山国劇蔹论文在线 http://www.paper.edu.cn Modeling the Interaction between Glycogen Synthase Kinase 3R(GsK-3B)and Its Non-ATP Competitive Inhibitors Department of Medicinal Chemistry, School of Pharmacy, Fudan University, Shanghai 201203, China The Genomics Research Center; Academia Sinica, 128 Academia Road, Section 2, Nankang, Taipei Department of Medicinal Chemistry, School of pharmacy, Shandong University, Jinan, Shandong 250012,Chia E-mail:cyl10@fudan.edu.cn(YChu),mli@sdu.edu.cn(M.Li) Abstract: Glycogen synthase kinase-3(GSK-3) plays an important role in a diverse number of regulatory pathways. GSK-3 inhibitors, particularly the non-ATP-competitive inhibitors, have been evaluated as promising drug candidates for a lot of unmet pathologies, such as Alzheimer's disease and diabetes. In this paper, a molecular docking study with the published GsK-3B crystal structure and receptor-based pharmacophore modeling of four highly active non-ATP-competitive GSK-3 inhibitors were performed by dOCK 5.4 and Catalyst 4.11, respectively. The results could provide an exquisite understanding on their mechanism of interaction within the non-ATP-binding pocket of GSK-3B, meanwhile the finding of the common properties shared by these pharmacological inhibitors of GsK-3B could be helpful to urther chemical optimization of such potent drug candidates Keywords: GSK-3B, non-ATP-competitive inhibitor, molecular dock, pharmacophore model Introduction In recent years, Glycogen Synthase Kinase-3(GSK-3) has been the focus of extensive medicinal hemistry efforts including for insulin resistance, Alzheimers disease(AD), stroke and bipolar disorders. Many GSK-3B inhibitors have been reported and reviewed in the literatures", such as maleimides, indirubins, paullones and hymenimaldisine derivatives. These compounds can decrease GSK-3 a thus to restore insulin responses including glucose uptake and glycogen synthesis in insulin signaling pathway, as well as decrease neurodegenerative markers and behaviora deficits in AD pathogenesis by reducing the production of AB, therefore reducing AB-induced neuronal cell death. Thus GsK-3 inhibitors clearly have clinical potential for the treatment of diabetes and AD Unfortunately, most of available inhibitors are bound to the ATP-binding pocket of GSK-3 Considering the aTP binding pocket is highly conserved in protein kinase, non-specific protein kinase inhibition by ATP site-directed inhibitors might have widespread effects. As a matter of fact, such GSK-3 inhibitors always showed many others kinases cross-activities though high selectivity of few were reported. This is the case of the great majority of GSK-3 inhibitors discovered until now diminishing their drug development possibilities I Financiall supported from Specialized Research Fund for the Doctoral Program of Higher Education(SRFDP No. 20070246089) Chinese Ministry of Education
1 Financiall supported from Specialized Research Fund for the Doctoral Program of Higher Education (SRFDP, No. 20070246089), Chinese Ministry of Education. - 1 - Modeling the Interaction between Glycogen Synthase Kinase 3β (GSK-3β) and Its Non-ATP Competitive Inhibitors Yong Chu 1 , Keng-Chang Tsai 2 , Deyong Ye 1 , Minyong Li 3,* 1 Department of Medicinal Chemistry, School of Pharmacy, Fudan University, Shanghai 201203, China 2 The Genomics Research Center, Academia Sinica, 128 Academia Road, Section 2, Nankang, Taipei 115 3 Department of Medicinal Chemistry, School of Pharmacy, Shandong University, Jinan, Shandong 250012, China E-mail: cy110@fudan.edu.cn (Y. Chu), * mli@sdu.edu.cn (M. Li) Abstract: Glycogen synthase kinase-3 (GSK-3) plays an important role in a diverse number of regulatory pathways. GSK-3 inhibitors, particularly the non-ATP-competitive inhibitors, have been evaluated as promising drug candidates for a lot of unmet pathologies, such as Alzheimer’s disease and diabetes. In this paper, a molecular docking study with the published GSK-3β crystal structure and receptor-based pharmacophore modeling of four highly active non-ATP-competitive GSK-3 inhibitors were performed by DOCK 5.4 and Catalyst 4.11, respectively. The results could provide an exquisite understanding on their mechanism of interaction within the non-ATP-binding pocket of GSK-3β, meanwhile the finding of the common properties shared by these pharmacological inhibitors of GSK-3β could be helpful to further chemical optimization of such potent drug candidates. Keywords: GSK-3β, non-ATP-competitive inhibitor, molecular dock, pharmacophore model Introduction In recent years, Glycogen Synthase Kinase-3 (GSK-3) has been the focus of extensive medicinal chemistry efforts including for insulin resistance 1-2, Alzheimer’s disease (AD)3-5, stroke and bipolar disorders6 . Many GSK-3β inhibitors have been reported and reviewed in the literatures7-8, such as maleimides, indirubins, paullones and hymenimaldisine derivatives. These compounds can decrease GSK-3 activity, thus to restore insulin responses including glucose uptake and glycogen synthesis in insulin signaling pathway9-11, as well as decrease neurodegenerative markers and behaviora deficits in AD pathogenesis12 by reducing the production of Aβ13, therefore reducing Aβ-induced neuronal cell death14. Thus GSK-3 inhibitors clearly have clinical potential for the treatment of diabetes and AD. Unfortunately, most of available inhibitors are bound to the ATP-binding pocket of GSK-3. Considering the ATP binding pocket is highly conserved in protein kinase, non-specific protein kinase inhibition by ATP site-directed inhibitors might have widespread effects. As a matter of fact, such GSK-3 inhibitors always showed many others kinases cross-activities though high selectivity of few were reported. This is the case of the great majority of GSK-3 inhibitors discovered until now, which is diminishing their drug development possibilities. 中国科技论文在线 http://www.paper.edu.cn
山国武论义在统 http://www edu.cn However, non-ATP-competitive GSK-3 selective inhibitors represent a more efficient pathway for providing real promising drugs for therapeutic intervention. Thiadiazolidinones (TDZDs)and halomethylarylketones(HMKs)were reported as first two families of non-ATP competitive GSK-3 inhibitors and both of them really do not show inhibition on others several kinases as PKA, PKC, CK-2 nd CDK1cyclin B. The privileged scaffold of TDZDs for the selective inhibition of GSK-3 has been revealed based on an extensive SAR study, and two binding modes were then put forward by mapping studies. This information in turn guided an optimization toward the inhibitory activity of TDZD Herein four highly active TDZD and HMK inhibitors 1-4 as depicted in Figure 1 were selected based on structure diversity and a docking study was performed with the published GsK-3B crystal structure(PDB code: 1Q3D)to provide an exquisite understanding of their mechanism of interaction within the non-ATP-binding pocket. The finding of the common properties shared by these pharmacological inhibitors of GSK-3B would be helpful to further optimize these potential drug candidates C O 2 Figure 1. The chemical structures of TDzD and hmK inhibitors 1-4 Materials and methods Molecular docking and structural optimization structures for these four inhibitors were refined using the pm3 method in the mopac 7 and assigned with AMl-BCC partial charges by the QuACPAC program. All partial charges on the atoms of the homology model were derived from AMBEr parameters Docking of the ligands into GSK-3B was performed by using DOCK 5.4. The active site included residues Arg 92 Arg96, Arg180, Lys205 and Tyr216 as recommended by literatures. 5, I7 After docking, MD simulations were carried out by using the CHARMM c33bl program- and a GBSW implicit solvation model- following similar procedures we reported elsewhere. The protein atoms were parameterized by CHARMM-GUl using the CHARMM22 force field.The surface tension coefficient(representing the non-polar solvation energy) was set to 0.03 kcal/mol A),which was consistent with literature precedents in the calculation of non-polar contributions in soluble proteins. 26 All bond lengths involving hydrogen atoms were fixed using the SHAKE algorithm 30.No cutoff for the non-bonded and GB energy calculations was used. In the simulation, temperature was at 300 K. Minimizations were carried out using 1500 steps of steepest descent, followed by Adopted Basis Newton-Raphson(ABNR) minimization until the root mean square gradient was less than 0.001 kcal/mol A. The whole system was then equilibrated for 50 ps, followed by another 10 ns of canonical ensemble (NVT)-MD simulation run. Finally, the ligand-receptor complexes were analyzed by
- 2 - However, non-ATP-competitive GSK-3 selective inhibitors represent a more efficient pathway for providing real promising drugs for therapeutic intervention. Thiadiazolidinones (TDZDs) and halomethylarylketones (HMKs) were reported as first two families of non-ATP competitive GSK-3 inhibitors and both of them really do not show inhibition on others several kinases as PKA, PKC, CK-2 and CDK1/cyclin B15-16. The privileged scaffold of TDZDs for the selective inhibition of GSK-3 has been revealed based on an extensive SAR study, and two binding modes were then put forward by mapping studies17-18. This information in turn guided an optimization toward the inhibitory activity of TDZDs. Herein four highly active TDZD and HMK inhibitors 1-4 as depicted in Figure 1 were selected based on structure diversity and a docking study was performed with the published GSK-3β crystal structure (PDB code: 1Q3D)19 to provide an exquisite understanding of their mechanism of interaction within the non-ATP-binding pocket. The finding of the common properties shared by these pharmacological inhibitors of GSK-3β would be helpful to further optimize these potential drug candidates. S N N O CH3 N S H S N O O N S N O O CH3 S Cl O Br Br 1 2 3 4 Figure 1. The chemical structures of TDZD and HMK inhibitors 1-4 Materials and Methods Molecular docking and structural optimization The 3D structures for these four inhibitors were refined using the PM3 method in the MOPAC 7 program 20 and assigned with AM1-BCC partial charges 21-23 by the QuACPAC program. All partial charges on the atoms of the homology model were derived from AMBER 8 parameters. Docking of the ligands into GSK-3β was performed by using DOCK 5.4 24. The active site included residues Arg92, Arg96, Arg180, Lys205 and Tyr216 as recommended by literatures.15, 17 After docking, MD simulations were carried out by using the CHARMM c33b1 program25 and a GBSW implicit solvation model26 following similar procedures we reported elsewhere.27 The protein atoms were parameterized by CHARMM-GUI28 using the CHARMM22 force field 29. The surface tension coefficient (representing the non-polar solvation energy) was set to 0.03 kcal/ (mol·Å2 ), which was consistent with literature precedents in the calculation of non-polar contributions in soluble proteins. 26 All bond lengths involving hydrogen atoms were fixed using the SHAKE algorithm 30. No cutoff for the non-bonded and GB energy calculations was used. In the simulation, temperature was at 300 K. Minimizations were carried out using 1500 steps of steepest descent, followed by Adopted Basis Newton-Raphson (ABNR) minimization until the root mean square gradient was less than 0.001 kcal/mol Å. The whole system was then equilibrated for 50 ps, followed by another 10 ns of canonical ensemble (NVT)-MD simulation run. Finally, the ligand-receptor complexes were analyzed by 中国科技论文在线 http://www.paper.edu.cn
山国到技论文在线 http://www.paper.edu.cn HBPLUS 3.06, LIGPLOT 4.223and Pymol 0.99 33 Docking-based pharmacophore modeling Based on the molecular docking and molecular dynamics results, an attempt to identify the hypothetical receptor-based pharmacophore model was made by using Hiphop algorithm implemented in Catalyst 4. 11 package. Such practice is the same as what we have used in the past for other modeling works. 4 +-33 In particular, HipHop algorithm finds common-feature pharmacophore models among a set of highly active compounds thus it carries out a 'qualitative model without the use of activity data, which represents the essential 3D arrangement of functional groups common to a set of molecules for interacting with a specific biological target In this hypothesis generation phase, the docking conformations of these four inhibitors were directly used as input coordinates without any structural minimization and conformational search. A default uncertainty factor of 3 for each compound was defined, and four chemical features, including hydrogen-bond acceptor(A), hydrogen-bond donor(D), aromatic ring(R)and hydrophobic(H) group, were selected to form the pharmacophore hypothesis using HipHop. A Principal number of 2 and MaxOmit Feat number of 0 was defined for the good mapping of all features of these compounds on a Hardware and sofhvare Molecular docking(doCk 5.4), binding analysis(HBPLUs 3.06 and ligplot 4.22)and visualization (PyMol 0.99) were carried out on a Linux workstation. MM calculations and MD simulations (CHARMM c33b1)were performed on URSA, a 160-processor computer based on the Power5+ processor and IBMs P series architecture. The pharmacophore modeling(Catalyst 4. 11)was executed on a SGI Origin 3800 workstation equipped with 48 x400 MHz MIPS R12000 processors Results and discussion The proposed binding conformation and schematic diagram of compound I are depicted in Figure 2 The symmetrical acyl groups herein are proposed to have multiple hydrogen bonding with Arg 96, Arg 180 and Lys 205, in the meanwhile the molecular scaffold of this compound is engaged in the hydrophobic interaction with Val 214 and Tyr 216. Moreover, the phenyl ring should have a T-stacking with Tyr 216
- 3 - HBPLUS 3.06 31, LIGPLOT 4.22 32 and Pymol 0.99 33. Docking-based pharmacophore modeling Based on the molecular docking and molecular dynamics results, an attempt to identify the hypothetical receptor-based pharmacophore model was made by using Hiphop algorithm implemented in Catalyst 4.11 package. Such practice is the same as what we have used in the past for other modeling works.34-35 In particular, HipHop algorithm finds common-feature pharmacophore models among a set of highly active compounds thus it carries out a ‘qualitative model’ without the use of activity data, which represents the essential 3D arrangement of functional groups common to a set of molecules for interacting with a specific biological target 36. In this hypothesis generation phase, the docking conformations of these four inhibitors were directly used as input coordinates without any structural minimization and conformational search. A default uncertainty factor of 3 for each compound was defined, and four chemical features, including hydrogen-bond acceptor (A), hydrogen-bond donor (D), aromatic ring (R) and hydrophobic (H) group, were selected to form the pharmacophore hypothesis using HipHop. A Principal number of 2 and MaxOmitFeat number of 0 was defined for the good mapping of all features of these compounds on a hypothesis model 37. Hardware and software Molecular docking (DOCK 5.4), binding analysis (HBPLUS 3.06 and Ligplot 4.22) and visualization (PyMol 0.99) were carried out on a Linux workstation. MM calculations and MD simulations (CHARMM c33b1) were performed on URSA, a 160-processor computer based on the Power5+ processor and IBM’s P series architecture. The pharmacophore modeling (Catalyst 4.11) was executed on a SGI Origin 3800 workstation equipped with 48×400 MHz MIPS R12000 processors. Results and Discussion The proposed binding conformation and schematic diagram of compound 1 are depicted in Figure 2. The symmetrical acyl groups herein are proposed to have multiple hydrogen bonding with Arg 96, Arg 180 and Lys 205, in the meanwhile the molecular scaffold of this compound is engaged in the hydrophobic interaction with Val 214 and Tyr 216. Moreover, the phenyl ring should have a π-stacking with Tyr 216. 中国科技论文在线 http://www.paper.edu.cn
山国到技论文在线 http://www.paper.edu.cn A B Figure 2. The simulated conformation(A)and interaction diagram(B)of compound I and GSK-3B For compound 2, it has similar binding pattern with compound 1 (Figure 3). The two ketone groups in compound 2 play roles of hydrogen bond acceptors for contacting with Arg 96, Arg 180 and Lys 205 The whole structure of this molecule is trapped into a hydrophobic pocket lined by gly 202, Ser 203 nd Tyr 216. It should be noted that the phenyl ring herein also contributes to a T-stacking with Tyr 216 A B Figure 3. The simulated conformation(A)and interaction diagram(B)of compound 2 and GsK-3B In the case of compound 3, the simulated results suggest that the ketone group forms three hydrogen bonds with Arg 180 and Lys 205, while the imine group engages in hydrogen binding with Arg 96 in GSK-3B(Figure 4). The computational results also propose that this compound inserts into a hydrophobic network clustered by Ser 203 and Tyr 216. In addition, the phenyl ring seems to build up a T-stacking interaction with Tyr 216
- 4 - Figure 2. The simulated conformation (A) and interaction diagram (B) of compound 1 and GSK-3β For compound 2, it has similar binding pattern with compound 1 (Figure 3). The two ketone groups in compound 2 play roles of hydrogen bond acceptors for contacting with Arg 96, Arg 180 and Lys 205. The whole structure of this molecule is trapped into a hydrophobic pocket lined by Gly 202, Ser 203 and Tyr 216. It should be noted that the phenyl ring herein also contributes to a π-stacking with Tyr 216. Figure 3. The simulated conformation (A) and interaction diagram (B) of compound 2 and GSK-3β In the case of compound 3, the simulated results suggest that the ketone group forms three hydrogen bonds with Arg 180 and Lys 205, while the imine group engages in hydrogen binding with Arg 96 in GSK-3β (Figure 4). The computational results also propose that this compound inserts into a hydrophobic network clustered by Ser 203 and Tyr 216. In addition, the phenyl ring seems to build up a π-stacking interaction with Tyr 216. 中国科技论文在线 http://www.paper.edu.cn
山国到技论文在线 http://www.paper.edu.cn A B Figure 4. The simulated conformation(A)and interaction diagram(B)of compound 3 and GSK-3B Figure 5 shows the proposed main interactions between GSK-3B and compound 4. In brief, this molecule is proposed to be captured by Val 214 and Tyr 216 through hydrophobic and T-stacking interaction. The ketone and chloride group forms a network of hydrogen bonds with Arg 96, Arg 180 and lys 205 A B Figure 5. The simulated conformation(A)and interaction diagram(B)of compound 4 and GsK-3B HipHop calculation produces ten hypotheses, and Hypol is the best significant pharmacophore hypothesis in this study, characterized by the highest ranking score(Table 1). All the ten hypotheses have the same features of one aromatic ring(r) and two hydrogen bond acceptors(A). In Hypol, it significantly that the docking conformations of these four compounds could be overlapped into three harmacophore points, in which the distances between two features are 3.96, 4.28 and 5.04 A( Figure 6), respectively Table 1. Results of the common feature hypothesis run* Hypothesis No Ranking score Direct hit Partial Hit RAA l08900 lIlI 0000 234 RAA 08.799 l111 0000 RAA 108.593 l111 RAA 108.388 l111
- 5 - Figure 4. The simulated conformation (A) and interaction diagram (B) of compound 3 and GSK-3β Figure 5 shows the proposed main interactions between GSK-3β and compound 4. In brief, this molecule is proposed to be captured by Val 214 and Tyr 216 through hydrophobic and π-stacking interaction. The ketone and chloride group forms a network of hydrogen bonds with Arg 96, Arg 180 and Lys 205. Figure 5. The simulated conformation (A) and interaction diagram (B) of compound 4 and GSK-3β HipHop calculation produces ten hypotheses, and Hypo1 is the best significant pharmacophore hypothesis in this study, characterized by the highest ranking score (Table 1). All the ten hypotheses have the same features of one aromatic ring (R) and two hydrogen bond acceptors (A). In Hypo1, it is significantly that the docking conformations of these four compounds could be overlapped into three pharmacophore points, in which the distances between two features are 3.96, 4.28 and 5.04 Å (Figure 6), respectively. Table 1. Results of the common feature hypothesis run* Hypothesis No. Composition Ranking Score Direct Hit Partial Hit 1 RAA 108.900 1111 0000 2 RAA 108.799 1111 0000 3 RAA 108.593 1111 0000 4 RAA 108.388 1111 0000 中国科技论文在线 http://www.paper.edu.cn