Halogen bonding enhances activity in a series of dual 5-HT₆/D₂ ligands designed in a hybrid bioisostere generation/virtual screening protocol

Abstract

A novel hybrid bioisostere generation/virtual screening method combined with narrowing of chemical space through similarity to compounds that are active at the second target was successfully applied for the development of structurally new dual 5-HT₆/D₂ receptor ligands. Consequently, a series of derivatives of the found hit 1d (N-[2-(dimethylamino)ethyl]-N-(2-phenylethyl)aniline) was synthesized. The most active 5-HT₆/D₂ ligands also showed affinity for 5-HT₇R and 5-HT_2AR. The para-chloroaniline derivative was identified as a potent dual 5-HT₆/5-HT₇ receptor antagonist (K_i = 24 nM and K_b = 30 nM, K_i = 4 nM and K_b = 1.4 nM, respectively). In the case of halogen-containing compounds, interesting structure–activity relationships were observed at 5-HT₆, D₂ and 5-HT₇ receptors, and the ligand–receptor complexes were subsequently examined using a molecular modelling approach that combined quantum-polarized ligand docking (QPLD) and Molecular-Mechanics-Generalized-Born/Surface Area (MM/GBSA) free-energy calculation, which permitted the identification of putative halogen binding pockets.

---

Introduction

New drugs can be designed or searched for in silico using a structure-based or ligand-based approach or both methods together.¹ Structure-based approaches require the structure (obtained either through crystallographic or NMR methods) of a biological target (e.g., protein) that is, in certain cases, difficult to obtain. Indeed, to date, the crystal structures of most serotonin and dopamine (except 5-HT_1BR, 5-HT_2BR and D₃) receptors have not been determined. In turn, ligand-based methods utilize structures of known compounds, both endogenous and exogenous, that function at a given biological target. Among the ligand-based methods is bioisosteric replacement, which includes the modification of the chemical structure of a drug in a way that a new compound with similar biological and modified pharmacokinetic properties is acquired.^2,3 Bioisosteric replacement is commonly used for the development of a series of analogues in hit-to-lead and/or lead optimization processes.

Virtual screening (VS) campaigns are usually aimed at identifying compounds that are active at one selected biological target. It is well known, however, that many drugs interact with more than one target, such as therapeutics that are used for mental disorders with a complex aetiology.^4–6 The design of ligands that display a given pattern of activities for CNS receptors is however a challenging task. Because the bioisostere generation/virtual screening method was successfully applied for the development of new 5-HT₆R ligands,⁷ we assumed that a similar approach combined with narrowing of the chemical space through similarity to ligands of the second target may result in the identification of structurally novel compounds that are active at two given receptors. This assumption could provide a starting point for rational design of ligands with a more complex profile of pharmacological activity. Thus, following the therapeutic properties revealed by agents of serotonergic and dopaminergic systems, the 5-HT₆ receptor was selected as a primary target,^10,11 whereas activity at dopamine D₂ receptor was additionally sought.^12,13 It should be mentioned, however, that the verification of the methodological approach, rather than focusing on development of drug candidates, was the main objective of the project. Nevertheless such dual 5-HT₆/D₂ activity profile has recently been reported for hybrid compounds that display antidepressant-like effects in animal models together with pro-cognitive properties.¹⁴ Moreover, pro-cognitive efficacy of 5-HT₆R antagonists is also suggested as an adjunctive for the primary function of dopamine D₂ receptors, in antipsychotic drug action.^8,9

Here, we report the design of new ligands with dual 5-HT₆R/D₂R biological activity through a multi-step in silico screening protocol. The developed group of N-[2-(dimethylamino)ethyl]-N-(2-phenylethyl)aniline derivatives was additionally tested for affinity for 5-HT_1A, 5-HT_2A and 5-HT₇ receptors. Among the group of developed derivatives, halogen-containing compounds showed interesting structure–activity relationships at 5-HT₆, D₂ and 5-HT₇ receptors, and thus their complexes were further investigated using hybrid quantum mechanic/molecular mechanic (QM/MM) methods.

Bioisostere query

The applied bioisosteric query was designed to identify 5-HT₆R ligands that are similar to any D₂R ligand and thus potentially exhibit the desired dual 5-HT₆R/D₂R activity (Fig. 1). Initially, a set of 4806 5-HT₆R ligands (K_i < 100 nM) extracted from the ChEMBL database was used for bioisostere generation with the PipelinePilot program.¹⁵ Next, the obtained group of 165 657 daughter structures was compared to a set of 3396 D₂R ligands that were also obtained from the ChEMBL database. Only 853 bioisosteres had a Tanimoto similarity (T_c) higher than 0.9, whereas 53 were identical to the corresponding D₂R ligands. These 906 structures that were daughters of the 95 parent structures were subsequently virtually screened for 5-HT₆R affinity. The screening protocol consisted of several steps that included Lipinski and Veber rules, physicochemical properties, ADME/Tox, 5-HT₆R pharmacophores and docking to 5HT₆R homology models.¹⁶ As a result, a group of 31 bioisosteres that were daughters of 20 5-HT₆R ligands was obtained. Visual inspection of these 20 parent structures revealed that they included only 8 chemical scaffolds (see ESI,† page 2). The specificity of the PipelinePilot program caused that none of the obtained scaffolds was unique because they were mostly simple derivatives of their parent compounds. The PipelinePilot utilizes only bioisosteric replacements that can be found in bioactive molecules (taken from the BIOSTER database¹⁷), and thus it has a limited ability to create a new unique chemical scaffold. However, the vBrood program¹⁸ provides analyses of the electro- and stereo-topological properties of molecular fragments and substitutes the fragment with another similar fragment according to established program settings. This procedure allows vBrood to produce dozens of thousands of new, unique chemical scaffolds, the majority of which are, however, nonsense. Eight scaffolds obtained in the first round of the VS protocol were therefore processed in vBrood and provided 26 791 bioisosteres, which were subsequently evaluated using the same virtual screening protocol against 5-HT₆R. One hundred and two bioisosteres, out of these 26 791, passed all VS filters, among which four compounds (1d–1g) were selected in terms of their synthetic accessibility and chemical scaffold novelty (Table 1). The chemical scaffolds (1a–1c) of the parents were synthesized rather than the parents themselves because this study was focused mainly on the effect on activity of the applied bioisosteric replacement. However, among all of the chosen bioisosteres, only 1d was active both at 5-HT₆R and D₂R. Thus, to investigate the structure–activity relationship, a group of its derivatives 4a–4t was obtained.

Fig. 1 Workflow of the applied protocol for developing a dual ligand that acts at 5-HT₆R and D₂R. From the structures of known 5-HT₆R ligands, a database of bioisosteres was generated (using a PipelinePilot program¹⁵), which was subsequently compared with D₂R ligands. The obtained structures with a Tanimoto coefficient > 0.9 were evaluated using the 5-HT₆R VS protocol. Structures that passed all filters were used in a second bioisostere generation step (using the vBrood program¹⁸), and the obtained structures were evaluated again using the same VS filters. The final hits were selected considering the synthetic accessibility and novelty of the chemical scaffolds.

Table 1 Synthesized daughter structures 1d–1g selected from 102 bioisosteres that passed through VS together with their parent compounds and main chemical scaffold

Parent structure	Chemical scaffold	Daughter structure

---

Synthesis

The parent compound 1c was synthesized according to a previously described method in patent US 2010/0120792 (Scheme 1).^19,20 The first step of the synthesis utilized the reaction of phenylhydrazine with N-methyl-4-piperidone in sulfuric acid, which produced N-methyl-2,3,4,5-tetrahydro-1H-pyrido[4,3-b]indole (1a). Next, the product was conjugated to phenylacetylene using KOH, (Bu₄N)₂SO₄ in DMSO and, finally, hydrogenated over 10% Pd/C in methanol to afford target product 1c.

Scheme 1 Synthesis of 2-methyl-5-(2-phenylethyl)-1H,2H,3H,4H,5H-pyrido[4,3-b]indole 1c.

The synthesis of bioisosteres 1d, 4a–4t consisted of two steps (Scheme 2). First, reaction of the appropriate anilines with an alkyl chloride (2-chloro-N,N-dimethylethylamine, 2-chloro-N,N-diethylethylamine or 2-chloro-N,N-dimethylpropylamine) in isopropanol with K₂CO₃ provided intermediates 3a–3u. Second, reaction of the intermediate with phenylacetaldehyde in DCE with triacetoxyborohydride yielded the final products 1d, 4a–4t.

Scheme 2 Synthesis of N-[2-(dimethylamine)ethyl]-N-(2-phenylehtyl)aniline derivatives 1d, 4a–4t.

Synthesis of 1b, 1e and 1f was achieved by the conjugation of 5d with 5i, 5j and 6b in acetonitrile with K₂CO₃ (Schemes 3 and 4). Intermediate 5d was synthesized by, first, bromination of ethyl acrylate, followed by conjugation to catechol, reduction of the ester group with LiAlH₄ and subsequent esterification with tosyl chloride. Intermediates 5i and 5j were obtained by the reaction of 2,4-dinitrochlorobenzene with isonicotinamide, followed by substitution with aniline. The obtained compound 5f was hydrogenated under pressure over platinum or palladium to generate 5g and 5h. The products were reduced using LiAlH₄ to produce 5i and 5j. Intermediate 6b was synthesized by Suzuki–Miyaura coupling of 4-bromobenzaldehyde with phenylboronic acid, followed by the reaction with hydroxylamine and subsequent reduction with NaBH₄. Unfortunately, the attempts to synthesize bioisostere 1g were unsuccessful, and the synthesis of parent 1c was also abandoned.

Scheme 3 Synthesis of (2,3-dihydro-1,4-benzodioxin-2-ylmethyl)[(1-phenyl-piperidin-4-yl)methyl]amine 1b and [(1-cyclohexylpiperidin-4-yl)methyl](2,3-dihydro-1,4-benzodioxin-2-ylmethyl)amine 1e.

Scheme 4 Synthesis of (2,3-dihydro-1,4-benzodioxin-2-ylmethyl)[(4-phenyl-phenyl)methyl]amine 1f.

Pharmacology

Radioligand binding assays were used to determine the affinity and selectivity profiles of the synthesized compounds for human serotonin 5-HT_1AR, 5-HT_2A, 5-HT₆R, 5-HT_7bR and D_2LR, which were stably expressed in HEK293 cells. This procedure was accomplished via the displacement of [³H]-8-OH-DPAT (187 Ci mmol⁻¹) for 5-HT_1AR, [³H]-Ketanserin (47.3 Ci mmol⁻¹) for 5-HT_2A, [³H]-LSD (85.2 Ci mmol⁻¹) for 5-HT₆R, [³H]-5-carboxyamidotryptamine (5-CT, 39.2 Ci mmol⁻¹) for 5-HT_7bR and [³H]-Raclopride (74.4 Ci mmol⁻¹) for D_2LR. Each compound was tested in triplicate at 7 to 8 different concentrations (10⁻¹¹ to 10⁻⁴ M). The inhibition constants (K_i) were calculated using the Cheng–Prusoff equation,²¹ and the results are expressed as the mean of at least two independent experiments (Table 2). The antagonistic properties of 4m on 5-HT₆R and 5-HT₇R were evaluated based on its ability to inhibit cAMP production induced by the agonist 5-CT (100 nM in the case of 5-HT₆R, 10 nM in the case of 5-HT₇R) in HEK293 cells overexpressing 5-HT₆ or 5-HT₇ receptors (Table 3). The experiments were performed in triplicate at 8 concentrations (10⁻¹¹ to 10⁻⁴ M).

Table 2 Binding affinities^a of the synthesized compounds 1a, 1b, 1d, 1e, 1f and 4a–4t

Cmpd.	n	Substitution				Ki [nM]
		R1	R2	R3	R4	5-HT6	D2	5-HT7	5-HT1A	5-HT2A
a Binding affinity, Ki, expressed as the average of at least two independent experiments; the maximum S.D. did not exceed 34% (see ESI, page 5); n.d. – not determined.
1a	—	—	—	—	—	22	77	8	217	29
1b	—	—	—	—	—	2280	210	430	10	n.d.
1d	1	Me	H	H	H	63	476	121	1682	233
1e	—	—	—	—	—	2675	1609	3428	22	n.d.
1f	—	—	—	—	—	3760	110	857	101	n.d.
4a	1	Et	H	H	H	51	266	205	7614	128
4b	2	Me	H	H	H	67	55	623	4486	228
4c	1	Me	H	H	F	23	37	31	5707	58
4d	1	Me	H	H	Me	25	33	10	482	70
4e	1	Me	H	H	CF3	61	274	30	5341	809
4f	1	Me	H	H	OMe	234	1147	141	>10 000	863
4g	1	Me	H	Me	H	264	n.d.	2253	605	777
4h	2	Me	H	H	F	38	80	38	8969	27
4i	1	Me	H	F	H	64	594	748	>10 000	312
4j	1	Me	Cl	H	Cl	597	1176	387	>10 000	472
4k	1	Me	H	OMe	H	355	1738	4512	>10 000	721
4l	2	Me	H	H	Me	83	108	24	2140	204
4m	1	Me	H	H	Cl	24	153	4	5960	212
4n	1	Me	Me	H	H	380	n.d.	1040	>10 000	348
4o	1	Me	Naphthyl		H	137	423	2037	4749	621
4p	1	Me	H	F	F	21	n.d.	114	>10 000	79
4q	1	Me	H	H	Br	90	505	19	n.d.	295
4r	1	Me	H	H	I	79	691	10	n.d.	372
4s	1	Me	Me	H	Me	347	1751	597	n.d.	518
4t	1	Me	H	Cl	H	150	675	353	n.d.	311

---

Table 3 Functional activity of 4m and reference compounds at 5-HT₆R and 5-HT₇R

Compound	Kb [nM]
	5-HT6R	5-HT7R
a A selective, potent 5-HT6R antagonist currently in phase II clinical trials.b A selective, potent 5-HT7R antagonist under development by GSK.
4m	30 (antagonist)	1.4 (antagonist)
SB-742457 (ref. 22)^a	0.5 (antagonist)	—
SB-269970 (ref. 23)^b	—	0.9 (antagonist)

---

Results and discussion

The T_c among 102 structures, obtained through tandem bioisosteric queries and the investigation of 5-HT₆R and D₂R ligand databases, ranged from 0.43 to 0.88. The obtained bioisostere 1d was active at both 5-HT₆ and D₂ receptors, validating the effectiveness of the designed bioisosteric query for creating a multi-receptor ligand (Table 2). Consequently, compound 1d was chosen as the lead for a SAR series (4a–4t).

Bioisosteres 1e and 1f were found to be inactive at 5-HT₆R, but they possessed moderate-to-low affinity for D₂R (1e K_i = 1609 nM, 1f K_i = 110 nM) and high-to-moderate affinity for 5-HT_1AR (1e K_i = 22 nM, 1f K_i = 101 nM). Interestingly, substitution of the phenyl group with cyclohexane (compounds 1b and 1e) resulted in a loss of activity at D₂R and only a 2-fold decrease in affinity for 5-HT_1AR. In contrast, substitution of the phenylpiperidine group with a biphenyl group (compounds 1b and 1f) resulted in a 2-fold increase in affinity for D₂R and a 10-fold decrease in affinity for 5-HT_1AR. This result suggests that aromatic interactions may be more important for ligand–receptor (L–R) interactions in D₂R compared with 5-HT_1AR.

Compared with the parent compound 1a, the hit 1d exhibited 3- to 15-fold lower affinity for all targets. However, the introduction of a fluorine and methyl at position 4 of the aniline aromatic ring (4c and 4d, respectively) resulted in compounds that possessed affinity comparable to the parent 1d for 5-HT₆R (4c K_i = 23 nM, 4d K_i = 25 nM) and were 2-fold more active at D₂R (4c K_i = 37 nM, 4d K_i = 33 nM).

An affinity similar to 5-HT₆R was also obtained for compounds 4m and 4p (K_i = 24 and K_i = 21, respectively), but they were 2-fold less active at D₂R compared with the parent compound. Compound 1a, in addition to its affinity for 5-HT₆R and D₂R, was also highly active at 5-HT₇R and 5-HT_2A (K_i = 8 and 29 nM, respectively) and moderately active at 5-HT_1A (K_i = 217 nM). Only one compound bearing the 3-(N,N-dimethylamino)propyl and 4-fluorophenyl group (4h), achieved comparable affinity for 5-HT_2A (K_i = 27 nM) but was less active at other targets. Compound 4m, in turn, was twice as active towards 5-HT₇ receptor (K_i = 4 nM) as the parent 1a, together with a comparable affinity for 5-HT₆R (K_i = 24 nM), but it exhibited 2- to 27-fold lower affinity for other targets. In addition, functional activity assays revealed that 4m possessed antagonistic properties at 5-HT₆R and 5-HT₇R, with the latter being comparable to the reference compound SB-269970 (Table 3). Analysis of structure–activity relationship in a series of 1d derivatives revealed a preference for halogens (chlorine, bromine and iodine), especially in the case of binding to 5-HT₇R. It appears that there may be a specific site in the 5-HT₇ receptor binding pocket that can form halogen bonds with a ligand. Among the three mentioned halogens, iodine forms the strongest halogen bonds.²⁴ However, the iodine-substituted compound 4r exhibited lower affinity for 5-HT₇R (K_i = 10 nM) than the chlorine-substituted compound 4m (K_i = 4 nM), although its affinity was higher than the bromine derivative 4q (K_i = 19 nM). This finding indicates that the putative halogen binding pocket may be large enough to accommodate chlorine but too small to fully accommodate larger halogens. Nevertheless, the halogen-bonding properties of iodine allowed it, in part, to counter this unfavourable steric hindrance and resulted in a higher affinity for 5-HT₇R than for bromine. A similar observation was obtained for the binding to 5-HT₆R; the chlorine-substituted compound (4m) belongs to derivatives with the highest 5-HT₆R affinity, together with fluorine and methyl-substituted compounds (4c, 4d). However, in this case, compounds (4q, 4r) that were substituted with heavier halogens, bromine and iodine, exhibited an approximately 4-fold lower affinity (K_i = 90 nM and K_i = 79 nM, respectively) than compound 4m, suggesting that the putative halogen binding pocket in 5-HT₆R is smaller than that in 5-HT₇R.

The binding affinity data suggested that a similar halogen binding pocket might also be present in D₂R. This hypothesis is supported by the observation that the introduction of a chlorine increases the affinity 3-fold compared with an unsubstituted hit 1d. However, introduction of a bromine and iodine resulted in a decrease in affinity, which might be explained by the potentially small volume of the putative halogen binding cavity. This hypothesis is supported by the observation that compounds containing small groups, such as methyl and fluorine (4d, 4c, respectively), showed the highest affinity for D₂R (K_i = 33 nM and 37 nM, respectively) and that compound containing a 4-trifluoromethyl group (3f) displayed a lower affinity (D₂R K_i = 274 nM) than those with a 4-methyl group (4d) – with van der Waals volumes equal to 39.8 and 21.6 ų for CF₃ and CH₃, respectively.²⁵

To gain insight into the observed changes in affinity of the obtained series of compounds, they were all re-docked to homology models of 5-HT₆, D₂ and 5-HT₇ receptors. The combination of QPLD from the Schrödinger Suite with MM-Generalized-Born/Surface Area (MM/GBSA) calculations was used to obtain L–R complexes because this approach is able to describe the anisotropy of the electron density around halogen atoms, which is a key feature during halogen bond examination. Only the top scored complexes, according to ΔG, were considered. The binding modes (Fig. 2) were proposed based on the mutual spatial arrangement of particular ligand fragments and 5-HT₆, D₂ and 5-HT₇ receptor binding pockets. All of the obtained L–R complexes exhibited very consistent binding modes (within each receptor type), in line with the findings of Kooistra et al.²⁶ and potentially associated with the similar binding pocket construction of these receptors (Table 4). In addition to the crucial charge-assisted hydrogen bond with Asp3.32, all of the docked ligands formed at least one specific aromatic interaction (CH–π or π–π stacking) with Phe6.51 or Phe6.52 residues. Additionally, the second aromatic ring of a ligand was usually targeted to Phe3.28 (D₂R/5-HT₇R) and Trp3.28 (5-HT₆R).

Fig. 2 Representative L–R complexes (top scores based on ΔG) of selected ligands with 5-HT₆, D₂, and 5-HT₇ receptors. Figures (A–C) show a comparison of docking poses for compounds 1d (orange) and 4m (yellow), whereas figures (D–F) show a comparison for compounds 4d (cyan) and 4f (magenta). Amino acids that were selected as crucial for the binding of the presented compounds are shown as sticks. A basic nitrogen atom forms a charge-assisted hydrogen bond with aspartic acid Asp3.32, an aniline aromatic ring (in the case of D₂R and 5-HT₇R complexes) interacts with Phe6.52, whereas in the case of 5-HT₆R, the same interaction is formed by a phenylethyl moiety. The ΔΔG [kcal mol⁻¹] value shows the difference between ΔG of complexes of a particular compound (4m, 4d, 4f) and an unsubstituted analogue 1d. In most cases, the ΔΔG values correspond to the binding affinity; the greatest energy gain was obtained by the introduction of a chlorine atom (4m, yellow), regardless of the receptor.

Table 4 Sequence alignment of the 5-HT₆, D₂ and 5-HT₇ receptors involving amino acids considered to be responsible for ligand interactions^a

a Identical amino acids in receptor binding pocket sequences are indicated by a grey background. Amino acids indicated in Fig. 1 are shown in bold.

---

The exchange of a methyl with a methoxy group in the para position resulted in a 9- (for 5-HT₆R) to 35-fold (D₂R) decrease in affinity. Analysis of the binding modes showed that the methoxy group occupied the same binding cavity as the methyl group (Fig. 2E), which formed many hydrophobic interactions that stabilized the L–R complex (e.g., Val5.39, Ser5.42, Phe6.52, and His6.55 in D₂R). It should also be noted that for the methoxy substituent, no hydrogen bonding with side chain amino acids was detected (e.g., Ser5.42 in D₂R), and thus it destabilized the L–R complex through unfavourable polar interactions.

The recognized halogen interaction appeared to possess a highly directional nature. To explain the 4- (D₂R) to 88-fold (5-HT₇R) decrease in activity by shifting the chlorine atom from position 4 to 3 in the phenyl ring, interaction spheres were plotted onto relevant backbone carbonyl oxygens (Fig. 3). In each case, the 4-chloro substituent was positioned within the energetically favourable areas of the sphere, whereas in the meta substitution, the 3-Cl atom pointed outside of the sphere, indicating that halogen bonding did not occur.

Fig. 3 A superposition of the top scored poses of 4-Cl (4m, yellow) and 3-Cl (4t, blue) derivatives against putative halogen binding pocket interaction spheres. The ΔΔG [kcal mol⁻¹] value shows the difference between the ΔG of complexes of a particular compound (4m, 4t) and an unsubstituted analogue 1d. The higher binding energy value for 3-Cl than for 4-Cl derivatives illustrates the highly directional nature of the identified halogen bond interaction. The methodology applied has been described by Wilcken et al.²⁷ Geometrical parameters for the halogen bonds are: 5-HT₆R d(Cl⋯O) = 3.7 Å, ∠(C–Cl⋯O) = 141°, D₂R d(Cl⋯O) = 3.2 Å, ∠(C–Cl⋯O) = 126°, 5-HT₇R d(Cl⋯O) = 2.9 Å, ∠(C–Cl⋯O) = 131°.

Conclusions

An average hit rate for virtual screening to identify new compounds with affinity for a given biological target seldom exceeds 10%. The multi-step protocol applied herein differed from standard virtual screening by the introduction of additional bioisosteric derivatization steps and a similarity filter for the ligand of the second target. These modifications allowed us to obtain new chemical scaffolds that would not be identified in commonly utilized virtual screening techniques for commercially available compound databases. Among 102 structures that fulfilled all of the VS filters, only four were chosen for further evaluation. Among them, one compound (1d) possessed the desired dual 5-HT₆R/D₂R activity, two (1e, 1f; bioisosteres of the same parent) were inactive at 5-HT₆R and the synthesis of one (1g) was unsuccessful. It is probable that the remaining screening hits might also exhibit activity similar to 1d, but their synthesis was determined to be more challenging.

The hybrid QM/MM analysis performed for the series of 1d derivatives allowed to identify a putative halogen binding pocket in 5-HT₆, D₂ and 5-HT₇ receptors. It must be stressed that halogen bonds in complexes with serotonin receptors have only recently been proposed in the case of 5-HT₆R²⁸ and 5-HT₇R²⁹ but with different anchoring points: the carbonyl oxygen of Pro4.60/Thr5.46 and Cys3.36(sulfur)/Thr3.37, respectively. Here, the characteristics of putative halogen binding pockets also differed between the investigated receptors (5-HT₆, D₂ and 5-HT₇) because the affinity gain resulting from the introduction of the halogen atom was not equivalent and depended on the volume of the halogen binding cleft; the most profound gain was observed for the chloro derivative 4m, which resulted in a 30-fold increase in affinity for 5-HT₇R (compared with the hit 1d).

Because the geometry of the recognized halogen bonds (see Fig. 3) is not optimal, they should be regarded rather as weak interactions. It has to be stressed, however, that receptor conformation is not optimized during Schrödinger QPLD protocol, thus detailed halogen bonding parameters (distance, sigma hole angle) and contribution of hydrophobic interaction would have required further studies including, e.g., full quantum mechanics optimization of L–R complexes.

Experimental

The ¹H NMR spectra were recorded using a Bruker Avance III HD 400 NMR spectrometer. Mass spectra were recorded using a TQD Waters LC/MS spectrometer with electrospray ionization. Substrates and solvents were purchased from Sigma-Aldrich and Apollo Scientific company and used without further purification.

Detailed synthetic, biomolecular and molecular modelling procedures are presented in ESI.†

General procedure 1 for the synthesis of N-[2-(dialkylamino)alkyl]anilines (3a–3u)

A 100 ml round bottom flask was charged with aniline (0.054 mol), appropriate (2-chloroalkyl)dialkylamine hydrochloride (0.018 mol), anhydrous potassium carbonate (0.11 mol) and isopropanol (50 ml). The reaction was maintained at reflux for 12 hours. After completion, the reaction mixture was filtered, the residuum was washed with isopropanol and the solvent was evaporated under reduced pressure. The product was purified by column chromatography on silica using ethyl acetate : methanol (9 : 1).

General procedure 2 for the synthesis of N-[2-(dialkylamino)alkyl]-N-(2-phenylethyl)anilines (1d, 4a–4t)

A 100 ml round bottom flask was charged with appropriate N-[2-(dialkylamino)alkyl]aniline (6 mmol), phenylacetaldehyde (12 mmol), triacetoxyborohydride (9 mmol) and dichloroethane (20 ml). The reaction was maintained at room temperature for 24 hours. After completion, distilled water (100 ml) was added, and the reaction mixture was extracted with dichloromethane (2 × 100 ml). The collected organic extracts were washed with brine, dried over anhydrous magnesium sulfate and evaporated under reduced pressure. The product was purified by column chromatography on silica using ethyl acetate : methanol (9 : 1).

Acknowledgements

The study was partially supported by the Polish-Norwegian Research Programme operated by the National Centre for Research and Development under the Norwegian Financial Mechanism 2009–2014 in the frame of the Project PLATFORMex (Pol-Nor/198887/73/2013) and by the National Science Center Grant No. DEC-2014/15/D/NZ7/01782.

References

1. G. Sliwoski, S. Kothiwale, J. Meiler and E. W. Lowe, Pharmacol. Rev., 2014, 66, 334–395.
2. L. M. Lima and E. J. Barreiro, Curr. Med. Chem., 2005, 12, 23–49.
3. S. Schann, K. Pompermayer, J. Feldman, B. Pfeiffer, P. Renard, E. Scalbert, P. Bousquet and J. Ehrhardt, J. Med. Chem., 2001, 44, 1588–1593.
4. R. Morphy and Z. Rankovic, J. Med. Chem., 2005, 48, 6523–6543.
5. D. H. Kim, M. J. Maneen and S. M. Stahl, Neurotherapeutics, 2009, 6, 78–85.
6. J. L. Medina-Franco, M. A. Giulianotti, G. S. Welmaker and R. A. Houghten, Drug Discovery Today, 2013, 18, 495–501.
7. J. Staroń, D. Warszycki, J. Kalinowska-Tłuscik, G. Satała and A. J. Bojarski, RSC Adv., 2015, 5, 25806–25815.
8. A. Nikiforuk, Rev. Neurosci., 2014, 25, 367–382.
9. K. Fijał, P. Popik and A. Nikiforuk, Psychopharmacology, 2014, 231, 269–281.
10. A. Wesołowska and A. Nikiforuk, Neuropharmacology, 2007, 52, 1274–1283.
11. A. Quiedeville, M. Boulouard, V. Da Silva Costa-Aze, F. Dauphin, V. Bouet and T. Freret, Rev. Neurosci., 2014, 25, 417–427.
12. A. Zhang, J. L. Neumeyer and R. J. Baldessarini, Chem. Rev., 2007, 107, 274–302.
13. N. Ye, J. L. Neumeyer, R. J. Baldessarini, X. Zhen and A. Zhang, Chem. Rev., 2013, 113, PR123–PR178.
14. M. Kołaczkowski, M. Marcinkowska, A. Bucki, J. Śniecikowska, M. Pawłowski, G. Kazek, A. Siwek, M. Jastrzębska-Więsek, A. Partyka, A. Wasik, A. Wesołowska, P. Mierzejewski and P. Bienkowski, Eur. J. Med. Chem., 2015, 92, 221–235.
15. W. A. Warr, J. Comput.-Aided Mol. Des., 2012, 26, 801–804.
16. R. Kurczab, M. Nowak, Z. Chilmonczyk, I. Sylte and A. J. Bojarski, Bioorg. Med. Chem. Lett., 2010, 20, 2465–2468.
17. I. Ujváry, Pestic. Sci., 1997, 51, 92–95.
18. BROOD Fragment Replacement for Medicinal Chemistry, Version 1.1.2 and 2.0.0, OpenEye Scientific Software, Inc., Santa Fe, NM, USA, 2010.
19. A. A. Ivashchenko, et al., U.S. 0120792, 2010.
20. Marvin was used for drawing, displaying and characterizing chemical structures, substructures and reactions, Marvin 6.2.2, 2014, ChemAxon (http://www.chemaxon.com).
21. Y.-C. Cheng and W. H. Prusoff, Biochem. Pharmacol., 1973, 22, 3099–3108.
22. N. Upton, T. T. Chuang, A. J. Hunter and D. J. Virley, Neurotherapeutics, 2008, 5, 458–469.
23. J. J. Hagan, G. W. Price, P. Je, N. J. Deeks, T. Stean, D. Piper, M. I. Smith, N. Upton, A. D. Medhurst, D. N. Middlemiss, G. J. Riley, P. J. Lovell, S. M. Bromidge and D. R. Thomas, Br. J. Pharmacol., 2000, 130, 539–548.
24. T. Clark, M. Hennemann, J. S. Murray and P. Politzer, J. Mol. Model., 2007, 13, 291–296.
25. F. Leroux, ChemBioChem, 2004, 5, 644–649.
26. A. J. Kooistra, S. Kuhne, I. J. P. de Esch, R. Leurs and C. de Graaf, Br. J. Pharmacol., 2013, 170, 101–126.
27. R. Wilcken, M. O. Zimmermann, A. Lange, S. Zahn and F. M. Boeckler, J. Comput.-Aided Mol. Des., 2012, 26, 935–945.
28. B. Benhamú, M. Martín-Fontecha, H. Vázquez-Villa, L. Pardo and M. L. López-Rodríguez, J. Med. Chem., 2014, 57, 7160–7181.
29. V. Canale, P. Guzik, R. Kurczab, P. Verdie, G. Satała, B. Kubica, M. Pawłowski, J. Martinez, G. Subra, A. J. Bojarski and P. Zajdel, Eur. J. Med. Chem., 2014, 78, 10–22.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra08714k

---