Design, Synthesis and Assay of Novel Methylxanthine–Alkynylmethylamine Derivatives as Acetylcholinesterase Inhibitors

Xanthine derivatives have been a great area of interest for the development of potent bioactive agents. Thirty-eight methylxanthine derivatives as acetylcholinesterase inhibitors (AChE) were designed and synthesized. Suzuki–Miyaura cross-coupling reactions of 8-chlorocaffeine with aryl(hetaryl)boronic acids, the CuAAC reaction of 8-ethynylcaffeine with several azides, and the copper(I) catalyzed one-pot three-component reaction (A3-coupling) of 8-ethynylcaffeine, 1-(prop-2-ynyl)-, or 7-(prop-2-ynyl)-dimethylxanthines with formaldehyde and secondary amines were the main approaches for the synthesis of substituted methylxanthine derivatives (yield 53–96%). The bioactivity of all new compounds was evaluated by Ellman’s method, and the results showed that most of the synthesized compounds displayed good and moderate acetylcholinesterase (AChE) inhibitory activities in vitro. The structure-activity relationships were also discussed. The data revealed that compounds 53, 59, 65, 66, and 69 exhibited the most potent inhibitory activity against AChE with IC50 of 0.25, 0.552, 0.089, 0.746, and 0.121 μM, respectively. The binding conformation and simultaneous interaction modes were further clarified by molecular docking studies.


Introduction
The natural methylxanthines caffeine 1, theobromine 2, and theophylline 3 ( Figure 1) are widely used in traditional medicines and also in the field of biological and medicinal chemistry [1]. Methylxanthines exert anti-inflammatory, antioxidative, and neuroprotective effects, so their consumption appears to be useful in the prevention of neurodegeneration [2]. Normally, four different mechanisms are proposed to mediate pharmacological methylxanthine activity at the cellular level: antagonism of adenosine receptors, modulation of GABA receptor action, acetylcholinesterase inhibition, and regulation of intracellular calcium levels [3][4][5][6][7]. In the past years, the impact of methylxanthines in neurodegenerative diseases has been extensively studied, and several new aspects have been elucidated [2,8,9]. Caffeine proved to be a relatively potent non-competitive inhibitor of acetylcholinesterase AChE [2,[10][11][12][13][14]. Caffeine acts as a reversible inhibitor of acetylcholinesterase. It binds outside the catalytic site of cholinesterase but in its proximity to the so called anionic site and inhibition, it is of a non-competitive type in the acetylcholinesterase case while caffeine exerts minimal affinity toward the second type of cholinesterase, butyrylcholinesterase, found in the blood plasma and serum [14]. Further studies have confirmed that caffeine 1 is an agonist of nAChRs [15,16]. All the reviewed data on the activity of natural methylxanthines stimulated the development of methods for modifications of the substituent in terase but in its proximity to the so called anionic site and inhibition, it is of a non-competitive type in the acetylcholinesterase case while caffeine exerts minimal affinity toward the second type of cholinesterase, butyrylcholinesterase, found in the blood plasma and serum [14]. Further studies have confirmed that caffeine 1 is an agonist of nAChRs [15,16]. All the reviewed data on the activity of natural methylxanthines stimulated the development of methods for modifications of the substituent in positions N-1, N-7, and C-8 of the xanthine core to find new compounds with selective action and pharmacological interest. Broad modifications of the substituent in position 8 of the xanthine core structure have resulted in high potency and selectivity for the A2A [17] and A2B [18] adenosine receptors. The synthesis and biological evaluation of xanthines with an N-benzyl-(piperidine or pyrrolidine) substituent and a methoxymethyl linker in N-7, C-8, or N-9 positions revealed dual inhibitors of acetylcholinesterase and butyrylcholinesterase [19,20]. Modification of caffeine at N-1 with a methylpentanoate moiety lead to pentoxyphylline, which exhibited selective inhibition of hAChE with no inhibition of hBuChE (IC50 > 50 μM) relative to the reference agent donepezil [21]. Connecting theophylline 3 with pyrrolidine substituent through a methylene chain of different lengths (three to seven carbon atoms) gave new 7-substituted theophylline derivatives that inhibited the AChE, of which the compound with the longest methylene chain showed the strongest effect. Electrophysiological studies showed that all compounds behave as agonists of the muscle and the neuronal α7 nAChR with greater potency than caffeine 1 [22].
In this work, we report the synthesis of new xanthine derivatives based on transformations of natural methylxanthines 1-3: 8-aryl(hetaryl) substituted, 8-(1,2,3-triazolyl)substituted, and substituted at the nitrogen in the side chain 8-propargylamino-based caffeine derivatives, 1-(4-aminobut-2-ynyl)theobromines, and 7-(4-aminobut-2-ynyl)theophyllines ( Figure 2) intended as AChE inhibitors. Broad modifications of the substituent in position 8 of the xanthine core structure have resulted in high potency and selectivity for the A 2A [17] and A 2B [18] adenosine receptors. The synthesis and biological evaluation of xanthines with an N-benzyl-(piperidine or pyrrolidine) substituent and a methoxymethyl linker in N-7, C-8, or N-9 positions revealed dual inhibitors of acetylcholinesterase and butyrylcholinesterase [19,20]. Modification of caffeine at N-1 with a methylpentanoate moiety lead to pentoxyphylline, which exhibited selective inhibition of hAChE with no inhibition of hBuChE (IC 50 > 50 µM) relative to the reference agent donepezil [21]. Connecting theophylline 3 with pyrrolidine substituent through a methylene chain of different lengths (three to seven carbon atoms) gave new 7-substituted theophylline derivatives that inhibited the AChE, of which the compound with the longest methylene chain showed the strongest effect. Electrophysiological studies showed that all compounds behave as agonists of the muscle and the neuronal α7 nAChR with greater potency than caffeine 1 [22].
In this work, we report the synthesis of new xanthine derivatives based on transformations of natural methylxanthines 1-3: 8-aryl(hetaryl) substituted, 8-(1,2,3-triazolyl)substituted, and substituted at the nitrogen in the side chain 8-propargylamino-based caffeine derivatives, 1-(4-aminobut-2-ynyl)theobromines, and 7-(4-aminobut-2-ynyl)theophyllines ( Figure 2) intended as AChE inhibitors. Based on the Sonogashira cross-coupling reaction, a synthetic route was designed for the preparation of 8-ethynylcaffeine 24. We found that the reaction of 8-bromocaffeine 5 with (trimethylsilyl)acetylene 25 in the presence of Pd(PPh3)4 (5 mol %) and CuI (5 mol%) as catalysts and Et3N (1.5 equiv.) as the base in toluene proceeds smoothly by Our main focus was the introduction of a variety of nitrogen-substituted aminopropargyl functional moieties in the C-8 position of methylxanthines ( Figure 1). The literature pathway for the preparation of those compounds follows the Pd-Cu-catalyzed cross-coupling reaction of nitrogen-substituted propargylamines with 8-iodo- [45] or 8-bromoxanthines [46,47]. However, none of the above methods has found broad application for the preparation of 8-(aminopropargyl)xanthine derivatives, possibly due to moderate yields and also the low availability of nitrogen-functionalized propargylamines.
The copper(I) catalyzed one-pot three-component (A 3 -coupling) reaction among terminal alkyne, formaldehyde, and amine (the Mannich reaction) has become a popular approach to synthesizing propargylamines [48,49]. Therefore, the sequence of the Sonogashira cross-coupling reaction of 3-bromocaffeine 5 and the A 3 -coupling reaction of 8-ethynylcaffeine 24 has been exploited in our studies (Scheme 3). The reaction of compound 24, aq. formaldehyde (3 equiv.) and diethylamine 34 (3 equiv.) in THF in the presence of copper(I) iodide or copper(I) chloride (0.1 equiv.) at 75 °C (bath) proceeds smoothly, and after 3 h the alkyne was almost consumed and the desired 8-(3-(diethylamino)prop-1-yn-1-yl)-1,3,7-trimethyl-3,7-dihydro-1H-purine-2,6-dione 35 was formed. By using copper (II) acetate monohydrate (0.1 equiv.) as the catalyst in the above conditions, compound 35 was isolated in a yield of 94% after column chromatography (Scheme 3). In this condition, 8-ethynylcaffeine 24 was reacted with secondary amines 36-38 to give the desired caffeine derivatives 39-41 in high isolated yields (57-96%). The reaction of pyrrolidine 42, azepane 43, azocane 44, homomorpholine 45, 4-methylpiperidine 46, morpholine 47, and N-substituted piperazines 48-50 with aq. formaldehyde and 8-ethynylcaffeine 24 in the presence of Cu(OAc)2 × H2O (conditions a) led to N-substituted 8-(1-(aminopropargyl))caffeine derivatives 51-59 in high yields (Scheme 3). All the caffeine derivatives were purified by column chromatography (chloroform:ethanol, solvent mixture). So we found that this A 3 -coupling approach allows for the direct installation of nitrogen-functionalized aminopropargyl substituent at the C-8 position, providing efficient access to different 8-(1-aminopropargyl)trimethylxanthines. It is well known that the reported reaction is thought to proceed through the alkyne activation forming a copper acetylide. After a nucleophilic addition on the intermediate formed by the reaction of formaldehyde and a secondary amine, the propargylamine derivative is obtained [48]. We performed also a one-pot deprotection-A 3 -coupling tandem procedure for obtaining compound 35. We found that the consecutive treatment of 8-(trimethylsilylethynyl)caffeine 26 in THF under argon flow with Bu4NF (1.1 equiv.), Our main focus was the introduction of a variety of nitrogen-substituted aminopropargyl functional moieties in the C-8 position of methylxanthines ( Figure 1). The literature pathway for the preparation of those compounds follows the Pd-Cu-catalyzed cross-coupling reaction of nitrogen-substituted propargylamines with 8-iodo- [45] or 8bromoxanthines [46,47]. However, none of the above methods has found broad application for the preparation of 8-(aminopropargyl)xanthine derivatives, possibly due to moderate yields and also the low availability of nitrogen-functionalized propargylamines.
The copper(I) catalyzed one-pot three-component (A 3 -coupling) reaction among terminal alkyne, formaldehyde, and amine (the Mannich reaction) has become a popular approach to synthesizing propargylamines [48,49]. Therefore, the sequence of the Sonogashira cross-coupling reaction of 3-bromocaffeine 5 and the A 3 -coupling reaction of 8ethynylcaffeine 24 has been exploited in our studies (Scheme 3). The reaction of compound 24, aq. formaldehyde (3 equiv.) and diethylamine 34 (3 equiv.) in THF in the presence of copper(I) iodide or copper(I) chloride (0.1 equiv.) at 75 • C (bath) proceeds smoothly, and after 3 h the alkyne was almost consumed and the desired 8-(3-(diethylamino)prop-1yn-1-yl)-1,3,7-trimethyl-3,7-dihydro-1H-purine-2,6-dione 35 was formed. By using copper (II) acetate monohydrate (0.1 equiv.) as the catalyst in the above conditions, compound 35 was isolated in a yield of 94% after column chromatography (Scheme 3). In this condition, 8-ethynylcaffeine 24 was reacted with secondary amines 36-38 to give the desired caffeine derivatives 39-41 in high isolated yields (57-96%). The reaction of pyrrolidine 42, azepane 43, azocane 44, homomorpholine 45, 4-methylpiperidine 46, morpholine 47, and N-substituted piperazines 48-50 with aq. formaldehyde and 8-ethynylcaffeine 24 in the presence of Cu(OAc) 2 × H 2 O (conditions a) led to N-substituted 8-(1-(aminopropargyl))caffeine derivatives 51-59 in high yields (Scheme 3). All the caffeine derivatives were purified by column chromatography (chloroform:ethanol, solvent mixture). So we found that this A 3 -coupling approach allows for the direct installation of nitrogen-functionalized aminopropargyl substituent at the C-8 position, providing efficient access to different 8-(1-aminopropargyl)trimethylxanthines. It is well known that the reported reaction is thought to proceed through the alkyne activation forming a copper acetylide. After a nucleophilic addition on the intermediate formed by the reaction of formaldehyde and a secondary amine, the propargylamine derivative is obtained [48]. We performed also a one-pot deprotection-A 3 -coupling tandem procedure for obtaining compound 35. We found that the consecutive treatment of 8-(trimethylsilylethynyl)caffeine 26 in THF under argon flow with Bu 4 NF (1.1 equiv.), amine 34, and aq. formaldehyde in the presence of Cu(OAc) 2 × H 2 O as the catalyst (0.1 equiv.) gave the desired caffeine derivative 35 in the isolated yield 44%. No products of alkyne deprotection reaction (compound 24 or dimeric butadiyne) were observed in this condition. It should be noted that caffeine derivatives 51, 52, and 56 are known compounds, which were synthesized by the Sonogashira reaction of 8-bromocaffeine with subsequently substituted alkynes (yield 38-57%) [46,47]. Using an A 3 -coupling reaction between 8-ethynylcaffeine 24, formaldehyde, and secondary amines seems to be more effective; the reaction proceeds with higher isolated yield by using more available and stable secondary amines as the key reagents. The stage of synthesis of ethynylamines necessary for a cross-coupling reaction can be problematic due to side reactions, as well as to purify. pounds, which were synthesized by the Sonogashira reaction of 8-bromocaffeine with subsequently substituted alkynes (yield 38-57%) [46,47]. Using an A 3 -coupling reaction between 8-ethynylcaffeine 24, formaldehyde, and secondary amines seems to be more effective; the reaction proceeds with higher isolated yield by using more available and stable secondary amines as the key reagents. The stage of synthesis of ethynylamines necessary for a cross-coupling reaction can be problematic due to side reactions, as well as to purify. To further investigate the significance of the propynyl linker on enzyme inhibition assay, we synthesized caffeine derivatives 60, 61 with N-methylpiperidinyl-or N-Boc-piperazinyl-substituent at the C-8 position (Scheme 4). Compounds 60, 61 (yield 78-88%), we prepared by the reaction of 8-bromocaffeine 5 with an excess (3 equiv.) of 4-methylpiperidine 46 or 1-Boc-piperazine 49 in DMF. To further investigate the significance of the propynyl linker on enzyme inhibition assay, we synthesized caffeine derivatives 60, 61 with N-methylpiperidinyl-or N-Bocpiperazinyl-substituent at the C-8 position (Scheme 4). Compounds 60, 61 (yield 78-88%), we prepared by the reaction of 8-bromocaffeine 5 with an excess (3 equiv.) of 4-methylpiperidine 46 or 1-Boc-piperazine 49 in DMF. To further investigate the significance of the propynyl linker on enzyme inhibition assay, we synthesized caffeine derivatives 60, 61 with N-methylpiperidinyl-or N-Boc-piperazinyl-substituent at the C-8 position (Scheme 4). Compounds 60, 61 (yield 78-88%), we prepared by the reaction of 8-bromocaffeine 5 with an excess (3 equiv.) of 4-methylpiperidine 46 or 1-Boc-piperazine 49 in DMF. Next, we concentrated our efforts on the design and synthesis of purine derivatives containing a nitrogen-substituted aminobut-2-ynyl residue in position N-1 or N-7 on the methylxanthine core. To our delight, we observed that the A 3 -coupling process was amenable to 1-(prop-2-yn-1-yl)-3,7-dimethylxanthine 62 obtained from the reaction of Next, we concentrated our efforts on the design and synthesis of purine derivatives containing a nitrogen-substituted aminobut-2-ynyl residue in position N-1 or N-7 on the methylxanthine core. To our delight, we observed that the A 3 -coupling process was amenable to 1-(prop-2-yn-1-yl)-3,7-dimethylxanthine 62 obtained from the reaction of theobromine 2 with propargyl bromide 63 under basic conditions (Scheme 5). The three-component reaction of 63 with aq. formaldehyde and secondary amines, including diisopropylamine 36, azocane 44, or 1-(2-(pyrrolidin-1-yl)ethyl)piperazine 50 in THF in the presence of copper(II) acetate monohydrate was very efficient, and occurred with the formation of nitrogen substituted 1-(4-(amino)but-2-yn-1-yl)-3,7-dimethyl-3,7-dihydro-1Hpurine-2,6-diones 64-66 (54-90%) (Scheme 5).  The composition and structure of all synthesized compounds were confirmed by IR, UV, 1 H and 13 C NMR spectroscopy, and mass spectrometry (HRMS). The purity of all compounds was checked additionally by thin-layer chromatography and elemental analysis. The 1 H and 13 C NMR spectra of all synthesized compounds agreed with their structures and contained one set of characteristic signals for the xanthine core and the corresponding substituent. The structures of 8-arylsubstituted and 8-triazolylsubstituted caffeine derivatives 18, and 32 were established by X-ray structure analysis. The refined molecules are shown in Figure 3. Aryl substituent on C8 in structure 18 has an interplane angle with caffeine core equaling 72.1º and the bond length of C8-C1ʹ 1.479(2) Å being longer than the same one in 32. The π…π-interactions of pyrimidine cycles form molecular dimers ( Figure 3A, right), having the following geometric parameters: the intercentroid distances equaling 3.639(1) and distances from the centroid to plain 3.354(1) Å.
Notably, the triazole ring in structure 32 lays practically in the plain of caffeine ( Figure 3B); the interplane angle between cycles is 1.3º, and the C8-C4 bond length equals 1.455 Å (the same as for a single bond in conjugated C=C-C=C [50]). Such an extended aromatic system and the presence of a hydroxy group lead to the formation of 1-D motifs along axes c ( Figure 3B, right) due to interactions of the π-system, with the intercentroid distances equaling 3.497(2)-3.581(2), and distances from centroids to plain laying in the interval of 3.337(2)-3.366(2) 2x Å, together with hydrogen bond O3A-H…N2′ with parameters H…N 2.21, O…N 2.912(8) Å and O-H…N 144º [51]. The composition and structure of all synthesized compounds were confirmed by IR, UV, 1 H and 13 C NMR spectroscopy, and mass spectrometry (HRMS). The purity of all compounds was checked additionally by thin-layer chromatography and elemental analysis. The 1 H and 13 C NMR spectra of all synthesized compounds agreed with their structures and contained one set of characteristic signals for the xanthine core and the corresponding substituent. The structures of 8-arylsubstituted and 8-triazolylsubstituted caffeine derivatives 18, and 32 were established by X-ray structure analysis. The refined molecules are shown in Figure 3. Aryl substituent on C8 in structure 18 has an interplane angle with caffeine core equaling 72.1º and the bond length of C8-C1 1.479(2) Å being longer than the same one in 32. The π . . . π-interactions of pyrimidine cycles form molecular dimers ( Figure 3A, right), having the following geometric parameters: the intercentroid distances equaling 3.639(1) and distances from the centroid to plain 3.354(1) Å.
Notably, the triazole ring in structure 32 lays practically in the plain of caffeine ( Figure 3B); the interplane angle between cycles is 1.3 • , and the C8-C4 bond length equals 1.455 Å (the same as for a single bond in conjugated C=C-C=C [50]). Such an extended aromatic system and the presence of a hydroxy group lead to the formation of 1-D motifs along axes c ( Figure 3B, right) due to interactions of the π-system, with the intercentroid distances equaling 3.497(2)-3.581(2), and distances from centroids to plain laying in the interval of 3.337(2)-3.366 (2)
Drawing a comparison among N-substituted 8-(1-aminopropargyl)caffeines 35, 39-41, 51-59, we can describe the following general SAR: For AChE inhibition, the presence of a cyclic substituent at the propargylamino moiety is essential. A remarkable increase in inhibition was observed for compound 53 with a bulk
The obtained data indicated that the AChE inhibitory activity of 8-aminopropynyl-, 1-, and 7-aminobutynylmethylxanthines is sensitive to the nature of the nitrogen functionalities in the substituent. To the best of our knowledge, this structural type of AChE
The obtained data indicated that the AChE inhibitory activity of 8-aminopropynyl-, 1-, and 7-aminobutynylmethylxanthines is sensitive to the nature of the nitrogen functionalities in the substituent. To the best of our knowledge, this structural type of AChE  Table 1).
The obtained data indicated that the AChE inhibitory activity of 8-aminopropynyl-, 1-, and 7-aminobutynylmethylxanthines is sensitive to the nature of the nitrogen functionalities in the substituent. To the best of our knowledge, this structural type of AChE inhibitor is presented herein for the first time, and our results confirmed the potential value of this class of biologically active substances.

Molecular Docking Study
With the aim of obtaining some information about the possible interactions of the methylxanthine derivatives with the enzyme, which, in turn, could be useful in the design of more potent inhibitors, a molecular modeling study was performed comparing with donepezil, a nanomolar AChE inhibitor. The calculation of the interaction energy of donepezil in XRD model coordinates (score in place mode) gave an estimated binding energy of −14.806 kcal/mol. Molecular docking of the minimized donepezil molecule to the AChE active site showed a very close value of −14.817 kcal/mol ( Table 2). Compounds  64, 68, 65, 69, and 55 show a theoretical affinity for the AChE active site at the level of donepezil. When comparing the conformations of donepezil in the XRD model and that obtained as a result of docking, the RMSD of the atomic coordinates was determined to be 0.363. Thus, the procedure of molecular docking under the conditions of the chosen AChE model makes it possible to quite accurately predict the position and conformation of the inhibitor. Molecular interactions of donepezil and active xanthine derivatives are shown in Figure 4. Analysis of the SAR of compounds differing in the position of attachment of the same type of substituent to the xanthine backbone of caffeine shows that, apparently, in the series of compounds with the diisopropylamino function: 64, 68 , and 39, the presence of a rotating bond in front of the propargylamine substituent has a significant effect on the theoretical affinity ( Figure 4E). The most theoretically affine compounds in this series (64 and 68) are modified by adding an aminobut-2-inyl linker to diisopropylamino substituent at positions 1 or 7 of the xanthine core ( Figure 4B,C), which allows the diisopropylaminobut-2-ynyl substituent to occupy a more advantageous position compared to compound 39 ( Figure 4D), where the rigid aminopropargyl substituent is attached directly to the 8-th position of the core. The presence of a rigid acetylene bond in compound 39 causes the only possible position of the linear part, and (isopropylamino)propynyl substituent acquired the possibility of the formation of hydrophobic interactions of isopropyl moiety with Trp-86. When comparing compounds 65 and 69, with a bulky azocane moiety at the but-2-ynyl substituent, it is more advantageous for binding to attach a substituent at position 1 of the xanthine backbone compared to position 7 ( Figure 4F-H). The differences in the estimated binding energies may be due to the fact that different carbonyl groups form a hydrogen bond with Phe295: in the case of derivative 65, this is the carbonyl oxygen in position 2 of the backbone, and for compound 69the carbonyl group in position C-6. The azocane substituent ensured the hydrophobic interaction with Trp86.
For compound 66 (pyrrolidinoethylpiperazinyl)butynyl substituent at position 1 of the backbone), the most energetically favorable conformation occurs at the "reverse" position of the methylxanthine core compared to the typical position for other active derivatives ( Figure 4I). With this orientation, no stacking occurs with amino acids characteristic of AChE inhibitors, and no hydrogen bond is formed with Phe295; however, the carbonyl group at C-2 forms a hydrogen bond with Glu202. In this case, it is worth considering other, less energetically favorable conformations in comparison with other active compounds. Hydrophobic interaction of substituent: piperazine ring with Phe-338 and pyrrolidine ring with Tyr341 and Trp286 are maintained ( Figure 4I). Compound 59 ((pyrrolidinoethylpiperazinyl)prop-2-ynyl substituent at position 8 of the backbone) does not form a hydrogen bond with Phe295 ( Figure 4J). The binding mode for the hAChE pocket is characterized by π-π interactions of the xanthine core with Tyr341. The internal ethylpiperazinyl substituent at the alkyne in the C-8 position formed hydrophobic interaction with Phe 338. The pyrrolidinoethylpiperazinyl substituent of this derivative is U-folded. Compared to donepezil, its structure is less flexible due to the presence of a rigid propargylamine linker ( Figure 4K). However, the pyrrolidine ring at the end of the substituent is able to establish the interaction with Trp86 ( Figure 4I).   The active site of AChE is saturated with aromatic amino acids, which provided the main type of non-covalent interactions with inhibitors-stacking and interactions of the π-systems of Trp286, Tyr341, Phe338, Tyr337, and Trp86 amino acids with π-systems and individual atoms of inhibitors. In this regard, the aromatic systems of the benzyl and dihydroindene rings of donepezil are actively involved in the interactions ( Figure 4A). Apparently, the hydrogen bond between the oxygen of the keto group associated with the dihydroindene ring and the amino acid residue Phe295 is important for the formation of a stable interaction between the inhibitor and the AChE active site. The xanthine core of the new derivatives is actively stacked with the Trp286 and Tyr341 π-systems (except for compound 66). A common feature of all the most active compounds is the ability of the carbonyl groups of the purine cycle to form a hydrogen bond with Phe295. Due to the high saturation of xanthine with polar atoms, such a bond can arise at any orientation of the purine in the active site of AChE, depending on the site of the linker attachment.
Analysis of the SAR of compounds differing in the position of attachment of the same type of substituent to the xanthine backbone of caffeine shows that, apparently, in the series of compounds with the diisopropylamino function: 64, 68, and 39, the presence of a rotating bond in front of the propargylamine substituent has a significant effect on the theoretical affinity ( Figure 4E). The most theoretically affine compounds in this series (64 and 68) are modified by adding an aminobut-2-inyl linker to diisopropylamino substituent at positions 1 or 7 of the xanthine core ( Figure 4B,C), which allows the diisopropylaminobut-2-ynyl substituent to occupy a more advantageous position compared to compound 39 ( Figure 4D), where the rigid aminopropargyl substituent is attached directly to the 8-th position of the core. The presence of a rigid acetylene bond in compound 39 causes the only possible position of the linear part, and (isopropylamino)propynyl substituent acquired the possibility of the formation of hydrophobic interactions of isopropyl moiety with Trp-86. When comparing compounds 65 and 69, with a bulky azocane moiety at the but-2-ynyl substituent, it is more advantageous for binding to attach a substituent at position 1 of the xanthine backbone compared to position 7 ( Figure 4F-H). The differences in the estimated binding energies may be due to the fact that different carbonyl groups form a hydrogen bond with Phe295: in the case of derivative 65, this is the carbonyl oxygen in position 2 of the backbone, and for compound 69 -the carbonyl group in position C-6. The azocane substituent ensured the hydrophobic interaction with Trp86.
For compound 66 (pyrrolidinoethylpiperazinyl)butynyl substituent at position 1 of the backbone), the most energetically favorable conformation occurs at the "reverse" position of the methylxanthine core compared to the typical position for other active derivatives ( Figure 4I). With this orientation, no stacking occurs with amino acids characteristic of AChE inhibitors, and no hydrogen bond is formed with Phe295; however, the carbonyl group at C-2 forms a hydrogen bond with Glu202. In this case, it is worth considering other, less energetically favorable conformations in comparison with other active compounds. Hydrophobic interaction of substituent: piperazine ring with Phe-338 and pyrrolidine ring with Tyr341 and Trp286 are maintained ( Figure 4I). Compound 59 ((pyrrolidinoethylpiperazinyl)prop-2-ynyl substituent at position 8 of the backbone) does not form a hydrogen bond with Phe295 ( Figure 4J). The binding mode for the hAChE pocket is characterized by π-π interactions of the xanthine core with Tyr341. The internal ethylpiperazinyl substituent at the alkyne in the C-8 position formed hydrophobic interaction with Phe 338. The pyrrolidinoethylpiperazinyl substituent of this derivative is U-folded. Compared to donepezil, its structure is less flexible due to the presence of a rigid propargylamine linker ( Figure 4K). However, the pyrrolidine ring at the end of the substituent is able to establish the interaction with Trp86 ( Figure 4I).
This molecular modeling helped to distinguish the key interaction of substituted methylxanthines and AChE. Molecular dynamics was performed to investigate the stability of the protein-ligand complexes in time (during 100 nanoseconds) Figure S2A-D, Supplementary Materials).

Donepezil.
During the simulation, the protein-ligand complex was stable, but interestingly, based on the RMSD-time relation, two segments of the simulation could be distinguished, with the first being from 5 to 38 ns and the second from 54 to 100 ns. It should be noted that it corresponds to the conformational change "chair to chair" of the piperidine fragment of the donepezil ligand ( Figure S2A, Supplementary Materials).
Compound 59. In general, the protein-59 complex was stable during the simulation; the insignificant deviations of ligand RMSD from 2 to 3 Å correspond to the conformational flexibility of the N-alkylpyrrolidine fragment of 59 ( Figure S2B, Supplementary Materials).
Compound 64. Similar to 59, the protein-64 complex was stable during the simulation as well. As could be seen from the RMSD-time relation, a ligand reversible conformation change occurred. Analyzing the trajectory of the protein-ligand complex, we concluded that it was a result of conformational interconversion of the N-diisopropyl fragment of 64.
In particular, a rotation took place across the -N-C-bond with a corresponding conformational change of the N-diisopropyl fragment from anti-gauche to gauche-anti ( Figure S2C, Supplementary Materials).
Compound 69. As could be seen, the protein-69 complex was stable for the first 35 nanoseconds of the simulation, and after that compound 69 partially left the binding pocket. During the period from 35 to 55 ns, the ligand was bound with protein by an azocyl fragment. It seemed to be a formation of π-cation interaction between the protonated tertiary amino group and the side chain of Trp286. On the other hand, the formation of hydrogen bonds between the Tyr77 hydroxyl group and the C6-carbonyl group of the purine core is supposed to be possible. During the following simulation, after 55 ns, the dissociation of the protein-ligand complex occurred. Interestingly, after 90 ns, the ligand binds an opposite side of the protein in the region of 61-64 amino acids with the following dissociation after 98 nanoseconds ( Figure S2D, Supplementary Materials).

Biochemical Method
AChE activity inhibition was evaluated by the assay described by Ellman et al. [52]. 5,5 -dithiobis(2-nitrobenzoic acid) (DTNB), AChE (AChE, E.C.3.1.1.7, Type V-S, lyophilized powder, from electric eel, 1000 units), and acetylthiocholine iodide were obtained from Sigma. A total of 100 µL of a solution containing 10 mM Tris-HCl pH7.5, 0.5 mM DTNB, and 1 µL of acetylcholinesterase at a concentration of 6.25 units × 10 −3 /mL was dropped into polystyrene wells in each. Then, 2 µL of solutions of inhibitors in DMSO were added. The control was supplemented with 2 µL of DMSO without inhibitors. Incubation was performed at room temperature for 15 min., then 5 min. at 4 • C on a planetary shaker. After that, 100 µL of a cooled solution containing 10 mM Tris-HCl pH7.5, 0.5 mM DTNB, and 1 mM acetylthiocholine iodide, prepared immediately before addition, was added to each well. The wells were placed in a reader (PerkinElmer 2103 Multilabel Reader), and the reaction was monitored at room temperature at a wavelength of 405 nm. Each inhibitor was applied at three concentrations in half-log increments (eg 10, 30, and 100) in triplicate. The measurements were carried out at a wavelength of 405 nm every 2 min. The 10th measurement was used for calculations (18 min. from the beginning of the reaction). For each well, the first measurement (background) was subtracted from the indicator of the last measurement, and the resulting numbers were used for further calculations.
Next, activity curves were built, where the point without inhibitor was taken as 100%. The points were used to build a trend line from which the inhibitor concentration where the activity was at 50% was calculated All constructions and determination of the mean and standard deviation were made in Excel 2007 and online services for solving equations. The curves for AChE inhibition by compounds 28, 64, 65, 66, and 70 are given in Supplementary Materials ( Figure S1A-E).

Molecular Modelling and Molecular Dynamic Procedures
Molecular modeling was carried out in the Schrodinger Maestro visualization environment using applications from the Schrodinger Small Molecule Drug Discovery Suite 2016-1 package [60]. Three-dimensional structures of the derivatives were obtained empirically in the LigPrep application using the OPLS3 force field [61]. All possible tautomeric forms of compounds, as well as various states of polar protons of molecules in the pH range of 7.0 ± 2.0 were taken into account. For calculations, the XRD model of human AChE with PDB ID 6O4W (resolution 2.35 Å) was chosen [62]. To model a possible mechanism of inhibition of the selected target, molecular docking of new compounds was performed at the binding site of donepezil of AChE in the Glide application [63]. The search area for docking was selected automatically, based on the size and physico-chemical properties of the inhibitor. The extra precision (XP) algorithm of docking was applied. Docking was performed in comparison with donepezil. The molecular structure of the inhibitor was obtained in the PubChem database and prepared in the LigPrep application. Non-covalent interactions of compounds in the binding site were visualized using Biovia Discovery Studio Client [64].
The molecular dynamic simulation was performed by using NAMD v. 2.14 [65]. Topology files for the ligands were generated by using SwissParam [66], and simulations were performed by using the CHARMM force field [67]. The protein-ligand complexes were solvated using a cubic box with periodic boundary conditions, with the minimal distance from the protein to the boundary being 20 Å. Water molecules, located in the protein pocket and taken into the account during docking procedure, were added as well into the systems to be simulated. The systems were neutralized by the addition of sodium or chlorine ions. Simulations were performed in conditions of 0.15 M NaCl solution. The systems were minimized, annealed up to 310 K, and equilibrated during 1 nanosecond in an NPT ensemble with the movable of the backbone being constrained. The following simulations were performed during 100 nanoseconds in an NVT ensemble at 310 K. System generations and calculations of the ligand root-mean-square deviations (RMSD) were performed in VMD [68]. Plots of changes in the RMSD of atomic coordinates over simulation time, obtained as a result of molecular dynamics research, are given in Supplementary Materials, Figure S1A-D.