Istanbul Technical University Graduate School of Science

ISTANBUL TECHNICAL UNIVERSITY  GRADUATE SCHOOL OF SCIENCE ENGINEERING AND TECHNOLOGY

IN SILICO DESIGN OF HERG NON-BLOCKER COMPOUNDS WITH RETAINED PHARMACOLOGICAL ACTIVITY USING MULTI-SCALE MOLECULAR MODELING APPLICATIONS

Ph.D. THESIS

Gülru KAYIK

Chemistry Department

Chemistry Programme

DECEMBER 2017

ISTANBUL TECHNICAL UNIVERSITY  GRADUATE SCHOOL OF SCIENCE ENGINEERING AND TECHNOLOGY

IN SILICO DESIGN OF HERG NON-BLOCKER COMPOUNDS WITH RETAINED PHARMACOLOGICAL ACTIVITY USING MULTI-SCALE MOLECULAR MODELING APPLICATIONS

Ph.D. THESIS

Gülru KAYIK (509112008)

Chemistry Department

Chemistry Programme

Thesis Advisor: Prof. Dr. Nurcan TÜZÜN Thesis Co-Advisor: Assoc. Prof. Dr. Serdar DURDAĞI

DECEMBER 2017

İSTANBUL TEKNİK ÜNİVERSİTESİ  FEN BİLİMLERİ ENSTİTÜSÜ

HERG BLOKER OLMAYAN FARMAKOLOJİK AKTİVİTESİ KORUNMUŞ BİLEŞİKLERİN ÇOK BOYUTLU MOLEKÜLER MODELLEME UYGULAMALARI İLE İN SİLİKO TASARIMI

DOKTORA TEZİ

Gülru KAYIK (509112008)

Kimya Anabilim Dalı

Kimya Programı

Tez Danışmanı: Prof. Dr. Nurcan TÜZÜN

Eş Danışman: Doç. Dr. Serdar DURDAĞI

ARALIK 2017

Gülru KAYIK, a Ph.D. student of İTU Graduate School of Science Engineering and Technology student ID 509112008, successfully defended the thesis entitled “IN SILICO DESIGN OF HERG NON-BLOCKER COMPOUNDS WITH RETAINED PHARMACOLOGICAL ACTIVITY USING MULTI-SCALE MOLECULAR MODELING APPLICATIONS”, which she prepared after fulfilling the requirements specified in the associated legislations, before the jury whose signatures are below.

Thesis Advisor : Prof. Dr. Nurcan TÜZÜN ......

Istanbul Technical University

Co-advisor : Assoc. Prof. Dr. Serdar DURDAĞI ......

Bahçeşehir University

Jury Members : Prof. Dr. Mine YURTSEVER ...... Istanbul Technical University

Prof. Dr. Kemal YELEKÇİ ...... Kadir Has University

Assis. Prof. Dr. Bülent BALTA ...... Istanbul Technical University

Assoc. Prof. Dr. Fethiye Aylin SUNGUR ...... Istanbul Technical University

Prof. Dr. Safiye ERDEM ...... Marmara University

Date of Submission : 26 October 2017 Date of Defense : 01 December 2017

vi FOREWORD

First of all, I would like to thank my Ph.D. thesis advisor Prof. Dr. Nurcan Tüzün and co-advisor Assoc. Prof. Dr. Serdar Durdağı for their kind concern, recommendations and supports during the course of my Ph.D. studies. I would like to present my acknowledgements to Istanbul Technical University Research Fund BAP (Project numbers: 38208 and 30492) and the National Center for High Performance Computing of Turkey (UHEM) under Grant 10982010 for supporting this thesis and providing the related computer resources. The numerical calculations reported in this thesis were partially performed at TUBITAK ULAKBIM, High Performance and Grid Computing Center (TRUBA resources). I would also like to thank The Scientific and Technological Research Council of Turkey (TUBITAK) for granting me the 2214-A Research Grant and providing financial support during the course of my Ph.D. thesis.

December 2017 Gülru KAYIK (Chemical Engineer)

vii

viii TABLE OF CONTENTS

Page

FOREWORD ...... vii TABLE OF CONTENTS ...... ix ABBREVIATIONS ...... xiii SYMBOLS ...... xv LIST OF TABLES ...... xvii LIST OF FIGURES ...... xix SUMMARY ...... xxv ÖZET...... xxvii 1. INTRODUCTION ...... 1 2. IN SILICO DESIGN OF NOVEL HERG-NEUTRAL SILDENAFIL LIKE PDE5 INHIBITORS ...... 3 2.1 Introduction ...... 3 2.2 Methods ...... 5 2.2.1 Molecular docking simulations ...... 6 2.2.2 Fragment-based de novo drug design & virtual screening ...... 7 2.2.3 MD simulations and post-processing MD analyses ...... 8 2.2.4 MD simulations of the target receptor: PDE5 in its apo state and bound with its inhibitors ...... 8 2.2.5 MD simulations of hERG K+ ion channel: Apo state and bound with PDE5 inhibitors ...... 9 2.2.6 Molecular Mechanics/Generalized Born surface area (MM/GBSA) calculations ...... 11 2.3 Results and Discussion ...... 11 2.3.1 Analysis of the key interactions of sildenafil with the target receptor (PDE5) ...... 13 2.3.2 Comparison of used docking tools in terms of predicting the binding positions of “Sildenafil” in the central cavities of hERG1 channel ...... 15 2.3.2.1 GOLD ...... 16 2.3.2.2 AutoDock ...... 21 2.3.2.3 MOE ...... 24 2.3.3 In silico Alanine mutagenesis study ...... 25 2.3.4 General statements on the binding energy predictions derived from GOLD, AutoDock and MOE ...... 27 2.3.5 Binding interactions of Vardenafil and Tadalafil with the hERG K+ ion channel ...... 27 2.3.6 Virtual screening results ...... 29 2.3.7 MM/GBSA analyses ...... 30 2.4 Conclusions ...... 30 3. INVESTIGATION OF PDE5/PDE6 AND PDE5/PDE11 SELECTIVE POTENT TADALAFIL-LIKE PDE5 INHIBITORS USING COMBINATION OF MOLECULAR MODELING APPROACHES, MOLECULAR FINGERPRINT-BASED VIRTUAL SCREENING

ix PROTOCOLS AND STRUCTURE-BASED PHARMACOPHORE DEVELOPMENT ...... 37 3.1 Introduction ...... 37 3.2 Methods ...... 41 3.2.1 Ligand and protein preparations ...... 41 3.2.2 Virtual library screening...... 41 3.2.3 Flexible molecular docking simulations ...... 41 3.2.4 Molecular Dynamics simulations ...... 42 3.2.5 Molecular Mechanics Generalized Born Solvation (MM/GBSA) Calculations ...... 42 3.3 Results and discussion ...... 43 3.3.1 Validation of the docking methodology ...... 43 3.3.2 Constructing the homology models of the catalytic domains of PDE6 (amino acid residues: 482-816) and PDE11 (amino acid residues: 587-910) .... 43 3.3.3 Binding affinity and binding pattern analysis of the hit compounds and tadalafil with PDE5, PDE6 and PDE11 ...... 47 3.3.4 MD simulations of apo and holo states of PDE5, PDE6 and PDE11 bound with the selected hit compounds (ZINC02120502 and ZINC16031243) and tadalafil ...... 59 3.3.5 MM-GBSA calculations ...... 62 3.3.6 hERG K+ ion channel activity of the compounds ...... 63 3.3.7 E-Pharmacophore studies ...... 65 3.4 Conclusions ...... 68 4. STRUCTURAL INVESTIGATION OF VESNARINONE AT THE PORE DOMAINS OF OPEN AND OPEN-INACTIVATED STATES OF HERG1 K+ CHANNEL ...... 71 4.1 Introduction ...... 71 4.2 Computational Methods ...... 73 4.2.1 Protein-Ligand docking calculations ...... 73 4.2.2 MD simulations ...... 74 4.2.3 Principal component analysis (Covariance analysis) of the MD trajectories ...... 75 4.3 Results and Discussion ...... 76 4.3.1 Protein-ligand docking calculations ...... 76 4.3.2 MD simulations ...... 79 4.3.2.1 General statements on the backbone and ligand RMSD evaluations and ‘Short Range’ energetics ...... 83 4.3.2.2 MD simulations initiated with GOLD/GoldScore fitness functions: analysis of the trajectories of vesnarinone-hERG1 K+ channel model in its open-inactivated state ...... 86 4.3.2.3 MD simulations initiated with GOLD/GoldScore fitness functions: analysis of the trajectories of vesnarinone-hERG1 K+ channel model in its open-state ...... 89 4.3.2.4 Principal Component Analysis (PCA) & comparison of time- dependent behaviors of apo states and vesnarinone-bound hERG1 systems . 93 4.3.2.5 MM/PBSA (Molecular Mechanics/Poisson Boltzmann surface area) calculations ...... 95 4.3.2.6 Extension of the MD simulations of hERG1-Vesnarinone complex systems ...... 102 4.4 Conclusions ...... 108

x 5. CONCLUSIONS AND RECOMMENDATIONS ...... 111 REFERENCES ...... 113 CURRICULUM VITAE ...... 127

xii

ABBREVIATIONS

B3LYP : Becke-Three Parameter Lee, Yang, Parr Density Functional cAMP : 3′, 5′-cyclic adenosine monophosphate CG : Conjugate Gradient cGMP : 3′, 5′-cyclic guanosine monophosphate CS : Conformational Search DFT : Density Functional Theory ED : Erectyle dysfunction EMEA : European Medicines Agency FDA : U.S Food And Drug Administration GBVI/WSA dG : Generalized Born Volume Integral/Weighted Surface Area hERG : human ether-à-go-go-related gene LGA : Lamarckian Genetic Algorithm LINCS : Linear Constraint Method LQTS : Long QT Syndrome MD : Molecular Dynamics MM-GBSA : Molecular Mechanics Generalized Born Solvation Area MM-PBSA : Molecular Mechanics Poisson-Boltzmann Solvation Area NPT : Isothermal-Isobaric Ensemble NVT : Isochoric-Isothermal Ensemble PDE11 : PDE11-phosphodiesterase 5-type enzyme PDE5 : PDE5-phosphodiesterase 5-type enzyme PDE6 : PDE6-phosphodiesterase 5-type enzyme PME : Particle Mesh Ewald QSAR :Quantitative Structure-Activity Relationships RMSD : Root mean square deviation RMSF : Root mean square fluctuation SD : Steepest Descent SAR : Structure-Activity Relationships TPSA : Topological Surface Area

xiii xiv SYMBOLS

Å : Angstrom Cmax : Maximum Concentration fs : femto second IC50 : Inhibitor concentration-half maximum MW : Molecular Weight nm : nanometer ns : nano second μM : micro molar μs : micro second ps : pico second K : Kelvin T : Temperature

xvi LIST OF TABLES

Page

Table 2.1 : Comparison of different docking programs: GOLD, MOE and AutoDock (“os” and “ois” stands for open-state and open-inactivated-state of hERG K+ ion channel, respectively. Binding Energies are given in kcal/mol). 18 Table 2.2 : Summary of interactions of sildenafil with the hERG channel. (Top 10 docking poses of GOLD program are considered in generating the specific interactions, Gray Area: hERG-Open-State, White Area: hERG- Open-Inactivated-State) ...... 22 Table 2.3 : In silico Alanine Mutagenesis Results. The binding free energies (ΔGbinding) are expressed in kcal/mol. The results are calculated according to the top-ranking pose generated with GOLD program...... 26 Table 2.4 : 2D structures and predicted binding free energies of the novel sildenafil analogs...... 31 Table 2.5 : Physicochemical properties of the molecules, calculated by MOE...... 34 Table 2.6 : Comparison of protein-ligand free energy results of Sildenafil and Frag_21 Molecules Using MM/GBSA Analyses...... 34 Table 3.1 : Predicted binding free energies (Chemscore.dG) and 2D structures of ZINC compounds against the principal target, PDE5 and off-target enzymes, PDE6 and PDE11 (binding scores are expressed in kJ/mol and calculated by Chemscore fitness function implemented in GOLD Docking Program). Chemscore.dG values were converted to calculated IC50 values -for the purpose of selectivity comparison-according to the formula; ΔGbinding =RTlnIC50, where T is taken as 300 K...... 48 Table 3.2 : Comparison of protein-ligand free energy results of tadalafil with selected hit compounds using MM/GBSA calculations...... 63 Table 3.3 : Predicted binding affinities of the selected compounds within the hERG K+ channel. Each compound was docked into the central cavities of the channel by GOLD docking software with ChemScore fitness function. Docking studies were realized by considering the two known conformational states of the channel. OS and OIS states stand for the open and open-inactivated states. dG.ChemScore values are expressed in kJ/mol. Tadalafil and the two selected potent and selective PDE5 inhibitor compounds ZINC02120502 and ZINC16031243 are shown in bold in the Table...... 66 Table 4.1 : Docking scores (Fitness) of vesnarinone binding inside the central cavities hERG channels, generated with GOLD program...... 83 Table 4.2 : Average Lennard-Jones (LJ), Average Coulomb (CL) and the total value of the averages of these energies between the hERG1 channel and vesnarinone, extracted from the whole simulation time (50 ns). ASP, ChemScore, CHEMPLP and GoldScore denote the fitness functions that served as the docking algorithms to deliver the starting input geometries for the MD simulations. The energies (expressed in kJ/mol unit), given in bold, represent the lowest values of their corresponding columns. .... 87

xvii Table 4.3 : MM-PBSA Results of Vesnarinone Binding to hERG Channels. Interaction Binding Energy (ΔGbind) components with their standart deviations are shown in kJ/mol...... 99 Table 4.4 : MM-PBSA Results of Vesnarinone Binding to hERG1 Channels (for GoldScore trajectories) during the 0.5 μs simulation time. Interaction binding energies with different components with their standart deviations are shown in kJ/mol...... 106 Table 4.5 : Comparison of docking poses and scores using model and cryo-EM structures...... 108

xviii LIST OF FIGURES

Page

Figure 2.1 : 2D chemical structures of sildenafil, vardenafil and tadalafil and their corresponding IC50 values for the hERG K+ ion channel...... 6 Figure 2.2 : Flowchart of the current study in designing the novel sildenafil-like molecules based on fragment replacement strategy in order to obtain promising molecules whose binding affinity to hERG K+ ion channel is decreased as the pharmacological activity against their therapeutic receptors, PDE5, is retained...... 9 Figure 2.3 : Virtual screening analysis. Binding energy values are plotted against the frequency (number of molecules in the corresponding energy range). Values are obtained at the end of the VS protocol via docking the sildenafil analogs against hERG channel in its open state by means of MOE software...... 10 Figure 2.4 : 3D structures of homology models of hERG K+ ion channels used in the current study. S1-S6 and S5-S6 domains are shown for the open state (a and b, left side) and open-inactivated state of the channel (a and b, right side), respectively...... 12 Figure 2.5 : (A) 3D ligand interactions diagram of the sildenafil generated with GOLD program using rigid target treatment. Bidentate hydrogen bondings interactions with the invariant Gln817 residue are shown with red-dashed lines. (B) Superposition of sildenafil orientations obtained with different docking programs is shown: White color represents the rigid receptor treatment; turquaz color represents the flexible receptor treatment; brown color represents the flexible residue treatment on the H-loop domain, obtained with GOLD); green color represents the MOE program binding pose prediction and yellow color represents the AUTODOCK program binding pose prediction. The coloring scheme is based on carbon atoms. (C) The group definition of sildenafil, used in the current work...... 14 Figure 2.6 : 2D protein-ligand interaction diagrams of top docking poses produced by three different programs are represented. Polar and hydrophobic resiues are circled in purple and green colors, respectively. Basic residues are shown with purple color with an exterior blue circle. Green- dashed arrow represents hydrogen bonding interaction between ligand and the side chain of a residue. Solvent-accessible surface area for ligand and receptor is shown with blue smudge and turquoise halo, respectively. Arene-arene and arene-hydrogen bonding interactions are shown with green colors. 3D protein–ligand interaction diagrams are also shown. .. 18 Figure 2.7 : Superpositions of the top ten docking solutions for sildenafil orientations at the central cavity of hERG1 (OS), generated by AutoDock and their corresponding binding energy predictions...... 23 Figure 2.8 : A schematic view of the top docking poses of vardenafil and tadalafil at the central cavity of hERG1 channel. Ribbon representations and side

xix chains of the channel are covered within the 2.5 Å space around the ligand. Hydrogen bonds are shown with red dotted line with their distances. Left and right columns represent the interactions with the open state and open-inactivated state of the hERG1 channel, respectively. ... 29 Figure 2.9 : Comparison of RMSD graphs of sildenafil and sildenafil_frag21 molecules at the PDE5 and hERG1 open and open inactivated state targets...... 34 Figure 3.1 : Flowchart of the current study in the effort for identifying novel and selective PDE5 inhibitors...... 40 Figure 3.2 : Electrostatic maps of the active sites of the enzymes (left panel). Blue, red and white colours represent positively charged, negatively charged and hydrophobic preferences built at the drug-binding cavity site (near 5 Å distance from the ligand). Tadalafil fulfills the positively charged electrostatic requirement created by the acceptor atom (Oɛ) of the invariant Glutamine side chain in each active sites via carrying a hydrogen-bond donating moiety (-NH) at the amide fragment (namely, Glutamine Switch). On the other hand, hydrophobic residues, Phe820, Phe776 and Trp 820, sandwich the ligand (namely, Hydrophobic Clamp). 2D ligand-protein interaction diagrams (right panel). Green arrows indicate hydrogen bonding interactions. Green and purple discs show hydrophobic and polar residues, respectively. Representations are created with MOE molecular modeling package...... 45 Figure 3.3 : Top docking poses of tadalafil with PDE enzymes. Only polar hydrogens are shown for clarity. Protein residues within 2.5 Å distance around tadalafil are depicted in the figures. Hydrogen bonds between amide hydrogen of tadalafil and Oɛ atom of invariant Glutamine amino acid residue are represented with red dashed lines...... 46 Figure 3.4 : Aminoacid sequence alignments of PDE6 andPDE11 over PDE5 catalytic site residues. Alignment procedure was achieved with BLOSUM62 matrix via MOE software...... 46 Figure 3.5 : Ramachandran plots of the homology models of PDE6 and PDE11. ... 47 Figure 3.6 : Superposition of PDE5, PDE6 and PDE11, illustrated with blue, cyan and red colors, respectively. The counterions, Zn2+ and Mg2+, are shown with blue circles at the metal binding side...... 53 Figure 3.7 : Contact energy profiles of the catalytic side residues that correspond to PDE5, PDE6 and PDE11. The x axis and y axis represent the aminoacid residues numbering and atom-atom contact pair energies in kcal/mol unit, respectively...... 54 Figure 3.8 : 2D protein-ligand interaction diagrams of the selected compounds (Table 3.1) with the catalytic site residues of PDE5...... 55 Figure 3.9 : Superposition of 27 selected compounds at the end of the docking simulations...... 56 Figure 3.10 : (A) Docked pose of ZINC02120502 at the active site of PDE6. (B) Docked pose of ZINC02120502 at the active site of PDE11. (C) Overlay of ZINC02120502 (tan) onto the crystal orientation [118] of tadalafil (blue). (D) Overlay of ZINC16031243 docked poses at the active site of PDE5 (blue), PDE6 (pink) and PDE11 (tan). Protein–ligand interaction diagrams of ZINC02120502 and ZINC16031243 with PDEs (shown in the bottom of the figure)...... 57

xx Figure 3.11 : Traces of protein backbone RMSD (root-mean-squared-deviation) evaluation during the whole production stages of the MD Simulations. 60 Figure 3.12 : The RMSD evaluation of the ligands during the simulation time...... 61 Figure 3.13 : Traces of hydrogen bonding interactions throughout simulation time (x and y axis represents the simulation time and distances between Oε atoms of Gln817 (PDE5), Gln773 (PDE6), Gln869 (PDE11) and indole fragments hydrogen in the ligands, respectively). Color codes: orange: PDE5+ZINC16031243; purple PDE11+tadalafil; blue: PDE5+ZINC02120502; red: PDE5+tadalafil; green: PDE6+tadalafil. .. 62 Figure 3.14 : Simulated structures of ZINC0210502 and ZINC16031243 at the substrate pockets of PDE5, PDE6 and PDE11...... 64 Figure 3.15 : Overlay of docking pose (blue) and representative structure of ZINC02120502 (white) in the catalytic pocket of PDE11...... 65 Figure 3.16 : Root-mean-square fluctuation (RMSF) values per residue during the MD simulations...... 67 Figure 3.17 : (Top) Derived top-scored six-sited (RRRHHH) E-pharmacophore model; (bottom) 176 000 compounds from Otava small-molecules database are screened against derived pharmacophore model and top- 1000 compounds that have high Fitness scores with these sites are then docked at the PDE5 binding pocket using Glide/SP (standard precision). Compounds that show high docking scores as well as high fitness scores are shown in the figure. 2D ligand interaction diagram of selected Otava compound (1094821) is also represented in the figure...... 69 Figure 4.1 : Topology of the hERG channel in its open-state conformation. The channel is a coassembled of four identical chains; each chain is shown with different colours (top). S5 and S6 helices line the pore domain of the channel. A close look at the some selected S6 and SF (selectivity filter) residues that are important for drug binding (bottom)...... 73 Figure 4.2 : Open state hERG Channel (S5 and S6 chains, only, represented with green ribbon) inserted in DPPC lipid-bilayer. Water and lipid molecules are illustrated with red and olive-orange lines. Potassium cations (K+) are located at the energetically favorable S0-S2-S4 sites at the selectivity filter (SF) and also pore domain (PD) region, depicted in pink spheres. 75 Figure 4.3 : The superposition of the top docking poses of vesnarinone, generated by four different scoring functions in the GOLD program...... 79 Figure 4.4 : 2D and 3D depiction of the interaction of vesnarinone with the aminoacid residues that line the pore domain and central inner cavity of the hERG1 channels. HOIS and HOS stands for the open-inactivated and open states of the hERG1 channel, respectively. The red, purple and green lines represent the hydrogen bonding interactions and π-π stacking interactions within the residues and vesnarinone at around 5 Å distance, respectively. Green, turquoise and gray colours indicate hydrophobic, polar and Glycine residues, respectively...... 80 Figure 4.5 : Backbone RMSD values of hERG1 open (top) and open-inactivated (bottom) states during the MD simulations...... 83 Figure 4.6 : Ligand RMSD traces during the production stages of the MD simulations (x and y axes represent the simulation time and ligand RMSD values in ns and Å units, respectively). (The least-square fit was done on protein Cα atoms for RMSD measurements.)...... 84

xxi Figure 4.7 : Ligand RMSD traces during the production stages of the MD simulations. X and y axes represent the simulation time and ligand RMSD values in ns and Å units, respectively. The minimum, maximum and average RMSD values are also shown on the graphs, in Å unit. (The least-square fit was done based on the starting position of vesnarinone at the begining of simulation in order to measure the conformational flexibility of the ligand.) ...... 85 Figure 4.8 : Short range energetics (van der Waals and Electrostatics) interactions between vesnarinone and hERG1 channel throughout the simulation time. Left and right panels represent the open and open-inactivated states of the channel, respectively; starting from the docking outputs of the ASP, ChemScore, CHEMPLP and GoldScore fitness functions (from top to down). CL and LJ are abbreviations for Coulomb (Electrostatics) and Lennard Jones (van der Waals), respectively...... 86 Figure 4.9 : The crucial residue-vesnarinone interactions of the initial geometry and MD-simulated structure of the GoldScore trajectory for the open- inactivated state of hERG1. Hydrogen bonding and π-π stacking interactions are depicted in the figure.The distances of stacking interactions are measured based on the centroids of the aromatic rings. 87 Figure 4.10 : Distances of remarkable intermolecular interactions between vesnarinone and hERG1 channels residues in its open-inactivated state, during the MD simulations. “C=O group’s oxygen atom” and “quinolin moiety’s oxygen atom” phrases refer to the atom groups in vesnarinone...... 88 Figure 4.11 : Distances of remarkable intermolecular interactions between vesnarinone and hERG channels residues in its open-state during the MD simulations. “Methoxy oxygens and quinolin moiety –NH group’s H atom” phrases refer to the atom groups in vesnarinone...... 90 Figure 4.12 : The crucial residue-vesnarinone interactions of the starting geometry and MD-simulated structure of the GoldScore trajectory for the open state of hERG1.Hydrogen bonding and π-π stacking interactions are depicted in the figure. The distances of stacking interactions are measured based on the centroids of the aromatic rings...... 91 Figure 4.13 : Snapshots from MD simulations (namely, ChemScore trajectory) of hERG1open state-vesnarinone complexed system. Vesnarinone and hERG1 are represented with sticks and ribbon presentations, respectively. Each chain is coloured with different colours and one of the chains of hERG is deleted, for clarity (top). Overlay of the snapshots from the simulations were shown. The representation was created with a color code, from red to blue according to the simulation time and only 100 frames (taken each 50 frames in the evenly distributed time course) are shown for clarity (bottom)...... 92 Figure 4.14 : Overlay of the snapshots from the MD simulations (namely, ChemScore trajectory) of hERG1 Open-inactivated state-vesnarinone complexed systems.Vesnarinone and hERG1 are represented with sticks and ribbon presentations, respectively. Each chain is coloured with different colours and two of the chains of hERG are deleted, for clarity (top). Snapshots from the simulations were shown. The representation was created with a color code, from red to blue according to the

xxii simulation time and only 100 frames (taken each 50 frames in the evenly distributed time course) are shown for clarity. (bottom)...... 93 Figure 4.15 : Distance between hydrogen atom of Ser649 and dihydroquinolin moiety’s oxygen atom of vesnarinone, in the ChemScore trajectory for the open state of hERG1 channel during the 50 ns simulation production period. Just at the beginning periods of the simulation (within the 0-5 ns time period), H-bond between these atoms was broken and never formed later for the rest of the simulation...... 94 Figure 4.16 : First ten eigenvalues of the protein backbone movement along the simulation time. Note that, slopes of the ChemPLP and GoldScore curves overlap for the hERG1 open inactivated state...... 95 Figure 4.17 : Per-residue energy decompositions of MM-PBSA binding energy calculations. (Contributions of each residue were presented based on the cumulative contributions from each chain.)...... 100 Figure 4.18 : The course of distances established between methoxy oxygens of vesnarinone and Ser649 H atom. The positions of Tyr652 and Phe656 residues at the beginning and end of the MD simulation. Vesnarinone is shown in magenta color...... 101 Figure 4.19 : Vesnarinone superposition for representative frames of GoldScore (green) and ChemPLP (red) trajectories in the hERG1 open inactivated state. Only polar hydrogen is shown for clarity...... 102 Figure 4.20 : Protein backbone RMSD traces during the 550-ns production stages of the MD simulations, started from the GoldScore docking output geometries. The overlay of vesnarinone at 50-ns and 550-ns are also depicted (below)...... 103 Figure 4.21 : Ligand RMSD traces during the 0.55 μs production stages of the MD simulations, started from the GoldScore docking output geometries. .. 104 Figure 4.22 : Short range energetics (van der Waals and Electrostatics) interactions between vesnarinone and hERG1 channel throughout the 0.55 μs simulation time (for GoldScore trajectories). CL and LJ represent the Coloumb (Electrostatics ) and Lennard Jones interactions (van der Waals), respectively. Average Coloumb and Lennard Jones interaction energies are -44.37 kJ/mol and -195.44 kJ/mol for the open state of hERG1 whereas average Coloumb and Lennard Jones interaction energies are -91.50 kJ/mol and -233.76 kJ/mol for the open inactivated state of hERG1...... 105 Figure 4.23 : Per-residue interaction energies of MM-PBSA binding energy calculations of vesnarinone binding at the hERG1 channels during the 0.5 μs simulation time. (Contributions of each residue were presented based on the cumulative contributions from each chain.)...... 106 Figure 4.24 : Alignment of hERG channel cryo-EM (cyan) and 3D model (open- state) (orange) structures...... 107 Figure 4.25 : Per-residue energy decompositions of MM-PBSA binding energy calculations of vesnarinone binding to open-state hERG1 channels (using Cryo-EM structure) during the 50 ns simulation time. (Energetic contributions of each residue for binding were presented based on the cumulative contributions from each chain.). (Residues between 635-668 are shown only, for clarity.) Note that, the calculated per-residue contribution energy profile demonstrates merely the preliminary outcome for 50 ns simulation time. The relatively high positive energetic

xxiii contribution to binding enthalpy–that came from polar Ser624 residues- need to be better refined via longer simulation time which may eventually further rectify the polar solvation energy component of ΔGbinding that also militate against binding process. For all that, with respect to the aminoacid residues that are situated at the pore domains of the channel, the relevant outcome still presents a qualitative data whether which residues tend to mostly contribute to binding; such as the bulky and aromatic Tyr652 residues on the S6 helix of the channel...... 109

xxiv

IN SILICO DESIGN OF HERG NON-BLOCKER COMPOUNDS WITH RETAINED PHARMACOLOGICAL ACTIVITY USING MULTI-SCALE MOLECULAR MODELING APPLICATIONS

SUMMARY hERG K+ ion channels are responsible in the regulation process of the action potential of human ventricular myocyte by contributing the rapid component of + delayed rectifier K current (IKr) component of the cardiac action potential. The direct inhibition of hERG channels/current raises cardiac diseases, i.e., serious life- threatening arrhythmias further leading to Torsades de pointes (TdP) and long QT syndrome (LQTS). Various drugs have been withdrawn from the drug markets such as terfenadine, cisapride, astemizol or restricted in their use icluding thioridazine, haloperidol, sertindole, and pimozide due to hERG-related life-threatening cardiac arrhythmias. In this thesis, molecular interactions between human ether-à-go-go-related gene (hERG) blocker compounds and pore domains of hERG1 K+ ion channels (in its voltage driven open and open-inactivated conformational states) together with their target proteins were investigated using computational modeling methods. In silico drug rehabilitation studies were conducted for the selected compounds (i.e., some phosphodiesterase 5-type enzyme (PDE5) inhibitors) as well with an aim of reducing their hERG1 blocking affinity while keeping their principal enzymes’ activity. Using computational approaches such as homology modeling, protein-ligand docking, Alanine mutagenesis studies, E-pharmacophore modeling, structure- and ligand- based virtual screening strategies (such as rapid screening of drug-like molecular database with molecular similarity approaches and fragment based in silico design of novel compounds using fragment libraries), molecular dynamics (MD) simulations, post-MD analyses via considering the detailed analysis of the trajectories-especially conformational behaviours of both the apo and holo systems-and post-MD computations ((e.g., Molecular Mechanics Poisson-Boltzmann Surface Area and Molecular Mechanics Generalized Born Surface Area (MM-PBSA and MM-GBSA, respectively)) for evaluating the binding energies of the studied ligands towards the hERG channels and also towards PDE enzymes have enabled to get insight into the important blocking elements of hERG channels in terms of drug association with its central inner cavities. Additionally, some drug-like hits (with reduced hERG blocking affinity and kept principal target activity) were presented in the consequence of the multi-step virtual screening protocols. Hence, the consequences of the derived outcomes from this thesis may be helpful in the effort for designing novel and safe drugs. The studied drugs, herein, (e.g., sildenafil-ViagraTM,vardenafil-LevitraTM, tadalafil- CialisTM and vesnarinone) are all PDE5 inhibitors used in the treatment of erectyle dysfunction except for vesnarinone which is used as a PDE3 inhibitor). Phosphodiesterase (PDE) type enzymes are important biological elements of human organisms and found in many tissues and organs. Their biological function in human

xxv organism is to regulate the cytoplasmic concentration levels of intracellular second messengers, 3’,5’-cyclic guanosine monophosphate (cGMP) and/or 3’,5’-cyclic adenosine monophosphate (cAMP) via catalytic degradation reactions occuring in their substrate pockets. At the end of the applied multi-scale modeling applications, tadalafil and sildenafil-like molecules having drug-like properties (tested using pharmacokinetic modeling applications) have been identified and proposed as novel and safe potential PDE5 inhibitors. In addition to the hERG direct blocking investigations, selectivity of tadalafil-like hits has also been tested against PDE6 and PDE11 isoenzymes which have been identified as the causes of the side-effects of the used PDE5 inhibitor drugs in the market. In summary, through the duration of this thesis, various computational modeling tools have been utilized and applied in the investigation of novel and safe drugs (with reduced side-effects towards hERG channels and also towards the related PDE- isoenzymes) whilst keeping their principal target activities. Besides, the critical protein-ligand interactions have been illuminated at the molecular level along with the dynamics of the studied drugs and their target proteins. This thesis comprises of three chapters that present three scientific articles that have been published at peer-reviewed journals. In the first article (Chapter 2), various computer aided drug-designing and computational molecular modeling techniques were used for the investigation of the action mechanisms of the FDA approved drugs, sildenafil-ViagraTM, vardenafil- LevitraTM and tadalafil-CialisTM with their target protein, PDE5 and also hERG1 channels. In addition, fragment-based virtual screening strategy was employed in order to obtain potent and safe sildenafil-like molecules. Second article (Chapter 3) focuses on the development of potent tadalafil-like molecules via combination of ligand-based screening and structure-based modeling protocols. hERG1 binding affinities of the selected compounds together with tadalafil were evaluated via flexible molecular docking computations. In addition, PDE5/PDE6 & PDE5/PDE11 selectivities of the compounds were studied and important structural binding patterns such as the critical residue-ligand interactions were highlighted. Third article (Chapter 4) deals with vesnarinone (used a PDE3 inhibitor agent)-hERG complex systems and time-dependent dynamical behaviour of vesnarinone at the pore domains of hERG channels in its open and open-inactivated states. By the use of molecular docking, MD simulations and detailed post-MD analysis computations, possible binding modes of vesnarinone within the central cavities of the channels were proposed. Also, crucial hERG residues in terms of vesnarinone binding were further highlighted which may help to design safe and novel drug-like molecules. In Chapter 5, overall interpretation of the results and potential further works are briefly presented.

xxvi HERG BLOKER OLMAYAN FARMAKOLOJİK AKTİVİTESİ KORUNMUŞ BİLEŞİKLERİN ÇOK BOYUTLU MOLEKÜLER MODELLEME UYGULAMALARI İLE İN SİLİKO TASARIMI

ÖZET

Bu tezde, bilgisayar destekli moleküler modelleme metodları kullanılarak, hERG blokörü olan bileşiklerin hERG1 K+ iyon kanalları (farklı membran potansiyellerinde açık ve açık-inaktif konformasyonel halleri göz önünde bulundurularak) ve ilgili hedef proteinleri ile olan etkileşimleri moleküler seviyede detaylı olarak incelenmiştir. İn siliko yöntemlerden yararlanılarak, literatürden seçilmiş bazı fosfodiesteraz tip 5 (PDE5) enzim inhibitörlerine yönelik olarak hERG1 kanalları için bağlanma afinitelerinin azaltılması amacı ile rehabilitasyon çalışmaları yapılmıştır. Aynı zamanda, bu bileşiklerin hedef proteinleri ile olan etkileşimlerinin aydınlatılması ve bağlanma enerjilerinin korunmasına yönelik olarak da hesaplamalar yapılmış ve incelemelerin sonuçları tezde sunulmuştur. Protein-ligant moleküler kenetlenme, homoloji modelleme, Alanin mutasyon çalışması, E- Farmakofor modelleme, yapısal ve ligant bazlı olarak sanal molekül kütüphanelerinin çeşitli teknikler ile tarama stratejileri, moleküler dinamik (MD) simülasyonlar, MD sonrası analizler (simülasyon trajektörilerinin detaylı olarak incelenmesi-özellikle apo ve holo sistemlerinin konformasyonal davranışları olmak üzere) ve ligantların kompleks yaptığı proteinlere (hERG1 iyon kanalı ve aynı zamanda PDE enzimlerine karşı olan) bağlanma enerjilerinin tayin edilmesi amacı ile MD sonrası yapılan hesaplamalar ((Moleküler Mekanik Poisson Boltzmann Yüzey Alanı (MM-PBSA) ve Moleküler Mekanik Genelleştirilmiş Born Yüzey Alanı (MM- GBSA)) neticesinde, hERG kanallarının merkez iç bağlanma bölgesinde gerçekleşen ilaç-benzeri küçük moleküller ile etkileşimlerinde rol oynayan önemli bloke edici elementler ve aminoasit rezidüleri ortaya konulmuştur. Aynı zamanda, uygulanan çok kademeli sanal tarama protokolleri sonucunda, bazı ilaç-benzeri moleküller (ana reseptör bağlanma afinitesi korunarak hERG bağlanma afinitesi düşürülmüş) sunulmuştur. hERG K+ iyon kanalları, insan kalp karıncığı kas hücrelerinin görev aldığı aksiyon potansiyelinin düzenlenmesi işleminde rol üstlenirler. Bu aşamada, hERG kanalları, hücre içi-dışı iyon konsantrasyon gradientlerinden kaynaklanan hücresel elektriksel potansiyeline bağlı olarak genel olarak üç aşamalı bir konformasyonel aşamadan (açık hal, açık-inaktif hal ve kapalı hal olmak üzere) geçerek potasyum akışını sağlarlar. hERG kanallarının içerisinden gerçekleşen potasyum iyonları akışı, hERG kanallarının sahip olduğu seçici filtrenin (SF) sahip olduğu aminoasit rezidülarının (S624, V625,G626, F627, G628) K+ iyonları ve su molekülleri ile oluşturabildiği koordinasyon motifi sayesinde gerçekleşir. Bu potasyum akışı, kalp aksiyon + potansiyelinin, IKr komponentine (geciktirilmiş düzeltici K akışı hızlı bileşeni) önemli bir katkı yapar. hERG kanallarının ve görev aldıkları akışın doğrudan inhibe edilmesi, ciddi ve ölümcül kalp hastalıklarına (örneğin Torsades de pointes (TdP) ve

xxvii uzun QT sendromu (LQTS)) neden olabilmektedir. Yıllar içerisinde, yol açtıkları kardiyak aritmiyaları sebebi ile birçok ilaç (bazı antihistaminler, antibakteriyeller, antipsikotikler ve antidepresanlar dahil olmak üzere) piyasadan çekilmiştir ve/veya kullanımları kısıtlandırılmıştır. Bu ilaçlara örnek olarak terfenadine, cisapride, astemizol, thioridazine, haloperidol, sertindole ve pimozide verilebilir. Bu yüzden, FDA (Amerikan Gıda ve İlaç Dairesi) yeni ilaç geliştirme aşamalarında yapılan testlerde ilaçların hERG taramasından geçirilmesini 2007 yılında zorunlu kılmıştır. Tezde, seçilen ilaçlara uygulanan in siliko rehabilitasyon çalışmalarının sonuçları, bu sebeple yeni ve güvenli ilaç tasarım çalışmalarına bir katkı olarak değerlendirilebilir. Bu tezde çalışılan ilaçlar (sildenafil-ViagraTM, vardenafil-LevitraTM, tadalafil- CialisTM ve vesnarinone)-vesnarinone haricinde (PDE3 inhibitörü)-PDE5 inhibitörüdürler ve erektil fonksiyon bozukluğunu tedavi amacı ile kullanılırlar. PDE enzimleri, insan organizmasında çok geniş doku ve organlara yayılmış ve birçok hastalığın da tedavisinde kullanılan ilaçların etki ettiği ana reseptör olma özelliğine sahiptirler. Bu enzimlerin biyolojik fonksiyonları, substrat (3’,5’-siklik guanozin monofosfat (cGMP) ve/veya 3’,5’-siklik adenozin monofosfat (cAMP)) bağlanma bölgelerinde gerçekleştirdikleri katalitik degradasyon reaksiyonları üzerinden bu substratların sitoplazmik konsantrasyon seviyelerini düzenlemektir. Bu tezde uygulanan çok kademeli moleküler modelleme teknikleri neticesinde, ilaç olma özelliğine sahip (farmakokinetik özellikleri moleküler modelleme uygulamaları ile tahmin edildikten sonra) yeni ve güvenilir (hERG1 ve PDE5 bağlanma enerjileri moleküler kenetlenme ve MM-PBSA ve MM-GBSA hesaplamaları ile kontrol edilmiştir) PDE5 inhibitörleri olarak tadalafil ve sildenafil benzeri moleküller ortaya konulmuştur. İlaçların, ana reseptörlerine karşı olan bağlanma afinitelerinin korunarak hERG bloke etme özelliklerinin düşürülmesi çalışmalarının yanısıra, PDE6 ve PDE11 enzimlerine karşı olan aktiviteleri de çalışılmıştır. Çünkü, PDE6 ve PDE11 enzimlerinin katalitik kısımlarının PDE5 ile aminoasit sekansı ve 3-boyutlu sekonder yapılarındaki benzerlikleri ve bu benzerliklerin neticesinde ise PDE5 inhibitörü olarak kullanılan ilaçların bu izoenzimlere karşı aktivite gösterdiği bilinmektedir. Bu izoenzimlere karşı olan bağlanma yatkınlığı neticesinde bazı yan etkiler (görme bozukluğu, kas ağrıları, vb.) görülebilmektedir. Özetle, tez süresince, birçok bilgisayar destekli modelleme araçları ile birlikte moleküler modelleme teknik ve yaklaşımlarından yararlanılmış, yeni ve yan etkileri azaltılmış (ana reseptöre olan-PDE5-etki mekanizması korunarak) güvenilir ilaç- benzeri moleküller ortaya konulmuştur. Bununla birlikte, önemli ve kritik protein- ligant etkileşimleri detaylı bir şekilde moleküler düzeyde çalışılmış ve tartışılmıştır. Moleküllerin ve etki ettikleri proteinlerin dinamik özellikleri (zaman içerisindeki konformasyonel değişimleri vb.) uygulanan MD simülasyon teknikleri ve MD sonrası trajektörü analizleri ile ortaya çıkarılmıştır. Bu tez, SCI ve SCI-Expanded indeksinde yeralan dergilerde yayını gerçekleştirilmiş üç adet makalenin derlenmesinden oluşturulmuştur. Yayınlanma tarihlerine göre makalelerin içerikleri ve amaçları şu şekilde detaylandırılabilir: Birinci makalenin konusu; sildenafil, vardenafil ve tadalafil ilaçlarının (FDA onaylı birinci jenerasyon PDE5 inhibitörleri), hedef reseptörleri olan PDE5 enzimi ve hERG1 kanal modelleri (açık ve açık-inaktif halleri) ile etkileşimlerinin birden fazla moleküler kenetlenme programı (esnek, rijit ve indüklenmiş protein-ligand kenetlenme yaklaşımları uygulanarak) kullanılarak incelenmesi, bağlanma özellikleri

xxviii ve bağlanma enerjilerinin ortaya konulması olarak özetlenebilir. Çalışmanın amacı ise, sildenafil molekülünün fragment-bazlı sanal tarama tekniği kullanılarak belirli bir fragmentin çıkarılıp bu fragment yerine eklenen fragmentler ile oluşturulan bir moleküler sanal kütüphanenin moleküler kenetlenme hesapları ile taranması ve MD simülasyon ile MM-GBSA analizleri neticesinde hERG afinitesi düşürülmüş ve PDE5 aktivitesi korunmuş sildenafil-benzeri moleküllerin açığa çıkarılması şeklinde verilebilir. Ayrıca, hERG kanalının bağlanma bölgesinden seçilmiş bazı aminoasit rezidüleri için sildenafilin bağlanmasına yönelik olarak Alanin mutasyon çalışması da yapılmıştır. Bu aşamadaki amaç hangi amoasitlerin sildenafilin bağlanmasına daha çok katkı yaptığının aydınlatılmasıdır. Seçilen rezidüler teker teker Alanin aminoasit rezidüsüne dönüştürülmüş ve moleküler kenetlenme çalışmaları elde edilen bu mutant hERG kanalları ile tekrarlanmıştır. Ortaya çıkan bulgular (moleküler kenetlenme skorları) ile doğal-tipteki (wild) sonuçlar karşılaştırılmış ve sildenafilin bağlanması bağlamındaki kritik rezidüler açığa çıkarılmıştır. İkinci makalede ise, diğer bir PDE5 inhibitörü olarak kullanılan tadalafil molekülü ele alınmıştır. Geniş bir ilaç-benzeri moleküler veri tabanına ligand-bazlı bir tarama tekniği uygulanarak (Tanimoto benzerlik katsayısı ile MACCS yapısal moleküler parmakizi araçlarının kombine edilmesi vasıtasıyla) tadalafile %80 oranında yapısal benzerlik gösteren moleküller filtrelenmiştir. Filtrenen bu moleküllerin, PDE5, PDE6, PDE11 ve hERG kanallarına esnek moleküler doking çalışmaları yürütülmüştür. Bundan önce, PDE6 ve PDE11 enzimlerinin deneysel üç boyutlu yapılarının henüz keşfedilememesi sebebiyle 3D homoloji modelleri yapılmıştır. Seçilen bazı moleküllerin MD simülasyonları gerçekleştirilmiştir. Buradaki temel amaç, filtrelenen tadalafil-benzeri moleküllerin PDE5/PDE6 ve PDE5/PDE11 seçiciliklerinin korunabilmesi ve ayrıca bazı hERG afinitesi düşürülmüş moleküllerin önerilmesidir. Bunun için, seçilen moleküller ile hERG kanallarının hem açık hem de açık-inaktif hallerine moleküler kenetlenme çalışmaları yapılmış ve sonuçlar yayınlanmıştır. Üçüncü makalede ise, vesnarinon molekülünün (PDE3 inhibitörü) hERG kanallarının iç boşluk kısmına (açık ve açık-inaktif halleri gözönünde bulundurularak) bağlanması ile ilgili olarak, değişik skorlama fonksiyonları kullanılarak yapılan moleküler kenetlenme çalışmaları sonuçlarının incelenmesi, birbirleri ile karşılaştırılması ve peşinden gerçekleştirilen MD simülasyonların ve MM-PBSA analizlerin yardımı ile vesnarinon molekülünün hERG kanalları ile olan etkileşimlerinin analiz edilmesi ve en olası bağlanma oriyantasyonunun açığa çıkartılması hedeflenmiştir. Elde edilen moleküler kenetlenme skorları, MM-PBSA hesaplamalarından gelen bağlanma enerjileri, simülasyonlardan elde edilen protein- ligant arasındaki Lennard-Jones (van der Waals) ve Elektrostatik (Coloumb) enerjileri ile karşılaştırılmıştır. Ayrıca her bir simülasyon yörüngesi için yapılan simülasyon süresindeki detaylı ligand/protein RMSD (karekök ortalama sapma değeri) ve PCA (Temel Bileşenler Analizi) analizleri ile elde edilen bulgular detaylandırılmıştır. Buradan elde edilen sonuçlar, hERG-ilaç molekülü etkileşimlerinin daha iyi anlaşılabilmesi, bu bağlamda kritik önem taşıyan aminoasit rezidülerinin aydınlatılması (hem protein-ligant kenetlenme hem de MM-PBSA sonrası yapılan rezidü başına dekompozisyon hesaplamaları vasıtasıyla) ve bunların neticesinde elde edilen bulgular ile beraber uygulanan in siliko modelleme tekniklerinin, hERG bağlanması rehabilite edilmek istenen moleküllere genel anlamda bir örnek teşkil edebilmesi bakımından önem taşıması olarak özetlenebilir.

xxix xxx 1. INTRODUCTION

The subject of this Ph.D. thesis is to investigate the molecular interactions between hERG blocker compounds and the central inner cavities of hERG K+ ion channels (in its open and open-inactivated conformational states which are voltage-dependent processes driven by the ion concentration/flux across the cell membrane bilayers) along with their corresponding main target (e.g., PDE5) and off-target proteins (e.g., PDE6 and PDE11) via various in silico techniques. Also, computer-aided drug design strategies have been employed with aim of identifying safe and novel drug- like molecules with reduced hERG blocking affinity and cross-reactivity with off- target PDE enzymes while keeping their principal target activities.

The main concept and purpose of the studies within this Ph.D thesis is about understanding and illuminating the blocking mechanisms and blocking elements of some selected PDE inhibitors (sildenafil,vardenafil, tadalafil and vesnarinone) towards hERG1 channel models and also within their main targets and further develop novel and safe drug-like hits by utilizing multi-scale molecular modeling applications.

Bringing into novel and safe drugs to drug market is a long time requiring process with its own rigour sides and expensive multi-step pre-clinical and clinical stages. Besides, thanks to the rapidly developing computational molecular modeling techniques, (such as virtual screening strategies of large molecular databases, protein-ligand docking, QSAR (Quantitaive-Structure Activity Relationships) developments and many other structure- and ligand-based methods in order to predict the accurate bioactive conformations and correct binding free energy trends of small molecules with their target proteins), cost-effective alternative routes could be supplemented to the early phases of the conventional drug-discovery (developing new molecules) and/or rehabilitation efforts of the undesired side-effects of existing drugs. Since 1990s, mandatory hERG toxicity screening tests have been incorporated into the new drug development steps by FDA (U.S Food and Drug Administration) and EMEA (European Medicines Agency). Computational screening techniques and

1 further in silico investigation of protein-ligand holo systems, as in this study, present a branch of useful scientific data in the area of a drug design field. Hence, the outcomes of this thesis which have been published in peer-reviewed journals may contribute to the design/rehabilitation efforts of hERG non-blocker compounds and also could be beneficial in the design of more potent and safe PDE enzymes.

This thesis comprises of three chapters that present three scientific articles that have been published at peer-reviewed journals.

In the first article (Chapter 2), various computer aided drug-designing and computational molecular modeling techniques were used for the investigation of the action mechanisms of the FDA approved drugs, sildenafil-ViagraTM, vardenafil- LevitraTM and tadalafil-CialisTM with their target protein PDE5, and also hERG1 channels. In addition, fragment-based virtual screening strategy was employed in order to obtain potent and safe sildenafil-like molecules.

Second article (Chapter 3) focuses on the development of potent tadalafil-like molecules via combination of ligand-based screening and structure-based modeling protocols. hERG1 binding affinities of the selected compounds together with tadalafil were evaluated via flexible molecular docking computations. In addition, PDE5/PDE6 & PDE5/PDE11 selectivities of the compounds were studied and important structural binding patterns such as the critical residue-ligand interactions were highlighted.

Third article (Chapter 4) deals with vesnarinone (used a PDE3 inhibitor agent)-hERG complex systems and time-dependent dynamical behaviour of vesnarinone at the pore domains of hERG channels in its open and open-inactivated states. By the use of molecular docking, MD simulations and detailed post-MD analysis computations, possible binding modes of vesnarinone within the central cavities of the channels were proposed. Also, crucial hERG residues in terms of vesnarinone binding were further highlighted which may help to design safe and novel drug-like molecules.

In Chapter 5, overall interpretation of the results and potential further works are briefly presented.

2 2. IN SILICO DESIGN OF NOVEL HERG-NEUTRAL SILDENAFIL LIKE PDE5 INHIBITORS1

2.1 Introduction

Cyclic nucleotide phosphodiesterase enzymes (PDEs) have functions in regulating the levels of intracellular second messengers, 3′,5′-cyclic adenosine monophosphate (cAMP) and 3′,5′-cyclic guanosine monophosphate (cGMP), via hydrolysis and decomposing mechanisms in cells [1]. These enzymes are abundant in diverse tissues and systems such as immune and cardiovascular systems in human body. They take essential roles in modulating various cellular activities such as memory and smooth muscle functions [2-4]. The diversity and abundancy of these enzymes in human body render them potential drug targets for many diseases (i.e. asthma, depression, erectile dysfunction, pulmonary disease) [5-7]. PDEs are classified into 11 superfamilies according to their biochemical functions, structural and kinetic profiles, or their target substrates. These superfamilies have more than 60 subtypes and isoforms that are generated by different promoters in distinct tissues. Specifically, while cGMP is decomposed by PDE5, PDE6, and PDE9; and cAMP is hydrolyzed by PDE4, PDE7, and PDE8; PDE1, PDE2, PDE3, and PDE10 hydrolyze both cAMP and cGMP. PDE5A is the only subtype of PDE5 which has been identified so far. It has four isoforms (PDE5A1–4) that are mainly expressed in the corpus cavernosum [8]. The catalytic domain of PDE5A (residues: 535–860) complexed with sildenafil, vardenafil, and tadalafil – the drugs that are used in the treatment of erectile dysfunction (ED) was first isolated in 2002 by Sung et al. [8].

In silico studies have been used in better understanding of the molecular determinants of drug-PDE5 binding interactions so far with a profound attention to the side effects of sildenafil (i.e. visual problems, headache, etc.) which stemmed from the cross-reactivity of this drug to the PDE6 and PDE11 enzymes, particularly.

1 This chapter is based on the article “Kayık, G., Tüzün, N. Ş., Durdagi, S. (2016). In silico design of hERG-neutral sildenafil-like PDE5 inhibitors. Journal of Bimolecular Structure and Dynamics, doi: 10.1080/07391102.2016.1231634.

3 Hence, enormous effort for developing “second generation PDE5 inhibitors” has been mainly focused on discovery of novel drugs with better selectivity against other enzymes within the PDE family, so far. For example, Huang et al. [9] performed homology modeling, molecular docking, and molecular dynamics (MD) simulations studies for understanding the better selectivity of tadalafil targeting to the PDE6 enzyme compared to sildenafil. Cichero et al. [10] worked on tadalafil analogs in order to gain a detailed perspective for understanding the key interactions between the catalytic side residues of PDE5 and PDE11 enzymes that are responsible for the cross-reactivity within these referred receptors. In addition to the structure-based design studies, ligand-based approaches were also applied for designing novel PDE5 inhibitors [11]. Many in vitro studies have also been carried out for the same aim. [12-20].

In addition to the aforementioned sanitary concerns, another unpleasant side effect of sildenafil that has been related to cardiovascular problems has also been considered.

Patch clamp experiments showed that the IC50 values of the human ether-à-go-gorelated gene (hERG1) potassium (K) ion channel blocking affinity of sildenafil, vardenafil, and tadalafil as 33, 12, and 100 μM, respectively [21]. In vitro studies highlight the concentration-dependent blockage of hERG channel by these drugs [21,22]. hERG channel is responsible for the regulation of the action potential of human ventricular myocyte by contributing the rapid component of delayed rectifier + K current (IKr) component of the cardiac action potential [23]. Blockage of the hERG channel by several drugs is known to cause a loss of function leading to serious life-threatening disorders such as long QT prolongation (LQTS). Several drugs including antihistamines, antibacterials, antipschotics, and antidepressants have been identified as LQTS inducer and some drugs have been either withdrawn from the market (terfenadine, cisapride, astemizol, etc.) or restricted in use (thioridazine, haloperidol, sertindole, and pimozide, etc.) due to life-threatening cardiac arrhythmias [24,25]. Although experimental information (IC50, Cmax, etc.) of these drugs against hERG channel [21,22] are available, the absence of the crystal structure of hERG ion channel in apo form as well as X-ray structures of these compounds with the channel limits the perception of molecular determinants on the interactions of these PDE5 inhibitors with the hERG channel. The mandatory incorporation of hERG assays on the developing stage of new drugs prior to their

4 marketing leads researchers to test their molecules in terms of hERG blockage. For instance, the newly developed two PDE5 molecules were tested for their ability to cross-react with PDE1-PDE4, PDE6, and PDE11; and also their inhibition potency of the hERG current was checked by in vivo analysis [26]. The investigation of the electrophysiological features into an another PDE5 inhibitor (ER-118,585) has underlined the risk of cardiac action potential duration according to the applied in vivo experiments [27].

The aim of the present study is to identify the interaction patterns and binding affinity predictions of the selected PDE5 inhibitors (sildenafil, vardenafil and tadalafil) against the hERG1 channel in its open and open-inactivated conformational states as well as attempting to identify PDE5 inhibitor analogs with lower binding affinity to hERG1 ion channel while keeping the pharmacological activity against its principal target PDE5 using in silico methods such as molecular docking, virtual screening, MD simulations, and Alanine mutagenesis studies. The IC50 values of these selected PDE5 inhibitors at the hERG1 channel as well as their 2D chemical structures are shown in Figure 2.1. In this sense, current study will particularly contribute to the understanding of the underlying molecular mechanism and molecular determinants (key residue-ligand interactions) that are responsible for hERG1 binding affinity in its two distinct conformational states (i.e. open and open- inactivated states).

2.2 Methods

2D structures of the ligands were drawn with the Builder tool in MOE molecular modeling package (MOE, 2015) [28]. Energy minimization and conformational search of the ligands were conducted with MMFF94x force field. Ligand and protein preparation steps were performed in the wash and protein preparation modules of the MOE program, respectively. The unmutated and full-length crystal structure of PDE5 (PDB ID: 2H42) [29] was used as target structure in this study. While co- crystallized ligand was removed from the active site of the enzyme, crystal water molecules within the binding pocket were retained for the docking simulations. Three different docking algorithms, namely GOLD [30], AUTODOCK [31], and MOE docking were used in the current study.

Figure 2.1 : 2D chemical structures of sildenafil, vardenafil and tadalafil and their + corresponding IC50 values for the hERG K ion channel.

2.2.1 Molecular docking simulations

GOLD (v.5.3.0): Consensus docking protocol was used to generate protein-ligand complexes with GOLD 5.3.0 software. In this respect, two docking scoring functions were combined: GoldScore (for docking) and ChemScore (for re-scoring). Ten amino acids within the binding cavity of PDE5 (Asn661, Asn662, Ser663, Tyr664, Ile665, Ile778, Phe786, Met816, Gln817 and Phe820) were handled with full flexibility during the docking simulations. The first five aminoacids belong to the H- loop (660–683) whereas the others are part of the Glutamine switch pocket (Q pocket) and Hydrophobic clamp pocket (P pocket) where the critical drug-protein interactions take place. The flexible dihedral angles of Phe656 and Tyr652 of hERG ion channel were allowed full rotation with 10o increment. The number of genetic algorithm (GA) run was set to 100. Search efficiency was set to its maximum value (200%)-exploring the search space as wide as possible-in order to increase the reliability of the docking results.

MOE (v.2014.09): Therapeutic receptor (PDE5) is treated rigid during the docking simulations. However, side chains of the central inner cavity of the off-target receptor (hERG) were tethered with a weight factor of 10 via induced fit docking protocol. Two different docking scores were utilized in order to evaluate the predicted binding free energy of the ligands to these receptors: 1-London dG and 2- GBVI/WSA dG (Generalized-Born Volume Integral/Weighted Surface Area dG). Triangle Matcher was chosen as the ligand placement methodology. MMFF94x force field is used to refine the free energy of binding in the second refinement step. 100 poses were generated in each re-scoring steps.

6 AUTODOCK (v.4.2): The active site of PDE5 enzyme was defined based on the sildenafil coordinates in the crystal structure allowing enough space for ligand movement during the docking pose prediction. The grid spacing of 0.3 Å was used in grid generation calculations for hERG K+ ion channel and PDE5. The Gasteiger charges were added to the receptors and ligands. The docking search parameters were used with their default values during the docking simulations except that 100 docking poses were generated with GA for each ligand and Lamarckian Genetic Algorithm (LGA) was used for the conformational search step. The aromatic Tyr652 and Phe656 aminoacid residues on S6 helix regions of the hERG potassium ion channel were set as flexible during docking simulations. However, the coordinates of the crystal structure of PDE5 enzyme were treated as rigid.

2.2.2 Fragment-based de novo drug design & virtual screening

The sulfone and methylpiperazine groups are truncated from the molecule and replaced with the fragment library of MOE. Scaffold replacement in MOE resulted more than 100.000 sildenafil analogs. Some physiochemical restrictions were applied as filter during the virtual library generation in order to give sufficient drug-like properties to the derived molecules. After the attachment of fragments to the remaining core of the sildenafil, the generated molecules were filtered according to their lipophilicity (SlogP), molecular weight (MW), and topological surface area (TPSA) values. These descriptors are calculated from the atomic connection of the molecules and can be rapidly used for the indication of the absorption and bioavailability features of the drug-like candidate molecules. The following criteria were used in designing the virtual library: MW <500; 40

7 performed by MMF94X force field with 0.001 kcal.mol−1.Å−2 RMSD gradient sensitivity. 10 conformers for each molecule were generated with 0.005 RMSD gradient by the Low-Mod MD method. The lowest energy conformer of each molecule was chosen to be docked into central cavity of the hERG1 ion channel in its open state with GOLD docking program. The flow chart for the whole VS protocol is summarized in Figure 2.2. For structure-based virtual screening, triangle matcher placement was used to generate 30 poses for each ligand combined with London dG rescoring function using MOE software. hERG open state conformation for the virtual screening protocol was used. Around 900 molecules were then selected (according to their relative low London dG energy of binding) for further docking simulations in the GOLD program at the end of the virtual screening step. The statistical analysis of the VS results in terms of binding energy prediction is shown in Figure 2.3.

2.2.3 MD simulations and post-processing MD analyses

MD simulations were performed for selected compounds by Gromacs 4.6.5 package [32].

2.2.4 MD simulations of the target receptor: PDE5 in its apo state and bound with its inhibitors

Energy minimizations of the systems were conducted by steepest descent (SD) integrator using an initial step size of 0.01 nm with 10.000 iterations and 1000 kJ/mol.nm minimum force tolerance. 5 ns position restrained dynamics in NPT ensemble with constraints on all bonds were applied prior to production MD stage. 50 ns production MD simulations were realized. 2 fs time-step was used for the simulations. Electrostatic interactions were calculated with Particle Mesh Ewald technique with a cut-off distance of 9.0 Å and van der Waals interactions were considered with a cut-off distance of 14.0 Å. Berendsen thermostat and Berendsen barostat were used as temperature and pressure coupling algorithms. The Linear Constraint method (LINCS) was used in order to fix all the bond lengths. Gromos43A1 force field was used for the simulations with leap-frog algorithm. Ligand topologies were prepared with the PRODRG Server [33]. Simulations were

8 conducted in cubic simulation boxes solvated with SPC waters and neutralized with Cl− ions.

Figure 2.2 : Flowchart of the current study in designing the novel sildenafil-like molecules based on fragment replacement strategy in order to obtain promising molecules whose binding affinity to hERG K+ ion channel is decreased as the pharmacological activity against their therapeutic receptors, PDE5, is retained.

2.2.5 MD simulations of hERG K+ ion channel: Apo state and bound with PDE5 inhibitors

S5–S6 domains of the hERG K+ ion channel were considered in MD simulations. CHARMM-GUI Service [34] was used for the preparation of the protein-membrane systems. hERG1 channels in its open state and open-inactivated state were inserted into lipid bilayer membrane that contains DPPC-type lipid molecules. After solvation of the systems with TIP3P waters and adding neutralizing ions, systems energy were minimized using steepest descent algorithm with Verlet cut-off scheme. 5000 iteration steps and 1000 kJ/mol.nm minimum force tolerance were used for this step. LINCS algorithm was used in order to put constraints on hydrogen bonds. The Particle Mesh Ewald method was used in order to treat the long-range electrostatics

9 with a cutoff distance of 12 Å. van der Waals interactions were also considered within 12 Å distance. Systems were equilibrated with six steps of default equilibration scheme, provided by CHARMM-GUI, using leap-frog integrator with Verlet cut-off scheme. Berendsen thermostat and Berendsen barostat were used during the equilibration period for the temperature and pressure control. Nose- Hoover and Parrinello-Rahman coupling algorithms were utilized in order to maintain 310 K and 1 atm simulation conditions, respectively, during the 50 ns production stages. CHARMM36 force field was used for the simulations.

Representative frames of the simulations were generated from the last 10 ns trajectories for the hERG/sildenafil, PDE5/sildenafil, and hERG1/sildenafil_frag21, PDE5/sildenafil_frag21 bounded systems and further subjected to MM-GBSA free energy calculations with Prime program [35,36].

Figure 2.3 : Virtual screening analysis. Binding energy values are plotted against the frequency (number of molecules in the corresponding energy range). Values are obtained at the end of the VS protocol via docking the sildenafil analogs against hERG channel in its open state by means of MOE software.

10 2.2.6 Molecular Mechanics/Generalized Born surface area (MM/GBSA) calculations

Protein-ligand free binding energies of the selected molecules were estimated using the MM-GBSA method, implemented to estimate the free energy of binding (ΔGbind) in Prime module of the Schrodinger’s molecular modeling package [35,36] based on the MD trajectory frames. Prime uses the VSGB 2.0 solvation model and the OPLS2005 force field to simulate the interactions. Trajectories derived from last 10 ns MD simulations are used for MM/GBSA calculations. For this aim, representative structures (i.e. the trajectory frame that has the smallest RMSD to the average structure) were used in calculations.

2.3 Results and Discussion

The discussion is mainly organized into two parts: First part involves analysis of the interactions of sildenafil, vardenafil, and tadalafil with the hERG1 channel in its open and open-inactivated states via three different docking programs. Second part focuses on structure-based virtual screening and involves an effort to discover novel sildenafil-like molecules as promising new drug candidates with less side effects (i.e. reduced hERG channel blocking affinity). We have used our previously published and experimentally partially validated homology models of hERG1 channel for this study [37] (Figure 2.4). The 3D models of hERG1 potassium ion channel models in open, open-inactivated, and closed states were generated by our group previously using Rosetta De Novo protein designing suite and further refined with all-atom MD simulations. hERG1 (KCNH2 or Kv11.1) is the name of a gene that encodes the α-subunit of a voltage-gated potassium channel. hERG1 channel is a tetramer consisting of four identical subunits where each subunit has six (S1–S6) transmembrane (TM) alpha helices [38]. S5 and S6 helices form the channel inner pore cavity (i.e. pore domain- PD) whereas the S1-S4 helices constitute the voltage-sensing domain of the channel. The channel has a selectivity filter region (i.e. S624, V625, G626, F627, G628) formed by a peptide linker between the S5 and S6 which selectively permeates the potassium cations across the cell membrane. Since the discovery of the drug-induced arrythmias caused by the blockage of hERG, there has been a great effort to resolve

11 the 3D structure of the channel and to understand the key components of drug and hERG channel interactions. Patch clamp electrophysiology techniques have been carried out to assess “hERG blocking activity” and it provides valuable data on drug- induced inhibition of hERG current, particularly in mammalian cells. The inhibition profiles of some known PDE inhibitors (i.e. vinpocetine, EHNA, vesnarinone, amrinone, milrinone, rolipram, dipyridamole and zaprinast) were examined with patch clamp experiments by Yunomae et al. [39]. They found that vinpocetine and vesnarinone markedly decreased the hERG current with IC50 values of 0.13 and 20.6 μM, respectively, at comparatively low concentrations [39]. Alanine scanning mutagenesis techniques have been widely used to identify the binding specificity of various channel blockers [40-49]. These studies have suggested many important residue-ligand interaction sites, so far. The consensus for hERG affinity generally suggests that Tyr652 and Phe656 residues that are located on the S6 helix are the most important blocking elements of hERG channel by their capability of several types of π-π stacking, cation-π, and H-bonding interactions. As a result of in silico studies, various pharmacophore requirements for hERG blockers could also have been emerged [50-52]. Generally, hydrophobic groups will make interactions with some specific residues (particularly, Tyr652 and Phe656 located at the bottom of the central cavity of the channel), a basic nitrogen at the center of the drug and multiple π-π stacking and hydrogen bonding interactions govern the high hERG blocking ability. On the other hand, molecular modeling techniques such as protein-ligand docking have been utilized during the hERG assays in order to overcome the hERG affinity via making substituent refinements on the lead compounds. An example is the CCR5 antagonist-maraviroc where in the course of its development stage, same procedure has been applied [53]. Molecular docking is actually one of the important basic tools in the structure-based field for the recognition of the inhibition of the hERG current by direct blockage of the channel [37,49,53,54-64].

Figure 2.4 : 3D structures of homology models of hERG K+ ion channels used in the current study. S1-S6 and S5-S6 domains are shown for the open state (a and b, left side) and open-inactivated state of the channel (a and b, right side), respectively.

12 2.3.1 Analysis of the key interactions of sildenafil with the target receptor (PDE5)

Sildenafil was docked into the drug-binding cavity of PDE5 using GOLD, MOE, and AUTODOCK docking tools. The superimposition of sildenafil conformers predicted by each docking runs was shown in Figure 2.5. All of the used docking programs predicted the hydrogen bonding interactions that are formed between amine hydrogen (–NH) and carbonyl oxygen (–C=O) groups of pyrazolopyrimidinone of sildenafil and Gln817 residue. GOLD and MOE top docking poses show that pyrazolopyrimidinone group of sildenafil has π-π stacking interactions with Phe820 and a π-H interaction is established by Val782 which is in good agreement with the crystal structure [29]. However, AUTODOCK top docking pose of the sildenafil interestingly shows hydrogen bonding interactions between i) Met816 and Phe820 with O=S=O functional group; ii) Leu765 and pyrazole moiety; and iii) Met816 and ethoxyphenyl group due to a slight shift in the predicted binding pose. Additionally, when partial flexibility is given to the enzyme at the H-loop site, a hydrogen bonding interaction was observed between the NH group of the methyl piperazine ring and Ser663 residue according to top docking pose of GOLD program, which is in good agreement with the experimentally determined bioactive conformation [29]. However, this interaction is absent in the other crystal structures of sildenafil complexed with PDE5 [8,65]. These differences in the binding orientation of sildenafil have been attributed to the chimerically hybridized mutant [65] and inactive form of the PDE5 enzyme [8] used in the assays. Besides, docking calculations herein show the accessibility of various conformational states that sildenafil may undergo around the single bond between sulfur and carbon that can be induced by the movements of the H-loop residues. Moreover, in addition to H-loop, when flexibility is given to the some selected aminoacids at the binding cavity in GOLD docking program (i.e. Gln817, Ile778, Phe786, Met816, and Phe820 residues), a hydrogen bonding interaction was formed between the oxygen which is attached to the ethoxyphenyl group of sildenafil and Gln817 residue with a distance of 1.82 Å. In this orientation, a hydrogen bonding interaction between the Nε2 of Gln817 and sildenafil is vanished and the hydrogen bonding interaction between Oε1 of Gln817 and -NH group of pyrazolopyrimidinone in sildenafil is conserved within a distance of 2.10 Å. However, the hydrogen bond network in the protein (Gln817-

13 Gln775-Ala767) is altered in this orientation, i.e. the hydrogen bonding interaction between Gln817 and Gln775 has disappeared.

Figure 2.5 : (A) 3D ligand interactions diagram of the sildenafil generated with GOLD program using rigid target treatment. Bidentate hydrogen bondings interactions with the invariant Gln817 residue are shown with red-dashed lines. (B) Superposition of sildenafil orientations obtained with different docking programs is shown: White color represents the rigid receptor treatment; turquaz color represents the flexible receptor treatment; brown color represents the flexible residue treatment on the H-loop domain, obtained with GOLD); green color represents the MOE program binding pose prediction and yellow color represents the AUTODOCK program binding pose prediction. The coloring scheme is based on carbon atoms. (C) The group definition of sildenafil, used in the current work.

The docking results of GOLD program generally imply a consensus that both

Chemscore, Goldscore and dGbinding values show higher scores in the case of rigid docking treatment or when partial flexibility is given only to the H-loop site. At this point, we would like to highlight an issue referred as self-docking and cross-docking [66] in the evaluation of reproducing crystallographic pose of the ligands in the molecular docking simulations. Since a receptor adopts a conformation induced by the electrostatic and hydrophobic environment created by ligand binding at the active site, it is sometimes not straightforward to obtain the bioactive conformation of other structurally diverse ligands when they are docked into the target structure. In the case of docking of tadalafil to PDE5, hereby, the crucial interactions at the active site of the enzyme could be reproduced only if the Q and P pocket residues were allowed to adopt side chain flexibility handled by the GOLD program.

14 The truncated moiety of sildenafil (R1 in Figure 2.5) is a highly flexible part of the molecule due to low rotational barrier around the single bond between sulfur and carbon atoms. There are three different orientations of these groups in the three different crystal structures obtained by different research groups: Sung et al. [8], Wang et al. [29], and Zhang et al. [65]. This moiety (R1) has clearly less significant role in stabilization of sildenafil at the active site of PDE5 as the docking calculations have shown. Actually, this sulfonamid moiety has been linked at the - para position of the ethoxyphenyl group (Ph-R2 in Figure 2.5), mostly with a goal of decreasing the lipophilicity of the molecule rather than increasing the binding affinity during the development stage of sildenafil [67]. On the other hand, the observed interaction of the ethoxy group of sildenafil and Gln817 via hydrogen bonding according to the GOLD-flexible pose may suggest an explanation to the higher affinity of this ethoxy substituted hit as compared to the others in a series of the several substituted species at the same position in the same study [67].

The overall docking results for binding energy predictions for sildenafil, vardenafil, and tadalafil for the principal target (PDE5) and off-target hERG1 channel are given in Table 2.1.

2.3.2 Comparison of used docking tools in terms of predicting the binding positions of “Sildenafil” in the central cavities of hERG1 channel

Detailed analysis of the docking outputs can serve as a valuable tool in rational drug design studies as the diversity of docking solutions in the pose prediction and the trend in binding energy predictions can suggest various important key residues and fragments of the drug that are responsible for high/low binding affinity. Especially, ambiguous picture of the drug binding poses in the hERG channel may be clarified at some level by comparative analysis of the results. In this study, we aim to gain both a consensus at some level with regard to hERG1 channel-PDE5 inhibitors interactions by means of using three docking tools and also benefit from the alternative outcomes of the various scoring functions having different algorithms utilized by three different software.

15 2.3.2.1 GOLD

Although the top 10 docking poses of GOLD program give diverse docking solutions for open-state (OS) and open-inactivated-state (OIS) of hERG1 channel, they generally share some common interactions at the pore domain (PD) of hERG1 channel as follows:

Open conformational state of hERG1

50% of docking poses show intermolecular hydrogen bonds, which were observed between the sulfonamid oxygen of the sildenafil with several residues (Thr623, Ser624, Ser649, Ala653, Ser660 and Phe656). The most populated interaction within the top docking pose was established by methylpiperazine moiety of the sildenafil via hydrogen bonding interactions. Met554, Thr623, Ser624, and Ser649 residues are involved in these interactions. In the top docking pose, this moiety also makes a π- cation interaction with Phe656 residue. The pyrazolopyrimidinone group of sildenafil that is responsible for the occurrence of invariant bidentate hydrogen bonding interactions in the Glutamine switch pocket, which mainly stabilize sildenafil and also vardenafil by mimicking the substrate (cGMP) binding to the PDE5 receptor, contribute to three hydrogen bonding interactions (top docking pose, 5th best pose and 10th best pose). Tyr652 and Phe 656 residues are involved in π-π stacking interactions with either ethoxyphenyl or pyrazole groups. Hydrogen bonding interactions between the pyrazole group and Leu622, Ser624 and Ser649 residues are observed in only two of the docking solutions. The top docking pose of sildenafil shows that the ligand is surrounded by polar residues (i.e. Ser649, Ser660, Ser624, Tyr652, Thr623) and hydrophobic residues (i.e. Phe656, Leu650, Ala653, and Leu622) (Figure 2.6). One of the sulfonamid oxygen of sildenafil contributes to a hydrogen bonding interaction with the side chain of Ser649 at a distance of 1.71 Å and there is also a π-cation interaction between the basic nitrogen of the piperazine ring and the phenyl ring of the Phe656 with a distance of 3.71 Å. The top docking solution gives a complex with predicted interaction energy of −7.37 kcal/mol. It must be noted that the top 10-docking poses indicate no direct interaction(s) between the ethoxyphenyl group of sildenafil and hERG1 channel PD residues.

16 Open-inactivated conformational state of hERG1

The binding free energy differences between the top docking poses of sildenafil in the open and open-inactivated states show tighter interactions of the ligand at the open-inactivated states (ΔΔGbinding = 3.23 kcal/mol). A similar result is observed for top-10 docking solutions, (ΔΔGbinding = 3.30 kcal/mol; based on average docking score results). It must be noted that the experimental observations also show that the binding affinity of a well-known hERG1 inhibitor dofetilide binds tightly to the hERG1 open-inactivated state compared to its open state. The docking results of sildenafil in the open and open-inactivated states clearly show that sildenafil has better binding affinity to the hERG channel in its open-inactivated rather than its open-conformational-state (Table 2.1).

The average predicted binding free energies of the ligands within the top-10 docking poses are calculated by means of the Botzmann probability distribution. As Table 2.2 clearly demonstrates, all the defined fragments make hydrogen bonding interactions with several residues. Similar to the open-state, the same groups of sildenafil (methylpiperazine and sulfonamid) are the dominant groups that are involved in H-bonding. Besides the van der Waals interactions, the relatively compact nature of the central cavity as compared to the open-state conformation makes the open- inactivated state residues more prone to form specific interactions both in terms of quantity and concurrence. For example, in the top docking pose, the proximity of each chain within the tetramer makes it possible for sildenafil to give rise to three effective H-bond interactions within two different chains. The calculated distances are found as 2.25 Å between Ser649 and sulfone oxygen; 1.70 Å between Ser624 and hydrogen atom of amide group of the pyrozolopyrimidinone; 1.63 Å between the proton of the sidechain of Ser649 and aromatic nitrogen of the pyrazole group. The geometry of the residues enables also π-π interaction between Tyr652 and ethoxyphenyl fragment. It is also noteworthy to mention that the stabilization of sildenafil is governed along the whole structural fragments by these interactions except the piperazine ring. The relatively high number of rotatable groups (i.e. 7 rotatable groups) makes sildenafil molecule very flexible and enables it to adopt various conformational degrees of freedom prior to stabilization in the binding pockets.

17 Table 2.1 : Comparison of different docking programs: GOLD, MOE and AutoDock (“os” and “ois” stands for open-state and open-inactivated-state of hERG K+ ion channel, respectively. Binding Energies are given in kcal/mol).

MOE DOCKING TOOLS GOLD London dG AUTODOCK Chemscore.dGbinding //GBVI/WSA dG dGbinding Sildenafil (PDE5) -10.36 -9.80//-9.41 -11.23 Sildenafil (hERG-os) -7.37 -9.50//-7.68 -7.48 Sildenafil (hERG-ois) -10.60 -8.97//-9.84 -10.16 Vardenafil (PDE5) -9.67 -10.71//-9.89 -11.3 Vardenafil (hERG-os) -7.30 -9.54//-7.80 -8.28 Vardenafil (hERG-ois) -10.76 -8.72//-9.43 -9.68 Tadalafil (PDE5) -9.57 -10.18//-6.90 -11.17 Tadalafil (hERG-os) -6.99 -9.88//-6.15 -9.45 Tadalafil (hERG-ois) -10.60 -9.19//-6.96 -11.75

Figure 2.6 : 2D protein-ligand interaction diagrams of top docking poses produced by three different programs are represented. Polar and hydrophobic resiues are circled in purple and green colors, respectively. Basic residues are shown with purple color with an exterior blue circle. Green-dashed arrow represents hydrogen bonding interaction between ligand and the side chain of a residue. Solvent-accessible surface area for ligand and receptor is shown with blue smudge and turquoise halo, respectively. Arene-arene and arene-hydrogen bonding interactions are shown with green colors. 3D protein–ligand interaction diagrams are also shown.

Figure 2.6 (continued) : 2D protein-ligand interaction diagrams of top docking poses produced by three different programs are represented. Polar and hydrophobic resiues are circled in purple and green colors, respectively. Basic residues are shown with purple color with an exterior blue circle. Green-dashed arrow represents hydrogen bonding interaction between ligand and the side chain of a residue. Solvent-accessible surface area for ligand and receptor is shown with blue smudge and turquoise halo, respectively. Arene-arene and arene-hydrogen bonding interactions are shown with green colors. 3D protein-ligand interaction diagrams are also shown.

The hydrogen bonding and stacking interaction capabilities get along with the promiscuous binding properties in the inner cavity residues of hERG channel in

20 terms of binding tendency and energy. This may partly explain the higher binding energy trend of sildenafil and structurally similar drug vardenafil (for the open state of the hERG channel) as compared to smaller and rigid molecule, tadalafil, predicted by the top docking solutions of GOLD and MOE programs (Table 2.1). The orientation of sildenafil (top pose) in the open-inactivated state differs from the orientation in the open state of the hERG channel in a way that it is elongated horizontally across the channel axis where it covers more volume at the central part of the inner cavity.

2.3.2.2 AutoDock

Open conformational state of hERG1

The cluster analysis of the docking conformations shows that there is a wide distribution of ligand poses that spread around the central cavity of hERG1 ion channel rather than localizing at one specific region of the channel. It is also noteworthy to mention that the predicted binding energy of the docking solutions in various clusters is very close to each other although they form different interaction patterns with the channel, which show that there are possible different binding modes of the ligands that yield very close predicted interaction energies. The superimposition of the top-10 docking solutions and their predicted binding free energies are given in order to illustrate the situation (Figure 2.7). Top-ranked docking pose gives a predicted binding energy of -7.48 kcal/mol where sildenafil is in close contact with Phe656, Ser649, Met645, Leu622, Tyr652, and Thr623 residues. In addition, AutoDock program was not able construct any H-bonding or π- π stacking interactions between the ligand and the target structure.

Open-inactivated conformational state of hERG1

A cluster analysis gives a similar situation that only three clusters have been formed with a mean binding energy of −5.16, −4.34, and −4.61 kcal/mol for the hERG1 channel in its open-inactivated state. However, there is a significant binding energy difference among the first-10 poses.

Table 2.2 : Summary of interactions of sildenafil with the hERG channel. (Top 10 docking poses of GOLD program are considered in generating the specific interactions, Gray Area: hERG-Open-State, White Area: hERG-Open-Inactivated-State) Fragment Definition Number of Interacting Residues Number of Interacting Number of Interacting of Sildenafil H-bonding π-cation Residues π-π Residues Interactions Interactions Interactions sulfonamide 7 THR623,SER624, - - - - SER649,ALA653,PHE656, SER660 methylpiperazine 10 MET554,THR623, 1 PHE656 - - SER624,SER649 pyrazolopyrimidinone 3 THR623,SER624, - - 4 PHE656 SER654 ethoxyphenyl - - - - 1 TYR652 pyrazole 3 LEU622,THR623, SER649 - - 4 PHE656 sulfonamide 7 THR623,SER624, - - - - TYR652 methylpiperazine 8 LEU622,SER624, 2 TYR652, - - MET645, PHE656 TYR652,ALA653 pyrazolopyrimidinone 4 SER624,SER649, PHE656 2 TYR652 2 TYR652 ethoxyphenyl - - - - 4 TYR652, PHE656 pyrazole 5 LEU622,SER624, SER649 2 SER624, 2 TYR652 TYR652

Figure 2.7 : Superpositions of the top ten docking solutions for sildenafil orientations at the central cavity of hERG1 (OS), generated by AutoDock and their corresponding binding energy predictions.

First three docking poses scores are given as: -10.46, -8.30, and -8.28 kcal/mol, respectively. Focusing on the top docking pose, ethoxy oxygen and Ser649 hydrogen make a H-bonding interaction with a distance of 2.05 Å where no such contribution was observed from the GOLD program docking results for both open and open- inactivated states. The best energetic pose according to the AutoDock program suggests that van der Waals interactions effectively contribute in stabilizing sildenafil at the central cavity of the channel in its open-inactivated state rather than specific interactions such that strong hydrogen bonding, π-π stacking, and π-cation interactions contrary to the GOLD docking program binding predictions.

Overall, comprehensive and detailed analysis of the docking solutions leads to suggestions that the ability of hERG K+ ion channel to accommodate the drugs at the pore domain with a widespread conformations is more likely when it is in the open- state, possibly due to its larger cavity volume compared to the open inactivated state. (i.e. surface area differences between open and open-inactivated states showed that open state has around 1.4-fold larger area compared to open-inactivated state. (Surface area is defined by within 10 Å proximity of Tyr652 and Phe656 residues).

23 2.3.2.3 MOE

We have adopted induced fit docking (IFD) protocol through the docking simulations via MOE software with the related PDE5 inhibitors in the current work. In this docking scheme, the final ligand pose prediction and its corresponding binding energy score are achieved by means of minimizing the ligand conformation and the residues at the defined binding region subsequent to ligand placement stage with the user defined alternative side chain treatments. In the tethering option utilized for this study, a constant weight factor of 10 is applied on each pocket residues that fall into 6 Å cut-off value within the defined binding pocket in order to put restrictions on the movements of side chains during the force field-based refinement step. A geometrical distance restriction was put on the atoms of the receptor by taking into account the various energetic terms that raise in consequence of the angle or dihedral shifts in order to incorporate these restrictions into the force-field-based energy function for the ultimate binding energy and pose prediction ranking.

Open conformational state of hERG1

The common interactions between the top ranked pose of MOE and GOLD were the one with the cation-π interaction between Phe656 and the amide nitrogen of the piperazine ring of sildenafil in the open-state conformation of hERG1. In addition, another aromatic interaction is established with the pyrazole ring of the sildenafil and the Tyr652 predicted by MOE docking pose for the open state conformation of hERG. The docking protocol applied here performs well in identifying various H- bonding interactions between the donor and acceptor groups within the protein and ligand and also various π-π stacking, π-cation, or π-H interactions between the aromatic moieties of sildenafil and hERG channel by checking at the top ten poses. The final scoring function (GBVI/WSA dG) estimates the free energy of binding of sildenafil to the open state of the channel with a 2.16 kcal/mol lower energy compared to the open-inactivated state, which is in agreement with the AUTODOCK and GOLD results in terms of energy trend.

Open-inactivated conformational state of hERG1

Similar to the previously described comparative interaction patterns, MOE docking results suggest also more specific interactions in open-inactivated state of the channel for sildenafil (Figure 2.6). Leu622 backbone atom (oxygen) forms H-bonding

24 interaction with one of the hydrogen atoms of the methyl group of piperazine fragment. Another H-bonding interaction occurs (i) between Ser624 and hydrogen atom of -NH on the pyrazolopyrimidinone group (2.23 Å); (ii) Ser649 hydrogen and the pyrazole group sp2 nitrogen within a distance of 2.35 Å (Figure 2.6). In addition, a π-H interaction occurs between Thr623 and the pyrazole ring. Results for the binding energy difference among the 100 poses are much more significant in the docking simulations for the open inactivated hERG1 model compared to the open- state hERG1 model.

2.3.3 In silico Alanine mutagenesis study

Single point Alanine mutagenesis study was performed with the GOLD program in order to understand the specific contributions of the residues to the binding affinity of sildenafil to the hERG1 channel. 17 residues from binding site cavity were selected and they were mutated with Alanine one by one and docking simulations were then performed with each individual mutant receptor and their corresponding docking scores were compared with the wild type (Table 2.3). The most significant binding energy attenuation is obtained for Phe656Ala mutation for the open-state hERG1 channel. The predicted binding energy of sildenafil was reduced from -7.37 to -5.94 kcal/mol which means a 19.4% of reduction in the docking score (Table 2.3). On the other hand, the channel in open-inactivated state is more sensible to single point mutations in terms of reduction in the docking scores. The predicted binding energy dramatically changes from -10.60 to -7.17 kcal/mol in the case of Tyr652Ala mutated channel for the open-inactivated state. The other significant attenuations are attained with the hERG (Met554Ala) and hERG (Ser624Ala) mutants. These results are consistent with the top docking poses (Figure 2.6) and also the number of specific interactions of these residues that are capable of forming (Table 2.2). Alanine scanning studies have also remarked the importance of Tyr652 and Phe656 aminoacid residues located on the pore domain in the context of direct blockage of the channels with various drugs (such as cisapride, maprotiline, dofetilide, terfenadine, quinidine, miconazole, etc.) [32,55-58,62-64,68].

Table 2.3 : In silico Alanine Mutagenesis Results. The binding free energies (ΔGbinding) are expressed in kcal/mol. The results are calculated according to the top-ranking pose generated with GOLD program.

Open- Open- Open-inactivated- Open-inactivated- state state state state

hERG dGbinding % Reduction dGbinding % Reduction Channel in Binding in Binding Residues Affinity Affinity 554A -7.71 - -9.29 12.3 %

621A -7.74 - -10.53 0.66 % 622A -8.28 - -10.13 4.43 % 623A -7.71 - -10.42 1.69 % 624A -8.00 - -9.52 10.18 % 625A -7.68 - -9.81 7.45 % 644A -8.22 - -10.82 - 645A -7.73 - -10.05 5.18 % 646A -8.27 - -10.08 4.90 % 648A -7.46 - -10.36 2.26 % 649A -7.39 - -10.63 - 650A -7.04 4.47 % -9.90 6.6 % 651A -8.07 - -11.67 - 652A -7.23 1.89 % -7.17 32.3 % 656A -5.94 19.4 % -10.28 3.01 % 659A -8.16 - -10.00 5.66 % 660A -8.80 - -10.61 - 664A -8.01 - -10.75 - Wild_Type -7.37 -10.60

26 2.3.4 General statements on the binding energy predictions derived from GOLD, AutoDock and MOE

As Table 2.1 demonstrates, all of the programs predict a lower binding energy (i.e. higher docking scores) value for sildenafil, vardenafil, and tadalafil when they are in the open-inactivated state of the channel as compared to open state. GOLD and MOE binding energy prediction trends of tadalafil are also in good agreement with the experimentally determined IC50 values of these drugs (Figure 2.1) where tadalafil has a higher IC50 value compared to both sildenafil and vardenafil. GOLD program results give the most significant binding energy differences (around 3 kcal/mol) of the three drugs within their top docking pose for the open and open-inactivated state. GBVI/WSA dG scoring function seems to refine the energetics of binding energy predictions of the default scoring scheme, London dG, when the results are evaluated in terms of predicting the higher IC50 value of tadalafil than the other two drugs.

2.3.5 Binding interactions of Vardenafil and Tadalafil with the hERG K+ ion channel

The 3D interaction diagrams of vardenafil and tadalafil inside the hERG1 K channel are shown in Figure 2.8. As the top docking poses for open state predicted by GOLD program suggest, vardenafil forms various effective H-bonding interactions with the pore domain residues of hERG1 channel. Ser649 and Ser624 are capable of forming five H-bonding interactions with sulfone group, piperazine ring, and pyrimidinone oxygen. On the other hand, there are three H-bonds formed in open-inactivated state between the Ser649 and basic nitrogen at the piperazine ring; between the Ser649 and pyrazole nitrogen; and between the pyrimidinone oxygen and Ser624. Besides the H-bonding interactions with the polar residues (Ser624, Ser649, particularly) non-polar interactions (i.e. van der Waals forces) seem to play crucial role in stabilizing vardenafil in the open-inactivated state (Table 2.1). Contrary to sildenafil, vardenafil has no π-π stacking or cation-π interactions between Phe656 and Tyr652. Although the molecular structure of vardenafil differs only slightly from sildenafil, the differences in binding modes with sildenafil (Figure 2.6) demonstrate how small modifications may have an effect on the binding modes in the hERG1 channel. However, their common stabilization interactions resemble each other when

27 compared to tadalafil, such that vardenafil is also capable of forming concurrent stacking and hydrogen bonding interactions with Thr623, Ser624, Ser649, Ser660, Tyr652, and Phe656 as observed from the top docking solutions generated by GOLD. Actually, vardenafil binding pose is very similar to sildenafil within their target receptor (PDE5). As GOLD docking program suggests vardenafil also makes bidentate hydrogen bonding interactions with the invariant Gln817 and the pyrimidonone ring stacks with Phe820 residue (in addition to Phe786 stacking of the ethoxy ring in the rigid receptor treatment). On the other hand, tadalafil constructs only one hydrogen bonding interaction between the oxygen of the Gln817 and similarly Phe820 stacks the aromatic moiety of the drug at the same aromatic position. The molecular basis for tadalafil having weaker binding activity than sildenafil or vardenafil against hERG1 channel may be grounded on its rigid and relatively small aromatic ring structure. A close look at its interactions with the hERG1 channel (Figure 2.8) shows that tadalafil makes close contact with Thr623, Ser649, Tyr652, Ala653, and Phe656 residues and it only forms a face-to-face π-π stacking interactions with Phe656 residue with a distance of 3.16 Å based on the center of the phenyl rings. Tadalafil binds more strongly in the open-inactivated state of the channel (dGbinding = −10.60 kcal/mol). An amphiphatic residue, Tyr652, is the key element for this predicted high binding affinity (Figure 2.8). Tyr652 residues, located at the two different chains, make a H-bonding interaction between either its side chain with the piperazinedione oxygen or its backbone oxygen with the hydrogen of -NH group on the indole ring of tadalafil. In addition, side chain of Tyr652 ring participates in π-π stacking interaction with a distance of 3.81 Å between the five membered ring indole moiety. The chiral carbon center with R configurations at the piperazinedione ring and also the aromatic ring adjacent to the piperazinedione and indole aromatic moities causes a shift in the orientation of Tyr652 residue and prevents a more parallel face-to-face interaction to avoid steric clashes between the protons of the referred aromatic rings of the ligand and the Tyr652 side chains. The presence of aromatic groups in the tadalafil enables various π-π, cation-π interactions with Phe656. Ser624 is the second important residue that stabilizes tadalafil through hydrogen bonding interactions in the open state of hERG1, observed in the top 10 docking poses. On the other hand, Tyr652 residue is siginificant in the context of various hydrogen bonds, π-π stacking, and cation-π interactions in the open-inactivated state of hERG1. The heterogenity in the binding

28 modes of many known hERG1 blockers in the pore domain of hERG1 channel addresses the difficulty of predicting the right docking pose for a drug, especially in the formidable cases where the crystal structure of a protein does not exist.

Figure 2.8 : A schematic view of the top docking poses of vardenafil and tadalafil at the central cavity of hERG1 channel. Ribbon representations and side chains of the channel are covered within the 2.5 Å space around the ligand. Hydrogen bonds are shown with red dotted line with their distances. Left and right columns represent the interactions with the open state and open-inactivated state of the hERG1 channel, respectively.

2.3.6 Virtual screening results

Various substitutions have been made at the R1 position (Figure 2.4) of sildenafil. The most promising molecules are compiled with their docking scores in Table 2.4 and their physicochemical properties are shown in Table 2.5.

29 2.3.7 MM/GBSA analyses

Selected compounds from molecular docking simulations (sildenafil and sildenafil_frag21) were used in MD simulations. Comparison of RMSD graphs of sildenafil and sildenafil_frag21 molecules at the PDE5 and hERG1 open and open- inactivated state targets is represented in Figure 2.9. RMSD results show similar patterns of both drugs and in all systems, RMSDs are converged after 20 ns production runs. Binding energies of these ligands are calculated by MM/GBSA analyses and Table 2.6 summarizes derived results. Although PDE5 binding results looks similar (slightly lower compared to sildenafil), hERG1 binding profile of sildenafil is higher than sildenafil_frag21 (especially at the open-state hERG1 channel). This can be interpreted as although sildenafil_ frag21molecule has similar predicted binding affinity with sildenafil at the PDE5 principal target, its predicted binding affinity to the hERG1 channel is lower compared to sildenafil. Thus, it can be considered as sildenafil_frag21 is a safe compound compared to sildenafil.

2.4 Conclusions

In summary, three PDE5 inhibitors (sildenafil, vardenafil and tadalafil) were screened for both principal (PDE5) and off-targets (hERG1 open and open- inactivated states) using three different docking programs (GOLD, AutoDock, and MOE). Based on detailed analyses of docking poses and predicted interaction energies, novel analogs of PDE5 inhibitors with lower predicted binding affinity to hERG channels without loosing their principal target activity were proposed. Detailed docking analysis as well as MD simulations helped us to better understand the PDE5 inhibitor-target binding interactions in the atomic level. Results of this study can be useful for designing of novel and safe PDE5 inhibitors with enhanced activity and other tailored properties.

30 Table 2.4 : 2D structures and predicted binding free energies of the novel sildenafil analogs.

31 Table 2.4 (continued) : 2D structures and predicted binding free energies of the novel sildenafil analogs.

32 Table 2.4 (continued) : 2D structures and predicted binding free energies of the novel sildenafil analogs.

33 Table 2.5 : Physicochemical properties of the molecules, calculated by MOE.

Ligand Rsynth SlogP TPSA Mw (%) (Å2) (g/mol) Sildenafil_frag_9 1.00 4.06 102.65 452.55 Sildenafil_frag_15 1.00 4.08 107.28 497.59 Sildenafil_frag_17 1.00 3.92 126.90 456.59 Sildenafil_frag_19 1.00 4.36 113.07 457.59 Sildenafil_frag_21 0.80 3.49 109.43 409.49 Sildenafil_frag_26 1.00 3.86 105.81 454.57 Sildenafil_frag_29 1.00 2.75 111.88 424.45 Sildenafil_frag_30 0.80 4.61 80.87 423.56 Sildenafil_frag_31 0.74 3.55 94.81 444.47 Sildenafil_frag_33 1.00 3.06 120.47 482.56 ZINC_77293467 1.00 3.31 88.74 390.87 ZINC_77293463 1.00 3.31 88.74 390.87 ZINC_77285660 1.00 3.37 85.58 388.85 ZINC_18203107 1.00 0.92 146.35 479.55 ZINC_18044545 1.00 1.91 126.12 463.55 Sildenafil 1.00 0.47 110.33 475.59 Table 2.6 : Comparison of protein-ligand free energy results of Sildenafil and Frag_21 Molecules Using MM/GBSA Analyses.

System ΔGMM/GBSA (kcal/mol) hERG1-Open/Sildenafil -53.46 hERG1-Open-Inact./Sildenafil -81.29 hERG1-Open/Frag_21 -31.62 hERG1-Open-Inact./Frag_21 -80.60 PDE5/Sildenafil -93.96 PDE5/Frag_21 -89.74

Figure 2.9 : Comparison of RMSD graphs of sildenafil and sildenafil_frag21 molecules at the PDE5 and hERG1 open and open inactivated state targets.

Figure 2.9 (continued) : Comparison of RMSD graph of sildenafil and sildenafil_ frag21 molecules at the PDE5 and hERG1 open and open inactivated state targets.

3. INVESTIGATION OF PDE5/PDE6 AND PDE5/PDE11 SELECTIVE POTENT TADALAFIL-LIKE PDE5 INHIBITORS USING COMBINATION OF MOLECULAR MODELING APPROACHES, MOLECULAR FINGERPRINT-BASED VIRTUAL SCREENING PROTOCOLS AND STRUCTURE-BASED PHARMACOPHORE DEVELOPMENT2

3.1 Introduction

PDE type enzymes are located in multiple tissues and organs of vertebrate systems in mammalian organisms [69]. Their essential biological function is to regulate the cytoplasmic levels of intracellular second messengers, cGMP and/or cAMP [70–72]. 21 different genes promote 11 isoenzymes of the PDE family. Of these enzymes, PDE5 has attracted a special attention over the years after its recognition as being the target enzyme in treating ED where millions of men suffer from this disease, worldwide. Since the revolutionary discovery of Sildenafil (ViagraTM) by Pfizer in 1998 [67], academic studies as well as clinical and pre-clinical industrial drug discovery programs have been widely carried out for more potent and selective PDE5 inhibitors. Although successful examples exist in the market, such as Tadalafil-CialisTM, Vardenafil-LevitraTM and Avanafil-StendraTM, main hurdle about PDE5 inhibitors has been focused on the circumvention of the undesired cross- reactivity with other PDE enzymes, especially towards PDE6 and PDE11 in the discovery and design studies [12, 13, 16, 19, 20, 73-90]. PDE5 enzymes can be found in diverse tissues in human body in addition to the corpus cavernosum; such as lung, brain, platelets, kidney and liver which raise the physiological importance of this enzyme and in turn leads PDE5 inhibitors as drugs for treating other diseases such as pulmonary hypertension (RevatioTM) and Raynaud’s disease (ViagraTM), as

2 This chapter is based on the article “Kayık, G., Tüzün, N. Ş., Durdagi, S. (2016), Investigation of PDE5/PDE6 and PDE5/PDE11 Selective Potent Tadalafil-like PDE5 Inhibitors Usıng Combination of Molecular Modeling Approaches, Molecular Fingerprint-Based Virtual Screening Protocols and Structure-Based Pharmacophore Development. Journal of Enzyme Inhibition and Medicinal Chemistry, 32, 311-330. doi: 10.1080/14756366.2016.1250756.

37 well. This convenient situation marks the possibility of evaluating PDE5 inhibitors in drug repositioning applications as sildenafil is presumably the one of the most important examples at this field.

Structural assembly of PDE5 is a homodimer that consists of two regulatory GAF domains (GAFA and GAFB) which are the allosteric binding regions for the enzyme substrate (cGMP), phosphorylation site (at Ser92 position) which takes role in the activation mechanism of the enzyme and catalytic site located at the C-terminal end of the protein (amino acid residues: 535-860) which contains the divalent metal (Zn2+ and possibly, Mg2+) binding domain. The mechanism of “PDE5 drug activity” in the corpus cavernosum starts with the competitive binding with cGMP at the active site of the enzyme, intervening the neurotransmitter nitric oxide (NO) mediated sexual stimulation which eventually reverts the smooth muscle contraction via depression of the intracellular Ca2+ concentration. On the other hand, PDE6 enzyme is the key effector enzyme for the phototransduction cascade in the rod and cone segments of the retina in the mammalian eyes. It has a function in visual transduction and respond to light via shifting mechanism from its inactivated to the activated states, regulated by its unique “γ-subunit” which is absent among other PDEs [91-95]. Due to the amino acid sequence and the secondary structural similarity of its catalytic domain with PDE5, first-generation PDE5 inhibitors (i.e., sildenafil and vardenafil) are also competitive inhibitors of PDE6. In addition, tadalafil has a better PDE5/PDE6 selectivity as compared to sildenafil and vardenafil [87, 96, 97]. Visual disorders such as functional blindness, blue (cynopsia) and blurred vision and enhanced light sensitivity have been attributed to the cross-reactivity with the PDE6 catalytic site, upon intake of PDE5 inhibitors in patients with ED. Apart from this reasonable basis for the foundation of vision-related side effects by consequence of direct inhibition of PDE6 which is in line with the location/function of PDE6; another hypothesis ascribed these side effects to the reactivity of these drugs targeting to PDE5 isoenzymes that are distributed in tissues other than the corpus cavernosum [98, 99]. In addition, tadalafil is a dual inhibitor of PDE5 and PDE11 enzymes which is thought to be the reason of back and muscle pain (myalgia) during the treatment of men with tadalafil since PDE11 enzymes are abundantly found in skeletal muscle cells [100, 101]. Although the catalytic site of PDE11 is the most similar one with PDE5, the absence of crystal structure together with the inadequate knowledge about

38 the physiological role of this enzyme in human body restrict the understanding of the mechanism of the inhibitory activity of this target.

Numerous scientific approaches appearing in medicinal chemistry area cover different experimental techniques along with computer-aided molecular modeling methods in the course of discovering more potent and selective PDE5 inhibitors. Especially, structure-activity relationship (SAR) studies are very helpful where distinct chemical synthesis routes guided by oral bioavailability, solubility, membrane permeability, toxicity, lipophilicity and other pharmacokinetic and physiochemical tests have given rise to novel potent compounds to emerge as PDE5 inhibitors [12, 13, 15, 19, 20, 75, 79-82, 84-88, 90, 102]. Scaffold hopping strategies have been implemented in designing structurally different-in terms of core architecture-novel compounds [16]. An unusual study was conducted by Pfizer in 2008, introducing chirality concept on sildenafil in order to improve PDE5/PDE6 selectivity [103]. On the other hand, in silico strategies including quantitative structure-activity relationship (QSAR) studies, pharmacophore hypothesis generations, virtual screening along with molecular docking and MD simulations have been emerged as popular techniques in identifying new leads and understanding the key concepts in protein–drug interactions [9, 10, 73, 83, 104-113].

Since the major challenge in designing novel PDE5 inhibitors is to decrease their cross-reactivity with PDE6 and PDE11, we attempt to identify potent tadalafil-like PDE5 inhibitors that have PDE5/PDE6 and PDE5/PDE11 selectivity as well as preserved principal target activity (i.e., comparable or higher binding affinity for PDE5 enzyme as compared to tadalafil). For this aim, similarity based virtual screening protocol is applied for the “clean drug-like subset of ZINC database” that contains more than 20 million small organic molecules. Another outcome of the present work is to illuminate the structural background for high/low binding tendencies to PDE5, PDE6 and PDE11 targets from molecular perspective. In this respect, we implemented step by step procedure in finding active hits as tadalafil-like compounds by combining ligand-based virtual screening (i.e., molecular fingerprint- based protocol) with structure-based modeling techniques (i.e., homology modeling, molecular docking and MD simulations). The flowchart of the current study is briefly illustrated in Figure 3.1. Since it is known that current PDE5 inhibitors block hERG1 K channels in concentration dependent manner, the cardiotoxicity prediction of the

39 hit molecules was also tested. Moreover, a novel approach for deriving structure- based pharmacophores (E-pharmacophore) was also applied for the selected hit compounds.

Figure 3.1 : Flowchart of the current study in the effort for identifying novel and selective PDE5 inhibitors.

40 3.2 Methods

3.2.1 Ligand and protein preparations

The crystal structure of PDE5 enzyme (PDB ID, 2H42) [29] was used as template protein for homology modeling and protein engineering procedures of catalytic domains of PDE6 and PDE11 targets. Protein preparations were handled with MOE molecular modeling package [28] after retrieving the bound-sildenafil and water molecules from PDE5 enzyme. Energy minimizations and conformational search (CS) of the ligands were realized with MMFF94X force field by means of MOE molecular modeling package. Low MD method was utilized in the CS step by generating 50 conformers for each molecule.

3.2.2 Virtual library screening

“Clean drug-like compound library” in the ZINC database (>20 million compounds) [114] was screened via molecular fingerprint and similarity search tools implemented in MOE. Molecular fingerprints of the molecules were computed with MACCS structural keys scheme. Subsequently, Tanimoto coefficient of each molecule in the database was calculated based on the characteristic MACCS structural keys of tadalafil. Tanimoto coefficient is defined as; TC=NAB/(NA+NB-NAB) where NAB represents the number of common MACCS structural key elements between molecule A and molecule B whereas NA and NB are the total number of MACCS structural key elements in molecule A and molecule B, respectively. MACCS structural keys include 166 structural elements that match the corresponding smart patterns in a compound. The applied similarity search strategy lies on an expectation that structurally similar molecules may also show similar biological activities [115, 116]. A coefficient value of 0.8 (Tanimoto coefficient) was chosen in the filtering step of the database which enabled us to obtain 1309 hit candidates that are further subjected to flexible molecular docking simulations.

3.2.3 Flexible molecular docking simulations

GOLD (v.5.3.0) program [30] was utilized for the docking simulations. ChemScore scoring function was used for generating the predicted binding energies of the

41 ligands within the targets. 10 amino acid residues at the drug binding region of PDE5, PDE6 and PDE11 targets were handled with flexibility by utilizing the rotamer library implemented in the GOLD docking program. Default settings were used for population and genetic operations steps. Early termination was switched-off and 20 docking poses were generated for each molecule.

3.2.4 Molecular Dynamics simulations

Gromacs v.4.6.5 package [32] was used for the MD simulations. Topology parameters for the ligands were prepared with PRODRG Server [33]. Partial charges of ligands were calculated with density functional theory (DFT) by 6-31G(d,p) basis set and B3LYP functional. Gaussian 09 program package [117] was used for this calculation. Production stages of the simulations were conducted in NPT at 310 K and 1 atm with periodic boundary conditions (PBC). GROMOS96 43A1 force field was used with leap-frog integrator. 2 fs time-step was used in simulations. SPC water model was used to solvate the systems in a cubic box and Na+ and Cl- ions were added to the systems as counter ions. Long-range electrostatic interactions were handled with PME algorithm. Energy minimizations with 5000 steps of steepest descent (SD) method followed by 5000 steps using conjugate gradient (CG) algorithm were performed for apo and holo state systems, using a threshold value of 100 kJ/mol.nm. Subsequently, two-steps restrained dynamics were applied in order to equilibrate the systems to the desired temperature and density: (i) systems were heated in NVT with a total of 0.1 ns simulations until the temperature reaches 310 K; (ii) then 2 ns NPT simulations were applied. Finally, 50 ns MD production runs were carried out without any restraints on atoms in an NPT ensemble. V-rescale temperature coupling scheme and Parrinello-Rahman pressure coupling scheme were used in order to control the temperature and pressure during the simulations, respectively. 5000 frames were collected through the 50 ns production runs for the post-processing MD analyses.

3.2.5 Molecular Mechanics Generalized Born Solvation (MM/GBSA) Calculations

Protein-ligand binding free energies of the selected hits as well as tadalafil were estimated using MM-GBSA method, implemented in Prime module of Schrodinger’s

42 molecular modeling package [35], based on the MD trajectory frames. Prime uses the VSGB 2.0 solvation model and the OPLS2005 force field to simulate the interactions.

3.3 Results and discussion

A molecular finger print-based virtual screening is performed for ZINC small molecules database in order to identify novel and potent PDE5 inhibitors. Molecular docking simulations were realized under “Protein H-bonding Constraint” (oxygen atom of the invariant Glutamine residues-Gln817, Gln773 and Gln869 in PDE5, PDE6 and PDE11, respectively-were enforced to participate in H-bonding interaction with any H-bond donor atom of the ligands) which augment the elimination of mis- docked outcomes. Thus, integration of ligand-based similarity screening protocol with constraint docking method was able to yield reasonable ligand orientations in the active site of the proteins and to analyze crucial active site residues-ligand interactions. Figure 3.2 points out that the dominant interaction to fulfill the receptor complimentary is the positively charged region (shown with blue meshes) around Gln817, Gln773 and Gln869 residues for PDE5, PDE6 and PDE11, respectively.

3.3.1 Validation of the docking methodology

Tadalafil is retrieved from its crystal-bound enzyme (PDB ID, 1XOZ) [118] and docked into the drug-binding region of PDE5, PDE6 and PDE11 targets by GOLD docking program. The ligand RMSD value yielded a value of <1.5 Å deviation (for PDE5) from its bioactive conformation at the crystal structure which supports the binding pose prediction power of docking methodology employed for the current study. Tadalafil exhibits monodentate hydrogen bonding interaction with invariant Glutamine residues of these enzymes as can be seen in Figure 3.3.

3.3.2 Constructing the homology models of the catalytic domains of PDE6 (amino acid residues: 482-816) and PDE11 (amino acid residues: 587-910)

In the absence of crystal structures of PDE6 and PDE11, homology models of the catalytic domains of these enzymes were constructed based on the available crystal

43 structure of PDE5 [29] as a template. This crystal structure (PDB ID, 2H42) was chosen as a template because it is not chimerically hybridized or mutated like the other PDE5 crystal structures [65, 118] in the literature and also it does not have any missing elements [8, 119]. The amino acid sequence of PDE6 (P16499) which belongs to the α-subunit of a human rod cell and PDE11 (Q9HCR9) were downloaded from Uniprot [120], respectively and further used in the sequence alignment procedure. MOE software was used in building the homology models with Amber99 forcefield. BLOSUM62 matrix was used in the sequence alignment step. 10 different homology models were built; the best homology model was selected among the intermediate models according to the Generalized Born/Volume Integral (GB/VI) methodology and further subjected to refinement and energy minimization procedures.

The RMSD values between PDE5 and the generated PDE6 and PDE11 are 0.46 Å and 0.90 Å, respectively, based on the Cα atoms. Sequence alignments of PDE5 with PDE6 and PDE11 are provided in Figure 3.4. PDE5 enzyme has a sequence similarity and identity of 64% and 42%; 68% and 47% for PDE6 and PDE11, respectively. 3D structures of PDE5 and PDE6 resemble each other except a clear difference in the β-hairpin domain of PDE6 (amino acid residues: Q687-M701) and its corresponding residues in PDE5 (amino acid residues: R739-L746) which is also observed in other homology model of PDE6 in the literature [9]. The reliability of the homology models is checked by Ramachandran’s plot (Figure 3.5). All torsional angles of amino acid residues of derived protein models are in either favored or allowed regions except one and three outliers that their torsional angles are slightly away from allowed regions for PDE6 and PDE11, respectively. In addition, the stereochemical qualities of the homology models are checked by protein geometry report module in MOE program, carefully, by computing the atom clashes, backbone bond length and angle violations and side chain rotamer outliers (rotamer strain energy cutoff value of 5 kcal/mol). No atom clashes and backbone bond length violations from the expected values (atom-atom pair repulsion energy cutoff value of 0.5 kcal/mol and Z-score value<5, respectively) were reported for the models.

Figure 3.2 : Electrostatic maps of the active sites of the enzymes (left panel). Blue, red and white colours represent positively charged, negatively charged and hydrophobic preferences built at the drug-binding cavity site (near 5 Å distance from the ligand). Tadalafil fulfills the positively charged electrostatic requirement created by the acceptor atom (Oɛ) of the invariant Glutamine side chain in each active sites via carrying a hydrogen-bond donating moiety (-NH) at the amide fragment (namely, Glutamine Switch). On the other hand, hydrophobic residues, Phe820, Phe776 and Trp 820, sandwich the ligand (namely, Hydrophobic Clamp). 2D ligand-protein interaction diagrams (right panel). Green arrows indicate hydrogen bonding interactions. Green and purple discs show hydrophobic and polar residues, respectively. Representations are created with MOE molecular modeling package.

Figure 3.3 : Top docking poses of tadalafil with PDE enzymes. Only polar hydrogens are shown for clarity. Protein residues within 2.5 Å distance around tadalafil are depicted in the figures. Hydrogen bonds between amide hydrogen of tadalafil and Oɛ atom of invariant Glutamine amino acid residue are represented with red dashed lines.

Figure 3.4 : Aminoacid sequence alignments of PDE6 andPDE11 over PDE5 catalytic site residues. Alignment procedure was achieved with BLOSUM62 matrix via MOE software.

Besides, one reported bond angle outlier (D609 for PDE11) and rotamer deviation (M702 for PDE6) are far away from the critical drug-binding cavity which in turn do not effect the docking experiments in the current work. Since the models are well aligned onto PDE5 (Figure 3.6), finally Zn2+ and Mg2+ counterions were embedded to the metal binding sites of PDE6 and PDE11 after overlaying the homology models on the PDE5 structure. Also, the comparison of the contact energy profiles (Figure

46 3.7) are plotted for the models and compared with PDE5 which show correlation with each other.

Figure 3.5 : Ramachandran plots of the homology models of PDE6 and PDE11.

3.3.3 Binding affinity and binding pattern analysis of the hit compounds and tadalafil with PDE5, PDE6 and PDE11

More than 20 million compounds were downloaded from ZINC database and these ligands were prepared at physiological conditions. Similarity analysis of these compounds with tadalafil were performed with molecular fingerprint and similarity search tools. At the end of the similarity search, we were able to filter out 1309 tadalafil-like molecules. These molecules were docked into the drug-binding cavities of PDE5, PDE6 and PDE11 in order to check their predicted binding energies at these targets. After completion of docking calculations, 27 molecules were presented, herein, based on their high docking score against the principal target- PDE5 and also their selectivity over PDE6 and PDE11 (Table 3.1). The common feature of these compounds is that all of them form strong hydrogen bonding interactions with Gln817 in PDE5 except for ZINC23055991, ZINC23183710 and ZINC32995890 compounds. This is due to the absence of polar hydrogens in ZINC23055991 and ZINC23183710. However, they still fitted very well to the substrate pocket in PDE5 (dG.Chemscore= -43.24 and -38.01 kJ/mol for ZINC23055991 and ZINC23183710, respectively) via mostly van der Waals interactions (for ZINC23055991) whereas via H-bonding interactions and π-H interaction (i.e. His613, Tyr612 and Phe786, respectively, for ZINC23183710). On the other hand, ZINC32995890 compound is

47 oriented in such a way that its 1-4 benzodioxine fragment makes a π–H interaction with Leu804 and two π-π stacking interactions by its 1,3-dioxo isoindolin-2-yl fragment with aromatic Phe820 residue.

Table 3.1 : Predicted binding free energies (Chemscore.dG) and 2D structures of ZINC compounds against the principal target, PDE5 and off-target enzymes, PDE6 and PDE11 (binding scores are expressed in kJ/mol and calculated by Chemscore fitness function implemented in GOLD Docking Program). Chemscore.dG values were converted to calculated IC50 values -for the purpose of selectivity comparison-according to the formula; ΔGbinding =RTlnIC50, where T is taken as 300 K.

48 Table 3.1 (continued) : Predicted binding free energies (Chemscore.dG) and 2D structures of ZINC compounds against the principal target, PDE5 and off- target enzymes, PDE6 and PDE11 (binding scores are expressed in kJ/mol and calculated by Chemscore fitness function implemented in GOLD Docking Program). Chemscore.dG values were converted to calculated IC50 values -for the purpose of selectivity comparison- according to the formula; ΔGbinding =RTlnIC50, where T is taken as 300K.

49 Table 3.1 (continued) : Predicted binding free energies (Chemscore.dG) and 2D structures of ZINC compounds against the principal target, PDE5 and off- target enzymes, PDE6 and PDE11 (binding scores are expressed in kJ/mol and calculated by Chemscore fitness function implemented in GOLD Docking Program). Chemscore.dG values were converted to calculated IC50 values -for the purpose of selectivity comparison- according to the formula; ΔGbinding =RTlnIC50, where T is taken as 300K.

50 Table 3.1 (continued) : Predicted binding free energies (Chemscore.dG) and 2D structures of ZINC compounds against the principal target, PDE5 and off-target enzymes, PDE6 and PDE11 (binding scores are expressed in kJ/mol and calculated by Chemscore fitness function implemented in GOLD Docking Program). Chemscore.dG values were converted to calculated IC50 values -for the purpose of selectivity comparison-according to the formula; ΔGbinding =RTlnIC50, where T is taken as 300K.

51 Table 3.1 (continued) : Predicted binding free energies (Chemscore.dG) and 2D structures of ZINC compounds against the principal target, PDE5 and off-target enzymes, PDE6 and PDE11 (binding scores are expressed in kJ/mol and calculated by Chemscore fitness function implemented in GOLD Docking Program). Chemscore.dG values were converted to IC50 values -for the purpose of selectivity comparison-according to the formula; ΔGbinding =RTlnIC50, where T is taken as 300K.

Figure 3.6 : Superposition of PDE5, PDE6 and PDE11, illustrated with blue, cyan and red colors, respectively. The counterions, Zn2+ and Mg2+, are shown with blue circles at the metal binding side.

Detailed protein-ligand interaction diagrams of each compound with three PDEs can be found in Figure 3.8. All of the molecules occupy the narrow and deep hydrophobic catalytic pockets of the enzymes (volume of around 300 Å3); the superpositions of the compounds can be seen in Figure 3.9. None of the compounds directly interacts with the metal atoms at the M site which is in line with the identified crystal structures of the PDE5, so far. We have especially focused on two ligands, namely ZINC02120502 and ZINC16031243, due to their high docking scores for PDE5 (-48.79 kJ/mol and -46.40 kJ/mol, respectively) and relatively low predicted binding affinities at the binding cavities of PDE6 and PDE11 targets (- 37.16 kJ/mol and -39.43 kJ/mol for ZINC02120502; and -34.96 kJ/mol and -38.49 kJ/mol for ZINC16031243 compounds) as compared to other hits. The key interactions of these two ligands within the target enzyme (PDE5) are summarized as follows: ZINC02120502 makes a π-π and π-H interaction with Phe820 and Val782 residues, respectively and a H-bonding interaction with Gln817 with its NH group within a distance of 2.07 Å (Figure 3.10).

Figure 3.7 : Contact energy profiles of the catalytic side residues that correspond to PDE5, PDE6 and PDE11. The x axis and y axis represent the aminoacid residues numbering and atom-atom contact pair energies in kcal/mol unit, respectively.

Figure 3.8 : 2D protein-ligand interaction diagrams of the selected compounds (Table 3.1) with the catalytic site residues of PDE5.

Figure 3.8 (continued) : 2D protein-ligand interaction diagrams of the selected compounds (Table 3.1) with the catalytic site residues of PDE5. It is in close contact with Met816, Phe786, Ile813, Leu804, Leu725, Ala779 residues and; Ile665 and Ser663 on the H-loop site residues via van der Waals interactions.

Figure 3.9 : Superposition of 27 selected compounds at the end of the docking simulations.

56 The orientation of this compound at the catalytic pocket of PDE5 resembles to the crystal orientation of tadalafil in many aspects, whereas the top-scored docked poses within the PDE6 and PDE11 differ significantly compared to its orientation in PDE5 and share common conformations and interaction patterns within each other in PDE6 and PDE11. ZINC02120502, is mainly sandwiched by the aromatic ring of Phe776 residue and Val738 via π-H bonding interaction at the binding pocket of PDE6. The other dominant interactions that stabilize the ligand in the substrate binding pocket are the hydrophobic interactions with Leu721, Ile724, Leu671, Val734, Ala723 and also ligand is surrounded by His559, Glu628, Asp720 and Gln731. ZINC02120502 is oriented at the binding cavity of PDE11 as in PDE6, i.e. its terminal alkyl chain, -

CH2–CH2–CH–(CH3)2-is pointed towards the metal site and its 4-ethoxy-3- methoxy-phenyl ring is clamped between the bulkier aromatic residue Trp872 (replaced with Phe776 in PDE6) and Val834 (replaced with Val738 in PDE6). The two ligands (ZINC02120502 and ZINC16031243) are mainly stabilized at the active sites of PDE6 and PDE11 by van der Waals interactions where they do not form any hydrogen bonding interactions with the active site residues (Figure 3.10).

Figure 3.10 : (A) Docked pose of ZINC02120502 at the active site of PDE6. (B) Docked pose of ZINC02120502 at the active site of PDE11. (C) Overlay of ZINC02120502 (tan) onto the crystal orientation [118] of tadalafil (blue). (D) Overlay of ZINC16031243 docked poses at the active site of PDE5 (blue), PDE6 (pink) and PDE11 (tan). Protein–ligand interaction diagrams of ZINC02120502 and ZINC16031243 with PDEs (shown in the bottom of the figure).

Figure 3.10 (continued) : (A) Docked pose of ZINC02120502 at the active site of PDE6. (B) Docked pose of ZINC02120502 at the active site of PDE11. (C) Overlay of ZINC02120502 (tan) onto the crystal orientation [118] of tadalafil (blue). (D) Overlay of ZINC16031243 docked poses at the active site of PDE5 (blue), PDE6 (pink) and PDE11 (tan). Protein–ligand interaction diagrams of ZINC02120502 and ZINC16031243 with PDEs (shown in the bottom of the figure). As Table 3.1 clearly demonstrates, all these 27 molecules show either higher or similar predicted binding affinities towards PDE5 as compared to tadalafil; also most of them show some selectivity against PDE6 and PDE11. (Binding scores were calculated by ChemScore fitness function implemented in GOLD docking program and ChemScore.dG values were converted to predicted IC50 values according to the

58 following formula; ΔGbinding=RTlnIC50 (T, 300 K) for the purpose of selectivity comparison.) ZINC16031243 is another compound that has a predicted IC50-ratio (PDE6/PDE5) value of 98.16 and predicted IC50- ratio (PDE11/PDE5) value of 23.84 with dG.Chemscore (PDE5) value of -46.40 kJ/mol. A close look at the superimpositions of this compound (ZINC16031243) with the proteins (Figure 3.10) shows that the crucial residues that stabilize these two compounds (ZINC16031243 and ZINC02120502) at the PDE6 and PDE11 drug pockets are located on the flexible loops of the proteins (namely, H-loop and M-loop) besides the Q pocket residues. The position of each molecule is dramatically shifted from its PDE5 docked pose. The importance of these two loops have also been pointed out by Cahill et al., earlier [96]. They proposed that M-loop residues together with its linked α-14 helix (Figure 3.10) are mainly responsible for the selectivity of tadalafil (PDE6/PDE5) in a comparative study with vardenafil; whereas Huang et al. [9] pointed out that the Q pocket residues, Val782 (PDE5)-Val738 (PDE6) and Leu804(PDE5)-Met760 (PDE6) play a critical role in binding affinity reduction of tadalafil (towards PDE6) as compared to sildenafil and vardenafil. However, it should be emphasized that the docked orientation of tadalafil differs in their generated homology models of PDE6 in these two studies. Moreover, an analysis of the dynamical behaviour of these loops by Zagrovic and van Gunsteren [112] showed a pattern of which they called “loop clamp” in order to explain the approaching of M and H loops to each other upon ligand binding to PDE5. Hence, by the motivation of the understanding of the structural aspects of the PDEs and ligand interactions, we further derived 50 ns classical all atom MD simulations trajectories.

3.3.4 MD simulations of apo and holo states of PDE5, PDE6 and PDE11 bound with the selected hit compounds (ZINC02120502 and ZINC16031243) and tadalafil

MD simulations were carried out in order to understand the structural and dynamical behaviour of the selected hits within the binding pockets of the enzymes. Backbone RMSD values during the whole simulations time show that studied systems did undergo in metastable states mostly after about 10 ns (Figure 3.11). Both apo and holo states of PDE5 systems have reasonably converged trajectories at around 3 Å which are slightly lower as compared to PDE6 and PDE11 systems (around 4 Å)

59 which can be expected owing to the construction of the models based on the PDE5 coordinates. It should be noted that, in the case of the PDE11+tadalafil system, the relative late convergence (around 30 ns) also corresponds to a sudden jump in the ligand RMSD trace due to a ligand conformational change around its one rotatable bond (Figure 3.12). The H-bonds that were established during the docking simulations between the oxygens atom of Gln 817, Gln773, Gln869 in PDE5, PDE6 and PDE11, respectively and -NH hydrogen of the common indole rings of the ligands were generally stable during the MD simulations (Figure 3.13), except for the ZINC16031243 in its complex with PDE5.

Figure 3.11 : Traces of protein backbone RMSD evaluation during the whole production stages of the MD Simulations.

In this case, this H-bond was broken at the beginning of the simulation due to the dramatic shifts in the dihedral position of the residue, Gln817. This residue is pointed towards the ligand according to the position of Gln775 and moreover, depends on its orientation whether Gln775 forms an intramolecular hydrogen bond with the donor or acceptor atoms of Gln817, as well. Simulation trajectories were clustered based on the clustering algorithm implemented in UCSF Chimera program (version 1.10.2) [121].

Figure 3.12 : The RMSD evaluation of the ligands during the simulation time.

Representative frames for the most populated cluster were chosen for each simulation and shown in Figure 3.14. Obviously, there is an energetic complimentary between the hydrogen bond network of the protein and the ligand-protein interactions. On the other hand, there is also considerable effect of orientations of water molecules-surrounded around the ligands-on the crucial residue-ligand interactions; e.g. the water molecule pulls the oxygen of Gln817 to itself causing a loss of a H-bond between ZINC02120502 and PDE5 during the simulation time (Figure 3.14). Besides, another H-bond is formed between Asn661 located on the H- loop and the oxygen of tetracyclic ring of ZINC02120502 in PDE5. It should be noted that the position of Phe820 in PDE5 kept its horizontal position according to the tetracyclic rings of each ligand, during the simulation time. Actually, the mentioned importance of rigid tetracyclic fragments in terms of high inhibition potency towards PDE5 was pointed out earlier [14].

Moreover, Asn607 makes a strong hydrogen bonding interaction within a distance of 1.84 Å with the same fragment of ZINC02120502 in PDE6, as well. Although the top docking poses of this compound overlay well within PDE6 and PDE11, the representative frame of ZINC02120502 in the catalytic pocket of PDE11 indicates an orientational change through the simulation time course due to a considerable conformational change in overall protein structure.

Figure 3.13 : Traces of hydrogen bonding interactions throughout simulation time (x and y axis represents the simulation time and distances between Oε atoms of Gln817 (PDE5), Gln773 (PDE6), Gln869 (PDE11) and indole fragments hydrogen in the ligands, respectively). Color codes: orange: PDE5+ZINC16031243; purple PDE11+tadalafil; blue: PDE5+ZINC02120502; red: PDE5+tadalafil; green: PDE6+tadalafil.

Specifically, H-loop migrates towards the Q-pocket residues and adopts a one-turn helix and the bend in the α-14 helix also lets the emergence of a H-bond with its Gln869 and indole hydrogen of the ZINC02120505 (Figure 3.15). Additionally, the indole ring is sandwiched by Phe838. In the case of PDE6+ZINC16031243 system, these motile loops are connected each other via an efficient H-bonding interaction via Tyr610 (H-loop) and Tyr816 (M-loop) within a 1.80 Å distance. Also a water molecule constructs a H-bond with Tyr816 which further gives rise to a unique, nearly overlapping conformational state of this part of the protein. These results point out that critical consideration should also be given to the residues on this highly mobile (which was confirmed by RMSF calculations, Figure 3.16) structural parts of the proteins (H and M loops) along with the Q pocket residues in terms of designing potent and selective PDE5 inhibitors.

3.3.5 MM-GBSA calculations

Predicted binding energies of the selected compounds as well as tadalafil are calculated by MM/GBSA calculations. Table 3.2 summarizes the derived results. Results verify higher predicted binding affinities of the selected hits from screening compared to tadalafil. Moreover, PDE5/PDE6 and PDE5/PDE11 selectivity profiles of these compounds are similar or higher than tadalafil.

62 3.3.6 hERG K+ ion channel activity of the compounds hERG (KCNH2 or Kv11.1) is the name of a gene that encodes the α-subunit of a voltage-gated potassium channel. hERG channels are expressed in various types of tissue and cell types such as heart muscles, brain and retina. Its topology consists of six transmembrane α-helices (S1–S6) where S5 and S6 helices form the channel inner cavity and S1–S4 helices constitute the voltage sensing domain. The function of the channel is to permeate potassium cations across the cell membrane via its unique selectivity filter (SVGFG). Since the inhibition of the potassium current by direct binding of the drugs at the central cavity causes abnormalities in the cardiac action potential [23] which may further lead LQTS, many drugs have been withdrawn from the drug market over the years or their usage has been restricted [25].

It was shown that tadalafil also inhibits the hERG channel by concentration dependent manner with an IC50 value of 100 μM [21]. In this study, for docking calculations, we used the refined structural models of hERG1 in different conformational states that were generated previously by our group and have been extensively validated in experimental and theoretical studies [37, 55-57, 61, 63, 64, 68, 122-124]. Accordingly, we report the in silico activities of the compounds with the channel (Table 3.3). ZINC02120502, ZINC16031243 and tadalafil show similar predicted binding affinities at the central cavities of the hERG K+ channel model.

Table 3.2 : Comparison of protein-ligand free energy results of tadalafil with selected hit compounds using MM/GBSA calculations.

Targets/compounds ΔGMM/GBSA Selectivity (kcal/mol) ratio PDE5+tadalafil -89.23 PDE6+tadalafil -75.63 1.18 PDE11+tadalafil -88.31 1.01 PDE5+ZINC02120502 -130.48 PDE6+ZINC02120502 -116.41 1.12 PDE11+ZINC02120502 -110.20 1.18 PDE5+ZINC16031243 -113.29 PDE6+ZINC16031243 -102.52 1.11 PDE11+ZINC16031243 -101.05 1.12

Figure 3.14 : Simulated structures of ZINC0210502 and ZINC16031243 at the substrate pockets of PDE5, PDE6 and PDE11.

Figure 3.15 : Overlay of docking pose (blue) and representative structure of ZINC02120502 (white) in the catalytic pocket of PDE11.

3.3.7 E-Pharmacophore studies

The structure-based pharmacophore modeling (E-pharmacophore) uses advantages of both ligand– and structure-based approaches by deriving energetically optimized structure-based pharmacophore models. For this aim, representative conformer from MD simulations of one of the selected hit compounds (ZINC02120502) is used for E-pharmacophore studies. Six-sited (RRRHHH) hypothesis was found as top-scored pharmacophore model. These main interactions were aromatic rings (labeled as “R”) and hydrophobic interactions (labeled as “H”, Figure 3.17). The RRRHHH hypothesis is then used for ligand screening of Otava Drug-like Green Collection (around 176000 compounds) using Glide/SP docking protocol from Schrodinger’s Maestro molecular modeling package [125] and top-1000 scored compounds were collected. Compounds that show high docking scores and high fitness scores are also shown in Figure 13.7.

65 Table 3.3 : Predicted binding affinities of the selected compounds within the hERG K+ channel. Each compound was docked into the central cavities of the channel by GOLD docking software with ChemScore fitness function. Docking studies were realized by considering the two known conformational states of the channel. OS and OIS states stand for the open and open-inactivated states. dG.ChemScore values are expressed in kJ/mol. Tadalafil and the two selected potent and selective PDE5 inhibitor compounds ZINC02120502 and ZINC16031243 are shown in bold in the Table.

Compounds dG.Chemscore dG.Chemscore (hERG-OS state) (hERG-OIS state) Tadalafil -32.22 -48.72 ZINC00490454 -34.97 -39.74 ZINC02092043 -35.76 -47.82 ZINC02093785 -32.8 -46.75 ZINC02120502 -35.04 -47.32 ZINC03024615 -29.46 -38.31 ZINC03024617 -33.34 -46.48 ZINC08204637 -32 -47.9 ZINC11692256 -34.06 -43.99 ZINC12360812 -32.71 -41.98 ZINC15955458 -36.54 -53.43 ZINC16031243 -35.12 -47.06 ZINC16042566 -32.95 -40.86 ZINC16043001 -40.52 -53.84 ZINC19020327 -30.57 -42.01 ZINC21986065 -29.05 -42.83 ZINC23055991 -35.17 -45.54 ZINC23183710 -33.42 -38.81 ZINC24891165 -30.73 -39.23 ZINC26772005 -37.63 -41.57 ZINC29158966 -34.38 -41.1 ZINC32995888 -33.84 -43.22 ZINC32995890 -32.06 -43.31 ZINC36055139 -34.8 -53.11 ZINC36210867 -35.18 -49.36 ZINC40146722 -32.34 -44.02 ZINC44448076 -37.69 -44.14 ZINC44448130 -36.76 -45.99

Figure 3.16 : Root-mean-square fluctuation (RMSF) values per residue during the MD simulations.

67 3.4 Conclusions

In this study, the similarity-based virtual screening protocol is applied for the ZINC small molecules database that contains more than 20 million small compounds. Based on Tanimato coefficient values, 1309 molecules from this database showed 80% or more structural similarity with the PDE5 inhibitor tadalafil. These compounds are then docked in PDE5 as well as structurally similar other isoforms PDE6 and PDE11. Results showed that 27 compounds have high predicted binding affinities towards the principle target, PDE5. Especially two hits (ZINC02120502 and ZINC16031243) from 27 compounds represented some selectivities against PDE6 and PDE11. Thus, these two compounds as well as tadalafil are used in classical MD simulations and post-processing MD analyses which showed some insights about their structural and dynamical behaviors at the studied targets. Finally, these hits are also tested at the hERG K+ channel models in order to predict their possible cardiotoxicity side effects. Selected two hits showed similar predicted binding affinities at the hERG channels with tadalafil. Moreover, a structure-based pharmacophores study was also applied for the selected hit compounds. Results of this study can be useful for designing of novel, safe and selective PDE5 inhibitors.

Figure 3.17 : (Top) Derived top-scored six-sited (RRRHHH) E-pharmacophore model; (bottom) 176 000 compounds from Otava small-molecules database are screened against derived pharmacophore model and top-1000 compounds that have high Fitness scores with these sites are then docked at the PDE5 binding pocket using Glide/SP (standard precision). Compounds that show high docking scores as well as high fitness scores are shown in the figure. 2D ligand interaction diagram of selected Otava compound (1094821) is also represented in the figure.

4. STRUCTURAL INVESTIGATION OF VESNARINONE AT THE PORE DOMAINS OF OPEN AND OPEN-INACTIVATED STATES OF HERG1 K+ CHANNEL3

4.1 Introduction

Vesnarinone is a selective phosphodiesterase 3-type enzyme (PDE3) inhibitor, acting as a cardiotonic regulator that has a clinical role in treating congestive heart failure [126]. PDE3 enzyme belongs to the large PDE enzyme family that has several members (PDE1-PDE11)-classified based on their substrate recognition, structural organization or biological functions [69–72]. PDEs are the vital target proteins in curing several diseases owing to their widely distributed feature in diverse tissues in human organism. PDE3 proteins are dual specific enzymes, responsible for the catalytic degradation of intracellular second messengers; cAMP and cGMP while they have higher affinity and turnover rate for cAMP rather than cGMP [127, 128]. PDE3 enzymes exist in muscle cells of many organs; such as liver, kidney and heart. Regarding the heart tissues, PDE3 inhibitors have been developed for the regulation of the cardiomyocytes via their positive inotropic effect and resulted action of cardiac contraction. Besides, the usage of PDE3 inhibitors in long-terms have been restricted due to causing sudden deaths associated with cardiovascular problems [129] and thus, studies have been focused on discovering more selective PDE3 inhibitors, so far. In addition, the positive inotropic effect of vesnarinone has also been contingently attributed to the inhibition of the hERG current (IKr). The mechanism of the related physiological routes and biophysical pathways of the corresponding actions have been described and explained well in detail, elsewhere [130]. In the light of whole cell patch clamp experiments, it was revealed that vesnarinone directly inhibits the hERG1 channels [39-43, 130]. Besides, investigation of the “voltage-dependent blockage”of vesnarinone by different

3 This chapter is based on the article “Kayık, G., Tüzün, N.Ş.,and Durdagi, S. (2017). Structural investigation of vesnarinone at the pore domains of open and open-inactivated states hERG1 K+ channel. Journal of Molecular Graphics and Modelling, 77, 399-412. doi: 10.1016/j.jmgm.2017.08.017

71 research groups resulted in two different conclusions regarding the strength of inhibition of hERG current according to the applied voltage. Katayama et al. found a concentration and voltage-dependent blockage where they observed a strong inhibition at more depolarized membrane voltages when the channel is in its open- state conformation [130]. However, another study by Kamiya et al. reported a voltage-independent block of the channel [41]. They unraveled some important aminoacid residues that have role in vesnarinone-blockage of the hERG channel using Alanine mutagenesis screening. However, regarding the voltage-dependent blockage, Menchaca et al. discussed and attributed this controversy, in terms of the experimental techniques applied in the hERG assays [40].

hERG1(KCNH2 or Kv11.1) channels are defined as the rapidly delayed inward- rectifier voltage-dependent ion channels that selectively permeate K+ ions from the inside of the cells to the outside, across the membrane bilayer. Thus, they provide significant contribution in the process of regulation of the electrical activity of cells (i.e., repolarization of cardiac action potential of human ventricular myocyte) [23]. Topology of hERG1 K+ channels comprise of six trans-membrane helices (S1-S6) where S1-S4 are the voltage-sensing regions owing to the mainly negatively charge amino acids in S1-S3 (i.e., Asp411 on S1, Asp456, Asp466 and Asp460 on S2, Asp501, Asp509 on S3) and positively charged amino acids (i.e., Lys525, Arg528, Arg531, Arg534, Arg537, Lys538 and Arg541 on S4) that undergo conformational changes responding to the driving force of the concentration gradient of the ion flux across the cell membranes. On the other hand, S5-S6 domains constitute the pore domain (PD) and line the putative binding pocket residues regarding notorious drug binding phenomenon that causes an inhibition of the hERG current and hence raises serious life-threatening arrhythmias, leading to Torsades de pointes (TdP) and LQTS [24, 25]. Topology of the hERG1 channel is illustrated in Figure 4.1.

In this study, molecular modeling techniques, i.e., protein-ligand docking, MD simulations, post-processing MD analysis techniques and principle component analysis (PCA) were employed in order to get insight into the vesnarinone-hERG1 channels (both open and open-inactivated states of hERG1 models) interactions, elaborately.

Figure 4.1 : Topology of the hERG channel in its open-state conformation. The channel is a coassembled of four identical chains; each chain is shown with different colours (top). S5 and S6 helices line the pore domain of the channel. A close look at the some selected S6 and SF (selectivity filter) residues that are important for drug binding (bottom).

4.2 Computational Methods

4.2.1 Protein-Ligand docking calculations

The three-dimensional coordinates of vesnarinone were retrieved from ZINC database [114, 131] and further energy minimized using MMFF94X force field with an RMS energy gradient threshold of 0.0001 kcal/mol by utilizing MOE molecular modeling package [28]. Subsequently, vesnarinone was docked into the central cavities of open and open-inactivated hERG1 channel models. Flexible molecular docking calculations (Phe656 and Tyr652 on each chain were handled with full flexibility) were performed with GOLD v.5.3.0 docking program [30] utilizing its four different docking algorithms; namely, ASP (Astex Statistical Potential),

73 ChemScore, GoldScore and CHEMPLP (Piecewise Linear Potential modified with ChemScore terms) for the binding pose predictions. 200-poses were generated in each docking runs. Efficiency of search was set to maximum (200%) and early termination was turned off. The top-docking poses, which were obtained from each scoring function, were served as the starting input geometries (total number of eight) for MD simulations.

4.2.2 MD simulations

All MD simulations were carried out with Gromacs (v.4.6.5) software [32]. PPM Server [132] was used in order to orient the hERG1 channel models with respect to its TM position. Subsequently, protein-lipid bilayer systems were prepared with CHARMM-GUI server [34]. Vesnarinone-hERG1 complexes as well as the apo forms of the channels were placed between DPPC type lipid bilayers (256 lipid molecules) and solvated with TIP3 waters. Additionally, PD of the channel was also solvated with pore waters. Three K+ ions were accommodated at S0-S2-S4 position at the selectivity filter (i.e. Ser624, Val625, Gly626, Phe627, Gly628) of the channel and one K+ was placed at the center of the PD as well. For the purpose of an illustration of the systems, an exemplary is presented in Figure 4.2. CHARRMM36 force field with a time step of 1 fs was used with leap-frog integrator throughout the simulations. Energy minimizations of the systems were done with SD algorithm until the maximum force on each particle reached a value lower than 1000 kJ/mol nm. Six stages of equilibration dynamics-two steps of NVT followed by four steps of NPAT (constant pressure, area, temperature ensemble) simulations were applied on the systems where restrained forces on protein, waters, membrane and ions were gradually reduced [133]. Production stages of the simulations were carried out in NPT at 310 K temperature and 1 atm pressure within a 50-ns total time. Verlet scheme was used with 1.2 nm cut-off value in order to treat the electrostatics and Lennard-Jones interactions between the particles. LINCS constraint algorithm was applied on the hydrogen-bonds during the equilibration and production stages. PME and PBC were used in the simulations. During the production steps, Nose-Hoover thermostat was used in order to maintain the temperature at 310 K and Parrinello- Rahman semi-isotropic coupling algorithm was used to control the pressure using a compressibility factor of 4×10−5. However, Berendsen algorithms were applied at the

74 equilibration steps for temperature and pressure control. Snapshots from the MD trajectories were collected every 10 ps (i.e., 5000 trajectory frames) for further analyses.

Figure 4.2 : Open state hERG Channel (S5 and S6 chains, only, represented with green ribbon) inserted in DPPC lipid-bilayer. Water and lipid molecules are illustrated with red and olive-orange lines. Potassium cations (K+) are located at the energetically favorable S0-S2-S4 sites at the selectivity filter (SF) and also pore domain (PD) region, depicted in pink spheres.

4.2.3 Principal component analysis (Covariance analysis) of the MD trajectories

Principal Component Analysis (PCA) is a commonly used statistical dimension- reduction technique to distinguish and interpret the MD trajectories in terms of revealing the correlated and concerted motions that mostly contribute to the significant conformational dynamics of protein-ligand systems [124, 134-139]. In the applied PCA technique, a covariance matrix of relevant atomic positions is constructed and diagonalized after removing the whole translational and rotational motions from the system. From the covariance matrix, the eigenvectors and their corresponding eigenvalues are further extracted where they represent the newly transformed, independent orthogonal coordinate systems and the quantitative measure of these variations (fluctuations of atom positions in time) in the conformational space, respectively. So that the collective motion of a protein is

75 transformed into individual components of which often correspond to the first few eigenvalues are called “principal components”. Thus, we employed PCA analysis on our trajectories in order to identify the dominant global correlated motions in the hERG1-vesnarinone systems as well as the apo forms of the channels.

4.3 Results and Discussion

Vesnarinone is a weak inhibitor of hERG1 channels that the experimental evidences reported hERG1 blocking IC50 values lie within a micromolar (μM) range [39, 41, 43, 130]. Since it has been a mandatory procedure to test the hERG1 blockage feature of the newly developing drugs, it has been a vital task to identify and understand the drug-hERG1 interactions. Numerous scientific studies covering experimental and in silico techniques have been published with this aim, so far [37, 40-54, 55-61, 63, 64, 68, 122, 123, 140, 141]. In this study, we have chosen vesnarinone, a PDE3 inhibitor, to elucidate the direct blockage mechanism of hERG1 and to predict the important key residues for hERG current blockage, regarding the channel binding phenomenon. It has been revealed so far that, vesnarinone can be captured by the canonical binding site i.e., the S6 domain and pore helix (pH residues) (Figure 4.1), including the notorious Tyr652, Phe656 (via π-π stacking interactions) and Thr623, Ser624 residues that face the central pore cavity. In addition, the employed MD analysis and docking calculations in this study have also identified the polar Ser649 residue as one of the crucial binding element (as well as the aromatic Tyr652 and Phe656 residues) while most of the top docking poses predicted H-bonding interactions between certain fragments of vesnarinone and Ser649, as discussed below. Although an Alanine mutagenesis study [41] revealed that Gly648 and Val659 residues (on S6 trans-membrane helix) are also important binding elements, we did not observe an important direct interaction between these residues and vesnarinone in our docking calculations (except ChemScore) since they mostly face the outer side of the pore domains of the channels and are relatively far away from the central inner cavities in our hERG1 homology models.

4.3.1 Protein-ligand docking calculations

Molecular docking calculations have been employed in order to elucidate the important aminoacid residues in the context of vesnarinone and hERG1 channel

76 interactions. The previously published 3D models of hERG1 channels from our group [37] -both in its open state and open-inactivated state-were used for the current study. Used 3D model structures of hERG1 channel were developed and vigorously validated for different conformational states of hERG1 using a multistep protocol. While the conserved elements were kept using multiple-template homology modeling utilizing available structures for Kv1.2, Kv1.2/2.1 chimera, and KcsA channels, the missing domains were modeled with the ROSETTA De Novo protein- designing suite and further refined with the MD simulations. The final ensemble of models was evaluated for consistency to the reported experimental data from biochemical, biophysical, and electrophysiological studies [37]. Four different fitness functions implemented in the GOLD program; namely, ASP, ChemScore, GoldScore and CHEMPLP have been utilized for pose prediction and further the top-scored binding positions that were attained from each scoring functions have been comparatively discussed. Figure 4.3 depicts the binding modes of vesnarinone at the central cavities of the hERG1 channel, attained from the docking calculations and Table 4.1 depicts their corresponding docking scores. For the open-inactivated state of the channel, interestingly, there is a resemblance at great extent in the binding orientation of vesnarinone along the central pore cavity, predicted by the four scoring functions (Figure 4.3, left panel). Figure 4.4 depicts the 2D protein-ligand interaction diagrams together with 3D representations of important elements of the hERG1 pore domain residues and vesnarinone association for each docking pose. In detail, GOLD/ASP Fitness function predicted two hydrogen bonding interactions between i) the sidechain of Ser649 (H of hydroxyl group) and dihydroquinolin group’s oxygen atom (2.42 Å); ii) the backbone oxygen of Leu622 and dihydroquinolin group’s hydrogen atom (1.99 Å). Moreover, on the same chain, a face-to-face π-π stacking interaction between Tyr652 and the dihydroquinolin ring was observed. In the GOLD/CHEMPLP top-docking pose, similarly, there is a H-bonding interaction within the Leu622 (1.75 Å) and dihydroquinolin group; however due to a slight tilt at this ring with respect to the former position, the previously observed H-bonding interaction within the Ser649 and dihydroquinolin moiety is vanished. In addition, the hydroxyl hydrogen on the side chain of Ser649 on another chain forms a H- bonding interaction with oxygen atom of the one of the methoxy group of vesnarinone. Also, two π-π stacking interactions are established via Tyr652 and Phe656 residues and dihydroquinolin at this predicted binding orientation. For

77 GOLD/ChemScore prediction, unlike from the previously observed binding poses, two edge-to-face π-π stacking interactions emerged within aromatic methoxybenzoyl ring and Tyr652 on different chains, in addition to the stacking between Phe656 and dihydroquinolin group. Two H-bonding interactions between Serine residues (Ser624 and Ser649) and methoxybenzoyl ring (2.06 Å) and dihydroquinolin group (1.97 Å) were further formed. In the GOLD/GoldScore pose, there is only one H-bonding interaction between the polar hydrogen of Ser649 and dihydroquinolin oxygen atom and two π-π stacking interactions between Tyr652 residues and methoxybenzoyl ring formed, similar to ChemScore prediction. On the other hand, for the open-state of the channel, GOLD/ASP yielded a H-bonding interaction (1.69 Å) between the sidechain hydrogen of Tyr652 and dihydroquinolin oxygen atom and double π-π stacking interactions within the methoxy group’s aromatic ring and Tyr652 and Phe656 residues. GOLD/CHEMPLP showed the same stacking interactions like the ASP fitness function’s prediction, with additional two H-bonding interactions between Ser649 and dihydroquinolin’s oxygen and methoxy’s oxygen (1.69 Å and 1.86 Å, respectively). GOLD/ChemScore showed two H-bonding interactions between Ser649 and dihydroquinolin, Leu622 and dihydroquinolin (1.83 Å and 1.83 Å) and a π-π stacking interaction within Tyr652 and dihydroquinolin. GOLD/GoldScore showed no H-bonding interactions between hERG1 channel residues and vesnarinone, however, it gave two π-π stacking interactions between the dihydroquinolin’s aromatic ring and Tyr652 and Phe656. Generally, it can be proposed from these results that; polar residue Ser649 and aromatic hydrophobic residues, Tyr652, Phe656 are the most prominent binding elements of the hERG1 channel in terms of direct blockage via vesnarinone. Also, as depicted in Table 4.1, ASPScore, ChemScore, ChemPLP and GoldScore scoring functions have both consensus predictions of relatively higher docking scores for the open-inactivated state of hERG1. Since it has been shown that [13, 41, 43] vesnarinone has an ability to access to the central inner cavity of the hERG1 channels; this consensus docking results also indicate that tighter interactions between vesnarinone and the hERG1 channel in its open-inactivated state as compared to the open-state may highly be probable. However, it is worth to mention at this point that, although molecular docking is a precious method for gaining a perspective of possible drug-binding positions in a receptor site-especially when the crystal structure of protein-ligand complex is absent-it is inadequate to handle the drug-binding phenomenon in terms

78 of time-dependent variations of the ligand-residue interactions or overall protein dynamics. Thus, we proceeded the current investigation with probing relatively long MD simulations, starting from 8 different top-docking poses for vesnarinone-hERG1 channel bound systems.

Figure 4.3 : The superposition of the top docking poses of vesnarinone, generated by four different scoring functions in the GOLD program.

4.3.2 MD simulations

In the absence of crystal structure of vesnarinone-hERG1 complex system, one of the major goals of this study is to figure out the probable binding orientation(s) of the drug at the pore domain of the channel, using computational modeling techniques. With this motivation, we generated 10 separate MD simulations: 500-ns MD production simulations time in total for holo states (4×2 for open and open

79 inactivated conformational states of the channels) starting from different docking conformations as well as the ligand-free apo forms (1×2) of the channels.

Figure 4.4 : 2D and 3D depiction of the interaction of vesnarinone with the aminoacid residues that line the pore domain and central inner cavity of the hERG1 channels. HOIS and HOS stands for the open-inactivated and open states of the hERG1 channel, respectively. The red, purple and green lines represent the hydrogen bonding interactions and π-π stacking interactions within the residues and vesnarinone at around 5 Å distance, respectively. Green, turquoise and gray colours indicate hydrophobic, polar and Glycine residues, respectively.

Figure 4.4 (continued): 2D and 3D depiction of the interaction of vesnarinone with the aminoacid residues that line the pore domain and central inner cavity of the hERG1 channels. HOIS and HOS stands for the open-inactivated and open states of the hERG1 channel, respectively. The red, purple and green lines represent the hydrogen bonding interactions and π-π stacking interactions within the residues and vesnarinone at around 5 Å distance, respectively. Green, turquoise and gray colours indicate hydrophobic, polar and Glycine residues, respectively.

At first, we have chosen the protein backbone and ligand RMSD evaluations along with the course of the strength of electrostatics (calculated based on Coulomb’s law) and van der Waals (modeled with Lennard Jones potential) interactions (within the hERG1 channels and vesnarinone) as the criteria in determining the most stable vesnarinone orientation and time-dependent dynamics at the pore domain of the hERG1 channel during the time of production runs. Although these energies are rough and do not represent the absolute binding free energies, they were in very good agreement with the backbone and ligand RMSD values in the context of quantifying the structural stability of ligand. These energy values were monitored for 5000 frames in the production stages of the MD simulations and calculated within a threshold value of 1.2 nm prior to applying the PME transformation on the particles of the systems.

82 Table 4.1 : Docking scores (Fitness) of vesnarinone binding inside the central cavities hERG channels, generated with GOLD program.

Docking Score Algorithms and Their Corresponding Scores

hERG Channel ASP Score ChemPLP ChemScore GoldScore Conformation Score Open State 45.48 75.28 35.66 66.81 Open-inactivated 62.73 90.17 38.42 79.28 State

4.3.2.1 General statements on the backbone and ligand RMSD evaluations and ‘Short Range’ energetics

Analyzing the hERG1 backbone RMSD values (Figure 4.5) shows that all trajectories reached almost a plateau after around 10-ns simulations. Trajectories- derived from simulations-initiated with top-scored GOLD/ASP and GOLD/ChemScore complexes have higher RMSD values than the apo state as compared to GOLD/ChemPLP and GOLD/GoldScore trajectories for the open-state hERG1 systems. In addition, the simulations initiated with the GOLD/GoldScore were stabilized significantly at lower RMSD values (around 5 Å) compared to other simulations.

Figure 4.5 : Backbone RMSD values of hERG1 open (top) and open-inactivated (bottom) states during the MD simulations.

83 On the other hand, for the open-inactivated state, except for the simulations initiated with GOLD/ChemScore, the behavior of the rest of the simulations resemble within each other more or less. Judging from the ligand RMSD traces (Figures 4.6 and 4.7), vesnarinone was stabilized more tightly (with minimum, maximum and average values of 0.28, 1.23 and 0.59 Å, respectively) at the central cavity of hERG1 during the whole simulation time, for the GOLD/GoldScore trajectory in the hERG open- state systems. Same argument holds for the open-inactivated systems for ChemPLP and GoldScore trajectories. The electrostatic and van der Waals energies (between hERG1 and vesnarinone) that were monitored during the simulation (Figure 4.8 and Table 4.2) also support the existence of strong interactions established at the PD with vesnarinone in the GOLD/GoldScore fitness functions that served as the docking algorithm to deliver the starting input geometries for the MD simulations.

Figure 4.6 : Ligand RMSD traces during the production stages of the MD simulations (x and y axes represent the simulation time and ligand RMSD values in ns and Å units, respectively). (The least-square fit was done on protein Cα atoms for RMSD measurements.).

Figure 4.7 : Ligand RMSD traces during the production stages of the MD simulations. X and y axes represent the simulation time and ligand RMSD values in ns and Å units, respectively. The minimum, maximum and average RMSD values are also shown on the graphs, in Å unit. (The least-square fit was done based on the starting position of vesnarinone at the begining of simulation in order to measure the conformational flexibility of the ligand.)

As Table 4.2 demonstrates, both CL, LJ and overall energies (CL+LJ) for the GOLD/GoldSscore trajectories are lower than the others while the most remarkable energy gaps occurred in the van der Waals constituent in respect of GOLD/GoldScore trajectory for the open-state hERG1 systems. The course of the absolute values of CL and LJ are illustrated in Figure 4.8. It must be noted that, a quantitative agreement between CL+LJ energy values and the docking scores (GOLD/GoldScore and GOLD/ChemPLP) (Table 4.1) were also observed such that tighter interactions occur in the vesnarinone-open-inactivated state bound systems as compared to the open-state systems. Since trajectories derived from the initial geometry of complex taken from GOLD/GoldScore fitness function have given considerable rigidness in both protein and ligand dynamics as compared to other trajectories, we analyzed it deeply in order to reveal the most important binding elements.

Figure 4.8 : Short range energetics (van der Waals and Electrostatics) interactions between vesnarinone and hERG1 channel throughout the simulation time. Left and right panels represent the open and open-inactivated states of the channel, respectively; starting from the docking outputs of the ASP, ChemScore, CHEMPLP and GoldScore fitness functions (from top to down). CL and LJ are abbreviations for Coulomb (Electrostatics) and Lennard Jones (van der Waals), respectively.

4.3.2.2 MD simulations initiated with GOLD/GoldScore fitness functions: analysis of the trajectories of vesnarinone-hERG1 K+ channel model in its open- inactivated state

It is observed that Ser624, Ser649 and Tyr652 residues are capable of forming strong H-bonding interactions and π-π stacking interactions with vesnarinone at the PD of the open-inactivated state of hERG1 during the simulation time. The initial and final trajectory frame geometries of the MD simulations are depicted in Figure 4.9 and the course of these interactions (distances) are illustrated in Figure 4.10.

86 Table 4.2 : Average Lennard-Jones (LJ), Average Coulomb (CL) and the total value of the averages of these energies between the hERG1 channel and vesnarinone, extracted from the whole simulation time (50 ns). ASP, ChemScore, CHEMPLP and GoldScore denote the fitness functions that served as the docking algorithms to deliver the starting input geometries for the MD simulations. The energies (expressed in kJ/mol unit), given in bold, represent the lowest values of their corresponding columns.

Open State hERG Open-inactivated State hERG Fitness Average Average LJ Energy Average Average LJ Energy Functions LJ Energy CL + LJ Energy CL + Energy CL Energy CL Energy Energy ASP -163.08 -45.71 -208.79 -164.08 -29.98 -194.06 ChemScore -168.78 -39.42 -208.20 -167.39 -23.68 -191.07 CHEMPLP -166.55 -44.93 -211.48 -225.58 -45.41 -270.99 GoldScore -206.20 -48.55 -254.75 -233.36 -91.38 -324.74

Figure 4.9 : The crucial residue-vesnarinone interactions of the initial geometry and MD-simulated structure of the GoldScore trajectory for the open-inactivated state of hERG1. Hydrogen bonding and π-π stacking interactions are depicted in the figure.The distances of stacking interactions are measured based on the centroids of the aromatic rings.

87 The methoxy ring in vesnarinone are squeezed by two aromatic Tyr652 residues (on A and C chains) via π-π stacking interactions. The C=O group’s oxygen makes a very effective H-bonding interaction between Ser624 (on chain C) which prevents vesnarinone translational movement along the central inner cavity and the aromatic pyrazole ring is captured by Ser649 and Tyr652 via H-bonding and π-π stacking interactions on the D chain.

Figure 4.10 : Distances of remarkable intermolecular interactions between vesnarinone and hERG1 channels residues in its open-inactivated state, during the MD simulations. “C=O group’s oxygen atom” and “quinolin moiety’s oxygen atom” phrases refer to the atom groups in vesnarinone.

Moreover, Tyr652 residues on chains A and B form intermolecular H-bonding, which also contribute to structural stability of protein at this bound configuration. The slight fluctuations, observed at around 40-ns in the ligand RMSD value (Figure 4.6 and Figure 4.7), correspond to the torsional movement of the planes of piperazine and quinolin moieties which did “not” result in the breaking of H-bonding pattern at that side. Ser624 was identified out earlier [41] as an important blocking element;

88 our results also support this finding and moreover, we propose a configuration for this interaction in the current work.

4.3.2.3 MD simulations initiated with GOLD/GoldScore fitness functions: analysis of the trajectories of vesnarinone-hERG1 K+ channel model in its open- state

The GOLD/GoldScore fitness function predicted two π-π stacking interactions between pyrazole ring and Tyr652 (on D chain) and Phe656 (on D chain). These stacking interactions were preserved throughout the simulation time. In addition, two π-π stacking interactions between Tyr652 and Phe656 (on C chain) and methoxy ring of vesnarinone were established during the simulation time. These π-π stacking interactions made a clamp like effect and squeezed vesnarinone from end to end in the pore cavity. Besides, three H-bonding interactions were also formed as follows: i) between Tyr652 (on D chain) and pyrazole hydrogen; and ii) between Ser649 (on A chain) and two of the methoxy oxygens. The course of the distances of the mentioned interactions can be seen in Figure 4.11. The initial and final frames of the trajectories were also represented in Figure 4.12. No H-bonding interactions within any fragments of vesnarinone and Ser624 residues were observed contrary to the GOLD/GoldScore open-inactivated state trajectories.

In order to recognize “vesnarinone motility” along the pore domain (whether it was diffused from the binding pocket or not), the RMSD values of vesnarinone were calculated based on least-square fit analysis on Cα atoms (Figure 4.6). This kind of least-square fit analysis enables to take into account the overall translational motion of a molecule as a whole. Inspection of the RMSD evaluations (Figure 4.6) points out that the most remarkable feature is the high RMSD of the trajectories coming from the GOLD/ChemScore fitness functions for both open and open inactivated states of the channel. Indeed, vesnarinone moved out of the central cavity in the GOLD/ChemScore-open-inactivated state (around 14 Å RMSD dif-ference); hence, we considered it an “unstable trajectory” and refer to it “a non-representative of vesnarinone dynamics” in the hERG1 channel pore domain while being its open- inactivated state. The snapshots within a time interval of 5-ns simulations of these trajectories are provided in Figures 13 and 14 for open and open-inactivated states, respectively. Analyzing the ChemScore open-state trajectories indicates that just at

89 the early stages of the simulation (<5 ns), the dihydroquinolin moiety shifted its torsional position and lie horizontally across the z-axis based on the trans-membrane. The distance between two atoms (H atom of Ser649 and dihydroquinolin moiety’s oxygen atom) was increased especially after breaking the H-bond between Ser649 and mentioned moiety of the ligand (Figure 4.15).

Figure 4.11 : Distances of remarkable intermolecular interactions between vesnarinone and hERG channels residues in its open-state during the MD simulations. “Methoxy oxygens and quinolin moiety –NH group’s H atom” phrases refer to the atom groups in vesnarinone.

90 Second dramatic shift occurs around 15-ns and the RMSD reaches to larger values. Similarly, for the GOLD/ChemScore open-inactivated state trajectories, two extremes were observed; (i) vesnarinone goes forward towards the near edge of the S5 domain and (ii) around 20 ns, also a large conformational change in the protein (Figure 4.5) accompanies the tottering of the ligand at the border of the protein.

Figure 4.12 : The crucial residue-vesnarinone interactions of the starting geometry and MD-simulated structure of the GoldScore trajectory for the open state of hERG1.Hydrogen bonding and π-π stacking interactions are depicted in the figure. The distances of stacking interactions are measured based on the centroids of the aromatic rings.

The established H-bond between the Ser624 at the base of the pore helix and ligand from the docking calculations was not observed during the whole stage of the production run; it was already diminished at the equilibration stages. Moreover, the H-bond within the pyrazole oxygen and Ser649, predicted by GOLD/ChemScore

91 docking, were not stable during the production stage; the distance of this bond (-NH hydrogen of pyrazole and -OH oxygen of Ser649) with the ligand and backbone RMSD values are in correlation with each other, suggesting that the orientation of vesnarinone in this case was not energetically favorable. These observations mark the importance of the complimentary nature/essence of MD simulations with regards to the used docking poses from docking simulations.

Figure 4.13 : Snapshots from MD simulations (namely, ChemScore trajectory) of hERG1open state-vesnarinone complexed system. Vesnarinone and hERG1 are represented with sticks and ribbon presentations, respectively. Each chain is coloured with different colours and one of the chains of hERG is deleted, for clarity (top). Overlay of the snapshots from the simulations were shown. The representation was created with a color code, from red to blue according to the simulation time and only 100 frames (taken each 50 frames in the evenly distributed time course) are shown for clarity (bottom).

Figure 4.14 : Overlay of the snapshots from the MD simulations (namely, ChemScore trajectory) of hERG1 Open-inactivated state-vesnarinone complexed systems.Vesnarinone and hERG1 are represented with sticks and ribbon presentations, respectively. Each chain is coloured with different colours and two of the chains of hERG are deleted, for clarity (top). Snapshots from the simulations were shown. The representation was created with a color code, from red to blue according to the simulation time and only 100 frames (taken each 50 frames in the evenly distributed time course) are shown for clarity. (bottom).

4.3.2.4 Principal Component Analysis (PCA) & comparison of time-dependent behaviors of apo states and vesnarinone-bound hERG1 systems

As expected, PCA analysis result shows that only first few eigenvectors contribute to the backbone movements of the channels mostly; the first 10 eigenvalues versus eigenvectors graphs indicate a rapid decay to zero in the eigenvalues before reaching the 10th eigenvector ranking (Figure 4.16).

Figure 4.15 : Distance between hydrogen atom of Ser649 and dihydroquinolin moiety’s oxygen atom of vesnarinone, in the ChemScore trajectory for the open state of hERG1 channel during the 50 ns simulation production period. Just at the beginning periods of the simulation (within the 0-5 ns time period), H-bond between these atoms was broken and never formed later for the rest of the simulation.

The highest first eigenvalues correspond to the trajectories derived from hERG1 open-state model used in GOLD/ChemScore algorithm, hERG1 open-state model used in GOLD/ChemPLP algorithm and the hERG open-inactivated state model used in GOLD/ChemScore algorithm, having first eigenvalues of 78.82, 81.17 and 103.97 nm2, respectively. The lowest overall eigenvalue in the hERG1 open state- vesnarinone complexed system belongs to the GOLD/GoldScore trajectory, having a trace of covariance matrix value of 55.43 nm2 where the others have 182.09 nm2, 173.62 nm2 and 164.90 nm2 for GOLD/ChemScore, GOLD/ASP and GOLD/ChemPLP open state trajectories, respectively. However, in the open- inactivated hERG1 system, except for the excluded trajectory (ChemScore) due to defining an “unstable binding pattern” based on the RMSD and interaction energetics discussions above, the GOLD/GoldScore, GOLD/ASP and GOLD/ChemPLP trajectories have close trace of covariance matrix values of 39.79, 35.56 and 44.90 nm2, respectively while the GOLD/ChemScore trajectory has a value of 143.47 nm2. These results highlight a correlation between the flexibility of the backbone motion of the hERG1 channels and the extent of conformational space of ligand during the simulation time. Also, comparison of the eigenvalues of the apo forms and holo systems (for GoldScore trajectories) highlights an effect of rigidity that was introduced to the backbones of hERG1 channels by vesnarinone. Visual inspection of the first and last frames of the first eigenvector movements of the apo forms and vesnarinone-bound systems show that the turret region (S5-P Extracellular Linker) fluctuate at most while the complexation of vesnarinone did not induce large

94 backbone deviations. Mostly, sidechain conformational motions govern the vesnarinone induced dynamics of the systems.

Figure 4.16 : First ten eigenvalues of the protein backbone movement along the simulation time. Note that, slopes of the ChemPLP and GoldScore curves overlap for the hERG1 open inactivated state.

4.3.2.5 MM/PBSA (Molecular Mechanics/Poisson Boltzmann surface area) calculations

MM-PBSA [142] is one of the most common place methods applied in the protein- ligand binding energy estimations [143–150]. Still, it has both advantages and questionable sides. Some of the questionable aspects are i) considering an implicit solvent model (continuum dielectric environment)-although almost all protein-ligand simulations are carried out in explicit water-, ii) handling the computation of entropy and iii) the resulting “large” binding energies which may cause high standard deviations. On the other hand, the advantages can be stated as it is cost-effective (not

95 computationally demanding) as compared to the alchemical methods; e.g thermodynamic integration (TI) or free energy perturbation (FEP). Its accuracy has several times stated to lie between these methods and docking scoring. The other important benefit of MM-PBSA approach is that it allows energy decomposition so that the specific residue-ligand interactions could be illuminated which is of the very vital issue in drug design studies, too. Moreover, it has a proven success in ranking the active and inactive ligands and reproducing the relative binding energies of experimentally known binding affinities, targeting the same protein.

In this study, we have used the g_mmpbsa program in conjunction with MmPbSaStat.py and MmPbSaDecomp.py python scripts [151, 152], designed for Gromacs simulation package, in order to calculate the vesnarinone binding affinities to hERG channels and discriminate the most energetically favorable vesnarinone- hERG conformation(s) and finally decompose the binding energies per residue in order to reveal out the critical residues which dominate to the association of vesnarinone and hERG channels. The binding free energy of a ligand to a protein is estimated given the functional form of MM-PBSA formalism in Eq. (1).

ΔGbind = GPL-complex- GP - GL

G = EMM + Gsolv - TS

EMM = Ebond + Eangle + Edihedral + Eimproper dihedral + Eel + EvdW (4.1)

Gsolv = Gsolv-pol + Gsolv-nonpol

In this equation; ΔGbind, GPL-complex, GP, GL represent the binding free energy of a ligand, free energy of protein-ligand complex and free energies of the sole protein and ligand in solvent, respectively. The free energy of each term (denoted as G) consists of three terms which are molecular mechanics (MM) potential energy (EMM) in vacuum, free energy of solvation (Gsolv) and TS term where T is the absolute temperature and S is the entropy. EMM comprises of bond, angle, dihedral and improper torsions’ energies (bonded components) which depend on the MM force field used and Eel (Electrostatic energy) and EvdW (van der Waals energy) which are calculated with Coulomb and Lennard Jones equations, respectively. However, in the adapted single trajectory approach, the difference in the average bonded components of the MM potential energy for the species-considering ligand-free and ligand-bound systems turns out to be zero due to the assumption that protein and ligand have the

96 same conformations before and after the protein-ligand complexation. Gsolv term has two components: Gsolv-pol and Gsolv-nonpol terms as they represent the polar and nonpolar solute-solvent interactions. The polar part of the solvation energy (Gsolv-pol) models the electrostatics interactions by solving the Poisson-Boltzmann Equation and the nonpolar part (Gsolv-nonpol) takes into account the rest of the solvent-solute interactions, including the cavity formation and the hydrophobic effect. Several implementations can be incorporated in the formula while computing the nonpolar part of the solvation energy, such as the commonly employed solvent accessible surface area (SASA) or solvent accessible volume (SAV) approaches. The entropic contribution, TΔS, is not implemented in the current MM-PBSA calculations; so the binding energies, herein, approximated to the enthalpy of binding rather than the free energy of binding energies. Last 40 ns (4000 frames, evenly taken in time intervals) of the trajectories were evaluated for the MM-PBSA calculations except for the GOLD/ChemScore trajectory for the hERG1 open state due its relative late convergence. The overall thermodynamic properties that MM-PBSA calculations yielded are summarized in Table 4.3. The most dominant stabilizing forces are the van der Waals forces whereas the Coulombic interactions have approximately 5-fold + less contributions in ΔGbind values for the open inactivated hERG1 K channel. Besides, ASP trajectory gave energetically less favorable binding dynamics, clearly. The polar solvation component of solvation energy militates against binding phenomenon as usual since it involves a payment of a penalty for making a solute cavity, etc. Analyzing the GOLD/GoldScore and GOLD/ChemPLP trajectories’ energetics on the basis of per-residue contributions (Figure 4.17) highlighted Tyr652, Ser649 and Phe656, Met645 residues as the most potent binding elements. These trajectories may be regarded as representations of equal probability in vesnarinone dynamics in the case of open-inactivated state of hERG channel. Note that, these MM-PBSA results are also in accord with the backbone, ligand RMSD values (Figure 4.5 and Figure 4.6, respectively), short range-energetics (Table 4.2) and also backbone PCA behavior (Figure 4.16). The main difference of GOLD/ChemPLP interaction patterns from the GOLD/GoldScore trajectory products are summarized as follows: No H-bond was formed between Ser624, Tyr652 residues and vesnarinone; however, vesnarinone was clamped by Tyr652 residues on four of the chains tightly and Phe656 residues were in close contact with vesnarinone in the whole simulation time.

97 On the other hand, two H-bonds were occasionally established between the methoxy oxygens and Ser649 (Figure 4.18). The similarity of the superposition of the vesnarinone orientations based on the representative frames for the two- aforementioned trajectory can be seen in Figure 4.19. (The representative frames were considered by taking the lowest RMSD value frame according to the average RMSD values calculated from last 40 ns). For the “open state hERG channel”, the free energy of binding (ΔGbind) of the GOLD/GoldScore trajectory is about 30 kJ/mol lower than the GOLD/ASP and GOLD/ChemScore trajectory and besides, the polar solvation component of the solvation energy militates against binding at most for the GOLD/ChemPLP trajectory which causes relatively unfavorable dynamics (around - 43 kJ/mol). The relatively high standard deviation at this component and finally at

ΔGbind value also imply the fluctuation of the binding energetics of vesnarinone association during the MD simulation time in the GOLD/ChemPLPL trajectory for the hERG-open state system (Table 4.3). Inspection of the GOLD/GoldScore per- residue energy decomposition results show that Phe656 and Tyr652 residues are crucial for binding process in the hERG-open state system.

Table 4.3 : MM-PBSA Results of Vesnarinone Binding to hERG Channels. Interaction Binding Energy (ΔGbind) components with their standart deviations are shown in kJ/mol.

Van der Waals Electrostatic Polar Nonpolar ΔGbind hERG Open inactivated Energy Energy Solvation Solvation State Energy Energy (SASA) GoldScore trajectory -243.410 ± -57.104 ± 188.128 ± -22.891 ± -135.277± 11.870 7.251 12.698 0.825 13.463 ASP trajectory -172.037 ± -29.716 ± 145.341 ± -20.214 ± -76.627 ± 9.604 8.146 17.421 0.827 15.922 ChemPLP trajectory -236.252 ± -43.000 ± 167.866 ± -23.199 ± -134.585 ± 11.746 9.030 15.705 0.851 13.449 hERG Open State GoldScore trajectory -213.127 ± -38.736 ± 155.883 ± -21.111 ± -117.091 ± 10.038 7.336 11.446 0.913 11.863 ASP trajectory -171.441 ± -29.412 ± 127.819 ± -20.746 ± -93.780 ± 9.891 10.905 17.121 0.888 11.665 ChemPLP trajectory -171.845 ± -42.933 ± 195.920 ± -24.476 ± -43.334 ± 10.236 9.512 36.624 4.933 34.184

ChemScore trajectory -180.872 ± -32.509 ± 144.371 ± -21.429 ± -90.439 ± 10.346 8.552 12.933 0.807 11.911

Figure 4.17 : Per-residue energy decompositions of MM-PBSA binding energy calculations. (Contributions of each residue were presented based on the cumulative contributions from each chain.).

100

Figure 4.18 : The course of distances established between methoxy oxygens of vesnarinone and Ser649 H atom. The positions of Tyr652 and Phe656 residues at the beginning and end of the MD simulation. Vesnarinone is shown in magenta color.

101

Figure 4.19 : Vesnarinone superposition for representative frames of GoldScore (green) and ChemPLP (red) trajectories in the hERG1 open inactivated state. Only polar hydrogen is shown for clarity.

4.3.2.6 Extension of the MD simulations of hERG1-Vesnarinone complex systems

We further extended the MD simulations of the 50-ns production time to 550-ns (11- fold of initial simulation time) for the open state and open inactivated states of hERG1 channel holo systems for Goldscore trajectories, due to their giving relatively more robust trajectories during the 50 ns simulation time as discussed above in order to see whether the systems’ observed binding patterns have changed, such as the binding energy and binding positions of vesnarinone to the pore domains of hERG1 channels and per-residue contribution to the binding process as well. Additionally, MM-PBSA computations were also employed for the last 0.5 μs of the simulations in order to check the ΔGbind energetics of vesnarinone binding to the pore domain of the channels as well. The backbone RMSD (Figure 4.20), ligand RMSD traces (Figure 4.21), short range energetics (Electrostatics and van der Waals interactions, Figure 4.22) and MM-PBSA binding energy and per-residue energy decomposition analysis calculations (Table 4.4 and Figure 4.23) are summarized as follows.

Inspection of the backbone RMSD graphs show that the maximum values slightly increased to 6.85 Å and 5.17 Å from 4.55 Å and 4.05 Å for the open-state and open- inactivated state of hERG1, respectively during the last 500 ns simulation time. Corresponding average RMSD values are 3.34 Å and 4.54 Å for the open-state and 3.17 Å and 3.76 Å for the open-inactivated states throughout 50-ns and 550 ns, respectively.

102

Figure 4.20 : Protein backbone RMSD traces during the 550-ns production stages of the MD simulations, started from the GoldScore docking output geometries. The overlay of vesnarinone at 50-ns and 550-ns are also depicted (below).

For the open-state case, fluctuations in the position of vesnarinone were observed within the 50-100 ns time interval and at around 150 ns and 350-ns as can be seen in the ligand RMSD graphs (Figure 4.21). However, the orientation of vesnarinone mainly remained the same during the last 500 ns. The overlay of vesnarinone structures at 50 ns and 550 ns are shown in Figure 4.20. Comparison of analysis of previous (50-ns) and extended simulations (550-ns) showed that, there is no dramatic change on the results when the simulations are extended. Alignment of frames between 50-ns and 550-ns at the open-state shows that bioactive conformation of the ligand at the binding pocket is quite stable (average RMSD difference between first 50-ns and rest of 500-ns is less than 1 Å).

103

Figure 4.21 : Ligand RMSD traces during the 0.55 μs production stages of the MD simulations, started from the GoldScore docking output geometries.

Although corresponding figure for open-inactivated state shows a slight positional change, RMSD difference between first 50-ns and rest of 500-ns is around 1 Å and 1.5 Å, based on the ligand-fit and Cα fit RMSD measurements, respectively.

MM-PBSA binding energy computations-considering the last 0.5 μs simulation time- showed similar per-residue energy computation results compared to the initial 50-ns simulation part such that the most potent residues in terms of vesnarinone binding were still Tyr652 and Phe656 residues for the open-inactivated and open state of hERG1, respectively, with an interaction energies of -30.07 kJ/mol and -16.15 kJ/mol which are in agreement with the initially observed (first 50 ns simulation time) most potent residues (Figure 4.17, -35.45 kJ/mol for Tyr652 and -13.40 kJ/mol for Phe656 for the open-inactivated and open-states of hERG1, respectively.)

104

Figure 4.22 : Short range energetics (van der Waals and Electrostatics) interactions between vesnarinone and hERG1 channel throughout the 0.55 μs simulation time (for GoldScore trajectories). CL and LJ represent the Coloumb (Electrostatics ) and Lennard Jones interactions (van der Waals), respectively. Average Coloumb and Lennard Jones interaction energies are -44.37 kJ/mol and -195.44 kJ/mol for the open state of hERG1 whereas average Coloumb and Lennard Jones interaction energies are -91.50 kJ/mol and -233.76 kJ/mol for the open inactivated state of hERG1.

Moreover, the overall ΔGbind values (Table 4.4) imply a slight change for the open- state hERG1 case such that vesnarinone bounded to the channels more tighter for the last 0.5 μs (ΔGbind = -129.301 kJ/mol) as compared to the first 50 ns (ΔGbind = - 117.091 kJ/mol). Also, inspection of the short-range electrostatics and Lennard-Jones energetics (Figure 4.22) show that vesnarinone still tightly bounded to the central inner cavities of the channels during the last 0.5 μs simulation time, generally with a relatively close interaction energy values with respect to the first 50 ns time (Table 4.2).

105 Table 4.4 : MM-PBSA Results of Vesnarinone Binding to hERG1 Channels (for GoldScore trajectories) during the 0.5 μs simulation time. Interaction binding energies with different components with their standart deviations are shown in kJ/mol.

hERG1 Van der Electrostatic Polar Nonpolar ΔGbind Open-inactivated Waals Energy Solvation Solvation State Energy Energy Energy (SASA) Goldscore -241.623 -65.257 198.633 -24.434 -132.681 trajectory (0.5 μs) ±13.344 ± 16.111 ±18.758 ± 0.898 ± 19.024

hERG1 Van der Electrostatic Polar Nonpolar ΔGbind Open Waals Energy Solvation Solvation State Energy Energy Energy (SASA) Goldscore -192.103 -33.745 117.209 -20.662 -129.301 trajectory (0.5 μs) ±12.270 ± 6.952 ±13.229 ± 0.984 ± 11.290

Figure 4.23 : Per-residue interaction energies of MM-PBSA binding energy calculations of vesnarinone binding at the hERG1 channels during the 0.5 μs simulation time. (Contributions of each residue were presented based on the cumulative contributions from each chain.).

106 However, taken into account for the short-range energetic values and MM-PBSA results together, an observed decrease in the MM-PBSA ΔGbind value as contrary to the short-range interactions between vesnarinone and hERG1 open-state mainly stemmed from the decrease in polar solvation energy component-with regards to the first 50 ns. This observation may be attributed to the slight conformational changes of vesnarinone during the last 0.50 μs that go along with the solvation process. On the other hand, for the open-inactivated case, ΔGbind value almost remained the same

(ΔGbind = -132.681 kJ/mol, Table 4.4) for the last 500 ns simulation period when compared to the intial 50 ns (ΔGbind = -135.277 kJ/mol, Table 4.3).

Recently, cryo-EM structure of hERG1 in open-state is deposited to the PDB by Rod MacKinnon group [153] (after completion of the current work). Rapid developments of Cryo-EM technology is likely to drive further refinement of hERG1 states and will have a major impact on the studies focusing on hERG1 channel/drug interactions. When we compare the hERG open-state cryo-EM and 3D hERG open-state model structures, RMSD between these two whole structures (TM and loop regions together) when they are aligned was calculated as 3.17 Å. The corresponding RMSD value for only TM domains was 2.08 Å. Figure 4.24 represents the alignment of cryo-EM and open state model of hERG1 structures (Structures were superimposed without performing any constrain or forcing alignment of both pair residues).

Figure 4.24 : Alignment of hERG channel cryo-EM (cyan) and 3D model (open- state) (orange) structures.

All of residue pairs aligned very well. Alignment score of whole structure and TM region were found as 0.42 and 0.17 (smaller is better). We also repeated docking

107 studies and MD simulations on the cryo-EM structure. In order to compare the top- docking poses of Vesnarinone at the pore domains of hERG using cryo EM and model structures, same protocol is applied in docking simulations.

Table 4.5 : Comparison of docking poses and scores using model and cryo-EM structures. CRYO-EM MODEL CHEMPLP CHEMPLP Phe557, Thr623, Ser624, Ser649, Met554, Leu622, Thr623, Ser624, Tyr652, Ala653, Phe656, Gly657, Leu646, Ser649, Leu650, Tyr652, Ser660, Ala661, Gln664 Ala653, Ile655, Phe656 Docking Score: 66.61 Docking Score: 75.28 ASP ASP Thr623, Ser624, Ser649, Tyr652, Leu622, Ser624, Ser649, Tyr652, Ala653, Phe656, Gly657, Ser660, Ala653, Phe656, Gly657, Ser660 Ala661, Gln664 Docking Score: 45.48 Docking Score: 34.94 GoldScore GoldScore Leu622, Thr623, Ser624, Val625, Thr623, Ser624, Leu646, Ser649, Met645, Gly648, Ser649, Tyr652, Leu650, Tyr652, Ala653, Phe656 Ala653, Phe656, Gly657, Ser660 Docking Score: 75.17 Docking Score: 66.81 ChemScore ChemScore Ser624, Tyr652, Ala653, Phe656, Met554, Leu622, The623, Ser624, Gly657, Ser660, Ala661, Gln664 Ser649, Leu650, Tyr652, Ala653, Phe656, Val659, Ile663, Gln664 Docking Score: 29.63 Docking Score: 35.66

When the docking poses of Vesnarinone at the model and cryo-EM structures were compared, it can be seen that in both cases, crucial amino acids of hERG channel at the pore domain are conserved in almost all docking algorithms: Thr623, Ser624, Ser649, Tyr652, Phe656. Table 4.5 shows amino acid residues near 4 Å from docking poses using hERG cryo-EM and model structures. Obtained top-docking scores from different GOLD algorithms are also similar using cryo-EM and model structures. Per-residue interaction analysis also derived similar crucial amino acids for binding of vesnarinone at the pore domain (Figure 4.25).

4.4 Conclusions

In the current study, the dynamics of vesnarinone at the inferred binding sites of hERG1 K+ channel in both open and open-inactivated states were investigated. The implied motivation is to understand the details of unwanted interactions of

108 vesnarinone and hence guide future drug development. To this end, we first employed docking/empirical scoring. To rectify the crude static binding poses that are typically generated by docking algorithms, MD simulations of top-scoring docking complexes were performed. MM/PBSA is also employed to get more accurate estimates of the binding enthalpy.

Figure 4.25 : Per-residue energy decompositions of MM-PBSA binding energy calculations of vesnarinone binding to open-state hERG1 channels (using Cryo-EM structure) during the 50 ns simulation time. (Energetic contributions of each residue for binding were presented based on the cumulative contributions from each chain.). (Residues between 635-668 are shown only, for clarity.) Note that, the calculated per-residue contribution energy profile demonstrates merely the preliminary outcome for 50 ns simulation time. The relatively high positive energetic contribution to binding enthalpy–that came from polar Ser624 residues-need to be better refined via longer simulation time which may eventually further rectify the polar solvation energy component of ΔGbinding that also militate against binding process. For all that, with respect to the aminoacid residues that are situated at the pore domains of the channel, the relevant outcome still presents a qualitative data whether which residues tend to mostly contribute to binding; such as the bulky and aromatic Tyr652 residues on the S6 helix of the channel.

The results revealed out the probable binding positions of vesnarinone and key- hERG1 residues that interact with the drug. Mainly, molecular docking outcomes were rectified via relatively long MD simulations, which put forward the most favorable/stable vesnarinone dynamics and orientations at the canonical drug binding site for both open and open-inactivated states of the hERG1 channels. As the result of MM-PBSA binding energy estimations, Met645, Ser649, Tyr652 and Phe656 residues-that are located on the S6 region of the channels-were identified as the crucial binding elements in terms of stabilization of vesnarinone. In the absence of

109 crystal structure of vesnarinone-bound hERG1 channels, the adapted multi-modeling methods and several post-MD analysis have been a considerable endeavor in illuminating the vesnarinone bound-hERG channels dynamics, as well. Also, current investigation of the time-dependent behaviors of the molecular systems and further MM-PBSA calculations subsequent to the elaborately and comparatively discussed docking results mark both the vitality and moreover ameliorated essence of the multi-scale modeling approaches in understanding the protein-ligand interactions which is one of the main significant outcomes of this study.

110

5. CONCLUSIONS AND RECOMMENDATIONS

Since the discovery of the vital function of hERG (human ether-à-go-go-related gene) channels in human organism at the mid 1990s, there have been great scientific investigations-contributed from both academic institutions and industrial foundations (especially pharmaceutical companies, etc.)-on the understanding of the hERG-drug interactions and rehabilitation efforts for the undesired activities of some drugs against the hERG channels. It has been discovered so far that the blockage/inhibition of hERG channels by drug molecules may lead to life-threatening disorders (mostly related to cardiovascular arrhythmias). Therefore, for the past years, many drugs covering a broad range of therapeutics, (i.e., antihistamines, antibacterials, antipschotics and antidepressants) have been either withdrawn from the market (e.g., terfenadine,cisapride, astemizol, etc.) or restricted in their uses (e.g., thioridazine, haloperidol, sertindole, and pimozide, etc.). The studied drugs, herein, are phosphodiesterase (PDE) type enzymes. Particularly, after the discovery of Sildenafil-ViagraTM by Pfizer in 1998 targeting the treatment of erectyle dysfunction, several studies have been carried out to discover more potent and safe PDE5 inhibitors.

Experimental works (biochemical, pharmacological, pharmacokinetic studies, in vitro and in vivo assays, and SAR (Structure Activity Relationships)) developments have provided enormously rich and valuable data on the understanding of hERG- drug interactions; however, in silico approaches as in this thesis introduce a considerable perspective from the point of view of molecular level which is indispensable in order to comprehend the protein-ligand interaction mechanisms.

This PhD thesis demonstrates various molecular modeling applications in the field of computer-aided drug design and investigation of protein-ligand systems. Overall, fragment-based de novo drug-design and molecular similarity-based virtual screening protocols combined with structure-based modeling approaches such as molecular docking, homology modeling, MD simulations, etc., have been used in order to design/rehabilitate sildenafil and tadalafil molecules. Cross-reactivity of tadalafil,

111 targeting to PDE6 and PDE11 enzymes, has been also considered. hERG blocking affinities and PDE activities of the sildenafil, vardenafil, tadalafil and vesnarinone molecules have been illuminated at molecular level and critical aminoacid residues in terms of ligand association at the catalytic regions of PDE enzymes and pore domains of hERG channels (in open and open-inactivated states) have been identified. Structural investigation of binding patterns, binding energetics and dynamics of vesnarinone at the central cavities of hERG channels have been explored via several molecular docking, elobarate MD simulations and post-MD analysis.

In the future, investigation of the structural features of biological systems via computer-based molecular modeling techniques will continue to survey further owing to the increasing computer power. This Ph.D study demonstrates several important findings and conclusions regarding the structural and dynamics aspects of the studied protein-ligand systems by the use of in silico techniques. Successful ones among the proposed potent and safe PDE5 inhibitors can be tested experimentally and may serve as potential drug leads in the future.

112

REFERENCES

[1] Beavo, J. A. (1995). Cyclic-nucleotide phosphodiesterases-Functional implications of multiple isoforms, Physiological Reviews, 75 (4), 725- 748. [2] Bender, A. T., and Beavo, J. A. (2006). Cyclic nucleotide phosphodiesterases: Molecular regulation to clinical use, Pharmacological Reviews, 58 (3), 488-520. [3] Goraya, T. A., and Cooper, D. M. F. (2005). Ca2+-calmodulin-dependent phosphodiesterase (PDE1): Current perspectives, Cellular Signalling, 17 (7), 789-797. [4] Soderling, S. H., and Beavo, J. A. (2000). Regulation of cAMP and cGMP signaling: New phosphodiesterases and new functions, Current Opinion in Cell Biology, 12 (2), 174-179. [5] Castro, A., Jerez, M. J., Gil, C., and Martinez, A. (2005). Cyclic nucleotide phosphodiesterases and their role in immunomodulatory responses: Advances in the development of specific phosphodiesterase inhibitors, Medicinal Research Reviews, 25 (2), 229-244. [6] Meng, F., Hou, J., Shao, Y. X., Wu, P. Y., Huang, M. N., Zhu, X. H., ..… Ke, H. M. (2012). Structure-based discovery of highly selective phosphodiesterase-9A inhibitors and implications for inhibitor design, Journal of Medicinal Chemistry, 55 (19), 8549-8558. [7] Menniti, F. S., Faraci, W. S., and Schmidt, C. J. (2006). Phosphodiesterases in the CNS: Targets for drug development, Nature Reviews Drug Discovery, 5 (8), 660-670. [8] Sung, B. J., Hwang, K. Y., Jeon, Y. H., Lee, J. I., Heo, Y. S., Kim, J. H., ..… Cho, J. M. (2003). Structure of the catalytic domain of human phosphodiesterase 5 with bound drug molecules, Nature, 425 (6953), 98–102. [9] Huang, Y. Y., Li, Z., Cai, Y. H., Feng, L. J., Wu, Y. N., Li, X. S., and Luo, H. B. (2013). The Molecular basis for the selectivity of tadalafil toward phosphodiesterase 5 and 6: A modeling study, Journal of Chemical Information and Modeling, 53 (11), 3044-3053. [10] Cichero, E., D’Ursi, P., Moscatelli, M., Bruno, O., Orro, A., Rotolo, C., ..… Fossa, P. (2013). Homology modeling, docking studies and molecular dynamic simulations using graphical processing unit architecture to probe the type-11 phosphodiesterase catalytic site: A computational approach for the rational design of selective inhibitors, Chemical Biology & Drug Design, 82 (6), 718-731.

113 [11] Eros, D., Szantai-Kis, C., Kiss, R., Keri, G., Hegymegi-Barakonyi, B., Kovesdi, I., and Orfi, L. (2008). Structure activity relationships of PDE5 inhibitors, Current Medicinal Chemistry, 15 (16), 1570-1585. [12] Bi, Y. Z., Stoy, P., Adam, L., He, B., Krupinski, J., Normandin, D., ..… Macor, J. E. (2001). The discovery of novel, potent and selective PDE5 inhibitors, Bioorganic & Medicinal Chemistry Letters, 11 (18), 2461-2464. [13] Bi, Y. Z., Stoy, P., Adam, L., He, B., Krupinski, J., Normandin, D., ..… Macor, J. E. (2004). Quinolines as extremely potent and selective PDE5 inhibitors as potential agents for treatment of erectile dysfunction. Bioorganic & Medicinal Chemistry Letters, 14 (6), 1577- 1580. [14] El-Gamil, D. S., Ahmed, N. S., Gary, B. D., Piazza, G. A., Engel, M., Hartmann, R. W., and Abadi, A. H. (2013). Design of novel betacarboline derivatives with pendant 5-bromothienyl and their evaluation as phosphodiesterase-5 inhibitors, Archiv Der Pharmazie, 346 (1), 23-33. [15] Mohamed, H. A., Girgis, N. M. R., Wilcken, R., Bauer, M. R., Tinsley, H. N., Gary, B. D., ..… Abadi, A. H. (2011). Synthesis and molecular modeling of novel tetrahydro-betacarboline derivatives with phosphodiesterase 5 inhibitory and anticancer properties, Journal of Medicinal Chemistry, 54 (2), 495-509. [16] Sakamoto, T., Koga, Y., Hikota, M., Matsuki, K., Murakami, M., Kikkawa, K., ..… Yamada, K. (2014). Design and synthesis of novel 5-(3,4,5- trimethoxybenzoyl)-4-aminopyrimidine derivatives as potent and selective phosphodiesterase 5 inhibitors: Scaffold hopping using a pseudo-ring by intramolecular hydrogen bond formation, Bioorganic & Medicinal Chemistry Letters, 24 (22), 5175-5180. [17] Shang, N. N., Shao, Y. X., Cai, Y. H., Guan, M., Huang, M. N., Cui, W. J., ….. Luo, H. B. (2014). Discovery of 3-(4-hydroxybenzyl)- 1- (thiophen-2-yl)chromeno [2,3-c]pyrrol-9(2H)- one as a phosphodiesterase-5 inhibitor and its complex crystal structure, Biochemical Pharmacology, 89 (1), 86-98. [18] Toque, H. A. F., Priviero, F. B. M., Teixeira, C. E., Perissutti, E., Fiorino, F., Severino, B., ..… De Nucci, G. (2008). Synthesis and pharmacological evaluations of sildenafil analogues for treatment of erectile dysfunction, Journal of Medicinal Chemistry, 51 (9), 2807- 2815. [19] Ukita, T., Nakamura, Y., Kubo, A., Yamamoto, Y., Moritani, Y., Saruta, K., ….. Omori, K. (2003). 1,7-and 2,7-naphthyridine derivatives as potent and highly specific PDE5 inhibitors, Bioorganic & Medicinal Chemistry Letters, 13 (14), 2341-2345. [20] Wang, G., Liu, Z., Chen, T. T., Wang, Z., Yang, H. Y., Zheng, M. Y., ..… Jiang, H. L. (2012). Design, synthesis, and pharmacological evaluation of monocyclic pyrimidinones as novel inhibitors of PDE5, Journal of Medicinal Chemistry, 55 (23), 10540-10550.

114 [21] Sarazan, R. D., Crumb, W. J., Beasley, C. M., Emmick, J. T., Ferguson, K. M., Strnat, C. A., and Sausen, P. J. (2004). Absence of clinically important HERG channel blockade by three compounds that inhibit phosphodiesterase 5-sildenafil, tadalafil, and vardenafil, European Journal of Pharmacolog, 502 (3), 163-167. [22] Geelen, P., Drolet, B., Rail, J., Berube, J., Daleau, P., Rousseau, G., ..… Turgeon, J. (2000). Sildenafil (viagra) prolongs cardiac repolarization by blocking the rapid component of the delayed rectifier potassium current, Circulation, 102 (3), 275-277. [23] Witchel, H. J. (2011). Drug-induced hERG block and long QT syndrome, Cardiovascular Therapeutics, 29 (4), 251-259. [24] Flagg, T. P., and Nichols, C. G. (2005). Sarcolemmal K-ATP channels: What do we really know?, Journal of Molecular and Cellular Cardiology, 39 (1), 61-70. [25] Sanguinetti, M. C., and Tristani-Firouzi, M. (2006). hERG potassium channels and cardiac arrhythmia, Nature, 440 (7083), 463-469. [26] Qiu, Y., Bhattacharjee, S., Kraft, P., John, T. M., Haynes- Johnson, D., Jiang, W., ..… Lundeen, S. (2006). JNJ-10280205 and JNJ- 10287069: novel PDE5 inhibitors as clinical candidates for erectile dysfunction, International Journal of Impotence Research, 18 (5), 477-483. [27] Mizuno, H., Adachi, H., Kimura, J., Sawa, Y., Kakiki, M., Lansdell, K., ..… Kerns, W. D. (2003). Cardiovascular assessment of ER-118585, a selective phosphodiesterase 5 inhibitor, Biological & Pharmaceutical Bulletin, 26 (12), 1661-1667. [28]Chemical Computing Group. (2015) Molecular Operating Environment, 2013.08. Montreal, QC. [29] Wang, H. C., Liu, Y. D., Huai, Q., Cai, J. W., Zoraghi, R., Francis, S. H., ….. Ke, H. (2006). Multiple conformations of phosphodiesterase-5 - Implications for enzyme function and drug development, Journal of Biological Chemistry, 281 (30), 21469-21479. [30] Jones, G., Willett, P., Glen, R. C., Leach, A. R., and Taylor, R. (1997). Development and validation of a genetic algorithm for flexible docking, Journal of Molecular Biology, 267 (3), 727-748. [31] Morris, G. M., Huey, R., Lindstrom, W., Sanner, M. F., Belew, R. K., Goodsell, D. S., and Olson, A. J. (2009). AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility, Journal of Computational Chemistry, 30 (16), 2785-2791. [32] Berendsen, H. J., van der Spoel, D., and van Drunen, R. (1995). GROMACS: A message-passing parallel molecular dynamics implementation, Computer Physics Communications, 91 (1-3), 43-56.

[33] Schüttelkopf, A. W., and Van Aalten, D. M. (2004). PRODRG:A tool for high-throughput crystallography of protein-ligand complexes, Acta Crystallographica Section D: Biological Crystallography, 60 (8), 1355-1363.

115 [34] Jo, S., Kim, T., Iyer, V. G., and Im, W. (2008). CHARMM-GUI: A web-based graphical user interface for CHARMM, Journal of Computational Chemistry, 29 (11), 1859-1865. [35] Rastelli, G., Rio, A. D., Degliesposti, G., and Sgobba, M. (2010). Fast and accurate predictions of binding free energies using MM-PBSA and MM-GBSA, Journal of Computational Chemistry, 31 (4), 797-810. [36] Schrödinger, P. (2014). (Version 3.5). New York, NY: LLC. [37] Durdagi, S., Deshpande, S., Duff, H. J., and Noskov, S. Y. (2012). Modeling of open, closed, and open-inactivated states of the hERG1 channel: Structural mechanisms of the state-dependent drug binding, Journal of Chemical Information and Modeling, 52 (10), 2760-2774. [38] Treptow, W., and Klein, M. L. (2012). Computer simulations of voltage-gated cation channels, Journal of Physical Chemistry Letters, 3 (8), 1017- 1023. [39] Yunomae, K., Ichisaki, S., Matsuo, J., Nagayama, S., Fukuzaki, K., Nagata, R., and Kito, G. (2007). Effects of phosphodiesterase (PDE) inhibitors on human ether-a-go-go related gene (hERG) channel activity, Journal of Applied Toxicology, 27 (1), 78-85. [40] Lees-Miller, J. P., Duan, Y. J., Teng, G. Q., and Duff, H. J. (2000). Molecular determinant of high-affinity dofetilide binding to HERG1 expressed in Xenopus oocytes: Involvement of S6 sites, Molecular Pharmacology, 57 (2), 367-374. [41] Kamiya, K., Mitcheson, J. S., Yasui, K., Kodama, I., and Sanguinetti, M. C. (2001). Open channel block of HERG K+ channels by vesnarinone, Molecular Pharmacology, 60 (2), 244-253. [42] Milnes, J. T., Crociani, O., Arcangeli, A., Hancox, J. C., and Witchel, H. J. (2003). Blockade of HERG potassium currents by fluvoxamine: Incomplete attenuation by S6 mutations at F656 or Y652, British Journal of Pharmacology, 139 (5), 887-898. [43] Sanchez-Chapula, J. A., Ferrer, T., Navarro-Polanco, R. A., and Sanguinetti, M. C. (2003). Voltage-dependent profile of human ether-a-go-go-related gene channel block is influenced by a single residue in the S6 transmembrane domain, Molecular Pharmacology, 63 (5), 1051-1058. [44] Kikuchi, K., Nagatomo, T., Abe, H., Kawakami, K., Duff, H. J., Makielski, J. C., ..… Nakashima, Y. (2005). Blockade of HERG cardiac K+ current by antifungal drug miconazole, British Journal of Pharmacology, 144 (6), 840–848. [45] Ferrer-Villada, T., Navarro-Polanco, R. A., Rodriguez-Menchaca, A. A., Benavides-Haro, D. E., and Sanchez-Chapula, J. A. (2006). Inhibition of cardiac HERG potassium channels by antidepressant maprotiline, European Journal of Pharmacology, 531 (1-3), 1-8. [46] Su, Z., Chen, J., Martin, R. L., McDermott, J. S., Cox, B. F., Gopalakrishnan, M., and Gintant, G. A. (2006). Block of hERG

116 channel by ziprasidone: Biophysical properties and molecular determinants, Biochemical Pharmacology, 71 (3), 278-286. [47] Kamiya, K., Niwa, R., Morishima, M., Honjo, H., and Sanguinetti, M. C. (2008). Molecular determinants of hERG channel block by terfenadine and cisapride, Journal of Pharmacological Sciences, 108 (3), 301-307. [48] Knape, K., Linder, T., Wolschann, P., Beyer, A., and Stary-Weinzinger, A. (2011). In silico analysis of conformational changes induced by mutation of aromatic binding residues: Consequences for drug binding in the hERG K+ channel, Plos One, 6 (12): e28778. doi: 10.1371/journal.pone.0028778. [49] Shin, D. S., Park, M. J., Lee, H. A., Lee, J. Y., Chung, H. C., Yoo, D. S., ..… Bae, M. A. (2014). A novel assessment of nefazodone-induced hERG inhibition by electrophysiological and stereochemical method, Toxicology and Applied Pharmacology, 274 (3), 361-371. [50] Cavalli, A., Poluzzi, E., De Ponti, F., and Recanatini, M. (2002). Toward a pharmacophore for drugs inducing the long QT syndrome: Insights from a CoMFA study of HERG K+ channel blockers, Journal of Medicinal Chemistry, 45 (18), 3844-3853. [51] Cavalli, A., Buonfiglio, R., Ianni, C., Masetti, M., Ceccarini, L., Caves, R., ….. Recanatini, M. (2012). Computational design and discovery of “Minimally Structured” hERG blockers, Journal of Medicinal Chemistry, 55 (8), 4010-4014. [52] Inanobe, A., Kamiya, N., Murakami, S., Fukunishi, Y., Nakamura, H., and Kurachi, Y. (2008). In silico prediction of the chemical block of human ether-a-Go-Go-related gene (hERG) K+ current, Journal of Physiological Sciences, 58 (7), 459-470. [53] Price, D. A., Armour, D., de Groot, M., Leishman, D., Napier, C., Perros, M., ..… Wood, A. (2006). Overcoming HERG affinity in the discovery of the CCR5 antagonist maraviroc, Bioorganic & Medicinal Chemistry Letters, 16 (17), 4633-4637. [54] Choe, H., Nah, K. H., Lee, S. N., Lee, H. S., Lee, H. S., Jo, S. H., ….. Jang, Y. J. (2006). A novel hypothesis for the binding mode of HERG channel blockers, Biochemical and Biophysical Research Communications, 344 (1), 72-78. [55] Durdagi, S., Subbotina, J., Lees-Miller, J., Guo, J., Duff, H. J., and Noskov, S. Y. (2010). Insights into the Molecular Mechanism of hERG1 Channel Activation and Blockade by Drugs, Current Medicinal Chemistry, 17 (30), 3514-3532. [56] Subbotina, J., Yarov-Yarovoy, V., Lees-Miller, J., Durdagi, S., Guo, J. Q., Duff, H. J., and Noskov, S. Y. (2010). Structural refinement of the hERG1 pore and voltage-sensing domains with ROSETTA-membrane and molecular dynamics simulations, Proteins-Structure Function and Bioinformatics, 78 (14), 2922-2934.

117 [57] Durdagi, S., Duff, H. J., and Noskov, S. Y. (2011). Combined receptor and ligand-based approach to the universal pharmacophore model development for studies of drug blockade to the herg1 pore domain, Journal of Chemical Information and Modeling, 51 (2), 463-474. [58] Durdagi, S., and Noskov, S. Y. (2011). Consistency of constructed hERG1 pore domain and pharmacophore models: A 3D-QSAR, molecular docking, and pharmacophore modeling study, Biochemistry and Cell Biology-Biochimie Et Biologie Cellulaire, 89, 266-267. [59] Durdagi, S., Deshpande, S., Duff, H., and Noskov, S. (2012). Development of atomistic models for closed, open and open-inactivated states of hERG1 channel using rosetta protein modeling suite and molecular dynamics simulations, Biophysical Journal, 102 (3), 679a-679a. [60] Durdagi, S., Randal, T., Duff, H. J., and Noskov, S. Y. (2013). Rehabilitation studies for withdrawn drugs from the market: derivation of non- hERG1 channel blocker cisapride analogues using multi-faceted approaches, Biophysical Journal, 104 (2), 266a-266a. [61] Dempsey, C. E., Wright, D., Colenso, C. K., Sessions, R. B., and Hancox, J. C. (2014). Assessing hERG pore models as templates for drug docking using published experimental constraints: The inactivated state in the context of drug block, Journal of Chemical Information and Modeling, 54 (2), 601-612. [62] Durdagi, S., Patterson, M., and Noskov, S. Y. (2014). Development and validation studies of universal pharmacophore models for hERG channel openers, Biophysical Journal, 106 (2), 15a-15a. [63] Guo, J. Q., Durdagi, S., Changalov, M., Perissinotti, L. L., Hargreaves, J. M., Back, T. G., ….. Duff, H. J. (2014). Structure driven design of novel human ether-A-go-gorelated- gene channel (hERG1) activators, Plos One, 9 (9), e105553. doi: 10.1371/journal.pone.0105553. [64] Guo, J. Q., Cheng, Y. M., Lees-Miller, J. P., Perissinotti, L. L., Claydon, T. W., Hull, C. M., ..… Duff, H. J. (2015). NS1643 interacts around L529 of hERG to alter voltage sensor movement on the path to activation, Biophysical Journal, 108 (6), 1400-1413. [65] Zhang, K. Y. J., Card, G. L., Suzuki, Y., Artis, D. R., Fong, D., Gillette, S., ….. Bollag, G. (2004). A glutamine switch mechanism for nucleotide selectivity by phosphodiesterases, Molecular Cell, 15 (2), 279-286. [66] Tuccinardi, T., Botta, M., Giordano, A., and Martinelli, A. (2010). Protein kinases: Docking and homology modeling reliability, Journal of Chemical Information and Modeling, 50 (8), 1432-1441. [67] Terrett, N. K., Bell, A. S., Brown, D., and Ellis, P. (1996). Sildenafil (VIAGRA(TM)), a potent and selective inhibitor of type 5 cGMP phosphodiesterase with utility for the treatment of male erectile dysfunction, Bioorganic & Medicinal Chemistry Letters, 6 (15), 1819- 1824. [68] Durdagi, S., Randall, T., Duff, H. J., Chamberlin, A., and Noskov, S. Y. (2014). Rehabilitating drug-induced long-QT promoters: In-silico

118 design of hERG-neutral cisapride analogues with retained pharmacological activity, Bmc Pharmacology & Toxicology, 15 (14), 1-15.doi: 10.1186/2050- 6511-15-14. [69] Wang, R., Burnett, A. L., Heller, W. H., Omori, K., Kotera, J., Kikkawa,K., ..… Peterson, C. A. (2012). Selectivity of Avanafil, a PDE5 inhibitor for the treatment of erectile dysfunction: implications for clinical safety and improved tolerability, Journal of Sexual Medicine, 9 (8), 2122-2129. [70] Corbin, J. D., Francis, S.H., and Webb, D. J. (2002). Phosphodiesterase Type 5 as a pharmacologic target in erectile dysfunction, Urology, 60 (2), 4- 11. [71] Rotella D. (2002). Phosphodiesterase 5 Inhibitors: Current Status and Potential Applications, Nature Reviews Drug Discovery, 1 (9), 674-682. [72] Rybalkin, S. D., Yan C., Bornfeldt, K. E., and Beavo, J. A. (2003). Cyclic GMP phosphodiesterases and regulation of smooth muscle function, Circulation Research, 93 (4), 280–291. [73] Srivani, P., Srinivas, E., Raghu. R., and Sastry, G. N. (2007). Molecular modeling studies of pyridopurinone derivatives-potential phosphodiesterase 5 inhibitors, Journal of Molecular Graphics and Modelling, 26 (1), 378–390. [74] Sakamoto, T., Koga, Y., Hikota, M., Matsuki, K., Mochida, H., Kikkawaet, K., ….. Yamada, K. (2015). 8-(3-Chloro-4- Methoxybenzyl)-8H- pyrido[2,3-d]pyrimidin-7-one derivatives as potent and selective phosphodiesterase 5 inhibitors, Bioorganic&Medicinal Chemistry Letters, 25 (7), 1431-1435. [75]Sakamoto, T., Koga, Y., Hikota, M., Matsuki, K., Murakami., M., Kikkawa, K., ….. Yamada, K. (2014). The discovery of avanafil for the treatment of erectile dysfunction: a novel pyrimidine- 5-carboxamide derivative as a potent and highly selective phosphodiesterase 5 inhibitor, Bioorganic&Medicinal Chemistry Letters, 24 (23), 5460- 5465. [76] Wang, Z., Zhu, D., Yang, X., Jianfeng, Li., Jiang, X., Tian, G., ….. Shen, J. (2013). The selectivity and potency of the new PDE5 inhibitor TPN729MA, Journal of Sexual Medicine,10 (11), 2790-2797. [77] Duan, H., Zheng, J., Lai, Q., Zheng, L., Tian, G., Wang, Z., Jianfeng, L., and Jingshan Shen, J. (2009). 2-Phenylquinazolin-4(3H)-One, a class of potent PDE5 inhibitors with high selectivity versus PDE6. Bioorganic & Medicinal Chemistry Letters, 19 (10), 2777-2779. [78] Choi, H., Lee, J., Kim, Y.H., Dai, S. I., Hwang, I., and Kim, S. J., ….. Lee, K.J. (2010). Discovery of potent, selective, and orally bioavailable PDE5 inhibitor: methyl-4-(3-chloro-4-methoxybenzylamino)-8-(2- hydroxyethyl)-7-methoxyquinazolin-6 ylmethylcarbamate (CKD 533), Bioorganic & Medicinal Chemistry Letters, 20 (1), 383-386. [79] Owen, D. R., Walker, J.K., Jon Jacobsen E., Freskos, J. N., Hughes, R. O., Brown, D.L., ….. Yung, Y. (2009). Identification, synthesis and SAR

119 of amino substituted pyrido[3,2b]pyrazinones as potent and selective PDE5 inhibitors, Bioorganic & Medicinal Chemistry Letters, 19 (15), 4088-4091. [80] Kim, Y. H., Choi, H., Lee, J., Hwang, I., Moon, S.E., Kim, S.J., ….. Choi, N. S. (2008). Quinazolines as potent and highly selective PDE5 inhibitors as potential therapeutics for male erectile dysfunction, Bioorganic & Medicinal Chemistry Letters, 18 (23), 6279-6282. [81] Arnold, N. J., Arnold, R., Beer, D., Bhalay, G., Collingwood, S. P., Craig, S., ….. Zurini , M. (2007). Potent and selective xanthine-based inhibitors of phosphodiesterase 5, Bioorganic & Medicinal Chemistry Letters, 17 (8), 2376-2379. [82] Giovannoni, M. P., Vergelli, C., Biancalani, C., Cesari, N., Graziano, A., Biagini, P., ….. Piaz, V.D. (2006). Novel pyrazolopyrimidopyridazinones with potent and selective phosphodiesterase 5 (PDE5) inhibitory activity as potential agents for treatment of erectile dysfunction, Journal of Medicinal Chemistry, 49 (17), 5363-5371. [83] Yang, G. F, Lu, H. T., Xiong, Y., and Zhan, C. G. (2006). Understanding the structure-activity and structure-selectivity correlation of cyclic guanine derivatives as phosphodiesterase-5 inhibitors by molecular docking, CoMFA and CoMSIA analyses, Bioorganic & Medicinal Chemistry, 14 (5), 1462-1473. [84] Xia G, Li J, Peng A, Lai, S., Zhang, S., Shen, J., ….. Ji, R,. (2005). Synthesis and phosphodiesterase 5 inhibitory activity of novel pyrido[1,2- e]purin-4(3H)-one derivatives, Bioorganic & Medicinal Chemistry Letters, 15 (11), 2790-2794. [85] Jiang, W., Alford, V.C., Qiu, Y., Bhattacharjee, S., John, T. M., Haynes- Johnson, D., ….. Sui, Z. (2004). Synthesis and SAR of tetracyclic pyrroloquinolones as phosphodiesterase 5 inhibitors, Bioorganic&Medicinal Chemistry, 12 (6), 1505-1515. [86] Pissarnitski, D. A., Asberom, T., Boyle, C.D, Chackalamannil, S., Chintala, M., Clader, J. W., ….. Xu, R. (2004). SAR development of polycyclic guanine derivatives targeted to the discovery of a selective PDE5 inhibitor for treatment of erectile dysfunction, Bioorganic & Medicinal Chemistry Letters, 14 (5), 1291-1294. [87] Daugan, A., Grondin, P., Ruault, C., Le Monnier de Gouville, A. C., Coste, H., Linger, J. M., ….. Labaudinie, R. (2003). The discovery of tadalafil: a novel and highly selective PDE5 inhibitor. 2:2,3,6,7,12,12a-hexahydropyrazino[10,20:1,6]pyrido[3,4-b]indole- 1,4-dione analogues, Journal of Medicinal Chemistry, 46 (21), 4533- 4542. [88] Maw, G.N., Allerton, C.M., Gbekor. E., and Million, W. A. (2003). Design, synthesis and biological activity of beta-carboline-based type-5 phosphodiesterase inhibitors, Bioorganic & Medicinal Chemistry Letters, 13 (8), 1425-1428.

120 [89] Jiang, W., Sui, Z., Macielag, M. J., Walsh, S. P., Fiordeliso, J. J., Lanter, J. C., ….. Clancy, J. (2003). Furoyl and benzofuroyl pyrroloquinolones as potent and selective PDE5 inhibitors for treatment of erectile dysfunction, Journal of Medicinal Chemistry, 46 (3), 441-444. [90] Wang, Y., Chackalamannil, S., Hu, Z., Boyle, C. D., Lankin, C. M., Xia, Y., ….. Wang, P. (2002). Design and synthesis of xanthine analogues as potent and selective PDE5 inhibitors, Bioorganic&Medicinal Chemistry Letters, 12 (21), 3149-3152. [91] Zhang, Z., and Artemyev, N. O. (2010). Determinants for phosphodiesterase 6 inhibition by its gamma-subunit, Biochemistry, 49 (18), 3862-3867. [92] Barren, B., Gakhar, L., Muradov, H., Boyd, K. K., Ramaswamy, S., and Artemyev, N. O. (2009). Structural basis of phosphodiesterase 6 inhibition by the C-terminal region of the gamma-subunit, EMBO Journal, 28 (22), 3613-3622. [93] Liu, Y.T., Matte, S.L, Corbin, J.D., Francis, S. H., and Cote, R. H. (2009). Probing the catalytic sites and activation mechanism of photoreceptor phosphodiesterase using radiolabeled phosphodiesterase inhibitors, Journal of Biological Chemistry, 284 (46), 31541-31547. [94] Zhang, X., Feng, Q., and Cote, R. H. (2005). Efficacy and selectivity of phosphodiesterase-targeted drugs in inhibiting photoreceptor phosphodiesterase (PDE6) in retinal photoreceptors, Investigative Ophthalmology & Visual Science, 46 (9), 3060-3066. [95] Granovsky, A. E., and Artemyev, N.O.A. (2001). A conformational switch in the inhibitory γ-subunit of PDE6 upon enzyme activation by transducin, Biochemistry, 40 (44), 13209-132015. [96] Cahill, K. B., Quade J. H., Carleton, K. L., and Cote, R.H. (2012). Identification of amino acid residues responsible for the selectivity of tadalafil binding to two closely related phosphodiesterases, PDE5 and PDE6, Journal of Biological Chemistry, 287 (49), 41406-41416. [97] Pissarnitski D. (2006). Phosphodiesterase 5 (PDE 5) inhibitors for the treatment of male erectile disorder: attaining selectivity versus PDE6, Medicinal Research Reviews, 26 (3), 369-395. [98] Kerr, N. M., and Danesh-Meyer, H. V. (2009). Phosphodiesterase inhibitors and the eye, Clinical Experimental Ophthalmology, 37 (5), 514-523. [99] Foresta, C., Caretta, N., Zuccarello, D., Poletti, A., Biagioli, A., Caretti, L., and Galan, A. (2008). Expression of the PDE5 enzyme on human retinal tissue: new aspects of PDE5 inhibitors ocular side effects, Eye (Lond), 22 (1), 144-149. [100]Makhlouf, A., Kshirsagar, A., and Niederberger, C. (2006). Phosphodiesterase 11: a brief review of structure, expression and function, International Jounal of Impotence Research, 18 (6), 501- 509. [101] Bischoff E. (2004). Potency, selectivity, and consequences of nonselectivity of PDE inhibition, International Jounal of Impotence Research, 16 (1), S11-S14.

121 [102] Ukita, T., Nakamura, Y., Kubo, A., Yamamoto,Y., Moritani,Y., Saruta,K., ….. Omori, K. (2001). Novel, potent, and selective phosphodiesterase 5 inhibitors: synthesis and biological activities of a series of 4-aryl-1-isoquinolinone derivatives, Journal of Medicinal Chemistry, 44 (13), 2204–2218. [103] Bunnage, M. E., Mathias, J.P., Wood, A., Miller, D., and Street, S. D. A. (2008). Highly potent and selective chiral inhibitors of PDE5: an illustration of Pfeiffer's rule, Bioorganic & Medicinal Chemistry Letters, 18 (23), 6033-6036. [104] Simon, A., Barabas, P., and Kardos, J. (2006). Structural determinants of phosphodiesterase 6 response on binding catalytic site inhibitors, Neurochemistry International, 49 (3), 215-222. [105] Chen, G., Wang, H., Robinson, H., Cai, J., Wan, Y., and Ke, H. (2008). An insight into the pharmacophores of phosphodiesterase-5 inhibitors from synthetic and crystal structural studies, Biochemical Pharmacology, 75 (9), 1717-1728. [106] Xiong, Y., Lu, H., and Zhan, C. (2008). Dynamic structures of phosphodiesterase-5 active site by combined molecular dynamics simulations and hybrid quantum mechanical/ molecular mechanical calculations, Journal of Computational Chemistry, 29 (8), 1259-1267. [107] Mittal, A., Paliwal, S., Sharma, M., Singh, A., Sharma, S., and Yadav, D. (2014). Pharmacophore based virtual screening, molecular docking and biological evaluation to identify novel PDE5 inhibitors with vasodilatory activity, Bioorganic & Medicinal Chemistry Letters, 24 (14), 3137-3141. [108] Li, Y., Wu, W., Ren, H., Wang, J., Zhang, S., Li, G., and Yang, L. (2012). Exploring the structure determinants of pyrazinone derivatives as PDE5 3HC8 inhibitors: an in silico analysis, Journal of Molecular Graphics and Modelling, 38, 112-122. [109] Tömöri, T., Hajdu, I., Barna, L., Lorincz, Z., Cseh, S., and Dormán, G. (2012). Combining 2D and 3D in silico methods for rapid selection of potential PDE5 inhibitors from multimillion compounds’ repositories: biological evaluation, Molecular Diversity, 16 (1), 59-72. [110] Antunes, J.E., Freitas, M.P, da Cunha, EFF., Ramalho, T. C., and Rittner, R. (2008). In silico prediction of novel phosphodiesterase type-5 inhibitors derived from sildenafil, vardenafil and tadalafil, Bioorganic Medicinal Chemistry, 16 (16), 7599-7606. [111] Yoo, J., Thai, K.M., Kim, D.K., Lee, J. Y., and Park, H. J. (2007). 3D- QSAR studies on sildenafil analogues, selective phosphodiesterase 5 inhibitors, Bioorganic Medicinal Chemistry Letters, 17 (15), 4271- 4274. [112] Zagrovic, B., and Van Gunsteren, W. F. (2007). Computational analysis of the mechanism and thermodynamics of inhibition of phosphodiesterase 5A by synthetic ligands, Journal of Chemical Theory and Computation, 3 (1), 301-311.

122 [113] Weeks, J. L., II, Corbin, J. D., and Francis, S. H. (2009). Interactions between cyclic nucleotide phosphodiesterase 11 catalytic site and substrates or tadalafil and role of a critical Gln-869 hydrogen bond, Journal of Pharmacolology and Experimental Therapeutics, 331 (1), 133-141. [114] Irwin, J. J., Sterling, T., Mysinger, M.M., Bolstad, E.S., and Coleman, R.G. (2012). ZINC: a free tool to discover chemistry for biology, Journal of Chemical Information and Modeling, 52 (7), 1757-1768. [115] Sheridan, R. P., and Kearsley, S. K. (2002). Why do we need so many chemical similarity search methods?, Drug Discovery Today, 7 (17), 903-911. [116] Willett P. (2006). Similarity-based virtual screening using 2D fingerprints. Drug Discovery Today, 11 (23-24), 1046-1053. [117] Frisch, M.J., Trucks, G.W., Schlegel, H.B., Scuseria, G.E., Robb, M.A., Cheeseman, J.R., ..... Fox, D.J. (2009). [Gaussian Inc.], Wallingford, CT. [118] Card, G. L., England, B. P., Suzuki, Y., Fong, D., Powell, B., Lee, B., ….. Zhang, K. Y. J. (2004). Structural basis for the activity of drugs that inhibit phosphodiesterases, Structure, 12 (12), 2233-2247. [119] Wang, H., Ye, M., Robinson, H., Francis, S. H., and Ke, H. (2008). Conformational variations of both phosphodiesterase-5 and inhibitors provide the structural basis for the physiological effects of vardenafil and sildenafil, Molecular Pharmacology, 73 (1), 104-110. [120] The UniProt Consortium. (2015). UniProt: a hub for protein information, Nucleic Acids Research, 43, D204–D212. [121] Pettersen, E. F., Goddard, T.D., Huang, C.C., Couch, G. S., Greenblatt, D. M., Meng, E. C., and Ferrin, T. E. (2004). UCSF Chimera-a visualization system for exploratory research and analysis, Journal of Computational Chemistry, 25 (13), 1605-1612. [122] Lees-Miller, J. P., Subbotina, J.O., Guo, J., Yarov-Yarovoy, V., Noskov, S. Y., and Duff, H. J. (2009). Interactions of H562 in the S5 helix with T618 and S621 in the pore helix are important determinants of hERG1 potassium channel structure and function, Biophysical Journal, 96 (9), 3600-3610. [123] Durdagi, S., Guo, J., Lees-Miller, J., Noskov, S.Y., and Duff, H. (2012). Structure-guided topographic mapping and mutagenesis to elucidate binding sites for the Herg1 potassium channel (KCNH2) activator- NS1643, Journal of Pharmacology and Experimental Therapeutics, 342 (2), 441-452. [124] Anwar-Mohamed, A., Barakat, K.H., Bhat, R., Noskov, S.Y., Tyrrell, D.L., Tuszynski, J.A., and Houghton, M. (2014). A human ether-a-go-gorelated (hERG) ion channel atomistic model generated by long supercomputer molecular dynamics simulations and its use in predicting drug cardiotoxicity, Toxicology Letters, 230 (3), 382-392.

123 [125] Sastry, G.M., Adzhigirey Day, T., Annabhimoju. R., and Sherman, W. (2013). Protein and ligand preparation: parameters, protocols, and influence on virtual screening enrichments, Journal of Computer Aided Molecular Design, 27 (3), 221-234. [126] Movsesian, M. (2016). Novel approaches to targeting PDE3 in cardiovascular disease, Pharmacology&Therapeutics, 163, 74-81. [127] Boyes, S., and Loten, E.G. (1988). Purification of an insulin-sensitive cyclic AMP phosphodiesterase from rat liver, European Journal of Biochemistry, 174 (2), 303-309. [128] Grant, P.G., and Colman, R.W. (1984). Purification and characterization of a human platelet cyclic nucleotide phosphodiesterase, Biochemistry-Us 23 (8), 1801-1807. [129] Sun, B., Li, H., Shakur, Y., Hensley, J., Hockman, S., Kambayashi, J., ….. Liu, Y. (2007). Role of phosphodiesterase type 3 A and 3 B in regulating platelet and cardiac function using subtype-selective knockout mice, Cellular Signaling, 19 (8), 1765-1771. [130] Katayama, Y., Fujita, A., Ohe, T., Findlay, I., and Kurachi, Y. (2000). Inhibitory effects of vesnarinone on cloned cardiac delayed rectifier K1 channels expressed in a mammalian cell line, Journal of Pharmacology and Experimental Therapeutics, 294 (1), 339-346. [131] Irwin, J.J. and Shoichet, B.K. (2005). ZINC-A free database of commercially available compounds for virtual screening, Journal of Chemical Information and Modeling, 45 (1), 177-182. [132] Lomize, M.A., Pogozheva, I.D., Joo, H., Mosberg, H.I., and Lomize A.L. (2012). OPM database and PPM web server: resources for positioning of proteins in membranes, Nucleic Acids Research, 40 (D1), D370- D376. [133] Jo, S. and Kim, T., and Im, W. (2007). Automated builder and database of protein/membrane complexes for molecular dynamics simulations, PLoS One, 2 (9), e880. [134] Tripathi, S., and Srivastava, G. (2016). Molecular dynamics simulation and free energy landscape methods in probing L215 H, L217 R and L225 M bI-tubulin mutations causing paclitaxel resistance in cancer cells, Biochemical and Biophysics Research Communications, 476 (4), 273- 279. [135] Bello, M., Fragoso-Vazquez, M.J., and Correa Basurto, J. (2016). Energetic and conformational features linked to the monomeric and dimeric states of bovine BLG, International Journal of Biological Macromolecules, 92, 625-636. [136] David, C.C., and Jacobs, D.J. (2014). Principal component analysis: a method for determining the essential dynamics of proteins, Methods in Molecular Biology, 1084, 193-226. [137] Srikumar, P.S., Rohini, K., and Rajesh, P.K. (2014). Molecular dynamics simulations and principal component analysis on human laforin mutation W32G and W32G/K87A, Protein Journal, 33 (3), 289-295.

124 [138] Liu, M., Wang, L., Sun, X., and Zhao X. (2014). Investigating the impact of asp181 point mutations on interactions between PTP1B and phosphotyrosine substrate, Scientific Reports, 4:5095. [139] Doss, G.P., Rajith, B., Chakraborty, C., NagaSundaram, N., Ali, S.K., and Zhu, H. (2014). Structural signature of the G719 S-T790 M double mutation in the EGFR kinase domain and its response to inhibitors, Scientific Reports, 4:5868. [140] Kayık, G., Tüzün, N.Ş., and Durdagi, S. (2017). In silico design of novel hERG-neutral sildenafil-like PDE5 inhibitors, Journal of Biomolecular Structure and Dynamics, 35 (13), 2830-2852. [141] Kayık, G., Tüzün, N.Ş., and Durdagi, S. (2017). Investigation of PDE5/PDE6 and PDE5/PDE11 selective potent tadalafil-like PDE5 inhibitors using combination of molecular modeling approaches, molecular fingerprint-based virtual screening protocols and structure- based pharmacophore development, Journal of Enzyme Inhibition and Medicinal Chemistry, 32 (1), 311-330. [142] Kollman, P.A., Massova, I., Reyes, C., Kuhn, B., Huo, S.H., Chong, L., ….. Cheatham, T.E. (2000). Calculating structures and free energies of complex molecules: combining molecular mechanics and continuum models, Accounts of Chemical Research, 33 (12), 889-897. [143] Petrou, A., Geronikaki, A., Terzi, E., Guler, O.O., Tuccinardi, T., and Supuran, C.T. (2016). Inhibition of carbonic anhydrase isoforms I, II, IX and XII with secondary sulfonamidesincorporating benzothiazole scaffolds, Journal of Enzyme Inhibition and Medicinal Chemistry, 31 (6), 1306-1311. [144] Tuccinardi, T., Zizzari, A.T., Brullo, C., Daniele, S., Musumeci, F., Schenone, S., .…. Botta, M. (2011). Substituted pyrazolo3,4- bpyridines as human A1 adenosine antagonists: developments in understanding the receptor stereoselectivity, Organic Biomolecular Chemistry, 9 (12), 4448-4455. [145] Da Silva Figueiredo Celestino Gomes, P., Chauvot De Beauchene, I., Panel, N., Lopez, S., De Sepulveda, P., Geraldo Pascutti, P., and Solary, E. (2016). Insight on mutation-induced resistance from molecular dynamics simulations of the native and mutated CSF-1R and KIT, PLoS One, 11 (7): e0160165. [146] Ren, W., Truong, T.M., and H.W. Ai. (2015). Study of the binding energies between unnatural amino acids and engineered orthogonal tyrosyl- tRNA synthetases, Scientific Reports, 5:12632. [147] Milella, L., Milazzo, S., De Leo, M., Vera Saltos, M.B., Faraone, I., Tuccinardi, T., .…. De Tommasi, N. (2016). α-Glucosidase and α- amylase inhibitors from arcytophyllum thymifolium, Journal of Natural Products, 79 (8), 2104-2112. [148] Wang, J.M., Morin, P., Wang, W., and Kollman, P.A. (2001). Use of MM- PBSA in reproducing the binding free energies to HIV-1 RT of TIBO derivatives and predicting the binding mode to HIV-1 RT of efavirenz

125 by docking and MM-PBSA, Journal of the American Chemical Society, 123 (22), 5221-5230. [149] Tuccinardi, T., Manetti, F., Schenone, S., Martinelli, A., and Botta, M. (2007), Construction and validation of a RET TK catalytic domain by homology modeling, Journal of Chemical Information and Modeling, 47 (2), 644-655. [150] Tuccinardi, T., Bertini, S., Martinelli, A., Minutolo, F., Ortore, G., Placanica, G .…. Macchia, M. (2006). Synthesis of anthranylaldoxime derivatives as estrogen receptor ligands and computational prediction of binding modes, Journal of Medicinal Chemistry, 49 (16), 5001-5012. [151] Kumari, R., and Kumar, R. (2014). Open source drug discovery, C.; Lynn, A. g mmpbsa-A GROMACS tool for high-Throughput MM-PBSA calculations, Journal of Chemical Information and Modeling, 54 (7), 1951-1962. [152] Baker, N.A., Sept, D., Joseph, S., Holst, M.J., and McCammon, J.A. (2001). Electrostatics of nanosystems: application to microtubules and the ribosome, Proceedings of the National Academy of Sciences, 98 (18), 10037-10041. [153] Wang, W., and MacKinnon, R. (2017). Cryo-EM structure of the open human ether-a-go-go-related K+ channel hERG, Cell, 169 (3), 422- 430.e10.

126

CURRICULUM VITAE

Name Surname : Gülru Kayık

Place and Date of Birth : İstanbul 21.05.1984

E-Mail : [email protected] , [email protected]

EDUCATION :

 B.Sc. : 2007, Istanbul University, Faculty of Engineering, Chemical Engineering Department  M.Sc. : 2011, Istanbul Technical University, Graduate School of Science Engineering And Technology, Polymer Science and Technology Programme

PROFESSIONAL EXPERIENCE AND REWARDS:

 University of PISA, Pisa, ITALY, Faculty of Pharmacy, Molecular Modelling & Virtual Screening Laboratory (June 2016-March 2017). Research Grant for Doctorate Students (2214-A) by the Scientific and Technological Research Council of Turkey (TÜBİTAK). Research Project Title: Rehabilitation of PDE5 Inhibitors: Minimization of hERG K+ Ion Channel-PDE5 Inhibitor Interactions Using Molecular Modelling Methods  Bahcesehir University, Faculty of Engineering and Natural Sciences, Istanbul, TURKEY - Teaching & Research Assistant-General Chemistry (2010-2015)

PUBLICATIONS, PRESENTATIONS AND PATENTS ON THE THESIS:  Kayık G., Tüzün N. S., Durdağı S. 2017. Structural investigation of vesnarinone at the pore domains of open and open-inactivated states of hERG1 K+ channel, Journal of Molecular Graphics and Modelling, 77, 399- 412. (article)  Kayık G., Tüzün N. S., Durdağı S. 2017. Investigation of PDE5/PDE6 and PDE5/PDE11 selective potent tadalafil-like PDE5 inhibitors using

127 combination of molecular modeling approaches, molecular fingerprint-based virtual screening protocols and structure-based pharmacophore development, Journal of Enzyme Inhibition and Medicinal Chemistry, 32 (1), 311–330. (article)  Kayık G., Tüzün N. S., Durdağı S. 2017. In silico design of novel hERG- neutral sildenafil-like PDE5 inhibitors, Journal of Biomolecular Structure and Dynamics, 35 (13), 2830-2852. (article)  Kayık G., Tüzün N. S., Durdağı S. 2017: On the Understanding of PDE Inhibitors-hERG K+ Ion Channel Interactions and Computational Design of Novel, Potent and Safe PDE5 Inhibitors. 5th International BAU Drug Design Congress, October 19-21, 2017 İstanbul-Turkey. (oral presentation)  Kayık G., Tüzün N. S., Durdağı S. 2015: Identification of Novel PDE5 Inhibitors with Reduced Side Effects by Homology Modeling, Virtual Screening and Docking Studies. IVEK 2nd International Convention of Pharmaceuticals and Pharmacies, November 27-29, 2015 İstanbul-Turkey. (poster presentation)  Kayık G., Tüzün N. S., Durdağı S. 2015: An In Silico Study on the Interactions of PDE5 Inhibitors with hERG1 K+ Ion Channel: Rehabilitating Sildenafil with Multi-Scale Molecular Modeling Methods. 3rd International BAU Drug Design Congress, October 1-3, 2015 İstanbul-Turkey. (poster presentation)

OTHER PUBLICATIONS, PRESENTATIONS AND PATENTS:

-Kayık G., Tüzün N. S. 2011: A DFT Study of Stereospecific Free Radical Polymerization of Acrylamide Derivatives. The Ninth Triennal Congress of the World Association of Theoretical and Computational Chemists-WATOC 2011, July 17- 22, 2011 Santiago de Compostela-SPAIN. (poster presentation)

-Kayık G., Tüzün N. S. 2010: A DFT Investigation of the Tacticity of Poly(N,N- Dimethylacrylamide) and the Effects of Tartrates on Stereoregularity of the Polymer Chain. 7th Congress on Electronic Structure:Principles and Applications, June 29- July 2, 2010 Oviedo-SPAIN. (poster presentation)

-Kurt B. Z., Gazioglu I., Dag A., Ekhteiari Salmas R. E., Kayık G., Durdagi S., Sonmez F. 2017. Synthesis, anticholinesterase activity and molecular modeling study of novel carbamate-substituted thymol/carvacrol derivatives, Bioorganic & Medicinal Chemistry, 25(4), 1352–1363. (article)

-Kayık G., Tüzün N. Ş. 2015. A Quantum Mechanical Study on the Propagation Kinetics of N-methylacrylamide: Comparison With N,N-Dimethylacrylamide in Free Radical Polymerization, Macromolecular Theory and Simulations, 24(3), 218-231. (article)

-Kayık G., Tüzün N. Ş. 2014. Stereoselective propagation in free radical polymerization of acrylamides: A DFT study, Journal of Molecular Graphics and Modelling, 49, 55-67. (article)

-Yucel B., Meral K., Ekinci D., Uzunoğlu G. Y., Tüzün N. Ş., Özbey S., Kazak C., Ozdemir Y., Sanli B., Kayık G., Dağdeviren M. 2014. Synthesis and

128 characterization of solution processable 6,11-dialkynyl substituted indeno [1,2-b] anthracenes, Dyes and Pigments, 100, 104-117. (article)

129