Introduction

Serotonin or 5-hydroxytryptamine (5-HT) is a monoamine neurotransmitter that has been linked to the regulation of mood, attention, aggression, aversive learning, impulse, and reward [1, 2]. A disrupted serotonin system can result in pathological conditions such as depression, Alzheimer’s disease, schizophrenia, and migraine [3]. Serotonin exerts its effect by interacting with a complex network of receptors, which was grouped into seven subfamilies (5-HT1–5-HT7) according to their structure, transduction, and function. These targets reside in the central and peripheral nervous systems (CNS/PNS), as well as a number of non-neuronal cells, such as in the gastrointestinal, blood, and endocrine systems [1, 4].

There are currently 14 known 5-HT receptor subtypes. Among them, 13 receptor subtypes belong to the G protein-coupled receptor (GPCR) superfamily, and 5-HT3 is classified as an ion channel receptor [5,6,7]. 5-HT receptors can be further grouped by the G proteins they are coupled with. The 5-HT1 and 5-HT5 subtypes are coupled with Gi/Go protein, which suppresses the activity of adenylyl cyclase (AC) and decreases the level of cyclic adenosine monophosphate (cAMP). In contrast, 5-HT4, 5-HT6, and 5-HT7 increase the activity of AC and increase the level of cAMP by coupling with the Gs protein [8]. The 5-HT2 receptor couples with Gq/G11 protein, which promotes the activity of phospholipase C and increases the levels of inositol trisphosphate and calcium cation (Ca2+) [9].

Various drugs that interact with 5-HT receptor subtypes are marketed for a wide range of serotonin-related indications [10, 11]. For example, vilazodone, a partial agonist of the 5-HT1A receptor, is indicated as a depression treatment; triptans such as sumatriptan and almotriptan act as selective 5-HT1B/1D agonists and are marketed as a migraine reliever; and risperidone, a commonly used antipsychotic, acts as a 5-HT2A/2C antagonist [12]. Moreover, 5-HT receptors are commonly associated with drug abuse due to the crucial roles they play in the development of addiction to various pharmaceutical and recreational drugs. Drugs such as cocaine, amphetamine, methamphetamine, and 3,4-methylenedioxymethamphetamine (MDMA) are known as psychostimulant drugs [13, 14]. When administered into the human body, psychostimulant drugs can interact with monoamine transporters and lead to an increase in the activity of extracellular 5-HT, dopamine, and noradrenaline in the brain. Increased 5-HT levels in combination with dopamine is a key mechanism of drug addiction [14]. Among the 7 subclasses (5-HT1–7) of serotonin receptors, 5-HT1A, 5-HT1B, 5-HT2A, and 5-HT3 are especially well known for their association with addictive substances, such as cocaine, amphetamine, methamphetamine, MDMA (ecstasy), morphine/heroin, cannabis, alcohol, and nicotine [15].

Wang et al. [16] reported the crystal structures of the human 5-HT1B receptor bound to its agonists ergotamine and dihydroergotamine. Their structures revealed that these ligands shared similar binding modes in 5-HT1B, occupying the orthosteric pocket and an extended binding pocket close to the extracellular loops. They also compared the structure with the 5-HT2B receptor and found that the 5-HT1B receptor displayed a 3 Å outward shift at the intracellular end of TM6, resulting in a more open extended pocket that may explain the subtype selectivity. To investigate the structural basis for biased signaling, Wacker et al. [17] reported the crystal structure of the human 5-HT2B receptor bound to ERG (Protein Data Bank (PDB) entry: 4IB4, resolution: 2.7 Å) and compared it with the 5-HT1B/ERG structure. These crystal structures provide a comprehensive structural basis for understanding the receptor–ligand interactions of the 5-HT family subtypes. However, due to high sequence similarity/identity among 5-HT receptors and the unavailability of the three-dimensional (3D) structures of the different 5-HT receptors, no reports systematically compare the key residues and selective residues involved in the binding pocket, making it difficult to design subtype-selective serotonergic drugs.

In the present work, we built 3D models for all 5-HT GPCRs based on the existing crystal structure of 5-HT1B/5-HT2B/5-HT2C using homology modeling. To explore our 3D models, we compared the binding results of 5-HT receptors with similar pockets to identify the causes of the variation in binding affinity. Our analysis indicated that the affinity variation may potentially be due to the selective residues located in the binding pocket. Moreover, we carried out molecular dynamics (MD) simulations for all the 5-HT receptors complexed with their reported ligands, and the results were consistent with previous findings. Finally, we conducted off-target prediction for Captagon using systems pharmacology analysis. The findings from the present work provide insights into the detail selectivity of the 5-HT receptor binding pockets, which could aid drug discovery and design of 5-HT receptor-related medication with high specificity to avoid potential addiction and drug abuse.

Materials and methods

Homology modeling and validation of 5-HT receptors

In this study, we used our previous homology model of human serotonin (5-HTs). Briefly, these models were constructed based on the crystal structure of 5-HT1B (PDB entry: 4IAQ/4IAR, resolution: 2.8/2.7 Å) [16], 5-HT2B (PDB entry: 4IB4, resolution: 2.7 Å) [17], and 5-HT2C (PDB entry: 6BQH, resolution: 2.7 Å) [18]. The sequences of the human GPCRs were obtained from UniProtKB/Swiss-Prot (https://www.uniprot.org/uniprot/). We truncated some residues from the N and C terminals. As the third intracellular loop (between helix 5 and helix 6) has a long, flexible sequence, we only kept ~15 residues. Then, Modeller 9.18 [19] was used to construct the homology models by (a) searching and selecting template(s) for the target protein, (b) conducting sequence alignment between the target and template(s), (c) adjusting the sequence alignment using the residue tables from the GPCR database (GPCRdb) [20] (http://gpcrdb.org/residue/residuetable), and (d) building and evaluating the homology models.

Docking study of ligands and receptors

We adopted the MOLCAD module implemented in SYBYL-X 1.3 to explore the potential binding pockets of receptors. The docking program Surflex-Dock GeomX (SFXC) in SYBYL-X 1.3 was used to construct receptor–ligand complexes in which the docking scores were expressed in −log10 (Kd) [21]. The main protocols or parameters of docking were addressed in our previous publications [22,23,24,25]. Briefly, the docking parameters were as follows: (a) the “number of starting conformations per ligand” was set to 10, and the “number of max conformations per fragment” was set to 20; (b) the “maximum number of rotatable bonds per molecule” was set to 100; (c) flags were turned on at “pre-dock minimization”, “post-dock minimization”, “molecule fragmentation”, and “soft grid treatment”; (d) “activate spin alignment method with density of search” was set to 9.0; and (e) the “number of spins per alignment” was set to 12.

MD simulations

The 5-HT receptors complexed with ligands were set up for MD simulation. For example, the 5-HT2b receptor complexed with inhibitor BF-1 system was put in 0.15 M NaCl solution and formed a cubic water box, which included 26,481 water molecules, 72 Na+ ions, and 82 Cl− ions. The initial configurations of the protein receptors and ligands were taken from docking studies. The sizes of the initial simulation boxes were ~100 Å × 100 Å × 100 Å. The other systems were set up with the same protocol.

The AMBER ff14SB force field [26] was applied to proteins. Water molecules were treated with the TIP3P (transferable intermolecular potential with 3 points) water model [27]. The partial atomic charges of ligands were derived using the semi-empirical with bond charge correction (AM1-BCC) method [28, 29]. The other force field parameters were obtained from GAFF in AMBER16 [29]. The residue topologies for ligands were prepared using the ANTECHAMBER module.

The MD simulations were carried out using the PMEMD.mpi and PMEMD.cuda modules in the AMBER16 [30,31,32] package. First, several minimization steps were conducted for the systems to avoid possible steric crashes. Then, each system was gradually heated from 0 K to 300 K during the heating stage and maintained at 300 K during the subsequent equilibrium and production stages. A time step of 2 fs was used for the heating stage, equilibrium stage, and the entire production stage. A periodic boundary condition was employed to maintain constant temperature and the pressure) ensembles. The pressure was set at 1 atm and controlled by the anisotropic (x-, y-, z-) pressure scaling protocol with a pressure relaxation time of 1 ps. The temperature was regulated using Langevin dynamics with a collision frequency of 2 ps−1 [33, 34]. The particle mesh Ewald (PME) method [35, 36] was adopted to handle long-range electrostatics and a 10 Å cutoff was set to treat real-space interactions. All covalent bonds involving hydrogen atoms were constrained with the SHAKE algorithm [37]. Each system was subject to a 50 ns MD simulation and the trajectory of simulated systems was saved every 100 ps.

Molecular mechanics/generalized born surface area (MM/GBSA) calculation

For the saved trajectories of MD simulations, the MM/GBSA [38,39,40,41,42,43,44] method was used to calculate the binding energies of receptors treated with different ligands. A total of 300 snapshots were extracted from each trajectory every 100 ps from 20 to 50 ns to calculate the mean binding energy. The formula is as follows:

$$\Delta {\it{E}}_{\mathrm{{bind}}} = \Delta {\it{E}}_{\mathrm{{MM}}} + \Delta {\it{E}}_{\mathrm{{SOL}}} = \Delta {\it{E}}_{\mathrm{{MM}}} + \Delta {\it{E}}_{\mathrm{{GB}}} + \Delta {\it{E}}_{\mathrm{{SA}}},$$

where △Ebind is the binding energy and △EMM denotes the sum of the molecular mechanical energies in vacuo and can be further divided into the contributions from electrostatic, van der Waals, and internal energies. This term could be computed using the molecular mechanics method. △ESOL is the solvation energy, which includes the polar solvation energy (△EGB) calculated with the generalized born (GB) approximation model [45, 46] and the non-polar part (△ESA) obtained by fitting solvent accessible surface area (SASA) [47] with the linear combinations of pairwise overlaps (LCPO) model [48, 49]. Additionally, the energies of each residue were decomposed into the backbone and side-chain atoms. The energy decomposition can be analyzed to determine the contributions of the key residues to the binding [50].

Hallucinogen-specific chemogenomics knowledgebase and systems pharmacology analysis

We constructed a hallucinogen-specific chemogenomics database [51] that can be used for target, off-target, additional identification, and systems pharmacology analysis of small molecules and their potential targets. Several in-house chemoinformatics tools were utilized, including TargetHunter, HTDocking, and the blood–brain barrier (BBB) predictor [52, 53]. HallucinogenPlatform (http://www.cbligand.org/hallucinogen/) collected 144 hallucinogen-related target proteins and 145 chemical compounds associated with these targets in 6721 assays and 23,598 references.

In the present work, we applied our HallucinogenPlatform and established chemoinformatics tools such as HTDocking to perform systems pharmacological analysis for Captagon. First, Captagon was docked into the target protein pockets. We matched these predicted target proteins to Captagon according to their docking scores. targets with higher docking scores may have higher binding affinities and therefore a greater chance of interacting with Captagon. Next, we mapped out a pharmacological network of interactions between drug compounds and target proteins at the molecular level [54, 55]. Cytoscape 3.4.0 [56] was used to generate, analyze, and visualize the network of targets and drugs/compounds, as described previously [55].

Results

Overview of the 5-HT receptor family binding pockets

Fig. 1 shows the 3D structures of 5-HT receptor variations constructed in the present study based on the crystal structures of 5-HT1B/5-HT2B/5-HT2C. The binding pocket and key amino acid residues associated with the binding pocket of the receptors are also indicated in Fig. 1. These binding pockets were generally composed of transmembrane helices III, V, VI, and VII. Being classified as receptors within the same family, 5-HT receptors share many commonalities in their amino acid residues. These conserved amino acid residues include aspartate (Asp) on helix III position 32 (3.32), tryptophan (Trp) on helix VI position 48 (6.48), phenylalanine (Phe) at helix VI position 51 (6.51) and 52 (6.52), and tyrosine (Tyr) on helix VII position 42 (7.42). These common amino acid key residues can be used to explain the similar functionality and affinity between these receptors. These key residues involved in the binding pocket of 5-HT receptors are listed in Table 1. The unique residues in individual receptors are highlighted in red and may play a key role in their selectivity (see below).

Fig. 1
figure 1

The three-dimensional (3D) structure, binding pocket, and key residues of 5-hydroxytryptamine (5-HT) receptors family. a The binding pocket and key residues at 5-HT1A. b The binding pocket and key residues at 5-HT1B. c The binding pocket and key residues at 5-HT1D. d The binding pocket and key residues at 5-HT1E. e The binding pocket and key residues at 5-HT1F. f The binding pocket and key residues at 5-HT2A. g The binding pocket and key residues at 5-HT2B. h The binding pocket and key residues at 5-HT2C. i The binding pocket and key residues at 5-HT4. j The binding pocket and key residues at 5-HT5. k The binding pocket and key residues at 5-HT6. l The binding pocket and key residues at 5-HT7. The binding pockets at 5-HT receptors are generally  formed by transmembrane helix III, V, VI, and VII

Full size image
Table 1 Key residues involved in the binding pocket of 5-hydroxytryptamine (5-HT) receptors
Full size table

5-HT1A and 5-HT7 bound with selective 5-HT1A antagonist NAD-299

The 5-HT1A and 5-HT7 are subtypes of the 5-HT serotonin receptor family that have different pharmacological mechanisms. To explore the differences in selectivity, we performed and compared molecule docking between these two receptors and NAD-299 (a selective 5-HT1A antagonist [57], Table 2).

Table 2 Pharmacological properties of important drugs or chemicals used in the present work (nM)
Full size table

As shown in Fig. 2, the binding pocket of 5-HT1A/5-HT7 was mainly composed of TM3, TM5, TM6, and TM7, and was defined by the reported key residues shown in Table 1. The binding pocket of 5-HT1A and 5-HT7 shared many similar residues, such as Asp116/Asp1623.32, Cys120/Cys1663.36, Ser199/Ser2435.43, Ala203/Ala2475.461, Trp358/Trp3406.48, Phe361/Phe3436.51, Phe362/Phe3446.52, and Tyr390/Tyr3747.42. NAD-299 resulted in similar conformations when docked into both receptors, with the chromane structure facing TM7 and four carbon rings facing TM5 and TM6. Although 5-HT1A and 5-HT7 share similar residues, their binding affinity with NAD-299 is very different. The binding affinity between 5-HT1A and NAD-299 is 0.59 nM (Table 2), whereas the binding affinity between NAD-299 and 5-HT7 is much weaker (1900 nM, Table 2). We suggest that this drastic difference in binding affinity is due to the slight variation in key residues related to the binding pocket of the 5-HT receptor subtypes. A major distinction in interactions that could contribute to this large difference in binding affinity is the ability to form a strong hydrogen bond between NAD-299 and the binding pocket. When docking NAD-299 to the binding pocket of 5-HT1A, three potential hydrogen bonds involving the oxygen on the amide group of NAD-299 were observed, with bond lengths of ~2.1 Å (with oxygen on Asp1163.32), ~3.4 Å (with nitrogen on Asn3867.38), and ~3.6 Å (with oxygen on Tyr3907.42), while only one possible hydrogen bond was observed in the binding pocket of 5-HT7 between the nitrogen on the amide group of NAD-299 and the single bonded oxygen on the carboxyl group of Asp1623.32 (~2.9 Å). The different interactions of the unique residues Asn386/Leu3707.38 greatly contributed to the different binding affinity. Compared to Asp1623.32, the lack of an additional nitrogen on Leu3707.38 in close vicinity of the oxygen on the amide group of NAD-299 also contributed to the inability to form a hydrogen bond. Aside from the difference in hydrogen bonds, the number of hydrophobic interactions between NAD-299 and the two binding pockets varies drastically. There were 14 hydrophobic interactions between NAD-299 and the binding pocket of 5-HT1A (some of which are not shown in Fig. 2) compared to only 6 hydrophobic interactions in the binding pocket of 5-HT7. The potential selective residues, including Ala3656.55 and Ala3837.35 on 5-HT1A, greatly contributed to the hydrophobic interaction. This difference in the number of hydrophobic interactions is another key factor in the selectivity of the binding pockets, which can also be explained by the difference in amino acid residues in the binding pocket and the variation in the conformations of similar residues.

Fig. 2
figure 2

The interaction between NAD-299 and 5-HT1A and 5-HT7. a The interactions between 5-HT1A and NAD-299. b The interactions between 5-HT7 and NAD-299. The binding affinity between 5-HT1A and NAD-299 is 0.59 nM, while the binding affinity between NAD-299 and 5-HT7 is only 1900 nM. Ala383/Arg3677.35 and Asn386/Leu3707.38 contributed to the selectivity. 5-HT 5-hydroxytryptamine

Full size image

5-HT1B, 5-HT1D, and 5-HT4 bound with selective 5-HT1B inverse agonist SB-236057

The 5-HT1 subgroup is one of the seven subgroups in the 5-HT receptor subfamily. Individual subtypes within the subgroup closely resemble each other structurally, functionally, and transductionally. To explore the selectivity among these receptors, we performed molecular docking between the selective 5-HT1B inverse agonist SB-236057 [58] (Table 2) and 5-HT1B /5-HT1D. We also performed docking between SB-236057 (Table 2) and 5-HT4 and compared the differences.

As shown in Fig. 3, the binding pocket of all three receptors as mainly composed of TM3, TM5, TM6, and TM7, and the reported key residues are shown in Table 1. The binding pose of SB-236057 in all three pockets was very similar (Fig. 3), with the oxadiazole structure and two benzene rings parallel to the transmembrane structure facing up and away from the residues and the tricyclic structure perpendicular to the transmembrane helixes.

Fig. 3
figure 3

The interaction between SB-236057 and 5-HT1B, 5-HT1D and 5-HT4. a The interactions between 5-HT1B and SB-236057. b The interactions between 5-HT1D and SB-236057. c The interations between 5-HT4 and SB-236057. The binding affinity between SB-236057 and 5-HT1B/5-HT1D is ~6.31 nM and ~501.19 nM, while the binding affinity between SB-236057 and 5-HT4 is ~3981.07 nM. 5-HT 5-hydroxytryptamine

Full size image

The 5-HT1B and 5-HT1D share all the key amino acid residues at the same locations on the transmembrane domains (Fig. 3). The binding pocket of 5-HT1B and 5-HT1D is mainly composed of Asp129/Asp1183.32, Ile130/Ile1193.33, Cys133/Cys1223.36, Tyr208/Tyr1975.39, Thr209/Thr1985.40, Ser212/Ser2015.43, Ala216/Ala2055.461, Trp327/Trp3146.48, Phe330/Phe3176.51, Phe331/Phe3216.52, Ser334/Ser3216.55, Asp352/Asp3397.35, Thr355/Thr3427.38, and Tyr390/Tyr3747.42. Although the binding pockets of 5-HT1B/5-HT1D consist of identical key residues, we found a large difference in the binding affinity between SB-236057 and 5-HT1B/5-HT1D (~6.31 nM and ~501.19 nM, respectively, Table 2). Although the residues in the binding pocket of 5-HT1B and 5-HT1D were consistent, we observed that the conformation of the side chain in some particular residues may play an important role in their selectivity. When docking SB-236057 into 5-HT1B, one hydrogen bond (3.9 Å) formed between Cys1333.36 and the oxygen on the tricyclic structure of SB-236057; another hydrogen bond (3.5 Å) formed between the single bonded oxygen on the carboxyl group of the Ser2125.43 and the nitrogen on the tricyclic structure of SB-236057. However, when docking SB-236057 to 5-HT1D, Ser2015.43 was the only residue capable of forming a hydrogen bond with SB-236057. The key difference was the conformation of Cys133/Cys1223.36 in the binding pocket. In the binding pocket of 5-HT1B, Cys1333.36 was in the conformation where the silicone faced SB-236057, making the hydrogen bond feasible, while in the binding pocket of 5-HT1D, Cys1223.36 faced away from the small molecule, making the formation of a hydrogen bond difficult.

The binding affinity between SB-236057 and 5-HT4 is ~3981.07 nM (Table 2), which is weaker than that of 5-HT1B. This difference in selectivity between 5-HT1B and 5-HT4 can be explained by the variation in binding residues. The main distinction in the key 5-HT1B and 5-HT4 residues include Ile130/Val1013.33, Cys133/Thr1043.36, Thr209/Ala1935.40, Ser212/Cys1965.43, Ser334/Asn2796.55, Asp352/Thr2957.35, and Thr355/Leu2987.38. As shown in Fig. 3, two potential strong hydrogen bonds can be observed in the binding pocket of 5-HT4 but, unlike 5-HT1B, 5-HT4 has a threonine (Thr1043.36) in place of the cysteine at helix III position 36 (3.36), which results in the formation of a hydrogen bond between the oxygen on the hydroxyl group of Thr1043.36 and the oxygen on the tricyclic structure of SB-236057. Compared to the hydrogen bond between Cys1333.36 and SB-236057 in the binding pocket of 5-HT1B, the hydrogen bond between Thr1043.36 and SB-236057 has a bond angle of almost 90 degrees, resulting in a weaker binding affinity.

5-HT7 and 5-HT2 subgroups bound with selective 5-HT2B receptor agonist BF-1

Each subtype in the 5-HT2 subgroup closely resembles each other structurally, functionally, and transductionally. There are three subtypes within the 5-HT2 subgroup: 5-HT2A, 5-HT2B, and 5-HT2C. The key residues for these receptors had been identified as shown in Table 1. The 5-HT2 subtypes shared similar residues. The variations among the key residues include: (1) 5-HT2B has methionine on helix V position 40 (Met2185.40), while the other two receptor subtypes have valine; (2) 5-HT2A has serine on helix V position 461 (Ser2425.461), while the other two receptor subtypes have alanine; and (3) 5-HT2B has glutamic acid on helix VII position 35 (Glu3637.35), while the other two receptor subtypes have asparagine acid. Although the subtypes within this subgroup are greatly similar, the selectivity among these subtypes is still reported. To explore it, we performed molecular docking between BF-1 [59] (the selective 5-HT2B receptor agonist) and all three 5-HT2 receptors. Additionally, we also docked BF-1 to 5-HT7 for further analysis.

The binding affinity of BF-1 (Table 2) is ~0.09 nM in 5-HT2B, ~2.82 nM in 5-HT2A, and ~22.91 nM in 5-HT2C. As shown in Fig. 4, the binding pose of BF-1 was almost the same in these receptors. To future explore this binding result, we looked at the binding pocket of each 5-HT2 receptor. All 5-HT2 receptors showed the ability to form two hydrogen bonds with BF-1 at the same position: one between oxygen on BF-1 and the oxygen on the Asp155/Asp135/Asp1343.32 (~3.7 Å, ~3.1 Å, and 2.9 Å respectively), and another between the same oxygen on BF-1 and the Tyr370/Tyr370/Tyr3587.42 (~3.4 Å, ~3.5 Å, and 3.4 Å respectively). Moreover, hydrophobic interaction played an important role in their selectivity. 5-HT2A, 5-HT2B, and 5-HT2C shared similar hydrophobic interactions. However, the residue at position 5.40 at 5-HT2B is Met2185.40 and different from that of 5-HT2A and 5-HT2C(Val235/Val2155.40), 5-HT2B was able to form an additional hydrophobic interaction, which Met2185.40 may contribute to the selectivity of 5-HT2B.

Fig. 4
figure 4

The interaction between BF-1 and 5-HT2B, 5-HT2A, 5-HT2C, and 5-HT7. a The interactions between 5-HT2B and BF-1. b The interactions between 5-HT2A and BF-1. c The interactions between 5-HT2C and BF-1. d The interactions between 5-HT7 and BF-1. The binding affinity of BF-1 is ~0.09 nM in 5-HT2B, ~2.82 nM in 5-HT2A, ~22.91 nM in 5-HT2C, and ~ 66.07 nM in 5-HT7. 5-HT 5-hydroxytryptamine

Full size image

The binding affinity between BF-1 and 5-HT7 is ~66.07 nM (Table 1). When comparing 5-HT2B and 5-HT7, we noticed seven different residues (Ser139/Cys1663.36, Met218/Thr2405.40, Val366/Leu3707.38) and interactions between them, resulting in different binding affinity of BF-1.

5-HT2A, 5-HT2B, 5-HT2C, and 5-HT1D subgroup bound with selective 5-HT2C receptor agonist SB-242084

As shown in Fig. 5, the binding pocket of 5-HT2A/5-HT2B/5-HT2C/5-HT1D mainly consisted of TM3, TM5, TM6, and TM7 (Table 1). Similar key residues among the binding pockets of these four receptors included Asp155/Asp135/Asp134/Asp1183.32, Phe339/Phe340/Phe327/Phe3176.51, Phe340/Phe341/Phe328/Phe3186.52, and Tyr370/Tyr370/Tyr358/Tyr3467.42. Even with many similarities, the binding affinities of SB-242084 [60] to these four receptors are extremely different. The binding affinity between 5-HT1D and SB-242084 is 398.11 nM (Table 2), whereas the binding affinity between SB-242084 and 5-HT2C is 1 nM (Table 2). Subtle differences were observed in the binding pockets of these four receptors, resulting in SB-242084 having different detailed interactions with 5-HT2A, 5-HT2B, 5-HT2C, and 5-HT1D.

Fig. 5
figure 5

The interaction between SB-242084 and 5-HT2C, 5-HT2B, 5-HT2A, 5-HT1D. a The interactions between 5-HT2C and SB-242084. b The interactions between 5-HT2B and SB-242084. c The interactions between 5-HT2A and SB-242084. d The interactions between 5-HT1D and SB-242084. The binding affinity between 5-HT1D and SB-242084 is 398.11 nM, while the binding affinity between SB-242084 and 5-HT2C is only 1 nM. 5-HT 5-hydroxytryptamine

Full size image

Fig. 5 illustrates the interaction between SB-242084 and these four 5-HT receptors. When docking SB-242084 to the receptor subtypes, the amino group of SB-242084 formed a hydrogen bond with residues in these receptors with different distances: Asp1343.32 (~2.6 Å) and Tyr3587.42 (~3.0 Å) in 5-HT2C; Asp1353.32 (~3.3 Å) and Tyr3707.42 (~3.7 Å) in 5-HT2B; Asp1553.32 (~2.5 Å) and Tyr3707.42 (~2.9 Å) in 5-HT2A; and Asp1183.32 (~3.1 Å) and Tyr3467.42 (~3.7 Å) in 5-HT1D. In addition to the common hydrogen bond mentioned above, the nitrogen atom on the pyridine ring of SB-242084 also formed hydrogen bonds with Asp1343.32 (~2.5 Å) and Ser1383.36 (~3.3 Å) in the binding pocket of 5-HT2C. This resulted in a total of four hydrogen bonds, making the binding pocket of 5-HT2C the most selective toward SB-242084. In addition to the hydrogen bonds, two residues Val3547.38 (~3.8 Å) and Val1353.33 (~4.1 Å) contributed to the strong hydrophobic interaction in 5-HT2C. In the binding pocket of 5-HT2B, one additional hydrogen bond between the nitrogen in terminal pyridine ring and Asn3446.55 (~4.3 Å) was observed. SB-242084 can also form hydrophobic interactions with Val1363.33 (~3.5 Å), Val3667.38 (~3.9 Å), and Phe2175.39 (~4.0 Å) in 5-HT2B. In the binding pocket of 5-HT2A and 5-HT1D, there were only two common hydrogen bonds noted above, leading to the lower affinity. For 5-HT2A, we observed four hydrophobic interactions between SB-242084 and the key residues (~3.7 Å and ~4.1 Å with Val1563.33, ~3.6 Å with Val3667.38, and ~4.1 Å with Ser2425.46), which was more than for 5-HT1D (~3.9 Å with Thr3427.38, ~3.4 Å with Ile1193.33, and ~3.8 Å with Ala2055.46).

5-HT2A and 5-HT6 receptor subtypes bound with selective 5-HT6 receptor agonist RO63-0563

As shown in Fig. 6, the binding pocket of 5-HT2A/5-HT6 was mainly composed of TM3, TM5, TM6, and TM7 (Table 1). Similar key residues were observed in the binding pockets, including Asp155/Asp1063.32, Val156/Val1073.33, Phe234/Phe1885.39, Val235/Val1895.40, Phe339/Phe2846.51, Phe340/Phe2856.52, Asn343/Asn2886.55, and Tyr370/Tyr3107.42. The binding affinity between 5-HT2A and RO63-0563 [61] was reported to be larger than 10,000 nM (Table 2), whereas the binding affinity between RO63-0563 and 5-HT6 was 12.3 nM (Table 2). Molecular docking was performed to explore the binding affinity, considering the number and distance of hydrophilic and hydrophobic interactions between RO63-0563 and these two 5-HT receptors.

Fig. 6
figure 6

The interaction between RO63-0563 and 5-HT6 and 5-HT2A. a The interactions between 5-HT6 and RO63-0563. b The interactions between 5-HT2A and RO63-0563. The binding affinity between RO63-0563 and 5-HT6 is 12.3 nM, while the binding affinity between 5-HT2A and RO63-0563 is larger than 10,000 nM. 5-HT 5-hydroxytryptamine

Full size image

Fig. 6 shows the selective 5-HT6 receptor agonist RO63-0563 binding with two different 5-HT subtypes, the 5-HT2A, and 5-HT6 receptors. When RO63-0563 is bound to 5-HT6 receptor, the secondary amines on the pyridine ring form hydrogen bonds with the Ser1935.43 (~3.3 Å) and Asn2886.55 (~3.0 Å); Asp3037.35 forms a hydrogen bond with the primary amino groups on the phenyl ring (~3.4 Å); the oxygen atom on sulfonyl forms hydrogen bonds with Asp1063.32 (~2.1 Å) and Tyr3107.42 (~4.0 Å); and the phenyl ring forms a hydrophobic interaction with Thr3067.38 (~3.0 Å). RO63-0563 can also interact with Val1895.40 (~3.3 Å) and Cys1103.36 (~3.1 Å) via hydrophobic interactions. When binding RO63-0563 to 5-HT2A, the 5-HT2A receptor can only interact with Phe2345.39 (~3.5 Å), Ser1593.36 (~3.5 Å), and Val3667.38 (~3.3 Å) in a hydrophobic manner, which agrees with the data from the literature. No strong hydrogen bond was observed between RO63-0563 and 5-HT2A, resulting in RO63-0563 having a significantly higher binding affinity to 5-HT6 than 5-HT2A.

5-HT7, 5-HT2A, 5-HT2B, and 5-HT1D receptor subtypes bound with selective 5-HT7 receptor agonist SB-656104

As shown in Table 1, similar key residues comprise the binding pocket of 5-HT1D/5-HT2A/5-HT2B/5-HT7 receptors, including Asp118/Asp155/Asp135/Asp1623.32, Phe317/Phe339/Phe340/Phe3436.51, Phe318/Phe340/Phe341/Phe3446.52, and Tyr346/Tyr370/Tyr370/Tyr3747.42. SB-656104 [62] has higher selectivity towards the 5-HT7 receptor than the other three receptors. The affinity between SB-656104 and 5-HT7 is 2.0 nM (Table 2), which is stronger than the affinity for 5-HT1D (25.12 nM, Table 2), 5-HT2A (63.10 nM, Table 2) and 5-HT2B (91.20 nM, Table 2). Observation of the detailed molecular interaction was made using molecular docking to provide an explanation for this phenomenon.

As shown in Fig. 7, the interactions between SB-656104 and the 5-HT1D, 5-HT2A, 5-HT2B, and 5-HT7 receptors differed subtly but distinctly. When docking SB-656104 to the 5-HT7 receptor, the nitrogen atom on the piperidine ring formed a hydrogen bond with one of the oxygen atoms on the carboxyl of Asp1623.32 (~3.5 Å); the hydroxyl of Ser2435.43 (~2.7 Å) and Tyr2305.30 (~3.8 Å) formed hydrogen bonds with oxygen atoms on the sulfonyl group of SB-656104; the nitrogen atom of Arg3677.35 (~3.0 Å) formed a hydrogen bond with the oxygen atom (ether bond) connecting the benzene ring and piperidine; and the nitrogen of indole in SB-656104 can form a hydrogen bond with the Val1633.33 (~3.4 Å). SB-656104 formed hydrophobic interactions with Ala2475.46 (~4.0 Å), Cys1663.36 (~3.1 Å), sTrp3406.48 (~3.4 Å), and Val1633.33 (~3.9 Å). SB-656104 formed five hydrogen bonds and four important hydrophobic bonds with 5-HT7, which is consistent with its higher affinity.

Fig. 7
figure 7

The interaction between SB-656104 and 5-HT7, 5-HT1D, 5-HT2A, and 5-HT2B. a The interactions between 5-HT7 and SB-656104. b The interactions between 5-HT1D and SB-656104. c The interactions between 5-HT2A and SB-656104. d The interactions between 5-HT2B and SB-656104. The affinity of SB-656104 with 5-HT7 is 2.0 nM, which is less than the affinity with 5-HT1D (25.12 nM), 5-HT2A (63.10 nM), and 5-HT2B (91.20 nM)

Full size image

In the binding pocket of 5-HT1D, the formation of two hydrogen bonds was observed when binding with SB-656104. One hydrogen bond formed between the oxygen atoms on the sulfonyl group of SB-656104 and Ser2015.43 (~3.0 Å) and the other hydrogen bond was observed between the five-membered ring and Ser3246.58 (~4.1 Å). SB-656104 also formed several hydrophobic interactions with Thr1985.40 (~3.3 Å), Cys1223.36 (~3.4 Å), and Trp3146.48 (~3.7 Å).

In the binding pocket of 5-HT2A, one hydrogen bond formed between the carboxyl group of Asp1553.32 (~3.4 Å) and the nitrogen atom on the piperidine. In addition, SB-656104 formed hydrophobic interactions with Val2355.40 (~3.4 Å), Val3667.38 (~3.5 Å), Val1563.33 (~3.7 Å), Ser1593.36 (~3.2 Å), and Trp3366.48 (~3.4 Å). In the binding pocket of 5-HT2B, one hydrogen bond formed between the nitrogen in the indole and Val1363.33 (~3.0 Å). In addition, five hydrophilic interactions with Val1363.33 (~3.5 Å), Ala2255.46 (~4.0 Å), Ser1393.36 (~3.4 Å), Asp1353.32 (~3.4 Å), and Trp3376.48 (~3.0 Å) are shown in Fig. 7. Considering the same number of hydrogen bonds and hydrophobic bonds with different bond distances, the binding affinity of SB-656104 and 5-HT2A is only slightly greater than that of 5-HT2B.

MD simulation between 5-HT receptors and ligands

To verify the proposed interactions and investigate the dynamic interactions, we carried out 50 ns MD simulations for six 5-HT receptors, 5-HT1A, 5-HT1B, 5-HT2B, 5-HT2C, 5-HT6, and 5-HT7, and their specific inhibitors.

For the 5-HT1A/NAD-299, 5-HT2B/BF-1, 5-HT2C/SB-24208, and 5-HT7/SB-656104 system, the root mean square deviation (RMSD) of the receptors and inhibitors fluctuated approximately 2.8 Å, 1.9 Å, 1.4 Å, 1.8 Å (black line in Fig. 8) and 1.8 Å, 1.8 Å, 0.8 Å, 2.6 Å (red line in Fig. 8) from 0 ns to 50 ns, respectively, indicating these systems remain quite stable during the simulation. For the 5-HT1B/SB-236057 complex, the RMSD of the protein slowly increased from ~1.4 Å at 0 ns to ~3.0 Å at 13 ns, then remained stable for the rest of simulation (black line in Fig. 8b); the RMSD of SB-236057 fluctuated around ~ 1.3 Å from 0 ns to 50 ns. For the 5-HT6/RO63-0563 system, the RMSD of 5-HT6 fluctuated approximately 1.8 Å and the RMSD of RO53-0563 increased from ~2.0 Å at 0 ns to ~2.4 Å at 12 ns, then sharply decreased to ~2.1 Å and fluctuated around that value for the remainder of the simulation.

Fig. 8
figure 8

The root mean square deviation (RMSD) of 5-HT (5-HT1A, 5-HT1B, 5-HT2B, 5-HT2C, 5-HT6, and 5-HT7) receptors complexed with compound for 50 ns molecular dynamics (MD) simulation. a The RMSD of 5-HT1A and NAD-299. b The RMSD of 5-HT1B and SB-236057. c The RMSD of 5-HT2B and BF-1. d The RMSD of 5-HT2C and SB-242084. e The RMSD of 5-HT6 and RO63-0563. f The RMSD of 5-HT7 and SB-656104.  All the systems are stable during the MD simulation. 5-HT 5-hydroxytryptamine

Full size image

All systems remained in a stable conformation for more than 30 ns, so 300 snapshots were extracted from 20 ns to 50 ns and used to calculate the mean binding energy. According to the binding energy of key residues, we determined the most important key residues and their contributions to receptor–ligand interactions. The residues at positions 3.32 and 3.33 (Table 1) in the third alpha helix significantly contribute to receptor–ligand binding in all systems except 5-HT6/RO63-0563 (Fig. 9e and Supplementary Table S5) due to their ability to form strong hydrogen bonds. In particular, Asp1343.32 (−8.868 kcal/mol, Supplementary Table S4) in 5-HT2c and Asp1623.32 in 5-HT7 (−6.793 kcal/mol, Supplementary Table S6) greatly contributed to the total binding of 5-HT2c/SB-242084 and 5-HT7/SB-656104 due to their strong electrostatic interactions (−11.362 kcal/mol/−10.807 kcal/mol, Supplementary Table S4/S6). Val1173.33 (−0.906 kcal/mol, Supplementary Table S1) in 5-HT1A, Ile1303.33 (−4.934 kcal/mol, Supplementary Table S2) in 5-HT1B, Val1363.33 (−3.884 kcal/mol, Supplementary Table S3) in 5-HT2B, Val1353.33 (−3.181 kcal/mol, Supplementary Table S4) in 5-HT2C, and Val1633.33 (−4.216 kcal/mol, Supplementary Table S6) in 5-HT7 greatly contributed to the binding of ligands in these residues. Interestingly, the binding between 5-HT6 and RO63-0563 (Fig. 9e) was mainly a result of residue Asp3037.35 (total energy: −8.921 kcal/mol, Supplementary Table S5) owing to electrostatic interaction (−9.033 kcal/mol, Supplementary Table S5), and Asn2886.55 (total energy: −2.454 kcal/mol, Supplementary Table S5) also contributed to the binding.

Fig. 9
figure 9

Free energy decomposition of key residues for 5-HT (5-HT1A, 5-HT1B, 5-HT2B, 5-HT2C, 5-HT6, and 5-HT7) receptors complexed with compounds during 50 ns molecular dynamics (MD) simulation. a The energy decomposition of key residues at 5-HT1A. b The energy decomposition of key residues at 5-HT1B. c The energy decomposition of key residues at 5-HT2B. d The energy decomposition of key residues at 5-HT2C. e The energy decomposition of key residues at 5-HT6. f The energy decomposition of key residues at 5-HT7. The contributions of key residues are consistent with our docking results. 5-HT 5-hydroxytryptamine

Full size image

Moreover, we performed 50 ns MD simulations for six other 5-HT receptors, 5-HT1D, 5-HT1E, 5-HT1F, 5-HT2A, 5-HT4, and 5-HT5, and their specific ligands. Fig. 10a–f shows the RMSD of the receptors and ligands. Our MD results show that the systems remained stable during the MD simulation. The RMSDs of all the receptors and ligands were approximately ~1.2 Å to 2.8 Å. For these six systems, 300 snapshots were extracted from 20 ns to 50 ns and used to calculate the mean binding energy, and the results are shown in Fig. 11. The MD results correlate well with the docking results. For example, Ile1193.33 and Ser3216.55 in 5-HT1D greatly contributed to the binding of SB-236057; Asp1033.32 and Ile1043.33 were important in the recognition of SB-242084 in 5-HT1E; Ile1043.33, Phe3106.52, and Glu3136.55 in 5-HT1F played key roles in the recognition of BF-1; Phe2355.40, Phe3406.52, and Leu3637.35 were important for BF-1 in 5-HT2A; Asp1003.32, Val1013.33, and Asn2796.55 were the key residues for SB-656104 in 5-HT4; and Asp1213.32, Val1223.33, and Tyr2005.39 played important roles for the binding of SB-656104 in 5-HT5.

Fig. 10
figure 10

The root mean square deviation (RMSD) of 5-HT (5-HT1D, 5-HT1E, 5-HT1F, 5-HT2A, 5-HT4, and 5-HT5) receptors complexed with compound for 50 ns molecular dynamics (MD) simulation. a The RMSD of 5-HT1D and SB-236057. b The RMSD of 5-HT1E and SB-242084. c The RMSD of 5-HT1F and BF-1. d The RMSD of 5-HT2A and BF-1. e The RMSD of 5-HT4 and SB-656104. f The RMSD of 5-HT5 and SB-656104. All the systems are stable during the MD simulation. 5-HT 5-hydroxytryptamine

Full size image
Fig. 11
figure 11

Free energy decomposition of key residues for 5-HT (5-HT1D, 5-HT1E, 5-HT1F, 5-HT2A, 5-HT4, and 5-HT5) receptors complexed with compounds during 50 ns molecular dynamics (MD) simulation. a The energy decompostion of key residues at 5-HT1D. b The energy decompostion of key residues at 5-HT1E. c The energy decompostion of key residues at 5-HT1F. d The energy decompostion of key residues at 5-HT2A. e The energy decompostion of key residues at 5-HT4. f The energy decompostion of key residues at 5-HT5. The contributions of key residues are consistent with our docking results. 5-HT 5-hydroxytryptamine

Full size image

Systems pharmacology analysis of Captagon using our hallucinogen knowledgebase

Captagon, the trademark name for the synthetic stimulant fenethylline, is a CNS stimulator with stronger and longer-lasting effects on aggression, detachment, cognitive enhancement, and alertness than one of its main metabolites, amphetamine [63,64,65,66,67,68].

Captagon was first synthesized in 1961 and can be metabolized into amphetamine (24.5% of oral dose) and theophylline (13.7% of oral dose). Captagon was banned in most countries by the 1980s due to its addictive properties and lack of medical use [63,64,65,66,67,68]. Captagon trafficking has been a big problem in the Middle East for many years. Between 2013 and the end of August 2016, Lebanon seized more than 70 million Captagon pills, worth $14 billion, which accounts for only 10% of the total production. In May 2017, French authorities reported confiscating approximately 300 pounds of Captagon, seizing 750,000 pills worth an estimated $1.7 million. In July 2018, Lebanon seized 3.7 million Captagon pills from Syria. Moreover, Captagon is demonized as a war drug: fighters on Captagon during combat may feel a sense of well-being, euphoria, and invincibility. It is clear that Captagon can be used to suppress pain and increase aggression in soldiers. The use of these drugs is not limited to soldiers and involves the civilian population in areas prevailing in hopelessness and helplessness.

First, we carried out the BBB prediction for Captagon. The BBB predictor was built by applying the SVM (Support Vector Machine) and LiCABEDS (Ligand Classifier of Adaptively Boosting Ensemble Decision Stumps) [69, 70] algorithms to four types of fingerprints (MACCS, PubChem, Molprint 2D, and FP2) of 1593 reported compounds [71]. Unity was used in our original publication and was replaced with PubChem Structure Fingerprint in a subsequent upgrade. The BBB predictor accepts a query compound with any given combination of algorithm and fingerprint as the input and predicts whether or not the query compound can move across the BBB to the central nervous system (CNS). In our scenario, the predictor was run eight times for all possible algorithm and fingerprint combinations. The results indicate that Captagon can move across the BBB to the CNS (as shown in Fig. 12).

Fig. 12
figure 12

Blood−brain barrier (BBB) prediction for Captagon using our established algorithms. a Adaboost algorithm with MACCS fingerprint. b Adaboost algorithm with PubChem fingerprint. c Adaboost algorithm with Molprint 2D fingerprint. d Adaboost algorithm with FP2 fingerprint. e SVM algorithm with MACCS fingerprint. f SVM algorithm with Pubchem fingerprint. g SVM algorithm with Molprint 2D fingerprint. h SVM algorithm with FP2 fingerprint

Full size image

Using our established knowledgebase and computational tools, we performed off-target prediction of Captagon with 5-HT receptors due to the lipophilicity of Captagon [72]. As shown in Fig. 13 (left), we mapped the prediction of these potential molecular targets of Captagon via the docking scores. Interesting, our results showed that most of the docking scores were higher than 7.0 due to the flexible structure of Captagon. Moreover, we suggest that 5-HT2c (docking score 8.84), 5-HT5a (docking score 8.91), and 5-HT7 (docking score 9.93) were the most promising targets for Captagon before metabolism. As an illustration of the detailed interactions between Captagon and 5-HT receptors, we explored the detailed docking pose of Captagon and 5-HT7, as shown in Fig. 13 (right). Our docking results showed that Asp1623.32 (2.5 Å) and Arg3677.35 (2.9-3.2 Å) in 5-HT7 formed strong hydrogen bonds with Captagon and several residues, including Val1633.33 (3.3 Å), Trp3406.48 (3.6 Å), and Phe3446.52 (4.7 Å), contributed to the hydrophobic or π–π interactions.

Fig. 13
figure 13

Systems pharmacology analysis of Captagon and the predicted targets using our established hallucinogen knowledgebase and HTDocking target identification program. a Predicted targets/off-targets for Captagon. b The detailed interactions between Captagon and 5-HT7. 5-HT2c (docking score: 8.84), 5-HT5a (docking score: 8.91), and 5-HT7 (docking score: 9.93) were the most promising targets for Captagon before metabolism

Full size image

To further validate the role of these residues, we carried out 50 ns MD simulations for 5-HT7 complexed with Captagon. Fig. 14a shows that the RMSDs of Captagon and 5-HT7 remained stable at ~2.8 Å and 1.9 Å during the MD. The energy decomposition is shown in Fig. 14b. Our results show that Asp1623.32 greatly contributed to the binding of Captagon, with electrostatic energy of −3.75 kcal/mol and total energy of −3.85 kcal/mol. Moreover, the total energy contribution of Val1633.33 and Arg3677.35 was −3.3 kcal/mol and −2.4 kcal/mol. These results are consistent with our docking results, indicating Captagon may bind to 5-HT7. Future experiments will be carried out to validate the interactions between Captagon and 5-HT receptors.

Fig. 14
figure 14

The root mean square deviation (RMSD) and energy decomposition of 5-HT7 and Captagon during 50 ns molecular dynamics (MD) simulation. a The RMSD of 5-HT7 and Captagon. b The energy decomposition of key residues at 5-HT7. Asp1623.32 contributed greatly to the binding of Captagon, with −8.2 kcal/mol for electrostatic and −6.8 kcal/mol for the total energy. Moreover, the total energy of Val1633.33 and Phe3446.52 were −3.5 kcal/mol and −2.4 kcal/mol. 5-HT 5-hydroxytryptamine

Full size image

Discussion

The residue at helix III position 32 (3.32) is a commonly conserved residue for the binding of GPCR receptors. In the present study, we found that Asp3.32 at this position of the 5-HT receptors greatly contributed to the formability of hydrogen bonds and the total binding energy, which is consistent with the literature. Moreover, Trp6.48, Phe6.51, and Phe6.52 at helix VI and Tyr7.42 at helix VII are conserved residues involved in the binding pocket of GPCR receptors. Our docking and MD results show the great energy contribution of these residues to the total binding energy between ligands and receptors. For example, phenylalanine residue (Phe6.51) played a key role in the total binding energy in 5-HT1A, 5-HT1B, 5-HT2B, and 5-HT2C, which is in line with the literature. Tyr7.42 interacted with small molecules inside the binding pockets of many 5-HT receptor subtypes, including 5-HT1A, 5-HT1D, 5-HT2A, 5-HT2B, 5-HT2C, and 5-HT6.

For selectivity, we found that Asn3867.38 (bold, Table 1), a unique residue found in 5-HT1A, contributed greatly to the formation of strong hydrogen bonds and the total binding energy. Moreover, three hydrophobic residues, Val1173.33, Asn3656.55, and Ala3837.35 (bold, Table 1), were unique in 5-HT1A, which may contribute to the selectivity via potential hydrophobic interactions. Based on our results, Met1033.33 in 5-HT1E (bold, Table 1), Ser1815.39 and Ala3337.38 in 5-HT1F (bold, Table 1) were unique residues, and these residues may contribute to the selectivity of these two receptors. Moreover, our docking results and MD results indicate that Ser2425.461 in 5-HT2A and Met2185.40/Glu3637.35 in 5-HT2B (bold, Table 1) could be major contributors to their selectivity. In addition, Thr1043.36 (bold, Table 1) is highly likely to be a major contributor to the selectivity of 5-HT4 due to its uniqueness and ability to form strong hydrogen bonds. Finally, Ser3217.35 in 5-HT5, Thr1965.461 in 5-HT6, and Arg3677.35 in 5-HT7 (bold, Table 1) were unique to each receptor and played essential roles in their binding processes.

The results from the hallucinogen knowledgebase, molecular docking study, and MD simulation performed on Captagon provide new information regarding the mechanism of Captagon. From the target prediction and interactions between Captagon and these 5-HT receptors, our results suggest that 5-HT2C, 5-HT5A, and 5-HT7 were the most promising targets for unmetabolized Captagon. We aim to conduct additional experiments in the future to explore the interactions between Captagon and 5-HT receptors in more detail.

Conclusion

In the present work, we built 5-HT receptor subtype three-dimensional models based on the published crystal structure of 5-HT1B/5-HT2B/5-HT2C through homology modeling. Based on the results of our molecular docking and MD simulation studies, we identified commonalities in uniqueness in residues that greatly contributes to the selectivity of each receptor. We hope the present work can be used to facilitate drug discovery of 5-HT-related medications to design compounds with high selectivity for each receptor subtype to decrease undesirable events such as addiction and side effects.