et drugs with dual effect as a DI receptor agonist and a D2 MODELING AND SIMULATION METHODS receptor antagonist would provide a new treatment for psy- strategy of modeling and simulation chotic diseases(4) An active compound, namely (-H-stepholidine(SPD), Experiments demonstrated that the sequences of aminergic receptors, viz. isolated from the Chinese herb Stephania, is to date the only dopamine, a-adrer drug with a dual effect as a DI receptor agonist and a D2 conservative within the TM domains, which dictate the common ligand- receptor antagonist(4). SPD has high affinity for DI- and suggest that different binding interactions of ligands to one receptor lead to D2-like receptors but low affinity for 5-HT2 receptors and different receptor conformations (10). The small cha a2-adrenoceptors(4). The dual action of SPD has been dem affect its interactions with the receptor and, hence, the receptor activa onstrated in both cortical and subcortical structures, includ tion(10). Thus, we faced two difficulties: 1), how to construct 3D models of the DI and D2 receptors, and 2), how to identify the binding conforma ing the mPFC. NAc, VTA, and basal ganglia dopamine tions of the two receptors with SPD. To solve these problems, we integrated systems(4). Moreover, clinic studies showed that SPD is several modeling and simulation methods in this study. The computational superior to perphenazine in antipsychotic efficacy. Unlike pipeline is outlined in Fig. SI in the Supplementary Material. Briefly.the perphenazine, SPd does not induce any EPSs(4), making computational flow is as follows SPD an attractive compound in studying the dual action 1. The 3D models of the DI and D2 receptors were constructed using a mechanism of a chemical for developing novel antipsychotic homology-modeling approach based on the x-ray crystal structure of drugs. However, the molecular basis of the dual action of bovine rhode SPD is obscure to date. Thus, exploring the dual action 2. Two Io-ns MD simulations were carried out respectively on the con- mechanism of SPD against DI and D2 receptors at the (palmitoyloleoylphosphatidylcholine) bilayer. atomic level is the first step toward developing more superior 3. To find the probable SPD binding conformations of DI and D2 recep- antipsychotic agents. Obviously, the 3D structures of DI and tors, SPD was docked into numerous minimum-energy conformations D2 receptors are essential to figuring out the dual action isolated from the MD trajectories of the two receptors by using the me mechanism. Unfortunately, these 3d structures are not cur- lecular docking approach, includin rently available energy between SPD and each selected conformation of the receptors. It has been widely recognized that molecular modeling Conformations of the two receptors with lowest binding free energies to SPD were selected as the initial structures for further simulations and simulation are an excellent complement to experiments 4. Two additional 10-ns MD simulations were conducted on the initial in explaining experimental results and providing clues for structures of SPD-DI and SPD-D2 complexes, resulting from the above further experiments. Furthermore, it may reveal information docking calculations, embedded in a hydrated POPC bilayer. that is not accessible by experiments(12). Recently, grea 5. The ligand-receptor interaction and the dual action mechanism of SPD were explored by analyzing all of the modeling and simulation results. success has been achieved in the field of structure predic tion of GPCRs(13-15). For instance, some algorithms ar apable of predicting the transmembrane(TM)structures (15-20). However, constructing appropriate conformational Homology modeling for the 3D models of D1 and states of the extracellular or intracellular domains of GPCRs D2 receptors is still a tough job. Molecular dynamics(MD) simulations Bovine rhodopsin, a member of the GPCR superfamily, was structur (21), taking advantage of iteratively tracking the trajectory determined at a resolution of 2.80 A(Protein Data Bank entry 1F88)(I f conformational change could be used to simulate the This x-ray structure of GPCRs provides a solid template for modeling the 3D binding process, to explore the binding conformation, and to structures of other GPCRs () To obtain a reliable map the binding mechanism at molecular and atomic levels pereceptorsdownloadedfromhttp://www. gpcr. org//tm were used. Residues were numbered according to the gener- To the best of our knowledge, no long timescale Md alized numbering scheme proposed by Ballesteros and Weinstein(22).To simulation has been carried out so far on drs, all this facilitate the comparison among the aligned residues in various GPCRs, the motivates us to carry out a computational study on DI and x 50 and residues within a given tM are then indexed relative to the 50 SPD on these two receptors. Thus, homology modeling, The MODELLER program encoded in Insightll (23)was employed to automated molecular docking, and MD simulations have assemble the 3D models of the a-helix bundles of DI and D2 receptors by been integrated in this study. A homology modeling using the x-ray crystal structure of bovine rhodopsin as a template. The approach was used to construct 3D structures of DI and FASTA program (24)was used to identify sequence homology through an D2 receptors. Automated docking was used to find possible in-house database(15)containing 700 loops and proteins with medium to high sequence identity. Clustalw (25)was then used to determine the binding sites for SPD against the receptors. The MD sim- fragments that have higher homology with the ulations, performed in a fully hydrated lipid bilayer envi- A reasonable fragment conformation was chosen from the top 10 candidates ronment, were carried out to address the following questions at have the lowest root mean-square(RMs) values and considerable How does SPD bind to the DI and D2 receptors? How does geometrical compatibility. The conserved disultide bond between residues SPD regulate the signal transe n of the DI receptor? Cys-3. 25 at the beginning of TM Ill and Cys_EL-2 in the middle of extracellular loop 2(EL-2)was also created. Energy minimizations of the What is the fund tal basis lying the dual action of models were carried out using the Discover module encoded in Insightll SPD? with the same parameters as that of our previous studies(12). Biophysical Journal 93(5)1431-1441drugs with dual effect as a D1 receptor agonist and a D2 receptor antagonist would provide a new treatment for psy￾chotic diseases (4). An active compound, namely ()–stepholidine (SPD), isolated from the Chinese herb Stephania, is to date the only drug with a dual effect as a D1 receptor agonist and a D2 receptor antagonist (4). SPD has high affinity for D1- and D2-like receptors but low affinity for 5-HT2 receptors and a2-adrenoceptors (4). The dual action of SPD has been dem￾onstrated in both cortical and subcortical structures, includ￾ing the mPFC, NAc, VTA, and basal ganglia dopamine systems (4). Moreover, clinic studies showed that SPD is superior to perphenazine in antipsychotic efficacy. Unlike perphenazine, SPD does not induce any EPSs (4), making SPD an attractive compound in studying the dual action mechanism of a chemical for developing novel antipsychotic drugs. However, the molecular basis of the dual action of SPD is obscure to date. Thus, exploring the dual action mechanism of SPD against D1 and D2 receptors at the atomic level is the first step toward developing more superior antipsychotic agents. Obviously, the 3D structures of D1 and D2 receptors are essential to figuring out the dual action mechanism. Unfortunately, these 3D structures are not cur￾rently available. It has been widely recognized that molecular modeling and simulation are an excellent complement to experiments in explaining experimental results and providing clues for further experiments. Furthermore, it may reveal information that is not accessible by experiments (12). Recently, great success has been achieved in the field of structure predic￾tion of GPCRs (13–15). For instance, some algorithms are capable of predicting the transmembrane (TM) structures (15–20). However, constructing appropriate conformational states of the extracellular or intracellular domains of GPCRs is still a tough job. Molecular dynamics (MD) simulations (21), taking advantage of iteratively tracking the trajectory of conformational change, could be used to simulate the binding process, to explore the binding conformation, and to map the binding mechanism at molecular and atomic levels. To the best of our knowledge, no long timescale MD simulation has been carried out so far on DRs. All this motivates us to carry out a computational study on D1 and D2 receptors to explore the mechanism of the dual action of SPD on these two receptors. Thus, homology modeling, automated molecular docking, and MD simulations have been integrated in this study. A homology modeling approach was used to construct 3D structures of D1 and D2 receptors. Automated docking was used to find possible binding sites for SPD against the receptors. The MD sim￾ulations, performed in a fully hydrated lipid bilayer envi￾ronment, were carried out to address the following questions: How does SPD bind to the D1 and D2 receptors? How does SPD regulate the signal transduction of the D1 receptor? What is the fundamental basis underlying the dual action of SPD? MODELING AND SIMULATION METHODS Strategy of modeling and simulation Experiments demonstrated that the sequences of aminergic receptors, viz. dopamine, a-adrenergic, b-adrenergic, and serotonin receptors, are highly conservative within the TM domains, which dictate the common ligand￾binding sites of these receptors (22). However, experimental data also suggest that different binding interactions of ligands to one receptor lead to different receptor conformations (10). The small changes in ligand structure may affect its interactions with the receptor and, hence, the receptor activa￾tion (10). Thus, we faced two difficulties: 1), how to construct 3D models of the D1 and D2 receptors, and 2), how to identify the binding conforma￾tions of the two receptors with SPD. To solve these problems, we integrated several modeling and simulation methods in this study. The computational pipeline is outlined in Fig. S1 in the Supplementary Material. Briefly, the computational flow is as follows: 1. The 3D models of the D1 and D2 receptors were constructed using a homology-modeling approach based on the x-ray crystal structure of bovine rhodopsin. 2. Two 10-ns MD simulations were carried out respectively on the con￾structed D1 and D2 receptor models, embedded in a hydrated POPC (palmitoyloleoylphosphatidylcholine) bilayer. 3. To find the probable SPD binding conformations of D1 and D2 recep￾tors, SPD was docked into numerous minimum-energy conformations isolated from the MD trajectories of the two receptors by using the mo￾lecular docking approach, including prediction of the binding free energy between SPD and each selected conformation of the receptors. Conformations of the two receptors with lowest binding free energies to SPD were selected as the initial structures for further simulations. 4. Two additional 10-ns MD simulations were conducted on the initial structures of SPD-D1 and SPD-D2 complexes, resulting from the above docking calculations, embedded in a hydrated POPC bilayer. 5. The ligand-receptor interaction and the dual action mechanism of SPD were explored by analyzing all of the modeling and simulation results. Homology modeling for the 3D models of D1 and D2 receptors Bovine rhodopsin, a member of the GPCR superfamily, was structurally determined at a resolution of 2.80 A˚ (Protein Data Bank entry 1F88) (17). This x-ray structure of GPCRs provides a solid template for modeling the 3D structures of other GPCRs (7). To obtain a reliable sequence alignment, 64 sequences of dopamine-type receptors downloaded from http://www. gpcr.org/7tm were used. Residues were numbered according to the gener￾alized numbering scheme proposed by Ballesteros and Weinstein (22). To facilitate the comparison among the aligned residues in various GPCRs, the most conserved residue in transmembrane X is given the index number X.50, and residues within a given TM are then indexed relative to the ‘‘50’’ position. The MODELLER program encoded in InsightII (23) was employed to assemble the 3D models of the a-helix bundles of D1 and D2 receptors by using the x-ray crystal structure of bovine rhodopsin as a template. The FASTA program (24) was used to identify sequence homology through an in-house database (15) containing 700 loops and proteins with medium to high sequence identity. ClustalW (25) was then used to determine the fragments that have higher homology with the loops of D1 and D2 receptors. A reasonable fragment conformation was chosen from the top 10 candidates that have the lowest root mean-square (RMS) values and considerable geometrical compatibility. The conserved disulfide bond between residues Cys-3.25 at the beginning of TM III and Cys_EL-2 in the middle of extracellular loop 2 (EL-2) was also created. Energy minimizations of the models were carried out using the Discover module encoded in InsightII with the same parameters as that of our previous studies (12). 1432 Fu et al. Biophysical Journal 93(5) 1431–1441