Investigating the Potency of Erythrina‒Derived Flavonoids as Cholinesterase Inhibitors and Free Radical Scavengers Through in silico Approach: Implications for Alzheimer’s Disease Therapy

Introduction

Alzheimer’s disease (AD) poses a significant global health challenge, affecting millions of people worldwide, with its prevalence increasing as age advances.¹ Several studies have shown that therapeutic interventions play a crucial role in addressing the cognitive decline associated with AD. During the treatment process, cholinesterase (ChE), including acetylcholinesterase (AChE) and butyrylcholinesterase (BuChE), serves as key molecular targets.² These enzymes are responsible for acetylcholine hydrolysis, which is essential for memory and cognitive functions. In addition, inhibition of ChE can elevate acetylcholine levels, potentially enhancing cholinergic neurotransmission and mitigating cognitive deficits in AD patients.³

The formation of free radical species is another important neurotoxic pathway in the development of AD.⁴ As individuals age, the formation and accumulation of free radical species in the body exacerbates oxidative stress and weakens cellular antioxidant defense mechanisms.⁵ Numerous studies have shown that antioxidants can either prevent or ameliorate AD.^6–8 Therefore, the discovery and development of medicinal compounds possessing both antioxidant and anti-ChE activities are essential for effective treatment.⁹

In line with these results, flavonoids are a diverse class of polyphenolic compounds found abundantly in fruits, vegetables, and medicinal plants.¹⁰ Flavonoids are recognized for their broad spectrum of health-promoting properties, including antioxidant and anti-inflammatory activities. In addition, these compounds have been extensively studied due to their potential therapeutic applications.¹¹ Flavonoids have also shown significant potential as anti-ChE agents and ChE activity inhibitors.^12,13 For instance, compounds isolated from Dorstenia and Polygonum genus showed inhibitory activity against AChE, with IC₅₀ values ranging from 5.93 to 8.75 µg/mL.¹⁴

The genus Erythrina has been reported to comprise medicinal plants rich in flavonoids, with approximately 471 compounds being isolated.^15,16 Several studies revealed that extracts from various Erythrina species, including E. corallodendron,¹⁷ E. caffra,¹⁸ E. velutina¹⁹ and E. variegata²⁰ have the potential to serve as therapeutic agents for AD by inhibiting AChE enzyme activity. However, there is no available literature on the potency of flavonoids derived from the genus Erythrina as AChE inhibitors.

In silico approaches, such as molecular docking and dynamic simulations, offer a powerful method to explore ligand-receptor interactions at the atomic level.²¹ Therefore, this study aims to determine the binding affinity, binding mode, and stability of flavonoids from the genus Erythrina with AchE using computational tools. Potential lead compounds were also identified for further experimental validation by elucidating the structural basis of flavonoid‒AChE interactions. The results of this study are expected to bridge the gap between natural product chemistry and computational biology, providing insights into the therapeutic potential of flavonoids from the genus Erythrina as ChE inhibitors to combat AD.

Materials and Methods

Molecular Docking System Preparation

The 3D X-ray crystal structure of human AChE in complex with galantamine (GNT) was downloaded from Protein Data Bank (PDB) (https://www.rscb.org, accessed on April 25th, 2024) with PDB ID of 4EYH. Furthermore, co-crystallized ligand was separated from the 3D structure of the protein and saved in PBD format, while the water molecules were removed from the ligand-free protein and saved in PDB format. Both the ligand and protein were processed using AutoDock tools (version 1.5.7), and saved in the pdbqt format.

In this study, the structures of the 471 flavonoids were retrieved from several literature reviews.^15,16 The protonated state of these flavonoids was predicted using Chemaxon MarvinSketch, and the 3D structures were optimized using the Merck molecular force field (MMFF94) and saved in the format. Finally, the 3D structures were further processed using AutoDock tools (version 1.5.7), and saved in pdbqt format.²²

Redocking and Virtual Screening

The prepared protein and co-crystallized ligand structures were subjected to a redocking procedure using AutoDock Vina (version 1.2.0). In addition, the grid box parameter was set to a size of 36×36 x 40 with a 0.375 Å spacing, which covered co-crystallized ligand and the largest flavonoid structure for docking. The docking poses generated by AutoDock Vina were separated using Vina Split, and the most preferred pose (with the lowest binding score) was superimposed onto co-crystallized ligand file. Subsequently, Root Mean Square Deviation (RMSD) was calculated using Biovia Discovery, which resulted in RMSD of 0.3403 Å. The docking parameters from redocking procedure were then applied to virtual screen 471 flavonoids against the human AChE protein.

Molecular Dynamics (MD) Simulation

According to previous studies, MD simulation for the best-hit compound was conducted using Amber20 with GPU acceleration.^23,24 The amino acid residues of AChE were parameterized using the ff19SB force field. Furthermore, the partial atomic charges of the ligands were determined using the Austin Model 1‒Bond Charge Corrections (AM1‒BCC) method in the antechamber module of Amber‒Tools21. Additional parameters for the ligand were sourced from the generalized amber force field 2 (GAFF2). AChE‒ligand complex was solvated using the tleap module in Amber-Tools21, with solvent box dimensions of 10 Å. In addition, a small number of Na⁺ and Cl^‒ ions were added to the system to achieve a 0.15 M salt concentration.

The simulation process began with 2 consecutive energy minimization steps. For the protein-ligand complex, a restraint of 25 kcal mol^–1Å^–2 was applied in the initial step, followed by 5 kcal mol^–1Å^–2 in the second step of minimization. After the minimization step, the temperature of the system was increased to 300 K under a 50‒ps NVT (Number‒Volume‒Temperature) condition and then switched to Number‒Pressure‒Temperature (NPT) condition. Subsequently, the system density was adjusted to 1 g.cm^–3 over a 50‒ps period. During subsequent NVT simulations, the restraint on the solute was gradually reduced by 1 kcal mol^–1 Å^–2 every 50 ps until it was fully eliminated.

In this study, the system was simulated at a temperature of 300 K under NPT conditions to produce a 100‒ns MD trajectory. Particle‒Mesh Ewald (PME) method was used to overcome long-range electrostatic interactions, while a 10‒Å cut‒off was applied for short‒range non‒bonded interactions. A SHAKE algorithm was used to maintain bond constraints, including hydrogen atoms. Simultaneously, the temperature and pressure of the system were controlled using a Langevin thermostat and Berendsen barostat, respectively.^23,24

Density Functional Theory (DFT) Study

Molecular structures of the hit compound and quercetin (used as a standard antioxidant) were optimized using DFT with the B3LYP method and the 6–311G++(d,p) basis set in the Gaussian 09 software. Following the optimization, vibrational frequency analysis confirmed that no imaginary frequencies were present in the optimized structures.²⁵

The global descriptive parameters, including the ionization potential (I), electron affinity (A), hardness (η), softness (S), electronegativity (χ), chemical potential (μ), electrophilicity index (ω), electron-donating power (ω‒), and electron accepting power (ω+), were calculated based on the vertical energy method using the following equations:²⁶

In silico Pharmacokinetics Prediction

The ADMETlab 2.0 web server (https://admetmesh.scbdd.com, accessed on May 3, 2024) was used to predict the pharmacokinetic properties (absorption, distribution, metabolism, and excretion) of the potential AChE inhibitor.

Results

Molecular Docking and Toxicity Screening

Molecular docking was used to screen the binding affinity of 471 flavonoids isolated from the genus Erythrina for AChE, with galantamine as the standard. Out of the 471 flavonoids, 88 exhibited lower docking scores compared to that of galantamine which exhibited docking score of −10.44 kcal/mol as cutoff (Table S1). These potential AChE inhibitors were evaluated for toxicity (Table S2), and 9 of these flavonoids showed a low to medium risk of toxicity and were listed in Table 1. Among these flavonoids, 2(S)‒5,7‒dihydroxy‒5′‒methoxy‒[2′′,2′′‒(3′′‒hydroxy)‒dimethylpyrano]‒(5′′,6′′:3′,4′) flavanone (89) and Abyssinoflavanone IV (83) (Figure 1) are the top hits which were further subjected to MD simulation, DFT study and pharmacokinetics prediction.

Table 1 List of 9 AChE Inhibitor Hits

Figure 1 Chemical structure of compounds 89 (left) and 83 (right).

Molecular Dynamics (MD) Simulation

Among these 9 hits, the first 2 were evaluated for further analysis using MD. The results could be described in terms of RMSD, Root Mean Square Fluctuation (RMSF), and Molecular Mechanics Generalized Born Surface Area (MMGBSA) free energy.

RMSD

RMSD plots of the Apo form, along with AChE‒89 and AChE‒83 complexes, were shown in Figure 2. The apo state of AChE exhibited the highest RMSD, with a median value of 1.783 Å and a median absolute deviation (MAD) of 0.321 Å. Both AChE‒83 and AChE‒89 complexes displayed the lowest RMSD (median = 1.525 Å, MAD = 0.114 Å and median = 1.558 Å, MAD = 0.126 Å, respectively) when compared to AChE‒GNT and AChE‒ACh complexes (Table S3).

Figure 2 RMSD plot comparison of Apo AChE Vs AChE‒ 83 complex (green) and AChE‒ 89 complex (blue).

RMSF

RMSF profiles of AChE‒89 and AChE‒83 complexes and the Apo form of AChE were shown in Figure 3. On close inspection of the 100 ns trajectories within several structural regions, including the triad, omega loop, and George entry, significant differences emerged between the apo (unbound) and ligand-bound forms of AChE. The amino acid residues in the triad, omega loop, and George entry exhibited lower B‒factors in the ligand-bound complex compared to the apo form of AChE (Table S4). This observation suggested that amino acids in these structural regions were more stable and exhibited less movement.

Figure 3 RMSF plot comparison of Apo AChE Vs AChE‒ 83 complex (green) and AChE‒ 89 complex (blue).

MMGBSA Free Energy Calculation

MMGBSA Analysis was a computational method extensively used in MD simulations to investigate binding affinities and interactions between molecules.²⁷ The ΔG^o_MMGBSA calculation from MD simulation trajectories revealed that galantamine (a well-known AChE inhibitor) had the lowest binding energy, followed by compounds 89, 83 and acetylcholine (a native substrate of AChE) (Table 2).

Table 2 Δgommgbsa Values of Acetylcholine‒, Galantamine‒, 89‒, and 83‒AChE Complex (Expressed as Median ± SD)

To evaluate whether there was a significant difference between the 4 complexes, Kruskal–Wallis rank-sum test was performed. Furthermore, the test was followed by Dunn’s multiple-comparison test, using the Bonferroni method as a post‒hoc test.²³ The test suggested that there was no statistical difference between the ΔG°_MMGBSA values of galantamine and both 89 and 83. However, both were significantly different from that of acetylcholine (Figure 4); therefore, compounds 89 and 83 were identified as potential AChE inhibitors.

Figure 4 The 2D representation of 89 (left) and 83 (right) interaction with active site of AChE.

Density Functional Theory (DFT)

According to previous studies on other flavonoid compounds, DFT study was conducted to evaluate the free‒radical scavenging potency of 89 and 83. The global descriptive parameters, including E_HOMO, E_LUMO, ΔE_gap, ionization potential (I), electron affinity (A), hardness (η), softness (σ), electronegativity (χ), chemical potential (μ), electrophilicity index (Ω), electron-donating power (ω^–), and electron‒accepting power (ω⁺), were used to assess the reactivity of 89 and 83 toward free radicals in comparison with Quercetin (a well‒known antioxidant) as a standard. The global descriptive parameter values for 89, 83, and Quercetin were listed in Table 3.

Table 3 Global Descriptive Parameter of Compound 89 and Quercetin as Standard

In silico Pharmacokinetics Prediction

Pharmacokinetic prediction was a crucial aspect of drug design and discovery, as it allowed for early estimation of the absorption, distribution, metabolism, and excretion (ADME) properties of potential drug candidates. This early estimation significantly reduced the fraction of pharmacokinetics-related failures in the clinical phase, which eventually enhanced the efficiency of drug development.²⁸ Therefore, after evaluating the potency of compound 89 as AChE inhibitor and radical scavenger, pharmacokinetic prediction of 89 and 83 was performed, and the results were summarized in Table 4.

Table 4 Pharmacokinetics Prediction of Compounds 89 and 83

Discussion

Out of the 471 flavonoids evaluated, 9 hits were identified with lower binding affinities compared to that of galantamine (GNT) and were predicted to exhibit low to medium toxicity. The top 2 hits, namely 2(S)‒5,7‒dihydroxy‒5′‒methoxy‒[2′′,2′′‒(3′′‒hydroxy)‒dimethylpyrano]‒(5′′,6′′:3′,4′) flavanone (89) and Abyssinoflavanone IV (83) (Figure 1) were further subjected to MD simulation to evaluate the stability of AChE upon binding to these compounds. This was followed by in silico pharmacokinetics and drug-likeness prediction to evaluate the potency of these compounds to be developed as potential drugs for AD therapy.

RMSD was a crucial metric for studying biomolecules during simulations and experiments. It quantified the average displacement of atoms from their initial positions and could provide insights into the dynamic behavior of a biomolecule as it interacted with other molecules (eg, a docked ligand). RMSD plot (Figure 2) indicated that both 89 and 83 stabilized AChE protein, as both AChE‒83 and AChE‒89 complexes displayed the lowest RMSD (median = 1.525 Å, MAD = 0.114 Å and median = 1.558 Å, MAD = 0.126 Å, respectively). A lower RMSD value indicated greater stability of the biomolecular structure during these interactions.²⁹ As comparison, RMSD values of both 83 and 89 were lower compared to that of AChE-quercetin complex in the literature (1.9 Å). Furthermore, it was noteworthy that Quercetin exhibited inhibitory activity against AChE with IC₅₀ of 4.59 ± 0.27 µM.³⁰ The stability of both AChE‒83 and AChE‒89 complexes was further supported by RMSF plot (Figure 3). RMSF quantified these fluctuations by calculating RMSD of the atomic positions from their equilibrium positions.³¹ RMSF plot served as a common representation of residues that had undergone substantial changes during MD simulation.³² For all AChE‒ligand complexes, except for the C-terminus and N-terminus, the other residues in AChE did not fluctuate significantly, including in the triad, omega loop, and George entry.

A 2D graphical representation of the interaction of AChE with 89 and 83 was shown in Figure 4. At the catalytic or acylation site of AChE, compound 89 interacted with His⁴⁴⁷ through Pi-Alkil, while compound 83 interacted with the same amino acid residue through a conventional hydrogen bond. The catalytic or acylation site comprised 3 essential amino acid residues (such as Ser²⁰³, His⁴⁴⁷, and Glu³³⁴). Each of these amino acid residues played a crucial role in the hydrolysis of acetylcholine. Ser²⁰³ acted as a nucleophile during substrate hydrolysis, His⁴⁴⁷ stabilized the transition state during the acylation process, and Glu³³⁴ assisted in substrate binding and catalysis.³³ Because 89 and 83 could interact with His⁴⁴⁷, these compounds could interrupt the catalytic function of AChE. Moreover, in the “anionic” subsite, compound 89 interacted with Trp⁸⁶ and Phe³³⁸, while compound 83 interacted with both Trp⁸⁶ and Tyr³³⁷. These amino acid residues played important roles in the binding and stabilization of the substrate in the active site as it was bound to the quaternary trimethylammonium choline moiety.³⁴

MMGBSA free energy calculations indicated that both 89 and 83 had significantly lower binding energies than ACh, which was the natural substrate of AChE. Although both 89 and 83 had higher binding energies than galantamine (a known AChE inhibitor), the median difference in binding energy during the 100 ns simulation was not significantly different (Figure 5). This result suggested that the binding of 89 or 83 was more favorable than that of ACh; therefore, inhibitory activity against AChE could be exhibited. As comparison with available literature data, the ΔG_bind of both 89 and 83 are lower than that of Quercetin (ΔG_bind = −21.0 ± 2.5 Kcal/mol)³⁵ which further support the potential inhibitory activity of 89 and 83 against AChE.

Figure 5 Box Plots of ΔGoMMGBSA for ligand binding AChE with significant test results from a Dunn test, a post hoc of Kruskal‒Wallis rank sum test.

In terms of free radical scavenging activity, the global descriptive parameter values of 89 and 83 were close to those of the Quercetin. However, in terms of electron affinity (EA) and electron-accepting power (ω⁺), the values were quite different. EA corresponded to the energy released when a neutral molecule captured an electron. The higher the EA value, the better the ability of a neutral molecule to accept electrons from a radical, thereby suggesting better radical scavenging activity.³⁶ Furthermore, EA of Quercetin was much higher than those of 89 and 83, implying that the Quercetin had a better ability to accept electrons compared to 89 and 83. Quercetin had a higher electron‒accepting power (ω⁺) than 89 and 83, which further supported the better ability of Quercetin to accept electrons. However, the higher E_HOMO and lower EA of 89 and 83 suggested that both compounds were good electron donors compared to quercetin. In this study, 89 and 83 could be potential radical scavengers owing to their electron‒donating activities.

Based on the results of MD simulation and DFT study, it was evident that 89 and 83 could be potential AChE inhibitors as well as free radical scavengers. Therefore, in the next step, in silico pharmacokinetic studies were conducted to predict the pharmacokinetic properties of 89 and 83. Pharmacokinetics referred to the study of how drug concentrations change over time following drug administration. This included processes, such as absorption, distribution, metabolism, and excretion (ADME).²²

Drug absorption was a critical aspect of pharmacokinetics that comprised the process by which a drug entered the bloodstream to exert its therapeutic effect.³⁷ In terms of drug absorption parameters, the prediction model suggested that both 89 and 83 exhibited high Caco-2, which were the human intestinal epithelial cells and MCDK permeability. High Caco-2 permeability indicated that a compound could readily cross the intestinal barrier and be absorbed into the systemic circulation. Meanwhile, a high MDCK permeability implied that a compound could traverse the renal tubular epithelium efficiently. Compounds with high Caco-2 and MDCK permeability were likely to be well‒absorbed and efficiently eliminated, making these compounds favorable candidates for drug development.³⁸ The prediction model also suggested that 89 and 83 were not substrates of P‒gp but could serve as P-glycoprotein (P‒gp) inhibitors. This result implied that the bioavailability of these compounds was not affected by P-gp activity, which acted as an efflux transporter and actively pumps xenobiotics into the lumen of the small intestine.²⁴

Pharmacokinetics drug distribution referred to the reversible transfer of a drug between different compartments of the body. Once the drug entered systemic circulation through absorption or direct administration, it was distributed in both interstitial and intracellular fluids.²² Based on the prediction model, 89 and 83 were reported to have a good volume distribution with a value of 0.04–20 L/kg. Blood-brain barrier (BBB) permeability was another parameter used to describe drug distribution in the human body. BBB was a selective, semi-permeable membrane composed of tight junctions. Furthermore, it served as a protective barrier that limited the entry of specific substances from the bloodstream into the central nervous system (CNS). When a drug crossed the BBB, it gained access to the brain tissue. This ability was essential for treating CNS disorders (eg AD and epilepsy). The prediction model suggested that 89 and 83 could cross BBB. The unbound fraction represented the proportion of free drugs in the plasma. It was the fraction of drug molecules that were not bound to plasma proteins (such as albumin). Only the unbound drug could be distributed to the tissues, while the bound drug remained in the bloodstream. The prediction model suggested that compound 89 had a medium unbound fraction. Nevertheless, both 89 and 83 were predicted to exhibit high plasma protein binding (>90%), which could have a slower elimination rate, affecting the dosing intervals.

In terms of drug metabolism, the prediction model suggested that 89 and 83 could act as CYP2C9 substrates, indicating that these compounds could be metabolized by this specific enzyme. In the human liver, was the predominant enzyme within the CYP2C subfamily and played a crucial role in metabolizing clinically relevant drugs in a narrow therapeutic range.³⁹

The drug excretion properties of 89 and 83 were evaluated in terms of drug clearance and half-life. Drug clearance represented the elimination rate of a substance divided by its concentration, and reflected the theoretical plasma volume from which the substance was entirely removed per unit of time. The prediction model suggested that compound 89 had moderate clearance, indicating that 89 was eliminated from the body at a moderate rate, and suggested a balance between rapid elimination and prolonged presence in the bloodstream. Meanwhile, 83 was predicted to have a high clearance, indicating that it could be eliminated from the body at a higher rate.

The half-life (T_1/2) represented the estimated time, which was required for its concentration or amount in the body to decrease to 50% of its initial concentration (50%). The prediction model suggested that compound 89 had a half-life lower than 83, suggesting that it could remain in the body for a longer duration before being eliminated compared to 89.

Conclusion

In conclusion, through a combination of molecular docking, in silico toxicity screening had led to the identified 2(S)-5,7-dihydroxy-5′-methoxy-[2′′,2′′-(3′′-hydroxy)-dimethylpyrano]-(5′′,6′′:3′,4′) flavanone (89) and Abyssinoflavanone IV (83) from the genus Erythrina as potential AChE inhibitors. MD simulation revealed the stability of both AChE‒83 and AChE‒89 complexes with RMSD of 1.525 ± 0.114 Å and 1.558 ± 0.126 Å, respectively. Both 83 and 89 exhibited lower binding energy compared to that of acetylcholine, further supporting the potency of 83 and 89 as AChE inhibitors. Moreover, DFT study suggested that both 83 and 89 exhibited higher E_HOMO and lower EA suggesting that both compounds were good electron donors compared to quercetin indicating that 89 and 83 could be potential radical scavengers through their electron-donating activity.

This study primarily relied on computational methods, the results could serve as s foundation to highlight the potency of 89 and 83 as both AChE inhibitors and free radical scavengers. Further investigation in wet laboratory experiments was warranted to validate their therapeutic potential for AD.