A Multilevel Study of Eupatorin and Scutellarein as Anti-Amyloid Agents in Alzheimer’s Disease

Today, Alzheimer’s disease (AD)—the most common neurodegenerative disorder, which affects 50 million people—remains incurable. Several studies suggest that one of the main pathological hallmarks of AD is the accumulation of abnormal amyloid beta (Aβ) aggregates; therefore, many therapeutic approaches focus on anti-Aβ aggregation inhibitors. Taking into consideration that plant-derived secondary metabolites seem to have neuroprotective effects, we attempted to assess the effects of two flavones—eupatorin and scutellarein—on the amyloidogenesis of Aβ peptides. Biophysical experimental methods were employed to inspect the aggregation process of Aβ after its incubation with each natural product, while we monitored their interactions with the oligomerized Aβ through molecular dynamics simulations. More importantly, we validated our in vitro and in silico results in a multicellular organismal model—namely, Caenorhabditis elegans—and we concluded that eupatorin is indeed able to delay the amyloidogenesis of Aβ peptides in a concentration-dependent manner. Finally, we propose that further investigation could lead to the exploitation of eupatorin or its analogues as potential drug candidates.


Introduction
Over the past decade, Alzheimer's disease (AD), along with dementia, has been classified as the seventh leading cause of death, according to the World Health Organization [1]. AD is characterized as an age-related neurodegenerative disorder with detrimental consequences for memory and cognitive abilities that eventually leads to death. It is widely accepted that one of the pathological hallmarks of the disease at a cellular level is the presence of senile plaques in the brain, which mainly consist of amyloid fibrils [2]. The main component of these extracellular deposits is the amyloid beta (Aβ) peptide, which is derived from the proteolytic degradation of amyloid precursor protein (APP). APP is a transmembrane protein that typically participates in synaptogenesis, neurite growth, and neuronal adhesion [3]. The predominant amyloid cascade hypothesis states that proteolytic enzymes called βand γ-secretases process APP and release the neurotoxic Aβ peptide. Aβ peptides function as monomers, aggregate into oligomers and, finally, self-assemble insulin fibril formation. In contrast, eupatorin is known for its anti-inflammatory action, but it has never been tested against Aβ 42 amyloidogenesis or AD [19]. These properties could possibly make them the ideal candidates for a multiple-target strategy.
In the current study, we conducted a comprehensive investigation using a multilevel approach encompassing in vitro, in silico, and in vivo methods to evaluate the ability of the two flavones-scutellarein and eupatorin-to inhibit or delay the Aβ 42 aggregation. While scutellarein has shown promising results in previous studies, we aimed at uncovering the interactions between the compound and the toxic oligomeric structure of Aβ 42 . Considering the structural similarity of eupatorin to scutellarein and the fact that it has not been studied before, we investigated the anti-amyloid profile of the former in vitro, in vivo, and in silico. First, we predicted their drug-likeness in silico, using SwissADME [20], and then we utilized biophysical experimental methods to study the effects of these natural products on Aβ 42 fibril formation. The compounds were subjected to molecular docking studies in conjunction with molecular dynamics (MD) simulations with an oligomeric structure of Aβ 42 to understand their interactions. Specific strains of Caenorhabditis elegans that overexpress the human Aβ were used to confirm the in vitro results. C. elegans has several advantages as a model organism, including its high degree of conservation with humans; thus, it is a valuable tool for studying neurodegenerative disorders at a complex, multicellular level [21]. This study demonstrates that eupatorin and scutellarein could be further examined as potential drug candidates against AD.

Peptide Synthesis and Disaggregation
Aβ 42 peptide (DAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVVIA) was produced and lyophilized by GeneCust© (Boynes, France), and the purity exceeded 95%, with the N-and C-termini being free. The peptide was dissolved in 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP) (Sigma©, St. Louis, MO, USA) at a concentration of 1 mg/mL to disaggregate any secondary structures formed during synthesis. The peptide solution was evaporated overnight at room temperature until a peptide-containing film was formed at the bottom of the Eppendorf tube. The peptide-containing films were stored at −20 • C. Just prior to use, each peptide-containing film was left at room temperature for 30 min.

Sample Preparation
Considering relevant studies, as well as the maximum solubility of each compound in DMSO, three different concentrations of the natural products were tested [22]. The Aβ 42peptide-containing films were mixed separately with the solutions of the natural products at three different molar ratios-1 Aβ 42 :1 natural product, 1 Aβ 42 :5 natural product and 1 Aβ 42 :10 natural product. Individual Aβ 42 solution was used as a control for all in vitro experimental assays. Depending on the experimental assay performed, the final peptide concentration was 10 µM or 40 µM. The final concentration of DMSO from the solutions of the natural products did not exceed 3%, so as to avoid altering the Aβ 42 aggregation kinetics [23]. The samples were placed in a 35 • C ultrasound water bath for 60 s; 4-(2hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (Sigma©, St. Louis, MO, USA) at a 0.1 M concentration and pH 7.5 was added to the final volume, and the samples were agitated at 37 • C for two hours. Finally, we incubated the natural product mixtures and the control at 37 • C for at least seven days.

Negative Staining and Transmission Electron Microscopy
The amyloid fibril formation of the Aβ 42 peptide in the presence and absence of eupatorin and scutellarein was studied using transmission electron microscopy (TEM). A 10 µL droplet of each 40 µM sample was placed on 400-mesh glow-discharged and carbon-coated copper TEM grids for 20 min. The grids were placed on a droplet of 2% (w/v) aqueous uranyl acetate for 50 s and then washed successively with three distilled water droplets to remove the extra stain. The excess water was removed with Whatman filter paper, and the grids were left to air-dry for a few seconds. The samples were inspected using a Philips Morgagni TM 268 Transmission Electron Microscope (FEI, Hillsboro, OR, USA) operated at 80 kV. Digital acquisitions were performed with an 11 megapixel sidemounted Morada CCD camera (Soft Imaging System, Muenster, Germany). The analysis of the electron micrographs was performed by ImageJ (National Institutes of Health (NIH), Bethesda, MD, USA) [24].

Congo Red Staining and Polarized Microscopy
A droplet from each 40 µM sample was placed on a glass slide and air-dried at room temperature. The sample was stained using 0.01 M Congo red (Sigma©, St. Louis, MO, USA) solution in PBS (137 mmol/L NaCl, 27 mmol/L KCl, 100 mmol/L Na 2 HPO 4 , 18 mmol/L KH 2 PO 4 , pH = 7.4) for approximately 20 min and washed several times with 90% ethanol. Then, the samples were left to air-dry for approximately 10 min and observed under bright-field illumination and between crossed polars, using a Leica MZ7.5 polarizing stereomicroscope (Leica Camera AG, Weltzar, Germany) equipped with a Sony a6000 camera (Sony, Tokyo, Japan). In addition, the natural products were stained separately to ensure that there was no birefringence.

Thioflavin T (ThT) Kinetic Assay
ThT fluorescence was measured at 37 • C in black 96-well plates with flat, clear bottoms, using a Tecan Spark microplate reader (Mannedorf, Switzerland). At first, Aβ 42 -peptidecontaining films were dissolved in DMSO, and HEPES was added to adjust the final volume to 1000 µL. Each well contained 10 µM freshly prepared Aβ 42 mixed with solutions of eupatorin or scutellarein at the three molar ratios (1:1, 1:5, 1:10), along with 25 µM ThT (Sigma©, St. Louis, MO, USA). The tops of the plates were sealed with microplate covers, and the fluorescence readings were performed through the bottom. A 444 nm filter was used for excitation, and a 484 nm filter was used for emission. For background measurements, ThT was diluted in HEPES. Each experiment was repeated three times, and the measurement lasted 40 h. ThT fluorescence was collected after 10 s of agitation at 270 rpm every 15 min. When using ThT concentrations higher than 5 µM, it is critical to take into consideration the background ThT fluorescence due to self-fluorescence [25]. Therefore, for the data analysis, ThT background fluorescence was subtracted from the sample readings at each time point, data were normalized on a scale of 0 to 100 arbitrary units, and the standard deviation for each point was calculated. Error bars in ThT fluorescence emission spectra represent the standard deviation of the triplicates. Data visualization was performed using the R statistical language [26] and the packages ggplot2, dplyr, ggthemes, extrafont, and ggpmisc via the integrated development environment RStudio [27].

In Vivo Assays in C. elegans
C. elegans is a small, free-living, and bacteria-eating soil nematode that typically lives for about 3 weeks at 20 • C in the laboratory. It is a nonhazardous and nonpathogenic animal that can be manipulated with standard safety rules. Several transgenic animals have been already produced and are widely used as models for AD. These animals overexpress the human Aβ peptide in their body-wall muscle cells, where the peptide gradually oligomerizes and aggregates; as a result, severe and fully penetrant age-progressive paralysis occurs. In our study, we utilized two commonly used C. elegans strains as models for AD, namely, CL2331 and CL4176. The former expresses the human Aβ 3-42 peptide fused to green fluo-rescent protein (GFP) in its body-wall muscle cells; the animals are gradually filled with Aβ aggregates that are visible through confocal microscopy [28]. The latter strain expresses the human Aβ 42 peptide in its body-wall muscle cells in a temperature-sensitive manner. When the temperature is upshifted, expression of the Aβ peptide is induced, resulting in paralysis of the animals within a few hours [29].

Phenotypic Characterization
For all assays, N2 animals laid eggs for 20-30 min on nematode growth medium (NGM) plates containing either 10 µg/mL eupatorin or DMSO (control). The following phenotypic characteristics were evaluated according to standard procedures, as previously described [30][31][32]: Pharyngeal pumping: At day 1 of adulthood, the number of pharyngeal pumps per minute was measured. Thirty-five animals per condition were scored.
Defecation assay: At day 1 of adulthood, the period in seconds from defecation to defecation (defecation cycle) was measured. Thirty-one animals per condition were scored.
Developmental timing: The duration of postembryonic development in hours from egg hatching to the L4 stage was recorded through frequent observation of the progeny. The experiment was repeated three times.
Fecundity assay: Single N2 L4 larvae were transferred onto NGM plates containing either 10 µg/mL eupatorin or DMSO. Each animal was transferred every two days to a fresh NGM plate containing fresh compound or DMSO. The progeny of each animal was scored at the L2-L3 larval stage. At least 8-10 animals per condition were scored.
Dauer formation: The progeny was maintained at 27 • C, and the number of animals at the dauer larval stage over the total number of animals was scored 72 h later. The experiment was repeated three times.

Paralysis Assay
CL4176 animals were synchronized on NGM plates with either eupatorin (concentrations: 1, 5, 10, and 20 µg/mL) or DMSO at 16 • C for 48 h before upshifting of the temperature to 25 • C. The paralyzed animals were scored 24 h after the upshift until the paralysis of all animals. The paralysis assay was repeated three times. The criteria to score nematodes as paralyzed were (1) the presence of halos of cleared bacteria around their heads, and (2) failure to undergo half-end body wave propagation upon prodding. Animals that died during the experiment were excluded, as described in our previous work [31,33]. The log-rank (Mantel-Cox) test was used to evaluate differences between paralysis curves and to determine p-values for all independent data. N in the paralysis figures represents the number of paralyzed animals. Median paralysis values are expressed as the mean ± SEM.

Confocal Analysis of Aβ Deposition
For scoring of the Aβ 3-42 aggregates, synchronized CL2331 animals were exposed to 10 µg/mL eupatorin or DMSO and grown at 20 • C (to induce the expression of the Aβ 3-42 peptide) until the L4 larval stage. Animals were mounted on 2% agarose pads on glass slides, anesthetized with 100 mM levamisole, and observed at room temperature using a Leica TSC SPE confocal laser scanning microscope (Leica Lasertechnik GmbH, Heidelberg, Germany). The LAS AF software (Leica Lasertechnik GmbH, Heidelberg, Germany) was used for image acquisition. Images focused in the anterior area of the nematodes were acquired with a 20x/0.70 objective. Quantification of Aβ 3-42 ::GFP was conducted using ImageJ [24]. The aggregates that were quantified were the ones contained within a region of 4 square units of the surface of the animal's body, starting from the mouth. The spots that were larger than 5 pixels were considered as aggregates. The quantification was performed automatically with the use of the function "analyze particles". The background of each region of interest was subtracted by 5 pixels, and the threshold was set automatically using the Otsu method. Thirty animals per condition were measured.

Statistical Analysis
The parametric two-tailed Student's t-test was used for the comparison of the means of two groups. The log-rank (Mantel-Cox) test was used to evaluate differences between paralysis curves and to determine p-values for all independent data. N in the paralysis figures represents the number of paralyzed animals. The data from all assays are depicted as the average of three independent experiments (unless otherwise indicated). The median paralysis values are expressed as the mean ± SEM (shown by error bars). Asterisks denote p-values as follows: ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns = not significant. Statistical analyses and graphs were produced using the GraphPad Prism 8 (GraphPad Software, Inc., San Diego, CA, USA) and Microsoft Office 365 Excel (Microsoft Corporation, Redmond, WA, USA) software packages. In order to assess the binding affinity and the interaction between the oligomerized Aβ 42 peptide and the flavones, molecular docking experiments were carried out. Particularly, protein-ligand docking was performed utilizing the publicly available software AutoDock Vina (Scripps Research, San Diego, CA, USA) [34], via the graphical user interface AutoDock Tools (ADT) [35]. Initially, the NMR-derived oligomeric structure of the Aβ 42 peptide (including residues 17-42) was retrieved from the Protein Data Bank (PDB ID: 2BEG) [36] and imported into the molecular graphics system PyMOL [37] to segregate the first model and to remove the hydrogens of the experimentally determined structure. Afterwards, with the use of AutoDock Tools, polar hydrogens and Kollman charges were added, and the AD4 atom type was assigned to the peptide, generating the structure file in pdbqt format. The 3D structures of the ligands scutellarein and eupatorin were retrieved from PubChem [38], and their respective PubChem CIDs are 5281697 and 97214. The ligands' structures were further processed with the use of the UCSF Chimera software [39] and the graphical interface Raccoon [40], in which Gasteiger partial charges were added and polar hydrogens were merged, resulting in the creation of the structure file for each ligand-including its active torsions-in pdbqt format. Subsequently, the proper coordinates of a rectangular parallelogram GridBox were appointed to include the whole oligomer of Aβ 42 , so that the ligands could freely bind to the most favorable region. The generated output files contained information regarding the coordinates for the first 10 conformations for each protein-ligand complex, ranked by the lowest binding affinity. Out of the 10 conformations of the complexes, the one with the lowest binding affinity was considered the most energetically favorable. Therefore, the subsequent in silico experiments were conducted on these complexes as starting points.

Molecular Dynamics Simulations
The two docked protein-ligand complexes were subjected to molecular dynamics simulations with the use of the GROMACS software package, version 2018.1 (University of Groningen, Groningen, Netherlands) [41,42]. The topology for the Aβ 42 peptide was generated with the all-atom additive CHARMM36 protein force field [43], which is considered to be the most suitable for protein-ligand MD simulations, whereas the ligands' topologies were generated with the CHARMM General Force Field (CGenFF) [44]-a "drug-like" force field with the relative parameters for small compounds. The water model employed for the protein-ligand complex was TIP3P (transferable intermolecular with three points) [45]-a 3-point water model-and the systems were placed in a cubic unit cell of the same periodic distance, which was defined at 1.2 nm. In addition, the complexes were solvated with the simple point charge (SPC) [46] water model, and 5 Na+ ions were added. The solvated, neutralized systems were subjected to energy minimization runs with the steepest descent algorithm, until the maximum force reached values below 10.0 kJ/mol. Two phases of equilibration took place with the application of position restraints to the non-hydrogen atoms of the protein-ligand complex. The first phase was conducted under a constant-volume (NVT) ensemble for 100 ps, and the temperature was maintained at 310 K, with the employment of a V-rescale thermostat [47]-a modified Berendsen thermostat, suitable for protein-nonprotein systems. The temperature was set at 310 K-just as in the in vitro experiments-to resemble the normal human body temperature. The second phase of equilibration was performed under a constant-pressure (NPT) ensemble for 100 ps, and the pressure was isotropically maintained at 1.013 bar (1 atm) with the use of a Berendsen barostat [48]. The equilibrated protein-ligand complexes were then set to be simulated for 500 ns at 310 K, in the absence of position restraints, while the Parrinello-Rahman barostat [49,50] was employed. Simulations were conducted with the use of the leapfrog algorithm for integration, and a 2 fs timestep was used. Long-range electrostatic interactions were modeled with the particle mesh Ewald (PME) method [51,52], while short-range non-bonded interactions were cut off at 1.2 nm. The linear constraint solver (LINCS) algorithm [53] was used for the constriction of hydrogen bonds in the entirety of the simulations, and periodic boundary conditions (PBCs) were applied in all directions.
The outputs of the simulations were analyzed with tools that are included in the GROMACS software package (University of Groningen, Groningen, Netherlands). In order to assess the potential structural changes of the oligomerized Aβ 42 peptide, the root-mean-square deviation and the root-mean-square fluctuation were calculated by the GROMACS modules rms [54] and rmsf [55], respectively. Moreover, with the integration of the DSSP algorithm [56], graphs were generated, depicting the variation in the number of peptide residues assigned to certain secondary structural elements during the course of the simulation. Snapshots of the two protein-ligand complexes were created with the molecular visualization system UCSF Chimera, and the acquired graphs were plotted using the R statistical language [26] via RStudio (Posit, Boston, MA, USA) [27] and the packages Peptides and MDPlot.

Computational Prediction of the Physicochemical Properties of the Natural Products
Initially, we conducted preliminary in silico research on the physicochemical properties of eupatorin and scutellarein. It is widely accepted that the early evaluation of ADMET (absorption, distribution, metabolism, excretion, and toxicity) parameters can provide useful guidelines for the preclinical stages of drug design and decrease pharmacokinetics-related clinical failures [57]. More precisely, physicochemical properties, lipophilicity, pharmacokinetics, drug-likeness, lead-likeness, and synthetic accessibility were computationally estimated using the SwissADME Web tool by the Swiss Institute of Bioinformatics (SIB) (Table S1, Supplementary File) [20]. First of all, it was predicted that eupatorin and scutellarein could be potential drug candidates, as both are characterized by drug-likeness and lead-likeness. This signifies that these plant-derived compounds have the structural and physicochemical properties of a drug and are suitable for further optimization with low synthetic difficulty. In detail, they were predicted to exhibit good aqueous solubility and high gastrointestinal absorption, which would help with oral administration. In addition, neither of the two is a substrate of P-glycoprotein-an active efflux transporter of biological membranes [58]. For these reasons, we decided to assess the inhibitory effects of these natural products on the aggregation of Aβ 42 in vitro.

The Effects of the Two Natural Products on the Amyloidogenicity of Aβ 42
In order to examine whether eupatorin and scutellarein could inhibit or delay the amyloidogenesis of Aβ 42 , we initially incubated the peptide individually as a control. Detailed processing of electron micrographs of the control sample showed that after seven days, Aβ 42 self-assembled into mature, unbranched fibrils of undefined length with a diameter of 8.9 nm (±1.68), as shown in Figures 1a and 2a. Indeed, the ThT fluorescence measurements confirmed that amyloid-like fibrils had formed at approximately eight hours (Figures 1b and 2b). It is worth noting that the lag phase was approximately an hour and a half, but we have also observed this before in one of our previous works [59]. Moreover, gel containing amyloid-like fibrils was stained with Congo red dye and observed under crossed polars of a polarized microscope. Congo red staining and polarized microscopy are among the first criteria to determine the amyloid properties of a protein in vitro or in histochemical studies, along with TEM. In the bright field, we observed that the sample was stained red, indicating that Congo red specifically binds to the sample, while under polarized light we observed the characteristic apple-green birefringence that amyloids typically exhibit, confirming the amyloid nature of the sample, as shown in Figure S2 in the Supplementary File. Of course, the Congo red staining is not as indicative as TEM for the identification of amyloid-like fibrils, but it is an amyloid-specific dye like thioflavin T, owing to its orientation between the beta-strands of amyloid fibers [60]. When incubated with scutellarein, the number of fibrils decreased as the concentr tion of the compound increased. It is worth mentioning that the dominant morpholog was not a network of amyloid-like fibrils, but amorphous aggregates at all three conce trations (red arrows in Figure 2c,e,g). In addition, the diameter of the fibrils was found be 8.2 nm (±0.8) at the 1:1 ratio, 7.1 nm (±0.2) at the 1:5 ratio, and 7.8 nm (±1.1) at the 1: ratio (Figure 2c,d,g, white arrows). The TEM results were consistent with the ThT kinet assays shown in Figure 2d,f,h, where a decrease in the maximal fluorescence signal b 34.2%, 12.09%, and 43.2% was measured at ratios of 1:1, 1:5, and 1:10, respectively. Th maximum signal was measured after 16 h-twice as long compared to Aβ42. Furthermor we observed a longer lag phase and a slower exponential phase in contrast to the contro Finally, the amorphous aggregates and the reduced number of amyloid-like fibrils cou explain the absence of apple-green birefringence when droplets from the co-incubatio samples were stained with Congo red and observed under crossed poles of the polarizin microscope ( Figure S3, Supplementary File). In the presence of eupatorin, Aβ 42 aggregation was significantly reduced according to the TEM observations. To be more specific, the best results were recorded at the ratios of 1:5 and 1:10. The grids were barely empty, hardly any fibrils were observed, and the dominant morphology was amorphous and prefibrillar aggregates (Figure 1c,g). These results are consistent with the absence of apple-green birefringence after Congo red staining ( Figure S3, Supplementary File). Furthermore, the effect of eupatorin was evident in the ThT kinetic assays (Figure 1d,f,h). At the 1:5 ratio, a sharp reduction in ThT fluorescence was already recorded in the first five hours (Figure 1d). Concerning the ratio of 1:1, the fibrils' morphology appeared no different than that of Aβ 42 alone. In Figure 1e, it is apparent that the peptide aggregated into amyloid-like fibrils and amorphous aggregates (Figure 1e, white and red arrows, respectively). More specifically, we observed elongated and unbranched fibrils of indefinite length that formed dense networks, and their diameter ranged from 8 to 9 nm (Figure 1e, white arrow). Lastly, only in this concentration did we record a light yellow-green birefringence, as expected based on the number of amyloid-like fibrils formed ( Figure S3, Supplementary File). Given these results, we concluded that eupatorin caused a dose-dependent reduction in Aβ 42 aggregation. 34.2%, 12.09%, and 43.2% was measured at ratios of 1:1, 1:5, and 1:10, respectively. The maximum signal was measured after 16 h-twice as long compared to Aβ42. Furthermore, we observed a longer lag phase and a slower exponential phase in contrast to the control. Finally, the amorphous aggregates and the reduced number of amyloid-like fibrils could explain the absence of apple-green birefringence when droplets from the co-incubation samples were stained with Congo red and observed under crossed poles of the polarizing microscope ( Figure S3, Supplementary File). When incubated with scutellarein, the number of fibrils decreased as the concentration of the compound increased. It is worth mentioning that the dominant morphology was not a network of amyloid-like fibrils, but amorphous aggregates at all three concentrations (red arrows in Figure 2c,e,g). In addition, the diameter of the fibrils was found to be 8.2 nm (±0.8) at the 1:1 ratio, 7.1 nm (±0.2) at the 1:5 ratio, and 7.8 nm (±1.1) at the 1:10 ratio (Figure 2c,d,g, white arrows). The TEM results were consistent with the ThT kinetic assays shown in Figure 2d,f,h, where a decrease in the maximal fluorescence signal by 34.2%, 12.09%, and 43.2% was measured at ratios of 1:1, 1:5, and 1:10, respectively. The maximum signal was measured after 16 h-twice as long compared to Aβ 42 . Furthermore, we observed a longer lag phase and a slower exponential phase in contrast to the control. Finally, the amorphous aggregates and the reduced number of amyloid-like fibrils could explain the absence of apple-green birefringence when droplets from the co-incubation samples were stained with Congo red and observed under crossed poles of the polarizing microscope ( Figure S3, Supplementary File).

The Impact of Each Natural Compound on the Tertiary Structural Stability of the Aβ 42 Oligomer
The results of molecular docking suggested that both compounds show good binding affinity, forming thermodynamically favorable complexes with the oligomer of Aβ 42 peptide. The lowest binding affinities for the peptide in complex with scutellarein and eupatorin exhibited values of −6.8 and −6.2 kcal/mol, respectively. These complexes were subsequently used for the conduction of molecular dynamics simulations. Among the simulated data analyzed, those considered to be of utmost importance were the number of residues participating in certain secondary structural elements of the Aβ 42 peptideprimarily beta-sheets and helices-and the possible change of the Aβ 42 peptide conformation by assessing the root-mean-square deviation (RMSD) and the root-mean-square fluctuation (RMSF) of the Aβ 42 peptide chains that form its oligomeric state.
As far as the secondary structure is concerned, all compounds caused a decrease to some extent in the number of amino acid residues of the Aβ 42 peptide involved in betastrands-which is inextricably linked to its pathogenic character-and, albeit temporarily, the formation of helices. Scutellarein triggered a steep decrease in the number of betastrands, with the number of residues in strands reduced by 23% when comparing the average numbers between the first and the last 5 ns of the simulation. Eupatorin followed in efficacy, with a stable decreasing tendency (12% reduction in the number of residues in beta-strands). Subsequently, the oligomer's conformational changes after ligand binding and the course of the simulation were evaluated by calculating the RMSD. Higher RMSD values indicate that the interaction between the compounds and the Aβ 42 oligomer leads to a less stable structure, proving their potency. As shown in Figure 3, scutellarein and eupatorin caused mild conformational changes, with maximum values around 1 nm.

The Effect of Eupatorin on C. elegans
Since our in vitro and in silico results confirmed that both flavones are able to prevent the fibrillogenesis of Aβ42, we proceeded with in vivo experiments on specific strains of C. elegans that are commonly used as animal models for Aβ accumulation and AD [28]. Since the anti-amyloid properties of scutellarein have already been investigated in vivo . Both compounds caused a decrease in the number of amino acids involved in betastrands. Simulation frames of the Aβ 42 oligomer at 0 ns and 500 ns. The Aβ 42 peptide is colored grey, and the inhibitors are light green. The first plot for each natural compound corresponds to the number of residues assigned to the B-sheet (green), A-helix (blue), π-helix (red), and 3-helix (grey) during the simulation, according to the DSSP calculations. The second plot depicts the root-mean-square deviation (RMSD) of the Aβ 42 oligomer during the course of the simulation. The root-mean-square fluctuation (RMSF) per residue for the five chains of the Aβ 42 oligomer is depicted in the third plot (RMSF: root-mean-square fluctuation, RMSD: root-mean-square deviation). RMSF analysis was employed to assess the flexibility of the peptide structure per amino acid residue and showed that ligand binding induces conformational changes to a greater extent in the C-termini of the peptide chains that form the oligomer. The complex with eupatorin exhibited less fluctuation along the entire length of all chains as compared to the other ligand, without exceeding 0.4 nm (Figure 3b). As far as the Aβ 42 peptidescutellarein complex is concerned, the chains that underwent substantial changes in their structural integrity-exhibiting higher mobility and being more flexible at the end of the molecular dynamic simulation-were D and E. The N-termini of these two chains showed greater fluctuation compared to the C-termini, and the effect of scutellarein was evident for all of the residues of chains D and E. However, chains A, B, and C were characterized by lower RMSF values, meaning that they were less flexible and were not extremely affected by their interaction with scutellarein (Figure 3a).

The Effect of Eupatorin on C. elegans
Since our in vitro and in silico results confirmed that both flavones are able to prevent the fibrillogenesis of Aβ 42 , we proceeded with in vivo experiments on specific strains of C. elegans that are commonly used as animal models for Aβ accumulation and AD [28]. Since the anti-amyloid properties of scutellarein have already been investigated in vivo by Gea-Gonzales et al. [61], we conducted in vivo assays with C. elegans only for eupatorin, which had never been tested before.
To evaluate the potential of eupatorin to confer protection against Aβ toxicity, we took advantage of a transgenic C. elegans strain (CL4176) that expresses the human Aβ 42 peptide in its body-wall muscle cells in a temperature-sensitive manner. When the temperature is upshifted, expression of Aβ peptide is induced, resulting in paralysis of the animals within a few hours [29]. We tested various concentrations of eupatorin ranging from 1 to 20 µg/mL; 10 µg/mL was the concentration that produced positive results (Figure 4a). Animals treated with eupatorin were paralyzed significantly later as compared to the control animals.
We sought to investigate whether this decelerated paralysis was due to lower levels of Aβ aggregates. We therefore took advantage of the CL2331 strain. Treatment with eupatorin resulted in a reduced number of Aβ aggregates (Figure 4b). In total, our results revealed a protective effect of eupatorin against Aβ toxicity.  3 Number of eggs that did not hatch. 4 Pumps in 1 min at day 1 of adulthood. 5 Duration of defecation cycle in seconds at day 1 of adulthood. 6 Percentage of animals becoming dauer larvae at 27 • C. Error bars denote the mean ± SEM; p denotes the p-value of Student's t-test, **** p < 0.0001.
Animals treated with eupatorin were paralyzed significantly later as compared to t control animals. We sought to investigate whether this decelerated paralysis was due to lower lev of Aβ aggregates. We therefore took advantage of the CL2331 strain. Treatment w eupatorin resulted in a reduced number of Aβ aggregates (Figure 4b). In total, our resu revealed a protective effect of eupatorin against Aβ toxicity.  C. elegans is an increasingly used model for toxicity testing, as it has various phenotypic characteristics that change upon toxic exposure [62]. We therefore evaluated the following phenotypic characteristics in wild-type nematodes in the presence of 10 µg/mL eupatorin and in control animals (treated with DMSO): developmental time, fertility, egg lethality, pharyngeal pumping, defecation rate, and dauer formation (Table 1). Phenotypic characteristics were found to be unaltered, with the exception of pharyngeal pumping, which was found to be significantly increased upon treatment with eupatorin. This increase was considered to be a positive outcome, since it has been associated with enhanced organismal healthspan [30,63,64]. In total, our results advocate for low eupatorin-dependent toxicity.

Discussion
In this study, we attempted for the first time to elucidate the mechanism of action of two rare flavones-namely, eupatorin and scutellarein-on the aggregation of Aβ 42 . Since many natural products today are tested as drugs, we posed a question about the drug-likeness of our compounds. Having established that, we conducted dose-dependent experiments to investigate whether these compounds could inhibit or at least delay the aggregation of the Aβ 42 peptide in vitro.
To begin with, the in silico evaluation of the structural and pharmacokinetic properties of the compounds led to the conclusion that they satisfy most of the criteria of drug candidates [65]. Regarding the in vitro results, both compounds were effective, since they prevented Aβ 42 from self-assembling into amyloid-like fibrils. This observation was verified by a decrease in the ThT fluorescence intensity at the 1:5 ratio in the presence of eupatorin, and at all ratios in the presence of scutellarein. Moreover, these results were confirmed by the absence of apple-green birefringence after Congo red staining. Consequently, we suggest that eupatorin and scutellarein are able to disrupt the aggregation of Aβ 42 .
Previous studies have demonstrated the anti-inflammatory, antioxidant, neuroprotective, and metal-chelating properties of scutellarein, which has been used as a scaffold for the development of multifunctional ligands for the treatment of Alzheimer's disease [66][67][68]. Recent studies confirm our in vitro results and show that scutellarein is able to extend the lifespan of the CL2355 strain (a strain with pan-neuronal expression of the human Aβ peptide [69]), while the number of Aβ 42 aggregates was found to be significantly reduced in the CL2331 strain [61]. For this reason, we focused our in vivo experiments on eupatorin, which had never been evaluated before. Concerning the effect of this compound as anti-amyloid agent in the two AD nematode models, our data indicate that this flavone can significantly reduce the number of Aβ aggregates and decelerate the paralysis that occurs upon accumulation of human Aβ peptide.
Our experimental results seemed to be consistent with those of molecular dynamic simulations, as both compounds caused noticeable alterations to the conformation of the Aβ 42 oligomer when comparing the 0 ns frames to the 500 ns frames. In particular, eupatorin contributed to the decrease in the number of residues in beta-strands, while a significant increase in the number of residues in helices was observed. Additionally, what stood out in the RMSF analysis was the flexibility of the hydrophilic N-terminus and the hydrophobic C-terminus at the end of the simulation. Prior studies have already noted the importance of the C-terminus in Aβ stability, as it is said to initiate the conformational change from α-helix to β-sheet and promote nucleation [70,71].
Our results, in addition to those of previous studies on other complementary biological properties of scutellarein and eupatorin, show that both are good candidates for the development of multifunctional Alzheimer's-disease-modifying agents. These are flavone aglycones and are seldom found in that form in natural sources; they are part of glycosides in plants, e.g., scutellarin. When scutellarin is ingested by humans, it is hydrolyzed into aglycones in the colon and is then absorbed as scutellarein, so scutellarein might be the real bioactive component in the body. Oral administration of scutellarein to rats showed that scutellarein had better neuroprotective effects than scutellarin, and it could attenuate neuronal injury by ischemia/reperfusion [72] and focal cerebral occlusion/reperfusion [73].
Concerning the blood-brain barrier distribution, several flavones have been found to be able to pass through it, such as genistein, isoliquiritigenin, and kaempherol [74,75]. In the future, we will aim to assess the BBB permeability of scutellarein and eupatorin by utilizing CaCo-2 and BBB cell models that have already been used by other groups to explore flavones' properties. In addition, further optimization could possibly enhance the pharmacokinetic properties of the compounds.

Conclusions
In this integrated study, we assessed the inhibitory effects of scutellarein and eupatorin on the aggregation of the Aβ 42 peptide in vitro, in silico, and in vivo. According to the experimental results and the molecular dynamics simulations, eupatorin exhibited an encouraging inhibitory effect on Aβ 42 amyloidogenesis. In the future, we suggest that this 5,3 -dihydroxy-6,7,4 -trimethoxyflavone or other optimized compounds with similar structures should be further tested and optimized as potential drug candidates against AD.