Transmissible spongiform encephalopathies (TSEs) or prion diseases are a group of rare neurodegenerative diseases. Uniquely, prion diseases occur familial, sporadic and infectious forms (Gambetti et al., 2003). According to the statistics, over eighty percent of clinical cases are recognized as sporadic (Appleby et al., 2009; Sun et al., 2020). Prion diseases are ascribed to misfolding of prion protein. Structurally, prion protein has two distinct isoforms such as Cellular form prion protein (PrP^C) and Scripe form prion protein (PrP^Sc). PrP^C is structurally rich with α-helix and is anchored to membranes by glycosyl phosphatidyl inositol (GPI). Under unknown conditions, PrP^C structurally converts into an infectious isoform—PrP^Sc with high β-sheet conformation. PrP^C is sensitive to proteinase K (PK), whereas PrP^Sc is resistant to PK. The existing PrP^Sc induces more PrP^C to replicate into PrP^Sc. Furthermore, PrP^Sc initiates a subsequent cascade of structural conversion to form insoluble amyloid PrP fibrils with abundant cross-β-sheets. The PrP^Sc spreads from the central nervous system (CNS) to other tissues via peripheral nervous system (PNS) (Gough and Maddison, 2010).

The misfolded PrP recruits the native PrP^C for propagation, and then expands the transmission region in the brain (Prusiner, 1998). In in vitro model and in vivo mouse model, PrP^Sc induces α-synuclein fibril formation and α-synuclein fibril also converts PrP^C to PrP^Sc (Masliah et al., 2012; Katorcha et al., 2017). Prion infection induces the decrease of α-synuclein expression and the hyperphosphorylation of p(S129)-α-synuclein, one hallmark of Parkinson’s disease, is colocalized with PrP aggregates (Chen et al., 2021). In the case of abnormal β-amyloid (Aβ) amyloidosis, infected prion patients or mice had Aβ deposition in brain without tau-pathology (Cali et al., 2018; Ezpeleta et al., 2019; Jankovska et al., 2021). In PrP^Sc-infected APP23 mice, the downregulation of the PDK1-TACE pathway triggers the pathway into beta-site amyloid precursor protein cleaving enzyme (BACE) cleavage, which produces Aβ40/42 and induces Aβ multimer formation, decreasing the survival rate (Ezpeleta et al., 2019). Shortly speaking, misfolding of PrP is not only the cause of prion diseases, but also the inducer of other diseases. Inhibiting the progression of PrP^Sc expansion in neurodegenerative diseases is a critical issue.

Scientists have been working on the development of anti-prion agents for decades (Gandini and Bolognesi, 2017). The function of an anti-prion agent includes inhibiting PrP^C-PrP^Sc conversion, inhibiting PrP^Sc polymerization, and promoting fibril degradation (Uliassi et al., 2022). It is difficult to interrupt the fibrillation once the protein misfolding is initiated. Therefore, many studies are focused on blocking the structural transition of amyloid protein into amyloid fibril. For prion diseases, the most common strategy is to stabilize the structure of PrP^C with chemical compounds. Under this strategy, the structure of PrP^C is critical. Several PrP^C structures have been solved by nuclear magnetic resonance spectroscopy (NMR) (Riek et al., 1996; Zahn et al., 2000). Some structure-based drug designs were conducted based on these NMR structures (Baral et al., 2014; Herrmann et al., 2015; Ishibashi et al., 2016). Subsequently, several anti-prion compounds, e.g. promazine, phenothiazine derivatives, and polythiophenes derivates were found effective in pharmacological studies (Baral et al., 2014; Herrmann et al., 2015; Ishibashi et al., 2016). However, these anti-prion compounds failed in clinical trials. These failures might be ascribed to the improper receptors (mammalian PrP^C protein monomers or yeast prion monomers) of these anti-prion compounds. It is considered to change the drug target of structure-based design from monomeric PrP to the structurally distinct mammalian PrP fibrils. This is the approach used in this study.

Solving the molecular structure of amyloid fibrils is very challenging. Fortunately, with the advantages of a revolutionary technique called cryo-EM, the three-dimensional structures of biological macromolecules with high resolution can be solved in a short time (Callaway, 2020). In 2020–2021, three structures of wild-type prion fibrils were solved by cryo-EM at a resolution within 3.5 Å, including fibrils of PrP_106–145 (fPrP_106–145, PDB 6UUR) converted with 4 M urea at pH 4.0 (Glynn et al., 2020), fibrils of PrP_170–229 (fPrP_170–229, PDB 6LNI) with 2 M guanidinium chloride (Gdn-HCl) at pH 7.4 (Wang et al., 2020) and fibrils of PrP_95–227 (fPrP_95–227, PDB 7LNA) from 263K mouse brain homogenate (Kraus et al., 2021). The cross-β-sheets and high β-sheet content from three identified prion amyloid structures are consistent with the early report (Wille et al., 2002). The high-resolution structure of fibrils can be used to resolve the interaction network contributing to fibril stability and can be used to design drugs for weakening the interaction to disrupt protein fibrils.

Polyphenols are a large family of natural compounds commonly found in plants. They are generally composed of multiple phenol units and are moderately water-soluble. Polyphenol molecules are categorized into five groups: phenolic acids, flavonoids, stilbenes, tannins, and lignans (Tijjani et al., 2020). It is found that polyphenols decrease fibril by inhibiting fibril formation or by fibril disaggregation with specific pathways depending on the proteins in Alzheimer’s and Parkinson’s diseases (Freyssin et al., 2018). In addition, these polyphenols serve as antioxidants in the cells to detoxify the oxidation of lipid, protein, and DNA from oxidative stress. The imbalance of production and scavenge of reactive oxygen species (ROS) or reactive nitrogen species (RNS) induces serious respond correlated to neurodegenerative diseases (Bhat et al., 2015). Quercetin, a kind of flavonoid, is absorbed by the small intestinal mucosa and then transported through blood vessels into the liver (Ulusoy and Sanlier, 2020). Quercetin is commonly found in our diet and has many benefits for anti-inflammation, anti-cancer, anti-neurological diseases, and diabetes mellitus (Ulusoy and Sanlier, 2020). Recently, we found that quercetin disaggregates mature prion fibrils, accelerates fibril clearance with PK, and reduces ROS levels in Neuro-2a cells (Yu and Lee, 2020). In addition, quercetin binding accelerates the formation of PK-sensitive and structure-less fibrils (Yu et al., 2022).

In drug development, structure-based drug design is the mainstream compared to ligand-based pharmacophore (Sharma et al., 2021). Structure-based drug design has resulted in the FDA’s approval of human immunodeficiency virus-1 protease inhibitors for acquired immunodeficiency syndrome, isoniazid for tuberculosis, and Pim-1 kinase inhibitors for cancer (Wlodawer and Vondrasek, 1998; Batool et al., 2019). A computer-aided drug design searches for the drug candidates that includes molecular docking and molecular dynamics (MD) simulation in a time saving way (Śledź and Caflisch, 2018). The protein-ligand interaction at the atomic level is resolved by MD simulation, and it is a complementary tool to NMR spectroscopy and X-ray crystallography.

In this study, we used the structure of fPrP_170–229 as the target for the screening of quercetin-like molecules in a pre-docking. To largely reduce the computing time, Lipinski’s rule of five (Lipinski, 2004) was applied to screen orally administrated and lipophilic drug candidates with pharmaceutical activity for curing the prion diseases in CNS, PNS, and other tissues. Based on the docking score and toxicity prediction, we minimized the number of target compounds from 1,300 compounds to 15 compounds (15 COMPs). The selective candidates were further evaluated for their toxicity and sequentially computed as shown in Figure 1.

The structure of mammalian prion fibril is extremely important for structure-based drug design. In accordance with the source of cryo-EM structures, fPrP_170−229 and fPrP_106−145 were converted from two different in vitro system, while fPrP₉₅₋₂₂₇ was converted from an in vivo system. Wang and colleagues found that fPrP_170−229 is structurally stabilized by a salt bridge formed by opposing subunits, resulting in the formation of hydrophobic cavity between contact surfaces (Wang et al., 2020). Different from Wang’s stacking model of fPrP_170−229 in neutral pH, Glynn and coworkers found that fPrP_106−145 is stacked in parallel and attached neighboring fibrils with the N-terminal Thr107-Leu125 (Glynn et al., 2020). After the cryo-EM results of fPrP_170−229, Requena and coworkers rebuilt the 4RβS model of fully glycosylated human PrP amyloid (Spagnolli et al., 2020). Through the tail end of C-terminal PrP^Sc in 4RβS model, PrP^Sc successfully induces structural conversion of PrP^C into PrP^Sc in MD simulation. To check the reliability of fPrP_170−229 from an in vitro system, we compared the fPrP_170−229 structure with the fPrP₉₅₋₂₂₇ structure from an in vivo 263K mouse model. The fPrP_170−229 has the similar secondary structure and tertiary structure in the comparison with fPrP₉₅₋₂₂₇ (PDB 7LNA, Supplementary Figure S9) (Kraus et al., 2021). As with fPrP₉₅₋₂₂₇, fPrP_170−229 exposes two glycosylation sites (Asn181 and Asn197) on the water-accessible surface. The E196K point mutation of PrP fibril (fPrP_175-217, PDB 7DWV, Supplementary Figure S10), was shown to have a large alteration in dimerized region (Wang et al., 2021). Two positive charge residues, K194 and K196, decrease the fibril stability by disrupting salt bridge directly and rearranging the local environment of interaction network (Wang et al., 2021). This evidence indicates the importance of the salt bridge of K194 and E196 for building up the fibril stability.

According to a previous study based on antibodies against different PrP regions, the flexible N-terminal tail mediates toxicity and globular C-terminal domain induces toxicity (Sonati et al., 2013). In the inherited prion diseases, the residues of point mutation are mainly located in the C-terminus (Prusiner, 1998). The infectious fragment of PrP^Sc has been identified as a part of the PK-resistant sequence (Wang et al., 2018). In the GPI-anchor free Scrapie infected mouse model, the highly PK-resistant fragment is the region from residue 179 to the C-terminal end (Vázquez-Fernández et al., 2012). In the comparison of protein misfolding cyclic amplification (PMCA) and Gdn-HCl induced conversion through hydrogen/deuterium exchange, the products from these two different conditions indicate the same highly protected region at residues 170–213 (Smirnovas et al., 2009). According to these studies, the C-terminal PrP has the most important role in prion infection and proteolytic resistance. In this study, the cryo-EM structure includes this highly protected region. Therefore, fPrP_170–229 is an ideal target to analyze the effect of drug candidates after their binding to the fibril core.

In MD simulation, the torsions of amino acid are recorded in the force field affecting the secondary structure formation. Currently, the bias of force fields has been found to tend to overestimate α-helix in the course of protein folding (Best et al., 2008). The MD simulation is based on the solid empirical data from experimental studies. For MD simulation of intrinsically disordered proteins (IDP) or fibrils, the force field is the most critical for the reliability of the results. Many force fields and explicit water models in MD simulation have been investigated in IDPs, such as Aβ and α-synuclein (Akbayrak et al., 2020). Akbayrak and colleagues have reviewed early-developed Amber force fields. In the use of the Amber99 force field, Aβ_16–22 favors the formation of helical structures in monomeric, dimeric, and trimeric forms. In contrast to the experimental findings, Aβ_16–22 cannot fold into β-sheets. Three explicit solvent models (SPC, TIP3P, and TIP4P) fold Aβ₄₀ into an α-helical structure when using the Amber03 force field. These problems were corrected in Amber99SB (a revised version of Amber99 (Hornak et al., 2006) along with TIP3P, as judged from the simulation of Aβ₄₂. Consistent with the experimental observation, α-synuclein forms an α-helical structure in the N-terminal region in MD simulation with Amber99*-ILDN or Amber99SB. In the MD tests of dimeric Aβ_16–22 fibrils, five newer force fields including Amber99-ILDN, Amber14SB, CHARMM22*, CHARMM36 and CHARMM36m were chosen for the study of amyloid peptide fibril-assembly (Man et al., 2019). In the modeling with Amber14SB, it was observed that there was a low transition in the secondary structure between anti-parallel β-sheet and parallel β-sheet in fibril dimer formation (Man et al., 2019). Shortly speaking, Amber14SB is a revised version to calibrate the side chain and backbone parameters of Amber99SB (Maier et al., 2015). Based on these studies, Amber14SB is suitable for resolving protein structures.

In the comparison of cryo-EM structure, similar Β’ value shown in Figure 2 indicates the suitability of the force field in this work. Solvent is critical for molecular interactions. The reliability of implicit solvent model has been investigated. Implicit solvent model has been used for simulation of paired helical filament 6 (PHF6) fragment from tau protein (Huang and Stultz, 2007) and Aβ fibril (Gurry and Stultz, 2014). The former study indicates that implicit solvent (generalized born model) is in good agreement with TIP3P explicit solvent model in Ramachandran plot and free-energy profile. In addition, the results obtained from implicit solvent model are consistent with the experimental observation of the structural change. The later study provides the elongation of Aβ fibril induced by Aβ peptide forming β-hairpin structure and stacking with fibril predicted based on the implicit model. This prediction is consistent with the results of explicit solvent model and experimental observation.

In the pathogenesis of prion diseases, the accumulation of oligomers or fibrils induces production of ROS, resulting in an inflammatory reaction with gliosis and neuron loss (Aguzzi and Zhu, 2017). Many studies have proposed various mechanisms of prion diseases, and it is widely believed that microglia activation is the early stage of prion diseases before neuron loss and spongiform biopsy, stepwise (Aguzzi and Zhu, 2017). It was reported that PrP_106–126 fibril activates microglia secreting pro-inflammatory cytokines (i.e.: TNF-α, IL-1β, IL-6) and upregulates NO synthase, which produces NO to induce neuronal cell death in BV2-primary microglia coculture (Kouadir et al., 2012). To decrease the neuronal cell death induced by microglia, the polyphenols are excellent in the elimination of ROS or RNS, resulting in a decrease in the cytokine secretion from microglia cells. The polyphenols from magnolia have been found to inhibit ROS and RNS production, and then to attenuate oxidative stress and inflammatory responses in neurons and microglia (Chuang et al., 2013). In PrP fibril treated Neuro-2a cells, the clearance of ROS and significantly increased cell viability were observed after quercetin treatment (Yu and Lee, 2020).

Previous studies have indicated that prion protein is stabilized by Glu196, forming intramolecular salt bridges with Arg156, His187 (Lee and Chang, 2019) or Lys194 (Sengupta et al., 2017). Based on the cryo-EM results (Wang et al., 2020), the intermolecular interaction between Lys194 and Glu196 is critical for fibril stability. The transition of Lys194 and Glu196 from intramolecular to intermolecular interaction is important to the PrP^C-PrP^Sc conversion. In our MD simulation, quercetin accelerates the low toxic and PK-sensitive fibril formation by weakening the fibril structure and interacting with PrP protein at Glu196 (Yu et al., 2022). The same binding residue consists of the MD result from the quercetin-fPrP_170–229 complex in this work. An early study has shown that 2-pyrrolidin-1-yl-N-[4-[4-(2-pyrrolidin-1-yl-acetylamino)-benzyl]-phenyl]-acetamide (GN8), a diphenylmethane derivative, interacts with Glu196, resulting in an effectively prolonged lifespan of PrP^Sc infected mice (Kuwata et al., 2007). Furthermore, E196K mutant induces the structural change extending the binding surface to a stable fibril structure (Wang et al., 2021). This evidence indicates that Glu196 is critical for PrP stability and drug-targeting. In this work, we have found some quercetin-like molecules interacting with the critical residues, including Lys194 and Glu196, to reduce β-strand content.

The PK-resistant fragments are located in the β-sheet region (Vázquez-Fernández et al., 2012). Our MD results indicate that some quercetin-like compounds have strong binding to the C-terminal region of fPrP_170–229 and this binding weakens the β-strand. In the nucleation-dependent protein polymerization theory, the nuclei are critical to the ongoing conversion. The nuclear morphology is dependent on the strains of PrP^Sc. Therefore, the structure of PrP fibrils is critical in PrP^C-PrP^Sc conversion (Bessen et al., 1995). Combining the MD results from this work and critical morphology role of PrP^Sc, we demonstrate that the change of fibril secondary structure affects the efficiency of fibril polymerization (Figure 8). Therefore, COMP-binding induced structural change is helpful in relieving the prion transmission.

Previously, many structure-based anti-prion drug developments were performed based on the structure of prion protein monomer. In this study, our target is prion fibrils rather than protein monomers. Thoughtfully considering the clinical potential of the searched molecules, we performed structure-based virtual screening, Lipinski’s rule of five, toxicity prediction, molecular docking and molecular dynamics to find potential anti-prion fibril flavonoids. Quercetin is well known to disaggregate many kinds of amyloid fibrils, including prion fibrils. Therefore, the effect of searched anti-prion fibril flavonoids is compared with quercetin. The selected five COMPs with pharmacological activity are eligible for Lipinski’s rule of five. Their binding energies to fPrP_170–229 are stronger than quercetin. The major binding site of COMPs with fPrP_170–229 is on the contacting surface of fPrP_170–229 for dimerization at Lys194 and Glu196. Most COMPs form strong interactions with fPrP_170–229 by van der Waals force. COMP 2A, 3A, 3B and 5A have strong electrical interactions with fPrP_170–229 contributed by strong or moderate hydrogen bonds. COMPs 3, 4, and 5 induce fPrP_170–229 to decrease the β-strand content significantly and to form helices or loops. The COMP 1 induces fPrP_170–229 to form helices. These COMPs have great potential for further in vitro and in vivo evaluation. Overall, this work takes a big stride in the treatment for prion diseases.