Computationally Assisted Lead Optimization of Novel Potent and Selective MAO-B Inhibitors

A series of dietary flavonoid acacetin 7-O-methyl ether derivatives were computationally designed aiming to improve the selectivity and potency profiles against monoamine oxidase (MAO) B. The designed compounds were evaluated for their potential to inhibit human MAO-A and -B. Compounds 1c, 2c, 3c, and 4c were the most potent with a Ki of 37 to 68 nM against MAO-B. Compounds 1c–4c displayed more than a thousand-fold selectivity index towards MAO-B compared with MAO-A. Moreover, compounds 1c and 2c showed reversible inhibition of MAO-B. These results provide a basis for further studies on the potential application of these modified flavonoids for the treatment of Parkinson’s Disease and other neurological disorders.


Introduction
Parkinson's Disease (PD) is one of the most prevalent neurodegenerative disorders [1], affecting over four million people worldwide. The causes of PD remain unknown. However, the disease is known to arise from interaction between environmental and genetic factors, resulting in progressive degeneration of neurons in the brain. Despite decades of research, the molecular pathways involved in neurodegeneration, the nature of the interaction, and the identity of the factors are poorly understood [2]. PD is characterized by the progressive loss of dopaminergic neurons. Most of the PD cases are sporadic, but rare familial forms have also been recognized. The common pathways underlying the pathogenesis of PD include oxidative stress, mitochondrial quality control, and protein degradation processes. Understanding the possible reasons behind these common processes provides various targets, which are therapeutically relevant to the discovery of disease-modifying treatments [1].

Evaluation of Pan Assay Interference Compounds (PAINS)
All the designed MAO-B inhibitors were analyzed per the recently published editorial [14] using the False Positive Remover [15] and the ZINC Pattern Identifier [16]. All compounds passed the filter and were not reported as covalent inhibitors or potential PAINS by any of these algorithms.

Monoamine Oxidase Inhibition Assay and Determination of IC 50 Values for Synthesized Compounds
To investigate the inhibitory effect of the acacetin 7-O-methyl ether analogs on human recombinant MAO-A and MAO-B, the kynuramine oxidation deamination assay was performed in 384-well plates as previously reported, with minor modifications [17]. A fixed single concentration of kynuramine substrate and varying concentrations of the test inhibitor were used to determine the IC 50 values. The kynuramine concentrations for MAO-A and -B were 80 µM and 50 µM, respectively. These concentrations of kynuramine were twice the apparent K M value for substrate binding [9,18]. The acacetin 7-O-methyl ether analogs were tested at the concentrations ranging from 0.001 µM to 100 µM for MAO-A and -B inhibition assays. Enzymatic reactions were performed in 50 µL of the assay mixture containing 0.1 M potassium phosphate buffer, pH 7.4. The inhibitors and acacetin 7-O-methyl ether analogs were dissolved in DMSO, diluted in the buffer solution, and pre-incubated at 37 • C for 10 min (the final concentration of DMSO was <1.0%). Reactions were initiated by the addition of 12.50 µL of MAO-A (to 5 µg/mL) or -B (to 10 µg/mL). The plate was incubated for 20 min at 37 • C, and the enzymatic reaction was terminated by the addition of 18.8 µL of 2N NaOH. Formation of the enzyme product, 4-hydroxyquinoline, was measured fluorometrically using a SpectraMax M5 fluorescence plate reader (Molecular Devices, Sunnyvale, CA, USA) with excitation (320 nm) and emission (380 nm) wavelengths and the Soft Max Pro program. The inhibition effects of enzyme activity were calculated as percent of product formation compared to the corresponding control (enzyme-substrate reaction) without inhibitors. Controls included the assays where the enzyme or the substrate was added after terminating the reaction to determine the interference with the fluorescence measurements. The IC 50 values for MAO-A and -B inhibition were calculated from the concentration-dependent inhibition curves using XLFit ® software.
The enzyme assay was performed at a fixed concentration of the substrate kynuramine (80 µM for MAO-A and 50 µM for MAO-B) and varying concentrations of the inhibitor/test analogs (0.01 µM to 100 µM) for MAO-A and (0.01 µM to 100 µM) for MAO-B. The doseresponse curves were generated using Microsoft ® Excel and the IC 50 values were calculated using XLfit software [9].

Enzyme Kinetics, Mechanism Studies, Analysis of Reversibility, and Binding Assays of Acacetin 7-O-Methyl Ether Analogs
For determination of the binding affinity of the inhibitor (Ki) with MAO-B, the enzyme assays were carried out at different concentrations of kynuramine substrate (1.90 µM to 500 µM) and varying concentrations of the test compound. Acacetin 7-O-methyl ether analogs: 1c, 2c, 3c, and 4c were tested at 0.015-0.500 µM for MAO-B to determine the K m and V max values of the enzymes in the presence of the inhibitor. The controls without inhibitor were also run simultaneously. Three sets of assays were performed at varying concentrations of the substrate for each experiment: one control without inhibitor and the others at two fixed concentrations of the inhibitor. The data were analyzed by double reciprocal Lineweaver-Burk plots to determine Ki (i.e., inhibition/binding affinity) values and the kinetic data, namely Km, Vmax, and Ki values, were computed by SigmaPlot 12 [9].
Most of the inhibitors produce inhibition of the target enzyme activity through formation of an enzyme-inhibitor complex. Formation of the enzyme-inhibitor complex may be accelerated in the presence of a high concentration of the inhibitor. The reversibility/irreversibility of binding of acacetin 7-O-methyl ether analogs with MAO-B was determined from the formation of the complex by incubating the enzyme with a high concentration of the inhibitor followed by extensive equilibrium-dialysis of the enzymeinhibitor complex. Recovery of the catalytic activity of the enzyme was determined by assay of the enzyme activity before and after equilibrium-dialysis. MAO-B (0.2 mg/mL protein) enzyme was incubated with each analog, namely acacetin (0.5 µM), 1c (1.5 µM), 2c (1.5 µM), 3c (1.5 µM), 4c (1.5 µM), and deprenyl (0.5 µM), in a total volume of 1 mL of potassium phosphate buffer (100 mM, pH 7.4). After 20 min of incubation at 37 • C, the reaction was stopped by chilling the tubes in the ice bath. All the samples were dialyzed against potassium phosphate buffer (25 mM; pH 7.4) at 4 • C for 14-18 h (with three buffer changes). The control enzyme (without inhibitor) was also run simultaneously using the same procedure and the activity of the enzyme was determined before and after the dialysis.
To analyze if the nature of MAO-B binding to the inhibitor was time-dependent, the enzyme was pre-incubated for different time periods (0-15 min) with the inhibitor concentrations, which exhibited approximately 70-80% inhibition. The inhibitor concentrations used to test time-dependent inhibition were acacetin (0.080 µM), 1c (0.100 µM), 2c (0.300 µM), 3c (0.300 µM), 4c (0.400 µM), and deprenyl (0.080 µM) with MAO-B (12.5 µg/mL). The controls, without inhibitors, were also run simultaneously. Activities of the enzyme were determined as described above, and the percentage of the remaining enzyme activity was plotted against the pre-incubation time to determine time-dependent inhibition.

Molecular Modeling Studies
The crystal structures of MAO-A (PDB ID: 2Z5Y with an atomic resolution of 2.17 Å) and MAO-B (PDB ID: 4A79 with an atomic resolution of 1.89 Å) were downloaded from the protein databank (www.rcsb.org, accessed on 12 January 2017). The protein structural files were prepared using the protein preparation wizard of Schrödinger suite [19]. Standard procedures were followed for protein preparation through assigning bond orders, adding hydrogens and removing original ones, creating zero-order bonds to metals, filling in the missing side chains, and deleting waters beyond 5 Å from the ligand. The ionization states were generated at pH 7.4 using Epik [20][21][22]. The next step was sampling water orientations and checking for possible protonation and tautomerization states at pH 7.4. Water molecules that did not display at least two hydrogen bonds with no-water residues were deleted. Finally, the structure was relaxed to remove all atomic clashes using OPLS3 force field [23].
The receptor grid was generated using Glide [24][25][26][27]. The binding pocket was identified by the ligand coordinates. Because of the importance of the FAD cofactor in directing the right pose in the substrate binding pocket [28][29][30] . The soft docking protocol was followed by softening the potential of nonpolar parts of the receptor (i.e., decreasing the van der Waals radius of nonpolar atoms to accommodate ligands). Ligands were prepared for docking using LigPrep [31]. This step was performed for acacetin 7-O-methyl ether and for the designed compounds as well. Possible protonation and tautomerization states were generated at a pH of 7.4 and the specified chiralities were retained. The lowest energy conformer of each ligand was kept. OPLS force field was employed in the preparation step. Acacetin 7-O-methyl ether and the designed compounds were docked into the receptor grids of MAO-A/B. The soft docking protocol [32] was followed by using flexible ligand sampling and the soft receptor grid. Standard docking precision was used to generate the best five poses for acacetin 7-O-methyl ether and the best pose for the other compounds.
The two most diverse poses of acacetin 7-O-methyl ether were used for molecular dynamics simulations and mapping of the active site water molecules. Each complex was prepared and solvated in an orthorhombic box of the TIP4P water solvation model using Desmond system builder [33,34]. The net charge on the protein was neutralized by adding the appropriate number of sodium ions. MD simulations were conducted using OPLS3 force field. The FAD cofactor was included in the MD studies and it was prepared by OPLS3 force field and charges. The cofactor showed a state penalty of 0.65 kcal/mol and a hydrogen bond count of 16 at a negative charge of −2. Desmond's minimization algorithm was used to relax the MD system for 2000 iterations. Desmond's algorithm involves a series of energy minimizations and short MD simulations to further relax the solvated protein before the MD production step, including two minimization steps and 4 short MD simulation steps for 12 ps, 12 ps, 24 ps, and 24 ps at 10 K, 10 K, 300 K, and 300 K, respectively, using the NPT ensemble. For the production step of the MD simulations, the NPT ensemble, the Nose-Hoover chain thermostat, and the Martyna-Tobias-Klein barostat [33,34] were used. The coordinates were saved at intervals of 25 ps, ending with 8000 frames, and the MD simulations were sampled over 200 ns. The active site water structure and thermodynamic stability were examined for the protein-ligand complexes using SZMAP [35][36][37]. Positive (unfavorable) and negative (favorable) free energy hydration sites of the protein-ligand complexes were identified by implementing the semi-continuous solvation model of SZMAP to map the surface of the protein-binding pocket.
flavone luteolin [43], while methoxylation of the same positions gained MAO-B selectivity in a manner similar to vetulin [11] and diosmetin [42], respectively. Monosubstitution of ring B on the flavone scaffold at C-4′ resulted in no selectivity towards any MAO as seen with apigenin [44], genkwanin [40], and acacetin [9]. Methylation of the C-7 position increased selectivity towards MAO-B in a manner similar to acacetin-7-O-methyl ether [11]. Based on this SAR analysis, oxygenation at the C-5 and C-7 positions of ring A is significant for MAO inhibition. Oxygenation of the C-3 position resulted in MAO-A selectivity, while substitution on C-4′ resulted in MAO-B selectivity.   The lack of a double bond between the C-2 and C-3 positions resulted in the depletion of MAO inhibition as seen in the flavanone naringenin [38]. The absence of substituents on ring B showed a slight preference for MAO-A in a manner similar to the flavone chrysin [39]. Monosubstitution on ring B of the flavonol scaffold decreased the SI of MAO-A versus MAO-B, improving the activity towards MAO-A in a manner comparable to the flavonols rhamnocitrin [40], 4 -O-methyl kaempferol, and 3, 4 -O-methyl kaempferol [41], while disubstitution of ring B resulted in the loss of MAO-A activity as seen with the flavonols quercetin [39] and isorhamnetin [42]. Dioxygenation at positions C-3 and C-4 on ring B resulted in the loss of selectivity towards MAO-B in a manner similar to the flavone luteolin [43], while methoxylation of the same positions gained MAO-B selectivity in a manner similar to vetulin [11] and diosmetin [42], respectively. Monosubstitution of ring B on the flavone scaffold at C-4 resulted in no selectivity towards any MAO as seen with apigenin [44], genkwanin [40], and acacetin [9]. Methylation of the C-7 position increased selectivity towards MAO-B in a manner similar to acacetin-7-O-methyl ether [11]. Based on this SAR analysis, oxygenation at the C-5 and C-7 positions of ring A is significant for MAO inhibition. Oxygenation of the C-3 position resulted in MAO-A selectivity, while substitution on C-4 resulted in MAO-B selectivity.

Molecular Dynamics (MD) Simulations of the Binding Modes of Acacetin and Acacetin 7-O-Methyl Ether
Acacetin 7-O-methyl ether was recently discovered to be a selective MAO-B inhibitor with 0.198 µM inhibition of MAO-B versus MAO-A (>100 µM) [11]. The binding characteristics of acacetin 7-O-methyl ether were studied to understand the basis of target selectivity. Molecular constraints were imposed on the system during docking simulations based on our previous investigations [9]. The chemistry of the acacetin 7-O-methyl ether was mapped with the binding site environment using molecular dynamics (MD) simulations and thermodynamic calculations to better define the most abundant binding mode. The same analysis was carried out for acacetin ( Figures S1-S73 Figure 3A) and in the second pose, the methoxyphenyl group is facing the cofactor (referred to as pose 2, Figure 3B). Pose 1 has a better docking score compared with pose 2. The average root mean square deviation (RMSD) of pose 1 was 3.2 Å with respect to 1.6 Å for the protein backbone over the course of the MD simulations.  Figure 3A) and in the second pose, the methoxyphenyl group is facing the cofactor (referred to as pose 2, Figure 3B). Pose 1 has a better docking score compared with pose 2. The average root mean square deviation (RMSD) of pose 1 was 3.2 Å with respect to 1.6 Å for the protein backbone over the course of the MD simulations.
The two possible binding modes for acacetin 7-O-methyl ether in the case of MAO-B were visualized and are shown in Figure 3. The first pose of acacetin 7-O-methyl ether has benzopyranone facing the cofactor (referred to as pose 1, Figure 3A) and in the second pose, the methoxyphenyl group is facing the cofactor (referred to as pose 2, Figure 3B). Pose 1 has a better docking score compared with pose 2. The average root mean square deviation (RMSD) of pose 1 was 3.2 Å with respect to 1.6 Å for the protein backbone over the course of the MD simulations.  showed hydrogen bonds through water molecules to Phe 99 (less than 5%), Gly 101 (11%), and Pro 102 (~5%). More hydrophobic contacts were observed for pose 1, while more hydrogen bonds and water bridges were traced for pose 2 ( Figure S75).
In conclusion, water molecules are more involved in the binding of pose 2 ( Figures  4-6). Several other important protein-ligand contacts for both poses were sampled over the course of MD simulations, including hydrophobic contacts, π-π stacking, hydrogen bonding, and interactions through water bridges with Leu 171, Leu 316, Tyr 326, Tyr 398, and Tyr 435 ( Figure 4).

Active Site Hydration of MAO-B: MD Simulations, Thermodynamics, and Ligand Designs
There are 10 crystallographic water molecules in the binding pocket of MAO-B, which would have significant effects on ligand recognition and ligand design. The exact location of active site water molecules was confirmed by thermodynamic calculations to assess their contributions in ligand binding and target selectivity. Analysis of the MD simulations revealed that an average of three water molecules were fluctuating near the cofactor. In the case of pose 1, the ligand interacts with the hydrogen bond to Pro 102 through a water molecule for~5% of the MD simulation time. On the other hand, pose 2 showed hydrogen bonds through water molecules to Phe 99 (less than 5%), Gly 101 (11%), and Pro 102 (~5%). More hydrophobic contacts were observed for pose 1, while more hydrogen bonds and water bridges were traced for pose 2 ( Figure S75).
In conclusion, water molecules are more involved in the binding of pose 2 (Figures 4-6). Several other important protein-ligand contacts for both poses were sampled over the course of MD simulations, including hydrophobic contacts, π-π stacking, hydrogen bond-ing, and interactions through water bridges with Leu 171, Leu 316, Tyr 326, Tyr 398, and Tyr 435 (Figure 4).   The active site water molecules ( Figure 5A) were also investigated by water mapping calculations to compute their free energy and predict their effect on target selectivity. The binding pocket of MAO-B has a hydration shell near the cofactor (which is similar to what was found during the MD) that overlaps well with the hydroxyl group of the benzopyranone of pose 1 ( Figure 5B). Therefore, to design high-affinity ligands, we should keep the hydroxyl group or modify the structure to have more polarity at the same position. A hydrophobic region (a positive free energy region) was noticed at the opposite site of FAD; and it covers the methoxyphenyl group of acacetin 7-O-methyl ether. To improve ligand binding, the polarity should be decreased at this position. The ligand chemistry of pose 2 did not fit well with the active site hydration map, while pose 1 showed a perfect match ( Figure 5C).
Based on the interaction profile of acacetin 7-O-methyl ether as well as the thermodynamic properties of MAO-B's active site, eight analogs were virtually designed to have small ether and/or amine modifications at C-4′ of the acacetin 7-O-methyl ether scaffold ( Figure 6) and docked into the active sites of MAO-A and -B using Glide (Table S1). The designed compounds were evaluated for their absorption, distribution, metabolism, and excretion (ADME) properties. They did not show any violation of the rules of five or three; they have a good balance of lipophilicity/hydrophilicity, and are expected to be orally bioavailable (Table S2). The compounds mapped well with the hydrophobic regions of the binding pocket while their polar chemistry fits with the small hydrophilic regions. Consequently, syntheses of five small ether derivatives were planned: propyl- The active site water molecules ( Figure 5A) were also investigated by water mapping calculations to compute their free energy and predict their effect on target selectivity.
The binding pocket of MAO-B has a hydration shell near the cofactor (which is similar to what was found during the MD) that overlaps well with the hydroxyl group of the benzopyranone of pose 1 ( Figure 5B). Therefore, to design high-affinity ligands, we should keep the hydroxyl group or modify the structure to have more polarity at the same position. A hydrophobic region (a positive free energy region) was noticed at the opposite site of FAD; and it covers the methoxyphenyl group of acacetin 7-O-methyl ether. To improve ligand binding, the polarity should be decreased at this position. The ligand chemistry of pose 2 did not fit well with the active site hydration map, while pose 1 showed a perfect match ( Figure 5C).
Based on the interaction profile of acacetin 7-O-methyl ether as well as the thermodynamic properties of MAO-B's active site, eight analogs were virtually designed to have small ether and/or amine modifications at C-4 of the acacetin 7-O-methyl ether scaffold ( Figure 6) and docked into the active sites of MAO-A and -B using Glide (Table S1).
The designed compounds were evaluated for their absorption, distribution, metabolism, and excretion (ADME) properties. They did not show any violation of the rules of five or three; they have a good balance of lipophilicity/hydrophilicity, and are expected to be orally bioavailable (Table S2). The compounds mapped well with the hydrophobic regions of the binding pocket while their polar chemistry fits with the small hydrophilic regions. Consequently, syntheses of five small ether derivatives were planned: propyl-ether (1c), isopropyl-ether (2c), isobutyl-ether (3c), propargylic-ether (4c), and methyl (cyclopropyl)ether (5c). The role of the bioisosteric change of nitrogen from oxygen was also scrutinized. Thus, propyl amine (6d), isopropyl amine (7d), and methyl (cyclopropyl)-amine (8d) were synthesized.
Compound 5 was prepared as seen in Scheme 2. 4-(cyclopropylmethoxy) benzaldehyde IV was synthesized by reacting (bromomethyl) cyclopropane III with 4-hydroxybenzaldehyde II in acetone and potassium carbonate. Chalcone 5a was prepared through Schmidt condensation by treatment of the cyclopropyl aldehyde IV with 2′-hydroxy-4′,6′dimethoxyacetophenone I and 50% NaOH in ethanol. The intermediate flavonoid 5b was obtained from the treatment of the chalcone with iodine crystals in a minimal amount of DMSO by undergoing cyclization. Several attempts to demethylate position C-5 were unsuccessful, thereby preventing the demethylation at C-5 and releasing the alkyl group to form the hydroxyl group at C-4′.

Synthesis of Modified Flavonoids 6-8
Compounds 6-8 were prepared as shown in Scheme 3 with an extra step involved in the formation of the brominated flavonoid at C-4′. Thus, the brominated chalcone 6a was produced by treating 2′-hydroxy-4′,6′-dimethoxyacetophenone I with 4-bromobenzaldehyde V through Schmidt condensation with 50% sodium hydroxide in ethanol. The brominated flavonoid 6b was prepared from the treatment of the chalcone with iodine crystals in a minimal amount of DMSO by cyclization. The flavonoids (6-8) c were prepared by treating compound 6b with sodium tert-butoxide, tris(dibenzylideneacetone) dipalladium, 1,1′-binaphthalene-2,2′-diylbis(diphenylphosphine), and the corresponding amines such as propylamine, isopropylamine, and (aminomethyl)cyclopropane in toluene. Finally, the selective demethylated flavonoids at position 5 (6-8) d were prepared with boron tribromide in dichloromethane (Scheme 3). With the exception of compound 5c, the evaluation of MAO-A and -B inhibition was performed on the final and intermediate compounds.   Table 2). The significant selectivity obtained for the oxygenated analogs clearly validated our design. Bioisosteric replacement of oxygen with nitrogen at position C-4′ diminished the selectivity of MAO-B considerably (see Table 2). The presence of the hydroxyl group at C-5 showed increased To obtain the target flavonoids (1-4) c, selective demethylation of position C-5 of (1-4) b was accomplished using boron tribromide (BBr 3 ) in CH 2 Cl 2 (Scheme 1).
Compound 5 was prepared as seen in Scheme 2. 4-(cyclopropylmethoxy) benzaldehyde IV was synthesized by reacting (bromomethyl) cyclopropane III with 4-hydroxybenzaldehyde II in acetone and potassium carbonate. Chalcone 5a was prepared through Schmidt condensation by treatment of the cyclopropyl aldehyde IV with 2 -hydroxy-4 ,6 -dimethoxyacet-ophenone I and 50% NaOH in ethanol. The intermediate flavonoid 5b was obtained from the treatment of the chalcone with iodine crystals in a minimal amount of DMSO by undergoing cyclization. Several attempts to demethylate position C-5 were unsuccessful, thereby preventing the demethylation at C-5 and releasing the alkyl group to form the hydroxyl group at C-4 .

Synthesis of Modified Flavonoids 6-8
Compounds 6-8 were prepared as shown in Scheme 3 with an extra step involved in the formation of the brominated flavonoid at C-4 . Thus, the brominated chalcone 6a was produced by treating 2 -hydroxy-4 ,6 -dimethoxyacetophenone I with 4-bromobenzaldehyde V through Schmidt condensation with 50% sodium hydroxide in ethanol. The brominated flavonoid 6b was prepared from the treatment of the chalcone with iodine crystals in a minimal amount of DMSO by cyclization. The flavonoids (6)(7)(8) c were prepared by treating compound 6b with sodium tert-butoxide, tris(dibenzylideneacetone) dipalladium, 1,1 -binaphthalene-2,2 -diylbis(diphenylphosphine), and the corresponding amines such as propylamine, isopropylamine, and (aminomethyl)cyclopropane in toluene. Finally, the selective demethylated flavonoids at position 5 (6-8) d were prepared with boron tribromide in dichloromethane (Scheme 3). With the exception of compound 5c, the evaluation of MAO-A and -B inhibition was performed on the final and intermediate compounds.  Table 2). The significant selectivity obtained for the oxygenated analogs clearly validated our design. Bioisosteric replacement of oxygen with nitrogen at position C-4 diminished the selectivity of MAO-B considerably (see Table 2). The presence of the hydroxyl group at C-5 showed increased selectivity towards MAO-B. Thus, the activity and selectivity are increased for the series as follows: hydroxyls at C-5 (1c-4c) > protected-methylated flavonoids at C-5 (1b-6b) > chalcones (1a-6a) > amine flavonoids at C-4 (6c,d-8c,d). Compounds (1)(2)(3)(4) c exhibited more than a thousand-fold SI (Table 2), equal to or more potent than safinamide (Table 2), and were considered for further studies to understand their mechanisms of inhibition. Figure S76 shows the inhibition dose-response for the potent analogs (1-4) c.  Table 3).

Evaluation of MAO-B Inhibition
Binding of modified flavonoids (1-4) c with human MAO-B affected K m (i.e., the affinity of the substrate for the enzyme) as well as V max (maximum enzyme activity) values, indicating the type of MAO-B inhibition by the analogs: flavonoid 1c (mixed/partially reversible), 2c (mixed/partially reversible), 3c (mixed/irreversible), and 4c (mixed/irreversible) (Table 3, Figure 7).  The binding characteristics of the modified flavonoids (1-4) c with MAO-B were examined by equilibrium dialysis to measure the dissociation of the enzyme-inhibitor complex ( Figure 8). MAO-B enzyme was incubated with the highest concentration of the analogs for 20 min at 37 °C to allow for binding of the inhibitor with the enzyme and formation of the enzyme-inhibitor complex. Then, the mixtures of the enzyme-inhibitor The binding characteristics of the modified flavonoids (1-4) c with MAO-B were examined by equilibrium dialysis to measure the dissociation of the enzyme-inhibitor complex ( Figure 8). MAO-B enzyme was incubated with the highest concentration of the analogs for 20 min at 37 • C to allow for binding of the inhibitor with the enzyme and formation of the enzyme-inhibitor complex. Then, the mixtures of the enzyme-inhibitor complex were dialyzed overnight at 4 • C using 25 mM KHPO 4 (pH 7.4) buffer. The enzyme activities were examined before and after dialysis. Incubation of MAO-B with 1.5 µM of modified flavonoids (1-4) c caused more than 70% inhibition of activity and only the enzyme activities of 1c (30%) and 2c (66%) were recovered after dialysis. Thus, the binding of 1c was partially reversible, while 2c was reversible with MAO-B ( Figure 8, Table 3).
To inspect the time-dependent binding inhibition of MAO-B, the enzyme was preincubated with the analog for 0 to 15 min at concentrations that caused nearly 40-80% inhibition depending on the analog ( Figure S77). For compounds 1c, 3c, and 4c, 70-80% inhibition was seen at the concentrations given in Figure S77. Meanwhile, compound 2c showed 30-50% of MAO-B enzyme inhibition. The control enzyme without inhibitor was also run concurrently. For validation, the MAO-B standards were run simultaneously for the time-dependent assay.

Computational Analysis of Enzyme-Inhibitor Interactions for Modified Flavonoids (1-4) c
Based on the remarkable experimental selectivity of MAO-B towards the designed compounds (1-4) c, further computational analysis was conducted. These analogs did not show good docking poses in MAO-A due to the clashes between the R group and the amino acids in the substrate binding site. The size and nature of the R group are essential for selective targeting. In general, these four acacetin 7-O-methyl ether analogs exploited the hydrophobic nature of the amino acids in the binding pocket of MAO-B, particularly Tyr 398 and Tyr 435, near the FAD cofactor by forming π-π stacking with their aromatic functionalities. The non-polar nature of the R group matches the hydrophobic nature of the Leu and Ile amino acids on the other side of the binding site. Considerable hydrophobic contacts were observed between the ligands and Leu 167, Leu 171, Ile 199, Tyr 236, and Phe 343. The ligands placed properly their polar chemistry to form strong hydrogen bonds with the backbones of Leu 171, Ile 199, and the side chain of Gln 206 (Figure 9). There is room in the binding pocket for water molecules to bridge the interactions between the ligands and amino acids, principally near the FAD.
Biomedicines 2021, 9, x FOR PEER REVIEW 21 of 25 complex were dialyzed overnight at 4 °C using 25 mM KHPO4 (pH 7.4) buffer. The enzyme activities were examined before and after dialysis. Incubation of MAO-B with 1.5 µ M of modified flavonoids (1-4) c caused more than 70% inhibition of activity and only the enzyme activities of 1c (30%) and 2c (66%) were recovered after dialysis. Thus, the binding of 1c was partially reversible, while 2c was reversible with MAO-B (Figure 8, Table 3). To inspect the time-dependent binding inhibition of MAO-B, the enzyme was preincubated with the analog for 0 to 15 min at concentrations that caused nearly 40-80% inhibition depending on the analog ( Figure S77). For compounds 1c, 3c, and 4c, 70-80% inhibition was seen at the concentrations given in Figure S77. Meanwhile, compound 2c showed 30-50% of MAO-B enzyme inhibition. The control enzyme without inhibitor was also run concurrently. For validation, the MAO-B standards were run simultaneously for the time-dependent assay.

Computational Analysis of Enzyme-Inhibitor Interactions for Modified Flavonoids (1-4) c
Based on the remarkable experimental selectivity of MAO-B towards the designed compounds (1-4) c, further computational analysis was conducted. These analogs did not show good docking poses in MAO-A due to the clashes between the R group and the amino acids in the substrate binding site. The size and nature of the R group are essential for selective targeting. In general, these four acacetin 7-O-methyl ether analogs exploited the hydrophobic nature of the amino acids in the binding pocket of MAO-B, particularly Tyr 398 and Tyr 435, near the FAD cofactor by forming π-π stacking with their aromatic functionalities. The non-polar nature of the R group matches the hydrophobic nature of the Leu and Ile amino acids on the other side of the binding site. Considerable hydrophobic contacts were observed between the ligands and Leu 167, Leu 171, Ile 199, Tyr 236, and Phe 343. The ligands placed properly their polar chemistry to form strong hydrogen bonds with the backbones of Leu 171, Ile 199, and the side chain of Gln 206 (Figure 9). There is

Conclusions
Based on our understanding of the SAR of the flavonoid skeleton and the molecular modeling studies, eight flavonoid analogs were designed. All the analogs did not demonstrate acceptable docking poses in the active site of MAO-A but they had good binding scores in MAO-B (Table S1)

Conclusions
Based on our understanding of the SAR of the flavonoid skeleton and the molecular modeling studies, eight flavonoid analogs were designed. All the analogs did not demonstrate acceptable docking poses in the active site of MAO-A but they had good binding scores in MAO-B (Table S1). They also fitted well with the thermodynamic properties of the MAO-B active site. The flavonoid compounds with O-propyl, O-isopropyl, O-isobutyl, and O-propargyl substituents showed excellent selectivity towards MAO-B, in the range of a 1200-to 3200-fold difference from MAO-A.
Rasagiline and selegiline are FDA-approved irreversible MAO-B inhibitors that exhibited a selectivity of 103-and 127-fold against MAO-A, while safinamide is a reversible MAO-B inhibitor with a selectivity of 1000-1500-fold towards MAO-B [8,9,46] The nitrogen amine series did not display any significant selectivity towards MAO-B. This could be attributed to the high electronegativity of oxygen compared with nitrogen. The study of the longer alkyl chain on C-4 should be explored. It is also worth pursuing the introduction of halogens at position 7, which might help improve the ADMET properties. Introduction of an electronegative atom such as nitrogen at position 8 would be interesting to understand the SAR of ring A in MAO activity. Another chemical modification that would help us to understand the activity of MAO-B versus MAO-A in the flavonoid scaffold that is worth exploring is the substitution of the B ring, introducing O-substituent groups at position C-3 or C-4 as on vetulin or diosmetin (see Table 1), which both show a slight preference towards MAO-B. Flavonoids have limited clinical use due to their significant challenges related to the pharmacokinetic and pharmacodynamic profiles. They have poor oral absorption along with extensive hepatic metabolism and low solubility, thereby resulting in poor bioavailability [12]. Modifying the chemical structure by substituting the flavonoid rings can help overcome the solubility issue. For example, the C-7 position on the A ring can be replaced with fluorine to improve its solubility and bioavailability.