to 10 M of recombinant MoPrP(89-230). Denaturation of truncated PrP in presence of DMSO alone was used as a negative control. In accordance with the ITC results, none of the 16 compounds significantly increased the T m of MoPrP(89-230), suggesting that they neither stabilized nor interacted with the folded domain of the protein (Fig. 3A and B).
To discount the possibility of compounds binding to the N-terminal unstructured region of PrP, we repeated the thermal-denaturation upshift assays using recombinant full-length MoPrP . The unfolding of MoPrP(23-230) was reversible (Supplemental Fig. S1C) and changes in protein stability in the presence of TMAO or GdnHCl could be detected (Supplemental Fig. S1D). None of the 16 cell-active antiprion compounds significantly increased the T m of MoPrP(23-230) (Fig. 3C and D).

Identification of compounds that interact with
PrP: Differential scanning fluorimetry.
The observation that none of the 16 antiprion compounds directly interacted with recombinant PrP can be interpreted in two ways. First, the analyzed chemical library may have been devoid of compounds that interact with PrP. Indeed, NMR and crystallographic studies indicate that the three-dimensional structure of PrP lacks deep cavities normally required for high-affinity binding interactions with small molecules, and as such, few molecules within an unbiased diversity chemical library are expected to bind PrP (32)(33)(34)(35)(36). Alternatively, general binding to PrP may be a poor correlate to antiprion activity and not all molecules that bind to PrP may exhibit antiprion activity. For example, a small molecule may need to bind and locally stabilize specific regions of PrP in order to inhibit its conversion to PrP Sc . To distinguish between these two possibilities, we screened the 2,160 compounds in the chemical library for their ability to interact with PrP using a highthroughput, in-vitro binding assay.
We employed an approach based on the differential scanning fluorimetry assay (DSF) (37). DSF monitors protein unfolding via a fluorescent dye that interacts preferentially with hydrophobic surfaces exposed during protein unfolding. DSF has been used to detect ligand binding (38), including PrP (39). To measure induced changes in the stability of PrP in a highthroughput fashion, we miniaturized the DSF assay and conducted a screen in a 96-well format (see Experimental Procedures, Fig. 4A).
The 2,160 compounds in the Microsource library were screened at a concentration of 100 M for direct binding to recombinant MoPrP(89-230). Melting curves of PrP in the presence and absence of compounds were recorded at 1°C increments from 40 °C to 90 °C. Typically, the melting curves could be divided into three regimes: the native baseline, the unfolding transition, and the denatured baseline (Fig. 4B). The T m value indicates the thermodynamic stability of the protein (31). Additionally, the slope and magnitude of the baselines provide data about the aggregation of the native and denatured states of the protein (37). In the absence of compounds, PrP has a stable native baseline, a transition at 65.9 ± 0.4 °C and a sloped denatured baseline (Fig.  4B). The slope in the denatured baseline likely reflects the aggregation of unfolded PrP at high temperatures. As expected, the addition of 200 mM GdnHCl decreased PrP stability (T m = 62.8 °C), whereas addition of 200 mM TMAO increased the protein stability (T m = 68.9 °C) (Fig. 4B).
As controls, each screened plate contained three wells with 200 mM GdnHCl and three wells with 1% DMSO alone (Fig. 4C). To evaluate the experimental noise in the screen, positive and negative controls were used to calculate the Z-factor (40) for each plate (Fig.  4D). Z-factors for all plates were above 0.5, indicating statistical robustness in the screening method. T m measurement and hit identification were performed by automated analysis of the data (Experimental Procedures). The distributions of T m values were plotted as a percentage of the relative population (Fig. 4E). Hits were defined as those for which the T m > 67.2 °C, derived from 3*SD. The distribution of T m values from all 2,160 compounds showed that no molecules in the library stabilized PrP (Fig. 4E), indicating none had measurable affinity for recombinant PrP(89-230).

Identification of compounds that increase the susceptibility of PrP Sc to proteolysis in vitro
Previous studies indicated that the antiprion activity of positively charged branched polyamines could be attributed to their ability to directly interact and disaggregate PrP Sc in acidic lysosomal compartments, rendering PrP Sc susceptible to proteolysis (12,41). We therefore analyzed the 16 identified antiprion compounds for this property by measuring their ability to increase the PK sensitivity of PrP Sc under acidic conditions in an in-vitro degradation assay.
Positively and negatively charged polyamidoamines, generation 4.0 and 4.5, were used as positive and negative controls, respectively. Compounds were added to PrP Sccontaining brain homogenates from RMLinfected mice and incubated for 2 h in acidic buffer (Experimental Procedures). Following incubation, the homogenates were digested with PK and the presence of protease-resistant PrP Sc was analyzed by Western blots. The results indicate that none of the 16 validated antiprion compounds disaggregated PrP Sc and increased the PK sensitivity of the prion-infected mouse brain homogenates in vitro (Fig. 5).

Identification of compounds that reduce PrP C expression levels
We next determined whether the antiprion activity of the 16 validated compounds could be due to induced changes in the expression level of PrP C . Uninfected N2a cells were treated with the compounds, after which semi-quantification of Western immunoblots was performed to determine PrP C levels. Of the 16 compounds, four appeared to decrease PrP C expression: Irigenin, 7-benzyl ether; 3-deoxy-3 -hydroxyangolensic acid methyl ester; amcinonide; and retinoic acid ( Fig  6). Amcinonide had the most pronounced effect, inducing a substantial decrease (>80%) in PrP C levels ( Table 1).
Identification of compounds that disrupt lipid rafts and PrP C localization -An alternative mechanism for decreasing PrP Sc formation is to alter the integrity of lipid rafts, membrane microdomains where PrP C is known to localize (42,43). We tested the 16 compounds for their ability to disrupt lipid rafts in uninfected N2a cells by detecting PrP and flotillin (a marker of lipid raft integrity (44) in detergent-resistant microdomains (DRMs), which were isolated by sucrose gradients using a flotation assay (see Experimental Procedures) (45). Whereas silver staining of the fractions collected from DMSOtreated cells showed that soluble proteins remained primarily at the bottom of the gradient (fractions 7 and 8), Western-blot analysis indicated that PrP colocalized with flotillin ( Fig.  7A) at the interface between 5% and 30% sucrose concentrations (Fraction 3). Active compounds were next tested, and all fractions from flotation assays were analyzed by Western immunoblotting for PrP and flotillin (Supplemental Fig S2). To compare directly the effects of the compound set, the levels of PrP and flotillin in fraction 3 of each gradient were compared on the same Western blot (Fig. 7B). Only lovastatin at 10 μM was able to disrupt lipid rafts (Fig. 7B, lane 10). Next, compoundinduced changes in PrP localization were analyzed by immunocytochemistry in uninfected N2a cells (Fig. 8A, Supplemental Fig. S3). Retinoic acid and lovastatin showed a diffuse signal within the cell that may be attributed to a change in PrP localization. More interestingly, irigenin, 7-benzyl ether; 3-deoxy-3hydroxyangolensic acid methyl ester; and amcinonide induced a general decrease of intracellular and membrane-localized PrP C . Because amcinonide induced the most substantial decrease in PrP C levels and this effect was dose dependent (Fig. 8B), we decided to further investigate its mechanism of action.

Effect of amcinonide on PrP C expression and degradation
We investigated whether amcinonide affects PrP C abundance at the level of transcription or protein stability. Cultured mouse N2a cells were treated with control solvent (0.5% DMSO) or amcinonide (50 μM) for 6, 24, 48 and 72 h. Total mRNA was isolated, and quantitative RT-PCR was performed to detect the expression levels of Prnp mRNA. Results indicated that amcinonide did not alter the level of Prnp mRNA (Fig. 8C). Next, we investigated whether amcinonide, added at its half-effective concentration (20 μM), increases the rate of PrP C degradation after the inhibition of translation by cycloheximide. Strikingly, while PrP C in untreated cells had a half-life of 12-18 h, its half-life decreased to less than 6 h with the addition of amcinonide (Fig 6D). The results indicate that amcinonide alters the PrP C expression levels by increasing its rate of intracellular degradation.

DISCUSSION
In protein aggregation diseases, including prion diseases, a common target for pharmacological intervention has been the aggregating molecule itself (46,47). While we were able to identify 16 nontoxic compounds that were active against prion formation in cell culture, none of the 16 molecules interacted with PrP C or PrP Sc in vitro, as evaluated by various biophysical techniques. These observations imply that at least one other cellular target-likely several given the diversity of the active molecules-modulates PrP conversion within the cell. Our results have broad implications for the potential mechanism of known antiprion compounds and the development of novel strategies for the discovery of effective therapeutics against prion diseases.
The set of active antiprion compounds was identified by a high-throughput, ELISAbased assay utilizing ScN2a cells. Sixteen compounds were confirmed by secondary Western-blot analysis. The results of three orthogonal, in-vitro assays (ITC, CD and DSF) suggest that none of the compounds directly interacted with recombinant PrP. Indeed, DSF analysis of the entire Microsource library was unable to identify any compounds that stoichiometrically bound to PrP with measurable affinity. The results are consistent with the reported crystal and NMR structures of PrP C , which indicate the lack of prominent structural clefts capable of ligand binding (32)(33)(34)(35). Various compounds have been reported to bind PrP C in vitro. However, these interactions have been shown to be nonspecific (e.g., anionic tetrapyrroles (48), irreproducible in alternative binding assays (e.g., Gn8 (10)), or have millimolar affinities (e.g., quinacrine (49)). More recently, it was reported that a cationic tetrapyrrole [Fe(III)-TMPyP] can specifically interact with the folded domain of human PrP with micromolar affinities, although the large size of the compound may preclude it as an effective brain-permeable clinical therapeutic (50). The results of our study suggest that a drug-discovery screen focused on the identification of specific PrP-interacting molecules is unlikely to identify many efficacious compounds.
Of 16 analyzed compounds, none rendered PrP Sc susceptible to proteolysis in vitro. The compounds exert their activity only in the context of intact cells, suggesting that they are acting on other targets involved in the formation and propagation of prions rather than with PrP Sc itself. One of the analyzed compounds, amcinonide, substantially reduced the cellular expression level of PrP C , providing a potential mechanism of action for its antiprion activity. Interestingly, a previous screen (26,39) identified an antiprion compound budesonide that is structurally similar to amcinonide. Future studies using an immunoprecipitation approach could be used to identify the molecular targets of this class of compounds and establish a link between their targets and PrP C expression.
Reductions in the expression level of PrP C represent a particularly promising mechanism of action for antiprion compounds. A compound that targets the expression of PrP C has the potential to remove the substrate for all pathogenic misfolded conformations of PrP and be efficacious against a range of prion strains. Several lines of evidence support the idea that reducing the expression level of PrP C can delay or prevent prion disease progression. In fact, one of the most compelling confirmations of the prion hypothesis was the demonstration that mice whose PrP gene has been deleted (Prnp 0/0 ) do not propagate infectious prions and are resistant to prion disease (51). Ablation of the neuronal PrP gene utilizing a cre-lox recombination strategy extended the incubation period of prion-infected mice (52). Using bigenic mice with PrP gene expression under the control of an inducible promoter, it was demonstrated that the downregulation of PrP can dramatically extend the lifespan of prioninfected mice (53).
Our screen identified two statin drugs: lovastatin and rosuvastatin. By inhibiting the cholesterol synthetic pathway, it had previously been shown that lovastatin alters the integrity of the lipid raft compartments, reducing the quantity of PrP C available for conversion to PrP Sc (42). Other hits, including isoliquiritigenin, irigenin, 4'-hydroxychalcone, genistein, daizein, triacetylresveratol, ethopropazine and chrysanthellin A, belong to chemical classes that had been shown previously to have antiprion activities (15,26,27). We also found four structurally novel antiprion compounds: 3deoxy-3 -hydroxyangolensic acid methyl ester, digitoxin, dehydrovariabilin and digitonin. Although the mechanisms of action of these compounds are unclear, our in-vitro analysis suggests that they are unlikely to interact directly with PrP C .
The key observation to emerge from this study is that none of the antiprion molecules active in cell culture is directly active on PrP itself. Given the presence of previously discovered antiprion compounds among our hits, this may be broadly true of molecules with antiprion properties. From a structural standpoint, this possibility is not unreasonable: the structure of PrP C reveals few if any pockets well suited to sequester a small molecule, at least in the native state. While the molecular targets of the antiprion molecules identified here remain to be determined, the compounds may act on some "druggable" targets, including proteins that modulate PrP expression, localization, and stability in the cell. The identification of these cofactors will advance our understanding of prion and other aggregative disorders as well as enable the optimization of anti-aggregative therapeutics against these debilitating diseases.

FOOTNOTES
This work was supported by grants from the National Institutes of Health (AG02132, AG10770, and AG021601) as well as by a gift from the G. Harold and Leila Y. Mathers Charitable Foundation. We thank A. Serban for providing recombinant MoPrP(89-230) and MoPrP , and F. Cohen (TPG Biotech, San Francisco) for helpful discussions.
The abbreviations used are: u C p , change in heat capacity; H, change in enthalpy; T m , change in melting point.  Table 1. Quinacrine (QA) and DMSO alone were analyzed as positive and negative controls, respectively. Apparent molecular weight markers are indicated in kilodaltons. (E) Quantification of PrP Sc band intensities relative to untreated controls. The gray, white and black bars indicate the relative intensities of the unglycosylated, monoglycosylated and diglycosylated bands, respectively.     5. Susceptibility of PrP Sc in prion-infected brain extract to protease digestion after exposure to 16 validated antiprion compounds. Brain homogenates of RML prion-infected mice were treated with 100 M of the indicated antiprion compounds for 2 h, then subjected to PK digestion and Western blot analysis. Lane numbers correspond to compound IDs in Table 1. PAMAM G4.0 and G4.5 were used as positive and negative controls, respectively. DMSO was also analyzed as a negative control. Apparent molecular weight markers are indicated in kilodaltons. Fig. 6. PrP C expression levels in uninfected N2a cells treated with the 16 validated antiprion compounds at the indicated concentrations for 5 days. Conjugated D13-HRP was used to detect PrP in Western immunoblots. Lane numbers correspond to compound IDs in Table 1. As a positive control, 3 M of quinacrine (QA) was used. As a negative control, N2a cells were incubated with DMSO.  Table 1). Blots were probed with anti-PrP (top) and anti-flotillin (bottom) antibodies. Apparent molecular weight markers are indicated in kilodaltons.