Inhibition of β-site amyloid precursor protein cleaving enzyme 1 and cholinesterases by pterosins via a specific structure−activity relationship with a strong BBB permeability

We extracted 15 pterosin derivatives from Pteridium aquilinum that inhibited β-site amyloid precursor protein cleaving enzyme 1 (BACE1) and cholinesterases involved in the pathogenesis of Alzheimer’s disease (AD). (2R)-Pterosin B inhibited BACE1, acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) with an IC50 of 29.6, 16.2 and 48.1 µM, respectively. The Ki values and binding energies (kcal/mol) between pterosins and BACE1, AChE, and BChE corresponded to the respective IC50 values. (2R)-Pterosin B was a noncompetitive inhibitor against human BACE1 and BChE as well as a mixed-type inhibitor against AChE, binding to the active sites of the corresponding enzymes. Molecular docking simulation of mixed-type and noncompetitive inhibitors for BACE1, AChE, and BChE indicated novel binding site-directed inhibition of the enzymes by pterosins and the structure−activity relationship. (2R)-Pterosin B exhibited a strong BBB permeability with an effective permeability (Pe) of 60.3×10−6 cm/s on PAMPA-BBB. (2R)-Pterosin B and (2R,3 R)-pteroside C significantly decreased the secretion of Aβ peptides from neuroblastoma cells that overexpressed human β-amyloid precursor protein at 500 μM. Conclusively, our study suggested that several pterosins are potential scaffolds for multitarget-directed ligands (MTDLs) for AD therapeutics.


Introduction
Alzheimer's disease (AD) is an age-related neurodegenerative disorder with characteristic clinical and pathological features, which are associated with the loss of neurons in certain brain areas, leading to memory impairment, cognitive dysfunction, behavioral disturbances, deficits in activities of daily living, and eventually death 1,2 . The symptoms of AD include dementia, apraxia, aphasia, depression, a short attention span, visuospatial navigation deficits, anxiety, and delusions. AD affects up to 5% of individuals aged above 65 years and increases to 20% in those above 80 years of age. Approximately 46.8 million individuals suffer from AD worldwide, accounting for an estimated annual societal economic cost of $818 billion in 2015, which is expected to increase to $1 trillion in 2018 and $2 trillion in 2030 3 .
Reflecting the multifactorial and complex etiology of AD, the histopathological hallmarks, such as amyloid βprotein (Aβ) deposits 4 ; dysfunctional signaling of acetylcholine (ACh) in certain areas of the brain 5 ; τ protein neurofibrillary tangles 6 ; metabolic pathways, such as those involving cAMP-responsive element binding protein (CREB) 7 ; oxidative stress 8 ; and inflammation 9 , appear to play significant roles. The two most common hypotheses include amyloid and cholinergic hypotheses.
The amyloid hypothesis suggests that the accumulation and oligomerization of Aβ peptide in the brain plays a critical role in AD pathogenesis 10 . It has been clearly established that the overproduction of Aβ by the aspartic protease BACE1 and subsequent oligomerization result in toxic amyloid oligomers inducing neurodegeneration. BACE1 cleaves β-amyloid precursor protein (APP) and forms approximately 90% of the Aβ peptides 11,12 . Endogenous BACE1 activity is increased in the brain of patients with sporadic AD 13 . In addition, emerging evidence shows significant elevation of BACE1 in the presence of other AD risk factors, such as traumatic brain injury, stroke, and cardiovascular events, which suggests that BACE1 is a stress-response protein and its activities increase during AD risk factor-related events 14 . However, all previously discovered natural and synthetic BACE1 inhibitors have failed clinical trials as therapeutic candidates for AD mainly due to adverse effects, such as ocular toxicity 15 and poor BBB penetration with low brain:plasma concentration ratios 16 . Currently, no BACE1 inhibitors have been approved for the clinical treatment of AD.
The cholinergic hypothesis proposes a massive loss of cholinergic neurons as a downstream phenotypic consequence in the pathogenesis of AD 1 . A decreased level of ACh, a neurotransmitter, in the brain plays a critical role in the progression of AD. ACh plays an important role in the cognitive mechanism and is hydrolyzed by AChE and BChE, leading to the loss of cognitive functions. With the progression of AD, the activity of AChE decreases, while that of BChE increases to compensate for the loss of AChE in an attempt to modulate ACh levels in cholinergic neurons and augment learning 17,18 . Both enzymes are therapeutic targets to combat cognitive deficits at different stages of AD with AChE in the early stage and BChE in the later stages 19 . In particular, AChE inhibitors, such as E2020 (donepezil), have become the drug of choice in the clinical management of AD. However, E2020 lacks BACE1-inhibitory activity and fails to provide pathogenetic treatment along with various adverse side effects 20 .
After several decades of research efforts, the treatment of AD continues to face a significant unmet need, with therapies based largely on the cholinesterase inhibitors rivastigmine, E2020, and galantamine, with the only exception an NMDA (N-methyl-D-aspartate) receptor antagonist, memantine 21 . However, these drugs only relieve AD symptoms for a short period of time without reversing disease progression. In recent years, the need for disease-modifying drugs for AD has been addressed via new approaches to design structures with different targets involved in the pathogenesis of AD, including multitargetdirected ligands (MTDLs), in line with the multifactorial and complex etiology of AD to produce the desired therapeutic efficacy 22,23 . Thus, MTDLs, such as dual-and multiacting anti-AD agents, represent a promising strategy for the treatment of AD, engaging different targets simultaneously as a hybrid.
Ferns belong to the botanical group Pteridophyta. Several ferns have been used in ethnopharmacy to treat various illnesses 24 . Alkaloids present in Huperzia serrata, particularly huperzine A, B, and R and 8-β phlegmariurine B, have been used for AD therapy as cholinesterase inhibitors 25 . Pteridium aquilinum (P. aquilinum) is distributed globally and has been widely consumed traditionally as foods in Korea and Japan. Pterosins, the major components of P. aquilinum, have been reported to be nontoxic to humans, although some of its components, such as the unstable glucoside ptaquiloside, are carcinogenic 26,27 . Recently, pterosin B was shown to prevent chondrocyte hypertrophy and osteoarthritis in mice at least in part via CREB activation 28 . In addition, pterosin derivatives, particularly pterosin A, showed pharmacological properties relevant to the prevention and treatment of diabetes and obesity in mice 29,30 . However, the anti-AD activity of pterosins has not been reported to date.
The present study was performed to characterize the anti-AD potential of pterosin derivatives by investigating their activities in vitro to inhibit BACE1, AChE and BChE as well as BBB permeability. In addition, the structure −activity relationship (SAR) was analyzed using molecular docking simulations to determine the molecular interactions between pterosins and the active sites of BACE1, AChE and BChE. Scientific (Ebsdorfergrund, Germany). β-secretase inhibitor III (β-SI) was purchased from Calbiochem (Darmstadt, Germany). The protease inhibitor mixture (a mixture of AEBSF, pepstatin A, E-64, bestain, leupeptin, and aprotinin) was purchased from Sigma-Aldrich (St. Louis, MO, USA). Enzyme-linked immunosorbent assay (ELISA) kits for Aβ40 and Aβ42 were obtained from IBL (Kunma, Japan). All chemicals and solvents used were reagent grade, purchased from commercial sources, and used as received.

Plant material
P. aquilinum was collected from the mountainous regions of Gapyeong-gun, Gyeonggido, Korea, and authenticated by Prof. Ki-Joong Kim (Korea University, Seoul). A voucher specimen was deposited in the laboratory of Prof. Gil Hong Park.
Extraction, fractionation, and identification of pterosin derivatives from P. aquilinum Pterosin derivatives were isolated from P. aquilinum using hot water followed by various chromatographic methods (Fig. 1). The hot water extract obtained by refluxing 250 g of the whole plants of P. aquilinum in 1.5 L H 2 O for 24 h in a steamer (OSK-2002, Red Ginseng Doctor, Well sosanaTM, Daewoong Pharmaceutical, Seoul, Korea) was initially partitioned with an equal volume of EtOAc and subsequently with n-BuOH. The repeated column chromatographic separation of the EtOAc-soluble fraction and the n-BuOH-soluble fraction resulted in the isolation of 15 pterosin derivatives with purities greater than 97%. The structures of the isolated compounds were identified via analysis of the spectral data, including MS, 1Dand 2D-NMR, as well as comparison of their data with the published values, including heteronuclear single quantum correlation and heteronuclear multiple bond correlation. The pterosin compounds were dissolved in Dimethyl Sulfoxide (DMSO) for use in the experiment.

In vitro BACE1 enzyme assay
The BACE1 FRET assay was carried out according to the manufacturer's instructions with slight modifications as previously described 31 . The assay was performed in a 381-well black plate using a multi-pipette. Readings were obtained two times, the first time at 0 min and the second time after 60 min incubation at room temperature (26°C) and stopping the reaction with BACE1 stop buffer, using a spectrofluorometer (Gemini EM; Molecular Devices, San Jose, CA, USA) at 545 nm (excitation) and 590 nm (emission). Quercetin was used as the positive control. % Inhibition calculation: [1 -(60 min value -0 min value/ control)] × 100.

In vitro cholinesterase enzyme assay
The inhibitory activities of pterosin derivatives against cholinesterases were measured using the method developed by Ellman et al. 32 . The reaction mixture contained 140 μL of sodium phosphate buffer (pH 8.0), 20 μL of test sample solution (final concentration of the compound 100 μM), and 20 μL of AChE or BChE solution, mixed and incubated for 15 min at room temperature. All test samples and the positive control (berberine) were diluted or dissolved in 10% analytical grade ethanol. The reactions were initiated following the addition of 10 μL of DTNB and 10 μL of acetyl thiocholine iodide for the AChE assay or butyryl thiocholine chloride for the BChE assay in 96-well microplates, which were incubated for 15 min. The hydrolysis of acetyl thiocholine iodide or butyryl thiocholine chloride was monitored by tracking the formation of yellow 5-thio-2-nitrobenzoate anion resulting from the reaction between DTNB and thiocholine released by the enzyme, using a microplate spectrophotometer (Molecular Devices, Sunnyvale, CA, USA) at 412 nm. The percent inhibition was calculated as (1−S/ E) × 100, where S and E represent the enzyme activities with and without the test sample, respectively.

Kinetic parameters of BACE1, AChE and BChE inhibition by pterosin derivatives and the inhibition mechanism
To determine the K i and the mode of enzymatic inhibition of the most active compounds against BACE1, AChE, and BChE, Dixon and Lineweaver−Burk plots were employed. The K i was determined by interpretation of the Dixon plot, with the value of the x-axis intercept taken as −K i . The Dixon plot is a graphical method [plot of 1/enzyme velocity (1/V) against inhibitor concentration (I)] for the determination of K i and the type of enzyme inhibition for the enzyme-inhibitor complex. Dixon plots for enzyme inhibition by pterosins were tested in the presence of different substrate concentrations: 150, 250, and 750 nM for BACE1, 100, 300, and 600 µM for AChE, and 200, 400, and 800 µM for BChE. Lineweaver−Burk plots were analyzed in the presence of different inhibitor concentrations: 0-125 µM for BACE1, 0-100 µM for AChE, and 0-150 µM for BChE.

Molecular docking
Docking studies were performed on BACE1, AChE, and BChE to understand the inhibition profile of the pterosin derivatives. The X-ray crystal structures of human BACE1 complexed with 2-amino-3-{(1r)-1-cyclohexyl-2-[(cyclohexylcarbonyl)amino]ethyl}-6-phenoxyquinazolin-3-ium (QUD) (PDB code: 2WJO) 33 , human AChE complexed with E2020 (PDB code: 4EY7) 34 and human BChE complexed with N-{[(3R)-1-(2,3-dihydro-1H-inden-2-yl) piperidin-3-yl]methyl}-N-(2-methoxyethyl)naphthalene-2-carboxamide (3F9) (PDB code: 4TPK) 35  Protein structures were held rigid, whereas ligands were treated as fully flexible during docking. Prior to docking, the protein and ligand structures were processed with AutoDock Tools (ADT) 1.5.6 37 . The cocrystallized ligands and water molecules were removed from the original PDBs. Polar hydrogen atoms were merged, and Kollman and Gasteiger charges were assigned to the protein structures. Gasteiger charges were added by default to the ligands for docking calculations. The number of rotatable bonds was set, and all torsions were allowed to rotate. For each enzyme, the AutoGrid program was employed to create a grid box of size 60 × 60 × 60 Å 3 with 0.375 Å spacing. The center of the grid was defined according to a recent study, which reported different sites for catalytic, mixed-type and noncompetitive BACE1, AChE, and BChE inhibitors 38 . Lamarckian genetic algorithm was used for the conformational search 39 . The docking protocol consisted of 100 runs, 25×10 5 energy evaluations and 27,000 iterations. Other docking parameters were set to the default values. Docked poses were selected on the basis of scoring functions and protein−ligand interactions. Binding interaction figures were generated using Discovery Studio 2017 R2.
PAMPA-BBB procedure PAMPA (parallel artificial membrane permeation assay) was used as a high-throughput assay to predict BBB permeation 40 . Porcine polar brain lipid (PBL) was used as an artificial membrane to predict BBB permeation. Initially, the test compound was dissolved in DMSO at 5 mg/ mL. This compound stock solution (10 µL) was diluted 200-fold in a universal buffer at pH 7.4 and mixed with a multi 8-channel pipette to obtain the secondary stock solution (final concentration 25 µg/mL). The secondary stock solution (200 µL) was placed in the donor wells. The filter membrane was coated with PBL in dodecane (4 mL volume of 20 mg/mL PBL in dodecane), and 5 µL of BBB-1 lipid solution was spread on PBL by pipette. The acceptor well was filled with 200 mL of Acceptor Sink Buffer. The acceptor filter plate was cautiously positioned on the donor plate to assemble a "sandwich" (comprising an aqueous acceptor, artificial lipid membrane and aqueous donor with test compound on the top, middle and bottom, respectively). The test compound diffuses from the donor well to the acceptor well through the lipid membrane. After 4 h of incubation at pH 7.4 and 25°C, the concentration of the compound in the acceptor, donor, reference and blank wells was estimated with a UV plate reader, Epoch Microplate Spectrophotometer (Bio-Tek Instruments, Inc., Winooski, VT, USA). The effective permeability (P e ) of the compounds was calculated using Pion PAMPA Explorer software. Samples were analyzed in quadruplets, and the average of the four runs was reported. To monitor the consistency of the analysis set, quality control standards were run with each sample set. Verapamil was employed as a high permeability standard (P e = 16 × 10 −6 cm/s).

Aβ ELISA
Aβ40 and Aβ42 ELISA was performed according to the manufacturer's instructions. The assay was conducted in a microplate coated with anti-human Aβ 35-40 mouse IgG (1A10) or anti-human Aβ 38-42 mouse IgG (44A3), respectively. The microplate was added with 100 μL conditioned media, incubated overnight at 4°C, and then washed with washing buffer. Horseradish peroxidaseconjugated anti-human Aβ 11-28 mouse IgG (82E1) was added into each well, incubated at 4°C for 1 h, and washed with washing buffer. 3,3′,5,5′-Tetramethylbenzidine as a chromogenic substrate was added to each well and incubated for 30 min. The reaction was stopped by the addition of 100 μL stop solution, and the absorbance at 450 nm was measured using a model 680 microplate reader (Bio-Rad, Hercules, CA).

3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay
Cytotoxicity was assessed using the MTT assay 41 . Briefly, the cells were seeded into a 96-well plate at a density of 1×10 5 cells per well in 100 µL of corresponding media and incubated at 37°C in an incubator under 5% CO 2 tension for 24 h. The culture media were then replaced with 100 µL of fresh serum-free media in the presence of varied concentrations of pterosin derivatives, and cells were incubated for an additional 24 h. The control was treated with an equal amount of DMSO present in the assay for a 5 mM concentration. Then, 100 µL of MTT solution (0.5 mg/mL in Phosphate Buffered Saline (PBS)) was added to each culture. To measure the proportion of surviving cells, the media were replaced with 100 µL of DMSO 2 h after the administration of MTT solution. The absorbance was measured at 570 nm using a spectrophotometric plate reader (Molecular Devices, Sunnyvale, CA, USA).

Statistical analysis
All results were presented as the mean ± standard deviation (SD) of triplicate or quadruple experiments. Statistical evaluation was performed by one-way analysis of variance (ANOVA). The analysis was performed using Graph Pad Prism 5.01 (Graph Pad Software, Inc., La Jolla, CA, USA). P < 0.05 was considered statistically significant.

Inhibitory activity of pterosin derivatives against BACE1, AChE, and BChE
To evaluate the anti-AD potential, the inhibitory activity of each pterosin compound against BACE1 and cholinesterases was evaluated by respective in vitro inhibition assays ( Table 1). All tested pterosin derivatives showed concentration-dependent inhibitory activities against BACE1 with a range of IC 50 values (half-maximum inhibitory concentration) of 9.74-94.4 μM, with the exception of (2S)-pterosin A and (2S)-pteroside P that were inactive at the concentrations tested, compared with the IC 50 of quercetin used as the positive control, which was 18.8 μM.
Collectively, most of the pterosin derivatives tested exhibited significant inhibitory activities against BACE1, AChE, and BChE simultaneously. The presence of the additional 2-hydroxymethyl-tetrahydro-pyran-3,4,5-triol group as in pteroside derivatives significantly increased the inhibitory activities against the enzymes. Moreover, the presence of the additional hydroxymethyl group at position-2 of the indanone ring of (2R)-pterosin B as in (2S)-pterosin A or the methyl group as in (3R)-pterosin D and pterosin Z decreased the inhibitory activities against the enzymes. In particular, the presence of the hydroxymethyl group at position-5 of the indanone ring as in (2S)-pterosin P decreased the inhibitory activity against BACE1.

Kinetic parameters of enzyme inhibition by pterosin derivatives
In an attempt to explain the mode of enzymatic inhibition of pterosin derivatives, we performed a kinetic analysis of BACE1 and cholinesterases for representative inhibitors ( Table 2 34 , and BChE complexed with 3F9 (PDB code: 4TPK, resolution: 2.70 Å) 35 were selected for docking. Initially, QUD, E2020, and 3F9 were extracted from crystal structures and redocked into the active sites of BACE1, AChE, and BChE, respectively. Subsequently, (2R,3R)-pteroside C, (3S)-pteroside D, (2R,3R)-pterosin C, (2R)-pteroside B, (2S,3R)-pterosin C and (2R)-pterosin B with the known mechanism of inhibition against BACE1, AChE and BChE were docked to determine their SAR. The docking results are summarized in Table 3. The SAR of the selected mixedtype and noncompetitive BACE1, AChE, and BChE inhibitors enabled the evaluation of novel binding site-directed inhibition of the enzymes by pterosins.
Our docking mode of E2020 was consistent with the experimentally determined binding mode previously reported with recombinant human AChE (rhAChE) (Supplementary Information 3) 34 . The root-mean-square deviation (RMSD) between the crystal and docked conformations of E2020 was 0.54 Å, which suggested the reliability of our docking setup in reproducing the experimental binding mode. In addition, the docked mode of E2020 led to a similar interaction as that of rhAChE-E2020. In our study, water molecules were removed from the crystal structure during docking; therefore, water-mediated interactions were not analyzed in the present study. Similarly, the docked modes of QUD and 3F9 were consistent with the available experimental data for BACE1 33 and BChE 35 , respectively ( Supplementary Information 3). The RMSDs between the crystal and docked conformations of QUD and 3F9 were 0.46 and 0.60 Å, respectively. Further, the binding sites of pterosin inhibitors were in agreement with a previous docking study that involved BACE1, AChE, and BChE 38 . However, the study used Tetronarce californica AChE (PDB code: 1ACJ), which contains slightly different residue numbers than human AChE due to variations in their sequences.

BACE1 docking
Based on the inhibition type and activity, (2R,3R)pteroside C and (2R)-pteroside B were selected to demonstrate the docked modes of mixed-type and noncompetitive BACE1 inhibitors, respectively. Figure 3a, b displays the docking models of (2R,3R)-pteroside C and (2R)-pteroside B, respectively. The interactions of the docked compounds inside the active site of BACE1 are shown in Fig. 4.

AChE docking
(2R)-Pteroside B and (2R,3R)-pterosin C were selected as representatives to demonstrate the docked modes of mixed-type and noncompetitive AChE inhibitors, respectively, due to their activities and type of AChE inhibition. Figure 5a, b illustrates the docking models of (2R)-pteroside B and (2R,3R)-pterosin C, respectively. The interactions of the docked compounds inside the active site of AChE are displayed in Fig. 6.
The docked pose of E2020 demonstrated a B.E. of −10.28 kcal/mol. As illustrated in Fig. 6a, the CO group of the indanone ring formed a hydrogen bond with the NH group of Phe295 at a distance of 1.70 Å. Trp286 and Tyr341 were involved in π-sigma interactions, whereas Trp86, Trp286, Tyr337, Phe338, and Tyr341 mediated hydrophobic interactions.  (Fig. 6b). The OH group of hydroxymethyl and the 3-OH group of the tetrahydro-pyrantriol ring displayed hydrogen bonds with the CO group of Asn87 at distances of 2.23 and 2.28 Å, respectively. Further, the 4-OH group formed a hydrogen bond with the CO group of Trp86 at a distance of 2.13 Å. In the case of (2R,3R)-pteroside C (Fig. 6c), the methyl group at position-2 of the indanone ring did not show a hydrophobic interaction with Tyr341 and thus exhibited comparatively lower activity than (2R)-pteroside B. However, Tyr341 maintained the hydrophobic interaction with the other part of the indanone ring as shown in that of (2R)pteroside B. The higher activity of (2S,3R)-pterosin C than (2R)-pterosin B was attributed to the existence of an additional OH group at position-3 of the indanone ring, which formed a hydrogen bond with Phe295 at a distance of 1.82 Å (Fig. 6d). The remaining interactions were comparable to (2R)-pterosin B interactions. As shown in Fig. 6e, the OH group of the hydroxyethyl group at position-6 of the indanone ring formed a hydrogen bond with Tyr124 at a distance of 2.48 Å. Tyr124, Trp286, Tyr337, Phe338, and Tyr341 contributed to the hydrophobic interactions. Compared with (2R)-pteroside B and (2R,3R)-pteroside C, (3S)-pteroside D exhibited dissimilar binding interactions due to the presence of the 2,2dimethyl group at the indanone ring (Fig. 6f). The 2,2dimethyl group significantly contributed to the distinct docked pose of (3S)-pteroside D. The 5-OH group of the tetrahydro-pyran-triol ring formed a hydrogen bond with Asp74 at a distance of 1.96 Å. The CO group of the indanone ring showed a hydrogen bond with Tyr72 at a distance of 2.97 Å. These interactions accounted for the low activity of (3S)-pteroside D.
The docked pose of (2R,3R)-pterosin C (noncompetitive AChE inhibitor) is displayed in Fig. 6g. (2R,3R)-Pterosin C (IC 50 = 23.2 µM) demonstrated a B.E. of −5.01 kcal/mol. The binding pocket of (2R,3R)-pterosin C comprised Tyr72, Trp286, Ser293, Phe297, and Tyr341, with two hydrogen bond interactions. One of the hydrogen bonds was formed between the 3-OH group of the indanone ring and the CO group of Ser293 at a distance of 2.03 Å. The second hydrogen bond was observed between the OH group of the hydroxyethyl group present at position-6 of the indanone ring and Tyr72 at a distance of 1.94 Å. Residues such as Tyr72, Trp286, Phe297, and Tyr341 participated in hydrophobic interactions.

BChE docking
Considering the activity levels and type of BChE inhibition, (2R,3R)-pteroside C and (2S,3R)-pterosin C were selected to demonstrate the docked modes of mixed-type  Fig. 8b, the 3-OH group of the indanone ring of (2R,3R)-pteroside C showed two hydrogen bonds with Gly78 and Tyr440 at distances of 2.87 and 2.89 Å, respectively. In the case of (3S)-pteroside D, the 2,2dimethyl group at the indanone ring affected the binding interactions of the 3-OH group (Fig. 8c). The 3-OH group formed only one hydrogen bond with Trp82 at a distance of 2.94 Å, which resulted in a comparatively lower activity of (3S)-pteroside D than (2R,3R)-pteroside C. As shown in Fig. 8d, (2R,3R)-pterosin C failed to produce a hydrogen bond with Ser287 due to the absence of a 2-hydroxymethyl-tetrahydro-pyran-3,4,5-triol group. Consequently, it yielded a lower activity than (2R,3R)-pteroside C. The lack of the OH group at position-3 of the indanone ring was found to be responsible for the very low activity of (2R)-pteroside B (Fig. 8e), which failed to form hydrogen bonds with Gly78 and Tyr440 similar to (2R,3R)-pteroside C.
The docked poses of (2S,3R)-pterosin C and (2R)-pterosin B (noncompetitive BChE inhibitors) are shown in Fig. 8f, g, respectively. These docked poses were contained in the cavity enclosed by Gly283, Leu286, Val288, Phe357, and Asn397. As per their activity levels, (2S,3R)-pterosin C (IC 50 = 44.3 µM) and (2R)-pterosin B (IC 50 = 48.1 µM) exhibited a B.E. of −5.40 and −5.06 kcal/mol, respectively. (2S,3 R)-Pterosin C was more potent than (2R)-pterosin B because of the presence of an additional OH group at position-3 of the indanone ring, which formed a hydrogen bond with the CO group of Asn397 at a distance of 2.05 Å (Fig. 8f). Other interactions were found to be similar to (2R)-pterosin B. As shown in Fig. 8g, the OH group in the hydroxyethyl group at position-6 of the indanone ring formed a hydrogen bond with Gly283 at a distance of 2.40 Å. The residues Leu286, Val288, and Phe357 participated in hydrophobic interactions.
Mixed-type inhibitors bind to both the free enzyme and the enzyme-substrate complex, which indicates that these compounds may bind to the catalytic site of each corresponding enzyme. Noncompetitive inhibitors bind to the allosteric site of the free enzyme or enzyme-substrate complex. A recent study suggested that competitive, mixed-type and noncompetitive inhibitors occupy different sites in the binding pockets of BACE1, AChE, and BChE 38 . During docking for the evaluation of the inhibitory mechanism of pterosin derivatives, the binding sites of the compounds were defined according to their type of inhibition. The docking results indicated that the binding sites of mixed-type and noncompetitive inhibitors for BACE1, AChE, and BChE partially overlap each other at each corresponding active site and were consistent with a previous report 38 . PAMPA-BBB, an in vitro artificial membrane permeability assay for the BBB, is one of the most reliable physicochemical screening tools in the early stage discovery of CNS-targeted drugs 40 . The PAMPA-BBB system models the transcellular passive diffusion of chemicals across the BBB and measures strictly passive transport mechanisms via an artificial lipid membrane on effective permeability (P e , cm/s). On the basis of the pattern established for BBB permeation prediction, compounds were classified into (i) "CNS+" (high BBB permeation predicted); P e (10 −6 cm/s) > 4.00, (ii) "CNS−" (low BBB permeation predicted); P e (10 −6 cm/s) < 2.00, and (iii) "CNS+/−" (BBB permeation uncertain); P e (10 −6 cm/s) from 4.00 to 2.00. Accordingly, (2R)-pterosin B, (2S)-pterosin P, and (2S)-pterosin A exhibited high BBB permeation with P e values of 60.3 × 10 −6 cm/s, 7.92 × 10 −6 cm/s, and 6.26 × 10 −6 cm/s, respectively (Table 4, Supplementary Information 4). The P e value of (2R)pterosin B was 1.7-fold higher than that of the CNS drug verapamil (P e = 34.6 × 10 −6 cm/s), which was used for the positive control. (2S,3R)-Pterosin C and (2R,3R)-pterosin C showed uncertain BBB permeation with P e values of 2.34 and 1.98, respectively. (2R,3R)-Pteroside C, (3S)pteroside D, and (2R)-pteroside B, which showed the most potent BACE1-and cholinesterase-inhibitory activities among the pterosin derivatives tested, exhibited a very low BBB permeability. The existence of the 2-hydroxymethyl-tetrahydro-pyran-3,4,5-triol group as in pteroside derivatives resulted in a remarkable decrease in the BBB permeability. Compared with (2R)-pterosin B, the additional presence of the OH group at position-3 of the indanone ring as in pterosin C, the hydroxymethyl group at position-2 of the indanone ring as in (2S)-pterosin A or the hydroxymethyl group at position-5 of the indanone ring as in (2S)-pterosin P also significantly reduced the BBB permeability. Considering an exceptionally high BBB permeability and the significant inhibition of BACE1, AChE, and BChE, (2R)-pterosin B may have the potential to exhibit a strong anti-AD activity. To investigate the function of (2R)-pterosin B and (2R,3R)-pteroside C in decreasing the excretion of Aβ from neuronal cells, we used a murine neuroblastoma cell line that stably overexpresses human APPswe. The cell line is a cellular model of AD characterized by the excessive secretion of Aβ40 and Aβ42. Toxic amyloid oligomers are formed from the two isoforms of Aβ peptide with different lengths. Aβ40 is the most abundant Aβ isoform in the brain 41 , while Aβ42 significantly increases with certain forms of AD 42 . Sandwich ELISA of Aβ40 showed that (2R)-pterosin B significantly reduced the amount of Aβ40 peptide secreted from the neuroblastoma cells into media up to 50% at 500 μM (P < 0.01) (Fig. 9a). Similarly, the secretion of Aβ42 peptide by the neuroblastoma cells significantly decreased in the presence of 500 µM of (2R,3R)-pteroside C (P < 0.05) (Fig. 9b). In conclusion, (2R)-pterosin B and (2R,3R)pteroside C significantly decreased the secretion of Aβ peptides from neuroblastoma cells at a concentration of 500 μM.

Discussion
The current study is the first investigation to evaluate pterosin derivatives as a series of novel scaffolds to provide MTDLs, which displayed significant inhibitory activities against BACE1, AChE, and BChE simultaneously in a dose-dependent manner. The molecular structures of the enzyme/inhibitor complexes were further predicted to simulate binding between the pterosin derivatives and BACE1, AChE, and BChE. These predictions facilitated the evaluation of binding site-directed inhibition of the enzymes. Furthermore, the docking results explained the SAR of selected mixed-type and noncompetitive BACE1, AChE, and BChE inhibitors. Both the in vitro evaluation and molecular docking data clearly indicated that specific pterosin compounds are potential lead compounds for the development of novel MTDLs for AD therapeutics via the Aβ and cholinesterase pathways.
In particular, (2R)-pterosin B exhibited the highest BBB permeation among commercially available drugs currently used for CNS diseases based on P e representing effective BBB permeability in vitro 40 . Therapeutic candidates for CNS diseases, including AD, must be able to permeate the BBB. Only compounds with a molecular weight smaller than 400-700 Da and lipophilicity have been shown to cross the BBB 44 . Compared with (2R)pterosin B, the additional 2-hydroxymethyl-tetrahydropyran-3,4,5-triol group as in pteroside derivatives, the presence of the 3-OH group as in pterosin C or the hydroxymethyl group at position-2 or 5 of the indanone ring as in (2S)-pterosin A or (2S)-pterosin P remarkably decreased the effective BBB permeability.
The present experiment used quercetin and berberine as positive controls for the BACE1 and cholinesterase assays, respectively. The flavonoid quercetin is a BACE1 inhibitor that exhibits novel pharmacophore features for AD 45 , as well as an inhibitor for AChE (IC 50 = 19.8 µM) 46 with a negligible influence on BChE activity 47 ; moreover, it effectively ameliorated AD pathology and tauopathy and protected cognitive and emotional functions in aged (21-24 months old) triple transgenic mouse models of AD (3xTg-AD) treated with 25 mg/kg via i.p. injection every 48 h for 3 months 48 . Quercetin was reported to exhibit effective BBB permeability, with a log P e of The main disadvantage of MTDLs as hybrid molecules is their high molecular weight, which conforms to Lipinski's rule of five 56 . Thus, the development of merged ligands with a small molar mass similar to a single compound is more difficult and tedious 16,22,23 . Despite the obstacles, several successful drug scaffolds that contain merged ligands addressing cholinesterases and Aβ oligomerization simultaneously have recently been reported. Bis-tacrines bearing a peptide moiety that specifically prevents surface sites of human AChE from Aβ binding were developed as MTDLs to combat AD 57 . Accordingly, the hybrid compounds bind the catalytic and peripheral sites of human AChE and act as potent inhibitors of both the catalytic and noncatalytic functions of AChE, interfering with the Aβ self-oligomerization process via its peripheral anionic site 58,59 . Further, the inhibition of both human BChE and fatty acid amide hydrolase (FAAH) as dual cholinesterase-FAAH inhibitors that target both cholinergic and endocannabinoid signaling resulted in improved neuronal transmission by ACh and a simultaneous reduction of neuroinflammation 60 . (2R)-Pterosin B exhibited significant inhibitory activities against BACE1, AChE, and BChE along with a remarkably high in vitro BBB permeability, which suggests the potential as a scaffold for MTDLs to suppress Aβ production and simultaneously enhance cholinesterase-mediated cognitive functions. Intriguingly, (2R)-pterosin B was demonstrated Fig. 9 The effects of (2R)-pterosin B and (2R,3R)-pteroside C on the cellular secretion of Aβ peptides. a Effect of (2R)-pterosin B on the secretion of Aβ peptides. APPswe-secreting neuroblastoma cells were treated with 12, 60, 250, or 500 µM of (2R)-pterosin B for 24 h, and conditioned media were collected in the presence of protease inhibitor. β-SI (10 µM) was used as positive control. Negative control included cells cultured in the absence of test compounds. Quantitative analysis of secreted Aβ40 and Aβ42 in the conditioned media was performed using sandwich ELISA. The means ± SD from three independent experiments are shown. The secreted Aβ40 peptides significantly decreased in the presence of 500 µM (2R)pterosin B. CON negative control, β-SI β-secretase inhibitor III, Aβ β-amyloid. *P < 0.05, **P < 0.01. b Effect of (2R,3R)-pteroside C on the secretion of Aβ peptides. APPswe-secreting neuroblastoma cells were treated with 12, 60, 125, 250, or 500 µM of (2R,3R)-pteroside C for 24 h, and conditioned media were collected in the presence of protease inhibitor. β-SI (10 µM) was used as positive control. Negative control included cells cultured in the absence of test compounds. Quantitative analysis of secreted Aβ40 and Aβ42 in the conditioned media was performed using sandwich ELISA. The means ± SD from three independent experiments are shown. The secreted Aβ42 peptide significantly decreased in the presence of 500 µM (2R,3R)pteroside C. CON negative control, β-SI β-secretase inhibitor III, Aβ β-amyloid. *P < 0.05 to activate CREB signaling to protect cartilage against osteoarthritic change 28 . Moreover, the metabolic pathways involving cyclic nucleotides, including cAMP and cGMP, play an important role in the pathogenesis of AD via CREB activation, which has been regarded as a molecular switch required for learning, memory and neuronal survival 7,61-63 . The role of (2R)-pterosin B in the activation of CREB signaling and its effects on cognitive functions and neuroprotection in the brain merit evaluation.
Importantly, the present cytotoxicity test based on the MTT assay supported prior observations that pterosin derivatives were nontoxic to humans and not carcinogenic 26 , although some are cytotoxic to cancer cells, such as HeLa cells 64 . Further, previous animal experiments demonstrated the biosafety of pterosin derivatives. A mouse model of osteoarthritis was injected with a large amount of pterosin B (15 μL of 900 μM pterosin B) into the intra-articular space of the knee joint three times per week for an extended period of 8-13 weeks, without adverse effects, while ameliorating osteoarthritis 28 . Further, pterosin A, administered orally in large amounts of 100 mg/kg/day for 4 weeks, displayed no significant adverse effects, while effectively improving glucose intolerance and insulin resistance in various mouse models of diabetes 29,30 . Additionally, the diabetes animal experiment demonstrated the good oral bioavailability of pterosin A. In our MTT-based cytotoxic test, (2R)-pterosin B exhibited biosafety comparable to (2S)-pterosin A (Supporting Information 5). Intriguingly, diabetes and insulin resistance have emerged as significant risk factors aggravating AD 65 . In this context, the antidiabetes potential of pterosin-based anti-AD agents is worth investigation.
Considerable genetic and molecular evidence supports that BACE1 is the rate-limiting enzyme in the production of Aβ and the crucial pathogenetic events that lead to AD 66 . Moreover, several AChE and BChE inhibitors, such as E2020, were approved for the treatment of mild to moderate AD, and the use of these drugs is beneficial in the treatment of AD symptoms [19][20][21] . Although the in vitro inhibitory capacities of (2R)-pterosin B against BACE1 (K i = 38.3 µM), AChE (IC 50 = 16.2 µM) and BChE (IC 50 = 48.1 µM) are substantially lower than those of a BACE1 inhibitor, AZD3293 (K i = 0.4 nM) in a clinical study 67 or the AChE inhibitor E2020 (IC 50 = 6.7 nM) approved for clinical use that share some structural features with (2R)pterosin B 68 , we suggest the possibility that pterosin may be administered up to dosages to reach effective concentrations to inhibit BACE1, AChE, and BChE in the brain of AD patients because of its biosafety and significantly high BBB permeability.
In conclusion, the currently available therapies for AD are only symptomatic. The MTDL approach is a very promising strategy for the treatment of AD due to its multifactorial etiology 22,23 . The structures of several pterosins suggest a potential biological and therapeutic role as a scaffold to provide new MTDLs for AD. Further, our results suggested that the molecular features of enzyme binding and BBB permeability of pterosins facilitate the structural modifications for the design of compounds with improved enzyme-inhibitory activities and BBB permeability for consideration as novel therapeutics for AD.