复旦大学：《药物设计学》课程教学资源（教学研究）基于配体结构的药物设计_Structure-based 3D-QSAR studies on thiazoles as 5-HT3 receptor antagonists

团购合买资源类别：文库，文档格式：PDF，文档页数：11，文件大小：608.2KB

ol model(2007)13:121-131 DOI10.1007/s00894-006-0131-1 ORIGINAL PAPER Structure-based 3D-QSAR studies on thiazoles as 5-HT3 receptor antagonists Li-Ping Zhu. De-Yong Ye. Yun tang Received: 2 December 2005/Accepted: 30 June 2006/Published online: 5 September 2006 C Springer-Verlag 2006 bstract Structure-based 3D-QSAR studies were per- Introduction formed on 20 thiazoles against their binding affinities to he 5-HT3 receptor with comparative molecular field Nausea and vomiting are major side effects associated with analysis(CoMFA) and comparative molecular similarity chemotherapy, radiotherapy, and operation [1, 2]. It is now indices analysis (CoMSIA). The thiazoles were well established that serotonin type 3(5-HT3) receptors, docked into the binding pocket of a human 5-HT3A present on vagal afferents in the Gl tract mucosa and in the homology model, constructed on the basis of the crystal brainstem centers, are involved in the vomiting reflex. structure of the snail acetylcholine binding protein Initiation of emesis is probably due to the release of (AChBP), using the GOLd program. The docked con- serotonin from enterochromaffin cells in the small intestine, formations were then extracted and used to build the 3D- which activates vagal afferent nerves via 5-HT3 receptors QSAR models, with cross-validated 2y values 0.785 and [3]. Delayed emesis may involve central 5-HT3 receptors 0.744 for CoMFA and CoMSIA, respectively. An addition- and/or serotonin stored in the enterochromaffin cells al five molecules were used to validate the models further, Specific 5-HT3 receptor antagonists such as Ondansetron giving satisfactory predictive r2 values of 0.582 and 0.804 Granisetron and Tropisetron block for CoMFA and CoMSIA, respectively. The results would probably by competitive inhibition at the 5-HT3 receptor be helpful for the discovery of new potent and selective sites centrally and peripherally [4] 5-HT3 receptor antagonists. Three subtypes of 5-HT3 receptors have been identified (1)5-HT3A, a neuronal receptor directly coupled to cation- Keywords Structure-based 3D-QSAR CoMFA selective channels; (2)5-HT3B, a regulatory subunit able to CoMSIA 5-HT3 receptor antagonists. Homology modeling modulate the intrinsic channel activity of 5-HT3A; and ( 3) 5-HT3c, a subunit that appears to modulate the 5-HT3 receptor responses [5]. The subunit 5-HT3A is functional omo-oligomeric while 5-HT3B is non-functional hetero- loop ligand-gated ion channels (LGICs), whose members LP.Zhu·D-Y.Ye)·Y.Tang share significant structural and functional homology to each Department of Medicinal Chemistry, School of Pharmacy, Fudan University other. Each subtype consists of five subunits, and each 138 Yixueyuan Road, subunit has a large extracellular N-terminal region and four putative transmembrane domains 6, 7]. Early experimental mail: dyyeashmu.edu.cn evidence showed that the ligand-binding site is located at Y Tang(<) the interface of two adjacent subunits. Unfortunately, date the three-dimensional (3D) structure of 5-HT3 recep- East China University of Science and Technology 130 Meilong Road tors has not been elucidated, which seriously limits our Shanghai 200237. China understanding of the antagonistic mechanism of the mail:ytang234@yahoo.com.cn receptor in detail. However, the structure of acetylcholine

ORIGINAL PAPER Structure-based 3D-QSAR studies on thiazoles as 5-HT3 receptor antagonists Li-Ping Zhu & De-Yong Ye & Yun Tang Received: 2 December 2005 /Accepted: 30 June 2006 / Published online: 5 September 2006 # Springer-Verlag 2006 Abstract Structure-based 3D-QSAR studies were performed on 20 thiazoles against their binding affinities to the 5-HT3 receptor with comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA). The thiazoles were initially docked into the binding pocket of a human 5-HT3A receptor homology model, constructed on the basis of the crystal structure of the snail acetylcholine binding protein (AChBP), using the GOLD program. The docked conformations were then extracted and used to build the 3DQSAR models, with cross-validated r2 cv values 0.785 and 0.744 for CoMFA and CoMSIA, respectively. An additional five molecules were used to validate the models further, giving satisfactory predictive r2 values of 0.582 and 0.804 for CoMFA and CoMSIA, respectively. The results would be helpful for the discovery of new potent and selective 5-HT3 receptor antagonists. Keywords Structure-based 3D-QSAR . CoMFA . CoMSIA . 5-HT3 receptor antagonists . Homology modeling Introduction Nausea and vomiting are major side effects associated with chemotherapy, radiotherapy, and operation [1, 2]. It is now well established that serotonin type 3 (5-HT3) receptors, present on vagal afferents in the GI tract mucosa and in the brainstem centers, are involved in the vomiting reflex. Initiation of emesis is probably due to the release of serotonin from enterochromaffin cells in the small intestine, which activates vagal afferent nerves via 5-HT3 receptors [3]. Delayed emesis may involve central 5-HT3 receptors and/or serotonin stored in the enterochromaffin cells. Specific 5-HT3 receptor antagonists such as Ondansetron, Granisetron and Tropisetron block nausea and vomiting, probably by competitive inhibition at the 5-HT3 receptor sites centrally and peripherally [4]. Three subtypes of 5-HT3 receptors have been identified: (1) 5-HT3A, a neuronal receptor directly coupled to cationselective channels; (2) 5-HT3B, a regulatory subunit able to modulate the intrinsic channel activity of 5-HT3A; and (3) 5-HT3C, a subunit that appears to modulate the 5-HT3 receptor responses [5]. The subunit 5-HT3A is functional homo-oligomeric while 5-HT3B is non-functional heteromeric. 5-HT3 receptors belong to the superfamily of Cysloop ligand-gated ion channels (LGICs), whose members share significant structural and functional homology to each other. Each subtype consists of five subunits, and each subunit has a large extracellular N-terminal region and four putative transmembrane domains [6, 7]. Early experimental evidence showed that the ligand-binding site is located at the interface of two adjacent subunits. Unfortunately, to date the three-dimensional (3D) structure of 5-HT3 receptors has not been elucidated, which seriously limits our understanding of the antagonistic mechanism of the receptor in detail. However, the structure of acetylcholine J Mol Model (2007) 13:121–131 DOI 10.1007/s00894-006-0131-1 L.-P. Zhu : D.-Y. Ye (*) : Y. Tang (*) Department of Medicinal Chemistry, School of Pharmacy, Fudan University, 138 Yixueyuan Road, Shanghai 200032, China e-mail: dyye@shmu.edu.cn Y. Tang School of Pharmacy, East China University of Science and Technology, 130 Meilong Road, Shanghai 200237, China Y. Tang (*) School of Pharmacy, East China University of Science and Technology, 130 Meilong Road, Shanghai 200237, China e-mail: ytang234@yahoo.com.cn

binding protein (AChBP) found in the snail Lymnaea stagnalis has been determined by X-ray crystallography. This protein is a member of the LGICs, and shares 19% homologous sequence with the extracellular domain of the 5-HT3A receptor [8]. Therefore, a homology model of the extracellular domain of the human 5-HT3A receptor was built based on the crystal structure of AChBP, and known selective antagonists were docked into the binding site to validate the model. During the course of this work, a similar study using a range of antagonists was published [9]. QSAR and 3D-QSAR have long been used to elucidate the mechanisms of drug action and for lead optimization. CoMFA (comparative molecular field analysis) [10] and CoMSIA (comparative molecular similarity indices analysis) [11] are two of the most widely used 3D-QSAR methods. In both approaches, molecular property fields are evaluated between a probe atom and each molecule of a dataset at the intersections of a regularly spaced grid. CoMFA calculates steric and electrostatic properties according to Lennard-Jones and Coulomb potentials, while CoMSIA considers five different similarity fields: steric, electrostatic, hydrophobic, hydrogen-bond donor, and hydrogen-bond acceptor properties. When the 3D-structure of a receptor is also available, ligand- and structure-based drug design methods can be combined. The structure-based 3DN S Ar NH N H3C N S Ar NH N N S X NH N H 1-19 20-23 24-25 Ar 1. Indol-3-yl (7.921) 2. o-MeOC6H4 (9.377) 3. C6H5 (8.745) 4. Quinolin-8-yl (9.000) 5. o-FC6H4 (8.712) 6. o-EtOC6H4 (9.022) 7. o-HOC6H4 (8.824) 8. o-MeC6H4 (8.061) 9. m-ClC6H4 (9.056) 10. m-MeOC6H4 (8.699) 11. m-FC6H4 (8.432) 12. m-BrC6H4 (8.699) 13. p-FC6H4 (8.310) 14. p-MeC6H4 (6.959) 15. p-BrC6H4 (6.569) 16. p-ClC6H4 (7.469) 17. 1-Methoxynaphth-2-yl (7.022) 18. 2-Methoxynaphth-1-yl (6.398) 19. 2,6-(MeO)2C6H3 (6.367) Ar 20. C6H5 (9.000) 21. o-MeOC6H4 (9.538) 22. Quinolin-8-yl (9.509) 23. Indol-3-yl (7.848) X 24. CH2 (8.481) 25. COOCH2CH2 (8.149) HN NH2 Scheme 1 Structures and actual pKi values of molecules used for 3D-QSAR studies [12, 13] 122 J Mol Model (2007) 13:121–131

J Mol Model(2007)13:121-131 12 Scheme 2 Sequence alignment AChBP(I) L DRADILYNIROTSRPDVIPTORD.RPVAVSVSLKFINILE of the extracellular domain 5-TA(32) PALLRLSDYLLTNYRKGVRPVRDWRKPTTVSIDVIVYAILN subunit with that of AchBP The crystal structure of AChBP AChBP(41) VNEITNEVDVVEWOOTTWSDRTLAWNSSHS.PDQVSVPIS 5-HT3A(73) VDEKNOVLTTYIWYROYWTDEFLOWNPEDFDNITKLSI D of AChBP(a-helix, red B-strand, blue; 310-helix, green) indicated A ChBP(SO) SLWVPDLAAYN.AISKPEVITPQLARVVSDGEVLYMPSIRO 5-TsA(I14) SIWVPDILINEEVDVGKSP.NIPYVYIRHOGEVONYKPLOV A ChBP(120) RESCDVSGVDTESG. ATCRIKIGSWTHHSREISVDPTTE 5-HTA(154) VTACSLDIYNFPFDVONCSLTFTSWLHTIQDINISLWRLP AChBP(l58) NSDDSEYESQYSRFEILDVTOKKNSVTYSCCPEAYEDVE 5-HT3A(194) EKVKSDRSVFMNQGEWELLGVLPYEREES. MESSNYYAEMK AChBP(97) VSLNFRKKGRSEIL 5HT3A(234) FYVVIRRRPLFYVV QSAR method is the result of such a combination, and it replacement of [HI-Tropisetron binding to NG-108-15 could provide more information for lead optimization cells. The compounds were divided into two sets: 20 In order to understand the antagonistic mechanism and molec randon guide the discovery of more potent ligands, a series of whereas the remaining five molecules served as an external highly potent and selective 5-HT3 receptor antagonists, test set reported by Nagel et al. [ 12, 13] were chosen to perform 3D-QSAR studies with both CoMFA and CoMSIA meth- Molecular modeling ods, based on their docking conformations on the structural model of human 5-HT3A receptor in this paper. The All molecular modeling studies were carried out on ar molecules chosen contain a thiazole moiety linking an R14000 SGI Fuel workstation using the molecular aromatic group and a nitrogenous basic region(Scheme 1); modeling software package SYBYL v6.9[14]. The three the thiazole group appears to be acting as a carbonyl dimensional model of the extracellular region of human bioisostere in this system. By mapping the CoMFA and 5-HT3A receptor was built using the module bIOPOlY- CoMSIA contour plots onto the structural model of the MER based on the crystal structure of AChBP determined eceptor, the results of this study might be conversely at 2.7 A(PDB entry code: 119B)[8]. The pentamer was validate the 3D-structural model of the receptor generated by replacing each amino acid of AChBP with the corresponding one of human 5-HT3A receptor's extracellu lar regions, ensuring the conformation of the templates Materials and methods backbone unchanged. The model generated was fixed using the Biopolymer command Fix Proline and Fix Sidechains, Sequence alignment to relieve all bad van der waals contacts. Then the mode was minimized using the AMBER4 1 force field [15] with a The sequence alignment of the extracellular domain of distance-dependent dielectric constant of 5.0 and a gradient human 5-HT3A receptor with AChBP is shown in Scheme 2 convergence value of 0.05 kcal mol in 2,000 cycles Atomic charges were calculated using the AMBER41 method. Then subunit a and b were extracted and used Dataset as the initial model for further docking studies. Sixteen 5-HT Twenty-five thiazoles were collected from literature docked into the binding pocket of the initial model, and reported by Nagel [12, 13]. The binding affinities of these the residues within 10 A of the ligand of the ligand-receptor compounds to the 5-HT3 receptor were evaluated with complex were minimized. Finally, the model was validated

QSAR method is the result of such a combination, and it could provide more information for lead optimization. In order to understand the antagonistic mechanism and guide the discovery of more potent ligands, a series of highly potent and selective 5-HT3 receptor antagonists, reported by Nagel et al. [12, 13] were chosen to perform 3D-QSAR studies with both CoMFA and CoMSIA methods, based on their docking conformations on the structural model of human 5-HT3A receptor in this paper. The molecules chosen contain a thiazole moiety linking an aromatic group and a nitrogenous basic region (Scheme 1); the thiazole group appears to be acting as a carbonyl bioisostere in this system. By mapping the CoMFA and CoMSIA contour plots onto the structural model of the receptor, the results of this study might be conversely validate the 3D-structural model of the receptor. Materials and methods Sequence alignment The sequence alignment of the extracellular domain of human 5-HT3A receptor with AChBP is shown in Scheme 2 [8]. Dataset Twenty-five thiazoles were collected from literature reported by Nagel [12, 13]. The binding affinities of these compounds to the 5-HT3 receptor were evaluated with replacement of [3 H]-Tropisetron binding to NG-108-15 cells. The compounds were divided into two sets: 20 molecules selected randomly to form the training set, whereas the remaining five molecules served as an external test set. Molecular modeling All molecular modeling studies were carried out on an R14000 SGI Fuel workstation using the molecular modeling software package SYBYL v6.9 [14]. The threedimensional model of the extracellular region of human 5-HT3A receptor was built using the module BIOPOLYMER based on the crystal structure of AChBP determined at 2.7 Å (PDB entry code: 1I9B) [8]. The pentamer was generated by replacing each amino acid of AChBP with the corresponding one of human 5-HT3A receptor’s extracellular regions, ensuring the conformation of the template’s backbone unchanged. The model generated was fixed using the Biopolymer command Fix Proline and Fix Sidechains, to relieve all bad van der Waals contacts. Then the model was minimized using the AMBER4.1 force field [15] with a distance-dependent dielectric constant of 5.0 and a gradient convergence value of 0.05 kcal mol−1 in 2,000 cycles. Atomic charges were calculated using the AMBER4.1 method. Then subunit A and B were extracted and used as the initial model for further docking studies. Sixteen 5-HT3 receptor antagonists, known as ‘setrons’, were docked into the binding pocket of the initial model, and the residues within 10 Å of the ligand of the ligand-receptor complex were minimized. Finally, the model was validated Scheme 2 Sequence alignment of the extracellular domain of human 5-HT3A receptor subunit with that of AChBP. The crystal structure of AChBP was taken from the PDB file (1I9B). Secondary structure elements of AChBP (α-helix, red; β-strand, blue; 310-helix, green) were indicated [8] J Mol Model (2007) 13:121–131 123

J Mol model(2007)13:121-131 by ProcHeck and Verify 3d(hTtp: //nihserver. mbi. ucla table was constructed from similarity indices calculated at edu/SAvS/, [16]) the intersections of a regularly spaced lattice(2 A grid)in The 16 setrons, compounds 2 and 24 were retrieved from CoMSIA. themDdrdatabasefromMdl(htTp://www.mdli.com).The 2D-structures were subsequently converted into 3D-struc- PLS analysis and validation of QSAR models tures with CoriNa(htTp: //www2. ccc. uni-erlangen. de/ software/corina/free struct. html). All the other 23 com- The CoMFA/CoMSIA fields combined with observed pounds were constructed based on the structure of com- biological activities(pki) were included in a molecular pound 2. All molecules were set in their unprotonated state spreadsheet, and partial least square(PLS) methods [20] and Gasteiger-Huickel charges were added in SYBYL were used to generate 3D-QSAR models. To check the statistical significance of the models, cross-validations were Ligand docking done to choose the optimum number of components(M)by means of the leave-one-out ( LOo)[21] procedure using the The binding site of the 5-HT3 receptor was defined as enhanced version of PLs, the SAMPLS method [221 atoms within a radius of 16 A of the Ca atom of Trp178 in subsequently used to derive the final QSAR models. The the binding pocket to ensure that most of the residues optimal numbers of components were selected on the basis critical for ligand binding verified/revealed by previous of the highest cross-validated correlation coefficient(2) experimental data were included. All molecules were which is defined as follows set with v2.2[17-19]. The default settings of GOLD were used, and 2=_2(Ypredicted-Yactmal) (1) no flipping was allowed CoMFA and comsia Where Predicted, Actual, and Ymean are predicted, actual and mean values of the target property (pki), respectively First,the docked conformations of the molecules in the The non-cross-validated PLS analyses were performed with training set were placed into a 3D cubic lattice with 2a a column filter value of 2.0. The CoMFA/CoMSIA results grid. CoMFA fields were generated using an sp carbon were interpreted graphically by field contribution maps probe atom carrying +1 charge to generate steric(Lennard- using theSTDEVxCOEFF' field type Jones potential) and electrostatic( Coulomb potential) fields To assess the predictive power of the 3D-QS CoMFA-Standard method in SYBYL. A 30 kcal mol-1 test set molecules were predicted. The predictive rrof the at each grid point. The CoMFA fields were scaled by the derived using the training set, biological activities value is calculated as follows energy cutoff was applied. The steric, electrostatic, hydrophobic, hydrogen-bond pred=(SD-PRESS) donor and acceptor CoMSIA fields were derived according to Klebe et al. [10]. A distance-dependent Gaussian type Where SD is the sum of squared deviations between the functional form was used. The default value of 0.3 was biological activity of the test set and the mean activity of used as the attenuation factor. Similar to CoMFA, a data training-set molecules, and PRESS is the sum of squared Fig. 1 (a)Overview of the entameric structure of the ini- D C tial model. In this presentation each monomer has a different clockwise with A-B. B-C. C-D D-E and E-A forming the plus Details of ligand binding site. especially the residues approved to be important for ligand B entation), are shown in theA E and B interface of the model

by PROCHECK and Verify_3D (http://nihserver.mbi.ucla. edu/SAVS/, [16]). The 16 setrons, compounds 2 and 24 were retrieved from the MDDR database from MDL (http://www.mdli.com). The 2D-structures were subsequently converted into 3D-structures with CORINA (http://www2.ccc.uni-erlangen.de/ software/corina/free_struct.html). All the other 23 compounds were constructed based on the structure of compound 2. All molecules were set in their unprotonated state and Gasteiger–Hückel charges were added in SYBYL. Ligand docking The binding site of the 5-HT3 receptor was defined as atoms within a radius of 16 Å of the Ca atom of Trp178 in the binding pocket to ensure that most of the residues critical for ligand binding verified/revealed by previous experimental data were included. All molecules were docked into the binding pocket with the program GOLD v2.2 [17–19]. The default settings of GOLD were used, and no flipping was allowed. CoMFA and CoMSIA First, the docked conformations of the molecules in the training set were placed into a 3D cubic lattice with 2 Å grid. CoMFA fields were generated using an sp3 carbon probe atom carrying +1 charge to generate steric (LennardJones potential) and electrostatic (Coulomb potential) fields at each grid point. The CoMFA fields were scaled by the CoMFA-Standard method in SYBYL. A 30 kcal mol−1 energy cutoff was applied. The steric, electrostatic, hydrophobic, hydrogen-bond donor and acceptor CoMSIA fields were derived according to Klebe et al. [10]. A distance-dependent Gaussian type functional form was used. The default value of 0.3 was used as the attenuation factor. Similar to CoMFA, a data table was constructed from similarity indices calculated at the intersections of a regularly spaced lattice (2 Å grid) in CoMSIA. PLS analysis and validation of QSAR models The CoMFA/CoMSIA fields combined with observed biological activities (pKi) were included in a molecular spreadsheet, and partial least square (PLS) methods [20] were used to generate 3D-QSAR models. To check the statistical significance of the models, cross-validations were done to choose the optimum number of components (N) by means of the leave-one-out (LOO) [21] procedure using the enhanced version of PLS, the SAMPLS method [22], subsequently used to derive the final QSAR models. The optimal numbers of components were selected on the basis of the highest cross-validated correlation coefficient r2 cv , which is defined as follows: r 2 cv ¼ P Ypredicted Yactual 2 Σð Þ Yactual Ymean 2 ð1Þ Where Ypredicted, Yactual, and Ymean are predicted, actual, and mean values of the target property (pKi), respectively. The non-cross-validated PLS analyses were performed with a column filter value of 2.0. The CoMFA/CoMSIA results were interpreted graphically by field contribution maps using the ‘STDEV×COEFF’ field type. To assess the predictive power of the 3D-QSAR models derived using the training set, biological activities of the test set molecules were predicted. The predictive r2 r2 pred value is calculated as follows: r 2 pred ¼ ð Þ SD PRESS =SD Where SD is the sum of squared deviations between the biological activity of the test set and the mean activity of training-set molecules, and PRESS is the sum of squared Fig. 1 (a) Overview of the pentameric structure of the initial model. In this presentation each monomer has a different color. Subunits are labeled anticlockwise, with A-B, B-C, C-D, D-E and E-A forming the plus and minus interface side. (b) Details of ligand binding site, especially the residues approved to be important for ligands binding (ball-and-stick representation), are shown in the A and B interface of the model 124 J Mol Model (2007) 13:121–131

J Mol Model(2007)13:121-131 Table 1 5-HT3 receptor antag between aromatic side-chains of the receptor (Trp178- onists selected for docking Group A Group B Tyr229, Tyr138-Tyr148); while the basic centers interact with Glu231 or Glul24(ionic interaction), and/or Trp85 Bemesetron (cation-7 interaction) of the receptor [9] Dolasetron Granisetron In our study, the docking conformations with highest Indisetror score of setrons (one per setron) fell broadly into two Galdansetron Palonosetron groups, which we have designated A and B (Table 1). Ramosetron However, the observations fitted the result recently pub- Tropisetron Zatosetron lished by Lummis et al. [24]. In group A, the azabicyclic g of setrons was located between Trp178 and Tyr229 and the aromatic rings lay near Phe221. In group B, the deviations between the actual and the predicted activities of orientation of setrons was reversed, and consequently the romatic rings was located between Trp178 and Tyr229 of components, the conventional correlation coefficient r2 and the azabicyclic ring lay near Phe221. Representatives and its standard error were also computed for each model. from each group are shown in Fig. 2 The final model was validated by PROCHECK and erify 3D. The results are shown in Fig. 3. 75.9% of the Results and discussion residues were located in the darkest core'regions(marked A, B, and L) in the Ramachandran Plot, which fitted the Sequence identity of the extracellular region of human majority of PDB (72.9% in most favored regions for 2.7 5-HT3 receptor with AChBP is about 19%. When the X-ray structures)[25]. As for the Verify_3D results, conservative replacements are considered, their sequence 71.70% of the final model residues had an averaged 3D- homology is beyond 30%, which could result in at least ID score >0.2, hile the initial model showed 62.26% of 80% identity with the secondary structure of AChBP [9]. the residues had an averaged 3D-lD score >0. 2. Above Homology modeling resulted in a B-sandwich structure the model could be accepted for further studies. ( Fig. 1)similar to that of AChBP. Twenty-five molecules extracted from Nagel et al.were docked into the binding pocket of the final model Ligand docking and validation of the model Conformations (one per compound) were selected manually in 3D-QSAR analysis, considering both the docking score Sixteen selective 5-HT3 receptor antagonists were docked and the conformation reliability(Fig 4). The superposition into the binding pocket of the initial model. The setrons showed reasonable fit to the binding pocket consisting of eported to date may be expressed with such a pharmaco- residues that had been proven to be critical for ligand phore: a carbonyl-containing side chain flanked by a binding. The imidazole ring of most ligands seemed to form lipophilic aromatic group and a nitrogenous basic moiety 7-T interactions with Trp85, Trpl78 and Tyr229 of the [23]. As mentioned above, a similar study using a range of receptor. The nh moiety of the imidazole ring of most antagonists was published, which suggested that the ligands donated hydrogen bonds to Tyr148 and Trp178 of aromatic groups of antagonists were supposed to intercalate the receptor. However, these observations differed from Fig. 2(a) The docked confor mation of Tropisetron together the bindin TYR148 5-HT3Al TYR148 TRP178 RPI case of setrons in NTYR229 TYR229 conformation of granisetron TRP85 together with the binding site of human 5-HT3A receptor, as is ASN123 ASNI23 the case of setrons in Group B GLU231 The ligand is shown in orange (All hydrogen atoms were omit ted for a better view) GLU124 PHE221 GLU124 PHE221 a)

deviations between the actual and the predicted activities of the test set molecules. In addition, the r2 cv, r2 pred and number of components, the conventional correlation coefficient r2 and its standard error were also computed for each model. Results and discussion Sequence identity of the extracellular region of human 5-HT3 receptor with AChBP is about 19%. When the conservative replacements are considered, their sequence homology is beyond 30%, which could result in at least 80% identity with the secondary structure of AChBP [9]. Homology modeling resulted in a β-sandwich structure (Fig. 1) similar to that of AChBP. Ligand docking and validation of the model Sixteen selective 5-HT3 receptor antagonists were docked into the binding pocket of the initial model. The setrons reported to date may be expressed with such a pharmacophore: a carbonyl-containing side chain flanked by a lipophilic aromatic group and a nitrogenous basic moiety [23]. As mentioned above, a similar study using a range of antagonists was published, which suggested that the aromatic groups of antagonists were supposed to intercalate between aromatic side-chains of the receptor (Trp178– Tyr229, Tyr138–Tyr148); while the basic centers interact with Glu231 or Glu124 (ionic interaction), and/or Trp85 (cation–π interaction) of the receptor [9]. In our study, the docking conformations with highest score of setrons (one per setron) fell broadly into two groups, which we have designated A and B (Table 1). However, the observations fitted the result recently published by Lummis et al. [24]. In group A, the azabicyclic ring of setrons was located between Trp178 and Tyr229, and the aromatic rings lay near Phe221. In group B, the orientation of setrons was reversed, and consequently the aromatic rings was located between Trp178 and Tyr229, and the azabicyclic ring lay near Phe221. Representatives from each group are shown in Fig. 2. The final model was validated by PROCHECK and Verify_3D. The results are shown in Fig. 3. 75.9% of the residues were located in the darkest ‘core’ regions (marked A, B, and L) in the Ramachandran Plot, which fitted the majority of PDB (72.9% in most favored regions for 2.7 Å X-ray structures) [25]. As for the Verify_3D results, 71.70% of the final model residues had an averaged 3D– 1D score >0.2, while the initial model showed 62.26% of the residues had an averaged 3D–1D score >0.2. Above all, the model could be accepted for further studies. Twenty-five molecules extracted from Nagel et al. were docked into the binding pocket of the final model. Conformations (one per compound) were selected manually in 3D-QSAR analysis, considering both the docking score and the conformation reliability (Fig. 4). The superposition showed reasonable fit to the binding pocket consisting of residues that had been proven to be critical for ligand binding. The imidazole ring of most ligands seemed to form π–π interactions with Trp85, Trp178 and Tyr229 of the receptor. The NH moiety of the imidazole ring of most ligands donated hydrogen bonds to Tyr148 and Trp178 of the receptor. However, these observations differed from Fig. 2 (a) The docked conformation of Tropisetron together with the binding site of human 5-HT3A receptor, as is the case of setrons in Group A, which was above-mentioned in the paper. (b) The docked conformation of Granisetron together with the binding site of human 5-HT3A receptor, as is the case of setrons in Group B. The ligand is shown in orange (ball-and-stick representation). (All hydrogen atoms were omitted for a better view) Group A Group B Alosetron Azasetron Cilansetron Bemesetron Dolasetron Granisetron Fabesetron Indisetron Galdansetron Palonosetron Lurosetron Ramosetron Ondansetron Ricasetron Tropisetron Zatosetron Table 1 5-HT3 receptor antagonists selected for docking studies J Mol Model (2007) 13:121–131 125

J Mol model(2007)13:121-131 Ramachandran plot 13 As点 180-135 in most favored regions IA.BL Fig. 4 The docked conformations of 20 molecules in the training set ed regions I-a, b, -l pl together with the binding site of human 5-HT3A receptor. The solvent accessible surface is indicated as a solid surface. (All hydrogen atoms were omitted. Iced an r2 of 0.986. These statistical indice at 2.0Azetrum reasonably high, indicating that the CoMFA model might have a strong predictive ability. The predicted affinities and the residual values of the training set and the test set are given in Table 3 The steric field descriptor explained 62.2% of the variance, and the proportion of electrostatic descriptor accounted for 37.8%. Therefore, the steric field had greater than the electrostatic field. The CoMfa contours were mapped into the ligand-receptor complex structure fig, 3 arThe Ramachandran plot of the final model produced by Table 2 Summary of results from the CoMFA and CoMSIA analysis resulted from Verify 3D. The vertical axis gives the average 3D-ID score for residues. The conformation of the residues which have got a bad score needs to be adjusted COMFA 0.78560.9860.148156.21 0.55920.7050.60020.293 above-mentioned Groups A and B, an novel interaction mode between antagonists. On the other hand, the m might reveal a CoMSIA (S+E+H) 0.79640.9340.30352.909 receptor and CoMSIA(S+E+H+D) 0.739 5 0.936 0.308 40.960 of the docking CoMSIA (S+E+H+A) 0.761 5 0.939 0.301 42.975 studies and 3D-QSAR would offer constructive suggestions COMSIA (S+EtH+D 0.744 50.9340.312 39.761 to the further rectification of the receptor model S, E, H, D, and A represent the steric, electrostatic, hydrophobe CoMFA hydrogen bond donor and acceptor property fields, respectively Leave one out(LOO) cross-validation correlation coefficient The CoMFA result is summarized in Table 2. The cross- "Optimum number of components Non-cross-validation correlation coefficient validated value, r2v, is 0.785, with an optimum number of d Standard error of estimate components 6. The non-cross-validated PLS analysis F-test value

above-mentioned Groups A and B, and thus might reveal a novel interaction mode between 5-HT3 receptor and antagonists. On the other hand, the results of the docking studies and 3D-QSAR would offer constructive suggestions to the further rectification of the receptor model. CoMFA The CoMFA result is summarized in Table 2. The crossvalidated value, r2 cv, is 0.785, with an optimum number of components 6. The non-cross-validated PLS analysis produced an r2 of 0.986. These statistical indices were reasonably high, indicating that the CoMFA model might have a strong predictive ability. The predicted affinities and the residual values of the training set and the test set are given in Table 3. The steric field descriptor explained 62.2% of the variance, and the proportion of electrostatic descriptor accounted for 37.8%. Therefore, the steric field had greater influence than the electrostatic field. The CoMFA contours were mapped into the ligand-receptor complex structure (Fig. 5). Fig. 4 The docked conformations of 20 molecules in the training set together with the binding site of human 5-HT3A receptor. The solventaccessible surface is indicated as a solid surface. (All hydrogen atoms were omitted.) Table 2 Summary of results from the CoMFA and CoMSIA analysis r 2 cva Nb r 2c SEEd Fe CoMFA 0.785 6 0.986 0.148 156.212 CoMSIA (S+E) 0.559 2 0.705 0.600 20.293 CoMSIA (S+E+H) 0.796 4 0.934 0.303 52.909 CoMSIA (S+E+H+D) 0.739 5 0.936 0.308 40.960 CoMSIA (S+E+H+A) 0.761 5 0.939 0.301 42.975 CoMSIA (S+E+H+D +A) 0.744 5 0.934 0.312 39.761 S, E, H, D, and A represent the steric, electrostatic, hydrophobic, hydrogen bond donor and acceptor property fields, respectively a Leave one out (LOO) cross-validation correlation coefficient b Optimum number of components c Non-cross-validation correlation coefficient d Standard error of estimate e F-test value PROCHECK B A L b a l p ~p ~b ~a ~l b ~b b ~b ~b -180 -135 -90 -45 0 45 90 135 180 -135 -90 -45 0 45 90 135 180 Ramachandran Plot apdb Phi (degrees) Psi (degrees) LYS 57 (A) ASP 103 (A) ASP 160 (A) ASP 167 (A) MET 223 (A) ASP 54 (B) ASN 104 (B) LEU 159(A) LEU 159 (B) LYS 195 (B) Plot statistics Residues in most favored regions [A,B,L] 297 76.9% Residues in additional allowed regions [a,b,l,p] 73 18.9% Residues in generously allowed regions [~a,~b,~l,~p] 9 2.3% Residues in disallowed regions 7 1.8% ---- ------ Number of non-glycine and non-proline residues 386 100.0% Number of end-residues (excl. Gly and Pro) 0 Number of glycine residues (shown as triangles) 10 Number of proline residues 26 ---- Total number of residues 422 Based on an analysis of 118 structures of resolution of at least 2.0 Angstroms and R-factor no greater than 20%, a good quality model would be expected to have over 90% in the most favored regions. ILE 161 (A) ILE 161 (B) ASP 54 (A) ASP 167 (B) LYS 195 (A) -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 1 41 81 121 161 201 3D-1D Averaged Score Final model Initial model b) a) Fig. 3 (a) The Ramachandran plot of the final model produced by PROCHECK. (b) Profile window plots for the initial and final models resulted from Verify_3D. The vertical axis gives the average 3D-1D score for residues. The conformation of the residues which have got a bad score needs to be adjusted 126 J Mol Model (2007) 13:121–131

J Mol Model(2007)13:121-131 Table 3 Exs" training set and test set for CoMFA and increase the activity. This is why compound 18 as well as imental pKi, predicted pk, and residual values of substitution of electropositive group at this position COMSIA 19, whose OCH3 group was around the blue region, have Predicted pk Residual values low activity. The presence of a red contour near the N-1 Compd Actual pk, CoMFA CoMSIA CoMFA CoMSIA atom of the imidazole ring corresponding to Tp178 of the receptor emphasized that an electronegative group is desirable at this position. Compounds occupying both blue T23568 9.377 9.3739 and red contours showed moderate activities. such as 23 8.454 and 24 8.9018.781-0.189-0.069 9.318 0.029 CoMSIA 8.061 8.344 8.346 0.283 8.432 8.4338.617-0.001-0.185 The results of CoMSIA are shown in Table 2. The CoMSIA 123456 8.784 9034-0.085-0335 model including S, E and H performed better than the other 8.310 8.233 7967 0.343 field combinations. The CoMSIA model that included all 6.959 7.605-0.035 -0.646 five fields was chosen for further analysis, as it would 6.650 provide comprehensive information especially when the 7.469 7.377 0.092 0.236 receptor was concemed. The contributions were 10.3, 21.3 7. 022 6. 992 7.095 0.030-0.073 33.0, 20. 1 and 15.3% for S, E, H, D, and A, respectively 0.056-0.03 6.430 6.272 0.06 0.095 The calculated activities and the residual values predicted 9,000 89318.641 0.069 0.359 by the CoMSIa model for the training set and the test set 21 9.538 9.637 9.383 -0.099 0.155 are given in Table 3 9.39395630.116 The steric field contour plots are shown in Fig. 6a. There 7.848 7.784 7 819 0.064 0.029 is a major white region around the C-2, 3 positions of the 8.481 8.469-0.0760.01 phenyl ring and a minor yellow steric contour near the C-4 Test set 0050 0099 position, which indicates that a large group around the 147 7921 7871 7.822 0.074 C-2, 3 positions and a small group near the C-4 position 8.824 8.752 8.560 0.072 0.264 would be beneficial to the binding affinity. This is supporte 8.699 0.283 8.846-0.584 -0.147 by the relatively higher affinities of compound 2, 6, 9, 11 8.149 8.1858.321 0.036 nd lower affinities of 14, 15, and 16 than compound 3 The CoMSIa electrostatic field contour plots are shown in Fig. 6b. White areas, positive charges favorable to The steric contour map(Fig. 5a) showed a green region affinity, corresponding to Tyr229 of the receptor, spread surrounding the C-2, 3 position of the phenyl ring around the thiazole and imidazole ring Compounds 18 an (Scheme 3), indicating that a bulky substituent is preferred 19 oriented their atom-N into the disfavored regions, shown to produce higher activity, which was consistent with the in white, resulting in bad binding affinity. The small yellow experiments that hydroxylated thiazoles were generally less region near the C-2 position of the phenyl ring indicates active than their alkoxylated counterparts. However, it that electronegative groups at this position may enhance the seemed that this region collided slightly with Glu224 and activity. Therefore, the activity of 2 was about ten-fold Ser225 of the 5-HT3A receptor, which meant the size of higher than that of 8, in which the -OCH, group replaced substituent was not unlimited, otherwise it would encounter the -CH3 group the receptor. There was another green region near the C-5. The hydrophobic contours are shown in Fig. 6c. The position of the imidazole ring, which could explain why white hydrophobic contour near the C-2, 3 position of the compound I was more active than 23. The yellow steric phenyl ring matches the green steric contour(Fig 6a). This contour near the C-4 position of the phenyl ring indicated indicates that any bulky group with lipophilic character is that any bulky substituent decreased activity, which should preferred at this position. The result agrees with the account for the activity differences of 4-halogenated contours of the CoMSIA steric field. The less active 18 thiazoles: 13(F-8.310), 16(Cl- 7.469)and 15(Br-, 6.569). and 19 orient their OCH, into the white contoured region, The CoMFA electrostatic contour plot is shown in whereas the more active 8 occupies the favored area with its Fig. 5b. There were two minor blue regions near the C-2 methyl substituent. The yellow hydrophilic contour near the position of the phenyl ring and the C-5 position of the C-4 position indicates that any lipophilic group at this imidazole ring, the former of which was located between position decreases the activity. Thus, the activity of Tyr148 and Tyr229 of the 5-HT3A receptor, indicating that molecule with F(13)or Cl(16)at this position is higher

The steric contour map (Fig. 5a) showed a green region surrounding the C-2,3 position of the phenyl ring (Scheme 3), indicating that a bulky substituent is preferred to produce higher activity, which was consistent with the experiments that hydroxylated thiazoles were generally less active than their alkoxylated counterparts. However, it seemed that this region collided slightly with Glu224 and Ser225 of the 5-HT3A receptor, which meant the size of substituent was not unlimited, otherwise it would encounter the receptor. There was another green region near the C-5’ position of the imidazole ring, which could explain why compound 1 was more active than 23. The yellow steric contour near the C-4 position of the phenyl ring indicated that any bulky substituent decreased activity, which should account for the activity differences of 4-halogenated thiazoles: 13(F-, 8.310), 16(Cl-, 7.469) and 15(Br-, 6.569). The CoMFA electrostatic contour plot is shown in Fig. 5b. There were two minor blue regions near the C-2 position of the phenyl ring and the C-5’ position of the imidazole ring, the former of which was located between Tyr148 and Tyr229 of the 5-HT3A receptor, indicating that substitution of electropositive group at this position would increase the activity. This is why compound 18 as well as 19, whose OCH3 group was around the blue region, have low activity. The presence of a red contour near the N-1’ atom of the imidazole ring corresponding to Trp178 of the receptor emphasized that an electronegative group is desirable at this position. Compounds occupying both blue and red contours showed moderate activities, such as 23 and 24. CoMSIA The results of CoMSIA are shown in Table 2. The CoMSIA model including S, E and H performed better than the other field combinations. The CoMSIA model that included all five fields was chosen for further analysis, as it would provide comprehensive information especially when the receptor was concerned. The contributions were 10.3, 21.3, 33.0, 20.1 and 15.3% for S, E, H, D, and A, respectively. The calculated activities and the residual values predicted by the CoMSIA model for the training set and the test set are given in Table 3. The steric field contour plots are shown in Fig. 6a. There is a major white region around the C-2,3 positions of the phenyl ring and a minor yellow steric contour near the C-4 position, which indicates that a large group around the C-2,3 positions and a small group near the C-4 position would be beneficial to the binding affinity. This is supported by the relatively higher affinities of compound 2, 6, 9, 11 and lower affinities of 14, 15, and 16 than compound 3. The CoMSIA electrostatic field contour plots are shown in Fig. 6b. White areas, positive charges favorable to affinity, corresponding to Tyr229 of the receptor, spread around the thiazole and imidazole ring. Compounds 18 and 19 oriented their atom-N into the disfavored regions, shown in white, resulting in bad binding affinity. The small yellow region near the C-2 position of the phenyl ring indicates that electronegative groups at this position may enhance the activity. Therefore, the activity of 2 was about ten-fold higher than that of 8, in which the –OCH3 group replaced the –CH3 group. The hydrophobic contours are shown in Fig. 6c. The white hydrophobic contour near the C-2,3 position of the phenyl ring matches the green steric contour (Fig. 6a). This indicates that any bulky group with lipophilic character is preferred at this position. The result agrees with the contours of the CoMSIA steric field. The less active 18 and 19 orient their OCH3 into the white contoured region, whereas the more active 8 occupies the favored area with its methyl substituent. The yellow hydrophilic contour near the C-4 position indicates that any lipophilic group at this position decreases the activity. Thus, the activity of molecule with F (13) or Cl (16) at this position is higher Table 3 Experimental pKi, predicted pKi and residual values of molecules used in the training set and test set for CoMFA and CoMSIA Predicted pKi Residual values Compd Actual pKi CoMFA CoMSIA CoMFA CoMSIA Training set 2 9.377 9.373 9.270 0.004 0.107 3 8.745 8.454 8.353 0.291 0.392 5 8.712 8.901 8.781 −0.189 −0.069 6 9.022 8.993 9.318 0.029 −0.296 8 8.061 8.344 8.346 −0.283 −0.285 9 9.056 8.974 8.725 0.082 0.331 11 8.432 8.433 8.617 −0.001 −0.185 12 8.699 8.784 9.034 −0.085 −0.335 13 8.310 8.233 7.967 0.077 0.343 14 6.959 6.994 7.605 −0.035 −0.646 15 6.569 6.650 6.654 −0.081 −0.085 16 7.469 7.377 7.233 0.092 0.236 17 7.022 6.992 7.095 0.030 −0.073 18 6.398 6.342 6.430 0.056 −0.032 19 6.367 6.430 6.272 −0.063 0.095 20 9.000 8.931 8.641 0.069 0.359 21 9.538 9.637 9.383 −0.099 0.155 22 9.509 9.393 9.563 0.116 −0.054 23 7.848 7.784 7.819 0.064 0.029 24 8.481 8.557 8.469 −0.076 0.012 Test set 1 7.921 7.871 7.822 0.050 0.099 4 9.000 8.926 8.809 0.074 0.191 7 8.824 8.752 8.560 0.072 0.264 10 8.699 9.283 8.846 −0.584 −0.147 25 8.149 8.185 8.321 −0.036 −0.172 J Mol Model (2007) 13:121–131 127

J Mol model(2007)13:121-131 TYR148 TYR148 TRP85 TRP178 TRP178 TRP85 GLU224 YASN123) SER177 GLU224 ASN123 SER177 THR176 7THRI76 TYR229 SER222 SER222 TYR229 PHE2 GLU231 PHE221 GLU231 b) Fig 5 Mapping the CoMFA contours in the active site of 5-HT yellow contours indicate receptor with compound 2 as an example. The hydrogen bonds are and (b)electrostatic fields: blue contour shown in broken red lines. The ligand is shown in pra fields; green indicate regions where electronegative groups increase activity.(All electropositive groups increase activity, w contours indicate regions where bulky groups increase activity, ydrogen atoms of the receptor were omitted) than that of the molecule with a CH3(14)group. There As for the receptor, the white contours should corre were several molecules for which the activities could not spond to the hydrophobic residues, whereas the yellow ones be interpreted by the hydrophobic contours alone. For should be near the polar residues. Actually, the major white example, the affinities of compounds 9, 1l and 12 should as well as the yellow contour are opposite Glu224 and be bad due to the halogen substituent near the white Ser225, while the minor white contour faces Trp178. We contoured area. In fact, their activities are moderately high. suggest that the position of Glu224 and Ser 225 might be All of the five fields' properties may be needed to describe rectified The hydrogen-bond donor and acceptor contours are colored contour near the NH group of the imidazole N2 indicates that a corresponding hydrogen bond acceptor on the 5-HT3A receptor might exist. This is consistent with the I fact that most ligands form hydrogen bonds with Tyrl4 nh their nh group (2 and 22) revealed high affinities, whereas those orienting their donor into the disfavored regions(18 N R possessing both cyan and purple contours were moderately active. such as 23 and 24 magenta region ns near and O implied that hydrogen-bond donors may exist at the corresponding R positions of the receptor's active site (Trp85, Tyr148 and Tyr229), consistent with ou observation clearly indicates that a hydrogen-bond accep- Scheme 3 Template for thiazoles, in which Ar is a phenyl substituent. tor near the magenta contours would increase the activity. Note that the atom numbering does not follow the IUPAC rules The Nh group and n of the imidazole ring at these

than that of the molecule with a CH3 (14) group. There were several molecules for which the activities could not be interpreted by the hydrophobic contours alone. For example, the affinities of compounds 9, 11 and 12 should be bad due to the halogen substituent near the white contoured area. In fact, their activities are moderately high. All of the five fields’ properties may be needed to describe the case. As for the receptor, the white contours should correspond to the hydrophobic residues, whereas the yellow ones should be near the polar residues. Actually, the major white as well as the yellow contour are opposite Glu224 and Ser225, while the minor white contour faces Trp178. We suggest that the position of Glu224 and Ser225 might be rectified. The hydrogen-bond donor and acceptor contours are shown in Fig. 7a,b, respectively. The presence of cyancolored contour near the NH group of the imidazole ring indicates that a corresponding hydrogen bond acceptor on the 5-HT3A receptor might exist. This is consistent with the fact that most ligands form hydrogen bonds with Tyr148 and Trp178. Molecules occupying the cyan contour with their NH group (2 and 22) revealed high affinities, whereas those orienting their donor into the disfavored regions (18 and 19) showed low affinities. Furthermore, molecules possessing both cyan and purple contours were moderately active, such as 23 and 24. Two magenta regions near N and O implied that hydrogen-bond donors may exist at the corresponding positions of the receptor’s active site (Trp85, Tyr148 and Tyr229), consistent with our homology model. This observation clearly indicates that a hydrogen-bond acceptor near the magenta contours would increase the activity. The NH group and N atom of the imidazole ring at these Fig. 5 Mapping the CoMFA contours in the active site of 5-HT3A receptor with compound 2 as an example. The hydrogen bonds are shown in broken red lines. The ligand is shown in orange. The contour plots (STDEV*COEFF) of the CoMFA: (a) steric fields; green contours indicate regions where bulky groups increase activity, whereas yellow contours indicate regions bulky groups decrease activity and (b) electrostatic fields; blue contours indicate regions where electropositive groups increase activity, whereas red contours indicate regions where electronegative groups increase activity. (All hydrogen atoms of the receptor were omitted) N S NH N R 1 2 3 4 1' 2' 3' 4' 5' R Scheme 3 Template for thiazoles, in which Ar is a phenyl substituent. Note that the atom numbering does not follow the IUPAC rules 128 J Mol Model (2007) 13:121–131

J Mol model(2007)13:121-131 10 COMFA Figure 8 shows the plots of actual versus predicted activity for both training set and test set. In almost all cases 9.5 of 3D-QSAR models, the predicted values fall close to the observed pKi values, deviating by not more than 0.5 logarithmic units except for compound 10. Finally, CoMFA and CoMSIA possessed distinct predictive power with respect to these five compounds(0582 vs 0.804) ◆ Training set r In this study, CoMFA and CoMSIA 3D-QSAR analy were carried out on the docked conformations of thiazoles as 5-HT3 receptor antagonists with the ori intention to validate the 3D-model of the extracellular 6.0 6.06.57075808.59095100 domain of the human 5-HT3A receptor. The QSAR models a Actual pKi may have good prediction capability in terms of y and r2 values. The well-correlated 3D-QSAR models showed that COMSIA it is reliable to construct the models on the docking conformations, which are close to pharmacophoric con- formations. The effects of the steric, electrostatic, hydro- phobic, and H-bond donor and acceptor fields around the docked conformations on their affinities were discussed in detail. According to these studies, some implications could .Training se be drawn to improve the activity and selectivity of thiazoles as 5-HT3 receptor antagonists, bulky hydrophobic substituent at the C-2, 3 position and small hydrophilic group at the C-4 position of the phenyl ring were preferred to produce higher activity. Besides, the eplacement of C-4 atom by N atom would result in comparative affinities theoretically. The results of 3D- QSAR were also consistent with the docking studies, which provided a tool to predict the affinity of related thiazoles, 606.570758.08.5909.510.0 b) and to guide further structural modification and synthesis of Actual pKi new potent and selective 5-HT3 receptor antagonists Actual versus predicted pKi of training set molecule and(b)CoMSIA 3D-QSAR The predicted Syntheses of new derivatives designed based on the 3D to the observed pKi values, by not more QSAR results are in progre References Validation of 3D-QSAR models I. Wolf H(2000) Scand J Rheumatol 29: 37-4 The predictive power of the CoMFA and CoMSIA 3D- 2. Haus U, Spath M, Farber L (2004)Scand J Rheumatol 33: 12-18 QSAR models was evaluated further by five additional 3. Karim F, Roerig SC, Saphier D(1996) Biochem Pharmacol 52:685-692 molecules in the test set. The results showed that a 4. Israili ZH (2001) Curr Med Chem--Central Nervous System moderately bulky hydrophobic substituent (i.e. OCH3, Agents 1: 171-199 5. Lummis SCR, Reeves DC(2002)Mol Membrane Biol 19: 11-26 OCH2CH3, and NHCH3 )at the C-2, 3 positions and a small 6 Boess FG,Beroukhim R,Martin IL(1995)JNeurochem hydrophilic group (i.e. OH, NH2)at the C-4 position of the 64:1401-1405 phenyl ring were preferred to produce higher binding 7. Boess FG, Steward LJ, Steele JA, Liu D, Reid J, Glencorse TA. affinities. The activities of corresponding pyridyl deriva- Martin IL (1997) Neuropharmacol 36: 637-647 tives, in which an N atom replaced C-4, were also 8 Brejc K, Van Dijk WI, Klaassen RV,Schuurmans M, van der Oost J, Smit AB, Sixma TK(2001) Nature 411: 269-276 heoretically high. The set of compounds is worthy of 9. Aksay G, Bikadi z, Simonyi MJ(2003)Recept Signal Transduct further studies Res23:255-270

Validation of 3D-QSAR models The predictive power of the CoMFA and CoMSIA 3DQSAR models was evaluated further by five additional molecules in the test set. The results showed that a moderately bulky hydrophobic substituent (i.e. OCH3, OCH2CH3, and NHCH3) at the C-2,3 positions and a small hydrophilic group (i.e. OH, NH2) at the C-4 position of the phenyl ring were preferred to produce higher binding affinities. The activities of corresponding pyridyl derivatives, in which an N atom replaced C-4, were also theoretically high. The set of compounds is worthy of further studies. Figure 8 shows the plots of actual versus predicted activity for both training set and test set. In almost all cases of 3D-QSAR models, the predicted values fall close to the observed pKi values, deviating by not more than 0.5 logarithmic units except for compound 10. Finally, CoMFA and CoMSIA possessed distinct predictive power with respect to these five compounds (0.582 vs. 0.804). Summary In this study, CoMFA and CoMSIA 3D-QSAR analyses were carried out on the docked conformations of 25 thiazoles as 5-HT3 receptor antagonists with the original intention to validate the 3D-model of the extracellular domain of the human 5-HT3A receptor. The QSAR models may have good prediction capability in terms of r2 cv and r2 values. The well-correlated 3D-QSAR models showed that it is reliable to construct the models on the docking conformations, which are close to pharmacophoric conformations. The effects of the steric, electrostatic, hydrophobic, and H-bond donor and acceptor fields around the docked conformations on their affinities were discussed in detail. According to these studies, some implications could be drawn to improve the activity and selectivity of thiazoles as 5-HT3 receptor antagonists, for example, moderately bulky hydrophobic substituent at the C-2,3 position and small hydrophilic group at the C-4 position of the phenyl ring were preferred to produce higher activity. Besides, the replacement of C-4 atom by N atom would result in comparative affinities theoretically. The results of 3DQSAR were also consistent with the docking studies, which provided a tool to predict the affinity of related thiazoles, and to guide further structural modification and synthesis of new potent and selective 5-HT3 receptor antagonists. Syntheses of new derivatives designed based on the 3DQSAR results are in progress. References 1. Wolf H (2000) Scand J Rheumatol 29:37–45 2. Haus U, Spath M, Farber L (2004) Scand J Rheumatol 33:12–18 3. Karim F, Roerig SC, Saphier D (1996) Biochem Pharmacol 52:685–692 4. Israili ZH (2001) Curr Med Chem—Central Nervous System Agents 1:171–199 5. Lummis SCR, Reeves DC (2002) Mol Membrane Biol 19:11–26 6. Boess FG, Beroukhim R, Martin IL (1995) J Neurochem 64:1401–1405 7. Boess FG, Steward LJ, Steele JA, Liu D, Reid J, Glencorse TA, Martin IL (1997) Neuropharmacol 36:637–647 8. Brejc K, Van Dijk WJ, Klaassen RV, Schuurmans M, van der Oost J, Smit AB, Sixma TK (2001) Nature 411:269–276 9. Maksay G, Bikádi Z, Simonyi MJ (2003) Recept Signal Transduct Res 23:255–270 CoMFA 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 Actual pKi Predicted pKi Training Set Test Set CoMSIA 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 Actual pKi Predicted pKi Training Set Test Set a) b) Fig. 8 Actual versus predicted pKi of training and test set molecules for (a) CoMFA and (b) CoMSIA 3D-QSAR models. The predicted values fell close to the observed pKi values, deviating by not more than 0.5 logarithmic units mostly 130 J Mol Model (2007) 13:121–131

点击进入文档下载页（PDF格式）

共11页，试读已结束，阅读完整版请下载

点击下载（PDF格式）

浏览记录

复旦大学：《药物设计学》课程教学资源（教学研究）基于配体结构的药物设计_Structure-based 3D-QSAR studies on thiazoles as 5-HT3 receptor antagonists