<%BANNER%>

Developing a Computational Model of Asparagine Synthetase-B Toward Rational Inhibitor Design

Permanent Link: http://ufdc.ufl.edu/UFE0041019/00001

Material Information

Title: Developing a Computational Model of Asparagine Synthetase-B Toward Rational Inhibitor Design
Physical Description: 1 online resource (140 p.)
Language: english
Creator: Humkey, Robert
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: Chemistry -- Dissertations, Academic -- UF
Genre: Chemistry thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: DEVELOPING A COMPUTATIONAL MODEL OF ASPARAGINE SYNTHETASE-B TOWARDS RATIONAL INHIBITOR DESIGN Robert N. Humkey (352) 219-5004 rhumkey@gmail.com Department of Chemistry Dr. Nigel G. J. Richards Ph.D. in Chemistry December 2009 The work in this dissertation represents a significant improvement in a Potential of Mean Force scoring potential for use in docking and in silico screening studies, which can be an important tool in the drug discovery process. This dissertation also presents a novel parameter set for the inclusion of a critical intermediate in the synthetase active site of asparagine synthetase-B enzyme in the CHARMM molecular modeling program. The new model of AS-B that was then optimized can be used in as a target for the in silico screening of compound libraries for the possible identification of novel inhibitors of AS. This model is also a step forward in development of a high-quality quantitative model of AS-B that could be used in the rational design of AS-B inhibitors.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Robert Humkey.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Richards, Nigel G.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0041019:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041019/00001

Material Information

Title: Developing a Computational Model of Asparagine Synthetase-B Toward Rational Inhibitor Design
Physical Description: 1 online resource (140 p.)
Language: english
Creator: Humkey, Robert
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: Chemistry -- Dissertations, Academic -- UF
Genre: Chemistry thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: DEVELOPING A COMPUTATIONAL MODEL OF ASPARAGINE SYNTHETASE-B TOWARDS RATIONAL INHIBITOR DESIGN Robert N. Humkey (352) 219-5004 rhumkey@gmail.com Department of Chemistry Dr. Nigel G. J. Richards Ph.D. in Chemistry December 2009 The work in this dissertation represents a significant improvement in a Potential of Mean Force scoring potential for use in docking and in silico screening studies, which can be an important tool in the drug discovery process. This dissertation also presents a novel parameter set for the inclusion of a critical intermediate in the synthetase active site of asparagine synthetase-B enzyme in the CHARMM molecular modeling program. The new model of AS-B that was then optimized can be used in as a target for the in silico screening of compound libraries for the possible identification of novel inhibitors of AS. This model is also a step forward in development of a high-quality quantitative model of AS-B that could be used in the rational design of AS-B inhibitors.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Robert Humkey.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Richards, Nigel G.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0041019:00001


This item has the following downloads:


Full Text

PAGE 1

DEVELOPING A COMPUTATIONAL MODEL OF ASPARAGINE SYNTHETASE-B TOWARDS RATIONAL INHIBITOR DESIGN By ROBERT N. HUMKEY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009 1

PAGE 2

2009 Robert N. Humkey 2

PAGE 3

To my grandfather, Charles Robe rt Ruyle, and my entire family 3

PAGE 4

ACKNOWLEDGMENTS This work was completed due in large part to the wonderful support I have received from all of my family and friends. I have to first thank my parents, John and Lynne Humkey, for all the love and support they have shown me in wh atever I have done. I w ould like to thank my brothers, Travis and Greg, and sister, Katie, fo r their continued encouragement along the way. I would also like to thank my a dvisor, Dr. Richards, and all of the Richards group, especially Sangbae Lee and Megan Meyer, for their encourag ement and dicussions that have helped with this work. I would also like to thank Lori Clark for her con tinued moral support and assistance guiding me through my time here at UF. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........7 LIST OF FIGURES.........................................................................................................................8 ABSTRACT...................................................................................................................................11 CHAPTER 1 INTRODUCTION................................................................................................................. .13 Acute Lymphoblastic Leukemia.............................................................................................13 Asparagine Synthetase.......................................................................................................... ..14 Glutamine-Dependent As paragine Synthetase................................................................15 Catalytic mechanism................................................................................................15 Enzyme structure......................................................................................................17 Intramolecular ammonia translocation.....................................................................22 Research Objectives............................................................................................................ ....23 2 PARAMETERIZATION OF A REACTION INTERMEDIATE OF ASPARAGINE SYNTHETASE-B FOR USE IN CHARMM.........................................................................26 Introduction................................................................................................................... ..........26 Methods..................................................................................................................................28 Parameterization Methodology.......................................................................................28 Parameterization Strategy................................................................................................30 Results and Discussion......................................................................................................... ..31 Model Selection and Initial Parameters...........................................................................31 Atom Types and Geom etric Parameters..........................................................................33 Partial Atomic Charges....................................................................................................34 Force Constants and Dihedral Rotation...........................................................................36 Vibrational Spectra for the Model Complex...................................................................44 Conclusions.............................................................................................................................46 3 DEVELOPMENT OF A COMPUTATIONAL MODEL OF ASPARAGINE SYNTHETASE-B...................................................................................................................49 Introduction................................................................................................................... ..........49 Methods..................................................................................................................................51 Simulated Annealing Methodolgy...................................................................................51 Simulated Annealing Procedure......................................................................................51 Model Systems................................................................................................................53 AspATP model system.............................................................................................53 5

PAGE 6

AspAMP model system.........................................................................................54 Results and Disscussion..........................................................................................................55 AspATP Model................................................................................................................55 Simulated annealing run 1........................................................................................58 Simulated annealing run 2........................................................................................60 Simulated annealing run 3........................................................................................65 AspAMP Model............................................................................................................68 Simulated annealing run 1........................................................................................71 Simulated annealing run 2........................................................................................75 Simulated annealing run 3........................................................................................78 Conclusions.............................................................................................................................81 4 ENHANCING THE PMF99 SCORING FUNCTION FOR MOLECULAR DOCKING AND VIRTUAL SCREENING..............................................................................................83 Introduction................................................................................................................... ..........83 Methods..................................................................................................................................85 Protein/Ligand Structure Preparation..............................................................................85 Docking Algorithm..........................................................................................................86 Results and Discussion......................................................................................................... ..87 The Enhanced PMF99 Scoring Function........................................................................87 Molecular Docking Algorithm........................................................................................91 Assessing the Performance of the ePMF99 Sc oring Function in Molecular Docking....93 Comparison to Other Dock ing/Scoring Algorithms........................................................97 Virtual Screening for Thymidine Kinase........................................................................98 Correlation of the ePMF99 Scor e and Biological Activity...........................................102 Conclusions...........................................................................................................................103 5 CONCLUSIONS.................................................................................................................. 109 Concluding Remarks............................................................................................................1 09 The Future of the AS-B Project fr om a Computational Perspective....................................110 Modeling the Inhibitors.................................................................................................111 Modeling the Residue Mutations...................................................................................114 APPENDIX CHARMM SIMULATED ANNEALING EXAMPLE INPUT FILES..............116 Heating Input File............................................................................................................. ....116 Equilibration Input File....................................................................................................... ..117 Annealing Input File........................................................................................................... ..118 Final Minimization Input File...............................................................................................126 LIST OF REFERENCES.............................................................................................................128 BIOGRAPHICAL SKETCH.......................................................................................................140 6

PAGE 7

LIST OF TABLES Table page 2-1 The bonds, angles and dihedrals to be parameterized with initial parameters...................32 2-2 Atomic charges and atom types for AspAMP and the model complex...........................33 2-3 Model complex, ab initio and CHARMM geometry data.................................................35 2-4 Vibrational frequencies for the model complex as calculated by HF and CHARMM......45 2-5 Final parameters for the model complex...........................................................................48 3-1 Final energies for the SA Runs for the AspATP model system........................................56 3-2 Final energies for the SA Runs for the AspAMP model system.....................................68 4-1 Comparison of rms deviation () for flexible docking.....................................................95 4-2 Accuracy in cross-docking of thymidine kinase inhibito rs to the 1kim active site...........98 4.3 Complete docking test set and calculated rmsd of docked ligand vs original crystal structure ligand............................................................................................................... ..105 5-1 Sulfoximine 4 model complex and CHARMM parameters necessary............................112 5-2 Initial CHARMM atom types and CHelpG charges assigned for the sulfoximine model complex.................................................................................................................1 12 5-3 AS-B synthetase domain mutant s with the putative relevance........................................115 7

PAGE 8

LIST OF FIGURES Figure page 1-1 The three reactions catalyzed by AS..................................................................................15 1-2 The bioconversion of aspartate into as paragine that is catalyzed by glutaminedependent asparagine synthetase.......................................................................................16 1-3 The putative reaction mechanism of ASNS that proceeds through the reaction intermediate, AspAMP....................................................................................................17 1-4 The AS-B monomer as depicted from the crystal structure...............................................18 1-5 The homodimer as depicted from the crystal structure of AS-B.......................................19 1-6 C1A AS-B model............................................................................................................. ..20 1-7 The intramolecular tunnel of AS-B....................................................................................21 2-1 The structures of the intermediate, AspAMP, and of the tetrahedral transition state for the ASNS catalyzed reaction........................................................................................27 2-2 AspAMP and the model complex with unique atom labels............................................30 2-3 The 2-dimensional ab initio energy surface for dihedral angle CB-CG-O3A-PA versus dihedral angle CG-O3A-PA-O5 of the model complex........................................38 2-4 The 2-dimensional CHARMM energy su rface for dihedral angle CB-CG-O3A-PA versus dihedral angle CG-O3A-PA-O5 of the model complex without energy contributions from the dihedrals in Table2-1.....................................................................39 2-5 Newman projections for the dihedr als O2A-PA-O3A-CG and PA-O3A-CG-OD1..........40 2-6 The 2-dimensional CHARMM energy su rface for dihedral angle CB-CG-O3A-PA versus dihedral angle CG-O3A-PA-O5 of the model complex with energy contributions from the dihedrals........................................................................................41 2-7 The 2-dimensional CHARMM energy su rface for dihedral angle CB-CG-O3A-PA versus dihedral angle CG-O3A-PA-O5 of the model complex using the final set of developed parameters.........................................................................................................43 3-1 Creation of AspAMP intermediate..................................................................................55 3-2 Synthetase active site interaction in the initial AspATP model system.............................57 3-3 Structural comparison of the initial stru cture of the AspATP model with the same structure heated to 600 K...................................................................................................58 8

PAGE 9

3-4 AspATP SA Run 1.............................................................................................................59 3-5 AspATP SA Run 2.............................................................................................................61 3-6 Structural alignments of the minimi zed structure at 390 K and the structure equilibrated at 300 K with the initial structure..................................................................62 3-7 A look at ATP binding interactions in SA Run 2..............................................................63 3-8 A look at free aspartate bindi ng interactions in SA Run 2................................................64 3-9 AspATP SA Run 3.............................................................................................................66 3-10 Structural comparison of the initial structure of the AspATP model with the minimized structure at 300 K.............................................................................................67 3-11 Synthetase active site interaction in the initial AspAMP model system.........................69 3-12 Structural comparison of the initial structure of the AspAMP model with the same structure heated to 600 K...................................................................................................70 3-13 AspAMP SA Run 1.........................................................................................................71 3-14 Active site comparison between the initial AspAMP model and the model minimized at 300 K............................................................................................................73 3-15 The shift of Glu-348...................................................................................................... .....74 3-16 AspAMP SA Run 2.........................................................................................................75 3-17 Active site comparison between the initial AspAMP model and the model minimized at 420 K and the model minimized after equilibration at 300 K.....................77 3-18 AspAMP SA Run 3.........................................................................................................79 3-19 Active site comparison between the initial AspAMP model and the model minimized at 300 K............................................................................................................80 4-1 A plot of the pairwise scores for protein/liga nd atom pair of NC-OD..............................91 4-2 Graphical visualization of optimal lig and poses for arabinose in the arabinosebinding protein...................................................................................................................96 4-3 Structures of ligands employed in cross-docking studies on thymidine kinase...............100 4-4 The effect of water on the bi nding of hpt and dT with TK..............................................102 4-5 Plot showing the correlati on of the ePMF99 score vs. logKi for the pyrimidine ( ) and purine ( ) analogs used in the cross-docking studies................................................103 9

PAGE 10

5-1 Structures of the AS reaction intermediate, AspAMP 1, transition-state 2, and known synthetase site inhibitors......................................................................................111 5-2 The 2-dimensional ab initio energy surface for dihedral angle CB-SA-NA-PA versus dihedral angle CA-CB-SA-NA of the sulfoximine model complex................................113 10

PAGE 11

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DEVELOPING A COMPUTATIONAL MODEL OF ASPARAGINE SYNTHETASE-B TOWARDS RATIONAL INHIBITOR DESIGN By Robert N. Humkey December 2009 Chair: Nigel G. J. Richards Major: Chemistry Although Acute Lymphoblastic Leukemia (ALL) ha s an initial treatment rate of nearly 95%, it claims the lives of over 1,000 people in th e US each year. These patients often develop resistance to L-asparaginase, a main chemotherape utic agent used in the treatment of ALL. There has been a long withstanding, inve rse correlation to patients re sistance to L-asparaginase and up-regulation of the enzyme, asparagine synthetase. Therefore, selectiv e potent inhibitors of asparagine synthetase (AS) may be usef ul for the clinical treatment of ALL. This work focuses on the development of a structural model of E. coli asparagine synthetase-B (AS-B); the glutamine-dependent bact erial form that is similar to human AS. Novel parameters were developed in CHARMM for incorporation of the AS reaction intermediate, aspartyl-AMP, into the AS-B model. This intermed iate has been successfully used as a model for inhibitor design. Therefore, mode ling this intermediate will provide understanding of the protein structure and active site interact ions, which will aid in the development of future inhibitors. Simulated annealing (SA) was used to arri ve at an optimized model of AS-B with glutamine bound in the glut aminase active site and -aspartyl-AMP, pyrophosphate and Mg2+ bound in the synthetase active site. The final mode l was heated from 0 K 600 K; cooled back 11

PAGE 12

12 down to 300 K; then equilibrated at 300 K before a final minimization. The optimized model had a very interesting side chain flip occur for Glu-348. The repositioned side chain would make the translocation of NH3 from the glutaminase site to the synthetase site more probable than the translocation is in the current AS-B crystal structure. The availability of a structural model for AS sets the stage for the application of in silico screening of virtual libraries to identify novel dugs for the treatm ent of asparaginase-resistant ALL. The validation of algorithms and potentials for molecular doc king is a necessary prelude to such efforts. Therefore, an enhanced Potent ial of Mean Force (PMF) scoring function was developed for in silico screening studies. The sum of the work presented here sets the stage for in silico screening or the rational design of inhibitors for AS.

PAGE 13

CHAPTER 1 INTRODUCTION Acute Lymphoblastic Leukemia Acute lymphoblastic leukemia (ALL) is a ca ncer of the blood, in which bone marrow produces large quantities of immature white bl ood cells. This over produc tion of useless white blood cells, in effect, crowds the properly f unctioning white and red blood cells along with platelets in the blood. ALL is most common in ch ildren below the age of 14, affecting three to four children per hundred thousand each year. The American Cancer Society estimates that in 2009, 5,760 new cases of ALL will be diagnosed and 1,400 deaths will be attributed to this form of leukemia.1 The enzyme L-asparaginase, which catalyzes th e hydrolysis of asparagine to aspartate and ammonia,2-4 is commonly used in conjunction with other chemotherapeutic drugs for the treatment of ALL. This usage is due, in large part, to the substantial evidence supporting an inverse correlation between the levels of intrac ellular asparagine biosynthesis and the drug susceptibility of T-cell leukemia. L-asparaginase administered alone can result in complete remission for 40 60% of ALL cases.5,6 When combined with other chemotherapeutic agents, the percentage of untreated A LL patients that experience complete remission jumps to 95%. Despite the apparent benefits, L-as paraginase treatment has limited clinical viability due to three key factors. The first limiting fa ctor is the wide variety of si de effects that result from Lasparaginase treatment.7 The second factor limiting clinical viability is the fact that many of the patients in remission experience relapses with L-asparaginase resistant tumors.7-10 The final limiting factor is the observation that L-asparagina se actually promotes th e growth of resistant tumors and raises their metastatic activity.5,11 The potential adverse e ffects of L-asparaginase 13

PAGE 14

treatment has limited its use for the treatment of A LL, despite estimates that 5 10% of all solid tumors, not solely ALL tumors, may be susceptible to this treatment.10 A major problem in the clinical treatment of relapsed ALL patients is the lack of understanding of L-asparaginase resistance on the molecular level.12,13 Early studies on Lasparaginase resistant leukemia patients found elevated levels of the enzyme asparagine synthetase (AS), which catalyzed the biosynthesis of asparagine in L-asparaginase resistant patients.14 This was the first insight into the correlation between the cellular upregulation of AS and L-asparaginase resistance in leukemia cells. Multiple studies since have confirmed this observation and strengthened the theory that AS is linked to L-asparaginase resistance in leukemia cells.9,15-18 A human MOLT-4 cell line was established from a 19-year-old male patient, diagnosed with ALL in 1971.19 This cell line has provided an excellent model for testing leukemia cells response to varying stimuli. It was found that short-term treatment of drugsensitive MOLT-4 leukemia cells with L-asparaginase resulted in elevated levels of AS, an effect that is not fully reversible.17 Further, work on the MOLT-4 cell line has provided direct evidence that the overexpression of AS is adequate for de velopment of L-asparagi nase resistance. When the drug-sensitive MOLT-4 cell line was transformed using a retrovirus containing the human AS gene under a constitutively active promot er, the resulting overexpression of AS was sufficient to induce L-asparaginase drug resistance in the MOLT-4 cell line.17 The apparent correlation between AS and L-asparaginase resistance in ALL,12,20-22 coupled with the fact that a nanomolar affinity inhibitor of AS has demonstr ated an ability to suppr ess proliferation of a drug-resistant MOLT-4 cell line makes AS a clinically interesting enzyme.23 Asparagine Synthetase Asparagine synthetase (AS), a member of the amidotransferase family of enzymes, catalyzes the biosynthesis of L-asparagine from L-aspartate, in an adenosine triphosphate (ATP) 14

PAGE 15

dependent reaction that uses either glutam ine or ammonia as a source of nitrogen.24 The bacterium Escherichia coli ( E. coli ) contains two separate genes that both code for AS. The first gene, asnA codes for E. coli asparagine synthetase-A (ASA) and catalyzes the ammoniadependent synthesis of asparagine.25-27 The second gene, asnB codes for E. coli asparagine synthetase-B (AS-B)28 and like human AS29 catalyzes the glutamine-dependent synthesis of asparagine, although AS-B is not ev olutionarily related to the human form of AS. All told, AS-B catalyzes three distinct reactions as shown in Figure1-1.30,31 Gln +H2O Glu+NH3ATP + Asp + NH3 Asn +AMP + PPiATP + Asp + Gln Asn + AMP + PPi + Glu (1) (2) (3) Figure 1-1. The three reactions cat alyzed by AS. Reaction (1) is th e hydrolysis of glutamine to glutamate and ammonia. Reaction (2) is the ammonia-dependent conversion of aspartate to asparagine, using ATP a nd producing AMP and pyrophosphate. Reaction (3) is the glutamine-dependent convers ion of aspartate to asparagine. The first residue of glutamine-dependent AS (ASNS), cysteine (Cys-1), mediates the glutaminase activity (reaction (1)) in the N-term inal domain. The sequence of the N-terminal glutamine amide transfer (GAT) domain, along with the distinguishing cysteine residue, places AS in the Class II, or N-terminal nucleophilic (Ntn), family of glutamine-dependent amidotransferases.32,33 Also included in the Ntn-family are glutamine 5-phosphoribosyl-1pyrophosphate amidotransferase (GPATase),34,35 glutamine fructose-6-phosphate amidotransferase (GFAT)36,37 and glutamine synthase.38-40 Glutamine-Dependent Asparagine Synthetase Catalytic mechanism As mentioned earlier, ASNS catal yzes the bioconversion of Laspartate into L-asparagine using ATP and glutamine as seen in Figure 12. ASNS has also shown the ability to use 15

PAGE 16

ammonia, hydroxylamine and hydrazine as its source of nitrogen under in vitro conditions.41 While the enzyme has been isolated from mammals,42-44 yeast,45 and bacteria,28,46 the only native mammalian enzyme studied in detail has been th at which is present in the bovine pancreas.44,47-49 Most of the characterization on the eukaryotic form of the enzyme, however, has been done on active, recombinant human ASNS.50 Recombinant forms of ASNS can also be obtained from E. coli31 and Vibrio cholerae .51 ASNS exhibits high glutaminase activity even in the absence of aspartate. This has resulted in several different suggestions for the order of substrate binding and product release.49,51-55 Based on an observed dependence of the glutamine:asparagine ratio on the initial glutamine concentration, one kinetic scheme suggested by Tesson, et al. predicts that the catalytic activity of synthetase and gl utaminase sites are only weakly coupled.52 This observation means ASNS is unlike the vast majority of the ot her amidotransferases, which all exhibit tightly coupled active sites. Recently, asparagine has b een shown to bind to the glutaminase site as a competitive inhibitor in both AS-B56 and the Vibrio enzyme.51 Therefore, asparagine can act as a means of regulating the glutamine-dependent activity of ASNS. +ATP +AMP+PPiCO2H3N OH O H CO2H3N NH2 O H GlnGlu NH3 Figure 1-2. The bioconversion of aspartate into asparagine that is catalyzed by glutaminedependent asparagine synthetase. The mechanism by which ASNS catalyzes the conversion of aspartate into asparagine is believed to occur via the reaction mechanism depi cted in Figure 1-3. As partate binds in the Cterminal synthetase active si te where the side chain carboxylate is activated by an Mg2+-ATP complex to form the -aspartyl-AMP intermediate ( AspAMP) and pyrophosphate (PPi). The 16

PAGE 17

AspAMP formation has been demonstrated by 18O isotope labeling experiments.44,53 While this is occurring in the synthetase site, glutamine is hydrolyzed to glutamate and ammonia in the Nterminal glutaminase site. The third and final r eaction involves the attack of ammonia, which is shuttled from the glutaminase site to the synthe tase site via the proposed intramolecular tunnel, on AspAMP yielding AMP and aspartate. Figure 1-3. The putative reaction mechanism of ASNS that proceeds through the reaction intermediate, AspAMP. Enzyme structure In an effort to elucidate the mechanism of AS NS, the structure of the Cysteine-1 to alanine (C1A) mutant of E. coli AS-B was crystallized in a ternar y complex with glutamine in the Nterminal glutaminase site and adenosine monophosphate (AMP) in the C-terminal synthetase site. The crystal structure (PDB code: 1ct9) is shown in Figure 1-4. The mutant lacks all glutaminase activity due to the replacement of th e first cysteine with alanine. This crystal structure was resolved to 2.0 .57 AS-B is comprised of 553 amino acids, with a molecular weight of 62.5 kDa, and is believed to functi on as a homodimer, i.e. two identical monomer units, with each monomer consisting of tw o domains, as represented in Figure 1-5.58 AS-B 17

PAGE 18

Figure 1-4. The AS-B monomer as depicted from the crystal structure (P DB Code: 1ct9). The glutaminase domain is colored in orange and the synthetase domain is colored in blue. Glutamine is present in the glutaminase active site and AMP is present in the synthetase active site. Both bound ligands ar e depicted in sphere representation and colored by element. possesses the N-terminal catalytic domain consiste nt of all Ntn amidotransferases, constructed from two layers of six-stranded anti-parallel -sheets that regulate the hydrolysis of glutamine (Gln) to glutamate (Glu) and ammonia.57 The Cterminal domain is the larger domain and 18

PAGE 19

Figure 1-5. The homodimer as depicted from the cr ystal structure of AS-B (PDB Code 1ct9). The N-terminal glutaminase domains are shown in orange and the C-terminal synthetase domains are shown in blue. The bound ligands are colored by element and shown in sphere representation. houses the synthetase site where aspartate (Asp ) is converted to asparagine (Asn), while consuming one molecule of ATP, yielding AMP and pyrophosphate (PPi).59 Interestingly, the Cterminal domain of AS-B does not resemb le that of the ammonia-dependent AS-A25-27,60,61 or GPATase,62-66 but rather the Cterminal domain of AS-B was found to be structurally homologous to guanosine-5-monophosphate synthetase (GMPS),67 argininosuccinate synthetase,68,69 ATP sulfurlase, 70,71 -lactam synthetase,72-74 and ThiI (4-thiouridine synthetase).75,76 Catalysis in all of these enzymes leads to the formation of PPi from ATP.59 The formation of PPi as a by-product, along with the conser ved amino acid motif SGGXDS, which is known as a pyrophosphatase loop, places these en zymes in the ATP pyrophosphatase family.59,77 19

PAGE 20

Disorder in the C-terminal domain of the AS-B cr ystal structure did not a llow two particular loop regions (Ala-250 to Leu-267 and Cys-422 to Ala-426) or the final 37 residues to be resolved, thus resulting in their absence in the crystal structure. The two unresolved loop regions were modeled onto the crystal structure using computa tional methods. The structure is depicted in Figure 1-6, as part of Dr Yun Dings dissertation,78 leaving only the final 37 residues absent. Figure 1-6. C1A AS-B model. The gl utaminase domain is depicted in orange and the synthetase domain is depicted in blue. The two loop re gions absent in the 1c t9 crystal structure have been modeled in (shown in green) and the last residue of the crystal structure (Gly-516) is shown in red in stick representation. The final 37 residues that remain absent would extend from Gly-516. Glutamine (in the glutaminase site) and AMP (in the synthetase site) are show n in sphere representation. 20

PAGE 21

The N-terminal glutaminase active site and the C-terminal synthetase active site are separated in the AS-B crystal structure by approximately 20 A solvent-inaccessible intramolecular tunnel, shown in Figure 1-7, is believed to connect th e two active sites for ammonia translocation. The putative tunnel is lined almost exclusively with hydrophobic residues (Met-120, Ile-142, Il e-143, Leu-232, Met-329, Ala-399 and Val-401), with only a few highly conserved polar residues found near the tunn el openings at the synt hetase site (Ser-346 and Glu-348) and at the glutaminase site (Arg-30) Because the majority of the identified tunnel residues are located in th e interior of the protein and associated with secondary structural Figure 1-7. The intramolecular tunn el of AS-B. This figure is a view looking down the tunnel from the synthetase site (bottom of pi cture in blue) with AMP present to the glutaminase site (top of pi cture in orange) with glutamine present. The residues believed to line the tunnel ar e shown in stick representation. 21

PAGE 22

elements, the current belief is that tunnel is always present and does not open and close through the catalytic cycle as in other amidotransferases. The presence, in the crystal structure, of water molecules at the interface of the two domains gives credence to the idea that free ammonia could potentially access the tunnel as the protein undergoes conformati onal changes during catalysis. Intramolecular ammonia translocation Ideally, it is assumed that the catalytic mech anism of AS-B would e xhibit a 1:1 ratio of Glu:Asn. The presence of an intramolecular tunnel should make it possible for every molecule of ammonia released from the hydrolysis of Gln to Gl u to be efficiently dire cted to the synthetase site for use in the conversion of Asp to Asn. Ho wever, an experimental and simulated study of the steady-state kinetics of AS-B found the catalytic ratio of Glu:Asn to be 1.8952. This demonstrates an inefficient coupling of the glutam inase and synthetase sites in ASNS that could be explained by flaws in translocation of amm onia via the intramolecular tunnel. To test the ability of exogenous ammonia to suppress the incorporation of nitrogen from glutamine hydrolysis, a competition experiment using E. coli AS-B was undertaken.56 This experiment utilized a new isotope-edited 1H NMR-based assay79 that has enhanced sensitivity for 15N NMR measurements. It is important to note that in these experiments only ammonium chloride was 15N labeled. The results of these experiments demonstrat ed that, in spite of glutamine concentrations upwards of 40mM, 15N incorporation into asparagine from 15NH3 still took place. A previous experiment had shown that 15N ammonia from solution could also attack the thioester intermediate in the glutaminase active site,80,81 even with saturating leve ls of ATP and aspartate. This observation led to the belief that exogenous ammonia can access the tunnel through conformational changes to the protein that rende r the tunnel accessible to solvent when ATP, aspartate and glutamine are all present. The mechanism by which this would occur is still unknown, but would hypothetically be the underl ying cause for the uncoupling of the 22

PAGE 23

glutaminase and synthetase activitites.52 The inefficiency of ASNS in coupling its separate active site activities makes it unique among the amidotransferases,82,83 which all demonstrate tightlycoupled glutaminase and synthetase activities. Interestingly, when aspartate is not present, the glutaminase activity of AS-B is enhanced almost 2-fold when ATP, AMPPNP (a non-hydrolysable ATP analog84), or AMP and PPi are bound in the synthetase site.31 As a control, it was shown that the glutaminase activity was not stimulated if only one of AMP or PPi was present in the assay.31 It was also proven in the same set of kinetic experiments that the highly cons erved Arg-30 plays an important role in the glutaminase stimulation. When Arg-30 was mutated to either alanine or lysine, the glutaminase activity was no longer stimulated.31 These results suggest that similar to GPATase,62 GFAT85,86 and GltS,40,87 AS-B possess inter-domain interactions that lead to conformational changes for the signaling of active site occupancy. What is still unc lear is whether these signals are transferred within an AS-B monomer or betw een separate monomers (glutaminase with synthetase) of the heterodimer unit as was depicted in Figure 1-5. The Arg-30 likely plays a prominent role in a hydrogen bonding network57 that orients the side chain of Asn-74, which is necessary for stabilization of the tetrahedral intermediate created during the glutam inase reaction via the oxyanion hole.88 This low-level active site communicati on may have evolved due to the ability of asparagine to compete for the glutaminase site with glutamine, hence lowering the detrimental cellular effects of the uniquely elev ated glutaminase activity of ASNS. Research Objectives The ultimate goal of the work on asparagine synthetase is to develop potent viable inhibitors of human form of the enzyme. The de velopment of viable inhibitors for ASNS could aid in the treatment of ALL, as well as be utilize d in the treatment of other solid tumors. As was discussed previously, inhibition of ASNS in drug-resistant ALL could provide those patients 23

PAGE 24

with a second opportunity for a successful treatmen t that is currently unavailable. Inhibitors of ASNS could also potentially have applications in the treatment of solid tumors in general.10 However, to date the structure of human aspara gine synthetase remains unresolved. The structure of the C1A mutant of E. coli AS-B57 is still the best available stru cture of ASNS and is believed to be similar to the human form of the enzyme. This structure still possess certain flaws, unresolved regions, and does not ha ve the synthetase reaction intermediate or transition state bound.Knowing the structure-function re lationships of the intermediate transition state or any of the known inhibitors will aid the progression of better inhibitor design. A better AS-B or human ASNS structure or model would be very useful in better understanding the specific interactions that are involved in the transiti on state stabilization and could be mimicked more directly in the rational design of inhibitors. The over-riding goal of this work is to furt her develop a rough mode l of AS-B that was originally undertaken by Dr. Yun Ding as part of her dissertation work.78 In order to refine an AS-B model, with the AspAMP synthetase intermediate bound to the synthetase active site, in the CHARMM modeling package,89,90 a new set of parameters must be developed. Another goal of this work was the testing and refinement of parameters for a molecular docking package. The availability of a refined docking algorithm w ould allow for the virtual screening of entire libraries of compounds with a model of AS-B. Ha ving the capability to virtually screen would aid in the identification, desi gn and direction for new or untested inhibitors of ASNS. The specific goals for this research were to (i) develop the necessary parameters for the synthetase reaction intermediate, AspAMP, for its inclusion in CHARMM; (ii) develop and refine a model of AS-B with glut amine in the glutaminase site and AspAMP, PPi and Mg2+ in the synthetase site to understand the stru cture-function relationship between AS-B and 24

PAGE 25

AspAMP; (iii) to explore the development of a new docking potential energy function that can be used in future virtual screening experiments. 25

PAGE 26

CHAPTER 2 PARAMETERIZATION OF A REACTION INTERMEDIATE OF ASPARAGINE SYNTHETASE-B FOR USE IN CHARMM Introduction The computational modeling of large-scale bi ological systems continues to grow in popularity, as the methods used to gain insight into these systems have developed. The ability of computer simulations to offer a greater understanding of the fundamental properties governing the molecules within these large systems has brought computational biochemistry wider acceptance. High-level quantum mechanical met hods will always be the gold standard for understanding chemical properties. Unfortunately, ev en with the dramatic increases in computer speeds and parallel programming algorithms, the current technologies still restrict ab initio and density functional theo ry calculations to well under 100 h eavy atoms. This limitation makes using pure quantum mechanical methods impossible when studying the likes of enzymatic reactions and protein folding. Instead, molecular dynamics (MD)91-93 and Monte Carlo94,95 simulations, which are governed by empirical mol ecular mechanics force fields, are used. These methods have the distinct advantage of computational efficiency, but rigorous parameterization of the potentials is crucial for realistic result s to be expected. The parameterization of the empirical potential is accomplished via small mode l compounds so that the parameters can then be transferred to larger molecules or proteins. In the past few decades, a number of empirical potential energy force fields have been extensively parameterized and tested for accuracy.90,96-103 The CHARMM force field89,90,104 was developed for use with large biomolecules, na mely proteins, lipids and nucleic acids. The continued parameterization of CHARMM is always done with the aim of improving accuracy without increasing the complexity of the potentia l energy function (that can be seen in Equation (2-1) below). By keeping the potential energy func tion static and optimizing the parameters or 26

PAGE 27

adding to the parameters, CHARMM has improve d accuracy without major increases to the computational time required to perform the calculations. CHARMM currently employs an allatom parameter approach for proteins,104 lipids,105 nucleic acids,106 carbohydrates,107 and more that has be shown to produce quality result s in a variety of di fferent simulations.108,109 One issue that arises in any parameter set is the lack of transferability of parame ters to any unique atom type, bond, angle, or dihedral. This is due to the basic philosophy of parameterization that seeks to maximize the reproducibility of target data for a particular set of compounds. In order to overcome this deficiency, substrate-specific para meters are continually developed by users for the direct application of their system, as is the case here. Figure 2-1. The structures of the intermediate, AspAMP, and of the tetrahedral transition state for the ASNS catalyzed reaction. The intermediate and transition state have been identified in the bioc onversion of aspartate into asparagine catalyzed by ASNS. The intermediate, AspAMP, and the tetrahedral transition state, shown in Figure 2-1, have been successf ully used as models for inhibitor design.23 Unfortunately, no structural information is available, crystallographic or otherwise, that would provide much needed information on the stru cture-function relationship between either AspAMP with ASNS or the transition state with ASNS. In order to develop a reliable model of the enzyme that contains either of these speci es, molecular dynamics must be used in the 27

PAGE 28

investigation. However, for this type of simulation to be performed in CHARMM, the uncommon bonds, angles and dihedrals within the AspAMP must first have parameters developed that can be integrat ed into the existing set of parameters that CHARMM currently employs. This chapter will describe the methods utilized to develop the parameters for AspAMP that heretofore were absent in th e parameter set used by CHARMM. Without the development of the aforementioned parameters, the investigation of the structure-function relationship of the AspAMP intermediate with ASNS would not be possible using the molecular mechanical methods within the CHARMM package. Methods Parameterization Methodology The parameters for the AspAMP intermediate were developed using the methods and philosophies adopted in the development of the CHARMM simulation program,90 wherever possible. The empirical energy function called in CHARMM is comprised of both intramolecular and intermolecular terms of the following form: nonbonded ij ji ij ij ij ij inpropers inp dihedrals angle UB UB bonds br qq rr K n K KSSKbbKRU1 6 12 2 0 2 0 2 0 2 04 cos1 (2-1) where Kb, KUB, K, K, and Kimp are the force constants for the bond, Urey-Bradley, angle, dihedral angle, and improper dihedral angl e, respectively. The bond length, Urey-Bradley 1,3distance, bond angle, dihedral angle, a nd improper torsion angle are represented by b, S, and respectively, with the subscript zero representing the equilibrium values for the individual terms. The Urey-Bradley term is included to im prove non-bonded 1,3-interac tions that tend to be repulsive. The last two terms strictly cover the bulk of the nonbonded interactions. In the 28

PAGE 29

Lennard-Jones 6 12 terms, is the radius when the Lennard-Jones energy is zero, is the minimum Lennard-Jones well depth, 21/6 is distance at the Lennard-Jones minimum, and rij is just the distance between atoms i and j The last term is the Coul omb contribution to the total energy. Here, qi and qj are the partial atomic charges on atoms i and j respectively, and 1 is the effective dielectric constant. The vector of the coordinates of the atoms, R when given, makes the distances and angles that are necessary to evaluate simple to determine. The LorentzBerthelodt combination rules are used in C HARMM to determine the Lennard-Jones parameters between pairs of different atoms. This is done for the calculation of ij values that are based on the geometric mean of i and j. The Lennard-Jones minimum, Rmin,ij, is calculated similarly as it is based on the arithmetic mean between Rmin,i and Rmin,j. For the current version of the CHARMM protein parameter set, CHARMM22,104 the effective dielectric constant, 1, is set at 1 in order to attain a balanced pa rameterization with regard to the electrostatic contributions to the energetics. If a high dielectric constant solvent is necessary, neutralized charged groups can be added to the system to add some of the effects of shielding.110 The nonbonded interactions are determined for all atoms that are separated by at least three bonds and no scaling is done on these interactions in CHARMM except in specific cases where there is scaling of the 1 4 LennardJones term (such as oxygen atoms, aliphatic ca rbons, and amide nitrogens). There is no term included for hydrogen bonding, as it has been dem onstrated that the Coulomb and Lennard-Jones terms can accurately model those interactions.111,112 CHARMM uses the TIP3P model113 for the representation of water in all calculations for co nsistency of the protein and solvent interactions. RU 29

PAGE 30

Parameterization Strategy The work presented here was focused on the development of parameters for the ASNS reaction intermediate, AspAMP. A model of E. coli AS-B with AspAMP bound in the synthetase site would be a great asset in expl oring the structure-func tion relationship of the enzyme with the intermediate. Determining what enzyme/intermediate interactions are important and resolving what residues play key roles in binding the intermediate and catalyzing the conversion of aspartate to as paragine could help not only to guide future mutagenesis experiments, but also would help with th e rational design of future inhibitors. The AspAMP intermediate is composed of one molecule of AMP and one free as partate. Parameters for both of those molecules exist in CHARMM. However, as can be seen in Figure 2-2, the area where these two molecules are joined is where CHARMM para meters must be developed for the unique bonds, angles, and dihedrals of AspAMP. Rather than using the entire AspAMP (with 49 atoms) in ab inito energy calculations to determine a tw o dihedral potential energy surface (PES) Figure 2-2. AspAMP and the model complex with uniqu e atom labels. This figure shows the AspAMP reaction intermediate and the model complex used to develop the parameters necessary to integrate AspAMP into CHARMM simulations. The model complex is based on the core region within the blue box on AspAMP. The hydrogen atoms in orange boxes on the model comple x were added to the model and are not a part AspAMP. 30

PAGE 31

and in other necessary ab initio calculations, a model complex (w ith only 15 atoms) that contains all of the critical atoms and connections needed to represent AspAMP accurately was used. The parameterization of AspAMP was particularly challe nging because there is limited crystallographic data on which AspAMP can be parameterized. Ideally, the necessary parameters would be developed with the goal of reproducing the crystall ographic geometries for the molecule of interest. Results and Discussion Model Selection and Initial Parameters The majority of the parameters for the AspAMP intermediate were transferred from the available parameters in the CHARMM package for AMP in the nucleic acid parameter set106 and the parameters for aspartate in the protein parameter set.104 The small set of bonds, angles and dihedrals that do not have existing parameters in CHARMM are related to the unique connection made between the carboxylate group on the side chain of aspartate a nd the phosphate group of AMP. A list of the bonds, angles and dihedrals that required parameters is summarized in Table 2-1. The model complex shown in Figure 2-2 wa s chosen to represent the unparameterized region within the AspAMP for computational efficiency when using ab initio methods for calculating the PES that was used to describe the critical dihedrals. The model was also used in the ab initio calculations that determined the vibrationa l spectra and partial atomic charges. This comprised the remainder of the target data fo r the parameterization. Th e CB and C5 carbons of AspAMP were chosen as the end caps for th e model and additional hydrogens were added to each carbon to create methyl groups. This common practice of enclosing the region of interest between methyl groups is done to describe more accurately the partial charges on the critical 31

PAGE 32

atoms, as partial charges on the methyl groups are constant and prevent long-range charge dispersion. Table 2-1. The bonds, angles and dihedrals to be parameterized with initial parameters. Structure Bonds Kb b0 CC-ON2 650.0a 1.316 Structure Angles K 0 KUB S0 P-ON2-CC 20.0b 128.2 35.0b 2.33b ON2-CC-CT2 70.0c 117.6 ON2-CC-OC 98.0d 120.3 Structure Dihedrals K N 4.1e 1 0.0 ON2-P-ON2-CC 3.5e 3 0.0 P-ON2-CC-CT2 11.2e 2 180.0 P-ON2-CC-OC 11.2f 2 180.0 C5' O5' O1A PA O2A O3A CG CB OD1 HB2 HB1 HB3 HC5' H5' H5'' ON3-P-ON2-CC 0.1g 3 180.0 32

PAGE 33

a Value taken from O-CC in acetamide; b Value taken from P-ON2-CN9 in methyl diphosphate; c Value taken from ON2-CC-CT2 in CoA; d Value taken from ON2-CC-OC in CoA; e Calculated from HF/6-31+G( d) potential energy surface (O5-PAO3A-CG vs PA-O3A-CG-CB); f Estimated to be same as P-ON2-CC-CT2; g Taken from ON3-P-ON2-CN7 in dimethylphosophate. The b0 and 0 were set to the distance and angles obtain from the ab initio minimized structure. Atom Types and Geometric Parameters The atom types were defined so that no ne w atom types would have to be introduced, without compromising any expect ed accuracy in reproducing the ab initio results. The only atom that would have greatly benefited from introduc ing a new atom type would have been the Table 2-2. Atomic charges and atom types for AspAMP and the model complex. AspAMP AspAMP Model Atom Type Charge Atom Type Charge Atom Type Charge N1 NN3A -0.74 H4 HN7 0.09 C5 CN8B 0.24 C2 CN4 0.50 O4 ON6B -0.50 H5 HN8 0.02 H2 HN3 0.13 C5 CN8B 0.23 H5 HN8 -0.01 N3 NN3A -0.75 H5 HN8 0.02 HC5 HN8 -0.01 C4 CN5 0.43 H5 HN8 -0.01 O5 ON2 -0.55 C5 CN5 0.28 O5 ON2 -0.55 PA P 1.52 C6 CN2 0.46 PA P 1.52 O1A ON3 -0.88 N6 NN1 -0.77 O1A ON3 -0.88 O2A ON3 -0.89 H61 HN1 0.38 O2A ON3 -0.89 O3A ON2 -0.61 H62 HN1 0.38 O3A ON2 -0.61 CG CC 1.01 N7 NN4 -0.71 CG CC 1.01 OD1 OC -0.74 C8 CN4 0.34 OD1 OC -0.74 CB CT2 -0.40 H8 HN3 0.12 CB CT2 -0.30 HB1 HA 0.10 N9 NN2 -0.05 HB1 HA 0.10 HB2 HA 0.10 C1 CN7B 0.16 HB2 HA 0.10 HB3 HA 0.10 H1 HN7 0.09 CA CT1 0.07 C2 CN7B 0.14 HA HB 0.09 H2 HN7 0.09 N NH3 -0.47 O2 ON5 -0.66 HN1 HC 0.31 H2 HN5 0.43 HN2 HC 0.25 C3 CN7 0.14 HN3 HC 0.26 H3 HN7 0.09 C CC 0.51 O3 ON5 -0.66 O OC -0.51 H3T HN5 0.43 OC OC -0.51 C4 CN7 0.16 The atom is the unique name given to each at om, while the type is the atom type recognized by CHARMM. 33

PAGE 34

bridging oxygen between the aspartate and AMP. However, for simplicity and consistency with the existing CHARMM parameters, the most similar atom type, a bridging oxygen in a phosphate group, was used. A full list of the atom types selected for the model complex as well as the AspAMP is summarized in Table 2-2. As mentioned previously, because there is no crystallographic data available for AspAMP, the equilibrium bond lengths and angles were set in accordance with the optimized geometry from an ab initio calculation. Multiple geometry op timizations were performed on both the model and the full AspAMP intermediate at the Hartre e-Fock (HF) level with a 6-31+G( d) basis set (i.e. HF/6-31+G( d) level). This and all other ab initio calculations were performed using the GAUSSIAN 03 program.114 HF was chosen as the model chemistry in an effort to develop parameters consistent with the parameteriza tion approach taken with CHARMM and in the development of parameters for other similar molecules.115 For the basis set, a diffuse function was added because of the negative charge carried by AspAMP and the model complex, and the single polarization function was added to help account for the presence of phosphorus. The geometric data is presented in Table 2-3. Partial Atomic Charges The partial charges calculated for the mode l complex were then transferred to the corresponding atoms in AspAMP. This procedure represents the largest deviation from the parameterization methodology generally employed in the development of CHARMM. Ordinarily, the partial charges in CHARMM are calculated by reproducing ab initio interaction energies and geometries betw een model compounds and water.111,112 The approach taken instead was the calculation of electro static potential charges according to the CHelpG scheme116 at the HF/6-31+G( d) level for the model complex in GAU SSIAN 03. The CHelpG scheme was 34

PAGE 35

Table 2-3. Model complex, ab initio and CHARMM geometry data. Bond Internal Coordinate CHARMM Ab Initio b1 PA-O1A 1.480 1.472 b2 PA-O3A 1.620 1.687 b3 CG-O3A 1.325 1.315 b4 OD1-CG 1.259 1.198 b5 CB-CG 1.529 1.509 b6 HB1-CB 1.111 1.083 b7 HB2-CB 1.110 1.082 b8 HB3-CB 1.110 1.082 b9 O2A-PA 1.481 1.468 b10 O5-PA 1.586 1.606 b11 C5-O5 1.431 1.401 b12 H5-C5 1.113 1.084 b13 H5-C5 1.113 1.084 b14 HC5-C5 1.111 1.082 b1b9 b1+b9/2.0 1.481 1.470 b2b10 b2+b10/2.0 1.603 1.647 Angle Internal Coordinate CHARMM Ab Initio a1 O3A-PA-O1A 109.6 106.3 a2 CG-O3A-PA 175.7 128.2 a3 OD1-CG-O3A 122.6 120.4 a4 CB-CG-O3A 114.8 117.6 a5 HB1-CB-CG 110.0 109.7 a6 HB2-CB-CG 109.5 109.9 a7 HB3-CB-CG 109.6 108.8 a8 O3A-PA-O2A 107.6 107.8 a9 O3A-PA-O5 103.5 96.0 a10 PA-O5-C5 116.6 120.5 a11 O5-C5-H5 111.7 110.8 a12 O5-C5-H5 111.2 107.1 a13 O5-C5-HC5 108.7 111.0 a1a8 a1+a8/2.0 108.6 107.0 Dihedral Internal Coordinate CHARMM Ab Initio d1 CG-O3A-PA-O1A -49.3 65.4 d2 OD1-CG-O3A-PA 133.1 -175.7 d3 CB-CG-O3A-PA -18.5 4.7 d4 HB1-CB-CG-O3A 177.7 -60.4 d5 HB2-CB-CG-O3A -62.3 57.3 d6 HB3-CB-CG-O3A 57.6 178.6 d7 O2A-PA-O3A-CG 78.3 -69.2 d8 O5-PA-O3A-CG -168.7 178.7 d9 C5-O5-PA-O1A -58.1 37.9 d10 H5-C5-O5-PA 55.7 66.4 d11 H5-C5-O5-PA -67.4 -174.8 d12 HC5-C5-O5-PA 174.4 -55.2 Bonds are measured in angstroms, while angles and dihedrals are measured in degrees. 35

PAGE 36

selected to calculate the partial charges on th e model complex because of its invariance to internal and external rotation. This makes CHelpG a good f it for charge calculation, as the overall focus of this work is essent ially a conformational-based study of the AspAMP intermediate within E. coli AS-B. The calculated partial charges were rounded to the nearest hundredth and yielded the appropria te total complex charge of -1. These charges were then integrated into the AspAMP and the charges on the extra hy drogens in the model were added to their respective carbons (HC5 C5 and HB3 CB). A full list of the charges on both the model complex and the AspAMP can be seen in Table 2-2. Force Constants and Dihedral Rotation The force constants for the bonds and angles were initially taken from already existing parameters in CHARMM. The idea was to use parameters for similar bonds and angles that had previously been rigorously para meterized. A full list of those parameters with the corresponding origins for each parameter is listed above in Ta ble 2-1. The (initial) dihedral parameters were derived from a variety of data, including the PES that was calculated using GAUSSIAN 03 at the HF/6-31+G( d) level, already existing dihedral parame ters in CHARMM, and from a dihedral model developed in Microsoft Excel. The periodicities, n, were the first parameters assigned for the dihedrals. The periodicities were assigned to each dihedral based on the ab initio PES, the geometries of the atom at the center of the dihedrals, and the Excel model. The dihedral force constants, K, were obtained by both calculations (from the PES) and through borrowing parameters from similar dihedrals already parameterized in CHARMM. The phase angles, were assigned from the PES, the Excel model, and Newman projections about the dihedrals. The parameterization of the dihedral began with the calculation of a 2-dimensional potential energy surface. A surface was compiled by a series of HF/6-31+G( d) level ab initio 36

PAGE 37

calculations, performed in GAUSSIAN 03, in wh ich the two primary dihedrals (CB-CG-O3APA(1) vs. CG-O3A-PA-O5(2)) were constrained while the rest of the structure was allowed to relax to the lowest energy conformation allowed by the specific set of dihedrals. To ensure adequate coverage, both dihe drals were increased in 10 increments from 0 to 360. This created a 2-dimensional energy surface comprised of 1,296 minimized model complex structures. The resulting ab initio PES is shown in Figure 2-3. As can be seen in the figure, the ab initio surface is symmetrical and has three minima. The global minimum seen at a 1_2 of 0_180 was set as zero. Two additional minima are seen at a 1_2 of 0 _70 and at a 1_2 of 170_60 with energies of 2.99 kcal/mol and 3.42 k cal/mol above the zero, respectively. The overall maxima for the surface is seen at a 1_2 of 120 _150 and has an energy of 10.40 kcal/mol above the zero. The ab initio surface was used to estimate the initial periodicity and force constants for the two primary dihedrals shown in Table 2-1. One-dimensional sections were originally taken from the surface to estimat e the periodicity of each dihedral. This approach yielded an initial guess of n = 2 for 1 (CB-CG-O3A-PA) and n = 1 and 3 for 2 (CG-O3A-PAO5). Using the equation 2 2 2 2 1 23cos1 cos11802cos122 21 11 k k kE (2-2) and three different combinations of energies with the associated 1_2 ( E ,1,2) from the ab initio surface, the force constants k11, k21, and k22 were estimated at 11.2, 4.1, and 3.5, respectively. As seen in Equation 2-2, the phase angles, for 1 and 2 were set at 180 and 0, respectively, from the ab initio surface. The initial paramete rs for the two supplementary dihedrals also required by CHARMM (PA-O3A-CG-OD1 and O2A-PA-O3A-CG) were estimated. The initial parameters for the supplementary dihedral PA-O3A-CG-OD1 was 37

PAGE 38

estimated to be the same as the CB-CG-O3A -PA dihedral parameters, while the initial parameters for the supplementary dihedral O2A-PA-O3A-CG were taken from ON3-P-ON2CN7 in dimethylphosphate. Figure 2-3. The 2-dimensional ab initio energy surface for dihedral angle CB-CG-O3A-PA versus dihedral angle CG-O3A-PA-O5 of the model complex. The global minimum energy (kcal/mol) was set to zero with a ll other energies offset equivalently. The dihedral parameterization began by dete rmining a baseline of all the forces contributing to the energy surface in CHARMM minus the contribution to the energy made by the dihedrals. The resulting surface can be seen in Figure 2-4. The initial dihedral parameters 38

PAGE 39

Figure 2-4. The 2-dimensional CHARMM energy surface for dihedral angle CB-CG-O3A-PA versus dihedral angle CG-O3A-PA-O5 of the model complex without energy contributions from the dihedrals in Table 2-1. In this figure and in the subsequent dihedral surfaces generated by CHARMM (Figure 2-4, Figure 2-6 and Figure 2-7), there are regions of solid blue where CH ARMM was unable to minimize the structure with the specific combination of dihedrals resulting in inco rrect energies that had to be set to zero for the remaining su rface to be appropr iately depicted. were then added to the energy calculation. Upon the addition of the initial parameters, it was clear that adjustments would be necessary to achieve better agreement of the 2-dimensional energy surfaces calculat ed by CHARMM with the ab initio surface. The first step was adding a second term to help describe the PA-O3A-CG-CB dihedral. Periodicities of both 1 and 3 were 39

PAGE 40

tested and only the term with n = 1 seemed to improve the empirical surface calculated by CHARMM. The adjustments of the force consta nt for this term eventually yielded a K value of 6.0 as the best fit. The two supplementary dihedr als were also very closely inspected. Newman projections looking down the two separate dihedrals of the lowest energy conformer of the model complex led to the alteration of the initial phase angles for the dihedral s in question. The force constants for the two supplementary dihedrals depi cted in Figure 2-5 were also changed. As was done with the term added to the PA-O3A-CG-CB dihedral, the force constants for these two dihedrals were adjusted to obtain the best agreement. This process produced a K for the O2APA-O3A-CG dihedral of 1.0 and a K for the PA-O3A-CG-OD1 dihedral of 0.3. The periodicities for these two dihedrals were estimate d in the belief that the periodicity of the PAO3A-CG-OD1 dihedral would mimic the primary periodicity of the main dihedral, PA-O3A-CGCB, and that the periodicity for the O2A-PA-O3A-CG dihedral would be similar to the ON3-PON2-CN7 dihedral in dimethylphosphate. The resulting PES calculated by CHARMM with all OD1-CG-O3A-PA CG-O3A-PA-O2A Figure 2-5. Newman projections for the di hedrals O2A-PA-O3A-CG and PA-O3A-CG-OD1. The first dihedral is looking down the O2 A-PA-O3A-CG dihedral with truncations made at O5 and CG for clar ity. The projection shows a 60 dihedral. The second dihedral is looking down the PA-O3A-CG-OD1 dihedral with truncations made at PA and CB for clarity. The projection shows a 0 dihedral. 40

PAGE 41

Figure 2-6. The 2-dimensional CHARMM energy surface for dihedral angle CB-CG-O3A-PA versus dihedral angle CG-O3A-PA-O5 of the model complex with energy contributions from the dihedrals. of the dihedral adjustments can be seen in Figure 2-6. As illustrated in Figure 2-6, this combination of parameters produced an ener gy surface in CHARMM that has a global minimum at the same 1_2 combination of 0 _180 and two other minima in the same general areas as the on the ab initio surface. However, the major deficiency of this energy surface is a local minimum in the area around a 1_2 of 120 _150 where the global maximum is located on the ab initio surface. This was considered a significant problem as the primary objective in reproducing energy surfaces in a parameterization is to match the overall shape of the energy surface. 41

PAGE 42

Reproducing the energy barriers within the surface is considered a secondary objective, as it is an extremely challenging feat to accomplish. Theref ore, the glaring omission of a maximum where the global maximum should be, prompted continui ng the parameterization efforts to achieve a better fit. Because the dihedral parameters were believed to be close to their optimal values, a second look was taken at the bond and angle parameters that were estimated from already existing parameters within CHARMM. Pavelites, et al determined that a negative force constant on a Urey-Bradley term added to the bridging bond in methyl diphosphate was necessary for CHARMM to be able to reproduce th e dihedral rotational energy surface.115 After reexamining the PA-O3A-CG bond angle, the initial estimates based on the P-ON2-CN9 angle in the same methyl diphosphate were replaced with the parameters for the P-ON2-P angle. This included the negative Urey-Bradley force constant that was critical in reproducing their ab initio energy surface. This change left a K of 15.0 with the same 0 as before, a KUB of -40.0, and a S0 of 2.8 as the parameters on the PA-O3A-C G angle of the model complex and AspAMP. Recalculating the CHARMM 2-dimensional energy surface while only changing the PAO3A-CG angle parameters yielde d the energy surface shown in Fi gure 2-7. The major change in the PES calculated by CHARMM between Figure 2-6 and Figure 2-7, is that now a maximum is in the area around a 1_2 of 120 _150. This is a critical shift in improving the overall fit of CHARMM PES with the equivalent ab initio surface. There is a loss of the third minimum at a 1_2 of 170 _60 in the surface calculated with the ne w set of parameters; however, this was believed to have been an acceptable tradeoff for having a maximum near the corresponding 1_2 of 120 _150 as seen in the ab initio surface in Figure 2-3. This final surface does retain the global minimum at a 1_2 of 0 _180 with a local minimum at a 1_2 of 0 _70. With this 42

PAGE 43

Figure 2-7. The 2-dimensional CHARMM energy surface for dihedral angle CB-CG-O3A-PA versus dihedral angle CG-O3A-PA-O5 of the model complex using the final set of developed parameters. This graph used the exact same parameters as Figure 2-6 with the exception of the changes made to the PA-O3A-CG angle parameters as described in the text. surface believed to be the best estimate of the ab initio surface to this point, further refinement to the force constants could be made by comparing the vibrational spectrum for the model complex as calculated in CHARMM with the target vibrational spectrum that was calculated for the model complex. 43

PAGE 44

Vibrational Spectra for the Model Complex The target vibrational frequency data for pa rameterizing the model complex was calculated via GAUSSIAN 03, and was calcu lated at the HF/6-31+G( d) level of theory to maintain consistency with all previous ab initio calculations performed for the model complex. For comparison of the ab initio vibrational frequencies with those calculated by CHARMM, a scaling factor was used to reduce the HF re sults, which are known to be systematically overestimated.117 A scale factor of 0.897 was used here in accordance with the findings of Scott and Radom for HF/6-31+G( d) level results.117 Table 2-4 shows the 39 vibrational frequencies calculated by HF and then scaled, as well as the vibrational fre quencies calculated by CHARMM. Generally, differences of 100 cm-1 or less are considered to be in relative agreement. As seen in Table 2-4, only four CH ARMM frequencies are more than 100 cm-1 different from the scaled ab initio frequencies. The overall rms deviation is 35.08 cm-1 with the average percent difference being 9.6%. The freque ncies below approximately 300 cm-1 are the most biologically relevant modes that normally occur in an MD simulation. These modes do minate the large-scale motions. The first eight frequenc ies, which comprise this lo w frequency region, all show good agreement with the ab initio results. The next ten vibra tional frequencies extend from approximately 300 cm-1 to approximately 1,000 cm-1, and also exhibit good overall fit. There are two frequencies (modes 15 and 17) that show a difference greater than 100 cm-1 from the corresponding ab initio modes. Although this difference is not ideal, because they first appear in the second half of the modes where the less biologically important freque ncies occur, the large difference is more acceptable. Again, only two outlie rs are seen in the remaining span of modes that stretch from 1,000 cm-1 to 3,000 cm-1. In all, only four fre quencies calculated by CHARMM are separated by more than 100 cm-1 from the corresponding frequencies calculated by HF. The good agreement in vibrational spectra demons trates a good level of parameterization. 44

PAGE 45

Table 2-4. Vibrational frequencies for the model complex as calculated by HF and CHARMM. Ab initio CHARMM Mode Frequencya (cm-1) Frequency (cm-1) ABS( AI CHARMM) 1 51.2 44.6 6.6 2 59.2 89.6 30.4 3 86.4 109.3 22.9 4 111.5 185.3 73.8 5 153.1 189.6 36.5 6 172.2 229.5 57.3 7 214.2 241.2 27.0 8 306.3 323.8 17.5 9 350.2 360.5 10.3 10 365.4 430.6 65.2 11 428.9 461.1 32.2 12 478.2 501.6 23.4 13 523.2 510.1 13.1 14 546.0 564.9 18.9 15 589.2 723.4 134.2 16 703.6 778.5 74.9 17 752.3 853.6 101.3 18 948.0 930.9 17.1 19 1018.5 1023.4 4.9 20 1063.5 1035.6 27.9 21 1071.1 1115.2 44.1 22 1099.5 1134.1 34.6 23 1163.4 1145.3 18.1 24 1190.4 1164.1 26.3 25 1259.9 1275.6 15.7 26 1289.3 1399.0 109.7 27 1403.1 1426.9 23.8 28 1452.3 1436.1 16.2 29 1453.9 1438.4 15.5 30 1455.0 1456.7 1.7 31 1473.8 1482.1 8.3 32 1487.7 1622.4 134.7 33 1743.9 1707.8 36.1 34 2880.6 2798.6 82.0 35 2897.8 2853.7 44.1 36 2934.1 2860.8 73.3 37 2952.6 2863.1 89.5 38 2964.9 2918.1 46.8 39 2975.6 2919.6 56.0 a Ab initio frequencies have been scaled by 0.897 as outlined in the text. An attempt to optimize the force constants fu rther using the vibrati onal spectra was made using the Automated Frequency Ma tching Method (AFMM) program.118 AFMM uses a Monte 45

PAGE 46

Carlo-like algorithm that adjusts parameters iteratively until the optimal fit of the vibrational spectra of CHARMM with the targ et vibrational spectra from ab initio methods is obtained.118 After multiple attempts to use AFMM to optimize the force constants, without a single example of improvement in the CHARMM energy surface, this method was abandoned and the previously reached set of parameters wa s kept as the final parameter set. Conclusions The parameterization of the model complex, and hence the AspAMP intermediate, centered primarily on the two unique dihedrals in this structure. The initial parameters for the one bond and three angles were taken from sim ilar structures. The equilibrium bond length and angles were taken from the globally minimum structure found in the HF/6-31+G( d) calculated PES (Figure 2-3). As was discus sed earlier, the partial atomic charges for the model complex were determined via the CHelpG scheme and are reported in Table 2-2. This leaves the absent dihedral parameters, which were major energy cont ributors to the complex, as the primary focus of the parameterization efforts. The final set of parameters for the mode l complex provides good agreement with the lowest energy conformation found with HF. Th e overall rms difference for the bonds is 0.017 with respect to the ab initio results, while the angles have an overall rms difference of 12.2 In general, CHARMM found slightly longer bonds than were seen from the ab initio structure with the exception of the two bridging oxygens (O5 and O3A). In both of these cases, the parameter set produced slightly shorter bonds than were seen in the ab initio structure. The largest absolute deviation of CHARMM with the ab initio result was only about 0.07 The rms difference in the angles is signifi cantly large at 12.2; however, this large rms can be attributed to a single significant outlier. As signified by the bolded va lue in Table 2-3, angle 2 (CG-O3A-PA) is found 46

PAGE 47

to be 175.7 on a bent oxygen rather than the value that is expected around 120 If this data point is removed, the rms difference for the angles drops significantly to 2.1 This means that if angle 2 were excluded, the larges t single deviation is only 7.5 Dihedral 2 (OD1-CG-O3A-PA) is also significantly off from its target, however, this large deviat ion can again be attributed to the previous angle (CG-O3A-PA) that was off. Overall, the lowest energy conformation produced for the model complex by CHARMM using the final parameter set demonstrates good agreement with the lowest energy conformer found at the HF/6-31+G( d ) level. Overall, the parameters developed here, in accordance with the CHARMM empirical force, do an adequate job of representing the confor mational properties of the model complex as established by ab initio means. The final set of parameters can be seen below in Table 2-5. 47

PAGE 48

48 Table 2-5. Final parameters for the model complex. Structure Bonds Kb b0 CC-ON2 650.0 1.316 Structure Angles K 0 KUB S0 P-ON2-CC 15.0 128.2 -40.0 2.8 ON2-CC-CT2 70.0 117.6 ON2-CC-OC 98.0 120.3 Structure Dihedrals K n 4.1 1 0.0 ON2-P-ON2-CC 3.5 3 0.0 6.0 1 180.0 P-ON2-CC-CT2 11.2 2 180.0 P-ON2-CC-OC 0.3 2 0.0 C5' O5' O1A PA O2A O3A CG CB OD1 HB2 HB1 HB3 HC5' H5' H5'' ON3-P-ON2-CC 1.0 3 60.0

PAGE 49

CHAPTER 3 DEVELOPMENT OF A COMPUTATIONAL MODEL OF ASPARAGINE SYNTHETASE-B Introduction The development of a computa tional model of a protein can offer valuable insights into both the structural and thermodynamic properties of a specific system. Traditionally, the goal of any protein structure investigati on, whether computational or experi mental, has been to identify the native conformation of the protein. This is believed to be the global minimum in potential energy of the system and the state that is bi ologically relevant. However, many enzymatic systems undergo conformational changes as the reaction progresses. Thus, the term global minimum can actually represent multiple protei n conformations depending on what substrates or ligands are bound. Identifying the global minimum of an enzymatic system at different stages of the reaction provides insight in to the conformational changes the system experiences, as well as the critical active site interactions at each st age. These important activ e site interactions can then be exploited in the rational design of targeted enzyme inhibitors. Various techniques have been developed for the purpose of conformational sampling with the goal of finding the global minimum conformation.119,120 Some of the approaches taken for structural prediction include ab initio simulated annealing,121 molecular dynamics-cluster analysis (MD-CA),122 relaxed annealing,123 pressure annealing,124 Monte Carlo with simulated annealing,125 Monte Carlo with molecular dynamics,126 molecular dynamics with simulated annealing,119 molecular dynamics with explicit water,127 multi-conformation simulated annealing pseudo-crystallographic re finement (MCSA-PCR),128 high temperature molecular dynamics,129 and replica exchange simulations130 to name a few. The common thr ead that appears in most of the above approaches in one way or anothe r is the simulated annealing (SA) method.131,132 49

PAGE 50

Asparagine synthetase is an exceptional candidate for structural optimization via simulated annealing. AS has proven to be a difficult enzyme to crystallize; to date, only a single crystal structure exists of a glutamine-depende nt prokaryotic form of AS (AS-B).57 While extremely valuable, as the only structural insight into glutamine-dependent AS, this structure still has many shortcomings. If modeled correctly, the improve the structure would provide a more nearly complete picture of the enzyme during catal ysis. Thorough understanding of the synthetase active site is of particular interest, because the roles of explicit residues in volved in catalysis will provide a detailed platform for rational design of inhibitors. Futhermore, the model could also be used to help understand experimental result s from site-directed mutagenesis studies. Unfortunately, the crystal structure of AS-B is incomplete as it does not have the coordinates for the residues of two small loop regions embe dded in the sequence, nor does it have the coordinates for the final 37 resi dues. The AS-B crystal structur e also only offers a limited perspective of the critical synthetase active site as it only has AMP, an eventual product, bound.57 The work presented here will expand on the initial efforts of Dr. Yun Ding towards the development of a model of AS-B.78 Two models of AS-B will be optimized. The first model will include Gln bound in the glutaminas e active site and Asp, ATP and Mg2+ bound in the synthetase active site of AS-B. This model represents AS-B in its presumed initial state, prior to the initiation of catalysis. The second model will show AS-B with Gln again bound in the glutaminase site and with the reaction intermediate AspAMP, PPi and Mg2+ bound in the synthetase active site. This model represents AS-B in the early stages of catalysis, after formation of the intermediate in the synthetase active site. 50

PAGE 51

Methods Simulated Annealing Methodolgy In SA, the system of interest is heated to a sufficiently high temperatur e to give the system enough kinetic energy to surpass all barriers in the potential energy landscape before being slowly cooled to the global minimum. If th e system reaches thermal equilibrium at each temperature, then the system can be compared to the Boltzmann distribution, which yields the most likely population of a state i with energy Ei at the temperature T : j kT E kT E ij ie Ne n. (3-1) Lower energy states become more probable as the temperature is lowered until it reaches absolute zero, where the system is in its lowest possible energy state. However, SA is in reality only an approximation, as an infini te number of time steps would be required and at each time step the system would be allowed to come to thermal equilibrium in order to reach the true global minimum.133 Cooling the system becomes the most critical aspect of a SA procedure. A logarithmic cooling scheme has been shown to be the most efficient at reaching the minimum energy conformations;134 however, it is still possible to reach the global minimum when not following a logarithmic cooling scheme if cooling is performed sufficiently slowly.135 Another option is running multiple SA experiments to increase the likelihood that the method finds the global minimum.136 Simulated Annealing Procedure Three different SA procedures were utilized to arrive at the global minimum conformations for the two model systems. They all started from the same minimized model systems, which are 51

PAGE 52

described in detail in a subse quent section. Long-range interac tions were calculated via the Particle-mesh Ewald (PME) method137-139 within the constructs of the periodic boundary condition (PBC)140,141 to avoid the improper use of cut-o ff distances. The setup for both the PBC and the calculation of the nonbonde d interactions remained unc hanged in every calculation presented here.The setup and calculation can be seen in any of the CHARMM example input files in the Appendix. The three di fferent SA methods all began from a single heating procedure; model systems were heated from 0 K to 600 K ove r the course of 240 ps with the temperature increasing 5 K every 2 ps at a c onstant pressure of 1 atm. As expected, in each case the volume of the system increased during the heating procedure. In the first set of SA experiments (SA Run 1) the heated systems were held at constant pressure (1 atm) at the peak temper ature (600 K) via the Nose-Hoover method,142,143 while the volume was allowed to expand as the temperatur e was distributed evenly throughout the system over 100 ps. The systems were then cooled linear ly from 600 K to 300 K over 500 ps at constant pressure (1 atm). Ten structures were collected at 50 ps intervals and at ea ch interval; the system was placed at constant temperature via the Nose-Hoover method for 0.1 ps to serve as a quenching mechanism to allow the temperature to quickly distribute evenly. In the second set of SA experiments (SA R un 2), the heated systems were immediately cooled after reaching 600 K by the same protocol described for the first set of SA experiments. Once the systems were cooled to 300 K, they were equilibrated over 100 ps with a constant pressure (1 atm) and at a constant temp erature (300 K), via the Nose-Hoover method. In the final set of SA experiments (SA Run 3), after heating, 100 ps of run time allowed the volume to expand dramatically as the pressure was held constant at 0.025 atm and the temperature was held constant by Nose-Hoover methods at 600 K. Th e systems were then cooled 52

PAGE 53

under high pressures (up to 1,000 atm) until the syst ems reached 300 K and were as close to their original volumes as possible. Again, in the cooli ng process, ten structures were collected at 50 ps intervals and quenching steps we re implemented as previously described. Every SA experiment finished with the minimization of the systems at each of the ten temperatures where structures were collected. The Appendix contains the input files for the heating, equilibration, cooling and final minimization for the SA Run 1 for the model system that contains the AspAMP intermediate. Only slight manipulations to these core input files and rearrangement of the order in which the files were run, as discussed above w ith the various SA protoc ols, were required to produce all of the remaining input files. Model Systems AspATP model system The first of the two model systems that were optimized in this work was taken from the work of a former group member, Dr. Yun Ding.78 This model consists of the E. coli glutaminedependent asparagine synthetase (AS-B) with glutamine bound in the N-terminal glutaminase active site and aspartate, ATP and Mg2+ bound in the C-terminal s ynthetase active site. The protein structure contains all residues from 1 516, including the two small loops from the interior of the protein that were unresolved in the crystal structure of AS-B (PDB code: 1ct9).57 The two unresolved loops were added as part of the dissertation work of Dr. Ding.78 As in the AS-B crystal structure, the first residue of the model is an alanine rather than the naturally occurring cysteine. The initial structural confor mation of the system was taken as the best structure as identified from the work of Dr. Di ng. There were 306 water molecules added to this complex based on their positions in the crystal structure. The complete complex (ASB/Gln/Asp/ATP/Mg2+/H2O) has a total charge of -20 (AS-B = -17; Gln = 0; Asp = -1; ATP = -4; Mg = +2). The complex was then solvated in a cubic box of water usi ng the TIP3 water model 53

PAGE 54

and NaCl was added to the system in the form of Na+ and Clions to neutralize the system. The system was then minimized to an energy convergence of 0.00001 with the SHAKE algorithm144 in use. For the rest of this work, this system setup will be referred to as the AspATP model system. AspAMP model system The second model system optimized in this work was initially built from the coordinates of the AspATP model complex described above. Usi ng the AspATP model as a starting structure, the AspAMP intermediate was built by manipulation of the aspartate and ATP structures. In the active site, aspartate was moved closer to ATP as the carboxylate side ch ain was rotated towards the -phosphate of ATP where it attacks to form the intermediate. A bond was then drawn to join the carboxylate tail of aspartate to an oxygen on the -phosphorus. Next, the bridging bond between the and -phosphates was broken to create PPi and the AspAMP intermediate. Aspartate was the only molecule to move during this procedure. Both the intermediate and PPi retained their initial coordinates from ATP in the AspATP model. Figure 3-1a below shows the original positions of the Asp and ATP and Figure 3-1b shows the newly created AspAMP and PPi. As in the AspATP model, the 306 crystal water molecules were added to this complex based on their positions in the crystal st ructure. The complete complex (ASB/Gln/ AspAMP/PPi/Mg2+/H2O) has a total charge of -20 (AS-B = -17; Gln = 0; AspAMP = 1; PPi = -4; Mg = +2). The complex was then solv ated in a cubic box of water using the TIP3 water model and NaCl was added to the system in the form of Na+ and Clions to neutralize the system. The system was then minimized to an energy convergence of 0.00001 with the SHAKE algorithm.144 For the remainder of this work, this system setup will be referred to as the AspAMP model system. 54

PAGE 55

Figure 3-1. Creation of AspAMP intermediate. (a) Original positions of the ATP and aspartate from the AspATP model. (b) The AspAMP intermediate created by the joining of aspartate and ATP as described in the text. Results and Disscussion AspATP Model The three simulated annealing runs that we re performed on the AspATP model system began from a single minimized initial structure and a single heating run to 600 K over 240 ps. This method was chosen to minimize the computa tional efforts required so that more annealing protocols could be tested. As was detailed more thoroughly in the previous section, the three runs then varied in their equilibration and cooling protoc ols, before all of the resulting structures were minimized. The resulting energies of the three co mplete SA protocols can be seen in Table 3-1. The resulting structures from the equilibrations of Run 1 and Run 3 were not minimized because the equilibrations were carried out at the peak temperature of 600 K and the resulting structures were so deviant from the initia l structures that a minimizatio n would not have produced final structures of any relevance. 55

PAGE 56

Table 3-1. Final energies for the SA Runs for the AspATP model system Energy (kcal/mol) SA Run 1 SA Run 2 SA Run 3 Initial -275,457.76 600 Ka -67,066.50 100 ps run at 600 Ka -44,075.66 -284,660.89b -44,800.98 570 K -281,372.96 -281,214.41 -281,372.96 540 K -282,930.30 -281,497.79 -281,533.30 510 K -281,943.36 -282,295.30 -281,943.36 480 K -283,983.00 -282,403.70 450 K -286,284.01 -284,966.86 -282,445.26 420 K -286,707.27 -285,230.17 -282,849.96 390 K -287,037.59 -285,303.97 -282,742.20 360 K -287,474.56 -283,631.96 -283,194.62 330 K -287,564.08 -283,955.49 -283,757.58 300 K -288,046.56 -284,453.55 -284,418.12 a Structures were not minimized; b This energy value was for the minimized final structure for the equilibration at 300 K, as no equilibration was done at 600 K in SA Run 2. The minimization at 480 K in SA Run 1 continually failed. The analysis for all the SA R uns began with a look at the initial structure. Figure 3-2 shows the potential hydrogen bonds that exist between active site re sidues of AS-B, free aspartate, and ATP in this model. Although this t ype of polar contact is only one piece in the assessment of the accuracy of mode l systems, it is a useful first step for comparison of the other structures. For clarity, the ATP and aspartate interactions have been split, rather than shown in the same view. Many of the residues shown interacti ng with the ATP have been identified as part of a common ATP binding sequence. Residue s Ser-234, Leu-237, Asp-238, and Ser-239 are all a part of the known pyrophosphatase loop (SGGXDS),59,77 while Lys-449 also has been proposed to have ATP binding implications in AS.145 A structural alignment was performed in PyMOL146 to compare the heated system to the initial system. PyMOL estimated an rsmd (based on the backbone coordinates of the C s) of 2.314 In the alignment (Figure 3-3), the major difference is that the opening at the synt hetase active site has grown. The loop shown in red on the heated 56

PAGE 57

Figure 3-2. Synthetase active site interaction in the initial AspATP model system. The ASB/ATP interactions are shown on the left with the AS-B/Asp interactions shown on the right. The AS-B is shown with the carbons colored green while both substrates are depicted with white carbons. The potential interactions are shown with the dashed yellow lines. The Mg2+ is also show as a sphere in the view on the left. For the remaining atoms O = red, N = blue, P = orange, and H = white. AspATP model (cyan) can moved away from th e active site. In addition, the entire protein structure has lost some tertiary structural feat ures with the added energy, as can be seen with both sides of the heated protein moving away from the synthetase active site. This results in an opening of the synthetase active site, which allows the free aspartate to migrate away from the initial position. The ability of the aspartate to move freely could indicate that a second Mg2+ is absent in the model. This Mg2+ would help anchor the aspartat e to prevent it from begin pushed away from ATP by charge-charge repulsion. This could also suggest th at the missing final 37 residues do serve a purpose in clos ing off the synthetase active site and restricting the position of the free aspartate. 57

PAGE 58

Figure 3-3. Structural co mparison of the initial structure of the AspATP model with the same structure heated to 600 K. The initial structure is in green, while the heated structure is in cyan and red. The synthetase substrates, aspartate, ATP and Mg2+, are shown in stick or sphere (Mg2+) representation with the ca rbons colored to match their respective system. Simulated annealing run 1 Upon reaching 600 K, the temperature was allowed to distribute evenly in the system for 100 ps, as described above. Finally, the system was cooled to 300 K over 500 ps, while structural snap-shots were written out in 50 ps, i.e. 30 K, increments. These 10 structures were minimized to obtain the final structures used in the anal ysis. Figure 3-4 shows four plots that track the 58

PAGE 59

evolution of the system throughout the annealin g process. The temperature, volume and energy plots were generated from the output files, while the rmsd plot was generated from the trajectory files created during the simulati on. As expected, the volume of the system fluctuated since the pressure was held constant (1 atm) and the temperature was regulated. The series of graphs shown in Figure 3-4 de monstrate that the si mulation progressed as planned. The interpretation of these results must th en be drawn from the final energies of the Figure 3-4. AspATP SA Run 1. Going clockwise from the top left, the first plot is a plot of the temperature of the system vs the time step. The second plot depicts the volume of the system vs the time step. The third plot show s the energy of the system vs time step. In this graph, the blue curve represents the kinetic ener gy of the system, while the orange curve represents the potential energy of the system, and fi nally the black curve follows the total energy (which is a sum of the kinetic and potential energies). The last plot shows the rmsd of the backbone atoms in the system as the simulation progresses, as compared with the initial structure vs the time step. 59

PAGE 60

system as the annealing progressed and from close visual inspection of the system. Table 3-1 shows the energies of the AspATP system over the course of the annealing. The energies are those of the system at each sp ecific temperature after the struct ures were minimized, effectively lowering each structure to 0 K and trapping them in the closest minima. During the constant run at 600 K for 100 ps of SA Run 1, the free aspartate migrated completely outside of the syntheta se active site, and was never able to return to the active site during the cooling process. Although these structures have the lo west overall energies for the AspATP system across all SA Runs, they cannot be regarded as valuable structures when aspartate is absent from the activ e site. Several constraints were tried in an attempt to pull the aspartate back into the active site; however, th e constraints were elevat ing the energy of the system to the point of crashing the calcul ation, rendering the cons traints not viable. Simulated annealing run 2 In the second simulated annealing protocol, the heated AspATP model system was cooled immediately, rather than equilibr ating at 600 K. The cooling process was the same as previously described, as 10 structures were collected during the cooling. Once the system reached 300 K, it was then equilibrated and the final structure of the equilibration was minimized by the same procedure as the 10 structures co llected from the cooling. Figure 3-5 shows the plots made from following the systems through this specific SA procedure. One distinguishable difference in the system from SA Run 1 to SA Run 2 is the tota l volume of the system. In this protocol, the volume did not expand as much as in the first run and the volume of the system ended up in the same range as where it began, rather than ending at a larger total volume. 60

PAGE 61

Figure 3-5. AspATP SA Run 2. Going clockwise from the top left, the first plot is a plot of the temperature of the system vs the time step. The second plot depicts the volume of the system vs the time step. The third plot show s the energy of the system vs time step. In this graph, the blue curve represents the kinetic ener gy of the system, while the orange curve represents the potential energy of the system, and fi nally the black curve follows the total energy (which is a sum of the kinetic and potential energies). The last plot shows the rmsd of the backbone atoms in the system as the simulation progresses, as compared with the initial structure vs the time step. While all of the structures were examined in comparison to the initial structure, the two which are the focus of this analysis are the lowest total energy structure (the minimization of the structure at 390 K) as seen in Table 3-1, and the final struct ure that was the result of a minimization after an equilibration period of 100 ps at 300 K. Both of these structures have low rmsd values when compared to the initial struct ure. PyMOL estimated the rmsd of the minimized structure at 390 K to be 1.651 in comparison to the initial stru cture, while the equilibrated 61

PAGE 62

structure had an rmsd of 1.578 Figure 3-6 shows the structural alignments of both structures with the initial AspATP system. In both cases, th e overall structure of th e proteins looks very similar to the original system. The major deviations occur in th e positions of the adenine group of the ATP and free aspartate. To examine the differences, the active sites were inspected in more detail. Figure 3-6. Structural alignmen ts of the minimized structure at 390 K and the structure equilibrated at 300 K with the initial structure. On the left the original structure is shown in green and the minimized structure at 390 K is shown in magenta, while on the right, the minimized structure of the equi libration run at 300 K is shown in yellow where the original structure remains in green. Figure 3-7 shows a detailed view of the ATP binding site for all three structures. In a comparison of the three structures, the pyrophosphatase loop (shown wi th residues Ser-234, Lue237, Asp-238 and Ser-239) has almost no change. This holds true for all of the residues surrounding the phosphate tail of ATP. The only large difference was in the position of the adenine moiety of ATP and the residues surrounding it. It rotates about the -phosphate group in each structure, shifting the residues around it as it moves. There is also one interesting side chain shift from the initial st ructure as well. In both the 390 K structure and the equilibrium structure, 62

PAGE 63

Figure 3-7. A look at ATP binding inte ractions in SA Run 2. The in itial structure is shown with green carbons, while the 390 K minimized structure shown with magenta carbons and the minimized structure of the equilibrati on at 300 K is shown with yellow carbons. The top view is the complete AS-B/ATP view that is then broken down into the residues only on the bottom left and th e ATP only on the bottom right. The noncarbon atoms are colored as follows: O = re d, N = blue, H = white, P = orange, and Mg2+ = color of model carbon. 63

PAGE 64

the side chain of reside Arg-447 has shifted to make a hydrogen bond with the phosphate tail of ATP. A closer look at the free aspartate and the residues that interact with it reveals a few interesting changes. Figure 3-8 shows shifts in residues and free aspartate that occurred in each Figure 3-8. A look at free aspart ate binding interactions in SA Run 2. The initial structure is shown with green carbons, while the 390 K minimized structure shown with magenta carbons and the minimized structure of th e equilibration at 300 K is shown yellow carbons. The top view is the complete AS-B/Asp/ATP/Mg2+ view that is then broken down into the residues only on the bottom left and the Asp/ATP/Mg2+ only on the bottom right. The non-carbon atoms are colo red as follows: O = red, N = blue, H = white, P = orange, and Mg2+ = color of model carbon. 64

PAGE 65

model. The major changes in residue s are seen in a side chain flip of Glu-352 and a large shift in Lys-376 as the adenine moiety of ATP shifts, pushing the aspartate and the Lys-376 back. In both the 390 K minimized structure and the equilibrated structure, the free aspartate is in a seemingly worse position for attack of ATP and fo rmation of the intermediate. This seems to be more evidence that something is missing in this model. Whether the important factor omitted is a second Mg2+ ion or the remaining 37 residues cannot be determined from these results. Simulated annealing run 3 In the third type of simulated annealing run, after the system was heated to 600 K, the temperature was then allowed to distribute evenly in the system under a reduced but still constant pressure of only 0.025 atm. This was done to al low the volume to expand even more. Increasing the volume of the system has the effect of lowe ring the density, and in principle should allow the protein more freedom to explore even more conformational space while at 600 K. By then cooling under higher pressure, the volume is forced down and can help the protein more readily adopt the lowest energy conformer.124 Figure 3-9 shows the progression of this SA run. At the end of 340 ps, this system had reached a larger vo lume than either of the previous SA runs. The first 250 ps of cooling were performed under a constant pressure of 1000 atm. This led to an instant and steep decrease in th e volume of the system. The pr essure was then reduced to 500 atm for the next 100 ps of cooling. The transiti on from 1000 atm to 500 atm is where a brief increase in the volume of the system is seen near the 600 ps mark. The pressure was then reduced again to 10 atm for 50 ps, where anothe r transition can be seen on the volume graph around 700 ps mark, before running the final 100 ps of cooling under a constant pressure of 1 atm. The initial system started with a volume of 778,688 3 and reached a maximum volume of approximately 1,380,471 3 after 340 ps of total simulation time before ending with a total system volume of approximately 700,605 3. Structures were collected at the same 10 points as 65

PAGE 66

Figure 3-9. AspATP SA Run 3. Going clockwise from the top left, the first plot is a plot of the temperature of the system vs the time step. The second plot depicts the volume of the system vs the time step. The third plot show s the energy of the system vs time step. In this graph, the blue curve represents the kinetic ener gy of the system, while the orange curve represents the potential energy of the system, and fi nally the black curve follows the total energy (which is a sum of the kinetic and potential energies). The last plot shows the rmsd of the backbone atoms in the system as the simulation progresses, as compared with the initial st ructure vs the time step. The small gap in the temperature, volume and energy graphs is because that particular output file was lost prior to extracting the necessary information. in the previous two SA runs and the energies of the corresponding minimized structures can be seen in Table 3-1. Much like what occurred in SA Run 1 for this system, allowing the temperature to distribute at 600 K opened the synt hetase active site and with no apparent interactions strong enough to hold the free aspartate in the active site; it migrated out. Figure 3-10 shows the 66

PAGE 67

structural alignment of the lowest energy struct ure from Table 3-1 for SA Run 3. This structure was the minimized structure acquired at 300 K. Py MOL was used to calculate the rmsd for the minimized structure at 300 K to be 1.606 in comp arison to the initial structure. The only major change in the overall structure is the C-termi nus, where the protein structure moved out away from the synthetase site creating a large opening. The free aspartate that has completely left the active site can be seen above the opening in th e active site, in the upper middle of Figure 3-10. Figure 3-10. Structural comparison of the initial structure of the AspATP model with the minimized structure at 300 K. The initial structure is in green, while the minimized structure at 300 K is in hot pink. The synthetase substr ates, aspartate, ATP and Mg2+, are shown in stick or sphere (Mg2+) representation with the carbons colored to match their respective system. The sum of these three SA runs on the AspATP model system suggests that there must be some piece of this model missing that would keep the free aspartate in the synthetase active site 67

PAGE 68

and in position for attack on ATP. As suggested earlier, the most plausible answer would be a second Mg2+ in this active site; however, the final 37 re sidues that are absent could also play a role in constraining aspartate to the active site. AspAMP Model The three simulated anneali ng runs performed on the AspAMP model system began from a single minimized initial struct ure and a single heati ng run to 600 K over 240 ps, just as was the case for the AspATP model system. As was detail ed more thoroughly in the Methods section, the three runs then varied in thei r equilibration and cooling protocol s, before all of the resulting structures were minimized in a consistent fashio n as the final step. The resulting energies of all three complete SA protocols can be seen in Table 3-2. Table 3-2. Final energies for the SA Runs for the AspAMP model system Energy (kcal/mol) SA Run 1 SA Run 2 SA Run 3 Initial -290,457.77 600 Ka -69,076.12 100 ps run at 600 Ka -46,136.79 -292,764.37b-46,115.30 570 K -290,097.90 -290,207.23 -289,884.32 540 K -290,189.82 -290,240.85 -290,375.42 510 K -292,308.88 -290,488.33 -290,555.66 480 K -293,666.30 -292,179.31 -291,091.17 450 K -294,375.28 -293,035.19 -291,478.55 420 K -294,859.69 -293,509.66 -291,637.92 390 K -294,925.12 -293,376.03 -292,007.39 360 K -295,462.21 -292,423.96 -291,941.47 330 K -295,938.30 -292,226.61 -292,373.55 300 K -295,918.78 -292,501.75 -292,762.35 a Structures were not minimized; b This energy value was for the minimized final structure for the equilibration at 300 K, as no equilibration was done at 600 K in SA Run 2. A closer examination of the initi al structure of AS-B with the AspAMP intermediate and PPi bound in the synthetase active site is shown in Figure 3-11. Because the AspAMP model 68

PAGE 69

Figure 3-11. Synthetase active site interaction in the initial AspAMP model system. The AS-B is shown with the carbon s colored green while the AspAMP intermediate is depicted with white carbons. The potential interac tions are shown with the dashed yellow lines. The Mg2+ is also show as a white sphe re. The other non-carbon atoms are colored as follows: O = red, N = blue, H = white, and P = orange. system was created from the AspATP model, most of the same inte ractions are seen here as in the previous system. A structural alignment was then performed in PyMOL to compare the initial structure with the structure after it had been heated to 600 K. PyMOL found the rmsd of the heated structure of 1.560 in comparison to the initi al structure. The structural alignment can be seen in Figure 3-12. As can be seen in the overlay and with the rmsd, there is not a great deal of variance between the heated and initial structure. Th is is in sharp contrast to the AspATP system that experienced a much larger struct ural change with the heating. In the AspAMP model, the AspAMP intermediate has enough interactions w ith the surrounding protein structure, shown in Figure 3-11, to stay close to th e position it started in. The PPi, on the other hand, is held in place 69

PAGE 70

because it is buried behind the AspAMP intermediate inside of the synthetase active site, as well as being locked down by interactions with the Mg2+. Figure 3-12. Structural comparison of the initial structure of the AspAMP model with the same structure heated to 600 K. The initial structure is in green, while the heated structure is in cyan. The synthetase substrates, AspAMP, PPi and Mg2+, are shown in stick or sphere (Mg2+) representation with the carbons colored to match their respective system. 70

PAGE 71

Simulated annealing run 1 The first simulated annealing run for the AspAMP model system followed the same protocol that was previously described, and for the first run for the AspATP model system. The system was first heated to 600 K over 240 ps at a constant pressure of 1 atm. After reaching 600 K, the temperature was allowed to distribute even ly in the system for 100 ps at a constant pressure of 1 atm. At that point, the system wa s cooled down to 300 K over the course of 500 ps Figure 3-13. AspAMP SA Run 1. Going clockwise from the top left, the first plot is a plot of the temperature of the system vs the time step. The second plot depicts the volume of the system vs the time step. The third pl ot shows the energy of the system vs time step. In this graph, the blue curve represen ts the kinetic energy of the system, while the orange curve represents the potential energy of the system, and finally the black curve follows the total energy (which is a su m of the kinetic and potential energies). The last plot shows the rmsd of the backbone atoms in the system as the simulation progresses, as compared with the initial structure vs the time step. 71

PAGE 72

with structures being collected every 50 ps. The collected struct ures were then minimized and the final energies of minimizations can be seen in Table 3-2. Figure 3-13 s hows the graphs of the temperature, volume, backbone rmsd and energy of the system as SA Run 1 progressed. The lowest energy structure was for the mi nimized structure at 300 K (Table 3-2). A structural alignment of this structure with the initial AspAMP structure in PyMOL gave an rmsd of 1.887 This was the lowest rmsd found between the initial structure and any of the minimized structures. Figure 3-14 shows the active site comparison between the initial AspAMP structure and the minimized structure at 300 K. As can be seen from this figure, there are a few significant differences. The AspAMP intermediate undergoe s a significant shift of the apsartyl tail. This shift seems to result in the sh ift of residues on the right side of the active site. Tyr-357 appears to have shifted out of the way of the intermediate tail, while the Arg-387, Asp384 and Lys-376 have shifted slightly because they are no longer within hydrogen bonding distance of the aspartyl tail. This shift in the tail also si gnificantly shifts Lys449 that is located at the top of the active site in Fi gure 3-14. As in the AspATP model system, the side chain of Arg447 shifts to interact with the PPi that is present in this model. Perhaps the most striking change is in resi due Glu-348 that is seen at the center of the active site (Figure 3-14). This Glu-348 has been extensively tested experimentally as part of the dissertation work by former group member, Dr. Jemy Gutierrez.147 This residue is located at the mouth of the intramolecular tunnel that connects the gluatminase ac tive site with the synthetase active site and is believed to be involved in the formation of the AspAMP intermediate or in active site communication. It is also possible that this residue acts as a gate for the passage of the NH3 formed in the glutaminase active site. In the a nnealed model, at all stages, the side chain of Glu-348 has flipped. As can be seen in Figure 3-15, the side chain f lip of Glu-348 seems to have 72

PAGE 73

Figure 3-14. Active site comparison between the initial AspAMP model and the model minimized at 300 K. The initial structure is shown with green carbons, while the minimized structure at 300 K is shown with magenta carbons. The top view is of the active site with the AspAMP intermediate, PPi and Mg2+ bound. The bottom left is the same view with the substrates removed for clarity and the bottom right is the only the substrates. The non-carbon atoms are co lored as follows: O = red, N = blue, H = white, P = orange, Mg2+ = model carbon color. positioned the negatively charges carboxylate tail away from the opening of the tunnel in the synthetase active site. Judging from the entire range of structures co llected, the shift of the side 73

PAGE 74

chain of Glu-348 is the most lik ely cause of the shift in the AspAMP intermediate tail shift that forces the other residues to shift as well. This sequence is suggested because in the series of models, the AspAMP intermediate tail gradually m oves away from Glu-348, which has flipped in every instance. Looking at this change in the remaining two series of SA runs should confirm this change in the AspAMP model. Figure 3-15. The shift of Glu-348. This view is looking down the intramolecular tunnel that connects the glutaminase site with glut amine bound (near) and the synthetase active site with the AspAMP intermediate bound (far) in AS-B. The original structure of the model is seen in green while the annealed model is shown in magenta. The Glu348 side chain flips from right to left away from the opening of the tunnel. 74

PAGE 75

Simulated annealing run 2 In the second simulated annealing run performed on the AspAMP model, the cooling stage was begun immediately afte r the system reached 600 K. This was the same process done for the SA Run 2 for the AspATP model. The syst em was then equilibrated at 300 K and all of Figure 3-16. AspAMP SA Run 2. Going clockwise from the top left, the first plot is a plot of the temperature of the system vs the time step. The second plot depicts the volume of the system vs the time step. The third pl ot shows the energy of the system vs time step. In this graph, the blue curve represen ts the kinetic energy of the system, while the orange curve represents the potential energy of the system, and finally the black curve follows the total energy (which is a su m of the kinetic and potential energies). The last plot shows the rmsd of the backbone atoms in the system as the simulation progresses, as compared with the initial structure vs the time step. 75

PAGE 76

the structures were minimized. The progression of th e SA run can be seen in the graphs in Figure 3-16. The system was held at a cons tant pressure of 1 atm for the en tire SA process. The energies of the minimized structures can be seen in Table 3-2. All of the minimized structures were then compared to the initial structure for the AspAMP model in PyMOL. The minimized struct ure taken at 420 K and the minimized results of equilibration at 300 K are the key structures for SA Run 2. These were chosen because the minimized structure at 420 K was found to be th e lowest overall energy structure for SA Run 2 (Table 3-2), while the structure that was minimi zed from the equilibration at 300 K was the final structure collected in this process. PyMOL was used to calculate an rmsd for the minimized structure at 420 K to be 2.006 in comparison to the initial st ructure, while the minimized structure from the equilibration run was found to have an rmsd of 1.609 A comparison of the synthetase active sites (Figure 3-17) shows much less variation than was seen in the model from SA Run 1. Both of the structures compared against the initial structure for SA Run2 have less variance in the active site in cont rast to the results from SA R un1 for the same model system. In Figure 3-17, the residues that showed the greate st deviation in SA R un 1 (Lys-449, Tyr-357 and Asp-279) are all closer to their in itial positions in SA Run 2. Even with what appears to be better structural agreement, Arg-447 and Glu-348 still experienced the same side chain flipping as seem in SA Run 1. In the case of the model minimized after equilibration at 300 K, the carboxylate of the AspAMP intermediate is within hydrogen bonding distance of Glu-348. The AspAMP from the minimized structure taken at 420 K is similiar to the result of SA Run 1, with the aspartyl tail shifted away from Glu-348. Th is aspartyl tail shift is the most significant difference between the two models examined in Fi gure 3-17. The remaining variations from the 76

PAGE 77

Figure 3-17. Active site comparison between the initial AspAMP model and the model minimized at 420 K and the model minimized after equilibration at 300 K. The initial structure is shown with gr een carbons, while the minimized structure at 420 K is shown with slate carbons and the minimized structure after equilibration is shown with yellow-orange carbons. The top vi ew is of the active site with the AspAMP intermediate, PPi and Mg2+ bound. The bottom left is the same view with the substrates removed for clarity and the bo ttom right is the only the substrates. The non-carbon atoms are colored as follows: O = red, N = blue, H = white, P = orange, Mg2+ = model carbon color. 77

PAGE 78

initial model are consistent betw een the two models produced in th e annealing process, as well as the other structures produced, but not shown here. These results provide evidence that residue Glu-348 undergoes a side chain f lip that is independent of AspAMP position. Simulated annealing run 3 In the third simulated annealing run, as was done for the AspATP model, after the system was heated to 600 K, the temperature was then a llowed to distribute evenly in the system under a reduced but still constant pressure of 0.025 atm. The system was then cooled under high pressure until it was at 300 K. The graphs in Figure 3-18 follow the progression of this SA run. At the end of 340 ps, this system reached a larger volume than either of the previous SA runs on the AspAMP model. The first 300 ps of cooling we re performed under a constant pressure of 1000 atm. This led to an instant and steep decrease in the volume of the system. The pressure was then reduced to 500 atm for the next 100 ps of cooling. The transition from 1000 atm to 500 atm is where a brief increase in the volum e of the system is seen near the 650 ps mark. The pressure was then reduced to 1 atm for the final 100 ps, where another transition can be seen on the volume graph around 750 ps. The initial sy stem started with a volume of 778,688 3 and reached a maximum volume of approximately 1,415,147 3 after 340 ps of total simulation time before ending with a total system volume of approximately 716,928 3. Structures were collected at the same 10 points as in the previous two SA runs and the energies of the corresponding minimized structures can be seen in Table 3-2. 78

PAGE 79

Figure 3-18. AspAMP SA Run 3. Going clockwise from the top left, the first plot is a plot of the temperature of the system vs the time step. The second plot depicts the volume of the system vs the time step. The third pl ot shows the energy of the system vs time step. In this graph, the blue curve represen ts the kinetic energy of the system, while the orange curve represents the potential energy of the system, and finally the black curve follows the total energy (which is a su m of the kinetic and potential energies). The last plot shows the rmsd of the backbone atoms in the system as the simulation progresses, as compared with the initial st ructure vs the time step. The small gap in the temperature, volume and energy graphs is because that particular output file was lost prior to extracting the necessary information until it was at 300 K. The lowest energy structure found in this annealing process was for the minimized structure at 300 K (Table 3-2). PyMOL was again used to calculate an rmsd of 2.379 for this structure in comparison to the initial structur e. A closer look at th e active site comparison between the minimized structure at 300 K and th e initial structure (Fi gure 3-19) reveals more similarities to the model produced in SA Run 1 th an those that resulted from SA Run 2 for this 79

PAGE 80

Figure 3-19. Active site comparison between the initial AspAMP model and the model minimized at 300 K. The initial structure is shown with green carbons, while the minimized structure at 300 K is shown with hot pink carbons.The top view is of the active site with the AspAMP intermediate, PPi and Mg2+ bound. The bottom left is the same view with the substrates removed for clarity and the bottom right is the only the substrates. The non-carbon atoms are co lored as follows: O = red, N = blue, H = white, P = orange, Mg2+ = model carbon color. model. Large shifts are s een in the residues surround the aspart yl tail of the AspAMP intermediate. This model actually has a slightly different position for the aspartyl tail, with it 80

PAGE 81

positioned coming out of the view, rather than ju st being shifted above or below the initial position. Despite the large deviations in active site residue positions, the Arg-447 and Glu-348 show side chain shifts in every structure pr oduced over the course of all three simulated annealing runs carried out on the AspAMP model. With these shifts being a constant across all of the simulations, it seems that these two residues in particular perfer to move to these new positions in this model. Conclusions Simulated annealing has been used for some tim e as an effective tool for the optimization of model systems. Here, two different model systems, AspATP model and AspAMP model, were optimized through three varying simulated a nnealing procedures. For both models, the first SA procedure produced the lowest energy structur es of all three anneal ings. However, these systems, in particular the AspATP model, expanded so gr eatly that they were never quite able to recover and ultimately produced higher rms deviati ons from their initial systems. The second set of simulated annealing runs performed on the two model systems returned the lowest rms deviations from the original structures of all th ree SA procedures. In both cases, the minimized structures from the equilibrations at 300 K appeared to be the be st overall optimized structures. The final SA procedure that included high pressures in the cooling phase seemed to produce structures and energies that were in between th e first two SA procedures in both structure rms deviations and in energies. The best structure for the AspATP model sy stem came as a result of the structure equilibrated at 300 K of SA Run 2. This system had the lowest overall rmsd in comparison to the initial structure and this entire SA run was the only one performed on AspATP in which the free aspartate did not leave the active site. However, ev en in the best structures, the position of the 81

PAGE 82

free aspartate is poorly positioned for attack on ATP to form the AspAMP intermediate. This result and the continued escape of the aspartate fr om the active site suggests that the model has shortcomings; it is either missing a Mg2+ or the final 37 residues that play a role in containing the free aspartate and preparing it for attack. Very recent unpublished work by Megan Meyer on the final 37 residues of AS-B found that the enzyme was still active even (activity was about half that of the wild-type enzyme) when the final 37 residues are removed. This may suggest that these residues are not critical for catalysis. This result in conjunction with the results of the SA would make the presence of a second Mg2+ the most likely explanation. The best structure for the AspAMP model system also came as a result of the structure equilibrated at 300 K of SA run 2. This structure has the lowest rms deviation in the protein structure from the initial structure of all the tr ials. One consistent cha nge across every SA run was the flipping of the side chain for the Glu-348 residue (Figure 3-15). One putative role of this residue is that it acts as a gate at the mouth of the intr amolecular tunnel connecting the glutaminase and synthetase site. The flip seems to position the car boxylate group, which could block the translocation of the NH3 by forming a hydrogen bond with it, away from the opening. This movement could allow the NH3 to pass by and attack the AspAMP intermediate to form the tetrahedral transition state more easily. This optimized model of AS-B with the AspAMP intermediate bound is believed to represent a significant improvement in any previous versions of the same model system. 82

PAGE 83

CHAPTER 4 ENHANCING THE PMF99 SCORING FUNCTION FOR MOLECULAR DOCKING AND VIRTUAL SCREENING Introduction Recently, the human genome project and effort s in high-throughput crystallography have led to a number high-resolution crystal structures that can be used in drug discovery projects. A routine first approach in drug discovery is the in silico screening of a databa se of molecules into a receptor of interest, using a scoring functi on to estimate the fit and possibly the binding affinity.148,149 Multiple docking programs are currentl y used in the biotech and pharmaceutical industries.150-159 Over the past decade, there have been multiple reports of virtual screening applications being successfully employed.160-162 Despite the success, there is still not one definitive methodology for accurate docking and scor ing across a wide array of protein/ligand complexes. The success of any docking study is constraine d by the choice of (i) the intramolecular potential energy description, or scoring function, and (ii) the al gorithm for exploring molecular conformations that are accessible to the ligand and, in principle, the residues defining the binding site in the drug target. While a number of soluti ons have been developed that address the second problem,163,164 the difficulty in constructing generally robust scoring functions is illustrated by the existence of the bewildering variety of potentials that have been implemented in software packages such as DOCK,150 GOLD,151 FlexX,152 Glide,165 and AutoDock.166 The situation is also complicated by the need to devise scoring func tions that accurately de scribe protein/ligand interactions while being suffici ently simple for very rapid ev aluation, especially given the number of structures th at must be examined for any ade quate sampling of the conformational space.167 Thus, while force field-based potentials may give very accurate estimates of interaction energies for the protein/ligand complex, they possess a relatively complicated mathematical 83

PAGE 84

form. In addition, including the effects of solvation and confor mational entropy changes into the calculated interaction enthalpies introduces an additional layer of complexity into the calculation.168 Given its speed and functional simplicity, a Potential of Mean Force (PMF) scoring potential is ideal for in silico screening experiments.169-172 Like other knowledge-based potentials, such as DrugScore173 or SMoG,174 the simple interaction-pair potentials in PMF are parameterized to reproduce experimentally observed structures of protein/ligand complexes, and therefore implicitly include solv ation and entropy effects that ar e difficult to model with force field-based strategies. On the other hand, the proc edure used to parameterize the pairwise PMF potentials is limited by the avai lability of crystallographic information. The inclusion of nonbonded intramolecular interactions within the ligand is also not accomplished in a straightforward manner in a PMF scoring function.171 Thus, the behavior of a modified PMF99 scoring function (ePMF99) in which the van der Waals terms in the original PMF99 implementation are replaced by pairwise non-bonded interactions was examined.175 As discussed below, this change permits the use of repulsion ra dii that are more approp riate for protein/ligand atom pairs that form hydrogen bonding or electrostatic interactions. The effects of this simple modification were demonstrated on the behavior of the ePMF99 potenti al in reproducing the structures of a well-defined test set of appr oximately 170 protein/ligand complexes taken from the Protein Data Bank.176 In addition, the new treatment of non-bonded repulsion was tested for improvements in the utility of the ePMF99 potential in cross-docki ng studies of thymidine kinase (TK) inhibitors.160 84

PAGE 85

Methods Protein/Ligand Structure Preparation All of the protein/ligand complexes were processed for subsequent docking experiments following an identical procedure to that reco mmended in previous efforts to assess the performance of the GLIDE scoring function.165 This procedure therefore yielded partially optimized structures from the original crysta llographic coordinates th at contained hydrogen atoms, with suitable adjustments to the protonatio n states of ionizable residues on the protein, optimization of tautomeric forms of histidine, and repositioning of reorientable hydrogen atoms to optimize hydrogen bonding intera ctions. Although explicit hydroge n atoms are ignored in the enhanced PMF99 scoring function, this procedur e was performed to ensure a valid comparison of results obtained using the enhanced PMF99 and GLIDE scoring functions. In addition, this procedure removes steric clashes within the pr otein/ligand complex. Processing to obtain all structures used in the ligand docking studies was performed using the Fi rstDiscovery suite of software packages for molecular modeling (V2.5; Schrdinger LLC, New York). Although details of the preparation procedur e have been described elsewhere, it is important to note that crystallographic water molecules are often remove d from the structure so as to increase the volume of the ligand binding pocket. When cofactor s were present in the crystal structure, they were assigned chemical bonds and formal charges consistent with standard Lewis structures, and treated as part of the protein in the docking st udies (i.e. their positions were fixed). After the addition of hydrogen atoms to the cofactors, prot ein residues, and ligand molecule, constrained energy minimizations were performed on the resulting complex, using Macromodel97 as implemented within the FirstDis covery V2.5 package. In these structural optimizations, which employed the MMFF94s force field,177 side chain hydroxyls (Ser, Thr and Tyr residues) and cysteine thiols were reoriented so as to op timize hydrogen bonding and remove steric clashes by 85

PAGE 86

strongly constraining all non-hydroge n atoms, and allowing hydrogen atoms to move freely in the absence of torsional interactions. Optimization is then repeated using weaker constraints until the rmsd of the non-hydrogen atom positions relativ e to their initial crys tallographic coordinates is less than 0.3 In preparation for docking, the crystal struct ure of TK complexed to thymidine (1kim)178 was processed following a protocol identical to that described above. Crystal structures of TK complexed to nine other purine and pyrimidine an alogs (Figure 4-3) were then employed in a standard superimposition procedure to obtain ini tial poses for the nine ligands within the 1kim TK active site. All ten ligands were then re-docke d into the rigid active site and their positions compared with those observed in the original crystal structures. Docking Algorithm Ligands were docked into their active sites in the protein structures using a Lamarckian genetic algorithm (LGA) similar to that im plemented within the AutoDock 3.0.5 package.166 Each search was performed for 50,000 generatio ns, or until the rmsd population fitness converged to a value less than 0.001 kcal/mol. E ach generation consisted of the following steps performed on the population of genomes (liga nd poses): selection us ing a roulette-wheel strategy,179 crossover (two-point), mutation (Cauchy-ba sed operator), elitis m, and local search. In these calculations, the probabilities of gene mutation (changes in ligand conformation) and gene crossover (generating a new ligand conformation by combining di hedral angles from multiple prior conformations) were defined as 0. 02 and 0.80, respectively. Local searches (SolisWest method)180 were performed on an individual with a probability of 0.06, and were performed for a maximum of 500 iterations. The population co mprised 100 individuals (ligand poses) and an elitism algorithm was employed in which the 10 best scoring individuals in each generation were preserved. All genetic algorithm parameter values correspond to those found to be optimal 86

PAGE 87

in prior studies employing the L GA search algorithm in AutoDock.166 On completion of the LGA search, the population was cleaned by cl ustering using a 0.3 rmsd cutoff in the coordinates of any two ligand conformations. Results and Discussion The Enhanced PMF99 Scoring Function As described in detail elsewhere,169-172 the PMF99 score for a given protein/ligand complex is computed by the addition of knowledge -based, pairwise atomic potentials for all intermolecular interactions, according to the expression: ij offcutrr kl ijrA score PMF )( (4-1) where Aij( r ) is the PMF99 interaction energy of a protein/ligand atom pair separated by a distance r that lies within a user-defined cutoff, rij cutoff. The PMF interaction energy Aij( r ) is calculated as: ij bulk ij seg j corrVol B ijr rfTkrA _ln (4-2) where kB is the Boltzmann factor, T is the absolute temperature and the volume correction factor 169,171 is included to account for the omi ssion of ligand/ligand interactions. j corrVolf_ rij segij bulk is the number density of ligand/pr otein atom pairs of type ij seen at an atom pair distance r is the number density of the atom pair type ij in the reference sphere that has a radius of 12 Although the magnitude of the PMF99 score likely has little physical me aning, it exhibits a correlation with experimental binding affinities for a diverse set of protein/ligand complexes.181184 Evaluating the PMF score is computationally efficient, however, because each pair potential can be pre-computed at 0.2 interval s and stored in a lookup table. 87

PAGE 88

As in calculations using empirical force fi elds, evaluating PMF99 scores requires the assignment of specific atom types. PMF99 cu rrently employs 34 ligand and 16 protein atom types, resulting in 544 unique intermolecular pair potentials.169 PMF04 has four significant changes in atom types as compared to PMF99. Th e first change was for the protein ring nitrogen atom type NR as applied to histidine ring nitr ogens. It was replaced for the more specific NA atom type for a nitrogen hydrogen bond acceptor and the already existing ND atom type for a hydrogen bond donating nitrogen. A ge neral atom type ME was introduced for metal ions of Zn, Ca, K, Mg, Mn and Fe. This atom type is onl y invoked for metal ions that are recorded as individual residues in the PDB. The ligand at om type OR for oxygen in a ring structure was merged into the oxygen in an ether bond atom type OE. The final significant change was the addition of the SO atom type for sulfur bonded to more than two atoms or to atoms other than carbon or hydrogen. These changes resulted in PM F04 having one additional protein atom type in comparison to the original PMF99. Thus, PMF04 has 578 unique intramolecular pair potentials.172 Since these numerical PMF potentials are obtained from crystal structures of protein/ligand complexes, they contain information about the free energy of interaction between all protein/ligand atom pairs as a function of distan ce, including the short distances that are typical for covalent bonds. On the other hand, the repulsiv e contribution to the pair wise potential is not modeled correctly because atom pairs do not get su fficiently close together in crystal structures used in parametrizing the interaction. A non-bonded repulsion term must therefore be added to the knowledge-based portion of the PMF to prevent atoms from being positioned at unphysically short distances. Standard implementations of the PMF99 scoring potentia l therefore employ van der Waals pairwise potentials from the AMBER fo rce field in order to overcome this limitation 88

PAGE 89

of the knowledge-based method.185 In the AMBER potential, however, each atom type is assigned a single fixed hardness and van der Waals ra dius to generate the re pulsive and attractive parts of the non-bonded potential, ir respective of the type of atom with which it forms a nonbonded interaction. Consequently, the non-bonded potential in the original version of PMF99 often does not describe hydrogen bonding inte ractions involving donor-acceptor, donor-donor, and acceptor-acceptor atoms equally well, and can minimize the importance of key hydrogen bonds in modeling protein/ligand complexes. There is a second problem in employing non-bonded repulsion potentials that are implemented in empirical force fields, such as AMBER, for modeling hydrogen bonds in knowledge-based PMF99 scoring function. This arises from the fact that in molecular mechanics methods,186 hydrogen bonding is represented as either a composite of electro static and van der Waals interactions, or by specific func tional forms, such as the 10,12-potential.101,104,187 As a result, standard repulsion/dispersion terms are ove rridden to ensure that atoms are placed at the correct hydrogen bond distance in any energy-minimized structure. The procedure for parameterizing PMF99 pair potentials cannot compensate for such errors in the AMBER repulsion potential, leading to hydrogen bonding distances that ma y be too long when compared with experimental data ( vide infra ). One solution to these problems is to empl oy specific 6,12 Lennard-Jones potentials for the pairwise interactions in the standard PMF99 im plementation, each of which can then be fit to account for all of the effects th at contribute to a non-bonded in teraction, i.e. van der Waals, electrostatics, and hydrogen bonds.175 This strategy permits hydrogen bonding and charge-charge interactions that would be pr ohibited on the basis of the van der Waals potential, as well as descriptions of donor-acceptor, donor-donor, and accep tor-acceptor pairwise in teractions that are 89

PAGE 90

of equal accuracy. This approach, which is implemented in the enhanced PMF99 scoring function, can be illustrated by c onsidering the hydrogen bonding in teraction between a positively charged amino group, such as that present on a ly sine side chain (NC), and the oxygen atom of a secondary alcohol (OD). Site-dir ected mutagenesis experiments ha ve established that such an interaction can contribute up to 3 kcal/mol of the ligand binding energy.188,189 Statistical analysis of structural data also supports the importance of this type of hydrogen bond, with the observed N-O distance being less than 3.8 .190,191 Plotting the AMBER non-bonde d potential for the NCOD interaction, however, shows that the repulsion term dominates if the heavy atoms are placed at a distance of less than 3 as seen in Fi gure 4-1. This is not a problem in force-field calculations because electrostatic terms that contribute to hydrogen bonding are also included, and the favorable charge-charge interactions rele ase sufficient energy to pay for the interatomic repulsion energy. Hence, the heavyatom N-O distance in optimized structures obtained using the AMBER force field reproduces expe rimental values. When the standard PMF99 potential for the same interaction is plotted (Figure 4-1), however, the effects of errors in the repulsive term in the AMBER 6,12 Lennard-Jones potential are not co rrected due to the absence of observed protein/ligand atom pairs that are placed too close together. As a resu lt, the standard PMF99 score becomes unfavorable at an N-O distance that is longer than that seen crystallographically for this type of hydrogen bonding interaction. On the other hand, in the enhanced PMF99 potential, the repulsion term is modified there by allowing the two heavy atoms to get closer together before the structure receives an unfavorable PMF99 score. This specific example illustrates the key problem with the original implementation of the non-bonded potential in the PMF99 method. The key role of the non-bonded potenti als is to ensure that bad structures in which atoms are too close together receive a high score. Unfortunately, flaws in the repulsion 90

PAGE 91

term can lead to good poses being scored unfavorably, resulting in their PMF99-derived free energy being higher than other incorrectly docked structures. Figure 4-1. A plot of the pairwise scores for protein/ligand atom pair of NC-OD. The original PMF99 score is shown in black with the updated PMF04 shown in green. The nonbonded correction is shown in blue, and the Amber vdW contributi on for the original PMF99 is shown in orange for comparison. Molecular Docking Algorithm Degrees of freedom in the docking consisted of translation of and rota tion about the center of mass of the ligand, and the dihedral angles of all rotatable bonds that we re not located in rings. Atoms in the protein were fixed at their crystallographic coordi nates during docking and scoring. All calculations in this study employed a Lama rckian genetic algorithm (LGA) search strategy for generating ligand conformations and their positio ns within the defined protein active sites. In this method, which was first implemented in the AutoDock 3.0.5 package,166 the Cartesian coordinates of each individual docked ligand co nformation (phenotype) are obtained from the degrees of freedom (genes) that are varied in the search algorithm, and the fitness is 91

PAGE 92

determined by the PMF99 score computed for th at complex. The population size was defined to contain 100 docked ligand conforma tions, and the 10 best scoring structures were used to generate the ligand population in the subse quent round of optimization. All simulations employed a maximum of 50,000 generations, with an energy convergence value of 0.001 kcal/mol. Although many values of variables such as population size, crossover rate, and elitism gave similar results for the complexes investigat ed, systematic investigations of the dependence of optimized ligand pose upon use r-defined parameters revealed the critical importance of defining the binding site volume in the protein, especially for solvent exposed sites. Thus, defining an active site region, or box, that is mu ch larger than the ligand creates an empty space that would normally be occupied by solvent and counter-ions outside the protein. In the absence of these molecules in the docking simulation, liga nds tend to occupy this empty space so as to avoid enclosure in the active site given that there is no energe tic penalty imposed by the PMF99 scoring function to prevent this. The use of a box defining the activ e site within which the ligand must be located overcomes this problem. On the other hand, the active site volume obviously needs to be large enough to accommodate the maximally extended conformation of the ligand because all relevant intermolecula r interactions must be included in the PMF99 scoring function. The initial ligand pose used in these docking experiments corr esponded to the starting pose in the cleaned crystal structure of the protein/ligand complex. While it has been argued that such a choice will bias the search and prevent explora tion of alternate ligand co nformations within the binding site,165 in many drug discovery studies there is a significant amount of structure-activity and mutagenesis data that provide some knowle dge of the initial ligand pose. Choosing the starting pose of the protein/ligand complex on this basis therefore enhances the probability that 92

PAGE 93

the final docked pose is at least as good as the initial pose by biasing the GA search process. On the other hand, since identifying bou nd conformations that may not be in accord with chemical intuition, the use of random ligand conformations to generate the starting complex was tested to examine if that alters the outcome of these doc king experiments. It was found that employing an initial conformation which differs significantly from the observed structure impacts only the number of steps required in the LGA search th at is required to positi on the ligand correctly within the defined active site. Assessing the Performance of the ePMF99 Scoring Function in Molecular Docking The evaluation of the combination of the enha nced PMF99 scoring function in conjunction with the LGA for docking was conducted with a test set of ove r 170 noncovalently bound cocrystallized protein/ligand complexes. The test set was chosen because it is large and encompassing as is the case in most evaluations of docking potential,148,164,192,193 and more specifically the test set was c hosen to closely mimic the published test set results of Glide,165 GOLD151 and FlexX.152 As in other studies, the strategy employed was to examine whether the combination of the ePMF99 scoring potential with the LGA could reproduce the crystal structures for each of the protein/ligand comple xes. Success was evaluated on the basis of the root-mean-square deviation (rmsd) in the coordi nates of the docked lig and relative to those observed experimentally. So as to simplify comparis ons with previous docking studies, all of the complexes were cleaned using a well-defined proc edure in which hydrogen atoms were added to the structure, with a subsequent constraine d energy minimization being used to remove physically unreasonable steric clashes.165 These calculations employed the MMFF94s force field,101 as implemented in the MacroModel software package.97 In these calibration studies, all crystallographic ordered water molecules present in the binding site of each complex were removed. This ensured that the ini tial protein structures were sim ilar to those employed in recent 93

PAGE 94

studies of the Glide algorith ms for molecular docking, simplif ying any comparison of results between the enhanced PMF and Glide Score potentia ls. In addition, the pres ence of these water molecules in virtual screening st udies may discriminate against lig ands that would displace them from the active site. Finally, removal of the waters aided the investigation of the importance of water-mediated, protein/ligand inte ractions, and the existence of alternate binding modes in the enlarged active site volume. For the test set of complexes, the ePMF99 potential employing specifically parameterized repulsion potentials gave a signi ficant improvement in the average RMSD over the original version of the PMF scoring function as seen in Tabl e 4-1. As can also be seen in the same table, PMF04 appears to underperform in comparison to ePMF99 and PMF99. This remained true when the behavior of the two scoring functions was assessed for ligands grouped according to the number of rotatable bonds (Table 4-1). PM F04 is suppose to represent a significant improvement to the PMF99 scoring function due to the reorganization of atom types and due to the increased PDB sample size from which the k nowledge-based potential is derived. There are two potential explanations for th e underachievement of PMF04 on th is particular test set. One possibility is that PMF04 is not implemented corr ectly in CAChe (the software package used for all the ePMF99, PMF99 and PMF04 docking simulati ons presented here). A second explanation for the apparent lack of improvement from PMF99 to PMF04 could be in the criteria used for the assessment rendered in this work. In all of the published original validation work by Muegge done on PMF99 and PMF04, he never assessed the validation of his PMF with rms deviations from the original crystal structures as is commonplace for the validation of most docking potentials. Rather, Muegge chose to present his validation in terms of the correlation between the PMF scores and experimental binding constants. Rough compar isons were made between the 94

PAGE 95

PMF04 scores of the final structures docked in CAChe and the PMF04 scores be estimated from Figure 8 of the Journal of Medicinal Chemistry paper PMF Scoring Revisited by Muegge. CAChe seemed to reproduce the PMF scores implied in this paper.172 Although very thin in quantity, this comparison seems to suggest that PMF04 has been implemented correctly in CAChe, and that PMF04 does not broadly represent an improvement to PMF99. Table 4-1. Comparison of rms deviation () for flexible docking 10 rotatable bonds 20 rotatable bonds all ligands av max av max av max Method rmsd rmsd rmsd rmsd rmsd rmsd PMF99a 2.91 13.29 2.96 13.29 2.98 16.32 PMF04b 3.29 15.74 3.59 15.74 3.51 15.74 ePMF99 1.47 7.67 1.58 7.67 1.52 7.67 Glide 1.34 7.55 1.70 11.88 1.98 13.10 GOLDc 2.62 14.01 3.01 14.01 3.16 14.01 FlexXd 3.07 12.55 3.51 13.41 3.73 15.54 The ePMF99 and Glide test set used for this comparison are identical and comprised of 120 protein/ligand complexes with fewer than 10 rotatable bonds, 159 complexes with fewer than 20 rotatable bonds, and 172 total complexes. a PMF99 only varies from ePMF99 in its exclusion of 2mcp (4 rotatable bonds) and 1ake (22 rotatable bonds); b PMF04 varies from ePMF99 in its exclusiong of 1byb (24 rotatable bonds), 1frp (7 rotatable bonds), 1hdy (0 rotatable bonds), 1lic (14 rotatable bonds), 1pbd (2 rotatable bonds), 1pph (9 rotatable bonds), 1slt (13 rotatable bonds), 1srj (4 rotatable bonds), 1tpp (5 ro tatable bonds), 1ulb (1 rotatable bond), 1xid (6 rotatable bonds), 2ak3 (6 rotatabl e bonds), 2xis (9 rotatable bonds), 3hvt (1 rotatable bond), 3tpi (7 rotatable bonds), 4fab (4 rotata ble bonds) and 8gch (9 rotatable bonds); c The GOLD test set contains 64 complexes ( 10 rotatable bonds), 78 complexes ( 20 rotatable bonds), and 85 total complexes that are all taken from the larger ePMF99 test set; d FlexX only varies from the ePMF99 test set in its exclusion of 1hdy (0 rotatable bonds). The values exhibited in the table above were calculated fro m values in Table 4-3. One complex that was initia lly poorly reproduced (2.87 rmsd) using the ePMF99 scoring function was that between the anomer of L-arabinose (abe_b) and the arabinose-binding protein (ABP) as visualized in Figure 4-2a. Th is was a particularly surprising result for two reasons. First, this small cyclic monosaccharide has relatively limited conformational freedom. Second, PMF calculations for the anomer of L-arabinose (abe_a) bound to ABP, which is observed in the same crystal structure,56,57 gave a predicted structure, Figure 4-2b, that was 95

PAGE 96

almost identical to the crysta l structure (0.20 rmsd). On in spection, the best docked pose for L-arabinose placed this ligand in an or ientation that was rotated by one atom with respect to the oxygen in the ring in its original crystal position (Figure 4-2a). Given the high proportion of hydrogen bonding groups in the liga nd, it was noticed that the sc oring error was associated primarily with the interactions between L-arabinose and the positively charged side chains of Lys-10 and Arg-151. As discussed previously, this problem is compensated for in the ePMF99 implementation if the nonbonded NC-OD and NC-O E potentials are correctly parameterized. Thus, the hydrogen bonding heavy atoms were pr evented from reaching their optimal hydrogen bonding distances. Adjusting the afor ementioned non-bonded parameters gave docked ligand poses that were almost identical to th e experimental structures for both the and anomers of the ligand (Figure 4-2b and Figure 4-2c, respectively). Figure 4-2. Graphical visualizat ion of optimal ligand poses fo r arabinose in the arabinosebinding protein. (a) Incorrectly docked -anomer of arabinose (turquoise) on the ligand at its crystallographic positio n (fuchsia). (b) Correctly docked -anomer (turquoise) relative to the ligand at its observed position (fuchsia) after optimization of the enhanced PMF scoring function. (c) Correctly docked -anomer (turquoise) relative to the ligand at its observed position (fuchsia) after optimization of the enhanced PMF scoring function. Protein at oms are rendered as ball-and-stick representations (C, grey; H, white, O, red; N, blue; S, yellow). 96

PAGE 97

Comparison to Other Docking/Scoring Algorithms The large test set used in this study was chos en so that a direct and adequate comparison could be made for the ePMF99/LGA dockings wi th other widely accepted and implemented methods for docking.148,160,161,194,195 A comparison was facilitated by the availability of results collected in recent studies of the Glide docking algorithm. In common with other studies, the comparisons are based solely on the rms deviation in the position of the docked ligand relative to that in the original crystal structure. The resu lts show that the ePMF99 scoring function yields docked structures that are as good as those obtained using Glide for fewer than 10 rotatable bonds. This group is the most releva nt to database screening applic ations that have the goal of finding relatively inflexible leads (Table 4-1).165 This same group of complexes vastly outperformed the results of GOLD and FlexX. For all ligands with fewer than 20 rotatable bonds, the ePMF99 scoring potential was found to outperform even Glide. That difference grew even larger for the entire test set, once the complexes that contained ligands with more than 20 rotatable bonds were added. The larger and mo re flexible the ligand, the more the ePMF99 scoring potential outpe rformed Glide, GOLD and FlexX, in general. On the whole, only approximately 34% of the docked complexes had rmsd values greater than the average total rmsd for all of the complexes of about 1.5 Of thos e, 73% of the poorly docked complexes contained metal ions, which PMF99 (or ePMF99) does not explicitly treat, while a separate 7% of the ligands that were poorly docked were primar ily hydrophobic in nature. Because ePMF99 was developed to improve hydrogen bond type intera ction, ligands that are primarily hydrophobic do not benefit from the changes made for the ePMF99 scoring function. This leaves only approximately 20% of the poorly docked ligands w ith no explicit explanation as to their poor results. 97

PAGE 98

The ePMF99 scoring function was able to adeq uately reproduce about 40% of the total metal-containing complexes, of which over half contained Ca2+ ions (which is a commonly added counter-ion resulting in some of the af orementioned ions not being located in the identified active site). The inability of PMF99 and PMF04 to treat hemeand FeS-groups led to the exclusion of all hemeand FeS-containing co mplexes from the test set. Three additional complexes (1imb, 1lna and 3cla) were also excluded from the test set because the metal ions they contained were not accounted for in the PMF atom types. The last eight complexes unaccounted for were not included due to errors in the docking or scoring procedure. Virtual Screening for Thymidine Kinase Having established that the ePMF99 potential performed well in reproducing the structures of known protein/ligand complexes, the next test wa s to examine its behavior in virtual screening studies using thymidine kinase (TK) as a mode l system. This choice was motivated by several factors. First, there are a substantial number of small molecule complexes of this enzyme for Table 4-2. Accuracy in cross-docking of thymid ine kinase inhibitors to the 1kim active site. Bioactivity rms deviation of Best Scoring Pose () Liganda Ki ( M) ePMF99 ePMF99/H2O Glide165 DOCK160 FlexX160 GOLD160 Surflex197 dT 0.2203 0.39 0.39 0.45 0.82 0.78 0.72 0.74 ahiu 0.40 0.52 0.54 1.16 0.88 0.63 0.87 idu 0.09204 0.33 0.55 0.35 9.33 1.03 0.77 1.05 mct 11.4205 0.90 0.65 0.79 7.56 1.11 1.19 0.87 hpt 26.6206 4.57 0.87b 1.58 1.02 4.18 0.49 1.90 dhbt 3.47 1.21 0.68 2.02 3.65 0.93 0.96 hmtt 30.9206 2.78 1.49 2.83 9.62 13.30 2.33 1.78 acv 170.0207 3.56 5.01 4.22 3.08 2.71 2.74 3.51 gcv 48.0208 2.24 2.88 3.19 3.01 6.07 3.11 3.54 pcv 1.5205 3.70 3.18 4.10 4.10 5.96 3.01 3.84 a Literature references for crys tal structures of these ligands bound to thymidine kinase: dT (1kim),178 ahiu (1ki6),208 idu (1ki7),208 mct (1e2k),199 hpt (1e2m),206 dhbt (1e2p),206 hmtt (1e2n),206 acv (2ki5),207 gcv (1ki2),208 pcv (1ki3).208 b Docking studies employed an additional water molecule in the ligand binding site of the TK crystal structure (1kim). See text for details. 98

PAGE 99

which high-resolution crystal structures are available, making it a popular system for validating scoring potentials a nd docking algorithms.160,165,196-198 In addition, these ligands exhibit a relatively wide range of activities as TK substrates or inhibitors (Table 4-2),199-201 thereby permitting an evaluation of the correlation between scoring differences and biological activity. Finally, TK presents a signifi cant challenge for validating in silico virtual screening strategies because (i) the active site is solvent accessible and can undergo conf ormational changes to accommodate a wide variety of structures, (ii) kn own ligands are nucleotides and/or nucleotide analogs, and (iii) the differential participati on of water molecules in mediating binding depending upon the structural class of ligand being docked into the active site.202 Coordinates are available for thymidine kinase complexed wi th ten different ligands.178,199,200,206-208 As in previous studies, and since the coordinates of the protein main chain atoms in all these structures are almost superimposable, the ePMF99 potential and LGA algorithm were used to dock the ten ligands into a single structure (1kim)178 and the X-ray pose of each ligand was inserted into this dT-bound c onformation of the TK active site for evaluation purposes. In preparing the initia l structures for thes e docking studies, the choice was made to remove all crystallographically observed water molecules from the TK active site so as to facilitate comparison of our docking results with those reported in studies employing the Glide software package.165 Since the TK site is designed to bind 2-deoxythymidine (dT), it was anticipated that the six pyrimidine-based inhibi tors (ahiu, mct, dhbt, idu, hmtt, hpt) would dock well within the 1kim structure. In contrast, the remaining three purine derivatives (acv, gcv, pcv) are larger and may require the side chains of th e protein to adopt alternate conformations (Figure 4-3). Initial docking experiments using the ePMF99 potential gave structures that were in good agreement with the crystallographic poses for only the substrate dT and f our related pyrimidine99

PAGE 100

acv O N HO N O O H dT O N HO N N N O HO NH2H gcv O N HO HO N N N O NH2H pcv N HO HO N N N O NH2H N N O O H mct N N O O H dhbt H HO HO O N HO N O O I H idu HO N N O O H hpt H HO N N O O I H O OH HO ahiu hmtt N H N O O H HO NN H O HO HO H O Figure 4-3. Structures of ligands employed in cross-docking studies on thymidine kinase. Compounds are named by the abbreviated ligand names. based inhibitors (ahiu, mct, idu, dhbt) (Table 42). The remaining pyrimidine derivatives (hmtt, hpt), however, differ from dT in lacking the deoxyribose ring and having a bulky substituent connected to C-6 rather than N1 (Figure 4-3). In the case of hmtt, which is the largest of the pyrimidine-derived ligands, the ke y problem that precludes locati ng its optimum pose in the TK active site is most likely associ ated with the conformation of th e Gln-125 side chain in the 1kim protein structure, as noted in similar crossdocking experiments performed using the Glide software package. Thus, the amide side chain of Gln-125 hydrogen bonds to hmtt active site. As 100

PAGE 101

a result, the LGA docking strategy failed to locate a ligand conformation in which the pyridimine ring of this compound forms the observed hydrogen bonding network with the side chains of residues Gln-125 and Arg-176. With this, the perf ormance of the docking/scoring algorithms was examined using an active site structure in which a third crystallographic water was retained for this problematic ligand (hpt), since such interacti ons were reported to be important in previous structural studies.202 Thus, for this ligand, an additional water molecule was placed in the 1kim active site at its position in the 1e2m crystal structure as were the three waters observed in the TK/hmtt complex (1e2n) (Figure 4-4a).206 However, a similar interaction cannot be formed when this active site residue is maintained in the c onformation observed when TK binds to its substrate dT (Figure 4-4b). Other scoring po tentials and docking algorithms al so fail to position this TK inhibitor within the 1kim crystal structure, conf irming the importance of active site flexibility in cross-docking studies. In the case of the hpt ligand, graphical vi sualization of the incorrectly docked structure showed that the bulky C-6 substituent was placed within an active site cavity that accommodates the deoxyribose moiety of dT in 1kim. The ligands hpt, dhbt and hmtt were docked into the resulting active site structure, which contained three water molecules. The third water molecule occupied a region of the active s ite that normally contains the deoxyribose ring of dT and the pyrimidine derivatives (Figure 4-4b). In this case, the use of the ePMF99 potential and LGA docking algorith m yielded a model of the TK/hpt complex with a much lower rmsd value (Table 4-2). Interestingly, the introduction of the third water molecule was not necessary for the docking of the dhbt and hmtt ligands beca use these exhibited in tramolecular hydrogen bonds to N-1 of the pyrimidine moiety thereby pr eventing any rotation of the ligand so as to occupy the cavity for the sugar ring of dT. 101

PAGE 102

Figure 4-4. The effect of water on the binding of hpt and dT with TK. (A ) Representation of the role of water molecules in mediating prot ein/ligand interactions in the complex between hpt and thymidylate kinase after docking the liga nd into the 1kim protein structure. Note that the additional water molecules occupy the space that is usually occupied by the deoxyribose ring of dT in the 1kim crystal structure. (B) The waterligand interactions seen in the TK/hpt complex are abse nt in the cognate complex containing dT, the natural substrate for the en zyme, so that protein residue side chains are directly hydrogen bonde d to the pyrimidine ring. Correlation of the ePMF99 Score and Biological Activity Given that the goal of cross-doc king studies is often to predic t molecular structures that exhibit biological activity, the extent to which the ePMF99 scores for the set of pyrimidine and purine ligands were correlated w ith binding affinity was examined. Although the size of the test set obviously is limited, the data showed a qual itative correlation of the ePMF99 scores and logKi for the pyrimidine-derived ligands (Figure 45). On the other hand, this was not the case for the purine-based analogs, alt hough this likely reflects the f act that these compounds (acv, pcv, gcv) were docked into a rigid active si te optimized for interactions with dT. 102

PAGE 103

Figure 4-5. Plot showing the correl ation of the ePMF99 score vs. logKi for the pyrimidine ( ) and purine ( ) analogs used in the cross-docking studies. Conclusions In summary, these studies sugge st that modification of the PM F99 scoring function so as to treat the repulsion term in a more physic ally reasonable manner permits considerable improvement in reproducing observed crystal struct ures. As noted previously in work employing the Glide algorithm,165 it is important to employ carefully pr epared initial structures if accurate ligand docking is to be attained using the LGA/ePMF99 strategy. Furthermore, the docked ligands of lowest rmsd when compared to the cr ystal structure do, in general, exhibit the best ePMF99 score. Importantly, the use of the ePMF 99 scoring potential yields optimized ligand poses that are comparable to those obtained by the scoring function implemented in Glide for the 103

PAGE 104

104 protein/ligand complexes examined in this study (Table 4-3). In addition, the ePMF99 performed well in cross-docking studies on TK for which extens ive structural and binding data are available for a variety of ligands. As noted in previous studies, the key problems encountered in modeling this system were associated with choosing the ap propriate number of active site water molecules, and the failure to include protein side chain flexibility during the LGA search. Thus, it is important to perform a series of calculations that systematically vary the number of active site waters since large changes in li gand structure result in the creati on of active site voids (when the protein is held rigid) that, if not filled by water, will bias the final docked pose of the ligand. Finally, as reported for the original implementation of the PMF99 scoring potential,169 obtaining a good correlation between the ePMF99 score and observed biological activity when flexible ligands are docked into rigid pr otein cavities requires the compar ison of ligands that do not differ radically in size and/or overall st ructure. On this point, it is noted that this problem does not seem solvable using the available scoring met hods, about which considerable controversy still seems to exist in the literature,148,194 since it primarily arises from failing to include active site flexibility during ligand docking.

PAGE 105

Table 4.3 Complete docking test set and calculated rmsd of docked ligand vs original crystal structure ligand. complex ePMF99 PMF99 PMF04 Glide GOLD FlexX complex ePMF99 PMF99 PMF04 Glide GOLD FlexX 121p (Mg) 5.72 5.66 7.83 1.57 n/a 1.29 1aaq 0.37 0.70 0.70 1.30 12.85 1.75 1abe 0.16 0.25 0.16 0.17 0.86 1.16 1abf 0.19 0.25 0.15 0.20 n/a 1.27 1acj 0.36 0.42 0.99 0.28 4.00 0.49 1acm (Zn) 4.73 5.41 3.28 0.29 0.81 1.39 1aco (FeS) n/a n/a n/a 1.02 0.86 0.96 1aha 0.16 2.74 4.80 0.11 0.51 0.56 1ake 0.39 n/a n/a 3.35 n/a 1.18 1apt 0.54 0.52 0.87 0.58 1.62 1.89 1atl (Zn) 2.79 2.72 4.63 0.94 n/a 2.06 1avd 0.59 3.18 3.46 0.52 n/a 1.22 1azm (Zn) 2.66 2.69 2.74 1.87 2.52 2.37 1baf 0.75 0.96 1.64 0.76 6.12 8.27 1bbp 0.73 8.72 2.24 4.96 n/a 3.75 1bma (Ca) 1.09 6.84 8.45 9.31 n/a 13.41 1byb 0.35 12.49 n/a 10.49 n/a 1.62 1cbs 0.52 0.75 0.84 1.96 n/a 1.68 1cbx (Zn) 1.66 6.29 6.24 0.36 0.54 1. 35 1cde 2.27 2.15 0.60 1.29 n/a 7.45 1cdg (Ca) n/a n/a n/a 3.98 n/a 4.87 1cil (Zn) 1.18 1.18 1.77 3.82 n/a 3.85 1com 0.92 0.60 2.80 3.64 n/a 1.62 1coy 0.68 0.97 1.05 0.28 0.86 1.06 1cps (Zn) 6.59 6.59 3.75 3.00 0.84 0.99 1ctr (Ca) 4.31 5.16 7.05 3.56 n/a 2.82 1dbb 0.30 0.29 7.28 0.41 1.17 0.81 1dbj 0.75 0.75 0.94 0.20 0.72 1.22 1dbk 0.68 0.67 4.33 0.47 n/a 0.76 1dbm 0.72 0.79 3.11 1.97 n/a 2.08 1did (Mn) 4.08 4.91 6.13 3.82 3.72 4.22 1die (Mg) 2.68 2.67 2.91 0.79 1.03 4.71 1dr1 (Ca) 0.33 0.37 0.68 1.47 1.41 5.64 1dwb 3.62 3.19 4.80 0.25 n/a 0.54 1dwc 0.73 0.84 5.38 0.87 n/a 1.19 1dwd 1.24 1.30 5.72 1.32 1.71 1.66 1eap 0.82 1.60 5.24 2.32 3.00 3.72 1eed 0.75 1.97 2.43 5.90 12.43 9.78 1ela (Ca) 4.52 5.15 2.77 1.60 n/a 9.71 1elb (Ca) 4.99 10.57 4.08 4.40 n/a 7.17 1elc (Ca) 6.82 6.94 5.85 8.22 n/a 4.74 1eld (Ca) 4.16 2.40 7.47 0.67 n/a 6.98 1ele (Ca) 4.89 4.32 4.52 2.52 n/a 10. 73 1epb 0.70 1.79 1.94 1.78 2.08 2.77 1eta 2.82 7.62 7.74 2.92 11.21 8.46 1etr 0.99 1.99 6.04 1.48 4.23 7.24 1fen 0.96 1.81 1.86 0.66 n/a 1.39 1fkg 1.10 1.21 2.03 1.25 1.81 7.59 1fki 0.10 0.29 0.14 1.92 0.71 0.59 1frp (Zn) 0.64 0.65 n/a 0.27 n/a 1.89 1ghb 0.37 0.40 5.80 1.89 1.45 1.33 1glp 5.27 7.68 7.36 0.34 n/a 0.47 1glq n/a n/a n/a 0.29 1.35 6.43 1hdc 3.60 4.64 4.39 0.58 10.49 11.74 1hdy (Zn) 2.57 2.65 n/a 1.74 0.94 n/a 1hef 1.05 16.32 6.68 5.30 1.87 15.32 1hfc (Zn) 6.43 3.83 7.90 2.24 n/a 2.51 1hgg n/a n/a n/a 2.10 n/a 10.05 1hgh n/a n/a n/a 0.28 n/a 4.14 1hgi n/a n/a n/a 0.28 n/a 0.97 1hgj n/a n/a n/a 0.18 n/a 3.98 1hri 1.07 2.36 1.25 1.59 14.01 10.23 1hsl (Cd) 0.26 0.26 0.32 1.31 0.97 0.59 1hti 1.74 1.83 1.72 4.40 n/a 1.54 1hvr 0.56 0.65 15.74 1.50 n/a 3.35 1hyt (Zn) 0.73 0.72 3.24 0.28 1.10 1.62 1icn 0.87 3.57 7.97 2.34 8.63 10.52 1ida 1.54 1.51 1.44 11.88 12.12 11.95 105

PAGE 106

Table 4-3. Continued. complex ePMF99 PMF99 PMF04 Glide GOLD FlexX complex ePMF99 PMF99 PMF04 Glide GOLD FlexX 1igj 0.58 10.19 5.32 1.30 9.42 7.17 1imb (Gd) n/a n/a n/a 0.89 n/a 4.71 1ivb (Ca) 0.25 0.75 0.45 4.97 n/a 1.29 1ivc (Ca) 2.24 2.25 1.86 1.94 n/a 2.21 1ivd (Ca) 0.56 3.09 3.42 0.72 n/a 5.42 1ive (Ca) 0.89 1.29 1.57 2.61 2.16 5.34 1ivf (Ca) 0.57 0.55 1.32 0.53 n/a 6.97 1lah 0.15 0.18 0.17 0.13 n/a 0.28 1lcp (Zn) 3.25 11.50 3.96 1.98 n/a 1.65 1ldm 0.45 8.32 8.60 0.30 1.00 0.74 1lic 3.07 3.55 n/a 4.87 10.78 5.07 1lmo 1.00 1.03 2.93 0.93 n/a 4.49 1lna (Co) n/a n/a n/a 0.95 n/a 5.40 1lst 0.14 0.12 0.18 0.14 0.87 0.71 1mbi (heme) n/a n/a n/a 1.68 n/a 0.47 1mcr 1.71 1.82 1.18 4.33 6.23 10.04 1mdr (Mg) 0.97 0.97 1.16 0.52 0.36 0.88 1mld 0.69 1.77 1.25 0.32 n/a 1.45 1mmq (Zn) 7.67 6.10 7.28 0.92 n/a 0.52 1mrg 0.12 0.41 3.99 0.30 n/a 0.81 1mrk 1.12 1.15 1.39 1.20 1.01 3.55 1mup (Cd) 1.42 1.42 2.01 4.37 3.96 3.82 1nco 1.22 9.66 9.84 6.99 n/a 5.85 1nis (FeS) n/a n/a n/a 0.97 4.29 1.41 1nsc (Ca) 0.38 0.38 1.23 1.21 n/a 2.12 1pbd 0.35 0.35 n/a 0.21 0.57 0.33 1pha (heme) n/a n/a n/a 0.69 1.24 n/a 1phd (heme) n/a n/a n/a 1.22 0.85 0.65 1phf (heme) n/a n/a n/a 1.14 n/a 4.23 1phg (heme) n/a n/a n/a 4.32 1.35 4.74 1poc (Ca) 1.12 0.88 1.29 5.09 1.27 9.25 1ppc (Ca) 2.44 5.42 6.45 7.92 n/a 3.05 1pph (Ca) 4.58 6.44 n/a 4.31 n/a 4.91 1ppi (Ca) 0.78 0.82 1.31 6.24 n/a 6.91 1ppk 0.52 0.69 0.69 0.45 n/a 1.54 1ppl 0.51 0.65 0.77 2.82 n/a 5.62 1ppm 0.92 1.15 1.11 0.62 n/a 8.27 1pso 1.19 1.68 9.41 13.10 n/a 1.61 1rbp 0.96 1.33 2.50 0.96 n/a 1.13 1rds 0.86 0.78 5.84 3.75 4.78 4.89 1rne 0.66 1.04 0.64 10.08 2.00 12.24 1rnt 1.09 1.03 1.05 0.72 n/a 1.90 1rob 0.36 0.37 4.41 1.85 3.75 7.70 1slt 1.44 1.62 n/a 0.51 0.78 1.63 1snc (Ca) 0.65 0.66 3.89 1.91 n/a 7.48 1srj 0.28 7.66 n/a 0.58 0.42 2.36 1stp 1.51 1.28 4.42 0.59 0.69 0.65 1tdb 2.49 3.31 4.22 1.46 10.48 10.10 1thy 3.20 2.71 3.56 2.31 n/a 2.67 1tka (Ca) 0.50 0.63 0.46 2.28 1.88 1.17 1tlp (Zn) 1.58 1.70 3.19 1.86 n/a 2.85 1tmn (Ca) 0.91 1.77 2.53 2.80 1.68 0.86 1tng (Ca) 2.64 2.63 6.22 0.19 n/a 1.93 1tnh (Ca) 2.43 2.46 6.46 0.33 n/a 0.56 1tni (Ca) 2.69 2.67 6.29 2.18 n/a 2.71 1tnj (Ca) 2.37 1.87 5.24 0.35 n/a 0.89 1tnk (Ca) 1.44 1.99 5.94 0.87 n/a 1.41 1tnl (Ca) 1.65 1.68 7.71 0.23 n/a 0.71 106

PAGE 107

Table 4-3. Continued. complex ePMF99 PMF99 PMF04 Glide GOLD FlexX complex ePMF99 PMF99 PMF04 Glide GOLD FlexX 1tph 0.57 10.41 8.03 0.20 n/a 1.50 1tpp (Ca) 2.45 2.54 n/a 1.12 0.43 1.11 1trk (Ca) 0.88 0.86 0.36 1.64 n/a 1.57 1tyl (Zn) 1.71 10.87 4.48 1.06 n/a 2.34 1ukz 1.95 4.79 3.26 0.37 n/a 0.94 1ulb 4.68 5.72 n/a 0.28 0.32 3.37 1wap n/a n/a n/a 0.12 n/a 0.57 1xid (Mn) 2.79 4.12 n/a 4.30 0.92 2.01 1xie (Mn) 2.66 2.72 2.46 3.86 0.69 1.94 2ack 0.68 0.71 1.02 0.97 4.99 2.21 2ada (Zn) 0.36 0.45 0.25 0.53 0.40 0. 67 2ak3 1.10 10.85 n/a 0.71 5.08 0.91 2cgr 0.50 0.32 0.40 0.38 0.99 3.53 2cht 0.30 0.48 3.82 0.42 0.59 4.58 2cmd 0.83 13.29 3.82 0.65 n/a 3.75 2cpp (heme) n/a n/a n/a 0.17 n/a 2.94 2ctc (Zn) 1.31 1.85 6.77 1.61 0.32 1.97 2dbl 0.31 3.12 9.04 0.69 1.31 1.49 2fox (4fxn) 0.31 12.94 1.25 0.44 n/a 1.04 2gbp (Zn) 0.18 0.20 0.16 0.15 n/a 0.92 2lgs (Mn) 1.72 4.44 6.13 7.55 n/a 4.63 2mcp 0.99 n/a 1.82 1.30 4.37 2.07 2phh 0.28 0.99 0.23 0.38 0.72 0.43 2pk4 0.40 0.55 1.01 0.86 1.34 1.66 2plv 1.72 1.95 1.78 1.88 13.92 7.85 2r04 0.73 0.86 1.68 0.80 n/a 12.55 2r07 0.71 11.82 12.42 0.48 8.23 11.63 2sim 0.70 1.04 1.57 0.92 0.92 1.99 2tmn (Zn) 1.11 1.11 1.24 0.58 n/a 5.16 2xis (Mg) 1.71 4.45 n/a 0.85 n/a 1.54 2yhx n/a n/a n/a 3.84 1.19 2.25 2ypi 1.04 8.26 1.42 0.31 n/a 1.22 3cla (Co) n/a n/a n/a 8.51 5.45 6.42 3cpa (Zn) 1.71 3.24 3.53 2.40 1.58 2.53 3hvt 0.48 0.78 n/a 0.77 1.12 10.26 3mth (Zn) 3.72 3.75 1.10 5.48 10.12 1.59 3ptb (Ca) 3.24 4.40 4.42 0.27 0.96 0.55 3tpi (Ca) 0.52 0.52 n/a 0.49 0.80 1.07 4aah (Ca) 0.39 0.45 0.93 0.30 0.42 5.93 4cts 0.64 0.66 0.55 0.19 1.57 1.53 4dfr (Ca) 1.12 1.06 1.23 1.12 1.44 1.40 4fab 2.55 5.58 n/a 4.50 5.69 4.95 4fbp 0.27 6.19 8.46 0.56 n/a 1.78 4hmg 0.55 0.66 4.70 0.78 n/a 5.74 4phv 0.32 0.40 0.51 0.38 1.11 1.12 4tim 0.59 0.67 1.47 1.32 n/a 4.09 4tln (Zn) 4.29 5.59 6.50 2.24 n/a 3.68 4tmn (Zn) 0.97 1.00 1.19 1.87 n/a 8.35 4ts1 0.50 6.31 0.97 0.85 n/a 1.41 5abp 0.17 0.37 0.21 0.21 n/a 1.17 5cpp (heme) n/a n/a n/a 0.59 n/a 1.49 5cts 0.59 1.40 0.36 0.28 n/a 11.61 5p2p (Ca) 1.19 1.25 1.13 1.82 1.55 1.00 5tim 2.60 2.61 3.89 0.58 n/a 1.99 5tmn (Zn) 2.47 1.97 2.39 2.43 n/a 4.38 6abp 0.16 0.29 0.24 0.40 1.08 1.12 6cpa (Zn) 0.68 0.85 6.81 4.58 n/a 6.61 6rnt (Ca) 1.39 1.14 8.41 2.22 1.20 4.79 6tim 0.72 0.91 0.91 1.73 n/a 1.60 6tmn (Zn) 2.13 2.19 8.96 2.66 n/a 5.10 7cpa (Zn) 0.93 0.99 1.56 4.14 n/a 9.11 7tim 0.49 0.48 1.45 0.14 0.78 1.49 107

PAGE 108

108Table 4-3. Continued. complex ePMF99 PMF99 PMF04 Glide GOLD FlexX complex ePMF99 PMF99 PMF04 Glide GOLD FlexX 8atc (Zn) 1.96 6.05 2.05 0.37 n/a 0.62 8gch 1.38 6.16 n/a 0.30 0.86 8.91 9hvp 1.01 0.76 1.25 2.68 n/a 15.54

PAGE 109

CHAPTER 5 CONCLUSIONS Concluding Remarks A great deal of progress has been made in identifying potent inhibitors for human AS, however, the absence of a well-defined structure or structures of AS has slowed these efforts. The development of a model of AS-B with the AspAMP intermediate bound at synthetase active site has the potential for aiding th e inhibitor design process in the future. Parameters have been developed in CHARMM for the AspAMP intermediate that allowed for a thoroughly optimized model of AS-B to be obtained. This model represents a significant improvement in the structural re lationship between the active site and the AspAMP intermediate. Significant side chain reorientati on was seen in the model, in which Arg-447 and Glu-348 flipped the positions of their side chains to form new interaction with the AspAMP intermediate. More significantl y, the Glu-348 side chain movement appears to have opened up the intramolecular tunnel that connects the glutamin ase active site with the synthetase active for NH3 translocation. This result coul d explain previous fi nding that suggest that Glu-348 acts as a gate for NH3 translocation and is important in AspAMP intermediate formation. The refinement of this model paves the way for the virtual screening of compound libraries on AS-B. One method that was proven to be very effective in both molecular docking and virtual screening was the LGA search algorithm coupl ed with the ePMF99 scoring function. The improvement made to the ePMF99 scoring function allows or allowed for more accurate representation of the non-bonded repulsion term in comparion to the standard PMF99. This docking and scoring combination was shown to work well when implemented in the CAChe program and could be used for the virtual scr eening of compound libraries on the newly refined AS-B model. 109

PAGE 110

The Future of the AS-B Project from a Computational Perspective The model of AS-B presented in this work paves the way for the refinement of a full quantitative model of the enzyme with an array of possible ligands bound in the active sites. In order to be able to confidently present a full quantitative model of AS-B, a few barriers must still be surpassed. First, a more detailed refinement of the parameters of the AspAMP intermediate should be accomplished. For this to be achieve d, the accurate assignment of the vibrational spectra calculated in CHARMM and via ab initio methods of the model complex is necessary. An accurate assignment of the vibrational spectra w ould assist in directing the refinement of the current parameters, so that the working 2-dimensional energy surface of the model complex calculated by CHARMM (Figure 2-7) would better reproduce the calculated ab initio surface (Figure 2-3). Second, building the final 37 residue s, whose purpose and structure remain unclear, on to the model is a challenging task that would yi eld a completed picture as to the structure of AS-B. Correctly modeling the tail of AS-B would answer the questions that remain concerning the importance of this region of the sequence. It is possible that these residue s play no role in the binding of substrates or mediation of catalysis, ho wever, until experiment or theory definitively answers that question, the impor tance of these residues cannot be taken for granted. A full quantitative model would be a good target for in silico library screening, as well as a good starting point for the docking and parameterizat ion of the known inhibitors of glutaminedependent AS, as seen in Figure 5-1. 110

PAGE 111

Figure 5-1. Structures of th e AS reaction intermediate, AspAMP 1, transition-state 2, and known synthetase site inhibitors. The N-acylsulfonamide 3209 was modeled after the AspAMP 1, where as the sulfoximine 423,210,211 and the sulfoximine derivative 6145 were modeled after the transition-stat e. L-cysteine sulfinic acid (CSA) 5210,212,213 is a competitive inhibitor with respect to aspartate. Modeling the Inhibitors Including the inhibitors in the model of AS-B will require a significant effort due to the need for novel parameters for each unique inhibi tor. The parameterization of the sulfoximine 4 have already begun, and as was detailed in Chapte r 2, the first steps take n were to identify the necessary parameters (Table 5-1) and calculate the first of two time-intensive ab initio PES for the corresponding model complex, seen in Fi gure 5-2. A preliminary CHARMM atom type 111

PAGE 112

Table 5-1. Sulfoximine 4 model complex and CHARMM parameters necessary. Bonds SA-CB PA-NA SA-OD1 SA-NA SA-CG Angles O5-PA-NA O2A-PA-NAa PA-NA-SA NA-SA-CG NA-SA-OD1 NA-SA-CB CG-SA-OD1 CG-SA-CB OD1-SA-CB SA-CG-HG1b SA-CB-HB1c SA-CB-CA HA-CA-HA2d Dihedrals O5-PA-NA-SA O2A-PA-NA-SAe PA-NA-SA-OD1 PA-NA-SA-CG NA-SA-CG-HG1f PA-NA-SA-CB NA-SA-CB-CA NA-SA-CB-HB1g a The O1A-PA-NA angle is included; b All three SA-CG-HG angles; c The SA-CB-HB2 angle is included; d All three HA-CA-HA angles; e The O1A-PA-NA-SA dihe dral is included; f All three NA-SA-CG-HG dihedrals; g The NA-SA-CB-HB2 dihedral is also included. The three dihedrals in bold are the three primary dihedrals of the m odel complex (and sulfoximine) that would be the focus of the parameter development. assignment for the model complex is shown in Table 5-2 with part ial charges calculated according to the CHelpG scheme116 at the HF/6-31+G( d) level in GAUSSIAN 03. This partial charge calculation was performed in the same manne r as the calculation of partial charges for the AspAMP 1 intermediate detailed in Chapter 2. Table 5-2. Initial CHARMM atom types and CH elpG charges assigned for the sulfoximine model complex. Atom Name Atom Type Charge Atom Name Atom Type Charge C5 CN8B 0.21 HG1 HA 0.17 H5 HN8 -0.02 HG2 HA 0.13 H5 HN8 -0.01 HG3 HA 0.16 HC5 HN8 0.01 OD1 OS -0.74 O5 ON2 -0.58 CB CT2 -0.02 PA P 1.70 HB1 HA 0.07 O1A ON3 -0.92 HB2 HA -0.01 O2A ON3 -0.98 CA CT1 -0.08 NA NN1C -0.97 HA HB 0.04 SA SS 1.34 HA2 HB 0.02 CG CT3 -0.53 HA3 HB 0.01 112

PAGE 113

The three consecutive primary dihedral s that are unique to the sulfoximine 4 inhibitor represent the largest barrier to incorporation of the sulfoximine 4 in the AS-B model. Calculating two separate 2-dimensional ab initio energy surfaces for the three ad jacent dihedrals is necessary for the accurate parameterization of this com pound. Figure 5-2 shows th e CA-CB-SA-NA versus CB-SA-NA-PA energy surface as calculated in GAUSSIAN 03 at the HF/6-31+G( d) level, as was done for the AspAMP 1. This energy surface was chosen as the first of the two possible Figure 5-2. The 2-dimensional ab initio energy surface for dihedral angle CB-SA-NA-PA versus dihedral angle CA-CB-SA-NA of the sulfoximine model complex. The global minimum energy (kcal/mol) was set to zero with all other energies offset relatively. 113

PAGE 114

surfaces calculated with the idea that the SANA-PA-O5 dihedral could be modeled off of the CB-O3A-PA-O5 dihedral from AspAMP 1 as an initial gue ss. Ideally, a second ab initio surface should be calculated for the adjacent di hedrals SA-NA-PA-O5 versus CB-SA-NA-PA in order to more accurately parameterize th e primary dihedrals of the sulfoximine 4 inhibitor. Developing a model with the sulfoximine 4 bound to AS-B would yield a stepwise picture of the enzyme as the reaction it catalyzes progr esses from bound substrates (Gln/Asp/ATP) to formations of the intermediate ( AspAMP 1) to tetrahedral transi tion-state (sulfoximine 4 mimics the transition-state) before the release of the final products. These models could then be used in much the same way a series of x-ra y crystal structures would be in showing the progression of catalysis and the critical residues and potential conformational changes involved. Fully refined parameters for any of the inhibito rs would also allow for an additional level of model validation, as a free energy perturbation (FEP) calculation could be run for comparison with the experimental binding constant for the particular inhibitor. Modeling the Residue Mutations The introduction of residue mutations into th e AS-B model is another avenue that should be explored in the future of this project. Seve ral mutations have been made in the AS-B enzyme that have aided in explaining the mechanism of catalysis. Mutations within the AS-B model could also help explain exactly what takes place in the enzyme as a result of the point residue mutations. Any mutation could result in a breakdown of the structure of AS-B or simply a breakdown of the interacti on between AS-B and the substrates or intermediate during catalysis. Table 5-3 lists important residue mutations that have be experimentally tested thus far. 114

PAGE 115

115 Table 5-3. AS-B synthetase domain mutants with the putative relevance Residue Mutations Tested Relevance Cys-523 Ala214 Aspartate binding Arg-325 Ala & Lys215 AspAMP formation Thr-322 Ala, Ser, Val & Tyr215 Aspartate binding or AspAMP formation Thr-323 Ala, Ile, Leu, Ser & Val215 Aspartate binding or AspAMP formation Ser-346 Ala & Thr147 Structurally important Glu-348 Ala, Asp & Gln147 AspAMP formation or ac tive site coordination Glu-352 Ala, Asp & Gln147 Stabalizing the aspartate -amino group Asp-384 Ala & Asn147 Catalysis Arg-387 Ala & Lys147 Catalysis Lys-449 Ala & Arg145 ATP binding or catalysis Without structural validation, it is difficult to prove unequivoca lly the specific role of any single residue. The mutations shown in Table 5-3 have not had their associated relevance proven. Glu-348 is the most well-characterized, as Dr Jemy Gutierrez was able to validate the importance of this residue through 18O-labeled transfer studies and 31P NMR assays. Molecular dynamics simulations of single mutations of the model would allow for a more definitive structural picture of how these mutations affect the structure of AS-B. They would also shed light on the how these mutations affect th e interaction of the protein with the AspAMP or aspartate, as many of these residues are believ ed to be important in substrate binding or intermediate formation.

PAGE 116

APPENDIX CHARMM SIMULATED ANNEALING EXAMPLE INPUT FILES Heating Input File GENERATED BY CHARMM-GUI (http://www.charmm-gui.org) on Apr, 24. 2009. INPUT FILE FOR EQUILIBRATION OF SOLVATED GLOBULAR PROTEIN !Read topology open read unit 10 card name top_all27_prot_na.rtf read rtf unit 10 card !Read parameters open read unit 20 card name par_all27_prot_na.prm read para unit 20 card !Stream in BAA top/par stream baa_stream.str !Read PSF and Coordinates open read unit 10 card name baa_model_mini_shake.psf read psf unit 10 card open read unit 10 card name baa_model_mini_shake.crd read coor unit 10 card Setup PBC (Periodic Boundary Condition) stream step3_pbcsetup.str open read unit 10 card name crystal_image.str CRYSTAL DEFINE @XTLtype @A @B @C @alpha @beta @gamma CRYSTAL READ UNIT 10 CARD !Image centering by residue IMAGE BYRESID XCEN 0.0 YCEN 0.0 ZCEN 0.0 sele resname TIP3 end IMAGE BYRESID XCEN 0.0 YCEN 0.0 ZCEN 0.0 sele ( segid @posid .or. segid @negid ) end Nonbonded Options nbonds atom vatom vfswitch bycb ctonnb 10.0 ctofnb 12.0 cutnb 16.0 cutim 16.0 inbfrq -1 imgfrq -1 wmin 1.0 cdie eps 1.0 ewald pmew fftx @fftx ffty @ffty fftz @fftz kappa .34 spline order 6 shake bonH para take reference values for bond lengths from param file !!**************** heating to 600K !!**************** 116

PAGE 117

! estimate Pmass from SYSmass (total system mass) [there could be problems with exreme values, such as Pmass << SYSmass or Pmass >> SYSmass scalar mass stat calc Pmass = int ( ?stot / 50.0 ) open write card unit 31 name "heat_600k_baa.rst" open write file unit 32 name "heat_600k_baa.dcd" DYNA CPT start time 0.001 nstep 240000 iseed 31415 firstt 0.0 finalt 600.0 teminc 5.0 ihtfrq 2000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ichecw 1 twindl -5.0 twindh +5.0 iasors 0 isvfrq 500 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunwrit 31 iuncrd 32 kunit -1 nsavc 100 !ie update nonbond list when any atom has move write out heated coordinates in CHARMM format open write unit 41 card name "baa_heat_600k.crd" write coor card unit 41 write out heated coordinates in pdb format open write unit 42 card name "baa_heat_600k.pdb" write coor pdb unit 42 STOP Equilibration Input File GENERATED BY CHARMM-GUI (http://www.charmm-gui.org) on Apr, 24. 2009. INPUT FILE FOR EQUILIBRATION OF SOLVATED GLOBULAR PROTEIN !Read topology open read unit 10 card name top_all27_prot_na.rtf read rtf unit 10 card !Read parameters open read unit 20 card name par_all27_prot_na.prm read para unit 20 card !Stream in BAA top/par stream baa_stream.str !Read PSF and Coordinates open read unit 10 card name baa_model_mini_shake.psf read psf unit 10 card open read unit 10 card name baa_model_mini_shake.crd read coor unit 10 card Setup PBC (Periodic Boundary Condition) stream step3_pbcsetup.str 117

PAGE 118

open read unit 10 card name crystal_image.str CRYSTAL DEFINE @XTLtype @A @B @C @alpha @beta @gamma CRYSTAL READ UNIT 10 CARD !Image centering by residue IMAGE BYRESID XCEN 0.0 YCEN 0.0 ZCEN 0.0 sele resname TIP3 end IMAGE BYRESID XCEN 0.0 YCEN 0.0 ZCEN 0.0 sele ( segid @posid .or. segid @negid ) end Nonbonded Options nbonds atom vatom vfswitch bycb ctonnb 10.0 ctofnb 12.0 cutnb 16.0 cutim 16.0 inbfrq -1 imgfrq -1 wmin 1.0 cdie eps 1.0 ewald pmew fftx @fftx ffty @ffty fftz @fftz kappa .34 spline order 6 We need to keep the water mol ecules rigid, and also want to keep covalent X-H bonds fixed, so use SHAKE shake bonH para take reference values for bond lengths from param file !!********************** equilibrate at 600K !!********************** estimate Pmass from SYSmass (total system mass) [there could be problems with exreme values, such as Pmass << SYSmass or Pmass >> SYSmass scalar mass stat calc Pmass = int ( ?stot / 50.0 ) open read card unit 31 name "heat_600k_baa.rst" open write card unit 32 name "eq_600k_baa.rst" open write file unit 33 name "eq_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100000 isvfrq 500 nprint 100 iprfrq 500 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 600.0 tmass 2000.0 firstt 600.0 finalt 600.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move write out equilibrated coordinates in CHARMM format open write unit 41 card name "baa_equi_600k.crd" write coor card unit 41 write out equilibrated coordinates in pdb format open write unit 42 card name "baa_equi_600k.pdb" write coor pdb unit 42 STOP Annealing Input File GENERATED BY CHARMM-GUI (http://www.charmm-gui.org) on Apr, 24. 2009. INPUT FILE FOR EQUILIBRATION OF SOLVATED GLOBULAR PROTEIN 118

PAGE 119

!Read topology open read unit 10 card name top_all27_prot_na.rtf read rtf unit 10 card !Read parameters open read unit 20 card name par_all27_prot_na.prm read para unit 20 card !Stream in BAA top/par stream baa_stream.str !Read PSF and Coordinates open read unit 10 card name baa_model_mini_shake.psf read psf unit 10 card open read unit 10 card name baa_model_mini_shake.crd read coor unit 10 card Setup PBC (Periodic Boundary Condition) stream step3_pbcsetup.str open read unit 10 card name crystal_image.str CRYSTAL DEFINE @XTLtype @A @B @C @alpha @beta @gamma CRYSTAL READ UNIT 10 CARD !Image centering by residue IMAGE BYRESID XCEN 0.0 YCEN 0.0 ZCEN 0.0 sele resname TIP3 end IMAGE BYRESID XCEN 0.0 YCEN 0.0 ZCEN 0.0 sele ( segid @posid .or. segid @negid ) end Nonbonded Options nbonds atom vatom vfswitch bycb ctonnb 10.0 ctofnb 12.0 cutnb 16.0 cutim 16.0 inbfrq -1 imgfrq -1 wmin 1.0 cdie eps 1.0 ewald pmew fftx @fftx ffty @ffty fftz @fftz kappa .34 spline order 6 We need to keep the water mol ecules rigid, and also want to keep covalent X-H bonds fixed, so use SHAKE shake bonH para take reference values for bond lengths from param file estimate Pmass from SYSmass (total system mass) [there could be problems with exreme values, such as Pmass << SYSmass or Pmass >> SYSmass scalar mass stat calc Pmass = int ( ?stot / 50.0 ) !!******************************************* Annealing down to 300K & 10 Quenching steps !!******************************************* !Step1 119

PAGE 120

open read card unit 31 name "eq_600k_baa.rst" open write card unit 32 name "run1_600k_baa.rst" open write file unit 33 name "run1_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 600.0 tmass 2000.0 firstt 600.0 finalt 600.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move open read card unit 31 name "run1_600k_baa.rst" open write card unit 32 name "ann1_600k_baa.rst" open write file unit 33 name "ann1_600k_baa.dcd" open write file unit 34 name "ann1_600k_baa.dvl" open write card unit 35 name "ann1_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 firstt 600.0 finalt 570.0 teminc -0.6 tstruc 600.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format open write unit 41 card name "baa_ann-final1.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final1.pdb" write coor pdb unit 42 !Step2 open read card unit 31 name "ann1_600k_baa.rst" open write card unit 32 name "run2_600k_baa.rst" open write file unit 33 name "run2_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 570.0 tmass 2000.0 firstt 570.0 finalt 570.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move open read card unit 31 name "run2_600k_baa.rst" open write card unit 32 name "ann2_600k_baa.rst" open write file unit 33 name "ann2_600k_baa.dcd" open write file unit 34 name "ann2_600k_baa.dvl" open write card unit 35 name "ann2_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 120

PAGE 121

firstt 570.0 finalt 540.0 teminc -0.6 tstruc 570.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format open write unit 41 card name "baa_ann-final2.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final2.pdb" write coor pdb unit 42 !Step3 open read card unit 31 name "ann2_600k_baa.rst" open write card unit 32 name "run3_600k_baa.rst" open write file unit 33 name "run3_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 540.0 tmass 2000.0 firstt 540.0 finalt 540.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move open read card unit 31 name "run3_600k_baa.rst" open write card unit 32 name "ann3_600k_baa.rst" open write file unit 33 name "ann3_600k_baa.dcd" open write file unit 34 name "ann3_600k_baa.dvl" open write card unit 35 name "ann3_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 firstt 540.0 finalt 510.0 teminc -0.6 tstruc 540.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format open write unit 41 card name "baa_ann-final3.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final3.pdb" write coor pdb unit 42 !Step4 open read card unit 31 name "ann3_600k_baa.rst" open write card unit 32 name "run4_600k_baa.rst" open write file unit 33 name "run4_600k_baa.dcd" 121

PAGE 122

DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 510.0 tmass 2000.0 firstt 510.0 finalt 510.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move open read card unit 31 name "run4_600k_baa.rst" open write card unit 32 name "ann4_600k_baa.rst" open write file unit 33 name "ann4_600k_baa.dcd" open write file unit 34 name "ann4_600k_baa.dvl" open write card unit 35 name "ann4_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 firstt 510.0 finalt 480.0 teminc -0.6 tstruc 510.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format open write unit 41 card name "baa_ann-final4.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final4.pdb" write coor pdb unit 42 !Step5 open read card unit 31 name "ann4_600k_baa.rst" open write card unit 32 name "run5_600k_baa.rst" open write file unit 33 name "run5_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 480.0 tmass 2000.0 firstt 480.0 finalt 480.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move open read card unit 31 name "run5_600k_baa.rst" open write card unit 32 name "ann5_600k_baa.rst" open write file unit 33 name "ann5_600k_baa.dcd" open write file unit 34 name "ann5_600k_baa.dvl" open write card unit 35 name "ann5_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 firstt 480.0 finalt 450.0 teminc -0.6 tstruc 480.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 122

PAGE 123

inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format open write unit 41 card name "baa_ann-final5.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final5.pdb" write coor pdb unit 42 !Step6 open read card unit 31 name "ann5_600k_baa.rst" open write card unit 32 name "run6_600k_baa.rst" open write file unit 33 name "run6_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 450.0 tmass 2000.0 firstt 450.0 finalt 450.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move open read card unit 31 name "run6_600k_baa.rst" open write card unit 32 name "ann6_600k_baa.rst" open write file unit 33 name "ann6_600k_baa.dcd" open write file unit 34 name "ann6_600k_baa.dvl" open write card unit 35 name "ann6_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 firstt 450.0 finalt 420.0 teminc -0.6 tstruc 450.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format open write unit 41 card name "baa_ann-final6.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final6.pdb" write coor pdb unit 42 !Step7 open read card unit 31 name "ann6_600k_baa.rst" open write card unit 32 name "run7_600k_baa.rst" open write file unit 33 name "run7_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 123

PAGE 124

HOOVER reft 420.0 tmass 2000.0 firstt 420.0 finalt 420.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move open read card unit 31 name "run7_600k_baa.rst" open write card unit 32 name "ann7_600k_baa.rst" open write file unit 33 name "ann7_600k_baa.dcd" open write file unit 34 name "ann7_600k_baa.dvl" open write card unit 35 name "ann7_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 firstt 420.0 finalt 390.0 teminc -0.6 tstruc 420.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format open write unit 41 card name "baa_ann-final7.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final7.pdb" write coor pdb unit 42 !Step8 open read card unit 31 name "ann7_600k_baa.rst" open write card unit 32 name "run8_600k_baa.rst" open write file unit 33 name "run8_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 390.0 tmass 2000.0 firstt 390.0 finalt 390.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move open read card unit 31 name "run8_600k_baa.rst" open write card unit 32 name "ann8_600k_baa.rst" open write file unit 33 name "ann8_600k_baa.dcd" open write file unit 34 name "ann8_600k_baa.dvl" open write card unit 35 name "ann8_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 firstt 390.0 finalt 360.0 teminc -0.6 tstruc 390.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format 124

PAGE 125

open write unit 41 card name "baa_ann-final8.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final8.pdb" write coor pdb unit 42 !Step9 open read card unit 31 name "ann8_600k_baa.rst" open write card unit 32 name "run9_600k_baa.rst" open write file unit 33 name "run9_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 360.0 tmass 2000.0 firstt 360.0 finalt 360.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move open read card unit 31 name "run9_600k_baa.rst" open write card unit 32 name "ann9_600k_baa.rst" open write file unit 33 name "ann9_600k_baa.dcd" open write file unit 34 name "ann9_600k_baa.dvl" open write card unit 35 name "ann9_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 firstt 360.0 finalt 330.0 teminc -0.6 tstruc 360.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format open write unit 41 card name "baa_ann-final9.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final9.pdb" write coor pdb unit 42 !Step10 open read card unit 31 name "ann9_600k_baa.rst" open write card unit 32 name "run10_600k_baa.rst" open write file unit 33 name "run10_600k_baa.dcd" DYNA CPT restart time 0.001 nstep 100 isvfrq 100 nprint 100 iprfrq 100 ntrfrq 100 PCONst pref 1.0 pmass @Pmass pgamma 20.0 HOOVER reft 330.0 tmass 2000.0 firstt 330.0 finalt 330.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunread 31 iunwrit 32 iuncrd 33 kunit -1 nsavc 100 !ie update nonbond list when any atom has move 125

PAGE 126

open read card unit 31 name "run10_600k_baa.rst" open write card unit 32 name "ann10_600k_baa.rst" open write file unit 33 name "ann10_600k_baa.dcd" open write file unit 34 name "ann10_600k_baa.dvl" open write card unit 35 name "ann10_600k_baa.ene" DYNA CPT start time 0.001 nstep 50000 firstt 330.0 finalt 300.0 teminc -0.6 tstruc 330.0 ihtfrq 1000 PCONst pref 1.0 pmass @Pmass pgamma 20.0 ieqfrq 0 ichecw 0 twindl -5.0 twindh +5.0 iasors 0 nprint 100 iprfrq 500 ntrfrq 100 inbfrq -1 ihbfrq 0 imgfrq -1 -1 for inbfrq/imgfrq means heuristic update iunrea 31 iunwrit 32 iuncrd 33 iunvel 34 kunit 35 nsavc 100 !ie update nonbond list when any atom has move write out Annealed coordinates in CHARMM format open write unit 41 card name "baa_ann-final10.crd" write coor card unit 41 write out Annealed coordinates in pdb format open write unit 42 card name "baa_ann-final10.pdb" write coor pdb unit 42 STOP Final Minimization Input File GENERATED BY CHARMM-GUI (http://www.charmm-gui.org) on Apr, 24. 2009. INPUT FILE FOR EQUILIBRATION OF SOLVATED GLOBULAR PROTEIN !Read topology open read unit 10 card name top_all27_prot_na.rtf read rtf unit 10 card !Read parameters open read unit 20 card name par_all27_prot_na.prm read para unit 20 card !Stream in BAA top/par stream baa_stream.str open read card unit 30 name baa_model_mini_shake.psf read psf card unit 30 open read card unit 10 name baa_ann-final10.pdb read coor pdb unit 10 resid Setup PBC (Periodic Boundary Condition) stream step3_pbcsetup.str open read unit 10 card name crystal_image.str CRYSTAL DEFINE @XTLtype @A @B @C @alpha @beta @gamma CRYSTAL READ UNIT 10 CARD 126

PAGE 127

127 !Image centering by residue IMAGE BYRESID XCEN 0.0 YCEN 0.0 ZCEN 0.0 sele resname TIP3 end IMAGE BYRESID XCEN 0.0 YCEN 0.0 ZCEN 0.0 sele (segid @posid .or. segid @negid ) end Nonbonded Options nbonds atom vatom vfswitch bycb ctonnb 10.0 ctofnb 12.0 cutnb 16.0 cutim 16.0 inbfrq -1 imgfrq -1 wmin 1.0 cdie eps 1.0 ewald pmew fftx @fftx ffty @ffty fftz @fftz kappa .34 spline order 6 energy coor copy comp We need to keep the water mol ecules rigid, and also want to keep covalent X-H bonds fixed, so use SHAKE shake bonH para take reference values for bond lengths from param file !!!!! MINIMIZATION allow water to adjust around protein cons harm force 50.0 sele segid asb end mini SD nstep 1000 tolgrd 0.01 cons harm force 20.0 sele segid asb end mini CONJ nstep 2500 tolgrd 0.01 cons harm force 10.0 sele segid asb end mini CONJ nstep 5000 tolgrd 0.01 cons harm force 5.0 sele segid asb end mini CONJ nstep 10000 tolgrd 0.001 cons harm force 0.0 sele segid asb end mini ABNR nstep 100000 tolgrd 0.00001 coor orient rms mass write out heated coordinates in CHARMM format open write unit 41 card name "baa_ann-final10_mini.crd" write coor card unit 41 write out heated coordinates in pdb format open write unit 42 card name "baa_ann-final10_mini.pdb" write coor pdb unit 42 STOP

PAGE 128

LIST OF REFERENCES (1) Cancer Facts and Figures, 2009, American Cancer Society, 2009. (2) Cooney, D. A.; Handschumacher, R. E. Annu. Rev. Pharmacol. 1970, 10, 421. (3) Swain, A. L.; Jaskolski, M.; Housset, D.; Rao, J. K. M.; Wlodawer, A. Proc. Natl. Acad. Sci. U. S. A. 1993, 90, 1474. (4) Aghaiypour, K.; Wlodawer, A.; Lubkowski, J. Biochemistry 2001, 40, 5655. (5) Tallal, L.; Tan, C.; Oettgen, H.; Wollner, N.; McCarthy, M.; Helson, L.; Burchenal, J.; Karnofsky, D.; Murphy, M. L. Cancer 1970, 25, 306. (6) Sutow, W. W.; Garcia, F.; Starling, K. A.; Williams, T. E.; Lane, D. M.; Gehan, E. A. Cancer 1971, 28, 819. (7) Terebelo, H. R.; Anderson, K.; Wiernik, P. H.; Cuttner, J.; Cooper, R. M.; Faso, L.; Berenberg, J. L. Am. J. Clin. Oncol. 1986, 9, 411. (8) 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. (9) Kiriyama, Y.; Kubota, M.; Takimoto, T.; K itoh, T.; Tanizawa, A.; Akiyama, Y.; Mikawa, H. Leukemia 1989, 3, 294. (10) Lobel, J. S.; O'Brien, R. T.; McIn tosh, S.; Aspnes, G. T.; Capizzi, R. L. Cancer 1979, 43, 1089. (11) Capizzi, R. L.; Bertino, J. R.; Skeel, R. T.; Creasey, W. A.; Zanes, R.; Olayon, C.; Peterson, R. G.; Handschumacher, R. E. Ann. Intern. Med. 1971, 74, 893. (12) Chakrabarti, R.; Schuster, S. M. Int. J. Pediatr. Hematol. Oncol. 1997, 4, 597. (13) Pieters, R.; Klumper, E.; Kaspers, G. J.; Veerman, A. J. Crit. Rev. Oncol. Hematol. 1997, 25, 11. (14) Haskell, C. M.; Canellos, G. P. Biochem. Pharmacol. 1969, 18, 2578. (15) Martin, J. K.; Sun, W.; Moraga-Amador, D.; Schuster, S. M.; Wylie, D. E. Amino Acids 1993, 5, 51. (16) Hutson, R. G.; Kitoh, T.; Amador, D. A. M. ; Cosic, S.; Schuster, S. M.; Kilberg, M. S. Am. J. Physiol. 1997, 272, C1691. (17) Aslanian, A. M.; Fletcher, B. S.; Kilberg, M. S. Biochem. J. 2001, 357, 321. 128

PAGE 129

(18) Aslanian, A. M.; Kilberg, M. S. Biochem. J. 2001, 358, 59. (19) Minowada, J.; Onuma, T.; Moore, G. E. J. Natl. Cancer Inst. 1972, 49, 891. (20) Verma, N.; Kumar, K.; Kaur, G.; Anand, S. Crit. Rev. Biotechnol. 2007, 27, 45. (21) Richards, N. G. J.; Kilberg, M. S. Annu. Rev. Biochem. 2006, 75, 629. (22) Su, N.; Pan, Y.-X.; Zhou, M.; Harvey Rich ard, C.; Hunger Stephen, P.; Kilberg Michael, S. Pediatr. Blood Cancer 2008, 50, 274. (23) Gutierrez, J. A.; Pan, Y.-X.; Koroniak, L.; Hiratake, J.; Kilberg, M. S.; Richards, N. G. J. Chem. Biol. 2006, 13, 1339. (24) Milman, H. A.; Cooney, D. A. Biochem. J. 1979, 181, 51. (25) Hinchman, S. K.; Schuster, S. M. Protein Eng. 1992, 5, 279. (26) Cedar, H.; Schwartz, J. H. J. Biol. Chem. 1969, 244, 4112. (27) Cedar, H.; Schwartz, J. H. J. Biol. Chem. 1969, 244, 4122. (28) Scofield, M. A.; Lewis, W. S.; Schuster, S. M. J. Biol. Chem. 1990, 265, 12895. (29) Horowitz, B.; Meister, A. J. Biol. Chem. 1972, 247, 6708. (30) Patterson, M. K., Jr.; Orr, G. R. J. Biol. Chem. 1968, 243, 376. (31) Boehlein, S. K.; Richards, N. G. J.; Schuster, S. M. J. Biol. Chem. 1994, 269, 7450. (32) Zalkin, H.; Smith, J. L. Adv. Enzymol. Relat. Areas Mol. Biol. 1998, 72, 87. (33) Smith, J. L. Biochem. Soc. Trans. 1995, 23, 894. (34) Smith, J. L. Curr. Opin. Struct. Biol. 1998, 8, 686. (35) Tso, J. Y.; Zalkin, H.; Van Cleemput, M.; Yanofsky, C.; Smith, J. M. J. Biol. Chem. 1982, 257, 3525. (36) Badet-Denisot, M. A.; Rene, L.; Badet, B. Bull. Soc. Chim. Fr. 1993, 130, 249. (37) Badet, B.; Vermoote, P.; Haumont, P. Y.; Lederer, F.; Le Goffic, F. Biochemistry 1987, 26, 1940. (38) Vanoni, M. A.; Curti, B. Cell. Mol. Life Sci. 1999, 55, 617. 129

PAGE 130

(39) Vanoni, M. A.; Edmondson, D. E.; Rescigno, M.; Zanetti, G.; Curti, B. Biochemistry 1991, 30, 11478. (40) van den Heuvel Robert, H. H.; Ferrari, D. ; Bossi Roberto, T.; Ravasio, S.; Curti, B.; Vanoni Maria, A.; Florencio Francisco, J.; Mattevi, A. J. Biol. Chem. 2002, 277, 24579. (41) Boehlein, S. K.; Schuster, S. M.; Richards, N. G. J. Biochemistry 1996, 35, 3031. (42) Andrulis, I. L.; Chen, J.; Ray, P. N. Mol. Cell. Biol. 1987, 7, 2435. (43) Hongo, S.; Sato, T. Biochimica et Biophysica Acta, Protein Structure and Molecular Enzymology 1983, 742, 484. (44) Luehr, C. A.; Schuster, S. M. Archives of Biochemistry and Biophysics 1985, 237, 335. (45) Dang, V.-D.; Valens, M.; Bolotin -Fukuhara, M.; Daignan-Fornier, B. Mol. Microbiol. 1996, 22, 681. (46) Reitzer, L. J.; Magasanik, B. J. Bacteriol. 1982, 151, 1299. (47) Mehlhaff, P. M.; Schuster, S. M. Arch. Biochem. Biophys. 1991, 284, 143. (48) Mehlhaff, P. M.; Luehr, C. A.; Schuster, S. M. Biochemistry 1985, 24, 1104. (49) Markin, R. S.; Luehr, C. A.; Schuster, S. M. Biochemistry 1981, 20, 7226. (50) Ciustea, M.; Gutierrez, J. A.; Abbatiell o, S. E.; Eyler, J. R.; Richards, N. G. J. Arch. Biochem. Biophys. 2005, 440, 18. (51) Fresquet, V.; Thoden, J. B.; Holden, H. M.; Raushel, F. M. Bioorg. Chem. 2004, 32, 63. (52) Tesson, A. R.; Soper, T. S.; Ciustea, M.; Richards, N. G. J. Arch. Biochem. Biophys. 2003, 413, 23. (53) Boehlein, S. K.; Stewart, J. D.; Walworth, E. S.; Thirumoorthy, R.; Richards, N. G. J.; Schuster, S. M. Biochemistry 1998, 37 13230. (54) Rognes, S. E. Phytochemistry (Elsevier) 1975, 14, 1975. (55) Hongo, S.; Sato, T. Arch. Biochem. Biophys. 1985, 238, 410. (56) Li, K. K.; Beeson, W. T.; Ghiv iriga, I.; Richards, N. G. J. Biochemistry 2007, 46, 4840. (57) 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. 130

PAGE 131

(58) Richards, N. G. J.; Schuster, S. M. Adv. Enzymol. Relat. Areas Mol. Biol. 1998, 72, 145. (59) Bork, P.; Koonin, E. V. Proteins: Struct., Funct., Genet. 1994, 20, 347. (60) Hinchman, S. K.; Heni koff, S.; Schuster, S. M. J. Biol. Chem. 1992, 267, 144. (61) Nakatsu, T.; Kato, H.; Oda, J. i. Nat. Struct. Biol. 1998, 5, 15. (62) Krahn, J. M.; Kim, J. H.; Burns, M. R.; Parry, R. J.; Zalkin, H.; Smith, J. L. Biochemistry 1997, 36, 11061. (63) Muchmore, C. R. A.; Krahn, J. M.; Kim, J. H.; Zalkin, H.; Smith, J. L. Protein Sci. 1998, 7, 39. (64) Chen, S.; Tomchick, D. R.; Wolle, D.; Hu, P.; Smith, J. L.; Switzer, R. L.; Zalkin, H. Biochemistry 1997, 36, 10718. (65) Kim, J. H.; Krahn, J. M.; Tomchick, D. R.; Smith, J. L.; Zalkin, H. J. Biol. Chem. 1996, 271, 15549. (66) Smith, J. L.; Zaluzec, E. J.; Wery, J. P.; Niu, L.; Switzer, R. L.; Zalkin, H.; Satow, Y. Science (Washington, DC, U. S.) 1994, 264, 1427. (67) Tesmer, J. J. G.; Klem, T. J.; Deras, M. L.; Davisson, V. J.; Smith, J. L. Nat. Struct. Biol. 1996, 3, 74. (68) Karlberg, T.; Collins, R.; van den Berg, S.; Flores, A.; Hammarstroem, M.; Hoegbom, M.; Holmberg Schiavone, L.; Uppenberg, J. Acta Crystallogr., Sect D: Biol. Crystallogr. 2008, D64 279. (69) Lemke, C. T.; Howell, P. L. Structure (Cambridge, MA, U. S.) 2001, 9, 1153. (70) Mougous, J. D.; Lee, D. H.; Hubbard, S. C. ; Schelle, M. W.; Vocadlo, D. J.; Berger, J. M.; Bertozzi, C. R. Mol. Cell 2006, 21 109. (71) Sun, M.; Leyh, T. S. Biochemistry 2006, 45, 11304. (72) Miller, M. T.; Bachmann, B. O.; Townsend, C. A.; Rosenzweig, A. C. Nat. Struct. Biol. 2001, 8, 684. (73) Bachmann, B. O.; Townsend, C. A. Biochemistry 2000, 39, 11187. (74) McNaughton, H. J.; Thirkettle, J. E.; Zha ng, Z.; Schofield, C. J.; Jensen, S. E.; Barton, B.; Greaves, P. Chemical Communications (Cambridge) 1998, 2325. 131

PAGE 132

(75) Waterman, D. G.; Ortiz-Lombardia, M.; Fogg, M. J.; Koonin, E. V.; Antson, A. A. J. Mol. Biol. 2006, 356, 97. (76) Mueller, E. G.; Palenchar, P. M. Protein Sci. 1999, 8, 2424. (77) Denessiouk, K. A.; Johnson, M. S. Proteins: Struct., Funct., Genet. 2000, 38, 310. (78) Ding, Y. Experimental and Theoretical Characterization of Asparagine Synthetase Inhibitors., University of Florida, 2002. (79) Hurd, R. E.; John, B. K. J. Magn. Reson. 1991, 91, 648. (80) Schnizer, H. G.; Boehlein, S. K.; Stewart, J. D.; Richards, N. G. J.; Schuster, S. M. Methods Enzymol. 2002, 354, 260. (81) Schnizer, H. G.; Boehlein, S. K.; Stewart, J. D.; Richards, N. G. J.; Schuster, S. M. Biochemistry 1999, 38, 3677. (82) Miles, B. W.; Raushel, F. M. Biochemistry 2000, 39, 5051. (83) Myers, R. S.; Amaro, R. E.; Luthey-Schulten, Z. A.; Davisson, V. J. Biochemistry 2005, 44, 11974. (84) Yount, R. G. Adv. Enzymol. Relat. Areas Mol. Biol. 1975, 43, 1. (85) Mouilleron, S.; Golinelli-Pimpaneau, B. Curr. Opin. Struct. Biol. 2007, 17, 653. (86) Mouilleron, S.; Badet-Denisot, M.-A.; Golinelli-Pimpaneau, B. J. Biol. Chem. 2006, 281, 4404. (87) van den Heuvel, R. H. H.; Svergun, D. I.; Petoukhov, M. V.; Coda, A.; Curti, B.; Ravasio, S.; Vanoni, M. A.; Mattevi, A. J. Mol. Biol. 2003, 330 113. (88) Boehlein, S. K.; Rosa-Rodriguez, J. G.; Schuster, S. M.; Richards, N. G. J. J. Am. Chem. Soc. 1997, 119, 5785. (89) MacKerel, A. D., Jr.; Brooks Iii, C. L. ; Nilsson, L.; Roux, B.; Won, Y.; Karplus, M. In CHARMM: The Energy Function and Its Para meterization with an Overview of the Program ; John Wiley & Sons: Chichester, 1998; Vol. 1; pp 271. (90) Brooks, B. R.; Bruccoleri, R. E.; Olafson, D. J.; States, D. J.; Swaminathan, S.; Karplus, M. J. Comput. Chem. 1983, 4, 187. (91) Alder, B. J.; Wainwright, T. E. J. Chem. Phys. 1957, 27, 1208. (92) Alder, B. J.; Wainwright, T. E. J. Chem. Phys. 1959, 31, 459. 132

PAGE 133

(93) McCammon, J. A.; Gelin, B. R.; Karplus, M. Nature (London, U. K.) 1977, 267, 585. (94) Metropolis, N.; Ulam, S. J. Am. Stat. Assoc. 1949, 44, 335. (95) Metropolis, N.; Rosenbluth, A. W.; Rose nbluth, M. N.; Teller, A. H.; Teller, E. J. Chem. Phys. 1953, 21, 1087. (96) Weiner, S. J.; Kollman, P. A.; Nguyen, D. T.; Case, D. A. J. Comput. Chem. 1986, 7, 230. (97) Mohamadi, F.; Richards, N. G. J.; Guida, W. C.; Liskamp, R.; Lipton, M.; Caufield, C.; Chang, G.; Hendrickson, T.; Still, W. C. J. Comput. Chem. 1990, 11, 440. (98) Vinter, J. G.; Davis, A.; Saunders, M. R. J. Comput. Aided Mol. Des. 1987, 1, 31. (99) Daura, X.; Mark, A. E.; Van Gunsteren, W. F. J. Comput. Chem. 1998, 19, 535. (100) Allinger, N. L.; Yuh, Y. H.; Lii, J. H. J. Am. Chem. Soc. 1989, 111, 8551. (101) Halgren, T. A. J. Comput. Chem. 1996, 17, 490. (102) Jorgensen, W. L.; Tirado-Rives, J. J. Am. Chem. Soc. 1988, 110, 1657. (103) Clark, M.; Cramer, R. D., III.; Van Opdenbosch, N. J. Comput. Chem. 1989, 10, 982. (104) MacKerell, A. D., Jr.; Bashford, D.; Bello tt, M.; Dunbrack, R. L.; Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.; Joseph-McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F. T. K.; Mattos, C.; Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E., III; Roux, B.; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub, J.; Watanabe, M.; Wiorkiewicz-Kuczera, J.; Yin, D.; Karplus, M. J. Phys. Chem. B 1998, 102, 3586. (105) Schlenkrich, M.; Brickmann, J.; Ma ckerell, A. D., Jr.; Karplus, M. Biological Membranes 1996, 31. (106) MacKerell, A. D., Jr.; Wior kiewicz-Kuczera, J.; Karplus, M. J. Am. Chem. Soc. 1995, 117, 11946. (107) Reiling, S.; Schlenkrich, M.; Brickmann, J. J. Comput. Chem. 1996, 17, 450. (108) Schmidt, J. M.; Brueschweiler, R.; Ernst, R. R.; Dunbrack, R. L., Jr.; Joseph, D.; Karplus, M. J. Am. Chem. Soc. 1993, 115, 8747. (109) Feng, M.-H.; Philippopoulos, M.; MacKerell, A. D., Jr.; Lim, C. J. Am. Chem. Soc. 1996, 118, 11265. 133

PAGE 134

(110) Lazaridis, T.; Karplus, M. Science (Washington, D. C.) 1997, 278, 1928. (111) Neria, E.; Fischer, S.; Karplus, M. J. Chem. Phys. 1996, 105, 1902. (112) Reiher, W. E., III. Theoretical Studies of Hydrogen Bonding, Harvard University, 1985. (113) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J. Chem. Phys. 1983, 79, 926. (114) Gaussian 03, R. C., Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, Jr., J. A. ; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S. ; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; and Pople, J. A.; Gaussian, Inc., Wallingford CT, 2004. (115) Pavelites, J. J.; Gao, J.; Ba sh, P. A.; Mackerell, A. D., Jr. J. Comput. Chem. 1997, 18, 221. (116) Breneman, C. M.; Wiberg, K. B. J. Comput. Chem. 1990, 11, 361. (117) Scott, A. P.; Radom, L. J. Phys. Chem. 1996, 100, 16502. (118) Vaiana, A. C.; Cournia, Z.; Costescu, I. B.; Smith, J. C. Comput. Phys. Commun. 2005, 167, 34. (119) Fiser, A.; Do, R. K. G.; Sali, A. Protein Sci. 2000, 9, 1753. (120) Fagerberg, T.; Cerottini, J.-C.; Michielin, O. J. Mol. Biol. 2006, 356, 521. (121) Doll, K.; Schoen, J. C.; Jansen, M. Journal of Physics: Conference Series 2008, 117, No pp given. (122) Fernandez-Lima, F. A.; Wei, H.; Gao, Y. Q.; Russell, D. H. J. Phys. Chem. A 2009, 113 8221. (123) Villani, V. J. Mol. Struct. 2004, 672, 17. 134

PAGE 135

(124) Hixson, C. A.; Wheeler, R. A. J. Chem. Theory Comput. 2009, 5, 1883. (125) Collura, V.; Higo, J.; Garnier, J. Protein Sci. 1993, 2, 1502. (126) Rapp, C. S.; Friesner, R. A. Proteins: Struct., Funct., Genet. 1999, 35, 173. (127) Rognan, D.; Zimmermann, N.; Jung, G.; Folkers, G. Eur. J. Biochem. 1992, 208, 101. (128) Ota, N.; Agard, D. A. J. Mol. Biol. 2001, 314, 607. (129) Bruccoleri, R. E.; Karplus, M. Biopolymers 1990, 29, 1847. (130) Feig, M.; Karanicolas, J.; Brooks, C. L. J. Mol. Graphics Modell. 2004, 22 377. (131) Kirkpatrick, S.; Gelatt, C. D.; Vecchi, M. P. Science 1983, 220, 671. (132) ern, V. Journal of Optimization Theory and Applications 1985, 45, 41. (133) Leach, A. R. Reviews in Computational Chemistry 1991, 2, 1. (134) Geman, S.; Geman, D. IEEE Transactions on Pattern Anal ysis and Machine Intelligence (PAMI) 1984, 6, 721. (135) Steinbach, P. J.; Brooks, B. R. Chem. Phys. Lett. 1994, 226, 447. (136) Gordon, H. L.; Kwan, W. K.; Gong, C.; Larrass, S.; Rothstein, S. M. J. Chem. Phys. 2003, 118, 1533. (137) Cheatham, T. E., III.; Miller, J. L. ; Fox, T.; Darden, T. A.; Kollman, P. A. J. Am. Chem. Soc. 1995, 117, 4193. (138) Essmann, U.; Perera, L.; Berkowitz, M. L.; Darden, T.; Lee, H.; Pedersen, L. G. J. Chem. Phys. 1995, 103, 8577. (139) Feller, S. E.; Pastor, R. W.; Ro jnuckarin, A.; Bogusz, S.; Brooks, B. R. J. Phys. Chem. 1996, 100, 17011. (140) Rapaport, D. C. The Art of Molecular Dynamics Simulation; Cambridge University Press: Cambridge, 1995. (141) Cramer, C. J. Essentials of Computational Chemistry ; John Wiley & Sons: New York, 2002. (142) Nose, S. J. Chem. Phys. 1984, 81, 511. 135

PAGE 136

(143) Hoover, W. G. Phys. Rev. A 1985, 31, 1695. (144) Ryckaert, J.-P.; Ciccotti, G.; Berendsen, H. J. C. J. Comput. Phys. 1977, 23, 327. (145) Ikeuchi, H.; Meyer, M. E.; Ding, Y.; Hiratake, J.; Richards, N. G. J. Bioorg. Med. Chem., In Press, Corrected Proof (146) DeLano, W. L. The PyMOL Molecular Gra phics System.; DeLano Scientific LLC: Palo Alto, CA, USA., 2008. (147) Gutierrez, J. A. Inhibition and Functiona l Characterization of As paragine Synthetase, University of Florida, 2006. (148) Cummings, M. D.; DesJarlais, R. L. ; Gibbs, A. C.; Mohan, V.; Jaeger, E. P. J. Med. Chem. 2005, 48, 962. (149) Kitchen, D. B.; Decornez, H.; Furr, J. R.; Bajorath, J. Nat. Rev. Drug Discovery 2004, 3, 935. (150) Ewing, T. J. A.; Makino, S.; Skillman, A. G.; Kuntz, I. D. J. Comput. Aided Mol. Des. 2001, 15, 411. (151) Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R. J. Mol. Biol. 1997, 267, 727. (152) Rarey, M.; Kramer, B.; Lengauer, T.; Klebe, G. J. Mol. Biol. 1996, 261, 470. (153) Abagyan, R.; Totrov, M.; Kuznetsov, D. J. Comput. Chem. 1994, 15, 488. (154) Abagyan, R.; Totrov, M. Curr. Opin. Chem. Biol. 2001, 5, 375. (155) Welch, W.; Ruppert, J.; Jain, A. N. Chem. Biol. 1996, 3, 449. (156) McMartin, C.; Bohacek, R. S. J. Comput. Aided Mol. Des. 1997, 11, 333. (157) Murray, C. W.; Baxter, C. A.; Frenkel, A. D. J. Comput. Aided Mol. Des. 1999, 13, 547. (158) McGann, M. R.; Almond, H. R.; Nicholls, A.; Grant, J. A.; Brown, F. K. Biopolymers 2003, 68, 76. (159) Venkatachalam, C. M.; Jia ng, X.; Oldfield, T.; Waldman, M. J. Mol. Graphics Modell. 2003, 21, 289. (160) Bissantz, C.; Folkers, G.; Rognan, D. J. Med. Chem. 2000, 43, 4759. (161) Charifson, P. S.; Corkery, J. J.; Murcko, M. A.; Walters, W. P. J. Med. Chem. 1999, 42, 5100. 136

PAGE 137

(162) Stahl, M.; Rarey, M. J. Med. Chem. 2001, 44, 1035. (163) Bursulaya, B. D.; Totrov, M.; Abagyan, R.; Brooks, C. L., III. J. Comput. Aided Mol. Des. 2003, 17, 755. (164) Taylor, R. D.; Jewsbury, P. J.; Essex, J. W. J. Comput. Aided Mol. Des. 2002, 16, 151. (165) 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.; Shenkin, P. S. J. Med. Chem. 2004, 47, 1739. (166) Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart, W. E.; Belew, R. K.; Olson, A. J. J. Comput. Chem. 1998, 19, 1639. (167) Tominaga, Y.; Jorgensen, W. L. J. Med. Chem. 2004, 47, 2534. (168) Hann, M. M.; Leach, A. R.; Harper, G. J. Chem. Inf. Comput. Sci. 2001, 41 856. (169) Muegge, I.; Martin, Y. C. J. Med. Chem. 1999, 42, 791. (170) Muegge, I. Perspectives in Drug Discovery and Design 2000, 20, 99. (171) Muegge, I. J. Comput. Chem. 2001, 22, 418. (172) Muegge, I. J. Med. Chem. 2006, 49, 5895. (173) Gohlke, H.; Hendlich, M.; Klebe, G. J. Mol. Biol. 2000, 295, 337. (174) DeWitte, R. S.; Shakhnovich, E. I. J. Am. Chem. Soc. 1996, 118, 11733. (175) Purvis, G. D., III. Calculating a potential of mean force (PMF) score of a protein-ligand complex; Fujitsu America, I., Ed. USA, 2005; pp 27. (176) Berman, H. M.; Westbrook, J.; Feng, Z. ; Gilliland, G.; Bhat, T. N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E. Nucleic Acids Res. 2000, 28, 235. (177) Halgren, T. A. J. Comput. Chem. 1999, 20, 720. (178) Brown, D. G.; Visse, R.; Sandhu, G.; Davies, A.; Rizkallah, P. J.; Melitz, C.; Summers, W. C.; Sanderson, M. R. Nat. Struct. Biol. 1995, 2, 876. (179) Michalewicz, Z. Genetic Algorithms + Data Structures = Evolution Programs, 3rd ed.; Springer-Verlag: New York, 1999. (180) Solis, F. J.; Wets, R. J. B. Mathematics of Operations Research 1981, 6, 19. 137

PAGE 138

(181) Muegge, I. Med. Chem. Res. 1999, 9, 490. (182) Muegge, I.; Rarey, M. Reviews in Computational Chemistry 2001, 17, 1. (183) Ha, S.; Andreani, R.; Robbins, A.; Muegge, I. J. Comput. Aided Mol. Des. 2000, 14, 435. (184) Muegge, I.; Martin, Y. C.; Hajduk, P. J.; Fesik, S. W. J. Med. Chem. 1999, 42, 2498. (185) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M., Jr.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A. J. Am. Chem. Soc. 1995, 117, 5179. (186) Bowen, J. P.; Allinger, N. L. Reviews in Computational Chemistry 1991, 2 81. (187) Kaminski, G. A.; Friesner, R. A. ; Tirado-Rives, J.; Jorgensen, W. L. J. Phys. Chem. B 2001, 105, 6474. (188) Fersht, A. R.; Shi, J. P.; Knill-Jones, J. ; Lowe, D. M.; Wilkinson, A. J.; Blow, D. M.; Brick, P.; Carter, P.; Waye, M. M. Y.; Winter, G. Nature (London, U. K.) 1985, 314, 235. (189) Fersht, A. R. Trends Biochem. Sci. 1987, 12, 301. (190) Mills, J. E. J.; Dean, P. M. J. Comput. Aided Mol. Des. 1996, 10, 607. (191) Thornton, J. M.; MacArthur, M. W.; McDonal d, I. K.; Jones, D. T.; Mitchell, J. B. O.; Nandi, L.; Price, S. L.; Zvelebil, M. J. J. M. Philosophical Transactions of the Royal Society of London, Series A: Mathematical, Physical and Engineering Sciences 1993, 345, 113. (192) Kellenberger, E.; Rodrig o, J.; Muller, P.; Rognan, D. Proteins: Struct., Funct., Bioinf. 2004, 57, 225. (193) Xing, L.; Hodgkin, E.; Liu, Q.; Sedlock, D. J. Comput. Aided Mol. Des. 2004, 18, 333. (194) Wang, R.; Lu, Y.; Wang, S. J. Med. Chem. 2003, 46, 2287. (195) Perola, E.; Walters, W. P.; Charifson, P. S. Proteins: Struct., Funct., Bioinf. 2004, 56, 235. (196) Jain, A. N. J. Med. Chem. 2004, 47, 947. (197) Jain, A. N. J. Med. Chem. 2003, 46, 499. (198) Verdonk, M. L.; Taylor, R. D.; Chessari, G.; Murray, C. W. I llustration of Current Challenges in Molecular Docking. In Structure-Based Drug Discovery 2007; pp 201. 138

PAGE 139

(199) Prota, A.; Vogt, J.; Pilger, B.; Perozzo, R. ; Wurth, C.; Marquez, V. E.; Russ, P.; Schulz, G. E.; Folkers, G.; Scapozza, L. Biochemistry 2000, 39, 9597. (200) Vogt, J.; Perozzo, R.; Pautsch, A.; Prota, A.; Schelling, P.; Pilger, B.; Folkers, G.; Scapozza, L.; Schulz, G. E. Proteins: Struct., Funct., Genet. 2000, 41, 545. (201) Russ, P.; Schelling, P.; Scapozza, L.; Fo lkers, G.; De Clercq, E.; Marquez, V. E. J. Med. Chem. 2003, 46, 5045. (202) Spadola, L.; Novellino, E.; Folkers, G.; Scapozza, L. Eur. J. Med. Chem. 2003, 38, 413. (203) Pilger, B. D.; Perozzo, R.; Alber, F.; Wurth, C.; Folkers, G.; Scapozza, L. J. Biol. Chem. 1999, 274, 31967. (204) Balzarini, J.; De Clercq, E.; Baumgartner, H.; Bodenteich, M.; Griengl, H. Mol. Pharmacol. 1990, 37, 395. (205) Schelling, P.; Folkers, G.; Scapozza, L. Anal. Biochem. 2001, 295, 82. (206) Wurth, C.; Kessler, U.; Vogt, J.; Schul z, G. E.; Folkers, G.; Scapozza, L. Protein Sci. 2001, 10, 63. (207) Bennett, M. S.; Wien, F.; Champness, J. N.; Batuwangala, T.; Rutherford, T.; Summers, W. C.; Sun, H.; Wright, G.; Sanderson, M. R. FEBS Lett. 1999, 443, 121. (208) Champness, J. N.; Bennett, M. S.; Wien, F.; Visse, R.; Summers, W. C.; Herdewijn, P.; De Clercq, E.; Ostrowski, T.; Jarvest, R. L.; Sanderson, M. R. Proteins: Struct., Funct., Genet. 1998, 32, 350. (209) Koroniak, L.; Ciustea, M.; Gutie rrez, J. A.; Richards, N. G. J. Org. Lett. 2003, 5, 2033. (210) Boehlein, S. K.; Nakatsu, T.; Hiratake, J. ; Thirumoorthy, R.; Stewart, J. D.; Richards, N. G. J.; Schuster, S. M. Biochemistry 2001, 40, 11168. (211) Koizumi, M.; Hiratake, J.; Nakatsu, T.; Kato, H.; Oda, J. I. J. Am. Chem. Soc. 1999, 121, 5799. (212) Foote, J.; Lipscomb, W. N. J. Biol. Chem. 1981, 256, 11428. (213) Parmentier, L. E.; Weiss, P. M.; O'L eary, M. H.; Schachman, H. K.; Cleland, W. W. Biochemistry 1992, 31, 6577. (214) Boehlein, S. K.; Walworth, E. S.; Schuster, S. M. Biochemistry 1997, 36, 10168. (215) Boehlein, S. K.; Walworth, E. S.; Richards, N. G. J.; Schuster, S. M. J. Biol. Chem. 1997, 272, 12384. 139

PAGE 140

BIOGRAPHICAL SKETCH Robert N. Humkey was born in Lubbock, Texas, and grew up in the small town of Versailles, Kentucky. He spent a very active youth with his parents and his two younger brothers and younger sister. In 1998, Robert began his under graduate studies in biochemistry at the University of San Diego. During his senior ye ar, an internship with a local pharmaceutical company introduced Robert to a higher level of research-driven chemistry that inspired him to pursue a doctorate in computati onal biochemistry. In 2002, he be gan the pursuit of his doctoral degree at the University of Florida, under the gu idance of Dr. Nigel G. J. Richards. Roberts graduate research was focused on using comput ational methods for drug discovery. While in pursuit of his doctorate, Robert also completed an MBA with concentrations in competitive strategy, marketing and finance from the Univer sity of Floridas Hough Graduate School of Business. 140