key: cord-0005417-tscdpmqc authors: Pradeep, H.; Rajanikant, G. K. title: A rational approach to selective pharmacophore designing: an innovative strategy for specific recognition of Gsk3β date: 2012-08-24 journal: Mol Divers DOI: 10.1007/s11030-012-9387-9 sha: 5cff34ecf431839b5a9abf4fbf3eacb2fd448a6b doc_id: 5417 cord_uid: tscdpmqc We propose a novel cheminformatics approach that combines structure and ligand-based design to identify target-specific pharmacophores with well-defined exclusion ability. Our strategy includes the prediction of selective interactions, developing structure, and knowledge-based selective pharmacophore models, followed by database screening and molecular docking. This unique strategy was employed in addressing the off-target toxicity of Gsk3β and CDKs. The connections of Gsk3β in eukaryotic cell apoptosis and the extensive potency of Gsk3β inhibitors to block cell death have made it a potential drug-discovery target for many grievous human disorders. Gsk3β is phylogenetically very closely related to the CDKs, such as CDK1 and CDK2, which are suggested to be the off-target proteins of Gsk3β inhibitors. Here, we have employed novel computational approaches in designing the ligand candidates that are potentially inhibitory against Gsk3β, with well-defined the exclusion ability to CDKs. A structure-ligand -based selective pharmacophore was modeled. This model was used to retrieve molecules from the zinc database. The hits retrieved were further screened by molecular docking and protein–ligand interaction fingerprints. Based on these results, four molecules were predicted as selective Gsk3β antagonists. It is anticipated that this unique approach can be extended to investigate any protein–ligand specificity. Herein, we introduce a novel computational screening strategy, which considers both the ligand features and receptor information for the identification of target-specific inhibitors. This rational approach includes deriving energy based pharmacophores [E-pharmacophores] from individual protein and using them to build selective pharmacophores, which are then sequentially employed to screen chemical databases. The selective pharmacophore model ensures that all essential features of each individual protein required for inhibition are conserved and taken into consideration during the sequential screening. Further, it also confirms that the final hits will have all essential features required for the selectivity towards single protein. To the best of our knowledge, this methodology has not been previously described to identify any selective inhibitors. In the present article, this unique strategy has been demonstrated by implementing on the classical example of glycogen synthase kinase 3β (Gsk3β) and cyclin dependent kinases (CDKs) off-target toxicity. Toxicity is a leading cause of attrition at all stages of the drug development process. Most of the safety-related attrition occurs at the preclinical level, earlier to which designing process might have exacted major portion of both time and money of the pharmaceutical companies [1] . Despite the tremendous advances in the science and technology of drug development, strangely there are no efficient approaches to tackle off-target effects of drugs. The consequences of off-target toxicity are increasing apace, prompting a desperate need of effective strategies against molecular cross-talking. Toxicity of drug is attributed by the molecular features that ensue in molecular cross-talk with different proteins rather than targeted protein. Hence, there is a burgeoning need to develop inhibitors that bind selectively to the target protein to address the toxicity related attritions [2, 3] . Gsk3β is a highly conserved evolutionary serine-threonine kinase. Historically, it was depicted as a central protein required in glycogen metabolism [4] . However, mounting evidence indicates that Gsk3β is a multifunctional enzyme involved in various crucial pathways such as the insulin signaling [5] and the Wnt pathway [6] , and plays a crucial role in the synaptic plasticity [7] and memory formation [8] . Most significant functions of this kinase are its role in the phosphorylation of apoptotic regulators and other cell death related factors, thus promoting the cellular degeneration [9, 10] . Hence, Gsk3β has emerged as one of the most attractive therapeutic targets for the development of selective antogonists as promising new drugs for numerous debilitating diseases such as Alzheimer's disease, ischemic stroke, chronic inflammatory processes, cancer and diabetes [11] [12] [13] [14] [15] . Most of the Gsk3β inhibitors developed thus far often show off-target effects due to serious limitations in their specificity. The low specificity of these inhibitors is due to the fact that the ATP binding site is highly conserved among diverse protein kinases. Even an extremely promising Gsk3β antagonist, can produce off-target effects by inducing changes in protein kinases, especially CDK1 or CDK2. CDK1 and CDK2 are archetypical kinases and central regulators that drive cells through G2 phase and mitosis, the inhibition of these key proteins has negative consequences on cell-cycle progression. As CDKs and Gsk3β belong to the same family of CMGC kinases, they share a high degree of sequence and structural homology [16] . Hence, a strategy should be implemented to develop Gsk3β selective inhibitors which perceptively exclude CDK1 and CDK2. To achieve these goals, we have (1) exploited the energetic analysis of protein-ligand docking to elucidate the key features of proteins/ ligand molecular interactions, (2) employed the knowledgebased coherent poses to derive a structure-based pharmacophore for Gsk3β and CDKs, (3) manually developed a selective pharmacophore by the visual analysis of all the three pharmacophores, and (4) applied the selective pharmacophore and virtual screening (VS) protocol to identify selective inhibitors of Gsk3β with the exclusion ability against CDKs (Fig. 1 ). The crystal structure of Gsk3β (PDB ID: 3GB2, 1Q41), and CDK2 (PDB ID: 3S2P) complexed with their respective antagonists were obtained from Protein Data Bank. Since, Fig. 1 Pipeline depicting the selective inhibitor designing the crystal structure of CDK1 is not available we built a homology model from the reported structure of CDK2 and employed in molecular modeling. The 3D structure of protein complexes was prepared using the protein preparation wizard tool (Schrodinger, LLC, 2010); water molecules were deleted, bond orders were assigned, hydrogens were added and metals were treated appropriately when present. Next, the orientation of the side chain structures of Gln and Asn was flipped if necessary to provide the maximum degree of H-bond interactions. The charge state of his residues was optimized. Finally, a restrained minimization of the protein structure was performed using the OPLS force field with backbone atoms being fixed [17] . Prepared protein structures were used to generate Glide scoring grids for the subsequent docking calculations. To each of the crystal structures of protein, a grid box of default size (20 × 20 × 20 Å) was centered on the corresponding active site position. Default parameters were used and no constraints were included during grid generation. The information of a binding pocket of a protein for its ligand is very important for revealing true binding mechanism and conducting mutagenesis studies [18] [19] [20] [21] . The binding pocket consists of a set of residues that have at least one heavy atom with a distance 5.0 Å from any heavy atom of a ligand. Such a criterion was originally used to define the binding pocket of ATP in the Cdk5-Nck5a* complex [22] that has later proved quite useful in identifying functional domains and stimulating the relevant truncation experiments [23] . The similar approach has also been used to define the binding pockets of many other receptor-ligand interactions important for drug design [19, [24] [25] [26] [27] . Here, the binding pockets were manually defined by either selecting the amino acid residues that span the kinase domain or selecting the ligand present within the crystal structure. Generation of selective pharmacophore model Recently developed E-pharmacophore approach of Schrödinger was employed in the construction of these energy-optimized pharmacophores. The E-pharmacophore [28, 29] coalesces the ligand and receptor-based techniques. Here, we have used the inhibitor library to probe the pharmacophoric features that are necessary for the selective inhibition of Gsk3β. This method has the advantage of combining pharmacophore perception with protein-ligand energetic terms to rank the importance of pharmacophore features. This combined approach attempts to take a step beyond simple contact scoring, by incorporating structural and energetic information obtained by docking. The library of 1,500 ligands for each protein was obtained from ChEMBL web site (https://www.ebi.ac.uk/chembldb/) and was used to probe the interaction types and chemically feasible geometries that are involved in the active site and ligand binding. The structures were annotated to expand protonation and tautomeric states using LigPrep [30] and ionization states at pH 7.0 ± 2.0 with Epik [31] . The generated grid files from the prepared receptors of Gsk3β and CDKs were used for the Glide_XP docking calculations. In Glide [32, 33] , the docking module "write XP descriptor information" option was enabled and rest of the parameters were kept as default. The XP scoring function was used to order the best ranked compounds and the specific interactions like π -cation and π − π stacking. In brief, the docking models of the inhibitors were refined using Glide XP, the Glide XP scoring terms were computed, and the energies were mapped into atoms. The E-pharmacophore sites were automatically generated with Phase [34] , using the default set of six chemical features: H-bond acceptor (A), H-bond donor (D), hydrophobic (H), negative ionizable (N), positive ionizable (P), and aromatic ring (R). In this manner, pharmacophores were derived for each of the proteins, representing Gsk3β, CDK1 and CDK2 and later were evaluated manually for the generation of selective models. The best pharmacophore hypothesis, AADDRRR was selected based on the careful observations of the feature scores and alignment of ligand structures to the generated hypotheses. The selected 3D pharmacophore hypothesis encompassed the following features: two H-bond acceptors The selective pharmacophore model (AADDRRR) was used as a query for retrieving potential inhibitors from the ZINC chemical database (13 million compounds). Virtual screening was carried out using zincPharmer [35, 36] that uses the pharmacophore to efficiently search the ZINC database of fixed conformers for pharmacophore matches. To accomplish the best 3D similarity search, we used constraints that included a maximum of 0.5 RMSD, obeying 15 rotatable bond cutoff and molecular weight range of 180-500 Dalton. The compounds that matched well with the pharmacophoric features of the AADDRRR hypothesis were retrieved as hits and were considered for molecular docking studies. Docking study was performed using Glide running on Windows XP. The Glide algorithm estimates a systematic search of positions, orientations, and conformations of the ligand in the enzyme-binding pocket via a series of hierarchical filters. The shape and properties of the receptor are symbolized on a grid by various dissimilar sets of fields that furnish precise scoring of the ligand pose. The potential hit compounds with lowest RMSD and standard molecular weights were considered for the docking analysis. All the hits were subjected to hydrogen additions, removal of salt, ionization and generation of low-energy ring conformations using LigPrep [37, 38] . The tautomers for each of these ligands were generated and optimized. Partial atomic charges were computed using the OPLS-2005 force field. The VS workflow in Maestro was used to dock and to score the lead-like compounds. In the first step, Glide was run in high-throughput virtual screen mode. 10 % of the top-scoring ligands were kept to move onto the Glide Single Precision (SP), stage. 5 % of the top-scoring leads from Glide-SP were retained and docked with Glide XP. Default values were accepted for van der Waals scaling and input partial charges were used. During the docking process, the G-score was used to select the best conformation for each ligand. G-score is based on the ChemScore, but includes a steric-clash term and adds buried polar atoms devised by the Schrodinger to penalize electrostatic mismatches: G-score = 0.065*vdW + 0.130*Coul + Lipo + Hbond + BuryP + RotB + site. To demonstrate the effectiveness of the described strategy, an additional docking experiment was carried out to compare the selective ability of our method with previously reported computational approaches for the identification of Gsk3β inhibitors. Here, we selected 15 potent ligands reported earlier for Gsk3β inhibition [39] [40] [41] [42] [43] [44] [45] [46] . These ligands were prepared using the LigPrep and conformational flexibility was sought through the ConfGen tool of the Schrodinger suite. Finally, the low-energy 3D structures of all the ligands were generated and employed in XP docking into the active sites of Gsk3β, CDK1, and CDK2. Generation of selective pharmacophore model The E-pharmacophore is a unique strategy, which blends the beneficial aspects of structure-based and ligand-based techniques. Mounting evidence indicates the advantage of incorporating protein-ligand interactions over the analysis with conventional ligand information alone. Here, we have used the inhibitor library for each protein to probe the pharmacophoric features that are necessary for the selective inhibition of Gsk3β [47] [48] [49] [50] . This method has the advantage of combining pharmacophore perception with protein-ligand energetic terms to rank the importance of pharmacophore features. This protein-ligand based approach attempts to take a step beyond simple contact scoring by incorporating structural and energetic information using the scoring function in glide extra precision (XP). Overall, 15 pharmacophore sites were predicted for each protein. The pharmacophore hypotheses of Gsk3β, CDK1, and CDK2 were analyzed carefully and as expected the pharmacophore hypotheses exhibited high degree of similarity with each other, especially at the base recognition part of the kinases (ADR of the all pharmacophore models) (Fig. 2) . From this, we understand that an ideal pharmacophore should possess these features, in addition to which we have to redact few discriminating features. By careful observation of the protein structure and pharmacophore, we concluded that the addition of donor and aromatic features 5 Å above to the base recognition features would be ideal exclusion strategy. The final hypothesis AADDRRR consists of two H-bond acceptors (A), two H-bond donor (D), and three aromatic ring (R) features. Inter-site distances between each feature are shown in Fig. 3 .The H-donor (D7) maps to the carboxylic group of the Asp133, while the H-bond acceptor (A4) maps to the peptidyl amino group of the Val135, indicating a strong influence on possible inhibition. Two aromatic features (R13 and R15) occupy the hydrophobic cleft in the base recognition site. While the acceptor (A5) and aromatic (R14) features are crucial for selectivity, spans ribose pocket of the active site (Fig. 4) . The ribose pocket of the Gsk3β is comparatively wider than both CDK1 and CDK2, hence the former can tolerate bulkier groups at this position when compared to CDKs. Thus, these energetically favorable features encompass the specific interactions required for the selectivity of the Gsk3β protein, and this information should prove helpful in the development of new Gsk3β inhibitors. Pharmacophore-based methods have been widely used in virtual screening [48, [50] [51] [52] [53] [54] . Application of structure-based pharmacophore in VS provides an advantage over the ligandbased pharmacophore approach as it uses the spatial information of the target protein for topological description of ligand-receptor interactions. It also provides an efficient alternative to docking-based VS, while continuing to represent specific ligand-protein interactions. Moreover, recent studies indicated that the structure-based pharmacophore approach provided more detailed information and accuracy in its description of ligand binding than ligand-based methods. VS was performed to identify possible lead compounds from the zinc database. The zinc chemical database with drug-like compounds consisting of over 13 million structurally diverse small molecules was screened with this query using the zincPharmer tool. A molecule which fits well with the pharmacophoric features, AADDRRR, was retrieved as a hit. Since the active site of Gsk3β is large, the hits may be larger in size and may violate the Lipinski's rule of five. To avoid possible inappropriate predictions, the molecular weight cutoff of 180-500 dalton range, RMSD of 0.5 and cutoff limit of 15 rotatable bonds filters were applied. Altogether, 3D structural query retrieved 1,206 compounds from the ZINC database. Molecular docking is a computational technique that tries to predict most desirable conformation of the ligand that fits into the receptor site of the protein. Docking is espe-cially useful as second step of the virtual screening for the identification of molecules that can fit into the pharmacophore. To further refine the retrieved hits and also to remove the false positives, VS hits were prepared using LigPrep and docked into the Gsk3β and CDKs using Glide. Further, the low-energy stereoisomers for each structure were generated; ultimately giving rise to 2534 isomers. Docking calculations were carried out using the VS workflow, a docking interface in the Schrödinger software. Finally, the binding modes of ten structures for each protein obtained by docking were ranked 5 ) exhibited all the conventional interactions that are required for the inhibition of kinase activity, viz. interfering with the ATP biding ability of Asp133 and Val135 and hydrophobic interactions with the hydrophobic cleft formed by Lys85, Val70, Ile62, Leu188, Cys199, and Val110. The hinge residues depicted H-bond interactions with ligand and occupy H-donor (D7) and Hbond acceptor (A4) of the pharmacophore. While, hydrophobic cleft maps to the two aromatic features (R13 and R15) of the AADDRRR hypothesis. However, the docking of these ligands to the CDKs protein did not show any of these interactions, suggesting the low affinity of these ligands to the CDK active site (Table 1b, c). By visualizing Fig. 6 , we can assume that the aromatic group (R14) faces steric clashes with Gly13, Glu 12, Asn133, Gln132, and Glu163 residues of CDKs, whereas R14 was well tolerated by Gsk3β. Because of the steric clashes, the ligands in the active site of the CDKs tend to disorient, ensuing low affinity of these ligands to the CDK active site. In order to demonstrate the effectiveness of our approach, we performed the comparative docking analysis on fifteen GSK3β ligands collected from previously reported computational work. All these ligands were employed in docking against GSK3β, CDK1, and CDK2. All 15 ligands showed nearly identical GlideScore for GSK3β, CDK1 and CDK2, suggesting a high affinity of these ligands to the CDKs active site as well (Table 2 ). In comparison, the evaluation of Glide-Scores of the ligands derived from our approach indicated a greater selectivity towards GSK3β (Table 1) . Many marvelous biological functions of proteins and their profound dynamic mechanisms can be revealed by studying their internal motions [55] [56] [57] [58] . Likewise, to accurately identify the ligand-binding pocket [59] and to really understand the interaction of a protein receptor with its ligand, we should consider not only the static structures concerned but also the dynamical information obtained by simulating their internal motions or dynamic process [59] [60] [61] [62] . We will make further efforts in this regard in our future study. Mounting evidence implicate that GSK3β plays an important role in the onset of many grievous human diseases. Recent data have further illustrated the contribution of Gsk3β activity to apoptosis, suggesting that the design and development of selective small molecule inhibitors of Gsk3β may be a good therapeutic option in the treatment of numerous human diseases, associated with apoptosis. In this study, a new protocol for selective lead identification for Gsk3β has been proposed. The protocol combines both ligand and structure-based approaches. A highly Nearly identical docking scores of the ligands for all three proteins (Gsk3β, CDK1 and CDK2) indicate their equal affinity towards the target proteins In pharmacophore studies, the spatial alignment of the compounds is usually one of the key steps to obtain meaningful result. Figure 7 shows the good alignment of docked ligands with the pharmacophore. The pharmacophore model was further used to screen potential compounds from the zinc database. The VS process fetched 1,206 ligands as hits, which were further filtered by docking analysis. Finally, four hits obtained after docking procedure demonstrated better docking score, Glide energy and greater selectivity towards Gsk3β. In conclusion, unifying the pharmacophore modeling, screening, molecular docking, and manual interpretation approach altogether, we have developed a new methodology to identify selective inhibitors for Gsk3β (Fig. 1) . Further, we have delineated the important interactions responsible for binding and selectivity of ligands to the active site more effectively. This protocol may be useful for the identification of specific lead molecules for other potential drug targets as well. The application of discovery toxicology and pathology towards the design of safer pharmaceutical lead candidates Off-target immune cell toxicity caused by AG-012986, a pan-CDK inhibitor, is associated with inhibition of p38 MAPK phosphorylation Single-chain antibody-based immunotoxins targeting Her2/neu: design optimization and impact of affinity on antitumor efficacy and off-target toxicity The multifaceted roles of glycogen synthase kinase 3beta in cellular signaling Conditional ablation of Gsk-3β in islet beta cells results in expanded mass and resistance to fat feeding-induced diabetes in mice Lithium suppresses astrogliogenesis by neural stem and progenitor cells by inhibiting STAT3 pathway independently of glycogen synthase kinase 3 beta The role of GSK-3 in synaptic plasticity Increased platelet GSK3B activity in patients with mild cognitive impairment and Alzheimer's disease Glycogen synthase kinase-3beta phosphorylates bax and promotes its mitochondrial localization during neuronal apoptosis Glycogen synthase kinase-3beta and the p25 activator of cyclin dependent kinase 5 increase pausing of mitochondria in neurons Lithium reduces Gsk3b mRNA levels: implications for Alzheimer Disease Glycogen synthase kinase-3β controls autophagy during myocardial ischemia and reperfusion Glycogen synthase kinase 3-β: a master regulator of tolllike receptor-mediated chronic intestinal inflammation Inhibition of glycogen synthase kinase-3β counteracts ligand-independent activity of the androgen receptor in castration resistant prostate cancer Similar pattern of peripheral neuropathy in mouse models of type 1 diabetes and Alzheimer's disease A hierarchical phosphorylation cascade that regulates the timing of PERIOD nuclear entry reveals novel roles for proline-directed kinases and GSK-3beta/SGG in circadian clocks Pharmacophore generation and atom-based 3D-QSAR of novel quinoline-3-carbonitrile derivatives as Tpl2 kinase inhibitors Structure and mechanism of the M2 proton channel of influenza A virus An in-depth analysis of the biological functional studies based on the NMR M2 channel structure of influenza A virus Energetic analysis of the two controversial drug binding sites of the M2 proton channel in influenza A virus Structural bioinformatics and its impact to biomedical science A model of the complex between cyclin-dependent kinase 5 and the activation domain of neuronal Cdk5 activator Identification of the N-terminal functional domains of Cdk5 by molecular truncation and computer modeling Binding mechanism of coronavirus main proteinase with ligands and its implication to drug design against SARS Study of drug resistance of chicken influenza A virus (H5N1) from homology-modeled 3D structures of neuraminidases Inhibitor design for SARS coronavirus main protease based on "distorted key theory Analysis of ligand binding to proteins using molecular dynamics simulations Novel method for generating structure-based pharmacophores using energetic analysis Energetic analysis of fragment docking and application to structure-based pharmacophore hypothesis generation LigPrep, version 2.5. Schrödinger, LLC Epik: a software program for pK a prediction and protonation state generation for drug-like molecules Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy Glide: a new approach for rapid, accurate docking and scoring. 2. Enrichment factors in database screening PHASE: a novel approach to pharmacophore modeling and 3D database searching Pharmer: efficient and exact pharmacophore search ZINC-a free database of commercially available compounds for virtual screening Computational identification of novel histone deacetylase inhibitors by docking based QSAR Identification of novel potential HIF-prolyl hydroxylase inhibitors by in silico screening 6-Amino-4-(pyrimidin-4-yl)pyridones: novel glycogen synthase kinase-3β inhibitors Discovery of potent and bioavailable GSK-3beta inhibitors Identification of small molecules that inhibit GSK-3beta through virtual screening Synthesis of new dipyrrolo-and furopyrrolopyrazinones related to tripentones and their biological evaluation as potential kinases (CDKs1-5, GSK-3) inhibitors Aminofurazans as potent inhibitors of AKT kinase From a natural product lead to the identification of potent and selective benzofuran-3-yl-(indol-3-yl)maleimides as glycogen synthase kinase 3beta inhibitors that suppress proliferation and survival of pancreatic cancer cells Novel GSK-3beta inhibitors from sequential virtual screening Pharmacophore modeling, quantitative structureactivity relationship analysis, and in silico screening reveal potent glycogen synthase kinase-3beta inhibitory activities for cimetidine, hydroxychloroquine, and gemifloxacin Docking and molecular dynamics study on the inhibitory activity of novel inhibitors on epidermal growth factor receptor (EGFR) 3D pharmacophore based virtual screening of A 2A adenosine receptor antagonists A pharmacophore model specific to active site of CYP1A2 with a novel molecular modeling explorer and CoMFA Virtual screening for SARS-CoV protease based on KZ7088 pharmacophore points Knowledge-based virtual screening of HLA-A*0201-restricted CD8+ T-cell epitope peptides from herpes simplex virus genome Toward the virtual screening of potential drugs in the homology modeled NAD+ dependent DNA ligase from Mycobacterium tuberculosis Virtual screening for finding natural inhibitor against cathepsin-L for SARS therapy Generation of pharmacophore and atom based 3D-QSAR model of novel isoquinolin-1-ones and quinazolin-4-ones as potent inhibitors of TNFα Low-frequency resonance and cooperativity of hemoglobin Collective motion in DNA and its role in drug intercalation Solitary wave dynamics as a mechanism for explaining the internal motion during microtubule growth Low-frequency collective motion in biomacromolecules and its biological functions Novel inhibitor design for hemagglutinin against H1N1 influenza virus by core hopping method Molecular dynamics studies on the interactions of PTP1B with inhibitors: from the first phosphate-binding site to the second one Molecular dynamics simulations of CYP2E1 Docking and molecular dynamics simulations of peroxisome proliferator activated receptors interacting with pan agonist sodelglitazar