sign in sign up
<<
Home , AMPAC, Homoquinolinic acid, Philanthotoxin, Willardiine

INFORMATION TO USERS

This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6 " x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

University Microfilms international A Bell & Howell Information Company 300 North Zeeb Road. Ann Arbor. Ml 48106-1346 USA 313/761-4700 800/521-0600

Order Number 9211140

P a rt 1 . Design, synthesis, and structure-activity studies of molecules with activity at non-NMDA glutamate receptors: Hydroxyphenylalanines, quinoxalinediones and related molecules. P a rt 2 . Computational studies on the interaction of the dopamine amino group and amino group replacements with carboxylate anions as a model for receptor interactions

Hill, Ronald Alan, Ph.D. The Ohio State University, 1991

Copyright ©1991by Hill, Ronald Alan. All rights reserved.

300 N. Zeeb Rd. Ann Arbor, MI 48106

PART 1: DESIGN, SYNTHESIS, AND STRUCTURE-ACTIVITY STUDIES

OF MOLECULES WITH ACTIVITY AT NON-NMDA GLUTAMATE RECEPTORS: HYDROXYPHENYLALANINES, QUINOXALINEDIONES

AND RELATED MOLECULES PART 2: COMPUTATIONAL STUDIES ON THE INTERACTION OF THE DOPAMINE AMINO GROUP AND AMINO GROUP

REPLACEMENTS WITH CARBOXYLATE ANIONS AS A MODEL FOR RECEPTOR INTERACTIONS

DISSERTATION

Presented in Partial Fullfillment of the Requirement for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

By Ronald A. Hill, B.S.Chem.

* * * * *

The Ohio State University

1991

Dissertation Committee: Approved by Duane D. Miller, Ph.D.

Robert W. Brueggemeier, Ph.D. Lane J. Wallace, Ph.D. Jan K. Labanowski, Ph.D. Duane D. Miller, Adviser College of Pharmacy Copyright by Ronald Alan Hill 1991 To Shelley, Sara, Laura, and Eric, and to the Chicken Pox, without which my graduate career would have been infinitely less stimulating... ACKNOWLEDGEMENTS

I would like to express my thanks to the following people: Professor Duane D. Miller, for his guidance and encouragement, and for his forbearance in allowing me to pursue many intellectual interests, to be creative (occasionally), make mistakes (more often), and in the process of it all become a better scientist. Dr. Jan K. Labanowski, also for his monumental guidance and patience, and for his good sense of humor which was often a necessity. The members of my dissertation committee, Dr. Miller, Dr. Labanowski, Dr. Breuggemeier, and Dr. Wallace. Professors Lane J. Wallace and Norman J. Uretsky, for many inspiring pharmacology discussions and tolerance of my many naive questions and observations. Dr. Jan Andzelm of Cray Research, Inc., without whose collaboration Part 2. of this thesis would be of substantially less worth. Dr. Lane Wallace, Rick Layer, Henry Tai, David Weinstein, Donna Supko, and David Willins, for their work on the pharmacological evaluation of the many compounds submitted to them. Dr. Shaun D. Black, who graciously allowed me the use of his laboratory, computers, and instrumentation, and supplied large quantities of high purity water for our HPLC systems. Jack Fowble and John Miller, who were "instrumental" in my chemistry efforts, and Bruce Posey, who provided key computer support. Joan Dandrea and Carol Settles, whose contributions are far too numerous to mention here. Kathy Brooks, Karen Patrick, and Jessica Pritchard, for their hard work in securing my paychecks (very important!), assisting with travel to meetings, and answering difficult questions. David Heisterberg and other staff members at the Ohio Supercomputer Center who provided assistance on the dopamine project and grudgingly put up with my frequent overindulgence in disk storage space on the Cray. Professors Robert Curley, Shaun Black, Raymond Doskotch, Robert Breuggemeier, Carter Olson, Alfred Staubus, Lane Wallace, Norman Uretsky, Jessie Au, Larry Robertson, Donald Witiak, Albert Soloway, and Dennis Feller, for sharing their knowledge, giving advice, providing support and assistance in securing scholarships and a job, and for making the College of Pharmacy a good place to team and do research. My research group colleagues and partners in crime, Seoung-Soo Hong, Yasser Ab-del Ghani, Katsuhisa Matsumoto, Jeff Christoff, Marc Harrold, Men Slavika, Akihiko Hamada, Yoshiya Amemiya, Vimon Tantishaiyakul, Mike Smar, and Carl Neidert, for their knowledge, encouragement, companionship, assistance, advice, support, and patience. My students, for making me a better teacher and a better person. The Ohio State University and the American Foundation for Pharmaceutical Education, without whose financial support this endeavor would have been impossible. Most importantly, my wife, who grudgingly accepted the necessity of this venture in the greater workings of die universe and took on many of my responsibilities in managing a household; my children, who somehow endured it all; and my parents, for their support and encouragement. VITA

March 30,1960 Bom - Athens, Ohio, USA. June-August 1980 Chemistry Assistant (GS-IV) Physical Sciences Laboratory Newark Aerospace Guidance and Metrology Center, Newark, OH May-August 1981 Chemist The Arcanum Corporation, Ann Arbor, MI May, 1982 B.S. Chemistry, University of Michigan, Ann Arbor, Michigan. May 1982 - Sep. 1986 Research Chemist The Upjohn Company, Kalamazoo, MI Sep. 1986 - Aug. 1987 University Fellow, The Ohio State University, Columbus, OH Sep. 1987 - Aug. 1988 Graduate Teaching Fellow The Ohio State University Sep. 1988 - Present Fellow of the American Foundation for Pharmaceutical Education and Graduate Research Associate, The Ohio State University

Sep. 1991 - Present Assistant Professor of Medicinal Chemistry School of Pharmacy, Northeast Louisiana University Monroe, LA PUBLICATIONS

Ronald A. Hill, Jan K. Labanowski, David J. Heisterberg, Duane D. Miller, "Formic Acid: Methylamine Complex Studied by the Hartree-Fock and Density Functional Approach", in Density Functional Methods in Chemistry (Springer-Verlag, New York, 1991) pp. 357-372.

R. A. Hill, B. G. Snider, and T. W. Rosanske, "A widely applicable automated sampling apparatus for dissolution testing", International Journal of Pharmaceutics, 36 (1987) 175-183. S. R. Cox, E. L. Harrington, R. A. Hill, V. J. Capponi, and A. C. Shah, "Bioavailability Studies with Ciglitazone in Beagles. 1. Effect of a Meal on the Bioavailability of Three Ciglitazone Dosage Forms", Biopharm. and Drug Dispos., 6 (1985) 67-80. FIELD OF STUDY Major Field: Pharmacy Studies in: Organic Medicinal Chemistry - Professor Duane D. Miller Computational Chemistry - Professors Duane D. Miller and Dr. Jan K. Labanowski TABLE OF CONTENTS

DEDICATION ii

ACKNOWLEDGEMENTS iii VITA v TABLE OF CONTENTS vii LIST OF TABLES x LIST OF FIGURES xiii LIST OF SCHEMES xviii PART 1: DESIGN, SYNTHESIS, AND STRUCTURE-ACTIVITY STUDIES OF MOLECULES WITH ACTIVITY AT AMPA GLUTAMATE RECEPTORS: HYDROXYPHENYLALANINES AND RELATED COMPOUNDS

CHAPTER I. INTRODUCTION 2 1.1 Background 2 1.2 Endogenous excitatory transmission 4 1.2.1 Identification of EAA 4 1.2.2 Receptor classification, physiology, and distribution 8 1.2.3 Metabolism of excitatory amino acids 19 1.2.4 Storage, release, and reuptake of EAA transmitters 2 2 1.3 Structure-activity relationships 24 1.3.1 NMDA receptors 24 a. Agonists 24 b. Competetive antagonists 26 c. Noncompetetive antagonists 28 1.3.2 Kainate receptors 30 a. Agonists 30 b. Antagonists 31 1.3.3 AP4 receptors 32 1.3.4 Metabotropic (ACPD) receptors 32 1.4 Excitatory amino acid 33 1.5 Therapeutic potential for excitatory amino acid modulation 37

H. STATEMENT OF THE PROBLEM AND OBJECTIVES 39 vii HI. CHEMISTRY 53 3.1 Hydroxyphenylalanines and related compounds 53 3.1.1 Substituted o- 53 3.1.2 Mononitro- and dinitro-w-tyrosines 63 3.1.3 Chiral synthesis of dinitro-o- 70 3.2 and 5-nitrowillardiine 71 3.3 2,3-Quinoxalinediones 75 3.4 pKa studies 93

IV. DISCUSSION OF BIOLOGICAL FINDINGS 98 4.1 Hydroxyphenylalanines and related compounds 98 4.2 Structure-activity studies of 1,4-dihydro- 2,3-quinoxalinediones 108 4.3 Potential link to flavin metabolism 112 V. EXPERIMENTAL 114

5.1 General procedures 114 5.2 pKa determinations 115 5.3 Synthesis of hydroxyphenylalanines and related compounds 116 5.4 Synthesis of 2,3-quinoxalinediones and related compounds 131

PART 2: COMPUTATIONAL STUDIES ON THE INTERACTION OF THE DOPAMINE AMINO GROUP AND AMINO GROUP REPLACEMENTS WITH CARBOXYLATE ANIONS AS A MODEL FOR RECEPTOR INTERACTIONS VI. INTRODUCTION AND STATEMENT OF THE PROBLEM 151

6.1 Dopamine pharmacology 151 6.2 Elements of a hypothesis that the charged form of dopamine is involved in binding to a carboxylate group in D2 receptors 156 6.3 Rationale for computational studies 161 6.4 Summary of objectives 169 VII. PRELIMINARY STUDIES AND HARTREE-FOCK THEORY 172 7.1 Introduction 172 7.2 Ab initio Hartree-Fock theory 173 7.3 Computational details 180 7.4 Results and discussion 181

Vm. PROTON AFFINITIES AND THE DENSITY FUNCTIONAL THEORY 195 8.1 Introduction 195 8.2 The density functional theory 197 viii 8.3 Computational details 201 8.4 Results and discussion 202 8.4.1 Molecular geometries 202 8.4.2 Proton affinities 208 IX. HYDROGEN BONDING IN FORMIC AC1D:METHYLAMINE COMPLEXES: COMPARISON OF HARTREE-FOCK AND DENSITY FUNCTIONAL APPROACHES 216 9.1 Introduction 216 9.2 Computational methods 218 9.2.1 General 218 9.2.2 FAMA complex 1 219 9.2.3 FAMA complex 2 222 9.2.4 I^M" bifurcated complex 223 9.3 Results and discussion 223 9.3.1 FAMA complex 1 223 9.3.2 FAMA complex 2 235 9.3.3 F+M' bifurcated complex 237 9.3.4 Summary of formic acid - methylammonium interactions 245

LIST OF REFERENCES 252 APPENDICES

A. UV ABSORBANCE MAXIMA AND ANALYTICAL WAVELENGTHS FOR pKa DETERMINATIONS 279 B.NONLIN 84 SUBROUTINE FOR pKa FITTING 280

ix LIST OF TABLES TABLE

1. Amino acid content of whole rat

2. Inhibition of [ 3H]-AMPA specific binding at 10 "4 M for various hydroxyphenylalanine analogs

3. Inhibition of specific [ 3H]-AMPA and [3H]-KAIN binding by various compounds

4. Inhibition of [ 3H]-AMPA and [3H]-CNQX specific binding by various quinoxaline-2,3-diones

5. Inhibition of [ 3H]-CNQX specific binding by riboflavin metabolites

6 . Inhibition of [ 3H]-acetycholine release from rat striatal slices by dopamine analogs 7. Apparent equilibrium binding dissociation constants for dopamine and analogs and the effect of NaCl

8 . Activity of chlorpromazine and analogs in D 2 receptor systems

9. Activity of sulpiride and analogs in D 2 receptor systems

10. Relative activity of acetylcholine homologs at muscarinic receptors; Relative activity of noradrenaline homologs at a- and p-adrenoceptors

11. Effect of basis set on various parameters calculated for acetate anion

12. Basis sets and calculation times for acetate anion 13. Effect of basis set on HOMO, LUMO, and total energies, and partial charges

14. Effect of basis set on selected bond lengths and angles in ethylammonium

15. Effect of basis set on computation times, energies, and partial charges for ethyltrimethylammonium 16. Effect of basis set on bond lengths, angles, and torsions for ethyltrimethylanunonium 17. Calculated and experimental proton affinities for formate and methylamine 18. Selected geometric parameters for formate, formic acid, and protonated formic acid 19. Selected geometric parameters for acetate, acetic acid, and protonated acetic acid 20. Selected geometric parameters for methanol and protonated methanol 21. Selected geometric parameters for ethanol and protonated ethanol 22. Selected geometric parameters for methylamine and methylammonium 23a. Electronic and vibrational energies 23b. Electronic and vibrational energies 23c. Electronic and vibrational energies 24. Vibrational frequency calculations for methanol using DFT (cm"1) 25. Vibrational frequencies calculated for methylamine using DFT (cm'1): effect of displacement distance and differencing method in DMol

26. Calculated vs. experimental proton affinities for various molecules 27. 3~21G*(*) optimized geometry for FAMA Complex 1 28. Total energies as a function of H5-N6 distance for FAMA Complex 1

29. Results of calculations with different DFT approaches 30. Comparison of energies and optimized geometries of FAMA Complex 1 calculated by MP2, HF, DGauss and DMol 31. Comparison of energies and optimized geometries of formic acid and methylamine calculated by MP2 and HF methods, DGauss (DG-1 basis) and DMol 32. Comparison of energies and optimized geometries of FAMA Complex 2 calculated by CADPAC, DGauss, and DMol 33. Comparison of energies and optimized geometries of formate:methylammonium bifurcated complex calculated by CADPAC, DGauss, and DMol 34. Summary of energies for the FAMA system LIST OF FIGURES FIGURES PAGE

1. Miscellaneous compounds discussed in the text 6 2. Kynurenate path of tryptophan metabolism 9

3. Prototypical compounds for EAA receptor subtypes 10

4. First-generation NMDA antagonists 11 5. Comparison of NMDA, QUIS and KAIN responses in CA1 hippocamapal pyramidal neurone 12

6 . Voltage-current relationships of EAA channels 14

7. Metabolic transformations of glutamate 2 0

8 . Alternate metabolic pathways to glutamate 21 9. Glutamate storage, release, and reuptake 23 10. NMDA agonists 25

11. Competetive NMDA antagonists 27

12. Noncompetetive NMDA antagonists acting at the PCP/MK-801 site 28

13. antagonists 29 14. Kainate agonists 30 15. Nonselective kainate antagonists 31 16. Compounds active at "AP4 receptors" 32

17, Compounds active at "metabotropic receptors" 34

18. AMPA antagonists 41 19. Glycine antagonists 42

20. Ring-modified analogs of quisqualate 44 21. SAR of AMPA analogs 45 xiii 22. Conformationally-restricted AMPA analogs 23. Comparison of the anionic structures of some AMPA agonists 24. Some substituted phenols with increased acidity 25. Glutamate homologs 26. Potential affinity and photoaffinity analogs of the 2,3- series 27. Rationale for the synthesis of "QXAA’s"

28. Chromatographic separation of nitro-o-tyrosines and degradation products in aqueous nitration

29.250 MHz !H-NMR spectrum of compounds 87 and 8 8

30. High-resolution El mass spectrum of compound 8 8

31. 500 MHz 2D 13C INADEQUATE spectrum of compound 8 8 32. Chromatographic separation of the three dinitro-m-tyrosines 90,91, and 92 33. Preparative reversed-phase chromatographic separation of dinitro-m-tyrosines 91 and 92

34. Reverse-phase HPLC of reaction mixture from Strecker synthesis of 131 35. Titration curve for the pKa determination for 3’-nitro-L-tyrosine using a UV spectrophotometric method 36. Comparison of ionization constants for phenolic groups in o-cresols and congeneric amino acids

37. Hammett plot for ionization of quinoxaline-2,3- diones 38. Comparison of the anionic structures of some AMPA agonists 39. Comparison of dinitro-o-tyrosine with 5-substituted willardiines

40. Proposed model for the AMPA receptor 41. Effect of AMPA receptor antagonists injected into the ventral pallidum on hypermotility elicited by systemic of stimulants 42. 6-azido-7-fluoroQX as a photoaffinity label 110 43. Some pathways for riboflavin metabolism 113 44. Partitioning of dopaminergic drugs into the receptor site 161 45. Chromatography of phenethylammonium and trimethylphenethylammonium on end-capped C- 8 packing 162

46. Helical wheel model of the ml mAChR 163 47. Synthetic acetylcholine receptor 166 48. Representative drug and receptor fragments 173 49. Valence shell MO diagram for Water 179 50. Effect of basis set on the energy of acetate HOMO 182 51. Effect of basis set on acetate C-O bond length 183 52. Effect of basis set on acetate O-C-O bond angle 183 53. Effect of basis set on acetate oxygen partial charge 184 54. Basis set dependence of Mulliken partial charge on N-H hydrogen in ethylammonium 187

55. Basis set dependence of Mulliken partial charge on in ethylammonium 188

56. Effect of basis set on the energy of the LUMO in ethylammonium 191 57. Comparison of Mulliken partial charges for ethylammonium and ethyltrimethylammonium 192 58. Geometry optimization for FAMA Complex 1 224 59. Dependence of interaction energy on the H5...N6 distance obtained by HF and MP2 methods 228

60. Basis set superposition error vs. H5...N6 distance in FAMA Complex 1 229 61. Dependence of interaction energy on the H5...N6 distance calculated by different DF approaches 231

xv 62. Comparison of interaction energies calculated by HF/MP2 and LSD approaches with and without gradient corrections 233 63. Geometry optimization for FAMA Complex 2 237 64. Optimized 3-21G*(*) geometry for the bifurcated formate:methylammonium complex 239 65. Comparison of energy vs. C-N distance for bifurcated and "linear" FAMA complexes 241

6 6 . Effect of basis set superposition error for the formate:methylammonium bifurcated complex with the 3-21G*(*) basis set 242

67. Effect of basis set superposition error for the formate:methylammonium bifurcated complex with the DZPP basis set 242

6 8 . Effect of basis set superposition error for the formate:methylammonium bifurcated complex with the 6-311++G** basis set 243 69. Effect of basis set superposition error for the formate: methylammonium bifurcated complex with the DZPF basis set and second order Moller-Plesset corrections (MP2) to the energies 243

70. Effect of basis set superposition error for the formate: methylammonium bifurcated complex with the 6-311++G** basis set and second order Moller-Plesset corrections (MP2) to the energies 244

71. Basis set superposition error vs. C1...N5 distance in the formate: methylammonium bifurcated complex using various basis sets at the Hartree-Fock level 245 72. Basis set superposition error vs. C1...N5 distance in the formate: methylammonium bifurcated complex using various basis sets and second-order Moller-Plesset corrections (MP2) to the energies 246

73. Dependence of interaction energy vs. C1...N5 distance in the formate:methylammonium bifurcated complex using 3-21G*(*) and 6-311G** basis sets at the Hartree-Fock level 246 74. Effect of diffuse functions on the interaction energy in the formate:methylammonium bifurcated complex at the Hartree-Fock level 247

xvi 75. Dependence of interaction energy on C1...N5 distance in the formate: methylammonium bifurcated complex by DMol (DNP) and DGauss (DZVP) implementations of the local spin density approximation 76. Comparison of interaction energies calculated by HF, HF/MP2 and DFT approaches with (NLSD) and without (LSD) gradient corrections to the local density approximation 77. Energies resulting from full geometry optimizations for fixed N-H distances using the 3-21G*(*) basis LIST OF SCHEMES SCHEME PAGE

I. Synthesis of 2’-hydroxy-5’-nitrophenylalanine and 5’-hydroxy-2’-nitrophenyIalanine 54 II. Synthesis of o- and p-nitro-o-tyrosines 55

ID. Attempted synthesis of BOC-dinitro-o-tyrosine 57

IV. Proposed mechanism for formation of 8 8 59 V. Syntheses of dinitro-o-tyrosine 63 VI. Alternate synthesis of o and p-nitro-o-tyrosines 64 VII. Synthesis of dinitro-m-tyrosines 65

Vm. Nitration of m-tyrosine 6 8 IX. Synthesis of 3’-hydroxy-2’,4’-dinitrophenylalanine and 5 ’ -hydroxy-2 ’ ,4 ’ -dinitropheny lalanine 69

X. Synthesis of 3’-hydioxy-2’,4’-dinitrophenylalanine and 3’-hydroxy-2’,6’-dinitrophenylaIanine 70 XI. Synthesis of (D)-o-tyrosine 72

XII. Synthesis of willardiine and 5-nitrowillardiine 73 Xm. Alternate nitrations of willardiine 74

XIV. Synthesis of 6-fluoroquinoxaline-2,3-diones 75 XV. Synthesis of quinoxalinediones 107.108.109. 110. and 111 78 XVI. Thermal decomposition of compound 111 79 XVD. Synthesis of 6-carboxyquinoxaline-2,3-diones 80 XVHI. Synthesis of 6-methoxyquinoxaline-2,3-diones 80 XIX. Proposed general synthesis of l-(2’amino-3 ’-carboxyethyl)-2,3-quinoxalinediones 81

XX. Methylation of quionxaline-2,3-diones 82 xviii XXI. Reaction of quionxaline-2,3-dione di-anion with with bromoacetaldehyde diethyl acetal in DMF XXII. Unsuccessful synthetic paths leading to 1 -(2 ’ -amino-3J -carboxyethyl)- 6 -fluoro-1,4-dihydro- quinoxaline-2,3-dione. XXm. Synthesis of l-(2’-amino-3’-carboxyethyl)- 1,4-dihydro-quinoxaline-2,3 -dione (127)

XXIV. Attempted synthesis of 122 from 2.3-bis(trimethylsilyloxy)quinoxaline XXV. Synthetic confirmation of the structure of 122 by an unambiguous route XXVI. Chemistry of the Strecker reaction on 130

XXVH. Attempted synthesis of l-(2’,2’-diethoxyethyl)- 1.4-dihydro-6-nitroquinoxaline-2,3-dione

XXVHI. Synthesis of NQXAA XXIX. Attempted synthesis of DNQXAA PART 1.

DESIGN, SYNTHESIS, AND STRUCTURE-ACTIVITY STUDIES OF MOLECULES WITH ACTIVITY AT NON-NMDA GLUTAMATE RECEPTORS: HYDROXYPHENYLALANINES, QUINOXALINEDIONES

AND RELATED MOLECULES

1 CHAPTER I INTRODUCTION

It is not surprising that, in an organ as complicated and sophisticated as the human brain, things sometimes go wrong and, since excitatory amino acids play such an important role in many of the sophisticated Junctions of the brain, neither is it surprising that perturbations o f their normal Junctions can have profound effects. In experimental situations such changes lead to impaired , abnormal behavior, convulsive activity and degeneration o f postsynaptic neurones. These observations and their potential application to our understanding and therapeutic manipulation o f psychoses, convulsive disorders and neurodegenerative diseases are amongst the most exciting in neurology today. -David Lodge, in Excitatory Amino Acids in Health and Disease (John Wiley, 1988).

1.1 Background Okamoto (1951)1 and Hayashi (1952)2 reported the excitatory responses of cerebral cortical cells to glutamate and aspartate obtained while studying epileptic phenomena.

Classic iontophoretic work by Curtis and Watkins in the early 1960’s3,4 revealed that glutamate and aspartate were powerful neuroexcitants, and subsequent examination of many compounds by these and other workers established structural requirements for potent activity. Further study of so-called "excitatory amino acids" proceeded at a moderate pace; by the late 1960’s, the possibility that endogenous excitatory amino acid was a significant process was widely discounted .5 This may have been in part because of the involvement of the most likely transmitter candidate, glutamate, in important pathways of intermediary metabolism, which was not consistent with the classical notion of specifically-synthesized molecules. Advances in autoradiography, , immunohistology, microscopy, and molecular biology coupled with evidence that excitatory amino acids serve as transmitters for a wide array of normal neurological functions and are implicated in many neurological disorders has brought about an explosive increase in excitatory amino acid research over the last decade.

A large portion of the in the mammalian appear to use excitatory amino acids as transmitters .6*7 Excitatory amino acids (EAAs) or related endogenous transmitters mediate normal excitatory synaptic transmission for sensory pathways such as the visual, auditory and olfactory systems, and in corticocortical and corticofiigal pathways ,8 including those involved in nociception, cardiovascular reflex, respiration, and motor control .9 The excitatory post-synaptic potentials (EPSPs) in these pathways are likely evoked through activation of AMPA or kainate (KAIN) receptors (vida infra). Excitatory amino acids also play a role in modulating synaptic transmission efficacy, and in modifying synaptic connections during development. In this capacity, NMD A receptors (yida infra ) co-localized with AMPA or KAIN receptors evidentally play a key role in 10 and long-term potentiation , 11 and thus in learning and . Excitatory amino acids appear to be involved in a number of neuronal growth and maturation processes, for example the onset of puberty in mammals . 12 EAA

" " 13 (ability to cause selective neuronal damage and death) is strongly implicated in the etiology of certain neurological dysfunctions (such as ) and neurodegenerative disorders. Currently, functional NMDA antagonists are being vigorously investigated for their therapeutic potential as cytoprotectants in -induced and other disorders .9,20 4

1.2 Endogenous excitatory amino acid transmission 1.2.1 Identification of EAA neurotransmitters The endogenous transmitters) have not been unequivocably identified for most of the synapses in the mammalian CNS which appear to use EAA transmission or modulation .7' 14 Studies by Kmjevic and Phillis (1963)15 on cortical cells provided early evidence that glutamate ( 1) and aspartate (2) might serve this role in at least some synapses. A major difficulty in establishing amino acid pathways of neurotransmission is the ubiquitous nature of these amino acids in nerve tissue. Glutamate is incorporated

Table 1. Amino Acid Content of Whole Rat Brain (pmoles/gram wet weight).

Glutamate 13.6 ± 0.4 Taurine 4.8 ± 0.3 4.4 ± 0.2 Aspartate 3.7 ± 0.2 GABA 2.3 ±0.1 Glycine 1.7 ±0.1 1.4 ±0.1 1.1 ± 0 .1 Lysine 0.4 ± 0.0 into peptides and , contributes to the regulation of osmolarity and levels (with glutamine), is a key component of , and serves as a precursor for

GAB A (y-aminobutyric acid, interestingly, an inhibitory neurotransmitter). Unlike most other neurotransmitters, glutamate is also involved in a number of important metabolic processes; for example, glutamate serves as a precursor for several Krebs cycle intermediates (section 1.2.3) and is involved in fatty acid synthesis . 16 Compared with other amino acids, glutamate is present at high concentrations in rat brain (Table l ) . 17 5

Historically, the demonstration of high-affinity synaptosomal glutamate and aspartate uptake 18 has provided a critical means of identifying nerve terminals which store these molecules and hence might utilize EAA transmission. Evidence for defining possible

EAA pathways includes :7,8,14' 17' 19 1) elicitation of similar responses (e.g. voltage dependence and effects of such as Mg++, Ca++ and C1‘) by stimulation of proposed afferent pathways and by application of the proposed transmitter (this is however entirely equivocable since, e.g., choline induces depolarizations at cholinergic synapses which are very much like those of acetylcholine, the endogenous transmitter20); 2 ) release of transmitter upon stimulation; 3) reduction in high-affinity uptake upon lesioning; 4) reduction in endogenous concentration following lesioning; S) retrograde transport of labelled amino acid, e.g. [3H ]-d -Aspartate (this has been used in combination with autoradiography on a microscopic scale); 6 ) immunohistochemical localization of the transmitter14,21 (difficult for glutamate due to its ubiquity) and/or thought to be involved in metabolism thereof (e.g. glutamate/aspartate transaminase (Asp-T), section

1.2.3); 7) the action of characterized EAA antagonists; 8 ) the augmentation of postsynaptic responses to exogenously-applied EAA’s by demonstrated reuptake inhibitors, e.g. dihydrokainate ( 6 ); 9) the presence of high-affinity binding sites for

[3H]-glutamate and other radioligands. All of these criteria are subject to caveats , 14 and the most convincing method, i.e. parallel intracellular recording of pre- and post-synaptic events, is technically demanding and thus of very limited use in mapping EAA pathways throughout the CNS . 14 A confounding consideration is that EAA response is primarily terminated by high-capacity reuptake systems; hence both local and regional differences in potencies for exogenously-applied agonists may be related to differences in uptake rather than true differences in receptor activation .22 6 o-

° T ° " ° y ° ' u n LcOj- l^COz- ° k .co,-

NH3+ CH3NH3+ n h 3+ 2 3 4 Asp NMDA HCA

O 0 0 HN-4 / m n rr J J O —< /“ C02- 0 0 J / \ ^ C 0 2- o ^ n' w T u I r 11 I CO,. C ^ c o w r V ^ N H - C 0 2H k- A ,H T 2 v n h T c ° 2' h 3c _ n h n h 3+ + ' ' J 5 6 7 8 Quisqualate Dihydrokainate NAAG Quinolinate

Figure 1. Miscellaneous compounds discussed in the text.

Based on combinations of the above criteria, glutamate and (to a lesser extent) aspartate are the likely endogenous transmitters at a large number of EAA synapses found throughout the CNS . 14-17-23 In invertebrate systems, glutamate is firmly established as the transmitter used at neuromuscular junctions in the crab, crayfish, and locust .24 The broader role of excitatory amino acids as peripheral transmitters in invertebrate systems likely accounts for the continuing discovery of natural products (e.g. plant and spider ) which express their activity at EAA synapses. In many EAA pathways in the mammalian CNS the evidence favors glutamate as the main neurotransmitter. A summary of the evidence for pathways proposed to use excitatory amino acid transmission has been compiled by McGeer, Eccles, and McGeer 17 [in the following list, those pathways where aspartate may play a key role are marked with a single asterisk (*), and those which may use both glutamate and aspartate are marked with a double asterisk (**)]. These pathways include (1) afferents from the cerebellum to the red nucleus and the thalamus; (2 ) afferents from the to the amygdala, cuneate nucleus, dentate gyrus, lateral geniculate nucleus** (involved in transduction of vision from retinal afferents), nucleus accumbens (see chapter II), olfactory tubercle, pontine nuclei, red nucleus, spinal cord, , substantia nigra, ventral tegmental area (see chapter II), and thalamus; (3) commissural/associational* and Schaffer collateral pathways in the ;** (4) afferents from the hippocampus to the hypothalamus, lateral septum, mammilary body, nucleus accumbens, bed nucleus of the stria terminals, and nucleus of the diagonal band; (S) the lateral olfactory tract;** ( 6 ) in the brainstem: auditory nerve afferents,** pathways from the inferior olive and pontine nucleus to the cerebellum, and afferents from the spinal trigeminal nucleus to the thalamus; (7) descending tracts, intemeurons,* and primary afferents in the spinal cord. The functional roles of most of these pathways are not clear.

There is also convincing evidence that homocysteate (HCA, 4) is an endogenous excitatory transmitter ; 19 for example, depolarization of rat cortex and hippocampal slices induces a Ca++-dependent release of HCA in the cerebellum. has been immunohistochemically localized primarily to glial elements both by light and electron microscopy , 19 and HCA may be released from glial cells upon depolarization or exchanged with extracellular glutamate which is taken up after release, subsequently acting on synaptic or extrasynaptic receptors . 19 HCA acts as an NMDA agonist (as demonstrated by response similarity to exogenous NMDA (3) and/or specific antagonism by NMDA antagonists) in the cat caudate nucleus, rat neocortex, mouse embryo hippocampus, cat dorso-lateral geniculate (visual stimulation pathway), and rat striatum. 19 Significant differences in the nature of physiological responses to activation by glutamate and HCA have been noted .24 N-acetyl-aspartyl-L-glutamate (NAAG, 7) is an endogenous acidic dipeptide which is convulsant and inhibits chloride-dependent [ 3H]-G1u binding to rat brain membranes .25

NAAG is present at very high levels in some CNS tissues (~1 nmol/mg wet ) .26 NAAG may be the primary transmitter exciting that part of the pyriform cortex which receives input from the lateral olfactory tract, since the endogenous excitations and those induced by NAAG, but not L-glutamate, are blocked by AP4 (11 ) . 27 NAAG also depolarizes hippocampal pyramidal cells, lateral septal , dissociated spinal neurons in culture, and cultured cerebellar neurons .28 NAALA-dipeptidase (NAALA =

N-acetylated a-linked acidic) is a chloride-dependent, membrane bound metallopeptidase localized to the brain and kidney which may play a key role in the synaptic degradation of NAAG .28 Glutamate derived from NAAG by NAALA-dipeptidase is subject to active Na+-dependent reuptake; thus NAAG transmission may be terminated in a fashion analogous to the termination of cholinergic transmission by acetycholinesterase. Interestingly, NAALA-dipeptidase is inhibited by quisqualate (5) with an IC 50 of 480 nM.

Quinolinate ( 8 ), a metabolite of tryptophan 190 (Figure 2), may serve as an endogenous transmitter at some synapses in the spinal cord and pyriform cortex .22,29 However, the depolarization-related release of quinolinate has not been demonstrated, and there appear to be no active uptake processes in the hippocampus, striatum, or cortex .29

1.2.2. Receptor classification, physiology, and distribution

Considering the ubiquity of EAA transmission in the CNS, it is not surprising that there are least rive types of excitatory amino acid receptors defined by radioligand binding, electrophysiological, and biochemical studies .6,30 Under current terminology 31 these are the NMDA, AMPA, kainate, AP4, and metabotropic (ACPD) receptors (see structures 9

NH NH ► NH N C02H NH Fe/Cu H ^ r 15 Kynurenic (B6) Acid

NH- uOH n O NH3+

|<;rY W^ c o 2- ► (B6 ) c £NH2 ^ ^ n h 2 NH- OH anthranilic acid (B6 )

O OH OH .CQ,- iP t N C 02H oc° 8 OH OH quinolinate 3-OH-anthranilic acid a, tryptophan pyrrolase; b, kynurenine formylase; c, kynurenine 3-hydroxylase; d, kynureninase; e, 3 -hydroxy anthranilate oxidase; B6 , depends on pyridoxal CoA 5-phosphate cofactor.

Figure 2. Kynurenate path of tryptophan metabolism.

Figure 3). Since the prototypical 3,4 compounds glutamate and aspartate excite most CNS neurons, it was initially thought that this excitation was nonspecific .6 Further study revealed regional differences in sensitivity to these two compounds in the CNS ,32 evidence that there were at least two major subtypes of EAA receptors. Historically, differences in the potency of agonists have been extremely important in defining EAA receptor subtypes. As new compounds displaying EAA activity were discovered, many from natural sources, classifying receptor subtypes became more complex. Glutamate, 10

co2- N O HO CH, C 02- r c o 2-

c h 3n h 3+ NH3+ + 3 9 10 NMDA AMPA KAIN HO O- p c o 2h V /

co2- Liu C 02-

n h 3+ n h 3+ 11 12 AP4 (Is, 3r)-ACPD

Figure 3. Prototypical compounds for EAA receptor subtypes. the probable transmitter at many EAA synapses, activates combinations of receptor subtypes to elicit different functional responses .8,24 Regional differences in response to various compounds can also be due to differences in uptake and sequestration as well as from intrinsic differences in receptor activation, and effects of some compounds may be expressed or enhanced by displacement (release) of endogenous .24,33 As a further complication, radioligand binding studies initially failed to correlate with the regional variations in physiological response .6,24

The classification of receptor subtypes was (and still is) hindered by the delayed discovery of antagonists for EAA receptors. The action of a-Aminoadipate (a-AA, 13), the first useful NMDA antagonist, was not characterized until the late 1970’s.35 This discovery was a major breakthrough, and was rapidly followed by the synthesis of the more-potent 14 (D-AP5, also known as 2-APV ) , 34,36 in which the co-carboxyl of a-AA was replaced by phosphonate. These compounds, particularly AP5, AP7, and the many 11

c

co2-

n h 3+ n h 3+ 13 14 D-aAA D-AP5

Figure 4. First-generation NMDA antagonists. analogs of these two compounds which have since been synthesized (see section 1.3.1), are the primary reason why NMDA pharmacology is much better understood than that of other EAA subtypes.

Electrophysiological criteria have been used to define the three classical EAA receptor types (c.f. Figure 5 ) : 22 (1) Quisqualate receptors (now referred to as AMPA receptors): onset of depolarization is rapid ( - 1 sec) and a plateau is quickly reached. -sensitive action potentials begin above a certain threshold. Repolarization is rapid after removal of the agonist (~5 sec). The response is mostly unaffected by AP5 (14) or (15. an endogenous tryptophan metabolite, see Figure 2). (2) Kainate receptors: onset of depolarization is slower (^5 sec) and action potentials again appear above threshold, but a real plateau is never reached and response termination is slow (>20 sec) after removing the agonist. The response is blocked by kynurenic acid but not AP5. (3) NMDA receptors: onset of depolarization is slower than in (1) and (2).

Large depolarizing shifts appear at a point below the threshold for (1) and (2), and "bursts" (action potentials each succeeded by a hyperpolarization) are superimposed upon these shifts. The bursts are then replaced by normal, tetrodotoxin-sensitive action 12

NMOA 5 QUIS 30 KAIN 4 0

A

- § § -'jl! mV .____ DAPV ?0

C ,111k!

KYNU 25 ■ a l l l 111111,11 H ,I

ULLiL 10 mV 5 sec

Figure 5. Comparison of NMDA, QUIS, and KAIN responses in CA1 hippocampal pyramidal neurone. (Reprinted with permission from 22). potentials. Response termination upon withdrawal of the agonist is similar to that of quisqualate receptors, and the action is sensitive to APS and kynurenic acid.

These receptor subtypes have also been characterized by differences in desensitization rates in whole cell voltage clamp of embryonic hippocampal neurons .37 AMPA and

QUIS responses showed rapid desensitization (time constant 30 ms) to plateau currents 20% of the peak levels. NMDA gave slower desensitization (time constant 250 ms) to plateau currents 50% those of peak, while KAIN and 5-bromo-willardiine (41) gave very little desensitization. L-Glutamate gave responses which were mixtures of AMPA/QUIS and NMDA components, and the NMDA component was absent in glycine-free solution and could be blocked by the addition of 1 mM Mg2* (see below). 13

NMDA, AMPA, and KAIN receptors are coupled to channels which, when activated, allow depolarization of post-synaptic membranes through an influx of , potassium, and, for NMDA receptors, ions .24*38*39 The depolarizations are not sensitive to tedrodotoxin .24 Although at one time it seemed that NMDA and non-NMDA receptors might both be coupled to a single ion channel, most studies suggest that NMDA and non-NMDA channels are distinct entities. Some lines of evidence are: (1) channel blockers, including divalent cations, selectively block NMDA currents ;40 (2) enzymatic treatment selectively inactivates NMDA receptors ;41 (3) Some central neurons possess NMDA receptors while others, which still respond to AMPA or KAIN, do not ;42 (4) it is possible to express functional NMDA receptors in frog oocytes independent of AMPA receptors,43,44and vice versa ;45 (5) radiation inactivation studies give different molecular target sizes for NMDA and AMPA receptors (MW ~125,000 and - 52,000 daltons, respectively ) ;46 and (6 ) functional non-NMDA receptors (which seem to be unitary AMPA/KAIN receptors, see below) have been isolated by affinity chromatography 47

A physiologically important effect is the voltage-dependent block of the NMDA ion channel by Mg2+, which results in an unusual decrease in membrane conductance with increasingly positive membrane potential (Figure 6 , curve A) . 24,39 A more normal dependence is seen with kainate or quisqualate (curve B). Stimulation by glutamate in membranes where both receptors are present results in a current-voltage relationship

(curve C) which is very flat,and this curve can be duplicated by combined application of

NMDA and KAIN.24 Such behavior effectively allows NMDA receptors to serve an amplification function ,48 and enhances bursting behavior 24 (see above). The voltage-dependent block of the NMDA channel by Mg2+ is consistent with a binding site 14 within the (open) channel pore; by contrast block by competetive antagonists (e.g. AP5 and Zn++), is voltage-independent and consistent with an extracellular agonist recognition site .38,40,49 The NMDA receptor channel requires glycine as a co-agonist for channel activation .44’50 There is also a site for binding non-competitive antagonists

nA + 2 -

-90 -60 -30 +30

mV

-2 -

Figure 6. Voltage/current relationships of EAA channels. After refs. 24 and 39. For explanation, see text. such as MK-801 (30, see section 1.3.1). Antagonism by these agents is use-dependent, perhaps because they bind to a site within the ion channel .38,40,51 A recent report indicates that NMDA receptor channels require two molecules of agonist for activation .52

At synapses where excitatory amino acids serve as conventional excitatory transmitters, the fast excitatory post-synaptic potentials (EPSP’s) are mediated by non-NMDA receptors . 14 These EPSP’s share characteristics with responses to exogenously administered non-NMDA agonists, such as voltage dependence (see above) and antagonism by selective antagonists. Interpretation of studies with antagonists is limited by lack of antagonist specificity, however, and most studies have hinged on differences 15 in the potencies of less specific agonists. Additional evidence for the role of non-NMDA receptors in mediating fast ESPS’s comes from comparisons to transmission at invertebrate neuromuscular junctions, where the neurotransmitter role of glutamate is more unequivocably established .24

Soire interesting observations have recently emerged h m gene cloning studies of

AMPA receptors .53,54 Different functional properties were conferred on four abundant non-NMDA (AMPA/KAIN) receptors (GluR A, B, C, D ) 54 depending on the incorporation by alternative splicing of one of two subunits, dubbed "flip" and "flop", in the main receptor protein. The distributions of the flip and flop versions of the receptors varied in the rat brain. The authors conclude that (1) this type of alternate splicing may contribute to regional variations in functional properties of EAA transmission; (2) alternate splicing in particular cells could serve to change synaptic efficacy (as in learning and memory functions); (3) malfunctions of such a system could contribute to the etiology of EAA-related diseases (section 1.4).

Foster and Fagg 33 have pointed out that "[radio] binding techniques are of value only if it has been demonstrated that the radioactive ligand selectively labels the physiological or pharmacological receptors of interest." There is only tentative correspondence between EAA receptors defined by physiological studies and those characterized in binding studies . 17 Difficulties in establishing correlations relate to high sensitivity of binding studies to differences in tissue preparation, buffers used, presence of ions, and removal of unbound radioactivity . 17,55 Thus, an early difficulty in the interpretation of L-[ 3H]-glutamate binding was the inclusion of chloride ions in buffer preparations ,6 which facilitates a high affinity glutamate uptake process .56 Also, the selectivity of available ligands obviously plays a key role in the usefulness of radioligand 16 binding. Johnson and Koemer 57 recently discussed the inherent theoretical limitations of binding studies using radiolabelled agonists for receptors mediating millisecond neurotransmission. Briefly, the association of a ligand (L) and its receptor (R) is described by the second order rate constant kt, and the dissociation by the first order rate constant k^. The equilibrium constant Kd is then given by equation 1.2. Johnson and

ki (1.1) L + R ^ LR (1.2) Kd = kj/kj k-i

Koemer point out that the off rate, k.i must be -10 3 s' 1 or greater to account for the observed firing rates of 200 Hz or more. The value of kj in many ligand-receptor systems is 10 6-107 M 'V 1; these rate constants correspond to a Kd of 10'3- lO -4 M. In binding studies, a washing step necessary to remove unbound radioligand requires seconds to minutes, during which time much of a ligand with k.i ~ 103 should be removed. The apparent Kd for excitatory amino acid agonists in physiological studies are 10-6 - 10-4 M; yet in binding studies compounds such as kainate (10) and AMPA (9) show Kd’s of 10 "9 - 10' 7 M. Although the coexistence of low and high affinity binding sites has been demonstrated in at least some CNS tissues, there is no reason to assume that this must be general.

The relationship of high affinity [ 3H]-KAIN binding (K

[3H]-AMPA and [3H]-KAIN binding sites are co-localized .47 [It is noteworthy however that this region has the highest density of [ 3H]-KAIN sites in the rat brain,58 is somewhat unique in having electrophysiological responses with high sensitivity to KAIN (Kd ~10 nM), and is extremely sensitive to the neurotoxic effects of kainate.] In Xenopus Laevis brain, octylglucoside extraction gave fractions having both QUIS and KAIN functional activity. Purification by several techniques gave a protein with 1:1 QUIS :KAIN activity in reconstituted systems .59 Radiation inactivation analysis in rat cortical membranes gave a molecular target size of 52,000 for a low-affinity (Kd=65 nM) [ 3H]-KAIN binding site, essentially the same as that found for [ 3H]-AMPA inactivation .60 Some electrophysiological studies seem to indicate a unitary receptor where AMPA and QUIS act as partial agonists and compete with the action of KAIN, a full agonist. Contrary to the usual situation for full agonists, however, kainate responses exhibit little desensitization, while resonses to the partial agonists AMPA and QUIS show rapid desensitization .61 The hypothesis has been advanced that in these systems, AMPA and KAIN receptors represent different conformations or functional states of the same protein .46,47,59

Nonetheless, the relative potencies of agonists at certain populations of neurons activated by kainate in rat spinal cord differ from those evoking QUIS/AMPA-type responses ,6,24 and C-fiber afferents in mammalian spinal cord have a population of KAIN receptors which are insensitive to AMPA .62 There is a substantial lack of overlap between the [3H]-KAIN and [3Hj-AMPA receptor populations measured autoradiographically, yet there is excellent correspondence between the regions labelled by [ 3H]-KAIN and the susceptibility of cells to the neurotoxic actions of kainate (see section 1.4 ) .63 Radiation inactivation studies also clearly demonstrate a high-affinity (Kd= 3.5 nM), Ca2+-sensitive binding site for [ 3H]-KAIN with a different molecular target size (MW ~ 70,000) from that of the low affinity site mentioned above .60 In elegant studies by Ascher and

Nowak 64 using whole-cell patch clamping of cultured mouse central neutrons (striatum, mesencephalon, and spinal cord), the conductance of channels carrying quisqualate-induced currents was 8 pS (picoSiemens), whereas that for kainate-induced currents was about 4 pS. Spectral noise analysis revealed time constants of 10-15 ms and 2-3 ms for quisqualate- and kainate-induced spikes. These results were found to vary somewhat from cell to cell but were largely indepent of the tissue origin.

It is of some interest that high affinity [ 3H]-KAIN binding sites in chick cerebellum are immunohistochemically localized to the Bergman glia ,47 and show very low affinity for glutamate. The significance of this localization is at present unknown, as is the possible generality of glial localization of some KAIN receptors in the mammalian CNS. Besides postsynaptic and glial receptors for kainate, there is evidence for a presynaptic receptor .65 Agonist binding to this receptor enhances the presynaptic release of glutamate, and it seems possible that in some tissues this population of kainate-preferring receptors may mediate the indirect unique to kainate agonists (section

1.4).

The AP4 and metabotropic receptors are less well-characterized. L-AP4 acts as an agonist in some tissue preparations, in particular the bipolar-ON cells in the retina .57 This may be a unique case, and in other tissues, especially the spinal cord, AP4 behaves 19 as an antagonist of endogeous EAA transmission, perhaps acting at presynaptic receptors .66,67 Activation of the "metabotropic" (ACPD) receptor increases cyclic GMP concentrations and phosphatidyl inositol turnover in brain slices and cultured neurons ,68,69,70 and activation of this receptor seems plausibly associated with longer-term changes such as synaptic plasticity and developmental functions .68

Autoradiographic studies with radioligands such as [ 3H]-glutamate (especially in combination with antagonists for specific receptor subtypes ) , 58,71 [3H]-NMDA,

[3H]-AMPA,72*73 [3H]-kainate,74,75 and [3H]-CNQX76 are proving useful in identifying CNS distributions of receptor subtypes, and in combination with other evidence may give clues about the functional roles of EAAs. For example, [ 3H]-NMDA labels regions of the brain which are highly "plastic", such as the CA1 region of the hippocampus, in which NMDA receptors have been shown to be involved in long-term potentiation .8

1.2.3 Metabolism of excitatory amino acids.

Glutamate and aspartate are not essential amino acids and do not readily cross the blood-brain barrier of mature animals in free form 7 (see however section 1.4). Hence glutamate used in the CNS must be synthesized de novo from glucose and other molecules which feed into the Krebs cycle, or recovered from molecules (such as peptides or proteins) which do cross the blood-brain barrier or are broken down within the CNS. The principle routes of glutamate synthesis in the mammalian CNS are (Figure

7) : 7,17 1) reductive animation of a-ketoglutarate from the Krebs cycle using the enzyme dehydrogenase (reversed); 2 ) transamination of a-ketoglutarate by aspartate using aspartate transaminase (Asp-T); hydrolysis of glutamine via glutaminase; synthesis from ornithine via A1 -pyrroline-5-carboxylie acid (P5C) and glutamic acid semialdehyde using the enzyme ornithine aminotransferase (Om-T) 20

O II -o 2c c c h 2c o 2- malate oxaloacetate citrate / GLU GLU fumarate t Asp-T cis-aconitate succinate

^ Asp Asp succinyl-CoA isocitrate O II -o 2c -c c h 2c h 2c o 2- a-ketoglutarate Glutamine + H+ + NADPH ^ . NH4+ + NADPH + H+ glutamate synthase ] f glutamate dehydrogenase NADP+ + H20 GLUTAMATE + NADP+ NH,+»3H I -o 2c c c h 2c h 2c o 2- Glutamate + ATP + NH4+ L ' n glutaminase K ADP + Ps + H+ h 2o n h S

-o 2c - c c h 2c h 2c o n h 2 glutamine

Figure 7. Metabolic transformations of glutamate. 21

(Figure 8 ). The relative significance of each of these metabolic pathways and the compartmentalization of glutamate are areas of active research. Immunohistochemical localization of glutamate metabolic enzymes or neurochemical lesioning studies investigating changes in metabolic enzyme activity have not produced conclusive evidence that any of these pathways is the primary source of the glutamate neurotransmitter pool, but it appears likely that different sources may predominate in different pathways .17 In many brain regions mitochondrial glutaminase may be of primary importance .7,77,78

NH,+ O O

NH3 NH ornithine Glutamate

NH2 t c ° 2‘

a: Ornithine aminotransferase (Om-T) b: A^pyrroline-Scarboxylic acid dehydrogenase c: glutamate semialdehyde oxidase d: proline oxidase

Figure 8 . Alternate metabolic pathways to glutamate.

The predominant metabolic source of CNS aspartate has not been identified .7 Aspartate can be obtained from glutamate via aspartate transaminase (Figure 7), or by hydrolysis of asparagine. 22 1.2.4 Storage, release, and reuptake of EAA transmitters.

Glutamate (and probably also aspartate) is stored in a vesicular synaptosomal subfraction and release is triggered by localized high concentrations of Ca++ in the synaptosomal cytoplasm .77,79 This Ca++-dependent release is dependent on the maintainance of high

ATP/ADP ratios. There appears to be some presynaptic inhibition of release ;80 although the details are unclear at present, there are apparently no NMDA- or AMPA-type presynaptic glutamate receptors .77,80 AP4 (11) is reported to inhibit glutamate release from synaptosomes in the hippocampus ;81 glutamate would presumably act at such presynaptic AP4 receptors ,80 and there may also be presynaptic inhibition by GABA (GABAB-type receptors) and acetylcholine (muscarinic receptors) in certain tissues. The use of glutamate as a neurotransmitter depends on the termination of action by a high-capacity uptake carrier in the synaptosomal membrane which is driven by the transmembranal sodium gradient 79 and can maintain a glutamate concentration differential of about 10 4 (inside/out). This carrier is rather nonspecific in its uptake of acidic amino acids, and probably transports either 2 Na+ ions and one H+, or 3 Na+, in with each glutamate molecule, with the concurrent release of one K+ ion 77 Thus, glutamate release is inhibited by high extracellular K+ concentrations. There is a similar transport system in glial cells .253 A second vesicular transporter is more specific and sequesters cytoplasmic glutamate into vesicles at concentrations as high as 100 mM, approximately 10 times that of the cytoplasm and 106 times the normal glutamate concentration in the synaptic cleft .77 Interestingly, this transporter does not take up aspartate into storage vesicles. The vesicular transporter for glutamate is dependent on the transmembrane potential generated by a Mg++-ATPase, but not ApH, in contrast to vesicular uptake of dopamine and GABA .79 Recently, f/wts-4-carboxy-L-proline was found to potently inhibit glutamate reuptake while showing minimal activation of post-synaptic receptors . 189 23

PRESYNAPTIC BULB

Glu GLIAL 100 mM CELL Gin Glu Glu T Glu 10 mM 100 mM Glu 100 mM 100 mM Glu 100 mM

POSTS YNAPTIC CELL

Figure 9. Glutamate storage, release, and reuptake. Glutamate is stored in vesicles (V) in the presynaptic bulb, and released in a Ca2+-dependent fashion. The released transmitter acts at postsynaptic receptors (R) to cause depolarization (non-NMDA) and modulatory (NMDA, ACPD) responses, and probably at presynaptic autoceptors (A) to regulate glutamate release. Transmitter is rapidly taken back up into the presynaptic bulb via a 3 Na+/Ghr transporter which concurrently transports one K+ out of the cytoplasm, and can theoretically maintain a concentration gradient of at least 2 0 ,0 0 0 :1. TTie glial cells have similar uptake transporters, and the glutamate which is taken up may be converted to glutamine by glutamine synthase (GS). This glutamine can be transported out of the glial cells, taken up into the terminals by a low affinity path, and converted back to glutamate by mitochondrial glutaminase (GA). 24

1.3 Structure-activitv relationships

Early examination of the activity of excitatory amino acid analogs proceeded without particular notions of EAA receptor subtypes. The iontophoretic work of Curtis and Watkins in the early 1960’s3*4 revealed that glutamate and aspartate were powerful neuroexcitants. Their examination of many other compounds established that the a-amino acid group was required, and a second acidic group positioned two or three carbon atoms from the glycine moeity was also necessary.

Structure-activity relationships of EAA agonists are not consistent throughout various regions of the CNS, likely due both to receptor subtype populations and differences in uptake and sequestration of various agonists. As examples :22 in the hippocampus both l - and D-homocysteate (4) behave as AMPA agonists, but in the striatum L-homocysteate gives NMDA-like responses which are blocked by AP5 (14): in the spinal cord L-glutamate appears to activate mainly AMPA-type receptors, whereas in the cerebral cortex it seems to be a mixed NMD A/non-NMD A agonist.

While this thesis is primarily concerned with AMPA receptors, a brief summary of structure-activity relationships at other EAA receptor types is given to show contrasts with AMPA-active compounds. A full discussion of the structure-activity relationships for AMPA receptors is given in Chapter n.

1.3.1 NMDA receptors a. Agonists

Many compounds have been tested for their ability to inhibit binding of d-[ 3H]-AP5, and an excellent summary of the resulting structure-activity relationships for agonists acting 3 2 1 4 16 d »l-NMDA D®L-Asp l»d -G1 u l > d HCA L-Ibotenate co2- co2-

COr H3N+ C° 2" H H 17 18 8 19 trans-2, 3-PDA 1r,3r-ACPD Quinolinate Homoquinolinate

Figure 10. NMDA agonists at NMDA receptors has recently appeared .82 Those compounds with significant activity are given in Figure 10. All of the compounds are either aspartate homologs (NMDA (3) trans-2,3-piperidinedicarboxylic acid (17). quinolinate ( 8 )) or glutamate homologs

(homocysteic acid (4), Q 6 ), (19)). is an endogenously-occurring tryptophan metabolite (Figure 2) which is less potent than the other molecules shown, but likely acts at either a subset of NMDA receptors or in some cases perhaps at its own receptor class . 190 Ibotenic Acid (IBO, 16) is a natural product isolated from the mushroom Amanita muscariaP The general structure-activity relationships are summarized as follows:

1) The optimum chain length for l amino acids corresponds to that of glutamate and HCA. For D amino acids, chain lengths longer than glutamate give antagonists (e.g. D-aAA,

see below); when the terminal group is carboxyl, the optimal chain length corresponds

to that of aspartate, whereas when the terminal group is sulfono or sulfino, the optimal

chain length is that of HCA. 26

2) In the l amino acids of glutamate length, the order of potency with substitution of the

co-terminal group is C02H > S03H > S02H » P03H (note that AP4 generally acts at

a different class of receptors than does NMDA). In l amino acids of aspartate length,

C02H * S02H » S03H > P03H. In d amino acids of glutamate length, S02H »

S03H > C02H » P03H, whereas in the D-aspartate series, COzH > S02H » S03H

> P 0 3H. The d , l tetrazole analogs of glutamate and aspartate are moderately potent

compounds .84 3) L-sulfoserine and L-sulfocysteine are moderately potent and almost equipotent with

glutamate, respectively. Heteroatom replacements in other analogs are much less

successful. 4) An a-methyl group greatly reduces the activity of glutamate and aspartate. A p-methyl group causes only moderate loss of activity in aspartate but a great reduction in glutamate, whereas a p-hydroxyl group is moderately tolerable in both. The y-position in glutamate may be substituted with methyl, methylene, fluoro, hydroxy,

or amino groups with retention of moderate to good activity.

5) Except for D-aspartate, N-alkylation causes large reductions in activity.

6 ) Removal of any of the basic or acidic groups causes almost complete loss of activity.

b. Competitive antagonists

Many competetive NMDA antagonists have recently been prepared .85,86 Virtually all of them are structural variations of AP-5 (14) and AP-7 (24); important examples are

CGS-19755 (21) and CPP (26). A series of cyclopropyl analogs of AP5 (not shown) were recently reported .87 p o 3h- H

c \ H co,- +/\ co2- H H 13 14 21 22 D-aAA D-AP5 CGS 19755 CGP 37849 Ki=13 pM Kj-0.5 pM Ki=0.18 pM Ki=0.035 pM

PO3H- PO3H- PO3H-

H + N C02- cnl +,n ; c o 2- +H,N H H H H 23 24 25 26 d -o AS D-AP7 2r-CPP 2R-CPPene K{=25 pM Ki-3.1 pM Kj=0.14 pM KpO.044 pM ICjq =0.8 |tM IC5o =0.08 pM

PO3H- PO,H- PO3H-

+H3N c o ,. +H,N +h 3n c o 7- 27 29 NPC451 IC50=61 pM IC50=1.0 pM IC50=3.3 pM

Figure 11. Competetive NMD A Antagonists. Values of K} and IC 50 are for [ 3H]-CPP binding. Ks values are from ref. 8 6 , IC50 values from reference 85 (thus a direct mathematical relationship does not exist) 30 31 MK-801 (PCP)

CH3

32 33 SKF 10047

Figure 12. Noncompetetive NMD A antagonists acting at the PCP/MK-801 site. c. Noncompetitive antagonists Besides die ions Zn2+ and Mg2+, which are physiologically important (section 1.2.2), a number of compounds non-competitively inhibit NMDA responses; most of these act at either the PCP/MK-801 site (Figure 12) or the glycine site (Figure 13). The synthesis or identification of compounds which act at these sites is a very active area of research. An excellent correlation was found between the ability of MK-801 (30). PCP (31).

(not shown) and SKF 10047 (32) to block NMDA responses in rat cortical slices and to inhibit specific [3H]-MK-801 binding in rat cortical membranes .49 Similar correlations have been found in rat behavioral paradigms .88 Nonetheless, PCP and related compounds produce some in vivo effects which are probably not mediated through NMDA antagonism, but rather through the high-affinity "a" () receptor .89 29

Glycine is required for NMDA responses in receptors expressed in frog oocytes from

cloning of rat brain m-RNA ,44 and much evidence indicates it is also required

endogenously. In the frog oocyte system, D-serine and D-alanine (but not the l forms) were almost as potent as glycine in potentiating NMDA responses. Antagonism of endogenous glycine binding is thus identified as a method of modulating NMDA activity. The compound HA-966 (34) is of historical importance but has low potency

and behaves as a . The quinoxalic acids (e.g 35 ) 90 are moderately potent at

O . ° H N 1 N CQ2H N C 02H HjN

34 35 36 HA966 X = I, Y = Cl (32 nM) X = Et, Y = Br (6 6 nM) COX X = Y = Br ( 8 6 nM) X = Y = Cl (200 nM) c o 2h X = I, Y = CH3(270nM) X = Et, Y = Cl(260 nM) X = Y = CHo (540 nM) 37 X = H, Y = Cl (560 nM) X = OH ( )-IC3 0 for[ 3H]-Gly X = OMe, OEt, O-nPr, O-iPr X = NH2 NHMe, NHEt, NHPr, NHCH2CH2Ph

Figure 13. Glycine antagonists. this site; the 6,7-dichloro analog shown is the most potent compound reported to date, but this compound also antagonizes kainate and AMPA responses .91,254 Leeson et. al.92 prepared a number of kynurenic acid analogs (36) and have identified several potent glycine antagonists with much reduced non-NMDA activity. Compounds in the indole-2- series (37) are somewhat less active, but optimization has 30

produced relatively potent glycine antagonists .93

1.3.2 Kainate receptors a. Agonists (KAIN, 10) and (38) are natural products from the seaweed Diginea simplex.M These compounds are still the most potent kainate agonists identified to date. Other agents which act at least in part at kainate receptors are quisqualate (5), acromelic acids A and B (39, 40), 5-bromo-willardiine (41), and glutamate. Changing the stereochemistry of any of the three stereocenters in KAIN reduces activity considerably, as does reducing the double bond in the isopropylidene side chain .48

O

n h 3+ 10 38 5 KAIN DOMOATE QUIS n

+ + NH3+ 39 40 41 ACRO-A ACRO-B 5-Br-WDN

Figure 9. Kainate agonists. 31 b. Antagonists: There are no compounds reported to block responses to kainate without affecting AMPA or NMDA responses. In the case of AMPA receptors, this may be in part due to the possibility that KAIN responses are in many cases mediated by AMPA receptors (section

1.2.2); however even in those brain regions where there are populations of KAIN receptors not affected by AMPA (e.g. afferent C-fibres), all of the compounds so far examined which block responses to kainate also block an AP5-sensitive (i.e. NMDA) component. Compounds which have proven useful in some studies include GAMS (42. see discussion Chapter II), kynurenic acid (15). a kainate-derived lactone (43 ) ,95 several benzoylpiperazine-dicarboxylic acid analogs (44 ) , 96,97 and CNQX (45).

O

+ + 42 43 44 GAMS R = I-, PhS- X = m-Cl, p-Cl, p -Br O H

H H 15 45 Kynurenic Acid CNQX

Figure 15. Nonselective kainate antagonists. 32

1.3.3 AP4 receptors

AP4 (11) displays diverse activities,57 acting most often as an antagonist of endogenous

transmission. In some cases at least, this antagonism is mediated through presynaptic receptors (see discussion section 1.2.3). L-AP4 (11) is more potent than D-AP4, and

L-phosphoserine (46) is less potent than AP4. Of several conformationally-restricted analogs incorporating cyclopentane and cyclohexane rings, 48 displayed activity most like AP4,98 and it was thus proposed that AP4 may bind in a folded conformation.84 OH O o cu ' o- I, H 11 — P p O ' " O - co2-

n h 3+

46 48 L-AP4 L-OPS

Figure 16. Compounds active at "AP4 receptors."

1.3.4 Metabotropic (ACPD) receptors The existence of excitatory amino acid receptors which are coupled to inositol phosphate metabolism has only recently been established (c. 1985);68 hence structure activity relationships are just now being studied. Research in this area is rapidly intensifying with the realization that the so-called "metabotropic" receptor is likely involved in long-term changes in synaptic efficacy and in neuronal growth and development processes. The only relatively selective agonist identified to date is 1 s,3r-ACPD (12. often called trans-ACPD in the literature, but the IUPAC nomenclature is ci5-( 1 s,3r)- I -aminocyclopentane-1,3-dicarboxylate).68,70,99 Quisqualate (5) is a partial agonist which also displays a number of other activities (see Chapter II), but QUIS is 33 more potent than ACPD in stimulating inositol phosphate metabolism .68 Ibotenic acid

(16) is comparably potent to ACPD in many systems but is also a partial agonist and an NMDA agonist (section 1.3.1). L-AP4 (11) is an ACPD antagonist in some tissue preparations, but is not very potent and is inactive in some cases. L-AP3 seems to be more selective than AP4 for metabotropic vs. ionotropic receptors ,68 and has no action at physiologically-defined AP4 receptors. L-AP3 is three to five times more potent than

D-AP3 in rat tissue slices.68 L-AP3 is probably a weak partial agonist and is inactive as an antagonist at high doses in some in vivo studies.70

1.4 Excitatory amino acid toxicity One of the most interesting aspects of excitatory amino acid physiology is the ability of many EAA agonists to cause nerve cell damage and death. Subcutaneous injections of glutamate were discovered to destroy the inner neural layers of the immature mouse retina

(1957).100 Later, Olney 101 and coworkers found that subcutaneous injections of monosodium glutamate (MSG) into newborn mice caused neuronal damage and death in brain regions known to have a poorly-developed blood-brain barrier. Adult animals treated with MSG at birth became severely obese with stunted skeletal development, and females were sterile. These disturbances were postulated to be the result of neurological damage in brain regions mediating endocrine functions .102 Only compounds which showed neuroexcitatory activity similar to that of glutamate (e.g. aspartate, cysteic acid and homocysteic acid) caused similar patterns of neuronal necrosis .103 occurs postsynaptically (dendritic and somatic destruction) while sparing axons, presynaptic terminals, and non-neuronal cells .104 Because of the weight of evidence, MSG was voluntarily removed from baby foods by manufacturers in the late 1970’s.105 34

Agonists CO,

c o ,- T co2 NH3+ n h 3+ 1 12 l-GIu 1s,3r-ACPD Partial Agonists O HO H N - f

CO,- o - V L co,- NH NH3+ 16 5 Ibotenic Acid QUIS Antagonists po3h- po3h-

COr CO,-

V + Y n h 3 nh 3+ 49 11 L-AP3 L-AP4 Figure 17. Compounds active at "metabotropic receptors."

Accumulating evidence 103 indicates a role for excitatory amino acids in brain damage associated with hypoglycemia, anoxia (as in stroke), physical trauma, epilepsy 106,107 and

Huntington’s chorea . 108 Quinolinic acid, an endogenous tryptophan metabolite (Figure 2), is strongly implicated in the etiology of Huntington’s disease , 109 possibly acting through receptor-mediated stimulation of somatostatin biosynthesis . 110 Elevated plasma 35

quinolinate levels have been observed in AIDS patients, and are correlated with

development of AIDS-related .249 A recent report indicates that brain degeneration associated with chronic liver failure may be at least in part the result of the

reduced glutamate uptake into perineuronal astrocytes . 111 Guam ALS/PD syndrome

(amyotrophic lateral sclerosis with Parkinsonism and Alzheimer’s-like ) has been

circumstantially linked to the comsumption of ( 5-N-methylaminoalanine (BMAA), a

natural component of cycad flour, by the Chamorro people of Guam during World War

II. 112>113 BMAA may exert its toxicity through NMDA receptors, with the requirement for

bicarbonate as a cofactor . 114 The disturbing implication is that consumption of the

produces some nerve damage which is not immediately symptomatic, but is evident only

when combined with normal age-related cell losses.

Much of the current research on glutamate toxicity concerns the role of NMDA receptors in acute ischemic damage (due to the therapeutic potential for treating , section

1.5) . 103 However, the actions of various receptor subtypes in causing excitoxic damage

may be integrative .6 The massive influxes of calcium which occur during neuronal insult

seem to play a key role in mediating cell death , 115,116 and oxidative stress may also be an

important factor.

Kainic acid (KAIN, 10) is a potent excitant of mammalian neurons (section 1.2), and injections into the rat striatum and hippocampus cause neurodegenerative patterns similar

to those observed in Huntington’s disease and status epilepticus, respectively . 117,118,119 Kainate toxicity in humans was clearly demonstrated by an incident in 1987, when about

150 Canadians became ill from eating mussels with high concentrations of domoic acid

(38).105 There were four fatalities, and indicated neuronal damage in the hippocampus and other areas of the limbic system . 119 Twelve survivors sustained 36 permanent memory impairment. Unaccountably, many who consumed the tainted mussels seemed to suffer no ill effects. Braitman and Coyle 120 have cited three major lines of evidence that the neurotoxic effects of KAIN are linked to specific receptors: (1) the neurotoxic effects of kainic acid analogs generally correlate with their affinity for the high-affinity [ 3H]-KAIN binding site ; 121,122,123'124 (2) neurons with the highest densities of high-affinity [ 3H]-KAIN receptors are most sensitive to the toxic effects of [ 3H]-KAIN (e.g. pyramidal cells in the hippocampus125); and (3) the development of neuronal vulnerablility to KAIN coincides with the developmental emergence of [ 3H]-KAIN binding sites in prenatal and juvenile animals . 126 KAIN binding sites are greatly enriched in isolated synaptic junctional fractions of rat forebrain, and are localized by autoradiography to terminal fields where KAIN acts as a potent . 127 The mechanism of kainate is not completely clear; however kainate analogs possess a unique ability to cause neuronal destruction remote from an injection site.

Kainate toxicity is thus at least partly indirect, and requires that excitatory afferents to vulnerable neurons are functionally intact . 118

AMPA receptors also appear to play an important role in some types of EAA-induced toxicity. Hence, consumption of p-N-oxalylaminoalanine (50, see chapter II), a naturally-occurring AMPA agonist found in the chickling pea, is evidentally the cause of human neurolathyrism . 128 In addition, NBQX (51, Chapter II) has been shown to exert a cytoprotective effect in a model for ischemia ; 129 this effect may be due to antagonism of glycine or kainate receptors, properties shown by virtually all quinoxalinediones tested so far (see Chapter II). 37

1.5 Therapeutic potential for excitatory amino acid modulation The many ways in which excitatoiy amino acids can mediate neurological damage and

cell death naturally suggests that EAA antagonists might be useful cytoprotective agents. Non-competetive NMDA antagonists such as MK-801 (30) bring dramatic reductions in

ischemic damage in rat 130 and gerbil models . 131 Indeed, clinical trials of NMDA antagonists are in the planning stages . 119 A major difficulty so far lies in developing proper experimental designs to demonstrate efficacy, and behavioral side effects of both competetive and non-competetive NMDA antagonists are a serious problem, though they may be of little consequence in such severe situations as stroke. Side effects include amnesia, confusion, and loss of muscle tone , 132,133 and also appear with potent competetive NMDA antagonists such as CPP (25).133 MK-801 and related compounds including PCP (31) have also been found to cause certain acute pathomorphological changes in adult rats at relatively low doses .251 DNQX (52)2S0 and NBQX (51)129 have been shown to have cytoprotectant properties; NBQX is systemically active at 30 mg/kg in rats (this is a substantial ). The effect may be due to the kainate antagonism.

The potential for the treatment of epilepsy with excitatory amino acid antagonists was recently discussed in detail . 135 AP5 (14), AP7 (24). and some other competitive NMDA antagonists (section 1.3.1b) are effective anticonvulsants when administered intraventricularly (i.c.v.) in a variety of animal models of epilepsy. These compounds were less effective systemically due to poor penetration of the blood-brain barrier.

MK-801 (30, section 1.3.1c) was more effective as an anticonvulsant than conventional antiepileptic drugs (e.g. diazepam, diphenylhydantoin and ), but behavioral side effects such as loss of muscle tone and ataxia were observed with effective doses . 135 The possible role of kainate receptors in epilepsy discussed in the previous section 38 indicates a potential therapeutic role for kainate antagonists . 104 Unfortunately, no selective and potent kainate antagonists have yet been reported.

Other therapeutic possibilities for EAA antagonists 11,136 include use as anxiolytics , 134 muscle relaxants, 137 antipsychotics 138,139 and in the treatment of neurodegenerative diseases (section 1.4). MK-801 can protect dopaminergic neurons in the rat substantia nigra from the toxic effects of MPTP (1 -methy 1-4-phenyl-1,2,3, 6 -tetrahydropyridine ) 140 and methamphetamine . 141 Considering the apparent role of NMDA receptors in synaptic plasticity and hence in learning and memory, NMDA agonists might be potentially useful nootropic agents . 137 The potential use of AMPA antagonists in the treatment of addiction is treated in detail in Chapter II. CHAPTER n STATEMENT OF THE PROBLEM AND OBJECTIVES

The functions of the wide distribution of receptors for all three of the major classes o f excitatory amino acid receptors, N-methyl-D-aspartate (NMDA), kainate and quisqualate, [.... are] unclear in the absence of specific antagonists against two o f the receptor classes. - J.S. Kelly and V. Crunelli in Excitatory Amino Acids in Health and Disease (John Wiley, 1988). There remain the quisqualate and kainate receptors, which now seem to often be assigned millisecond excitatory roles by default. We emphasize that [CNQX] is the only antagonist with micromolar potency for excitations evoked by these compounds. All other antagonists have low potency and specificity. We suggest that these receptors be considered kainate-preferring and quisqualate-preferring subtypes and that questions of possible heterogeneity be addressed by seeking more specific antagonists. [...] Lack of potent pathway-specific antagonists is the major obstacle to further progress in studies of millisecond excitatory transmission. [...] We feel justified in reiterating this well-worn appeal. Specifically, novel kainic acid, , N-acetyl-L-aspartyl-L-glutamic acid, and kynurenic acid analogues seem to us to be particularly needed ifjurther advances are to be made in the area o f excitatory amino acid neurotransmission. - R. L. Johnson and J. F, Koemer, Journal of Medicinal Chemistry. 31 (1988) 2057-2066. As discussed in the previous chapter, excitatory amino acid (EAA) transmission is pervasive in the mammalian CNS. Exploiting EAA pathways for therapeutic intervention thus presents strong demands for selectivity. There are no drugs presently in use which have been definitively demonstrated to modulate EAA action; there are thus no empirical leads as there have been in many other drug classes. Those agents so far discovered which have been crafted by evolution to intervene in EAA transmission are toxins, e.g. those found in spiders 142 and plants (see section 1.3). This is not surprising since EAAs play a large role in invertebrate neurotransmission, serving for example as the primary transmitter at the neuromuscular junction. The possible role of amino acid toxins in causing long-term neurological damage in humans was discussed in the 39 40 previous chapter. On the other hand, many of today’s useful therapeutic agents are toxins used in low doses, for example the cardiac glycosides.

Recent molecular biology studies (i.e. cloning and expression of genes obtained from low-stringency screening of genomic libraries) indicate the probable existence of a rich variety of chemically distinct non-NMDA receptor subtypes .47,53,143 The concentration of mRNA for various subtypes varies regionally in the CNS. Diversity among receptor subtypes, relative numbers of subtypes, uptake and storage systems, and metabolic processes is to be expected in a such a ubiquitous transmitter system. This makes the study of EAA pathways exceedingly complex, but at the same time offers the promise of selectivity through interaction with narrow subpopulations of receptors in target regions of the CNS. Molecular biology techniques are now suggesting the basis for some of the empirically-obtained selectivity in other drug classes (such as dopaminergic modulators, see Chapter VI).

Much of EAA pharmacology has been elucidated through the use of selective agonists, many of which are natural products. The availability of a variety of NMDA antagonists has brought about major advances in the study of NMDA pharmacology. Many functional or competetive NMDA antagonists have now been discovered or synthesized and are being intensively investigated for therapeutic potential (see sections 1.3.1 and 1.5). There is however a conspicuous lack of antagonists for non-NMDA EAA receptors. The quest to understand the workings of the brain as well as possible therapeutic applications discussed in the Chapter I make the discovery of such compounds an urgent task. 41

GAMS (42) has been shown to inhibit the electrophysiological effects of both AMPA and KAIN in some studies . 144,145 Some specificity may be obtained, for example intra-accumbens injections of GAMS can inhibit the hypermotility response to intra-accumbens AMPA at doses which do not affect kainate-elicited hypermotility ; 146 similar effects were obtained in the ventral pallidum . 147 On the other hand, Davies and

Watkins33,148 found that GAMS was more potent in antagonizing responses to KAIN than quisqualate (QUIS) in dorsal horn and Renshaw cells in the cat spinal cord. GAMS

N H ^ S03- OEt 1 H o 2n N _ 0 A. A h 2n s o ^ C O , . C 02Et .R x x x , I I H n h 3+ n h 2 45R-CN CNQX 51 42 57 52 R*N0 2 DNQX NBQX GAMS GDEE Figure 18: AMPA antagonists. is only moderately potent, and does not inhibit specific [ 3H]-AMPA binding at concentrations as high as 100 pM . 149 The mechanism of action of GAMS thus remains unknown, although it has recently been reported that GAMS inhibits the specific binding of [ 3H]-CNQX to rat cortical membranes . 55 This result was not duplicated by Wallace et. a/ . 149 Several spider venom components were originally reported to be functional non-NMDA antagonists , 150 but are now thought to antagonize NMDA receptors .249 The quinoxaline-2,3-diones CNQX (45) and DNQX (52),129,151 and and more recently NBQX (51),129,152 have been introduced as competetive antagonists of AMPA receptors ;45,152,153,154,155,156 however DNQX and CNQX are about an order of magnitude less potent in their inhibition of specific [ 3H]-AMPA binding and much less 42 potent in their inhibition of specific [ 3H]-KAIN binding than are AMPA and KAIN, respectively. With the possible exception of NBQX, there is little specificity among the reported quinoxaline-2,3 -diones between KAIN and AMPA receptors in either binding or physiological studies .129,151,152 [3H]-CNQX has been introduced as a radioligand for

H Cl^.N^O 02N

CITT l ^ ^ N ^ O H NO,H 53 6,7-diClQX MNQX H Cl ra v 35 7-CI-KYN

Figure 19. Glycine antagonists. binding studies, and exhibits saturable and reversible binding to rat cortical membranes with an apparent KD of 39 nM .55 AMPA inhibition of [ 3H]-CNQX binding is biphasic, suggesting that CNQX binds with similar affinity to two different sites. Radiation inactivation studies give the same molecular target size for the [ 3H]-CNQX binding site as for [ 3H]-AMPA, but different than for the high affinity [ 3H]-KAIN binding site .55

Some members of this class, especially 6,7-dichloroQX (53 ) 91,254 and MNQX (54) , 152 but also including CNQX and DNQX ,92,157 antagonize the glycine potentiation of

NMDA responses .55 This may be due to their structural relationship to the kynurenic acids (e.g.55 ) 92,156 and quinoxalic acids (e.g. 35 ) , 154 some of which are NMD A/glycine antagonists (section 1.3.1). The quest for potent non-NMDA antagonists which have 43 either AMPA or kainate specificity and no NMDA activity thus continues.

While several areas of therapeutic potential for AMPA antagonists were discussed in Chapter I, biological studies of our compounds have been directed in particular towards the possible role of AMPA receptors in the nucleus accumbens and ventral pallidum in mediating reinforced drug-seeking behavior ("addiction"). The nucleus accumbens is a forebrain region which may serve as a link between the limbic and motor systems, thus being part of the neural circuitry involved in converting emotions into behavior . 158 In this context, the desire (an emotion) for the rewarding effects of an "addicting" drug would produce the behavior of seeking and obtaining the drug, and the rewarding effect of the drug would reinforce this behavior (conditioning). The dopaminergic neurons originating in the ventral tegmental area (VTA) and terminating in the nucleus accumbens seem to provide a key pathway mediating this behavioral pattern . 159 Hence, lesioning of this pathway with 6 -hydroxydopamine or administration of dopamine receptor antagonists in the nucleus accumbens blocks the behavioral manifestations associated with the administration of a number of addictive drugs. These include conditioned place preference, drug self-administration by lever pressing, and hypermotility responses to systemic or intra-accumbens amphetamine, cocaine, and p opioid receptor agonists ( l o w doses ) .160,161,162 EAA agonists such as NMDA, kainate, and AMPA injected into the nucleus accumbens stimulate hypermotility , 163 and this response is blocked by corresponding antagonists to each receptor subtype . 146,164,165

Although the relationship between addictive behavior and the hypermotility response elicited by these drugs is not entirely clear, the neural substrates that mediate the rewarding behavior of psychostimulant drugs also seem to mediate stimulation of locomotion, and this response may represent goal-directed behavior . 166 AMPA receptors seem to be particularly important for the effects of amphetamine and cocaine, since DNQX (52) or GAMS (42) injected into the nucleus accumbens or ventral

pallidum were found to block the stimulated locomotion produced by systemic

amphetamine, and DNQX blocks the hypermotility response to cocaine . 191 There is a

recent report that intra-accumbens GDEE (57). a weak AMPA antagonist, also blocks the

hypermotility response to amphetamine . 167 All of this evidence indicates that AMPA

antagonists which could act selectively in the nucleus accumbens might be useful therapeutic agents for addictive disorders.

O O O HN-f' NO 0^N-0 HO-^CH, 0=%.NH

NH3+ NH3+ n h 3+ n h 3+

5 9 58 5? QUIS AMPA Figure 20. Ring-modified analogs of quisqualate.

With the exception of our work (Chapter IV), there are presently no amino acid AMPA

antagonists reported in the literature which competetively inhibit [ 3H]-AMPA binding. We set out to add to the understanding of structure-activity relationships for AMPA

agonists in order to discover new lead compounds which might, through structural modifications, be converted to antagonists. The most potent AMPA agonists identified to date have 5-membered heterocycles in place of the y-carboxylate group of glutamate,

i.e. AMPA (9 ) 168 and quisqualate (5). Structural modifications of the quisqualate

heterocyle (Figure 20) to give the hydantoin (58) and 3,5-dioxo-l,2,4-triazolidine analog

(59) resulted in complete loss of activity in depolarizing the locust neuromuscular junction . 169 The pKa of the hydantoin was found to be 8.2; hence low acidity is one 45 possible explanation given by Boden et. a l for the inactivity of 58 (this explanation is unsatisfying given the high activity of willardiine, see discussion below and in Chapter IV). The pKa of the triazolidinedione 59 was found to be 4.7, so low acidity could not be invoked in this case. Quisqualic acid (the neutral form) was found in X-ray studies to assume a pyramidal geometry at the nitrogen to which the amino acid side chain is attached, whereas in the hydantoin 58 the attachment was planar. Calculations also indicated a preference for a planar ring attachment in 59. The authors proposed that the high activity of QUIS relative to the other two compounds was due to this difference in ring junction geometry, and that the unexpectedly high activity of d-QUIS was a result of pyramidal inversion. This explanation seems highly unsatisfying considering the sp 2 ring attachment in AMPA (9) (unfortunately AMPA was not tested in this system), and a more likely explanation may be substantial differences in the electronic structure of the heterocyclic anions.

h o ~ * n -o N -0 N -0 -o?c N O O CH, HO ^ s j/^ C H 3 H O ^ s ^ R O h , co2- co,- NO K y C O r NH3+ n h 3+ O ^ S j ^ C H 3 nh 3+ k ^ C 0 2 60 R= t-Bu 64 61 R= CH2Br n h 3+ AMOA 62 R= Ph 63 AMNH Figure 21. SAR of AMPA analogs.

Structural modifications at the 5 position of AMPA (Figure 21) have given agonists

(60-61)152,170,171 or partial agonists (62).170 This position seems relatively insensitive to 46 substitution and is probably external to the binding cleft, since an affinity column linked through the 5-methyl group of AMPA was very effective in receptor purification studies. 172 Recently, compounds 63 and 64 have been reported to be weak AMPA antagonists, though the activity of 64 is very low . 173 A number of cyclic

("conformationally restricted") AMPA analogs have also been reported (Figure

22) . 152’174’175 Compounds 65 ("5-HPCA") and 6 6 ("7-HPCA") are AMPA agonists,

O — N

H N + H 65 66 5-HPCA 7-HPCA

OH

67 68 4-HPCA 6 -HPCA

Figure 22. Conformationally-restricted AMPA analogs. whereas 67 ("4-HPCA) and 6 8 ("6 -HPCA") are inactive at any EAA receptor subtype.

Neither 65 or 6 6 are as potent as AMPA, but X-ray studies of these compounds and AMPA do provide some sense of the active conformation of this molecule.

Because modifications of the five-membered ring compounds have so far not produced antagonists and the amino acid portion of the molecule is known to be very resistant to modification , 176 we decided to concentrate on 6 -membered ring systems. Willardiine (70) is a fJ-(l-) substituted alanine which is almost equipotent with glutamate in binding ([ 3H]-AMPA, see Chapter IV) and physiological studies . 177 It was also reported that 6 -hydroxydopa (141) potently depolarizes frog spinal neurons . 179,180 We postulated that the phenylate anion corresponding to the ortho (2 ’-) hydroxyl group in these compounds might serve as a bioisostere for the y-carboxyl group in glutamate, as the exocyclic oxygen in AMPA presumably does (Figure 23). This is not a new idea, since

OH

NH3+ n h 3+ n h 3+ 141 1 9 6 -OH-DOPA GLU AMPA O

n h 3+ n h 3+ 70 5 WDN QUIS Figure 23: Comparison of the anionic structures of some AMPA agonists.

Curtis and Watkins (I960)3 tested the activity of o-tyrosine (70) in their early SAR studies. They found, however, that 70 could not depolarize cat dorsal hom intemeurones, Renshaw cells, or motoneurones. The pKa of the o-tyrosine phenolic hydroxyl is 9-10,181 hence the lack of activity is not surprising; however willardiine has a pKa above 9 (chapters in, IV), and yet retains good biological activity. Nitrotyrosines

73 and 74 were studied for the pH-dependence of their uptake in Ehrlich ascites tumor cells via the glutamic acid uptake system .183 The rate of uptake was well correlated with 141 70 71 72 6-OH-DOPA o-TYR ro-TYR /j-TYR

n o 2 OH

0 2n HO HO C 02- C02-

NH3+ n h 3+ co2- 73 74 75 n = 3 or 5 R=H or Cl

the degree of protonation of these two compounds in pH-dependence studies, as it was for glutamate. More recently, phenyl rings have been used as bioisosteric replacements

of carboxyl groups in developing artificial sweeteners. Lipinski et. a/ . 184 synthesized

the NMDA antagonists 75, in which the ortAo-mercaptophenols are evidentally isosteric with the terminal phosphonate group of AP7 (20). Curiously, the sulfone and sulfoxide analogs were much less active; the opposite might be expected from the relative acidities of these analogs.

The compounds DL-ortho-, meta- and para-tyrosine (70-72). obtained commercially,

were unable to inhibit specific [ 3H]-AMPA binding in rat forebrain homogenates (see

Chapter IV). Since the phenyl ring in o-tyrosine (70) appeared to offer a flexible template on which to build potential antagonists, our objective was to try to increase the activity via increased acidity (lower pKJ. Some ways in which this might be 49

OH 0 2N

HO HO pKa-9.2 pKa-7.2 pKa-7.2

H 0 2N

HO HS pKa-4.1 pKa-7.3 PK.-6.5

Figure 24. Some substituted phenols with increased acidity.

accomplished are shown in Figure 24. Nitro groups were chosen as a starting point.

Homoibotenic acid (HIBO) is not a very potent inhibitor of [ 3H]-AMPA specific binding in rat brain homogenates, but 5-bromo-HIBO (76) and 5-methyl HIBO (77) are

approximately equipotent with glutamate, only about an order of magnitude less potent

than AMPA in competing for [ 3H]-AMPA specific binding in rat brain

homogenates . 185,186 p-Oxalylaminoalanine (BOAA, 50), a neurotoxin discussed in

Chapter I,shows comparable activity. These compounds are glutamate homologs (Figure 25), with an additional atom between the amino acid and the terminal acidic groups. Substituted m-tyrosines with increased acidity are similar homologs, and we synthesized and tested several such compounds.

As discussed above, several quinoxaline-2,3-diones (e.g. 45, 51, 52, and 78) are reported to be potent, competetive AMPA antagonists . 129,151,152,188 Affinity label analogs of these compounds might be useful in receptor isolation and characterization, particularly since the strong intrinsic fluorescence characteristics might obviate the need for radiolabelling. 50

HO HO -o o -N = N r HO O NH co2- V n h 3+ n h 3+ NH3+ NH3+ 76 77 50 71 Br-HTOO Me-HIBO BOAA m-TYR

Figure 25. Glutamate homologs with AMPA activity.

Such affinity labels might also be useful as irreversible inhibitors of AMPA and/or kainate receptors in vivo. It was therefore our objective to synthesize compounds such as those shown in Figure 26, where analogs 81 (R = CH3, F, N02, C02\ ...) are potential photo affinity labels and would require photoactivation by UV light. Syntheses were

H H O ° * N 0 N xO :x°o H n o 2 h

45 R=CN CNQX 51 78 52 R=N02 DNQX NBQX designed in such a way as to make additional analogs easily obtainable through a similar route, and to make intermediates which would provide further structure activity information.

The quinoxaline-2,3-diones are not amino acids and are an order of magnitude less potent than AMPA or quisqualate in inhibiting specific [ 3H]-AMPA binding to rat brain homogenates. It should therefore be possible to make antagonists which are much more potent, since the rate theory of receptor binding predicts that antagonists should 51

H H H

x c h 2c o n :

H H H 7? 80 81 X=Br, I, Cl

Figure 26. Potential affinity and photoaffinity analogs of the 2,3-quinoxalinedione series.

generally exhibit higher binding affinities than agonists (for an excellent discussion of these kinetic considerations in neuiotransmitter systems, see ref. 57). We were intrigued by the resemblence of the quinoxalinedione anion 82 to that of BOAA, 50 (Figure 27), and set out to synthesize compounds having the general structure 83. At the time we began this work there was no information regarding the effects of N-substitution on the activity of the quinoxaline-2,3-diones, though it is now known that monomethylation

results in only slight loss of activity, whereas N,N-dimethylation abolishes activity .92

The latter would be expected if the activity depends on the formation of an anion (Figure

27). Assuming biological activity was retained or enhanced, we would then proceed to

compare the structure-activity relationships within this new series to those of the parent compounds.

NH3+ 50 83

Figure 27. Rationale for the synthesis of "QXAA’s". 52

Summary of Ob jectives: Briefly, the objectives of this research were: 1) to further investigate the structural requirements for AMPA receptor binding,

particularly with respect to the isostere for the y-carboxylate group of glutamate, and

determinants of AMPA vs. selectivity; 2) to use this information to attempt to develop AMPA antagonists; 3) to synthesize potential affinity and photoaffinity analogs of the quinoxaline-2,3-diones, and father investigate structure-activity relationships in this series. CHAPTER m CHEMISTRY

I have not failed; I have only discovered 10,000 ways that didn’t work - Thomas Edison.

3.1 Hydroxyphenylalanines and related compounds 3.1.1 Substituted o-tyrosines

The syntheses of 5 ’-hydroxy-2 ’ -nitro-D jL-phenylalanine (74 ) 192 and

2’-hydroxy-5’-nitro-D,L-phenylalanine (73 ) 193 have been previously reported (Scheme I). We wanted to obtain both 73 and 83 for biological testing. The synthesis of

3’-nitrotyrosine was reported by Dall’Asta and Farrario , 194 and these conditions were adapted for the nitration of commercially-avialable D^-o-tyrosine (Scheme II). A small amount of the dinitro product was obtained with this procedure, and we attempted to isolate all three products. The N-acetyl and N-trifluoroacetyl derivatives were difficult to separate on silica owing to the high polarity of the former and lability of the latter. The t-butyloxycarbonyl-protected derivatives 84 and 85 were readily prepared using

BOC-ON195 (Scheme II) and were separable by thin layer chromatography (TLC) using

7% CH3OH/7% isopropanol/0.7% AcOH/CHCl 3 (Rf of 84 - 0.6, Rf of 85 = 0.4, and Rf for the dinitro analog = 0.25). Acetic acid was added to reduce band tailing caused by the free carboxylic acid. was the optimum weak solvent but could be substituted with dichloromethane, while attempts to eliminate either of the two alcohols gave poor separations. It was found both on TLC and later on a preparative column that under conditions of high sample loading the Rf of the dinitro compound increased to 53 54 SCHEME I. Synthesis of 2’-hydroxy-5’-nitrophenylalanine and 5*-hydroxy-2'-nitrophenylalanine.

NO, NO, NBS CH cat. ROOR CH3° CH, CCL CH2Br

NO NO AcNHCH(COzEt)2 48% HBr HO ► CO,- NaOEt EtOH NHAc 73

HO ch 3o TpL NO, co2- CH, nh 3+ 74

0.4-0.5, rendering the separation useless. The preparative chromatography was optimal with a somewhat weaker mobile phase and slightly higher acetic acid concentration, 8 % isopropanol/2% CH 3OH/l% AcOH/CHC13. No more than about one gram of mixture dissolved in ethyl acetate or 5% isopropanol/CHCl 3 was chromatographed on a 51 mm ID x 39 cm L Michel-Miller column (Ace Glass) packed with 40-63 pm silica (E. Merck). After eluting the two mononitro compounds the eluent strength was increased to 30% methanol/10% isopropanol to elute the dinitro compound. These conditions gave essentially baseline resolution of all three BOC derivatives. 55

SCHEME n. Synthesis of o- and p-nitro-o-tyrosines.

^ , N 0 2 1) 30% TFA cHacit^ HOjf i r L C02H 2)IEXChrom. k .C 0 2- 1) 23% HN03 « (75%) * r i l 3 S3 NHBOC NH3+ JUJJ 2 h/0 °C HO XI prtr o _ ------►<' Silica suica 23% from 73 T* 2 2 )IE X O ro m X P mMI,atoS- ° 't>rrosme NHa+ 3) BOC-ON Et-,N °2N ^ 1) 30% TFA ° 2N 131,1 I p ) CH2C12 H20/Dioxane HO ks ,C 02H 2)Et3N 0 . NHBOC (83%) 84 NH3+ BOC-ON =" tf N 7.2% from o-tyrosine

The two BOC-protected mononitro compounds were readily crystallized from ethyl acetate/hexane, but attempts to isolate the small quantities of dinitro derivative as a pure

solid were unsuccessful. Deprotection of 84 and 85 proceeded in high yields using 30%

trifluoroacetic acid in methylene chloride . 196 The trifluoracetic acid salt of the p~N0 2

compound 73 crystallized during this deprotection and was isolated directly. This salt could be converted to the free zwitterion by cation exchange chromatography. Isolation

of the o-N0 2 compound 83 required neutralization with triethylamine at the end of the deprotection reaction, which gave the free zwitterion directly and in pure form. The overall yields obtained using the method of Scheme II were not high, particularly for the o-N0 2 compound 83; however, only three synthetic steps were required instead of six if the two compounds had been synthesized seperately as in Scheme I. 56

Catalytic La (III ) 197 was added in an attempt to bring about selective ortho nitration, using the same conditions otherwise. While this was unsuccessful, this catalyst has been successfully applied only under conditions of phase transfer , 198 which cannot be accomplished with underivatized amino acids. Conditions were later developed (see below) which gave a greater proportion of ortho nitration.

To attempt to increase the yield of the dinitro compound 87 the reaction conditions were varied. Heating with the same nitric acid concentration (23%) resulted in degradation of the amino acids, as noted by reversed-phase HPLC (c.f. chromatogram Figure 28). The reaction time was extended to 24 hours in 23% HN0 3 and another 24 hours in 32%

HN03 at room temperature, and the dinitro compound 87 was then isolated (in less than 10% yield) by anion exchange separation from the mononitro compounds 73 and 83.

This anion exchange separation took advantage of the difference in the pKa of the mononitro (~ 6-7) and dinitro (3.4) compounds, where the former were selectively eluted from the dimethylhydroxyethylamine resin with pH 5.0 acetate buffer. Increasing the nitric acid concentration to 40% or 50% increased the ratio of dinitro to mononitro products but did not increase the yield. An attempt to nitrate the BOC-protected

/?-N02-o-tyrosine 85 using catalytic La3+ (c.f. ref. 199) gave only the deprotected starting material (Scheme HI). After treating o-tyrosine with 32% aqueous HN0 3 for 2.5 h at

0°C followed by 60% HN0 3 for 1 h at 0°C and warming this solution to room temperature over the next 7 h, none of the three amino acids remained. One product, shown to be 8 8 (see below), was isolated from this mixture in 32% yield by extraction with 10% CH3CN/EtOAc.

The product 8 8 might be formed by both pathways given in Scheme IV; monitoring the progress of this reaction by HPLC (Figure 28) did not readily reveal which of these paths 57 SCHEME m. Attempted synthesis BOC-dinitro-o-tyrosine. NO

n o 2 HO

c o 2h NHBOC

NHBOC NO 85 HO

predominated, and rigorous kinetic analysis was not attempted. It is probable that the replacement of the amino group by hydroxyl occurs through diazotization of the amine by HONO (nitrous acid) present in 60% nitric acid; such chemistry is known in the generation of a-hydroxy acids , 197 and forms the basis for the van Slyke method of determining free amino groups in amino acids, peptides, and proteins. The formation of a-fluoro, a-chloro, and a-bromo acids may also be accomplished in this manner .200*201

Studies to verify this mechanism have not yet been carried out

The formula weight and elemental composition of 8 8 were established by mass spectroscopy (El) and elemental analysis, respectively, and this information combined with the proton and 13C NMR indicated the structure to be either 8 8 or 89, where 89 could have been formed by elimination of the (presumed) diazo group and 1,4-readdition of water. The striking similarity of the !H-NMR spectrum to that of the amino acid, particularly the diasteriotopic separation of the (later confirmed) methylene protons strongly suggested that 8 8 was the correct structure (Figure 29). A major fragmentation 58

NO HO ,CO,H HO COjH

OH NO,

HO CO,-

HO

20 16 12 8 4 0 min. Figure 28. Chromatographic separation of nitro-o-tyrosines and degradation products in aqueous nitration. Conditions: 5 pm end-capped C- 8 , 4.6 mm ID x 15 cm L 8 % CH3CN / 0.1% trifluoroacetic acid/H20 from 0 to 8 min. Linear gradient 8 % to 70% CH3CN from 8 to 18 min. 70% CH3CN from 18 to 23 min. Flow rate: 2.0 ml/min. Detection: UV at 254 nm

pathway in the mass spectrum (Figure 30) was loss of water, though a peak corresponding to the loss of H 02CCHO was observed. "COLOC" ( 13C-*H multiple bond correlation) studies 202*203 using various mixing times in the pulse sequence were ultimately ambiguous in assigning coupling between the side chain carbons and

6 ’-proton of the aromatic ring and coupling between the aliphatic protons and the appropriate aromatic carbons. Part of the difficulty resulted from the unusually small coupling constants observed for these interactions. Ultimately the carbon-carbon connectivity was verified using a 13C INADEQUATE (double quantum transfer) experiment (Figure 31).204 This required a 54 hour collection time with ca. 1 g/4 ml of 59

SCHEME IV: Proposed mechanism for formation of 88.

60% HNO 2h/0°C 0,N NO NO 5h/R.T. HO HO HO CO,H NH3+ OH

88

NO- NO HO HO HO

N2+ OH N2+

o 2n 0 2N NO,

HO 9 2 or HO f i r \ c o 2h H0 ^ C 0 2H HO 88 89 sample in a 10 mm NMR tube.

Conditions were sought to minimize the pathway leading to the formation of 8 8 in order to increase the yields of dinitro-o-tyrosine, 87. The starting material was adsorbed from alcoholic ammonium hydroxide solution onto anhydrous sulfate ( 1 0 -fold excess by weight), which was then ground to a fine powder and dried in an oven. This material was incubated for 7 h at room temperature in a chamber containing an open dish 60

•HO,

HD' ,CO.H

OK

JO,

Figure 29.250 MHz 1H-NMR spectrum of compounds 87 and 8 8 . of fuming nitric acid, giving the desired product in ca. 44% yield and free of mononitrated products (Scheme V). Despite this success, the procedure required substantial amounts of costly fuming nitric acid, and the reaction proved difficult to scale. It was found that o-tyrosine could be nitrated in high yield using 85% nitronium tetrafluoroborate 205 in reagent grade acetonitrile at 0°C (Scheme V). The compound proved difficult to isolate as the free acid, undoubtedly due to the low pKa (3.4). The material isolated by cation exchange chromatography, eluting with ammonium hydroxide, appeared very pure by ^-N M R and reversed-phase HPLC; however, the elemental analyses were inappropriate, indicating a mixture of free acid and ammonium salt. A sample of this material which was dried in vacuo with heating gained water with time, as evidenced by changes in the elemental analysis without evidence of degradation in the analytical profile (!H-NMR, HPLC). The compound was ultimately isolated as 61

234

Figure 30. High resolution El mass spectrum of compound 8 8 .

the hydrated ammonium salt by reducing the volume of an aqueous ammonia solution to the point of crystallization, then adding isopropanol and . Salt formation

was verified by the presence of a broad peak in the 'H-NMR centered near 7.5 ppm, and

integrating for the ammonium protons and the RNH3+ protons from the amino acid

(de-DMSO). The unusual hydration state ( 5/ 2 H20) was supported by molecular weight

confirmation with fast atom bombardardment (FAB) mass spectroscopy. A separately-prepared lot of material assumed the same hydration state, which was found to be stable for at least one year.

Because high yields of 87 were obtained using N0 2+BF4\ the mononitro compounds 73 and 83 were again synthesized using one equivalent of this reagent added slowly at 1

1

' 1 5- NO-

HO 1 * 1 ___ OH

•

HERTZ ' I " ...... ■ |---- I ■ I...... I...... 1...... 1...... 160 140 120 100 60 60 40 PPM

Figure 31.500 MHz 2D 13C INADEQUATE spectrum of compound 8 8 . 63

SCHEME V: Syntheses of dinitro-o-tyrosine.

1) Adsorbed dry on MgS0 4 over fuming HN03 (44%) o 2N n o 2 or HO N 0 2BF4 in dry CH3CN NH4+ _Q •5/2h 2o co2- 2h/0°C (6 8 %) C 02H ► n h 3+ 2) Cation Exch. Chrom. NH3+

-20°C (Scheme VI). The formation of some dinitro product was observed by HPLC, hence some starting material remained unreacted, and it was difficult to decide when to

stop adding the N 0 2+BF4'. The solid mixture isolated from this reaction contained 33:22:45 83:73:unreacted starting material on a molar basis, representing 22% conversion to the o-nitro and 14% conversion to p-nitro compounds. The

BOC-derivatized starting material eluted between the two mononitro products using the TLC and preparative systems described above (Rf * 0.4, 0.5,0.6 for 85, BOC-S.M., and 84); however this did not degrade the separation of - 2 g of BOC-derivatized mixture on the column described for Scheme II due to the sharp band shapes for 84 and the

BOC-protected starting material. The yield of 84, the o-nitro product, was substantially improved by this method, otherwise there was no particular advantage. At lower temperatures it might be possible to drive the mononitrations further to completion while avoiding substantial dinitration.

3.1.2 Mononitro- and dinitro-m-tyrosines

Due to the successful dinitration of o-tyrosine, and interesting biological results obtained for compound 87, we decided to attempt the nitration of m-tyrosine. While it was expected that 90 would be obtained almost exclusively, three products (90-92) were observed by HPLC (see chromatogram Figure 32) and by ‘H-NMR of the solid product SCHEME VI: Alternate synthesis ofo - and p-nitro-o-tyrosines. NO NO HO HO 1 2 % conv. 17% yield 73 14% from o-TYR NHBOC 1) 1 eq. N 0 2+BF4- 40 min./-20°C 1) BOC-ON ► CO; i" HO 2) IEX Chrom. HO 2) Silica n h 3+ COo- Chromat. 15% conv. 22% NH,+ 2 1 % yield NHBOC from o-TYR

HO CO,- HO 30% NHBOC 65

SCHEME VII: Synthesis of dinitro-m-tyrosines. NO o - n h 4+

18% 1) n o 2b f 4 NO, n h 3+ OH c h 3c n x . o - n h 4+

27h/0°C o 2n 91 CO, C-18 2,4-DNmT _ co,- chrom. NH3+ 2) Cation * .. V Exch. Chrom. 2 ' * NH3+ O- NH4+ 90 0 ,N 4,6-DNmT

7% n h 3+

2,6-DNmT

mixture recovered from a cation exchange column. The isolation of all three products was facilitated by the discovery that the anhydrous ammonium salt of 90 had a much

lower aqueous solubility than the other two products and could be crystallized from

solution almost quantitatively, leaving a mixture of the other two products. Fractional

crystallization failed to readily separate the remaining two compounds, but the mixture could be obtained as a solid and separated in 50 mg portions on a 47 mm ID x 37 cm L Michel Miller medium pressure LC column packed with C-18-derivitized 40-63 pM silica and eluted with 15% CH3CN/1% CF3C 0 2H/H20. The separation is shown in

Figure 33. Some material had to be recycled, but the separation was actually quite good considering the modest separation obtained on an analytical column. The yields of 91 and 92 shown in Scheme VII represent isolated yields after preparative chromatography, and the actual yields were higher. Depending on the exact conditions used in the nitration, especially the rate at which the N0 2+BF4' was added, either 90 or 91 was the 66

OH

o 2h co,-

,0H

0 ,N

HH3+

KMu u .

0 4 8 min. Figure 32. Chromatographic separation of the three dinitro-m-tyrosines 90,91, and 92. Conditions: Column: 5pm end-capped C- 8 , 4.6 mm x 15 cm L Mobile phase: 4% CH3CN / 0.1 % CF3C02H / H20 , 2.0 ml/min. Detection: 254 nm UV. About 2.5 pg of each compound injected. predominant product. The unexpectedly 91, in which the aromatic ring is

1,2,3,4-tetrasubstituted, evidentally reflects the propensity of N0 2+BF4' to effect ortho nitration (Scheme VI). The formation of 1,2,3,4-substituted 92, also unexpected, seemed most likely to arise from initial o-nitration to form the 1,2,3-substituted intermediate 94 (Scheme Vm), however further experiments indicate that this is not exclusively the case (see below). 67 NO HO HO 0 ,N

NH NH

20 40 00 80 Figure 33. Preparative reversed-phase chromatographic separation of dinitro-m-tyrosines 91 and 92. Conditions: 15% CH 3CN/ 0.2% CF3C 02H / H20 - 20 ml/min. Column: 40-63 pM C-18, 47 mm ID x 38 cm L. Detection: UV @ 280-310 nm; 0.1 cm cell. 60 mg/1 ml sample injected.

It was impossible to choose the structural assignments between 91 and 92, since the !H-NMR spectra were nearly identical. Because the preparative separation of these compounds was not practical on a large scale, it was envisioned that they might be synthesized (unambiguously) by further nitrating the two mononitro compounds 74 and

93 to give mixtures of 90 with each of the other two dinitrated products (as in Scheme

VIII). As noted above, the synthesis of 74 was reported in the literature (see Scheme I).

Compound 93 was synthesized by the same route, through the known intermediate

93a ,257 and nitrated to give the expected mixture of 90 and 91. These compounds were separated, without resorting to chromatography, by fractional crystallization alone

(Scheme IX). In this nitration, the HBr salt could not be used because additional (unidentified) products were produced in substantial yield. Also, more than one SCHEME VIII: Nitration of m-tyrosine. NO OH

91 NO OH

co2- 90

OH

CO,-

92 SCHEME IX: Synthesis of 3’-hydroxy-2’,4’-dinitrophenylalanine and 5’-hydroxy-2’,4’-dinitrophenylalanine.

N 02 NO, N 0 2 OCH, AcNHCH(C02Et)2 i5'°CHi " . S y ► t y * cat. PhC02OCOPh NaOEt CH, CC14 CH2Br EtOH I (C02Et)2 18 hr reflux 10 min. R.T. NHAc 93a 36% 71%

n o 2 NO NO A . OH O- NH4+ i ) 2 m n o 2b f 4 48% HBr (2 eq.) ^

V ° r d[y0™iCN r + 21% 35% 91% NH3+ ! fa n h 3+ 2) Cation exchange 93 chromatography 92 3) Fractional recryst. 70 equivalent of N 0 2 +BF4 " was needed to ensure complete reaction of the starting material 93, which was otherwise difficult to separate from the product 91. The required excess seemed to depend on the rate of NC> 2 +BF4 "addition, in that more rapid addition required less nitrating agent to complete the reaction. The known compound 74 was nitrated to give 90 and 92 (Scheme X).

SCHEME X: Synthesis of S’-hydroxy^’^’-dinitrophenylalanine and 3’-hydroxy-2*,6*-dinitrophenylalanine. O- NH4+

H O ^ \ ° 2N ^ CH30 V\] 1)2M N02BF4 XT*] S ^ n o 2 - ► V S o 2 L r n (2eq.) 90 NH3+ CH3 T " l u 2' dry CH3CN NH3+ 0°C 74 1 hr o- n h 4+ 2) Cation Exch. o 2n fx Chroma tog. ‘ co2- 92 NH3+

3.1.3 Chiral synthesis of dinitro-o-tyrosine

Because of the interesting biological activity of dinitro-o-tyrosine (87). namely in vivo antagonism of AMPA-stimulated locomotion in the nucleus accumbens and ventral pallidum, it is critical to compare the enantioselectivity of this activity with that established for the agonists AMPA, quisqualic acid, willardiine and glutamate (see discussion Section 4.1), namely l » D.252 The work is still in progress, but it is envisioned that each enantiomer of o-tyrosine will be nitrated without loss of chirality using nitronium tetrafluoroborate in dry acetonitrile (c.f. section 3.1.1). o-Tyrosine has been resolved chromatographically 206 on an analytical scale, and preparatively by 71

fractional crystallization using chiral binaphthyl phosphoric acid .207 Other potentially

useful approaches include addition of organometallics to chiral serine lactones ,208

N-protected p-iodoalanine esters ,209 aziridine-2-carboxyIates ,210 or chiral nickel

complexes of dehydroalanine 211 Alkylation of benzyl halides or tosylates with chiral

glycine equivalents is a widely-exploited approach to the synthesis of similar

molecules .212,213,214 Maurer et. aL215 have reported the synthesis of D-dopa using L-serine as a chiral educt. The first three steps in the corresponding synthesis of

(d )-o-tyrosine have been completed (Scheme XI). It will be critical to determine

whether the nitration conditions (using N0 2+BF4‘) will preserve the chirality at the a carbon. If so, the L-enantiomer whould be readily synthesized starting from inexpensive

D-serine.

3.2 Willardiine and S-nitrowillardiine For reasons discussed in sections 2.1 and 4.1, we wanted to obtain both willardiine and S-nitrowillardiine for biological testing. Willardiine (69), a natural product first isolated from Acacia willardiana, has been synthesized by several routes .196,216,217 Dewar and

Shaw216 reported the unambiguous synthesis of this natural product and prepared the natural (-) enantiomer by crystallization with (+)-a-methylphenethylamine. The synthesis of Martinez and Lee , 198 based in part on that of Dewar and Shaw 216 (Scheme

XII), is the most straightforward and highest-yielding. This synthesis is unusual 218 in effecting exclusively 1-monoalkylation of uracil (with some dialkylation); given the moderate excess of sodium hydride, it is evidentally the monoanion (as opposed to the dianion) which alkylates in this fashion. Our yield in preparing the diethyl acetal 95 was

50% (based on uracil) compared with 37% reported by Martinez and Lee, with the modifications that twice the volume of DMF was used due to the low solubility of the uracil anion, the reaction time was extended, and the extraction in the workup was done SCHEME XI: Synthesisof (D)-o-tyrosine.

OH OH A HO S 6 eq O o C H , PhS02Cl Li h ! VT » H'-Js, H +H3Nj *w fO-,. C 02- sat. sat. Na-iCO-iNa2C 0 3 Ducn phS02NH mu »* ^ c ° 2‘ THF PhS02NH -78°C O OCH3 76% 43% L-serine

Et3SiH 0 2 48% HBr ► OCH3 HO c f 3c o 2h Pt02 Reflux HO <’0 ,- NHS02Ph H NH S02Ph

(36%)

K> 73

SCHEME XII. Synthesis of willardiine and 5-nitrowillardiine.

0 i)uar X? 4^f 0 N 2) BrCH2CH(OEt)2 * \ v/ 0Et ^ ° H I 4 h A 50% ^ 85% H 95

1) KCN O O n o 2 " V s h n ^ FuRmt g h n V 4 h/60° C h N° 3 o N O N'

2) 6 N HC1 k / C° 2- h 2S ° 4 * \ ^ C°2- 3 h reflux 49% I R T - 67% ^ ION HQ NH3 3 1 h reflux 59 96

O O 1NHC1 O u

hn 'YN°2 ™'VNO j A JJ 1) 1 .2 eq. NaH 1 .5 h O N ► u " 1 OH H DMF k OEt , \ / ^ 2) BrCH2CH(OEt) 2 / \ octf. OH A 40% OEt ' x 85% 4 M NHXl 99 A 98 with 2 :1 :ethylacetate instead of ethyl acetate alone.

Compound 96, 5-nitrowillardiine, was prepared as a synthetic intermediate by Martinez,

Lee, and Goodman 219 by nitration of willardiine with fuming nitric acid in sulfuric acid in 8 6 % yield. We readily repeated this preparation (Scheme XII). Willardiine was not 74 nitrated with nitronium tetrafluoroborate in acetonitrile, under conditions similar to those described in the previous section. This was surprising due to a report that uracil nucleosides could be nitrated with a 0.5 M solution of this reagent in sulfolane .220

Willardiine was indeed nitrated with 0.5 M N02+BF4‘ in sulfolane (Scheme XIII), but

Scheme XIII: Alternate nitrations of willardiine.

O 1 eq. N0 2+BF4- l] ► No Reaction o * V c h 3c n COr 0°c - R.T. S t" 24 h n h 3+

2 no2*bf4- S no, 2 ^ 0 5 ^ + unreacted S.M. k. / C02- Sulfolane i C 02- T 4 days/R.T. J NII3+ NH3+ the yield was less than with the Scheme XII procedure and reaction was not complete in

4 days, giving a product that was contaminated with willardiine which was difficult to remove.

Martinez, Lee and Goodman 219 report that the synthesis of S-nitrowillardiine from

5-nitrouracil using the same path as for willardiine failed in the Strecker reaction of the aldehyde (Scheme XII). We were also unable to obtain 5-nitrowillardiine by this route. The aldehyde 97 is unusual in that the !H-NMR in dg-DMSO indicates exclusively the hydrated form 98. Further study did not reveal the reason for the failure of this Strecker reaction, but it is not too surprising considering the complex chemistry observed in the

Strecker reaction of 130 (section 3.3). 75

O O

3.3 2.3-Quinoxalinediones

The synthesis of 2,3-quinoxalinediones was initially begun in an attempt to develop affinity and/or photoaffinity analogs of the compounds DNQX (52) and CNQX (45), as discussed in Chapter n. The syntheses were designed in such a way that additional analogs would be prepared as intermediates and could be tested for biological activity to further elucidate important structure-activity elements in this class of compounds.

SCHEME XIV: Synthesis of various 6-fluoroquinoxaline-2)3-diones. H

N O p WU2 HCl/EtOH * ► < NH2 2. NaOH THF Reflux H 3!h02CC02H iq2 101 78% ca. 1 0 0 % H H 80%HNO3 F'T

j 0 4 H 2. NH.OH 105 2. NaN3 91% 83% H H

NaOAc H20 H H H 106 100 85% 65% 76

Our first target compound was 6-azido-7-fluoro-1,4-dihydroquinoxaline-2,3-dione, 100. It was anticipated and later confirmed that this compound might be synthesized from

6 -fluoroquinoxaline- 2 ,3 -dione, 1 0 1 , a compound reported in the patent literature ,221 by a

nitration/ reduction/ diazotization sequence (Scheme XIV). Quinoxaline-2,3-diones are

typically obtained from the corresponding diamines by reaction with oxalyl chloride,

oxalic acid, or diethyloxalate .222 The required 4-fluorophenylenediamine (102) was readily obtained by stannous chloride reduction of the commercially-available 4-fiuoro-2-nitroaniline and isolated as the hydrochloride or, more conveniently, the oxalate salt; the salt could then be reconverted to the free diamine at the time of use. The quinoxalinedione was not obtained from this diamine using oxalyl chloride under

several sets of conditions; instead a mixture of products was evident by HPLC. No

significant amount of product was formed using oxalic acid in boiling benzene, or

diethyl oxalate in refluxing ethanol .223,224 The desired product was obtained in the latter

procedure when the ethanol was allowed to boil off to leave a thick paste of the diamine and diethyloxalate, though the yield was less than 50%. After trying several variations on this procedure, it was discovered that almost quantitative yields were obtained using

an eightfold excess of diethyl oxalate and an equivalent volume of THF, and refluxing

for several days. This procedure proved to be very convenient and very general; for example 3,4-dimethyl-1,2-phenylenediamine was readily converted to 6,7-dimethylquinoxalinedione, 103, and a number of other examples appear in the following schemes.

Nitration of 6 -fluoroquinoxalinedione 101 did not proceed in 70% aqueous nitric acid, possibly due to the extremely low solubility of this compound; however increasing the nitric acid concentration through the addition of an equal volume of fuming nitric acid 77 gave 104 almost quantitatively after two hours at 0°C. Pure product was isolated simply by pouring the reaction mixture on ice. The same nitration conditions were also applied to three other quinoxaline-2,3-diones (Schemes XV and XVIII), which also gave nearly quantitative yields.

Stannous chloride reduction of 104 was chosen rather than hydrogenation since this material is poorly soluble in or water, and the HC1 salt of the product could be isolated directly by filtration from the chilled reaction mixture in high yield. Reductions of various nitroquinoxalinediones (Schemes XV and XVIII) gave comparable results.

These products could be converted to the free diamines by heating in alcoholic ammonium hydroxide.

Diazotization of the amine 105 in HC1 solution and displacement with sodium azide readily gave 6-azido-7-fluoroquinoxalinedione 100. The comparable displacement using

CuCN to make the 6 -cyano analog was unsuccessful under several sets of conditions.

The acetamide 106 was synthesized because the a-haloacetami des are potential affinity labels (see Chapter II). Acetylation using acetic anhydride in aqueous sodium acetate required repeated treatments with acetic anhydride to achieve complete acetylation, probably due to the low solubility of the amine. Compound 106 was poorly active in competing for specific [ 3H]-AMPA binding in rat brain homogenates, hence none of the a-haloacetamides were actually synthesized (see discussion section 4.2).

The known 225 6 -nitroquinoxalinedione 107 was obtained almost quantitatively from the parent quinoxalinedione (Scheme XV) using the same nitration conditions as for the fluoro analog. This compound had been previously prepared in high yield using one 78

SCHEME XV: Synthesis of quinoxalinediones 107. 108, 109, 110, and 111.

H H H

. 0 80% HNQ 3°2N 'Y ^ 'T IV 0 SnC12 ^ ” 2N "|

H H CH3CONH N ^ O CH3CONH O

AC2O O 80% ^ ^ 3 O2N NaOAc 1 0 9 H lh/0»C ^ H

2 98% 8 6 %

H NaN02 b f 4- + n 2 . N ^ O 108 — —— ► HBF4 XXX i n H 55% equivalent of potassium nitrate in sulfuric acid .225 Reduction and acetylation were also accomplished as in Scheme XIV, but here the acetylation proceeded much more readily.

The nitration of 6 -ace tamidoquinoxalinedione 109 in 1:1 fuming:conc. HN0 3 was rapid. HPLC gave evidence of further nitration and N-acetyl cleavage with extended reaction times, hence the reaction had to be carefully monitored to ensure complete reaction of the starting material (which would otherwise be difficult to remove from the product) but avoid secondary reactions of the product.

Stable diazo salts of many aromatic amines have been obtained using appropriate counterions, namely BF4‘, PF6', SiF62‘ and SbF6' (for a review see ref. 226). These salts 79

SCHEME XVI: Thermal decomposition of compound 111.

H H BF4- +N2 N ^ O HO N ^ O

H H H (?) 111 101 112

smoothly decompose as solids to give aryl fluorides at temperatures ranging to well over 100°C. They also decompose in solution to give products dependent on the nucleophilic

species present and characteristic of a phenyl cation intermediate .227 Pure, relatively stable bisulfate salts have also been prepared by reaction of arylamines with nitrosylsulfuric acid in acetic acid and (when necessary) precipitation of the products

with ether.228 The diazo substituent was of interest because the most active AMPA antagonists identified to date are also acidic (see section 3.4), and N2+ is perhaps the strongest known electron-withdrawing group which can be attached to aromatic rings.

This group is also a potential affinity label. Compound 111 was prepared as a test compound to determine 1) the stability of the solid salt; 2) the ability of the N2+ group to lower the pKa of the parent compound; and 3) the biological activity, in particular the tolerance of a formal positive charge in this part of the molecule. It was crystallized as a hydrated 1:1 mixture with NaBF 4 (based on fluoride analysis). This compound was further characterized by decomposition in a melting point tube (Scheme XVI), which gave a mixture of 1 0 1 and an unidentified compound, possibly 6 -hydroxyquinoxaline

1 1 2 from reaction with hydration water (such reactions are well documented in the literature) .226 80 SCHEME XVH: Synthesis of 6-carboxyquinoxaline-2,3-diones.

H HOiC v u HO-jC m n 2 2 Et02CC02Et 2 ^ - T 5 — VNI0 «•» 72 h Reflux 92% H

113

H Fuming H 02C . N ^ ,0 HNO, - - - => 3 h/0°C 0 2N ' i ^ N ^ 0 2 83% H 114

The known 229 6 -carboxyquinoxalinedione 113 was prepared in high yield by reacting 3,4-diaminobenzoic acid with diethyl oxalate in refluxing THF (Scheme XVII). Since

further nitration of 6 -nitroquinoxalinedione 107 gives DNQX (52, i.e. the 6,7-dinitro

compound ) ,225 we anticipated 7-nitration of 113 to give 114. This compound was in fact SCHEME XVIH: Synthesis of 6-methylquinoxaline-2,3-diones. H H

H3C fiT ™ ’ ao^QOiB H3C o 80%mro, 3 ^ ^ N H , THF 3 h/0°C* °2 N 72 h Reflux „ 100% H 115 11*

H SnCl, H3C HCl/EtOHr q . + h 3n XXn10h*0 Reflux H 88% 117 81 obtained, as evidenced by the sharp 5 and 8 proton peaks in the *H-NMR which gave no evidence of the meta coupling observed for this class of compounds in all other cases when it occurred.

Several 6 -methylquinoxalinediones (115,230 116,231 117) were synthesized using the same methods as described above (Scheme XVHI).

SCHEME XIX: Proposed general synthesis of l-(2’-amino-3’-carboxyethyl)-2,3-quionoxalinediones. H H

R f Y \ ° DMF w R’^^N ^O 1N HC1 ^ R’^^N ^O R’^^N^O BrCH.CHfOEt).BrCH2CH(OEt) 2 Lk ^ O OEt E t A ^___ CHO H A T OEt 82

1) NaCN NH4C1

2) HC1 k ^ - C ° 2

A 83 NH^+

We anticipated that compounds of the general structure 83 might be obtained using methods (Scheme XIX) similar to those used in the synthesis of willardiine discussed in section 3.2. It was known that l,4-dihydroquinoxaline-2,3-dione

(2,3-dihydroxyquinoxaline, 127) and other compounds in this class could be N-dimethylated using dimethylsulfate in aqueous sodium or potassium hydroxide

(Scheme XX) ,225*233 but no information on reactions of quinoxalinedione anions in DMF was discovered. In a recent literature report 92 (not available at the time this work was 82

SCHEME XX: Methylation of quinoxaIine-2,3-diones.

H a i 3 Me2SQ4 ^ 2NNaOH ^ * * 0 127 H ^ CH3

OM H CH3 °*N Y \ N'f° M(*s°4, 0jN •lfV’V 0 0 2N 1NKOH o2n 52 H CH3 H P ^ CH» 3 2eqNaH , H3C y ^ r * * '118 f ° H,C CH3I HjC i U l NX 0 103 H DMF 48% CH, JJ "3C v ^ -^ ° w o, ",c 'Ip r \ cl NaocH3H’c ^ . n och , h3c -^J'N't0 H,c ‘'HjC - ^ n '^ oc Hj 103 H 119

H HjC T ^ r ^ - i NaHTOMP HjC ' r^ r N -'1 30% H A HJC y ^ N ^ O

H jC *"H jC *"h ,C V u CH,I 3 . u AcOH 3 O H CH3 . . . CH3 120 121 planned), the dianion of 6,7-dimethylquinoxalinedione 103 generated with sodium hydride in DMF reacted with methyl iodide to give the N,N-dimethyl compound 118 which was distinct from the 0,0-dimethyl compound 119 generated by a different path

(Scheme XX); however the yield was only moderate, so that other substitution products, though not mentioned in the report, might have been formed. The same authors also report the synthesis of the 1 -monomethylated compound 1 2 1 by oxidation of the corresponding 2-oxo-1,2-dihydroquinoxaline 120 (Scheme XX), a procedure which has long been documented in the literature .232,234 83

SCHEM£ XXI: Reaction of quinoxaline-2,3-dione dianion with bromoacetaidehyde diethyl acetal in DMF.

N ^ O

OEt _J 2.4 eq. NaH N . 0 N O N ^ 0 BrCH2CH(OEt)2N| ?

127 H

N O N .^ O

We attempted to react the dianion of quinoxaline-2,3-dione generated with NaH in DMF with one equivalent of bromoacetaidehyde diethyl acetal (Scheme XXI). The desired product 1 2 2 (confirmed by comparison to authentic product obtained by another route) was one of four major products formed in the presence of unreacted starting material; the other products, though not characterized, were likely various O- and N-substituted compounds, all having higher Rf values than 122 but similar fluorescence characteristics on TLC. The 6 -fluoro analog 101 did not react under similar conditions, evidentally due to the poor solubility of the anion. When DMSO was added as a co-solvent, modest amounts of at least four products were observed by reverse-phase HPLC. This situation is even more complex than that in Scheme XXI due to the asymmetry introduced by the fluorine atom. The 6-fluoro-7-nitro analog 104 gave a mixture of products which could 84 not be isolated.

Attempts were made to synthesize an appropriate precursor which could later be

converted to the 1,2 -diamine and cyclized to the quinoxaline dione. 4-Fluoro-2-nitroaniline (123) did not react with bromoacetaidehyde diethylacetal under

the conditions shown in Scheme XXII. The anion of this aniline generated with sodium

SCHEME XXH: Unsuccessful synthetic paths leading to l-(2’-amino-3’-carbo xyethyl)-6-fluoro-l,4-dihydroquinoxaline-2v3-dione.

BrCH2CH(OEt) 2 cat? ' m T ► NO REACTION ^ ^ n h 2 CH2C12 aq. NaHC 0 3 /Na2C0 3 123 24 h R.T. A 2 h(-C H 2Cl2> 72 h R.T. 24 h reflux

NaH/DMF 02 w^ N H BrCH2CH(OEt) 2 k j.O E t cat? 123 OEt A 124

NO NO- Fn/>N 0 2 NaH/DMF NCOCH, NCOCH- ^ N H C O C H , BrCH2CH(OEt) 2 k .O E t cat? OEt k CHO 125 A 126 hydride in DMF generated a mixture of products upon reaction with bromoacetaidehyde diethyl acetal. The major product was isolated by column chromatography as an oil, and exhibited an *H-NMR spectrum appropriate for 124, but was colorless (the starting 85

SCHEME XXHI: Synthesis of l-(2’-amino-3’-carboxyethyl)- l,4-dihydroquinoxa!ine-2,3-dione (127).

H

N ^ O BrCH2CH(OEt) 2 N ^ O 2MKOH > O c t H 30% EtOH cat. Nal 96% 127 A OEt 2-3 weeks 27% Conversion 122 130 H

1) NaCN NH,C1 ► 2) 7 N HC1 co2- A <20% NH3+

131 material is quite yellow) and was unchanged by hydrogenation on a Pd/C catalyst. The identity of this compound was not determined. The anion of N-acetyl-4-fluoro-2-nitroaniline (125) also gave a mixture of products. The major product, isolated by column chromatography and obtained as an oil, exhibited an

!H-NMR spectrum appropriate for 126 (but was not further characterized). An attempt to hydrolyze this acetal was unsuccessful, possibly due to concurrent amide hydrolysis and subsequent polymerization.

It was found that 2,3-dihydroxyquinoxaline (127) did react with bromoacetaidehyde diethylacetal (although very slowly) in refluxing 2 N sodium hydroxide. The yields were improved by the addition of 20-30% ethanol (Scheme XXHI), though it still required about two weeks to achieve 27% conversion to the desired product. linger reaction 86 times did not improve this conversion. The workup developed for this reaction was quite simple, using 2 :1 benzene:ethyl acetate to exclude almost all of the highly insoluble unreacted starting material in the extraction, followed by direct crystallization of the product from these extracts. With some effort, much of the unreacted 127 could be recovered.

It was not readily evident whether this reaction gave N- or O-substitution, and potential spectroscopic means of structure characterization (e.g. ^C-NMR ,92 UV,233a or IR spectroscopy233*5) required the synthesis of model compounds. We were also interested in developing a synthetic procedure to make 122 in a shorter time and in higher yields. It was reported that 2,3-trimethylsilyloxyquinoxaline (128) could be prepared by treatment with hexamethyldisilazane and N,N-dimethylated with methyl iodide by heating 10 hours in an autoclave at 125°C.235 It was later reported that 128. and the 6,7 dimethyl- and 6,7 dichloro- analogs, could be N-ribosylated with l-0-Acetyl-2,3,5-tri-0-benzoyl-p-D-ribofuranose in dry methylene chloride with boron trifluoride etherate at room temperature in 90 minutes .234 The solid bis(trimethylsilyl) derivative 128 was readily prepared in neat hexamethyldisilazane (Scheme XXIV), and could be handled in the open air; however this material was found to readily revert to 127 in polar solvents containing only traces of water. For example, heating briefly in dg-DMSO transferred from a sealed ampule directly into an oven-dried NMR tube caused extensive desilylation of this material to quinoxaline-2,3-dione, whereas the material was stable to heating in dry CDC13. No reaction occurred overnight at room temperature with bromoacetaidehyde diethyl acetal (neat, unpurified). Warming gave a flocculant white precipitate which was isolated and determined to be quinoxaline-2,3-dione (127). Some reaction did occur overnight at 120-130°C in neat bromoacetaidehyde diethyl acetal which had been washed with 10% NaOH, dried over 87

SCHEME XXIV. Attempted synthesis of 122 from 2,3-bis(trimethyIsiIyIoxy)quinoxaline.

H «CH3)3Si)2NH ^ . N OSi(CHs),

Q: n X 0 - n r * W o sOSi(CH i , 3) 3 H Reflux

127 128 H neat N ^ O H BrCH2CH(OEt) 2 ► ox 12?A 30°C k / ° E t 24 hr oxk 122 OEt + other products

MgS 0 4, and distilled under argon into an oven-dried vessel containing 4A molecular

sieves (Scheme XXIV). The reaction mixture became very dark, and there were several

other products formed under these conditions along with substantial amounts of unreacted starting material (128) or quinoxaline-2,3-dione, 127 (these could not be distinguished by TLC because the silylated material was cleaved on silica); hence, the products were not isolated. This reaction also did not provide unequivocable proof of the structure of 122. since under these harsh conditions quinoxaline-2,3-dione generated by

desilylation of the starting material might itself react to give O-substitution.

The pathway shown in Scheme XXV gave 122. which was shown to be identical to the material produced as in Scheme XXIII by acquiring the !H-NMR of an admixture. The unreacted diamine could be almost quantitatively removed from the mixture of oxalate salts after the Erst step by crystallization in diethyl ether. Since the diamine intermediate 129 appeared to proceed to 122 almost quantitatively, optimizing the Erst 88

SCHEME XXV: Synthetic confirmation of the structure of 122 by an unambiguous route.

OX- NH

NH 1) 0.5 eq. OEt u u BrCH2CH(OEt)2 NH OEt OEt C l ------129 122 ^ N H 2 1:1 H20/Dioxane OEt Reflux 2) (COOH)2 OEt Et^O OEt THF NH reflux 24 hr NH OEt OEt

step of this reaction should provide a reasonable synthesis of this material starting from inexpensive o-phenylenediamine.

The diethyl acetal 122 was readily hydrolyzed in high yield in dilute (Scheme XXHI), though yields were reduced if heating was continued for more than 10 minutes and heating for less time resulted in a product contaminated with starting material, which was deleterious in the next step. In DMSO solution the !H-NMR revealed an equilibrium between the free aldehyde (130) and the hydrated form, with the latter predominating. 89

SCHEME XXVI: Chemistry of the Strecker reaction on 130.

H H N ^ O 1) NaCN N O NH4CI Unident a x H,Q Intermed. ax k CN 130 CHO 133 i excess NH3+ CN- SLOW ■1 NHCl A t H H H N^.0 ax a n : N O H ax CN CO-, 132 OH 127 131 n h 3+ ~ quantitative quantitative with 2 eq. (4 M) CN" from 133

The Strecker reaction exhibited complex chemistry (Scheme XXVI), and the yields of amino acid were not high, ranging from 15 to 20%. The Strecker sythesis in general gives modest yields (for a review, see ref. 236), although a similar Strecker gave willardiine (69) in 50% yield (Scheme XU). The reaction was run under a variety of conditions, and extensive monitoring of the progress of the reaction by HPLC (Figure 34) and isolation of side products for identification has revealed some of the non-productive processes which occur. The aldehyde 130 is poorly soluble in water; hence, reaction of the solubilized product occurs with an excess of cyanide and ammonium present. When the reaction was run with two equivalents of sodium cyanide at 4 M concentration, the cyanohydrin 132 was produced almost quantitatively, while Uhidem. 90 Umdera. ImeimetS.

A: t=35 min. B: t=4.5 h 1 drop/4 ml 1 drop/4 ml 5 pi 0.16 AUFS 5 pi 0.16 AUFS

10 min Figure 34. Reverse-phase HPLC of reaction mixture from Strecker synthesis of 131. Conditions: 4% CH 3CN/0.1% CF3C0 2H/H20 , 2.0 ml/min. Column: S pm end-capped C- 8 , 4.6 mm ID x 15 cm L, Det: UV at 254 nm. one equivalent of CN' generally gave much smaller amounts of this material. The cyanohydrin 132 was identified by elemental analysis, FAB mass spectroscopy,

13C-NMR, and infrared spectroscopy. When the starting material was dissolved in

DMSO and added to a 6 M solution of NaCN and ammonium citrate (pH 9), the unknown intermediate (Scheme XXVI) was formed in large amounts, but it could not be further converted to a-aminonitrile 133 by heating or addition of more ammonium hydroxide. This intermediate has not yet been isolated and identified, but this will be necessary to provide a complete picture of this reaction. Some heating was necessary to obtain the a-amino nitrile, but heating significantly accelerated the formation of 91 quinoxaline-2,3-dione (127) from the unknown intermediate. The amino acid could not be synthesized without formation of 127 as a side product; fortunately, sufficient dilution of the reaction solution and extensive washing of the cation exchange column during isolation of the amino acid allowed 127 to pass through the column (otherwise it could not be removed from the product). The Strecker reaction proceeded comparably well using or ammonium citrate below pH 9. Additional ammonium hydroxide seemed to facilitate conversion of the unidentified intermediate to quinoxaline-2,3-dione, while the use of sodium rather than seemed to favor cyanohydrin formation.

SCHEME XXVH. Attempted synthesis of l-(2’^ ’-diethoxyethyl)-l,4-dihydro- 6-nitroquinoxaline-2,3-dione.

n M H B.CH2CH(OEt) 2 0 2N i n ° 2Nv^vN^° 2MNaOH XJX a t . \X > H 20% MeOH \ QEt

107 OEt o 2n

o x

134 (structure uncertain)

Structure-activity studies of the quinoxaline-2,3-diones (section 4.3) indicate that 6 - or 6,7-substitution with a variety of groups substantially enhances activity of these compounds, with nitro and cyano groups giving the highest activity. We therefore planned to synthesize nitro-substituted derivatives of 131. Since it was reported that

6 -nitroquinoxaline could be N,N-dimethylated with dimethyl sulfate in aqueous base ,237 92 this compound was allowed to react with bromoacetaidehyde diethyl acetal in refluxing 2 N NaOH (as was successfully carried out for 127 in Scheme XXIII). This did not give the desired product (Scheme XXVII), but gave another product tentatively identified as 134 (based on solubility behavior and ^-NMR) in very high yield. Since this compound is not reported in the literature, further characterization of this product is necessary.

SCHEME XXVm: Synthesis of NQXAA.

H H

r ^ V * ' f 0 0.2 M N02+BF4- ° 2N ^ ' f 0 W o < yC 02- 23°C k^C C V n h 3+ n h 3+ 131 135 structure unconfirmed

Considering the difficulties with the Strecker reaction of 130 (Scheme XXVI), it seemed more reasonable to obtain nitro substitution by nitrating 131; this work is still in progress and was hindered by the limited availability of starting material. The amino acid 131 was nitrated on a test scale (38 mg) with 2.4 equivalents of nitronium tetrafluoroborate (0.2 M) in sulfolane (Scheme XXVm). About 20 mg of product (probably 135) was isolated from this reaction by ion exchange chromatography and ciystallization. The

!H-NMR was appropriate for the proposed structure, but the material was somewhat impure and was not further characterized. On a somewhat larger scale, dinitration was attempted using fuming nitric acid in concentrated sulfuric acid, conditions under which willardiine was nitrated in high yield (see section 3.2). Initially there formed what appeared by HPLC to be one major product, presumably 135; further reaction led to two major products (Scheme XXIX), tentatively identified as 136 and 137. These were isolated as a solid mixture which contained about 2:1 136:137 (proposed structures) by :H-NMR. Since these compounds are well separated by HPLC, preparative reverse phase chromatography should provide both of these materials for biological testing in the near future.

SCHEME XXIX. Attempted synthesis of DNQXAA, H

Fuming HNO, conc. H.SO,

The products in this mixture have not yet been 3 separated and structures unequivocably confirmed 137 (see text for discussion). — - 1

3.4 pK, studies.

This project is fundamentally concerned with the effects of functional group acidity on the biological activity of excitatory amino acid analogs; hence, pKa determinations were an important aspect of this project. Since we needed to measure the pKa of many compounds, flexibility was an important consideration. Additionally, for some compounds only milligram quantities were available for study. Fortunately, most of the compounds were amenable to study using UV spectrophotometric methods. Since we 94

were primarily concerned with ionizations involving aromatic systems, this method had

the additional advantage of measuring the sought pKa values in the presence of

additional functional groups (e.g. amino acids) which might interfere with potentiometric determinations. Finally, rigorous mathematical treatment and interpretation of data from

the photometric method is more straightforward than for potentiometric methods .238

General procedures are available for spectrophotometric determination of pKa values .238 We modified these methods (section S. 1.2) in that, instead of preparing a standard buffer series spanning the appropriate range (ApH = pKa + 2) for each compound, material was

dissolved in 0.01 N sodium hydroxide/0.09 M KC1 and carefully titrated with 0,1 N hydrochloric acid. The ionic strength was thus maintained at 0.1 M, which is important because ionization constants are strongly dependent on ionic strength. This particular

ionic strength was chosen as being relevant to physiological systems. Less than a

milligram of compound was required. The mathematical method used to analyze the data is developed in Appendix 2, and a FORTRAN subroutine for the program

NONLIN84239 (Appendix 3) was written to fit the data to these equations. A typical

titration curve is shown in Figure 35).

The ionization constants for the phenolic group of the nitrotyrosines we have studied are given in Table 2. (section 4.1). Unexpectedly, the values measured for 73, 83, and 87

(Figure 36) were nearly a log unit lower than the literature values 240 for the corresponding o-cresols 139,138, and 140. The ionization constants for these o-cresols were measured under the same conditions, and were in excellent agreement with the literature values. Hence, the amino acid side chain must exert a substantial electron-withdrawing effect on the aromatic ring in these compounds, probably through the a bond network. 95 0.5S0 0.500 0.460 eE 0.400 a N ♦ 0.860 % 0 0.800 0 s 0.260 A

1 0.200

> 0.160 3

0.000 .0 0 2. DO 8.DO 4.1DO 5.1DO 6 . DO 7. DO 8. 00 0 . DO 10 ,00 11 .00 12. pH Figure 35. Titration curve for the pKa determination for 3’-nitro-L-tyrosine using a UV spectrophotometric method.

We also studied the ionization of a number of quinoxaline-2,3-diones (Table 4, section

4.2). The acidity of these compounds was essentially predicted by Hammett substituent constants (Figure 37 ) .241 Many of the points on the plot involve at least one nitro substituent, which strongly influences the slope of the plot. In 6,7-disubstituted compounds involving a nitro substituent, the nitro group may have to rotate out of the plane of the aromatic ring due to steric and electronic interactions, and this will affect the reliability of predictions based on substituent constants (however some combined substituent constants for disubstituted rings were available). The presence of two ionizable protons on the quinoxalinedione ring system sometimes rendered the choice of am and op for each substituent ambiguous; in Figure 37, nitro groups were always assigned to op. Also, op substituent constants for phenols and anilines are generally 96

NO NO

HO HO HO HO

NH NH NH OH 83 88 pK^e.6 pKa = 6.4 pKa = 3.4 pKa = 4.2

0,N o 2n NO, NO,

HO' HO HO CH3 CH3 CH3 138 139 140 pK, = 7.7 PK» = 7.2 pKa = 4.4

Figure 36. Comparison of ionization constants for phenolic groups in o-cresols and congeneric amino acids. different than for other compounds, but a set of these constants for quinoxalines does not exist. pKa (exper.) 10 6 7 8 9 Figure 37. Hammett plot for ionization of of ionization for plot Hammett 37. Figure 0.0 quinoxaiine-2,3-diones. 4 . 0 0.8 "*■**p 1.2 64 .6 1 = p 1.6 2.0 97 CHAPTER IV DISCUSSION OF BIOLOGICAL FINDINGS

4.1. Hydroxyphenylalanines and related compounds We initiated structure activity studies of tyrosine analogs based on several reports that

6 -hydroxydopa (141) could depolarize neurons through activation of a quisqualate-preferring receptor population in frog spinal neurons (see discussion Chapter n). 179,180 We postulated that the phenolate anion could serve as a bioisostere for the y-carboxylate of glutamate. Such an isostere might provide a more flexible template on

OH V

n h 3+ n h 3+ n h 3+ n h 3+

141 70 71 6 -OH-DOPA o-TYR w-TYR which to build potential A MPA antagonists. A comparison of the anionic structures for several compounds which bind to AMPA receptors is illustrated with AMPA (9), willardiine (69) and quisqualate (5) in Figure 38. Willardiine (WDN) contains a

6 -membered heterocyclic ring in place of the glutamate y-carboxyl, as does

6 -hydroxydopa. We began our test of this hypothesis using commercially-available DL-o-tyrosine (70), DL-m-tyrosine (71) and DL-p-tyrosine (72), which have phenolic groups corresponding to those at each of the positions of 6 -OH-dopa. None of these

98 99

OH OH ^N -O 0 ^ \ ^ ~ C H 3 ^ - c c -

n h 3+ n h 3+ n h 3+ 141 1 9 6-OH-DOPA GLU AMPA

O N CO,-

n h 3+ NH3+

WDN

Figure 38. Comparison of the anionic structures of some AMPA agonists. compounds showed significant competition for specific [3H]-AMPA binding in rat cortex homogenates (Table 2), perhaps due to the weak anionic character of the phenolic groups compared with the y-carboxyl of glutamate (pKa=4.3) or the heterocyclic rings of

AMPA (pKa=4,8) or QUIS (pKa=4.5). We postulated that an anion might be required for molecules to bind to AMPA receptors (but see below), and therefore synthesized 2’-hydroxy-3’-nitro-phenylalanine (83). 2’-hydroxy-5’-nitro-phenylalanine (73). and

2,-hydroxy-3\5’-dinitrophenylalanine (87) (section 3.1.1). We also synthesized the a-hydroxy acid 88 (section 3.1.1). Some of these compounds could inhibit specific [3H]-AMPA binding (Table 2), with dinitro-o-tyrosine (87) giving the highest activity.

For the series oTYR (70), oNoTYR (83), pNoTYR (73), and DNoTYR (87), the pKa values suggest a good correlation between extent of ionization at physiological pH and affinity for the AMPA binding site. Hence the dinitro compound DNoTYR, 87, which should be completely ionized in the binding assays, was the most active. The a-hydroxy 100

Table 2. Inhibition of [3H]-AMPA specific binding at 10-4 M R for various hydroxyphenylalanine analogs.

Rb X * ^ * 5 R2 X R$ V co2- [3H» COMPOUND r2 r 3 R4 R5 R« X Y pKa Inhib.

70 o-TYR OH HH H H C n h 3+ 0,0,4 71 m-TYR H OH HH H C n h 3+ 0 ,0 , 0 72 p-TYR H H OH H H C n h 3+ 0 ,0 ,0 141 6-OH-DOPA OH H OH OH H C n h 3+ -50 73 pNoTYR OH H H n o 2 H C n h 3+ 6.4 20,57 83 0 N0 TYR OH n o 2 HH H C n h 3+ 6.6 37,35 87 DNoTYR OH n o 2 H n o 2 H C n h 3+ 3.4 100 88 a-OH-DNoT OH n o 2 H n o 2 H C o h 4.2 0 142 3-NTYR H n o 2 OH H H C n h 3+ 6.8 0,30 143 DNTYR H n o 2 OH n o 2 H C n h 3+ 3.2 0 ,0 93 4-NmTYR H OH n o 2 H H C n h 3+ 0 90 4,6-DNmTYR H OH no 2 H n o 2 C n h 3+ 3.8 -50 92 2,6-DNmTYR n o 2 OH HH n o 2 C n h 3+ 3.6 -50 91 2,4-DNmTYR n o 2 OH n o 2 H H C n h 3+ >90 69 WDN =0 H =0 H HN n h 3+ 9.3 100 96 5 -NO2-WDN =0 H =0 n o 2 HN n h 3+ 6.4 100

*% Inhibition of specific 3H-AMPA binding by 100 pM test compound. Replicate results represent determinations in different tissue preparations. ND = not determined. n h 3+ n h 3+ n h 3+ o h 83 73 87 88 oNoTYR pNoTYR DNoTYR acid 88 was inactive, re-emphasizing the importance of the amino acid functionality.

The rationale for synthesizing the m-tyrosine analogs 90, 91, and 92 (Chapter II) was the substantial binding and functional activity of p-oxalylaminoalanine (BOAA, SO).128,187 4-bromo-homoibotenic acid (Br-HEBO, 76),185,186 and 4-methyl-homoibotenic acid (Me-HIBO, 77)185,186 at AMPA receptors. L-BOAA antagonizes specific 3H-AMPA binding with potency (IC^ = 1 pM) comparable to L-Glutamate, is fairly selective for AMPA vs. KAIN receptors in binding studies,187 and shows some selectivity for cortical AMPA receptors over spinal chord AMPA receptors in mice.128 The compounds 50, 76, and 77are apparently all agonists. Of the m-tyrosine analogs which we synthesized and tested, 91 exhibited substantial activity, but we do not yet know whether this compound is an agonist or antagonist.

We obtained 3 ’-nitro-L-tyrosine (142) and 3 \5 ’ -dinitro-L-ty rosine (143) from commercial sources (the racemic compounds were unfortunately not available). While 3’-nitrotyrosine gave some inhibition of [3H]-AMPA specific binding in one test, the dinitro compound gave no evidence of activity. If the enantioselectivity of this series

(see below) turns out to be the reverse of that for agonists (as it is for NMDA receptors, section 3.3.1), it will be desirable to synthesize the D-enantiomeis (or at least the 102

NO NO HO HO

NO- CO,-

n h 3+ NH3+ NH3+

90 91 92 4,6-DNmTYR 2,4-DNmTYR 2,6-DNmTYR

-O HO HO o — N H,C OX NH Vco2- n h 3+ NH3+ NH,+ 50 76 77 BOAA Br-HIBO Me-HIBO

racemates) of these compounds for repeat screening.

For those compounds showing good activity in the [3H]-AMPA screening studies, potencies as inhibitors of both [3H]-AMPA and [3H]-kainate specific binding were determined (Table 3). The ratio of the 50% inhibition values (IC50) for specific binding of the two radioligands provides an index of the relative selectivity of these compounds for AMPA vs. KAIN receptors. Glutamate shows very little apparent AMPA/KAIN selectivity, while QUIS and CNQX are somewhat more active as [3H]-AMPA inhibitors.

Willardine, 69, is surprisingly potent considering that the 1-substituted uracil ring has a pKa of about 9.3. We predicted that the 5-nitro group of 96 would, via increased acidity, confer as much as two orders of magnitude increase in potency vs. willardiine; at pH 7.4, willardiine (pKa= 9.3) is about 1% ionized, compared with 90% for 5-N02-WDN 103

Table 3. Inhibition of specific 3H-AMPA and 3H-KAIN binding by various compounds.

COMPOUND [3H]-AMPAIC5o(pM) [3H]-KAIN IC50 (pM)

1 l-GLU 0.83 1 9 AMPA 0.05 1 0 “ 5 QUIS 0.03“ 0.15“ 69 WDN 2 .0 90 96 5-N02-WDN 1.1 > 1 0 0 87 DNoTYR 3 > 1 0 0 91 2,4-DNmTYR 13 10 KAIN 2 0 “ 0.005“ 45 CNQX 0.32 1 .6

“Data from T. Honore et. al., Science 241:701 (1988)

(pKa=6.4). However, 5-N02-WDN (96) was only slightly more active than WDN (Table

3). One possible explanation for this result is that the acidity of willardiine is enhanced in the receptor site. This could be readily effected by the interaction of an electronegative hydrogen bond donor with the 4’-oxo group in willardiine, as shown in Figure 40, sketch B, where group Z-H could be a lysine, arginine, or histidine residue.

This scenario also offers an explanation for the increased activity of 6 -hydroxydopa

(141) compared to o-, m-, and p-tyrosine (70-72).

Willardiine and 5-nitrowillardiine were also found to be quite selective for AMPA vs. KAIN receptors based on the difference in radioligand displacement (Table 3). This is of interest due to reports indicating the opposite Junctional selectivity for

5-bromowilIardine ,37’243 i.e. KAIN>AMPA. DNoTYR (87) was not potent enough as an inhibitor of specific [ 3H]-KAIN binding to provide an IC 50 value, giving only 40% inhibition at 0.1 mM concentration, but there was at least a 20-fold difference from the

IC50 for [ 3H]-AMPA binding. The structures of WDN, 5-NOr WDN, 5-Br-WDN, and DNoTYR 5-N02-WDN WDN 5-Br-WDN pKa=3.4 pKa-6.4 pKa-9.3 pKa=? AMPA>KAIN AMPA»KAIN AMPA »K A IN KAIN>AMPA** "Functional data only. Figure 39: Comparison of dinitro-o-tyrosine with 5-substituted willardiines.

DNoTYR are compared in Figure 39.

DNoTYR (87) appears to be an AMPA antagonist (see below). Several hypotheses might explain this unexpected result. Recent pH-dependence and chemical modification (diethylpyrocarbonate) studies suggest that a histidine residue may be critical for binding

the y-carboxyl of glutamate to AMPA receptors.249 In enzymes, histidine imidazole

groups frequently appear to be involved in proton transfer. Proton transfer to or from the histidine imidazole might induce rapid conformational changes in the AMPA receptor protein, thereby modulating the conductance state of the ion channel. For agonists such as GLU, AMPA, QUIS, and WDN, there are two heteroatoms (0 ,0 in GLU, 0,N in the others) which could support hydrogen bonds and proton transfers; one possible configuration is depicted in Figure 40, where the imidazole is arbitrarily positioned to interact with the willardiine nitrogen in diagram B). In DNoTYR (87). there is a carbon in place of the second heteroatom which could not participate in hydrogen bonding or proton transfer. The nitro group would be incorrectly positioned to accept a proton, and 105

Figure 40. Proposed model for the AMPA receptor. P signifies attachment to the receptor protein. might also prevent conformational changes due to steric hindrance, since this space is not occupied in the aforementioned agonists.

The above hypotheses depend on the assumption that the nitrotyrosine analogs interact with AMPA receptors in the same manner as AMPA and WDN. A critical contribution to the binding would then be from the amino acid group (note that the a-hydroxy acid 88 is inactive. As discussed in section 3.1.3, this is one important reason for synthesizing the enantiomers of DNoTYR (87). If this compound shows the opposite enantioselcctivity from AMPA (for which L »d ) ,252 the compound may be binding with a different orientation or at a different (possibly overlapping) receptor. Certain invertebrate neurons are depolarized by 3’-carboxy-phenylalanine (145). apparently through activation of non-NMDA receptors.244 In the nitro-o-tyrosines 73, 83, and 87, a nitro group can always be in the position corresponding to the 3’-carboxy group in 145. 106

/ o O

145 83 73 3-C02-Phe oNoTYR pNoTYR

and it is possible that the nitro group could interact with the receptor in a similar manner, but without causing receptor activation.

Most of these compounds have yet to be studied in functional assays; however,

DNoTYR (87) injected into the rat ventral pallidum (VP) antagonizes the hypermotility response mediated by AMPA receptors in the nucleus accumbens and VP (Figure 41). The order of activity for AFQX (105. section 4.2), DNoTYR (87). and CNQX (45) correlated well with the relative ability of these drugs to inhibit specific [3H]-AMPA binding in rat brain homogenates. DNoTYR, unlike NMDA, was unable to stimulate or reduce acetylcholine release in rat striatal slices, indicating selectivity for non-NMDA receptors.

To summarize, of the dinitrotyrosines tested for inhibition of specific [3H]-AMPA binding in rat brain homogenates, the ortho hydroxy group confers the highest activity, as expected. One dinitro-m-tyrosine gave moderate activity, while dinitro-p-L-tyrosine, which is quite acidic and thus fully ionized at physiological pH, is inactive. Although DNoTYR (87) is two orders of magnitude less potent than AMPA or quisqualate in inhibiting specific [3H]-AMPA binding to rat brain homogenates, and is about an order of magnitude less potent than CNQX in blocking the AMPA-mediated hypermotility 107

L. 3 2 5 0 0 j -C Amph u V 1 m g /k g Q. w 2 0 0 0 -■ c3 O Amph Amph o 1 5 0 0 - ■ 1 m g /k g 1 m g /k g £ 1 0 0 0 ■■ % a: 5 0 0 - 2 uO 0 - - s Amph Amph Amph ONQX DNoT AFQX (5 nmole) (5 nmole) (5 nmole) 52 87 105

Figure 41. Effect of AMPA receptor antagonists injected into the ventral pallidum on hypermotility elicited by systemic injection of stimulants. Microinjections of 1 pg DNQX, 5 pg GAMS, 1 pg AFQX, 1 pg DNoTYR, or vehicle were made into the ventral pallidum of anesthetized rats. Immediately following this injection, the rats were given 0.5 mg/kg amphetamine. Ambulatory locomotor activity was monitored during a one hour period immediately following recovery from anesthesia. Control nypermotility (systemic stimulant; vehicle in the nucleus accumbens) was 2,000 to 3,000 locomotor counts with the doses of stimulants used, and baseline motility in the absence of any drugs was approximately 500 locomotor counts. Data are expressed as mean + s.e.m., and most groups contained 6-8 animals. 108 response in the rat ventral pallidum, we have shown that the phenolate anion can serve as a bioisostere for the Y-carboxylate of glutamate. The AMPA/KAIN selectivity also appears to be substantially higher than for the 2,3-quinoxalinediones (see section 4.2).

This molecule provides a novel, somewhat flexible template from which to design more potent non-NMDA antagonists.

4.2 Structure-activity studies of 2.3-quinoxalinediones The 2,3-quinoxalinediones are not amino acids, but are competetive antagonists of specific binding and functional responses to both AMPA and kainate (KAIN). We synthesized a number of compounds to examine how the nature of the 6 and 7 substituents affects antagonism of radioligand binding (Table 4) and functional activity. As with the compounds discussed in the previous section, there appears to be some relationship between binding activity and pKa, and in this set the most acidic compounds are most active; however, several compounds (e.g. 103, 113) which have pKa>9 are surprisingly active. In addition, the diazonium compound 111, which is predicted to have a pKa of about 6.3 (lower than both DNQX and CNQX) is essentially inactive.

Evidentally, the binding site will not tolerate a formal positive charge para to the anionic group. This is not particularly surprising since the most active compounds have a nitro group in this position, which should build up significant negative potential, whereas ionization of the quinoxaline ring in 111 produces an electroneutral compound. This may explain the potency of the 6-carboxy compound (113) even though the pKa is greater than 9. Surprisingly, 6-carboxy-7-nitro-QX (114) is almost inactive. The substantial activity of 6,7-dimethyl-QX (103) vs. almost complete lack of activity in QX (127) emphasizes the importance of the 6 and 7 substituents. 109

Table 4. Inhibition of [3H]-AMPA and [3H]-CNQX specific binding by various quinoxaline-2,3-diones.

H H N. NH,SO 1 t x Y e c u : H CO, H 131 NH, 51

ic50(pM) COMPOUND r2 pK. Ri [3H]-AMPA [3H]-CNQX

52 DNQX n o 2 n o 2 6.7 0.32 NT 45 CNQX CN n o 2 6.9 0.30 0.20 104 FNQX F n o 2 7.8 1.4 116 MeNQX c h 3 n o 2 4.3 107 NQX H n o 2 8.2 5.7b 114 COzNQX co2- n o 2 8.3 >100 100 AzFQX n 3 F 8.6 15 105 AFQX n h 2 F 9.2 100 127 QX H H 9.5 II 103 DMQX c h 3 c h 3 13 23 113 CO,QX co2- H 9.2 A 23 101 FQX F H 9.1 I I 106AcAFQX CH3CONH F 8.7 I I HOAcANQX c h 3c o n h n o 2 7.8 A 3.9 111 N2QX n 2+ H (6.2) I 131 QXAA (see structure) 0.69 26 51 NBQX (see structure) 0.15*

Various concentrations of compounds were tested for ability to compete for 3H-AMPA and 3H-CNQX binding sites in homogenates of rat brain. NT - not tested; I * <50% competition at 100 pM. A = 100% competition for specific binding at 100 pM concentration, but lower concentrations have not been tested. () pKa predicted from Hammett substituent constant •Data from Sheardown e t al., Science v. 247 (1990) 571. bData from Leeson e t al., J. Med. Chem.. 34 (1991) 1243. 110

100

o> | 80 FAQX + UV Light S < a. 1 60 I x n C c 4° o 3 *o £ 20 N

0

Figure 42.6-azido-7-fluoroQX as a photoaffinity label.

One of our compounds, AzFQX (100). was able to irreversibly inactivate AMPA binding sites in a photoaffinity experiment (Figure 42). Although this compound is not potent enough to be useful for receptor identification and purification, it does indicate possiblities for such experiments if the potency can be increased through structural modifications (the fluoro substituent is clearly not ideal). While we were planning to make a-halo acetamides as potential affinity labels, the inactivity of 106 initially discouraged us; however our subsequent synthesis and testing of 110 indicates that reasonable potency might be attainable.

QXAA (131). which is 1-subsdtuted with an amino acid side chain, is quite active (Table 4) in antagonizing both 3H-AMPA and 3H-CNQX specific binding, while QX (127) is essentially inactive at 100 pM. Interestingly, it has been recently reported that I l l

Table 5. Inhibition of [3H]-CNQX specific binding by riboflavin metabolites.

Competition for 3H-CNQX Binding Compound ICj,, (pM)

CH2-(CHOH),-CH2OH I H H:,c N 1 _ NH ^ O H,c Ribonavin >>1## o H H ' Y^V,N V N 'f . O n l < s M A m Lumichrome 17 H3C N " y f H O

DMQX 23 CH,CH3 ^ N| 0 H

6-CQX 19 ■ * o;i: H

N-monomethylation of 6,7-dimetliylquinoxaIine (103) results in a significant reduction of glycine activity, while non-NMDA activity is only slightly reduced.92 While we expected that compounds such as QXAA should be potent AMPA antagonists, 107 displayed potent convulsant activity upon direct intracerebroventricular injection into rats. Synthesis of additional analogs with 6 and 7 substituents will determine whether agonist activity is maintained, or whether these compounds will be converted to partial agonists or antagonists. 4.3 Potential link to flavin metabolism

The compound 6,7-dimethyIQX (103) is a known mammalian metabolite of riboflavin

(Figure 43).245 Compounds 146 and 147 have also been isolated from the urine of rabbits administered high doses of riboflavin.246 Lumichrome (148) and CO 2QX (113. a known247,248 compound which could be easily synthesized in one step) were found to have significant activity (Table 5). Since almost nothing seems to be known about the metabolism of flavins in the mammalian CNS, the significance of these observations is not evident but warrants further investigation. CH2-(CHOH)3-CH2OH CH2-(CHOH)3-CH2OH CH2-(CHOH)3-CH2OH H3^ .N. n1 n" ^-0 0 n H3C ,N ^ Oo H-.C — - ‘Cs H,C YXlt-V 'N'A)fN,H HjC YT" ' ^ J' n1'A Co !h H,C N X O o H Riboflavin

I H H .N .^ 0 N ^ O

HjC r rY NT NY ° — ► HlC Y Y " ' f ° h 3c n x h , c ■J^ A N' ' \ ' N h h , c /V ' h a y i 148 0 103 H Lumichrome 4 I H O H H02C

h o c " Y x Y 2 XIX h 3c h 3c H H H 147 146

Figure 43. Some pathways for riboflavin metabolism. Compounds not enclosed in brackets have been identified as peripheral riboflavin metabolites. None of these compounds (except riboflavin) have been reported to be present in the mammalian CNS. CHAPTER V EXPERIMENTAL

5.1 General procedures Melting points were determined on a Thomas-Hoover melting point apparatus; emergent stem corrections were not applied. Unless otherwise noted, !H-NMR spectra were acquired using an IBM AF/250 spectrometer (250 MHz). Some 13C-NMR and other special experiments were conducted with an IBM AF/270 spectrometer (270 MHz). Infrared spectra were obtained using a Laser Precision Analytical RFX-40 FTIR or a Beckman 4230 dispersive infrared spectrophotometer. High resolution electron impact (El) mass spectra were obtained at The Ohio State University Chemical Instrumentation Center using a Kratos MS-30 spectrometer, while fast atom bombardment (FAB) mass spectra were acquired using either a VG 70-250S or Finnigan MAT-90 spectrometer on "magic bullet" or 3-nitrobenzyl alcohol, using DMSO as an additional solvent when necessary. High performance liquid chromatography was performed on a Beckman 421 system equipped with Model 112 pumps and a Model 153 fixed wavelength 254 nm detector, or a Beckman System Gold equipped with a Model 167 scanning UV-visible detector, Model 406 analog interface, and Model 110B pumps. Unless otherwise noted, chromatograms were obtained on an IBM 5 pm EC-C8 column, 4.6 mm ID x 15 cm L, using a mobile phase flow rate of 2.0 ml/min, and detection was by ultraviolet absorption using a fixed wavelength 254 nm detector. Ultraviolet spectra were acquired with a Kontron UVEKON 860 or an IBM Model 9420 UV-visible spectrometer. Elemental analyses were performed by Galbraith Laboratories, Inc., Knoxville, TN or Oneida Research Services, Inc., Whitesboro, NY.

Instead of attempting to obtain anhydrous materials for all compounds, the general approach was to allow compounds to equilibrate with the atmosphere for a period of time (usually at least several weeks) before obtaining elemental analyses. In this way stable hydration states were usually obtained, which allowed for routine handling 114 115 without the necessity of excluding the sample from ambient moisture. In some instances elemental analyses were repeated after allowing time to elapse in order to confirm the stability of the hydrated crystalline state.

5.2 pKB determinations A spectrophotometric method was used to determine the proton dissociation constants given for the compounds in section 3.4. This method is an adaptation of published methods.238 Compound (1-2 mg) was weighed into a 25 ml volumetric flask and dissolved in 0.01 N NaOH + 0.09 M KC1 to make the ionic strength » 0.1. Sonication was used to aid dissolution if necessary. The pH of the resulting solutions was about 11.8. This stock solution was further diluted with 0.01 M NaOH + 0.09 M KC1 to give a solution with an optical density of about 1 AU, which typically required a two-fold to four-fold dilution. The absorbance spectrum of the compound was determined. For those compounds with 10

5.3 Synthesis of hydroxyphenylalanines and related compounds

0,N

NH^OC(CH3)3 NH^OC(CH3)3 co 2ho co 2ho

84 85

N-(t-butyIoxvcarbonyl)-2’-hydroxv-3,-nitro-p>L-phenylalanine (84) and N-(t-butvloxycarbonvl)-2’-hvdroxv-5,-nitro-Dj.-phenylalanine (85) (METHOD A): In a 50 ml round-bottomed flask, 5.0 g (27.6 mmol) of 2’-hydroxy-D,L-phenylalanine (D,L-o-tyrosine) were suspended in 20.7 ml of water, and 5.2 ml of 50% nitric acid were added over 30 min. at room temperature with stirring. The solution was cooled in an ice bath, and an additional 11.0 ml of 50% nitric acid were added dropwise over 2 hours. The solution was allowed to warm slowly to room temperature with stirring during 5 h, then placed in a refrigerator (-5°C) overnight. Water (50 ml) was added, and the mixture was extracted with benzene (2 X 100 ml). The aqueous fraction was neutralized with Na2C03 and again extracted twice with 100 ml portions of benzene. The aqueous solution was evaporated to dryness on a rotary evaporator, using isopropanol for azeotropic removal of the last portion of water. The solid residue was taken up in 100 ml of 1.2 N HC1 (a black solid remained undissolved) and eluted through a short 1" ID column containing 5 g of polystyrene/divinylbenzene beads (BIORAD BIOBEADS SM-16, 20-50 mesh) supported on a layer of glass wool and cotton and covered with a layer of sand. The solution was diluted 10-fold and eluted through cation exchange resin (BIORAD AG50W-X12, sulfonic acid type, hydrogen 117 form, 200-400 mesh), 2.5 cm ID X 9 cm L. The column was washed with water until the effluent was pale yellow (about 1500 ml), then eluted with 0.75 N ammonium hydroxide. The water and ammonia were evaporated on a rotary evaporator to dryness, again using isopropanol for azeotropic removal of the final portion of solvent, to give 2.57 g of solid which was estimated by *H-NMR to contain 83:73 ( ortho:para) in about a 3:7 ratio. A small amount (<10 mol%) of dinitro-o-tyrosine (87) was also formed.

A 1.5 g portion (ca. 6 .6 mmol) of the above mixture was dissolved in 100 ml of 1:1 dioxane-water, and 4 ml (30.3 mmol) of triethylamine and 3.59 g (14.6 mmol) 2 -ferf-butyloxycarbonyloxyimino- 2 -phenylacetonitrile ("BOC-ON " )195 were added. The solution was allowed to stir 16 h at room temperature, 200 ml of water were added, and the solution was extracted twice with 2 0 0 ml portions of ethyl acetate to remove unreacted BOC-ON and 2-hydroxyimino-2-phenylacetonitrile (the byproduct generated by the reaction with BOC-ON). The aqueous phase was acidified with 5% citric acid (75 ml) and 1.2 N HC1 (20 ml), and extracted twice with 150 ml portions of ethyl acetate. The extracts were dried over MgS04, the drying agent was removed by filtration, and the volume was reduced to about 20 ml on a rotary evaporator. This solution was divided into two equal portions, and each fraction was chromatographed on fresh silica (40-63 pm, E. Merck) dry-packed in a 51 mm ID X 43 cm L Michel-Miller column (Ace Glass, Inc.) and pre-equilibrated with chloroform. The column was eluted with 8 % isopropanol/2% methanol/0.7% acetic acid in chloroform, pumping at 20 ml/min. The separation was monitored by optical extinction at 280 nm with a flow cell (ISCO UV-5A), and by TLC (7% IpOH/7% MeOH/0.7% HOAc in CHC13). Under these conditions essentially baseline resolution was obtained, and appropriate fractions were combined. The resulting solutions were concentrated to about 2 0 % of the original volumes using a rotary evaporator, several volumes of were added (for azeotropic removal of the acetic acid), and the solutions were evaporated just to dryness (rotary evaporator). The residues were taken up in ethyl acetate and triturated with hexane to give crystals which were collected after one day at ambient temperature, washed with several portions of hexane, and dried in a stream of air. The products from both separations were combined to give 0.345 g of 84 (yellow crystals, 6 .6 % from o-tyrosine) and 0.883 g of 85 (pale yellow crystals, 16.8% from o-tyrosine). 118

Analytical data for 84: d.p. 139-140°C. !H NMR ((^-acetone, TMS): 6 8.04 (d, 1H, J*=8 .1 Hz, peak broadening due to meta coupling to 6 ■ 7.67 proton, ArH), 7.67 (d, 1H, Jofiho = 6 8 Hz, Jmela-lH z, ArH), 7.03 (asymmetric triplet, 1H, ArH), 6.18 (bd, 0.3 H (some solvent exchange), J=7.9 Hz, NH), 4.59 (bdd, 1H, J=4.7, 9.8 Hz, a-H), 3.44 (dd, 1H, Jyic=4.7 Hz, Jgcm=13.6 Hz, CH2), 3.03 (dd, 1H, ^ = 9 .8 Hz, Jgefn“ 13.6 Hz, CHJ, 1.31 (s, 9H,-C(CH3)3). IR (KBr, cm'1): 3339, 3092 (N-H overlapped with Ar-OH and COOH), 2980 (aliphatic CH), 1719 (sh, C 02H C=0), 1660 (carbamate C=0), 1608 (Ar C=C), 1542 (N=0 in NOj), 1453 (C-N), 1253 (COzH C-O), 1163 (phenol C-O). HPLC (5pm IBM RP-phenyl, 4.6mm x 15 cm L, 3 min. @ 10% MeOH/0.1% CF 3CO2H/0.2% CH3C 0 2H/H20 , linear gradient over 15 inin. to 21% MeOH, see also general procedures) showed about 2% contamination with 85, assuming equal area response factors, and no other significant peaks. Analysis for Ci 4Hi8 N20 7 : calculated C, 51.53; H, 5.56; N, 8.59; found C, 51.60; H, 5.63; N, 8.46.

Analytical data for 85: d.p. 205°C. JH-NMR (d^-acetone, TMS): 8 8.13 (d, 1H, Jffleta=2.7 Hz, ArH), 8.03 (dd, 1H, Jmcta=2.7 Hz, J ^ S . 9 Hz, ArH), 7.03 (d, 1H, Jonho=8-9 Hz, ArH), 6.26 (bd, 0.44H (some solvent exchange), J=7.2 Hz, NH), 4.54-4.62 (broad m, 1H, a-H), 3.41 (dd, 1H, Jvic=4.6 Hz, Jgem=13.6 Hz, CHJ, 3.00 (dd, 1H, Jvfc-9.8 Hz, Jgem=T3.6 Hz, CHz), 1.31 (s, 9H, -C(CH3)3). IR (KBr, cm"1): 3368 (bd, OH from C 02H and ArOH, and NH), 2986 (aliphatic C-H), 1730 (sh, C 02H O O ), 1684 (carbamate C=0), 1594 (Ar C=C), 1524 (N=0 in NOz), 1341 (N-O in NOz), 1287 (C02 C-O), 1165 (phenol C-O). HPLC (same conditions as for 84) indicated less than 1% contamination of the product with 84, again assuming equal area response factors. Analysis for CI 4H 18 N20 7: calculated C, 51.53; H, 5.56; N, 8.59; found C, 51.34; H, 5.56; N, 8.45.

N-(t-butyIoxycarbonyl)-2,-hydroxy-3,-nitro-D,L-phenylalanine (84) and N-(t-butyIoxycarbonyl)-2’-hydroxy-5’-nili o-D,L-phenylalanine (85) (METHOD B): DL-o-Tyrosine (2.94 g, 16.2 mmol) was suspended in 50 ml of reagent grade acetonitrile in a 100 ml round-bottomed flask, and the suspension was blanketed with dry argon gas to exclude moisture. The suspension was cooled to -18°C in a dry ice/sat. NaCl bath, and 85% nitronium tetrafiuoroborate (2.53 g, 16.2 mmol) w as added over 30 min. in small portions with stirring. The addition was done in such a way that about 1/3 of the material was added over the the first 15 min. and the rem aining 2/3 119 over the remaining time, with increasing rapidity (during this time the solid N0 2+BF4" was maintained in a test tube with periodic additions of dry argon gas. Stirring was continued for 30 min. longer with continued cooling, and the reaction was quenched by pouring the mixture into 470 ml of water and adding 28 ml of 1.2 N HC1. After standing overnight the solution was filtered to remove a brown precipitate and eluted through a cation exchange column (2.5 cm ID x 8 cm L, DOWEX 50X8-400, 200-400 dry mesh, sulfonic acid type, hydrogen form, prewashed with 200 ml of 6 N HC1 and 100 ml of water). The column was washed with 1500 ml of water and eluted with 1000 ml of 0.3 N ammonium hydroxide, collecting the product fraction from the point when the eluent became deep yellow. The water and ammonia were evaporated on a rotary evaporator to dryness. The residue was suspended in 10 ml of water and the grayish solid was collected, washed with two 10 ml portions of isopropanol and dried in a stream of air to a constant weight of 2.26 g. ^-N M R (DC1/DMSO) showed the mixture to be 33:22:45 83:73:DL-o-tyrosine, corresponding to 22%, 14%, and 30% yields.

A portion of this mixture (2.03 g, ca. 9.0 mmol) was dissolved in 130 ml of 1:1 H20:dioxane and 4.8 ml (36 mmol) of triethylamine. "BOC-ON" (see method A, 4.43 g, 18.0 mmol) was added and the reaction was allowed to stir overnight. Water (300 ml) was added and the mixture was extracted with two 300 ml portions of ethyl acetate to remove the 2-hydroxyimino-2-phenylacetonitrile by-product. The yellow aqueous layer was acidified with 100 ml of 5% citric acid and 30 ml of 1 N HC1. The products were extracted with two 2 0 0 ml portions of ethyl acetate, and the combined extracts were shaken with 100 ml of saturated NaCl and dried over MgS04. The extracts were filtered and concentrated with a rotary evaporator to about 5 ml (probably mostly dioxane) and diluted to 40 ml with chloroform. This solution was separated in a single portion as in Method A, and the two products were crystallized as before to give 0.73 g of 84 (15% from o-tyrosine) and 0.58 g of 85 (12% from o-tyrosine). 2,-hydroxv-5,-nitro-DX-phenylalanine (73): Compound 85 (0.50 g, 1.5 mmol) was suspended in 14 ml of dichloromethane in a 25 ml erlenmeyer flask, and treated with 6 ml of trifluoroacetic acid while stirring. The starting material dissolved, and stirring was continued for 3 h at room temperature. The white precipitate which formed during this time was collected, washed with dichloromethane, and dried in a stream of air. Yield: 0.42 g (80% as the trifluoroacetate salt), d.p. 208-209°C. Reverse phase HPLC (see general procedures for column, detection, eluting 8 min. @ 8 % CH3CN/0.1% CF3CO2H/H2O, then linear gradient to 70% CH3CN over 10 min.) showed the material was free of significant quantitities of 83 (less than 0.5% was difficult to quantitate because 83 eluted in the peak tail of 73), and no significant amounts of any other compounds were detected. Since the elemental analysis was in slight disagreement with the expected results, a quantity of the material was further purified by dissolving 230 mg in 90 ml of 0.1 N HC1 and eluting this solution through a 2.5 cm ID X 3.5 cm L column of 200-400 mesh AG50X12 sulfonic acid-type cation exchange resin (BIORAD, hydrogen form). The column was washed with 600 ml of deionized water, and the product was eluted with 550 ml 0.75 N ammonium hydroxide. The water and ammonia were removed on a rotary evaporator to a volume of about 15 ml. Isopropanol (50 ml) was added and the volume was again reduced to about 20 ml. The product was collected by filtration, washing with isopropanol and diethyl ether. Recovery from the column was about 94%. d.p.- 246-256°C; lit .193 252-257°C dec. JH-NMR (DC1/D20 , DSS): 5 8.15 (s, 1H, ArH, overlapped with 6=8.13), 8.13 (dd overlapped with 6=8.15 proton, 1H, Jmeta=2.8 Hz, ArH), 7.04 (dd, 1H, J 0rtho=:2.8 Hz, .^ = 1 .6 Hz, ArH), 4.48 (dd, 1H, J=5.8, 7.3 Hz, a-H), 3.44 (dd, 1H, ^ = 5 .8 Hz, Jgem= 1 4 -5Hz, CHj), 3.27(dd,lH, ^ = 7 .3 Hz, Jgem=14.5 Hz, CHj). IR (KBr): 3400 (bd, OH), 3185 (bd, OH), 2988 (bd, NH3+), 1582 (C02' asymm. stretch, N=0 in NO 2), 1480 (N-H sym. bend), 1326 (N-0 in NO 2), 1287 (C02‘ sym. stretch). MS (FAB): m/e 121

263 (M+l). Analysis for C 9 H!0N2O5: calculated C, 47.79; H, 4.46; N, 12.39; found C, 47.61; H, 4.65; N, 12.25.

HO

C02-

83

2’-hydroxy-3>nitro-D.L-phenylalanine (83): Compound 84 (200 mg, 0.61 mmol) was suspended in 7 ml of dichloromethane, and treated with 3 ml of trifluoroacetic acid with stirring. Stirring was continued for 7 hours at room temperature, and the reaction was neutralized using about 30 ml of 10% triethylamine in methylene chloride. The precipitated product was collected, washed with a small portion of methylene chloride and dried in air to give 116 mg (83%) of yellow powder in two crops, d.p. 236-237°C. ‘H-NMR (DCI/D20 , DSS): 6 8.07 (dd, 1H, Jmcta- 1.3 Hz, Jojtho-8.5 Hz, ArH), 7.67 (dd, 1H, Jmeta=1.3 Hz, JOItho=7-4 Hz, ArH), 7.10 (asymm. triplet, 1H, ArH), 4.24 (asymm. triplet, 1H, a-H), 3.37 (dd, 1H, ^ = 6 .9 Hz, Jgcm=14.0 Hz, CH^ 3.24 (dd, 1H, ^ = 6 .9 Hz, Jgcm=14.0 Hz, CH 2). IR (KBr, cm 1): 3420 (bd, OH), 3211 (bd, OH), 3070 (bd, NH3+), 1663 (C02H C=0 (?)), 1608 (bd, N0 2 N=0, C02' asym. stretch, Arom. C=C), 1544 (N-H bend), 1399 (C02* sym. stretch), 1357 (N0 2 N-O), 1147 (C-NH3+ stretch). MS (FAB): m/e 227 (M+l). Reverse phase HPLC (see general procedures for column, detection, eluting 8 min. @ 8 % CH3CN/0.1% CF3C02H/H20 , then linear gradient to 70% CH3CN over 10 min.) showed the 1st crop contained less than 0.5% of 73. assuming equal area response factors. Analysis for C 9 H10N2O5: calculated C, 47.79; H, 4.46; N, 12.39; found C, 47.45; H, 4.63; N, 12.14. 122

2-hvdroxv-3-(2’-hvdroxv-3,.5’-dinitrophenyl)propionic acid ( 88 ); A 25 ml round-bottomed flask was charged with 6 ml of water and 5 ml of conc. nitric acid with cooling in an ice bath and stirring, and 1.00 g (5.52 mmol) of DL-o-tyrosine were added in small portions during 10 min. Stirring was continued for 2.5 h, and then 11 ml of fuming nitric acid were added over 15 min. with stirring and continued cooling. After an additional hour, the ice bath was removed and the solution was allowed to slowly come to room temperature. After a total of 11 h had elapsed, the reaction was diluted to 400 ml with water and extracted with two 400 ml portions of 10% CH3CN in ethyl acetate. The extracts were shaken with 200 ml of saturated sodium chloride solution, dried over MgS04, filtered, and evaporated to a volume of about 20 ml on a rotary evaporator. Hexane was added slowly, with warming, until cloudiness barely persisted, and the solution was allowed to stand for 22 h. Some additional hexane was added, and crystallization was allowed to continue for one more day at ambient temperature. The product was collected and washed with a small portion of 1:5 ethyl acetate:hexane to yield 0.48 g (32%) of pale yellow crystals, m.p. 140-141°C after recrystallization from ethyl acetate/hexane. *H-NMR (d*-acetone, TMS): 6 8.85 (d, 1H, J=2.8 Hz, ArH), 8.53 (d, 1H, J=2.8 Hz, ArH), 4.57 (dd, 1H, J-4.2, 8.9 Hz, a-H), 3.46 (dd, 1H, 1 ^ 4 .2 Hz, Jgem=14.2 Hz, CH 2), 3.11 (dd, 1H, ^ - 8 . 9 Hz, Jgem=14.2 Hz, CH^. 13C-NMR (d^-acetone): 175 ppm (s, C^O), 158.1 (s, =Cr -OH), 140.1 (s, =C 5’-NC>2), 134.2 (s, =C3’-N0 2 ), 133.0 (d, =C6’-H), 131.9 (s, C1’), 120.7 (d, =C*'-H), 69.7 (d, Ca H), 35.2 (t, C3H2). IR (KBr, cm 1): 3480 (bd, COO-H), 3200 (bd, O-H), 1750 (C=0 in COzH), 1615 (N=0 in NOz), {1559, 1530, 1470,1432} (arom. C=C), 1348 (N-0 in NO 2), 1267 (C-O in C02H), 1091 (phenolic C-O). MS (El): m/e 272.025 (calc. 272.xxx, M+), 254 (-H20), 227 (-C02H), 198 (-H0 2CCHO), 76 (base). Analysis for G>H 8 N20 8 V2H20: calculated C, 38.49; H, 3.22; N, 9.97; found C, 38.69; H, 3.00; N, 9.85. 123

NO

HO n h 3 5/2h 2o

87

Z’-Hydroxv-S’^ ’-dinitro-DX-phenylalanine. ammonium salt, s/7 hydrate (87): A suspension of 0.50 g (2.76 mmol) of D,L-o-tyrosine in 8 ml of reagent grade acetonitrile was cooled with an ice bath to 0-5°C, and 1.0 g (6.4 mmol) of 85% nitronium tetrafluoroborate was added in small portions over 10 min. with continued cooling and stirring. Cooling and stirring was continued for 8 h, and the reaction was quenched by the addition of 120 ml of water. After standing overnight, the dark brown solid formed was removed by filtration, and the filtrate was eluted through a 2.5 cm ID by 6 cm L cation exchange column (BIORAD AG50W-X12, 200-400 mesh, sulfonic acid type, hydrogen form), washing with 2500 ml of water. The product was eluted with 0.4 N ammonium hydroxide, collecting the product fraction from the point where the eluent became deep orange and continuing the elution until the eluent was nearly colorless. The volume of the solution was reduced to about 10 ml using a rotary evaporator, then isopropanol (1 0 0 ml) was added and the suspension was heated briefly on a steam bath. After cooling, diethyl ether (500 ml) was slowly added and the flask was allowed to stand in a refrigerator overnight. The product was recovered by filtration, washed with a small amount of isopropanol followed by diethyl ether, and dried in air to give 0.625 g ( 6 8 %) of fluffy, yellow-orange powder. Melting point: heating at about 5°C/min., the material visibly changed form ca. 160°C, discolored > 180°C, rapid dec. > 190°C. ‘H-NMR (D20 , DSS): 6 8.78 (d, 1H, J=3.1 Hz, ArH), 8.05 (d, 1H, J= 3.1 Hz, ArH), 4.02 (dd, 1H, J= 3.6,7.7 Hz, Ho), 3.27 (dd, 1H, 3.6 Hz, Jgem- 14.6 Hz, CHJ, 3.00 (dd, 1H, I*,.- 7.7 Hz, Jgem» 14.6 Hz, CH2). In dg-DMSO there was a broad peak centered - 7.5 ppm (7H, NH3+ and NH^). IR (KBr, cm'1): 3410 (bd, H 20 , N H /), 3161 (bd, NH3+), 1608 (C02' asym. stretch), 1560 (N=0 in NOj), 1476 (NH bend), 1402 (C-O sym. stretch in C 02 ), 1334 (NO in NOj), 1253 (phenolic C-O). MS (FAB): m/e 272 (M+l), 226 (-NO^. Analysis for NH 3 -5/2H20 : calculated C, 32.46; H, 5.13; N, 16.82; found C, 32.74; H, 5.17; N, 16.47. The product from a 124 separate reaction gave the following analysis: C, 32.52; H, 5.01; N, 16.56.

N 02 n o 2 HO n h 3 n h 3 n ° 2 7/2H20 r NH3+ n h 3+ co2- co2- co2-

90 92 91

5 ’-hydroxv-2’ .4 ’-dmitro-DX-phenylalanine. ammonium salt (90) and 3>-hydroxy-2,,6,-dinitro-ptL-phenvlalanine, ammonium salt, 7/2 hydrate (92): Commercially-obtained 3-hydroxy-D,L-phenylalanine (m-tyrosine, 1.00 g, 5.52 mmol) was suspended in 15 ml of reagent grade acetonitrile, blanketed with argon, and cooled in an ice bath. This suspension was treated with solid 85% nitronium tetrofluoroborate (2.07 g, 13.2 mmol) in small portions over 10 min. with constant stirring and cooling in ice. The resulting deep red solution was stirred 2 h more in the ice bath, and then allowed to come slowly to room temperature for another 5.5 hours with stirring. The reaction was placed in a freezer at -15°C overnight. The following day, the solution was allowed to return to room temperature with stirring over two hours, then diluted with water to 500 ml. The solution was passed through a column of Dowex 50X8-200 (100-200 dry mesh, sulfonic acid-type, hydrogen form, prewashed with 200 ml 6 N HC1 and 100 ml water), 2.5 cm ID X 7 cm L. The column was rinsed with 500 ml of water, then eluted with 1000 ml of 0.3 N ammonium hydroxide, collecting the product fraction from the point where the eluent became basic and yellow-orange in color. The water and ammonia were removed on a rotary evaporator (bath temperature 55°C) to a volume of about 20 ml, whereupon 90 crystallized. This suspension was allowed to stand in a refrigerator for 2 weeks (a day is probably sufficient based on additional work). The solid (nearly pure compound 90) was removed by filtration and the mother liquor was set aside. The product was washed with two 1-2 ml portions of water (which were not added to the mother liquor) and dried in a stream of air for 10 min. The product was further dried by allowing it to stand in air overnight, then placing it in a dessicator for 3 days to give 428 mg (26.9%) of 90 as the anhydrous ammonium salt. 125

The product was recrystallized from a minimal volume of hot water. Melting behavior: heating at about 5°C/min., the yellow-orange salt slowly discolored at temperatures above 200°C, and rapidly turned dark brown at ~ 245°C. !H-NMR (DC1/D 20 , DSS): 6 9.02 (s, 1H, Ar-H), 7.33 (s, 1H, Ar-H), 4.52 (asymm. triplet, 1H, a-CH), 3.77 (dd, 1H, J=7.6, 13.8 Hz, CH 2), 3.58 (dd, 1H, J=7.5, 13.8 Hz, CHz). IR (KBr): 3170 (bd, N H / and NH3+), 1618 (C-O asym. stretch in C 02 ), 1560 (N=0 in NOz), 1511 (N-H bend), 1460, 1443 (arom. C=C), 1400 (C-O sym. stretch in C0 2 ), 1306 (N-O), 1250 (phenolic C-O). Analysis for C 9 H9 N3O7 NH3: calculated C, 37.51; H, 4.20; N 19.44; found C, 37.37; H, 4.20; N 19.06. The amounts of the other two in this material were quantitated by HPLC (see general procedures), using authentic standards due to the differences in response factors; after recrystallization there was found to be < 0.5% of 91 and < 0.5% of 92 (mobile phase 8 % CH3CN/0.1% CF3C 0 2H/H20).

The mother liquors from above, containing 91 and 92 and some residual 90, were slowly triturated with isopropanol to a volume of about 400 ml, and diethyl ether (150 ml) was then added. The suspension was evaporated (rotary evaporator) to about 50 ml, then 150 ml of isopropanol and 150 ml ether were added, and the crystallization was allowed to proceed overnight at room temperature. The solid was collected by filtration, washed with 30 ml of ether, and air dried to yield 633 mg of material which was estimated to contain 2:1 91:92 (molar basis) by ^-NM R. This solid was separated by preparative reversed phase chromatography as follows: ten 60 mg portions were each dissolved in 1 ml of 1.2 N HC1 and injected onto a 47 mm ID X 38 cm L Michel-Miller column (Ace Glass, Inc.) packed with 40-63 pM silica (E. Merck) which had been derivatized with octadodecyl-trichlorosilane (Cl 8 ). The compounds were eluted with 15% acetonitrile/1% trifluoroacetic acid in water at 20 ml/min, collecting 20 ml fractions and monitoring the separation by UV detection at 280-310 nm. The fractions were analyzed by injecting 20 pi on an HPLC system (see general procedures), eluting with 4% acetonitrile/0.1% trifluoroacetic acid in water at 2.0 ml/min, with UV detection at 254 nm (fixed wavelength, gain = 0.02 or 0.04 AUFS). Some fractions were combined and recycled (repeating the ion exchange chromatography to again isolate solid); in general it was necessary to recycle only one or two fractions containing both 91 and 92 per run, along with several fractions in the tail of the 91 peak which contained residual 90. Hence an eleventh separation on 50 mg of recovered material was performed in addition to the ten primary separations. Those fractions containing mixtures were discarded in this final separation. The 126 fractions containing 92 were combined, and HC1 was added to make the solution 0.1 N. This solution was eluted through a 2.5 cm ID X 7 cm L Dowex AG50X8-200 cation exchange column as above, again washing with water and eluting with 0.3 N ammonium hydroxide. The water and ammonia were evaporated on a rotary evaporator to a damp cake (bath temp. 50°C), 10 ml of water were added, and the solution was triturated with isopropyl alcohol to the point of initial precipitation. Crystallization was allowed to continue in the freezer for one week, and the product was collected, washed with isopropanol and dried in a stream of air, to give 130 mg (6.7%) of 92, d.p. 170-180°C. !H-NMR (D 20 , DSS): 6 8.07 (d, 1H, J=9.6 Hz, Ar-H), 6.58 (d, 1H, J=9.6 Hz, Ar-H), 3.92 (dd, 1H, J=6.9, 8.7 Hz, a-CH), 3.50 (dd, 1H, J=6.9, 14.3 Hz, P-CH^, 3.04 (dd, 1H, J=8.7,14.3, p-CH*). IR (KBr): 3404 (bd, H 20, NH4+), 3127 (bd, NH3+), 1590 (C02- asym. stretch, N=0 in NOj), 1555 (N=0 in NOj), 1524 (N-H bend), 1402 (C-O sym. stretch in C02'), 1300 (phenolic C-O stretch, N-0 in NO^. MS (FAB): m/e 272 (M+l). Analysis for G jH ^O ? NH 3 7/2H20: calculated C, 30.77; H, 5.45; N, 15.95; found C, 31.12; H, 5.51; N, 15.57. The amount of the other two isomers was quantitated vs. authentic standards using HPLC, giving ~ 0.35% of 91 and -0.04% of 90.

The fractions containing mostly 91 were combined, and HC1 was added to make the solution 0.1 N. Cation exchange isolation was performed as for 92. The water and ammonia were removed on the rotary evaporator to a final volume of about 5 ml, and then isopropanol was added just to the point of cloudiness. Crystallization was allowed to proceed slowly in the freezer. The product (91) was recovered by filtration, washed with isopropanol, and dried in a stream of air to give 306 mg (approx. 19%) which was contaminated with 92 (3% on a molar basis) and 90 (1% on a molar basis). This material was not purified further because an alternate synthesis was in progress (see below).

The total isolated yield for the three compounds was 53%. 127

NO

O 149

Diethyl-l-acetamido-Z-O’-methoxy^’-nitrobenzvD-l.l-ethanedicarboxylate (149): Sodium metal (1.77 g, 77.0 mmol), purified by melting in boiling toluene, was dissolved in 375 ml absolute ethanol in a flame-dried 500 ml R.B. flask (in two trials, one using Mg°-dried ethanol 256 and one using absolute ethanol directly from the bottle, the yields were the same). After the sodium had completely dissolved, acetamidomalonic acid diethyl ester (15.2 g, 70.0 mmol) was added and the solution was heated to 60°C with an oil bath, protected from moisture under dry argon. This solution was treated with 17.22 g (70.0 mmol) 4-bromomethyl-2-methoxy- 1-nitrobenzene 257 with strong stirring. The reaction was rapid, and after 4-5 minutes stiong precipitation occurred and the suspension quickly became an unstirrable cake, whereupon heating was discontinued. After another 15 min,, the reaction flask was placed in a refrigerator. After 3 days, 50 ml of ethanol were added and the cake was broken up and recovered by filtration. The cake was resuspended in 400 ml of ethanol and stirred vigorously for one hour. The solid was again recovered by filtration, washed with another 50 ml of ethanol in two portions, and dried in a stream of air for one hour to give 149, 18.98 g (70.9%), m.p. 165.0- 165.5°C (lit257 163-164°C.) 'H-NMR (CDC13, TMS) 6 7.76 (d, 1H, 1 ^ 0 = 8 .3 Hz, Ar-H), 6.73 (d, 1H, Jmeta=1.3 Hz, Ar-H), 6 .6 6 (dd, 1H, ^ 0=8 .3 Hz, JmeU=1.3 Hz, Ar-H), 6.59 (s, 1H, NH), 4.20-4.36 (m, 4H, CH2), 3.90 (s, 3H, ArOCH3), 3.73 (s, 2H, AiCHj), 2.04 (s, 3H, COCH3), 1.31 (t, 6 H, J=7.1 Hz, CH3). MS (El): m/e 382.136 (calc. 382.138, M+), 323 (-CH 3CONH2), 276, 43 (base, CH3CO+). Analysis for Ci7H22N20 8: calculated C, 53.40; H, 5.80; N, 7.33; found C, 53.24; H, 5.73; N, 7.24. Semiquantitative TLC (7% EtOH/CHCl 3 on silica GF254) using bracketed standards and short-wave UV detection indicated that approximately 1 g of product remained in the combined mother liquors, along with some unreacted benzyl bromide and 128

acetamidomalonic acid diethyl ester (detected by strong I 2 complexation).

NO HO

93

3,-hvdroxY-4,-nitro-p.L-phenylalanine, (93): A suspension of 9.56 g (25.0 mmol) of 149 in 95 ml of 48% HBr was brought to gentle reflux in a 250 ml R.B. flask. After 20 hours heating was discontinued, and after several more hours the HBr salt had crystallized from solution. The suspension was placed in the freezer for 3 days, and the salt was recovered by filtration on a glass frit with no washing to give a damp cake. This cake was dissolved in 50 ml of water and titrated to pH 5 with conc. ammonium hydroxide while stirring and cooling in an ice bath. The endpoint of the titration is evident by flashes of orange upon further NH 4OH addition. After overnight refrigeration, another 50 ml of water were added, and the product was recovered by filtration, washing with 40 ml of ice-cold water and 20 ml of room-temperature water, and drying by air suction and then 18 h on a vacuum pump to give 4.31 g (76.3%) of light yellow crystals, d.p. 229°C (at 2°C/min from 226°Q. !H-NMR (DC1/D 20 , DSS): 6 8.13 (d, 1H, Jortho=8-7 Hz, ArH), 7.15 (d, 1H, Jmcta = l/7 Hz, ArH), 7.14 (dd, 1H, Jonho=8.7 Hz, Jmeta=1.7 Hz, ArH), 4.48 (dd, 1H, J=7.5, 6.1 Hz, CHo), 3.47 (dd, 1H, Jgem=14.5 Hz, JviC=6.1 Hz, CH2), 3.26 (dd, 1H, Jgem=14.5 Hz, ^ = 7 .5 Hz, CHj). MS (FAB): m/e 227 (M+l), 181 (-NO^. Analysis for C 9 H 10N2O5: calculated C, 47.79%; H, 4.46%; N, 12.39%; found C, 47.12%; H, 4.38%; N, 12.23%.

In a separate trial (1/2.5 scale), the hydrolysis took only 5 h (monitored by reversed-phase HPLC), and the HBr salt was isolated by washing the isolated filter cake (see above) three times with isopropanol, air drying, and placing the product on a vacuum pump at about 3 torr for 3.5 h. The *H-NMR spectrum (D 20) was essentially 129

identical to that of the free zwitterion in DC1/D 20 . ER (KBr, cm*1): 3001 (bd, NH3+), 1738 (C=0), 1626 (N-H bend), 1590 (N=0 in NOj), 1532 (arom. C=C), 1486 (N-H bend), 1443 (arom. C=C), 1325 (N-O in NO 2), 1261 (PhOH C-O), 1202 (C02H C-O). Analysis for C 9 Hn BrN20 5: calculated C, 35.20%; H, 3.61%; N, 9.12%; found C, 35.25%; H, 3.56%; N, 9.04%.

NO- NO HO HO NH NH->+

C 02- 90 91

S’-hydroxy-l’^ ’-dinitro-DL-phenylalanine, ammonium salt (90) and 3>-hydroxy-2>.4*-dinitro-PL-phenylalanine, ammonium salt, dihydrate (91): Compound 93 (2.26 g, 10 mmol) was suspended in 10 ml of reagent grade acetonitrile, blanketed with dry argon gas to protect the reaction from moisture, and cooled to 0-4°C in an ice bath. Nitronium tetrafluoroborate (85%, 3.91 g, 25 mmol) was added during 5 min. with continued cooling of the solution. After 2.7 h, the contents were poured into 1000 ml of 0.1 N HC1. This solution was eluted through a cation exchange column (DOWEX AG5-X8, RSO 3H type, 100-200 mesh, 2.5 cm DD x 7.5 cm L, prewashed with 200 ml 6 N HC1 and 200 ml of water). An unidentified side product, which eluted much earlier than the products using reverse phase HPLC (see general procedures, 4% CH3CN/0.1% CF3C 0 2H/H20 , 2.0 ml/min.), was not reatined by the ion exchange column. The column was washed with 600 ml of water and eluted with 0.3 N ammonium hydroxide (1 0 0 0 ml total); collection of the product fraction began when the eluent became deep orange and basic. The water and ammonia were evaporated with a rotary evaporator (bath temp. = 55°C) to a damp residue. Water (10 ml) was added and the suspension was placed in a refrigerator for 5 days. The solid was removed by filtration and the mother liquor was set aside. The solid was washed with two additional portions of water, dried in a stream of air for several hours, then allowed to stand in air and continue to dry to a constant weight over 4 days to give 0.75 g (26%) 130 of 90, which was further purified by crystallization from hot water (see above). The mother liquor was triturated with isopropanol just to the point of persistent cloudiness, and placed in a freezer (-15°C). After three weeks, the product (deep orange crystals) was collected, washed with 8 ml of isopropanol in 2 portions and dried 45 min. in a stream of air to give 0.27 g of ammonium salt dihydrate, dec. slow > 170°C, rapid > 185°C. This crop contained 0.9% 90 quantitated by HPLC vs. an authentic standard (see general procedures, 8 % CH3CN/0.1% CF3C 0 2H/H20 at 2.0 ml/min.). A second crop was obtained by concentrating the mother liquor on a rotary evaporator to about 1 ml, again triturating with isopropyl alchol and allowing several days for crystallization (42.1 mg), total 11%. ^-N M R (D 20 , DSS): 6 7.96 (d, 1H, J=8.9 Hz, ArH), 6.38 (d, 1H, J=8.9 Hz, ArH), 3.98 (dd, 1H, J=5.2, 9.0 Hz, a-H), 3.20 (dd, 1H, JviC=5.2 Hz, Jgem«14.8 Hz, CH 2), 2.92 (dd, 1H, JviC=9 0 Hz, Jgem = 1 4 -8 Hz, CH^. The spectrum in d^-DMSO gave a broad peak centered near 7.5 ppm (NH4+ + RNH3+). IR (KBr, cm '1): 3424 (bd, NH4+), 3114 (bd, NH3+), 1606 (C02' asym. stretch, N=0 in NO 2), 1560 (N=0 in NO2), 1523 (N-H bend), 1489 (N-H bend or arom. C=C), 1338 (N-0 in NO 2), 1257 (phenolic C-O). MS (FAB): m/e 272 (M+l). Analysis for C^H 9 N30 7 NH3 •2H20 : calculated C, 33.34; H, 4.97; N, 17.28; found C, 33.76; H, 4.69; N, 17.33.

HO

NH

150

(S)-a-(N-(Phenylsulfonvl)amino)-P-hydroxy-2-methoxypropiophenone (151): A 500 ml boiling flask was charged with N-Plienylsulfonyl-L-serine (4.91 g, 20.0 mmol) and heated 1 h in an oven at 130°C, then placed under vacuum at 2 torr while cooling for an additional hour. Dry argon was admitted to the evacuated flask, and 320 ml of sodium-distilled THF were added. A 1000 ml boiling flask was charged with 160 ml of additional dry THF and 23.14 g (120.0 mmol) of 2-bromoanisole and cooled to -78°C under dry argon. n-Butyllithium (2.5 M in hexane, 53 ml = 133 mmol) was added 131 slowly over 10 min. This solution was stirred for one hour at -78°C, and the suspension of N-phenylsulfonyl-L-serine was added by cannulation over 50 min. After an additional 30 min., the dry ice bath was removed and the solution was allowed to warm slowly over 1.2 h, after which time all of the starting material had reacted (TLC: 7% EtOH/CHCl3). The solution was slowly poured into 500 ml of ice-cold 1.2 N HC1 while stirring in an ice bath, and this mixture was extracted with 1400 ml of diethyl ether in three portions. The ether extracts were washed twice with 300 ml portions of of 1/2-satd. NaHC03, once with 250 ml satd. NaCl, and dried over MgS04. After filtering, the volume was reduced on a rotary evaporator to about 50 ml, and trituration with hexane gave crystals. After several days in a freezer, the product was collected, washed with hexane, and dried in air to give 2.89 g (43.1%), m.p. 107.2-109.6°C. ‘H-NMR (CDC13, TMS): 6 7.89-7.84 (m, 2H, ArH), 7.57 (dd, 1H, J=7.8, 1.8 Hz, ArH), 7.54-7.40 (m, 4H, ArH), 7.01-6.90 (m, 2H, ArH), 6.20 (bd, 1H, J=7.0 Hz, NH), 5.11 (ddd, 1H, J=7.0,4.2, 3.5 Hz, CH), 3.96 (ddd, 1H, J-11.6, 6.9, 3.4 Hz, CH2), 3.72 (ddd, 1H, J=11.6,6 .8 ,4.2 Hz, CH^, 2.36 (t, 2H, J=6.8 Hz, OH). MS (El): m/e 336.091 (calc. 336.083, M+), 200 (PhS0 2NH=CHCH2OH+), 135 (base, Ph(OCH3)CO+). Analysis for C16H17N05S: calculated C, 57.30; H, 5.11; N, 4.18; found C, 57.40; H, 4.96; N, 4.13.

5.4 Synthesis of 2,3-quinoxaIinediones and related compounds

(C 0 2H)2

l,2-diamino-4-fluoro-benzene, oxalate salt (102): A 1000 ml round-bottomed flask was charged with 5.0 g (32 mmol) of 4-fluoro-2-nitroaniline, which was dissolved in 300 ml of absolute ethanol. Stannous chloride dihydrate (36.0 g, 160 mmol) was dissolved in 36 mol of conc. HC1 and added to the above solution. The mixture was brought to reflux for 1 h and allowed to cool. Water (550 ml) was added and the ethanol was removed by distillation on a rotary evaporator. The aqueous solution was adjusted to pH 8 with conc. ammonium hydroxide with strong stirring in an ice bath 132 over 15 min. (neutralization with 50% NaOH gives comparable results). The resulting thick white slurry was extracted with diethyl ether (350 + 350 + 250 ml), and the extracts were dried over MgS04. The drying agent was removed by filtration, and a solution of 8 .1 g (64 mmol) of oxalic acid in 120 ml of isopropanol was added, slowly at first to allow good nucleation. After two days at room temperature, the product was collected, washed with ether, dried in air to give 5.70 g (82.3%) of silver-gray needles. Recrystallization from methanol gives silvery plates, dec. 160-170°C to a white solid, m.p.>275°C. !H-NMR (CD3OD/d6-DMSO, TMS): 5 6.91 (dd, 1H, Jortho = 8-6 Hz, JHF.meta= 5.7 Hz), 6.52 (dd, 1H, Jmeta=2.8 Hz, Jhf, oitho = 10-6 Hz), 6.34 (sextet, 1H, Jmeta= 2-8 Hz> J«iho= 8-6 Hz, JHF,ortho=8.5 Hz). Analysis for Q H yF IV C ^C V calculated C, 44.45; H, 4.20; N, 12.96; found C, 44.62, H, 4.25; N, 13.30.

H

" Y ® N O I H 101

6-fluoro-l,4-dihydroquinoxaline-2,3-dione. (101): Compound 102 (2.16 g, 10.0 mmol) was partitioned into 400 ml of diethyl ether from 250 ml of 10% NaOH, and the sodium hydroxide layer was further extracted with 350 ml of diethyl ether. The solution was dried over MgS0 4 and the solvent was distilled off on a rotary evaporator. The white diamine was taken up in 20 ml of THF and transferred to a 50 ml round-bottomed flask. Diethyl oxalate (10 ml, 74 mmol) and 2 drops of acetic acid were added, and the flask was purged with argon and brought to reflux under a positive pressure of argon. After several hours, solid product began to precipitate. Reflux was continued a total of 64 h, then the condenser was removed and the THF was allowed to distill off over 6 h. The reaction was allowed to cool and 20 ml of ether were added. The product was recovered by filtration, resuspended and stirred vigorously in 100 ml of ether, and again recovered by filtration, washing with 100 ml of diethyl ether in several portions. The product was allowed to stand in air for two days to give 1.71 g (95.0%) of an off-white powder. Recrystallization from ethanol gave white crystals, 133

ra.p. > 350°C. !H-NMR (dg-DMSO, TMS): 6 = 11.97 (bs, 1H, NH), 11.93 (bs, 1H, NH), 7.12 (dd, 1H, Jortho=8-6 Hz, Jh f^ u -5 .3 Hz, H8), 6.97-6.87 (m, 2H, H5, H7). IR (KBr, c m 1): {3170, 3060} (bd, NH), {2952, 2840} (Ar C-H), 1695 (s, C=0), 1632 (sh, NH bend), {1523, 1508} (Arom. C=C), 1387 (s, C-N stretch), 1258 (C-F). MS (FAB): m/e 181 (M+l). Analysis for C 8 H5FN20 2: calculated C, 53.34; H, 2.80; N, 15.55; found C, 53.51, H, 2.82; N, 15.27.

6-fluoro-1.4-dihydro-7-nitro-quinoxaline-23-dione (104): A 50 ml round-bottomed flask was charged with 15 ml of 70% HN0 3 which was cooled in an ice bath, and 15 ml of fuming nitric acid were cautiously added with stirring. After 5 min. of continued stirring, 2.0 g (11.1 mmol) of 101 were added in small portions over 22 min. Stirring was continued 10 min. more in the ice bath and the reaction was allowed to warm to room temperature over 1 h with continued stirring. The mixture was then poured onto 2 0 g of ice, a further 10 g of ice were added, and the mixture was placed in a freezer overnight. An additional 70 g of ice were added, 25 ml of 50% NaOH was slowly added over 1 h with stirring in an ice bath, and the mixture was returned to the freezer overnight. The product was collected by filtration, washed with small portions of water, isopropanol, and diethyl ether, and air-dried to give 104 (2.27 g, 90.8%). Recrystallization from ethanol gave pale yellow crystals, dec. to solid > 250°C. XH-NMR (d^-DMSO, TMS): 6 12.38 (bs, 1H, NH), 12.11 (bs, 1H, NH), 7.86 (d, 1H,

W ^ r 7 1 Hz), 7.07 (d, 1H, Jh f,orth o-12.0 Hz). IR (KBr, cm'1): 3100 (bd, N-H), {2940, 2813} (Ar C-H), 1693 (C=0), 1541 (N=0 in NOz), 1400 (C-N), 1326 (N-O in N 02), 1310 (sh, C-F). MS (FAB): 226 (M+l). Analysis for C g H ^ C ^ : calculated: C, 42.68; H, 1.79; N, 18.66; found C, 42.64; H, 1.81; N, 18.55. 134

H i N

N i H

105

6-amino-7-fluoro-l,4-dihydroquinoxaline-2,3-dione (105): Tin (II) chloride dihydrate (1.13 g, 5 mmol) was dissolved in 1.13 ml conc. HC1. This solution was added to a suspension of compound 104 (225 mg, 1 mmol) in 25 ml of ethanol, and the mixture was brought to reflux for 1 h, giving a clear greenish solution. The reaction was allowed to cool, 30 ml of water were added, and the ethanol was distilled off using a rotary evaporator. The remaining mixture was diluted to 100 ml to give a solution which was eluted through a cation exchange column (BIORAD AG-50X12, 200-400 mesh, sulfonic acid type, protonated form, 2.5 cm ID x 5 cm L). The column was washed with 500 ml of water to remove the tin salts, and then eluted with 700 ml 0.9 M ammonium hydroxide. The water and ammonia were evaporated on a rotary evaporator to a small volume, when crystals formed. These were collected by filtration, washed with isopropanol and then diethyl ether, and dried in a stream of air to give 165 mg (85%) of a light yellow powder, m.p. > 350°C. NMR (DMSO, TMS): 6 11.73 (bs, 1H, NH), 11.60 (bs, 1H, NH), 6.78 (d, 1H, Vonho-H -5 Hz), 6.55 (d, 1H, Hz), 5.12 (bs, 2H, NHJ. IR (KBr, cm 1): {3438, 3318} (amine N-H), {3160, 3067} (bd, N-H), {2960,2837} (Ar C-H), 1663 (C-O), 1510 (C=C), 1459 (N-H sciss.), 1380 (C-N), 1303 (C-NH2 str.), 1238 (C-F). MS (FAB): 196 (M+l). Analysis for Q H e l * ^ : calculated C, 49.24; H, 3.10; N, 21.53; found C, 48.73; H, 3.26; N, 21.41. 135

H i

H

152

6-amino-7-fluoro-l<4-dihvdroquinoxaline-2t3-dione. HC1 salt monohvdrate (152): Tin (II) chloride dihydrate (7.02 g, 31.1 mmol) was dissolved in 7 ml of conc. HC1, and this solution was added to a suspension of 104 (1.40 g, 6.22 mmol) in 55 ml of ethanol. The mixture was brought to reflux, giving a clear greenish solution. After cooling, the suspension was placed in the freezer for 3 days. The product was collected by filtration, washed with 25 ml of ethanol and two small portions of ether, and dried in air to give 1.22 g (79%) of a yellow powder, d.p. 150°C to another solid, m.p. >270°C. ^-NM R (de-DMSO, TMS): 6 12.02 (bs, 2H, NH), 7.14 (d, 1H, JnF.meur™ Hz), 7.03 (d, 1H, JHF,oitho=l 1-0 H2)* - 6.5 (very broad singlet, 5H, NH3+ and HzO). Analysis for C8 H6FN30 2 HC1 H20: calculated C, 38.49; H, 3.63; N, 16.83; found C, 38.51; H, 3.29; N, 16.78.

CHiCONH

6-acetamido-7-fluoro-l,4-dihvdroquinoxaline-2.3-dione. (106): Compound 152 (0.50 g, 2.00 mmol) was suspended in 5 ml of water in a 25 ml round-bottomed flask, and 0.23 ml (0.24 g, 2.4 mmol) of acetic anhydride was added in one portion with stirring. After 5 min., 1.1 ml of a sodium acetate solution (2.65 g/15 ml) were added. After 45 min. another 0.23 ml of acetic anhydride were added, followed 10 min. later 136 by another 1.1 ml portion of sodium acetate solution. Similar additions were done at 2.5, 3, and 3.5 h after the start of the reaction. The course of the reaction was monitored by reversed phase HPLC (see general procedures). After another 30 min (4 h from the start), 6 ml of 1 N HC1 were added and stirring was continued for two more hours. The suspension was placed in the refrigerator overnight, and the solid was collected by filtration, washing with water. The sticky cake, which smelled of acetic acid, was boiled in 1900 ml of ethanol for several hours (the product did not completely dissolve), and after cooling the crystals were collected, washed with a portion of ethanol, and dried in a stream of air to give 406 mg (85%) of a light yellow powder, m.p. >270°C (slight discoloring was evident at 240°C). *H NMR (dg-DMSO): 8 11.87 (bs, 2H, ring NH), 9.71 (bs, 1H, amide NH), 7.74 (d, 1H, Jhf, meta” 7.8 Hz, ArH), 6.92 (d, 1H, Jhf, ortho = 11.8 Hz, ArH), 2.07 (s, 3H, CH3). IR (KBr, c m 1): 3329 (amide NH), {3195, 3078} (ring NH), 1692 (overlapped C-O’s), 1623 (sh, N-H bend), 1552 (C=Q, 1397 (sh, C-N), 1381 (C-N), 1292 (Ar-N), 1259 (C-F). MS (FAB): 238 (M+l). Analysis for CioH 8 FN3 0 3: calculated C, 50.64; H, 3.40; N, 17.72; found C, 50.55; H, 3.57; N, 17.43.

6-azido-7-fluoro-1.4-dihydroquinoxaline-2.3-dione. (100); Compound 152 (0.5 g, 2.00 mmol) was suspended in 1:1 conc. HCl:methanol (50 ml), and the suspension was chilled in an ice/salt bath to -12°C. NaN0 2 (0.48 g, 6.5 mmol) was dissolved in 10 ml of H20 and added over 15 min. with stirring, and the resulting lime-green suspension was stirred 2 h more at £ 10°C. NaN3 (0.21 g, 3.2 mmol) was dissolved in 20 ml of water and chilled to -10°C in the ice/salt bath. The suspension of diazo salt was slowly added to the sodium azide solution with stirring over 50 min. Stirring was continued in the ice/salt bath for 90 min., during which time the fume hood was darkened and the 137 reaction was covered with foil. The product was collected, washed with small portions of ice water, isopropanol, and diethyl ether, and dried in a stream of air while shielded from light to give 0.31 g (70%) of light gray crystals, d.p. 130-140°C to another solid. The material was homogenous by reversed-phase HPLC (see general procedures). !H-NMR (dfi-DMSO): 11.97 (bs, 1H, NH); 11.87 (bs, 1H, NH), 6.97 (d, 1H, Jnp <** 0=11-8 Hz), 6.91 (d, 1H, JHF,meta=7.8 Hz). IR (KBr, cm*1): {3144, 3062} (bd, N-H), {3062, 2953, 2827} (Ar C-H), 2120 (N3), 1695 (C-O), 1510 (C-C), 1378 (C-N), 1353 (C-N3), 1307 (C-F). MS (FAB): m/e 222 (M+l), 194 (-N2). Analysis for C 8 H4FN50 2: calculated C, 43.45; H, 1.82; N, 31.67; found C, 43.47; H, 1.96; N, 31.64.

6-carboxv-l,4-dihvdroquinoxaline-2,3-dione (113): A 250 ml boiling flask was charged with 2.74 g (18.0 mmol) of 3,4-diaminobenzoic acid, 36 ml of reagent grade THF, 19.4 g (133 mmol) of diethyl oxalate, and 5 drops of glacial acetic acid. The mixture was brought to a gentle reflux with stirring. After 48 hours the reaction was allowed to cool, and 130 ml of diethyl ether were added. The suspension was stirred vigorously for several minutes and the product was collected by filtration, washing with three 40 ml portions of diethyl ether (each time resuspending the product) to remove the excess diethyl oxalate. The product was dried in a stream of air for 15 min. to give 3.60 g of solid. This solid was redissolved in 500 ml of 2.5% aqueous sodium hydroxide with warming on a steam bath. The solution was filtered and allowed to cool to room temperature, then chilled in an ice bath. The solution was slowly neutralized with 12 N HC1 (about 40 ml), then another 50 ml of 12 N HC1 were added. After an hour of stirring with continued cooling, the product was collected by filtration, washing with 0.1 N HC1 (50 ml), water (20 ml), ethanol (2 x 20 ml), and diethyl ether (2 x 20 ml). The material was dried in a stream of air, then allowed to stand for one week exposed to the open air to give 3.70 g (91.7%) as the monohydrate, m.p. >270°C. 138

JH-NMR (ds-DMSO): 8 12.15 (s, 1H, NH), 12.03 (s, 1H, NH), 7.71 (d, 1H, Jmeta= 1.7 Hz, ArH), 7.42 (dd, 1H, J ^ - 8.4 Hz, Jmeta = 1.7 Hz, ArH), 7.16 (d, 1H, 1 ^ = 8.4 Hz, ArH). Analysis for C^HgNjCV^O: calculated C, 48.22; H, 3.60; N, 12.50; found C, 47.89; H, 3.51; N, 12.59.

H

I H

114

6-carboxv-l,4-dihvdro-7-nitro-quinoxaline-2^-dione, (114): Fuming nitric acid (9 ml) in a 25 ml round-bottomed flask was chilled in an ice bath, and 6 -carboxyquinoxalinedione 113 was added in one portion with stirring. The reaction was continually maintained at ice bath temperature and monitored by reversed-phase HPLC (see general procedures). After 2 h, the reaction sluny (probably containing solid anhydrous product) was slowly added to 63 g of ice. At first the solid redissolved, then crystals began to form. The mixture was placed in the freezer overnight, then the solid was collected by filtering, washing with 5 ml of icewater. The solid was twice resuspended in 10 ml, then 5 ml of ice water, dried in a stream of air, and allowed to stand for two weeks exposed to air to yield 0.90 g (83% as the monohydrate). A portion of this material was recrystallized from hot 5:1 watenmethanol, then from dilute ammonium hydroxide, acidifying with HC1 to give light yellow needles, m.p. >290°C. *H-NMR (de-DMSO): 8 12.30 (bs, 1H), 12.25 (bs, 1H), 7.61 (s, 1H), 7.51 (s, 1H). IR (KBr, cm*1): 3441 (bd, H 20), 3234 (bd, N-H), 1707 (bd, all O O ), 1559 (N=0 in NO2), 1386 (C-N), 1352 (N-O), 1293 (C-O). MS (FAB): m/e 252 (M+l). Analysis for C 9 H5N30 6 H20: calculated C, 40.16; H, 2.62; N, 15.61; found C, 40.05; H, 2.45; N, 15.42. l,4-dihydro-6-nitro-quinoxaline-213-dione, (107): A 500 ml erlenmeyer flask was charged with 100 ml of conc. nitric acid (69-71%) which was then chilled in an ice bath, and 100 ml of fuming nitric acid were slowly added with stirring. After several minutes of continued cooling, 9.72 g (60.0 mmol) of 2,3-dihydroxyquinoxaline was added in portions over 5 min. with vigorous stirring. During this addition the temperature never rose above 5°C due to the slow dissolution of the starting material. After 2 hours, the suspension was poured over 600 g of ice, allowed to stand 30 min., and placed in a refrigerator overnight. The product was collected, washed with three 30 ml portions of water (resuspending the solid each time), dried 1 h in a stream of air, and allowed to equilibrate for several days with the open atmosphere to give 12.83 g (95.0%) of pale yellow crystals, m.p. >270°C (lit .258 ). *H-NMR (dg-DMSO): 6 12.35 (bs, 1H, NH), 12.15 (bs, 1H, NH), 7.94 (dd, 1H (half under 6-7.92 proton), Jmeta=2.0 Hz, ArH7), 7.92 (s, 1H, should be doublet with J-2.0 Hz, but masked due to the other half of the 6 7.94 signal, ArH5), 7.23 (d, 1H, Jortho-7.8 Hz, ArH 8 ). Analysis for C 8 H5N30 4H 20: calculated C, 42.68; H,3.13; N, 18.66; found C, 42.80; H, 3.13; N, 18.83. 140

h >n XX„I01h H2o H

108

6-amino-I,4-dihydroquinoxaline-2.3-dione, (108): Compound 107 (2.07 g, 9.19 mmol) was suspended in 100 ml of anhydrous ethanol, and 11.3 g (SO.O mmol) of stannous chloride dihydrate dissolved in 11.5 ml conc. HC1 were added. The mixture was brought to reflux, and after 5 hours another 2.3 g (10.2 mmol) of stannous chloride dihydrate dissolved in 2.3 ml conc. HC1 were added. After a further 1.5 h of reflux (total 6.5 h), the suspension was allowed to cool and placed in a freezer overnight. The product was collected, washed with 50 ml of ethanol in two portions and dried in a stream of air to give 1.87 g of solid HC1 salt. This solid was dissolved in 900 ml of hot 1:1 2 M ammonium hydroxide:ethanol, filtered, and the ethanol, ammonia, and water were evaporated on a rotary evaporator to a volume of about 100 ml. The solid was collected, washed with two 2 0 ml portions of water and 2 0 ml of isopropanol, dried in a stream of air and allowed to stand 3 days to give 1.36 g (80%) of a greenish-yellow powder, m.p. >270°C. . !H-NMR (dg-DMSO): 6 11.67 (bs, 1H, N-H), 11.56 (bs, 1H, N-H), 6.80 (d, 1H, ^ = 9 . 1 Hz, H 8 ), 6.3 Hz (overlapping s+d, 2H, H5 and H7), 5.07 (s, 2H, NH2). Analysis for C gfyN aC V ^O : calculated C, 51.61; H, 4.33; N, 22.57; found C, 51.27; H, 3.95; N, 22.51. 141

CHiCONH

6-acetamido-1.4-dihydroquinoxaline-2.3-dione. (109); Compound 108 (0.708 g, 4.00 mmol) was suspended in 10 ml of water with stirring in a 50 ml round-bottomed flask and warmed to S0°C. Acetic anhydride (0.8 ml, 8.3 mmol) was added with stirring. Almost immediately the suspension became thick and milky-white. Another 10 ml of water were added and the mixture was swirled vigorously. Twenty-five minutes after adding the acetic anhydride, 3.6 ml of sodium acetate solution (18 g/100 ml) were added with vigorous mixing. Reversed-phase HPLC showed the reaction to be nearly complete. Fourty minutes after the first addition of acetic anhydride, a second 0 .8 ml portion of acetic anhydride was added while heating on a steam bath with vigorous swirling. After another 10 min., a second 3.6 ml portion of sodium acetate solution was added. After 30 min. further, the suspension was diluted with water to a volume of 200 ml and allowed to stand at room temperature for 2.5 h. The solid product was collected, washed with 50 ml of water, 50 ml of isopropanol, and 50 ml of ether, and dried in a stream of air to give 0.86 g (99%) of off-white powder, m.p. >270°C. !H-NMR (DMSO): 6 11.88 (bs, 1H, quinoxaline NH), 11.82 (bs, 1H, quinoxaline NH), 9.98 (bs, 1H, amide NH), 7.60 (d, 1H, Jmeta=2.0 Hz, H5), 7.17 (dd, 1H, Jortho-9.8 Hz, Jmela-2.0 Hz, H7), 7.02 (d, 1H, ^ - 9 . 8 Hz, H 8 ), 2.01 ppm (s, 3H, CH3). Analysis for C 10H9 N3O3 H2O: calculated C, 50.63; H, 4.67; N, 17.71; found C, 50.21; H, 4.34; N, 17.62. 142

CHiCONH

6-acetamido-l,4-dihydro-7-nitroquinoxaline-2,3dione. (110): A 10 ml R.B. flask was charged with 3 ml of conc. nitric acid (69-71 %) and chilled in an ice bath. Fuming nitric acid (3 ml) was slowly added, and after several minutes solid compound 109 was added in small portions over 10 min. Cooling and stirring were continued for 3 h (HPLC showed the reaction was essentially complete after two hours), and the reaction mixture was poured onto 30 g of ice. An additional 20 ml of water were used to rinse the remaining residue in the reaction vessel onto the ice, and the suspension of product was placed in the freezer overnight. The ice was allowed to melt and the product was collected, washed with water and dried in a stream of air. After standing in a dessicator overnight, the cake was resuspended in 30 ml of isopropanol, the solid chunks were crushed, and the liquid was suctioned off. This process was repeated twice more with isopropanol and once with diethyl ether, and the product was dried in a stream of air for 10 min. After 1 h of standing in air the yield was 0.67 g (85.6%) of the monohydrate as bright yellow crystals, m.p. >270°C. *H-NMR (DMSO): 6 12.27 (bs, 1H, N-H4), 12.02 (bs, 1H, N-Hl), 10.21 (bs, 1H, amide NH), 7.74 (s, 1H, H5), 7.55 (s, 1H, H 8 ), 2.07 (s, 3H, CH3). MS (FAB): ip/e 265 (M+l). Analysis for C 10H8 N4O5 HzO: calculated C, 42.56; H, 3.57; N, 19.85; found C, 41.94; H, 3.67; N, 20.13. 143

BF4-+ N 2

■72H20 NaBF4

111

6-Diazo-l,4-dihvdroquinoxaline-2,3-dione, 3/?hydrate. ( I ll) : Compound 108 (0.376 g, 2.00 mmol) was suspended in 20 ml of 1:1 48% HBF 4:CH3OH and chilled to 0-4°C in an ice bath. Sodium nitrite (0.41 g, 6.0 mmol) was dissolved in 5 ml of water, this solution was chilled in an ice bath and added in one portion. After 16 h at 0°C, the reaction was not complete (HPLC, 4% CH 3CN/0.1% CF3C 0 2H/H20 , 2.0 ml/min). Another 10 ml of 48% HBF 4 and another 0.41 g of NaN0 2 (again dissolved in 5 ml of water) were added. After a further 1.5 h of stirring the reaction was essentially complete. Isopropanol (80 ml) was slowly added with continued cooling, and the reaction vessel was placed in a freezer at -15°C overnight. The crystals which had formed were collected, washed with two 10 ml portions of ice-cold ethanol, one 5 ml portion of ice water, and two more 10 ml portions of cold ethanol. The yield after drying in a stream of air was 0.46 g (55.2%) as a pale yellow hydrated 1:1 complex with NaBF4. The product was redissolved in ice-cold water and stirred 30 min., filtered, and reprecipitated with four volumes of isopropanol. After several days in the freezer the crystals were collected, washed with two portions of isopropanol, and dried in a stream of air. The product gave one peak by HPLC with detection at 254 run (see above and general procedures). Decomposition point: no decomposition was evident below 100°C, rapid dec. > 150°C. The solid from this decomposition gave two peaks by HPLC, one of which coeluted with authentic 101. ‘H-NMR (d^-DMSO, TMS): 6 12.70 (bs, 2H, NH), 8.33 (dd, 1H, Jonho-8.9 Hz, Jmeta=2.2 Hz, H7), 8.23 (d, 1H, Jmeta=2.2 Hz, H7), 7.44 (d, 1H, 1 ^ * 8 .9 Hz, H 8 ). IR (KBr, cm'1): 2268 (s, N2+), 1711 (vs, C=0), 1591 (N-H bend), 1502 (C=C), 1370 (C-N), 1083 (B-F). Analysis for C8 H8 BF4N40 2 NaBF4 3/2H20: calculated C, 23.28; H, 1.95; N, 13.57; F, 36.82; found C, 22.97; H, 1.60; N, 13.23; F, 37.07. 144

1.4-dihvdro-6-methyl-7-nitro-quinoxaline-2.3-dione. (116): A 500 ml Erlenmeyer flask was charged with 165 ml of conc. (69-71 % nitric acid) which was chilled in an ice bath, and 165 ml of fuming nitric acid were slowly added. After a few minutes, 16.7 g (95.6 mmol) of solid M-dihydro^-methylquinoxaline^.S-dione 261 were added over 15 min. with stirring. The temperature rose, but never got above 18°C during this addition. After 45 min. the mixture was poured onto 1000 g of ice, and two hours later the product was collected, washed with two portions of water and two portions of isopropanol (resuspending the product the first time), and dried 30 min. in a stream of air to give 21.14 g (100%) of light yellow powder, m.p. >305°C. !H-NMR (d^-DMSO, TMS): 6 12.26 (bs, 1H, N-H4), 12.07 (bs, 1H, N-Hl), 7.82 (s, 1H, H 8 ), 7.04 (s, 1H, H5), 2.52 (s, overlaps with DMSO, CH3). IR (KBr, cm'1): 3147 (bd, N-H), 1727 (C=0), 1708 (C=0), {1632, 1604} (N-H bends), 1539 (N=0), 1508 (C=C), 1406 (C-N), {1343, 1322 (C-N, N-O)}. MS (FAB): m/e 222 (M+l). Analysis for C9 H7N30 4: calculated C, 48.88; H, 3.19; N, 19.00; found C, 48.78; H, 3.08; N, 19.03. 145

H I O

O OCH2CH3

o c h 2c h 3

122 l-(2’,2’-diethoxvethvl)-1.4-dihydroquinoxaline-2.3-dione (122): In a 500 ml round-bottomed flask, 2,3-dihydroxyquinoxaline (19.46 g, 120 mmol) was suspended in 30% ethanolic 2 M sodium hydroxide (500 mmol). Sodium iodide (1.80 g, 12 mmol) was added and the reaction was brought to reflux. Bromoacetaldehyde diethyl acetal (35.5 g, 180 mmol) was added to this clear solution and reflux was continued for 9 days, when 65 ml of 10% ethanolic 2 M sodium hydroxide were added. Reflux was continued for another 7 days. During this time, a solid formed on the sides of the flask. The contents of the flask were diluted to 1000 ml with hot water and adjusted to pH 6 with conc. phosphoric acid. The workup was continued in two portions: 500 ml of this suspension was extracted with three portions of 2:1 benzene-ethyl acetate (700 ml, 800 ml, 300 ml). The combined organic layers were washed with water (2 X 200 ml) and brine (100 ml), and dried over MgS04. The volume was reduced to about 50 ml on a rotary evaporator to give crystals, and the crystallization was allowed to continue overnight at room temperature. The product was collected by filtration and washed with two small portions of benzene. This workup was repeated for the other half of the reaction mixture, for a combined yield of 8.85 g (26.7%), m.p. 181-183°C. *H NMR (250 MHz, dg-acetone): 10.90 (bs, 1H, NH), 7.59-7.53 (m, 1H, ArH), 7.31-7.23 (m, 1H, ArH), 7.21-7.15 (m, 2H, ArH), 4.86 (t, 1H, J=5.3 Hz, CH), 4.29 (d, 2H, J=5.3 Hz, CH^, 3.80-3.68 (m, 2H, CHJ, 3.59-3.46 (m, 2H, CH^, 1.08 (t, 6 H, J= 7.0 Hz, CH3). IR (KBr): 3151 (bd, NH), 3066 (NH), {2974, 2928, 2864} (C-H), 1690 (C=0), 1604 (sh, N-H bend), 1512 (C=C), 1398 (C-N), 1068 (ether C-O). MS (El): m/e 278.1272 (calc. 278.1267, M+), 233 (M+-CH 3CH20), 205 (M=233 - C ^ ) , 103 (base, +CH(OCH2CH3)2). Analysis for Cj 4HlgN 20 4: calculated C, 60.42; H, 6.52; N, 10.07; 146 found C, 60.41; H,6.54; N, 10.16.

H O H20 O

V H 130 l,4-dihvdro-l-(2’-oxoethyl)QuinoxaIine-2,3-dione (130); Compound 122 was suspended in 60 ml of 1.2 N HC1 in a 100 ml R.B. flask, and heated 10 min. on a steam bath, giving a clear solution. The solution was allowed to cool slowly to room temperature for 1.5 hours, when crystals began to form. The flask was placed in a freezer overnight, and the product was collected by filtration, washed with water (15 ml), and dried in air to give 130 (2.93 g, 87.9%), dec. 223°. !H-NMR (dg-DMSO, TMS): 6 12.05 ppm (bs, 1H, NH), 9.68 (s, 0.15H, CHO), 7.51 (asyrnrn. t, 1H, J-3.8 Hz), 7.21-7.13 (m, 3H, ArH), 6.09 (bd, 1.8H, J-4.7 Hz, C(OH)2, exchangeable with added D20), 5.15 (bs, 1H, CH), 4.10 (d, 2H, J=5.4 Hz, CH^. IR (KBr, cm 1): 3371 (bd, N-H), 1678 (overlapped C=0’s), 1508 (C=C), 1405 (C-N). MS (El): 204.057 (calc. 204.053, M+), 176 (M+ - CO), 162 (M+ - C ^ O ) , 119 (Ph(NH 2)NCH+, base). Analysis for C 10H8 N2O3*H2O: calculated C, 54.06; H, 4.54, N 12.61; found C, 54.51; H, 4.55; N, 12.77. 147

l-(2>-amino-2,-carboxyethyI)-l,4-dihvdroqumoxaline-2,3-dione (131); Ammonium chloride (0.128 g, 2.4 mmol) and potassium cyanide (0.156 g, 2.4 mmol) were dissolved in 1.2 ml of water in a 5 ml round-bottomed flask. Solid 130 (0.408 g, 1.84 mmol) was added, giving a thick, unstirrable paste, and warming was noted. The solution was further wanned on the steam bath for 5 min., during which time some darkening of the solution was observed, then heating was discontinued and the reaction was stirred overnight. Heating was resumed for an additional 50 min., and then the contents of the flask were transferred to a 25 ml R.B. flask with 10 ml of 7 N HC1. The solution was brought to reflux overnight (HPLC analysis shows 1-2 h is sufficient, but the product does not appear to degrade under these conditions). The solution was diluted with water to 1500 ml and eluted through a 2.5 cm ID x 6 cm L cation exchange column (Dowex AG-50W, 100-200 mesh, hydrogen form). The column was washed with 500 ml of water and eluted with 1500 ml of 0.3 N ammonium hydroxide. The elution was monitored by spotting aliquots on a GF 254 TLC plate; the product gave a deep blue fluorescent spot under short-wave UV light viewed against the fluorescent background. The water and ammonium hydroxide were distilled off on a rotary evaporator to a volume of about 1 ml. Isopropanol (20 ml) was added to crystallize the product, and the solvent was evaporated to a damp cake. Another 20 ml of isopropanol were added and the solution was again evaporated to a volume of about 3 ml. Diethyl ether was added, and after a few minutes the product was collected, washed with diethyl ether, and dried in a stream of air to give 90 mg (15.0%) of an off-white solid, d.p. 200-201°C. JH-NMR (DMSO, TMS): 6 = 7.57-7.53 (m, 1H, ArH), 7.39-7.35 (m, 1H, ArH), 7.28-7.24 (m, 2H, ArH), 4.68 (d, 2H, J=7.3 Hz, Hp), 4.18 (t, 1H, J= 7.3 Hz, Ho). IR (KBr, cm 1): (3514, 3413} (bd, H 20 , NH3, COOH), 3213 (NH), 1698 148 (quinoxaline C=0), 1659 (sh, quinoxaline C=0), 1629 (sh, C 02* asymm. str.), 1600 (sh, N-H bend), 1506 (arom. C=C and N-H symmetric bend), 1402 (C02* symm. str.), 1380 (C-N), 1363 (C-N). MS (FAB): m/e 250 (M+l). Analysis for CnH nN aO ^H jO : calculated C, 43.57; H, 5.65; N 13.86; found C, 43.45; H, 5.70; N, 13.78.

H i

OH

132

l-(2’-CYano-2>-hvdroxv)-1.4-dihYdroquinoxaline-2,3-dione, (132): Compound 130 (1.78 g, 8.00 mmol) was suspended in 6.0 ml of water and warmed 10 min. on a steam bath, then allowed to cool to about 40°C. NH 4C1 (1.72 g, 32.2 mmol) and NaCN (0.780 g, 15.9 mmol) were dissolved/suspended in 2.0 ml HzO. The thick suspension of starting material was added during 5 minutes time. After completing this transfer, the mixture was heated 5 min. on a steam bath. After another 40 min., 8 ml of water were added. Although formation of the cyanohydrin was almost quantitative at this point (HPLC, 4% CH 3CN/0.1%CF3CO2H/H2O, 2.0 ml/min.), the mixture was allowed to stand another 17 h. About 2 ml of the mixture were removed (the remainder was reserved for additional experiments), from which the solid was collected by filtration, washed with several portions of water, and dried in a stream of air and then for several hours on a vacuum pump at 2 torr to a constant weight. Yield: 204 mg of an off-white powder, d.p. 196-200°C. *H-NMR (DMSO, TMS): 6 13.0-9.0 (s, 1H, NH), 7.6-7.S (m, 1H, ArH), 7.2-7.1 (m, 3H, ArH), 4.90 (t, 1H, J=6.7 Hz,

COMPUTATIONAL STUDIES ON THE INTERACTION OF

THE DOPAMINE AMINO GROUP AND AMINO GROUP REPLACEMENTS WITH CARBOXYLATE ANIONS AS A MODEL

FOR RECEPTOR INTERACTIONS

150 CHAPTER VI INTRODUCTION AND STATEMENT OF THE PROBLEM

Not all persons would be equally believed, Demerzel. A mathematician, however, who could back his prophecy with mathematical formulas and terminology, might be understood by no one and yet believed by everyone. -The Emperor Cleon I in Prelude to Foundation, by Isaac Asimov (Bantam, 1988)

6.1 Dopamine pharmacology Dopamine (153) is a key neurotransmitter in the mammalian central nervous system (CNS). Dopamine deficiencies resulting from the degeneration of neurons which have cell bodies in the substantia nigra and nerve terminals in the caudate nucleus and putamen are clearly related to the movement disorders syptomatic of Parkinson’s disease. Dopamine prodrugs such as l-DOPA (154) and dopamine receptor agonists which enter the CNS are thus used to relieve the symptoms of this disease .262 Deficiencies in dopaminergic transmission have also been linked to tardive diskinesias.263 On the other hand, dopamine antagonists have been developed for the treatment of psychoses, in particular , which seem to be related to excessive dopaminergic stimulation is some brain regions .264 The potency of most currently-used "typical" antipsychotics correlates with their ability to block dopamine D 2 receptors (see below for a discussion of DA receptor subtypes ) .265 In positron emission tomography (PET) studies, it was demonstrated that 70-80% of D 2 receptors in the striatum were occupied with therapeutically effective doses of drug .266 Examples of potent D 2 antagonists are raclopride (155) and remoxipride (156).

151 153 Dopamine

OH O OMeO N‘ H N » "cnCV'"3 c h 2c h 3 CH2CH3 Cl Br 154 155 156 l-DOPA Raclopride Remoxipride CH, HO HO NCH NCH N HO HO N-

H 157 158 159 SCH 23,390 SCH 39,166 Clozapine

Though Dj antagonists such as SCH-23390 (157) and SCH-39166 (158)267 are in preclinical or clinical study for schizophrenia, rodent studies reveal that D} antagonists can induce catalepsy and movement impairments .265 Treatment with D2 antagonists used as "typical" antipsychotics may cause so-called "extrapyramidal" side effects, such as dyskinesias and rigidity 264 These are most likely mediated by the same regions of the brain (in particular the dorsal striatum) where degeneration occurs in Parkinson’s disease 268 Atypical antipsychotics such as clozapine (159) have a reduced incidence of such side effects, which has been postulated to be due to inhibition of both Dj and D 2 receptors (but see below ) .269 Dopaminergic transmission is clearly involved in the 153 behavioral responses to abused drugs such as amphetamine, including cocaine addiction .270,271 Peripheral DA! dopamine receptors are located in vascular smooth muscle cells, where they promote vasodilation, and also in the renal tubules where they promote sodium excretion .268 Presynaptic DA 2 receptors are located in the terminals of postganglionic sympathetic neurons, and activation of these receptors inhibits release of norepinephrine. Based on these actions, dopamine is used in the treatment of circulatory shock and heart failure, and peripherally-acting dopamine agonists are being studied for their use in hypertension and other cardiovascular disorders .268

Postsynaptic dopamine receptors have been subtyped as Dj or D 2.272 Activation of Dt receptors stimulates cAMP production by adenyl cyclase through linkage to a pertussis-toxin sensitive G-protein. By contrast, activation of D 2 receptors generally inhibits cAMP production .268 The affinity of dopamine for Dj receptors is low (- pM) compared to D 2 receptors (~ nM), and occupancy of Dt receptors at a will therefore be high only immediately after release of large amounts of transmitter.

Dopamine D 2 receptors are also located presynaptically; these are now frequently referred to as dopamine autoceptors, and their activation inhibits release of dopamine from presynaptic cells .273 Recently, a new dopamine receptor subtype, D3, was characterized through gene cloning studies and transfection into COS-7 and

Chinese hamster ovary (CHO) cells .274 The protein sequence was 52% homologous with a reported D2A receptor, and no coupling with cyclic AMP (cAMP) formation was observed. There were distinct differences in brain distributions of the messenger RNA

(mRNA) for D 2 and D3 receptors determined by autoradiography, and the binding affinities differed for a number of dopaminergic agents as determined by inhibition of specific [ 123I]-sulpiride binding (CHO cells). The possibility has been raised 268 that at least some of the antipsychotic effects previously associated with D 2 receptor blockade 154

may be due to antagonism of D 3 receptors, especially in the corpus striatum (including

the olfactory tubercle, islands of Callejae and nucleus accumbens) which was found to be

enriched with D3 receptors. These areas are linked to the limbic system, which is

important in converting emotions into behavioral responses .268 D3 receptors are much

less abundant and more restricted in distribution than D 2 receptors. Recently, D 4 and D5

receptors have been cloned and expressed, and mRNA distributions characterized .275,276 Both of these receptors also display more restricted abundance and distribution

compared with D} or D 2 receptors. The D 4 receptor is of the same family as D 2 and D3 receptors, and is of particular interest to medicinal chemists because it displays a high affinity for the atypical antipsychotic drug clozapine .271,275 Levels of mRNA coding for the D4 receptor in monkey brain were highest in the frontal cortex, which seems to function abnormally in schizophrenia ,271 and low in the basal ganglia, which is important in the control of movement; this coupled with the relatively low affinity of clozapine for the Dj and D2 receptor may help explain the low incidence of extrapyramidal side effects (EPS). Because of the frequent occurance of potentially fatal but fully reversible agranulocytosis during clozapine therapy, this drug can only be used with frequent blood monitoring, but it is nonetheless used because of its uniquely low incidence of EPS and efficacy in psychoses resistant to therapy with other drugs. Fuller characterization of the

D4 receptor may aid in the search for new agents not having this serious flaw. The D 5 receptor 276 is of interest primarily because, despite its physical and pharmacological similarities to the Dt receptor, its affinity for dopamine is ten times higher, predicting a much higher receptor occupancy at synapses in vivo and a possible role for this receptor in regulating D 2 activity. D 5 binding affinities for most otheT agents tested were similar to that of the Dj receptor. Activation of the Ds receptor, as with Db stimulates adenyl cyclase. The mRNA for D 5 receptors was found to be expressed in limbic regions of the brain, such as the frontal cortex, striatum, hippocampus, and hypothalamus, but (in contrast to D} receptor mRNA) was almost absent in the basal ganglia. and D 5 distributions overlapped strongly in the olfactory tubule, olfactory bulb, caudate-putamen and nucleus accumbens. The higher affinity for dopamine together with the stimulation of adenyl cyclase suggest a role in the maintainance of dopaminergic tone (balance) and arousal .276

153 160 Dopamine

N(CH3)2 NH(CH3)2+

168 DMDA O O

O 162 • i 164

161 163

HO S(CH3)2+ C r ~ HO 166 N(CH3)3+ Se(CH3)2+

165 167 DAN* DASe+ 156

6.2 Elements of a hypothesis that the charged form of dopamine is involved in binding to a carboxylate group in D? receptors At physiological pH, dopamine (DA) may exist in either the uncharged amine (153) or the charged ammonium form (160). This question has been addressed by the synthesis and study of permanently-charged and permanently uncharged analogs of DA .277 In these studies, the ammonium group of DA was replaced with neutral methylsulfide (161). methylselenide (162). methylsulfone (163) and methylsulfoxide groups (164) or charged trimethylammonium (DAN*), dimethylsulfonium (DAS+), or dimethylselenonium (DASe+) groups (165-167).278’279 The permanently-charged compounds showed significant agonist activity in a D 2 system (Table 6 ) ,280,281 and were able to inhibit the binding of [ 3H]-spiperone (169. a selective D 2 antagonist) in a manner qualitatively similar to conventional DA agonists (Table 7 ) .281 In contrast, the

Table 6. Inhibition of 3H-acetylcholine release from rat striatal slices by dopamine analogs.

Drug IC 50 (pM) 95% Conf. Limits

Dopamine (153) 0 .1 2 (0.07-0.23)

DMDA8 (168) 0.16 (0.07-0.33) DAN* (165) 8.5 (5.4-13.0)

DAS* (166) 29.7 (15.8-55.9) DASe* (167) 9.0* (5.9-13.6)

Striatal slices were prepared from mice pretreated with a-methyltyrosine (depletes endogenous dopamine) and 300 pM a-methyltyrosine was present in the medium along with 10 pM cocaine (inhibits DA reuptake). aDMDA = N,N-dimethyldopamine. *This result was obtained without cocaine in a seperate study; however cocaine had little effect on the activity of DAN* or DAS*. 157

Table 7. Apparent equilibrium binding dissociation constants ______for dopamine and analogs and the effect of NaCl .281 No NaCl

KH(pM) Kl (mM) %H

DA (153) 0 .0 1 0 + 0 .0 0 2 0.35 + 0.04 60 + 3 DMDA (168) 0.025 + 0.007 0.72 + 0.17 51 + 3 DAN+ (165) 2.63 + 0.52 70.6 + 10.7 53 + 5 DAS+ (166) 1.28 + 0.28 76.9 + 36.5 55 + 7 DASe+ (167) 0.54 + 0.15 22.9 + 5.8 62 + 10

125 mM NaCl

k h (mM) KL(pM) %H DA (153) 0 .0 1 1 + 0 .0 0 1 0.96 + 0.003 24+1 DMDA (168) 0.118 + 0.023 1.73+ 0.17 26 + 5 DAN+ (165) 0.34 +0.16 160 +74 10+1 DAS+ (166) ND 46.8 + 7.4 0 DASe+ (167) 4.5 +0.15 152 + 99 30 + 6 Kh and represent high and low affinity dissociation constants determined by inhibition of 3H-spiperone specific binding in rat forebrain homogenates. %H is the percentage of high affinity binding sites detected in the mathematical analysis. ND - not detected. permanently uncharged methylsulfide, methylselenide and sulfoxide analogs were inactive.282 These results strongly suggest that it is the cationic form of a DA agonist that activates the D2 receptor 283,284 The permanently-charged analogs (165 and 166.

DAN+and DAS+) could not be shown to inhibit [ 3H]-SCH 23,390 (157) binding, which would be indicative of affinity for Dj receptors (however this is not surprising since dopamine itself has only micromolar affinity for Dj receptors ) .285

Dopamine antagonists also exist in solution in charged and uncharged molecular forms, having an amine group which presumably occupies the same region of space in the receptor cavity as that in DA .286 Permanently charged analogs 171 and 172 of the dopamine antagonist chlorpromazine (170) were able to inhibit the specific binding of

[3H]-spiperone in rat striatal membranes (Table 8 ).M7 These compounds also behaved as 158 o NH CCCH^N N

169 Spiperone ooa. a:cia

'N(CH3)2

170 171 172 Chlorpromazine CPZ-N+ CPZ-S+ dopamine D 2 antagonists in that they could reverse the apomorphine-induced inhibition of K+-evoked [ 3H]-acetylcholine release in striatal slices (Table 8 ).288 Chlorpromazine is a relatively unselective Dt/D 2 antagonist The permanently charged analogs CPZ-N*

(171) and CPZ-S+ (172) could also inhibit specific [3H]-SCH-23,390 binding in rat striatal homogenates (indicative of binding at Dt receptors); the values were 12,300 + 1600 nM and 2750 + 370 nM, respectively, compared with 6.2 + 2.0 nM for chlorpromazine 285 The permanently uncharged methylsulfide analog of chlorpromazine was too insoluble in aqueous solution to allow for proper biological testing; hence the permanently charged pyrrolidinium analogs 174 and 175. and the tetrahydrothiophenium analog 176 were synthesized, and their activities compared with that of sulpiride (173) and the uncharged tetrahydrothiophene (177)289 The permanently-charged analogs 174.

175. and 176 could reverse the inhibition of K+-evoked release of [ 3H]-acetylcholine by apomorphine (Table 9 ) ,289 while the uncharged analog 177 was inactive. Compound 176 159

Table 8. Activity of chlorpromazine and analogs in D2 receptor systems

Compound Ki (nM)* KB(nM T

CPZ (170) 1.2 ±0.3 72 CPZ-N+ (171) 2150 ±230 1350

CPZ-S+ (172) 280 ± 18 770

*Kj = equilibrium dissociation constant determined by inhibition of specific [ 3H]spiperone binding in rat striatal membranes. **Kb determined for antagonism of apomorphine-induced inhibition of [ 3H]-acetylcholine release in rat striatal slices.

and sulpiride (173) could also inhibit the specific binding of [ 3H]-spiperone.

The ability of permanently-charged agonists and antagonists to bind to the dopamine D 2 receptor (and activate it in the case of agonists), combined with the inactivity of

uncharged compounds, suggests interaction with an anionic site on the D 2 receptor .284

The compound 2-ethoxy-l-ethoxycarbonyl-l,2-dihydroquinoline (EEDQ) irreversibly

blocks the actions of dopaminergic agonists of Dt and D 2 receptors, which suggests that

the anionic group is a carboxylate .290,291 Williamson and Strange 292 have supplied

additional evidence that a carboxyl group is important in binding basic drugs to D 2

receptors: 1) the nonspecific protein modification agents N-acetylimidazole,

5,5’-dithiobis(2-nitrobenzoic acid), 1,2-cyclohexanedione and acetic anhydride had no specific effect on [ 3H]-spiperone binding (indicating no participation by tyrosine hydroxyl, free sulfhydryl, arginine or primary amine groups in ligand binding). 2) DCC

(dicyclohexylcarbodiimide) potently reduced the number of [ 3H]-spiperone binding sites, 160

O h 2n s o 2 NH

OCH

173 Sulpiride

h 2n s o 2 2n s o 2 O H N H NH H C + o c h 3 3 c h 2c h 3 3 (CH2CH3)2

174 175 SP-NMeEt+ SPNEt,+

O h 2n s o 2 h 2n s o 2 S I + x £o c r h 3 c h 2c h 3

176 177 SP-SEt+ SP-S

and this was preventable by incubation with D2-selective compounds. This reagent reacts

preferentially with carboxyl and sulfhydryl groups and to a lesser extent with tyrosine

hydroxyl groups; since modification by other sulfhydryl- and tyrosine-specific reagents

failed to produce effects, this was taken as evidence that the primary effect of DCC was

on a carboxylate group. Additionally, DCC has been shown to modify carboxyl groups in hydrophobic domains of transmembranal receptor proteins, e.g. the p-adrenergic receptor, providing circumstantial evidence that the carboxyl group involved in ligand binding is located within the membrane lipid bilayer. 3) The dependence of specific

[3H]-spiperone binding on pH indicates the importance of an ionization for a functional group with pKa - 5.2. 4) The amino acid sequence of the rat D 2 receptor 293 contains Asp 161

Table 9. Activity of sulpiride and analogs in D2 receptor systems

Compound KB(mM)* Kj (mM)**

Sulpiride (173) 0.32 0.0055 ± 0.0006

SP-NMeEt+ (174) 1.1

SP-NEt^ (175) 1 .2 SP-SEt+ (176) 5.0 5.2 ± 0.2 SP-S (177) inactive

*Kb = equilibrium dissociation constants for the inhibition of apomorphine antagonism of K+-evoked [3H]-acetylcholine release from rat striatal slices.

**Kj for inhibition of specific [ 3H]-spiperone specific binding in rat striatal homogenates.

80, Asp 114 and Glu 95 in putative transmembrane domains. Asp 80 is conserved in all G-protein-linked receptors whereas Asp 114 is conserved in those that bind catacholamines .294,295 Consistent with these observations, the sequences of the D 3 and

D4 receptor contain Asp 110 and Asp 115, respectively, while the D 5 receptor (of the "Di family") contains Asp 120 .274,275,276

6.3 Rationale for computational studies In all of the studies with permanently-charged analogs of dopamine and dopamine antagonists, the activity of these analogs is lower than that of the parent amines in both binding affinity and physiological effect (Tables 6-9). There are a number of possible reasons for this result. It is generally recognized that the observed activity of a drug in vivo is dependent on ( 1) its concentration in the vicinity of the receptor and ( 2 ) the value of the free energy change (AG) due to the formation of the receptor-drug complex. Regarding the first factor, the evidence discussed above indicates that the binding site for 162

H O ^

H O ^ y ^ N H ; •HO H O ^ HO

HO>^ Y^N M e3+ HO NMe3+ HO1^ 1 * h o £ > .

Figure 44. Partitioning of dopaminergic drugs into the receptor site. dopamine agonists is likely to be well within the membrane on the receptor protein; hence the drug molecules might have to enter a relatively hydrophobic environment, and the effective drug concentration at the receptor site would depend on the efficiency of partitioning into the membrane. This possibility has been raised previously .284 For an amine, some percentage of the molecules will be in the neutral uncharged form, depending on the pKa of drug molecule. This neutral form could partition into the 163 membrane, as depicted in Figure 44. At the binding site, it might then abstract a proton from a bound water or some acidic group on the peptide and reprotonate. By comparison, the permanently-charged compounds cannot deprotonate, and the charged form would have to partition into the membrane. Trimethylammonium [-N(CH3)3+] has an aromatic substituent lipophilicity constant of rt = -5.96, compared with -1.23 and 0.18 for -NH 2 and -N(CH3) 2 respectively .296 The partitioning kinetics for tetramethylammonium are clearly different than for ammonium: a chromatogram of phenethylammonium and trimethylphenethylammonium on a reverse phase column (Figure 45) shows evidence of these differences in the band shape. The observed band

N(CH3)3+

Conditions: 20% CH3CN/0.1%CF3CO2H/H2O 2 .0 ml/min. 5 pM End-capped C -8 column 15 cm L x 4.6 mm ID 254 nm

min. Figure 45. Chromatography of phenethylammonium and trimethylphenethylammonium on end-capped C-8 packing. broadening in the peak for the quaternary compound is not likely due to interaction with unprotetected silanol groups, since the peak for the amine is quite sharp; hence the band broadening must be due to slow partitioning into and/or out of the lipophilic stationary phase on the column. 164

Figure 46. Helical wheel model of the m l mAChR. Heavy circles represent potential hydrogen-bonding residues, smaller circles represent non-polar residues. Reprinted with permission from ref. 298.

Figure 46 shows a helical wheel model of the Ml muscarinic acetylcholine receptor with the putative transmembrane helices viewed as if looking down onto the membrane surface from outside the cell .297 There is evidence for drug binding to residues on several segments within the membrane. The heavy dots indicate possible hydrogen bonding residues; hence this model shows a somewhat hydrophilic environment established by the many hydrogen bonding residues within the membrane, with many hydrophobic residues facing the outside of this arrangement toward the lipid bilayer. It is not clear whether a "pore" in the membrane is established by this arrangement, or whether drug must diffuse through a relatively hydrophobic environment in order to reach the binding site. It is notable, however, that the agonist for muscarinic receptors is acetylcholine (178). a quaternary ammonium compound. If a pore is formed, the architecture of it might provide an additional means to achieve selectivity by limiting access to the binding site. Sequence homology between cholinergic and catecholamine receptors is relatively high, particularly in the putative transmembrane regions where the 165

drug binding site most evidentally lies ,298 so it is possible that access of the permanently-charged analogs to the dopamine receptor may be facile.

The free energy of binding of drugs to receptors (AG) is composed of enthalpic (AH) and entropic (TAS) components which can be equally important in the expression of drug activity. For many drugs which bind to dopamine receptors, temperature-dependence studies indicate that the binding is driven by entropy (exceptions are the substituted benzamides such as sulpiride (173). where the binding is dominated by the enthalpy component ) .299 Ions are "water structure breakers," hence solvation of ions is generally exergonic (AGCO); the pairing of ions to form an ionic bond in aqueous solution is generally endergonic (AG>0), since the total solvation shell for an ion pair is smaller than for individually solvated ions .331 This is essentially the reverse of "hydrophobic binding." The major contributions to the binding of a catecholamine would then be from the hydrogen bonding of the catechol hydroxyl groups to serine hydroxyl groups 292 (a AH component) and from hydrophobic (and/or "x-stacking") interactions of the catechol ring and methylene groups (primarily a AS component). In the case of the former, hydrogen bonds formed by the catechols should be little different in energy from those formed in aqueous solution outside of the receptor cavity. This hydrophobic bonding must also be sufficiently strong to overcome the unfavorable entropy changes associated with binding, i.e. the "freezing out" of rotational and translational degrees of freedom.

Unfortunately, the temperature dependence for permanently-charged analogs has not yet been studied, and the interpretation of such studies is not always straightforward .300

An interesting discussion of receptor specificity in adrenergic and cholinergic receptors recently appeared .301 A comparison of the binding of acetylcholine analogs to muscarinic receptors (Table 10a) with that for similar compounds to adrenergic receptors 166

Table 10a. Relative activity of acetylcholine homologs at muscarinic receptors.*

CH3C0 2CH2CH2NMe3+ 1

-NHMe^ 0 .0 2

-NH2Me+ 0 .0 0 2

-n h 3+ 0 .0 0 0 1

Table 10b. Relative activity of noradrenaline _____ homologs at«- and p-adrenoceptors.*

(3,4-OH)-PhCHOH-CH2NMe3+ <0 .0 0 1 <0 .0 0 1

-NHMe2+ 0.025 0 .0 0 1

-NH2Me+ 1 1

-n h 3+ 0.5 0.005

-NH2Et 0.08 1

-NH2-iPr 0 .0 0 0 1 0.16

* Adapted from reference 301.

(Table 10b) reveals large differences in the order of activity for various amine and quaternary ammonium analogs. Burger 301 proposed that the selectivity of cholinergic receptors for the quaternary group of acetylcholine is attained through an ionic bond reinforced by non-polar interactions. Support for this hypothesis comes from the study of synthetic "receptors" for quaternary ammonium compounds .302,303,304 One such synthetic receptor (179) was found to bind acetylcholine (178) with Kd = 50 pM in aqueous buffer. In this receptor molecule, esterification of the carboxyl groups did not significantly reduce binding affinity for organic cations, but the carboxyl groups contributed to improved aqueous solubility of the synthetic "receptor". Significantly, NMR studies indicated that for compound 180. the quaternary ammonium group was

i 167

R = CO,', CO,Et 179 R

CH, +Me,N Me3C NMe,+

180 Acetylcholine

NMe3+

AAG°so, = -65 kcal/mol

Figure 47. Synthetic acetylcholine receptor. preferentially bound within the receptor cavity even though solvation effects alone should greatly favor the binding of the t-butyl group (AAG°soi « -65 kcal/mol by statistical perturbation theory using Monte Carlo simulation). This preference is indicative of the large energies available through cation-n interactions. Other synthetic receptors 301,302 have incorporated anionic groups; acetylcholine is bound in one of these systems with Kd = 2 mM, but ammonium compounds are bound more effectively (c.f. 168

Table 10a and 10b!), as is the case with a synthetic receptor for dopamine .305 Cohen et. al. have discussed the lack of evidence for an anionic group binding directly to

acetylcholine in acetylcholinesterase ;306 one important aspect is that uncharged analogs of acetylcholinesterase antagonists potently inhibit the enzyme. Daugherty and

Stauffer302 also point out that in the X-ray structure of an immunoglobulin Fab fragment bound to phosphocholine, the first solvation shell for the quaternary cation is composed of electron-rich tyrosine and tryptophan aromatic rings.

Miller and Uretsky 284 proposed that the reduced activity of the permanently-charged analogs compared with the parent amines might be due to the ability of the protonated amines to form ionic bonds reinforced by hydrogen bonding, whereas the permanently-charged analogs would have no protons available to form traditional hydrogen bonds. The energy of a reinforced ionic bond is estimated to be AG = -10 kcal/mol vs. -5 kcal/mol for simple ionic bonds .302 This energy difference corresponds to a factor of 10 3 in binding affinity; a comparison with Tables 7, 8 and 9 shows this estimate accounts quite well for the observed relative binding affinities. Besides differences in the nature of the bond itself, it is not unreasonable to suspect that steric factors might be important. The D 2 receptor is fairly tolerant of bulk at the amine nitrogen, since N,N-dimethyl dopamine and dopamine are nearly equipotent in inhibiting

[3H]-spiperone specific binding in rat striatal membranes (Table 7). However in the quaternary ammonium analog there is an additional methyl group, and in the sulfonium and selenonium analogs the size of the sulfur and selenium atoms is larger than the nitrogen in the amine and quaternary ammonium analogs. Steric crowding at the anionic site might require adjustments in the conformations of the drug and/or receptor to accomodate the added bulk; these adjustments would detract from binding energy.

i 169 The biological systems in which binding and activity are studied are very complex. Moreover, only circumstantial evidence for receptor structure is available, and the nature of these membrane-bound receptor proteins makes it unlikely that any useful three-dimensional structure will be determined. It is therefore desirable to study

simplified systems in order to determine individual contributions to binding interactions. The studies reported herein concentrate on the interaction of the (presumed) cationic end of dopamine and various charged analogs with the receptor carboxyl group.

6.4 Summary of objectives The objectives of this work may be summarized as follows:

1) Determine the computational requirements for model dimers needed to study the interactions, beginning with acetate or formate as models for the aspartate carboxyl group and methylammonium, ethylammonium, and ethyltrimethylammonium as models for dopamine and the quaternary ammonium analog of dopamine (addressed in Chapter VII). Other groups, such as dimethylsulfonium, dimethylselenonium, guanidinium, and were planned for later study. Emphasis was placed on first

principles techniques of , namely Hartree-Fock (HF) and Density

Functional Theory (see Chapter VII), since semi-empirical models give adequate

descriptions only for "typical" bonding situations. 2) Address the question: "what is the nature of an ionic-reinforced hydrogen bond?" 3) Identify possible geometries for the interaction of a carboxyl group with amines

(Chapter IX) and other groups. Hartree-Fock and extended HF methods generally perform very well in predicting geometries without assumptions as to, e.g.,

particular tautomeric forms or bond orders ,307 and hence would be well suited to these studies if the size of the required computations proved to be manageable. The performance of Density Functional Theory (DFT) in this area has not been 169

characterized,308 and comparing the performance of DFT to HF and extended HF

methods was a major objective of the project. 4) Calculate relative interaction energies for carboxyl groups with amines and various amine replacements (chapters IX and X). These would be values for AE or AH in vacuo. The environment in the receptor cavity is clearly not that of a vacuum

(however for an extremely lipophilic receptor cavities it could be a fair approximation); neither is it the same as in aqueous solution, since any solvation

that occurs at the receptor likely consists of highly specific interactions with bound waters or functional groups from amino acid side chains on the receptor protein. The interaction energies which we planned to calculate would clearly not be values of AG for binding of this portion of a drug molecule to the receptor from free solution, which is likely to be almost energy-neutral or even endergonic (see discussion above). In addition, these calculations cannot take into account the

orientation of the entire drug molecule in the receptor which results from a balance between steric interactions and flexibility of the receptor cavity. They can however

provide valuable insight into possible optimal arrangements of the cationic and anionic groups of the drug and receptor, respectively. They can also provide information about the energetic penalties in cases where ideal configurations are

compromised by steric constraints. 5) Since experimental in vacuo interaction energies are not available for any of the

dimers we planned to study, we also calculated proton affinities using the various

HF, extended HF, and DFT techniques, since in this case experimental results are available for direct comparison (Chapter VIII).

Longer-term objectives for the project include the use of first principles calculations to derive force field parameters for use in simulations. Such 171 simulations would be useful in determining entropic (AS) components of binding. All of these studies would provide for more understanding of underlying interactions of drug-receptor interactions, thus providing a basis for rational prediction of amino group replacements which might confer greater activity and/or selectivity. CHAPTER VII

PRELIMINARY STUDIES AND HARTREE-FOCK THEORY

Anyone who is not shocked by quantum theory has not understood it. -Niels Bohr (1885-1962)

7.1 Introduction The rationale for performing model computational studies of drug and receptor fragments for dopamine and dopamine analogs was outlined in the previous Chapter. Semi-empirical methods do not perform well for calculating aminexarboxyl interactions ,309 though progress is being made in other hydrogen bonded systems with programs using the MNDO approach with AMI and PM3 Hamiltonians, i.e. AMPAC and MOPAC.310,311,312,313 Also, semi-empirical parameters do not exist for many of the molecular fragments we planned to study, such as quaternary ammonium, sulfonium and selenonium (see Chapter VI). High level quantum theory, however, is known to provide an adequate treatment of hydrogen bonding and other interactions for a variety of molecular pairs .314,315,316 In order to evaluate the computational demands of ab initio methods for the large systems we planned to study, we examined several key fragments which we felt would be minimal acceptable representations (Figure 48). We particularly concentrated on the anions and cations, which are known to require more extensive theoretical treatment than neutral molecules. These studies also provide data necessary for the derivation of force field paramaters for these systems, which will be used in the

172 Acetate Acetic Acid

H H H H H H H H H H N N \ H H H Ethylammonium Ethylamine

H H

H Ethyltrimethylammonium

Figure 48. Representative drug and receptor fragments. future for molecular simulations.

7.2 Ab initio Hartree-Fock theory A brief introduction to quantum molecular orbital theory as applied to molecules is presented here, so that it may be contrasted with the density functional approach in the next chapter. Hie reader is referred to references 317, 318, and 319 for more detailed 174 treatments. Atomic units are used in this presentation because they simplify the equations. In this system the unit of length is the Bohr radius (sq = 0.5292 A), the unit of mass is the resting mass of the electron (m^ = 9.1095 x 10 ' 27 kg), and the unit of charge is the electron (or proton) charge (e = 1.60219 x 10 ' 19 Coulombs).

The quantum theory of atoms and molecules dates to the early 1900’s, and in principle makes use only of fundamental physical constants, i.e. masses and charges, and physical laws governing their behavior (this is what is meant by ab initio - from first principles).

Under the Bom-Oppenheimer approximation, nuclear and electron motions are treated as separable because the time scale for nuclear motion is very much longer than the time scale for electron motion; i.e., on the time scale of nuclear motions such as vibration or , adjustments of electronic configuration are essentially instantaneous (see ref. 319, pp. 316-318). The electronic portion of the time-independent Schrddinger equation

(equation 7.1) gives the energy E as an observable eigenvalue (observable means a quantity which can be measured) of the Hamiltonian operator H operating on a stationary wave function, nr=Y(xlfX2,---,*n)» describing the system.

(7.1) H\y = E\|/

The jq are electron coordinates including both space and spin components. The

Hamiltonian operator consists of a kinetic energy operator T, and potential energy operators Vne and V ^ , so that eqn. 7.1 can be rewritten as in eqn. 7.2; the individual operators are given in eqns. 7.3-7.5.

(7.2) (T + Vne + Vee)V = EV

(7.3) T = E , (-VjV,2).

where V2 = + d^ld y 2 + cP/dz? 175

(7.4) Vne = Zi>a(Za/ria)

(7.5) Vee = Zi

Vne is the coulomb attraction term for the positive nuclei and negative electrons, whereas VM is the electron-electron repulsion operator. The total energy W (equation 7.6) must include an energy term for nuclear-nuclear repulsion, which is simply given by equation

(7.7), where Ra and Rp are the intemuclear distances and Za and Zp are the charges on the nuclei being considered in the sum. Note that in eqns. 7.5 and 7.7, the sums are constructed to avoid counting any interactions twice.

(7.6) W = E + Vnn

(7.7) Vun = £(x

In principle, these equations are exact except for errors introduced by the Bom-Oppenheimer approximation and the neglect of relativity effects; the latter are small for first and second row atoms but become larger in metallic elements. However, the exact wave functions cannot be found for molecules more complicated than H2+ (i.e., systems having more than one electron) because the electronic variables cannot be separated in the resulting differential equations, and therefore analytical solutions are not possible. There are actually any number of eigenfunctions \|/k with eigenvalues Ek which solve equation 7.1, and for exact solutions these correspond to the ground state energy and the energies of all possible excited states. For many problems of chemical and biochemical interest, we are interested in ground state behavior. Although we do not know the exact wave function for the ground state, the "variational principle" lets us find an approximation to both the energy and the wave function at any degree of accuracy, provided the computational power is available. For a system in a state with wave 176 function y, which does not necessarily solve equation 7.1, the "expectation value" of the

energy is given by eqn. 7.8.

(7.8) E[y] = Jy*Hycbc/Jy*ydjc,

where y* is the complex conjugate of y.

The square brackets in eqn. 7.8 indicate that y determines E, so that E is a functional of y (this is relevant for the discussion in chapter VIII). The variational principle states that

E[y] 2: E0 (for a proof, see ref. 319, Chapter 8 ); this means in practice that we can take an appropriate guess at the wave function, and keep improving it (e.g. by changing adjustable parameters in the function) according to the extremum condition of equation 7.9. The resulting energy will then be an upper bound to the ground state energy.

(7.9) 5 (fy*Hycbc -Ejy*ydjc) = 0

If y in equation 7.9 were the exact ground state wave function y 0 (and therefore an eigenfunction of the hamiltonian operator H), then eq. 7.9 would be equivalent to eqn.

7.1, and the energy, E, would be the ground state energy Eq .

The way in which equation 7.9 is applied to molecules makes use of the Hartree-Fock

Self-Consistent-Field (SCF) method, which was first developed for treating multi-electron atoms .319 In atoms an approximate wave function is written as a product of atomic wave functions, the form of which are usually based on the exact (non-relativistic) solutions for the hydrogen atom. The total wave function y for the multielectron system is written as a Slater determinant (eqn. 7.10) involving one-electron functions j. The Slater determinant satisfies a fundamental property of electrons (which belong to the fermion particle family), namely that the wave functions be antisymmetric with respect to exchange of the coordinates for any two electrons. In practical 177

n (*i)

1 ♦1(^2) $l(x2) ••• (7.10) Vh f " — « • ■ /N 4 • •

♦ * •

+i(*n )4>2(* n ) — ] contain adjustable parameters which are optimized according to the variational principle described above. For most molecular calculations, the

University of Chicago, and independently by Hall .321

(7.11) i = EjCjXj

The x/s comprise a complete set called basis functions. Originally, functions having the form of hydrogen-like atomic orbitals (Slater-type orbitals, or STO’s) were used for the

Xj’s (i.e. s functions, p functions, d functions, etc.); however the form of the integrals (see below) is inefficient for computation by electronic computers, and many modem quantum chemistry programs use functions.

The Hartree-Fock approximation is a method for minimizing the Slater determinant according to equation 7.9. Briefly, the exact Hamiltonian operator H in equation 7.9 is replaced by the Fock operator F, and the wavefunctions y are replaced with The energy expectation value is (eqns. 7.12-7.15):

(7.12) EnptvHF] = Jvhf*^Vhp^*/Jvhf*'Khf^

= Z A + V2 V ij-K ij)

(7.13) ^ = J«j)i*(x)[-V2V2 + v(x)Mx)dx 178

(7.14) Jy= Jc|>i(* 1)(lii*(j: 1)[l/rl2](t»j*(jf2)(f)j(ac2)dxidx 2

(7.15) Ky = fi*(xi)texdx2

The J terms are classical coulomb repulsion, while the K terms are non-classical exchange integrals. It is the evaluation of these integrals in a HF problem which consumes the bulk of the processor time and memory requirements. The constants Cj in equation 7.11 are adjusted in an iterative fashion to improve the overall description of the wave function yjjp, thereby minimizing the energy in eqn. 7.12. The orbital energies

€j are then obtained by solving the resulting set of Fock differential equations (7.16).

(7.16) Fi(x) = Sj€y(J>j(af), where F = H + J - K

(7.17) €y = JV Ffcdjr = Hj + Ej (Jy-Ky)

The €y are the Hartree-Fock orbital energies, and the total energy of the system is given by eqn. 7.18:

(7.18) WHF = r i e1-Vee + Vnn where is the total electron-electron repulsion energy which must be subtracted because it is counted twice in the orbital energy sum. It is just these values of Wnp that are reported in our calculations in the next section. Most readers will be familiar with the qualitative results of such calculations. Orbitals such as those shown for H20 in

Figures 48 and 49 are obtained.

The Hartree-Fock approximation is basically incorrect in that it calculates the interaction of a given electron in the averaged field of all other electrons (including itself); this 179

lai = 1.000(01s)+0.015(02s1)+0.003(02pz)-0.004(H1ls+H2ls)

2a, = -0.027(0ls)+0.820(02si)+0.132(O2pz)+0.152(Hj ls+H 2 ls)

lb 2 = 0.624(O2py)+0.424(H1 ls-H 2 ls)

3a, - -0.026(01s)-0,502(02s1)+0.787(02pz)+0.264(H1ls+H2ls)

lb , = 0 2 px

Figure 48. Five lowest SCF molecular orbitals found for minimal basis set treatment of symmetrical H20. (Pitzer and Merrifield, J. Chem. Phys., 52 (1970) 4782.

_L 20 2 bz 10 / V ^ N . 0

/ / lb , \ 4 - -10 / - df- ____ H ,ls H21s -20 02p, 02py2d „02pz''s" 02o, \ ;-/^ ' ' / / >t>2 / — -30 / ' t

0 2 s — -40 2 a, -50 Figure 49. Valence shell MO diagram for Water. self-consistent approximation is made in order to allow separation of variables in the differential equations of motion for the electrons. In a real system, however, electron motions are correlated (that is, the electrons avoid each other to lower the coulomb repulsion energy). Methods have been developed to attempt to recover some of this correlation energy; the Hartree-Fock wave functions are generally used as a starting 180 point for these approximations, and thus these methods are often referred to as "extended

Hartree-Fock". The popular Moller-Plesset approximation ,318,322 uses

Rayleigh-Schrddinger perturbation theory. This treatment has been computationally refined to a high degree, but suffers from the fact that it is not variational (not subject to the extremum condition in equation 7.9), thus it is sometimes possible to get energies lower than the true energy of the system. Still, for studying complexes this method has the advantage of "size-consistency" (for definition see reference 318).

7.3 Computational details

For the computations discussed in this chapter, we used the program CADPAC 323 running on a Cray Y-MP8/864 supercomputer. Basis sets were selected from the library included with the program. Molecular fragments were built and optimized with the program SYBYL ,324 using the "cleanup" routines to obtain reasonable starting geometries. For acetate and ethylammonium, these geometries were used directly for further optimization with the minimal STO-3G basis set in CADPAC. For ethyltrimethylammonium, further minimization was performed using the MAXIMIN molecular mechanics routine with default parameters (i.e. the Tripos force field 325 and conjugate gradient minimer). The geometry obtained with this routine was compared with that obtained by the minimizer in ALCHEMY 326 and the ANNEAL routine in SYBYL. There were no significant differences in the resulting geometries. The

STO-3G optimization with the BFGS algorithm in CADPAC failed to converge for ethyltrimethylammonium, so the initial HF optimization was performed with the 3-21G basis.

Full geometry optimizations were done with CADPAC using each basis set studied, and the Broyden-Fletcher-Goldfarb-Shanno (Newton-Raphson) minimization routine .327 181 Geometry optimizations were terminated when the largest component of the energy gradient was less than 1 x 10 *4 hartree/bohr. In general, the coordinates from geometry optimization at one level of theory were used to begin optimization at the next higher level.

Geometry minimizations usually make use of the electrostatic theorem, which is a special case of the general Hellman-Feynman theorem (eqn. 7.18). The result of the electrostatic theorem is that the force on a nucleus k at coordinates xk, yk, zk is given by equation 7.19, where i, j, and k are the unit vectors in the x, y and z directions.

(7.18) dEJd\ =JVn*( 0 H/dX)yndT,

where t are all spatial and spin coordinates.

(7.19) Fk = -i(aU/dXk) -j(0U/0yk) - k( 0 U/0 zk), where

(7.20) dU/dq^ = /v e*[(aVc^0qk)+(aV 1J 0 q k)]veldrel

In eqn. 7.20, Vn,, is the same as in equation 7.7, and Ve|= VM + V^, as defined in equations 7.4 and 7.5. In practice the theorem means that if the wavefunction is known

(the theorem works for Vhf )» the forces on the nuclei can be evaluated analytically. These forces can be used to direct nuclei to new coordinates, and the process is repeated until satisfactory convergence is achieved.

7.4 Results and discussion The total energy for acetate (Table 11) decreases rapidly in going from the minimal basis set (STO-3G) to 6-31G, but seems to converge for larger basis sets (0.01 atomic units = 6 kcal/mol). Significantly, the HOMO (highest occupied molecular orbital) energy continues to decrease for larger basis sets (Table 11, Figure 50), and as expected changes 182 -0.1200 321G -0.1300'

-0.1400

-0.1500 631G

S -0.1600

5 -0.1700 DZP DZPP

-0.1800 631GE 631 GEE

110 130 Number of Bosla Punetlona

Figure 50. Effect of basis set on the energy of acetate HOMO.

significantly with the addition of "diffuse" functions (6-31GE basis) to heavy atoms (in

this case, there are two essentially degenerate HOMOs on oxygen). This clearly should affect bonding, and the proton affinity for acetate (see Chapter VTII) agrees best with

experimental values when diffuse functions are included.32* Bond lengths and angles appear to be treated adequately at the 6-31G level (Table 11 and Figure 51), but even the

minimal basis set gives a good representation of the C-H bond length and a surprisingly good treatment of the O-C-O bond angle (Figure 52). The latter is probably fortuitous,

since diffuse functions are generally needed for proper bond angle descriptions in

anions .329 Dihedral angles (not shown) are essentially invariant for all basis sets (within

0.2°). The partial charges obtained by Mulliken population analysis 330 show no clear trend with varying basis set (Table 11, Figure 53), but the way in which the electrons are partitioned between atoms in this analysis is highly dependent on the nature of the basis set (note the difference between DZP and 6-31GE in Figure 53), and the use of this u —C—O Bond Anglo C—O Bond Length (Angrtroma) 128.0 128.5 1.230 1.235 1.240 1.240 1.245 1.250 130.0 1.2S5 1.260 1.260 1.265 130.51 1.270 10 Figure 52. Effect of basis set on acetate O-C-O bond angle., Figure 51. Effect of basis set on acetate C-O bond length. bond C-O acetate on set basis of Effect 51. Figure ST03G ST03G —t— 30 631G 321G 321G 631G DZ ubr f ol Funeftona Boala of Number ubro Bsa Funeftona Basla of Number 0 0 SO 70 50 - 4 - 631G* 631G** • • Z DZPP DZP • • H 631GE 3G 631 GEE 631GE • • 110 110 631 GEE 130 130 183 184 -0.45

ST03G -0.30

-0.55

S -0.60 DZP DZPP

-0.65

S -0.70

631G* 631G** -0.75

-0.80; 110 130 Number of Baala Funetlena

Figure 53. Effect of basis set on acetate oxygen partial charge. method of population analysis is declining. It is also clear that including polarization

(p-type) functions on the methyl hydrogens in this molecule makes very little difference in any of the parameters examined, and it should be possible to omit these functions in further calculations. A summary of the basis sets used for acetate, and the computer time requirements, are given in Table 12.

For ethylammonium cation, the minimal (STO-3G) and 3-21G basis sets substantially exaggerated the C-N and N-H bond lengths (Table 14), which would almost certainly affect binding behavior. The H-N-H bond angle is reduced almost 1° when polarization

(p-type) functions are included on these hydrogens ( 1° seems fairly insignificant, but makes a substantial difference in, e.g., calculated vibrational frequencies). Increasing the size of the basis set appears to make the hydrogens substantially more positive (Figure 54) and the nitrogen more negative (Figure 55). It is notable that with the 185

Table 11. Effect of basis set on various parameters calculated for acetate anion. Hs 03 % / H«— C,—C2 / \ Hi ° 4

Total HOMO LUMO Basis Energy Energy Energy **02-04 ** 01-02 **01-115 Z.03-C2-0 Set (au) (au) (au) (A) (A) (A) (degrees) STO-3G -224.0483 0.07109 0.65730 1.263 1.631 1.088 130.5 3-21G -225.9331 -0.12695 0.45763 1.251 1.575 1.088 129.8 6-31G -227.1228 -0.15547 0.41826 1.262 1.540 1.087 128.3 6-31G* -227.2251 -0.15989 0.43142 1.235 1.554 1.090 129.5 6-31G** -227.2299 -0.15988 0.43129 1.235 1.554 1.090 129.5 DZP -227.2837 -0.17258 0.44127 1.239 1.551 1.089 129.2 DZPP -227.2891 -0.17242 0.44184 1.239 1.551 1.090 129.2 6-31GE -227.2575 -0.18703 0.23697 1.235 1.548 1.090 128.7 6-31GEE -227.2624 -0.18741 0.19721 1.235 1.547 1.091 128.7

Mulliken Partial Charge on: Set Cl C2 03 H5

STO-3G -0.23 +0.23 -0.50 0 .0 0 3-21G -0.69 +0.81 -0.80 +0.16 6-31G -0.52 +0.70 -0.77 +0 .1 2 6-31G* -0.55 +0.72 -0.76 +0 .1 1 6-31G** -0.40 +0.71 -0.76 +0.07 DZP -0.65 +0.44 -0.62 +0.15 DZPP -0.47 +0.43 -0.62 +0.09 6-31GE -0.81 +0 .6 6 -0.74 +0.21 6-31GEE -0.76 +0.58 -0.75 +0.21 Table 12. Basis sets and calculation times for acetate anion.

Basis Functions on Total # Tine Per Optimization . i ■ of Basis Iteration Time Basis Set C,0 H Functions (sec) (sec)

ST0-3C Is 2s 2p(3) Is 23 17 300

3-21G Is 2s 2p(3x2) ls(2) 42 23 420

6-31G Is 2sp(4x2) 42 43 390 ID -- ls(2) 6-31G* 66 140 1840 +d(6) 6-31G** D- +P(3) 75 238 1190

DZ ls(2) 2s(2) 2p(3x2) ls(2) 46 100 1010 DZP Is (2) 70 250 1750 +d(6) DZPP 3 - +P(3) 79 360 2160

6-31GE 2sp(4x2) 1(4) ls(2) 106 700 5600 6-31GEE D - [ l s d(6x2) +s +p(3x2) 127 1500 9000

(3600 sec - 1 hr processor time - Cray YMP) 187

0.52 031 GE 0.50 031GEx 0.48 0310* « •DZP P 0.40 6\ '8 z a 3210 6 0.44 ”5 0.42 ro a. 031 G*x c 0.40+ a0 B 0.38 DZPx 0.30 0.34 ST03G 0.32 0.30 10 30 SO 70 90 110 Numb0r of Basil Function* Figure 54. Basis set dependence of Mulliken partial charge on N-H hydrogen in ethylammenium.

Mulliken partitioning scheme, the nitrogen is always negative, and with the largest (and presumably most reliable) basis sets the nitrogen carries a charge of nearly -1. While the total Hartree-Fock energies reach a reasonably constant value with basis sets larger than

6-31G*, adding diffuse functions greatly lowers the energies of the two lowest unoccupied molecular orbitals (Table 13, Figure 56). No other orbitals are greatly affected. This might be expected to greatly alter binding behavior, and we later found this to be the case (Chapter IX).

We performed a limited series of calculations on ethyltrimethylammonium cation. This molecule is very large for treatment by ab initio methods, with 6 heavy atoms and 14 hydrogens; the modest 6-31G** calculation required 160 basis functions. Each iteiation of the geometry optimization required 3750 seconds of processor time (Table 15), and the total optimization (4 iterations) needed 20,000 seconds. The optimization using the 188

-0 .3 "STOUT

-0.4. DZPx • -0 .5 ■_ a a - 0 .6 631 G*x ■3 ? -0 .7 OZP L DZ e 321G & - 0.8 631G* | 6*1 G x -0 .9 631GEx ■ - 1.0 631 GE

- 1.1 10 30 50 70 90 110 Number of Bette Function* Figure 55. Basis set dependence of Mulliken partial charge on nitrogen in ethylammonium.

DZP basis required 16 iterations to converge, thus consuming 40,000 seconds of processor time. The C-N bond lengths were shortened significantly when polarization (d-type) functions were added to these atoms, otherwise there were no large differences in geometry among the basis sets examined (Table 16). As was the case for acetate and ethylammonium, the nature of the basis set significantly affected the Mulliken partial charges (Table 15); however the nitrogen was again always negative, with the hydrogens carrying the positive charge distributed among them. In Figure 57, we compare the Mulliken partial charges for ethylammonium and ethyltrimethylammonium calculated using the 6-31G** basis set. In ethylammonium, the positive charge is highly concentrated on the three ammonium hydrogens, whereas in ethyltrimethylammonium the charge is distributed over 11 hydrogens vicinal to the quaternary nitrogen. 189

Table 13. Effect of basis set on HOMO, LUMO and total energies, and partial charges. Hio H9 I H4 \ I / C2— N,+ / ^ He Cj Hu H5 A H 7

Total HOMO LUMO Mulliken Partial Charges Basis Energy Energy Energy Set (hartrees) (hartrees) (hartrees) C l C2 N3 H4

STO-3G -133.0496 -0.6843 +0.2323 -0.19 +0 .0 2 -0.34 +0.32 3-21G -133.8864 -0.7029 +0.0025 -0.63 -0.23 -0.83 +0.44 6-31G -134.5658 -0.6996 -0.0288 -0.49 -0.17 -0.87 +0.46 6-31G* -134.6165 -0.7028 -0.0288 -0.53 -0 .2 2 -0.85 +0.47 6-31G*X“ -134.6122 -0.7005 -0.0281 -0.46 -0.26 -0.62 +0.40 6-31G*NH*b -134.6268 -0.7024 -0.0284 -0.52 -0.23 -0.62 +0.40 6-31G** -134.6349 -0.7009 -0.0281 -0.38 -0.13 -0.63 +0,40 DZ -134.5854 -0.6998 -0.0152 -0.52 -0 .2 2 -0.77 +0.45 DZP -134.6374 -0.7028 -0.0154 -0.52 -0.24 -0.75 +0.46 DZPXC -134.6319 -0.7007 -0.0144 -0.53 -0.24 -0.48 +0.37 DZP-NHP*1 -134.6471 -0.7026 -0.0146 -0.52 -0.24 -0.48 +0.34 DZPP -134.6567 -0.7006 -0.0140 -0.33 -0 .1 0 -0.49 +0.37 631GEe -134.6208 -0.7027 -0.0931 -0.79 -0.24 -1.03 +0.52 631GEXf -134.6165 -0.7004 -0 .1 1 2 1 -0.46 -0.37 -0.96 +0.49 631GE-NHE8 -134.6208 -0.7027 -0.0930 -0.79 -0.25 -1.03 +0.52 “Polarization functions on C2, N3, H4, H5, and HI 1. bPolarization functions on C l, C2, N3, H4, H5, and HI 1 cAs in note a. dAs in note b. “Diffuse s and p functions, and two sets of polarization functions on all heavy atoms. fAs in note e, except extra functions only on nitrogen, and diffuse s functions on H4, H5 and HI 1. 8 As in note f, except extra functions are also added to C2. 190

Table 14. Effect of basis set on selected bond lengths and angles in ethylammonium. H M / C 2 - N +

H — C , ri 'H A H H

Basis Rci-C2 ^C 2-N 3 R N3-H4 Z C A N a Z.C2N3H4 AH4N3H5 Z.C1C2N3H4 Set (A) (A) (A) deg. deg. deg. deg. STO-3G 1.517 1.516 1.042 110.3 1 1 1 .6 107.5 179.9 3-21G 1.527 1.560 1.018 109.5 1 1 1 .1 108.4 179.9 6-31G 1.519 1.537 1.009 1 1 0 .2 111.3 108.1 179.9 6-31G* 1.518 1.520 1 .0 1 2 110.3 111.9 107.4 179.9 6-31G*Xa 1.516 1.519 1 .0 1 0 110.3 111.7 107.6 179.9 6-31G*NH*b 1.518 1.518 1 .0 1 0 110.4 1 1 1 .8 107.5 179.9 6-31G** 1.517 1.518 1 .0 1 0 110.4 111.7 107.5 179.9 DZ 1.526 1.542 1 .0 1 1 110.4 1 1 1 .1 108.3 179.9 DZP 1.520 1.520 1 .0 1 2 110.4 111.7 107.6 179.9 DZPXC 1.520 1.520 1 .0 1 0 110.4 1 1 1 .6 107.7 179.9 DZP-NHPd 1.520 1.519 1 .0 1 0 110.5 1 1 1 .6 107.7 179.9 DZPP 1.520 1.518 1 .0 1 0 110.5 1 1 1 .6 107.7 179.9 631GEe 1.517 1.516 1.009 110.3 111.7 107.5 179.9 631GEXf 1.517 1.516 1 .0 1 0 110.4 1 1 1 .6 107.7 179.9 631GE-NHB 1.517 1.516 1.009 110.3 111.7 107.5 179.9 “-gSee notes Table 13. LUMO Energy (harlreea) Figure 56. Effect of basis set on the energy of the LUMO in ethykunmonium. - -0.08 -0.06 -0.04. - ■°-,Vo 0.10 0.02 0.00 0.02 3G 831G*831(S»x 631G 321G • • • • Z Z DZPx DZP DZ • • • • 0 0 0 110 90 70 50 Number ef ef Number Baala Functions 631 GE 631GEx 191 192

8+0.19 h 6+019 H I 5+0.19 \ J 6+0.19 6 +0.19 H H c H \! / , N. 6-0.60 „ 6+0.19 / C-"-H H 6 C _ 5 +0.19 I H 5+0.19 I 5+0.19 H 6+0.19

8+0.21 H 6+021 H 5+0.40

N 5 -0.63

5+0.40

8+0.40 H

Figure 57. Comparison of Mulliken partial charges for ethylammonium and ethyltrimethylammonium. Table 15. Effect of basis set on computation times, energies, and partial charges for ethyltrimethylammonium. \rHl3 Hio H9 C4 H14 \ f / C2 N, H 18 / M 2 IV \

H 15

# Time Total LUMO Mulliken partial charge on: Basis # Basis per Energy Energy Set Shells Functions Iter, (sec) (hartrees) (hartrees) C l C2 N3 C4 H ll 6-31G 46 82 660 -251.61255 0.0240 -0.50 -0.12 -0.77 -0.27 +0.23 6-31G^ 52 118 1550 -251.72076 0.0253 -0.53 -0.17 -0.57 -0.34 +0.24 6-31G** 6 6 160 3750 (4)* -241.74295 0.0251 -0.38 -0.07 -0.60 -0.18 +0.19 DZ 64 8 8 1350 (18)* -251.64257 0.0368 -0.59 -0.15 -0.33 -0.40 +0.24 DZP 70 124 2500(16)* -251.75381 0.0397 -0.57 -0.20 -0.17 -0.43 +0.24 ♦Iterations to convergence. Table 16. Effect of basis set on bond lengths, angles, and torsions for ethyltrimethylammonium.

H|3

\ l Hio H9 C4 H 14 M / Co N , n H 18 / *-12 H6 — c x IV H 19 H 8 H? I H 15

Bond Lengths (Angstroms) ______Bond Angles (degrees) ______Torsion Angles (degrees) odais Set ®C1-C2 ®C2-N3 RN3-C4 ®C4-H11 C1C2N3 C2N3C4 C5C3C12 H9C2H10 H11C4H13 C1C2N3C4 H6C1C2N3 6*31G 1.522 1.531 1.510 1.078 115.6 107.8 109.5 108.3 109.9 -179.3 -179.6 6-31G* 1.522 1.518 1.497 1.079 115.9 107.9 109.4 108.1 109.9 -179.3 -179.6 6-31G** 1.521 1.518 1.497 1.080 115.9 107.9 109.4 108.1 109.9 -179.3 -179.6 DZ 1.529 1.537 1.517 1.078 115.7 107.8 109.5 108.3 109.8 -179.3 -179.6 DZP 1.524 1.518 1.499 1.079 116.1 107.8 109.4 108.1 109.9 -179.4 -179.6

VO-P- CHAPTER VHI PROTON AFFINITIES AND THE DENSITY FUNCTIONAL THEORY

Truth suffers from too much analysis - Ancient Fremen Saying (from Dune Messiah, by Frank Herbert).

8.1 Introduction The interaction energies between the formate anion and either methylammonium (Chapter DC) or tetramethylammonium cation (to be reported later) calculated relative to the charged species at infinite seperation are large, for the former on the order of 1 2 0 kcal/mol. The reason for these large energies (which are in vacuo, not solution energies) is the high proton affinity of the anion; experimentally, for formate AH = 348.5 kcal/mol .332 There are no experimental energies or geometries for gas phase interactions of formate with methylammonium or tetramethylammonium, but gas phase proton affinities have been measured for many molecules .333,334 Interaction energies and vibrational frequencies can be calculated by ab initio methods and used to compute proton affinities for direct comparison with experimental values.

Hartree-Fock theory and post-Hartree-Fock correlated methods sometimes give systematic errors in the calculation of proton affinities (Table 1 7 ) 335»336 Proton affinities of anions are generally reproduced well provided that diffuse functions are used

(represented by + in Table 17, as in 3-21+G and 3-21++G, where the former indicates diffuse functions on heavy atoms only and the latter on hydrogens as well ) .328,337

Evidentally the effect of these functions is to provide a better represention of the electron

195 196

Table 17. Calculated and experimental proton affinities for formate and methylamine335,336

Proton Affinity (kcal/mol)

Basis Set Formate Anion Methylamine

STO-3G 47 8 269 3-21G 374 237 4-31G 357 232 4-31G* (*)a 371 232 6-31G 358 231 3-21+G 350 228 3-21++G 347 229 6-31++G 348 228 6-31G* 364 229 6-31G*(*)a 368 231 6 -31G** 368 231 6 -31+G* 352 228 6-31G* (*)a,b 230 MP2/6-31G* (*)b,c 230 MP3/6-31G* (*)b-c 231 MP4/6-31G* (*)b,c 230

Exptl.d 353 222

aPolarization (p) function added to acidic proton. From ref. 336. cMPn indicates use of M0ller- Plesset perturbation theory carried to order n (see text). ^Experimental energies (see Table 26) corrected for zero-point energy and to 0° K (see text). distribution in the anion by providing function space further from the oxygen nuclei, thus lowering the electron-electron repulsion energy. The best calculations for methylamine are still in substantial disagreement with the experimental result (this could be cause for reinvestigating the experiment, but see below). There also appears to be no significant effect of electron correlation based on calculations using Moller-Plesset theory (Table

17).

For reasons discussed in Chapter IX, we wished to compare the results of density functional theory (DFT) to that of Hartree-Fock theory in calculations on hydrogen 197 bonded systems. Since experimental results were available for proton affinities, and proton affinities bear relevance to interactions between formate or acetate anions and organic cations, we undertook a limited study comparing the performance of density functional theory to standard Hartree-Fock methods for calculating proton affinities. We were unsure what to expect, since DFT calculations involving hydrogen are known to be problematic ;338'339 this is due to the treatment of the electron density as a homogeneous electron gas in the local density approximation (see below). Results obtained by ourselves (Chapts. VIE and IX) and others 340 demonstrate that corrections to this approximation yield significant improvements. To our knowledge these are the first reported calculations of proton affinities using DFT.

8.2 The density functional theory A brief introduction to density functional theory is included here. The reader is referred to references 317, 341, and 342 for further details. Density functional theory (DFT) is a special, rigorously proven 343 theory applicable to the ground state of atoms and molecules. Many properties of interest to organic chemists are ground state properties.

The theory has been used by physicists to study problems in solid state and materials sciences, such as electron band structure in conducting solids ,344 and, more recently, the properties of glasses .339 DFT, like Hartree-Fock theory, is ab initio in the sense that only fundamental constants (masses, charges) and coordinates are required as input; both theories do, however, require similar approximations in practical computational methods. DFT is of interest primarily because the computational demands are at least one or two orders of magnitude less than for the Hartree-Fock (HF) approach. In HF theory the size of the calculation scales in principle as - N4, where N is the number of basis functions or heavy atoms; in practical applications, where some integrals are eliminated by distance cutoffs, it may scale more like ~ N 3,345 but post-HF methods 198 which recover even some of the electron correlation energy (see below) scale at least as - N5. By contrast, DFT calculations scale in principle as N3, and in practice sometimes

less depending on the implementation .317,344 DFT also explicitly includes terms for electron correlation (see below), so the computational advantage may be as large as -

N2- N3. Relativistic effects are readily included in DFT for calculations on metals,

whereas this is difficult with HF theory .317,339 Density functional theory has been applied to small organic molecules only recently, and commercially-available programs have appeared only in the last 2-3 years. A serious limitation in applying the method is thus the much smaller base of experience compared with traditional HF methods; hence for those calculations which cannot be compared to experiment, the best recourse is to make comparisons with high-level HF methods where the performance has been characterized for related systems. Unfortunately, this is not always possible because the size of the computations using HF theory may be prohibitive.

A comparison 341 between the Hartree-Fock equations and the Kohn-Sham equations (8.1

- 8 .S) 343 illustrates their similarity and fundamental differences. In HF theory, the energy is determined by the wave function, y, and is a Junctional of it (see discussion

Chapt. VII); in contrast, in the Kohn-Sham treatment the energy is a functional of the electron density (eqn. 8.1). The terms in the Fock one-electron Hamiltonian operator are similar to those in the Kohn-Sham equations, except for the inclusion of the exchange-correlation operator Exc in the latter (eqn. 8.2). In DFT, minimization of the energy based on the variational principle (Chapter VII) adjusts the electron density, p(r), rather than the wave function, y, as in HF theory (eqn. 8.4). Besides the computational advantages this provides, there are conceptual advantages ,317 since the electron density is a physical observable (i.e. it is just what is measured in, e.g., X-ray diffraction experiments). The orbital energies e, are then computed in a comparable manner (eqn. 199

Kohn-Sham Hartree-Fock

(8.1) E = E[p>R a ] E = E[vt RoJ

(8.2) E = T[p] + U[p] + E ^ p ] E • J\K*{Ei[ti + v(*)] + E i.jd/ry)}^ dr

1 (8.3) p(r) = Eocc U r)\2

(8.4) 0E/0p = 0 SE/dy = 0

(8.5) [- 1/ 2V2 + Vc (r) + pix(r)](t>i [-V2V2 + Vc (r) + Mxc(r)] * = -€ifo

8.5). The potential energy term U in eqn. 8.2 is expanded in eqn. 8 .6 , so that the similarity to eqns. 7.4-7.5 is evident (the nuclear repulsion energy Vm is generally included in U, but is omitted in eqn. 8 .6 for analogy to the HF equation above). The total energy W is then given by eqn. 8.7 (which is the same as eqn. 7.6) in both cases.

(8 .6 )U = -Ea {p(r,)[Za/(Ra-r1)]} + '/jlpfrJpO-HI/Ciyrj)])

(8.7)W -E + Vm

The key difference in the theories is most evident in equation 8.3: it is the electron density which is the fundamental variable, rather than the wave function, and it is the density which is optimized in the self consistent field procedure.

There are any number of unique wave functions, not necessarily from ground states, which may satisfy the extremum condition of equation 8.4. The constrained-search 200

definition for the density functional 346 (primarily attributed to Levy317) resolves this

issue (eqn. 8 .8 ).

(8.8) FhkIPq] =JV0*(T + VeJVodT = min(y^poJV*(T +Vee)f dr

This equation indicates that in a search over all antisymmetric wave functions y that give

the ground state density p0 (as in eqn. 8.3), f h k (the Hohenberg-Kohn functional) obtains the minimum expectation value of the kinetic and electronic potential energy,

which results when y = y0. The proof is outlined by Parr .317

Equations 8.1 - 8.5 are exact, however the appropriate mathematical form of the correlation-exchange operator Exc is not known, and much of current theoretical research in DFT deals with improving the representation of this operator. A reasonable

approximation for many problems is the local density approximation (LDA, 317 eqn. 8.9),

which is based on the known behavior of a uniform electron gas .347

(8.9) Exc[p]»Jp(r)exc[p(r)]dr

A number of analytical forms have been developed. The program DMol (which we used

in these studies) incorporates the potential of von Barth and Hedin , 348 whereas the

program DGauss uses that of Vosko, Wilk, and Nusair .349 A complete description of

these potential functions is beyond the scope of this thesis.

In certain situations, the local density approximation breaks down. This appears to be a particular problem for hydrogen, which might be expected due to its sparse electron density. Corrections to the correlation-exchange operator for the inhomogeneity of the electron gas have been proposed by Becke ,350,351,352 Perdew,353 Stoll ,354 and others. 201

Some of these "non-local" or "gradient" corrections have been implemented in the program DGauss, apparently to great advantage as we shall see. These corrections are straightforward incorporations of analytical functions into the correlation-exchange operator, and thus do not significantly increase the computational demands of the method.

8.3 Computational details Starting geometries for various molecules were obtained from the literature (acetic acid, protonated acetic acid, formic acid, protonated formic acid, methanol, protonated methanol, ethanol, and protonated ethanol ) 355 or were obtained with molecular mechanics (formate, acetate, methylammonium, and methylamine) using SYBYL modeling software (version 5.3, Tripos Associates, St. Louis, Missouri). All

Hartree-Fock and HF/MP2 356,357 calculations were performed with GAUSSIAN90358 running on a CRAY Y-MP8/864 supercomputer. Geometry optimizations at the HF and

MP2 levels were terminated when the largest component of the energy gradient was less than 1.0 x 10*4 Hartree Bohr'1, and made use of the Schlegel or Fletcher-Powell optimization schemes. Density Functional (DF) calculations were performed starting with the same set of coordinates using the program DMol , 359,360 which uses the von

Barth and Hedin local spin density (LSD) potential .348 For these trial calculations we selected the double-numerical atomic basis sets supplied by the program (similar to double-^ in HF methodology) augmented with polarization (d-type) functions. Default parameters were selected, with the following exceptions: for calculations, the MESH parameter was set to FINE; the parameters FASCF and FBSCF (mixing coefficients) were set to 0.2 to achieve satisfactory SCF convergence; the SMEAR parameter (charge smearing at the Fermi level during SCF optimization) was set to 0.02 to enhance convergence. Geometry optimizations were terminated when the largest component of 202

the energy gradient was less than 1.0 x 10 "4 Hartree Bohr1, and made use of

Broyden-Goldfarb-Fletcher-Powell (Newton-Raphson) optimization scheme. The Hessian matrices required for derivation of vibrational frequencies were calculated by

the method of finite differences, with nuclear displacements of 0.01 A and single-point differencing.

Calculations were also performed starting with these same coordinate sets with the program DGauss (these calculations were performed for us by J.W.A. at Cray

Research361). DGauss incorporates the LSD potential of Vosko et. al .349 The

LSD-optimized Gaussian basis set DG-2 345’362’363 was used, which is a valence double-? basis with polarization functions (DZVPP). This set has a pattern (721/51/1), shorthand

for contracted basis sets with three s, two p and one d contractions. The s-type contractions have 7, 2, and 1 primitive Gaussian-type orbitals; similarly, the p-type contractions contain 5 and 1 primitive Gaussians, while the d functions contain only one Gaussian primitive. Full geometry optimizations were performed in the Local Density Approximation with default DGauss options, i.e. triple zeta fitting for the electron

density and default grid selection. Gradient corrections to the correlation-exchange

operator were made according to the method of Becke and Perdew 362 in order to

calculate corrected energies; however, calculation of energy gradients (as eqn. 7.20) for

gradient-corrected densities has not yet been implemented, so that these corrections can only be applied to fixed molecular geometries (i.e., geometry optimizations were performed under the LSD approximation). 203 8.4 Results and discussion 8.4.1 Molecular geometries In comparing the proton affinities calculated using different methods, it is important to critically examine the optimized geometries. For some of the molecules we studied, experimental gas phase geometries are available (Tables 18-22). Bond lengths and angles calculated with the HF/MP2 procedure seem to agree best overall with

experimental values. The DMol implementation of DFT generally gives better

agreement than the HF/MP2 treatment for C-0 and C=0 bond lengths, while giving modest agreement for C-C and C-N bonds and significantly overestimating C-H, N-H, and O-H bond lengths. DGauss similarly overestimates R-H bond lengths, so the effect is evidentally due to (known) deficiencies in the description of hydrogen by the local

density approximation (c.f. refs. 359 and 364, but see also ref. 345 for a case where C-H

bond lengths are in good agreement with experiment). Bond angles calculated by

various methods are reasonably consistent, though not always in agreement with experiment; where there are significant differences between methods, the DFT results are usually close to those obtained by HF/MP2 calculations (note in particular the C-O-H

bond angles in neutral and protonated formic acid, acetic acid, ethanol and methanol

(Tables 18-21). There are several notable instances of disagreement in bond angles calculated by DGAUSS and DMol, namely the H-O-H angles in protonated methanol

and ethanol, and the H-N-H angle in methylammonium. There are also significant

differences in the torsion angles obtained by various methods for ethanol and protonated ethanol (Table 21). The overestimation of N-H and O-H bond lengths by DFT should

certainly affect the calculated proton affinities, and we are currently studying the nature of this effect. Table 18. Selected geometric parameters for formate, formic acid, and protonated formic acid. O

O

Method Rc-o Rc-H Z.O-C-0

HF DZPP 1.236 1.122 130.8 MP2 DZPP 1.267 1.131 130.7 DGAUSS 1.261 1.154 130.4 DMol 1.258 1.149 130.5

- Exptla 1.24-1 .25 126.3 O-H H-

Method Rc-o RC-O Rc -h Ro-h Z.OCOZ.COH

HF DZPP 1.184 1.323 1.086 0.950 124.9 109.2 MP2 DZPP 1.215 1.353 1.095 0.972 125.2 106 .6 DGAUSS 1.212 1.342 1.114 0 .987 124 .9 106 .4 DMol 1.205 1. 341 1.103 0 .996 124.8 106 .0

Exptlb 1.204 1.342 1.097 0.972 106 .3

He \ 05 / H 4 - Cl + \ 02 H3

Z.OCO Method RC-02 Rc -h R02-H3 R05-H6 Z.C02H3 Z.C05H6

HF DZPP 1.247 1.080 0.960 0.956 120.8 116.0 116 .3 MP2 DZPP 1.271 1.088 0 .982 0 .978 120 .6 113.8 114 . 0 DGAUSS 1.267 1.107 0.997 0.991 120.2 114 .1 114 .7 DMol 1.262 1.094 1.004 0.999 120.7 114.4 114.0 aExptl. results as in T. Clark, J. Chandrasekhar, G. Spitznagel P.von Rague Schleyer, J. Comp. Chem., 4 (1983) 294-301. bExptl. results as in von Nagy-Felsobuki and Kimura, J. Phys. Chem., vol. 94 (1990) 8041-8044. 205 Table 19. Selected geometric parameters for acetate, acetic acid, and protonated acetic acid. Hs 02

H4~C3— Ci - i \ He 07

Method Rc-c Rc-o Rc-H4 Z.OCO Z.H5CH6

HF DZPP 1. 549 1.238 1.090 129 .2 110 . 5 MP2 DZPP 1.560 1.267 1.095 129 .5 110.5 DGAUSS 1.550 1.263 1.108 129 .2 110 .9 DMol 1.529 1.262 1.104 128.9 110 .6 Exptl® 1.53 1.25 - 125.7 - He 0 2 — H3 1 / H5— C4— Cl i ^ H7 08 n ^ O n o Method Rc-c Rc-o Rc-o RC-H5 r o -h Z.OCO 1 Z.COH

HF DZPP 1.503 1.189 1. 333 1.084 0.949 122.2 125.7 108 . 5 MP2 DZPP 1.504 1.220 1.364 1.090 0.97 2 122.5 126.5 105 .7 DGAUSS 1.491 1.218 1.352 1.105 0.987 121.9 125.3 104 .7 DMol 1.47 3 1.212 1.353 1.102 0.994 121.8 125.6 104 . 8 Exptlb 1.494 1.209 1.357 1. 090 0.970 121. 8 126.2 105.9

H9 \ H7 08 / Hs— Ca — Ci + \ He 0 2 H3

Method RC-02 Rc-c R C-H5 R02-H3 Z.OCO L e e 02 Z.C02H3 Z.C08H9

HF DZPP 1.258 1.488 1. 07 8 0.958 117 .7 119 . 0 115.0 115.9 MP2 DZPP 1.281 1.482 1. 085 0 .981 116 .7 118. 8 112 .7 113.5 DGAUSS 1.282 1.460 1 . Ill 0.994 116.3 118.9 112.8 114.2 DMol 1.277 1.441 1.107 1.000 116 . 3 118. 8 112 .6 113.4

Exptlc 1. 265 1.480 118 . 3 118.2 aExptl. results as in Clark et. a l ., J. Comp. Chem., 4 (1983) 294. bExptl. results as in von Nagy-Felsobuki and Kimura, J. Phys. Chem., vol. 94 (1990) 8041. cExptl. results as in Hehre et. al., Ab Initio Molecular Orbital Theory (John Wiley, New York, 1986) p. 214. 206 Table 20. Selected geometric parameters for methanol ana protonated methanol. H5 \ C — 02 \

Method Rc-o RC-H5 Ro-h Z.COH Z.H4CH6 Z.H5COH

HF DZPP 1.402 1.082 0.943 110.0 108.7 179 .9 MP2 DZPP 1.424 1. 089 0.963 108.0 108.8 180.0 DGAUSS 1.411 1.105 0.975 108.6 108.3 179 .9 DMol 1.419 1. 099 0.985 108.0 109 . 0 180.0

Exptl8 1.421 1. 094 0.963 108.0

H5 \ H4 Ci-— 02 / 1 + H7 I H6

Method Z.COHZ.HOH AH5COH4 Rc-o RC-H5 Ro-h

HF DZPP 1.510 1. 077 0 .957 116 .7 111.7 68.0 MP2 DZPP 1.514 1. 085 0 .977 114.0 109 .6 63.5 DGAUSS 1.488 1.100 0 .990 114.4 110.7 64.6 DMol 1.496 1.094 1.000 112.4 108.4 61.2

aExptl. results as in von Nagy-Felsobuki and Kimura, J . Phys. Chem., 94 (1990) 8041.

8.4.2 Proton affinities The proton affinity of a molecule B is defined as the negative of the molar enthalpy change at 298.15°K (-A H°) for the reaction B + H+ —* BH+, or in the case of anions, A' + H+ —> AH. To calculate these values from theory for gas phase reactions, we may in most cases obtain adequate results assuming ideal gas behavior: A / / 0 = AE° - RT. The internal energy E(T) of one mole of gas consisting of nonlinear polyatomic molecules is given by eqn. 9.1 ,365 where n is the number of atoms in the molecule, ZPE (eqn. 9.2) is 207

Table 21. Selected geometric parameters for ethanol and protonated ethanol. H7 H3 He \ C i— C2 H’" J X04 H® uf

Z.CCO Z.COH Z.H7CCO Z.CCOH Method Rc-c Rc-o Ro-h

HF DZPP 1.522 1.405 0.943 112.7 109.9 177.5 61.1 MP2 DZPP 1.522 1.429 0.965 112.7 107.5 176.5 59.4 DGAUSS 1.511 1.415 0.977 112.9 107.9 177.4 55.4 DMol 1.498 1.422 0.986 112.9 107.7 177.4 55.0

Exptl® 1.530 1.425 0.945 108.5

H6 Hs H 3

\ H 3 f H 7 Ci-— C i H io j 0 + 4- He H 9 H 5 C l

Method Rc-c Rc-O Ro-h Z.COH5 Z.CCO Z.HOH

HF DZPP 1.506 1.545 0 .956 116.5 107.4 111.4 MP2 DZPP 1.503 1.546 0 .977 113.1 107.1 109.4 DGAUSS 1.486 1.525 0.991 112.5 108.0 110.3 DMol 1. 474 1.532 1 . 000 110.3 107.0 108.6

Method Z.H8C1C204 Z.C1C204H5 Z.C1C204H6

HF DZPP 177 .1 52.4 178.5 MP2 DZPP 177 .1 52.4 17 8.5 DGAUSS -179.0 59.9 -173.2 DMol 177 .6 46.3 167.9 aExptl. results as in von Nagy-Felsbuki and Kimura, J. Phys. Chem., vol. 94 (1990) 8041. 208 Table 22. Selected geometric parameters for methylamine and methylammonium. He H3 \ H4 C2- Ni' Hfl Hi

Method Z.CNHZ.H6CN Z.HNH Z.H6CNH4 nc-N Kc-IK r n -h

HF DZPP 1.455 1.091 1 . 000 110.9 114.6 107 .4 59 .6 MP2 DZPP 1.467 1.098 1. 015 109 .8 115.1 106 .2 58 .3 DGAUSS 1.449 1.114 1. 024 111.0 116 . 2 107 .2 59 .6 DMol 1.453 1.109 1.029 110 .3 115.4 106 .7 58.3

Exptl8 1.471 1.099 1.010 110.5 119.7

Hi Hs \ C2- N l Hej \ H3 Hs

Method Kc-N Kc-H r n -h Z.CNHZ.HCN Z.HNH Z.H7CNH3

HF DZPP 1.507 1.079 1.010 111.3 108.2 107 .6 180. 0 MP2 DZPP 1.510 1.086 1.024 111.4 108 .1 107 .5 180.0 DGAUSS 1.490 1.100 1.037 111 .6 109 . 0 107 . 3 180.0 DMol 1.488 1.094 1.038 111 .7 108.7 110.3 179 .7 aExptl. results as in Hehre et. al., Ab Initio Molecular Orbital Theory (John Wiley, New York, 1986), p. 148. 209

(9.1) Ea(T) = %RT + % RT + ZPE + E vib(T) + Eelec H H L- r - J 1------,------1 Efrons Erot Etfi,

3n-6 N/iv,- (9.2) ZPE = 2 i-1 3n-6 „ NAv,(eN/,vf/R7) (9.3) £ v,b(T) (i

the zero point energy and E ^ iT ) is a temperature-dependent portion of vibrational energy which results from population of vibrational levels according to the Boltzmann distribution. N = 6.0221367 * 1023 molecules/mole, h is Planck’s constant, R is the ideal gas constant, v,- are calculated vibrational frequencies, and T is the absolute temperature. The change in energy occuring during protonation of 1 mole of gas at 298.15 K, AE° -

£ ah +(298.15 K) - £ A(298.15 K) - £H+(298.15 K), will then consist of the following components: AE°rot- the change of rotational energy resulting from the protonation. Since a proton does not contribute to rotational energy, this term is nonzero only if the parent molecule is linear and the protonated molecule is nonlinear (or, conceivably, vice-versa). AE°trans - the change in energy associated with a change in translational degrees of freedom. Joining the separate species B and H+ to form a single molecule results in the loss of three degrees of translational freedom, A£°frartiS = - 3/2£T, »0.889 kcal/mol at 298.15 K.

AE°vib - the change in energy associated with internal vibrations within products and substrates. Only the change in zero point energy (A ZPE) is significant, while AE'yjt, is much less than 1 kcal/mol for the molecules we studied, and is included here only for completeness. AEPeiec - the change in the electronic energy during reaction. In our case it is the difference between ground state energies (electronic and nuclear) taken from the quantum calculations with full geometry optimizations for the protonated and parent molecule (the electronic energy of a proton is zero). 210

The working equation for the proton affinity is equation 9.4.

(9.4) PA = -AE°elec - AZPE - A + 5/2RT

Since real vibrational frequencies are anharmonic while those obtained in calculations are harmonic, empirical corrections (average scaling factors) are sometimes applied to the frequencies .366 We did not scale the frequencies in our calculations, as applying the standard scaling factors gave poorer agreement with experimental frequencies. The vibrational frequencies obtained using density functional theory were generally in better agreement with experiment than those obtained with HF or HF/MP2 calculations, in accord with the results of other workers (see ref. 364 and Tables 24 and 25 below). The overestimation of R-H bond lengths discussed above would be expected to cause systematic errors in these frequencies (perhaps giving fortuitously better agreement with experimental values), and a detailed examination is in progress.

The total energies, zero point energies, and £ ’vi &298 values are given in Tables 23a-c. Values for ZPE and are not given for the DGauss NLSD (gradient-corrected) results because the frequencies were calculated using the local spin density approximation (LSD or LDA). The values of EeUc are given mainly for reference rather than for direct comparison, since very small differences in the computational method can have large effects on this value (note that 0.1 hartree 13 63 kcal/mol). The zero point energies are quite consistent between the two implementations of density functional theory, while the DFT results are generally lower than those obtained by MP2/HF theory using the double-^ basis set, reflecting the differences in the frequency calculations discussed above. 211

Table 23a. Electronic and vibrational energies.

Molecule: CH3COO' CH3COOH

p 29t Method E elec ZPE Evib298 E elec ZPE Evib

HF/DZPP -227.288105 32.57 0.94 -227.871825 41.94 0.96 MP2/DZPP -227.969663 31.01 1.03 -228.547581 39 .70 1.09 DMol DFT -226.87 8791 29 .21 1.00 -227.431587 37 .64 1.05 DGauss LSD -226.797767 29 .47 -227.357678 37 .67 1.17 DGauss NLSD -228.583853 -229.151437

E°ejec in hartree , ZPE and Eyft, in kcal/mol.

Molecule: CH3COOH2+ HCOCf

p 298 p 298 Method E°elec ZPE ^ b E elec ZPE Evib

HF/DZPP -228.192110 50 .32 1.13 -188.236185 14 . 03 0.06 KP2/DZPP -228.857807 47 .85 1.18 -188.764907 13 . 04 0.09 DMol DFT -227 .730382 44 .79 1.32 -187.903715 12 .13 0.11 DGauss LSD -227 .662718 45.26 1.22 -187.838977 12.57 0.10 DGauss NLSD -229.462849 -189.256199

E°elec in hartree, ZPE and E^j, in kcal/mol.

We expended considerable effort examining the effects of various parameters in DMol on calculated vibrational frequencies. We are still analyzing the results, which will be reported in another paper, but representative data are given in Tables 24 and 25. The importance of a properly-converged geometry is evident in Table 24, where an imaginary vibrational frequency is obtained for methanol when DMol is used to calculate vibrational frequencies for a geometry optimized using DGauss. We have not assigned any of these frequencies to vibrational modes, pending installation of a program to accomplish this task.

The proton affinities are overestimated for all molecules by the HF treatment with the DZPP basis set (double-^ with polarization functions on both heavy atoms and 212

Table 23b. Electronic and vibrational energies.

Molecule: HCOOH HCOOH2+

Method E elec ZPE Evib298 E elec ZPE Evib29®

HF/DZPP -188.814337 23.29 0 .17 -189.118947 31.92 0.23 MP2/DZPP -189.338509 21.65 0.21 -189.633006 30. 03 0.27 DMol DFT -188.449596 20.39 0.24 -188.727812 28.18 0.33 DGauss LSD -188.393595 20.60 0.23 -188.679666 28.45 0.31 DGauss NLSD -189.819241 -190.1127 47

E °elec hartree, ZPE and in kcal/mol.

Molecule: CH3OH CH3O H 2+

Method E°elec ZPE E vib298 E elec ZPE Evib298

HF/DZPP -115.074058 34.65 0 .28 -115.382222 43 . 08 0.49 MP2/DZPP -115.431888 33.18 0.29 -115.735752 41. 51 0.46 DMol DFT -114.862138 30.96 0.34 -115.145634 39.25 0.48 DGauss LSD -114.820143 31.19 0.31 -115.110564 39 . 27 0.46 DGauss NLSD -115.753584 -116.050971

E °elec in hartree, ZPE and E^j, in kcal/mol. hydrogen); the overestimation is especially large for the two carboxylate anions and methylamine, in accord with previous work (c.f. Table 17). The best HF results are for protonation of methanol and ethanol. Second order Moller-Plesset perturbation theory

(MP2) improves the calculation for all molecules studied. Density functional theory (DFT) using the local density approximation (or local spin density, LSD) substantially underestimates the experimental results. There is a sizeable discrepancy between DMol and DGauss for all the molecules studied; such discrepancies were not observed in our studies of hydrogen bonding (Chapter IX), and in this regard the differences in R-H bond lengths obtained with DGauss and DMol (Tables 18-22) may be an important factor.

Adding gradient corrections using the Becke-Perdew method 362 (labelled as NLSD = non-local spin density in Table 26) improves the esimate considerably; for the anions and methylamine, the estimates are a substantial improvement over any calculations 213

Table 23c. Electronic and vibrational energies.

Molecule: CH3CH2OH CH3CH2OH2+ v 298 Method E °clec ZPE E vib298 E °elec ZPE vib

HF/DZPP -154.122480 53.88 0.81 -154.438663 62 . 02 1.11 MP2/DZPP -154.634459 51.79 0.85 -154.945575 59 .78 1.10 DMol DFT -153.837216 48.56 0.94 -154.130723 56 .23 1.32 DGauss LSD -153.777156 48.70 0.94 -154.076033 56 .34 1.16 DGauss NLSD -155.078652 -155.385157

E °elec hartree, ZPE and in kcal/mol.

Molecule: CH3NH2 CH3NH3+

p 298 p 298 Method E elec ZPE Evib E °elec ZPE E vib

HF/DZPP -95.240290 43.05 0.32 -95.605062 53.30 0.35 MP2/DZPP -95.590232 41.47 0.34 -95.953177 51.20 0.37 DMol DFT -95.068884 38.85 0.39 -95.410276 48.43 0.37 DGauss LSD -95.029013 39 . 03 0.39 -95.377146 48.39 0.38 DGauss NLSD -95.881481 -96.237388

E °elec in hartree, ZPE and E^b in kcal/mol. reported in the literature to date. It is likely the improved agreement with experiment when gradient corrections were applied results from improvements in the description of the electron density for hydrogen, and hence improved estimates of the energy of the protonated species (we are still analyzing these results). The proton affinities for oxygen bases (i.e. acetic and formic acids, methanol and ethanol) are underestimated by 2-3 kcal/mol, which is however well within experimental error. 214

Table 24. Vibrational frequency calculations for methanol using DFT (cm*1).

DMol converged to Geometry: DGAUSS <0.001 <0.0002 DGAUSS hartree bohr'* Hr ** Method: DMol DMol DMol DGAUSS Exper. '

-137 283 282 329 999 1014 1014 1029 1090 1064 1068 1100 1033 (C-0 str.) 1116 1108 1111 1120 1060 (ch3 rock) 1302 1315 1317 1311 1345 (O-H bend) 1413 1400 1397 1411 1455 (CH3 s-deform) 1440 1440 1440 1422 1477 (ch3 d-deform) 1454 1452 1455 1438 1477 (ch3 d-deform) 2871 2919 2919 2904 2844 (ch3 sym str) 2917 2973 2975 2963 2960 (ch3 d-str) 2978 3040 3033 3044 3000 (ch3 d-str) 3767 3653 3646 3747 3681 (O-H str)

*Anharmonic frequencies. The assignment of calculated frequencies to vibrational modes for comparison with experimental values has not been completed. 215 Table 25. Vibrational frequencies calculated for methylamine using DFT (cm'1): effect of displacement distance and differencing method in DMol. ______★ ith Method: DMol DGAUSS E x p e r . * N D I F F :a 1 1 1 2 2 2 VIBDIF:b 0 . 005 0 . 01 0.02 0. 005 0.01 0 . 02

289 282 265 301 299 292 311 268 806 820 817 812 819 816 780 780 939 937 936 938 934 934 935 1044 1051 1066 1061 1059 1069 1062 1086 1130 1124 1124 1128 1122 1124 1125 1128 1195 1295 1294 1294 1296 1294 1295 1280 1419 1399 1389 1395 1405 1394 1400 1387 1430 1462 1452 1449 1454 1449 1447 1424 1473 1466 1464 1460 1464 1462 1460 1449 1485 1610 1612 1611 1610 1610 1612 1588 1623 2888 2886 2870 2889 2890 2882 2893 2820 2986 2983 2976 2985 2985 2983 3004 2961 3020 3019 3021 3021 3022 3026 3047 2985 3373 3372 3371 3372 3373 3373 3448 3361 3451 34 52 3459 3449 3448 3449 3538 3427

*Anharmonic frequencies. **The assignment of calculated frequencies to vibrational modes has not been made. aNumber of nuclear displacements used to calculate energy gradients. N u c l e a r displacement, in angstroms.

Table 26. Calculated vs. experimental proton affinities for various molecules.

Molecule: CH3C O O ‘ H C OO" CH3NH2 CH3COOH HCOOH CH3OH ch3ch2oh Method Proton Affinity (kcal/mol)a

HF/DZPP 357 .5 354 . 0 219.2 193.0 183 . 0 185.3 190 .6 MP2/DZPP 354 .5 351.8 218 .6 187 . 0 176 .9 182. 8 187 .6 DMol DFT 339 .0 334 .7 205.3 180.7 167 . 3 170.1 176 .7 DGauss LSD J 343 .1 340.5 209 .7 184 .4 172.2 174 . 6 180 . 3 DGauss NLSD 347 .9 345 .7 214.6 188.4 176 .8 179 .0 185.1

EXPER. 348.5b 345.2b 214.1° 190.2° 178. 8° 181.9° 188.3°

a -

It was merely a mathematical demonstration. It made a point of interest to mathematicians, but there was no thought in my mind of its being useful in any way. -Hari Seldon, in Prelude to Foundation, by Isaac Asimov (Bantam, 1988) 9.1 Introduction Hydrogen bonding plays an essential role in biochemical recognition and in stabilizing

molecular arrangements in biological systems. The proximity of amino and carboxyl groups is believed to be important for the binding of many ligands with receptors. The discussion in Chapter VI focussed on dopamine analogs in which the amine moeity is replaced by other functional groups. Evidence was presented that the amine group in

dopamine interacts with a carboxyl group on the receptor protein, and that it does so in the protonated (charged) form. To understand the trends in activity of these analogs and potentially predict new bioisosteric replacements for the amine group, we undertook to study the interactions of these various groups with a carboxyl group using quantum mechanical methods. Semi-empirical approaches are known to perform rather poorly in such systems ,309,312 while traditional ab initio methods seem to provide an adequate description of hydrogen bonding when large basis sets are used and electron correlation is taken into account .314,315,316,367,368

216 217 The potential merits of the Density Functional Theory (DFT) approach for descriptions of ground state chemical phenomena were discussed in the previous chapter. Systematic studies of hydrogen bonding using DFT have not been performed, and the treatment of

hydrogen atoms by DFT is known to be problematic .338 DFT will almost surely be extensively applied to hydrogen bonding in the near future given the advent of widely-available computer programs and improvements in the formulation of the theory. We chose to study hydrogen bonding in the formic acid:methylamine and formate: methylammonium complexes due to our interest in dopamine. In choosing such a

system, for which no experimental results are available, we are in a sense trying to run before learning to walk; on the other hand, we can estimate the reliability of our results from 1) the known behavior of extended Hartree-Fock methods for treating hydrogen

bonding (see above), and 2 ) the performance of various models in the study of proton affinities (Chapter VIII). Given the critical role of hydrogen bonding and proton transfers

in biological systems ,369 developing an approach which can reliably treat various hydrogen bonds using reasonable computational effort and without dependence on

empirical parameterization is highly desirable.

The amine:carboxyl system was studied previously using ab initio methods at the

Hartree-Fock level by Hadzi and co-workers 335,370 and by Sapse and Russell .371 They found that the lowest-energy form of the isolated C0 2H...NH2 complex is the neutral one (i.e. without proton transfer from carboxyl to amino group), in agreement with infrared studies of trimethylamine and acetic acid complexes .372 This is in contrast to guanadinium:carboxylate interactions, which have been extensively studied by ab initio methods; the energetically preferred form seems to be a bifurcated complex wherein the proton resides with the guanadinium moiety .371,373,374 The reported binding energy for a neutral bifurcated hydrogen bond in an acetic acid:methylamine complex was -11.7 and 218

-7.5 kcal/mol for 6-31G and 6-31G* basis sets .371 For a near-linear geometry the binding energy in formic acid:methylamine was -17.6, -14.8, -14.0, and -11.8 kcal/mol for

STO-3G, 4-31G, 6-31G, and 4-31G** basis sets, and the estimated basis set superposition errors (see discussion below) for these values were 6 .0 , 2 .2 , 2 .0 and 2 .0 kcal/mol, respectively .335

9.2 Computational methods 9.2.1 General Starting geometries of formate (F), methylammonium (M+), formic acid (FA) and methylamine (MA) were obtained with molecular mechanics using SYBYL modeling software (version 5.3, Tripos Associates, St. Louis, Missouri), and then optimized at the

Hartree-Fock level using the program CADPAC 323 and various basis sets as described below. All Hartree-Fock and HF/MP2 356'357 calculations were performed either with

CADPAC323 or GAUSSIAN90358 running on a CRAY Y-MP8/864 supercomputer.

Geometry optimizations at the HF and MP2 levels were terminated when the largest component of the energy gradient was less than 1.0 * 10 "4 Hartree Bohr'1.

Density Functional (DF) calculations were performed using the programs DMol 359,360 and

DGauss (the DGauss calculations were performed by J.W.A. at Cray Research ,361 since the program has only recently been released for sale). DMol generates high-quality atomic basis sets for use in calculations, and in all cases we used double numerical basis sets (similar to double-^ in HF methodology) augmented with polarization (d-type) functions as supplied by the program. 219 9.2.2 FAMA Complex 1

From formate anion and methylammonium cation , each at a geometry optimized using the

6-31G** basis set with the program CADPAC, a complex was constructed as follows

(Figure 58A): the hydrogen bond in 03-H5-N6 was made linear (180.0°), with H5 and N 6

in the plane of formate (established by 04-C1-03); the length of the hydrogen bond was

(somewhat arbitrarily) set to 1.6 A; the angle C1-03-H5 was set to 120°; N 6 and C7 were placed in the 03-C1-04 plane in such a way that C7 was remote fiom 04; hence Cl, H2,

04, and 03 in formate were coplanar with H5, N 6 , and C7 in methylammonium. The

geometry of this complex was then optimized by the BFGS (Broyden-Fletcher-Goldfarb-Shanno) algorithm, a Newton-Raphson optimization scheme,

using a 3-21G basis set 375,376 augmented with polarization functions only on atoms Cl,

03, 04, N6 , H5, H8 , H9, i.e. 6 d-type functions with exponent 0.8 for C, N and O and a single p-type function with exponent 1.1 for H. Throughout the rest of this chapter, this basis set collection will be referred to as 3-21G*(*). This approach was feit to give a reasonable description of the valence shells on the "business end" of each partner in the

complexes, while offering some economy for the geometry optimizations compared with

basis sets having a larger number of gaussian functions .377 Geometry optimizations were

performed both in cartesian (X,Y,Z) coordinates and in internal coordinates (for the latter,

bond lengths and angles are the optimized variables).

To evaluate the dependence of the interaction energy on the distance H5—N 6 , the

geometry of the 3-21G*(*)-optimized complex (Figures 58C,D) was varied by setting

this distance to 0.8, 1.0, 1.2, 1.4, 1.6, 1.833 (the optimal value from 3-21G*(*)

calculations), 2 .0 , 2 .2 , 2.5, 3.0, 4.0, 6 .0 , 8 .0 , and 1 2 .0 A, while keeping all other bond lengths and valence and torsional angles constant (see Table 27). The total energy for these rigid molecular arrangements was obtained by HF calculations with 3-21G*(*) and 220

double-C (DZ) basis sets 378 with polarization functions on all atoms (single p-type

functions with exponent 1.0 for H atoms and 6 d-type functons with exponents 0.8 for C

and N and 0.9 for O, respectively). The latter set of basis functions will be referred to as DZPP throughout the remainder of this chapter. Interaction energies ( Eb) were then calculated for each distance point as equation 9.1, where EFM denotes the total energy of

(9.1) Eh = E fm -E f -E m the formic acid:methylamine (FAMA) pair, and EFan& EM are the energies of formic acid and methylamine, respectively, calculated for their geometries in the 3-21G*(*)-optimized complex. No geometry optimizations were performed in this case. The basis set superposition error ( BSSE) 379»380 in the interaction energies for each distance point was estimated by the counterpoise method 380 (see discussion below). In this method, seperate energy calculations were performed at each distance point, but with the nuclei and electrons of either formic acid or methylamine removed, leaving "ghost atoms" with only the basis functions present (i.e, the energy of each monomer was calculated with the full dimer basis). New interaction energies Eb (adjusted to account for estimated BSSE) are again calculated using equation 9.1, but EF and EM in this case correspond to the energies of formic acid and methylamine with the ghost atoms present. Binding energies (at optimized geometry) were also calculated, but in this case the energies EFM, EF, and EM were those from full geometry optimizations for formic acid, methylamine, and the complex, respectively. In this case it is not possible to calculate BSSE, since the geometries of isolated molecules are different from those in the complex.

Density Functional (DF) calculations were performed for the same set of coordinates using the program DMol .359*360 This program uses the von Barth and Hedin local spin density (LSD) potential .348 We used default parameters for calculations, but the 221

integration mesh was modified so that integration points extended out from each center by

15 Bohr, the maximum allowed by the program, and the angular sampling frequency was increased accordingly. The parameters FASCF and FBSCF (mixing coefficients) were set to 0.4, since this resulted in acceptable convergence of the SCF procedure.

Calculations were also performed on these same coordinate sets with the program DGauss.

This program incorporates the LSD potential of Vosfco et. al .349 Two LSD-optimized

Gaussian basis sets362,345 were explored with DGauss in order to test for BSSE effects; both were valence double-^ with polarization functions (DZVPP), but differed in representation of the core orbitals. The first set (DG-1) had a pattern (621/41/1*), shorthand for a basis set of three s, two p and one d contractions. The s-type contractions have 6 , 2, and 1 primitive Gaussian-type orbitals; similarly, the p-part is a contraction of 4 and 1 primitive Gaussians. The second basis set (DG-2) had a pattern (721/51/1) and differed in the number of primitive Gaussians representing core orbitals.

Full geometry optimizations were performed with DMol and DGauss for formic acid, methylamine, and the complex starting from 3-21G*(*) geometries. Default values were used with DMol, except the MESH parameter was set to FINE. The default values of 0.7 and 0.4 for FASCF and FBSCF did not result in satisfactory convergence, and both were reduced to 0.2. DNP sets were again used, and LMAX (number of multipolar functions for fitting the density) was set to 4 for C, N, and O atoms and 3 for H. For DGauss, the

DG-1 basis set and default DGauss options were chosen, i.e. triple zeta fitting set and default grid selection. Geometry optimizations in both cases used the method of analytic derivatives (see discussion Chapter VIII). Neither DGauss nor DMol made use of gradient corrections to the local density approximation (Chapt. VIII) in the geometry optimizations. 222

9.2.3 FAMA Complex 2 A complex was constructed from formic acid and methylamine separately optimized with DMol using the DNP (double numeric with polarization basis set as follows (see atom numbering Figure 63A): N 6 was placed 1.50 A from H5 with an 03-H5-N6 angle of

130.47 degrees (calculated so that HI, C2 and N 6 were co-linear) and in the plane of formic acid (N6-H5-03-C2 torsion 0°). C7 was then placed to make a 130° angle with the H1-C2-N6 line (somewhat arbitrarily chosen) and such that the C7-N6-C2-03 torsion was

90°, i.e. the C2-N6-C7 plane was perpendicular to the plane of formic acid. Finally, H 8 was also projected approximately 90° to the formic acid plane by making the

H5-C7-N6-H8 torsion 122.53°.

The geometry of this complex was then optimized using DMol, again with the DNP basis and the BFGS (Broyden-Fletcher-Goldfarb-Shanno) algorithm. MEDIUM mesh and default LMAX values were chosen, and FASCF and FBSCF were again set to 0.2. Additional geometry optimizations were done with CADPAC using 3-21G*(*) and

DZP(P) basis sets; the DZP(P) basis is similar to the 3-21G*(*) basis in having polarization (d-type) functions on all heavy atoms but only on the RNH 2 and -C02H hydrogens (H5, H 8 , H9). DGauss optimizations (DG-1 basis) were performed starting from both the DZP(P)-optimized geometry and from the DMol-optimized geometries (they were significantly different - see Results and Discussion). Interaction energies were calculated as for FAMA Complex 1. Distance-dependence studies for this complex are still in progress. 223 9.2.4 F* M~ Bifurcated Complex

A bifurcated complex was constructed from 3-21G*(*)-optimized formate anion and methylammonium cation as in Figure 64. H7 and H 8 were set equidistant from 03 and

04 at 1 .6 A by placing NS on the H2-C1 projection at a distance of 3.24 A. The C1-N5-C6 plane was made perpendicular to the formate plane, and the C1-N5-C6 angle was set to 55°, which made H2, Cl, 0 3 ,0 4 , N5, H7, and H 8 all coplanar. The geometry of this complex was then optimized using CADPAC with the 3-21G*(*) basis, at first holding the monomer (formate and methylammonium) geometries constant, and subsequently allowing full relaxation. Optimizations were performed both in internal and cartesian coordinates. We again used the BFGS (Broyden-Fletcher-Goldfarb-Shanno) algorithm. The distance dependence of the interaction was studied by varying the C1*«N5 distance while keeping all other intermolecular and intramolecular constants fixed at the geometries of the 3-21G*(*)-optimized monomers. Additional geometry optimizations were done with DMol using the DNP basis, DGauss with the DG-1 basis, and Gaussian 90 with the 6-311++G** basis set.

9.3 Results and discussion

9.3.1 FAMA Complex 1 Full geometry optimizations with the 3-21G*(*) basis set at the HF level in internal and cartesian coordinates gave essentially the same final geometry (Figure 58C,D, and Table

27). This geometry was, however, substantially different from the starting one in that the proton from the ammonium group migrated to the formate oxygen as expected, resulting in an arrangement similar to that described by HadoSdek and Hadzi .335 The H-bond was not quite linear and assumed an 03-H5-N6 angle of 163° (Figure 58D). If the internal coordinates of the molecules were held fixed, so that the proton could not migrate from the 224

>H12 B

07 m o R12

' 03 H5 .0 - - C7 H ll H2 H2 H8 N6 H10 04 0 4 HS

c D H12 H12 0 3 H5 03 H9 H10 H ll C7 H5 H2 C l m H10 H2 Cl H9 0 4 HS 0 4

Figure 58. Geometry optimization for FAMA Complex 1. A: Starting geometry constructed as described in the text. B: Geometry obtained with internal coordinates for each monomer held fixed. C: Fully-optimized geometry using the 3-21G*(*) basis. Atom numbering for the text discussion is shown. D: A different view of the fully-optimized complex.

N6 to 03, the molecules assumed the configuration shown in Figure 58B. Presumably the distortion of the H-bond far from linearity results from effort to 1) allign the dipole moments, and 2) to achieve some electrostatic and overlap interactions between 04 and the two hydrogens H 8 and H9, although a rigorous analysis has not yet been completed.

This geometry is similar to one identified by Hado££ek and Hadii .363 225

Table 27.3-21G*(*) optimized geometry for FAMA Complex 1. Bond lengths are expressed in A and angles in degrees.

Bond lengths Bond angles Torsion angles

C1-H2 1.094 H2-C1-03 110.42 H2-C1-03-H5 180.00 C l-0 3 1.310 H 2-C 1-04 123.44 04-C1-03-H5 0.00 C l-0 4 1.186 03-C1-04 126.13 C1-03-H5-N6 0.00 03-H 5 0.974 C 1-03-H 5 108.05 03-H5-N6-C7 180.00 H5-N6 1.833 03-H 5-N 6 163.35 03-H5-N6-H8 53.31 N6-C7 1.472 H5-N6-C7 129.74 03-H5-N6-H9 -53.32 N6-H8 1.005 H5-N6-H8 98.98 H5-N6-C7-H10 -58.80 N6-H9 1.006 H5-N6-H9 99.00 H5-N6-C7-H11 180.00 C7-H10 1.083 C7-N6-H8 110.73 H5-N6-C7-H12 58.83 C7-H H 1.087 C7-N6-H9 110.72 H8-N6-C7-H10 63.32 C7-H12 1.083 H8-N6-H9 104.76 H8-N6-C7-H11 -57.86 N6-C7-H10 109.12 H8-N6-C7-H12 179.05 N6-C7-H11 113.90 H9-N6-C7-H10 179.07 N6-C7-H12 109.13 H9-N6-C7-H11 57.89 H10-C7-H11 108.33 H9-N6-C7-H12 -63.30 H10-C7-H12 107.85 H11-C7-H12 108.34 226

The total energy as a function of H5—N 6 distance obtained by HF calculations shows a minimum at 1.8 - 2.0 A, depending on the method of calculation (Table 28, Fig. 59). The interaction energies in Figure 59 were calculated using eqn. 9.1, and the energy zero

corresponds to EF + Em for isolated formic acid and methylamine molecules at their geometries in the 3-21G*(*) optimized complex (not the optimized geometries for the individual molecules). The DZPP and DZPP/MP2 results suggest that the optimum

H5«»N6 distance is about 1.9 A and the corresponding interaction energy is around 10

kcal mol'1. The basis set superposition errors are plotted as a function of H5*»N6 distance in Figure 60. The BSSE with second order Moller-Plesset (MP2) calculations with the

DZPP basis amounts to about 3 kcal mol ' 1 at the minimum energy distance, while for the HF calculations with the same basis set the BSSE is much smaller, less than 1 kcal mol'1.

The BSSE effect is relatively small for the DZPP basis even at close range; however it is much larger with the 3-21G*(*) basis and increases rapidly at short distances with the MP2 treatment. The explanation for the MP2 behavior can be found by examining

energies and coefficients of the HOMO (highest occupied molecular orbital) and LUMO

(lowest uncoccupied molecular orbital) for formic acid and methylamine with and without "ghost" basis functions (the former comprising the full dimer basis set). For both formic

acid and methylamine, the energies of the corresponding HOMO’s (and lower-lying orbitals) calculated with and without the additional basis functions were similar and the coefficients for the ghost functions were very small; however the LUMO’s for both formic acid and methylamine were substantially lowered in energy when ghost functions were included, and the coefficients of the ghost functions were very significant. Since in the second order the excitations to LUMO contribute most significantly to the MP2 correction to the total energy of the molecule, the calculated MP2 energies of formic acid and methylamine with ghost functions are considerably lower than those without, and as a 227 Table 28. Total energies as a function of H5" N6 distance for FAMA Complex 1. Energies in hartrees (1 hartree - 627.51 kcal/mol). The BSSE columns refer to calcu­ lations in which the basis functions and centers were the same as for the "complex" but nuclear charges were placed only on real atoms. The 00 row represents total energies for formic acid and methylamine calculated without the addition of ghost atoms.

d [A | Com plex | Formic (BSSE) | M ethylam ine (BSSE) HF 3-21G (polarisation function* on Cl, 03, 04, N6, H5, and H9) 0.800 -282.2684382 -187.8422810 •94.7409442 1.000 -282.4575668 -187.8410366 -94.7392584 1.200 -282.5359981 -187.8399431 -94.7382043 1.400 -282.5695321 -187.8386934 -94.7376184 1.600 -282.5825630 -187.8373426 -94.7371506 1.833 •282.5860864 • 187.8358511 -94.7363435 2.000 -282.5850421 -187.8349464 -94.7357412 2.200 -282.5822218 -187.8340863 -94.7351583 2.500 -282.5771349 -187.8332053 -94.7341689 3.000 -282.5700004 -187.8325430 -94.7320359 4.000 -282.5643668 -187.8323626 -94.7301396 6.000 -282.5628192 -187.8323597 -94.7300686 8.000 -282.5625473 -187.8323597 -94.7300686 12.000 -282.5624482 -187.8323597 -94.7300686 OQ -187.8323597 -94.7300686 HF DZP ba*i* let with polarisation function* on all atom* 0.800 -283.7527918 -188.8154796 -95.2417233 1.000 -283.9414618 -188.8152362 -95.2415791 1.200 -284.0187932 -188.8150373 •95.2414687 1.400 -284.0518493 -188.8148997 -95.2412830 1.600 -284.0653372 -188.8148007 -95.2410510 1.833 -284.0700179 -188.8147040 -95.2408339 2.000 -284.0700479 -188.8146397 -95.2407190 2.200 -284.0685636 -188.8145676 -95.2406083 2.500 -284.0655114 -188.8144744 •95.2405083 3.000 •284.0611919 -188.8143432 -95.2404710 4.000 -284.0566921 -188.8141444 -95.2402647 6.000 •284.0546907 -188.8141076 -95.2401402 8.000 -284.0543721 •188.8141076 -95.2401401 12.000 •284.0542644 • 188.8141076 -95.2401401 oo -188.8141076 •95.2401401 MP2 DZP baaia *et with polarisation function* on all atom* 0.800 -284.6565704 -189.3480416 -95.6060717 1.000 -284.8413707 -189.3466897 -95.6029886 1.200 -284.9147298 -189.3456710 -95.6003464 1.400 -284.9441343 -189.3449009 -95.5983918 1.600 -284.9547199 -189.3443330 -95.5970781 1.833 -284.9570022 -189.3436335 -95.5961221 2.000 -284.9557851 -189.3435425 -95.5956277 2.200 -284.9531825 -189.3432568 -95.5951443 2.500 -284.9489466 -189.3429380 -95.5946002 3.000 -284.9434190 -189.3425901 -95.5940407 4.000 -284.9379820 -189.3422248 -95.5933728 6.000 -284.9358250 -189.3421672 -95.5931791 8.000 -284.9354910 -189.3421672 -95.5931790 12.000 -284.9353711 -189.3421672 -95.5931790 oo -189.3421672 -95.5931790 Interaction Energy [kcal/mol] -20 -10 DZPP basis sets. Calcuations with Calcuations CADPAC. basis sets. DZPP 3-21G*(*)and is shown for error superposition set basis distance obtained by HF and MP2 methods. The effect of of effect The methods. MP2 and HF by obtained distance Figure 59. Dependence of interaction energy on the H5~N6 H5~N6 the on energy interaction of 59.Dependence Figure 12 3 .. Dsac [A] Distance N...H 4 BSSE MP2 DZP DZP MP2 DZP MP2 BSSE DZP HF F DZP HF *(*) G BSSE 1 2 - 3 HF F -1 «(«*) 3-21G HF 228 229

2 .0 0 0 t

0.000 g+——-i i ♦ i a o - 2.000

—4.000 1 0) O - 6.000

Figure 60. Basis set superposition error vi i. H5™N6 distance in FAMA Complex 1. The optimal 3-21G* [*) distance, 1.83 A, is marked with arrows.

result the calculated interaction energy appears larger if the BSSE correction is not included.

The LSD calculations by DMol and DGauss gave similar results for the optimum of the

interaction energy and the corresponding distance, namely -19 kcal mol ' 1 and 1.6 A,

respectively (Figure 61, Table 29). These values disagree considerably with those

obtained by the HF/MP2 calculations (Figure 62). Gradient corrections to the LSD energies were only available in DGauss, and only for rigid geometries. The Becke-Perdew corrections for two different basis sets gave very similar values for the optimum distance and corresponding interaction energy, i.e. 1.8 A and 12 kcal mol'1. The Becke-Stoll corrections were only calculated with the DG-1 basis set; the energy minimum was - 9 kcal mol *1 near 2.0 A. The Becke-Perdew corrections closely reproduce the MP2 results (Figure 62), whereas the Becke-Stoll corrections result in a Table 29. Results of calculations with different DFT approaches. Total energies (in Hartrees) of the molecular arrangement derived from 3-21G*(*) optimized complex as a function of H5 N 6 distance. Distances are given in Angstroms.

dfr»...jva D M o l 1 D G a u s s 3 D G a u s s 3 LSD 3 LSD Becke-Stoll4 0.800 •283.2732442 -283.1679315 -285.3870636 1.000 •283.4445384 -283.3463971 -285.5740960 1 . 2 0 0 -283.5101007 -283.4159256 -285.6521537 1.400 -283.5345387 -283.4419731 -285.6866404 1.600 -283.5415423 -283.4494973 -285.7011342 1.833 -283.5408469 -283.4486787 -285.7074617 2 . 0 0 0 -283.5376894 -283.4456232 -285.7088223 2 . 2 0 0 -283.5333614 -283.4412407 -285.7078829 2.500 -283.5273672 •283.4349614 -285.7051159 3.000 -283.5205093 -283.4275672 -285.7022478 4.000 •283.5149494 •283.4214682 -285.6977108

6 . 0 0 0 -283.5130614 -283.4196324 -285.6954231 8 . 0 0 0 -283.5128124 -283.4191288 -285.6951405 1 2 . 0 0 0 -283.5127422 -283.4190123 -285.6950408 HCOOH -188.4479132 -188.3915076 -189.8236360 CH 3 NH 3 -95.0648210 -95.0274799 -95.8713837 D G a u s s 3 D G a u s s 6 D G a u s s 6 Becke-Perdew® LSD Becke-Perdew 9 0.800 -285.4167163 -283.1922663 -285.4433321 1.000 -285.5982520 -283.3711538 -285.6251044 1 . 2 0 0 -285.6710343 -283.4403778 -285.6973920 1.400 -285.7009482 •283.4662743 -285.7270324 1.600 -285.7114990 -283.4735820 -285.7374377 1.833 •285.7141582 -283.4724816 -285.7400187 2 . 0 0 0 -285.7134722 -283.4693001 -285.7393342 2 . 2 0 0 -285.7106762 -283.4647722 -285.7366041 2.500 •285.7061422 -283.4585133 -285.7321756 3.000 -285.7019023 -283.4512311 -285.7279409 4.000 -285.6975518 -283.4457046 -285.7238110 6 . 0 0 0 -285.6956007 -283.4440900 -285.7219708 8 . 0 0 0 -285.6952987 -283.4435581 -285.7216556 1 2 . 0 0 0 -285.6951907 •283.4434194 -285.7215466 HCOOH -189.8148552 -188.4070742 -189.8317654 CH 3 N H j -95.8803125 -95.0363224 -95.8897596 ^DG-1 basis set 3Version 1.2 of DMol 4LSD with gradient corrections 350,354 5DG-2 basis set 6LSD with gradient corrections 351'353 Interaction Energy [kcal/mol] -20 -10 10 0 distance calculated by different DF approaches. See text text See corrections. approaches. H5™N6 gradient DF the on various the of different energy by discussion for calculated interaction f o distance Dependence 61. Figure 1 .. Dsac [A] Distance N...H 2 3 4 ♦ o DG-1 DGauss ■ □ + DGauss DG-1 DG-1 DGauss + + DGauss DG-2 -P DG-2 B NLC DGauss LSD DGauss DG-2 DG-2 DGauss L B-P B NLC DGauss DG-1 DG-1 DGauss -S B NLC LSD LSD DNP DMol 231 232 hydrogen bond which is too long and too weak (Figure 61). It is known that gradient corrections to LSD energies are essential in predicting the strength of certain chemical bonds 381,382 (also see Chapter Vlll). The importance of these corrections in studying the hydrogen bond in water has been noted .340

At the moment, neither DMol nor DGauss provide for inclusion of ghost atoms to check for basis set superposition errors. There are however some indications that BSSE effects should not be large. The curves obtained by DGauss for two different basis sets (DG-1 and DG-2) are very similar. Since these basis sets differ only in the size of core orbitals, which are mainly responsible for BSSE, one can expect the BSSE to be small in the LSD

Gaussian-type calculations. In the HF treatment, BSSE energies were quite small when using the DZPP basis set; a similar basis is used in the program DMol, though the theoretical approaches are quite different. Also, DMol uses high-quality atomic basis sets, which should provide excellent treatment for the core orbitals and minimize the possibility of basis set superposition errors .345 On the other hand, it is unclear whether the exchange-correlation operator (which has no analogy in the Hartree-Fock treatment, see equation 8.2) will introduce BSSE effects.

The fully-optimized geometries and corresponding energies for formic acid, methylamine, and their complex obtained by DMol and DGauss starting from the 3-21G*(*) geometry

(within the local spin density (LSD) approximation) are compared to those obtained by

CADPAC at HF and MP2 levels using the DZPP basis set in Tables 30 and 31. The most significant differences are in the description of the hydrogen bond. The overall length of the H-bond (03*»N6) from LSD calculations is 0.12 A and 0.3 A shorter than that obtained in the HF/MP2 and HF calculations, respectively; also, in the LSD geometry H5 is 0.2 and 0.4 A closer to N 6 than in the MP2 and HF geometries, respectively. Since Interaction Energy [kcal/mol] -20 -10 Figure 62. Comparison of interaction energies calculated by HF/MP2 10 0 1 .. Dsac [A] Distance N...H and LSD approaches with and without gradient corrections. 2 3 4 us -2 G D auss G D ♦ DQas DQ-2 Q D auss Q D O + DMol DNP DNP DMol + ▼ MP2 BSSE BSSE MP2 ▼ L B-P B NLC LSD LSD DZP 233 Table 30. Comparison of energies and optimized geometries of FAMA complex 1 calculated by MP2, HF, DGauss and DMol. Bond lengths in A, angles in degrees, energies in hartrees.

MP2 DZP HF DZP D G au u LSD DMol LSD

Formic acid:methylaxnine complex

Energy -284.9616978 -284.0713031 -283.4593640 -283.5472272 C 1-H 2 1.097 1.088 1.118 1.116 C l - 0 3 1.332 1.309 1.314 1.315 C l - 0 4 1.225 1.191 1.230 1.229 0 3 -H S 1.013 0.971 1.086 1.096 H 5-N 6 1.692 1.874 1.501 1.492 N 6-C 7 1.473 1.462 1.459 1.458 N 6-H 8 1.017 1.002 1.030 1.033 N 6-H 9 1.017 1.002 1.030 1.033 C7-H10 1.090 1.084 1.103 1.103 C7-H 11 1.094 1.088 1.107 1.107 C7-H12 1.090 1.084 1.103 1.103 H 2 -C 1 -0 3 110.55 111.08 112.20 112.20 H2-C1-04 123.46 123.12 122.38 121.98 0 3 - C 1 - 0 4 125.99 i25.80 125.41 125.81 C 1 -0 3 -H 5 107.30 110.57 105.52 105.28 03-H5-N6 171.73 175.23 165.19 165.49 H 5-N 6-C 7 122.59 121.39 126.75 126.89 H5-N6-H8 103.64 103.23 99.18 99.18 H5-N6-H9 103.65 103.23 99.33 99.19 C 7-N 6-H 8 110.30 110.79 112.38 112.45 C 7-N 6-H 9 . 110.30 110.80 112.36 112.44 H8-N6-H9 104.82 106.15 103.90 103.64 N 6-C 7-H 10 108.81 109.30 109.69 109.50 N 6-C 7-H 11 113.66 113.58 113.70 114.03 N6-C7-H12 108.81 109.30 109.75 109.50 H10-C7-H11 108.70 108.36 108.14 108.13 H 10-C 7-H 12 108.02 107.76 107.27 107.33 H 11-C 7-H 12 108.70 108.36 108.08 108.14 H2-C1-03-H5 180.00 180.00 180.00 180.00 04-C1-03-H5 0.00 0.00 0.00 0.00 C1-03-H5-N6 0.00 0.00 0.00 0.00 03-H5-N6-C7 180.00 180.00 180.00 180.00 03-H5-N6-H8 54.60 55.24 52.55 52.75 G3-H5-N6-H9 •54.66 -55.17 -53.30 -52.80 H5-N6-C7-H10 -58.73 •58.86 -58.73 -58.70 H5-N6-C7-H11 180.00 180.00 180.00 180.00 H5-N6-C7-H12 58.74 58.86 58.88 58.72 H8-N6-C7-H10 63.61 62.36 62.78 63.03 H8-N6-C7-H11 -57.66 -58.78 -58.43 -58.25 H8-N6-C7-HI2 -178.92 180.00 180.00 180.00 H9-N6-C7-H10 178.92 180.00 180.00 180.00 H9-N6-C7-H11 57.65 58.77 58.34 58.28 H9-N6-C7-H12 -63.61 -62.38 -62.84 -63.02 235

LSD geometry optimizations by DMol and DGauss give very similar structures, it seems likely that the discrepancies between traditional Hartree-Fock and DFT calculations can only be remedied by including gradient corrections in geometry optimizations, an option which is not yet available. The binding energy is also severely overestimated without gradient corrections: -9.9, -14.2, -22.3 and -21.9 kcal mol -1 for HF/DZPP, MP2/DZPP,

LSD-DGauss and LSD-DMol, respectively. These binding energies were not corrected for BSSE, but the magnitudes of these errors for the HF and HF/MP2 results should be similar to those at the minimum in the distance study, namely 1 kcal mol ' 1 and 3 kcal mol'1, respectively.

9.3.2 FAMA Complex 2 This complex was constructed to be similar that studied extensively by Hadzi et, a l'335,370,363 p u y ge o m e t r y optimizations were performed using two different basis sets at the Hartree-Fock level, namely DZP(P) (see section 2.3) and 6-311++G**, and also using

DMol and DGauss LSD-DFT implementations. The two HF calculations find substantively the same geometries (Table 32), and the two LSD-DFT calculations agree closely. There were however substantial differences in between the geometries found by the HF treatments and those obtained using DMol or DGauss (Table 32, Figure 63); in the latter, the 04-H9 distance (1.93 A) allows a second hydrogen bonding interaction, making a "bifurcated" complex. In contrast, the 04-H9 distance in the DZP(P) and 6-311++G** calculations (2.79 A) is much larger, and the contribution to the binding energy from this interaction should be much less (c.f. Figure 59). We have not yet performed HF/MP2 calculations on this complex, but in the computations presented in Chapter VIII it was found that DFT geometries were generally in closer agreement with HF/MP2 than with

HF geometries. Table 31. Comparison of energies and optimized geometries of formic acid and methylamine calculated by MP2 and HF methods, DGauss (DG-1 basis) and DMol. Bond lengths in A, angles in degrees, and total energy in Hartrees.

M P2D Z P HF DZP D C a u ia LSD DM ol LSD

Formic acid

Energy -189.3453762 -188.8150210 -188.3943201 -188.4452214 C 1-H 2 1.094 1.086 1.114 1.112 C l - 0 3 1.352 1.323 1!342 1.343 C l - 0 4 1.215 1.184 1.211 1.211 0 3 -H 5 0.972 0.950 0.988 0.998 H2-C1-03 109.25 110.48 109.49 109.64 H 2 -C 1 -0 4 125.55 124.58 125.66 125.33 0 3 - C 1 - 0 4 125.20 124.94 124.86 125.03 C 1 -0 3 -H 5 106.54 109.19 106.34 105.98 H2-CI-03-H5 180.00 180.00 180.00 180.00 04-C1-03-H5 0.00 0.00 0.00 0.00

Methylamine

Energy -95.5936433 -95.2405088 -95.0295531 -95.0671446 N 6-C 7 1.468 1.456 1.451 1.456 N 6-H 8 1.015 1.001 1.025 1.032 N 6-H 9 1.015 1.001 1.025 1.032 C 7-H 10 1.091 1.085 1.105 1.105 C7-H 11 1.097 1.092 1.114 1.113 C 7-H 12 1.091 1.085 1.105 1.105 C 7 -N 6 -H 8 109.62 110.91 110.87 109.60 C 7 -N 6-H 9 109.82 110.91 111.00 109.59 H 8-N 6-H 9 106.01 107.39 107.10 105.17 N 6-C 7-H 10 108.87 109.32 109.45 109.28 N 6-C 7-H 11 115.06 114.53 116.24 115.65 N 6-C 7-H 12 108.88 109.32 109.50 109.29 H 10-C 7-H 11 108.16 108.03 107.50 107.74 H 10-C 7-H 12 107.45 107.37 106.20 106.79 H 11-C 7-H 12 108.17 108.03 107.47 107.74 H8-N6-C7-H10 63.59 61.75 62.64 64.28 H8-N6-C7-H11 -57.97 -59.61 -59.34 -57.45 H8-N6-C7-H12 180.00 179.02 178.67 180.00 H9-N6-C7-H10 180.00 -179.01 -178.45 180.00 H9-N6-C7-H11 58.00 59.63 59.56 57.48 H9-N6-C7-H12 -63.58 -81.74 -62.42 -64.26 237 H12 H12

C7 - * Hll

mo C2 HIO f H9 H I • HI N6 H5 H5 H8 03 H8

c

f H12 04 i

• C7 H ll C2 HI %PH9 HIO 7N 6 0 3 H5 H8

Figure 63. Geometry optimization for FAMA Complex 2. A: Starting geometry constructed as described in the text B: Fully-optimized geometry obtained with DMol with the DNP basis. C: Fully-optimized geometry obtained with CADPAC DZP(P) basis.

9.3,3 F* M' Bifurcated Complex The optimized 3-21G*(*) geometry for the bifurcated complex is shown in Figure 64.

There were only minor differences between the starting geometry (ree section 2.4) and the fully-optimized geometry. The geometries obtained using DFT in both the DGauss and DMol implementations were also very similar (Table 33). We were surprised when we identified this complex wherein the ammonium and formate essentially retain their charged geometries, and at the time we began this project such a complex had not been reported in the literature, although subsequently Koller and Hadzi identified a similar 238

Table 32. Comparison of energies and optimized geometries of FAMA Complex 2 calculated by CADPAC, DGauss and DMol. See Figure 63 for atom numbering. Bond lengths in A, bond angles in degrees, energies in hartrees. HF HP DFT DFT DZP(P) 6-311++G** DGauss DG-1 DMol DNP

Energy: -284.06663569 -284.08685958 -283.46134121 -283.55537139 -285.72164219* C1-H2 1.088 1.087 1.117 1.107 C2-03 1.309 1.307 1.309 1.306 C2-04 1.192 1.185 1.236 1.230 03-H5 0.971 * 0.967 * 1.117* 1. 1 2 0 * C2-N6 3.425 * 3.446* 2.954 * 2.959 * H5-N6 1.870* 1.890* 1.429 * 1.431 * 04-H9 2.786 * 2.789 * 1.927 * 1.932 * N6-H8 1.002 1.001 1.026 1.027 N6-H9 1.003 1.002 1.045 1.043 N6-C7 1.463 1.460 1.458 1.456 C7-H10 1.083 1.084 1.103 1.099 C7-H11 1.088 1.088 1.107 1.102 C7-H12 1.083 1.084 1.104 1.100 03-C 2-04 125.8 125.8 125.9 125.8 H1-C2-03 111.1 111.1 112.8 113.0 H1-C2-04 123.1 123.1 121.2 121.2 C2-03-H5 110.5* 110.7* 105.7 * 105.8 * 03-H5-N6 172.2 * 172.6 * 162.4 * 162.4 * H5-N6-C7 114.5* 116.9* 115.0* 115.2* H5-N6-H9 96.0 95.2 89.5 89.4 H5-N6-H8 116.6 114.5 116.4 115.9 107.4 107.1 109.3 108.8 C7-N6-H8 110.6 110.8 112.5 113.0 C7-N6-H9 110.4 110.8 111.7 112.0 C2-N6-C7 117.4* 122.7 * 121.2 * 121.2 * N6-C7-H12 109.2 109.2 108.7 108.5 H10-C7-H11 108.4 108.4 108.5 108.7 H10-C7-H12 107.9 107.8 107.6 108.0 H11-C7-H12 108.3 108.3 108.1 108.3 04-C2-03-H5 0.0 0 0.05 0 .02 -0.05 H1-C2-03-H5 180.0 180.0 180.0 180.0 C2-03-H5-N6 -5.8* -7.3* -7.8* -8.4* 03-H5-N6-C7 -97.7 * -104.0 * -103.3 * - 102.6 * “Gradient-corrected (NLSD) energy. *Key parameters defining the interaction of formic acid with methylamine. 239

H12 HIO CO J ' 0 3 jO H 7 H l l H2 C l

N5 H9 0 4 H8

Figure 64. Optimized 3-21G*(*) geometry for the bifurcated formate:methylammonium complex.

configuration in their semi-empirical studies using AM-1 and MNDO/H methods .309 In

an attempt to optimize this complex using the 6-311++G** basis, the final geometry was that of FAMA Complex 2 (as in Figure 63). Since Table 34 indicates that the energy of the F+M- bifurcated complex is above that of the combined energies for separate formic acid and methylamine at the HF level (but not with MP2 treatment), this optimization must be repeated with higher levels of theory before firmer conclusions can be drawn.

The geometry of the bifurcated complex allows for a closer C-N approach than the neutral hydrogen bond, 2 .8 A vs. 3.3 A (Figure 65), and even though this complex is energetically less favorable than the neutral complexes in vacuo (Table 34), this interaction geometry could be relevant in solution, where charged forms are favored.

We examined the effects of varying the C-N distance, keeping all other inter- and intra-molecular parameters fixed. Because of the closer approach of the two monomers, the effects of basis set superposition error ( BSSE) are larger than in FAMA Complex 1

(Figures 6 6 - 72). For the 3-21G*(*) basis, the BSSE amounts to almost 14 kcal mol ' 1 at 240

Table 33. Comparison of energies and optimized geometries of formate:methylammmonium bifurcated complex calculated by CADPAC, DGauss, and DMol. Bond lengths in A, angles in degrees, energies in hartrees.

HF DFT DFT 3-21G*(*) DGauss DZVP DMol DNP

Energy: -282.56389820 LSD: -283.45906262 -283.55341591 NLSD: -285.71456932 C1-H2 1.104 1 .120 1.110 Cl-03 1.240 * 1.271 1.267 Cl-04 1.240 * 1.271 1.266 03-H 7 1.584 * 1.467 1.480 04-H 9 1.584 * 1.466 1.484 C1-N5 2.835 * 2.811 2.809 N5-C6 1.482 1.460 1.456 N5-H7 1.066* 1.132 1.122 N5-H8 1.006 1.027 1.027 N5-H9 1.066* 1.133 1.120 C6-H10 1.080 1.103 1.097 C6-H11 1.080 1.103 1.098 C6-H12 1.080 1.103 1.098 03-C 1-04 127.5 126.4 126.6 H2-C1-03 116.2 116.8 116.7 H2-C1-04 116.2 116.8 116.7 C1-03-H7 103.5 103.1 103.2 C1-04-H9 103.5 103.1 103.2 03-H7-N5 146.7 * 149.5 * 149.0 * 04-H9-N5 146.8 * 149.5 * 148.8 * H7-N5-H9 91.6 88.1 89.0 C1-N5-C6 H7-N5-C6 113.1 113.8 114.4 H8-N5-C6 113.0 113.5 114.0 H9-N5-C6 113.1 113.8 114.4 H10-C6-H11 109.8 108.9 109.1 H10-C6-H12 109.4 108.4 108.7 H11-C6-H12 109.4 108.4 108.7

03-C1-04-H9 - 0 .6 -0.7 - 1.6 C1-04-H9-N5 4.9 5.6 2.5 04-H9-N5-C6 109.6 108.3 114.1 H9-N5-C6-H10 -170.6 -168.7 -169.6 * Denotes key parameter defining interaction geometry. 241

-282.25

-282.30

-282.35 g t o -282.40 -C ^ -282.45 c Ui -282.50

-282.55

-282.60 C1-N5 Distance (Angstroms)

Figure 65. Comparison of energy vs. C-N distance for bifurcated and "linear" FAMA complexes. Calculations using 3-21G*(*) basis set. the energy minimum (Fig. 6 6 ); still, when corrected for BSSE, the potential curves for

3-21G*(*) and 6-311G** are very similar (Figure 73), which in a sense validates this method of correcting for BSSE for these calculations even though it is known in the literature that this approach sometimes breaks down .383,384 The BSSE's are much smaller for the DZPP basis (Figure 67) and almost negligable using the 6-311++G** basis (Figure

6 8 ). As with the neutral complex the BSSE is increased when energies are corrected for part of the electron correlation energy using second order Moller-Plesset perturbation theory (MP2) (Figures 69 and 70); the errors are smaller with the large 6-311++G** basis than with the DZPP basis, yet these calculations indicate that even larger basis sets may be required to give optimized geometries free from the effects of BSSE. This is unfortunate since the MP2/6-311++G** calculations are already very time consuming in this modestly-sized system. Interaction Energy (kca l/mol) Interaction Energy (kca l/mol) 1 - -100 -140 1 - -130 -120 -110 100 0 -1 90 -9 basis s e t Calculations performed with the program CADPAC. DZPP program the the ith with w the performed for complex r Calculations rro e bifurcated t e s basis superposition set f basis o formate:methylammonium Effect 67. Figure fonuate:m ethylam m onium bifurcated complex w ith the 3-21G*(*) 3-21G*(*) the ith w complex bifurcated onium m ethylam fonuate:m basis set. Calculations performed with the program CADPAC. program the with performed Calculations set. basis Figure 0 9 - 110 6 6 . Effect o f basis set superposition e rro r for the the for r rro e superposition set basis f o Effect . . 26363.8 3.6 2.6 2.4 2.4 A P C -1K) 32G<) BSSE 3-21G«<*) ’ « CADPAC 3-21GK*) 2.6 A P C ZP ■ DZPP «■ BSSE CADPAC DZPP C1— (Angstroms) RadiusN5 C1— (Angstroms) RadiusN5 2.8 2.8 . . 3.4 3.2 3.0 3.0 . . 3.6 3.4 3.2 . 4.0 3.8 4.0 242 Interaction Energy (kcal/mol) -1 130 3 -1 -120 0 0 1 - 90 -9 the energies. Calculations formate:methylammonium using CADPAC. bifurcated complex with the DZPP basis set and second Figureorder Moller-Plesset 69. Effect of basis corrections set superposition (MP2) to error for the basis set. Calculations formate:methylammonium performed bifurcatedwith the complex program with GAUSSIAN the 6-311++G** 90. Figure 68. Effect of basis set superposition error for the 2. 24 . 28 . 32 . 36 . 4.0 3.8 3.6 3.4 3.2 3.0 2.8 2.6 2.4 12?.2 2.4 A P C ZP12 DZPP/MP2 BSSE CADPAC DZPPA1P2 9 6- +G* - —1+G* BSSE 6—311++G«* *- 1++G** 1 -3 6 G90 . 2.8 2.6 1N Rdu (Angstroms) Radius C1-N5 1N Rdu (Angstroms) Radius C1-N5 . 3.2 3.0 . 3.6 3.4 . 4.0 3.8 -9 0 244

„E -100 "5

L.& g -1 1 0 UJ

0 1 -1 2 0 *E

- 1 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 C1-N5 Radius (Angstroms) G90 6-311 ++G**/MP2 6-311 ++G«*/MP2 BSSE Figure 70. Effect of basis set superposition error for the formate:methylammonium bifurcated complex with the 6-311++G** basis set and second order Moller-Plesset corrections (MP2) to the energies. Calculations using GAUSSIAN 90.

Adding diffuse functions gives smaller interaction energies (Figure 74), as expected based on proton affinity calculations (section 8.1). DFT interaction energies calculated using the DMol and DGauss implementations of the LSD approximation are in very close agreement (Figure 75). As for Complex 1, these interaction energies are substantially higher than that obtained with any of the Hartree-Fock or HF/MP2 treatments (Figure 76).

In contrast to the Complex 1, even with the gradient corrections the DFT calculations give a larger interaction energy than any of the HF or HF/MP2 treatments. Particularly noteworthy is the sizeable discrepancy (9 kcal/mol) between the gradient-corrected ("NLSD") results and the MP2/6-311++G** interaction energies, since our proton affinity studies (Chapter VIII) suggest that the 6-311++G** results should be the most reliable estimate of the interaction energy for this complex. The agreement between the DGauss NLSD results and the MP2/DZPP results is much closer (4 kcal/mol), which might suggest that diffuse functions are needed in the DFT basis. This did not appear to be the 245

I \

- 1 0

-2 0

-2 5 C1—N5 Radius (Angstroms) ♦ CDPC 3-21 G*<*) CADPAC DZPP ♦ G90 6 - 3 1 1++G G90 6—311G** -*■ G90 6 - 3 1 1++G«* Figure 71. Basis set superposition error vs. C1~N5 distance in the formate:methylammonium bifiircated complex using various basis sets at the Hartree-Fock level. Calculations with CADPAC or GAUSSIAN 90 (G90).

case in calculating proton affinities, but this complex is clearly demanding from a computational standpoint and may therefore be a good test system for further development of the DFT methodology. In studying this complex we did not test for the effects of basis

set superposition error using different core orbital representations in DGauss; however the

apparent overestimation of the energy is likely not due to BSSE effects (see discussion section 9.3.1). Unfortunately, the discrepancy between the 6-311++G**/MP2 results and the NLSD (gradient-corrected) results obtained with DGauss means we do not know which answer is really "right"

9.3.4 Summary of formic acid - methylammonium interactions

We have identified several interaction geometries for formic acid and methylammonium.

These and others have been recently proposed .309-335*369,370,371 Energetically the neutral complexes are favored by as much as 10 kcal mol ' 1 (Table 34) over the F+ M' Bifurcated 246

6 £ - 1 0 Ori oe- - i s 9- w•M - 2 0 - c8 n -25 *5a CD -30 C1—N5 Radius (Angstroms)(Angstroms) CADPAC DZPP/MP2 «■ G90 6—311++G/MP2 G90 6-311 G*^/MP2 -*• G90 6-311++G«*/MP2 Figure 72. Basis set superposition error vs. C1~N5 distance in the formate:methylammonium bifbrcated complex using various basis sets and second order Maller-Plesset corrections (MP2) to the energies. Calculations with CADPAC or GAUSSIAN 90 (G90). -90

1 \ -1 00

£ —110 UJ c o

s -1 2 0 ■6

- 1 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 C1-N5 Radius (Angstroms) ♦ CADPAC 3-21GK*) BS «* G90 6-311G«* BSSE Figure 73. Dependence of interaction energy vs. CI-N5 distance in the formate:methylammonium bifurcated complex using 3-21G*(*) and 6-311G** basis sets at the Hartree-Fock level. Calculations with CADPAC (3-2IG+(*)) or GAUSSIAN 90 (G90) (6-311G*+). Interaction Energy (kca l/mol) Interaction Energy (kca l/mol) DGauss (DZVP)in the formate:methylammonium implementations of the bifurcatedlocal spin complex density approximation. bv DMol (DNP) and Figure 75. Dependence of interaction energv on C1~N5 distance Fock level. Calculationsin the formate: methylammoniumwith GAUSSIAN bifurcated 90 (G90).complex at the Hartree- Figure 74. Effect of diffuse functions on the interaction energy 4 ' - 0 2 -1 -110 -100 -1 130 3 -1 - 0 0 1 - 90 -9 80 -8 120 0 1 1 ' 311G** 1 -3 6 . 2.6 2.4 2.4 M l ZP DGAUSS DZVP LSD DMol DZPP 2.6 C1—N5 (Angstroms) Radius 1N Rdu (Angstroms) Radius C1-N5 . 30 . 3.4 3.2 3.0 2.8 . 30 . 3.4 3.2 3.0 2.8 311+ "" 6—311++G<* "0" 1++G 1 -3 6 . . 4.0 3.8 3.6 . 3.8 3.6 4.0 247 Interaction Energy (kcal/mo!) " 5 0 1 - -125- 145 4 -1 -45--, -25 - —1 +*/P G ZP S DG DZVP NLSD DG DZVP LSD 6—311 ++G**/MP«- 65 gradient corrections to the local density approximation. HF, HF/MP2Figure by and 76. DFT Comparison approaches of withinteraction (NLSD) energies and without (LSD) A P C ZP•- 9 6-311++G** 690 •o- CADPAC DZPP * 1N Rdu (Angstroms) Radius C1-N5 DZPP/MP2 (Angstroms) 249 Complex. This is to be expected in a vacuum given the higher proton affinity of the formate (or acetate) anions compared to methylamine (Chapter VIII). A set of

calculations was performed wherein the N-H bond distance was held at fixed values while allowing otherwise full geometry optimizations using the 3-21G*(*) basis set (Figure 77). These results corroborate (but with a more extensive treatment) results obtained by other

workers .335 The neutral complex is clearly favored, and the charged form (which is not the bifurcated complex however) constitutes only an inflection point on the curve. Binding of drug to receptor clearly does not take place in vacuo , however, and in solution phases at neutral pH the ionic forms of most aliphatic amines and carboxylates are preferred. Drug-receptor interactions consist of highly specific hydrophobic and hydrogen bonds between the drug molecule, amino acid side chains on the receptor protein, and bound water molecules. This is not the environment of free aqueous solution nor what would be encountered in a hydrophobic phase such as hexane. The possible effects of solvation have been modelled in FAMA complexes by appropriately-placed point charges 370 or a reactive solvent field model .335 These models generally favor proton transfer to the nitrogen to give charged complexes. Hence, the calculated in vacuo energetics may have minimal relevance to binding in the real biological system. Continued improvement and characterization of the DFT approach offers the promise of improved receptor models, since calculations may be performed on much larger systems

(including, e.g., solvent water and additional receptor fragments). Table 34. Summary of Energies for the FAMA System.

Geometry: 3-21G*(*) DMol DGauss DZVP Distances (A) Entity 3-21G*(*> DZPP DZPP/MP2 DNP LSDNLSD O-N O-H N-H

Separate +148 +134 +136 +128 +130 +130 h c o 2- + CH 3NH3+

Separate (0 ) (0) (0 ) (0 )(0 )(0 ) h c o 2h + CH 3NH2

Bifurcated +14 +7.2 -1.2 -22 -23 -8.7 2.54 1.58 1.07 F+M' Complex (2.51) (1.48) (1.12)

FAMA Complex 1 -8.7 -9.9 -10.7 -18 -23 2.78 0.97 1.83 (2.57) (1.10) (1.49)

FAMA Complex 2 c c c -23 -24 -13.1 2.84b 0.97b 1.83b (2.52) (1.12) (1.43) “Corrected for estimated basis set superposition error bDZP(P) geometry ( ) DMol geomtry. ‘Optimization not yet performed with this basis set -282.520 -t i i t- i i i 251

-282.530

-282.540

I -282.550 o x

| -282.560 e •-A -282.570 9 -2B2.580 /

-282. 8 1.0 1.2 1.4 . 1.6 1.8 2.0 2.2 2.4 N-H Distance (Angstroms) Figure 77. Energies resulting from ftill geometry optimizations for fixed N-H distances using the 3-21G*(*) basis. Hie geometries of the two limiting forms are shown. For reference, 0.1 Hartree = 62.7 kcal mol'1. BIBLIOGRAPHY 1. Okamoto, S., "Epileptogenic action of glutamate directly applied into the brain of animals and inhibitory effect of proteins and tissue emulsions on its action", J. Physiol. Soc. Jpn., 13 (1951) 555-562.

2. Hayashi, T. "A physiological study of epileptic following cortical stimulation in animals and its application in the human clinic", Jpn. J. Physiol., 3 (1952) 46-64.

3. Curtis, D.R. and J.C. Watkins, "The excitation and depression of spinal neurons by structurally-related amino acids", J. Neurochem., 6 (I960) 117-141. 4. Curtis, D.R. and J.C. Watkins, "Acidic amino acids with strong excitatory actions on mammalian neurones", J. Physiol, 166 (1963) 1-14. 5. Anonymous, "Excitatory Amino Acids ‘Flying High’", Tocris Neuramin, September, 1988.

6 . Monaghan, D.T., R.J. Bridges, C.W. Cotman, "The Excitatory Amino Acid Receptors: Their Classes, Pharmacology, and Distinct Properties in the Function of the Central Nervous System", Annu. Rev. Pharmacol. Toxicol, 29 (1989) 365-402. 7. McGeer, P.L., and E.G. McGeer, "Amino Acid Neurotransmitters", in Basic Neurochemistry, 4th edit., G. Siegel, B. Agranoff, R.W. Albers, P. Molinoff, eds. (Raven Press, New York, 1989), Chapter 15, pp. 323-328.

8 . Cotman, C.W., D.T. Monaghan, C.P. Ottersen, and J. Storm-Mathisen, "Anatomical organization of excitatory amino acid receptors and their pathways", Trends Neurol. Sci., 70(1987) 273-280. 9. Johnson, G., "Recent Advances in Excitatory Amino Acid Research", Annu. Rep. Med. Chem., 24 (1989) 41-50.

10. Collingridge, G.L., and W. Singer, "Excitatory amino acid receptors and synaptic plasticity", Trends Pharmacol. Sci., 11 (1990) 290-296. 11. a. Brown, T.H., P.F. Chapman, E.W. Kairiss, and C.L. Keenan, "Long-Term Synaptic Potentiation", Science, 242 (1988) 724-728. b. Collingridge, G.L., and T.V.P. Bliss, "NMDA receptors - their role in long-term potentiation", Trends in Neurol. Sci., 10 (1987) 288-293. 12. Ojeda, S., "The Mystery of Mammalian Puberty: How Much More Do We Know?", Perspectives in Biology arid Medicine, 34 (1991) 365-383. 13. Olney, J.W., "Endogenous Excitotoxins and Neuropathological Disorders", in Excitatory Amino Acids in Health and Disease (John Wiley, New York, 1988), 337-351. 252 253 14. Headley, P.M., and S. Grillner, "Excitatory amino acids and synaptic transmission: the evidence for a physiological function", Trends Pharmacol. Sci., 11 (1990) 205-11. 15. Kmjevic, K., and J.W. Phillis, "Iontophoretic studies of neurons in the mammalian cerebral cortex", J. Physiol (London), 165 (1963)274-304. 16. Umbarger, E.H., "Utilization of Amino Acids", in Biochemistry, ed. G. Zubay (Addison Wesley, Reading, MA, U.S.A., 1986), Chapter 23. 17. McGeer, P.L., "Putative Excitatory Neurons: Glutamate and Aspartate", in Molecular Neurobiology o f the Mammalian Brain , 2nd edit., P.L. McGeer, J.C. Eccles, and E.G. McGeer, eds. (Plenum Press, New York, 1987), Chapter 6 , pp. 175-195. 18. Wofsey, A.R., M.J. Kuhar, and S.H. Snyder, "A unique synaptosomal fraction which accumulates glutamic and aspartic acids in brain tissue", Proc. Natl Acad. Sci. U.S.A., 68 (1971), 1102-1106. 19. Cuenod, M., Kim Quang Do, P. Streit, "Homocysteic Acid as an endogenous excitatory amino acid", Trends Pharmacol Sci., 11 (1990) 477-478.

20. Robinson, M.B., and J.T.Coyle, "Glutamate and related acidic excitatory neurotransmitters: from basic science to clinical application", FASEBJ., 1 (1987) 446-455. 21. Ottersen, O.P. and J. Storm-Mathisen, "Localization of amino acid neurotransmistters by immunocytochemistry", Trends Neurol Sci., 10 (1987) 250-255. 22. McLennan, H., "The Pharmacological Characterization of Excitatory Amino Acid Receptors", in Excitatory Amino Acids in Health and Disease (John Wiley, New York, 1988), 1-11.

23. Foster, A.C., and P.J. Roberts, "High Affinity L-[ 3H]-Glutamate Binding to Postsynaptic Receptor Sites on Rat Cerebellar Membranes", J. Neurochem., 31 (1978) 1467-77. 24. Mayer, M.L., and G.L. Westbrook, "The Physiology of Excitatory Amino Acids in the Vertebrate Central Nervous System", Progress in Neurobiology, 28 (1987) 197-276. 25. Zaczek, R., K. Koller, R. Cotter, D. Heller, and J.T. Coyle, "N-Acetylaspartylglutamate: An endogenous peptide with high affinity for a brain "glutamate" receptor", Proc. Natl. Acad. Sci. USA, 80 (1983) 1116-1119. 26. McGeer, P.I., and E.G. McGeer, "Amino Acid Neurotransmitters", in Basic Neurochemistry, 3rd edit., G.J. Siegel, R.W. Albers, B.W. Agranoff, and R. Katzman, eds., (Little, Brown, and Co., Boston, 1981) pp. 233-253.

27. FFrench-Mullen, J.M.H.. K. Koller, R. Zaczek, J.T. Coyle, N. Hori, and D.O. Carpenter, "N-Acetylaspartylglutamate: Possible role as the neurotransmitter of the lateral olfactory tract", Proc. Natl. Acad. Sci. USA, 82 (1985) 3897-3900. 254

28. Robinson, M.B., R.D. Blakely, R. Conto, J.T. Coyle, "Hydrolysis of the Brain Dipeptide N-Acetyl-L-aspartyl-L-glutamate: Identification and Characterization of a Novel N-Acetylated a-Linked Acidic Dipeptidase Activity from Rat Brain", J. Biol Chem., 262 (1987) 14498-14506.

29. Stone, T.W., and N.R. Burton, "NMDA Receptors and Ligands in the Vertebrate CNS", Progress in Neurobiology, 30 (1988) 333-368. 30. Lodge, D., and G. Collingridge, "Les agents provocateurs: a series on the pharmacology of excitatory amino acids", Trends in Pharmacol Sci., 11 (1990) 22-24. 31. Trends in Pharmacol Sci., 11 (1990) January 1990 Supplement. 32. McLennan, H., R.D. Huffman, and K.C. Marshall, "Patterns of excitation of thalamic neurones by amino-acids and by acetylcholine", Nature, 219 (1968) 387-388.

33. Foster, A.C., and G.E. Fagg, "Acidic Amino Acid Binding Sites in Mammalian Neuronal Membranes: Their Characteristics and Relationship to Synaptic Receptors", Brain Research Reviews, 7(1984) 103-164. 34. Davies, J.D., A.A. Francis, A.W. Jones, J.C. Watkins, "2-Amino-5-phosphonovalerate (2-APV), a potent and selective antagonist of amino acid-induced and synaptic excitation", Neurosci. Lett., 21 (1981) 77-81. 35. Hall, J.G., H. McLennan, and H.V. Wheal, "The actions of certain amino acids as neuronal excitants", J. Physiol, 272 (1977) 52-53. 36. Biscoe, T.J., R.H. Evans, A.A. Francis, M.R. Martin, J.C. Watkins, J. Davies, A. Dray, "D-a-aminoadipate as a selective antagonist of amino acid-induced and synaptic excitation of mammalian spinal neurones", Nature, 270 (1977) 743-745.

37. M.L. Mayer and L. Vylicky, Jr., "Concanavalin A selectively reduces desensitization of mammalian neuronal quisqualate receptors", Proc. Natl. Acad Sci. USA, 86 (1989) 1411-1415. 38. Ascher, P., and L. Nowak, "Electrophysiological studies of NMDA receptors", Trends Neurol Sci., 10 (1987) 284-285. 39. MacDermott, A.B., and N. Dale, "Receptors, ion channels and synaptic potentials underlying the integrative actions of excitatory amino acids", Trends Neurol. Sci., 10 (1987) 280-283.

40. MacDonald, J.F., and L.M. Nowak, "Mechanisms of blockade of excitatory amino acid receptor channels", Trends Pharmacol Sci., 11 (1990) 167-172.

41. Akaike, N., M. Kaneda, N. Hori, and O.A. Krishtal, "Blockade of N-methyl-D-aspartate response in enzyme-treated rat hippocampal neurons", Neurosci. Lett., 87 ( 1988)75-79. 42. Llano, I., A. Marty, J.W. Johnson, P. Ascher, and B.H. Gahwiler, "Patch-clamp recording of amino acid-activated responses in ‘organotypic’ slice cultures", Proc. N atl A cad Sci. USA, 85 (1988) 3221-3225. 255

43. Llerma, J., L. Kushner, D.C. Spray, M.V.L. Bennett, and R.S. Zukin, "mRNA from NCB-20 cells encodes the N-methyl-D-aspartate/phencyclidine receptor: A Xenopus oocyte expression study", Proc. Natl Acad. Sci. USA, 86 (1989) 1708-1711. 44. Dingledine, R., and N.W. Kleckner, "Requirement for Glycine in Activation of NMDA-Receptors Expressed in Xenopus Oocytes", Science, 241 (1988) 835-837. 45. Verdoom, T.A., N.W. Kleckner, and R. Dingledine, "N-Methyl-D-Aspartate/Glycine and Quisqualate/Kainate Receptors Expressed in Xenopus oocytes: Antagonist Pharmacology", Molecular Pharmacology, 35 (1989) 360-368. 46. Honore, T., "Excitatory Amino Acid Receptor Subtypes and Specific Antagonists", Medicinal Research Reviews, 9 (1989) 1-23. 47. Barnard, E.A., and J.M. Henley, "The non-NMDA receptors: types, protein structure and molecular biology", Trends Pharmacol. Sci., 11 (1990) 500-507. 48. Hansen, J J., and P. Krogsgaard-Larsen, "Structural, Conformational, and Stereochemical Requirements of Central Excitatory Amino Acid Receptors", Medicinal Research Reviews, 10 (1990) 55-94. 49. Willetts, J., R.L. Balster, and J.D. Leander, "The behavioral pharmacology of NMDA receptor antagonists", Trends Pharmacol Sci., 11 (1990) 423-428. 50. Thomson, A.M., V.E. Walker, and D.M. Flynn, "Glycine enhances NMDA-receptor mediated synaptic potentials in neocortical slices", Nature (London), 338 (1989) 422-427.

51. MacDonald, J.F., and L.M. Nowak, "Mechanisms of blockade of excitatoiy amino acid receptor channels", Trends Pharmacol Sci., 11 (1990) 167-172.

52. Javitt, D.C., M J. Frusciante, and S.R. Zukin, "Rat Brain N-Methyl-D-aspartate Receptors Require Multiple Molecules of Agonist for Activation", Molecular Pharmacology, 37 (1990) 603-607. 53. Sommer, B., K. Keinanen, T.A. Verdoom, W. Wisden, N. Bumashev, A. Herb, M. Kohler, T. Takagi, B. Sakmann, P.H. Seeburg, "Flip and Hop: A Cell-Specific Functional Switch in Glutamate-Operated Channels of the CNS", Science, 249 (1990) 1580-1585.

54. Keinanen et. al., "A Family of AMPA-Selective Glutamate Receptors", Science, 249 (1990) 556-560. 55. Honore, T., J. Drejer, E.O. Nielsen and M. Nielsen, "Non-NMDA Antagonist 3H-CNQX Binds with Equal Affinity to Two Agonist States of Quisqualate Receptors", Biochem. Pharmacol., 38 (1989) 3207-3212.

56. Pin, J.-P., J. Bockaert and M. Recasesn, "The Ca 2+/C l' dependent L-[3H]glutamate binding: a new receptor or a particular transport process?", FEBS Letters, 175 (1984) 31-36. 57.Johnson, R.L. and J.F. Koemer, "Excitatory Amino Acid Neurotransmission", J. Med. Chem., 31 (1988) 2057-2066. 256

58. Greenamyre, J.T., J.M.M. Olson, J.B. Penney, Jr., and A.B. Young, "Autoradiographic Characterization of N-Methyl-D-Aspartate-, Quisqualate-, and Kainate-Sensitive Glutamate Binding Sites", J. Pharmacol. Exper. Therap., 233 (1985) 254-263. 59. Henley, J.M., A. Ambrosini, P. Krogsgaard-Larsen, and E.A. Barnard, "Evidence for a Single Glutamate Receptor of the Ionotropic Kainate/Quisqualate Type", The New Biologist, 1 (1989) 153-158.

60. Honore, T., J. Drejer, and M. Nielsen, "Calcium Discriminates Two [ 3H]Kainate Binding Sites with Different Molecular Target Sizes in Rat Cortex", Neuroscience Letters, 65 (1986) 47-52. 61. Pin, J.-P., B.J. Van Vliet, and J. Bockaert, "Complex interaction between quisqualate and kainate receptors as revealed by measurement of GABA release from striatal neurons in primary culture", Eur. J. Pharmacol - Molec. Pharmacol Section, 172 (1989) 81-91. 62. Evans, R.H., S.J. Evans, P.C. Pook, and D.C. Sunter, "A comparison of excitatory amino acid antagonists acting at primary afferent C fibres and motoneurones of the isolated spinal cord of the rat", Br. J. Pharmacol, 91 (1987) 531-537,

63. Monaghan, D.T., R.J. Bridges, C.W. Cotman, "The Excitatory Amino Acid Receptors: Their Classes, Pharmacology, and Distinct Properties in the Function of the Central Nervous System", Annu. Rev. Pharmacol Toxicol., 29 (1989) p. 376.

64. Ascher, P., and L. Nowak, Quisqualate- and Kainate-activated Channels in Mouse Central Neurones in Culture", J. Physiol, 399 (1988), 227-245.

65. a. Potashner, S.J., and D. Gerard, "Kainate-enhanced release of D-[ 3H]aspartate form cerebral cortex and striatum: reversal by baclofen and ", J. Neurochem., 40 (1983) 1548-1557. b. Ben-Ari, Y., M. Gho, Long-lastmg modification of the synaptic properties of rat CA3 hippocampal neurons induced by kainic acid", J. Physiol, 404 (1988) 365-384. c. Fisher, R.S., B.E. Alger, "Electrophysiological mechanisms of kainic acid-induced epileptiform activity in the rat hippocampal slice", J. Neurosci. 4 (1984) 1312-1323.

6 6 . Gannon, R.L., L.T. Baty, and D.M. Tertian, "L(+)-2-Amino-4-phosphonobutyrate inhibits the release of both glutamate and dynorphin from guinea pig but not rat hippocampal and mossy fiber synaptosomes", Brain Res., 495 (1989) 151-155. 67. Anson, J., and G.G.S Collins, "Possible presynaptic actions of 2-amino-4-phosphonobutyrate in rat olfactory cortex", Br. J. Pharmac., 91 (1987) 753-761.

6 8 . Schoepp, D., J. Bockaert and F. Sladeczek, "Pharmacological and functional characteristics of metabotropic excitatory amino acid receptors", Trends Pharmacol Sci., 11 (1990)508-515.

69. Murphy, S.N., and R.J. Miller, "Two Distinct Quisqualate Receptors Regulate Ca2+ Homeostasis in Hippocampal Neurons ", Molecular Pharmacology, 35 (1989) 671-680. 257

70. Schoepp, D.D., B.G. Johnson, C.R. Salhoff, J.W. McDonald, and M.V. Johnston, "In Vitro and In Vivo Pharmacology of trans-md ci5,-(+)-1 -Amino-1,3-Cyclopentanedicarboxylic Acid: Dissociation of Metabotropic and Ionotropic Excitatory Amino Acid Receptor Effects", J. Neurochem ., 56 (1991) 1789-1796.

71. Monaghan, D.T., D. Yao and C.W. Cotman, "L-[3H]Glutamate binds to kainate-, NMDA-, and AMPA-sensitive binding sites: an autoradiographic analysis", Brain Res., 340 (1985) 378-383. 72. Olsen, R.W., O. Szamraj, and C.R. Houser, "[3H]AMPA binding to glutamate receptor subpopulations in rat brain", Brain Res., 402 (1987) 243-254. 73. Rainbow, T.C., C.M. Wieczorek, and S. Halpain, "Quantitative autoradiography of binding sites for [3H]AMPA, a structural analogue of glutamic acid", Brain Res., 309 (1984) 173-177.

74. Monaghan, D.T., and C.W. Cotman, "The Distribution of [3H]Kainic Acid Binding Sites in Rat CNS as Determined by Autoradiography", Brain Res., 252 (1982) 91-100.

75. Unnerstall, J.R., and J.K. Wamsley, "Autoradiographic Localization of High-Affinity [3H]Kainic Acid Binding Sites in the Rat Forebrain", Eur. J. Pharmacol, 86 (1983) 361-371. 76. Nielsen, E.O., J. Drejer, J.-H. J. Cha, A.B. Young, and T. Honore, "Autoradiographic Characterization and Localization of Quisqualate Binding Sites in Rat Brain Using the Antagonist [3H]6-Cyano-7-Nitroquinoxaline-2,3-Dione: Comparison with (R,S)-[3H] a-Amino-3 -Hydroxy-5 -Methy 1-4-Isoxazolepropionic Acid Binding Sites", J. Neurochem., 54 (1990) 686-695.

77. Nicholls, D., and D. Attwell, The release and uptake of excitatory amino acids", Trends Pharmacol. Sci., 11 (1990) 462-468.

78. Palaiologos, G., L. Hertz, and A. Schousboe, "Evidence that Aspartate Aminotransferase Activity and Ketodicarboxylate Carrier Function Are Essential for Biosynthesis of Transmitter Glutamate", J. Neurochem., 51 (1988) 317-320. 79. a. Maycox, P.R., J.W. Hell, and R. Jahn, "Amino acid neurotransmission: spotlight on synaptic vesicles", Trends Neurol. Sci., 13 (1990) 83-87. b. Naito, S., and T. Ueda, "Characterization of Glutamate Uptake into Synaptic Vesicles", J. Neurochem., 44 (1985) 99-109. 80. Martin, D., G.A. Bustos, M.A. Bowe, S.D. Bray, and J.V. Nadler, "Autoreceptor Regulation of Glutamate and Aspartate Release from Slices of the Hippocampal CA1 Area", J. Neurochem., 56 ( 1991) 1647-1655.

81. Gannon, R.L., L.T. Baty, and D.M. Terrian, "L(+)-2-amino-4-phosphonobutyrate inhibits the release of both glutamate and dynorphin from guinea pig but not rat hippocampal mossy fiber synaptosomes", Brain Res., 495 (1989) 151-155. 258

82. J.C. Watkins and H.J. Olverman, "Structural Requirements for Activation and Blockade of EAA Receptors", in Excitatory Amino Acids in Health and Disease (John Wiley, New York, 1988), 13-45.

83. Ethnopharmacologic Search for Psychoactive Drugs (eds. D.H. Efirom, B. Holmstedt, N.S. Kline), U.S. Health Svc. Pub. No. 1645 (1967) p. 416. 84. Koemer, J.F., R.L. Johnson, R.K. Freund, M.B. Robinson, and S.L. Crooks, "Structure-Function Relationships for Gamma-Substituted Glutamate Analogues on Dentate Granule Cells", Brain Res., 272 (1983) 299-309. 85. Bigge, C.F., J.T. Drummond, G. Johnson, T. Malone, A.W. Probert, Jr., F.W. Marcoux, L.L. Coughenour, and L.J. Brahce, "Exploration of Phenyl-Spaced 2-Amino-(5-9)-phosphonoalkanoic Acids as Competetive N-Methyl-D- Antagonists", J. Mea. Ckem., 32 (1989) 1580.

86 . Watkins, J.C., P. Krogsgaard-Larsen, and T. Honore, "Structure-activity relationships in the development of excitatory amino acid receptor agonists and competetive antagonists", Trends Pharmacol Sci., 11 (1990)25-33. 87. Wong, E.H.F., J.A. Kemp, T. Priestley, A.R. Knight, G.N. Woodruff, and L.L. Iversen, "The anticonvulsant MK-801 is a potent N-methyl-D-aspartate antagonist", Proc. Natl. Acad. Sci. USA, 83 (1986) 7104-7108. 88 . Snell, L.D., and K.M. Johnson, "Phencyclidine: Behavioral Correlates of NMDA Antagonism", in Excitatory Amino Acids in Health and Disease (John Wiley, New York, 1988), 261-273. 89. Church, J., "Neuromodulatory effects of dextromethorphan: role of NMDA receptors in responses", Trends Pharmacol Sci., 11 (1990) 146-147. 90. Kessler, M., T. Terramani, G. Lynch, M. Baudry, "A Glycine Site Associated with N-Methyl-D-Aspartic Acid Receptors: Characterization and Identification of a New Class of Antagonists", / Neurochem., 52 (1989) 1319-1328. 91. Frey, P., D. Bemey, P.L. Herrling, W. Mueller, and S. Urwyler, "6,7-Dichloro-3-hydroxy-2-quinoxaihiecarboxylic acid is a relatively potent antagonist at NMDA and kainate receptors", Neurosci. Lett., 91 (1988) 194-198. 92. Leeson, P.D., R. Baker, R.W. Carling, N.R. Curtis, K.W. Moore, B.J. Williams, A.C. Foster, A.E. Donald, J.A. Kemp, and G.R. Marshall, "Kynurenic Acid Derivatives. Structure-Activity Relationships for Excitatory Amino Acid Antagonism and Identification of Potent and Selective Antagonists at the Glycine Site on the N-Methyl-D-aspartate Receptor", J. Med. Chem., 34 (1991) 1243-1252.

93. Gray, N.M., M.S. Dappen, B.K. Cheng, A.A. Cordi, J.P. Biesterfeldt, W.F. Hood, and J.B. Monahan, "Novel Indole-2-carboxylates as Ligands for the -Insensitive N-Methyl-D-aspartate-Linked Glycine Receptor", / Med. Chem., 34 (1991) 1283-1292.

94. Murayama, K., S. Morimura, Y. Nakamura, and G. Sunagawa, "Synthesis of Pyrrolidine Derivatives, n. Synthesis of Kainic Acid and Its Derivatives by Wittig Reaction." Z Pharm. Soc. Japan (Yakugaku Zasshi), 85 (1965) 757-765. 259

95. Goldberg, O., A. Luini and V.L. Teichberg, "Lactones Derived from Kainic Acid: Novel Selective Antagonists of Amino Acid-induced Na+ Fluxes in Rat Stiatum Slices", Neuroscience Letters, 22 (1981) 187-191. 96. Davies, J., A.W. Jones, M.J. Sheardown, D. A.S. Smith, and J.C. Watkins, "Phosphono Dipeptides and Piperazine Derivatives as Antagonists of Amino Acid-Induced and Synaptic Excitation in Mammalian and Amphibian Spinal Cord", Neurosci. Lett., 52 (1984) 79-84.

97. Ganong, A.H., A.W. Jones, J.C. Watkins, and C.W. Cotman, "Parallel Antagonism of Synaptic Transmission and Kainate/Quisqualate Responses in the Hippocampus by Piperazine-2,3-Dicarboxylic Acid Analogs", J. Neurosci., 6 (1986) 930-937. 98. Crooks, S.L., M.B. Robinson, J.K. Koemer, R.L. Johnson, "Cyclic Analogues of 2-Amino-4-phosphonobutanoic Acid (APB) and Their Inhibiition of Hippocampal Excitatory Transmission and Displacement of [3H]-APB Binding", J. Med Chem., 29 (1986) 1988-1995. 99. Curry, K., D.S.K. Magnuson, H. McLennan, and M.J. Peet, "Excitation of rat hippocampal neurones by the stereoisomers of cis- and trans- 1 -amino-1,3-cyclopentane dicarboxylate", Can. J. Physiol. Pharmacol, 65 (1987) 2196-2201.

100. Lucas, D.R., and J.P. Newhouse, "The toxic effect of sodium L-glutamate on the inner layers of the retina", AMA Arch. Ophthalmol, 158 , (1957) 193-201.

101. Olney, J.W., "Brain Lesions, Obesity, and Other Disturbances in Mice Treated with Monosodium Glutamate", Science, 164 (1969) 719-721. 102. Olney, J.W., "Glutamate-induced Neuronal Necrosis in the Infant Mouse Hypothalamus", J. Neuropathol Exper. Neurol, 30 (1971) 75-90.

103. Rothman, S.M., and J.W. Olney, "Excitotoxicity and the NMDA receptor", Trends Neurol Sci., 10 (1987) 299-302.

104. Olney, J.W., "Endogenous Excitotoxins and Neuropathological Disorders", in Excitatory Amino Acids in Health and Disease (John Wiley, New York, 1988), 337-351.

105. Barinaga, M., "Amino Acids: How Much Excitement is Too Much?", Science, 247 (1990) 20-22. 106. Dingledine, R., C. J. McBain, and J.O. McNamara, "Excitatory amino acid receptors in epilepsy", Trends Pharmacol. Sci., 11 (1990), 334-338. 107. Pedder, S.C.J., R.I. Wilcox, J.M. Tuchek, D.D. Johnson, and R.D. Crawford, "Quisqualate receptors in epileptic fowl: the absence of coupling between quisqualate and N-methyl-D-aspartate receptors", Eur. J. Pharmacol., 175 (1990) 85-91. 108. Young, A.B., J.T. Greenamyre, Z. Hollingsworth, R. Albin, C. D’Amato, I. Shoulson, J.B. Penney, "NMDA Receptor Losses in Putamen from Patients with Huntington’s Disease", Science, 241 (1988) 981-983. 260 109. Stone, T.W., J.H. Connick, N.R. Burton and D.A.S. Smith, "Quinolinic and Kynurenic Acids as Endogenous Ligands at NMDA Receptors", in Excitatory Amino Acid Transmission (Alan R. Liss, 1987) 147-152.

110. Patel, S.C., D.N. Papachristou, and Y.C. Patel, "Quinolinic Acid Stimulates Somatostatin Gene Expression in Cultured Rat CorticiU Neurons", 7. Neurochem., 56 (1991) 1286-1291. 111. Butterworth, R.F., O.L. Le, J. Lavoie, and J.C. Szerb, "Effect of Portacaval Anastomosis on Electrically Stimulated Release of Glutamate from Rat Hippocampal Slices", J. Neurochem., 56 (1991) 1481-1484. 112. Lewin, R., "Environmental Hypothesis for Brain Diseases Strengthened by New Data", Science, 237 ( 1987) 483-484. 113. Spencer, P.S., P.B. Nunn, J. Hugon, A.C. Ludolph, S.M. Ross, D.N. Roy, and R.C. Robertson, "Guam Amyotrophic Lateral Sclerosis-Parksinsonism-Dementia Linked to Plant Excitant Neurotoxin", Science, 237 (1987) 517-522. 114. Weiss, J.H., and D.W. Choi, "Beta-N-Methylamino-L-Alanine Neurotoxicity: Requirement for Bicarbonate as a Cofactor", Science, 241 (1988) 973-975.

115. Bames, D.M., "Sex and Violence in Neuroscience", Science, 242 (1988) 1509-1510. 116. Bames, D.M., "NMDA Receptors Trigger Excitement", Science, 239 (1988) 254-256. 117. Foster, A.C., E.E. Mena, D.T. Monaghan, and C.W. Cotman, "Synaptic localization of kainic acid binding sites", Nature, 289 (1981) 73-75. 118. Coyle, J.T., "Kainic acid: insights into excitatory mechanisms causing selective neuronal degeneration", in Selective Neuronal Death (Ciba Foundation Symposium 126), (Wiley, Chichester, 1987) 186-203.

119. Meldrum, B., and J. Garthwaite, "Excitatory amino acid neurotoxicity and neurodegenerative disease", Trends Pharmacol Sci., 11 (1990) 379-387. 120. Braitman, D.J., and J.T. Coyle, "Inhibition of 3H-Kainic Acid Receptor Binding by Divalent Cations Correlates with Ion Affinity for the Calcium Channel", Neuropharmacology, 26 (1987) 1247-1251. 121. Winn, P., T.W. Stone, M. Latimer, M.H. Hastings and A.J.M. Clark, "A comparison of excitotoxic lesions of the basal forebrain by kainate, quinolinate, ibotenate, N-methyl-D-aspartate or quisqualate, and the effects on toxicity of 2-amino-5-phosphonovaleric acid and kynurenic acid in the rat", Br. J. Pharmacol., 102 (1991) 904-908. 122. Zaczek, R., and J.T. Coyle, "Excitatory Amino Acid Analogues: Neurotoxicity and Seizures", Neuropharmacol, 21 (1982) 15-26. 123. Schwarcz, R., D. Scholz, and J.T. Coyle, "Structure-activity Relations for the Neurotoxicity of Kainic Acid Derivatives and Glutamate Analogues", Neuropharmacol, 17 ( 1978) 145-151. 261

124. Coyle, J.T., "Kainic acid: insights into excitatory mechanisms causing selective neuronal degeneration", in Selective Neuronal Death (Ciba Foundation Symposium 126), (Wiley, Chichester, 1987) 186-203. 125. Nadler, J.V., B.W. Perry, and C.W. Cotman, "Intraventricular kainic acid preferentially destroys hippocampal pyramidal cells", Nature, 271 (1978) 676-677. 126. Coyle, J.T., "Neurotoxic Action of Kainic Acid",/. Neurochem., 41 (1983) 1-11. 127. Foster, A.C., E.E. Mena, D.T. Monaghan and C.W. Cotman, "Synaptic localization of kainic acid binding sites", Nature, 289 (1981) 73-75. 128. Ross, S.M., D.N. Roy, and P.S. Spencer, "P-W-Oxalylamino-L-Alanine, Action on Glutamate Receptors",/. Neurochem., 53 (1989) 710-715.

129. Sheardown, M.J., E.0. Nielsen, A.J. Hansen, P. Jacobsen, T. Honore, "2,3-Dihydroxy-6-nitro-7-sulfamoyl-benzo(0quinoxaline: A Neuroprotectant for Cerebral Ischemia", Science, 247 ( 1990) 571-574. 130. Gill, R., A.C. Foster, and G.N. Woodruff, "Systemic Administration of MK-801 Protects Against Ischaemic Neuropathology in Rats", Br. J. Pharmacol., Proceedings Suppl., 91 (1987) 31 IP. 131. Kemp, J.A., A.C. Foster, and Erik H.F. Wong, "Non-competetive antagonists of excitatory amino acid receptors", Trends Neurol Sci., 10 (1987) 294-298. 132. Lehman, J., J.A. Schneider, and M. Williams, "Excitatory Amino Acids and Mammalian CNS Function", Ann. Rep. Med. Chem., 22 (1987) 31-40.

133. Tricklebank, M.D., L. Singh, R.J. Oles, C. Preston, and S.D. Iversen, "The behavioural effects of MK-801: a comparison with antagonists acting non-competitively and competitively at the NMDA receptor", Eur. J. Pharmacol, 167 (1989) 127-135. 134. Guimaraes, F.S., A.P. Carobrez, J.C. DeAguiar, and F.G Graeff, "Anxiolytic effect in the elevated plus-maze of the NMDA AP7 microinjected into the dorsal periaqueductal grey", Psychopharmacol, 103 (1991) 91-94. 135. Patel, S., A.G. Chapman, M.H. Millan and B.S. Meldrum, "Epilepsy and Excitatory Amino Acid Antagonists", in Excitatory Amino Acids in Health and Disease (John Wiley, New York, 1988), 353-378. 136. Meldrum, B., "Possible therapeutic applications of antagonists of excitatory amino acid neurotransmitters", Clinical Science, 68 (1985) 113-122.

137. Lehman, J., J.A. Schneider, and M. Williams, "Excitatoiy Amino Acids and Mammalian CNS Function", Ann. Rep. Med. Chem., 22 (1987) 31-40.

138. Abbott, A., "5-HT3 antagonists and ligands for dopamine D1 and autoreceptors offer new leads for antipsychotic drugs", Trends Pharmacol Sci., 11 (1990) 49-51. 262

139. Komhuber, J., "Glutamate and schizophrenia", Trends Pharmacol Sci., 11 (1990) 357. 140. Turski, L., K. Bressler, K.-J. Rettig, P.-A. Ldschmann, and H. Wachtel, "Protection of substantia nigra from MPP+ neurotoxicity by N-methyl-D-aspartate antagonists", Nature, 349 (1991) 414-418. 141. Sonsalla, P.K., W.J. Nicklas, and R.E. Heikkila, "Role for Excitatory Amino Acids in Methamphetamine-Induced Nigrostriatal Dopaminergic Toxicity", Science, 243 (1989) 398-400. 142. Lodge, D., and K.M. Johnson, "Noncompetetive excitatory amino acid receptor antagonists", Trends Pharmacol. Sci., 11 (1990) 81-86. 143. Monaghan, D.T., R.J. Bridges, C.W. Cotman, "The Excitatory Amino Acid Receptors: Their Classes, Pharmacology, and Distinct Properties in the Function of the Central Nervous System", Annu. Rev. PharmacoL Toxicol, 29 (1989) 365-402. 144. Jones, A.W., D.A.S. Smith and J.C. Watkins, "Structure activity relations of dipeptide antagonists of excitatory amino acids", Neuroscience 13 (1984) 573-581. 145. Davies, J., R.H. Evans, A.W. Jones, D.A.S. Smith and J.C. Watkins, "Differential activation and blockade of excitatory amino acid receptors in the mammalian and amphibian central nervous system", Comp. Biochem. Physiol 72C (1982) 211-224. 146. Shreve, P.E., and N.J. Uretsky, "Role of Quisqualic Acid Receptors in the Hypermotility Response Produced by the Injection of AMPA Into the Nucleus Accumbens", Pharmacology, Biochemistry and Behavior, 30 (1988) 379-384. 147. Shreve, P.E., and N.J. Uretsky, "AMPA, Kainic Acid, and N-Methyl-D-Aspartic Acid Stimulate Locomotor Activity After Injection Into the Substantia Innominata/Lateral Preoptic Area", Pharmacology, Biochemistry and Behavior, 34 (1989) 101-106.

148. Davies, J., and J.C. Watkins, "Depressant Actions of Y-D-GlutamylaminomethylSulfonate (GAMS) on Amino Acid-Induced and Synaptic Excitation in the Cat Spinal Cord", Brain Res., 327 (1985) 113-120.

149. Watkins, J.C., P. Krogsgaard-Larsen and T. Honore, "Structure-activity relationships in the development of excitatory amino acid receptor agonists and competetive antagonists", Trends Pharmacol. Sci., 11 (1990) 25-33. 150. Johnson, G., "Recent Advances in Excitatory Amino Acid Research", Annu. Rep. Med. Chem., 24 (1989) 41-50. 151. Honore, T., S.N. Davies, J. Drejer, E. Fletcher, P. Jacobsen, D. Lodge, F. Nielsen, "Quinoxalinediones: Potent Competetive Non-NMDA Glutamate Receptor Antagonists", Science, 241 (1988) 701-703. 152. Watkins, J.C., P. Krogsgaard-Larsen, and T. Honore, "Structure-activity relationships in the development of excitatory amino acid receptor agonists and competetive antagonists", Trends Pharmacol Sci., 11 (1990) 25-33. 263 153. Blake, J.F., R.G. Yates, M.W. Brown, and G.L. Collingridge, "6-Cyano-7-nitroquinoxaline-2,3-dione as an excitatory amino acid antagonist in area CA1 of rat hippocampus", Br. J. Pharmacol., 97 (1989) 71-76. 154. Frey, P., D. Bemey, P.L. Herrling, W. Mueller, and S. Urwyler, "6,7-Dichloro-3-hydroxy-2-quinoxalinecarboxylic acid is a relatively potent antagonist at NMDA and kainate receptors", Neurosci. Lett., 91 (1988) 194-198. 155. Blake, J.F., M.W. Brown, and G.L. Collingridge, "CNQX blocks acidic amino acid induced depolarizations and synaptic components mediated by non-NMDA receptors in rat hippocampal slices", Neurosci. Lett., 89 (1988) 182-186. 156. Birch, P.J., C.J. Grossman, and A.G. Hayes, "Kynurenate and FG9041 have both competetive and non-competedve antagonist actions at excitatory amino acid receptors", Eur. J. Pharmacol, 151 (1988) 313-315.

157. Kessler, M., T. Terramani, G. Lynch and M. Baudry, "A glycine site associated with N-methyl-D-aspartic acid receptors: characterization and identification of a new class of antagonists", J. Neurochem., 52 (1989) 1319-1328. 158. Mogenson, G.J., D.L. Jones, and C.Y. Yim, "From motivation to action: functional interface between the limbic system and the motor system", Progress in Neurobiol, 14 (1980) 69-97. 159. Wise, R.A., "The role of reward pathways in the development of drug dependence", Pharmacol Ther., 35 (1987) 227-263. 160. Hoffman, D.C., "The use of place conditioning in studying the neuropharmacology of drug reinforcement", Brain Research Bulletin, 23 (1989) 373-387. 161. Swerdlow, N.R., and G.F. Koob, "Separate neural substrates of the locomotor-activating properties of amphetamine, heroin, caffeine, and corticotropin releasing factor (CRF) in the rat", Pharmacol Biochem. and Behav., 23 (1985) 303-307.

162. Shreve, P.E., and N.J. Uretsky, "Effect of GABAergic Transmission in the Subpallidal Region on the Hypermotility Response to the Administration of Excitatory Amino Acids and into the Nucleus Accumbens", Neuropharmacology, 27 (1988) 1271-1277. 163. Donzanti, B.A., and N.J. Uretsky, "Effects of excitatory amino acids on locomotor activity after bilateral microinjection into the rat nucleus accumbens: possible dependence on dopaminergic mechanisms", Neuropharmacology, 22 (1983) 971-981. 164. Donzanti, B.A., and N.J. Uretsky, "Antagonism of the hypermotility response induced by excitatory amino acids in the rat nucleus accumbens", Arch. Pharmacol, 325 (1984) 1-7.

165. Donzanti, B.A., and N.J. Uretsky, "Magnesium selectively inhibits N-methyl-aspartic acid-induced hypermotility after intra-accumbens injection", Pharmacol, Biochem. and Behav., 20 (1984)243-246. 264 166. Wise, R.A., and M.A. Bozarth, "A psychomotor stimulant theory of addiction", Psychol. Rev., 94 (1987) 469-492.

167. Pulvirenti, L., N.R. Swerdlow, and G.F. Koob, "Microinjection of a glutamate antagonist into the nucleus accumbens reduces psychostimulant locomotion in rats", Neurosci. Lett., 103 (1989) 213-218.

168. Krogsgaard-Larsen, P., T. Honore, J.J. Hansen, D.R. Curtis, and D. Lodge, "New class of glutamate agonist structurally related to ibotenic acid", Nature, 284 (1980) 64-66. 169. Boden, P., B.W. Bycroft, S.R. Chhabra, J. Chiplin, P.J. Crowley, R.J. Grout, T.J. King, E. McDonald, P. Rafferty, and P.N.R. Usherwood, "The Action of Natural and Synthetic Isomers of Quisqualic Acid at a Well-Defined Synapse", Brain Research, 385 (1986) 205-211.

170. Hansen, J. J., and P. Krogsgaard-Larsen, "Structural, Conformational, and Stereochemical Requirements of Central Excitatory Amino Acid Receptors", Medicinal Research Reviews, 10 (1990) 55-94. 171. Lauridsen, J., T. Honore, and P. Krogsgaard-Larsen, "Ibotenic Acid Analogues. Synthesis, Molecular Flexibility, and in Vitro Activity of Agonists and Antagonists at Central Glutamic Acid Receptors", J. Med. Chem., 28 (1985) 668-672. 172. Henley, J.M., A. Ambrosini, P. Krogsgaard-Larsen, and E.A. Barnard, The New Biologist, 2 (1989) 153-158. 173. Krogsgaard-Larsen, P., J.W. Ferkany, E.0. Nielsen, U. Madsen, B. Ebert, J.S. Johansen, N.H. Diemer, T. Bruhn, D.T. Beattie, and D.R. Curtis, "Novel Class of Amino Acid Antagonists at Non-N-methyl-D-aspardc Acid Excitatory Amino Acid Receptors. Synthesis, in Vitro and in Vivo Pharmacology, and ", J. Med Chem., 34 (1991) 123-130. 174. Madsen, U., K. Schaumburg, L. Brehm, D.R. Curtis, and P. Krogsgaard-Larsen, "Ibotenic Acid Analogues. Synthesis and Biological Testing of Two Bicyclic 3-Isoxazolol Amino Acids", Acta Chemica Scandinavica, B, 40 (1986) 92-97. 175. Nielsen, E.0., U. Madsen, K. Schaumburg, L. Brehm, and P. Krogsgaard-Larsen, "Studies on receptor-active conformations of excitatory amino acid agonists and antagonists", Eur. J. Med Chem. - Chim. Ther., 21 (1986) 433-437. 176. J.C. Watkins and H.J. Olverman, "Structural Requirements for Activation and Blockade of EAA Receptors", in Excitatory Amino Acids in Health and Disease (John Wiley, New York, 1988), 13-45. 177. Slevin, J.T., J.F. Collins, T. Coyle, "Analog interactions with brain receptor labelled by [3H]-kainic acid", Brain Res., 265 (1983) 169-172.

178. a. Agrawal, S.G., and R.H. Evans, "The primary afferent depolarizing action of kainate in the rat", Br. J. Pharmacol, 87 (1986) 345-355. b. Evans, R.H., A.W. Jones, and J.C. Watkins, J. Physiol (London) 308 (1980) 71P. 265

179. Biscoe, T J., R.H. Evans, P.M. Headley, M.R. Martin, and J.C. Watkins, "Structure-Activity Relations of Excitatory Amino Acids on Frog and Rat Spinal Neurones", Br. J. Pharmacol, 58 (1976) 373-382. 180. Bledsoe, Jr., S.C., D.M. Chihal, R.P. Bobbin, and D.N. Morgan, "Comparative Actions of Glutamate and Related Substances on the Lateral Line of Xenopus Laevis", Comp. Biochem. Physiol, 75C (1983) 199-206. 181. Kiss, T. B. Toth, "Microscopic processes of deprotonation of some tyrosine derivatives", Magy. Kern. Foly., 87(1981) 374-380. 182. Lidak, M.Yu., I.V. Dipan, R.A. Paegle, and Ya.P. Stradyn, "Protolysis of some a-amino-p-(l -pyrimidyl)propionic acids and their analogs", Khimiya Geterotsildicheskikh Soedinenii (1972) 708-712. (English translation, pp. 644-648.

183. Garcia-Sancho, J., A. Sanchez, and H.N. Christensen, "Role of Proton Dissociation in the Transport of Acidic Amino Acids by the Ehrlich Ascites Tumor Cell", Biochimica et. BiophysicaActa, 464 (1977) 295-312. 184. Lipinsld, C.A., and B.L. Chenard, "Acidic Isostere Design: Synthetic Strategies and Recent P ogress in Understanding Electronic Properties and Metabolic Stability", Pestic. Sci., 29 (1990) 227-140. 185. Krogsgaard-Larsen, P., T. Honore, J.J. Hansen, D.R. Curtis, and D. Lodge, "New class of glutamate agonist structurally related to ibotenic acid", Nature, 284 (1980) 64-66. 186. Nielsen, E.0., U. Madsen, K. Schaumburg, L. Brehm, and P. Krogsgaard-Larsen, "Studies on receptor-active conformations of excitatory amino acid agonists and antagonists", Eur. J. Med Chem. - Chim. Ther., 21 (1986) 433-437. 187. Bridges, R.J., M.M. Kadri, D.T. Monaghan, P.B. Nunn, J.C. Watkins, and C.W. Cotman, "Inhibition of [3H]a-amino-3-hydroxy-5-methyl-4-isoxazole- propionic acid binding by die excitotoxin p-N-oxalyl-L-a,p-diaminopropionic acid", Eur. J. Pharmacol, 145 (1988) 357-359. 188. Jacobsen, P., F.E. Nielsen, T. Honore, "Preparation and forumuiations of benzo- and pyridoquinoxalines as excitatory neurotransmitter antagonists", Eur. Pat. Appl. EP 283,959 (1987). 189. Bridges, R.J., M.S. Stanley, M.W. Anderson, C.W. Cotman, and A. Richard Chamberlin, "Conformationally Defined Neurotransmitter Analogues. Selective Inhibition of Glutamate Uptake by One Pyrrolidine-2,4-dicarboxylate Diastereomer", J. Med Chem., 34 (1991) 717-725. 190. Stone, T.W., and J.H. Connick, "Quinolinic Acid and Other in the Central Nervous System", Neuroscience, 15 (1985) 597-617. 191. Personal communication, Dr. Lane J. Wallace, May, 1990.

192. McCord, T.J., J.L. Kreps, J.N. Hubbard, and A.L. Davis, "The Rearrangement of 3 -Amino-3,4-dihydro-1 -hydroxycarbostyril to 3-Amino-3,4-dihydro-6-hydroxycarbostyril",y. Heterocycl Chem., 6(1969)937-939. 266

193. Burckhalter, J.H., and V.C. Stephens, "Synthesis of Phenylalanine Analogs as Antimetabolites", J. Am. Chem. Soc., 73 (1951), 56-58. 194. Dall’Asta, L.and P. Ferrario, "Synthetische Versuche in der Gruppe der cytotoxisch wirksamen Substanzen: Darstellung von d l and L-3-[Di-(p-chlorathyl)-amino]-4-hydroxyphenylalanine", Helv. Chim. Acta, 45 (1962) 1065-1071. 195. Itoh, M., D. Hagiwara, and T. Kamiya, "A New tert-Butyloxycafbonylating Reagent, 2-tert-Butylcarbonyloxyinino-2-phenylacetonitrile", Tet. Lett., 49 (1975) 4393-4394. 196. Bodzansky, The Practice o f Peptide Synthesis, 2nd edit. (Pergamon Press, New York, 1986). 197. Fuson, R.C., Advanced Organic Chemistry (John Wiley & Sons, New York, 1950) p. 524. 198. Martinez, A.P., and Lee, W.W., "An Improved Synthesis of Willardiine and l-(2’,2’-Diethoxyethyl)uracil", J. Org. Chem., 30 (1965) 317-318. 199. Weinstock, J., D.E. Gaitanopoulos, O.D. Stringer, R,G. Franz, J.P. Hieble, L.B. Kinter, L.B., W.A. Mann, K.E. Flaim, and G. Gessner, "Synthesis and Evaluation of Non-Catechol D-l and D-2 Dopanine Receptor Agonists: Benzimidazol-2-one, Benzoxazol-2-one, and the Highly Potent Benothiazol-2-one 7-Ethylamines", J. Med. Chem., 30 (1987) 1166-1176. 200. March, J., Advanced Organic Chemistry: Reactions, Mechanisms, and Structure (John Wiley & Sons, New York, 1985) pp. 386-387. 201. Olah, G.A., J. Shih, and G.K. Surya Prakash, "Preparation of a-Bromo- and a-Chlorocarboxylic Acids from a-Amino Acids", Helv. Chim. Acta, 66 (1983) 1028-1030.

202. Derome, A.E., Modem NMR Techniques for Chemistry Research (Pergamon Press, New York, 1987) p. 254. 203. Kessler, H., C. Griesinger, J. Zarbock, and H.R. Looslie, J. Magn. Res., 57 (1989) 331-336. 204. Duddeck, H., and W. Dietrich, Structure Elucidation by Modem NMR - A Workbook (Springer-Verlag, New York, 1989) pp. 25-28. 205. Kuhn, S.J., and G.A. Olah, "Aromatic Substitution. VII. Friedel-Crafts Type Nitration of Aromatics", J. Amer. Chem. Soc., 83 (1961) 4564.

206. Guebitz, G. F. Juffman, W. Jellenz, "Direct Separation of amino acid enantiomers by high performance ligand exchange chromatography on chemically-bonded chiral phases", Chromatographia, 16 (1982) 103-106. 267 207. Gamier-Suillerot, A., J.P. Albertini, A. Collet, L.Faury, J.M. Pastor, L.Tosi, "Spectrophotomoetric Studies of the Copper(E)-D-o-Tyrosine Complex. Assignment of the 330-nm Dichroic Band in Copper(H) and Iron(m) Transferrins", J. Chem. Soc. Dalt. Trans. (1981) 2544-2549. 208. Arnold, L.D., R.G. May, J.C. Vederas, "Synthesis of Optically-Pure a-Amino Acids via Salts of a-Amino-p-propiolactone", J. Amer. Chem. Soc., 110 (1988) 2237-2241. 209. Bajgrowicz, J.A., A. A1 Hallaoui, R. Jacquier, Ch. Pigiere, Ph. Viallefont, "Les Organocuprates Dans Une Nouvelle Synthese D’Aminoacides Enantiomeriquement Purs", Tetradhedron, 41 (1985) 1833-1843. 210. Baldwin, J.E., R.M. Adlington, and N.G. Robinson, "Nucleophilic Ring Opening of Aziridine-2-carboxylates with Wittig Reagents; An Enantioefficient Synthesis of Unsaturated Amino Acids", J. Chem. Soc., Chem. Commun. (1987) 153-155.

211. Belokon, Y.N., A.S. Sagyan, S.M. Djamgaryan, V.I. Bakhmutov and V.M. Belikov, "Asymmetric Synthesis of P-Substituted a-Amino Acids via a Chiral Nin Complex of Dehydroalanine", Tetrahedron, 44 (1988) 5507-5514. 212. Haemers, A., L. Mishra, I. Van Assche, and W. Bollaert, "Asymmetric synthesis of amino acids by enantio- and diastereodifferendating reactions", Pharmazie, 44 (1989) 97-109. 213. Williams, R.M., P.J. Sinclair, D. Zhai, and D. Chen, "Practical Asymmetric Syntheses of a-Amino Acids through Carbon-Carbon Bond Constructions on Electrophilic Glycine Templates", J. Amer. Chem. Soc., 110 (1988) 1547-1557. 214. Schollkopf, U., "Asymmetric Syntheses of Amino Acids via Metalated Bis-Lactim Ethers of 2,5-Diketopiperazines", Pure Appl Chem., 55 (1983) 1799-1806.

215. Maurer, P.J., H. Takahata, H. Rapoport, "a-Amino Acids as Chiral Educts for Asymmetric Products. A General Synthesis of D-a-Amino Acids from L-Serine." J. Amer. Chem. Soc., 106 (1984) 1095-1098. 216. Dewar, J.H., and G. Shaw, "Purines, Pyrimidines, and Imidazoles. Part XVII. A Synthesis of Willardiine", J. Chem. Soc. (London), (1962) 583-585.

217. Murakoshi, I., F. Ikegami, Y. Yoneda, H. Ihara, K. Sakata, and C. Koide, "Biomimetic Synthesis of Heterocyclic p-Substituted by Pyridoxal 5’-Phosphate-Catalyzed Chemical Reactions", Chem. Pharm. Bull., 34 (1986) 1473-1478. 218. For a summary of the N-alkylation of , see D.J. Brown, The Pyrimidines, Interscience Publishers, Inc., New York, 1962. 219. Martinez, A.P., W.W. Lee, and L. Goodman, "DL-Willardiine Mustard", J. Med. Chem., 11 (1968) 60-62. 220. Huang, Guang-Fu, and P.F. Torrence, "Nitration of Pyrimidine Bases and Nucleosides by Nitronium Tetrafluoroborate. Synthesis of 5-Nitro-2’-deoxyuridine", J. Org. Chem., 42 (1977) 3821-3824. 268 221. Kadin, S.B., R. Sarges, "Triazoloquinoxalines as and antifatigue agents", Eur. Pat. Appl. EP 107455 A l, 2 May 1984. (CA, 101(19):171285e) 222. St. Clair, R.L., T.D. Thibault, U.S. Patent 3,992,378 (Eli Lilly & Co., 16 Nov. 1976). (Referenced in CA, 56:89890m)

223. Okofor, C.O., "Studies in the Heterocyclic Series XV. Synthesis and Reactions of the First Triazophenoxazine Ring System", J. Heterocyclic Chem., 16 (1979) 1025. 224. Stevens, J.R., K. Pfister, 3rd, and F.J. Wolf, "Substituted Sulfaquinoxalines. I. The Isolation and Synthesis of 3-Hydroxy-2-sulfanilamidoquionoxaline and of Related Quinoxalines", J. Amer. Chem. Soc., 68 (1946) 1035-1039.

225. Cheeseman, G.W.H., "Quinoxalines and Related Compounds. Part VI. Substitution of 2,3-Dihydroxyquinoxaline and its 1,4-Dimethyl Derivative", J. Chem. Soc. (1962) 1170-1176. 226. Suschitzky, H., "The Balz-Schiemann Reaction", in Adv. Fluorine Chem., 4 (1965) 1-30. 227. a) Swain, C.G., J.E. Sheats, D.G. Gorenstein, and K.G. Harbison, "Mechanism of Formation of Aryl Fluorides from Arenediazonium Fluoborates", J. Amer. Chem. Soc., 97 (1975) 799-800. b) [same authors], "Aromatic Hydrogen Isotope Effects in Reactions of Benzenediazonium Salts", Ibid., 791-795. c) Swain, C.G., J.E. Sheats, and K.G. Harbison, "Nitrogen Isotope Effects in th Hydrolysis of Benzenediazonium Salts", Ibid 1,796-798. d) Swain, C.G., J.E. Sheats, and K.G. Harbison, "Evidence for Phenyl Cation as an Intermediate in Reactions of Benzendiazonium Salts in Solution", Ibid. 783-790.

228. Piercey, M.R., and E.R. Ward, "The Preparation of Pure Solid Diazonium Sulphates", J. Chem. Soc. (1962) 3841-3843. 229. Ross, J.M. (E.I. DuPont de Nemours & Co.), U.S. Pat. 3,296267 (Jan 3., 1967). Referenced (with m.p.) in Chemical Abstracts, 66 (1967) 66739s. See also refs. 247 and 248.

230. Honore, T., P. Jacobsen, F. Elmelund, L. Naerum, (Ferrosan) "Preparation and testing of 2,3 (lH,4H)-quinoxalinediones as neuroleptics", Eur. Pat. Appl. EP 315,959 (17 May 1989). 231. Sastry, C.V.R., M. Jogibhukta, V.S.H. Krishnan, K.S. Rao, K. Vemana, D.R. Shridhar, R.M. Tripathi, R.K. Verman, R. Kaushal, Indian J. Chem., Sect. B., 27B (1988) 1110- 1112. 232. Cheeseman, G.H.W., "Quinoxalines and Related Compounds. Part V. Some Experiments with l,2-Dihydro-2-oxoquinoxalines", J. Chem. Soc. (1961) 1246-1251.

233. a. Cheeseman, G.W.H., "Quinoxalines and Related Compounds. Part IV. The Fine Structure of the 2- and 3- Hydroxyquinoxalines and 2-Amino- and 2-Mercapto-quinoxaline", J. Chem. Soc. (1958) 108-113. See also: Newbold and Spring, J. Chem. Soc. (London) (1948) 519. b. Cheeseman, G.W.H., A.R. Katritzky, and S. 0ksne, "The Infrared Spectra of Polycyclic Heteroaromatic Compounds. Part n. Quinoxaline-2,3-diones." J. Chem. Soc. (1961) 3983-3985. 269

234. Kazimierczuk, Z., and W. Pfleiderer, "Synthese and Eigenschaften von l-(p-D-Ribofuranosyl)-2-chinoxalinonen und -2,3-chinoxalindionen", Liebigs Ann. Chem. (1982) 754-761. 235. Birkofer, L., A. Ritter and H-P. Kiihlthau, "Alkylierungen and Glykosidierungen fiber Silyl-Derivate", Chem. Ber., 97 (1964) 934-945. 236. Weinges, K., K.-P. Klotz, and H. Droste, "Optisch aktiv 5-Amino-4-phenyl-l,3-dioxane und deren Einflu(3 auf die Stereoselekdvitat der asynunetrischen Strecker-Synthese", Chem. Ber., 113 (1980) 710-721.

237. Cheeseman, G.W.H., "Quinoxalines and Related Compounds. Part VI. Substitution of 2,3-Dihydroxyquinoxaline and its 1,4-Dimethyl Derivative", J. Chem. Soc. (1962) 1170-1176.

238. Albert, A., and E.P. Seijeant, The Determination o f Ionization Constants - A Laboratory Manual , 3rd. edit. (Chapman and Hall, New York, 1984).

239. Metzger, C.M., and D.M. Weiner, Program NONLIN84, Version V02-A), Statistical Consultants, Inc., 1984. 240. Dean, J.A., Lange’s Handbook o f Chemistry, 12th edit. (McGraw-Hill, New York, 1979) pp. 5.12-5.41. 241. Hansch, C., A. Leo, S.H. Unger, K.H. Kim, D. Nikaitani, and E.J. Lien, "‘Aromatic’ Substituent Constants for Structure-Activity Correlations", J. Med. Chem., 16 (1973) 1207-1216, and references therein. 242. Martinez, A.P., W.W. Lee, and L. Goodman, "DL-Willardiine Mustard", J. Med. Chem., 11 (1968) 60-62.

243. Agrawal, S.G., and R.H. Evans, "The primary afferent depolarizing action of kainate in the rat", Br. J. Pharmacol, 87 (1986) 345-355.

244. Walker, R. J., and C. Jane Roberts, "Neuropharmacological Studies on the Receptors Mediating Responses to Carbachol, Amino Acids and Octopamine on Limulus and Hirudo Central Neurons", Comp. Biochem. Physiol., 77C (1984) 371-380. 245. Saito, I., and T. Matsuura, "Chemical Studies on Riboflavin and Related Compounds. I. Oxidation of Quinoxaline-2,3-diols as a Possible Model for the Biological Decomposition of Riboflavin", Biochemistry, 6 (1967) 3602-3608.

246. Flavins and flavoproteins: proceedings of the Eight International Symposium (Brighton, England, July 9-13,1984). 247. Gilbert, E.E., "Perhaloketones XIX (1). Derivatives of Benzo-N-heterocycles", J. Heterocyclic Chem., 6 (1969) 483-490. 248. Kleb, K.G., E. Siegel, and K. Sasse, "New Reactive Dyes", Angew. Chem. internal. Edit., 3 (1964) 408-416. 270

249. Lehmann, J., J.C.R. Randle, and I J. Reynolds, "Excitatory amino acid receptors: NMDA modulatory sites, kainate cloned, and a new role in AIDS", Trends Pharmacol Sci., 11 (1990) 1-3. 250. Patel, J., W.C. Zinkand, A.B. Klika, T.J. Mangano, R.A. Keith, and A.I. Salama, "6,7-Dinitroquinoxaline-2,3-Dione Blocks the Cytotoxicity of N-Methyl-D-Aspartate and Kainate, but Not Quisqualate, in Cortical Cultures", J. Neurochem., 55 (1990) 114-121. 251. Olney, J.W., J. Labruyere, M.T. Price, "Pathological Changes Induced in Cerebrocoxtical Neurons by Phencyclidine and Related Drugs", Science, 244 (1989) 1360-1362.

252. Hansen, J.J., J. Laundsen, E. Nielsen, and P. Krogsgaard-Larsen, "Enzymic Resolution and Binding to Rat Brain Membranes of the Glutamic Acid Agonist a-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic Acid", J. Med Chem., 26 (1983) 901-903. 253. Barbour, B., H. Brew and D. Attwell, "Electrogenic glutamate uptake in glial cells is activated by intracellular potassium", Nature, 335 (1988) 433-435. 254. Birch, P.J., C.J. Grossman and A.G. Hayes, "Antagonist profile of 6,7-dichloro-3-hydroxy-2-quinoxalinecarboxylate at excitatory amino acid receptors in the neonatal rat spinal cord", Eur. J. Pharmacol, 163 (1989) 127-131. 256. Benet, L.Z., and J.E. Goyan, "Potentiometric Determination of Dissociation Constants", J. Pharm. Sci., 56 (1967) 665-680. 257. Ismail, I.A., D.E. Sharp, and M.R. Chedekel, "Synthesis of Benzothiazoles. a-Amino-(4-hydroxy-6-benzothiazolyl)propionic Acid", J. Org. Chem., 45 (1980) 2243-2246.

258. m.p. 345-347 (dec.): Asai, M., "Chemotherapeutics. VI. Oxidation of quinoxaline with nitric acid", Yakugaku Zasshi, 79 (1959) 260-262. 260. Sastry, C.V.R., Rao, K.S., Krishnan, V.S.H., K. Rastogi, M.L. Jain, Indian J. Chem., Sect. B, 28B (1989) 888-890. 262. Aminoff, M.J., "Pharmacologic Management of Parkinsonism and Other Movement Disorders", in Basic and Clinical Pharmacology, 3rd edit., B.G. Katzung, ed. (Appleton and Lange, Norwalk, CN, 1987) pp. 306-314. 263. Neuroleptics: Neurochemical, Behavioral and Clinical Perspectives, J.T. Coyle, S.J. Enna (eds), (Raven Press, New York, 1983).

264. Miller, D.D., and N.J. Uretsky, "Dopaminergic Agonists and CNS Receptors", Advances in CNS Drug-Receptor Interactions, 1 (1991) 75-129. 265. Abbott, A., "5-HT3 antagonists and ligands for dopamine D1 and autoreceptors offer new leads for antipsychotic drugs", Trends Pharmacol Sci., 11 (1990) 49-51. 271 266. Farde, L., F-A. Wiesel, C. Halldin and G. Sedvall, "Central D2-dopamine receptor occupancy in patients treated with antipsychotic drugs", Arch. Gen. Psychiatry, 45 (1988) 1-78. 267. Abou-Gharbia, M., and J.A. Moyer, "Novel Antipsychotic Agents", Annu. Rep. Med. Chem., 25 (1990) 1-10. 268. Horn, P.T., and M.B. Murphy, "New Dopamine Receptor Agonists in Heart Failure and Hypertension - Implications for Future Therapy", Drugs, 40 (1990) 487-492. 269. Friedman, E. G. Gianutosos, and J. Kuster, J. Pharmacol Exp. Ther., 7 (1983) 226.

270. Wise, R.A., and M.A. Bozarth, "A psychomotor stimulant theory of addiction", Psychol Rev., 94 (1987) 469-492. 271. Lewin, D.I., "Dopamine Receptors: Five and Counting", J. NIH Research, 3 (1991) 58-59. 272. Snyder, S.H., "The dopamine connection", Nature, 347 (1990) 121-122. 273. Jean, J.C., L.D. Wise, T.G. Heffner, T.A. Pugsley, and L.T. Meltzer, "Dopamine Autoreceptor Agonists as Potential Antipsychotics. 2. (Aminoalkoxy)-4//-l-benzopyran-4-ones", J. Med. Chem., 34 (1991) 248-256.

274. Sokoloff, P., B. Giros, M.-P. Martres, M.-L. Bouthenet, and J.-C. Schwartz, "Molecular cloning and characterization of a novel dopamine receptor (D3) as a target for neuroleptics", Nature, 347 ( 1990) 146-151.

275. Van Tol, H.H.M., J.R. Bunzow, H.-C. Guan, R.K. Sunahara, P. Seeman, H.B. Niznik, and O. Civelli, "Cloning of the gene for a human dopamine D4 receptor with high affinity for the antipsychotic clozapine", Nature, 350 (1991) 610-614.

276. Sunahara, R.K., H.-C. Guan, B.F. O’Dowd, P. Seeman, L.G. Laurier, G. Ng, S.R. George, J. Torchia, H.H.M. Van Tol, and H.B. Niznik, "Cloning of the gene for a human dopamine D5 receptor with higher affinity for dopamine than Dj", Nature, 350 (1991) 614-619. 277. Miller, D.D., M. Harrold, R.A. Wallace, L.J. Wallace, and N.J. Uretsky, "Dopaminergic drugs in the cationic form interact with D2 dopamine receptors", Trends Pharmacol. Sci., 9 (1988) 282-284. 278. Chang, Y.-A., J. Ares, K. Anderson, B. Sabol, R.A. Wallace, T. Farooqui, N. Uretsky, and D.D. Miller, "Dopamine Agonists: Effects of Charged and Uncharged Analogues of Dopamine", J. Med. Chem., 30 (1987) 214-218. 279. Anderson, K., A. Kuruvilla, N. Uretsky, and D.D. Miller, "Synthesis and Pharmacological Evaluation of Sulfonium Analogues of Dopamine: Nonclassical Dopamine Agonists", J. Med. Chem., 24 (1981) 683-687. 272 280. Farooqui, T., R. A. Wallace, D.D. Miller, and N.J. Uretsky, "Effect of the Inhibition of Dopamine Uptake on the Dopamine- and Dimethyldopamine-induced Inhibition of the Potassium-Evoked Release of [3H] Acetylcholine from Striatal Slices", Gen. Pharmac., 20 (1989) 731-736. 281. Wallace, R.A., T. Farooqui, L. Wallace, J. Ares, Y.A. Chang, D. Miller and N. Uretsky, "Interaction of Permanently Charged Analogs of Dopamine with the D-2 Dopaminergic Receptor", Biochem. Pharmacol, 36 (1987) 3903-3910.

282. Wallace, R.A., T. Farooqui, L. Wallace, J. Ares, Y.A. Chang, D. Miller and N. Uretsky, "Interaction of Permanently Uncharged Dopamine Analogs with die D-2 Dopaminergic Receptor", Biochem. Pharmacol, 37 (1988) 2077-2084. 283. Miller, D.D., and N.J. Uretsky, "Advances in dopaminergic drug research", Pharmacy International, 7 (1986) 95-98. 284. Miller, D.D., M. Harrold, R.A. Wallace, L.J. Wallace, and N.J. Uretsky, "Dopaminergic drugs in the cationic form interact with D2 dopamine receptors", Trends Pharmacol Sci., 9 (1988) 282-284. 285. Wallace, R.A., L. Wallace, M. Harrold, D. Miller, and N.J. Uretsky, "Interaction of Permanently Charged Chlorpromazine and Dopamine Analogs with the Striatal D-l Dopaminergic Receptor", Biochem. Pharmacol, 38 (1989) 2019-2025. 286. Horn, A.S., and S.H. Snyder, Proc. Natl Aca. Sci. U.S.A., 68 (1971) 2325. 287. Harrold, M.W., Y.-A. Chang, R.A. Wallace, T. Farooqui, L.J. Wallace, N. Uretsky, and D.D. Miller, "Charged Analogues of Chlorpromazine as Dopamine Antagonists", J. Med. Chem., 30(1987) 1631-1635. 288. Harrold, M.W., R. A. Wallace, T. Farooqui, L.J, Wallace, N. Uretsky, and D.D. Miller, "Synthesis and D2 Dopaminergic Activity of Pyrrolidinium, Tetrahydrothiophenium, and Tetrahydrothiophene Analogs of Sulpiride", J. Med. Chem., 32 (1989) 874-880.

289. Harrold, M.W., R.A. Wallace, T. Farooqui, L.J. Wallace, N. Uretsky, and D.D. Miller, "Synthesis and D2 Dopaminergic Activity of Pyrrolidinium, Tetrahydrothiophenium, and Tetrahydrothiophene Analogs of Sulpiride", J. Med. Chem., 32 (1989) 874-880. 290. Arndt, J., J. Hyttel, E. Meier, Europ. J. Pharmacol, 155 (1988) 37. 291. Hess, E.J., G. Battaglia, A.B. Norman, I. Creese, Mol Pharmacol., 31 (1987) 50.

292. Williamson, R.A., and P.G. Strange, "Evidence for the Importance of a Carboxyl Group in the Binding of Ligands to the D2 Dopamine Receptor", J. Neurochem., 55 (1990) 1357-1365. 293. Bunzow, J.R., H.H.M. Van Tol, D.K. Grandy, P. Albert, J. Solon, M. Christie, C. A. Machida, K. A. Neve, and O. Civelli, "Cloning and expression of a rat D2 dopamine receptor cDNA", Nature, 336 (1988) 783-785. 273 294. Strader, C.D., I.S. Sigal, and R.A.F. Dixon, "Genetic approaches to the determination of sturcture-fiinction relationships of G protein-coupled receptors", Trends in Pharmacol Sci., 10 (1989) December 1989 supplement, 26-30. See also J. Biol. Chem., 263 (1988) 10267-10271.

295. Strader, C.D., I.S. Sigal, and R.A.F. Dixon, "Structural basis of p-adrenergic receptor function", FASEBJ., 3 (1989), 1825-1832. 296. Hansch, C., A. Leo, S.H. Unger, K.H. Kim, D. Nikaitani, and E.J. Lien, "‘Aromatic’ Substituent Constants for Structure-Activity Correlations", J. Med. Chem., 16 (1973) 1207-1216. 297. Hulme, B.C., N.J.M. Birdsall, and N.J. Buckley, "Muscarinic Receptor Subtypes", Annu. Rev. Pharmacol, 30 (1990) 633-673. 298. Hulme, B.C., C.A.M. Curtis, M. Wheatley, A. Aitken, and A.C. Harris, "Localization and structure of the muscarinic receptor ligand binding site", Trends Pharmacol Sci., 10 (1989) December suppl., pp. 22-25. 299. Kilpatrick, G. J., N. El Tayar, H. Van de Waterbeend, P. Jenner, B. Testa, and C.D. Marsen, "The thermodynamics of agonist and antagonist binding to dopamine D2 receptors., Mol. Pharmacol., 30 (1986) 226-234. 300. Miklavc, A., D. Kocjan, J. Mavri, J. Koller, and D. Hadii, "On the Fundamental Difference in the Thermodynamics of Agonist and Antagonist Interactions with p-Adrenergic Receptors and die Mechanism of Entropy-Driven Binding", Biochem. Pharmacol, 40 (1990) 663-669. 301. Burgen, A.S.V., "Some considerations of receptor specificity", Trends Pharmacol. Sci., 10 (1989) December suppl., pp. 1-3. 302. Zimmerman, J.J., and S. Feldman, "Physical-Chemical Properties and Biologic Activity", in Principles of Medicinal Chemistry, ed. W.O. Foye (Lea and Febiger, Philadelphia, 1989).

303. Dhaenens, M., L. Lacombe, J.-M. Lehn, J.-P. Vigneron, "Binding of Acetylcholine and other Molecular Cations by a Macrocyclic Receptor Molecule of Speleand Type", J. Chem. Soc. Chem. Commun. (1984) 1097-1099. 304. Schneider, H.-J., D. Guttes, U. Schneider, "Host-Guest Complexes with Water-Soluble Macrocyclic Polyphenolates Including Induced Fit and Simple Elements of a Proton Pump", J. Am. Chem. Soc., 110 (1988) 6449-6454. 305. Behr, J.-P., J.-M. Lehn and P. Vierling, "182. Molecular Receptors. Structural Effects and Substrate Recognition in Binding of Organic and Biogenic Ammonium Ions by Chiral Polyfunctional Macrocyclic Polyethers Bearing Amino-Acid and Other Side Chains", Helv. Chim. Acta, 65 (1982) 1853-1867. 306. Cohen, S.G., D.L. Lieberman, F.B. Hasan, J.B. Cohen," 1-Bromopinacolone, an active-site-directed covalent inhibitor for acetylcholinesterase", J. Biol Chem., 257 (1982) 14087-14092. 274

307. Hehre, W.J., L. Radom, P.v.R. Schleyer, J.A. Pople, Ab Initio Molecular Orbital Theory (John Wiley, New York, 1986).

308. Parr, R.G., and W. Yang, Density Functional Theory o f Atoms and Molecules (Oxford University Press, New York, 1989). 309. Koller, J., and D. Hadii, "MNDO/H and AMI Calculations on the Formic Acid-Methylamine Complexes", J. Mol Struct. (THEOCHEM), 200(1989)533-541. 310. Khalil, M., R.J. Woods, D.F. Weaver, and V.H. Smith, Jr., "An Examination of /n/ermolecular and /n/ramolecular Hydrogen Bonding in Biomolecules by AMI and MNDO/M Semiempirical Molecular Orbital Studies", J. Comput. Chem., 12 (1991) 584-593.

311. Vos, R.J., R. Hendriks, and F.B. van Duijneveldt, "SCF, MP2, and CEPA-1 Calculations on the OH -0 Hydrogen Bonded Complexes (H20)2 and (H20- H2CO)", J. Comput. Chem., 11 (1990) 1-18. 312. Goldblum, A., J. Comp. Chem., 8 (1987) 835. 313. Buemi, G., F. Zuccarello, and A. Raudino, J. Molec. Struct. (THEOCHEM), 164 (1988) 379. 314. Del Bene, J.E., "Molecular orbital studies of the complexes (AHL)2H+ formed form ammonia, water, hydrogen fluoride, phosphine, hydrogen sulfide, ana hydrogen chloride", J. Phys. Chem., 89 (1985) 3669-3674. 315. Hehre, W.J., L. Radom, P.v.R. Schleyer, J.A. Pople, Ab Initio Molecular Orbital Theory (John Wiley, New York, 1986) pp. 215-223. 316. Del Bene, J.E., "An Ab Initio Molecular Orbital Study of the Structures and Energies of Neutral and Charged Bimolecular Complexes of NH3 with the Hydrides AH„ (A = N, O, F, P, S, and Cl)", J. Comp. Chem. 10 (1989) 603-615.

317. Parr, R.G., and W. Yang, Density Functional Theory of Atoms and Molecules

(Oxford University Press, New York, 1989), Chapter 1.

318. Szabo, A., and N.S. Ostlund, Modem Quantum Chemistry (Macmillan Press, New York, 1982). 319. Levine, I.N., Quantum Chemistry, 3rd edit. (Allyn and Bacon, Inc., Boston, 1983), Chapters 13 and 15. 320. Roothan, C.C.J., "New developments in molecular orbital theory", Rev. Mod. Phys., 23 (1951) 69-89. 322. SzczQsniak, M.M., and S. Scheiner, "Moller-Plesset treatment of electron correlation effects in (HOHOH)'", J. Chem. Phys., 77 (1982) 4586-4593. 275 323. Amos, R.D., and J.E. Rice, CADPAC 4 - Cambridge Analytic Derviative Package, issue 4.0 for Cray X-MP, Cambridge, UK, 1987. 324. Program SYBYL, version 4.1, Tripos Associates, Inc., St. Louis, MO, USA. 325. Clark, M., R.D. Cramer HI, and N. Van Opdenbosch, J. Comput. Chem., 10 (1989) 982-1012.

326. Alchemy II, Tripos Associates, Inc., St. Louis, MO, USA. 327. c.f. W.H. Press, B.P. Flannery, S.A. Teukolsky, W.T. Vetterling, Numerical Recipies, the Art o f Scientific Computing (Cambridge University Press, New York, 1986). 328. Clark, T., J. Chandrasekhar, G. Spitznagel, and P.v.R. Schleyer, "Efficient Diffuse Function-Augmented Basis Sets for Anion Calculations. III. The 3-21+G Basis Set for First-Row Elements, Li-F", J. Comp. Chem., 4 (1983) 294-301. 329. Hehre, W.J., L. Radom, P.v.R. Schleyer, J.A. Pople, Ab Initio Molecular Orbital Theory (John Wiley, New York, 1986). 330. Mulliken, R.S., "Molecular Complexes and Their Spectra. VI. Some Problems and New Developments", Reel Trav. Chim. Pays-Bays, 75 (1956) 845-852. 331. Martin, Y.C., Quantitative Drug Design, A Critical Introduction (Marcel Dekkar, Inc., New York, 1978) Chapt. 3, pp. 33-58. 332. Bartmess, J.E., and R.T. Mclver, Jr., "The gas-phase acidity scale", in Gas Phase Ion Chemistry, ed. M.T. Bowers (Academic Press, NY, 1979), pp. 87-121. 333. Lias, S.G., J.F. Liebman, R.D. Levin, "Evaluated gas phase basicities and proton affinities of molecules; heats of formation of protonated molecules" J. Phys. Chem. Ref. Data, 13 (1984) 695-808. 334. a. Dixon, D.A., and S.G. Lias, "Absolute Values of Gas Phase Proton Affinities and Basicities of Molecules: A Comparison Between Theory and Experiment", Molecular Structure and Energetics, 2 (1987) 269-313.

335. HadoSdek, M. and D. Hadii, "Proton transfer in the HCOOH-CH 3NH2 complex. Ab initio study with various basis sets and solvent reaction field", J. Mol Struct. (THEOCHEM), 198 (1989) 461-473. 336. Del Bene, J.E., M.J. Frisch, K. Raghavachari, and J.A. Pople, "Molecular Orbital Study of Some Protonated Bases", J. Phys. Chem., 86 (1982) 1529.

337. Hehre, W.J., L. Radom, P.v.R. Schleyer, J.A. Pople, Ab Initio Molecular Orbital Theory (John Wiley, New York, 1986) pp. 310-321. 338. Dunlap, B.I., in Density Functional Methods in Chemistry, J.K. Labanowski and J.W. Andzelm, eds. (Springer Verlag, New York, 1991) p. 2.

339. Teter, M.P., "Researchers at Coming Glass Make Breakthroughs in Modelling Quantum Chemistry", Chemical Design and Automation News, 5 (1990) p.l. 276

340. Andzelm, J., J. Mertz, personal communication.

341. Wimmer, E., "Density Functional Theory for Solids, Surfaces, and Molecules: From Energy Bands to Molecular Bonds", in Density Functional Methods in Chemistry, J.K. Labanowski and J.W. Andzelm, eds. (Springer Verlag, New York, 1991) p. 7-31.

342. Manual for program DMol, Version 2.0, BIOSYM Technologies, San Diego, CA, USA, 1990), Chapter 1. 343. Kohn, W., and L.J. Sham, "Self-Consistent Equations Including Exchange and Correlation Effects", Phys. Rev., Series A, 140 (1965) 1133-1138. 344. Borman, S., "Density-Functional Theory Gams Following Among Chemists", Chemical and Engineering News (April 9,1990), 22-30. 345. Dixon, D.A., J. Andzelm, G. Fitzgerald, E. Wimmer, and P. Jasien, "Benchmark and Testing of the Local Density Functional Method for Molecular Systems", in Density Functional Methods in Chemistry, J.K. Labanowski and J.W. Andzelm, eds. (Springer Verlag, New York, 1991) pp. 33-48. 346. Levy, M., "Electron densities in search of Hamiltonians", Phys. Rev. A, 26 (1982) 1200-1208. 347. Parr, R.G., and W. Yang, Density Functional Theory of Atoms and Molecules (Oxford University Press, New York, 1989), Appendix E. 348. von Barth, U., and L. Hedin, "Local exchange-correlation potential for the spin-polarized case. I" J. Phys., C, 5 (1972) 1629-1642. 349. Vosko, S.H., L. Wilk, and M. Nusair, "Accurate spin-dependent electron liquid correlation energies for local spin density calculations: a critical analysis" Can. J. Phys., 58(1980) 1200-1211.

350. Becke, A.D., "Correlation energy of an inhomogenous electron gas: a coordinate-space model", J. Chem. Phys., 88 (1988) 1053-1062.

351. Becke, A.D., "Density-functional exchange-energy approximation with correct asymptotoic behavior", Phys. Rev. A, 38 (1988) 3098-3109. 352. Becke, A.D., in: The Challenge o fd and f Electrons: Theory and Computations, eds. D.R. Salahub, M. Zemer, ACS Symposium Series No. 394 (American Chemical Society, Washington, D.C., USA, 1989), p. 166. 353. Perdew, J.P., "Density-functional approximation for the correlation energy of the inhomogeneous electron gas." Phys. Rev. B, 33 (1986) 8822-8824.

354. Stoll, H., C.M.E. Pavlidou, and J. Preuss, "On the calculation of correlation energies in the spm-density functional formulation", Theor. Chim. Acta, 49 (1978) 143-149. 355. von Nagy-Felsobuki, E.I., and K. Kimura, "Ab Initio Proton Affinities of HCOOH, CH3COOH, CH3OH, and QHsOH", J. Phys. Chem., 94 (1990) 8041-8044. 277

356. Moller, C., and M.S. Plesset, "Approximate treatment for many-electron systems", Phys. Rev., 46 (1934) 618-622.

357. Pople, J.A., J.S. Binkley, and R. Seeger, "Theoretical Models incorporating electron correlation", Int. J. Quantum Chem. Symp., 10 (1976) 1. 358. GAUSSIAN 90, Revision H, M.J. Frisch, M. Head-Gordon, K. Raghavachari, M. Robb, J.S. Binkley, C. Gonzalez, C.F. Melius, J. Baker, R.L. Martin, L.R. Kahn, J.J.P. Stewart, S. Topiol, and J.A. Pople, Gaussian, Inc., Pittsburgh, PA, 1990. 359. Program DMol, Version 2.0, BIOSYM Technologies, Inc., San Diego, CA, USA, 1990.

360. Delley, B., "DMol Methodology and Applications", in Density Functional Methods in Chemistry, J.K. Labanowski and J.W. Andzelm, eds. (Springer Veriag, New York, 1991) p. 101-107. 361. Andzelm, Jan W., Cray Research, Inc., Mendota Heights, MN, USA 55120

362. Andzelm, J. "DGauss: D en sity Functional - Gaussian Approach," in Density Functional Methods in Chemistry, J.K. Labanowski and J.W. Andzelm, eds. (Springer Veriag, New York, 1991) p. 155-174.

363. HadoSdek, M. and D. Hadii, "Proton transfer in the HCOOH-CH 3NH2 complex. Ab initio study with various basis sets and solvent reaction field", J. Mol. Struct. (THEOCHEM), 198 (1989) 461^73.

364. Salahub, D.R., R. Fournier, P. Mlynarski, I. Papai, A. St-Amant, and J. Ushio, "Gaussian-based Density Functional Methodology, Software, and Applications", in Density Functional Methods in Chemistry, J.K. Labanowski and J.W. Andzelm, eds. (Springer Veriag, New York, 1991) p. 77-100.

365. McQuarrie, D.A., Statistical Mechanics (Harper & Row, New York, 1976) p. 201. 366. Hehre, W.J., L. Radom, P.v.R. Schleyer, J.A. Pople, Ab Initio Molecular Orbital Theory (John Wiley, New York, 1986) p. 260 (see pp. 226-260). 367. Vos, R.J., R. Hendriks, and F.B. van Duijneveldt, "SCF, MP2, and CEPA-1 Calculations on the OH..O Hydrogen Bonded Complexes (H 20 )2 and (H20-H 2C 0)", J. Comp. Chem., 11 (1990) 1-18.

368. Szczqsniak, M.M., and S. Scheiner, Moller-Plesset treatment of electron correlation effects in (HOHOH) ",J. Chem. Phys., 77(1982) 4586-4593.

369. a. Hadzi, D., "Quantum chemical modeling of some biological mechanisms involving proton transfer", Pure & Appl. Chem., 61 (1989) 2175-2183. b. Hadzi, D., "Proton Transfers in Biological Mechanisms", J. Mol. Struct., 177 (1988) 1-21. 370. Koller, J., M. HadoSdek, and D. Hadli, "An Ab Initio Moledel Study of the Medium Effect on Proton Transfer in the Methylamine-Formic Acid Complex", J. Mol Struct. (THEOCHEM), 106 (1984) 301-307. 278

371. Sapse, A.M., and C.S. Russell, "Theoretical Studies of the Binding of Methylamine and Guanididne to Carboxylate", J. Mol Struct. (THEOCHEM), 137 (1986) 43-53. 372. Kulbida, A.I., and V.M. Schreiber, "Infrared Study of Some Complexes with a Strong Hydrogen Bond at High and Low Temperatures", J. ofMolec. Struct., 47 (1978) 323-328. 373. Sapse, A.M., and C.S. Russell, "Ab Initio Calculations of Guanidinium-Carboxylate Interaction", Int. J. Quantum Chem., 26 (1984) 91-99. 374. Fulscher, M.P., and E.L. Mehler, "Ab Initio Interaction Potential of Guanidine-Formic Acid in the Neutral and Charged States", J. Molec. Struct. (THEOCHEM), 165 (1988) 319-327.

375. Binkley, J.S., J.A. Pople, W.J. Hehre, J. Am. Chem. Soc., 102 (1980) 939. 376. Harihan, P.C., J.A. Pople, "Effect of d-fiinctions on molecular orbital energies for hydrocarbons", Chem. Phys. Lett., 16 (1972) 3099-3108. 377. Hehre, W.J., L. Radom, P.V.R. Schleyer, J.A. Pople, Ab Initio Molecular Orbital Theory (John Wiley, New York, 1986) p. 77. 378. Dunning, T.H., and P.J. Hay, in Modem (Ed. H.F. Schaefer HI), vol. 3 (Plenum Press, New York, 1977) p. 1. 379. Davidson, E.R., and D. Feller, "Basis Set Selection for Molecular Calculations", Chem. Rev. 86 (1986) 681-696. 380. Boys, S.F., and F. Bemardi, "The calculation of small molecular interactions by the differences of separate total energies. Some procedures with reduced errors", Mol. Phys., 19 (1970) 553. APPENDIX A UV ABSORBANCE AND ANALYTICAL WAVELENGTHS FOR pKa DETERMINATIONS

O R*

rY2 x i " r*

COj- UV Maxima (nm) Analytical COMPOUND Rj R , R4 Rj R* X Y pK, Base Formb AcidFonn* Wavelength

73 pNoTYR OH H H NO, H c NHj* 6.4 413 324 413 B 0N0TYR OH NO, H H H c NH,* 6.6 430.291 353,280 430 i^DNoTYR OH NO, H NO, H c NH,* 3.4 380.40Ssh 263.29Ssh 380 m a-OH-DNoT OH NO, H NO, H c OH 4.2 267 371 142 3-NTYR H NO, OH H H c NH,* 6.8 429,284,232 354,276 429 I S DNTYR H NO, OH NO, H c NHj* 3.2 444 355,247 444 M 4,6-DNmTYR H OH NO, H NO, c NHj* 3.8 360,270 265 360 92 2,6-DNmTYR NO, OH H H NO, c NH,* 3.6 381 302 381 §9 WDN -O H -O H H N NH,* 9.3 264 259 258 g 5*NOj-WDN - o H < 0 NO, H N NH,* 6.4 217,326 237JO 1 326

*pH 11.8 ^pH 1.5-1.8

279 APPENDIX B NONLIN84 SUBROUTINE FOR pK„ FITTING

SUBROUTINE: SUBROUTINE COMPUT(F,X,WT,DTA,NVAR,IF,IOB,P,NP)IFLAG,Z,DZ,NDER) IMPLICIT REAL* 8 (A-H,0-Z) COMMON /CONST/ NOBS(10),CON(50),NTOT,NF,NCON,IWRITE>IPAR, - IDATAJPLOT,IPARMJSECO DIMENSION P(20),DTA(10),Z(10),DZ(20) REAL* 8 PKA, A_ACID, A_BASE If "Subroutine to fit Single pKa Curve from Spectrophotometric " Determination to the Equation: " pH - pKa + log [(Abs-A_acid)/(A_base-Abs)] II " rearranged to: ll Abs - [A_acid+(A_base* 10*(pH-pKa»]/[l+10“(pH-pKa>] l l " (Hence, Absorbance is the dependent (y) " and pH the independent (x) variable) "TEMPORARY VARIABLE ASSIGNMENTS PKA*P(1) A_ACID-P(2) A_BASE=*P(3) PH -X L-PH-PKA LL=L*DLOG( 10) TTTL"DEXP(LL)

"FUNCTION DEFINITIONS IF (IFLAG .EQ. 3) THEN F-ABS F=(A ACID+(A BASE*TTTL))/(1+TTTL) 1=0 ENDEF

"WRITE STATEMENTS IF (I .EQ. 0 .AND. IFLAG .EQ. 11) THEN WRTTE(6,100) ’ ’ WRITE(6,100) ’RESULTS FOR FIT OF SPECTROPHOTOMETRIC pKa DATA’ WRTTE(6,100) ’______’ WRITE(6,101) ’ pKa - PKA 280 281 WRITE(6,102) ’ Absorbance of Acid Form = A_ACED, ’ AU’ WRITE(6,102) ’ Absorbance of Base Form - ’, A_BASE, ’ AU’ 1= 1+1 ENDIF RETURN 100 FORMAT (’O’, 10X, A) 101 FORMAT (’O’, 10X, A, F5.3) 102 FORMAT (’O’, 10X, A, F5.4.A) END DATA INPUT:

TITLE pKa Fit for 5-N02-WDN at 326nm METH2 WEIGHT 0 CONV IE-6 ITER 50 NPAR 3 INIT 6.45 0.450 1.173 LOWE 5.9 0.400 1.223 UPPE 7.0 0.500 1.123 NCON 0 NSEC 0 NFUN 1 NOBS 17 DATA 8.35 1.172 7.93 1.154 7.20 1.060 7.07 1.035 6.99 1.011 6.81 0.963 6.71 0.921 6.57 0.868 6.47 0.831 6.35 0.779 6.22 0.731 6.11 0.689 5.73 0.580 5.2 0.494 4.69 0.465 4.18 0.453 3.86 0.451 BEGIN FINISH