Phosphodiesterase 9A
2.2. Materials and Methods
In silico approach is the best way to screen drug-like candidate from the huge
dataset. In this study, xanthine based compounds were screened from ZINC database for PDE9A inhibition. Figure 2.1 illustrates the methodology used for screening best compounds from existing ones.
2.2.1. Structural analysis of active site of PDE9A for in silico studies
The structure of PDE9A catalytic domain (2HD1) was extracted from protein data bank (PDB) and its active site composition was analyzed. The structure based pharmacophore analysis was performed using PyMol and Discovery Studio Visualizer to understand the amino acid residues within 5 Å regions of substrate (cGMP) in cGMP- PDE9A complex. 5 Å regions around substrate were selected as this range covered all active site residues. This study was carried out by superimposition of all PDEs over PDE9A to understand the role of corresponding amino acid residues in the active site pocket of various members of PDE superfamily (Singh and Patra, 2014).
2.2.2. Initial screening of xanthine based derivatives from ZINC database using Lipinski rule of five
Xanthine derivatives belong to a class of nitrogen based alkaloids. Derivatives such as Caffeine, Theophylline, IBMX etc. are known for their non-selective inhibition of PDEs (Hatzelmann et al., 1995). The structural resemblance of xanthine with substrate cGMP makes xanthine ring a suitable target for PDEs. Hence, in this study “xanthine ring” was taken as scaffold to search specific inhibitors for PDE9A from non commercial ZINC database [http://zinc.docking.org/]. A total of 2055 xanthine derivatives available in the ZINC database were extracted for initial pharmacophore screening. The files for virtual screening were generated with the help of raccoon, a graphical interface for processing ligand libraries in different formats (PDB, multi-structure MOL2 and PDBQT), multiple receptor conformations (e.g. relaxed complex experiments) and flexible residues. These molecules were then prepared in Raccoon using Lipinski filter. The shortlisted compounds were subjected to virtual screening. Figure 2.2 illustrates the structure of
Figure 2.2 Structure of ‘Xanthine ring’
2.2.3. Macromolecule files preparation for virtual screening and other in silico studies
Upon consideration of entireness and resolution of structures available in RCSB protein data bank, three dimensional crystal structures of coding domain of PDE1B, PDE2A, PDE3B, PDE4D, PDE5A, PDE7A, PDE8A, PDE9A and PDE10A were extracted from the data bank (http://www.rcsb.org; PDB ID are 1TAZ, 1Z1L, 1SOJ, 1ZKN, 1RKP, 3DBA, 3G3N, 3ECM, 2HD1, 4DFF respectively). The water molecules were initially removed and hydrogen molecules were added. The protein ‘pdb’ files were prepared in Swiss-pdb Viewer which helped in analyzing the protein thoroughly and preparing the macromolecule file for virtual screening and docking by removing hetero atom including ligands and water molecules. For virtual screening, protein pdbqt file was prepared in raccoon. For docking, protein pdbqt file was prepared in AutoDock Tool (ADT) by removing polar hydrogen followed by addition of non-polar hydrogen, computation of Gasteiger charges and merging of non-polar hydrogen.
2.2.4. Virtual Screening for selection of specific inhibitors for PDE9A
Virtual screening of shortlisted 1480 compounds was carried in Autodock 4.2.
The preparation of grid file was required to understand the shape and property of the receptor under different sets of fields. Grid files were generated by covering all residues
The parameter used for generation of grid file in ADT was in x, y and z direction with resolution of 0.253Å. The docking files were 90×90×90 points generated in ADT. The parameters for docking file generation was 20 GA run, 150 population sizes, 27000 maximum numbers of generation and 25000000 maximum numbers of evaluations.
Finally, separate folder for each compound having macromolecule (pdbqt) files, ligand (pdbqt) files, grid (gpf) files and docking (dpf) files were generated. The virtual screening was carried out in CentOS Linux system with the help of scripts for generating grid .glg file and docking .dlg file.
2.2.5. Molecular docking of screened compounds obtained from virtual screening
Molecular docking was used to check the interaction pattern and further computational validation of compounds shortlisted with lowest free energy of binding.
Out of 1480 compounds, the 10 top hits were selected for molecular docking. The parameters used were 100 GA run, 300 population size, 27000 maximum numbers of generation and 25000000 maximum numbers of evaluations. Docking was carried out in Autodock 4.2 and ADT. Molecular docking of all ten compounds was carried out in CentOS Linux system. The interaction patterns were studied in PyMol software and Discovery Studio 3.1.
2.2.6. Comparative studies of top four screened compounds with other PDEs
PDE9A has certain unique features in its structure which makes this enzyme more specific towards cGMP. But it also has certain structural similarity with some of
drug discovery targeting PDE9A specific inhibitor. The comparative docking studies of selected compounds with other members of PDE superfamily were important to determine the potency and selectivity of screened inhibitors towards PDE9A. These comparative studies were performed by molecular docking in Autodock 4.2. with the same parameters as mentioned in the above section.
2.2.7. Molecular Dynamic Simulation of the best compound obtained from screening against ZINC database
After performing docking based virtual screening, MD simulation was carried out to equilibrate the protein structure and to understand the stability and compactness of the protein-ligand complex. Selected compounds obtained from docking and comparative studies were used as initial structures for MD simulation. MD simulation was carried out using GROMACS 4.5 software (Pronk et al., 2013) and GROMOS 53a6 force-field (Oostenbrink et al., 2004). Automated Topology Builder (ATB) server was used to generate topology and force field parameters of the ligand (Malde et al., 2011). The protein-ligand complex was solvated with simple point charge (SPC) water molecules in a 0.9 nm cubic box and chlorine ions were used to neutralize the whole system. Energy minimization of protein-ligand complex was carried out using the steepest descent algorithm until it converged with a force tolerance of 100 kJ mol−1 nm−1. After minimization, the protein-ligand complex was equilibrated at 300K through a stepwise heating protocol in the NVT ensemble using 2.0 fs integration time and 500 ps.
Equilibration step followed 500 ps in the NPT ensemble with 1 bar pressure by restraining the position of protein and ligand molecule. Finally, MD simulation of protein-ligand complex was performed for a timescale of 6 ns under periodic boundary
barostat (Parrinello, 1981) were used to control the temperature and pressure with a temperature and pressure coupling time constant of 1.0 ps. Particle-mesh Ewald method was used to calculate non-bonded interactions with a cut-off of 0.9 nm (Essmann et al., 1995). During simulation all bonds with hydrogen atom was constrained by using the LINCS algorithm (Hess et al., 1997).
2.2.8. Drug likeness and ADMET properties of screened inhibitors
Drug likeness is a required property for any inhibitor on certain decided parameters to qualify for being taken up as future drug. The best compound obtained from in silico studies was subjected to study drug-likeness properties. These properties includes rule of 5, leadlike rule, CMC like rule, MDDR like rule, WDI (World Drug Index) and blood brain barrier (BBB) permeability. PreADMET software was used to calculate the drug likeness, ADMET (adsorption, distribution, metabolism, excretion and transport) and toxicity properties of the selected compound [preadmet.bmdrc.kr].