Design, Synthesis, and Biological Evaluation of a Series of 5- and 7-Hydroxycoumarin Derivatives as 5-HT1A Serotonin Receptor Antagonists

We have designed and synthesized a series of 60 new 5- and 7-hydroxycoumarin derivatives bearing the piperazine moiety with the expected binding to 5-HT1A and 5-HT2A receptors. Molecular docking of all investigated compounds revealed subnanomolar estimates of 5-HT1AR Ki for three ligands and 5-HT2AR Ki for one ligand as well as numerous low nanomolar estimates of Ki for both receptors. Intrigued by these results we synthesized all 60 new derivatives using microwave-assisted protocols. We show that three new compounds show a relatively high antagonistic activity against the 5HT1A receptor, although lower than the reference compound WAY-100635. These compounds also showed relatively low binding affinities to the 5-HT2A receptor. We also provide a detailed structure–activity analysis of this series of compounds and compare it with previously obtained results for an exhaustive series of coumarin derivatives.


Introduction
N-arylpiperazine-containing ligands are a large class of chemical compounds with various known biological activities, such as enzyme inhibition, antibacterial, antineoplastic, and anticancer properties, as well as adrenergic and serotonin receptor inhibition [1][2][3][4][5][6][7]. This last activity is particularly prominent for this family of compounds, as even some of its simplest members, such as 1- (3-chlorophenyl) piperazine or m-trifluoromethylphenylpiperazine, are known to interact with serotonin receptors [4,8]. The high affinity of these systems to 5HT receptors stems from the highly basic nitrogen atom of the piperazine, which is able to form strong interactions with the conserved acidic amino acids in the GPCR transmembrane domain of these proteins [9]. In order to be effective as 5HT receptor antagonists or agonists, such compounds require in their structure, however, also a relatively bulky moiety connected usually to the N-arylpiperazine via a flexible aliphatic linker. Such a design principle has been first realized in buspirone, a 5-HT 1A receptor agonist, which also has moderate activity against other 5HT receptors and selected dopamine receptors, and is followed until today with the goal of finding new agonists/antagonists of 5HT receptors [10][11][12]. 5HT receptors proteins modulate the release of many neurotransmitters, therefore are an important target for a variety of drugs, including antipsychotics, antidepressants, hallucinogens anorectics, and antimigraine agents [13][14][15][16].
Among many groups considered as the bulky moiety connected to N-arylpiperazine coumarin derivatives have gained some attention, particularly after the investigations of Chen et al., who showed that selected N-arylpiperazines connected to coumarins in position 7 via a (CH 2 ) 4 linker have nanomolar Ki values toward 5-HT 1A and 5-HT 2A receptors [17,18]. Inspired by these works we have expanded the family of potential serotonin agonists/antagonists based on the same design principle by introducing different arylpiperazine derivatives of 7-hydroxycoumarin, some of which showed subnanomolar affinities to 5-HT 1A receptor and low nanomolar affinities to 5-HT 2A receptor [19,20]. Later we have also obtained a series of arylpiperazine derivatives of 5-hydroxycoumarin [21][22][23]. We showed that the highest, subnanomolar affinities for 5-HT 1A receptor were associated with the presence of the acetyl group in the C-6 position at the coumarin ring and the substituents in the 2 or 3 position in the phenyl ring of piperazine. Finally, in 2020 we designed a new series of arylpiperazinyl derivatives of 6-acetyl-5-hydroxy-4,7-dimethylcoumarin, which also showed very low nanomolar affinities toward 5HT 1A and 5HT 2A but also low affinities to the D 2 receptor [24]. In these studies we noticed that the length of the alkyl linker (three-carbon versus four-carbon) had little impact on the obtained Ki values, since the affinities for specific serotonin receptors for analogous compounds containing the same arylpiperazinyl fragments, differing only in the length of the alkyl linker, were very similar. It is worth noting that this finding is not based on single cases but on a large number of cases, showing a clear tendency for this particular length of the linker (Figure 1). To conclude our search for new biologically active compounds in this series as well as to gain even more knowledge of the structure-activity relationships we have designed two new series of arylpiperazine derivatives of 5-hydroxycoumarins and 7-hydroxycoumarins. These series were designed based on the aryl substituents giving the highest affinities in our previous studies, but with different lengths of the alkyl linkers, consisting of either two or five CH 2 moieties. In this study we have used molecular docking to homology models of 5-HT 1A and 5-HT 2A receptors followed by microwave-assisted protocols to synthesize all 60 compounds. We also performed functional activity studies for the 5-HT 1A receptor, as well as 5-HT 2A receptor affinity studies.

Docking Studies
The results of the Ki estimates obtained from the computational studies are presented in Tables 1 and 2. While for the starting compounds 1-6 the Ki values were estimated at 56-922 nM for 5-HT 1A and above 1 µm for 5-HT 2A receptor, some of the functionalized derivatives show nanomolar or even subnanomolar affinities. In particular, there are three new compounds with the estimated Ki versus 5-HT 1A receptor below 1 nm (1a, 6b, and 6g) and one for 5-HT 2A receptor (5j). As there are many other compounds with the estimated Ki below 10 nM we decided to synthesize all of these systems to verify their 5-HT 1A/2A receptor affinities. We also decided to perform a detailed analysis of the predicted binding poses for 5-HT 1A receptor, as the Ki values for the 5HT 2A receptor are usually less favorable and our previous studies showed that this class of compounds in most cases binds stronger to 5-HT 1A receptor than to 5-HT 2A receptor [19][20][21][22][23][24]. Of the three compounds predicted to have subnanomolar affinities to 5-HT 1A receptor (1a, 6b, 6g) all have the crucial interaction between the basic nitrogen atom of the piperazine group with the conserved D116 of the receptor binding site (see Figure 2). As one can see the poses of these three systems are also very similar, with the coumarin part going deep into the binding pocket, toward transmembrane (TM) helix 7 and the arylpiperazine extending toward transmembrane helix 4. Apart from the salt bridge to D114 from transmembrane helix 3, 1a is predicted to form also hydrogen bonds between the methoxy moiety of the arylpiperazine and S199 (TM4) as well as between the coumarin part and N392 (TM7). On the other hand 6b is predicted to be additionally stabilized by the interaction between the F atom of the arylpiperazine and S199 (TM4), while 6g by the hydrogen bond between the oxygen atom of the linker and Y390 (TM7). It is also worth mentioning that these poses are similar to our previously predicted poses for the coumarin derivatives with three CH 2 moieties. On the other hand for some derivatives with four CH 2 groups we predicted a different orientation of the ligand, where the coumarin part extends toward TM helix 4, while the arylpiperazine part goes deep into the pocket and interacts with the residues located on TM7 [24].

5-HT 1A Receptor Activity
After purification via column chromatography, all newly synthesized compounds were subjected to in vitro evaluation of their functional activity for the 5-HT 1A receptor, as well as 5-HT 2A receptor affinity studies. Since in our previous study similar coumarin derivatives showed high affinities to 5-HT 1A receptor and low to 5-HT 2A receptor we decided to employ in this study a functional assay to establish the potency and efficacy of 5-HT 1A binding of our series of compound. The major advantage of this approach over determining only receptor affinity is the ability to predict intracellular consequences of receptor binding, leading to either receptor activation, blockage, or alteration of constitutive activity. Moreover, measures of affinity may not correspond to drug potency, owing to the possible existence of a receptor reserve [27]. Also, for ligands displaying functional bias, measuring one distinctive activation pattern allows for the prediction of the therapeutic usefulness of the drug in question [28]. Thus, functional characteristics are of major importance for any drug discovery program which strives for in vivo evaluation of compound activity. For the 5-HT 2A receptor we expected, on the hand, low affinities and decided to perform standard receptor affinity studies.
In the family of 5-hydroxy-4,7-dimethylchromen-2-one (A) derivatives, compounds with a five carbon linker were much more active than those with a two carbon linker. Comparing the systems containing the same piperazinyl part within this family, we can see that the five carbon linker derivatives have always a higher activity than the one with two CH 2 moieties, e.g., EC 50 = 29.4 ± 7.3 nM for 1a and EC 50 = 1881 ± 427 nM for 2a; EC 50 = 980 ± 207 nM for 1b and no activity for 2b; EC 50 = 1698 ± 358 nM for 1c and EC 50 = 19130 ± 2363 nM for 2c, etc. For 6-acetyl-5-hydroxy-4,7-dimethylchromen-2-one (B) derivatives, which differ from the A family in the presence of an additional acetyl group at the position C-6 of the coumarin ring, derivatives with the five-carbon linker had higher activity than those with the two-carbon linker, when they contained 2-methoxyphenyl (see 3a and 4a), 3-methoxyphenyl (see 3c and 4c), 2,5-dimethylyphenyl (see 3d and 4d), 3bromophenyl (see 3g and 4g), or 2,5-dimethylyphenyl moiety (see 3h and 4h). On the other hand, derivatives with the two-carbon linker showed higher activity than those with the five-carbon linkers, when they contained the 2-fluorophenyl (see 3b and 4b), 3-fluorophenyl (see 3e and 4e), 2-bromophenyl (see 3f and 4f), or 2,3-dichlorophenyl moiety (see 3i and 4i). Such a difference may stem from the fact that a longer alkyl linker may maximize the interactions of the ligand with the receptor's residues of different transmembrane regions for all derivatives apart from selected B derivatives, which due to the presence of the additional acetyl moiety makes the ligand too large for bulkier arylpiperazines to find optimal interactions in the binding site. Molecular docking studies suggest that upon anchoring to D116 coumarin derivatives can extend both toward transmembrane regions 4 and 7 to find favorable interactions within the binding site. The two-carbon linker makes such an extension impossible, lowering in most cases the potency of the antagonist. Finally, for 8-acetyl-7-hydroxy-4-methylchromen-2-one (C) derivatives, all compounds with the five-carbon linker (5a-5i) showed higher antagonistic activities than their twocarbon linker counterparts (6a-6i), with the exception of 8-acetyl-7-(2-[4-(2-cyanophenyl) piperazin-1-yl]ethoxy)-4-methylchromen-2-one (6j), which showed a higher activity (EC 50 = 7804 ± 1876 nM) than its analogue, 8-acetyl-7-(5-[4-(2-cyanophenyl)piperazin-1yl]penthoxy)-4-methylchromen-2-one (5j) (EC 50 = 13,860 ± 2059 nM). We can speculate that the higher antagonistic activities are a result of a similar structural feature as in the A family, due to a different position of the acetyl moiety in the C family, which lowers the volume of these derivatives with respect to the B family.
The moderate agreement between the experimental and theoretical K i values for 5HT 2A receptor warrant a short comment. While the predicted K i values for the newly synthesized set of coumarins derivatives are usually in the low nanomolar range, the experimental values are usually closer to micromolar values. The most likely explanation of these discrepancies is the combination of the imperfection of our computational model of the 5HT 2A receptor, particularly in the binding site part and the limited accuracy of the computational methods. The second problem is very well-known in the scientific community, as it has been shown that while Autodock and other similar programs can identify the correct binding poses, they often have problem is predicting correct bonding affinities [29]. As for the accuracy of homology models of GPCRs, they certainly can be improved by resorting to more sophisticated methods, such as e.g., using multiple templates or going beyond the homology modelling, and we are planning to make use of these new methods in the future [30][31][32]. Nevertheless the most 2D schematic representations of the predicted binding sites for the selects, most interesting coumarins derivatives are presented in the Supplementary Materials. Taking compound 1j as the example we can suggest, that this compound is able to perfectly fit into the binding site of the 5HT 2A receptor, keeping the salt bridge to D155, while retaining the coumarin part in the hydrophobic region of the binding site and the piperazine part in the hydrophilic one. This is not true for this compound binding to the 5HT 1A receptor, as the salt bridge to D116 forced 1j to move the coumarin part into a more hydrophilic region, lowering the affinity to the receptor. Additionally, 1j in the binding site of the 5HT 2A receptor is stabilized by two hydrogen bonds and a π-π stacking interactions with F340.

Materials and Methods
All starting materials were purchased from Aldrich or Merck and used without further purification. Microwave oven Plazmatronika 1000 was used (http://www.plazmatronika. com.pl (accessed on 27 December 2020)). Melting points were determined with ElectroThermal 9001 Digital Melting Point apparatus and are uncorrected. High resolution mass spectra were recorded on Quattro LCT (TOF). 1 H NMR, 13 C NMR spectra in solution were recorded at 25 C with a Varian NMRS-300 spectrometer, and standard Varian software was employed. The calculated shielding constants were used as an aid in an assignment of resonances of 13 C atoms. Chemical shifts d [ppm] were referenced to TMS. TLC was carried out using Kieselgel 60 F254 sheets and spots were visualized by UV e 254 and 365 nm.

Conclusions
Sixty new aryl-piperazinyl derivatives of 5-hydroxy-4,7-dimethylchromen-2-one (A), 6acetyl-5-hydroxy-4,7-dimethylchromen-2-one (B) and 8-acetyl-7-hydroxy-4-methylchromen-2-one (C) were designed, synthesized, and evaluated in silico and experimentally for their 5-HT 1A and 5-HT 2A receptor-binding affinities. Figure 3 present summary of the results for the most active compounds. Five compounds showed high antagonistic activities against the 5-HT 1A receptor (1a, 3a, 4a, 5a, and 5b), though lower than WAY-100635, the reference 5HT 1A antagonist, while three compounds showed moderate affinity for 5-HT 2A receptors (5i, 1j, and 5g) with respect to ketanserin. The designed derivatives had two-or five-carbon alkyl linkers between coumarin and arylpiperazinyl moiety. The studies showed that the new compounds showed less profound binding for the tested serotonin receptors than the derivatives containing three-or four-carbon linkers, which we described in our previous works. While the differences in 5HT 1A activities between the three-carbon or four-carbon linker derivatives were minimal, further shortening or lengthening of the linker quite significantly lowered the potency of coumarin derivative to bind to this receptor. Overall, the results for the series of 5-and 7-hydroxycoumarin derivatives obtained in this and our previous investigations on this topic provide an exhaustive structure-activity relationship database, which can be used in future search for novel agents acting on serotonin receptors, either based on coumarin derivatives or other organic scaffolds.