复旦大学：《药物设计学》课程教学资源（教学研究）基于靶点结构的药物设计_Discovery of a Novel 5-HT2A Inhibitor by Pharmacophore-based Virtual Screening

CHEM. RES CHINESE UNIVERSITIES 2011, 27(4), 655-660 Discovery of a Novel 5-HT2A Inhibitor by Pharmacophore-based Virtual Screenin ng XIONG Zi-jun, DU Peng, LI Bian, XU Li-li, ZHEN Xue-chu2"and FU Wei' 1. Key Laboratory of Smart Drug Delivery, Ministry of Education PLA, Department of Medicinal Chemistry School of pharmacy, Fudan University Shanghai 201203, P R China: 2 State Key laboratory for Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai 201203, P R China: 3. Department of Pharmacology, College of Pharmacy, Soochow University, Suchou 215123, P R China Abstract The serotonin 2A(5-HT2A) receptor has been implicated in several neurological conditions and potent HT2A antagonists have therapeutic effects in the treatment of schizo phrenia and depression. In this study a potent novel 5-HT2A inhibitor 05245768 with a Ki value of (593.89+34.10)nmol/L was discovered by integrating a set of computational approaches and experiments(protein structure prediction, pharmacophore-based virtual screening, au- tomated molecular docking and pharmacological bioassay ) The 5-HT2A receptor showed a negatively charged bin- ding pocket. The binding mode of compound 05245768 with 5-HT2A was obtained by gold docking procedure, which revealed the conserved interaction between protonated nitrogen in compound 05245768 and carboxylate group of D3.32 at the active site of 5-ht2a Keywords Pharmacophore model; Serotonin 2A receptor; Database search, Virtual screening; Molecular docking Article ID1005-9040(2011)-04-655-06 1 Introduction nterested in the discovery of potent 5-HT2A inhibitor, with eventual aim to develop new SGa drugs for the treatment of Schizo phrenia is a severe and chronic mental illness, as- diseases affecting central neuronal system such as schizo phre- sociated with high prevalence, which affects ca. 0.5%--1.5% nia. In particular, our research is focused on the application of of the worldwide population. Symptoms of schizo phrenia are molecular simulation and modeling methods in the rational classified as either positive or negative symptoms. The former design of new blocking agents of 5-HT2A receptor. includes hallucinations delusions and chorea etc . the latter In this study, homology modeling, automated flexible mo- involves apathy, cognitive impairment, tardive dyskinesia and lecular docking, pharmacophore model building,database academic handicap. Among them deficit in both attention and searching and biological test were integrated to investigate the emory is the most prominent features of schizo phrenia. The action mechanism of 5-HT2A, design and discover inhibitors of first generation antipsychotic(FGA)drugs are characterized by 5-HTA. Five types of chemical probes were used to map the high affinity for dopamine(DA)D2 receptors. The application active site by means of grid calculation, Lipinski's Rule of Five of FGa drugs is very limited because of their high incidence of was used to filter out the non-drug like redundancies, and the extrapyramidal side-effects(EPS). The second generation anti- interaction cluster analysis in combination with scoring func psychotic(SGA) drugs display reduced EPS liability and en- tion was used to guide the final selection. Biological tests were hance therapeutic efficacy. Therefore, SGA drugs are mainly employed to filter active compounds. Molecular docking was used in current clinics. Compared to FGA drugs which mainly used to reveal the receptor-ligand interaction and help to design block dopamine D2 subtype receptor, most of SGa drugs shov more potent inhibitors potent activity against 5-HT2A receptor. The involvement of 5-HT2A receptor in the pharmacological profile of SAG drugs 2 Modeling and Simulation Method has been supported by tremendous amounts of biological pharmacological and clinical studiesl2-l. It has been widely 2.1 Strategy of Modeling and simulation accepted that high ratio of pKi between antipsychotic agents Experiments demonstrate that the sequences of aminergic targeting on 5-HT2A and Dz receptors reflects the atypical pro- receptors, viz. dopamine, a-adrenergic, B-adrenergic, and sero- file of SGA drugs. It could decrease the incidence of adverse tonin receptors, are highly conservative within the transmem- EPS, cognitive deficits, hyperprolactemia, and negative symp- brane(TM) domains, which indicates the common ligand- toms. Most of SGa drugs show that the ratio is >1.12. We are binding sites of these receptor 6). However, experimental data *Correspondingauthors.E-mail:weifuuh@gmail.com;xczhen@mail.shcncaccn Received May 11, 2011; accepted June 9, 2011 Supported by the National High Technology Research and Development Program of China(No 2009AA02Z308), the Major State Basic Research Development Program of China(No 2010CB912601) and the National Natural Science Foundation of China No.20702009)

CHEM. RES. CHINESE UNIVERSITIES 2011, 27(4), 655—660 ——————————— *Corresponding authors. E-mail: weifuuh@gmail.com; xczhen@mail.shcnc.ac.cn Received May 11, 2011; accepted June 9, 2011. Supported by the National High Technology Research and Development Program of China(No.2009AA02Z308), the Major State Basic Research Development Program of China(No.2010CB912601) and the National Natural Science Foundation of China (No.20702009). Discovery of a Novel 5-HT2A Inhibitor by Pharmacophore-based Virtual Screening XIONG Zi-jun1 , DU Peng1 , LI Bian1 , XU Li-li1 , ZHEN Xue-chu2,3* and FU Wei1* 1. Key Laboratory of Smart Drug Delivery, Ministry of Education & PLA, Department of Medicinal Chemistry, School of Pharmacy, Fudan University, Shanghai 201203, P. R. China; 2. State Key Laboratory for Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai 201203, P. R. China; 3. Department of Pharmacology, College of Pharmacy, Soochow University, Suzhou 215123, P. R. China Abstract The serotonin 2A(5-HT2A) receptor has been implicated in several neurological conditions and potent 5-HT2A antagonists have therapeutic effects in the treatment of schizo phrenia and depression. In this study, a potent novel 5-HT2A inhibitor 05245768 with a Ki value of (593.89±34.10) nmol/L was discovered by integrating a set of computational approaches and experiments(protein structure prediction, pharmacophore-based virtual screening, au￾tomated molecular docking and pharmacological bioassay). The 5-HT2A receptor showed a negatively charged bin￾ding pocket. The binding mode of compound 05245768 with 5-HT2A was obtained by GOLD docking procedure, which revealed the conserved interaction between protonated nitrogen in compound 05245768 and carboxylate group of D3.32 at the active site of 5-HT2A. Keywords Pharmacophore model; Serotonin 2A receptor; Database search; Virtual screening; Molecular docking Article ID 1005-9040(2011)-04-655-06 1 Introduction Schizo phrenia is a severe and chronic mental illness, as￾sociated with high prevalence, which affects ca. 0.5%―1.5% of the worldwide population[1]. Symptoms of schizo phrenia are classified as either positive or negative symptoms. The former includes hallucinations, delusions and chorea etc.; the latter involves apathy, cognitive impairment, tardive dyskinesia and academic handicap. Among them deficit in both attention and memory is the most prominent features of schizo phrenia. The first generation antipsychotic(FGA) drugs are characterized by high affinity for dopamine(DA) D2 receptors. The application of FGA drugs is very limited because of their high incidence of extrapyramidal side-effects(EPS). The second generation anti￾psychotic(SGA) drugs display reduced EPS liability and en￾hance therapeutic efficacy. Therefore, SGA drugs are mainly used in current clinics. Compared to FGA drugs which mainly block dopamine D2 subtype receptor, most of SGA drugs show potent activity against 5-HT2A receptor. The involvement of 5-HT2A receptor in the pharmacological profile of SAG drugs has been supported by tremendous amounts of biological, pharmacological and clinical studies[2―4]. It has been widely accepted that high ratio of pKi between antipsychotic agents targeting on 5-HT2A and D2 receptors reflects the atypical pro￾file of SGA drugs. It could decrease the incidence of adverse EPS, cognitive deficits, hyperprolactemia, and negative symp￾toms. Most of SGA drugs show that the ratio is >1.12[5]. We are interested in the discovery of potent 5-HT2A inhibitor, with the eventual aim to develop new SGA drugs for the treatment of diseases affecting central neuronal system such as schizo phre￾nia. In particular, our research is focused on the application of molecular simulation and modeling methods in the rational design of new blocking agents of 5-HT2A receptor. In this study, homology modeling, automated flexible mo￾lecular docking, pharmacophore model building, database searching and biological test were integrated to investigate the action mechanism of 5-HT2A , design and discover inhibitors of 5-HT2A. Five types of chemical probes were used to map the active site by means of grid calculation, Lipinski’s Rule of Five was used to filter out the non-drug like redundancies, and the interaction cluster analysis in combination with scoring func￾tion was used to guide the final selection. Biological tests were employed to filter active compounds. Molecular docking was used to reveal the receptor-ligand interaction and help to design more potent inhibitors. 2 Modeling and Simulation Method 2.1 Strategy of Modeling and Simulation Experiments demonstrate that the sequences of aminergic receptors, viz. dopamine, α-adrenergic, β-adrenergic, and sero￾tonin receptors, are highly conservative within the transmem￾brane(TM) domains, which indicates the common ligand￾binding sites of these receptors[6]. However, experimental data

CHEM RES CHINESE UNIVERSITIES VoL 27 also suggest that different binding interactions of ligands to one 5.0.1 71 d to different receptor conformations. Small changes in ligand structure might affect its interactions with 2.3 Active Site Mapping and Pharmacophore receptor/I. Thus, we faced two difficulties: how to construct Model Building 3D models of 5-HT2A receptors and how to design an effective strategy to find active inhibitors of 5-HT2A receptor. To solve The binding pocket of 5-HT2A was determined by CASTp ogram!s, which was consistent with the conservative active methods and biological assay for virtual screening in this study. sites of GPCR and mutation experiments of 5-HT2A recepto Briefly, the workflow was as follows: ()the 3D model of 5-HT2A receptor was constructed via a homology modeling of the protein for energetic interactions with the following approach widely used based on the X-ray crystal structure of probes: negative ionizable(Co0), positive ionizable(NI' B2-adrenergic receptor s),(2)the grid calculation was carried hydrogen-bond acceptor(O), hydrogen-bond donor(NI)and out to map the active site of 5-HT2A receptor and Catalyst module encoded in Discovery Studio 3.0 program! 9I was used the grid calculations, grid points were superimposed to identify to build the pharmacophore model of 5-HT2A receptor; (3) clusters of positions. Members of each identified cluster were search available chemical database to find the computational combined into one pharmacophore feature so that the center of active compounds; (4)dock potentially active compounds into the pharmacophore feature was at the geometric center of the the active site of 5-hT2A and investigate the interaction mode, members in the cluster. The tolerance volumes of the pharma- ) score function was used to rank the hits, and more poten- cophore features were determined by the radius of gyration for the members in the cluster plus the van der Waals radius of the tially active compounds were further selected based on their atom represented by the probe. In the pharmacophore model g energies,(6) the candidates were pur chased for biological test; (7) the active hits were docked into developed in this study, the clusters of the positive ionizable 5-HT2A receptor to elaborate the binding mode probe that were within a distance of 0.3 nm from the most con- servative residue D3. 32 were selected and the clusters of hy- 2.2 Homology Modeling of 5-HT2A Receptor drogen-bond donor(HBD) probe within a hydrogen-bonding distance from the conserved s5.46 were selected. Since the B2-Adrenergic receptor, a member of G protein-coupled active pocket of 5-HT2A receptor was totally negatively resolution of 0.240 nm(PDB entry 2RHI)S. This X-ray struc- ted. At last, a pharmacophore model was buli probe were omit- receptor(GPCR)superfamily, was structurally determined at a charged, all the clusters of negative ionizable probe were omit ture of GPCRs provided a solid template for modeling the accurate 3D structures of other aminergic GPCRs To obtain a 2.4 Database Searching and Compound Selecting reliable sequence alignment, 32 sequences of serotonin type of receptors(downloadedfromhttp://www.gpcr.org/7tm)were Database searching and fitting of the resulting hits to the used. Residues were numbered according to the generalized pharmacophore model were performed on a Linux Pentium numbering scheme proposed by Ballesteros and Weinstein! ol workstation. The final pharmacophore model was imported into To facilitate the comparison among the aligned residues in ed residues in CATALYST module encoded in Discovery Studio 3.0 program for searching gold Collection of Asinex 3D databases of che- various gPCrs. the most conserved residue in transmembrane X(TM-X) was given the index number X50, and the residues mical compounds via the"best flexible"search routine. Lipins- within a given TM were then indexed relatively to the 50 ki's Rule of Five was applied to reducing the number of re position. turned hits. Finally 7 compounds were selected for biological The Modeller 9v2 program 2 was employed to assem- test based on fit values, G-score of docking, interaction with important residues inside active site and structural diversity. ble the 3D model of 5-HT2A receptor by means of the X-ray The obtained pharmacophore model will be tested by the crystal structure of B2-adrenergic receptor as a template. The FASTA programwas used to identify sequence homology through the latest PDB database and Clustalwllal was then used 2.5 Test of Pharmacological Profiles and molecu to determine the fragments that have higher homology with the loops of 5-HT2A receptor. A reasonable fragment conformation lar Modeling and docking was chosen from the top 10 candidates that had the lowest root Spiperone was selected as the positive control; all the mean square(RMS) values and considerable geometrical com- tested compounds and control drugs were dissolved in patibility. The conserved disulfide bond between residues DMsO(dimethyl sulfoxide) to a concentration of 0.01 mol/L Cys325 at the beginning of TM-3 and Cys EL-2 in the middle and further diluted by deionized water to 100 umol/L. Add 10 of the extracellular loop 2(EL-2) was also created. Energy mi- HL of each of the compounds to be tested and the radioactive nimization of the model was carried out via Syby169 soft- ligand respectively with 80 uL of receptor protein to a reaction arelI51. Procheck 3.5.4 program[l6) was employed to check the tube, and the reaction was incubated for 30 min on a water-bath accuracy of the model. To further validate the accuracy of the at 37C. Then, the reaction mixture was immediately placed odel, a 5-HT2A antagonist ketanserin was docked into the onto an ice- bath to terminate the reaction. In the Millipore cell 5-HT2A model via the flexible docking program GOLd sample collector, a rapid filtration was performed throu

656 CHEM. RES. CHINESE UNIVERSITIES Vol.27 also suggest that different binding interactions of ligands to one receptor lead to different receptor conformations. Small changes in ligand structure might affect its interactions with receptor[7]. Thus, we faced two difficulties: how to construct 3D models of 5-HT2A receptors and how to design an effective strategy to find active inhibitors of 5-HT2A receptor. To solve these problems, we integrated several modeling and simulation methods and biological assay for virtual screening in this study. Briefly, the workflow was as follows: (1) the 3D model of 5-HT2A receptor was constructed via a homology modeling approach widely used based on the X-ray crystal structure of β2-adrenergic receptor[8]; (2) the grid calculation was carried out to map the active site of 5-HT2A receptor and Catalyst module encoded in Discovery Studio 3.0 program[9] was used to build the pharmacophore model of 5-HT2A receptor; (3) search available chemical database to find the computational active compounds; (4) dock potentially active compounds into the active site of 5-HT2A and investigate the interaction mode; (5) score function was used to rank the hits, and more poten￾tially active compounds were further selected based on their scores and their binding energies; (6) the candidates were pur￾chased for biological test; (7) the active hits were docked into 5-HT2A receptor to elaborate the binding mode. 2.2 Homology Modeling of 5-HT2A Receptor β2-Adrenergic receptor, a member of G protein-coupled receptor(GPCR) superfamily, was structurally determined at a resolution of 0.240 nm(PDB entry 2RH1)[8]. This X-ray struc￾ture of GPCRs provided a solid template for modeling the accurate 3D structures of other aminergic GPCRs. To obtain a reliable sequence alignment, 32 sequences of serotonin type of receptors(downloaded from http://www.gpcr.org/7tm) were used. Residues were numbered according to the generalized numbering scheme proposed by Ballesteros and Weinstein[10]. To facilitate the comparison among the aligned residues in various GPCRs, the most conserved residue in transmembrane X(TM-X) was given the index number X.50, and the residues within a given TM were then indexed relatively to the “50” position. The Modeller 9v2 program[11,12] was employed to assem￾ble the 3D model of 5-HT2A receptor by means of the X-ray crystal structure of β2-adrenergic receptor as a template. The FASTA program[13] was used to identify sequence homology through the latest PDB database and ClustalW[14] was then used to determine the fragments that have higher homology with the loops of 5-HT2A receptor. A reasonable fragment conformation was chosen from the top 10 candidates that had the lowest root mean square(RMS) values and considerable geometrical com￾patibility. The conserved disulfide bond between residues Cys3.25 at the beginning of TM-3 and Cys_EL-2 in the middle of the extracellular loop 2(EL-2) was also created. Energy mi￾nimization of the model was carried out via Sybyl6.9 soft￾ware[15]. Procheck 3.5.4 program[16] was employed to check the accuracy of the model. To further validate the accuracy of the model, a 5-HT2A antagonist ketanserin was docked into the 5-HT2A model via the flexible docking program GOLD 5.0.1[17]. 2.3 Active Site Mapping and Pharmacophore Model Building The binding pocket of 5-HT2A was determined by CASTp program[18], which was consistent with the conservative active sites of GPCR and mutation experiments of 5-HT2A receptor. Then, the GRID22 program[19] was used to map the active sites of the protein for energetic interactions with the following probes: negative ionizable(COO– ), positive ionizable(N1+ ), hydrogen-bond acceptor(O), hydrogen-bond donor(N1) and hydrophobic probes(DRY). For each of the five probes used in the grid calculations, grid points were superimposed to identify clusters of positions. Members of each identified cluster were combined into one pharmacophore feature so that the center of the pharmacophore feature was at the geometric center of the members in the cluster. The tolerance volumes of the pharma￾cophore features were determined by the radius of gyration for the members in the cluster plus the van der Waal’s radius of the atom represented by the probe. In the pharmacophore model developed in this study, the clusters of the positive ionizable probe that were within a distance of 0.3 nm from the most con￾servative residue D3.32 were selected and the clusters of hy￾drogen-bond donor(HBD) probe within a hydrogen-bonding distance from the conserved S5.46 were selected. Since the active pocket of 5-HT2A receptor was totally negatively￾charged, all the clusters of negative ionizable probe were omit￾ted. At last, a pharmacophore model was built. 2.4 Database Searching and Compound Selecting Database searching and fitting of the resulting hits to the pharmacophore model were performed on a Linux Pentium workstation. The final pharmacophore model was imported into CATALYST module encoded in Discovery Studio 3.0 program for searching gold Collection of Asinex 3D databases of che￾mical compounds via the “best flexible” search routine. Lipins￾ki’s Rule of Five was applied to reducing the number of re￾turned hits. Finally 7 compounds were selected for biological test based on fit values, G-score of docking, interaction with important residues inside active site and structural diversity. The obtained pharmacophore model will be tested by the pharmacological test. 2.5 Test of Pharmacological Profiles and Molecu￾lar Modeling and Docking Spiperone was selected as the positive control; all the tested compounds and control drugs were dissolved in DMSO(dimethyl sulfoxide) to a concentration of 0.01 mol/L and further diluted by deionized water to 100 μmol/L. Add 10 μL of each of the compounds to be tested and the radioactive ligand respectively with 80 μL of receptor protein to a reaction tube, and the reaction was incubated for 30 min on a water-bath at 37 °C. Then, the reaction mixture was immediately placed onto an ice-bath to terminate the reaction. In the Millipore cell sample collector, a rapid filtration was performed through

XIONG Zi-jun et al. GF/B glass fiber filter paper and washing was performed 3 model of 5-HT2A receptor was assembled taking the X-ray times with 3 mL of eluent(50 mmol/L Tris-HCL, pH=7.7), dry- crystal structure of B2-adrenergic receptor as a template. Step- ing was conducted in the microwave oven for 8-9 min, then wise energy minimizations of the homology 5-HT2A model was the filter paper was transferred into a 0.5 mL centrifuge tube, to carried out for 500 steps with Kollman All-Atom force field hich 500 uL of soluble scintillation fluid was then added. and the SYBYL program. The final model was verified by After 30 min of rest in darkness, the radioactivity was mea- professional structure validation program Procheck 3.5.4. The sured. Then pharmacological parameters: percentage inhibition, Procheck statistics shows that 96.1% of the residues of the ICso and Ki were calculated. Concentrations were measured 5-HT2A model were either in the most favored or in the addi every two tubes, two independent tests of each compound were tionally allowed regions of the Ramachandran map(Fig. 2) conducted suggesting that the overall main chain and side chain struc- The inhibitor was docked into the 5-ht2a receptor via the tures were all reasonable flexible docking program GOLD 4.12. Before docking, the (A) final conformation of the ligand was obtained after 1000 steps of energy minimization via the Tripos force field and optimiza tion at the dFT/B3LYP/6-311G level by means of Gaussian 2009/201 Instead of the default 10600 genetic algorithm(GA) runs were performed to fully take into account the flexibility of the ligand. For each GA run, the default Ga settings were used with no early termination allowed and internal ligand energy offset turned on. Finally, GOLD-Score was used as the fitness function 3 Results and discussion Fig2 3D model of 5-HT2A receptor(A)and ramchandran 3.1 3D Structure of 5-HT2A Receptor The different colored areas indicate disallowed( white), generously allowed (light yellow), additional allowed(yellow)and most favored(red)regions Sequence alignment( Fig. 1)indicates that the sequence entity and similarity are 39.0% and 62. 4%, for the TMs be TM-I TM-2 B2-Adrenerg IVLATFCNLVITATAKEI HIOEKUSAL TAWIILTI Ar EL-I TM-3 EL-2 Fig3 Binding pocket of 5-HT2A model obtained B2-Adrer 5-HT2A TKAFL from CAsTp program 3.2 Active Site of 5-HT, Model TM-5 IL-3 TM-6 The active site recognized by Castp program is shown in B2-Adrenergic Fig 3, which is in good agreement with the available mutation experiments of 5-HT2A and other GPCRs 21 22).The conserved TM-7 active site is surrounded by TM-3, TM-5, TM-6 and TM-7 and OISGNPLIECR covered by EL-2. A hydrophobic cluster, composed of residues 5-HT2A KESCNEDVIGALLNPVVICYLSSASLVYTL w6.48. F6.51 and F6.52 is located at the active site and Fig1 Sequence alignments of p2-adrenergic(2RHI) 3.3 Antagonistic Conformation of 5-HT2A with 5-HT2A generated by Clustalw The asterisk(*) indicates strict identity;a In order to obtain the antagonistic conformation of similarity, the do.)indicates a semi-conserved similarity(sequence identity 5-HT2A, the well-known antagonist ketanserin, which has been is 39.0%, similarity is 62. 4% in the TM region between 5-HT2A and p2-adr). shown to have high affinity for 5-HT2a in an amount of availa- The most conserved residues X 50 for each TM are marked by a black star ble mutation experiments, was docked into the active site of under the sequences. The seven transmembrane helices( TM), extracellular loops(EL) and intracellular loops(IL) are marked above the sequences. 5-hT2a by the advanced docking program GOLD 4. 12, which tween B2-adrenergic receptor and 5-HT2A receptor. Based on considers the flexibility of both receptor and ligand. The key the high homology revealed by sequence alignments, the 3D residues of 5-HT2A involved in the interaction between 5-HT2A

No.4 XIONG Zi-jun et al. 657 GF/B glass fiber filter paper and washing was performed 3 times with 3 mL of eluent(50 mmol/L Tris-HCl, pH=7.7), dry￾ing was conducted in the microwave oven for 8―9 min, then the filter paper was transferred into a 0.5 mL centrifuge tube, to which 500 μL of soluble scintillation fluid was then added. After 30 min of rest in darkness, the radioactivity was mea￾sured. Then pharmacological parameters: percentage inhibition, IC50 and Ki were calculated. Concentrations were measured every two tubes, two independent tests of each compound were conducted. The inhibitor was docked into the 5-HT2A receptor via the flexible docking program GOLD 4.12. Before docking, the final conformation of the ligand was obtained after 1000 steps of energy minimization via the Tripos force field and optimiza￾tion at the DFT/B3LYP/6-311G level by means of Gaussian 2009[20]. Instead of the default 10600 genetic algorithm(GA) runs were performed to fully take into account the flexibility of the ligand. For each GA run, the default GA settings were used with no early termination allowed and internal ligand energy offset turned on. Finally, GOLD-Score was used as the fitness function. 3 Results and Discussion 3.1 3D Structure of 5-HT2A Receptor Sequence alignment(Fig.1) indicates that the sequence identity and similarity are 39.0% and 62.4%, for the TMs be￾Fig.1 Sequence alignments of β2-adrenergic(2RH1) with 5-HT2A generated by ClustalW The asterisk(*) indicates strict identity; a colon(:) indicates a conserved similarity; the dot(.) indicates a semi-conserved similarity(sequence identity is 39.0%, similarity is 62.4% in the TM region between 5-HT2A and β2-adr). The most conserved residues X.50 for each TM are marked by a black star under the sequences. The seven transmembrane helices(TM), extracellular loops(EL) and intracellular loops(IL) are marked above the sequences. tween β2-adrenergic receptor and 5-HT2A receptor. Based on the high homology revealed by sequence alignments, the 3D model of 5-HT2A receptor was assembled taking the X-ray crystal structure of β2-adrenergic receptor as a template. Step￾wise energy minimizations of the homology 5-HT2A model was carried out for 500 steps with Kollman All-Atom force field and the SYBYL program. The final model was verified by a professional structure validation program Procheck 3.5.4. The Procheck statistics shows that 96.1% of the residues of the 5-HT2A model were either in the most favored or in the addi￾tionally allowed regions of the Ramachandran map(Fig.2), suggesting that the overall main chain and side chain struc￾tures were all reasonable. Fig.2 3D model of 5-HT2A receptor(A) and Ramchandran plots of 5-HT2A model(B) The different colored areas indicate disallowed(white), generously allowed (light yellow), additional allowed(yellow) and most favored(red) regions. Fig.3 Binding pocket of 5-HT2A model obtained from CASTp program 3.2 Active Site of 5-HT2A Model The active site recognized by CASTp program is shown in Fig.3, which is in good agreement with the available mutation experiments of 5-HT2A and other GPCRs[21,22]. The conserved active site is surrounded by TM-3, TM-5, TM-6 and TM-7 and covered by EL-2. A hydrophobic cluster, composed of residues W6.48, F6.51 and F6.52, is located at the active site and sup￾posed to participate in a coordinated conformational change. 3.3 Antagonistic Conformation of 5-HT2A In order to obtain the antagonistic conformation of 5-HT2A, the well-known antagonist ketanserin, which has been shown to have high affinity for 5-HT2A in an amount of availa￾ble mutation experiments, was docked into the active site of 5-HT2A by the advanced docking program GOLD 4.12, which considers the flexibility of both receptor and ligand. The key residues of 5-HT2A involved in the interaction between 5-HT2A

658 CHEM RES CHINESE UNIVERSITIES VoL 27 and ketanserin are shown in Fig 4, which indicates that ketan- serin mainly makes contacts with the residues in TM-3, TM-7 and EL-2. Closer inspection of this docking study reveals that ketanserin binds to 5-hT2a by three main interactions. The quinazolinedione fragment penetrates into the hydrophobic pocket formed by TM-6 and TM-7, it is in close contact with Y7. 43 and F6.51. The fluorine atom is halogen bonded to the hydroxyl group of S5.43. However, the most important and conserved interaction remains a salt bridge between the proto- nated nitrogen of the piperidine ring and the carboxy late group of D3. 32. Our docking result is in agreement with most site directed mutagenesis experiments/ 21 22), confirming the accura- Fig 5 Positively charged cluster detected by y of our model. The obtained antagonistic conformation of positive probe NI The red sphere represent cluster of positive probe NI, it forms electrostatic 5-HT2A will be used in the building of pharmacophore model. "sandwich"" interaction with the side chains of D3.32 and F6.51 S5.46. Two big hydrophobic clusters were detected by probe DRY pocket formed by residues F6.51 and F6.52 at the active site and the other is close to hydrophobic residues of 13.29 and Leu228. These hydro- phobic features are thus chosen since they can significantly stabilize the complex. A small negatively charged cluster was omitted due to two reasons: ) the interaction energy with the S5.43 residues at the active site of protein is relatively high, (ID) the binding pocket of the protein has a tremendous negative poten- 743 tial, it repulses the negatively charged ligand and prevents entrance. Finally, a pharmacophore model was generated( Fig. 6) It contains one positive ionizable group, two hydrophobic groups, and one HBd group. As shown in Fig. 6(B), the dis- tances from positive ionizable group, which locates in the cen- ter of the active site. to the hbd. hydri. and hydra are 0.560, 0.832 and 0.997 nm, respectively. They distribute dif- ferent sites at the active site. It shows that our model is quite Fig 4 Structure of ketanserin(A)and the interaction node between 5-HT2A receptor and its anta- gonist ketanserin(B) The key residues involved in the interactions are shown. The hydrogen bonds are shown by dashed lines, ketanserin is represented by sticks and TEOHHYDR2 the key residues if they are the active site of 5-HT2A are shown in stick 3.4 Pharmacophore Model generation HBD The program GRiD 22 was used to generate an interaction ap of the active site residues in the antagonistic conformation with five different probes(DRY, COo, NI,O and NI)ex- 46HBI plained in the methods section. It resulted in a total of 5 inte raction maps. For each probe used, the resulting combined map was overlaid on the active site of 5-HT2A. This helped to iden model, a strategy was used to select probes that would com- plement essential conserved groups at the active site The essential residues at the active site of 5-HT2A are Fig 6 Four-feature pharmacophore model with 6 ex- conserved S5.43 and S5.46. The D3. 32 is essential for the clusion volumes(A)and pharmacophore model binding of the protonated amino group of ligand, S5.43 and at the active site of 5-HT2A(B) S5. 46 may form the conserved interaction with the hydroxyl(A) The hydrophobic features are shown in cya group of the ligand A positively charged cluster was detected sizable feature in red color, the hydrogen bond donor feature in purple by positive probe NI, and it formed electrostatic"sandwich" color and the exclusion volumes in grey;(B)5-HT2A is shown in fanc interaction with the side chains of D3.32 and F6.51(Fig. 5). The cartoon representation. The hydrophobic features(HYDRI and HYDR2)are shown as cyan color meshes, the positive ionizable feature(POS)as a red HBd cluster was detected by o probe and it formed hydrogen mesh, and the HBD as purple mesh, and the key residues if they are at the ond interaction with the hydroxyl group of the side chain of active site of 5-HTzA are shown in stick. The unit of the distances is nm

658 CHEM. RES. CHINESE UNIVERSITIES Vol.27 and ketanserin are shown in Fig.4, which indicates that ketan￾serin mainly makes contacts with the residues in TM-3, TM-7 and EL-2. Closer inspection of this docking study reveals that ketanserin binds to 5-HT2A by three main interactions. The quinazolinedione fragment penetrates into the hydrophobic pocket formed by TM-6 and TM-7, it is in close contact with Y7.43 and F6.51. The fluorine atom is halogen bonded to the hydroxyl group of S5.43. However, the most important and conserved interaction remains a salt bridge between the proto￾nated nitrogen of the piperidine ring and the carboxylate group of D3.32. Our docking result is in agreement with most site￾directed mutagenesis experiments[21,22], confirming the accura￾cy of our model. The obtained antagonistic conformation of 5-HT2A will be used in the building of pharmacophore model. Fig.4 Structure of ketanserin(A) and the interaction mode between 5-HT2A receptor and its anta￾gonist ketanserin(B) The key residues involved in the interactions are shown. The hydrogen bonds are shown by dashed lines, ketanserin is represented by sticks and the key residues if they are the active site of 5-HT2A are shown in stick. 3.4 Pharmacophore Model Generation The program GRID 22 was used to generate an interaction map of the active site residues in the antagonistic conformation with five different probes(DRY, COO– , N1+ , O and N1) ex￾plained in the methods section. It resulted in a total of 5 inte￾raction maps. For each probe used, the resulting combined map was overlaid on the active site of 5-HT2A. This helped to iden￾tify the conserved positions of each probe. In developing the model, a strategy was used to select probes that would com￾plement essential conserved groups at the active site. The essential residues at the active site of 5-HT2A are conserved S5.43 and S5.46. The D3.32 is essential for the binding of the protonated amino group of ligand, S5.43 and S5.46 may form the conserved interaction with the hydroxyl group of the ligand. A positively charged cluster was detected by positive probe N1+ , and it formed electrostatic “sandwich” interaction with the side chains of D3.32 and F6.51(Fig.5). The HBD cluster was detected by O probe and it formed hydrogen bond interaction with the hydroxyl group of the side chain of Fig.5 Positively charged cluster detected by positive probe N1+ The red sphere represent cluster of positive probe N1+ , it forms electrostatic “sandwich” interaction with the side chains of D3.32 and F6.51. S5.46. Two big hydrophobic clusters were detected by probe DRY. One is close to the hydrophobic pocket formed by residues F6.51 and F6.52 at the active site and the other is close to hydrophobic residues of I3.29 and Leu228. These hydro￾phobic features are thus chosen since they can significantly stabilize the complex. A small negatively charged cluster was omitted due to two reasons: (I) the interaction energy with the residues at the active site of protein is relatively high; (II) the binding pocket of the protein has a tremendous negative poten￾tial, it repulses the negatively charged ligand and prevents its entrance. Finally, a pharmacophore model was generated(Fig.6). It contains one positive ionizable group, two hydrophobic groups, and one HBD group. As shown in Fig.6(B), the dis￾tances from positive ionizable group, which locates in the cen￾ter of the active site, to the HBD, HYDR1, and HYDR2 are 0.560, 0.832 and 0.997 nm, respectively. They distribute dif￾ferent sites at the active site. It shows that our model is quite Fig.6 Four-feature pharmacophore model with 6 ex￾clusion volumes(A) and pharmacophore model at the active site of 5-HT2A(B) (A) The hydrophobic features are shown in cyan color, the positive(POS) ionizable feature in red color, the hydrogen bond donor feature in purple color and the exclusion volumes in grey; (B) 5-HT2A is shown in fancy cartoon representation. The hydrophobic features(HYDR1 and HYDR2) are shown as cyan color meshes, the positive ionizable feature(POS) as a red mesh, and the HBD as purple mesh, and the key residues if they are at the active site of 5-HT2A are shown in stick. The unit of the distances is nm

XIONG Zi-jun et al. reliable and reasonable order to be considered for further analysis and docking studies As a result, 371 compounds were identified by the model with 3.5 Database Searching and Candidate Inhibitors an average fit value of 2.367. In order to reduce false positive Identifying hits and nondrug- like compounds, several selection criteria Model shown in Fig. 6(A)was used to search Asinex Gold were used to filter out redundancies. An initial filtering was Collection database, which contains approximate 20000 performed according to Lipinski's Rule of Five by means of compounds, via the best flexible search method implemented in Pipeline Pilot software: molecular weight(Mw)<500, hydro- CATALYST module encoded Discovery Studio 3.0 program phobic constant Alog P<5, number of HBA<10, number HBD<5, and the number of rotatable bonds( RBsk10. As a The hit compounds from database were then fitted to the phar- result, the numbers of hits of the model reduced to 184.Then macophore model with CATALYST module encoded Discovery the scoring function G-Score, fit value, scaffold diversity and Studio 3.0 program. For each compound, a fit value was re- visual observation were considered generally. Finally, 7 hits turned that represented how well the compound fit to the phar- macophore model Compounds had to map the four features were selected and purchased for pharmacological assay. The ore mo fit value of more than 2.0 in relative drug-like properties of the 7 hits are listed in Table I Drug-like properties, scores and fit values of the top 7 compounds Compund II weight Number of HDs Number of HAs ALO Number of RB G-Score Fit value Awo1l19 431.86 3.891 8 47.9510 2.646 456.52 12 50.2230 2.909 305245768 42.7878 466.57 473983 01511644 2202 5686668 3.945 7 46.1974 2.462 03008103 12.49 7 51.2033 89l 01817599 479.53 2.217 3.6 Pharmacological Profiles compounds had potent inhibitory activity against 5-HT2A with inhibitory ratios above 80%. Among them, compound Our pharmacological test(Table 2)shows that tw 05245768 was the most potent with a percentage inhibition of Table 2 Biological test results of virtual screening hits 94.59% and a Ki value of (593.89*34.10) nmol/L. Fig.7 Compound shows the fitness of this compound to the pharmacophore rate(%)(nmol-L-) model and its fit value is 3.076, which verifies the rationality Awol119 X人心F of our pharmacophore model 47.2 HYDRI HYDR2 05245768 4.59593.89±34.10 01176572 28.6 HBD Fig 7 Compound 05245768 mapped with the pharmacophore model Pharmacophore features are colored the same as those in Fig. 6. l511644 12.24 3.7 Receptor-ligand Interaction Our docking studies suggest that there are three main im- portant interactions: salt bridge, hydrophobic and hydrogen ond interactions[Fig 8(A) bridge and hydrophobic 27.5 interactions are similar to those of ketanserin in the binding pocket of 5-HT2A. The salt bridge is found between the proto- nated nitrogen atom of the ligand and the carboxylate group of the conserved D3. 32 and the hydrophobic interaction is formed 0l817599 1981 between the benzopyrimidine ring and a hydrophobic cluster formed by w3. 28, 13.29 and F6.51. The cavity formed by the hydrophobic cluster in 05245768-5-HT2A complex is slightly larger than that in ketanserin-5-HT2A complex. Compound

No.4 XIONG Zi-jun et al. 659 reliable and reasonable. 3.5 Database Searching and Candidate Inhibitors Identifying Model shown in Fig.6(A) was used to search Asinex Gold Collection database, which contains approximate 200000 compounds, via the best flexible search method implemented in CATALYST module encoded Discovery Studio 3.0 program[23]. The hit compounds from database were then fitted to the phar￾macophore model with CATALYST module encoded Discovery Studio 3.0 program. For each compound, a fit value was re￾turned that represented how well the compound fit to the phar￾macophore model. Compounds had to map the four features in the pharmacophore model with a fit value of more than 2.0 in order to be considered for further analysis and docking studies. As a result, 371 compounds were identified by the model with an average fit value of 2.367. In order to reduce false positive hits and nondrug-like compounds, several selection criteria were used to filter out redundancies. An initial filtering was performed according to Lipinski's Rule of Five by means of Pipeline Pilot software: molecular weight(MW)<500, hydro￾phobic constant AlogP<5, number of HBA<10, number of HBD<5, and the number of rotatable bonds(RBs)<10. As a result, the numbers of hits of the model reduced to 184. Then the scoring function G-Score, fit value, scaffold diversity and visual observation were considered generally. Finally, 7 hits were selected and purchased for pharmacological assay. The relative drug-like properties of the 7 hits are listed in Table 1. Table 1 Drug-like properties, scores and fit values of the top 7 compounds No. Compund ID Molecular weight Number of HDs Number of HAs ALogP Number of RBs G-Score Fit value 1 AW01119 431.86 1 5 3.891 8 47.9510 2.646 2 02987623 456.52 1 6 4.373 12 50.2230 2.909 3 05245768 409.48 2 8 2.925 7 42.7878 3.076 4 01176572 466.57 2 6 3.118 8 47.3983 2.718 5 01511644 449.34 0 6 3.945 7 46.1974 2.462 6 03008103 412.49 2 6 1.556 7 51.2033 2.891 7 01817599 479.53 1 8 2.217 9 53.4037 2.982 3.6 Pharmacological Profiles Our pharmacological test(Table 2) shows that two Table 2 Biological test results of virtual screening hits Compound ID Structure Inhibitory rate(%) Ki/ (nmol·L–1) AW01119 81.35 NT* 02987623 47.2 NT 05245768 94.59 593.89±34.10 01176572 28.6 NT 01511644 12.24 NT 03008103 27.5 NT 01817599 19.81 NT * NT: not tested. compounds had potent inhibitory activity against 5-HT2A with inhibitory ratios above 80%. Among them, compound 05245768 was the most potent with a percentage inhibition of 94.59% and a Ki value of (593.89±34.10) nmol/L. Fig.7 shows the fitness of this compound to the pharmacophore model and its fit value is 3.076, which verifies the rationality of our pharmacophore model. Fig.7 Compound 05245768 mapped with the pharmacophore model Pharmacophore features are colored the same as those in Fig.6. 3.7 Receptor-ligand Interaction Our docking studies suggest that there are three main im￾portant interactions: salt bridge, hydrophobic and hydrogen bond interactions[Fig.8(A)]. The salt bridge and hydrophobic interactions are similar to those of ketanserin in the binding pocket of 5-HT2A. The salt bridge is found between the proto￾nated nitrogen atom of the ligand and the carboxylate group of the conserved D3.32 and the hydrophobic interaction is formed between the benzopyrimidine ring and a hydrophobic cluster formed by W3.28, I3.29 and F6.51. The cavity formed by the hydrophobic cluster in 05245768-5-HT2A complex is slightly larger than that in ketanserin-5-HT2A complex. Compound

CHEM. RES CHINESE UNIVERSITIES VoL 27 05245768 differs with ketanserin in that its hydroxyl tail forms molecular docking, it guided the optimization of the ligand by a hydrogen bond with $5.46 instead of the aforemen tioned enhancing the lipophilicity of its aromatic ring. More com- halogen bond between the fluorine atom and $5.43. This result pounds as designed will be purchased or synthesized for fur- is consistent well with previous studies 24,251. From the docking ther biological test to discover more effective inhibitors of result, we can predict that in order to keep or increase the acti- 5-HT2A receptor. In this study, we developed a pharmaco- vity, the protonated nitrogen atom of the ligand is necessary for phore based virtual screening method by integrating a set of the electrostatic interaction, the hydroxyl group should also be computational approaches and acological test. The reserved for the hydrogen bond interaction, with regard to results described in this paper that this method is phenyl ring, it can be substituted by lipophilic groups such as very efficient for identifying novel bioactive agents. This methyl, isopropyl, phenyl and halogen for stronger hydropho- workflow can also be applied in the study of hit identification bic interaction with the hydrophobic pocket formed by w3.28, and lead optimization. 13.29 and Leu228[ Fig 8(B) (A) Reference [2] Horacek J, Pharmacopsychiatry, 2000, 33(Suppl. 1), 34 [3] Aghajanian G. K, Marek G. J, Brain. Res. Rev., 2000, 31(2/3), [4] Reynolds G. P, J Psychopharmacol., 2004, 18(3), 340 F [5] Meltzer H. Y, Matsubara S, Lee J C.,J. Pharmacol. Exp. Thera- peat,1989,25/(1),238 [6] Huang E. S, Protein. Sci., 2003, 12(7), 1360 [7] Dunn M. F, Handbook of Proteins, 2007, 2, 678 18] Cherezov v, Rosenbaum D M, Hanson M. A, Rasmussen S. G F, Thian F. S, Kobilka T. S, Choi H. J, Kuhn P, Weis I. W ian K. K, et al, Science, 2007, 318(5854), 1258 [9] Accelrys Software, Discovery Studio 3.0 Edition, San Diego, CA, 0] Ballesteros J. A, Weinstein H, Meth. Neurosci [11] Sali A, Blundell T. L, J. Mol. Biol, 1993, 234(3), [12] Fiser A, Sali A,, Meth. Eneymol, 2003, 374, 46 Fig 8 Binding modes of compound 05245768 with 13 Pearson W.R., Meth. Ensymol, 1990, 183, 63 . hT2A(A)and hydrophobic pocket around [14 Thompson J. D, Higgins D G, Gibson T J, Nucl. Acids. Res. the benzyl ring of compound 05245768 formed 1994,2(22),4673 by w3.28, 13.29 and Leu228(B) [15] SYBYL Manual, Tripos Inc, 2001 The key residues involved in the interactions are shown. The salt bridge [16] Laskowski R. A, Mac Arthur M. W, Moss D. S, Thornton J. M, J. and hydrogen bond are shown by dashed lines, compound 05245768 is Appl. Crystallogr, 1993, 26(2),283 presented by sticks and the key residues are shown in stick if they are [17 Jones G, Willett P, Glen R. C, Leach A. R, Taylor R, J. Mol. Bi- the active site of 5-HT2A. The unit of the distances is nm. ol,1995,245,43 4 Conclusions [18 Binkowski T. A, Naghibzadeh S, Liang J, Nucl. Acids. Res, 2003, 3l(13),3352 We reported herein the 3D structure of 5-HT2A receptor [19] Leach C, Br J. Clin. Psychol., 1988, 27( Part 2), 173 with the newly resolved high resolution X-ray structure of [20] Frisch M. J, Trucks G. w,Schlegel H B, Scuseria G. E, Robb M B2-adrenergic receptor as template. The structure was verified A, Cheeseman J. R, Scalmani G, Barone V, Mennucci B, Pe- G al. Gaussian 09. Revision AI. Gaussian Inc model was developed based on the grid calculation in which allingford CT. 2009 five types of chemical probes ctive site 21] Shapiro D. A, Kristiansen K, Kroeze W.K., Roth B. L, Mol and the cluster analyses were performed to select the phar- Pharmacol.,2000,58(5),877 122 Wurch T, Palmier C, Pauwels P J, Biochem. Pharmacol., 2000 for the gold collection of asinex database which contains ca 5909),1l17 200000 molecules. The combination of Lipinski's Rule of 23] Bao H J, Tang Y L, XuX. J, Xiang J F, Zheng Z H, Lu X.H., Five with fit value and G-Score was used to filter redundan- Chem. J. Chinese Universities, 2010, 31(50,938 cies and false positives. Finally, 7 molecules were obtained [24 Bruno A, Guadix A. E, Costantino G, J. Chem. Inf Model, 2009 and purchased for the biological binding assay. A potent 496),1602 5-hT2A inhibitor 05245768 with a Ki value of (593.89*34.10) 5] Dezi C, Brea J, Alvarado M, Ravina E, Masaguer C. F, Loza mol/L was found I, Sanz F, Pastor M J. Med. Chem., 2007, 50(14), 3242 reasonable receptor-ligand revealed by

660 CHEM. RES. CHINESE UNIVERSITIES Vol.27 05245768 differs with ketanserin in that its hydroxyl tail forms a hydrogen bond with S5.46 instead of the aforemen tioned halogen bond between the fluorine atom and S5.43. This result is consistent well with previous studies[24,25]. From the docking result, we can predict that in order to keep or increase the acti￾vity, the protonated nitrogen atom of the ligand is necessary for the electrostatic interaction, the hydroxyl group should also be reserved for the hydrogen bond interaction, with regard to the phenyl ring, it can be substituted by lipophilic groups such as methyl, isopropyl, phenyl and halogen for stronger hydropho￾bic interaction with the hydrophobic pocket formed by W3.28, I3.29 and Leu228[Fig.8(B)]. Fig.8 Binding modes of compound 05245768 with 5-HT2A(A) and hydrophobic pocket around the benzyl ring of compound 05245768 formed by W3.28, I3.29 and Leu228(B) The key residues involved in the interactions are shown. The salt bridge and hydrogen bond are shown by dashed lines, compound 05245768 is represented by sticks and the key residues are shown in stick if they are the active site of 5-HT2A. The unit of the distances is nm. 4 Conclusions We reported herein the 3D structure of 5-HT2A receptor with the newly resolved high resolution X-ray structure of β2-adrenergic receptor as template. The structure was verified by a professional program Procheck 3.5.4. A pharmacophore model was developed based on the grid calculation in which five types of chemical probes were used to map the active site and the cluster analyses were performed to select the phar￾macophore feature. The obtained model was used to search for the Gold collection of Asinex Database which contains ca. 200000 molecules. The combination of Lipinski’s Rule of Five with fit value and G-Score was used to filter redundan￾cies and false positives. Finally, 7 molecules were obtained and purchased for the biological binding assay. A potent 5-HT2A inhibitor 05245768 with a Ki value of (593.89±34.10) nmol/L was found. A reasonable receptor-ligand interaction was revealed by molecular docking, it guided the optimization of the ligand by enhancing the lipophilicity of its aromatic ring. More com￾pounds as designed will be purchased or synthesized for fur￾ther biological test to discover more effective inhibitors of 5-HT2A receptor. In this study, we developed a pharmaco￾phore based virtual screening method by integrating a set of computational approaches and pharmacological test. The results described in this paper indicate that this method is very efficient for identifying novel bioactive agents. This workflow can also be applied in the study of hit identification and lead optimization. Reference [1] Abi-Dargham A., Laruelle M., Eur. Psychiatr., 2005, 20(1), 15 [2] Horacek J., Pharmacopsychiatry, 2000, 33(Suppl. 1), 34 [3] Aghajanian G. K., Marek G. J., Brain. Res. Rev., 2000, 31(2/3), 302 [4] Reynolds G. P., J. Psychopharmacol., 2004, 18(3), 340 [5] Meltzer H. Y., Matsubara S., Lee J. C., J. Pharmacol. Exp. Thera￾peut., 1989, 251(1), 238 [6] Huang E. S., Protein. Sci., 2003, 12(7), 1360 [7] Dunn M. F., Handbook of Proteins, 2007, 2, 678 [8] Cherezov V., Rosenbaum D. M., Hanson M. A., Rasmussen S. G. F., Thian F. S., Kobilka T. S., Choi H. J., Kuhn P., Weis I. W., Brian K. K., et al., Science, 2007, 318(5854), 1258 [9] Accelrys Software, Discovery Studio 3.0 Edition, San Diego, CA, 2010 [10] Ballesteros J. A., Weinstein H., Meth. Neurosci., 1995, 25, 366 [11] Sali A., Blundell T. L., J. Mol. Biol., 1993, 234(3), 779 [12] Fiser A., Sali A., Meth. Enzymol., 2003, 374, 461 [13] Pearson W. R., Meth. Enzymol., 1990, 183, 63 [14] Thompson J. D., Higgins D. G., Gibson T. J., Nucl. Acids. Res., 1994, 22(22), 4673 [15] SYBYL Manual, Tripos Inc., 2001 [16] Laskowski R. A., MacArthur M. W., Moss D. S., Thornton J. M., J. Appl. Crystallogr., 1993, 26(2), 283 [17] Jones G., Willett P., Glen R. C., Leach A. R., Taylor R., J. Mol. Bi￾ol., 1995, 245, 43 [18] Binkowski T. A., Naghibzadeh S., Liang J., Nucl. Acids. Res., 2003, 31(13), 3352 [19] Leach C., Br. J. Clin. Psychol., 1988, 27( Part 2), 173 [20] Frisch M. J., Trucks G. W., Schlegel H. B., Scuseria G. E., Robb M. A., Cheeseman J. R., Scalmani G., Barone V., Mennucci B., Pe￾tersson G. A., et al., Gaussian 09, Revision A.1., Gaussian Inc., Wallingford CT, 2009 [21] Shapiro D. A., Kristiansen K., Kroeze W. K., Roth B. L., Mol. Pharmacol., 2000, 58(5), 877 [22] Wurch T., Palmier C., Pauwels P. J., Biochem. Pharmacol., 2000, 59(9), 1117 [23] Bao H. J., Tang Y. L., Xu X. J., Xiang J. F., Zheng Z. H., Lu X. H., Chem. J. Chinese Universities, 2010, 31(5), 938 [24] Bruno A., Guadix A. E., Costantino G., J. Chem. Inf. Model., 2009, 49(6), 1602 [25] Dezi C., Brea J., Alvarado M., Raviña E., Masaguer C. F., Loza M. I., Sanz F., Pastor M., J. Med. Chem., 2007, 50(14), 3242