Discovery of Novel Anti-prion Compounds Using In Silico and In Vitro Approaches

Prion diseases are associated with the conformational conversion of the physiological form of cellular prion protein (PrPC) to the pathogenic form, PrPSc. Compounds that inhibit this process by blocking conversion to the PrPSc could provide useful anti-prion therapies. However, no suitable drugs have been identified to date. To identify novel anti-prion compounds, we developed a combined structure- and ligand-based virtual screening system in silico. Virtual screening of a 700,000-compound database, followed by cluster analysis, identified 37 compounds with strong interactions with essential hotspot PrP residues identified in a previous study of PrPC interaction with a known anti-prion compound (GN8). These compounds were tested in vitro using a multimer detection system, cell-based assays, and surface plasmon resonance. Some compounds effectively reduced PrPSc levels and one of these compounds also showed a high binding affinity for PrPC. These results provide a promising starting point for the development of anti-prion compounds.

time-dependent manner with an IC 50 below 1 μ M, by inducing cellular clearance of PrP Sc3 . In addition, phenothiazine, statin, and quinacrine provide attractive options because they have been approved by the United States Food and Drug Administration for use in other diseases 7,9 . However, these drugs were shown to be ineffective against prion disease in rodents 10,17 . The toxicity of anti-prion compounds and their inability to cross the blood-brain barrier has limited their successful application 18 .
In cell culture systems, anti-prion compounds are generally assessed by monitoring the levels of protease-resistant PrP Sc using proteinase K (PK) digestion followed by western blotting. As this screening approach is fairly time-consuming and semi-quantitative, we employed a highly quantitative high-throughput misfolded protein detection assay (multimer detection system; MDS) to screen compounds for anti-prion efficacy. This competition assay uses a magnetic bead-conjugated capture antibody and a horseradish peroxidase (HRP)-conjugated detection antibody, with overlapping epitopes to achieve specific detection of multimers (such as PrP Sc ), and not monomers (such as PrP C ). The T2 and 3E7 prion antibodies employed by the MDS recognize amino acids 147-152 and 140-160, respectively, of the PrP sequence 19 .
Although PrP C and PrP Sc usually have identical primary amino acid sequences, it has been shown that the conversion from PrP C to PrP Sc causes a substantial change in the secondary protein structure at various locations, including the factor X-binding site, the hotspot binding site, and the unstructured N-terminal binding site [20][21][22] . Several computational and biophysical studies have targeted these significant regions and used well-known anti-prion compounds to demonstrate stabilization of the secondary structural changes 23,24 . Anti-prion compounds that have been identified by different research groups possess diverse scaffolds and similar inhibitory activities, highlighting the need for clarification of the structure-activity relationship (SAR).
The recent development of structure-based virtual screening supported by docking simulations has facilitated effective in silico screening of the interactions between chemical compounds and their target proteins, which can contribute to the identification of a desired activity from a large database of chemicals that are structurally different from known active compounds, reducing the time and cost of identifying chemical hits 25,26 . Using the structure of PrP C -GN8 (a known anti-prion compound), a 3D pharmacophore model was generated and compounds were docked into the prion hotspot to determine their potential binding mode, which enabled the selection of a small number of molecules for in vitro testing. In total, 37 compounds were assessed by MDS assay, in scrapie-infected mouse neuroblastoma N2a (ScN2a), in PrP C -overexpressing N2a (L2-2B1) cells, and by surface plasmon resonance (SPR) direct-binding experiments.

Results
Virtual Screening. The overall discovery steps employed in the present study are shown in Fig. 1a.
Ten pharmacophore models were generated using the receptor-ligand interaction protocols in Discovery Studio (DS) and the best was chosen using the Genetic Function Approximation (GFA) model (Fig. 1b). To generate the receptor-ligand interaction-based pharmacophore model, a well-defined anti-prion compound (GN8) was docked into the PrP C hotspot (1AG2), as shown in Supplementary Figure 1a 27,28 . Although several crystal and nuclear magnetic resonance (NMR) structures are available for PrP C , it should be noted that only the NMR of PrP C (1AG2) with GN8 structural details has been thoroughly characterised 23 . Because the pharmacophore model was based on the previously reported NMR structure of PrP C -GN8, we expected it to include two important hydrogen bonds from Glu196 and Asn159. As anticipated, the pharmacophore model included five features: the two key hydrogen bonds, two hydrophobic features (H), two hydrogen bond acceptors (HBAs), and one hydrogen bond donor (HBD) (Fig. 1b). To validate this pharmacophore model, we also mapped a previously reported anti-prion compound, GJP49.
Virtual screening of our in-house chemical database with the selected pharmacophore model yielded 1110 compounds. Cluster analysis was performed to filter these compounds, and 682 diverse compounds were selected based on their structural similarities and differences. The compounds were further subjected to visual inspection, resulting in the selection of 37 drug-like compounds for further evaluation.
PrP C Compound Docking. The highest binding energy of the 37 hit compounds was measured using the AutoDock score. All compounds were docked into the PrP C hotspot in the same way, with a similar binding mode. Fifty docked conformations were obtained per compound and their best binding energy (kcal/mol) is listed in Table 1. To illustrate the interaction, the binding modes of the PrP C -compound interactions are shown in Fig. 2, which illustrates a compound anchored at the PrP C hotspot (Fig. 2a) and interacting with specific amino acids via multiple hydrogen bonds and hydrophobic interactions (Fig. 2b).

MDS Assay of PrP Sc Inhibition.
Our primary in vitro screening test employed the MDS enzyme-linked immunosorbent assay (ELISA) to quantify PrP Sc formation. Recombinant PrP was exposed to each of the 37 hit compounds or 1 μ M of quinacrine as a positive control. As shown in Fig. 3, quinacrine significantly reduced the formation of PrP Sc . The effects of the hit compounds in this assay were not always concentration-dependent. Some compounds actually increased PrP Sc formation at a concentration of 20 μ M (BMD42-01, 12, 19, 20, and 30). Twelve compounds (BMD42-03, 06, 07, 08, 10, 16, 23, 26, 28, 29, 31, and 35) exhibited > 50% inhibition of PrP Sc formation at both concentrations tested (5 μ M and 20 μ M). Based on this primary screening, we selected 7 compounds with statistically significant effects (BMD42-06, 23, 26, 29, 31, 33, and 35) for secondary cell culture screening. One of the selected compounds (BMD42-31) showed outstanding inhibitory effects in the MDS assay and even out-performed the positive control, quinacrine. Cytotoxicity Assay. We used a commercially available cytotoxicity assay to identify suitable treatment concentrations for each compound. Seven compounds were tested at six different concentrations between 0.5 μ M and 1 mM (Fig. 4) in ScN2a and L2-2B1 cells. Most of the compounds caused severe cytotoxicity at concentrations above 200 μ M. BMD42-06 showed the lowest cytotoxicity of the 7 compounds tested. Concentrations between 0.5 and 30 μ M were considered optimal, and concentrations of 5 and 20 μ M were thus selected for further study. There were no observable differences in cell viability between the ScN2a and L2-2B1 cell lines. We observed an obvious cytotoxic effect of quinacrine at above 2 μ M in both cell lines.

Compound Effects on PrP C and PrP Sc Propagation in Cultured Cells. ScN2a cells are infected
with the Rocky Mountain Laboratory scrapie prion strain and persistently produce PrP Sc . This cell model is widely used to screen anti-prion candidate compounds. We first used a transfected cell line over-expressing PrP C to examine anti-prion effects. This cell line facilitated the observation of changes in the amount of PrP C , relative to a normal neuronal cell. Figure 5 shows representative immunoblots for PrP Sc and PrP C or total PrP obtained from ScN2a and L2-2B1 cells exposed to different concentrations of the 7 test compounds, quinacrine (1 μ M), or dimethyl sulfoxide (DMSO). Compounds with anti-prion activity would reduce the protein levels. The 7 compounds selected by MDS primary screening revealed a variety of inhibitory effects on PrP Sc levels in ScN2a cells (Fig. 5a). The amount of PrP Sc was concentration-independently reduced by BMD42-06 and 35. The extent of this reduction equaled the effect of quinacrine ( Fig. 5a lane 1). BMD42-23, 29 and 33 also reduced PrP Sc levels, whereas BMD42-26 and 31 only produced minor inhibitory effects. However, the reduced levels were not detected in L2-2B1 cells during this study, irrespective of whether the cells were exposed to hit compounds or controls (Fig. 5b). In ScN2a cells, total PrP levels were reduced in the presence of BMD42-35 but were unaffected by quinacrine, DMSO, or the other 6 hit compounds (Fig. 5c,g). In L2-2B1 cells, PrP C was inhibited by 6 compounds, but not by BMD42-29 (Fig 5d,h). Thus, had only ScN2a cells been used, inhibitory effects against PrP C may not have been found for other compounds. Among them, BMD42-35 showed the strongest inhibitory effects in the cell-based assay. Quinacrine completely inhibited the formation of PrP Sc , but not PrP C , at 1 μ M.
SPR Measurement of Compound-PrP C Binding Affinities. The 37 hit compounds identified by VS were tested by SPR to quantify their direct PrP C binding ability. Multiple strategies were explored for the immobilization of PrP C to the surface of a high-density sensor chip. Direct measurements of quinacrine binding were repeatedly performed to confirm the stability and function of the coated chips. SPR revealed that the majority of the compounds interacted directly with PrP C . The apparent affinities were determined for 23 compounds (Table 1, dissociation constant rate; K D ). The remaining 14 compounds were tested, but did not bind. These compounds showed consistent results in the MDS assay. Sensorgram curves revealed that similar to previously reported GJP derivatives, our compounds showed single and specific binding to PrP C . In particular, BMD42-29 showed rapid association and slow dissociation rates (Fig. 6). This compound showed at least 2-fold tighter binding to PrP C than the other hit compounds. Notably, these findings correlated with those of the computational docking study, where BMD42-29 had the highest AutoDock binding energy (− 7.87 kcal/mol) of the hit compounds tested (Table 1). BMD42-35 was found to be effective in the cell-based assay, but did not show binding in the SPR assay.
Binding Mode and Selectivity. The predicted binding modes for the 37 hit compounds were analyzed to investigate the structural basis of their selectivity. We grouped the compounds by their inhibition efficacy to examine the SAR. Although BMD42-35 was one of the most active compounds in the cell-based assay, it showed no binding in the SPR assay. Several of the other hit compounds showed independent anti-prion activity in cell-based and SPR assays (BMD42-2, 23, 29 and 33). Comparison of the scaffolds of these 4 compounds revealed diverse structural features. Notably, these most active compounds  Student's t-test was used and differences were considered statistically significant at a P value of < 0.05.
included three sulfonamide compounds (BMD42-23, 29, 33), two thiazole compounds (BMD42-02 and 23) and one benoxazole compound (BMD42-29). The 2-amino thiazole scaffold present in BMD42-23 had previously been widely studied in the context of anti-prion drug discovery and had even been used in a clinical trial for prion disease 29 . We next investigated the binding mode of each compound with PrP C , bearing in mind that two strong hydrogen bonds from Asn159 and Glu196 play important roles in inhibitory activity (Fig. 2). In addition to the high affinity compounds, 10 more compounds (BMD42-31, 26, 16, 03, 04, 10, 22, 07, 34,  In addition, their binding modes predicted a single hydrogen bond and hydrophobic contacts with the hotspot residues of Glu196/ Asn159/Lys194/Glu160/Leu130, which may contribute to stabilization of the PrP C structure. In addition, we grouped 7 moderately active compounds (BMD42-17, 30, 19, 24, 36, 14, and 12) and elucidated their binding with hotspot residues. The chemical structures of these compounds confirmed that the presence of sulfonamide and amide scaffolds (BMD42-24 and 14) produced potent anti-prion effects.
To confirm the above findings, we investigated two less active compounds, BMD42-01 and 05 ( Supplementary Fig. 2). These compounds had no hydrogen bonding or hydrophobic interactions with key hotspot residues, as their structures differed from those of the most active compounds. In contrast, one of the most active compounds (BMD42-29) showed strong hydrogen bonding at the helix regions of Asn159 and Glu196, and hydrophobic interactions with Leu130 and Arg156. These findings emphasized the importance of these hydrogen bonds and hydrophobic interactions for efficient anti-prion effects. Compounds lacking interaction with Asn159 and Glu196 were unable to inhibit PrP Sc formation ( Supplementary Fig. 2).

Discussion
Research aiming to develop treatments for prion diseases has led to the identification of a range of compounds such as quinacrine and doxycycline that can selectively block conversion of PrP C to PrP Sc30 . However, these compounds have not provided useful treatments for patients. To discover and validate novel non-toxic and effective anti-prion drugs, we studied a structurally diverse series of compounds with anti-prion activity using an in silico approach. Use of the structures of PrP C -hit compound complexes enabled analysis of compound localization at the hotspot sites and interactions with key residues in the previously determined NMR PrP C -GN8 complex.
The three most effective strategies for discovering anti-prion compounds include screening for inhibitory effects on PrP Sc accumulation, inactivation of endogenous PrP C (as the substrate for prion conversion), and the enhancement of PrP Sc degradation 31 . Inhibition of in vitro PrP Sc accumulation represents a primary target for prion disease therapy 16,32 .
Expression of PrP c in host neurons is required for PrP Sc replication and disease progress. As such, persistently infected cell lines acting as a host for PrP Sc are frequently used to study prion-related cellular processes, and also to screen for effective anti-prion compounds 33 . ScN2a cells have been used extensively as a relevant model for the study of prion diseases 34 . In the present study, considerable reductions in PrP C levels were only observed in L2-2B1 cells, possibly because the cells had higher initial levels of PrP C . The L2-2B1 cell line was therefore very useful for analyzing this aspect of the compounds' activities. Quinacrine was cytotoxic in both cell lines at concentrations above 2 μ M. This result was consistent with a previous study in which cytotoxicity was tested in the presence of 0.02-200 μ M quinacrine, and the optimal concentration found to inhibit PrP Sc formation was below 4 μ M 7 .
We performed a highly quantitative and precise protein misfolding detection assay for the primary screening of anti-prion compound efficacy. The results of this MDS assay facilitated the selection of 7 potent hit compounds (BMD42-06, 23, 26, 29, 31, 33, and 35). Although BMD42-31 was more effective than quinacrine in the MDS assay, it was excluded during our secondary cell-based screening as it produced only a minor reduction in PrP Sc level, although the PrP C level was reduced at 20 μ M. Chemical modification of this compound may increase its efficacy. The MDS assay results appeared to indicate that BMD42-06, 23,26,29,31,33, and 35 conformationally stabilized PrP C by reducing its aggregation. BMD42-06, 23,26,31,33, and 35 also decreased the level of PrP C . The mechanisms underlying this effect may involve PrP C consumption as a substrate for aggregation; a reduction in the PrP mRNA level; or protein degradation caused by compound-PrP C binding. The results indicated that in ScN2a cells, BMD42-06, 23, 29, 33, and 35 reduced the amount of PrP Sc , either by interfering with the conformational interaction between PrP C and PrP Sc , or by interacting directly with PrP Sc .
Recently, Kamatari and co-workers classified anti-prion compounds according to four potential molecular mechanisms of action: (i) specific conformational stabilization of PrP C ; (ii) nonspecific stabilization, including interference with the conformational interaction between PrP C and PrP Sc , in addition to hotspot binding; (iii) promotion of PrP C aggregation and precipitation, thus reducing the amount available for conversion to PrP Sc ; or (iv) interactions with molecules other than PrP C , such as PrP Sc or membrane proteins 35 . The compounds identified in the present study showed multiple mechanisms of action, based on their effects on PrP C aggregation, reduction, and PrP Sc propagation, as well as their binding affinities. Placing our results in the context of Kamatari's classification, BMD42-29 may act by specific stabilization of PrP C , as evidenced by its strong binding affinity. BMD42-06, 23, 29, 33, and 35 may act by nonspecific blockade of PrP C conversion to PrP Sc , or by interacting with PrP Sc , as evidenced by their weak binding affinities and the observed reduction in the level of PrP Sc or PrP C aggregate. BMD42-06, 23, 26, 31, 33, and 35 may act by reducing the amount of PrP C by reducing PrP expression, or stimulating PrP degradation. Further mechanistic studies are required to investigate exactly how these compounds alter PrP C levels.
Our results suggest that BMD42-29 is an optimal compound, exhibiting PrP Sc inhibition and a stronger binding affinity than other anti-prion compounds reported to date. However, it did not produce a marked reduction in PrP C levels, possibly indicating that it stabilized PrP C and inhibited its pathological conformation change to PrP Sc . BMD42-35 produced the highest decrease in PrP Sc and PrP C levels, although it showed a lower binding affinity, which suggests that BMD42-35 acts by either stabilizing or eliminating PrP C and may control PrP C non-specifically. This contradictory result may relate to problems with SPR assays reported by earlier studies, including the unlimited increase in ProteOnGLH sensor responses, low solubility of compounds, and slow binding kinetics 36 .
It has been proposed that the hydrogen bonds from stand S1 and helix B may prevent the conversion of PrP C to PrP Sc21 . Collectively, our compounds showed strong hydrogen bonds at Asn159 (stand S1) and Glu196 (helix B) within PrP C . The most active compounds occupied the hydrophobic area highlighted in the pharmacophore model generated from the PrP C -GN8NMR structure. BMD42-29, which was active in both cell-based and SPR-based assays, and BMD42-35, which was most active in the cell-based assay, both interacted with conserved PrP C hotspot residues, indicating the importance of the two hydrogen bonds and the hydrophobic environment; this was predicted by the pharmacophore model.
Further elucidation of the mechanisms of action of BMD42-29 and BMD42-35 will provide unique tools to study the mechanism of prion replication. The approach used in the present study may provide in vitro screening data that are more highly predictive of in vivo activity, contributing to the rational design of novel and effective anti-prion treatments.

Methods
Pharmacophore Development. The protein data bank NMR structure of PrP C (1AG2) was used as the template for pharmacophore generation 37 . DS software was used to map the active site of PrP C and critical residues were identified based on the PrP C -GN8 NMR data 10 . Subsequently, the receptor-ligand pharmacophore generation protocols were used to generate a quantitative model of PrP C . The protocols resulted in 10 3D pharmacophore hypotheses and the highest ranked of these was selected.
Virtual Screening. The 3D pharmacophore hypothesis was used to extract chemical compounds from our in-house database of 700,000 compounds. The screening processes were performed using the best flexible conformation search method in DS, with standard settings. Molecular Docking. Molecular docking calculations were performed using the AutoDock tool. To validate the AutoDock, we docked the known anti-prion compounds, GN8 and GJP49, into the PrP C hotspot (1AG2). A grid of 63, 57, and 53 points in the x, y, and z directions was constructed on the center of the PrP C hotspot mass. A default setting grid spacing of 0.375 Å and a distance-dependent function of the dielectric constant were used for the energetic map calculations. The docked compounds were subjected to 50 runs of the AutoDock Lamarckian genetic algorithm, with 500,000 steps of energy evaluation and default values for the other parameters. Cluster analysis was performed on the results using 1.0 Å. The docking pose analysis was conducted for the first pose of the most populated cluster in the AutoDock output. purchased from Princeton Biomolecular Research. BMD42-20 was purchased from Synthon-lab. BMD42-25 was purchased from Sigma Aldrich. BMD42-31 was purchased from Uorsy. The compounds were dissolved in DMSO, diluted to produce a 50 mM stock solution, and stored at -20°C. ScN2a cells were derived from N2a cells obtained from the ATCC and were generously provided by Dr. Ryu, Hanyang University, Korea. The L2-2B1 subclone cell line overexpressing PrP C derived from N2a was established by Dr. Kim, Korean National Institute of Health. MDS Assay. The MDS assay kit was supplied by People Bio Inc. and performed according to the instructions, with minor revision 19 . Briefly, each compound was pre-treated with the reaction buffer containing 50 ng of recombinant PrP, 1% Triton X-100, 10% Blockace and Tris-buffered saline containing 0.1% (vol/vol) Tween 20 (TBST) in 2 mL screw cap tubes. DMSO (0.1% vol/vol) was used as a negative control and quinacrine (1 μ M) (Sigma Aldrich, Q3251) was the positive control. The mixture was incubated with continuous shaking for 3 h at 37 °C. The 3E7 PrP antibody (2 μ g), conjugated to magnetic beads, and the HRP-conjugated PrP T2 antibody (8 μ g) were added to the pre-incubated mixture. After incubation for 1 h under the same conditions, the beads were separated and washed three times with TBST using a magnetic particle concentrator (Invitrogen, 120.20D). The assay signal was developed by adding Supersignal ELISA pico chemiluminescence substrate (Pierce, 37070) and quantified using a VICTOR3 microplate reader (Perkin Elmer, 1420-032).
Cell Culture and Treatments. ScN2a and L2-2B1 cells were seeded in six-well plates (2 × 10 5 cells/well) and incubated in Opti-MEM (Gibco, 31985) containing 10% fetal bovine serum, 1% penicillin-streptomycin (Gibco, 15140) and 2 mM l-glutamine (Gibco, 25030) with the indicated concentrations of the compounds in 5% CO 2 at 37 °C. The stock compound solutions (50 mM) were diluted in Opti-MEM. Control cell cultures were treated with DMSO only (0.1% v/v). Each compound was tested in duplicate in three independent experiments.
Western Immunoblotting. Protein samples were separated by SDS-polyacrylaminde gel electrophoresis for 2 h at 150 V at 4 °C, using 12% gels. Proteins were then transferred to a polyvinylidenefluoride membrane using the i-Blot system (Invitrogen) for 7 min. The membrane was blocked for 1 h at room temperature in 5% (wt/vol) skim milk in PBS containing 0.1% (vol/vol) Tween 20 (PBST) and then incubated overnight at 4 °C with the PrP antibody, SAF-32 (1:200) (Cayman Chemicals, 189720) diluted in 10% (vol/vol) blocking buffer. After washing with PBST, the membrane was incubated for 1 h at room temperature with HRP-conjugated anti-mouse IgG (1:2,000) (DAKO, P0260). The signals were visualized using ECL (Elpis Biotech, EBP-1073) and the protein bands were scanned by Image Scanner III (GE). The relative band densities are shown as the volume intensity/mm 2 , relative to the β -actin band density. Blots used for PrP C detection were stripped and then re-probed using a β -actin antibody (1:5000) (Cell Signaling, 4970). Cytotoxicity Assay. Cytotoxicity was evaluated using a commercial kit (CellTiter 96 Non-Radioactive Cell Proliferation Assay, Promega, G4001), following the manufacturer's instructions. ScN2a and L2-2B1 cell suspensions (5,000 cells) were plated into each well of a 96-well plate. After adding the compounds at six different concentrations between 0.5 μ M and 1 mM or the control, the cells were incubated for 6 days. Dye solution was then added to each well and incubated for up to 4 h. Solubilization solution/stop mix was then added and cell viability was measured using an ELISA plate reader (Bio-Rad) at a wavelength of 570 nm. Cytotoxicity was expressed as a percentage of the signal observed in untreated control cells. or high K a ), and stable complex formation (slow 'off rate' or low K d ), in accordance with the equation, K D = K d /K a . Statistical analysis. Each experiment was repeated a minimum of three times. The one-way analysis of variance with the Tukey-Kramer procedure was used. Differences were considered as significant at P < 0.05.