Derivatives of the β-Crinane Amaryllidaceae Alkaloid Haemanthamine as Multi-Target Directed Ligands for Alzheimer’s Disease

Twelve derivatives 1a–1m of the β-crinane-type alkaloid haemanthamine were developed. All the semisynthetic derivatives were studied for their inhibitory potential against both acetylcholinesterase and butyrylcholinesterase. In addition, glycogen synthase kinase 3β (GSK-3β) inhibition potency was evaluated in the active derivatives. In order to reveal the availability of the drugs to the CNS, we elucidated the potential of selected derivatives to penetrate through the blood-brain barrier (BBB). Two compounds, namely 11-O-(2-methylbenzoyl)-haemanthamine (1j) and 11-O-(4-nitrobenzoyl)-haemanthamine (1m), revealed the most intriguing profile, both being acetylcholinesterase (hAChE) inhibitors on a micromolar scale, with GSK-3β inhibition properties, and predicted permeation through the BBB. In vitro data were further corroborated by detailed inspection of the compounds’ plausible binding modes in the active sites of hAChE and hBuChE, which led us to provide the structural determinants responsible for the activity towards these enzymes.

. Structures of selected bioactive Amaryllidaceae alkaloids.
Nowadays the portfolio of drugs used for treatment of AD is narrow [14], and the search for new and effective therapy of the disease is urgently needed. As a part of our ongoing research on Amaryllidaceae alkaloids being potential drugs in the treatment of neurodegenerative diseases, we have carried out modifications of the structure of the β-crinane type alkaloid haemanthamine (1). The obtained derivatives were evaluated for their potency for the inhibition of enzymes connected with the potential treatment of AD (hAChE, hBuChE, and GSK-3β). Drugs targeting the CNS need to be able to pass through the blood-brain barrier (BBB), and thus the potential of the selected derivatives to penetrate this was also studied. In vitro data were further corroborated by detailed inspection of the drugs' plausible binding modes in the active sites of hAChE and hBuChE, which allowed us to establish some preliminary structure-activity relationships.

Synthesis of Haemanthamine Derivatives 1a-1m
The structural aspects of haemanthamine (1) allowed us to prepare novel chemical entities by derivatization of its free hydroxyl group and to inspect the structure-activity relationship of the novel derivatives. Structural modifications of the haemanthamine scaffold have not been previously described in the literature. Only a few marginal studies exist, one of them reporting alkaline degradation of oxohaemanthamine and oxodihydrohaemanthamine to afford haemanthamine [32]. The scarcity of the literature further underlies a report dealing with stereoselective tyramine labeling, which was used to detect enzymatic hydroxylation of haemanthamine [33]. Noteworthy, some aliphatic esters of haemanthamine or dihydrohaemanthamine have recently been proposed to alter fermentation in the rumen [34]. The chemical modifications of 1 were possible because of its large scale isolation from Narcissus pseudonarcissus cv. Dutch Master. The preparation of 1 derivatives was inspired by lycorine, another biologically active Amaryllidaceae alkaloid. Lycorine itself exerted very low inhibitory activity against cholinesterases, but its acylated and etherified analogues possessed potent inhibitory activities against both cholinesterases [35,36]. In this context, aliphatic and aromatic substituents (esters) were explored in order to establish a structure-activity relationship (SAR). All the structural modifications related to 1 are shown in Scheme 1. The substitutions were carried out on the hydroxyl group at C-11. Compounds 1a-1e were obtained by acylation with corresponding Nowadays the portfolio of drugs used for treatment of AD is narrow [14], and the search for new and effective therapy of the disease is urgently needed. As a part of our ongoing research on Amaryllidaceae alkaloids being potential drugs in the treatment of neurodegenerative diseases, we have carried out modifications of the structure of the β-crinane type alkaloid haemanthamine (1). The obtained derivatives were evaluated for their potency for the inhibition of enzymes connected with the potential treatment of AD (hAChE, hBuChE, and GSK-3β). Drugs targeting the CNS need to be able to pass through the blood-brain barrier (BBB), and thus the potential of the selected derivatives to penetrate this was also studied. In vitro data were further corroborated by detailed inspection of the drugs' plausible binding modes in the active sites of hAChE and hBuChE, which allowed us to establish some preliminary structure-activity relationships.

Synthesis of Haemanthamine Derivatives 1a-1m
The structural aspects of haemanthamine (1) allowed us to prepare novel chemical entities by derivatization of its free hydroxyl group and to inspect the structure-activity relationship of the novel derivatives. Structural modifications of the haemanthamine scaffold have not been previously described in the literature. Only a few marginal studies exist, one of them reporting alkaline degradation of oxohaemanthamine and oxodihydrohaemanthamine to afford haemanthamine [32]. The scarcity of the literature further underlies a report dealing with stereoselective tyramine labeling, which was used to detect enzymatic hydroxylation of haemanthamine [33]. Noteworthy, some aliphatic esters of haemanthamine or dihydrohaemanthamine have recently been proposed to alter fermentation in the rumen [34]. The chemical modifications of 1 were possible because of its large scale isolation from Narcissus pseudonarcissus cv. Dutch Master. The preparation of 1 derivatives was inspired by lycorine, another biologically active Amaryllidaceae alkaloid. Lycorine itself exerted very low inhibitory activity against cholinesterases, but its acylated and etherified analogues possessed potent inhibitory activities against both cholinesterases [35,36]. In this context, aliphatic and aromatic substituents (esters) were explored in order to establish a structure-activity relationship (SAR). All the structural modifications related to 1 are shown in Scheme 1. The substitutions were carried out on the hydroxyl group at C-11. Compounds 1a-1e were obtained by acylation with corresponding anhydrides (Scheme 1) [26]. The hydroxyl group was also acylated with differently substituted benzoyl chlorides affording the corresponding esters 1f-1m (Scheme 1). The yields of the reactions exceeded 65 % in all cases. anhydrides (Scheme 1) [26]. The hydroxyl group was also acylated with differently substituted benzoyl chlorides affording the corresponding esters 1f-1m (Scheme 1). The yields of the reactions exceeded 65 % in all cases. Scheme 1. Preparation of derivatives 1a-1m from haemanthamine.

Biological Profile of Haemanthamine and Haemanthamine Derivatives 1a-1m
The primary goal of the development of 1 derivatives was to investigate them as novel chemical probes with potential higher cytotoxic activity against tumor cell lines and compare them with the parent compound 1, while maintaining low toxicity to healthy cells. As reported previously, 1 displayed pronounced cell growth inhibitory activities against a variety of tumor cells [25,30,31]. This action is presumably mediated by inhibiting early stages of ribosome biogenesis via p53-dependent antitumoral stress response [37]. Up to date, there are only a few studies devoted to drug design and biological activities of the derivatives of 1 [25,26]. Interestingly, 1 was found to be completely inactive to both cholinesterases. In the first step we decided to prepare a library of six aliphatic (compounds 1a-1e) and six aromatic esters (compounds 1g-1m). Out of these compounds, only two (1a, 1c) have been described previously [26], while the others have been synthesized for the first time in the current study. The antiproliferative effect of 1a-1m at 10 μM single-dose treatment regimen was assessed using WST-1 tetrazolium salt assay. Interestingly, none of the derivatives displayed any considerable toxicity against the tested human cell lines (Jurkat, MOLT-4, A549, HT-29, PANC-1, A2780, HeLa, MCF-7, SAOS-2 and MRC-5) during 48 h of exposure, which is in striking contrast with 1 that elicited potent cytotoxic and apoptotic effects. Mean values for each cell line proliferation are shown in Supplementary material. In the next step, the low cytotoxicity of the prepared derivatives led us to investigate the interaction of 1a-1m with cholinesterases, and GSK-3β. Interestingly, some derivatives exerted promising inhibitory activities against cholinesterases (Table 1) and GSK-3β. All aliphatic esters 1a-1f displayed either weak (1 mM ≥ IC50 ≥ 200 μM) or no (IC50 ≥ 1000 μM) inhibition activity in the cholinesterase assay. The presence of an aromatic acyl-, as in compounds 1g-1m, seems to be crucial for the hAChE and hBuChE inhibitory potency. The most active inhibitors of hAChE were 11-O-(4-nitrobenzoyl)haemanthamine (1m) and 11-O-(2-methylbenzoyl)haemanthamine (1j), both displaying two-digit micromolar IC50 values ( Table 1). The unsubstituted analogue 11-Obenzoylhaemanthamine (1g) revealed reduced potency by one order of magnitude. Analogue 11-O-(2-methylbenzoyl)haemanthamine (1j) showed a non-selective profile for both cholinesterases in the micromolar scale. Kinetic analysis of 1j and 1m was used for the determination of their inhibition pattern against both hAChE and hBuChE. Results are shown using Lineweaver-Burk plots ( Figure 2) Scheme 1. Preparation of derivatives 1a-1m from haemanthamine.

Biological Profile of Haemanthamine and Haemanthamine Derivatives 1a-1m
The primary goal of the development of 1 derivatives was to investigate them as novel chemical probes with potential higher cytotoxic activity against tumor cell lines and compare them with the parent compound 1, while maintaining low toxicity to healthy cells. As reported previously, 1 displayed pronounced cell growth inhibitory activities against a variety of tumor cells [25,30,31]. This action is presumably mediated by inhibiting early stages of ribosome biogenesis via p53-dependent antitumoral stress response [37]. Up to date, there are only a few studies devoted to drug design and biological activities of the derivatives of 1 [25,26]. Interestingly, 1 was found to be completely inactive to both cholinesterases. In the first step we decided to prepare a library of six aliphatic (compounds 1a-1e) and six aromatic esters (compounds 1g-1m). Out of these compounds, only two (1a, 1c) have been described previously [26], while the others have been synthesized for the first time in the current study. The antiproliferative effect of 1a-1m at 10 µM single-dose treatment regimen was assessed using WST-1 tetrazolium salt assay. Interestingly, none of the derivatives displayed any considerable toxicity against the tested human cell lines (Jurkat, MOLT-4, A549, HT-29, PANC-1, A2780, HeLa, MCF-7, SAOS-2 and MRC-5) during 48 h of exposure, which is in striking contrast with 1 that elicited potent cytotoxic and apoptotic effects. Mean values for each cell line proliferation are shown in Supplementary Material. In the next step, the low cytotoxicity of the prepared derivatives led us to investigate the interaction of 1a-1m with cholinesterases, and GSK-3β. Interestingly, some derivatives exerted promising inhibitory activities against cholinesterases (Table 1) and GSK-3β. All aliphatic esters 1a-1f displayed either weak (1 mM ≥ IC 50 ≥ 200 µM) or no (IC 50 ≥ 1000 µM) inhibition activity in the cholinesterase assay. The presence of an aromatic acyl-, as in compounds 1g-1m, seems to be crucial for the hAChE and hBuChE inhibitory potency. The most active inhibitors of hAChE were 11-O-(4-nitrobenzoyl)haemanthamine (1m) and 11-O-(2-methylbenzoyl)haemanthamine (1j), both displaying two-digit micromolar IC 50 values ( Table 1). The unsubstituted analogue 11-O-benzoylhaemanthamine (1g) revealed reduced potency by one order of magnitude. Analogue 11-O-(2-methylbenzoyl)haemanthamine (1j) showed a non-selective profile for both cholinesterases in the micromolar scale. Kinetic analysis of 1j and 1m was used for the determination of their inhibition pattern against both hAChE and hBuChE. Results are shown using Lineweaver-Burk plots ( Figure 2) for 1j [38]. For the results of kinetic analysis for 1m, readers are kindly referred to the Supplementary Material. The K m and V m values were calculated from the Lineweaver-Burk plot. The values of K m and V m for reactions in the presence of the tested alkaloids were decreased compared with the values for the reaction in their absence. From the results obtained it could be concluded that both derivatives act via a mixed mechanism of inhibition. Corresponding K i values of 60.1 ± 0.68 µM and 12.5 ± 0.45 µM were determined for hAChE and hBuChE, respectively.  Selected haemanthamine derivatives (compounds 1g, 1j and 1m) were screened for their GSK-3β inhibition potency at a concentration of 10 μM. Since 1m revealed the highest inhibition potency in the tested concentration scale, its IC50 was further determined demonstrating a slightly less potent profile compared with the native Amaryllidaceae alkaloids masonine, caranine, and narcimatuline [39].
The crucial part in the process of AD therapeutic drug development is the ability of the drug to reach the CNS, thus crossing the BBB. The screening for BBB penetration in early drug discovery programs provides important information for compound selection [40]. The majority of CNS drugs enter the brain by transcellular passive diffusion, thus, the evaluation of ADME (absorption, distribution, metabolism, excretion) properties is of importance to reduce attrition in the development process. The parallel artificial membrane permeability assay (PAMPA) was selected as a useful technique to predict passive diffusion through biological membranes [41], employing a brain lipid porcine membrane. The in vitro permeability (Pe) through the lipid membrane of commercial drugs was determined together with compounds 1g, 1j and 1m (Table 1). An assay validation was made comparing the reported permeability values of the commercial drugs with our experimental data (see Figure S1, Supplementary Material). A good correlation between experimental and described values was obtained Pe (exptl) = 0.9079 (bibl) − 0.696 (R 2 = 0.974). Following the pattern established in the literature for BBB permeation prediction we could classify compounds under the survey (1g, 1j and 1m) as centrally active having Pe values > 4.00 × 10 −6 cm s −1 (Table 1). However, the PAMPA assay uses artificial membranes to observe passive membrane permeability, thus neglecting the special characteristics of the BBB. For this reason we also exploited another methodology calculating logBB for active haemanthamine derivatives 1g, 1j and 1m. LogBB is the most common numeric value describing permeability across the BBB. It is defined as the logarithmic ratio between the concentration of a compound in the brain (Cbrain) and blood (Cblood) [42]. LogBB values of the calculated compounds vary from 0.007 to 0.285. Compounds with logBB > 0.3 readily cross the BBB, while those with logBB < −1.0 are only poorly distributed to the brain [43]. The obtained results for both the PAMPA assay and logBB calculation indicated that the novel derivatives represent compounds with potential to permeate through the BBB. Selected haemanthamine derivatives (compounds 1g, 1j and 1m) were screened for their GSK-3β inhibition potency at a concentration of 10 µM. Since 1m revealed the highest inhibition potency in the tested concentration scale, its IC 50 was further determined demonstrating a slightly less potent profile compared with the native Amaryllidaceae alkaloids masonine, caranine, and narcimatuline [39].
The crucial part in the process of AD therapeutic drug development is the ability of the drug to reach the CNS, thus crossing the BBB. The screening for BBB penetration in early drug discovery programs provides important information for compound selection [40]. The majority of CNS drugs enter the brain by transcellular passive diffusion, thus, the evaluation of ADME (absorption, distribution, metabolism, excretion) properties is of importance to reduce attrition in the development process. The parallel artificial membrane permeability assay (PAMPA) was selected as a useful technique to predict passive diffusion through biological membranes [41], employing a brain lipid porcine membrane. The in vitro permeability (Pe) through the lipid membrane of commercial drugs was determined together with compounds 1g, 1j and 1m (Table 1). An assay validation was made comparing the reported permeability values of the commercial drugs with our experimental data (see Figure S1, Supplementary Material). A good correlation between experimental and described values was obtained Pe (exptl) = 0.9079 (bibl) − 0.696 (R 2 = 0.974). Following the pattern established in the literature for BBB permeation prediction we could classify compounds under the survey (1g, 1j and 1m) as centrally active having Pe values > 4.00 × 10 −6 cm s −1 (Table 1). However, the PAMPA assay uses artificial membranes to observe passive membrane permeability, thus neglecting the special characteristics of the BBB. For this reason we also exploited another methodology calculating logBB for active haemanthamine derivatives 1g, 1j and 1m. LogBB is the most common numeric value describing permeability across the BBB. It is defined as the logarithmic ratio between the concentration of a compound in the brain (C brain ) and blood (C blood ) [42]. LogBB values of the calculated compounds vary from 0.007 to 0.285. Compounds with logBB > 0.3 readily cross the BBB, while those with logBB < −1.0 are only poorly distributed to the brain [43]. The obtained results for both the PAMPA assay and logBB calculation indicated that the novel derivatives represent compounds with potential to permeate through the BBB.

Docking Studies
In order to understand better the structural basis of hAChE and hBuChE inhibition by the new semi-synthetic alkaloids reported herein, we have conducted a computational study to identify a binding mode allowing us to account for the experimental SAR. All the compounds were firstly submitted to a screening method employing docking with flexible residues with 20 repetitions for each ligand 1a-1m counting both cholinesterases, i.e., hAChE (PDB ID: 4EY7) and hBuChE (PDB ID: 4BDS) [44,45]. The selection of these enzymes is in accordance with the in vitro testing where enzymes of human origin were also used. The binding energies for the top-scored docking poses are collected in Table 2. The results for hAChE nicely fit those obtained in vitro, even though this has to be taken with precaution since no clear correlation has been established between the scoring functions from in silico and data from in vitro [46]. In the case of hAChE inhibition, more potent inhibitors, i.e., those possessing IC 50 s in the two to three digit micromolar range (compounds 1g-1m), displayed binding energies between −12.7 kcal/mol to −13.4 kcal/mol. This stems from additional hydrophobic-type interaction, mainly mediated by π-π interactions with Tyr337 or Trp86, given by the presence of an attached aryl moiety to the basic core of 1. A similar trend can be found when dealing with in silico results for hBuChE (Table 2). It is apparent that potent enzyme inhibition is linked to the presence of an aryl appendage attached to 1, which is also reflected by the calculations for 1g-1k (binding energy estimates ≥ −12.8 kcal/mol). With the exception of 1m, this is due to aryl-aryl interactions mediated between the arylcarbonyl moiety of the ligands and Phe329 (this type of interaction is missing in the case of aliphatic derivatives 1a-1f) and by other less-specific interactions, mostly with catalytic triad residues.
Next, we turned our attention to two of the most pronounced cholinesterase inhibitors, namely 1j and 1m. The rationale for their selection lies in the fact that 1j was the most active hBuChE and non-specific inhibitor, whereas 1m exerted the highest affinity and selectivity for hAChE. In accordance with in vitro testing, these ligands were docked into enzyme cavities of human origin. All 20 poses retrieved from calculation via Autodock Vina software mostly preserved the putative binding modes coined to the catalytic anionic site (CAS) region of both cholinesterases (Table 2). This applies to both 1j (−13.8 kcal/mol for hAChE and −13.7 kcal/mol for hBuChE) and 1m (−13.4 kcal/mol for hAChE and −11.7 kcal/mol for hBuChE). As expected, 1j ( Figure 3A,B) is anchored within the CAS region at the bottom of the cavity gorge of hAChE. The 2-methylbenzoyl moiety is sandwiched between Tyr337 (parallel π-π stacking, 3.7 Å) and Tyr341 (distorted π-π stacking, 3.7 Å). binding modes coined to the catalytic anionic site (CAS) region of both cholinesterases ( Table 2). This applies to both 1j (−13.8 kcal/mol for hAChE and −13.7 kcal/mol for hBuChE) and 1m (−13.4 kcal/mol for hAChE and −11.7 kcal/mol for hBuChE).
a only top-scored binding energy is shown.
As expected, 1j ( Figure 3A and B) is anchored within the CAS region at the bottom of the cavity gorge of hAChE. The 2-methylbenzoyl moiety is sandwiched between Tyr337 (parallel π-π stacking, 3.7 Å) and Tyr341 (distorted π-π stacking, 3.7 Å).  Some π-alkyl interaction can also be observed between Phe338 and the 2-methylbenzoyl moiety. Linking the carbonyl from the ester group gave rise to a hydrogen bond with the hydroxyl from Tyr124 (3.1 Å). The core of the ligand is located in the proximity of aromatic residues Trp86 and Tyr133 via van der Waal's interaction. More importantly, His447 enabled the formation of a cation-π bond with a protonated nitrogen from the tetrahydroisoquinoline scaffold of 1j (3.7 Å). Another residue from the catalytic triad, namely Ser203, displayed van der Waal's contact with the polycyclic aliphatic cage of 1j. Compound 1m (Figure 3C,D) revealed a different binding pose in the hAChE active site compared with 1j, with the heamanthamine core of 1m protruding the out of the gorge towards the peripheral anionic site (PAS). This structural alteration is plausibly the result of several interactions responsible for the ligand docking, namely i) hydrogen bond formation between the 4-nitro group of the benzoyl moiety and the hydroxyl from Tyr133 (3.0 Å and 2.9 Å), ii) π-π stacking with Tyr341 (3.9 Å) and Phe297 (3.8 Å) in a parallel and T-shaped manner, respectively, iii) CH-π and cation-π forces with Phe338 and iv) parallel π-π interaction between the 4-nitrobenzoyl moiety and Trp86 (3.7 Å). The catalytic triad in this case remained unaffected; only Ser203 seems to be employed in some non-specific hydrophobic interactions with 1m. In general, it can be concluded that a major discrepancy in the binding pose is given by the presence of the nitro group in the 4-nitrobenzoyl moiety. It is noteworthy that both 1j and 1m arranged in completely different binding poses compared with galantamine in the hAChE active site (PDB ID: 4EY6), but still fitted well in the CAS region [44]. This finding could be a promising starting point in terms of designing novel ligands derived from haemanthamine.
Next, we inspected the determinants responsible for the binding differences between 1j-hBuChE ( Figure 4A,B) and 1m-hBuChE ( Figure 4C,D) complexes. Similar binding patterns for both ligands can be observed in the hBuChE active site, i.e., i) the tetrahydroisoquinoline moiety imposed face-to-face π-π stacking to Trp82 at distances of 3.7 Å and 4.3 Å for 1j and 1m, respectively, ii) a salt bridge between Asp70 (2.2 Å and 2.6 Å for 1j and 1m, respectively) and a protonated nitrogen of the haemanthamine core, as well as iii) a benzoyl moiety residing in the proximity of Phe329 via T-shaped π-π contacts (3.7 Å for both ligands). The possible explanation of the different affinities of each ligand may be the engagement of the catalytic triad residues in the ligand-enzyme lodging, as well as the overall hydrophobic contribution. In this regard, all three catalytic triad residues, i.e., Glu197, Ser198 and His438, are affected by 1j, whereas contact with Glu197 is missing in the case of the 1m-hBuChE complex. On the other hand, the nitro group of 1m provided a conventional hydrogen bond with the hydroxyl from Ser198 (2.9 Å). As indicated above, other hydrophobic contributors like π-alkyl interaction with Tyr332 presumably provided better fitting in the hBuChE active site. In this respect, 1j can be considered a more balanced, non-selective cholinesterase inhibitor worthy of further studies. Some π-alkyl interaction can also be observed between Phe338 and the 2-methylbenzoyl moiety. Linking the carbonyl from the ester group gave rise to a hydrogen bond with the hydroxyl from Tyr124 (3.1 Å). The core of the ligand is located in the proximity of aromatic residues Trp86 and Tyr133 via van der Waal's interaction. More importantly, His447 enabled the formation of a cation-π bond with a protonated nitrogen from the tetrahydroisoquinoline scaffold of 1j (3.7 Å). Another residue from the catalytic triad, namely Ser203, displayed van der Waal's contact with the polycyclic aliphatic cage of 1j. Compound 1m (Figure 3C and D) revealed a different binding pose in the hAChE active site compared with 1j, with the heamanthamine core of 1m protruding the out of the gorge towards the peripheral anionic site (PAS). This structural alteration is plausibly the result of several interactions responsible for the ligand docking, namely i) hydrogen bond formation between the 4nitro group of the benzoyl moiety and the hydroxyl from Tyr133 (3.0 Å and 2.9 Å), ii) π-π stacking with Tyr341 (3.9 Å) and Phe297 (3.8 Å) in a parallel and T-shaped manner, respectively, iii) CH-π and cation-π forces with Phe338 and iv) parallel π-π interaction between the 4-nitrobenzoyl moiety and Trp86 (3.7 Å). The catalytic triad in this case remained unaffected; only Ser203 seems to be employed in some non-specific hydrophobic interactions with 1m. In general, it can be concluded that a major discrepancy in the binding pose is given by the presence of the nitro group in the 4-nitrobenzoyl moiety. It is noteworthy that both 1j and 1m arranged in completely different binding poses compared with galantamine in the hAChE active site (PDB ID: 4EY6), but still fitted well in the CAS region [44]. This finding could be a promising starting point in terms of designing novel ligands derived from haemanthamine.
Next, we inspected the determinants responsible for the binding differences between 1j-hBuChE ( Figure 4A,B) and 1m-hBuChE ( Figures 4C,D) complexes. Similar binding patterns for both ligands can be observed in the hBuChE active site, i.e., i) the tetrahydroisoquinoline moiety imposed face-toface π-π stacking to Trp82 at distances of 3.7 Å and 4.3 Å for 1j and 1m, respectively, ii) a salt bridge between Asp70 (2.2 Å and 2.6 Å for 1j and 1m, respectively) and a protonated nitrogen of the haemanthamine core, as well as iii) a benzoyl moiety residing in the proximity of Phe329 via T-shaped π-π contacts (3.7 Å for both ligands). The possible explanation of the different affinities of each ligand may be the engagement of the catalytic triad residues in the ligand-enzyme lodging, as well as the overall hydrophobic contribution. In this regard, all three catalytic triad residues, i.e., Glu197, Ser198 and His438, are affected by 1j, whereas contact with Glu197 is missing in the case of the 1m-hBuChE complex. On the other hand, the nitro group of 1m provided a conventional hydrogen bond with the hydroxyl from Ser198 (2.9 Å). As indicated above, other hydrophobic contributors like π-alkyl interaction with Tyr332 presumably provided better fitting in the hBuChE active site. In this respect, 1j can be considered a more balanced, non-selective cholinesterase inhibitor worthy of further studies.

General Experimental Procedures
All solvents were treated by using standard techniques before use. All reagents and catalysts were purchased from commercial sources (Sigma Aldrich, Prague, Czech Republic) and used without purification. The NMR spectra were obtained in CDCl3 and CD3OD at ambient temperature on a VNMR S500 (Varian, Palo Alto, CA, USA) spectrometer operating at 500 MHz for 1 H and 125.7 MHz for 13 C. Chemical shifts were recorded as δ values in parts per million (ppm) and were indirectly referenced to tetramethylsilane (TMS) via the solvent signal (CDCl3-7.26 ppm for 1 H and 77.0 ppm for 13 C; CD3OD-3.30 ppm for 1 H and 49.0 ppm for 13 C). Coupling constants (J) are given in Hz. For unambiguous assignment of 1 H and 13 C signals, 2D NMR experiments, namely gCOSY, gHSQC, gHMBC and NOESY were measured using standard parameter settings and standard pulse programs provided by the producer of the spectrometer. ESI-HRMS spectra were obtained with a Synapt G2-Si hybrid mass analyzer of a quadrupole-time-of-flight (Q-TOF) type (Waters, Manchester, Great Britain) coupled to a Waters Acquity I-Class UHPLC system. The EI-MS were obtained on an Agilent 7890A GC 5975 inert MSD operating in EI mode at 70 eV (Agilent Technologies, Santa Clara, CA, USA). A DB-5 column (30 m × 0.25 mm × 0.25 μm, Agilent Technologies) was used. The temperature program was: 100-180 °C at 15 °C/min, 1 min hold at 180 °C, and 180-300 °C at 5 °C/min and 5 min hold at 300 °C; detection range m/z 40-600. The injector temperature was 280 °C. The flowrate of the carrier gas (helium) was 0.8 mL/min. A split ratio of 1:15 was used. TLC was carried out on precoated silica gel 60 F254 plates (Merck, Darmstadt, Germany). Compounds on the plate were observed under UV light (254 and 366 nm) and visualized by spraying with Dragendorff´s reagent.

General Experimental Procedures
All solvents were treated by using standard techniques before use. All reagents and catalysts were purchased from commercial sources (Sigma Aldrich, Prague, Czech Republic) and used without purification. The NMR spectra were obtained in CDCl 3 and CD 3 OD at ambient temperature on a VNMR S500 (Varian, Palo Alto, CA, USA) spectrometer operating at 500 MHz for 1 H and 125.7 MHz for 13 C. Chemical shifts were recorded as δ values in parts per million (ppm) and were indirectly referenced to tetramethylsilane (TMS) via the solvent signal (CDCl 3 -7.26 ppm for 1 H and 77.0 ppm for 13 C; CD 3 OD-3.30 ppm for 1 H and 49.0 ppm for 13 C). Coupling constants (J) are given in Hz. For unambiguous assignment of 1 H and 13 C signals, 2D NMR experiments, namely gCOSY, gHSQC, gHMBC and NOESY were measured using standard parameter settings and standard pulse programs provided by the producer of the spectrometer. ESI-HRMS spectra were obtained with a Synapt G2-Si hybrid mass analyzer of a quadrupole-time-of-flight (Q-TOF) type (Waters, Manchester, Great Britain) coupled to a Waters Acquity I-Class UHPLC system. The EI-MS were obtained on an Agilent 7890A GC 5975 inert MSD operating in EI mode at 70 eV (Agilent Technologies, Santa Clara, CA, USA). A DB-5 column (30 m × 0.25 mm × 0.25 µm, Agilent Technologies) was used. The temperature program was: 100-180 • C at 15 • C/min, 1 min hold at 180 • C, and 180-300 • C at 5 • C/min and 5 min hold at 300 • C; detection range m/z 40-600. The injector temperature was 280 • C. The flow-rate of the carrier gas (helium) was 0.8 mL/min. A split ratio of 1:15 was used. TLC was carried out on precoated silica gel 60 F254 plates (Merck, Darmstadt, Germany). Compounds on the plate were observed under UV light (254 and 366 nm) and visualized by spraying with Dragendorff's reagent.

Preparation of Haemanthamine Derivatives
Haemanthamine (1) used for preparation of semisynthetic derivatives has been isolated from fresh bulbs of Narcissus pseudonarcissus cv. Dutch Master (unpublished results, for 1 H-, 13 C-NMR, HPLC, and HPLC-MS spectra see Supplementary Material).

General Procedure for Acylation of 1 Using Anhydrides
To a solution of haemanthamine (1) in 3 mL of dry pyridine, 2.5-5 eq. of the corresponding anhydride was added. The mixture was stirred at room temperature until the starting material disappeared (TLC detection). The solvent was then evaporated and the residue was purified by preparative TLC using either cHx:Et 2 NH 9:1 or cHx:To:Et 2 NH 45:45:10 to afford corresponding esters 1a-1e.

11-O-Butanoylhaemanthamine (1f)
The procedure described above was followed using butyryl chloride (50 µL, 0.480 mmol) and a catalytic amount of DMAP (2 mg). After 12 h stirring at 80 • C, the solvent was evaporated and the residue was purified by preparative TLC to yield 54 mg (89%) of 1f as a colorless oil. 1

11-O-(2-Methylbenzoyl)haemanthamine (1j)
The procedure described above was followed using 2-methylbenzoyl chloride (50 µL, 0.383 mmol) and a catalytic amount of DMAP (2 mg). After 8 h stirring at 80 • C, the solvent was evaporated and the residue was purified by preparative TLC to yield 45 mg (65%) of 1j as an amorphous white solid. The procedure described above was followed using 3-methoxylbenzoyl chloride (50 µL, 0.344 mmol) and a catalytic amount of DMAP (2 mg). After 8 h stirring at 80 • C, the solvent was evaporated and the residue was purified by preparative TLC using To:Et 2 NH 9:1 as eluent to yield 69 mg (95%) of 1k as a pale white oil. 1

hAChE and hBuChE Inhibition Assay
The hAChE and hBuChE activities were determined using a modified method of Ellman with acetylthiocholine iodide (ATChI) and butyrylthiocholine iodide (BuTChI) as substrates, respectively [47].

AChE/BuChE Inhibition Mechanism
The procedure for determination of the inhibition mechanism was similar to that for determination of IC 50 , with a difference in that uninhibited and inhibited processes were observed using three different concentrations of ATChI and BuTChI (20 µM, 40 µM, 60 µM). The dependence of absorbance (412 nm) vs. time was measured and the reaction rate was calculated for all reactions (uninhibited and inhibited). Then, a Lineweaver-Burk plot was constructed and Km and Vm values were [38].

CNS Penetration: In Vitro Parallel Artificial Membrane Permeability Assay
The PAMPA assay was performed as previously described [48].

WST-1 Cytotoxicity Assay
The WST-1 (Roche, Mannheim, Germany) reagent was used to determine the cytostatic effect of the test compounds. WST-1 is designed for the spectrophotometric quantification of cell proliferation, growth, viability and chemosensitivity in cell populations using a 96-well-plate format (Sigma, St. Louis, MO, USA). The principle of WST-1 is based on photometric detection of the reduction of tetrazolium salt to a colored formazan product. The cells were seeded at a previously established optimal density (30,000 Jurkat, 25,000 MOLT-4, 500 A549, 1500 HT-29, 2000 PANC-1, 5000 A2780, 500 HeLa, 1500 MCF-7, 2000 SAOS-2 and 2000 MRC-5 cells/well) in 100 µL of culture medium, and adherent cells were allowed to reattach overnight. Thereafter, the cells were treated with 100 µL of either corresponding alkaloids or doxorubicin stock solutions to obtain the desired concentrations and incubated in 5% CO 2 at 37 • C. WST-1 reagent diluted 4-fold with PBS (50 µL) was added 48 h after treatment. Absorbance was measured after 3 h incubation with WST-1 at 440 nm. The measurements were performed in a Tecan Infinite M200 spectrometer (Tecan Group, Männedorf, Switzerland). All experiments were performed at least three times with triplicate measurements at each drug concentration per experiment. The viability was quantified as described previously according to the following formula: (%) viability = (A sample − A blank )/(A control − A blank ) × 100, where A is the absorbance of the employed WST-1 formazan measured at 440 nm [49]. The viability of the treated cells was normalized to the viability of cells treated with 0.1% DMSO (Sigma-Aldrich) as a vehicle control.

Statistical Analysis
The descriptive statistics of the results were calculated and the charts made in either Microsoft Office Excel 2010 (Microsoft, Redmond, WA, USA) or GraphPad Prism 5 biostatistics (GraphPad Software, La Jolla, CA, USA). In this study, all of the values were expressed as arithmetic means with SD of triplicates (n = 3), unless otherwise noted. The significant differences between the groups were analyzed using the Student's t-test and a P value ≤ 0.05 was considered statistically significant.

Molecular Modeling Studies
Two structures of hAChE and hBuChE were gained from RCSB Protein Data Bank-PDB ID: 4EY7 (crystal structure of hAChE) and 4BDS (crystal structure of hBuChE) [44,45]. All receptor structures were prepared by DockPrep function of UCSF Chimera (version 1.4) and converted to pdbqt-files by AutodockTools (v. 1.5.6) [50,51]. Flexible residues selection was based on previous experience with either hAChE or the spherical region around the binding cavity [52][53][54]. Three-dimensional structures of ligands were built by Open Babel (v. 2.3.1), minimized by Avogadro (v 1.1.0) and converted to pdbqt-file format by AutodockTools [55]. The docking calculations were made by Autodock Vina (v. 1.1.2) with the exhaustiveness of 8 [56]. Calculation was repeated 20 times for each ligand and receptor and the best-scored result was selected for manual inspection. The visualization of enzyme-ligand interactions was prepared using The PyMOL Molecular Graphics System, Version 2.0 (Schrödinger LLC, Mannheim, Germany). 2D diagrams were created with BIOVIA, Discovery Studio Visualizer, v 17.2.0.16349 (2016) (Dassault Systèmes, San Diego, CA, USA).

Conclusions
In conclusion, twelve derivatives were prepared derived from the β-crinane-type alkaloid haemanthamine. All semisynthetic derivatives were studied for their inhibitory potential of enzymes connected with AD, and for cytotoxicity. Using in silico methods, we were able to identify two plausible binding modes for 1j and 1m in hAChE and hBuChE to provide structural insights into the SAR for the two highlighted compounds reported herein. The potential to penetrate through the BBB of the active derivatives was also studied. The observations made herein should pave the way for the structure-based optimization of novel haemanthamine analogues potentially applicable in neurodegenerative processes. Promising biological activities were demonstrated by 1j, 1m and 1g, and these compounds will be used as lead-structures in the development and optimization of more potent drugs for the treatment of AD in the near future.