This item is only available as the following downloads:
1 MOLECULAR DYNAMICS SIMULATIONS AND VIRTUAL SCREENING TO IDENTIFY POTENT INHIBITORS OF HUMAN ASPARAGINE SYNTHETASE By LAMEES HEGAZY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILL MENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Lamees Hegazy
3 To the martyrs and wounded of the Egyptian revolution
4 ACKNOWLEDGMENTS First and foremost, I thank the almighty God for g iving me so many blessing s and perseverance to complete this journey Special thanks to the martyrs and wounded of the Egyptian revolution. Your sacrifice s for the sa ke of justice and freedom will always be remembered I would like to express my deepest gratitude to my advisor, Dr. Nigel Richards for his continuous help and support even after leaving the University of Florida. I could not accomplish this work without his encouragement, guidance and effort throughout the cours e of this work I am grateful to Dr. Adrian Roitberg for his support and acceptance to chair the committee after Dr. Nigel Richards Indiana University Purdue University I would like to extend my gratitude to all my committee members: Dr. D a vid Os trov, Dr. J on Stewart, Dr. Steve Bruner, Dr. Reb e cca Butcher and former member Dr. Kenneth Merz for their assistance and all of their contribution. I would also like to thank former and current members of the Richards group. Special thanks to Dr. Kathryn Williams for he r help with thesis editing I cannot express my appreciation to my father and my mother for their love, prayers, unconditional support, and believing in me. I also thank my brothers Sherief and M u ha mm ad and my sisters A laa and Sarah for thei r continuous love and support. I have been blessed with two little angles Albaraa and Darine who brought joy and happiness to my life. Finally, I would like to thank my husband Bahaa Elgendy His love support, patience and encouragement were definitely th e foundation upon which the past five years of my life have been built. Thank you will never be enough.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURE S ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 17 Asparagine Synthetase Structure and Function ................................ ...................... 17 Asparagine synthetase kinetic mechanism ................................ ............................. 21 Asparagine Synthetase Inhibitors ................................ ................................ ........... 24 Research Objectives ................................ ................................ ............................... 26 2 OPTIMIZED CHARMM GENERAL FORCE FIELD PARAMETERS FOR ACYLPHOSPHATE AND N PHOSPHONOSULFONIMIDOYL FUNCTIONAL GROUPS ................................ ................................ ................................ ................ 29 Background ................................ ................................ ................................ ............. 29 Molecular Mechanics Force Fields ................................ ................................ ......... 30 CGenFF Potential Energy Function ................................ ................................ .. 31 CGenFF Parametr Optimization Protocol ................................ ......................... 35 Computational Details ................................ ................................ ............................. 36 Results and Discussion ................................ ................................ ........................... 38 Identification of Missing Parameters ................................ ................................ 38 Charge Optimization ................................ ................................ ......................... 40 Optimization of Equilibrium Values and Force Constants for Bond Lengths and Angles ................................ ................................ ................................ .... 44 Optimization of Dihedral Terms ................................ ................................ ........ 50 Molecular Dynamics Simulations for Model Compounds in Aqueous Solution ................................ ................................ ................................ ......... 54 Summary ................................ ................................ ................................ ................ 57 3 MOLECULAR DYNAMICS SIMULATIONS OF UNLIGANDED ASPARAGINE ................................ ................ 60 Background ................................ ................................ ................................ ............. 60 Molecular Dynamics Simulation ................................ ................................ .............. 62 Computational Details ................................ ................................ ............................. 66 Results and Discussion ................................ ................................ ........................... 72 Root Mean Square Deviation (RMSD) ................................ ............................. 72
6 Root Mean Square Fluctuations (RMSF) ................................ .......................... 72 Mg 2+ Coordination ................................ ................................ ............................ 74 Hydrogen Bonds ................................ ................................ ............................... 79 Behavior of Important Residues in the Active Site of Unliganded ASNS and 2+ Complex. ................................ ...................... 86 Glu 348 ................................ ................................ ................................ ...... 86 Glu 352 ................................ ................................ ................................ ...... 89 Asp 384 ................................ ................................ ................................ ...... 92 Lys 3 76 ................................ ................................ ................................ ...... 92 Lys 429 ................................ ................................ ................................ ...... 97 Ser 346 ................................ ................................ ................................ ...... 97 Ser 234 and Ser 239 ................................ ................................ ................. 97 Lys 449 ................................ ................................ ................................ .... 103 Loop 443 452 Mobility ................................ ................................ .................... 108 Summary ................................ ................................ ................................ .............. 112 4 DISCOVERY OF A NEW INHIBITOR OF HUMAN ASPARAGINE SYNTHETASE IDENTIFIED BY STRUCTURE BASED VIRTUAL SCREENING 115 Background ................................ ................................ ................................ ........... 115 Structure Based Virtual Screening ................................ ................................ ........ 117 Computational Details ................................ ................................ ........................... 120 Virtual Screening Using GL IDE ................................ ................................ ...... 120 Virtual Screening Using DOCK ................................ ................................ ....... 122 Results and Discussion ................................ ................................ ......................... 123 Summary ................................ ................................ ................................ .............. 136 5 CONCLUSIONS AND FUTURE WORK ................................ ............................... 138 Concluding Remarks ................................ ................................ ............................. 138 Future Work: Free Energy Calculations to Estimate the Difference in Binding Free Energy between (R) and (S) Isomers of an N phosphonosulfonimidoyl Based Inhibitor ................................ ................................ ................................ ... 142 APPENDIX A QM AND MM OPTIMIZED VIBRATIONAL SPECTRA FOR N PHOSPHONOSULFONIMIDOYL MODEL COMPOUND. ................................ .... 151 B CHARMM TOPOLOGY AND PARAMETER FILE FOR ACYLPHOSPHATE AND N PHOSPHONOSULFONIMIDOYL MODEL COMPOUNDS ....................... 154 CHARMM Topology and Parameter File for Acylphosphate Model ...................... 154 CHARMM Topology and Parameter File for N phosphonosulfonimidoyl Mo del .... 156 LIST OF REFERENCES ................................ ................................ ............................. 158 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 167
7 LIST OF TABLES Table page 2 1 Comparison of HF/6 31G(d) and CHARMM interaction energies and interaction distances before and after optimization for the acylphosphate molecule. ................................ ................................ ................................ ............ 42 2 2 Comparison of HF/6 31G(d) and CHARMM interaction energies and interaction distances before and after optimization for the N phosphonosulfonimidoyl molecule. ................................ ................................ ..... 42 2 3 Op timized atomic partial charges for atoms in the acylphosphate and N phosphonosulfonimidoyl molecules. ................................ ................................ ... 43 2 4 Comparison of the CHARMM optimized geometry for missing parameters with the QM geom etry for the acylphosphate molecule. ................................ ..... 46 2 5 Comparison of the CHARMM optimized geometry for missing parameters with the QM geometry for the N phosphonosulfonimidoyl molecule. .................. 47 2 6 Vibrational spectra computed for the acylphosphate molecule at the scaled MP2 level and with optimized CGenFF parameters ................................ ........... 48 2 7 New bonded interaction parameters assigned for the acylphosphate moiety. .... 58 2 8 New bonded interaction parameters assigned for the N phosphonosulfonimidoyl moiety. ................................ ................................ ......... 59 3 1 Hydrogen bonding interactions of aspartyl AMP (BAA) with active site residues. ................................ ................................ ................................ ............. 82 3 2 Hydrogen bonding interactions of the p yrophosphate group (PPi) with amino acid residues in the active site. ................................ ................................ ........... 83 3 3 Hydrogen bonding interactions between the residues in the active site in MD ................................ ................................ ......... 84 3 4 Hydr ogen bonding interactions between the residues in the active site in MD simulations of unliganded ASNS. ................................ ................................ ...... 85 4 1 Docking score of the ten selected hits from GL IDE virtual screening, their Lipinski properties and their residual activity values. ................................ ........ 124 4 2 pKas of amide hydrogens of some s ulfonamide derivatives in water ............... 132 4 3 Docking score of some selected hits from DOCK virtual screening .................. 135
8 A 1 Vibrational spectra computed for the N phosphonosulfonimidoyl molecule at the scal ed MP2 level and with optimized CGenFF parameters ........................ 151
9 LIST OF FIGURES Figure page 1 1 Overview of the reaction catalyzed by glutamine dependent asparagine synthetase. ................................ ................................ ................................ ............ 20 1 2 Crystal structure of AS B ................................ ................................ ........................ 20 1 3 Proposed kinetic model for glutamine dependent asparagine syn thetase. ........... 23 1 4 Chemical structures of hASNS identified inhibitors. ................................ ............... 25 1 5 Conjugate bases of the functionalized acylsulfamate 2 and sulfamide 3 .............. 25 1 6 Pairwise sequence alignment between the E. coli and human asparagine synthetases. ................................ ................................ ................................ ........... 28 2 1 CGenFF par ameters optimization process ................................ ............................ 34 2 2 Parts of the molecules that require extensive opt imization are enclosed in boxes ................................ ................................ ................................ .................... 39 2 3 The equilibrium geometry and atomic lab els used in the text and tables .............. 39 2 4 Water model compound complexes ................................ ................................ ...... 41 2 5 Potential energy scans on the acylphosphate molecule dihedrals ......................... 51 2 6 Potential energy scans on the N phosphonosulfonimidoyl molecule dihedrals ...... 52 2 7 MD trajectory data showing that the phosphate moiety in the acylphosphate model undergoes rotation during the simulation ................................ .................... 55 2 8 MD trajectory data showing that the phosphate m oiety in the acylphosphate model undergoes rotation during the simulation. ................................ ................... 56 3 1 Structure based sequence alignment of LS and ASNS ................................ ....... 67 3 2 Stereoview of the AS LS active sites ................................ ........................ 68 3 3 Comparison between Chemical reactions of AS B and L S ................................ 70 3 4 Root mean square deviation of backbone atoms for unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ .. 73 3 5 RMS F of backbone atoms for first independent trajectory simulations for unliganded and ligand bound asparagine synthetase enzyme ............................. 75
10 3 6 RMSF of backbone atoms for the second independent traject ory simulations for unliganded and ligand bound asparagine synthetase enzyme .............................. 76 3 7 RMSF of backbone atoms for third independent trajectory simulations for unliganded and ligand bound asparag ine synthetase enzyme .............................. 77 3 8 Enzyme regions with changes in RMSF of backbone atoms greater than 0.5 are colored. ................................ ................................ ................................ ............ 78 3 9 Focu sed view of the Mg 2+ ................................ ........... 80 3 10 Hydrogen bonding interactions of AspAMP with active site residues. ................. 80 3 11 Interactions of PPi with active site residues.. ................................ ........................ 83 3 12 Glu 1 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 87 3 13 Glu 2 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 88 3 14 Glu 352 1 torsio n change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ .. 90 3 15 Glu 352 2 torsion change during MD simulations of the unliganded and ligand bound aspara gine synthetase enzyme ................................ ................................ .. 91 3 16 Asp 384 1 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 93 3 17 Asp 384 2 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 94 3 18 Lys 376 1 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 95 3 19 Lys 376 2 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ .. 96 3 20 Lys 1 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 98 3 21 Lys 4 2 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 99 3 22 1 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 100
11 3 23 Ser234 1 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 101 3 24 Ser239 1 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 102 3 25 Lys 449 1 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 104 3 26 Lys 449 2 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase en zyme ................................ ................................ 105 3 27 Lys 449 3 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 106 3 28 Lys 449 4 torsion change during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ................................ ................................ 107 3 29 The closed and open forms of loop 443 452 in ASNS. ................................ ....... 110 3 30 Structural overlay of the conserved lysine residue AS B, CMA AMP i LS and LS ... 110 3 31 Distance between K449 N and D351 C during MD simulations of the unliganded and ligand bound asparagine synthetase enzyme ........................... 111 4 1 Overlay of the re int ermediate using DOCK and the nativ e ................................ ................................ ................... 124 4 2 2D predicted binding poses of the ten selected hits from GLIDE virtual screening. ................................ ................................ ................................ ............ 125 4 3 Glutamine dependent production of PPi in the presence of the inhibitor NSC 605322 ................................ ................................ ................................ ................. 128 4 4 Ammonia dependent production of PPi in the presence of the inhibitor NSC 605322 ................................ ................................ ................................ ................. 129 4 5 Putative interactions of compound NSC605322 with ASNS active site residues. 131 4 6 Overlay of the docking pose of ASNS inhibitor complex and the ASNS ................................ ................................ .............................. 131 4 7 Conjugate base of ionized sulfonamide functional group of compound NSC605322. ................................ ................................ ................................ ........ 132 4 8 2D predicted binding poses of some hits identified by virtual screening using DOCK. ................................ ................................ ................................ ................. 134
12 5 1 Chemical structures of hASNS N phosphonosulfonimidoyl based inhibitors. ...... 141 5 2 Overview of the reaction catalyzed by glutamine dependent asparagine synthetase. ................................ ................................ ................................ .......... 141 5 3 Putative interactions of the four diastereoisomers with key res idues in the ASNS active site ................................ ................................ ................................ 143 5 4 Electrostatic potential in the synthetase active site of the AS B model ............... 145 5 5 Thermodynamic cycle used for the calculation of relative free energy of binding of (R) and (S) diatereoisomers of the phosphonosulfonimidoyl inhibitor ........... 147 5 6 P athway for the transmutation between the ( R ) and ( S ) phosphonosulfo nimidoyl inhibitor ................................ ................................ ......... 150
13 LIST OF ABBREVIATIONS ALL Acute lymphoblastic leukemia AMP CPP methyleneadenosine triphosphate AS B Escherichia coli gluta mine dependent ASNS ASNase L asparaginase ASNS Glutamine dependent asparagine synthetase ATP Adenosine triphosphate CEA N 2 (carboxyethyl) L arginine CGenFF CHARMM general force field DGPC Deoxyguanidinoproclavaminic acid hASNS Human aspargine synthetase HTS High throughput screening LBVS Ligand based virtual screening LJ Lennard Jones MD Molecular Dynamics MM Molecular mechanics PES Potential energy surface PPi Inorganic pyrophosphate QM Quantum mechanics RMSD Root Mean Square Deviation RMSF Root Mean Sq uare Fluctuations SBVS Structure based virtual screening UB Urey Bradley distances vdW van der Waals interactions VS Virtual screening
14 Beta aspartyl adenosine monophosphate LS Beta lactam synthetase
15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy MOLECULAR DYNAMICS SIMULATIONS AND VIRTUAL SCREENING TO IDENTIFY POTENT INHIBITORS OF HUMAN ASPARAGINE SYNTHETA SE By Lamees Hegazy December 2013 Chair: Adrian Roitberg Major: Chemistry Asparagine synthetase (ASNS) which mediate s the biosynthesis of L asparagine, has become of increas ing interest as a drug target. Overexpression of ASNS was observed in ASNase resistant leukemia cell lines and inhibitors of human asparagine synthetase were able to suppress the proliferation of ASNase resistant leukemia cells. Human asparagine synthetase is also involved in other forms of cancer, and was shown to be a predictive biomarker of L asparaginase activity in ovarian cancer cell lines. In addition, recent work suggested that ASNS is up regulated in the castration resistant stage of prostate cancer (CRPC) where knockdown of ASNS in asparagine depleted media le d to growth inhibition in in both androgen responsive lymph nodes and castrati on resistant prostate cancer cell lines Thus, depletion of asparagine using ASNS inhibitors may be a new approach for treatment of CRPC. In this study, new parameters that describe the acylphosphate and N phosphonosulfonimidoyl functional groups were dev eloped following the standard protocols of CHARMM general force field In addition, molecular dynamics simulations were performed on ASNS complexed with a key reaction intermediate The ligand free
16 ASNS was also simulated to detect related conformational c hanges as a result of intermediate binding. A model representative of the average structure was further used in virtual screening calculations to identify potent ASNS inhibitors. A total of 500,000 compounds from the NCI chemical database were docked to th e ASNS active site A new lead inhibitor of hASNS was identified by t hese docking studies This lead inhibitor has a new structural motif that should pe rmit the development of compounds with improved cell permeability than those previously reported.
17 CHAPTE R 1 INTRODUCTION Asparagine Synthetase Structure and Function A spar a gine synthetase is an ATP dependent enzyme that catalyzes the synthesis of L asparagine from L aspartate with g lutamine or ammonia as the nitrogen source ( Figure 1 1 ) 1 Asparagine synthetase catalyzes three reactions: 1. Gln + H 2 O Glu + NH 3 2. ATP + Asp + NH 3 AMP + PPi + Asn 3. ATP + Asp + Gln AMP + PPi+ Asn + Glu Reaction 1 is the glutaminase reaction in which g lutamine is hydrolyzed to glutamate and am monia. Reactions 2 and 3 are the ammonia and glutamine dependent reactions of asparagine synth et ase respectively. The Escherichia coli enzyme, asparagine synthetase B (AS B) and the human enzyme (hASNS) employ glutamine as the primary source of nitrogen. Glutamine dependent asparagine sythetase catalyze s the transfer of glutamine amide nitrogen to another substrate ( Figure 1 1) and is classified as Ntn or class II amidotransferase, because it has an N terminal conserved cysteine residue 2 Other members of this class include glutamine fructose 6 phosphate amidotransferase(GFAT) 3 glutamine phosphoribosyl pyrophosphate amidotransferase (GPAT) 4 and glutamine synth et ase. 5 Asparagine synthetase is important as a drug target. Overexpression of ASNS is observed in ASNase resistan t leukemia cell lines 6 and several studies suggest that inhibiting ASNS activity represents a possible approach for treating ASNase resistant acu te lymphoblastic leukemia (ALL) 7 8 9 The enzyme L asparaginase (ASNase), in combination with other drugs, has been commonly used in the chemotherapy of ALL 10 11 The molecular basis of ASNase chemotherapy is not completely understood,
18 but it is believed that the malignant lymphocytes produce asparagine at lower level s than normal lymphocytes. Therefore, ASNase treatment causes asparagine reduction in both nor mal and malignant cells N ormal cells are able to produce asparagine in adequate amount to respon d to this asparagine depletion. H owever, cancer cells produce asparagine in lower rate leading to suppression of the growth of leukemia cells 7 12 and continuous treatment with ASNase leads to remission 13 C hemotherapy of ALL using Escherichia coli L asparaginase in combination with other drugs has been very successful with remission rate s of 95% 10 14 However, 12% of patients suffer from development of drug resistance during the treatment leading to the failure of treatment protocol 15 16 17 Several studies have reported a correlation between ASNase resistance and elevated expression level s of glutamine dependent asparagine synthetase (ASNS), which catalyz es the biosynthesis of L asparagine ( Figure 1 1) 18 21 Kilberg et al have shown that asparagine synthetase expression alone is sufficient to induce L asparaginase r esistance in MOLT 4 human leukaemia cells. In their study the drug resistance phenotype was induced when parental MOLT 4 cells were transduced by ASNS expressing virus 6 In addition, a potent inhibitor of huma n asparagine synthetase (hASNS), compound 1 suppressed the proliferation of ASNase resistant leukemia cells ( Figure 1 4 ) 22 This finding supports the hypothesis that inhibitors of ASNS can be used in drug disc overy for treatment of ALL. Hum an asparagine synthetase is also involved in other forms of cancer and was shown to be a predictive biomarker of L asparaginase activity in ovarian cancer cell lines 23 24 In addition, recent work suggest ed that ASNS is up regulated in the castration resistant stage of prostate cancer (CRPC) in which k nockdown of ASNS by
19 small interfering RNAs in asparagine depleted media le a d s to gro wth inhibition in both androgen responsive lymph node s and castration resistant prostate cancer c ell lines Thus depletion of asparagine using ASNS inhibit ors may be a new approach for treatment of CRPC 25 Stu dying the structural and kinetic properties of the h ASNS has been difficult due to low availability and purification problems Alternatively, s tudies on Escherichia coli glutamine dependent ASNS (AS B) have provided detailed information about the structure and reaction mechanism of the enzyme. The N terminal cysteine residue is conserved among the Ntn amid otransferase family and is the nucleophile for hydrolysis of glutamine. Therefore, the C1A mutant of the enzyme AS B does not have glutaminase act ivity. I t does however has high glutaminase affinity and was crystal l ized with glutamine bound in the N terminal glutaminase domain and AMP bound in the C terminal synthase domain with resolution of 2.0 (Figure 1 2 ) 1 As expected from previous studies using monoclonal antibodies of bovine ASNS and from sequence alignment, the crystal structure of C1A mutant of AS B (PDB 1CT9) showed that AS B has two distinct domains 1 Each domain has a separate active site, the N terminal glutaminase domain and the C terminal synth etas e domain. The N terminal active site catalyzes the glutamine hydrolysis to glutamate and ammonia. The N terminal active site catalyzes the glutamine hyd rolysis to glutamate and ammonia. The N structure is similar to that of the N terminal amidohydrolase superfamily 26 27
20 Figure 1 1. Overview of the reaction catalyzed by glutamine dependent asparagine synthetase. Figure 1 2 Crystal structure of AS B (PDBID; 1CT9). Glutaminase domain is shown in purple and syntha se domain in brown. Glutamine and AMP are shown in ball and stick representation. I mage rendered in CHIMERA 28
21 The C terminal domain uses ATP and Mg 2+ as cofactors and catalyzes the conversion of aspartate to aspartyl AMP P). The C Terminal active site tertiary structure is observed in ATP pyrophosphatases such as GMP synthetase 29 30 arginosuccinate synthetase 29 lactam synthetase 31 and carbenam synthetase 32 The two active sites are sep arated by a solvent inaccessible tunnel through which ammonia moves from the glutaminase domain to the synth et ase domain. 1 Asparagine synthetase kinetic mechanism An u nderstanding of the steady state kinetics of the glutamine dependent ASNS, and identification of key residues involved in the bindin g of the intermediate sp AMP will aid in the design of selective inhibitors of ASNS. Asparagine synthetase is characterized by high glutaminase activity in the abs ence of aspartate. In addition, glutamine production is not strictly coordinated with asparagine production 33 34 S tudies using human asparagine synthetase (hASNS) incubated with glutamine in the absence of aspartate showed that the glutaminase activity increases. 34 This behavior was also observed in the Escherichia coli ASNS (AS B) where the Glu/Asn ratio increased with increasing concentration of L glutamine, reaching a limiting value of approximately 1.8 35 The weak coordination between the glutaminase and synthase domains is unusual and contr ast s with other gluta mine depen dent amidotransferases such as glutamine 5 phosphoribosylpyrophosphate amidotransferase (GPATase), in which higher glutaminase activity is observed in the presence of substrates or reactive intermediates in the synthase domain 36 37 A nother important characteristic of ASNS behavior is the lack of ATP/PPi exchange. 38 Some of class II glutamine amidotransfera se subfamily of enzymes such as GPATase 39 and
22 xanthylate synthetase exhibit this behavior 40 41 The absence of ATP/PPi exchange le ads to the assumption that PPi is released as the final product from the enzyme. This hypothesis is supported by the crystal structures of the structural ly lactam synthetase bound with methyleneadenosine 5 triphosphate AMP/PPI) and lactam synthetase complex ed with the product deoxyguan idinoproclavaminic acid (DGPC) both of them show that PPi is located deep in the active site 42 This contrasts with the ASNS from Vibrio c holerae where PPi release was reported to occur before glutamine binding 43 The amino acid residues lining the binding pocket of PPi are also conserved in AS B and oth er A TP pyrophosphatase enzymes The characteristic sequence SGGXDS is referred to as the P P motif 44 Several kinetic mechanisms have been proposed for AS B. However kinetic simulations, showed that only one kinet ic model was consistent with all observed experimental data for AS B (F igure 1 3 ) and was also able to reproduce the glutamine d ependent Glu/Asn stoichiometry at saturating levels of aspartate and ATP. 45 R ate constants for each of the steps in this kinetic model were assigned from literature data on the steady state kinetics of AS B 34 46 The production of P exhibi ts burst kinetics in the absence of a nitrogen source, 47 where the rate of formation exceeds that of its hydrolyti c breakdown and product release. Therefore the turnover number was assigned to k 7 which include s the steps involved in the release of asparagine, AMP, PPi, and glutamate from the enzyme active sites (Figure 1 3 ) The rate constant k 3 was adjust ed by numerical simulation to imitate the experimental Glu/Asn stoichiometry
23 Figure 1 3 Proposed kine tic model for glutamine dependent asparagine synthetase. Reprinted with permission from Archive of Biochemistry and Biophysics vol 413. Tesson, A.R, Soper, T.S., Ciusteau, M., Richards, N.G.J. Pages 23 31. Copyright 2003.
24 In this kinetic model glutamine binds to the ternary complex E.ATP.ASP to yield the quaterna ry complex E.ATP.Asp.Gln, from which glutamine can be hydrolyzed to t hat the two active sites are wea kly coupled befo re the co mmits the enzyme to aspar agine formation, in agreement with the previously proposed model 35 48 The PPi is released as the final product there by rationalizing the lack of ATP/PPi exchange. This kinetic model also indicates that ASNS must bind with high affinity and stable analogs of this intermediate and the transition state may be potent ASNS inhibitors (Figure 1 1). Asparagine Synthet ase Inhibitors Kinetic studies suggested that the intermediate must be stabilized within the active site and that analog ue s for the ASNS intermediate or the transition state should act as tight binding inhibitors. 45 A n N phosphonosulfonimidoyl inhibitor 1 that is a stable analogue of the ASNS transition state inhibited hASNS with nanomolar potency ( Figure 1 4 ) 22 In addition, this inhibitor slowed the growth of ASNase resistant leukemia cells when used in high concentrations (100 1000 M) and when L asparagine was reduced i n the external medium by ASNase 49 These findings validated the hypothesis that inhibitors o f human asparagine synthetase (hASNS) can be used in the discovery of therapeutic drugs to treat ALL. According to one hypothesis N phosphonosulfonimidoyl 1 i s needed in high concentrations t o suppress the growth of ASNase resistant leukemia cells because the charged groups prevent its entry in to the cell 22 Further studies with acyl adenylate analogue 2 in which the phosphate group is replaced by a sulfamate group, inhibited hASNS with micromolar potency 50 Further kinetic studies to investigate their inhibi tory effect of compounds 3 and 4 showed that
25 Figure 1 4 Chemical structures of hASNS identified inhibitors Figure 1 5 Conjugate bases of the functionalized acylsulfamate 2 and sulfamide 3
26 compound 3 has much reduced inhibitor activity compared to compound 1 while compound 4 did not show any inhibitory effect at all 51 These results suggested that the negatively charged phosphate group is essential to inhibit hASNS and it was hypothesized that the catalytically important lysine (Lys 449 in AS B and Lys 466 in hASNS) make s a critical electrostatic interaction with th e negatively charged oxygen of the phosphate group 51 This suggestion is supported by the presence of resonance forms of compound s 2 and 3 ( Figure 1 5 ) which stabilize the negative charge on the sulfamate oxyg en when the acidic hydrogen of NH dissociates thereby allowing compound 2 and 3 to inhibit hASNS 51 Therefore, a negatively charged group that mimic the phosphate group should be considered in the future desig n of new asparagine synthetase inhibitors. Research O bjectives The main goal of this research i s i dentification of new lead inhibitors that can be further optimized as drugs to treat acute lymphoblastic leukemia (ALL). Prior identification of the nanomolar N phosphonosulfonimidoyl based inhibitor ( compound 1 ) is an important step in this process However, N phosphonosulfonimidoyl 1 is limited in its clinical utility, due to the high concentration necessary to exert its biological activity and the difficulty of its synthesis Therefore it is necessary to discover new leads that are specific and easier to synthesize and which have higher potency. Availability of t he structure of ASNS bound with the intermediate or transition state will aid in understanding st ructure activity relationships between the enzyme and bound intermediate and will facilitate efforts towards the design of new leads. Currently, t here is no available x ray crystal structure of human asparagine synthetase and the bacterial X ray crystal s tructure of AS B with AMP bound in the active site is the best available
27 structure. The active site residues involved in binding the substrate and intermediate are conserved in both enzymes (Figure 1 6) The goal of this work is to refine a model of the E. Coli ASNS sp AMP complex using CHARMM software However, there are no accurate CHARMM general force field (CGenFF ) parameters to represent the acyl phosphate functional grou p of the sp AMP intermediate In order to refine this model, new CGenFF parameters for the acylpho sphate functional group of the sp AMP intermediate must be developed. This model should be further refined using computational methods, such as molecular dynamics simulations. The refined model will then be used as the docking receptor in virt ual screening of chemical libraries to identify new lead inhibitors of hASNS. The specific goals of this research are the following: 1) D evelopment of new CGenFF parameters for the acyl phosphate functional group of the sp AMP intermediate. 2) Simulation s of the ASNS sp AMP complex by m olecular dynamics simulations to generate a refined model of ASNS with the bound intermediate and to understand the structure function relationship between ASNS and the bound intermediate. 3) Structure based v irtual scree ning using the refined model of ASNS sp AMP complex as the docking receptor to identify new lead inhibitors of human asparagine synthetase.
28 Figure 1 6 Pairwise sequence alignment between the E. coli and human asparagine synthetases. Residues involved in binding the substrate or the intermediate are marked by a star.
29 CHAPTER 2 OPTIMIZED CHARMM GENERAL FORCE FIELD PARAMETER S FOR ACYL PHOSPHATE AND N PHOSPHONOSULFONIMIDOYL FUNCTIONAL GROUPS Background Computational chemistry uses mathematical models impl emented via computer programs to investigate physical and chemical properties such as energy, dipole moment, charge distribution, vibrational frequencies and reactivity. The accuracy of the calculated properties depends on the quality of the computational methods used. Accurate c omputational methods include quantum mechanical (QM) methods where applying theory directly to compute chemical properties of interest. However, QM methods require extensive computer time; therefore their use is limited to chemical systems with no more than 100 atoms. On the other hand molecular mechanical (MM) methods use empirical equations, which implement experimental parameters with simple mathematical equations that can be applied on large systems such as proteins Empirical e nergy functions, unlike quantum mechanics, do no t include the nuclei and electrons in the energy function used to describe the system. R ather the atoms are the smallest particles in the empirical energy equation. In addition, compared to QM, the empirical energy equation s use simpler intra and inter molecular terms with highly optimized parameters to describe the mo lecular interactions. Therefore e mpirical energy calculations can be performed on very large biological systems in reasonable computer time an d with high accuracy. Examples of widely used empirical energy functions are CHARMM 53 AMBER 54 55 and OPLS AA 56 58 Reprinted in part with permission from Hegazy, L.; Richards, N. G. J. J. Mol. Model. 2013 doi: 10.1007/s00894 013 1990 x
30 G lutamine dependent asparagine synthetase (ASNS) wa s identified recently as a biomarker for ovarian cancer and it is also involved in the development of prostate cancer 24 25 In addition, elevated levels of ASNS have been observed in ASNase resistant acute lymphoblastic leuke mia 18 21 Glutamine dependent asparagine synthetase catalyzes the synthesis of asparagine from aspartate in the presen ce of glutamine as the predominant nitrogen source. A n N phosphonosulfonimidoyl based inhibitor which mimics the ASNS transition state, inhibited hASNS with nanomolar potency ( Figure 1 4 C hapter 1 ) 59 and was able to suppress the proliferation of asparaginase resistant MOLT 4 cells. C aspartyl AMP A spAMP) and the N phosphonosulfonimidoyl inhibitor have been hindered by the lack of parameters tha t accurately describe both functional groups. In this research, parameters A spAMP and the N phosphonosulfonimidoyl moieties were developed following the CHARMM general force field protocol The newly developed parameters will facilitate the detailed c omputational studies of inhibitors with N phosphonosulfonimidoyl derivatives and acyladenylat ion catalyzing enzymes such as tRNA aminoacyl synthetases 60 glutamine dependent NAD+ synthetas e, 61 62 and aminoa cyl tRNA transamidating enzymes 63 65 Molecular Mechanics Force Fields Mole cular mechanics use s potential energy functions to describe molecular interactions of large molecules. P arameters are included in the mathematical equations and the accuracy of the molecular interactions derived from these equations depends mainly on the q uality of the derived parameters. Together the mathematical equation and the parameters comprise a force field. There are many force fields available for
31 different types of chemical systems depending on the way the parameters were developed and optimized. For example there are extended atom force fields and additive all atom force fields 66 In extended atom force fields, the hydrogen atoms are not represented explicitly but are included with the atoms to which they are connected and the v an der Waals parameters and charges are modified to account for the deleted hydrogen atoms. The omitted hydrogen atoms are mainly bonded to elements with low electron negativities, while polar hydrogen atoms involved in hydrog en bonding are treated explicitly. This approach reduces the number of atoms and hence computer time significantly However because of improvements of c omputer resources, the additive all atom force fields, which represent all hydrogen atoms explicitly, ar e m ore commonly used to treat chemical and biological systems. CHARMM 52 AMBER 54 55 and OPLS AA 56 57 are widely used additive all atom force fields. CGenFF Potential Energy Function In this work, p arameterization process followed the CHARMM general force field (CGenFF) protocol. 67 CGenFF is an organic force field developed mainly to simulate drug like molecules in a biological environment represented by the CHARMM additive force fields The potential energy function of CGenFF is the s ame as that of the CHARMM force field for biological molecules thereby providing properly balanced electrostatic and van der Waals (vdW) interactions between t he ligand and the biomolecule. The potential energy function is a mathematical equation that des cribes the various intramolecular and intermolecular physical interactions of a particular system. The potential energy function, V(r) used in CGenFF is 67
32 The CGenFF potential energy function contains intramolecular or internal terms and intermolecular or external terms. The internal terms describe bon ds, valence angles and dihedral energies while external terms describe nonbo nded interactions such as electrostatic and v an der Waals ( vd W) interactions. The potential energy function contain s terms that are obtained from the known structure (X ray crystal structure, NMR structure or computationally modeled structures) such as bo nd lengths, b, valence Bradley or (1,3) distance, x and the distance between atoms r ij The remaining terms are called parameters and they are specific for each atom based on its type and the type s of the other atoms covalently bonded to it. Therefore, it is the quality of those parameters that affect the accuracy of the results calculated by the potential energy function. The parameters are accurate if they are able to reproduce target data o btained experimentally or calculated by quantum mechanical methods. These parameters are b 0 0 0 and x 0 which are the equilibrium values for bond lenght s, angles, improper dihedral angles and Urey Bradley distances, respective force constants ; K constant, multiplicity and phase of the dihedral a ngle ; q i and q j are the partial atomic charge s of atom s i and j, respectively ; D is the dielectric constant ; ij is the well depth (T he geometric mean between i and j ) Rmin, i,j is the radius in th e Lennard Jones (LJ)
33 6 12 term (T he arithmetic mean betwe en Rmin, i and Rmin, j ); and r ij is the distance between i and j. Bond stretching and angle bending are represented by harmonic potentials si nce the biological simulations we re performed at room temperature where the bonds and angles are close to equilibr ium. The Urey Bradley term (1 3 distance) is a quadratic function of the distance, x, between the first and third atoms connected by two bonds. The improper dihedral angle term is also a quadratic function, w in the case where three connected atoms l,i,k and fourth atom j connected to the central atom i, the improper dihedral angle is iljk. Both Urey Bradley and improper dihedral terms are used in limited cases when better vibrational spectra are needed such as o ut of plane motion 67 The dihedral or torsion function is a cosine function which model s the barriers and minima encountered for a rotation around the central bond of four atoms connected by three covalent bonds (1, 4 interactions) T he force constant, K indicates the height of the barrier; t he periodicity, n, is the number of barriers or minima per rotation; and t he is the location of the minima. The dihedral angle can be represented by a sum of dihedral terms with differe nt force constants, multiplicities an d phase s (Fourier series), thereby allowing accurate reproduction of the dihedral profile of the target data. The nonbonded interactions a re represented by Lennard Jones (LJ) and electrostatic potential s The LJ potential describes the interatomic potenti al between two atom s ; w here ij is the depth of the potential well and is the geometric mean of i and j. R min i,j is the distance at whi ch the LJ potential is minimum, evaluated as the arithmetic mean between Rmin, i and Rmin, j ; and r ij is the distance between i and j
34 Figure 2 1 CGenFF parameters optimization process a dapted from Vanommeslaeghe, K.; Hatcher, E.; Acharya, C.; Kundu, S.; Zhong, S.; Shim, J.; E. Dari an, O. Lopes, P.; Vorobyov, I; Mackerell, A. D., Jr. J. Comput. Chem. 2010 31 671.
35 The (R minij /r ij ) 12 term describes the interatomic Pauli short range repulsion while (R minij /r ij ) 6 descri bes the long range attraction. The electrostatic or Coulombic interact ions is the other term of the nonbonding interactions where D is the dielectric constant, q i and q j are the partial atomic charge s of atom s i and j, respectively. CGenFF Parametr Optim ization Protocol The aim of the CGenFF parameter optimization methodol ogy is accuracy and simplicity so that new drug like molecules can be optimized in reasonable computational time 67 Therefore, although the CHAR MM bimolecular force fields use experimental target data supplemen ted with QM data, CGenFF use s mainly QM results as target data with empirical scaling factors to mak e the QM data appropriate for bulk phase s ( Figure 2 1 ) Identification of missing parameters in the molecule of interest and selecti on of a model compound is the first step in the parameter optimization process. The model compound must be in its equilibrium geometry. Next, intermolecular parameters are optimized This involves optimization of partial atomic charges and Lennard Jones, LJ ( vd W) parameters. CGe nff encompass es a wide range of diverse atom types with experimentally driven LJ parameters. Therefore straight assignment of LJ parameters from already existing parameters is suitable LJ parameters optimization is limited in cases where the partial atom ic charge optimization could not be achieved within convergence criteria Partial atomic charges are optimized to reproduce the interaction energies and distance s of the model water complexes determined from QM calculations at the HF/6 31G(d) level This t heory level is used by the CHARMM bimolecular force filed when optimizing the charges of biological molecules like proteins and nucleic acids Therefore use of HF/6 31G(d) rather than a higher theory level in the parameter
36 optimization of drug like molecu les provide s balanced nonbonding interactions between the ligand and its biological environment 67 After optimization of the intermolecular parameters, the intramolecular parameters for bonds, angles and dihedrals are considered Bond and valence angle equilibrium parameters are optimized to reproduce the QM equilibrium geometry. The force constan ts for bonds, angles, UB distances and improper dihedral angles are optimized to reproduce the QM calculated vibrational spectra Finally force constants, multiplicit ies and phase s of the dihedral angles are optimized to reproduce the minima and barrier he ights of the QM potential energy surface (PES) Because t here is a strong correlation between the parameters, following optimization of the dihedrals, validation of the new parameters is performed by repeating the charge optimization calculations and re op timizing the charges if necessary. If charges are re optimized then the optimization of the intramolecular parameters i s repeated until the results converge 67 Computational Details The ParamChem web interfac e was used to assign atom types and to make initial guess es for charges and parameters 67 70 Molecular mechanics energy minimizations were performed using CHARMM ve rsion 35 53 Energy minimization employed 200 steps of the conjugate gradi e nt and Newton Raphson algorithm s to a n RMS gradient of 10 5 kcal/ (mol ) and with infinite cut off distance. 53 Vibrational analyses were performed with the VIBRAN and MOLVIB modules in CHARMM 71 All quantum mechanics calculations were performed with Gaussian 09 72 The QM geometry optimization for the acylphosphate and N phosphonosulfonimidoyl model compounds were performed using MP2/6 31+G(d) level of theory. 73 74 I nteractions with
37 water molecules were modeled using a series of monohydrate complexes with a water molecule (TIP3P geometry) 75 at each H bond donor and acceptor group with the model compounds in their MP2/6 31G (d) geometries The QM calculations were then performed on the monohydrate complexes using HF/6 31G(d) level of theory 76 77 The interaction energy was calculated a s the difference between the total energy of the complex and the sum of the monomer energies. The QM vibrational frequencies were calculated using MP2/6 31G(d) and scaled by 0.943 78 A p otential energy scan (PE S) on each selected dihedral was calculate d using the scan keyword (Opt = ModRedundant) with 15 steps and optimization at each increment at the MP2 /6 31+G(d) theory level. Single point energy calculations using MP2/cc pVTZ level was then employed for each optimized structure. As a further validation of the new CGenFF parameters obtained for the acyl phophate and N phosphonosulfonimidoyl model compounds, molecular dynamics (MD) simulations of these molecules in aqueous solution we re performed Each model co mpound (Figure 2 3 ) was solvated in an octahedral box (28 x 28 x 28 ) of TIP3P water molecules. 75 The solvated models were energy minimized by steepest descent (SD) and adopted basis N ewton Raphson (ABNR) algorithms. Periodic boundaries were used in all MD simulations, with the particle mesh Ewald method used to obtain electrostatic energies. 79 The Leapfrog algorithm was used to integrate equations of motion with 1fs time step. The SHAKE algorithm 80 was used to constrain the length of covalent bonds to hydro gen atoms to their equilibrium values. After heating to 300 K (30 ps), each system was equilibrated for a further 40 ps in the c anonical ensemble (NVT), CHARMM version 36b1 was performed in the
38 i sobaric isothermal ense mble(NPT) ensemble ( 16 ns for the acylphosphate model and 17 ns for the N phosphonosulfonimidoyl compound ). The Hoover thermostat was used to maintain the temperature around the average 57 81 Results and Discussion Ident if i cation of M issing P arameters The first step in paramet erization process is t he identification of the missing parameters an d selection of a model compound to be parameterized. The ParamChem web site based on CGenFF, was used to add atom types, assign ch arges and parameters to the missing parameters in the molecule by analogy to existing parameters in CGenFF 67 69 The ParamChem program produces high penalty scores to missing parameters so that parts of the molecule to be parameterized can be identified. The structure of t he selected compound subjected for param eter optimization should be in its equilibrium geometry a spartyl AMP and N phosphonosulfonimidoyl inhibitor were uploaded as mol2 format to the ParamChem utility and missing parameters were identifie d and represented by model compounds ( Figure 2 2 ) with e ach model compound at a global minimum on the potential energy surface. Different initial conformations of each molecule w ere optimized using MP2/6 31+G(d). The lowest energy structure from this conformational search was then subjected to frequency calculation to confirm that there were no imaginary frequencies, and thereby confirming th at the lowest energy structure corresponds to the equilibrium geometry (Figure 2 3 ).
39 A B Figure 2 2 Parts of the molecules that require extensive optimization are enclosed in box es A spAMP B) N phosphonosulfonimidoyl inhibitor A B Figure 2 3 The equilibrium geometry and atomic labels used in the text and tables A) Acylph osphate molecule (B) N phosphonosulfonimidoyl molecule
40 Charge Optimization Partial atomic c harges we re then optimized to reproduce the QM interaction ener gies and distances of the model single water complexes. W ater molecules (TIP3P geometry) 75 are placed at each H bond donor and acceptor group using the model compounds in their MP2/6 31G(d) geometr ies Five monohydrates were generated for each model compound and each H bond was optimized while fixing other degrees of freedom using Hf/6 31G( d) ( Figure 2 4 ). The interaction energy was calculated as the difference between the total energy of the complex and the sum of the monomer energies. Although higher theory level will provide more accurate results, using the Hartree Fock level of theory wi ll be consistent with t he CHARMM bimolecular force field and thereby providing well balanced non b onded interactions 67 The aliphatic hydrogen atom s were assigned a standard value of +0.09 with t he methyl and methylene carbons given the same values as assigned by the ParamChem tool to allow transferability 67 70 In addition, overall charge of the molecule ha s to add 1 and O2 and O7 we re assigned the same charge. Then MM charges were adjusted manually until the QM and MM interaction energies were wi thin satisfactory agreement. MM interaction energies should be within 0.2 k cal/mol of the corresponding QM interaction energy and MM interaction distances sh ould be 0.2 shorter than the respective QM interaction distances.
41 A B Figure 2 4 Water model compound complexes A ) Acyl phosphate molecule B ) N phosphonosulfonimidoyl molecule with O3 in the trigonal position. A monohydrate complex was used in each calculation.
42 Table 2 1 Comparison of HF/6 31G(d) and CHARMM interaction energies and interaction distances before and afte r optimization for the acylphosphate molecule. Interaction Energies (Kcal/mole) Interaction Distances(A) QM MM Optimized QM MM Optimized 11.36 12.17 11.5 1.89 1.70 1.70 11.82 11.42 11.49 1.89 1.70 1.70 7.82 11.6 5 7.81 2.03 1.70 1.77 7.74 5.46 7.95 2.18 2.57 1.91 5) O10..HOH 8.33 8.83 8.54 1.96 1.74 1.76 Average Difference 0.0 5 RMS Difference 0.2 1 Average Absolute Deviation 0.18 Table 2 2. Comparison of HF/6 31G(d) and CHARMM interact ion energies and interaction distances before and after optimization for the N phosphonosulfonimidoyl molecule. Interaction Energies (Kcal/mole) Interaction Distances(A) QM MM Optimized QM MM Optimized 13.0 15.5 12.8 1.83 1.64 1.71 2) O7 HOH 10.0 14.1 11.3 1.87 1.70 1.73 10.0 12.4 10.8 1.95 1.7 1.76 HOH(Trigonal) 9.95 8.3 8.15 2.02 1.74 1.81 HOH 10.0 15.2 10.1 2.07 1.84 1.99 5) O12.. HOH 7.8 6.6 8.0 1.98 1.82 1.77 Average Difference 0. 4 4 RMS Difference 0.5 3 Average Absolute Deviation 0.5 2
43 Table 2 3. Optimized atomic partial charges for atoms in the acylphosphate and N phosphonosulfonimidoyl molecules. Acylphosphate model N phosphonosulfonimidoyl model Atom Type Charge Atom Ty pe Charge P1 PG1 1.30 P1 PG1 0.20 O2 OG2P1 0.71 O2 OG2P1 0.46 O3 OG303 0.46 O3 OG303 0.28 C4 CG331 0.17 C4 CG331 0.17 H5 HGA3 0.09 H5 HGA3 0.09 H6 HGA3 0.09 H6 HGA3 0.09 O7 OG2P1 0.71 O7 OG2P1 0.46 O8 OG305 0.38 N8 NG2D1 0 .38 C9 CG2O2 0.34 C9 CG321 0.02 O10 OG2D1 0.48 H10 HGA3 0.09 C11 CG331 0.27 S11 SG3O2 0.12 H12 HGA3 0.09 O12 OG2P1 0.42 H13 HGA3 0.09 C13 CG331 0.11 H14 HGA3 0.09 H14 HGA3 0.09 H15 HGA3 0.09 H15 HGA3 0.09 H16 HGA3 0.09 H17 HGA2 0 .09 H18 HGA2 0.09 C19 CG331 0.27 H20 HGA3 0.09 H21 HGA3 0.09 H22 HGA3 0.09
44 As shown in T able 2 1, the interaction energies and distances of the model water complexes for the acylphosphate molecule are in excellent agreem ent with scaled QM data and satisf y the convergence criter ia. The interaction distances are almost 0.2 smaller than QM interaction distances, which is excellent. For the N phosphonosulfonimidoyl molecule, Table 2 2 shows that the interaction energies and distances for the optimized parameters compar e well with the QM data except the hydrogen bonds of O7 and O3 with water. The inter action energy of O7 with water is more favorab le than the QM value by 1 .3 k cal/mol The QM interaction energy value of the h ydrogen bond for the O7 water interaction i s smaller than the interaction energy value of O2 water hydrogen bond. It was not possible to optimize to convergence criteria while satisfying the constraint of keeping the charge s equal on O2 and O7. The N phosp honosulfonimidoyl molecule is crowded with different functional groups which make charge optimization more difficult than in the acylphosphate molecule However, the MM optimized charge s are significantly improved compared to the one s initially assigned by analogy. The charge optimization for of O3 was carried out using two conformations ( the trigonal geometry and lone pair position ) and a compromise between the two positions wa s considered during charge optimization. In general, the overall charge optimiza tion for both the acylphosphate and N phosphonosulfonimidoyl molecules yielded MM interaction energies and distances for the individual model water complexes that wer e in good agreement with the QM data ( T able 2 3) Optimization of Equilibrium Values and F orce Constants for Bond Lengths and Angles Following optimization of the partial atomic charges, intramolecular or bonded parameters are optimiz ed. Bonded parameters include equilibrium values and force
45 constants of bond lengths valence angles and torsion angles The equilibrium terms of bond lengths and angles for missing parameters were optimized to reproduce the ab initio equilibrium geometry. After optimization of the equilibrium terms, the MM optimized bond lengths and angles were in very close agreem ent with the QM data (Tables 2 4 and 2 5 ). Force constan ts for bonds, angles, Urey Bradley distances improper dihedrals and dihedrals containing terminal hydrogen atom s we re optimized based on the QM vibrational spectra. QM frequencies w ere calculated us ing MP2/6 31G(d) and were scaled by 0.943 to be appropriate for the bulk phase 78 The vibrational spectra are composed of the frequencies and their assignments as well as their numerical values. MOLVIB module in CHARMM was used to generate the potential energy distribution (PED) for the QM and MM spectra and to assign the frequenci es to the different contributions of the normal modes 71 82 Manual adjustment of the force constants was performed until agreement between the QM and MM spectra was obtained. In general, i t is very difficult to achieve quantitative convergence between the QM and MM vibrational spectra because the normal modes contribute differently to a given frequency in the two methods Generally, the MM vibrational frequencies sho uld be within 5% of their corresponding QM frequencies especially the lowest MM vibrational modes because these are the normal modes that involve conformational change s in the molecule. In addition, it is better for the MM vibrational frequencies to be lo wer than the QM frequencies rather than vice versa to generate a more flexible molecule and to make conformational change s during MD simulations more flexible. 67
46 Table 2 4 Comparison of the CHARMM optimized ge ometry for missing parameters with the QM geometry for the acylphosphate molecule. MP2/6 31+G* CHARMM Difference Bonds() p1 o8 1.75 1.75 0.00 O8 C9 1.34 1.34 0.00 Angles() O3 P1 O8 94.12 94.3 0.18 O2 P1 O8 105.99 107.6 1.61 O7 P1 O8 108.07 107.6 0.47 C9 O8 P1 124.67 124.95 0.28 O10 C9 O8 119.82 119.55 0.27 Dihedrals ( o ) P1 O8 C9 C11 2 2 0 P1 O8 C9 O10 178 178 0 O2 P1 O8 C9 68 66 2 O7 P1 O8 C9 68 68 0 O3 P1 O8 C9 179 177 2 O8 P1 O3 C4 69 71 2 Improper Torsions ( o ) C9 C11 O10 O8 0.1 0.0 0.1
47 Table 2 5 Comparison of the CHARMM optimized geometry for missing parameters with the QM geometry for the N phosphonosulfonimidoyl molecule. MP2/6 31+G* CHARMM Difference Bonds() p1 n8 1.75 1.75 0.00 n8 s11 1.53 1.53 0.00 Angles() O3 P1 N8 99.79 99.79 0.00 O2 P1 N8 109.0 110.62 1.62 O7 P1 N8 108.3 106.8 1.5 C9 S11 N8 112.17 111.8 0.37 P1 N8 S11 120 120 0 N8 S11 C13 111 111 0 N8 S11 O12 116 116 0 Dihedrals ( ) N8 S11 C9 C19 176 169 7 O2 P1 N8 S11 24 23 1 O7 P1 N8 S11 161 155 6 O3 P1 N8 S11 85 90 5 P1 N8 S11 O12 171 171 0 P1 N8 S11 C9 49 48 1 P1 N8 S11 C13 67 67 0 N8 P1 O3 C4 71 74 3
48 Table 2 6 Vibrational spect ra comput ed for the acylphosphate molecule at the scaled MP2 level and with optimized CGenFF a parameters MP2/6 31+G(d) scaled by a factor 0.943 CGenFF Freq b Assign (%) Assign (%) Assign (%) Freq b Assign (%) Assign (%) Assign (%) 67.0 tdOPOC (98) 68.2 tdCCO P (65) tdOPOC(29) 81.8 tdCCOP (73) tdCOPO (23) 79.8 tdCOPO (47) tdOPOC (36) scPO4' (7) 100.5 tdCOPO (55) tdCCOP (33) scPO4 (9) 106.5 tdOPOC (33) tdCOPO (32) tdCCOP (23) 140.2 tdPOCH (69) dPOC9 (14) scPO4 (11) 141.2 tdPOCH (57) scPO4 (31) dPOC9 (8) 180.2 dPOC9 (53) tdPOCH (24) tdCOPO (12) 207.9 tdPOCH (29) dPOC9 (18) rPO4 (16) 194.3 tdPOCH (89) 239.4 dPOC4 (33) dPOC9 (19) rPO4 (14) 219.0 dPOC4 (41) twPO4 (18) scPO4 (8) 268.0 tdPOCH (90) 295.4 scPO4 (38) twP O4 (20) dCCO (12) 289.4 dPOC4 (31) scPO4' (19) twPO4 (9) 329.7 twPO4 (36) dPOC4 (19) saOP (12) 323.4 dCCO (28) twPO4 (24) dPOC9 (21) 343.8 dCCO (19) ssOP (17) saOP (16) 334.8 twPO4 (38) wPO4 (28) dCCO (9) 414.2 scPO4 (42) wPO4 (20) rPO4 (11) 417.2 s sOP (30) scPO4 (24) rPO4 (20) 460.0 wPO4 (32) dCCO (24) rC=O (15) 471.9 wPO4 (36) dPOC4 (19) twPO4 (14) 505.7 scPO4 (21) wPO4 (16) dPOC4 (15) 499.9 scPO4 (44) dCCO (27) dPOC9 (6) 515.2 rPO4 (54) scPO4 (13) ssOP (10) 537.0 rC=O (39) saOP (12) scPO4 (1 1) 564.4 tiOCOC (79) rCH3 11 (14) 564.3 tiOCOC (90) 685.5 sCC (34) rC=O (25) saOP (16) 671.9 saOP (47) ssOP (11) rPO4 (9) 723.7 saOP (38) ssOP (32) sO3C (9) 679.6 ssCC (34) ssOP (18) rC=O (11) 906.9 sCC (33) sO8C (26) rC=O (11) 952.4 sO8C (34) s sPO (16) r'CH3C11 (11) 993.8 r'CH3C11 (38) ssPO (18) rCH3C11 (13) 996.4 r'CH3C11 (35) ssPO (34) rCH3C11 (10) 1035.2 sO3C (59) ssPO (25) r'CH3C11 (10) 1023.3 sO3C (23) ssPO (18) r'CH3C11 (13) 1043.3 rCH3C11 (53) r'CH3C11(20) tiOCOC (20) 1040.1 sO3C (5 4) ssPO (16) ssOP (8) 1048.6 ssPO (46) sO3C (27) ssOP (9) 1046.7 rCH3C11 (64) r'CH3C11 (19) ad'CH3C11(10) 1135.9 rCH3C4 (72) r'CH3C4 (25) 1139.0 rCH3C4 (40) r'CH3C4 (37) ad'CH3C4 (19) 1157.5 r'CH3C4 (66) rCH3C4 (24) 1144.8 r'CH3C4 (39) rCH3C4 (34) adCH3C4 (24) 1243.9 saPO (93) 1187.6 saPO (92) wPO4 (6)
49 Table 2 6 Continued. MP2/6 31+G(d) scaled by a factor 0.943 CGenFF Freq b Assign (%) Assign (%) Assign (%) Freq b Assign (%) Assign (%) Assign (%) 1258.8 sO8C (45) sCC (13) rC=O (12) 1232. 3 sO8C (33) sCC (30) rC=O (23) 1373.0 sdCH3C11 (86) sCC (8) sO8C (5) 1387.2 sdCH3C11 (98) 1421.8 sdCH3C4 (99) 1429.7 ad'CH3C4 (58) adCH3C4 (23) r'CH3C4 (15) 1447.1 ad'CH3C4 (92) rCH3C11 (6) 1434.3 adCH3C11 (91) 1460.0 adCH3C11 (91) 1451.9 ad'CH3C11(88) rCH3C11 (8) 1465.0 ad'CH3C4 (92) ad'CH3C4 (34) 1469.4 adCH3C4 (52) ad'CH3C4 (21) rCH3C4 (21) 1488.9 adCH3C4 (59) adCH3C4 (36) 1615.8 sdCH3C4 (88) sO3C (11) 1673.5 sC=O (84) 1741.2 sC=O (88) sCC (5) 2914.5 ssCH3C4 (100) 2854.1 ssCH3C4 (100) 2936.5 ssCH3C11(100) 2913.1 saCH3'C4 (71) saCH3C4 (29) 3001.6 saCH3C4 (99) 2915.8 ssCH3C11(100) 3012.5 saCH3C4 (99) 2917.3 saCH3C4 (71) saCH3'C4 (29) 3023.1 saCH3C11 (76) ssCH3C11 (24) 2973.4 ssCH3'C11 (75) saCH3C11 (25) 3041.8 ssCH3C11 (76) saCH3C11 (24) 2975.9 saCH3C11 (75) ssCH3'C11 (25) a Optimized vibrational contributions from interactions for which parameters have been developed in this study are shown in bold font ; s stands for bond stretching with the variations ss and sa for symmetric and asymmetric stretching, respectively ; d means angle deformation with the variations sd and ad for symmetric and asymmetric deformation, respectively ; td and ti stand for torsional and improper torsion deformation, respectively ; sc stands for scissoring, r for rocking ; w for wagging and tw for twisting. b Frequencies are expressed in units of cm 1
50 The o ptimized vibrational spectra for the acyl phosphate molecule compare very well with the QM spectra (Table 2 6). Likewise, t he optimized vibrational spectra for the N phosphonosulfonimidoyl molecule are in good agreement with the QM vibrational spectra (Table A 1 in the appendix ) Optimization of Dihedral Terms Dihedral parameters to be optimized include force const ants, multiplicit ies and phase s Di hedrals are optimized to reproduce the QM adiabatic potential energy surface (PES) especially the minima and barrier heights. The phase is usually assigned 0 or 180 so that different stereoisomers have same dihedral ene rgy with the same parameters. The selected dihedral wa s fixed at a specific value while constraining the other dihedrals at their equilibrium values and the angle was increased by 15 with optimization after each increment until the complete surface wa s o btained. The QM adiabatic potential energy surface (PES) was calculated on the global minimum structure using MP2/6 31+G(d), followed by single point calculations using MP2/cc pVTZ on the optimized geometries. The t arget for torsion optimization was the re production of the overall QM potential energy surface (PES) especially the minima and barrier heights H owever, in some cases it is impossible to reproduce the overall PES. In such situation, it is essential to reproduce the minima rather than barrier he ights, because barriers over than 8 k cal/mol will not be populated during MD simulations. The potential energy scans on the acylphosphate and N phosphonosulfonimidoyl dihedrals are shown (F igures 2 5 and 2 6 )
51 A B C Figure 2 5 Potential energy sc ans on the acylphosphate molecule dihedrals. QM PES (red), optimized (black) and initial (blue) MM PES. A) C11 C9 O8 P1 PES B) C9 O8 P1 O3 PES C) O8 P1 O3 C4 PES. Interaction labels correspond to the atom numbers in F igure 2 3
52 A B Figure 2 6 Poten tial energy scans on the N phosphonosulfonimidoyl molecule dihedrals. QM PES (red), optimized (black) and initial (blue) MM PES. A) C19 C9 S11 N8 PES B) C 13 S11 N8 P1 PES C) S11 N8 P1 O3 PES D) N8 P1 O3 C4 PE S Interaction labels correspond to the atom numbers in F igure 2 3 :
53 C D Figure 2 6 Continued
54 It is very important when performing the potential energy surface scan that the conformation be the same in QM and MM calculations. For the acyl phosphate molecule, the MM geometry after minimization underwent conformational change that wa s different from the equilibrium QM geometry making comparison of the QM and MM PES scans unreliable. Therefore, the MM geometry was constrained to the QM conformation during the CHARMM dihedral scan. The initial MM dihedral potential energy surfaces before optimization show ed large deviations from the QM torsion PES especially for the acylphosphate molecule. The O8 P1 O3 C4 dihedral for the acyl phosphate molecule ( F igure 2 5) was not a mi ssing parameter. Ho wever t o improve the agreement of this dihedral with the QM one, a new atom type was assigned to O8 s o that dihedral optimization could be performed. In general, it is preferable to assign a new atom type rather than modifying already existing parameters to pre serve the consistency of the force field and to allow transferability. This new atom type was assigned the same Lennard Jones parameters as O3. After optimization, the MM PES show ed significant improvement. In summary, the newly developed parameters were a ble to reproduce the QM geometries, vibrational spectra, dihedral PES and interaction energies of the model compounds with water. The topology and pa rameter files for the acylphosphate and N phosphonosulfonimidoyl models are provided (A ppendix B ) Molecula r Dynamics Simulations f or Model Compounds i n Aqueous Solution As a further validation of the new ly developed CGenFF parameters m olecular dynamics (MD) simulations of the acylphosphate and N phosphonosulfonimidoyl model compounds in aqueous solution were performed. R otation of the phosphate group was observed (Fig ures 2 7 and 2 8 ), and no major bond length or bond angle distortions occurred during the simulation.
55 A B C Figure 2 7 MD trajectory data showing that the phosphate moiety in the acylphospha te model undergoes rotation during the simulation. A) O2 P1 O8 C9 B) O3 P1 O8 C9, C) O7 P1 O8 C9. Dihedral angles are labeled with the atom numbers shown in F igure 2 3
56 A B C Figure 2 8 MD trajectory data showing that the phosphate moiety in the acy lphosphate model undergoes rotation during the simulation. A) O2 P1 N8 C9 B) O3 P1 N8 C9 C) O7 P1 N8 C9. Dihedral angles are labeled with the atom numbers shown in F igure 2 3
57 The temperature, pressure and total energy were monitored throughout the si mulations with no large fluctuations were observed In addition, the torsion angles for which new parameters had been developed fluctuated about values corresponding to minima on the potential energy surface Therefore these data suggests that these CGenF F parameters will be suitable for use in the simulated annealing in silico docking and free energ y perturbation calculations that will be undertaken as part of the drug discovery efforts. Summary New force field parameters for the N phosphonosulfonimidoy l and acylphosphate functional groups were developed. The parameterization process followed the CGenFF protocol with the ab initio results as the target data for the optimization The charges were optimized to reproduce the QM interaction energies and dist ance s for the model water complexes. The results indicate d that the new parameters were able to reproduce the QM geometries, vibrational spectra, potential energy surface s for the dihedrals and th e molecule water interactions. The new parameters were furth er validated by molecular dynamics simulations of the N phosphonosulfonimidoyl and acylphosphate models in aqueous solution The newly developed parameters will enable computational studies using molecular mechanics approaches, such as molecular dynamics s imulations and free energy calculations, on molecules containing the same functional groups using the CHARMM force field.
58 Table 2 7. New bonded interaction parameters assigned for the acylphosphate moiety. a b eq equilibrium bond distance () and K b force constant (kcal/mol/ 2 ). b eq equilibrium bond angle ( o ) and K force constant (kcal/mol/rad 2 ). c K torsional potential (kcal/mol), o ) of the torsion, respectively. d K improper dihedral potential (kcal/mol/rad 2 o equilibrium improper dihedral angle ( o ). Coordinate Atom Types Optimized Parameters Bonds b eq a K b P1 O8 PG1 OG305 1.78 170 C9 O10 CG2O2 OG305 1.34 230 Bond Angles eq b K x UB K x P1 O8 C9 PG1 OG305 CG2O2 121.5 70 O2 P1 O8 OG2P1 PG1 OG305 103.0 60 O3 P1 O8 OG303 PG1 OG305 90.8 60 O7 P1 O8 OG2P1 PG1 OG305 103.0 60 O8 C9 O10 OG305 CG2O2 OG2D1 118.0 70 2.26 160 O10 C9 C11 OG305 CG2O2 CG331 104.0 30 2.33 5 Dihedral Angles K c n P1 O8 C9 O10 PG1 OG305 CG2O2 OG2D1 1.30 1 180 P1 O8 C9 O10 PG1 OG305 CG2O2 OG2D1 2.60 2 180 P1 O8 C9 C11 PG1 OG305 CG2O2 CG331 3.80 1 180 P1 O8 C9 C11 PG1 OG305 CG2O2 CG331 1.60 2 180 O2 P1 O8 C9 OG2P1 PG1 OG305 CG2O2 0.10 3 0 O3 P1 O8 C9 OG303 PG1 OG305 CG2O2 0.10 2 180 O3 P1 O8 C9 OG303 PG1 OG305 CG2O2 0.10 3 0 C4 O3 P1 O8 CG331 OG30 3 PG1 OG305 1.47 2 0 C4 O3 P1 O8 CG331 OG303 PG1 OG305 0.70 3 0 Improper Torsion K d o O8 O10 C11 C9 OG305 OG2D1 CG331 CG2O2 56 0
59 Table 2 8. New bonded interaction parameters assigned for the N phosphonosulfonimid oyl moiety. a b eq equilibrium bond distance () and K R force constant (kcal/mol/ 2 ). b eq equilibrium bond angle ( o ) and K force constant (kcal/mol/rad 2 ). c K torsional potential (kcal/mol/rad 2 ) ; o ) of the torsion, respectively Coordinate Atom Types Optimized Parameters Bonds Atom Types b eq a K b P1 N8 PG1 NG2D1 1.72 100 N8 S11 NG2D1 SG3O2 1.53 400 Bond Angles Atom Types eq b K P1 N8 S11 PG1 NG2D1 SG3O2 113.0 30 O2 P1 N8 OG2P1 PG1 NG2D1 106.0 50 O3 P1 N8 NG2D1 PG1 OG303 98.8 94 O7 P1 N8 OG2P1 PG1 NG2D1 106.0 50 N8 S11 C9 NG2D1 SG3O2 CG321 114.3 65 N8 S11 O12 NG2D1 SG3O2 OG2P1 119.0 65 N8 S11 C13 NG2D1 SG3O2 CG321 114.0 79 Dihedral Angle s Atom Types K c n P1 N8 S11 O2 PG1 NG2D1 SG3O2 OG2P1 2.50 1 180 P1 N8 S11 O2 PG1 NG2D1 SG3O2 OG2P1 1.00 2 0 P1 N8 S11 C9 PG1 NG2D1 SG3O2 CG321 1.00 2 0 P1 N8 S11 C13 PG1 NG2D1 SG3O2 CG331 1.00 1 0 P1 N8 S11 C13 PG1 NG2D1 SG3O2 CG331 0.60 2 0 O2 P1 N8 S11 OG2P1 PG1 NG2D1 SG3O2 0.50 4 0 O3 P1 N8 S11 OG303 PG1 NG2D1 SG3O2 1.80 1 0 O3 P1 N8 S11 OG303 PG1 NG2D1 SG3O2 3.00 2 0 N8 P1 O3 C4 NG2D1 PG1 OG303 CG331 0.40 1 0 N8 P1 O3 C4 NG2D1 PG1 OG303 CG331 0.80 2 0 N8 P1 O3 C4 NG2D1 PG1 OG303 CG331 0.35 3 0 N8 S11 C9 C19 NG2D1 SG3O2 CG321 CG331 1.40 1 180 N8 S11 C9 C19 NG2D1 SG3O2 CG321 CG331 0.001 3 0 N8 S11 C9 H17 NG2D1 SG3O2 CG321 HGA2 0.16 3 0 N8 S11 C13 H14 NG2D1 SG3O2 CG331 HGA3 0.18 3 0
60 CHAPTER 3 MOLECULAR DYNA MICS SIMULATIONS OF UNLIGANDED ASPARAGINE SYNTHETASE AND ASNS / ASPAMP COMPLEX Background Proteins are not static structures; instead they are in constant motion. The energy landscape can be used to describe a prot and different protein conformations are populated according to their energies. 83 84 Molecular dynamic s (MD) simulation is a computational technique widely used to study biologica l macromolecules like proteins and to obtain valuable insights in to their dynamical behavior including both the fast motions of the amino acid side chains and the slow conformational changes of active site loops. 8 5 86 MD simulations also provide time average properties of biological systems such as density, interaction energies and entropies. The main goal of this research is the identification of potent lead inhib itors for human asparagine synthetase (h ASNS ) through virtual screening of large libraries of chemical compounds. This is very challenging because t he protein receptor conformation used in the docking study is crucial to the outcome 87 89 For e xample; the conformations of an apo enzyme may not be adequate for docking new ligands, because of wrongly positioned side chains or loops that block ligand access to the binding site. I ncorporatin g protein flexibility in the docking process through molecular dynamics (MD) simulations plays an important role in improving the docking calculations and thereby increasing the chances of predicting reliable lead inhibitors. Therefore, use of one represen tative protein structure of that best matches the most probable protein conformation observed after simulations or dock ing the ligand into different protein conformations can reduc e the risks associated with docking to an inadequate protein
61 conformation 90 91 Therefore it is important to prepare the structure of the target protein carefully before considering the docking process. Currently there exists only one crystal structur e of glutamine dependent asparagine synthetase the Escherichia coli enzyme (AS B) with a resolution of 2 (Figure 1 2 Chapter 1 ) 1 Various attempts to obtain crystal structures for hASNS or AS B complexed wit aspartyl sp AMP) have been unsuccessful 92 Therefore, the re is incomplete information on the important catalytic residues in the synthetase active site because the available crystal structure does not include the substrate or the intermediate and ha s only the byproduct adenosine monophosphate (AMP) bound i n the synthetase active site. A computational model was developed previously in the Richards group where aspartyl AMP intermediat e PPi and one Mg 2+ ion were modeled in the synthetase ac tive site using the very highly lactam synthetase complex ed with the adenylated substrate (CEA), ATP analog (AMP CPP) and one Mg 2+ ion (PDB ID 1JGT ) (Figure 3 2). 92 94 The lack of sp AMP intermediate limited efforts towards the refinement of the structure of sp AMP/PPi/Mg 2+ complex. However, after developin sp AMP, as described in detail in Chapter 2 sp AMP/PPi/Mg 2+ model w as refined using molecular dynamics simulations. The ligand free ASNS was also simulated to detect related conformational change s as a result of inte rmediate binding and to identify key active site residues. The results of simulations were also compared to already available experimental data on mutagenesis and other experiments.
62 Molecular Dynamics Simulation Molecular dynamics simulation is a widely used computational technique that explores microscopic information such as atomic positions and velocities. Statistical mechanics provide s the mathematical equations that relate the microscopic information to macroscopic properties such as pressure, energ y and heat capacit y In statistical mechanics, averages of experimental observables are expressed as ensemble averages. An ensemble is a collection of molecular systems that have similar thermodynamic states such as temperature, pressure, volume and numbe r of particles. Each system is described by the phase space of the system which includes specific positions, r, and momenta, P, for all particles 95 There exist different kinds of ensembles ; for example, microcanonical ensemble s canonical ensemble s and isobaric isothermal ensemble s In microcanonical ensemble (NVE) thermodynamic state of the system is characterized by a fixed number of atoms, N, a fixed volume, V, and a fixed energy. In Canonical ensemble s (NVT), the systems have fixed number of atoms, N, fixed volume, V, and a fixed temperature, T. I sobaric isothermal ensemble s (NPT) are composed of systems that have fixed number of atoms, N, fixed pressure, P, and fixed temperature, T. In statistical mechanics, the phase space average of a property of interest, A, for a canonical ensemble, is the integral over all phase space. 96 (3 1) P(r, p) is the probability of finding the system at every point in phase space and is given by
63 (3 2 ) w here E(r, p) is the total energy of the sys tem k B temperature and Q is the canonical ensemble partition function (3 3 ) Calculati on of ensemble averages using Equation 3 3 is extremely difficult especially taking in to account every possible state of the system. Alternatively, in molecular dynamics simulations, the average value for a property of interest is calculated as the time average attained when the motion of a single point through phase space is followed as a function of time. 66 The time average is expressed as (3 4 ) According to the ergodic hypothesis ; when the trajectory of the system of interest is allowed to evolve indefinitely, that system will cover all phase space, so both the time average and ensemble average will be equal ( Equation 3 5). 95 (3 5 ) T he refore, if the system of interest is allowed to evolve in time until sufficient representative conformations are generated, experimentall y relevant information can be calculated. Molecular dynamics simulations are based on solving or eq uation of motion
64 (3 6 ) w here F i is the force acting on particle i, m i is the mass of the particle i, a i is its acceleration given i, with respect to time, t. The force can also be obtained from the gradient ( ) of the potential energy function, U ( r ) as (3 7 ) Combining Equation 3 6 and Equation 3 7 gives the acceleration : (3 8 ) The potential energy, U(r) is a very complicated function of all atomic coordinates of the system It ca not be solved analytically and must be evaluated numerically. There are numerous algorithms used in the int egra tion of equations of motion: for example, the Verlet, leap frog and velocity Verlet algorithm s. These algorithms use Taylor expansion to approximate positions, velocities and acceleration. 66 The leap frog algor ithm was used in the present study In this algorithm, t he velocities are calculated at the midpoints of the time steps and it can be written as (3 9 ) (3 10 ) e time step. The size of the time step is an important factor in the integration quality. The smaller the e
65 of does not allow sufficient sampling of phase space in a reasonable computational time. The proper time step should be 0.05 of the time involved in the fastest motion in the system. For example, in proteins the fastest motion is the stretching vibration of the bond of hydrogen with heavy atoms (X H) This occur at around 3000cm 1 with a period of about 10 fs thus the time step size should no t exceed 0.5fs. 6 This value of time step size is too small to sample enough of the ensemble phase space in a reasonable computer time. To overcome this obstacle, the SHAKE algorithm is used to constrain X H bo nds to their equilibrium values. T hus freezing those vibrat fs) and hence offer s more integration stability. Initial coordinates are assigned from experimentally determined x ray crystallography and NMR experiments or are generated by computer modeling. The stru cture is the further submitted to minimization to relieve nonbonded interactions as well as bond lengths and angle distortions before carrying out the molecular dynamics simulations. The initial velocities are randomly assigned from a Gaussian or Maxwell Boltzmann distribution (Eq uation 3 11) 6 which gives the probability that an atom, i, with mass, m i has a velocity, v, at specific temperature, T (3 11 ) It is advisable to start the simulations with a gradual heating phase starting from an in itial low temperature and the n to heat gradually until the desired temperature is reach ed This gradual heating process avoid s accidental assignment of high velocities to a group of atoms thereby creating a 66
66 Following the heating phase, the system is equilibrated for several hundred picoseconds. The system p roperties such as total energy pressure and temperature are monitored during this phase to ensure that the system is free of unpredictable fluctuat i ons. After equilibrating the system, the system is submitted for production simulations which are allowed to run for nanoseconds, depending on the system size and the computer time available, and the generated trajectory is stored for further analysis. Com putational Details ASNS complex ed and one M g 2+ ion was computationally modeled previously 51 94 This model used the very highly similar enzyme l actam LS) complexe d with its substrate, N 2 ( carboxyethyl) L arginine (CEA) triphosphate (AMP CPP) an ATP analog, and one Mg 2+ ion in the active site (PDBID 1JGT) and the M g 2+ ion in the active site of ASNS However severa l x ray crystallographic snapshots of the LS catalytic cycle were resolved afterwards and a ll showed two Mg 2+ ions instead of only one Mg 2+ ion as in LS /CEA/AMP CPP complex 42 Beca use t he residues involved in binding of the ATP and Mg 2+ ions are conserved between AS B and LS (Figure 3 1 ) we decided to modify the original model and another Mg 2+ ion was modeled on the opposite side of inorganic pyrophosphate (PPi) by superimposing lactam synthetase ( PDBID 1MBZ) and ASNS active sites (Figure 3 2 ) Lactam synthetase converts N 2 (carboxyethyl) L arginine (CEA) to deoxyguanidinoproclavaminic acid (DGPC) in the presence of ATP and two Mg 2+ ions
67 Figure 3 1. Structure based s equence alignment of LS and ASN S. 2001, rights managed by nature publishing group 93
68 A B Figure 3 2 Stereoview of the AS LS active sites A) The AS B model active site and tw o Mg 2+ ions bound. B ) T he LS (PDBID 1MBZ) active site with CMA AMP, PPi and two Mg 2+ ions bound (PDBID 1MBZ) The intermediates and key residues are shown in stick representation. Mg 2+ ions are shown as green spheres.
69 Both AS LS facilitate inter mediate formation by adenylation, but AS B LS, lactam ring (Figure 3 3). Two systems were designed for simulati ons : the unliganded ASNS and the sp AMP/PPi/Mg 2+ complex. To sample adequate conformational space in reasonable computer time e ach system was simulated as three independent trajectories as opposed to a single long trajectory. E ach trajectory was ass igned a different initial velocity by using a different seed for the random number generator 97 The hydrogen atoms were then added to the resulting model using the HBUILD utility in CHARMM. 52 The pKa for charged amino acid residues side chains were checked using PROPKA web interface 98 100 Charged amino acids side chains were modeled at their usual protonation states at physiological p H. The topology a nd parameter files for the enzyme used the CHARMM27 biological force field. sp AMP were optimized following a standard protocol (Chapter 2). The system was min imized first in vacuum for 50 steps using steepest descent (SD) followed by 50 steps of the adopted basis Newton Raphson (ABNR) algorithm H armonic restraints were applied on the intermediates and protein backbone atoms with force constant of 50 k cal/mol/ 2 The harmonic restraints on the protein atoms were then turned off while keeping the restraints on the Mg 2+ ions and the coordinated water molecules T he system was then minimized again for 100 steps using the ABNR algorithm. The system was then solvate d in an octahedral box (92 x 92 x 92 ) of TIP3P water molecules 75 T he ions were added to give a system of zero net charge. Coulombic interactions were treated by t he particle mesh Ewald m ethod 101 with a cut off value of 12 .
70 Figure 3 3 Comparison between Chemical reactions of AS B and LS. A ) AS B B LS
71 The Lennard Jones (LJ) interactions were treated by a force switching function 103 over the range 10 14 In the presence of periodic boundary conditions, 79 each system was minimized for 500 steps using the SD algorithm and 5000 steps using the ABNR algorithm. The minimized structure was then gradually heated for 30ps. The Leapfrog algorithm was used to integrate equations of motion with a 1fs time step. The SHAKE algorithm 80 was used to constrain the length of coval ent bonds to hydrogen ato ms to their equilibrium values. E ach system was then equilibrated at 300 K using a canonical ensemble (NVT) over 40ps. The restraints on the Mg 2+ ions and coordinating water molecules were maintained during minimization in solvent, heating and the initial 40 ps of the equilibration period, after which the harm onic restraints were released Final equilibration and production simulations of the system were performed using an isothermal isobaric (NPT) ensemble. The Hoover thermostat wa s used to maintain the temperature around the average 57 81 The total simulation time for each system was 30 ns incl uding heating and equilibration time. The simul ations were performed using CHARMM version 36b1. CHARMM analysis tools were used to calculate root mean square deviation s (RMSD) and root mean square fluctuations (RMSF) and to monitor hydrogen bonds and torsion angles An equilibration time of 1ns was ex cluded from each trajectory during the analysis resulting in a total production simulation time of 27 ns The root mean square deviation (RMSD) which is a measure of the distance difference between two atoms averaged over all atoms i s a frequently used tool to assess the stability of the
72 simulations The RMSD is defined by Eq uation 3 12, where i is the distance difference between two equivalent atoms (3 12 ) Root mean square fluctuation measures the devia tion of a specific atom, x i from a reference position of the same atom (average position in our case ) averaged over time. The RMSF give s information about the struc ture mobility and is defined as following : (3 12 ) Hydrogen bond analysis was performed using default criteria with a d onor acceptor distance cutoff of 2.4 and no angle cut off value The conformations of the residues in the protein were monitored using 1 and 2 dihedral angles 1 corresponds to the dihedral angle of N C in the amino acid residue of interest 2 corresponds to C Results and Discussion Root Mean Square Deviation (RMSD) The RMSD of the back bone atoms for each independent trajectory was calculated as described above ( F igure 3 4 ). The starting geom etry of production simulations was used as the reference structure RMSD did not exceed 2. 0 thereby assessing the simulations stability. Root Mea n Square Fluctuations (RMSF) RMSF were calculated per residue backbone atoms for both the free and liganded enzyme independent trajectories ( Figures 3 5 through 3 8 ).
73 A B C Figure 3 4 Root mean square deviation (RMSD) of backbone atoms for unligan ded (green) and ligand bound (red) asparagine synthetase enzyme. A) MD1. B) MD2. C) MD3.
74 The RMSF difference between the liganded and unliganded enzyme for each independent trajectory reveal s changes in fluctuations in many backbone regions Enzyme regions with RMSF difference s greater than 0.5 are colored for clarity Each trajectory exhibit s different flexibilty regions. Regions which are far from the active site will not be considered further The loop extending in the region 442 452 exhibit s reduced f luctuations in two simulations due to presence of the intermediate in the active site This loop bears the catalytically important residue Lys449 The loop 442 452 moves towards the inside of the active site forming a more close d active site in the ASNS/ A sp /PPi/Mg 2+ complex (Figure 3 8 ) Mg 2+ Coordination The two Mg 2+ ions displayed approximately octahedral coordination in all three simulations (Figure 3 9 ). The Mg 2+ ion (Mg1) ha s direct contacts with the terminal oxygen atoms of the pyrophosph ate (PPi), one oxygen atom from the sp AMP, a one side chain carboxylate oxygen from each of the Asp 351 and Asp 238 residues, and one water molecule. The second Mg 2+ ion (Mg2) is located on the opposite side of the pyrophosphate group and has direct contacts with oxygen atoms from each of terminal, bridging terminal group s and three water molecules. In the third independent MD simulations, Mg2 is observed to coordinate with the side chain carboxylate of Asp 279. Mg2 has indirect sp AMP phosphate group through a bridging water molecule which is observed in all three simulations of the ASNS/ sp AMP/PPi/Mg 2+ complex Mg 2+ ions stabilize the negative charge on PPi especially du ring the adenylation reaction. This behavior of Mg 2+ coordination was LS/ CMA AMP / PPi / Mg 2+ complex (PDBID 1MBZ) (Figure 3 2 ) 42
75 A B Figure 3 5 RMSF of backbone atoms for first independen t trajectory simulations (MD1) (A) RMSF for unliganded (red) and ligand bound (black) enzyme. (B) RMSF difference between liganded and unliganded simulations.
76 A B Figure 3 6 RMSF of backbone atoms for the second independent trajectory simulations (MD2) A) RMSF for unliganded (red) an d ligand bound (black) enzyme. B) RMSF difference between liganded and unliganded simulations.
77 A B Figure 3 7 RMSF of backbone atoms for third independent trajectory simulations (MD3) (A) RMSF for unliganded (red ) and ligand bound (black) enzyme. B) RMSF difference between liganded and unliganded simulations.
78 A B C Figure 3 8 Enzyme regions with changes in RMSF of backbone atoms greater than 0.5 are colored. Red indicates increased fluctuations in the A SNS/BASP/PPi/Mg 2+ complex and blue indicates reduced fluctuations in the ASNS/ A sp /PPi/Mg 2+ complex. A) MD1. B) MD 2. C) MD 3.
79 Hydrogen Bonds The stability of hydrogen bonding interactions over all simulations w as also investigated (Tables 3 1 and 3 2). Table 3 1 shows the ave rage occupancy and average life time for the hydrogen bonds aspartyl AMP and active site residues in the three independent simulations for the sp AMP/PPi/Mg 2+ complex. sp AMP forms three very stable hydrogen bonds with the carboxylate side chains of Glu 348, Glu 352 and Asp 384. sp AMP is stabilized by a hydrogen bond to Lys 376. One of the phosphate oxygen atoms coordinates with one Mg 2+ ion (Mg1). The other phospha te oxygen on the opposite side makes a hydrogen bond with a bridging wate r molecule that coordinates with the other Mg 2+ ion (Mg2). This bridging water molecule is stable in the three simulations sp AMP/PPi/Mg 2+ complex In kinetic studies to i n vestigate the inhibitory effect of transition state analogs, compounds 3 a nd 4 on human ASNS showed that compound 3 has much less inhibitor activity than compound 1 whilecompound 4 did not show any inhibitory effect ( Figures 1 4 and 1 5 in C hapter 1 ) These results suggested that the negatively charged phosphate group is essent ial to inhibit hASNS and it was hypothesized that the negatively charged oxygen of the phosp hate group make s a critical electrostatic interaction with the catalytically important Lys 449 in compound s 2 and 3 producing a resonance stab ilized negative char ge on the sulfamate 51 However, according to the MD simulation results, the negative charge of the phosphate group makes critical electrostatic interactions by coordinating to the Mg 2+ ions and stabilizing the overall integrity of the active site for subsequent asparagine formation. Therefore, the negative resonance form s of the sulfamate group s in compounds 2 and 3 are able to mimic
80 Figure 3 9 Focused view of the Mg 2+ ions coordinating residues. Coordina t e bonds are indicated by purple pseudobonds. Figure 3 10. Hydrogen bonding interactions of A sp AMP with active site residues. Hydrogen bond s are indicated by blue lines.
81 sp AMP intermediate and hence exhibit their inhibitory effect s Lys 449 is observed to interact with the acyl phosphate oxygen of sp AMP through a bridging water molecule. One of the adenosine hydroxyls interact s with the backbone atoms of Gly 347 and Leu 232 and t he other hydroxyl makes bifurcated hydrogen bonds with two pyrophosphate terminal oxygen atoms The adenine nitrogen and ami no group make hydrogen bonds to the backbone of Val 272. The PPi binding pocket called the PP motif is conserved in many enzymes such as argininosuccinate synthetases, asparagin e synthetases, and A TP sulfurylases and is characterized by the sequence SGGXDS. 44 From the hydrogen bonding analysis, PPi forms stable hydrogen bonds within the PP motif with the side chain of Ser 234, backbone of Leu 237 and bot h backbone and side chain atoms of Ser 239 (Figure 3 1 1 ) The absence of ATP/PPi sp AMP intermediate formation is irreversible) and the very stable interactions of PPi deep within the PP motif all suggest that PPi is the last product to be released 104 LS, PPi bind s deep ly in the active site cleft and kinetic studies suggested that PPi is the last product released 42 105 Stability of hydrogen bonding interactions between amino acid residues within 7 sp AMP were also investigated (Tables 3 3 and 3 4). There is no significant difference in the hydrogen bonding pattern in the ASNS active site in presence and absence of intermediate binding That is due to the presence of a set of charged amino acids that make a network of electrostatic interactions and keep the active side residues positioned in an orientation r eady for binding of the substrate.
82 Table 3 1 Hydrogen bond ing interactions of aspartyl AMP (BAA) with ac t ive site residues. H Bond Average occupancy % ( Average life time, p s ) MD 1 MD 2 MD 3 BAA H47 OE1 E348 100 ( 4500 ) 100 (9000) 61 (2741) BAA G347 75 ( 12 ) 61 (7) 86 (19) BAA L232 23 ( 6.8 ) 0.2 (4) 59 (12) BAA G S346 58 ( 21 ) 88 (25) 24 (51) BAA V272 17 ( 57 ) 19 (15) 93 (111) BAA V272 80 (52) 69 (24) 5 (105) BAA V272 62 (7) 83 (15) 71 (9) BAA Y357 2 (3) 72 (150) BAA Y357 30 (9) 0.2 (3) BAA K376 94 (356) 96 (222) 7 (65) BAA H48...OE1 E352 93 (77) 100 (1515) 36 (18) BAA H48...OE2 E352 64 (16) 57 (7) 44 (16) BAA H49...OD1 D384 99 (998) 100 (1135) 100 ( 1228 ) *BAA K449 23 ( 14 ) 6 (10) 2 (8) *BAA K449 17 ( 10) 23 (11) 33 (14) *BAA Y357 7 (4) 5 (3) 29 (4) *BAA L232 34 (7) 33 (5) 6 (4) *BAA G347 18 (4) 19 (4) 1 (3) *BAA G347 60 (6) 40 (5) 23 (7) *BAA D351 10 (17) 2 (3) *BAA Y357 21 (5) 9 (3) 27 (4) *BAA Y359 12 (9) 8 (3) *BAA K449 52 (37 ) 60 (48) 22 (13) *BAA Y357 10 (4) 9 (3) 7 (3) *BAA OOC D384 48 (13) 41 (10) 100 (20) *BAA R387 46 (6) 17 (5) 25 (5) H bonding interactions through bridged water molecules are indicated by an asterisk ( )
83 Table 3 2 Hydrogen bond i ng interactions of the p yrophosphate group (PP i ) with amino acid residues in the active site. H Bond Average occupancy ,% ( Average life time p s ) MD 1 MD 2 MD 3 PP i K449 62 (60) 3 (9) 1 (8) PP i K449 36 (14) 58 (30) 89 (35) PP i K42 9 29 (15) 30 (5) 1 (5) PP i S234 99 (1496) 93 (36) 100 (9000) PP i L237 55 (7) 85 (18) 92 (32) PPi O3G...HZ3 K429 20 (415) 100 (4549) 100 (9000) PPi O2B... HN S239 99 (640) 100 (4549) 100 (2260) PPi O2B...HG1 S239 100 (4499) 100 (9000) 100 (9000) *PP i G347 23 (3) 7 (3) 8 (3) *PPi K449 17 (16) 55 (50) 20 (15) *PPi E450 19 (6) 83 (19) 68 (19) *PPi R433 10 (10) 87 (25) 77 (15) *PPi OE2 E428 100 (78) 0.3 (6) 7 (5) *PPi K429 79 (945) H bonding interactions through bridged water molecules are indicated by an asterisk ( ) Figure 3 1 1 Interactions of PPi with active site residues. Hydrogen bonding interactions are indicated by blue lines The PP motif is colored in green.
84 Table 3 3. Hydrogen bonding interactions between the residues in the active site in MD sp AMP. H Bond Average occupancy % (Average lif e time, p s) MD 1 MD 2 MD 3 E348 NH S346 28 (4) 27 (3) 30 (4) E348 OE1 G34 9 89 (25) 91 (29) 63 (12) E348 OE2...HH 21 R387 97 (126) 28 (16) 60 (3) E352 COO E352 90 (100) 99 (376) 45 (32) E352 NH3 K376 27 (240) 98 (3000) 33 (120) E352 K376 75 (759) 20 (85) E352 COO Y357 71 (25) 8 (550) 90 (50) D 384 NH OD2 D384 71 (10) 73 (16) 73 (11) D3 84 HN A388 89 (25) 36 (4) 92 (38) D384 K376 40 (10) 87 (8) D384 R387 100 (312) 70 (80) 99 (70) D384 NH3 K376 69 (142) 80 (52) K376 A379 98 (17) 87 (21) 49 (12) K376 E372 82 (17) 89 (28) 94 (44) K376 OE1 E352 99 (100 0) 99 (3751) 50 (220) K376 OOC D 384 76 (148) 90 (62) L237 I241 83 (16) 87 (21) 91 (30) K429 OOC E428 95 (64) 99 (2700) 99 (1130) K429 L432 19 (4) 36 (4) 23 (3) K429 R433 75 (13) 88 (21) 52 (6) K429 OOC D238 99 (120) 99 (3000) 60 (10) K429 OOC D351 90 (195) 99 (4550) 99 (9000) K449 OOC E428 73 (11) 90 (60) 89 (26) K449 OOC D351 92 (145) 99 (60) 60 (30) S346 E348 28 (4) 27 (3) 30 (4) S346 R400 56 (7) 60 (6) 50 (6) S234 S239 61 (7) 6 (4 ) 73 (10) S234 S234 26 (4) 2 23 (4) S234 G236 7 (5) 62 (7) 6 (5) S239 S234 61 (7) 6 (4) 73 (10) S239 G235 20 (3) 64 (7) S239 S243 92 (31) 98 (142) 98 (118) S239 S243 60 (41) 99 (13) 2 (17)
85 Table 3 4. Hydrogen bonding interactions between the residues in the active site in MD simulations of unliganded ASNS H Bond Average occupancy % (Average life time, p s ) MD 1 MD 2 MD 3 E348 S346 31 (4) 34 (4) 3 (3) E348 HN G34 9 5 (10) E348 OE2 ... NH2 R387 70 (40 ) 300 (18) 63 (45) E348 OE1...NH2 R387 50 (35) 270 (16) 70 (40) E352 HN E352 97 ( 96) 80 (19) 99 (357) E352 K376 100 (2998) 80 (1300) 99 (4500) E352 K376 75 (759) E352 COO HH Y357 99 ( 50 ) 92 (73) 34 (4) D384 D384 4 ( 3) 37 (8) 10 (3) D3 84 A388 79 (17) 20 (4) 86 (24) D384 K376 2 (3) 125 (15) 9 (10) D384 R387 100 (75) 95 (32) 99 (183) D384 K376 83 (80) 28 (40) 98 (948) K376 A379 78 (13) 84 (18) 80 (13) K376 E372 93 (39) 89 (28 ) 90 (28) K376 OOC E352 100 (2998) 70 (1232) 99 (3700) K376 OOC D 384 83 (80) 20 (30) 99 (1000) L237 I241 82 (19) 96 (79) 80 (15) K429 E428 80 (40) 99 (389) 99 (1500) K429 L432 30 (4) 21 (4) 40 (4) K429 R433 90 (28) 9 7 (89) 72 (10) K429 OOC D238 99 (132) 97 (111) 99 (9000) K429 OOC D351 99 (153) 99 (50) 99 (2000) K4 49 2 E428 51 (15 ) K44 9 OOC D351 57 ( 71 ) 99 (200) K44 9 OOC D 279 16 (100) 90 (40) S346 E348 31 (4) 33 (4) 3 (2.8) S346 R400 80 (15) 77 (13) 94 (44) S346 OOC E348 23 (40) 7 (9) S234 S239 5 (3) 4 (3) 18 (4) S234 L232 41 (11) 55 (8) S234 S234 11 (3) 9 (3) 16 (3) S234 G 23 6 18 (4) 19 (4) 18 (4) S239 S234 5 (3) 4 (3) 18 (4) S239 H G 1 OG S234 41 (14) 52 (15) 54 (8) S239 G235 38 (6) 43 (6) 10 (3) S239 G236 5 (3) 10 (3) 4 (3) S239 S243 98 (128) 98 (155) 97 (102) S239 S243 82 (27) 64 (52) 86 (65)
86 The hydrogen bonding interactions together with confor mational behavior for each residue will be discussed in detail in the following section Behavior of Important Residues in t he Active Site of Unliganded ASNS and the ASNS sp AMP/PPi/ Mg 2 + Complex Glu 348 1 and 2 torsions of Glu 348 which is conserve d among all AS NS enzymes 1 exhibit different conformational behavior in the presence and absence of the intermediate (Figures 3 1 2 and 3 1 3 ) In the unliganded ASNS, the Glu 348 side chain sample s different con formations with the carboxylate group forming hydrogen bonds with the guanidine group of Arg 387 and the backbone of Glu 348 making a hydrogen bond with the hydroxyl group of Ser 346. In the ASNS / sp AMP complex, the 1 and 2 torsions of Glu 348 change t heir conformation to orient Glu 348 into the active site with one of the carboxylate oxygen s making a hydrogen bond with the amino group of sp AMP and the backbone of Gly 349. The other carboxylate oxygen still hydrogen bond s with the guanidine group of A rg 387 and the hydroxyl group of Ser 346. O xygen 18 isotope transfer experiments have show n that E 348A and E348Q AS B mutants were not able to form the Asp AMP intermediate although their glutaminase activity is retained as the wild type enzyme 104 O n the other hand E348D AS B mutant wa s able to form a sp AMP intermediate, but the Glu:Asn ratio for the E348D increased to a value of 18:1 compared to a value of 7.2:1 in the wild type enzyme This indicates that the glutaminase and synthetase active sites became more uncoupled in the mutant enzyme. These results suggest that Glu 348 has an essential role in catalyzing the formation of the sp AMP intermediate and may be playing a role albeit weakly, on
87 A B F igure 3 1 2 Glu 348 1 torsion change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) ; and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 1 tor sion. B) Probability distribution of 1 torsion. 1 torsion for the ASNS and 3 (red, black and blue respectively) have the same dihedral angle value.
88 A B Figure 3 1 3 Glu 348 2 torsion change during MD simul ations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) ; and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 2 torsion. B) Probability distribution of 2 torsion. No 1 torsion for the ASNS and 3 (red, black and blue respectively) have the same dihedral angle value.
89 coordinating the glutaminase and synthetase active sites by acting as a general base to facilitate the attack of am Therefore, it is proposed that m utating Glu 348 to aspartate changes the length of the side chain and leads to a conformation that is not proper for ammonia translocation and proper positioning of the intermediate in the active site 104 Base d on these simulations, Glu 348 adopts a specific conformation in the ASNS/ sp AMP simulations and the side chain carboxylate form s a stable hydrogen bond with the amino group of the sp AMP intermediate This explains the role of Glu 348 in stabilizing the intermediate in the synthetase active site. The coordination of the Glu 348 to the two active site s could not be understood by these simulations F urther MD simulations are need ed with glutamate bound in the glutaminase active site in addition to sp AMP and ammonia molecule present in the synthetase active site Glu 348 in AS B corresponds to Tyr 348 in the evolutiona ll l actam s ynthetase enzy LS) (Figure 3 2 ). Both Glu 348 and Tyr 348 lie in the same positions in the active sites of AS B and LS respectively It ha s been shown that Tyr 348 together with Glu LS form s a catalytic dyad to deprotonate the secondary amine of the aden ylated intermediate and catalyze intramolecular ring closure ( Figure 3 3 ) 106 Glu 352 The Glu 352 1 torsion adopts slightly different conformation s in MD simulations sp AMP complex (Figur es 3 1 4 ) The 2 torsion of the sp AMP complex populates the same dihedral as the 2 torsion of the unliganded enzyme at 290 and also populates another dihedral angle at about 50 (Figures 3 1 5 ) The 2 sp AMP complex adopts only one conformation at about 50
90 A B Figure 3 1 4 Glu 352 1 torsion change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) ; and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan an d yellow respectively) A) Time series of 1 torsion. B) Probability distribution of 1 torsion.
91 A B B Figure 3 1 5 Glu 352 2 torsion change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 2 torsion. B) Probability distribution of 2 torsion.
92 The side chain carboxylate of Glu 352 in the unliganded ASNS forms hydrogen bonds to its backbon e atoms and to the side chains of Lys 376 and Tyr 357. The carboxylate group of Asp 352 preserves the sam e hydrogen bonding interactions but with less occupancy It is also involved in a key hydrogen bonding interaction with the amino sp AMP. Asp 384 Asp 384 1 torsion did not change its conformation in the presence of the intermediate. However, a small conformational change is observed for 2 torsion for the spAMP complex with the dihedral angle primarily populated at 70 This dihedral an gle is also sampled by the unliganded enzyme. The 2 torsion for Asp 384 of the unliganded enzyme is primarily populated at about 125 (Figures 3 1 6 and 3 1 7 ). This minor conformational change is due to presence of electrostatic interactions of both sides of the Asp 384 carboxylate with the amino group side chain of Lys376 and the guanidine group of Arg 387, thereby clamping Asp 384 in place and ready for interaction with the substrate. Lys 376 Two close conformations for the 1 torsion are sampled by both the sp AMP complex and ligand free ASNS. In the case of the 2 torsion, a dihedral angle of about 250 is primarily populated in sp AMP simulations 1 and 3. However, MD 2 of the sp AMP complex together with simulations 1 and 3 of the liga nd free ASNS populates 2 torsion are populated at about 120 (Figures 3 1 8 and 3 1 9 ) Despite the minor change in populations of 2 torsion, the hydrogen bonding interactions between Lys 376 and the carboxylate group s of both Glu 352 and Asp 384 are reta ined Together they form a network of hydrogen bonding interactions holding
93 A B Figure 3 1 6 Asp 384 1 torsion change during MD simulations of the ASNS ASPAMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 1 torsion. B) Probability distribution of 1 torsion.
94 A B Figure 3 1 7 Asp 384 2 torsion change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 an d 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 2 torsion. B) Probability distribution of 2 torsion.
95 A B Figure 3 1 8 Lys 376 1 torsion change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 1 torsion. B) Probability distribution of 1 torsio n.
96 A B Figure 3 1 9 Lys 376 2 torsion change during MD simulations in ASNS A sp AMP complex simulations ; 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 2 torsion. B) Probab ili ty distribution of 2 torsion
97 them in a position set to interact with the intermediate Lys 376 interact ing with the carboxylate group of the intermediate and the carboxylate group s of both Glu 352 and Asp 384 intera ct ing with the amino group Lys 429 The 1 torsion did not change its conformation due to presence of the intermediate. The 2 torsion of Lys sp AMP complex is populated primarily at 360 while it exhibits different conformations in the unlig anded ASNS simulations indicating that it is sampling more conformations (Figures 3 20 and 3 2 1 ). Lys 429 shows the same hydrogen bonding interactions with neighboring residues in sp AMP complex. The side chain of Lys 429 u ndergoes hydrogen bonding interactions with Asp 238, Asp 351 and inorganic pyrophosphate (PPi). Ser 346 Ser 346 side chain hydroxyl group displays a hydrogen bonding interaction with sp AMP and it also interacts by hydrogen bonding with the backbone of G lu 348 and Arg 400 in the sp AMP/PPi/Mg 2+ and ligand free ASNS simulations. N o conformational change was 1 torsion of Ser 346 (Figure 3 2 2 ). Ser 234 and Ser 239 Both Ser 234 and Ser 239 occur on the PP motif (Figur e 3 1 1 ). Both residues form hydrogen bonding interactions with pyrophosphate ( PPi ) thereby stabilizing PP i within the PP sp AMP/PPi/Mg 2+ simulations Ser 234 side chain forms also hydrogen bonding interaction with Gly 236 and Ser 234 Se r 239 side chain displays hydrogen bonding interactions with Gly 235, Gly 236 and Ser 234.
98 A B Figure 3 20 Lys 429 1 torsion change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and u nliganded ASNS ; simulations ; 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 1 torsion. B) Probability distribution of 1 torsion.
99 A B Figure 3 2 1 Lys 429 2 torsion change during MD simulations of the ASNS sp AMP complex ; si mulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively). A) Time series of 2 torsion. B) Probability distribution of 2 torsion
100 A B Figure 3 2 2 Ser346 1 torsio n change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 1 torsion. B) Probability distrib ution of 1 torsion
101 A B Figure 3 2 3 Ser234 1 torsion change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respecti vely) A) Time series of 1 torsion. B) Probability distribution of 1 torsion.
102 A B Figure 3 2 4 Ser239 1 torsion change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 1 torsion. B) Probability distribution of 1 torsion.
103 1 conformational change for Ser 234 and Ser 239 in the absence and presence of intermediate in the ASN S active site (Figures 3 2 3 and 3 2 4 ). Lys 449 Lys 449 is conserved among all asparagine syntheta se enzymes. Investig ation of the 1 2 3 4 torsions of Lys 449 (Figure s 3 2 5 3 2 8 ) shows that the 4 torsion in two independent simulations of the ASNS / sp AMP/PPi/Mg 2+ complex is primarily populated at 250 The 4 torsion of Lys 449 in the unliganded ASNS sample s different conformations indicating its ability to move freely. Lys 449 makes direct hydrogen bonding interactions with the inorganic pyroph osphate (PPi) and with phosphate oxygen of the sp AMP inte rmediate through bridging water molecules (Tables 3 1 and 3 2). The side chain amino group is also involved in hydrogen bonding interactions with the side chain carbox ylate of Asp 351. In the ligan d free ASNS, a hydrogen bonding interaction is observed between the side chain amino group of Lys 449 and the carboxylate side ch a in of Asp 279 (Table 3 4). Mutati on of Lys 449 in to arginine or alanine resulted in complete loss of synthetase activity in dicating the importance of Lys 449 for enzyme catalysis. 92 Previous studies on a set of sulfonamide inhibitors have shown that the resonance stabilized negative charge of the sulfonamide group mimics the negative charge on the intermediate and is critical for inhibitor binding ( Figures 1 4 and 1 5 in C hapter 1). It was sp AMP is involved in a critical electrostatic interaction with Lys 449. Howe ver, according to MD simulation results, Lys 449 forms stable hydrogen bonding interactions primarily with PPi and the carboxylate side chain of Asp 351 in the ASNS/ sp AMP complex.
104 A B Figure 3 2 5 Lys 449 1 torsion change during MD simulations of the ASNS A sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively) A) Time series of 1 torsion. B) Probability distribution of 1 torsion.
105 A B Figure 3 2 6 Lys 449 2 torsion change during MD simulations of the ASNS sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively). A) Time series of 2 torsion. B) Probability distribution of 2 torsion
106 A B Figure 3 2 7 Lys 449 3 torsion change during M D simulations of the ASNS sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively). A) Time series of 3 torsion. B) Probability distribution of 3 torsi on
107 A B Figure 3 2 8 Lys 449 4 torsion change during MD simulations of the ASNS sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively). A) Time series of 4 torsion. B) Probability distribution of 4 torsion
108 In addition, Lys 449 occurs on the loop 443 452 which is observed to have reduced fluctuations when the intermediate is bound in two simulations of the ASNS/ sp AMP/PPi/Mg 2+ complex and is observed to clo se upon ligand binding (Figure 3 2 9 ). Therefore, the catalytic importance of Lys 449 m ay be due to its role in retaining key hydrogen bonding interactions in the ASNS active site in addition to its mediating the loop conformational change a s discussed b e l o w Lys 449 corresponds to Lys LS and it occurs on a similar loop compris ing residues 444 LS ( Figure 3 2 ). This loop closes tightly on the LS and it ha s been shown that the conformational change of this loop is partially rate limiting 42 107 In addition, Lys 443 is observed to form hydrogen bonding interaction s with the carbonyl of the intermediate CMA and the product DGPC (Figure 3 30 ) 42 It was later shown by kinetic, mutagenesis, and density functional theory (DFT) studies that Lys lactam formation 108 However, this interaction is not observed in the sp AMP simulated model sp AMP intermediate is shorter with its carboxyl group hydrogen bond ing with Lys 376 and repelling the carbonyl from Lys 449 (Figure 3 10) Lys 376 is LS The absence of Lys LS and the additional electrostatic interaction of Lys 44 3 with the carbonyl of the CMA intermediate may be LS active site to accommodate the long er substrate (CEA) and formation of the lactam (Figure 3 2 ) Loop 443 452 Mobility Loop 443 452 is observed in two independent simulations to encompass reduced fluctuations when the intermediate is bound (Figure 3 8 ). This loop carries the catalyticall y important residue Lys 449. Movement of the loop extending from residue
109 443 to residue 452 wa s monitored by measuring the distance between the side chain nitrogen of Lys 449 and the side chain carboxylate of Asp 351 over all simulation trajectories (Figur e 3 3 1 ). The distance in two simulations of the ASNS/ sp AMP/PPi/Mg 2+ system remains at 3 where Lys 449 is clamped in position by hydrogen bonding interactions with the side chain carboxylate of Asp 351 and inorganic pyrophosphate oxygen atoms. This state represents the closed form of the loop and prevents the hydrolysis of the intermediates. The distance becomes slightly longer, approximately 4.5 in the third MD simulation of ASNS/ sp AMP/PPi/Mg 2+ In the ligand free ASNS simulations, the loop is observed to open widely to about 10 and Lys 449 flips down to make a hydrogen bond with the side chain carboxylate of Asp 279 thereby exposing the active site to more solvent. The loop is observed in the third independent MD simulation of th e ligand free ASNS to be semi closed around 4. 0 because the hydr ogen bonding between Lys 449 and Asp 351 is maintained T his explains the similar RMSF of this loop in spAMP and ligand free ASNS. Most probably if this trajectory is allowed to run for a longer time this loop will experience open states as in the other two independent unliganded ASNS s imulations. The behavior of loop 443 452 explains previous muta genesis studies where mutating Lys 449 to arginine retained the glutaminase activity of the enzyme while the synthetase activity was lost. 92 Mutating Lys 449 in to alanine make s the enzyme lose both its glutami nase and synthetase activity. According to MD simulation results, removal of Lys 449 by mutation may lead to los s of important hydrogen bonding interactions with the side chain carboxylate of Asp 351 and PPi This would lead to a widely open
110 Figure 3 2 9 The c losed ( p urple) and open ( y ellow) forms of loop 443 452 in ASNS Figure 3 30 Structural overlay of the conserved lysine residue (K449 in AS B, K443 in sp AMP intermediate in AS B (blue), CMA AMP intermediate LS (pink). Hydrogen bonds between K443 and the carbonyl and phosp hate oxygen s of the LS are shown as blue lines. In contrast, K449 in ASNS does not hydrogen bond with the carbonyl and phosphate oxygen of the sp AMP intermediate.
111 A B Figure 3 3 1 D istance between K449 N and D351 C dur ing MD simulations of the ASNS sp AMP complex ; simulations 1, 2 and 3 ( r ed, black and blue respectively) and unliganded ASNS ; simulations 1, 2 and 3 ( p urple, cyan and yellow respectively). A) Distance time series. B) Probability distribution
112 state of loop 443 452 with the active site becoming more exposed to solvent possibly lead ing to hydrolysis of the intermediates and hence loss of activity. A s mentioned earlier, a similar loop compris ing residues 444 453 closes tightly on the adenylated intermedi LS and it has been shown that the conformational change of this loop is partially rate limiting 42 107 Summary Molecular dynamics simulations of the ASNS sp AMP/PPi/Mg 2+ complex and the unliganded ASNS have been performed. The MD simulation results revealed important interactions in the active site that mediate intermediate binding. The adenosine group binds in a well defined ATP binding pocket, where the adenine undergoes hydrogen bonding interactions with the Val 272 backbone atoms. The adenosine hydroxyls forms hydrogen bonds with the side chain of Ser 346, the backbone of Gly 347 and inorganic pyrophosphate (PPi). The phosphate oxygen atoms experience critical electrostatic interactions by coordinating to the Mg 2+ ions. The amino sp AMP intermediate bind with the substrate (aspartate) binding residues in which the amino group is stabilized by electrostatic interactions with the carboxylate side chains of Glu 348, Asp 384 and Glu 352. The sp AMP is stabilized by hydrogen bonding with Lys 376. According to kinetic results the absence of ATP/PPi exchange and the very low ), PPi i s proposed to be the last product released from the active site. Inorganic pyrophosphate lies in a deep binding pocket called the PP motif and is stab ilized by hydrogen bonds with Ser 234, Ser 239 and Leu 237 within the PP motif a s well as with Lys 429 and L ys 449. The Mg 2+ ions exhibit approximately octahedral coordination in all
113 o f the simulations and ar e very important in balanc ing the negative charge of the PPi and facilitat ing asparagine formation and prevent ing sp AMP formation. Based on structural similarity and sequence alignment, the sp AMP intermediate, PPi and Mg 2+ ions were modeled in the ASNS active site based on the LS/CMA AMP/PPi/Mg 2+ crystal structure (1MBZ). Both enzymes catalyze int ermediate formation by an adenylation reaction. AS B catalyzes intermolecular amide bond LS which catalyzes intramolecular amide bond fo rmation to form a lactam ring ( Figure 3 3 ). LS active si te is elongated to accommodate the longer intermediate (Figure 3 2 ), the behavior of the sp AMP /PPi/Mg 2+ model resembles that of LS/CMA AMP/PPi/Mg 2+ complex. The ATP binding residues, PP motif, and Mg 2+ coordination residues, in addi tion to the presence of a catalytic loop that carries a conserved lysine residue are similar in the active s ite s of both enzymes. However, MD simulations reveal changes in active site residues that bind to the substrate. For example, Asp 382 and Glu 348 in AS B are replaced by Glu 382 and Tyr LS. Asp 382 and Glu 348 in AS sp AMP intermediate. Also Glu 348 plays a role in coordinating the glutaminase and synthetase active sites, albeit weakly, by acting as a gen eral base to facilitate the attack of ammonia on the sp AMP intermediate. On the other hand Glu 382 and Tyr LS a re reported to act through acid/ base catalysis to deprotonate the secondary amine of the intermediate and facilitate ring closure. An other difference revealed by the MD simulations is the behavior of a conserved lysine residue (Lys 449 in AS B and Lys LS ) This residue in both enzymes occurs on a cat alytic loop that closes upon intermediate binding and
114 prevents LS, a stable hydrogen bond between the lysine side chain and the intermediate carbonyl wa s observed in the X ray crystal structure s snapshots and it was shown to stab i l ize lactam formation. However, based on the MD simulatio n results, electrostatic interaction between Lys 449 and the carbonyl sp AMP intermediate is not observed and its catalytic importance is primarily to mediate loop closure by electrostatic interactions with Asp 251 and PPi when the intermediate is bound in the active site. That is due to the negative carboxylate of sp AMP which repels the carbonyl oxygen and binds with the side chain of Lys 376, LS. In summary, the MD simulations results reveal important information abou t active sp AMP active site was compared to known data about the LS active site. Finally, through molecular modeling and MD simulations, we were able to develop sp AMP intermediate bound in the active site after various unsuccessful efforts to obtain a crystal structure of this complex. A model representative of the average st ructure will be used further in virtual screening efforts to discover potent ASNS inhibitors.
115 CHAPTER 4 DISCOVERY OF A NEW INHIBITOR OF HUMAN ASPARAGINE SYN THETASE IDENTIFIED BY STRUCTURE BASED VIRTUAL SCREENING Background The core of the drug discovery process is the identif i cation of new inhibitors (leads) which show weak affinities against a particular biological t arget and the op timization of the leads in to m ore potent and effective drugs 109 High throughput screening (HTS) is the experimental screening of a large number of drug like compounds for their biochemical acti vity against the biological target. In recent years, automated systems have emerged which can screen tens of thousands of compounds per day. However, despite the recent advances in HTS, certain problems are associated with its use in random screening The high cost associated with assays and synthesis of lots of compounds, in addition to the low chance of identifying new hits contribute to these problems. As an a lternative, virtual screening (VS) is the computational screening of very large chemical libra ries through molecular docking. The screened compounds are ranked according to different criteria such as their electrostatic and v an der Waals complementarity with the target receptor to identify new drug candidates (hits). The interplay between HTS and VS is very important because VS generate s a more focused library for subsequent experimental testing and biochemical assays. The advances in high performance computing centers have made the virtual screening more affordable and time efficient 109 Identifying true hits and rejecting false ones through virtual screening of a large chemica l library is very challenging. Towards achieving successful virtual screening, the
116 binding site flexibility of the target rec eptor together with exploration of the ligand conformation should be considered 110 Molecular dynamics (MD) simulations are widely used to study the dynamic behavior of biological macromolecule s MD simulati ons can provide valuable information on the flexibility of the targe t protein upon ligand binding. The three dimensional structure of the target protein bound with the ligand is very important in docking studies. The ligand free receptor may not be adequa te in virtual screening because of the incorrectly positioned amino acid side chains or inadequate loop conformation that can hinder ligand access and positioning in the active site. 110 In a study where several ligands were docked into a series of x ray crystal structure complexes for a particular target, the best results were obtained when a ligand was docked in the native crystal structure conformation while poor results occur when the ligand was docked into a crystal structure of a different complex. 111 Therefor e it is essential to carefully prepare the target receptor before performing the docking calculations. Inhibitors of hASNS have been shown to suppress th e proliferation of resistant MOLT 4 leukemia cells. 51 22 59 However because the report ed hASNS inhibitors are needed in high concentration their cli nical utility is limited The virtual screening in the following study attempts to identify new leads with more potent and effective physiochemical properties that can inhibit human asparagine synthetase (hASNS). The target receptor used in this study is a representative of the average structure from MD simulations explained in C hapter 3. The NCI database was virtually screened using GLIDE (Grid based Ligand Docking with Energetics) software. 112 Ten hits were
117 id entified a nd subjected to kinetic assays to determine their activity against hASNS. Compound NSC605322 inhibited hASNS and was confirmed as a new lead inhibitor of hASNS. Structure Based Virtual Screening Virtual screening is the computational screening of large chemical libraries to identify new drug leads that can exhibit activity against a particular target. There are two main appro aches used in virtual screening: structure based virtual screening (SBVS) and ligand based virtual screening (LBVS). LBVS is designed to find compounds simil ar to known active compounds, while SBVS involves rapid docking of many chemical compounds into the active site of a biological target. Docking is the process of bringing two molecules together to a form a stable complex ; f or example, docking a small molec ule in the active site of an enzyme. The molecules are further scored and ranked according to scoring functions. Thus the quality of the scoring function, in terms of accuracy and efficacy to predict the binding affinity of each docked structure is very important 113 Three different types of conformational search es are employed by docking software 114 Programs such as GLIDE, 112 AFFINITY and ICM perform conformational search on each docked molecule in the binding pocket using stochastic search methods such as Monte Carlo (MC) and simulated annealing methods. Other software s for instance FRED and SLIDE perform conformational search on each molecule before docking it in the active site. The different generated conformations for each molecule are then docked rigidly in the active site and the rotation and translation of the rigid conformers are considered. The third category of conformational search makes use of an incremental construction algorithm. In incremental construction, the
118 rotating bonds of the ligand are broken to produce fragments. The fragments are then docked rigidly at differen t favorable positions in the binding site with the larger fragment as the starting base fragment and the entire ligand is reassembled. Software s that perform incremental construction include DOCK and HOOK. 114 Scoring functions are approximate methods to estimate the binding affinity between the docked ligand and the receptor based on ligand complementarity such as electrostatic and v an der Waals interactions with the active site. Because s coring functions ten d to be sufficiently simple to screen large chemical library in reasonable computer time they often suffer from poor accuracy. However, the scoring functions are in continuous development of to yield accurate binding without sacrificing accuracy There ar e different types of scoring functions i mplemented in docking programs, such as force field based, empirical based, and knowledge based scoring functions, as well as consensus scoring. 114 116 Force field based scoring functions sum up the electrostatic and vd W contributions between the ligand and receptor active site. 117 The parameters for force field based scoring functi ons are obtained from both quantum mechanical calculations and experimental data. For example, the energy parameters for DOCK are obtained using the AMBER force field. 118 120 Solva tion effects are usually represented using implicit solvent models such as the Poisson Boltzmann surface area (PB/SA) 121 123 model and the generalized Born surface area (GB/SA) model s 124 126 Empirical scoring functions estimate the binding free energy between receptor and ligand as a set of weighted energy components 127 (4 1 )
119 w here i represents different energy components such as electrostatic, hydrogen bond, desolvation, entropy and vd W. The coefficients are determined by calibra tion against a set of experimentally known receptor ligand strictures and their binding affinities using regression analysis. 128 134 The empirical energy function d epends on the training set used in deriving the weighting coefficients. Well known docking programs use empirical energy functions such as FlexX and GLIDE. The Glide Score is a modified version of the ChemScore function as follows: 135 (4 2 ) The G coefficients are obtained by multiple linear regressions 135 Distances or angles that lie within specific limits are given a full score by the functions f g and h and a partial score is given for distances and angles that lie outside those limits. 135 and are the deviations of the hydrogen bonds distances and angles from th eir ideal values respectively. Knowledge based scoring function is another kind of scoring function and is also referred to as statistical potential based scoring functions. A statistical potential is derived based on intermolecular distances observed in experimentally known
120 structures. Frequently observed nearby intermolecular interactions between different atom types are proposed to be energetically advantageous and contribute favorably to the binding affinity. 136 138 Because each scoring function has its own imperfections and advantages, a recent trend is of using consensus scoring has been introduced. 139 Consensus sc oring combines scores from different scoring functions so that increasing the probability of finding true hits. An example of consensus scoring is X Cscore. 139 140 Computational Details The first step in structure based virtual screening is the preparation of the protein receptor structure. A representative of the average structure from the molecular dynamics simulations described in C hapter 3 was used as the target for this virtual screening study. Solvent was removed except for the water molecules coordinated with the Mg 2+ ions As was mentioned in C hapter 3, i norganic pyrophosphate (PPi) is stabilized by many hydrogen bonding interactions in a deep binding pocket c alled the PP motif and covered with a catalytic loop Ac cording to kinetic data and the absence of ATP/PPi exchange, it is believed that PPi is the last product released from the active site. In addition, absence of the PPi will lead to loss of hydrogen bo nding interactions between the key residue Lys 449 and PPi which will result in a wide open state of the catalytic loop (residues 443 452) with the active site becoming more exposed to solvent, possibly leading to hydrolysis of the intermediates and hence loss of activity (Figure 3 29). Therefore, PPi and Mg 2+ ions were kept in the active site during docking calculations. Virtual Screening Using GLIDE editing capabilities were used to examine and adjust the formal charge s and the atom types for the metal
121 ions and the cofactors. Following protein preparation, the receptor grid which represents the position and size of the active site wa s generated. GLIDE uses two boxes for receptor grid generation (both having the same center): t he inner box (ligand diameter midpoint box) and the enclosing box (outer box). The l igand center which is the midpoint between the most widely separated atoms lies within the inner box during site point search The ligand can move outside of the inner box but must be enclosed within t he enclosing box The enclosing box should be large enough to contain the ligand and inner box lengths. The relationship between the lengths of the outer enclosing box, E, inner box, B and li gand size, L is: Following receptor grid AspAMP was extracted from the ASNS A spAMP model and re docked in the active site. The bond orders and sp AMP ligand were adjusted us ing LigPrep modul e in Maestro. GLIDE produced a docking binding pose that closely resembled the binding pose of the MD simulated ASNS sp AMP complex and there by verifying the docking approach. The docking score of the native ligand wa s 4.9Kcal/mol. The docking score of t he previously identified inhibitor (compound 1 Figure 1 4) was also 4.9Kcal/mol. Three downloaded from the website (http://cactus.nci.nih.gov/download/nci/) and processed with LigPr ep software (available from Schr dinger). 135 Flexible docking in in which the docked ligands undergo conformational search generating a set of conformations for each ligand was then performed A site point se arch was executed for each conformation to find best orientation and position of the ligand within the active site. Orientations that display ed steric clashes with active site atoms we re excluded. The n
122 rotation of each orientation wa s followed and hydroge n bonding interactions of the ligand conformation with the active site residues we re scored with modified version of the ChemScore empirical scoring function 141 The best selected poses (200 400) wer e subjected to energy minimization using the energy function for the OPLS AA force field followed by a Monte Carlo sampling procedure in which torsion angles of rotamer groups were examined to improve sampling and scoring 141 Finally the minimized poses we re ranked using Glide Score Standard Precision scoring function. A total of 500,000 compounds from scores. However, the be st docked pose for each ligand wa s selected by using a model energy score called Emodel, which is a composite of GlideScore, the molecular mechanics interaction energy between the ligand and receptor and the strain energy of the ligand. Glide Score is a modified version of the ChemSco re function (Equation 4 2). 135 Virtual Screening Using DOCK Virtual screening of two million drug like compounds from the ZINC database 33 was performed using DOCK6 143 The ligand wa s oriented within the active site using a negative surface image of the receptor binding site prepared using the DMS tool in UCSF Chimera 28 The SPHGEN tool within the DOCK pac kage wa s used to create overlapping spheres to represent the binding site and describe the shape of the molecular surface. 144 The size and po sition of the active site grid wa s def ined by the program SHOWBOX and the grid wa s generated by the program Grid tool in DOCK. Flexible docking wa s used, in which the conformational space for the docked molecules wa s sampled using the anchor and grow incremental algorithms. 145 The anchor which undergoes rigid orientation within the active site was first identified as the largest rigid
123 substructure of the ligand followed by identifica tion of the flexible parts. An chor orientations we re optimized u sing the grid scoring function and simplex minimizer. The AspAMP was extracted from the ASNS A spAMP model and re docked in the active site. DOCK generated a docking pose that closely resembled the native sp A MP in the activ e site, there by verifying the docking approaches (Figure 4 1). The dock ing score of the native ligand wa s 98Kcal/mol. Drug like compounds from the ZINC database were then docked in the ASNS active site and scored using the grid based scoring function, whi ch is based on the non bonded terms of the molecular mechanics force field to approximate the interaction energies. Results and Discussion Three dimensional structures from the NCI database (500, 000 compounds) were docked in the active site of asparagine s ynthetase (ASNS) using GLIDE. In addition two million drug like compounds from the ZINC database were virtually screened using DOCK. Molecules with docking score s higher than the score of the native ligand were inspected for their drug likeness In 1997 Christopher A. Lipinski formulated a rule to evaluate drug likeness called rule of five (RO5) with respect to physicochemical factors linked with aqueous solubility and intestinal permeability. 146 The rule of five (RO5) defines four physicochemical characteristics of 90% of orally active drugs:(1) m olecular weight less than 500; (2) H bond donors fewer than 5; (3) H bond acceptors fewer than 10; (4) Partition function (log P) less than five Compounds that do not satisfy at least three of those four parameters are less likely to be oral active drugs and ca nnot pass phase II of clinical trials. Therefore it is important to keep this rule in mind during drug discovery process. Compounds that are similar to other enzyme substrates or peptides
124 Figure 4 1. Overlay of the re sp AMP intermediate using DOCK (green) and the native sp AMP (pink) Table 4 1. Docking score of the ten selected hits from GLIDE virtual screening, their Lipinski pro perties and their residual activity values. Compound Docking score ( k cal/mol) H bond acceptors H bond donors Log P Molecular weight (g/mol) Residual activity (%) NSC605322 9.7 6 2 4.34 474.96 65 NSC333462 9.6 12 5 5.7 650.72 91 NSC8805 9.5 7 4 1.74 3 86.38 71 NSC9219 9.4 7 5 1.99 302.24 25 NSC18719 9.2 7 6 0.19 302.71 89 NSC43663 8.9 7 6 0.5 268.27 93 NSC352274 8.8 8 5 0.27 386.4 91 NSC36398 8.2 7 5 0.18 304.3 39 NSC1678 8.3 8 6 4.05 238.19 94 NSC369 6.9 6 0 1.58 408.49 94 The residua l activity test was performed by lab member Yong mo Ahn. Reaction mixtures for residual activity consisted of human ASNS (4 g), 0.5 mM ATP, 10 mM L aspartate, either 25 mM L glutamine or 100 mM NH 4 Cl as nitrogen source, and the pyrophosphate reagent (350 L) dissolved in 100 mM EPPS buffer, pH 8.0, containing 100 M of each compound, 10 mM MgCl 2 and 10% DMSO (1 mL total volume).
125 NSC605322 NSC333462 NSC8805 NSC9219 NSC18719 NSC43663 Figure 4 2. 2D predicted binding poses of the ten select ed hits from GLIDE virtual screening
126 NSC352274 NSC369 8 NSC1678 NSC36 9 Figure 4 2. Continued.
127 or glycosides were excl uded because of specificity concerns. It has been also observed that compounds with more th an 10 rotatable bonds correlate wit h decreased rat oral bioavailability. 147 Therefore compounds wit h more than 10 rotatable bonds we re excluded. A f inal set of ten hits form GLIDE virtual screening w as selected for experimental testing against hASNS inhibition (Figure 4 2, T able 4 1). All in vitro experiments were performed by lab member Yong mo Ahn. The ten selected hits were subjected to a residual activity test to approximate the effect of inhibitors on PPi production under steady state condit ions in which PPi formation is coupled to NADH consumption (340 nm). The residual activity of hASNS(%) is defined as the quotient of the PPi production rate of the assay enzyme in the presence of 100 M inhibitor divided by the pyrophosphate production rat e by the same enzyme in the absence of inhibitor. The l ower the residual activity value, the more p otent is the inhibitor For example; the previous identified nan o molar inhibitor, compound 1 ( Figure 1 4 C hapter 1) ha s a residual activity of 20 %, however compound 3 with micro molar inhibition potency has a residual activity of 60% A standard residual activity below 75% was set for further kinetic analysis experiments. Compounds NSC8805, NSC921 9 NSC605322 and NSC363 9 8 had residual activities less than 75 % (Table 4 1), but only compound NSC605322 was subjected to further analysis because compounds NSC8805, NSC921 9 and NSC363 9 8 were shown to inhibit the coupling enzymes by control experiments in the pyrophosphate reagent used for the continuous assay The inhibitory effect of compound NSC605322 on glutamine dependent hASNS was assayed by measuring the rate of pyrophosphate (PPi) production 148
128 Figure 4 3 Glutamine dependent production of PPi i n the presenc e of the inhibitor NSC 605322, (0 100 M). open squares, 0 M; open circles, 10 M; open diamonds, 25 M; open triangles, 40 M; filled squares, 60 M; filled circles, 80 M; filled diamonds 100 M. Experiment performed by Yongmo Ahn
129 Figure 4 4 Ammoni a dependent production of PPi in the presence of the inhi bitor NSC 605322, (0 80 M). open squares, 0 M; open circles, 10 M; open diamonds, 25 M; open triangles, 40 M; filled squares, 60 M; filled circles, 80 M. Experiment performed by Youngmo Ahn
130 Compound NS C605322 showed slow tight binding inhibition 149 (Figures 4 3 and 4 4) with an overall inhibition constant of 1.6 M when ammonia was the nitrogen source and 2.2 M when L Gln was employed as the ni trogen source. Although compound NSC605322 is much weaker inhibitor than compound 1 NSC605322 provides key structural features and can be optimized further to a more potent lead inhibitor with improved cell permeability The docking pose of compound NSC6 05322 in the active site of ASNS was investigated (Figure 4 5 ). Overlay between A s p AMP complex is shown (F igure 4 6 ) The predicted binding pose of compound NSC605322 employs a docking pose similar to that of the native intermediate sp AMP. The a cridine moiety is settled in the adenine binding pocket and makes hydrophobic contacts with hydro phobic amino acid side chains. Acridine based pharmacophores were reported to show biological activ ities such as intercalati on int o DNA, as well as anticancer and antimicrobial activities 150 151 Therefore, new drugs based on acridine derivatives are regularly designed and synthesized as antit umor and antibacterial therapeutic agents. 152 153 The sulfonamide group occupies the same position as the phosphate group. The sulfonamide group of the inhibitor wa s predicted by Epik 154 (A program within the GLIDE package that predict s the pKa of the ionizable groups in ligands ) during docking calculations to be deprotonated. The p K a values for the amide nitrogen of some sulfonamide derivatives are reported to be around 5.5 7 .0 (Table 4 2 ). Therefor ; the pKa of compound NSC605322 is likely in the range 6 .0 7 .0 T hus under the conditions used in the assay, pH 8.0, the amide nitrogen will be mainly deprotonated. The sulfona mide deprotonated form is also stabilized by resonance with the pyridine aromatic ring.
131 Figure 4 5 Putative interactions of compound NSC605322 with ASNS active site residues. Figure 4 6 Overlay of the docking pose of ASNS inhibitor complex (blue backbone atoms) and the ASNS sp AMP complex ( b rown backbone atoms).
132 Table 4 2 pKa s of amide hydrogen s of some sulfonamide derivatives in water. 155 R pKa 6.33 6.71 5.77 5.19 5.41 6.22 Figure 4 7 Conjugate base of ionized sulfonamide functional group of compound NSC605322
133 Therefore the sulfonamide group can bind as its conjugate base and coordinates with Mg 2+ ions, the (Figure 4 7 ). This result is consistent with a previous study on hASNS inhibitors ( Figures 1 4 and 1 5 Chapter 1) in which only compound s 2 and 3 were able to inhibit hASNS because they contain an ionized functional group that can mimic the charged phosphate group while co mpound 4 did not exert any inhibitory effect because it does no t have an ionized functional group. In addition, the m olecular dynamics simulations study performed on the sp AMP/PPi/Mg 2+ complex have shown that Lys 449 occurs on a catalytic loop that closes tightly in the presence of the intermediate (Figure 3 2 9 Chapter 3 ) Loop closure is mediated mainly by hydrogen bonding interaction of Lys 449 side chain with PPi and the side chain carboxylate of Asp 351. Based on the docking results, the negativ ely charged sulfonamide group of c ompound NSC605322 in its ionized form makes electrostatic interactions with the side chain of Lys 449 (Figure 4 5 ). This electrostatic binding may increase the probability of loop closure on the active site and result in t ighter binding of the inhibitor Compound NSC605322 is the only hit from the experimentally tested set that has an ionized functional group that can mimic the charged phosphate group. Although no experiments have been performed to test the inhibition activ ity of selected hit compounds from DOCK, many of the identified hits contain a sulfo namide moiety thereby emphasizing the importance of the presence of a functional group that can mimic the native intermediate phosphate group (Figure 4 8 ). Therefore, this functional group should be considered in further design and optimization of lead inhibitors for hASNS.
134 ZINC12433053 ZINC49413960 ZINC65085571 Figure 4 8 2D predicted binding poses of some hits identified by virtual screening using DOCK.
135 ZI NC56918962 ZINC22914737 Figure 4 8 .Continued. Table 4 3 Docking score of some selected hits from DOCK virtual screening Compound Docking score ( k cal/mol) ZINC12433053 85 ZINC49413960 83 ZINC65085571 82 ZINC56918962 84 ZINC22914737 84
136 In addition, t he protonated nitrogen of the pyridine ring forms a hydrogen bond with the carboxylate group of Glu 348 which is a key residue in the active site and undergoes a sp AMP intermediate. Glu 348 was reported to be important for intermediate binding and for mediating coordination of the glutaminase and synthetase active sites. 104 Summary Virtual screening to identify new inhibitors of human asp aragine synthetase was performed. Two million drug like molecules from the ZINC database were docked in the active site of hASNS using DOCK. In addition 500,000 compounds from the NCI chemical database were docked using GLIDE. Ten hits were selected form t he NCI database residues. A new lead inhibitor of hASNS (NSC605322) was discovered. Kinetic experiments showed that compound NSC605322 inhibits hASNS with an overall inhibit ion constant of 1.6 M w ith ammonia as the nitrogen source and 2.2 M with L Gln as the nitrogen source. Further experiments such as cell based assays are required to test the effect of compound NSC 605322 on the growth of cancer cell lines for example, M OLt 4 leukemia cell lines The new inhibitor has an ionized sulfonamide group that mimics the charged phosphate group in coordinating with Mg 2+ ions. In addition the newly identified inhibitor makes electrostatic interactions with the key residues Lys 449 and Glu 348. The discovery of a lead inhibitor is the first step in the drug discovery process. Focused libraries of no vel compounds based on compound NSC605322 with acridine and sulfonamide moieties should be synthesized to gain better understanding of th e structure activity relationships (SAR) responsible for the inhibitory effect against
137 hASNS. For example, the pyridyl group may be replaced with other electron withdrawing hetero cyclic groups such as thiazole, oxazole, thiadiazole and pyrazole. In additi on the acridine ring can be modified to contain H bond donor and acceptor groups th at can undergo electrostatic interactions sp AMP adenine with the backbone atoms of Val 272. For further discovery of new hASNS inhibitors, kin etic s experiments to test the inhibitory effect of selected hits from the DOCK virtual screening calculations can be performed Also, structure based virtual screening of new focused libraries that contain ioni zable sulfonamide moieties and maintain electr ostatic interactions with key residues should be considered Another approach is to perform ligand based virtual screening based on the identified hit to provide a library of structurally similar compounds that can be tested experimentally. In summary, a new micromolar inhibitor of hASNS was discovered using structure based virtual screening. This lead inhibitor has a new structural motif that should pe rmit the development of compounds with improved cell permeability than those previously reported.
138 CHAP TER 5 CONCLUSIONS AND FUTURE WORK Conclu ding Remarks The ability of the identified hASNS N phosphonosulfonimidoyl based inhibitors ( Figure 5 1) to suppress the proliferation of MOLT 4 leukemia cells with nanomolar potency proved that the hASNS inhibitors c an be used in the treatment of acute lymphoblastic leukemia. However, th e se inhibitors are difficult to synthesize and high concentration s a re needed for cell internalization Thus new hASNS inhibitors with more drug like properties are needed Structure based virtual screening of large chemical libraries is a widely used technique to identify new lead inhibitors by docking many drug like compounds to the biological target of interest. Incorporating protein flexibility in the docking process plays an impor tant role in improving the docking calculations and thereby increasing the chances of predicting reliable lead inhibitors. To reduce the risks associated with docking to an inadequate protein conformation one rep resentative protein structure having the mo s t probable protein conformation is used, or the ligands are docked to different protein conformations The absence of good parameters that describe the acylphosphate and N phosphonosulfonimidoyl functional groups precluded previous computational efforts t o study ASNS bound with the intermediate or the inhibitor in the active site. New parameters that describe the acylphosphate and N phosphonosulfonimidoyl functional groups were optimized following the CGenFF protocol. The parameters are compatible with th e CHARMM additive all atom force field for biological molecules. The newly optimized charges were able to reproduce the interaction energies and distanc es for the model water complexes calculated by QM calculations In addition, QM
139 geometries, vibrational spectra and dihedral potential energy scans were reproduced well by the newly developed parameters. The newly developed parameters will enable computational studies using molecular mechanics approaches, such as molecular dynamics simulations and free energ y calculations, on enzymes that catalyze acylphosphate formation and inhibitors with N phosphonosulfonimidoyl functional groups. After developing new CHARMM parameters for the acylphosphate functional group in the intermediate sp sp AMP/PPi /Mg 2+ complex w as modeled and refined using molecular dynamics simulations. The ligand free ASNS was also simulated to detect related conformational change s as a result of intermediate binding and to identify key active site residues. The MD simulation res ults revealed important interactions in the active site that mediate intermediate binding. The adenosine group binds in a well defined ATP binding pocket. The phosphate oxygen atoms undergo critical electrostatic inte ractions by coordinating with the Mg 2+ ions. The amino group is stabilized with electrostatic interactions with the carboxylate side chains of Glu 348, Asp 384 and Glu spAMP is stabilized by hydrogen bonding with the side chain of Lys 376. Inorganic pyrophosphate lies in a deep binding pocket called the PP motif where it is stabilized by various hydrogen bond ing interactions According to kinetic results the absence of ATP/PPi exchange and the very low PIX of i norganic pyrophosphate (PPi), PPi was proposed to b e the last product released from the active site. A catalytic loop that carries a conserved lysine residue (Lys 449) is observed to close up on the intermediate and prevent its hydrolysis. The behavior of the sp AMP active site was compared to known d ata for the
140 LS active site. Finally, through molecular modeling and MD simulations, we were able to provide a reliable model of ASNS with the sp AMP intermediate bound in the active site after various unsuccessful efforts to obtain a crystal structure o f this complex. A model of the sp AMP /PPi/Mg 2+ complex representative of the average structure was further used in virtual screening calculations to identify potent ASNS inhibitors. A total of 500,000 compounds from the NCI chemical database were do cked to the ASNS active site using GLIDE. Ten hits were selected based on agreement with the presence of electrostatic interaction s with key active site residues. From this group, a new lead inhibitor of hASNS (NSC605322) was ident ified Compound NSC605322 inhibited hASNS with an over all inhibition constant of 1.6 M w ith amm onia as the nitrogen source and 2.2 M w ith L Gln as the nitrogen source. The new inhibitor has an ionized sulfonamide group that mimics the charged phosphate g roup in coordinating with Mg 2+ ions. In addition the newly identified inhibitor interacts with the key residues Lys 449 and Glue 348. The discovery of a lead inhibitor is the first step in the drug discovery process. Focused libraries of novel compounds based on compound NSC605322 acridine and sulfonamide moieties should be synthesized to gain better understanding of the structure activity relationships responsible for the inhibitory effect against hASNS
141 Figure 5 1. C hemical structures of hASNS N phosphonosulfonimidoyl based inhibitors. Figure 5 2 Overview of the reaction catalyzed by glutamine dependent asparagine synthetase.
142 Future Work: Free Energy Calculations to Estimate the Difference in Binding Free Energy between (R) and ( S ) Isomers of a n N phosphonosulfonimidoyl B ased I nhibitor A sulfoximine based inhibitor 1 of human asp aragine synthetase (hASNS) is analogue of ASNS transition state and was able to inhibit hASNS wi th nanomolar potency ( Figure 5 1 ). In addition, this inhibitor slowed the growth of ASNase resistant leukemia cells when used in high concentrations and when L asparagine wa s reduced in the external medium by ASNase 49 It was hypothesized that compound 1 i s needed in high concentrations to suppress the growth of AS Nase resistant leukemia cells, because t he charged groups prevent its entry into the cell 22 In order to further investigate the relative contribution of the other charged groups to the inhibitory effect of compound 1 the carboxyl and amino groups of compound 1 we re removed and c ompounds 4 and 5 were synthesized and tested experimentally for the inhibition of hASNS. 59 Compound 4 which has no net charge at cellular pH, inhibited hASNS in higher potency than compound 1 In addition, Compound 4 slowed the growth of leukemia ASNase resistant cell s with IC 50 = 0.1 mM and in the absence of ASNase in the external medium This concentration is a factor of 10 lower than the IC 50 of compound 1 59 Compound 4 has an overall inhibition constant (K I ) of 8 nM when L glutamine was u sed as the nitrogen source. The K I value for N phosphonosulfonimidoyl 4 increased four fold under the same conditions when ammonia was used instead of L glutamine. Contrary to t he higher inhibitory effect of amino N phosphonosulfonimidoyl 4 the carboxy N phosphonosulfonimidoyl 5 was a relati vely poor hASNS inhibitor with K I values of 800 and 300 nM when the nitrogen source was glutamine or ammonia, respectively.
143 Figure 5 3 Putative interactions of the four diastereoisomers with key residues in the ASNS active site. (A and B) are the ( R ) and ( S ) dia stereoisomers of compound 4 respectively. (C and D) are the ( R ) and ( S ) dia stereoisomers of compound 5 respectively. Color scheme: C, grey; H, white; N, blue; O, red; P, orange; S, yellow. Reprinted with permission from Ikeuchi a, H.; Ahn, Y.; Otokawa, T.; Watanabe, B., Hegazy, L.; Hiratake, J.; Richards, N. G. J. Bioorg. Med. Chem. 2012 20 5915.
144 C ompounds 4 and 5 were employed as a mixture of diastereoisomers at the sulfur stereogenic center. Prev ious work showed that that there is a strong dependence on the stereochemistry of tight binding sulfoximine inhibitors 156 157 For example, the (R) configuration o Glutamylcysteine Synthetase GCS) acted as a potent slow binding inhibitor while the (S) sulfoximine showed a simple reversible inhibition with much lower potency. 156 The molecular interactions of the diastereoisomers of compounds 4 and 5 within the synthetase active site were qualitatively investigated. Because of the structur al resemblance of compounds 4 and 5 with the ASNS transition state, each isomer was placed i n the synthetase active site in the same orientatio n of the transition state. Each model was solvated in a TIP3 P octahedral box followed by energy minimiz ation using the CHARMM27 force field. 59 The diastereoi somer in which the methyl substituent on the stereogenic sulfur points towards the ammonia tunnel corresponds to the ( R ) configuration of both compounds 4 and 5 (Figure 5 3 ). It is believed that the methyl group in the ( R ) configuration mimics the ammonia aspartyl AMP ( Figure 5 2 .) T herefore we believe that the ( R ) configuration of both compounds 4 and 5 is the true inhibitor. The critical electrostatic interaction with Lys 449 was maintained in both diastereoisomer s of com pound 4 The amino group of both isomers also makes a hydrogen bond with Glu 348 which is a key residue in the ASNS active site and is aspartyl AMP intermediate 104 However, t he amino group in the stereo isomers of compound 4 undergoes electrostatic interactions with different carboxylate side chains.
145 Figure 5 4 E lectrostatic potential in the synthetase active site of the AS B model rendered on the van der Waals surface of residues located within 0.4 nm of aspartyl AMP intermediate. Regions of positive and negative electrostatic potential are colored blue and red, respectively. The potential was computed using the Poisson Boltzmann equation using all atoms i n the uncomplexed protein, and wa s visualized using the VMD software package. Quantitative values at 300 K can be computed by multiplying those shown in the legend by 0.026 J/mol/C. ). Reprinted with permission from Ikeuchi a, H.; Ahn, Y.; Otokawa, T.; Wata nabe, B., Hegazy, L.; Hiratake, J.; Richards, N. G. J. Bioorg. Med. Chem. 2012 20 5915.
146 In case of the carboxy N phosphonosulfonimidoyl 5 the key residue Glu 348 reorients and makes a hydrogen bond with other structures. The Lys 449 electrostatic intera ction with the phosphate group is maintained only in the ( R ) configuration. Examination of the electrostatic potential map of the active site revealed that the active site has mainly negative electrostatic potential with the exception of only two portions with positive potential stabilizing the negative charge s of the phosphate and carboxylate groups (Figure 5 3 ). Therefore the higher affinity towards hASNS inhibition of compound 4 over compound 5 may be due to the overall charge and stability within the a ctive site Experimental efforts to obtain compound 4 in an optically pure form were unsuccessful. Alternatively, computational methods will be used to e stimate the free energy of binding. A f ree energy cycle is a widely used computational technique to ca lculate the difference in binding free energy of two different ligands to a given receptor. The thermodynamic cycle employed in this study is shown in F igure 5 5 which describes the binding of the (R) and (S) phosphonosulfonimidoyl inhibitor to ASNS. Fr ee energy changes 3 4 describe the binding process and can be studied experimentally However, they are very difficult to calculate computationally Free energy changes 1 2 correspond to alchemical nonphysical, mutation between the two iso mer s in th e aqueous and enzyme bound states a state function and therefore path independent, both the vertical and horizontal legs provide routes to the difference in the binding free energies of the (R) and (S) N pho sphonosulfonimidoyl inhibitors to the enzyme (Equation 5 1) (5 1)
147 Figure 5 5 Thermodynamic cycle used for the calculation of relative free energy of binding of (R) and (S) diatereoisomers of the phosp honosulfonimidoyl inhibitor (compound 4 ) to ASNS. G 3 and G 4 are the binding free energy changes of the two diastereoisomers to ASNS while G 1 and G 2 correspond to the two hypothetical reactions that convert the (R) phosphonosulfonimidoyl into (S) pho sphonosulfonimidoyl when bound with ASNS and in aqueous environment respectively. The direction of the alchemical transformation (doted) arrows indicate s that the change in free energy is calculated by taking the difference between the free energy of the s tate at the end and that at the beginning
148 The free energy change for the alchemical transformation of the (R) isomer into the (S) isomer is calculated by the introduction of intermediate hybrid states coupled by an independent variable The chimeric change is performed by the gradual vanishing of specific atoms in the initial state, (R) appearance of the necessary atoms in the (S) configuration through intermediate free i between any two successive states is given by: (5 2) where the ensemble average is calculated as a time average over the Hamiltonian trajectory, (5 3) To avoid convergence errors, the initial and final states should be very similar and should be kept small. Therefore, the alchemical pathway for the transformation of the (R) isomer into the (S) isomer will be performed on several step s (Figure 5 6): (1) In the first step (S1) of transformation, the methyl group of the ( R ) inhibitor will be mutated into a hydroxyl group, (2) In S2, The hydroxyl group will be deprotonated and mutated into a negatively charged oxygen; (3) In S3, the nega tively charged oxygen of the N phosphonosulfonimidoyl inhibitor will be mutated into a hydroxyl group in the (S) position of the stereospecific sulfur; (4) In the final step, S4, the hydroxyl group at the (S) position, will be mutated into a methyl grou p at the (S) position. The calculated free energies converge better when the atoms disappear instead of being added. Therefore,
149 the transmutation step (S4) will be computed in the reverse direction, i. e, the methyl group on the (S) position will be mutat ed in to a hydroxyl group on the (S) position and the calculated free energy sign will be reversed when calculating the total transmutation free energy. The total transmutation free energy is the summing of the for each mutation step We propose that th e free energy of deprotonation of the hydroxyl derivatives of both diastereoisomers are equal in either aqueous or enzyme bound states and their values cancel out when calculati ng the total transmutation free energies (steps S2 in the forward direction and S3 in the backward direction ) Thus, only two transmutation step calculations will be required (steps S1 in the forward direction and S4 in the backward direction). T he f ree energy perturbation calculations will be carried out using the TSM module in CH ARMM. Detailed parameter optimization for the N phosphonosulfonimidoyl functional group for accurate representation of its chemical and physical behavior using computational calculations is explained in detail in C hapter 2. Also, b oth the (R) and (S) met hyl derivatives of the N phosphonosulfonimidoyl inhibitor (compounds 4 and 5 ) have been modeled and equilibrated following the same equilibration procedure described in C hapter 3. However, the presence of the hydroxyl substituent of the N phosphonosulfonim idoyl inhibitor requires careful parameter validation and optimization. P arameters that represent the hydroxyl derivative of the N phosphonosulfonimidoyl functional group will be calculated following the same parameter optimization protocol described in de tail in C hapter 2. will be calculated using E quation 5 3. A very small step increment ( = 0.05 ) will be
150 used with equilibration and simulation of each chimeric state. The diff erence in binding free energy will be computed using E quation 5 1 Figure 5 6 Pathway for the transmutation between the ( R ) and ( S ) phosphonosulfonimidoyl inhibitor. Free energy perturbation calculations for S1 will be performed in the forward direction. Free energy perturbation calculations for S 4 will be performed in the backward direction. S2 and S 3 free energies are proposed to cancel each other
151 APPENDIX A QM AND MM OPTIMIZED VIBRATIONAL SPECTRA FOR N PHOSPHONO SULFONIMIDOYL MODEL COMPOUND Table A 1 Vibrational spectra computed for the N phosphonosulfonimidoyl mo lecule at the scaled MP2 level and with optimized CGenFF a parameters MP2/6 31+G(d) scaled by a factor 0.943 CGenFF Freq b Assign (%) Assign (%) Ass ign (%) Freq b Assign (%) Assign (%) Assign (%) 62.6 tdOPNS (98) tdNSC9H (7) 41.1 tdOPNS (89) tdC4OPN (6) 81.4 tdPNSC9 (53) tdC4OPN (30) scO3PN8 (11) 70.6 tdPNSC9 (59) tdC4OPN (38) 100.2 tdNSC9H (70) tdPNSC9 (15) 85.7 tdNSC9H (50) tdC4OPN (30) tdPNSC9 (9) 114.9 tdC4OPN (52) tdPNSC9 (32) 99.6 tdPNSC9 (38) tdNSC9H (29) dPNS (18) 151.4 dPNS (35) tdPOC4H (20) dC9SN (18) 154.9 dPNS (51) tdPOC4H (26) dC9SN (17) 168.4 tdPOC4H (67) dPNS (16) tdPNSC9 (7) 182.8 tdPOC4H(24) dPNS (24) dC13SN (14) 196.1 scSC9C19 (35) dPNS (24) dC9SN (17) 207.2 dC9SN (29) twPO2O7 (23) rPO2O7 (14) 218.1 tdSCC19H (65) dPOC (7) dC13SC9 (5) 219.1 rPO2O7 (16) twPO2O7 (16) scSC9C19 (16) 224.7 dPOC (35) twPO2O7 (14) tdSCC19H(11) 229.7 tdSCC19H(38) tdNSC13H (28) tdPOC4H (9) 250.6 dC13SN (36) twPO2O7 (28) dC13SC9 (9) 240.3 dC13SN (36) tdNSC13H (12) wPO2O7 (12) 259.7 tdNSC13H (84) 258.2 tdNSC13H(35) tdSCC19 H (30) dPOC (10) 281.6 dC13SC9 (25) twPO2O7 (14) rSO' (11) 277.6 dPOC (52) twPO2O7 (13) tdNSC13H (10) 290.7 scSC9C19 (25) scO3PN8 (10) rSO' (8) 285.8 scSC9C19(37) rSO (20) sC9S (10) 319.2 dC13SC9 (25) sN8P (21) dC9SN (10) 335.8 dC13SC9 (62) rSO (13) wPO2O7 (4) 340.0 scO3PN8 (30) rSO' (20) rSO (14) 376.6 scO3PN8 (36) dC13SN (20) wPO2O7 (14) 387.2 scPO2O7 (16) wPO2O7 (15) dC9SN (11) 383.3 rSO' (54) rSO (14) 419.6 dC13SN (29) rSO (24) twPO2O7 (7) 389.7 dC13SN (24) wPO2O7 (20) scSC9C19 (14) 450.1 wPO2O7 (24) rSO' (13) rPO2O7 (11) 418.2 scO3PN8 (29) rSO (18) wPO2O7 (13) 491.1 scPO2O7 (37) dPOC (18) scO3PN8 (15) 455.3 scPO2O7 (31) rPO2O7 (2 5) dPOC (13) 515.6 wPO2O7 (36) rPO2O7 (25) scPO2O7 (14) 482.4 scPO2O7 (37) dPOC (19) twPO2O7 (14) 594.5 sC9S (34) sC13S (33) sN8P (12) 580.2 sN8P (27) sC9S (17) sC13S (13) 669.2 sO3P7 (67) wPO2O7 (5) 645. 3 sC9S (31) sO3P7 (17) sC13S (9) 701.5 sC13S (28) sC9S (25) rCH2C9 (11) 686.5 sO3P7 (48) sN8P (16) wPO2O7 (14) 720.7 sN8P (34) sC13S (13) rCH2C9 (11) 719.7 sC13S (58) sC9S (17) rCH2C9 (5)
152 Table A 1 Continued. MP2/6 31+G(d) scaled by a factor 0.943 CGenFF Freq b Assign (%) Assign (%) Assign (%) Freq b Assign (%) Assign (%) Assign (%) 781.4 rCH2C9 (34) rCH3C19' (23) sC13S (15) 821.5 rCH2C9 (70) rCH3C19' (14) tdNSC9H (5) 958.7 rCH3C13 (58) rCH3C19 (11) rCH3C13' (6) 941.0 rCH3C13' (81) sN8S (9) 964.6 rCH3C13' (53) sC9C19 (21) rCH3C19 (6) 973.7 rCH3C13 (87) adCH3C13 (6) 977.3 sC9C19 (19) rCH3C13 (19) rCH3C19 (15) 993.5 rCH3C19 (42) sC9C19 (30) w CH2C9 (9) 1003.4 ssPO (81) 1003.2 ssPO (66) sO3C4 (8) 1028.4 sO3C4 (59) sN8S (14) sS=O (12) 1027.7 rCH3C19' (45) sN8S (14) rCH3C13' (7) 958.7 rCH3C13 (58) rCH3C19 (11) rCH3C13' (6) 941.0 rCH3C13' (81) sN8S (9) 964.6 rCH3C13' (53) sC9C19 (21) rCH3C19 (6) 973.7 rCH3C13 (87) adCH3C13 (6) 977.3 sC9C19 (19) rCH3C13 (19) rCH3C19 (15) 993.5 rCH3C19 (42) sC9C19 (30) wCH2C9 (9) 1003.4 ssPO (81) 1003.2 ssPO (66) sO3C 4 (8) 1028.4 sO3C4 (59) sN8S (14) sS=O (12) 1027.7 rCH3C19' (45) sN8S (14) rCH3C13' (7) 1037.4 sO3C4 (34) sS=O (25) sN8S (12) 1038.1 sO3C4 (67) sdCH3C4 (9) ssPO (8) 1048.5 sC9C19 (43) rCH3C19 (34) sc SC9C19(7) 1043.8 sC9C19 (34) rCH3C19' (31) rCH3C19 (11) 1072.9 rCH2C9 (18) rCH3C19' (17) sN8S (17) 1052.7 sN8S (29) sC9C19 (15) rCH3C19 (12) 1135.9 rCH3C4 (81) r'CH3C4 (15) 1142.7 rCH3C4 (74) ad'CH3C4 (17) adC H3C4 (10) 1154.2 r'CH3C4 (77) rCH3C4 (14) 1154.0 r'CH3C4 (76) adCH3C4 (17) 1181.1 sS=O (42) sN8S (35) 1175.2 saPO (92) rPO2O7 (5) 1212.8 saPO (82) twCH2C9 (5) 1222.7 twCH2C9 (74) sS=O (10) 12 46.0 twCH2C9 (60) rCH3C19 (19) 1244.2 sS=O (66) twCH2C9 (13) 1287.3 wCH2C9 (56) sdCH3C13(28) 1365.1 wCH2C9 (73) scCH2C9 (11) 1311.3 sdCH3C13(67) wCH2C9 (22) 1400.4 sdCH3C13(80) adCH3C13'(10) 1374.5 sdC H3C19(94) 1416.4 sdCH3C19(79) adCH3C19' (8) 1418.7 sdCH3C4 (90) scCH2C9 (6) 1424.4 adCH3C19'(49) adCH3C19(26) sdCH3C19(15) 1423.7 adCH3C13(82) sdCH3C4 (5) rCH3C13 (5) 1432.5 ad'CH3C4 (66) adCH3C4 (17) r'CH3C4 (12) 1432.0 scCH2C9 (93) adCH3C13' (6) 1438.5 adCH3C13(37) adCH3C13'(26) adCH3C19'(15) 1454.5 adCH3C13'(86) scCH2C9 (6) 1444.3 adCH3C19(33) adCH3C19'(29) adCH3C13'(21)
153 Table A 1 Continued. a Optimized vibrational contr ibutions from interactions for which parameters have been developed in this study are shown in bold font ; s stands for bond stretching with the variations ss and sa for symmetric and asymmetric stretching, respectively ; d means angle deformation with the v ariations sd and ad for symmetric and asymmetric deformation, respectively ; td and ti stand for torsional and improper torsion deformation, respectively ; sc stands for scissoring ; r for rocking ; w for wagging and tw for twisting. b Frequencies are expresse d in units of cm 1 MP2/6 31+G(d) scaled by a factor 0 .943 CGenFF Freq b Assign (%) Assign (%) Assign (%) Freq b Assign (%) Assign (%) Assign (%) 1462.6 adCH3C4 (80) ad'CH3C4 (15) 1458.8 scCH2C9 (92) wCH2C9 (8) 1472.7 adCH3C19(70) adCH3C19'(21) rCH3C19'(6) 1462.4 adCH3C13(44) adCH3C13'(36) sdCH3C13 (8) 1480.4 adCH3C19'(71) adCH3C19(21) rCH3C19 (6) 1484.1 adCH3C4 (55) rCH3C4 (21) ad'CH3C4 (13) 1489.3 ad'CH3C4 (80) adCH3C4 (14) 1615.4 sdCH3C4 (89) sO3C4 (10) 2894.2 ssCH3C4 (98) 2847.9 ss CH3C13(57) saCH3C13'(41) 2918.0 ssCH3C19 (94) 2854.0 ssCH3C4 (99) 2924.6 ssCH3C13 (68) saCH3C13'(31) 2855.9 ssCH3C9 (98) 2945.5 ssCH3C9 (99) 2894.1 saCH3C9 (99) 2974.9 saCH3C4 (87) saCH3C4 (11) 2902.4 ssCH3C 19 (98) 2992.8 saCH3C19(83) saCH3C19 (9) 2913.4 saCH3C4' (73) saCH3C4 (26) 2995.1 saCH3C4 (88) saCH3C4 (12) 2914.9 saCH3C13 (81) saCH3C13' (20) 3005.8 saCH3C9 (81) saCH3C19'(13) saCH3C19(9) 2918.5 saCH3C4 (74) saCH3C4' (26) 3 023.4 saCH3C19'(76) saCH3C9 (17) 2922.1 ssCH3C13(45) saCH3C13'(41) saCH3C13( 18) 3038.8 saCH3C13(99) 2958.2 saCH3C19(98) 3043.7 saCH3C13'(67) ssCH3C13(27) 2960.8 saCH3C19'(98)
154 APPENDIX B CHARMM TOPOLOGY AND PARAMETER FILE FOR ACYLPHOSPHATE AND N PHOSPHONOSULFONIMIDOYL MODEL COMPOUNDS CHARMM Topology and Parameter File for Acylphosphate Model Toppar stream file generated by CHARMM General Force Fie ld (CGenFF) program version 0.9.6 beta For use with CGenFF version 2b7 * Topologies generated by CHARMM General Force Field (CGenFF) program version 0.9.6 beta 36 1 !optimized parameters for acetyl phosphate (Lamees Hegazy) OG305 should be assig ned the same mass and Lenn ard Jones parameters by the user in the main CGenFF topology and parameter files RESI ampa 1.000 GROUP CHARGE ATOM P1 PG1 1.300 ATOM O2 OG2P1 0.710 ATOM O3 OG303 0.460 ATOM C4 CG331 0.170 ATOM H5 HGA3 0.090 ATOM H6 HGA3 0.090 ATOM O7 OG2P1 0.710 ATOM O8 OG305 0.380 New atom type, same VdW parameters as OG303 ATOM C9 CG2O2 0.340 ATOM O10 OG2D1 0.480 ATOM C11 CG331 0.270 ATOM H12 HGA3 0.090 ATOM H13 HGA3 0.090 ATOM H14 HGA3 0.090 ATOM H15 HGA3 0.090 BOND P1 O2 BOND P1 O3 BOND P1 O7 BOND P1 O8 BOND O3 C4
155 BOND C4 H5 BOND C4 H6 BOND C4 H15 BOND O8 C9 BOND C9 O10 BOND C9 C 11 BOND C11 H12 BOND C11 H13 BOND C11 H14 IMPR C9 C11 O10 O8 END read param card flex append BONDS CG2O2 OG305 230.00 1.3400 OG305 PG1 170.00 1.7800 ANGLES CG331 CG2O2 OG305 30.00 104.00 05.00 2.32600 OG2D1 CG2O2 OG305 70.00 118.00 160.00 2.25760 CG2O2 OG305 PG1 70.00 121.50 OG2P1 PG1 OG305 60.00 103.000 OG303 PG1 OG305 60.00 90.80 DIHEDRALS OG305 CG2O2 CG331 HGA3 0.0000 3 0.00 CG331 CG2O 2 OG305 PG1 3.8000 1 180.00 CG331 CG2O2 OG305 PG1 1.6000 2 180.00 OG2D1 CG2O2 OG305 PG1 1.3000 1 180.00 OG2D1 CG2O2 OG305 PG1 2.6000 2 180.00 CG2O2 OG305 PG1 OG2P1 0.1000 3 0.00 CG 2O2 OG305 PG1 OG303 0.1000 2 180.00 CG2O2 OG305 PG1 OG303 0.1000 3 0.00 CG331 OG303 PG1 OG305 1.4700 2 0.00 CG331 OG303 PG1 OG305 0.7000 3 0.00 IMPROPERS CG2O2 CG331 OG2D1 OG305 56. 0000 0 0.00 END RETURN
156 CHARMM Topology and Parameter File f or N phosphonosulfonimidoyl Model Toppar stream file generated by CHARMM General Force Field (CGenFF) progra m version 0.9.6 beta For use with CGenFF version 2b7 * Topologies generated by CHARMM General Force Field (CGenFF) program version 0.9.6 beta 36 1 !optimized parameters for N phosphonosulfonimidoyl functional group RESI msxn 1.000 GROUP ATOM P1 PG1 0.200 ATOM O2 OG2P1 0.460 ATOM O3 OG303 0.280 ATOM C4 CG331 0.170 ATOM H5 HGA3 0.090 ATOM H6 HGA3 0.090 ATOM O7 OG2P1 0.460 ATOM N8 NG2D1 0.380 ATOM C9 CG321 0.020 ATOM H10 HGA3 0.090 ATOM S11 SG3O2 0.120 ! ATOM O12 OG2P1 0.420 ATOM C13 CG331 0.110 ATOM H14 HGA3 0.090 ATOM H15 HGA3 0.090 ATOM H16 HGA3 0.090 ATOM H17 HGA2 0 .090 ATOM H18 HGA2 0.090 ATOM C19 CG331 0.270 ATOM H20 HGA3 0.090 ATOM H21 HGA3 0.090 ATOM H22 HGA3 0.090 BOND P1 O2 BOND P1 O3 BOND P1 O7 BOND P1 N8 BOND O3 C4 BOND C4 H5 BOND C 4 H6 BOND C4 H10
157 BOND N8 S11 BOND C9 S11 BOND C9 H17 BOND C9 H18 BOND C9 C19 BOND S11 O12 BOND S11 C13 BOND C13 H14 BOND C13 H15 BOND C13 H16 BOND C19 H20 BOND C19 H21 BOND C19 H22 END read param card flex append BON DS NG2D1 PG1 100.00 1.7220 NG2D1 SG3O2 400.00 1.5320 ANGLES PG1 NG2D1 SG3O2 30.00 113.00 NG2D1 PG1 OG2P1 50.00 106.00 NG2D1 PG1 OG303 94.00 98.80 CG321 SG3O2 NG2D1 65.00 114.30 CG331 SG3O2 NG2D1 79.00 114.00 NG2D1 SG3O2 OG2P1 65.00 119.00 DIHEDRALS CG331 CG321 SG3O2 NG2D1 0.0010 3 0.00 CG331 CG321 SG3O2 NG2D1 1.4000 1 180.00 HGA2 CG321 SG3O2 NG2D1 0.1600 3 0.00 HGA3 CG331 SG3O2 NG2D1 0.1800 3 0.00 !SG3O2 NG2D1 PG1 OG2P1 0.5000 2 0.00 SG3O2 NG2D1 PG1 OG2P1 0.5000 4 0.00 SG3O2 NG2D1 PG1 OG303 3.0000 2 0.00 SG3O2 NG2D1 PG1 OG303 1.8000 1 0.00 CG321 SG3O2 NG2D1 PG1 1.00 2 0.0 OG2P1 SG3O2 NG2D1 PG1 1.00 2 0.0 OG2P1 SG3O2 NG2D1 PG1 2.50 1 180.0 CG331 SG3O2 NG2D1 PG1 0.60 2 0.0 CG331 SG3O2 NG2D1 PG1 1.00 1 0.0 CG331 OG303 PG1 NG2D1 0.4000 1 0.00 CG331 OG303 PG1 NG2D1 0.8000 2 0.00 CG331 OG303 PG1 NG2D1 0.3500 3 0.00 IMPROPERS END RETURN
158 LIST OF REFERENCES 1. Larsen, T. M.; Boehlein, S. K.; Schuster, S. M.; Richards N. G. J.; Thoden, J. B.; Holden, H. M.; Rayment, I. Biochemistry 1999 38 16146. 2. Zalkin, H. a.; Smith, J. L. Adv. Enzymol. Rela t. Areas Mol. Biol. 1998 72 87. 3. Badet, B.; Vermoote, P.; Haumont, P. Y.; Lederer, F.; Le Goffic, F. Biochemistry 1987 26 1940. 4. Tso, J. Y.; Zalkin, H.; Van Cleemput, M.; Yanofsky, C.; Smith, J. M. J. Biol. Chem. 1982 257 3525. 5. Van den Heuvel, R. H. H .; Curti, B.; Vanoni, M. A.; Mattevi, A. Cell. Mol. Life Sci. 2004 61 669. 6. Aslanian, A. M.; Fletcher, B. S; Kilberg, M. S. Biochem. J. 2001 357 321. 7. Cooney, D. A.; Handschumacher, R. E. Annu. Rev. Pharmacol. 1970, 10 421. 8. Chakrabarti, R.; Schuster, S. M. Int. J. Pediatr. Hemotol./Oncol. 1997 4 597. 9. Chakrabarti, R.; Wylie, D. E.; Schuster, S. M. J. Biol. Chem. 1989 264 15494 10. Muller, H. J., BooS, J. Crit. Rev. Oncol. Hematol. 1998 28 97. 11. Pieters, R.; Hunger, S. P.; Boos, J.; Rizzari, C.; Silverman, L.; Baruchel, A. Cancer 2011 117 238. 12. Ueno, T.; Ohtawa, K.; Mitsui, K.; Kodera, Y.; Hiroto, M.; Matsushima, A.; Inada, Y.; Nishimura, H. Leukemia 1997 11 1858. 13. Richards, N. G.; Kilberg, M. S. Annu. Rev. Biochem. 2006 75 629. 14. Tallal, L.; Tan, C.; Oet tgen, H.;Wollner, N.; McCarthy, M.; Helson, L.; Burchenal, J.; Karnofsky, D.; Murphy, M. L. Cancer 1970 25 306. 15. Barr, R.D.; DeVeber, L.L.; Pai, K.M.; Andrew, M.; Halton, J.; Cairney, A. E.; Whitton, A. C. Am. J. Pediatr. Hematol.Oncol 1992 14 136. 16. T erebelo, H. R.; Anderson, K.; Wiernik, P. H.; Cuttner, J.; Cooper, R. M.; Faso, L.; Berenberg, J. L. Am. J. Clin. Oncol. 1986 9 411. 17. Cancer 1979 43 1089. 18. Prager, M. D.; Bachynsk y, N. Biochem. Biophys. Res. Commun. 1968 31 43.
159 19. Kiryama, Y.; Kubota, M.; Takimoto, T.; Kitoh, T.; Tanizawa, A.; Akiyama, Y.; Mikawa, H. Leukemia 1989 3 294. 20. Leslie, M.; Case, M. C.; Hall, A. G.;Coulthard, S. A. Br. J. Haematol. 2006 132 740. 21. Manning J. M.; Moore, S.; Rowe, W. B.; Meister, A. Biochemistry 1969 8 2681. 22. Koizumi, M.; Hiratake, J.; Nakatsu, T.; Kato, H.; Oda, J. I. J. Am. Chem. Soc. 1999 121 5799. 23. Lorenzi, P. L.; Reinhold, W. C.; Rudelius, M.; Gunsior, M.; Shankavaram, U.; Bussey, K. J.; Scherf, U.; Eichler, G. S.; Martin, S. E.; Chin, K.; Gray, J. W.; Kohn, E. C.; Horak, I. D.; Von Hoff D. D.; Raffeld M.; Goldsmith P. K.; Caplen, N. J.; Weinstein, J. N. Mol. Cancer Ther. 2006 5 2613. 24. Lorenzi, P. L.; Llamas, J.; Gunsior, M.; Ozbun, L.; Reinhold, W. C.; Varma, S.; Ji, H.; Kim, H.; Hutchinson, A. A.; Kohn, E. C.; Goldsmith, P. K.; Birrer, M. J.; Weinstein, J. N. Mol. Cancer Ther. 2008 7 3123. 25. Sircar, K.; Huang, H.; Hu, L. M.; Cogdell, D.; Dhillon, J.; Tzelepim, V.; Efstathiou, E.; Ko umakpayi, I. H.; Saad, F.; Luo, D. J.; Bismar, T. A.; Aparicio, A.; Troncoso, P.; Navone, N.; Zhang, W. Am. J. Pathol. 2012 180 893. 26. Oinonen, C.; Rouvinen, J. Protein Sci. 2000 9 2329. 27. Brannigan, J, A.; Dodson, G.; Duggleby, H. J.; Moody, P. C. E.; Smi th, J. L.; Nature 1998 378 416. 28. Pettersen, E. F.; Goddard, T. D.; Huang, C. C.; Couch, G. S.; Greenblatt, D. M.; Meng, E. C.; Ferrin, T. E. J. Comput. Chem. 2004 25 1605. 29. Lemke, C. T.; Howell, P. L. Structure 2001 9 1153. 30. Tesmer, J. J. G.; Klem, T. J .; Deras, M. L.; Davisson, V. J.; Smith, J. L. Nat. Struct. Biol. 1996 3 74. 31. Bachmann, B. O.; Li, R.; Townsend, C. A. Proc. Natl. Acad. Sci. USA. 1998 95 9082. 32. Miller, M. T.; Gerratana, B.; Stapon, A.; Townsend, C. A.; Rosenzweig, A. C. J. Biol. Chem. 2 003 278 40996. 33. Horowitz, B.; Meister, A. J. Biol. Chem. 1972 247 6708. 34. Boehlein, S.K.; Schuster, S.M.; Richards, N.G.J. Biochemistry 1996 35 3031. 35. Boehlein, S. K.; Stewart. J. D.; Walworth, E. S.; Thirumoorthy, R.; Richards, N. G. J.; Schuster, S. M. Biochemistry 1998 37 13230.
160 36. Kim, J.H.; Krahn, J.M.; Tomchick, D.R.; Smith, J.L.; Zalkin, H. J. Biol. Chem. 1996 271 15549. 37. Van den Heuvel, R. H. H.; Ferrari, D.; Bossi, R. T.; Ravasio, S.; Curti, B.; Vanoni, M. A.; Florencio, F.J.; Mattevi, A.; J. Bio l. Chem. 2002 277, 24579. 38. Boehlein, S. K.; Stewart. J. D.; Walworth, E. S.; Thirumoorthy, R.; Richards, N. G. J.; Schuster, S. M. Biochemistry 1998 37 13230. 39. Hartman, S. C.; Buchanan, J. M. J. Biol. Chem. 1958 233 451. 40. Abrahms, R.; Bentley, M. Arch. B iochem. Biophys. 1959 79 91. 41. Moyed, H. S.; Magasinik, H. S. J. Biol. Chem 1957 226 351. 42. Miller, M. T.; Bachmann, B. O.; Townsend, C. A.; Rosenzweig, A. C. Proc. Natl. Acad. Sci. USA 2002, 99 14752. 43. Fresquet, V.; Thoden, J. B.; Holden, H. M.; Raushel, F. M. Bioorg. Chem. 2004 32 63. 44. Bork, P.; Koonin, E. V. Proteins Struct. Funct. Genet. 1994 20 347. 45. Tesson, A. R.; Soper,T. S.; Ciustea, M.; Richards, N.G.J. Arch. Biochem. Biophys. 2003, 413 23. 46. Boehlein, S. K.; Walworth, E. S.; Richards, N. G. J.; S chuster, S. M. J. Biol. Chem 1994 269 26789. 47. Boehlein, S. K.; Nakatsu, T.; Hiratake, J.; Thirumoorthy, R.; Stewart, J.D.; Richards, N.G.J.; Schuster, S.M. Biochemistry 2001 40 11168. 48. Boehlein, S.K.; Walworth, E.S.; Richards, N.G.J.; Schuster, S.M. J. Biol. Chem. 1997 272 ,12384. 49. Gutierrez, J. A.; Pan, Y. X.; Koroniak, L.; Hiratake, J.; Kilberg, M. S.; Richards, N. G. J. Chem. Biol. 2006 13 1339. 50. Koroniak, L.; Ciustea, M.; Gutierrez, J. A.; Richards, N. G. Org. Lett. 2003 5 2033. 51. Ikeuchi, H.; Meyer, M. E.; Ding, Y.; Hiratake, J.; Richards, N. G. Bioorg. Med. Chem 2009 17 6641. 52. Hegazy, L.; Richards, N. G. J. J. Mol. Model. 2013 doi: 10.1007/s00894 013 1990 x
161 53. Brooks, B. R.; Brooks, C. L.; Mackerell, A. D., Jr.; Nilsson, L., Petrella, R. J.; Roux, B .; Won, Y.; Archontis, G.; Bartels, C.; Boresch, S.; Caflisch, A.; Caves, L.; Cul, Q.; Dinner, A. R.; Feig, M.; Fischer, S.; Gao, J.; Hodoscek, M.; Im, W.; Kuczera, K.; Lazaridis, T.; Ma, J.; Ovchinnikov, V.; Paci, E.; Pastor, R. W.; Post, C. B.; Pu, J. Z. ; Schaefer, M.; Tidor, B.; Venable, R. M.; woodcock, H. L.; Wu, X.; Yang, W.; York, D. M.; Karplus, M. J. Comput. Chem. 2009 30 1545. 54. Case, D. A.; Cheatham, T. E.; Darden, T.; Gohlke, H.; Luo, R.; Merz, K. M. Jr.; Onufriev, A.; Simmerling, C.; Wang, B.; Woods, R. J. Comput. Chem. 2005 26 1668. 55. Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A. J. Am. Chem. Soc. 1995 117 5179. 56. Kaminski, G.; Friesner, R. A.; Tirado Rives, J.; Jorgensen, W. L. J. Phys. Chem. B 2001 ; 105 6474. 57. Price, M. L. P.; Ostrovsky, D.; Jorgensen, W. L. J. Comput. Chem. 2001 22, 1340. 58. Damm, W.; Frontera, A.; Tirado Rives, J.; Jorgensen, W. L. J. Comput. Chem. 1997 18 1955 59. Ikeuchi a, H.; Ahn, Y.; Otokawa, T.; Watanabe, B., Hegazy, L.; Hiratake, J.; Richards, N. G. J. Bioorg. Med. Chem. 2012 20 5915. 60. Agarwal. V.; Nair, S. K. Med. Chem. Commun 2012 3 887. 61. LaRonde LeBlanc, N.; Resto, M.; Gerratana, B. Nat. Struct. Mol. Biol 2009 16 4 21. 62. Duckworth, B. P.; Nelson, K. M.; Aldrich, C. C. Curr. Top. Med. Chem. 2012 12 766. 63. Curnow, A. W.; Hong, K. W.; Yuan, R.; Kim, SI.; Martins, O.; Winkler, W.; Henkin, T. M.; Soll, D. Proc. Natl. Acad. Sci. USA 1997 94 .11819. 64. Horiuchi, K. Y.; Harpel, M R.; Shen, L.; Luo, Y.; Rogers, K. C.; Copeland, R. A. Biochemistry 2001 40 : 6450. 65. Oshikane, H.; Sheppard, K.; Fukai, S.; Nakamura, Y.; Ishitani, R.; Numata, T.; Sherrer, RI.; Feng, L.; Schmitt, E.; Panvert, M.; Blanquet, S.; Mchulam, Y.; Soll, D.; Nurek i, O. Science 2006 312 1950. 66. Becker, O. M., Mackerell, A. D., Roux, B.; Watanabe, M., Ed. Computational Biochemistry and biophysics; Marcel Dekker: New York USA, 2001
162 67. Vanommeslaeghe, K.; Hatcher, E.; Acharya, C.; Kundu, S.; Zhong, S.; Shim, J.; E. Dari an, O. Lopes, P.; Vorobyov, I; Mackerell, A. D., Jr. J. Comput. Chem. 2010 31 671. 68. Vanommeslaeghe, K.; Mackerell, A. D., Jr. J. Chem. Inf. Model. 2012 52 3144. 69. Retrieved from https://www.paramchem.org/ 70. Vanommeslaeghe, K.; Raman, E. P.; Mackerell, A. D ., Jr. J. Chem. Inf. Model. 2012 52 3155. 71. Kuczera, K.; Wiorkiewicz, J.; Karplus, M. CHARMM. Harvard University. 1993. 72. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, B.; P etersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian, H. P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenb., Gaussian, Inc., Wallingford CT. 2009. 73. Francl, M. M.; Pietro, W. J.; Hehre W. J.; Binkley, J. S.; Gordon, M. S.; DeFrees, D, J.; Pop le, J. A. J. Chem. Phys. 1982 77, 3654. 74. Head Gordon, M.; Pople, J. A.; Frisch, M. J. Chem. Phys. Lett. 1988 153 503. 75. Jorgensen, w. l.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J. Chem. Phys. 1983 79 926. 76. Ditchfield, R.; Hehre, W. J .; Pople, J. A. J. Chem. Phys. 1971 54 724. 77. Frisch, M. J.; Pople, J. A.; Binkley, J. S. J. Chem. Phys 1984 80 3265. 78. Scott, A. P.; Radom, L. J. Phys. Chem. 1996 100 16502. 79. Allen, M. P., Tildesley, D. J. Computer Simulation of Liquids ; Clarendon Press : Oxford, 1987. 80. Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. J. Comput. Phys. 1977 23 327. 81. Shuichi, N.; Klein, M. L. Mol. Phys. 1983 50 1055. 82. Pulay, P.; Fogarasi, G.; Pang, F.; Boggs, J. E., J. Am. Chem. Soc. 1979 101 2550. 83. Frisch, M. J.; Pople J. A.; Binkley, J. S. J. Chem. Phys 1984 80 3265. 84. Vinson, V. J. Science 2009 324, 197. 85. Karplus, M.; McCammon, A. Nat. Struct. Biol. 2002 9 646. 86. Lee, G. M.; Craik, C. S. Science 2009 324 213.
163 87. Snow, C. D.; Sorin, E. J.; Rhee, Y. M.; Pande, V. S. An nu. Rev. Biophys. Biomol. Struct. 2005 34, 43. 88. McGovern, S. L.; Shoichet, B. K. J. Med. Chem. 2003 46 2895. 89. Erickson, J. A.; Jalaie. M; Robertson, D. H.; Lewis, R. A.; Vieth, M. J Med. Chem. 2004 47, 45. 90. Murray, C. W.; Baxter, C. A.; Frenkel, A. D. J. Comput. Aided Mol. Des. 1999 13 547. 91. Sugita, Y., Okamoto, Y. Chem. Phys. Lett., 1999 314 ,141. 92. Meyer, M. E. Inhibition, Characterization, and Crystallization of Glutamine Dependent Asparagine Synthetase, University of Florida, Gainesville, FL, 2010. 93. Mil ler, M. T.; Bachmann, B. O.; Townsend, C. A.; Rosenzweig, A. C. Nature 2001 8 684. 94. Ding, Y. Experimental and Theoretical Characterization of Asparagine Synthetase Inhibitors. Ph.D. Dissertation, University of Florida, Gainesville, FL, 2002. 95. Mcquarrie, D. A. Statistical mechanics ; University Science Books: Sausalito, CA, USA, 2000. 96. Cramer, C. J. Essentials of computational chemistry, 2 nd ed ; John Wiley & Sons: England, U.K., 2004. 97. Hoover, W. G. Phys. Rev. A 1985 31 1695. 98. Li, H.; Robertson, A. D.; Jensen J. H. Proteins 2005 61 704. 99. Bas, D. C.; Rogers, D. M.; Jensen, J. H. Proteins 2008 73 765. 100. Olsson, M. H. M.; Sndergard, C. R.; Rostkowski, M.; Jensen, J. H. J. Chem. Theory Comput. 2011 7 525. 101. Sndergaard, C. R.; Olsson, M. H. M.; Rostkowski, M .; Jensen, J. H. J. Chem. Theory Comput. 2011 7 2284. 102. Darden, T. A.; York, D.; Pedersen, L. G. J. Chem. Phys. 1993, 98 10089. 103. Steinbach, P. J.; Brooks, B. R. J. Comput. Chem. 1994 15, 667. 104. Meyer, M. E.; Gutierrez, J. A.; Raushel, F. M.; Richards, N. G. J. Biochemistry 2010 49 9391. 105. Bachmann, B. O.; Townsend, C. A. Biochemistry 2000 39 11187.
164 106. Raber, M. L.; Arnett, S. O.; Townsend, C. A.; Biochemistry 2009 48 4959. 107. Raber, M. L.; Arnett, S. O.; Townsend, C. A. J. Biol. Chem. 2009 284,217. 108. Raber, M. L.; Castillo, A.; Greer, A.; Townsend, C. A. ChemBioChem 2009 10 2904. 109. Waszkowycz, B.; Perkins, T. D. J.; Sykes, R. A.; Li, J., IBM Syst. J. 2001 40 360. 110. Alonso, H.; Bliznyuk, A. A.; Gready, J. E., Med. Res. Rev. 2006 26 531. 111. Murray, C. W.; Baxter, C. A.; Frenkel, A. D., J. Comput. Aided. Mol. Des. 1999, 13, 547 112. Glide, version 5.5, Schrdinger, LLC, New York, NY, 2009. 113. Lyne, P. D. Drug Discov. Today 2002 7 1047. 114. Reddy, A. S.; Pati, S. P.; Kumar, P. P.; Pradeep, H. N.; Sastry, G. N., Curr. Protein Pept. Sci. 2007 8 329. 115. Stahl, M.; Rarey, M. J. Med. Chem. 2001 44 1035. 116. Perola, E.; Walters, W.P.; Charifson, P.S. Proteins: Struct. Funct. Bioinf 2004 56 235. 117. Halgren, T.A. J. Comput. Chem. 1996 17 520. 118. Meng, E. C.; Shoichet, B. K.; Kuntz, I. D. J. Comput. Chem. 1992 13 505. 119. Weiner, S. J.; Kollman, P. A.; Case, D. A. J. Am. Chem. Soc. 1984 106 765. 120. Weiner, S. J.; Kollman, P. A.; Nguyen, D. T.; Case, D. A. J. Comput. Chem. 1986 7 230. 121. Rocchia, W.; Sridharan, S.; Nicholls, A.; Alexov, E.; Ch iabrera, A.; Honig, B. J. Comput. Chem. 2002 23 128. 122. Grant, J. A.; Pickup, B. T.; Nicholls, A. J. Comput. Chem. 2001 22 608. 123. Baker, N. A.; Sept, D.; Joseph, S.; Holst, M. J.; McCammon, J. A. Proc. Natl. Acad. Sci. USA. 2001 98 10037. 124. Still, W. C.; Te mpczyk, A.; Hawley, R. C.; Hendrickson, T. J. Am. Chem. Soc. 1990 112 6127. 125. Hawkins, G. D.; Cramer, C. J.; Truhlar, D. G. Chem. Phys. Lett. 1995 246 122.
165 126. Qiu, D.; Shenkin, P. S.; Hollinger, F. P.; Still, W. C. J. Phys. Chem. A 1997 101 3005. 127. Huang, S .; Grinter, S. Z.; Zou, X. Phys. Chem. Chem. Phys. 2010 12 12899. 128. Bhm, H. J. J. Comput. Aid. Mol. Des 1994 8 243. 129. Jain, A. N. J. Comput. Aid. Mol. Des. 1996 10, 427. 130. Head, R. D.; Smythe, M. L.; Oprea, T. I.; Waller, C. L.; Green, S. M.; Marshall, G. R. J. Am. Chem. Soc. 1996 118 3959. 131. Eldridge, M. D.; Murray, C. W.; Auton, T. R.; Paolini, G.V.; Mee, R. P.; J. Comput. Aid. Mol. Des. 1997 11 425. 132. Bhm, H. J. J. Comput. Aid. Mol. Des. 1998 12 309. 133. Wang, R.; Gao, Y.; Lai, L.; J. Mol. Model. 1998 4, 379. 134. Wanga, R.; Laib, L.; Wanga S. J. Comput. Aid. Mol. Des. 2002 16 11. 135. Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.; Klicic, J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shelley, M.; Perry, J. K.; Shaw, D. E.; Francis, P.; Shenk in, P. S. J. Med. Chem. 2004, 47, 1739. 136. Muegge, I. Perspect. Drug Discov. Des. 2000 20 99. 137. Koppensteiner, W. A.; Sippl, M. J. Biochemistry (Moscow) 1998 63 247. 138. DeWitte, R. S.; Shakhnovich, E. I. J. Am. Chem. Soc. 1996 118 11733. 139. Charifson, P. S.; C orkery, J. J.; Murcko, M. A.; Walters, W. P. J. Med. Chem. 1999 42 5100. 140. Clark, R. D.; Strizhev, A.; Leonard, J. M.; Blake, J. F.; Matthew, J.B. J. Mol. Graph. Model 2002 20 281. 141. Eldridge, M. D.; Murray, C. W.; Auton., T. R.; Paolini, G. V.; Mee, R. P J. Comput. Aided Mol. Des. 1997 11 425. 142. Jorgensen, W. L.; Maxwell, D. S.; Tirado Rives, J J. Am. Chem. Soc 1996 118 11225. 143. Lang, P. T.; Brozell, S. R.; Mukherjee, S.; Pettersen, E. F.; Meng, E. C.; Thomas, V.; Rizzo, R. C.; Case, D. A.; James, T. L .; Kuntz, I. D. RNA 2009 15 1219. 144. Kuntz, I. D.; Blaney, J. M.; Oatley, S. J.; Langridge, R.; Ferrin, T. E.; J. Mol. Biol. 1982 161 269.
166 145. Ewing, T. J. A.; Makino, S.; Skillman, A. G.; Kuntz, I. D. J. Comput. Aided. Mol. Des 2001 15 411. 146. Lipinski, C. A .; Lombardo, F.; Dominy, B. W.; Feeney, P. J. Adv. Drug Deliv. Rev. 1997 23, 3. 147. Veber, D. F.; Johnson, S. R.; Cheng, H.; Smith, B. R.; Ward, K. W.; Kopple, K. D. J. Med. Chem 2002 45 2615. 148. O'Brien, W. E. Anal. Biochem. 1976 76 423. 149. Morrison, J. F., a nd Walsh, C. T. Adv. Enzymol. Relat. Areas Mol. Biol. 1988 61 201. 150. Albert, A. The Acridines. Their preparation, physical, chemical and biological properties and uses ; Edward Arnold publishers Ltd: London, 1966. 151. Magidson, J.; Grigorowski, A. M. Chem. Ber. 1936 69 396. 152. Charmantary, F.; Demnynck, M.; Carrer, D. J. Med. Chem. 2003 46 967. 153. Patel, M. M.; Mali, M. D.; Patel, S. K. Bioorg. Med. Chem. Lett. 2010 20 6324. 154. Shelley, J. C.; Cholleti, A.; Frye, L. L.; Greenwood, J. R.; Timlin, M. R.; Uchimaya, M. J. Comp. Aided. Mol. Des. 2007 21 ,681. 155. Sanli, S. S.; Altun, Y. K.; Sanli, N.; Alsancak, G.; Beltran, J. L. J. Chem. Eng. Data 2009 54 3014. 156. Tokutake, N.; Hiratake, J.; Katoh, M.; Irie, T.; Kato, H.; Oda, J. Bioorg. Med. Chem. 1998 6 1935. 157. Hiratake, J. Chem. Rec. 2005 5 209.
167 BIOGRAPHICAL SKETCH Lamees Hegazy was born in Dakahlia, Egypt in August of 1984. She received her Bachelor of Science degree in biochemistry with honor s from Mansoura University in May of 2005 Lamees finished the graduate courses at Mansoura University in 20 06 and then moved to the United States of America where she commenced her Ph.D. in computational biochemistry in January of 2009 under the supervision of Prof. Nigel Richards During her course of study Lamees has been focusing on modeling and discovery of lead inhibit ors of asparagine synthetase. S he was recognized with the certificate of outstanding achievement two times L amees graduated from University of Florida in December of 2013.