Human Serotonin 5-HT_2C G Protein-Coupled Receptor Homology Model from the β₂ Adrenoceptor Structure: Ligand Docking and Mutagenesis Studies

Abstract

Activation of the serotonin (5-hydroxytryptamine, 5-HT) 5HT_2C G protein-coupled receptor (GPCR) is proposed as novel pharmacotherapy for obesity and neuropsychiatric disorders. In contrast, activation of the 5-HT_2A and 5-HT_2B GPCRs is associated with untoward hallucinogenic and cardiopulmonary effects, respectively. There is no crystal structure available to guide design of 5-HT_2C receptor-specific ligands. For this reason, a homology model of the 5-HT_2C receptor was built based on the crystal structure of the human β₂ adrenoceptor GPCR to delineate molecular determinants of ligand–receptor interactions for drug design purposes. Computational and experimental studies were carried out to validate the model. Binding of N(CH₃)₂-PAT [(1R, 3S)-(−)-trans-1-phenyl-3-N,N-dimethylamino-1,2,3,4-tetrahydronaphthalene], a novel 5-HT_2C agonist/5-HT_2A/2B inverse agonist, and its secondary [NH(CH₃)-PAT] and primary (NH₂-PAT) amine analogs were studied at the 5-HT_2C wild type (WT) and D3.32A, S3.36A, and Y7.43A 5-HT_2C point-mutated receptors. Reference ligands included the tertiary amines lisuride and mesulergine and the primary amine 5-HT. Modeling results indicated that 5-HT_2C residues D3.32, S3.36, and Y7.43 play a role in ligand binding. Experimental ligand binding results with WT and point-mutated receptors confirmed the impact of D3.32, S3.36, and Y7.43 on ligand affinity.

Introduction

Ligands that activate the serotonin 5-hydroxy-tryptamine (5-HT_2C) G protein-coupled receptor (GPCR) may be therapeutic for obesity, psychoses, psychostimulant addiction, and other neuropsychiatric disorders [1, 2]. Clinically, there is essentially no tolerance for activation of 5-HT_2A and 5-HT_2B receptors that can lead to untoward hallucinogenic and cardiopulmonary effects, respectively. The 5-HT₂ GPCR family shares approximately 80% sequence homology [3] and have a common primary signaling pathway, complicating 5-HT_2C receptor-specific agonist development.

GPCRs putatively share a three-dimensional (3D) structure consisting of a bundle of seven transmembrane (TM) alpha helices, connected by alternating intracellular and extracellular loops, with the N-terminus in the extracellular domain and C-terminus in the intracellular domain. To date, GPCR crystal structures have been reported for bovine rhodopsin (bRho) [4–8], opsin, [9, 10], turkey β₁ adrenoceptor (β₁AR) [11], human β₂ AR (β₂AR) in an inactive state [12–15], human A_2A adenosine receptor (AA₂AR) [16], a nanobody-stabilized active state of the β₂AR [17], and the structure of an irreversible agonist–β₂AR complex [18]. The 3D structures of the serotonin GPCR families (5-HT₁, 5-HT₂, 5-HT₄, 5-HT₅, 5-HT₆, and 5-HT₇) remain elusive, as is the case for most other GPCRs. A recent review summarizes the GPCRs structure known to date [19].

All ligands for aminergic neurotransmitter GPCRs contain a basic amine moiety. The endogenous neurotransmitters have a primary amine moiety, but, ligands with secondary or tertiary amine groups also are known, including, N(CH₃)₂-PAT [(1R, 3S)-(−)-trans-1-phenyl-3-N,N-dimethylamino-1,2,3,4-tetrahydronaphthalene], a novel 5-HT_2C agonist/5-HT_2A/2B inverse agonist [20]. Mutagenesis studies indicate that the fully conserved aspartate residue D3.32 of aminergic neurotransmitter GPCRs interacts with the positively charged amine moiety of endogenous agonists and other ligands at physiological pH; this ionic interaction is required for ligand binding [21–23]. In 5-HT₂ receptors, there is a conserved serine residue S3.36, approximately one helical turn from D3.32. Mutagenesis experiments and molecular modeling studies (based on the bovine rhodopsin structure) of the 5-HT_2A receptor [24, 25] suggest that the 5-HT primary amine moiety forms a hydrogen bond to S3.36 and D3.32. Mutation of the 5-HT_2A S3.36 residue to alanine (S3.36A) decreases affinity of primary amine more than secondary amine ligands, whereas, affinity of tertiary amines is barely affected. For the 5-HT_2C receptor, the effects of the D3.32A, S3.36A, and Y7.43A single point mutations on ligand binding have been reported recently (Canal et al., submitted for publication; Refs. [26–28]). Such experimental ligand binding and mutagenesis studies have been used to validate the accuracy of GPCR homology models, including, for the 5-HT_2C receptor. In this study, we describe the building of a 5-HT_2C model based on homology to the human β₂AR and analyze its structure using molecular dynamics (MD) simulations and ligand docking techniques. Results are presented that compare ligand docking outcomes to experimental binding results using several tertiary, secondary, and primary amine ligands at wild type (WT) and D3.32A, S3.36A, and Y7.43A point-mutated 5-HT_2C receptors.

Methods

MODEL BUILDING

A homology model of the human serotonin 5-HT_2C receptor was built based on the crystal structure of the β₂AR/T4-lysozyme chimera (Protein Data bank entry 2rh1) [14]. The 5-HT_2C native sequence was aligned to the β₂AR sequence using ClustalW multiple sequence alignment [29, 30]. Point mutations were performed as needed and the gaps were analyzed, followed by the appropriate sequence additions and deletions to match the 5-HT_2C receptor amino acid sequence. The TM domains were built using the Biopolymer module of Sybyl 8.1 (Tripos International, St. Louis, MO). The inverse agonist carazolol, present in the crystal structure of the human hβ₂AR (2rh1), was deleted and the resulting 7-TM bundle was optimized using Tripos force field [31]. Regions outside the 7-TM bundle such as the 5-HT_2C N-terminus, C-terminus, extracellular and intracellular loop residues were built using loop database PRODAT in Sybyl 8.1. The disulfide bond formed between the cysteine amino acids C3.25 (at the end of TM domain 3) and C6.31 (in extracellular loop 3) in the β₂AR structure was conserved in the 5-HT_2C model. The crude model was minimized using the Powell method implemented in Sybyl with Tripos force field [31] and AMBER charges [32]. The resulting model was inserted into a rectangular box containing a pre-equilibrated 1-palmitoyl-2-oleyl-sn-glycero phosphatidyl choline (POPC) bilayer [33] that consisted of 200 lipid molecules, forming a rectangular patch, and, 5,483 water molecules covering the head groups on each side of the bilayer. The system containing the WT 5-HT_2C receptor model within the simulated membrane contained 37,775 atoms. The system was relaxed using the Tripos force field to a gradient 0.05 kcal/Å mol, prior to MD simulation in the POPC membrane. MD simulation conditions were time run 5000 ps, time step 1 fs, with snapshots collected every 5 fs. Other parameters were the constant number of particles (N), volume (V) and temperature (T) NVT canonical ensemble, 300 K temperature, Boltzmann initial velocities, and nonbonded cutoff set at 8 Å. Constraints for alpha carbons in the TM domains were used. Subsequently, the constraints were removed for a 1000 ps MD simulation run. The final model was obtained from the median structure after clustering analysis of the frames from the last 10 ps of the MD simulation and optimized using the Tripos force field to a convergence of 0.05 kcal/ Å mol. The 3D molecular models of the D3.32A, S3.36A, and Y7.43A 5-HT_2C mutated receptors were obtained by point mutation of the WT 5-HT_2C receptor followed by relaxation and MD procedures as described above.

LIGAND DOCKING

Ligands used in this study were the tertiary amines lisuride, mesulergine, (1R, 3S)-(−)-trans-1-phenyl-3-N,N-dimethylamino-1,2,3,4-tetrahydronaphthalene (N(CH₃)₂-PAT) [20] its secondary amine analog NHCH₃-PAT, and its primary amine analog NH₂-PAT, as well as, the primary amine endogenous agonist 5-HT (Fig. 1). The ligand structures were built as monocations (protonated amines) using HyperChem 8.0 (Hyper-Chem (TM) Professional 8.0, Hypercube, Gainesville, FL) and structures were optimized using PM3 model Hamiltonian to a gradient of 0.01 kcal/Å mol.

FIGURE 1.

Ligands used in 5-HT_2C docking studies.

Ligands were prepositioned in the binding pocket by performing rigid docking with the PatchDock server [34]. The low-energy–high-score solutions were analyzed to select the initial configuration, ensuring the essential interaction between the carboxylate oxygen of receptor residue D3.32 and the ligand protonated amine moiety [21, 22]. The initial ligand–receptor complex configuration was used for flexible ligand docking with Flexidock in Sybyl 8.1. Flexidock uses an algorithm to probe the conformational space defining possible interactions between the ligand and its putative binding site. The binding site was defined by assigning residue D3.32 as a definitive binding site interaction point, and including residues within a 7 Å radius. Structure preparation was carried out prior to docking studies assigning AMBER [32] charges for the protein and Gesteiger-Marsili [35] charges for the ligand. Rotatable bonds in the ligand and the side chains of residues defining the receptor putative active site were screened for optimal positioning of the ligand and side chains in the conformational space; remaining residues were frozen during docking. Default FlexiDock parameters were set at 80,000 generation. The best docking solution, according to the highest FlexiDock score, was minimized using the Tripos force field to a gradient 0.05 kcal/Å mol, prior to MD simulation. The selected high-score pose of the docked ligand was subjected to a MD simulation run for 500 ps, with other parameters the same as above. The final structure of the ligand docked into the receptor was obtained from the average of last 10 ps of the dynamics simulation.

SATURATION AND COMPETITION BINDING STUDIES

Saturation and competition binding studies were carried out as previously described (Canal et al., submitted for publication). The WT, D3.32A, S3.36A, and Y7.43A 5-HT_2C receptors were radiolabeled with [³H]-mesulergine (Perkin-Elmer), having specific activity of 92 Ci/mmol. Test ligands were: 5-HT hydrochloride (Alfa Aesar), lisuride hydrogen maleate (gift from Dr. Elaine Sanders-Bush (Vanderbilt University, Nashville, TN). Synthesis of N(CH₃)₂-, NH(CH₃)-, and NH₂-PAT was performed according to previous laboratory methods [20, 36] to obtain the hydrochloride salt of the trans-stereoisomers [ligand docking studies considered only the active (1R, 3S) enantiomer]; ligand structures are given in Figure 1. The cDNA encoding the human unedited WT 5-HT_2C receptor was obtained from University of Missouri-Rolla (UMR) cDNA Resource Center (Rolla, MO). The D3.32A, S3.36A, and Y7.43A point-mutated 5-HT_2C receptors were generated by polymerase chain reaction (PCR) using standard procedures (Canal et al., submitted for publication). For saturation binding, [³H]-mesulergine was used to obtain K_D and B_max values. Competitive displacement experiments were used to determine IC₅₀ values. Ligand affinity is expressed as an approximation of K_i values by conversion of the IC₅₀ data using the equation K_i = IC₅₀/1 + L/K_D where L is the concentration of radioligand having affinity K_D [37]. Comparisons of K_i and K_D values between WT, S3.36A, and Y7.43A 5-HT_2C receptors were performed using one-way analysis of variance (ANOVA)s with Tukey’s multiple comparison post hoc tests.

Results and Discussion

HOMOLOGY MODELING

The sequence of the WT 5-HT_2C receptor was aligned to the sequence of the template structure β₂A receptor using ClustalW. Sequence alignment showing the TM domains is shown in Table I. Conserved residues are indicated in bold, and reference residues are labeled according to the standard Ballesteros nomenclature for GPCRs [23]. This nomenclature allows for comparison of GPCRs having different residue sequences. The most conserved residue in a given TM domain is labeled Xη.50, where X is the one-letter amino acid code and η is the TM helix (TMH) number; the other amino acids in each TMH are numbered according to this reference residue. Despite the low overall sequence identity between the β₂A and 5-HT_2C receptors (~30%), the majority of highly conserved residues among the 5-HT₂ and β₂A GPCRs are in the TM helices, thus allowing for accurate alignment.

TABLE I.

Alignment of 5-HT_2C and β₂ADR receptor sequences using ClustalW.

Conserved residues are indicated in bold. Reference residues are labeled according to Ballesteros nomenclature [23].

Star(*) indicated the boundaries of TM domains.

Conserved residue sequences among 5-HT₂ receptors and the β₂A receptor in the TM domains were used to verify the alignment, and included: the GNXLVI motif in TM I, including the reference residue N1.50, the TNYF, SLAXAD motifs in TM II, including the reference residue D2.50, the DVL motif in TM III, including the essential residue D3.32, and the TASI and DRY motifs, the KA motif and W4.50 in TM IV, the FXXPLXIM motif in TM V, including P5.50, the WXPFFIXNI motif in TM VI, including residues W6.48, P6.50, and the WIGY and NPLXY motifs in TM VII, including residue P7.50.

The alignment in Table I was used to generate the 3D model of the WT 5-HT_2C receptor. The initial WT 5-HT_2C model, generated as described in the Model Building section, was optimized in vacuum and subsequently equilibrated using MD simulation in the POPC membrane. The resulting TM bundle structure of the WT 5-HT_2C receptor is shown in Figure 2, generated with PyMOL 1.3 [38]. TMH are spectrum color-coded, from blue for TM I, to red for TM VII. In Figure 2, panel A shows the TM bundle oriented with the extracellular domain on top and the intracellular domain at the bottom. Panel B is the view from the extracellular domain down into the receptor cavity. The overall structure of the D3.32A, S3.36A, and Y7.43A point-mutated 5-HT_2C receptors (not shown) was similar to the WT 5-HT_2C receptor.

FIGURE 2.

Three-dimensional 5-HT_2C model structure: Panel A: TM domains with the extracellular domain on top and the intracellular domain at the bottom. Panel B: View from the extracellular domain showing the binding site cavity. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

At the end of the process, the 5-HT_2C model was analyzed using PDBsum in PROCHECK 3.6.2 [39, 40]. Figure 3 shows the Ramachandran plot for the 5-HT_2C model; PROCHECK statistics are reported in Table II. Analysis of the Ramachandran plot results show that 96.9% of residues are in the allowed regions, 89.1% are in the most favored regions (Psi angle values vary from −180° to 0°; Phi angle values vary from 0° to 180°; bottom-left quadrant, Fig. 3), and 7.8% are in additional allowed regions; supplementary 2.7% are in generously allowed regions, confirming the structure to a reasonable model.

FIGURE 3.

Ramachandran Plot of the 5-HT_2C homology model generated with PROCHECK. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

TABLE II.

Procheck statistics of TM residues in the 5-HT_2C homology model.

	Number of residues	% tage
Most favored regions	228	89.1
Additional allowed regions	20	7.8
Generously allowed regions	7	2.7

A description of the residues contained in each TM domain is given in Table III. The TM spanning bundle is composed of seven helices (TMH I–TMH VII) of variable length (23–34 residues), with the longest being TMH III that orients diagonally with respect to the whole TM bundle (Fig. 2). TMH VII is the shortest (23 residues) and continues to helix VIII (12 residues) that orients nearly perpendicular to the TM spanning bundle.

TABLE III.

5-HT_2C transmembrane domain residues.

Domain	Residues in TM	Reference residue
TMH I	P49-M80, 31 residues	N71 (N1.50)
TMH II	A87-Y116, 29 residues	D99 (D2.50)
TMH III	P123-R157, 34 residues	R152 (R3.50)
TMH IV	R168-V191, 23 residues	W179 (W4.50)
TMH V	P212-Q244, 32 residues	P226 (P5.50)
TMH VI	N305-L336, 31 residues	P326 (P6.50)
TMH VII	E347-L370, 23 residues	P365 (P7.50)
TMH VIII	K373-C385, 12 residues

The overall structure of the 7-TM bundle is similar to the previously reported rhodopsin-based 5-HT_2C homology model [20]; however, some differences are found mainly arising from the template receptor structure used. For example, there was a more pronounced kink in TM V near P5.50 in the Rho-based model compared to the β₂AR-based 5-HT_2C model, a stronger helix kink in TM VI near P6.50 in the β₂AR-based 5-HT_2C model compared to the Rho-based model, and a small helix in the extracellular domain loop connecting TM IV and TM V for the βAR-based 5-HT_2C model that is not present in the Rho-based model.

LIGAND BINDING

No specific binding of [³H]-mesulergine to the D3.32A point-mutated 5-HT_2C receptor was detected (data not shown), consistent with results previously reported (Canal et al., submitted for publication). Values for ligand binding (K_i) at WT, S3.36A, and Y7.43A 5-HT_2C receptors are shown in Table IV. The S3.36A mutation caused significant reduction in affinity (higher K_i values) for the primary amine ligands 5-HT and NH₂PAT, and, modestly decreased affinity of the secondary and tertiary amines NHCH₃-PAT and N(CH₃)₂-PAT, but, did not affect affinity of the tertiary amine lisuride. The Y7.43A mutation caused a large reduction in affinity for 5-HT and a moderate reduction for the secondary and tertiary amines, NHCH₃-PAT and N(CH₃)₂-PAT. In contrast, no effect was observed for the primary amine NH₂-PAT, and a moderate increase in affinity was found for the tertiary amine lisuride (Canal et al., submitted for publication). These experimental mutagenesis results provide important information regarding ligand–receptor molecular interactions. Thus, ligand affinity is impacted when, for example, hydrogen and/or electrostatic binding interactions are lost due to residue mutation to alanine. The experimental results were used to validate molecular interactions of the ligands in the 5-HT_2C binding pocket proposed from molecular modeling and ligand docking experiments.

TABLE IV.

K_i values of ligands tested at WT, S3.36A, and Y7.43A 5-HT_2C receptors.

Ligand	WT	Ki (nM)		Fold change relative to WT
		S3.36A	Y7.43A	S3.36A	Y7.43A
N(CH3)2-PAT	80	183	238	2.3	3.0
NHCH3-PAT	183	564	505	3.1	2.8
NH2-PAT	1,800	8,992	1,736	5.0	1.0
5-HT	7	89	112	12.4	15.6
Lisuride	10	9	6	0.9	0.6

LIGAND DOCKING

Residues observed in the putative binding pocket of the 5-HT_2C receptor model [Fig. 4(A)] include, W3.28, I3.29, L3.31, D3.32, V3.33, S3.36, T3.37, S5.43, F5.47, M6.47, W6.48, C6.49, F6.51, F6.52, N6.55, N7.36, V7.39, W7.40, G7.42, and Y7.43. With no ligand bound, the model showed that the para-hydroxy group of the Y7.43 residue was in close proximity (3.35 Å) to the carboxylate group of D3.32 [Fig. 4(A)]; this close interaction (3.38 Å) was also observed in the template β₂AR structure bound to the inverse agonist carazolol [12–14]. This result suggests that Y7.43 plays a role in stabilizing the negative charge of the D3.32 carboxylate; however, this result may vary depending on the ligand structure and conformation/activation state of the receptor. On binding of 5-HT [Fig. 4(B)], the distance between the Y7.43 para-hydroxy group and D3.32 carboxylate was shortened to 2.50 Å. The 5-HT protonated amine group formed a hydrogen bond to the D3.32 carboxylate group (1.85 Å) and also to S3.36 (2.20 Å), as well as, to Y7.43 over a longer distance (3.12 Å). The 5-hydroxy group of 5-HT was near N6.55 (4.28 Å) and D5.35 (5.31 Å), likely forming electrostatic interactions, but too far for hydrogen bonding. An alternative 5-HT pose (not shown) was found in which the protonated amino moiety of 5-HT is able to form hydrogen bond interactions with the side chain of D3.32 (1.85 Å), S3.36 (2.08 Å), and Y7.43 (3.10 Å), while the 5-hydroxy group was in close proximity to Y7.43 para-hydroxy group (2.46 Å). In the S3.36A and Y.743A 5-HT_2C receptor models (not shown), hydrogen bond interactions between 5-HT and residues S3.36 and Y7.43 were not present, consistent with the lower affinity of 5-HT at the S3.36A and Y7.43A point-mutated 5-HT_2C receptors compared to the WT receptor (Table IV).

FIGURE 4.

(A) Unbound WT 5-HT_2C receptor and (B) serotonin (5-HT) docked to WT 5-HT_2C receptor. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

The effect of the 5-HT_2C S3.36A point mutation on binding affinity of primary, secondary, and tertiary amine ligands studied here was similar to that observed for the 5-HT_2A S3.36A receptor [24, 25]. Protonated primary amines (5-HT, NH₂-PAT) having three hydrogen atoms available for interactions with D3.32 and S3.36 were more sensitive to mutation of these residues as compared to secondary (NHCH₃-PAT) and tertiary amine [N(CH₃)₂-PAT, lisuride] ligands (Table IV).

For the Y7.43A mutation, factors other than or in addition to hydrogen bonding apparently come into play, depending on the ligand. For the primary amine 5-HT, the Y7.43A mutation produced an even greater loss of affinity than the S3.36A mutation. For the PAT analogs, however, the tertiary and secondary amines were more affected than the primary amine by the Y7.43A mutation. As results for the S3.36A mutation demonstrate, if hydrogen bonding was the primary interaction between Y7.43 and the PAT analogs, the Y7.43A mutation should negatively impact affinity of the primary amine more than the secondary and tertiary amines. Thus, the experimental binding results for the Y7.43A point-mutated receptor suggest the possibility of other interactions. For example, in the case of N(CH₃)₂-PAT, the ligand docked at 5-HT_2C receptor such that aromatic interactions with the Y7.43 side chain were possible. In contrast, the primary amine NH₂-PAT docked in a position that did favor π–π interactions with the Y7.43 side chain. Given that the affinity of N(CH₃)₂-PAT is greater than that of NHCH₃-PAT and NH₂-PAT at the WT 5-HT_2C receptor (Table IV), it is likely that the PAT N-methyl groups not only increase the basicity of the amine nitrogen but also form van der Waals and perhaps other interactions that lead to a tight fit in the binding pocket. Unfortunately, MD simulations using the Y7.43F point-mutated 5-HT_2C receptor indicated loss of electrostatic interactions between F7.43 and the D3.32 side chain, which allowed for greatly increased rotation of the F7.43 side chain that confounded interpretation.

Figures 5, 6, and 7 show N(CH₃)₂-PAT, NHCH₃-PAT, and NH₂-PAT, respectively, docked at the WT 5-HT_2C binding pocket. For N(CH₃)₂-PAT, NHCH₃-PAT, and NH₂-PAT (respectively): the protonated amino groups form hydrogen bonds with D3.32 (1.84, 1.97, 1.93 Å) with S3.36 (3.27, 2.84, 2.62 Å) and Y7.43 (2.96, 2.78, 3.70 Å), albeit, over a longer distance. Docking poses show similar configurations of N(CH₃)₂-PAT and NHCH₃-PAT, where the aromatic ring in the PAT bicycle structure is close to the aromatic ring of Y7.43 side chain. Moreover, for N(CH₃)₂-PAT, the bicycle aromatic ring is parallel to the Y7.43 aromatic ring, suggesting π–π stacking interactions. However, for NH₂-PAT, the bicycle aromatic ring is not positioned for π–π stacking interactions with Y7.43. Loss of π–π stacking interactions for N(CH₃)₂-PAT and NHCH₃-PAT apparently accounts for the reduced affinity of these analogs at the Y7.43A point-mutated receptor, whereas, no change in affinity was observed for NH₂-PAT (Table IV).

FIGURE 5.

N(CH₃)₂-PAT docked in WT 5-HT_2C receptor. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

FIGURE 6.

NHCH₃-PAT docked in WT 5-HT_2C receptor. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

FIGURE 7.

NH₂-PAT docked in WT 5-HT_2C receptor. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 8 shows lisuride docked at the 5-HT_2C binding pocket. The lisuride configuration in the binding pocket allows for formation of a hydrogen bond to D3.32 (1.56 Å), but not with S3.36 (7.21 Å), or to Y7.43 (3.95 Å). Consistent with these modeling results, affinity of lisuride at the S3.36A point-mutated receptor was not different than the WT receptor (Table IV). Also in the molecular model, a hydrogen at the amide side chain of N6.55 is in close proximity to the carbonyl oxygen of lisuride (1.72 Å), likely, providing an additional stabilizing interaction that may explain the moderate increase in affinity of lisuride at the point-mutated compared to WT 5-HT_2C receptor (Table IV) (see supporting information).

FIGURE 8.

Lisuride docked in WT 5-HT_2C receptor. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Conclusions

The 5-HT_2C GPCR homology model built from the human β₂AR crystal structure was tested using computational techniques and ligand docking results were validated by comparison to results obtained from point-mutated receptor binding experiments. The model structure was stable after unrestrained MD simulations in lipid bilayer membrane. PROCHECK statistics show that the majority of residue side chains are in allowed regions. The overall structure of the 7-TM bundle was similar to the previously reported rhodopsin-based 5-HT_2C homology model [20]; however, some differences were apparent due to the different template structure used. The human 5-HT_2C homology model herein explains the experimental binding results of the non-selective 5-HT₂ agonist 5-HT and the selective 5-HT_2C agonist N(CH₃)₂-PAT [20]. The carboxylate of 5-HT_2C residue D3.32 was found to be in close proximity to the para-hydroxy phenyl group of Y7.43, which suggests that a hydrogen bond between these amino acids is involved in binding pocket stabilization. Additionally, it appears that ligand binding modulates the distance between the 5-HT_2C D3.32 and Y7.43 side chains and that this distance varies, depending on the ligand chemical structure.

While results here document the usefulness of combining mutagenesis and molecular modeling/ computational chemistry approaches to predict GPCR ligand affinity, extending such results to predict function is fraught with complexity. A critical assumption of current GPCR signaling theory is that there exists a heterogeneous population of active and inactive receptor conformations. However, the molecular determinants governing ligand binding that lead to stabilization of agonist versus inverse agonist GPCR conformation states are poorly understood. Moreover, it is now well-established that the same ligand may stabilize both agonist and inverse agonist conformations of the same GPCR, and, that the same GPCR may activate multiple signaling pathways [41–43]. Accordingly, current GPCR drug discovery efforts may reliably be guided by computationally predicted ligand affinity; however, functional outcome(s) largely must still be experimentally determined.