In Vitro and In Silico Evaluation of Cholinesterase Inhibition by Alkaloids Obtained from Branches of Abuta panurensis Eichler

Alkaloids are natural products known as ethnobotanicals that have attracted increasing attention due to a wide range of their pharmacological properties. In this study, cholinesterase inhibitors were obtained from branches of Abuta panurensis Eichler (Menispermaceae), an endemic species from the Amazonian rainforest. Five alkaloids were isolated, and their structure was elucidated by a combination of 1D and 2D 1H and 13C NMR spectroscopy, HPLC-MS, and high-resolution MS: Lindoldhamine isomer m/z 569.2674 (1), stepharine m/z 298.1461 (2), palmatine m/z 352.1616 (3), 5-N-methylmaytenine m/z 420.2669 (4) and the N-trans-feruloyltyramine m/z 314.1404 (5). The compounds 1, 3, and 5 were isolated from A. panurensis for the first time. Interaction of the above-mentioned alkaloids with acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) enzymes was investigated in silico by molecular docking and molecular dynamics. The molecules under investigation were able to bind effectively with the active sites of the AChE and BChE enzymes. The compounds 1–4 demonstrated in vitro an inhibitory effect on acetylcholinesterase with IC50 values in the range of 19.55 µM to 61.24 µM. The data obtained in silico corroborate the results of AChE enzyme inhibition.


Introduction
The Menispermaceae family has a wide geographic distribution, mainly in tropical and subtropical regions of the world [1]. Some species have been used by indigenous people for a long time to prepare curare poison [2]. The Abuta genus is native to tropical Central and South America, where it is represented by more than 30 species, some of which have been used in traditional medicine [1,3].
In the present study, we isolated the lindoldhamine isomer 1, palmatine 3 and the N-trans-feruloyltyramine 5 for the first time from A. panurensis, while the other two alkaloids (stepharine 2 and 5-N-methylmaytenine 4) were isolated by us earlier [10]. This is also the first report on the occurrence of 1 and 5 in the Abuta genus. The potential of alkaloids isolated as AChE inhibitors was demonstrated by in vitro studies. Interactions of the alkaloids in question with AChE and BChE enzymes were investigated in silico by molecular docking to investigate the affinity of the alkaloids at the active sites of AChE and BChE. The chemical structures of the isolated alkaloids are shown in the Figure 1. Lindoldhamine was previously reported in Menispermaceae family in the genera Triclisia [12], Abuta [22], and Albertisia [23]. Lindoldhamine isomer 1 was first reported in arrow and dart poison of the Matis tribe [24]. Bisbenzyltetrahydroisoquinoline alkaloids represent a large and important class of natural products [1,25] possessing a variety of pharmacological properties, including the acetyl-and butyrylcholinesterase inhibitory activity [11][12][13]26,27].
Therefore, all the alkaloids in question possess a common feature-they are inhibitors of predominant cholinesterases. The AChE and BChE are involved in the hydrolysis of a neurotransmitter, acetylcholine. The AChE is highly selective for the neurotransmitter hydrolysis, whereas the BChE can metabolize different substrates. In the brains of Alzheimer's disease patients, the activity of AChE tends to decrease whereas that of BChE increases. Consequently, cholinesterase inhibitors that suppress both AChE and BChE may provide a better therapeutic response rather than AChE-only selective agents [50][51][52].

Spectral Data
The lindoldhamine isomer (1) was isolated as a light yellow amorphous solid (3.6 mg). 1  The HMBC ( Figure S5) spectrum shows six quaternary carbons. The presence of C7-O-C11 bond was supported by the HMBC spectrum according to the characteristics observed in the chemical environment of the carbons in question δ C 144.5 (C-11') and δ C 145.6 (C-7). The data suggest that the compound 1 is more likely an isomer of lindoldhamine with head-to-tail bonding.
The MS/MS ( Figure S7) spectrum of the compound 1 is in agreement with the above NMR data. The ion m/z 569 showed a loss of 17 Da (m/z 552) attributed to that of a hydroxyl [M-OH] + in a C ring of the bisbenzyltetrahydroisoquinoline skeleton. The peak at m/z 178 can be considered as a diagnostic ion in the skeleton of benzyldihydroisoquinoline alkaloids [12,22,53]. Therefore, the compound 1 was identified as an isomer of lindoldhamine.
The molecular formula of the compound 2 was determined by HRMS in a positive mode as C 18 H 20 NO 3 . The MS/MS spectrum of the m/z 298 ion showed sequential losses of 17 Da (m/z 281) and 15 Da (m/z 266), and a loss of 31 Da (m/z 281 → 250) ( Figure S16); which are consistent with aporphine alkaloids containing adjacent methoxyls in the ring A and a non-substituted ring D [54,55].
In agreement with the above, the HMBC experiments ( Figure S14) demonstrated signals at δ H 7.16 (dd; 3 and 10 Hz; H-8) and δ H 7.01 (dd; 3 and 10 Hz; H 12) having a J 3 -coupling to the carbon at δ C 186.6 (C-10). The proaporphine skeleton was also established following long-distance 1 H-13 C couplings of the signals at δ H 6.89 (s, H-3) and those of the methoxyls at δ H 3.61 (s, 3H) and δ H 3.82 (s, 3H) with the carbons at δ C 144.6 (C-1) and δ C 154.7 (C-2), thus, confirming the existence of two substitutions in the A ring. Therefore, the compound (2) was elucidated as being the proaporphine alkaloid stepharine.
The molecular formula of the compound 3 was determined by HRMS in positive mode as C 14 Figure S26). In the aliphatic region of the 13 C NMR spectra ( Figure S28), the presence of a methyl (δ C 42.09) and seven methylene carbons (δ C 24-55 ppm) was confirmed by DEPT-135 ( Figure S36). Additionally, the HMBC spectra ( Figure S32) revealed two carbonyls in the structure and showed correlations of the -NH group (δ H 8.12) with the above-mentioned carbonyl signals (δ C 165.27 and δ C 165.32), thus confirming the presence of two cinnamamide moieties. The methyl group (δ C 42.09) directly linked to the nitrogen in the aliphatic chain can be confirmed in the HMBC with the long-distance proton couplings between the δ H 2.16 (s, NCH 3 ) and the methylene carbons at δ C 55.26 (C-4) and δ C 57.12 (C-6).
The MS/MS spectrum of the compound 4 is in agreement with the above NMR data. The ion m/z 420 showed sequential losses of 218 Da (m/z 202), 14 Da (m/z 188), and 57 Da (m/z 131) ( Figure S38), typical for the fragmentation pattern of cinnamic acid amides [56,57]. Therefore, the compound was identified as 5-N-methylmaytenine or 1,10-di-E-cinnamamide of 5 N-methylspermidine.
The molecular formula of the compound 5 was determined by HRMS in the positive mode as C 18 H 20  and δ H 3.46 (m; H-8 ) and one methoxy-group at δ H 3.88 (s; 4-OCH 3 ) were also found ( Figure S40). According to the HMBC experiment, δ H 7.44 (d; 16Hz; H-7) correlated at two and three bonds with the carbon signals at δ C 123.4 (C-1), δ C 111.1 (C-2), and δ C 169.3 (C-9) indicating the presence of a trans-feruloyl moiety ( Figure S44). The signal of δ H 2.75 (m; H-7') correlated with the signals at δ C 130.9 (C-1 ) and δ C 42.1 (C-8 ) that is characteristic to the presence of ethylbenzene moieties. Additionally, the signal δ H 3.46 (H-8') was assigned to a proton of the imino-group in a tyramine moiety. Coupling of the H-8' signal with those at δ C 130.9 (C-1') and δ C 169.3 (C-9) observed in the HMBC enabled us to identify the compound as N-trans-feruloyltyramine.
All compounds were first isolated by us from A. panurensis either previously (2, 4) [10] or in the present work (1, 3, 5). They were identified by a comparison of the obtained spectral results with the data reported in the literature. The isomer of lindoldhamine alkaloid 1, and the N-trans-feruloyltyramine 5 were found for the first time in Abuta genus, while palmatine 3 was isolated for the first time from A. panurensis.

Acetylcholinesterase Inhibition Assay
The inhibitory activity of the lindoldhamine isomer, stepharine, palmatine, 5-Nmethylmaytenine, and neostigmine (positive control) toward the AChE enzyme showed the inhibition percentage above 70%, whereas N-trans-feruloyltyramine was practically inactive (the inhibition percentage 19.88 ± 0.09%). The alkaloids isolated from A. panurensis revealed promising IC 50 values of 19.55-61.24 µM, the neostigmine (positive control) is an efficient short-term reversible inhibitor of the AChE enzyme that demonstrated the IC 50 of 3.72 µM (Table 1).

Binding with Acetylcholinesterase of Tetronarce californica (TcAChE)
The X-ray structure of the AChE from the electric ray T. californica was found in the RCSB Protein Data Bank under 6H12 code. The analysis showed that the majority of ligands were stable within the simulation and did not change their position with respect to their induced fit docking (IFD) pose.
Neostigmine was used as a positive control in the AChE binding assay. Results of molecular dynamics (MD) calculations for this ligand revealed a pose similar to that proposed previously by our docking studies [10]. The benzene ring of neostigmine forms a stacking interaction with Phe330, while an ammonia cation forms a salt bridge with the carboxyl group of Glu199 and π-cation interaction with Trp84 at the anionic site ( Figure 2). The docking and further MD study revealed that all active ligands with positively charged ammonia nitrogen formed a stable π-cation and π-π interactions with one of the aromatic amino acids of the external part of the AChE binding pocket (see Figure 3). The lindoldhamine isomer 1 (see Figure 3) has the highest binding affinity to AChE of −115 kcal/mol, whereas other ligands show similar affinities in the range of −70 to −80 kcal/mol (Table 1). This can be explained by the presence of 4 aromatic rings and two charged ammonia nitrogens in the molecule possessing quite high conformational mobility. Surprisingly, despite the overall aromatic character of the binding pocket, only one stacking interaction with Trp84 remained stable throughout the MD run. Instead, the π-cation interactions and hydrogen bonds with -OH groups were predominantly responsible for the binding of 1. Moreover, the compound also formed a steady ionic contact with the Glu199 moiety of the AChE.
Stepharine 2 (see Figure 3) forms one stacking interaction with Tyr334, a π-cation interaction with Trp279, and hydrogen bonds with Phe288 and Tyr121. Quite similar interactions were formed between palmatine 3 and AChE viz., stacking with Tyr334, hydrogen bonds between -OMe groups and Phe288, and π-cation interactions with Tyr121 and Trp279. Rigidity of the structure did not allow it to attain a more favourable binding pose.
N-methylmaytenine 4 forms three stacking interactions with Phe75, Trp84, and Tyr334, a π-cation interaction with Tyr121, and several hydrogen bonds with protein residues and surrounding water molecules. The structure and conformational mobility of 4 led to the second-highest affinity in the series under investigation.
Only one phenyl ring of N-trans-feruloyltyramine 5 formed stacking interaction with Phe290, and hydrogen bonds were formed only with the surrounding water molecules leading to a relatively low binding affinity.
In general, we observe better experimental IC 50 values for more conformationally flexible molecules. 5-N-methylmaytenine 4 has the highest count of rotatable bonds among the series and the lowest IC 50 value. The binding affinity of the less flexible lindoldhamine isomer 1 is slightly lower according to the experimental results. The most rigid ligand stepharine 2 has both the lowest molecular mechanics/generalized Born surface area continuum solvation method (MM/GBSA) energy and experimental activity. Despite possessing a quite rigid scaffold, palmatine 3 showed a good correlation of the predicted binding activity with the experimental IC 50 values. Presumably, the aromatic character of the structure allows it to form favorable hydrophobic interactions, and that may be one of the reasons explaining the inactivity of N-trans-feruloyltyramine 5. It lacks π-cation interaction with the AChE aromatic residues due to the absence of charged ammonia nitrogen in the structure. In line with the described trend, neostigmine has low predicted binding affinity. Nevertheless, it shows the strongest acetylcholinesterase inhibition. MM/GBSA is known to successfully rank binding affinities [63]; however, a limited number of alkaloids isolated from A. panurensis along with close values of the predicted energies for some complexes might affect the ranking accuracy.

Binding with Human Butyrylcholinesterase (BChE)
The X-ray structure of human BChE was retrieved from the RCSB Protein Data Bank under 6EP4 code [64]. The binding site of the BChE contains only four aromatic residues compared to ten in the binding pocket of AChE. (Figure 4). Therefore, interactions of aromatic molecules with the BChE binding site could be slightly weaker ( Table 1).
The majority of compounds under investigation form stacking and π-cation interactions with Trp80 at the choline-binding pocket of the acylation site as well as hydrogen bonds with deeply buried and highly conserved Glu195 residue (see Figure 5). At the peripheral site, the Asp68 forms hydrogen bonds with some of the alkaloids in question, either directly (lindoldhamine isomer 1) or through a water molecule (stepharine 2 and 5-N-methylmaytenine 4). The N-feruloyltyramine 5 is the only ligand with a different interaction pattern: it stretches between the acyl-and choline-binding pockets and forms a weak π-stacking with Phe396 and a hydrogen bond between OH-group and a His436 moiety. The larger predicted binding energy of palmatine 3 with BChE compared to AChE can be explained by the different strength of the π-stacking interactions in both complexes: the one with the Trp80 moiety of BChE was preserved during the MD run, whereas stacking with the Tyr334 in the AChE complex was more labile, according to the MD simulation. Despite fewer aromatic residues in the binding pocket of BChE compared to AChE, calculated binding affinities do not differ substantially. Results of the MD calculations show that polar interactions, including hydrogen bonds and π-cation interactions, survive more readily than hydrophobic ones. As there is no clear difference in the number and strength of polar interactions formed with BChE compared to AChE, the binding affinities to those receptors have close values.
Currently, natural products constitute one of the main sources of the AChE inhibitors used as active compounds to treat central and peripheral nervous system damages as well as to alleviate symptoms of neurodegenerative diseases [14,65,66]. The alkaloids isolated by us from A. panurensis belong to several different groups, such as bisbenzyltetrahydroisoquinoline (1), proaporphine (2), protoberberine (3), polyamine (4), and amide alkaloids (5). According to the literature, bisbenzyltetrahydroisoquinoline and protoberberine alkaloids exhibit a moderate AChE enzyme inhibition potential with IC 50 [15]. Polyamine alkaloids such as putrescine, spermidine, spermine, cadaverine, and their derivatives are present ubiquitously in all living cells [67]. In particular, they can act on receptors of the central nervous system related to neurodegenerative processes [68,69]. For example, spermidine decreases significantly the AChE activity, oxidative stress, and neuroinflammation in a cerebral hippocampus [70]. According to a previous report [49], N-trans-feruloiltyramine was inactive towards the AChE enzyme, the results of the present study confirm those findings, the alkaloid did not reach 20% of the enzyme inhibition. Taking into a consideration the values of IC 50 determined for the alkaloids under investigation together with a comparison of those with the IC 50 values reported in the literature for the same groups of alkaloids, the compound 4 appears to be the most promising, because it showed the lowest IC 50 value. Beyond a regulation of synaptic acetylcholine levels, BChE also plays an important role in neurodegenerative disease progression [51,71]. With respect to the BChE inhibition, the bisbenzylisoquinoline [11,72,73] and protoberberine [74,75] alkaloids are usually more potent inhibitors of BChE than AChE. This fact is in a good agreement with the Gibbs free energies of binding in protein-ligand complexes found here by MD simulations for 1 and 3. The aporphine alkaloids may inhibit both AChE and BChE; preferential activity toward one enzyme or another strongly depends on a particular alkaloid structure [76,77].

Chemicals
Reagents and HPLC-grade solvents were purchased from Tedia Company (Fairfield, OH, USA) and Merck KGaA (Darmstadt, Germany) and used as supplied. P.A. (nuclear)grade solvents were purified by standard procedures used in natural products chemistry. An ultrahigh-purity water was obtained by Milli Q system (Millipore, Bedford, MA, USA).

Plant Material
The authors declare that a specific permission from the National Institute of Ama- The branches collected were dried at room temperature (ca. 20 • C) for 10 days. Subsequently, the vegetal material (1.4 kg of branches) was crushed in a knife mill and stored in a refrigerator until use.

Extraction
Dried and crushed plant material was subjected to an acid-base extraction [54]. The crushed branches (300 g) were macerated with a mixture of 10% solution of NH 4 OH (2 L) and CH 2 Cl 2 (2 L) at room temperature (20 • C) for 72 h. The material was stirred every 24 h. The organic phase (1.5 L) was transferred to a separatory funnel with a 10% solution of acetic acid (2 L) and shacked manually. Then, the acidic aqueous phase was transferred to another vessel and the pH was adjusted to 10.0 by using the NH 4 OH solution, and extracted with CH 2 Cl 2 (2 × 300 mL). The organic phase was separated, concentrated on a rotary evaporator under reduced pressure, and dried with anhydrous sodium sulfate, resulting in the alkaloid fraction (280 mg).

HPLC-APCI-MS Analysis
HPLC-APCI-MS analyses were performed on an Acella chromatograph (Thermo Scientific, Waltham, MA, USA) coupled to a triple-quadrupole mass spectrometer model TSQ Quantum Acess ® (Thermo Scientific), equipped with an atmospheric pressure chemical ionization (APCI) source, and operated in positive mode with monitoring in the range of m/z 100-800. The instrument was equipped with a Surveyor LC Pump Plus, a Surveyor Autosampler Plus, a Rheodyne injection valve (25 µL), and a Luna C18 column (150 × 4.60 mm, 5 µm) (Phenomenex-Torrance, Torrance, CA, USA), and the mass spectrometer was operated simultaneously with a Surveyor PDA Plus diode array detector (DAD). The mobile phase was composed of B (methanol) and A (formic acid 1% v/v in

Docking and MD Procedures
Missing residues Ala536-Phe565 in the published structure of AChE (PDB ID: 6H12) [60] and the the residues Val377-Gln380 in the published structure of BchE (PDB ID: 6EP4) [64] were modelled by using Modeller web service [78]. Then, completed AChE and BchE structures were prepared by employing the Protein Preparation Wizard of the Schrödinger package [79][80][81][82], and all the water and other small molecules and ions were removed. Conformations of the alkaloids 1-5 and their ionization states were determined by using LigPrep [83] at pH 7.4 by using Epik [80] force field OPLS3e [84].
The calculations proceeded as shown in Figure 6. First, the ligands were docked by using the Induced Fit Docking (IFD) protocol of the Schrödinger software package [82,85]. In contrast to the standard docking protocols that assume a rigid receptor structure, the IFD tailors both a receptor (including a backbone) and a ligand structure to allow for the best binding affinity [82,[85][86][87]. The top five poses for each ligand were further analyzed for binding free-energy perturbations by the molecular mechanics method by using Prime MM/GBSA (molecular mechanics energies combined with the generalized Born and surface area continuum solvation) [82]. The poses with higher MM/GBSA binding energy were chosen; these are referred to as binding affinities throughout the text. Finally, to verify the stability of the binding poses, we conducted 10 ns molecular dynamic simulations by employing a Desmond package of the Schrodinger one [88,89]. Simulation conditions: explicit water TIP4PEw, 0.15M NaCl, T = 300 K, forcefield OPLS4. The binding poses and interactions of the protein-ligand complexes were analyses using the MD calculation results. The MD simulation reports generated by Schrodinger are available in the supporting information.

Acetylcholinesterase Inhibition Assay
An acetylcholinesterase enzyme from an electric ray Tetronarce californica (Sigma-Aldrich, St. Louis, MO, USA) was used for the experiments. The in vitro AChE inhibition assay was performed in 96-well microplates according to the methodology proposed by Ellman et al. [90] and Senol et al. [91] with some modifications. The alkaloid fraction and alkaloids were tested at concentrations of 2. A thiocholine formed by the enzymatic hydrolysis of the AChI, reacts with the DTNB giving rise to a yellow 5-mercapto-2-nitrobenzoate anion. Concentration of the latter in each well was measured as absorbance at 405 nm by using a 96-well microplate reader ELx800 TM (Bio Tek Instruments, Inc., Winooski, VT, USA). The transformation was moni- Finally, to verify the stability of the binding poses, we conducted 10 ns molecular dynamic simulations by employing a Desmond package of the Schrodinger one [88,89]. Simulation conditions: explicit water TIP4PEw, 0.15M NaCl, T = 300 K, forcefield OPLS4. The binding poses and interactions of the protein-ligand complexes were analyses using the MD calculation results. The MD simulation reports generated by Schrodinger are available in the supporting information.

Acetylcholinesterase Inhibition Assay
An acetylcholinesterase enzyme from an electric ray Tetronarce californica (Sigma-Aldrich, St. Louis, MO, USA) was used for the experiments. The in vitro AChE inhibition assay was performed in 96-well microplates according to the methodology proposed by Ellman et al. [90] and Senol et al. [91] with some modifications. The alkaloid fraction and alkaloids were tested at concentrations of 2. A thiocholine formed by the enzymatic hydrolysis of the AChI, reacts with the DTNB giving rise to a yellow 5-mercapto-2-nitrobenzoate anion. Concentration of the latter in each well was measured as absorbance at 405 nm by using a 96-well microplate reader ELx800 TM (Bio Tek Instruments, Inc., Winooski, VT, USA). The transformation was monitored for 30 min at 5-min intervals. The inhibition curve was plotted as the inhibition percent vs. concentration. All assays were performed in triplicate. Statistical analysis was performed with GraphPad Prism 7.0 ® software.

Conclusions
This is the first report on the isolation of the lindoldhamine isomer 1, palmatine 3, and the N-trans-feruloyltyramine 5 from A. panurensis, and two other alkaloids, stepharine 2 and 5-N-methylmaytenine 4, that we isolated earlier from the same species. Inhibitory activity of those alkaloids towards the AChE enzyme was evaluated by spectrophotometry and a molecular docking study. The compounds in question bind effectively to the enzyme active site and demonstrate promising inhibitory potential, except the compound 5 which was inactive. Alkaloids 1-4 were capable of interacting with both anionic and peripheral subsites, thus demonstrating higher AChE inhibition potential. Our results suggest that the alkaloids 1-4 could be used as reversible AChE inhibitors in the treatment of neurological disorder manifestations. However, more research is necessary to better investigate the complete pharmacological potential and toxicological profile of these compounds. Gibbs free binding energies in the BChE-alkaloid complexes assessed by MD simulations suggest that lindoldhamine isomer (1) and palmatine (3) are the most promising BChE inhibitors among the series investigated.  Table S1: NMR chemical shifts of lindoldhamine isomer Table S2: NMR chemical shifts of stepharine, Table S3: NMR chemical shifts of palmatine, Table S4: NMR chemical shifts of 5-N-methylmaitenine, Table S5: NMR chemical shifts of N-trans-feruloyltyramine, and MD simulation reports generated by Schrödinger software on AChE and BChE interactions with alkaloids.