Arylbenzofurans from the Root Bark of Morus alba as Triple Inhibitors of Cholinesterase, β-Site Amyloid Precursor Protein Cleaving Enzyme 1, and Glycogen Synthase Kinase-3β: Relevance to Alzheimer's Disease.

Cholinesterase, β-site amyloid precursor protein cleaving enzyme 1 (BACE1), and glycogen synthase kinase-3β (GSK-3β) are the three main enzymes responsible for the early onset of Alzheimer's disease (AD). The main aim of the present study was to delineate and accentuate the triple-inhibitory potential of arylbenzofurans from Morus alba against these enzymes. Overall, the enzyme inhibition assays demonstrated the prominence of mulberrofuran D2 as an inhibitor of AChE, BChE, BACE1, and GSK-3β enzymes with IC50 values of 4.61, 1.51, 0.73, and 6.36 μM, respectively. Enzyme kinetics revealed different modes of inhibition, and in silico modeling suggested that mulberrofuran D2 inhibited these enzymes with low binding energy through hydrophilic, hydrophobic, and π-cation interactions in the active site cavities. Similarly, in Aβ-aggregation assays, mulberrofuran D2 inhibited self-induced and AChE-induced Aβ aggregation in a concentration-dependent manner that was superior to reference drugs. These results suggest that arylbenzofurans from M. alba, especially mulberrofuran D2, are triple inhibitors of cholinesterase, BACE1, and GSK-3β and may represent a novel class of anti-AD drugs.

Introduction

In Alzheimer’s disease (AD), cognition progressively declines due to neurodegeneration. AD affects the majority of elderly people and thus has been regarded as the fourth leading cause of death worldwide. Neuropathologically, AD is characterized by the formation of amyloid-β peptide (Aβ) and neurofibrillary tangles (NFTs), along with the degeneration of basal forebrain cholinergic neurons.[1] AD autopsy studies in the 1970s unveiled significant losses of cholinergic cells in the basal forebrain and cholinergic enzymes in the cortex. These findings, along with the well-known contribution of the cholinergic system to memory and cognitive skills, led to the “cholinergic hypothesis”, and since then, the cholinergic neurotransmitter system has become known for its role in the cognitive decline seen in AD.[2]

Additionally, the deposition of amyloid protein via the amyloid cascade leads to neurodegeneration and ultimately senile dementia,[3] as described in the “amyloid hypothesis”.[4] Under normal conditions, amyloid-β is cleaved from amyloid precursor protein (APP) by β- and γ-secretases and is released outside the cell, where it is rapidly degraded. However, under pathological conditions (e.g., a reduced metabolic ability to degrade Aβ), Aβ peptides accumulate, contributing to Aβ amyloid fibril formation. The assembly of Aβ amyloid fibrils progresses to senile plagues, causing neurotoxicity and induced tau pathology, thus triggering neuronal cell death.[5]

Another hypothesis is the “tau hypothesis”, in which tau protein is regarded as the principle causative agent of AD. Tau is a microtubule-associated protein that regulates the stability of tubulin. In AD, tau accumulates in neuronal cells in a hyperphosphorylated state, forming what are commonly known as neurofibrillary tangles (NFTs). This impairs the ability of tau to bind microtubules, which reduces their stability and regulation,[6] ultimately leading to tauopathy and cognitive decline.[7]

The three hypotheses on AD described above have gained much attention, and the treatment approaches for AD include the use of cholinesterase (AChE and BChE) inhibitors, secretase (β- and γ-secretase) inhibitors, and potential kinase inhibitors (for GSK-3β, CDK5 and MARK).[8,9] Despite the long history of the cholinergic and amyloid hypotheses, there have been disappointments during the discovery and development of therapeutics for AD, due to failures in different stages of clinical trials; thus, interest has shifted toward the tau hypothesis. Nevertheless, regarding the former two hypotheses, researchers have done their best to discover novel therapeutics by natural or semisynthetic routes, so we have been gifted with several novel dual inhibitors. Still, we researchers are not satisfied with these discoveries and are thus focusing on triple inhibitors or multi-target-directed ligands (MTDLs). Due to multifactorial etiology of AD and the fact that there is no effective treatment till date besides drugs alleviating associated symptoms, molecules designed to hit simultaneously different key targets of the complex pathological network (known as MTDLs) emerge as a more realistic alternative.[10] Compared to “one-molecule, one-target” with presumed selectivity and inadequacy for multigenic diseases,[11] discovery of multitarget small molecules that can modulate diverse receptors or enzymatic systems in AD would lead to a new and more efficient treatment approach.[12,13] In addition, combination therapy has issues of drug–drug interactions, off-target adverse effects, poor patient compliance, and high development costs, and MTDLs have inherent advantages over these issues.[14] Herein, we continue our efforts to discover novel phytochemicals from natural sources for the treatment of AD. We have evaluated the triple-inhibitory potential of arylbenzofurans from Morus alba against cholinesterase, β-site amyloid precursor protein cleaving enzyme 1 (BACE1), and glycogen synthase kinase-3β (GSK-3β), which address the three aforementioned hypotheses.

Different parts of the M. alba Linn plant, especially the root bark, leaves, fruit, and twigs, have been used in traditional Chinese medicine as remedies for impaired vision, hypertension, various metabolic disorders (e.g., diabetes), arthritis, and fevers. Diverse groups of phytochemicals have been isolated from M. alba, including terpenoids, alkaloids, flavonoids (such as anthocyanins and chalcones), phenolic acids, stilbenoids, and coumarins.[15] However, Diels–Alder-type adducts (formed by the Diels–Alder reaction between the α,β-olefinic moiety of a chalcone and an isoprene moiety) are the most representative compounds of the genus Morus.[16] Many of these compounds exhibit various biological activities, including antioxidant, anti-inflammatory, antifungal, antimicrobial, antitumor, antihypotensive, antidiabetic, and anti-Alzheimer’s disease activity.[16−19] In our recent study on the root bark of M. alba, we discovered moracin derivatives that functioned as dual inhibitors of BACE1 and cholinesterase.[20] Similarly, in a recent study by Xia et al.,[21] 10 Diels–Alder-type adducts, which are abundant in M. alba, have been characterized as multitargeted agents for Alzheimer’s disease. Besides Diels–Alder-type adducts, 2-arylbenzofurans represent the other group of active phytochemicals that are commonly substituted by prenyl and geranyl groups at different positions. Numerous studies have suggested an important role of prenyl and geranyl groups in enzyme inhibition.[22] In addition, it was reported that the prenyl, geranyl, and/or farnesyl group increase cellular permeability.[23] Therefore, 2-arylbenzofurans substituted by prenyl, geranyl and/or farnesyl group at different positions could be promising targets in neurodegenerative diseases. Thus, in the present study, we have evaluated the triple-inhibitory potential of arylbenzofurans from M. alba (Figure ), having prenyl, geranyl, and/or farnesyl group at different positions, against cholinesterase, BACE1, and GSK-3β.

Figure 1

Structures of compounds isolated from root barks of M. alba Linn.

Results

Inhibition of Multiple Enzyme Targets by Arylbenzofurans

The structures of the arylbenzofurans isolated from the root bark of M. alba Linn are shown in Figure . The potential of each compound to inhibit the tested enzymes by 50% is listed in Table . As shown in the table, 1, 2 and 4 strongly inhibited AChE. Among them, 2 was the most potent, with an IC50 value of 4.61 ± 0.08 μM, followed by 4 (IC50, 13.92 ± 1.98 μM) and 1 (IC50, 23.05 ± 1.15 μM). Interestingly, 3 and 5 did not have observable effects. However, for BChE inhibition, all of the test compounds except 4 were active. Again, with the lowest IC50 value (1.51 ± 0.09 μM), 2 was a stronger inhibitor than the other compounds against BChE. The IC50 values for 1, 3, and 5 were 6.12, 64.51, and 19.14 μM, respectively. Comparing the inhibitory activities of the compounds against the cholinesterases, 4 was selective toward AChE, 3 and 5 were selective toward BChE, and 1 and 2 exhibited dual inhibition, although 2 had multifold higher activity than that of 1 and was the most potent compound overall.

Table 1

Cholinesterase, β-Site Amyloid Precursor Protein Cleaving Enzyme 1 (BACE1), and Glycogen Synthase Kinase-3β (GSK-3β) Inhibition by Compounds Isolated from M. alba, along with the Enzyme Kinetics of Mulberrofuran D2

	IC50 (μM)a
compounds	AChE	BChE	BACE1	GSK-3β
mulberrofuran D (1)	23.05 ± 1.15	6.12 ± 1.41	3.74 ± 1.01	8.12 ± 0.66
mulberrofuran D2 (2)	4.61 ± 0.08	1.51 ± 0.09	0.73 ± 0.03	6.36 ± 0.74
mulberrofuran H (3)	>200	64.51 ± 3.10	1.04 ± 0.78	4.41 ± 0.30
morusalfuran B (4)	13.92 ± 1.98	>200	2.03 ± 0.34	7.57 ± 0.61
sanggenofuran A (5)	>200	19.14 ± 1.20	5.64 ± 0.27	3.59 ± 0.14
berberineb	1.33 ± 0.01	34.57 ± 0.74
quercetinb			3.38 ± 0.05
luteolinb				2.18 ± 0.13

IC50 is expressed as the mean ± standard deviation (SD) from triplicate experiments.

Berberine, quercetin, and luteolin were used as positive controls.

The inhibition type was determined from the Lineweaver–Burk plot, and the inhibition constant was determined from the Dixon plot

Inhibition constants at constant adenosine 5-triphosphate (ATP) and substrate concentrations, respectively.

The BACE1 inhibition results demonstrated that all of the test compounds were potent at low micromolar concentrations. With a submicromolar IC50 value (0.73 ± 0.03 μM), 2 was the most active, followed by 3 (1.04 ± 0.78 μM), 4 (2.03 ± 0.34 μM), 1 (3.74 ± 1.01 μM), and 5 (5.64 ± 0.27 μM). Compounds 2–4 had better activity than the reference drug quercetin (IC50, 3.38 ± 0.05 μM). Similarly, the IC50 values of 1–5 against GSK-3β were less than 10 μM, indicating that these compounds were potent inhibitors of this enzyme. Luteolin was used as a natural reference drug, as it has been reported to be a potent inhibitor of GSK-3β.[24] Upon review of the inhibition patterns of 1–5 against the tested enzymes, 1 and 2 were found to be triple inhibitors of cholinesterase, BACE1, and GSK-3β, with 2 being the more active compound, having lower IC50 values than those of 1 for all of the enzymes.

Enzyme Kinetics of Arylbenzofurans

To determine the enzyme inhibition mechanisms, we analyzed the enzyme kinetics of the respective enzymes at different substrate concentrations. Based on the IC50 values of enzyme inhibition, the inhibition mode and enzyme kinetic parameters of 2 were investigated via Lineweaver–Burk plots and Dixon plots (Table and Figures and 3). The kinetic results of enzyme inhibition revealed that 2 was a noncompetitive inhibitor of AChE and BACE1, with Ki values of 1.40 and 0.45 μM, respectively. For BChE inhibition, 2 exhibited a mixed-type inhibition mode, with a Ki value of 0.78 μM. In the noncompetitive mode of enzyme inhibition (Figure D,F), the Vmax value increases in a dose-dependent manner without changing the Km value. In the case of BChE inhibition (Figure E), the plots of 1/V versus 1/[S] produced a family of straight lines with a common intercept in the second quadrant, indicating mixed-type inhibition.

Figure 2

Dixon plots and Lineweaver–Burk plots for AChE (A, D), BChE (B, E), and BACE1 (C, F) inhibition, respectively, by mulberrofurans D2.

Figure 3

Dixon and Lineweaver–Burk plots for GSK-3β inhibition by mulberrofuran D2, (A, C) at constant ATP and (B, D) at constant GSM substrate, respectively.

To determine the mechanism of GSK-3β enzyme inhibition, we assessed the ability of 2 to competitively replace ATP and the substrate GSM (Figure ). Under the conditions of either a constant substrate (25 μM) or a constant ATP (1 μM) concentration, the Lineweaver–Burk plots revealed that all of the lines intersected at the same point on the y-axis (Figure C,D), suggesting that Km increased with increasing concentrations of 2, while 1/Vmax did not change. The data demonstrated that 2 competed with both the substrate and ATP at their respective binding sites, with Ki values of 3.09 and 5.54 μM, respectively.

In Silico Molecular Docking Simulation of Enzyme Inhibition

Compound 2, the triple inhibitor of cholinesterase, BACE1, and GSK-3β, was selected for further molecular docking simulations, due to its high inhibitory activity against these enzymes (Figures –6). Tables and 3 display the binding energies of 2 and the reference compounds in the active site cavities of the respective enzymes, along with various interacting residues. As shown, 2 bound to the active site cavity of AChE (PDB code 1EVE) with lower binding energy than the reference compounds (−11.66 kcal/mol) by forming hydrophobic bonds with Trp84, Tyr334, and Trp279 and an H-bond with Tyr121 (Figure A–C). These are prime interacting residues at the peripheral anionic site (PAS) responsible for allosteric inhibition. Besides this, other hydrophobic interactions with Tyr70, Phe330, and His440 were observed, which probably stabilize 2 within the binding cavity. Compared to the noncompetitive reference inhibitor donepezil, 2 displayed two additional H-bond interacting residues (Arg289 and Tyr121), in addition to a common residue (Phe288).

Figure 4

Molecular docking of AChE (A) and BChE (D) binding with mulberrofuran D2 along with positive controls. Chemical structures of galantamine, donepezil, tacrine, 5i, and mulberrofuran D2 are shown as red, blue, black, pink, and green stick, respectively. Closeup of the mulberrofuran D2 binding pose in the active site of AChE (B, C) and BChE (E, F).

Figure 6

Molecular docking of glycogen synthase kinase (GSK)-3β binding with mulberrofuran D2 along with positive controls (A). Chemical structures of AMP–PNP, andrographolide, VP0.7, and mulberrofuran D2 are shown as red, blue, white, and green sticks, respectively. Magnesium ions are shown as brown spheres. Closeup of the mulberrofuran D2 binding pose in the ATP-binding site (B, C) and substrate-binding site (D, E), respectively.

Table 2

Binding Sites and Docking Scores of Mulberrofuran D2 in Acetylcholinesterase, Butyrylcholinesterase, and β-Site Amyloid Precursor Protein Cleaving Enzyme 1

compounds	target enzyme (binding energy, kcal/mol)	H-bond interacting residues (no. of H-bonds)	hydrophobic interacting residues	electrostatic interacting residues
galantaminea	AChE (−9.32)	Gly119 (1), Ser200 (2), His440 (1), Glu199 (1)	π–alkyl: Trp84; π–π T-shaped: Phe331; π–σ: Phe330, Trp233	attractive charge: Asp72; π–cation: Phe330
donepezila	AChE (−10.22b)	Phe288 (1)	π–alkyl: Phe330; π–π stacked: Trp84, Trp279; π–σ: Tyr334, Phe331, Trp279
tacrinea	BChE (−10.22b)	His438 (1)	π–alkyl: Ala328; π–π stacked: Trp82	π–cation: Trp82; attractive charge: Glu197
5ia,c	BChE (−6.23)	Tyr440 (1), Trp430 (1), Trp82 (1)	π–alkyl: Ala328; π–π stacked: Phe329	π–anion: Asp70
QUDa	BACE1 (−10.22b)	Asp228 (3), Asp32 (1), Gly230 (1), Thr231 (2)	alkyl: Val332; π–π T-shaped: Tyr71, Phe108
sinensetina	BACE1 (−7.30)	Trp76 (1), Tyr198 (1), Asn37 (1), Ile126 (1)	π–alkyl: Val69; π–π T-shaped: Phe108
mulberrofuran D2	AChE (−11.66)	Phe288 (1), Arg289 (1), Tyr121 (1)	π–alkyl: Tyr70, Trp84, Phe330, Tyr334, His440; π–π T-shaped: Phe330, Tyr334; π–π stacked: Trp279
	BChE (−10.12)	Gln67 (1), Asp70 (1), Asn68 (1), His438 (1)	π–alkyl: Pro84, Tyr440, Tyr332, Phe329, Trp82; alkyl: Ala328, Met437, Pro285; π–π stacked: Trp82
	BACE1 (−10.06)	Ser36 (1), Asn37 (1), Tyr198 (1)	alkyl: Val69; π–alkyl: Tyr71, Trp76, Phe108, Tyr198, Ile126; π–π stacked: Tyr71

Positive control.

RMSD values: 0.88 Å for donepezil, 0.35 Å for tacrine, and 0.58 Å for QUD.

1-[2-(2,4-Dichlorophenyl)-2-phenylmethoxyethyl]imidazole.

Table 3

Binding Sites and Docking Scores of Mulberrofuran D2 in Glycogen Synthase Kinase-3β

compounds (binding energy, kcal/mol)	H-bond interacting residues (no. of H-bonds)	hydrophobic interacting residues	electrostatic interacting residues
AMP–PNPa (−7.75b)	Lys85 (1), Val135 (2), Lys183 (1), Asp133 (1), Gly63 (1), Asn64 (2), Gly65 (1)	π–alkyl: Val70, Ala83, Leu188, Cys199, Ala83, Val135	electrostatic bond: MG1002, MG1003, Lys85, Lys183
andrographolidea (−8.17)	Gln89 (1), Lys94 (2), Arg96 (2), Arg180 (2), Lys292 (1)	alkyl: Pro294, Arg96; π–alkyl: Phe93, Phe293
VP0.7a (−6.75)	Arg209 (1), Leu207 (1), Ser236 (1), Thr235 (1)	alkyl: Val240, Pro331; π–alkyl: His173, Arg209, Leu207; amide-π stacked: Val208, Arg209
mulberrofuran D2 (−9.78c, −9.87c)	Asp133 (1), Val135 (1)	alkyl: Leu188, Ile62; π–alkyl: Tyr134, Ala83; π–σ: Val70, Leu188,	π–cation: MG1002
	Gln89 (1), Asp90 (1), Lys94 (1)	alkyl: Val87, Arg96, Val263; π–alkyl: Pro294, π–π T-shaped: Phe67

Phosphoaminophosphonic acid–adenylate ester (AMP–PNP), andrographolide, and N′-dodecanoyl-1-ethyl-4-hydroxy-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (VP0.7) were used as positive controls for ATP-competitive, substrate-competitive, and allosteric inhibition, respectively.

RMSD value: 0.98 Å.

Binding poses of mulberrofuran D2 in ATP- and substrate-binding sites, respectively.

Similarly, 2 exhibited mixed-type inhibition of BChE (PDB code 4BDS), with a low binding energy (−10.12 kcal/mol) that was similar to that of the reference allosteric inhibitor 5i (1-[2-(2,4-Dichlorophenyl)-2-phenylmethoxyethyl]imidazole)[25] but lower than that of the reference catalytic inhibitor tacrine. The H-bond interacting residues included Gln67, Asp70, Asn68, and His438, while the major hydrophobic residues responsible for BChE inhibition by 2 were Trp82, Ala328, Pro84, Pro285, Met437, Phe329, Tyr440, and Tyr332 (Figure D–F).

X-ray crystallography has been used to determine the structures of numerous BACE1 complexes, and we employed 2WJO for our docking study (Figure ). As shown in Table , QUD was used as a reference catalytic inhibitor of BACE1, while sinensetin was used as an allosteric inhibitor.[26] Bridging with three H-bond interacting residues (Asn37, Ser36, and Tyr198), 2 bound to the active allosteric site of BACE1 with very low energy (−10.06 kcal/mol). The other hydrophobic interacting residues (Val69, Tyr71, Trp76, Phe108, Tyr198, and Ile126) enhanced the binding affinity by increasing stability.

Figure 5

Molecular docking of BACE1 (A) binding with mulberrofuran D2 along with positive controls. Chemical structures of QUD, sinensetin, and mulberrofuran D2 are shown as red, blue, and green sticks, respectively. Closeup of the mulberrofuran D2 binding pose in the active site of BACE1 (B, C).

Similarly, for GSK-3β enzyme inhibition (Table ), 2 formed H-bonds with Gln89, Asp90, and Lys94 within the substrate pocket, with lower binding energy (−9.87 kcal/mol) than that of the reference substrate-competitive inhibitor andrographolide (−8.17 kcal/mol). Other hydrophobic interacting residues that helped to stabilize ligand 2 within this pocket were Arg96 and Pro294, similar to the reference inhibitor (Figure ). Additionally, in the ATP-binding pocket, the major H-bond interaction involved Asp133 and Val135 residues, which are prime interacting residues for ATP-competitive inhibitors. Furthermore, the aromatic ring exhibited a π–cation interaction with MG1002, which accounted for the higher binding affinity of 2 (−9.78 kcal/mol) than that of the ATP-competitive reference inhibitor phosphoaminophosphonic acid–adenylate ester (AMP–PNP). Ala83, Leu188, and Val70 stabilized ligand 2 within the ATP-binding pocket in a similar manner to the reference inhibitor. Overall, the docking study revealed that ligand 2 bound to the active site cavity of each enzyme with lower energy and higher binding affinity than those of the reference inhibitors.

Effects on the Aβ Aggregation

Since 2 showed the highest potency for AChE inhibition, it was selected to assess its ability to inhibit self-induced and AChE-induced amyloid-β aggregations. As shown in Figure , 2 showed moderate effect on self-induced Aβ1–42 aggregation inhibition, which was concentration-dependent. At 20 μM concentration, 2 inhibited the self-induced Aβ1–42 aggregations by 60%. Curcumin was used as a reference drug, which inhibited the aggregation by 43% at 20 μM. Similarly, in the AChE-induced amyloid aggregation assay, 2 inhibited the Aβ1–40 aggregations by 83% at 20 μM. Donepezil was used as a reference drug, which inhibited AChE-induced amyloid aggregation by 35% at 50 μM concentration. Activity of 2 even at 10 μM concentration was better than that of reference compounds in Aβ aggregation assays.

Figure 7

Results from the thioflavin T assay. (A) Inhibition of Aβ1–42 self-induced aggregation and (B) AChE-induced Aβ1–40 aggregation by mulberrofuran D2 and reference compounds, curcumin and donepezil. The data represent the mean ± SD of three independent experiments. Different letters in graphs are significantly different with Duncan’s test at p < 0.05.

Druglikeness and Absorption, Distribution, Metabolism, and Excretion (ADME) Prediction

Druglikeness was predicted for all of the tested 2-arylbenzofuran derivatives. The results suggested that these compounds have good druglike properties, as they adhered to the MDDR-like rule and Lipinski’s rule (Table ). However, in the case of 4, though it exhibited druglike properties according to the MDDR-like rule,[27] it did not obey Lipinski’s rule.[28]

Table 4

Druglikeness and ADME Characteristics as Determined by PreADMETg

	druglikeness		ADME characteristics
compounds	MDDR-like rule	Lipinski’s rule	log Po/wa	PPBb	HIAc	in vitro Caco-2 permeability (nm/s)d	in vitro MDCK cell permeability (nm/s)e	in vivo BBB penetration ([brain]/[blood])f
mulberrofuran D (1)	druglike	suitable	8.84	100	94.23	32.18	0.044	11.34
mulberrofuran D2 (2)	druglike	suitable	8.28	100	95.77	31.01	0.044	12.13
mulberrofuran H (3)	mid-structure	suitable	4.74	100	91.57	5.82	0.052	1.01
morusalfuran B (4)	druglike	violated	10.66	100	94.81	44.49	0.078	13.63
sanggenofuran A (5)	druglike	suitable	8.98	100	95.85	47.86	0.047	13.41

The log of the coefficient of solvent partitioning between 1-octanol and water.

Plasma protein binding (<90% represents weak binding, and > 90% represents strong binding).

Human intestinal absorption (HIA) (0–20% is poorly absorbed, 20–70% is moderately absorbed, and 70–100% is well-absorbed).

0–10 nm/s is low permeability, 10–100 nm/s is medium permeability, and >100 nm/s is high permeability.

Permeability across Madin–Darby canine kidney (MDCK) cells.

<0.1 is low absorption by the central nervous system, 0.1–2.0 is middle absorption, and >2.0 is high absorption.

Lipinski’s rule: an orally active drug has no more than one violation of H-bond donors (≤5), H-bond acceptors (≤10), molecular weight (≤500 Da), and log P (≤5). MDDR-like rule: the MDDR-like rule describes a molecule as druglike or non-drug-like on the basis of the number of rings, rigid bonds, and rotatable bonds.

The ADME prediction results of these compounds revealed an excellent percentage of plasma protein binding (100%), a high percentage of human intestinal absorption (HIA) (>90%), and good lipophilicity (log Po/w value range of 4.74–10.66) (Table ). The blood–brain barrier (BBB) penetration values ([brain]/[blood]) were >10%, except for 3 (1.01%). Since human epithelial colorectal adenocarcinoma cells (Caco-2) and Madin–Darby canine kidney (MDCK) cells are commonly used to assess intestinal permeability in early drug discovery in vitro, we used these cell models to predict the permeability of our test compounds. The compounds exhibited low permeation across MDCK cells but medium permeation across Caco-2 cells, except for 3, which exhibited a low value. All of these predicted results will be helpful for the optimization of druglike properties.

Discussion

The most common form of neurodegenerative disease is Alzheimer’s disease. It is characterized by amyloid-β and tau protein deposition. Various proteases, α-, β-, γ-, and δ-secretases, have been reported to be involved in APP cleavage via either amyloidogenic or nonamyloidogenic pathways.[29,30] However, β-secretase mediates the initial and rate-limiting process of Aβ generation. Monomeric and oligomeric forms of this protein in the brain parenchyma aggregate through a nucleation-dependent cascade, forming mature β-sheet-rich protein structures called amyloid fibrils. These amyloid fibrils stack upon one another and form conserved amyloid, propagating the pathology and neurotoxicity. Hence, inhibitors of β-secretase prevent amyloidogenesis. Our results revealed five arylbenzofurans with potent inhibitory activity against BACE1 at low micromolar concentrations. Test compounds 2–5 were more potent than the reference compound berberine. Interestingly, 2 exhibited the most prominent activity against BACE1, with an IC50 value <1 μM. Enzyme kinetics and molecular docking studies revealed 2 as a competitive inhibitor with lower binding energy than that of the reference compounds. To validate our docking results, we used QUD, a potent catalytic inhibitor of BACE1, as a reference.

The cholinergic system is the most important neurotransmitter system involved in learning and memory. Acetylcholinesterase (AChE) hydrolyzes acetylcholine (ACh) to choline and acetate, while butyrylcholinesterase (BChE) acts as a coregulator of AChE activity. These esterases also have nonenzymatic functions associated with Aβ aggregation and NFT formation in the AD brains of mice and humans.[31,32] Our cholinesterase inhibition assays demonstrated that three compounds were active against AChE and four were active against BChE. Among them, 1 and 2 were active against both cholinesterases, but 2 was a stronger inhibitor, as its IC50 value was one-fifth that of 1.

Insight into various essential structural elements and modes of ACh processing can be obtained from the X-ray structures of AChE cocrystalized with its ligands. The presence of a narrow but long hydrophobic gorge, 20 Å deep inside, with a catalytic triad consisting of Glu334, His447, and Ser203, is an astounding structural feature of AChE.[33] In addition, a peripheral anionic site (PAS), which comprises the other set of aromatic residues (namely, Asp74, Tyr72, Tyr124, Tyr286, and Tyr341) at the rim of the gorge, provides a binding site for allosteric inhibition.[34] So far, two mechanisms of AChE inhibition have been proposed, besides an exceptional case of backdoor inhibition, in which inhibitors induce a conformational change in Trp86 at the catalytic triad at a depth of 20 Å and interfere with the opening and closing of the product-releasing channel.[35] The two well-explained AChE inhibition mechanisms are as follows: (a) competitive inhibition via interaction with the catalytic triad inside the gorge at a depth of 20 Å and (b) allosteric inhibition via π–cation or π–π interactions with PAS, thus blocking the entry of the substrate into the gorge or preventing the exit of products.[36]

In our docking study of AChE, 2 exhibited hydrophobic interactions with Tyr70, Trp279, Trp84, Phe330, Tyr334, and His440 and H-bond interactions with Phe288, Arg289, and Tyr121. Trp84 is the main binding residue of the catalytic anionic site in the vicinity of the catalytic triad, while Tyr70, Tyr121, Tyr334, and Trp279 are the main peripheral anionic sites at the gorge rim.[37] Thus, from our kinetics and docking results, we can conclude that 2 allosterically inhibited AChE through hydrophobic interactions with Tyr70, Tyr121, Trp84, Tyr334, and Trp279, while other residues stabilized this inhibitor at the binding site. The two extra H-bond interacting residues of 2 (Arg289 and Tyr121) in addition to the common residue Phe288 might explain why 2 had a higher binding affinity than that of the reference inhibitor donepezil.

Similarly, at the BChE active site, 2 bound with Tyr332 within the anionic hinge that prevents choline from entering the enzyme cavity. A recent study by Murakawa et al.[38] revealed that Phe329 is a common binding site for active BChE inhibitors. The authors also described that Asp70 in the hinge and Gly116 within the oxyanion cavity form another active site for BChE where inhibitors may bind. Our docking results also revealed hydrophobic interactions of 2 with Phe329 and Asp70.

The cholinesterases AChE and BChE share 65% amino acid sequence homology but differ in their substrate specificity and susceptibility to various inhibitors. The anionic site of AChE, comprising Trp84, Tyr130, Phe330, and Phe331, is responsible for substrate binding with π–cation interactions. However, in BChE, the key tryptophan residue (Trp82) is conserved, while Phe330 is replaced by Ala328. Thus, inhibitors that interact strongly with Phe330 have a higher affinity for AChE than for BChE, due to the replacement of Phe330 by Ala328 in the latter.[39] However, in our study, 2 interacted with Phe330 at the anionic site of AChE and with Ala328 in BChE and was further stabilized by other hydrophobic interactions, thereby demonstrating itself as a potent allosteric inhibitor of both cholinesterases.

Among various secretases, β-secretase is a prime enzyme catalyzing the rate-limiting step in the production of Aβ, a principal pathological component of AD. Although all of the tested arylbenzofurans were potent inhibitors of BACE1, we further evaluated 2 for its enzyme kinetics and mechanism via in silico docking, considering its submicromolar IC50 value. The numerous H-bond interacting residues and the low binding energy of 2 demonstrated its high binding affinity for the active site cavity. Various X-ray crystallography studies have revealed the structures of numerous BACE1 complexes,[40−42] indicating that BACE1 is covered by a flexible antiparallel β-hairpin (flap) that controls substrate access to the active site and determines the geometry of the catalytic process; the inhibitor-bound form is tightly packed in a closed conformation. However, the apo (substrate-free) structure of BACE1 is in an open conformation, indicating the necessity of a conformational change upon inhibitor or substrate binding.[43]

Ligand 2 exhibited hydrophobic interactions with allosteric residues Val69, Tyr71, Trp76, Phe108, and Ile126 of BACE1, without involving the catalytic dyad. A study by Kumar et al.[44] revealed that a catalytic Asp dyad is responsible for the cleavage of APP by an acid–base mechanism. Since BACE1 is a membrane-bound aspartyl protease with a catalytic Asp dyad, direct inhibition via these aspartyl residues (Asp32 and Asp228) may have deleterious effects.[45] This emphasizes the need to discover allosteric inhibitors that bind to regions other than the conserved active site of BACE1. Ligand 2 formed a dual hydrophobic and hydrophilic interaction with the allosteric Tyr198 residue of the enzyme. This accumulation of evidence, together with the allosteric inhibition of BACE1 by 2 through allosteric residues Asn37, Tyr198, Trp76, Val69, and Phe108, demonstrates the mechanism whereby 2 prevented APP cleavage. The additional interactions of 2 with Ser36 and Tyr71 enhanced its binding affinity and stability compared to those of sinensetin, the reference allosteric inhibitor.

The other important and emerging hypothesis besides the cholinergic and amyloid hypotheses is the tau or GSK-3 hypothesis of AD. According to this hypothesis, enhanced Aβ production and memory impairment are due to the overexpression of GSK-3. GSK-3 is a kinase protein and a promising drug target in AD, due to its pathophysiological involvement in tau protein phosphorylation or hyperphosphorylation. The inhibition of GSK-3 prevents neuronal destruction resulting from tau pathology in AD. In the present study, all of the tested arylbenzofurans potently inhibited GSK-3β, with 50% inhibitory concentrations less than 10 μM. Although 1 and 4 were stronger inhibitors of GSK-3β, 2 was selected for enzyme inhibition and molecular modeling studies, as it had the lowest IC50 values against the cholinesterase and BACE1 enzymes. In GSK-3β enzyme kinetics studies, 2 competed with both the substrate and ATP and was characterized as a competitive inhibitor. An X-ray crystallographic structure of GSK-3β (PDB ID: 1PYX) was selected to take into consideration the potential induced-fit effects upon ligand binding, as this structure corresponds to the native enzyme for substrate recognition.[24] The docking results of 1PYX demonstrated that −OH groups of 2 formed H-bonds with GSK-3β residues Gln89, Asp90, and Lys94 within the substrate pocket. Similarly, within the ATP-binding site, the two −OH groups established H-bond interactions with Asp133 and Val135 and an aromatic ring formed a π–cation interaction with MG1002. These residues were previously reported as backbone residues for ATP-competitive inhibitors of GSK-3β.[46] In Aβ-aggregation assays employing the ThT fluorescence method, 2 showed a concentration-dependent inhibition of self-induced and AChE-induced aggregation of Aβ1–40 and Aβ1–42 superior to reference controls. Since AChE promotes aggregation of Aβ-peptide fragments by forming a complex with the growing fibrils,[47] inhibitory effect of 2 on Aβ-aggregation represents a preventive action on senile plaque formation.

Despite the fact that AD is a progressive neurodegenerative disease, there is no clinically accepted therapy to cure it. The US Food and Drug Administration (FDA) has approved drugs (donepezil, galantamine, memantine, and rivastigmine) that at their best provide only marginal benefits, thus demonstrating the urgent need to explore other molecular entities as future drug candidates for AD. A closeup view of the molecular framework of these FDA-approved drugs indicates that most of them have heterocyclic ring systems, highlighting the potential of heterocyclic compounds as potent drug candidates. Likewise, heterocyclic compounds have been evaluated for the treatment of neurological and psychological disorders and benzofurans and indole derivatives have been suggested for the treatment of AD.[48] In support of this, our ADME and druglikeness prediction results demonstrated that the compounds we tested were druglike molecules with excellent human intestinal absorption (>90%) and 100% plasma protein binding. Thus, these arylbenzofuran scaffolds from the root bark of M. alba Linn, especially 2, could be promising drug candidates in the management of AD. However, detailed in vivo studies will be of the utmost importance to support our findings. Similarly, in recent studies by Zhang et al.,[29] δ-secretase was found to cleave APP and regulate the pathogenesis of AD, and its inhibition improved cognitive function in mouse models of AD. Thus, it would be worth examining the effects of these compounds on δ-secretase.

In the present study, we identified and characterized arylbenzofurans (1–5) from M. alba Linn as triple inhibitors of cholinesterase, BACE1, and GSK-3β by in vitro enzyme kinetics assays and in silico molecular simulations. The triple inhibition mode of this class of compounds, especially 2, explains their multiple-targeting properties. Due to their unique inhibition mechanism, these compounds have specific advantages over the reported single and/or dual inhibitors, and the results of ADME and druglikeness predictions further support the possibility of drug development. Still, in vivo studies remain to be conducted.

Materials and Methods

Chemicals and Reagents

Acetylcholine iodide (ACh), butyrylthiocholine chloride (BCh), electric eel AChE (EC 3.1.1.7), horse serum BChE (EC 3.1.1.8), 5,5′-dithiobis[2-nitrobenzoic acid], quercetin, and berberine were purchased from E. Merck, Fluka, or Sigma, unless otherwise stated. BACE1 fluorescence resonance energy transfer (FRET) assay kits (β-secretase) were purchased from Pan Vera Co. (Madison, WI). Human recombinant GSK-3β was obtained from ProSpec (ProSpec-Tany TechnoGene Ltd., Ness-Ziona, Israel). The GSM substrate mimicking muscle glycogen synthase was purchased from Merck Millipore (Millipore Corporation, CA). Kinase-Glo kits were obtained from Promega (Promega Corporation, Madison, WI). Adenosine 5-triphosphate (ATP) disodium salt hydrate, ammonium hydroxide, 4-(2-hydroxyethyl) piperazine-1-ethanesulfonic acid (HEPES), ammonium acetate, formic acid, ethylene glycol-bis(aminoethylether)-N,N,N,N-tetra acetic acid tetra sodium salt (EGTA), dimethyl sulfoxide (DMSO), ethylenediaminetetraacetic acid (EDTA), amyloid-β protein fragments (Aβ1–40/42), thioflavin T, 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP), and magnesium acetate tetrahydrate were purchased from Sigma-Aldrich (St. Louis, MO). All chemicals and solvents from commercial sources were of reagent grade.

Isolation of Compounds

A MeOH extract (1.2 kg) of the root bark of M. alba Linn was suspended in distilled water (H2O) and successively partitioned with n-hexane, CH2Cl2, and EtOAc. The CH2Cl2 fraction (215.2 g) was then subjected to silica gel CC and eluted with a gradient of 0–100% MeOH in CHCl3 to afford 15 fractions (F1–F15). Fraction F2 (5.5 g) was chromatographed by silica gel CC with n-hexane–acetone (100:0 to 0:100, gradient, v/v) as the eluent to afford 10 subfractions (F2.1–F2.10). Subfraction F2.7 (225 mg) was passed over a silica gel column with n-hexane–EtOAc (10:1, v/v) as the mobile phase, yielding compounds 1 (8 mg) and 2 (11 mg). Subfraction F2.9 (280 mg) was chromatographed on a reversed-phase C18 (RP-C18) silica gel column with acetonitrile–H2O (1.5:1, v/v) as the mobile phase to yield compound 3 (24 mg). Fraction F3 (6.5 g) was subjected to repeated silica gel CC (CH2Cl2–MeOH, 100:0 to 0:100, gradient, v/v) and purification on a RP-C18 silica gel column with 70% MeOH, yielding compounds 4 (58 mg) and 5 (23 mg). These compounds were identified as mulberrofuran D (1), mulberrofuran D2 (2), mulberrofuran H (3), morusalfuran B (4), and sanggenofuran A (5) through the comparison of the spectral data with those reported in the literature.[49,50] Compounds were purified on a reversed-phase C18 (RP-C18) gel CC using different mobile phases as follows: 60% acetonitrile for compounds 1–3 and 70% MeOH for compounds 4 and 5. The purity of the isolated compounds was estimated to be >98% based on the proton and carbon NMR spectra.

Cholinesterase Enzyme Inhibition Assay

The in vitro ChE-inhibitory potential of the arylbenzofurans isolated from the root bark of M. alba Linn was determined through our previously reported procedure.[51] ACh and BCh were used as substrates to examine the inhibition of AChE and BChE, respectively. The test samples and the positive control (berberine) were each dissolved in 10% analytical-grade DMSO. The reactions were conducted in triplicate in a 96-well microplate. A VersaMax microplate reader (Molecular Devices, Sunnyvale, CA) was employed to monitor the hydrolysis of ACh or BCh at 412 nm. The percent inhibition of ChE was calculated as (1 – S/E) × 100, where S and E are the enzyme activities in the presence or absence of inhibitors, respectively. The IC50 values were calculated from the log dose–inhibition curve.

BACE1 Enzyme Inhibition Assay

Inhibition of the BACE1 enzyme was evaluated in vitro with human recombinant β-secretase and a BACE1 fluorescence resonance energy transfer (FRET) assay kit. All of the experimental conditions and procedures were similar to those in our previous report.[52] Quercetin was used as a reference control. The percent (%) inhibition of BACE1 was calculated from the following equation: % inhibition = [1 – (A60 – A0)/(C60 – C0)] × 100%, where A0 and C0 are the initial fluorescences of the test and control samples, respectively, and A60 and C60 are the final (after 60 min) fluorescences of the test and control samples, respectively. The IC50 values were calculated from the log dose–inhibition curve.

GSK-3β Enzyme Inhibition Assay

The inhibition of the GSK-3β enzyme was evaluated in vitro with hGSK-3β and a Kinase-Glo reagent kit. The procedure reported by Baki et al.[53] was used, with some modifications.[54] To initiate the reaction, 5 μL of the sample, 5 μL of ATP (1 μM final concentration), 5 μL of 50 μM GSM and 5 μL of 20 ng GSK-3β were mixed per well in a 384-well black plate. The test samples were serially diluted in an assay buffer containing 50 mM HEPES, 1 mM EDTA, 1 mM EGTA, and 15 mM magnesium acetate at pH 7.5. The final concentration of DMSO in each reaction mixture was less than 5%. The reaction mixture was then incubated at 37 °C for 30 min, followed by the addition of 20 μL of the Kinase-Glo reagent as a stop solution. Once the reaction was terminated, glow-type luminescence was recorded. The method was validated with luteolin as a reference compound causing maximum inhibition, and the maximum enzyme activity was determined in the absence of inhibitors (5 μL of sample replaced by 5 μL of buffer).

Enzyme Kinetics

The assay method described in the enzyme inhibition section was used to evaluate the kinetics of AChE inhibition by mulberrofuran D2 (2.5, 5, and 10 μM) in the presence of varying concentrations of ACh (0.15, 0.3, and 0.6 nM). For BChE enzyme kinetics, mulberrofuran D2 was used at concentrations of 0.2, 1, and 5 μM, with different concentrations of BCh (0.3, 0.6, and 0.9 nM). Similarly, mulberrofuran D2 (0.5, 1, and 2 μM) was reacted with different substrate concentrations (250, 375, and 750 nM) following the assay method for BACE1 enzyme inhibition described above. The assay method in the enzyme inhibition section was used to evaluate the enzyme kinetics of GSK-3β with varying concentrations of ATP (2, 1, and 0.5 μM) or substrate (50, 25, and 12.5 μM). Mulberrofuran D2 was used at concentrations of 2.5, 5, and 10 μM. The double reciprocal plots of the enzyme kinetic data were used to derive a Lineweaver–Burk plot. Ki values were calculated from the Dixon plots.

Docking Studies

Molecular docking analysis was carried out in AutoDock 4.2.[55] X-ray crystallographic structures of AChE–donepezil (PDB ID: 1EVE),[56] BChE–tacrine (PDB ID: 4BDS),[57] and BACE1–QUD (PDB ID: 2WJO)[58] complexes were obtained from the RCSB Protein Data Bank (PDB) website (http://www.rcsb.org/) at resolutions of 2.50, 2.10, and 2.50 Å, respectively. The X-ray crystallographic structure of human GSK-3β complexed with a known ATP-competitive inhibitor, phosphoaminophosphonic acid–adenylate ester (AMP–PNP), was obtained from the RCSB Protein Data Bank (PDB ID: 1PYX) at a resolution of 2.4 Å.[59] Rotatable bonds in the inhibitors and positive controls were assigned by AutoDockTools (ADT). The structures of mulberrofuran D2, 1-[2-(2,4-dichlorophenyl)-2-phenylmethoxyethyl]imidazole (5i), and N′-dodecanoyl-1-ethyl-4-hydroxy-2-oxo-1,2-dihydroquinoline-3-carbohydrazide (VP0.7) were created in Chem3D Pro software (v12.0, CambridgeSoft Inc., Cambridge, MA). The structures of galantamine, sinensetin, and andrographolide were obtained from NCBI PubChem, with the respective CIDs of 9651, 145659, and 5318517. Energy minimization was carried out for each ligand by means of a molecular mechanics 2 (MM2) force field. AutoDock 4.2 was used for docking simulations, and grid maps that included the catalytic site and allosteric site were generated in the AutoGrid program. The protocol for rigid and flexible ligand docking comprised 15 independent genetic algorithms. Docking results were visualized and analyzed with PyMOL (v1.7.4, Schrödinger, LLC, Cambridge, MA) and Discovery Studio (v17.1, Accelrys, San Diego, CA).

Self-Induced Aβ1–42 Aggregation Inhibition Assay

Mulberrofuran D2, which showed the highest potency for AChE inhibition, was selected to assess its ability to inhibit self-induced Aβ1–42 aggregations via the thioflavin T (ThT) fluorescence method.[60] Briefly, Aβ1–42 lyophilized from 2 mg/mL hexafluoro-2-propanol (HFIP) solution was dissolved in DMSO to obtain 0.5 mM stock solution. Then, the stock solution of Aβ1–42 was diluted by 10-fold in 0.05 M phosphate buffer solution containing 0.1 M NaCl (pH 7.5). To 30 μL of the amyloid solution, 30 μL of different concentrations of mulberrofuran D2 (5, 10, and 20 μM) and/or curcumin (20 μM) was added. After 24 h incubation at 37 °C, 240 μL of a 5 μM thioflavin T in 50 mM glycine–NaOH buffer (pH 8.5) was added, and after 5 min, the fluorescence emission at 490 nm (ex = 446 nm) was measured. The fluorescence intensities were compared and the percent inhibition due to the presence of the inhibitor was calculated by the following formula: 100 – (Fi/Fc × 100), where Fi and Fc are the fluorescence intensities obtained for Aβ1–42 in the presence and in the absence of inhibitors, respectively.

AChE-Induced Aβ1–40 Aggregation Inhibition Assay

For the AChE-induced Aβ1–40 aggregation inhibition assay, aliquots of 2 μL of Aβ1–40 were incubated for 24 h at room temperature in 0.215 mM sodium phosphate buffer (pH 8.0) at a final concentration of 230 μM. For coincubation experiments, 16 μL of AChE (final concentration of 2.3 μM, Aβ1–40/AChE molar ratio of 100:1) and AChE in the presence of 2 μL of the tested inhibitor (final concentration 5, 10, and 20 μM) in 0.215 M sodium phosphate buffer (pH 8.0) solutions were added. Blanks containing Aβ1–40 alone, human AChE alone, and Aβ1–40 plus tested inhibitors in 0.215 sodium phosphate buffer (pH 8.0) were prepared. After incubation, 180 μL of 5 μM thioflavin T in 50 mM glycine–NaOH buffer (pH 8.5) was added. Each assay was run in triplicate, and donepezil was used as a reference drug. The detection method was the same as above.

Druglikeness and ADME Prediction

Druglikeness predictions were carried out with PreADMET. This web-based server can be used to calculate absorption, distribution, metabolism, and excretion (ADME) data and build a druglikeness library in silico.

Statistical Analysis

All of the results are presented as the mean ± standard deviation (SD) of triplicate experiments. One-way analysis of variance and Duncan’s test (v23.0, Systat Inc., Evanston, IL) were used to calculate the statistical significance. A p-value <0.05 was considered significant.