Electrostatic Interactions in Chemistry and Biology Citation Gallivan, Justin Patrick (2000) Electrostatic Interactions in Chemistry and Biology. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/hj9z-mp47. https://resolver.caltech.edu/CaltechTHESIS:08252025-211023743 Abstract Electrostatic interactions such as hydrogen bonds, salt bridges, and cation-π interactions play a large role in structural biology. A major goal of this thesis is to build upon previous studies of the cation-1r interaction to further understand its role in biological systems. Put simply, we wish to understand how, when, and why Nature uses cation-π interactions. We begin by highlighting a cation-1r interaction important in the binding of acetylcholine to the nicotinic acetylcholine receptor (nAChR). By combining ab initio calculations and molecular neurobiology, we provide compelling evidence that a cation-π interaction is a major determinant of the recognition of acetylcholine by the nAChR. We then ask a broader question: To what extent does Nature use cation-π interactions within protein structures? By surveying the protein databank, we demonstrate that energetically significant cation-π interactions are quite common within protein structures. To explain why, we ask what advantages cation-π interactions have over other noncovalent interactions commonly found in proteins. Using quantum mechanical calculations, we study the strengths of cation-π interactions and salt bridges in both water and in a range of organic solvents. The results suggest that cation-π interactions maintain their strength over a wide range of solvents, whereas the strength of a salt bridge is severely attenuated when it is placed in a high-dielectric solvent. We then turn our attention to a different type of electrostatic interaction - the interaction between water and hexafluorobenzene. We find that in the gas phase, water binds to hexafluorobenzene in a geometry in which the lone pairs of electrons located on the oxygen are directed towards the π-system of the aromatic. This surprising result is easily explained using electrostatics. In addition, we present computational studies of the triphenylene···perfluorotriphenylene "supramolecular synthon." Finally, we return to the nAChR. A challenge in the study of integral membrane proteins is determining their transmembrane topology. Here we present a potentially general method for determining not only the transmembrane topology of a functional neuroreceptor expressed in a living cell, but also the surface accessibility of individual amino acids. Item Type: Thesis (Dissertation (Ph.D.)) Subject Keywords: (Chemistry) Degree Grantor: California Institute of Technology Division: Chemistry and Chemical Engineering Major Option: Chemistry Thesis Availability: Public (worldwide access) Research Advisor(s): Hoffmann, Michael R. Thesis Committee: Dervan, Peter B. (chair) Hoffmann, Michael R. Rees, Douglas C. Bercaw, John E. Defense Date: 22 May 2000 Record Number: CaltechTHESIS:08252025-211023743 Persistent URL: https://resolver.caltech.edu/CaltechTHESIS:08252025-211023743 DOI: 10.7907/hj9z-mp47 Default Usage Policy: No commercial reproduction, distribution, display or performance rights in this work are provided. ID Code: 17645 Collection: CaltechTHESIS Deposited By: Benjamin Perez Deposited On: 28 Aug 2025 09:28 Last Modified: 28 Aug 2025 09:35 Thesis Files PDF - Final Version See Usage Policy. 71MB Repository Staff Only: item control page

Electrostatic Interactions in Chemistry and Biology Thesis by Justin Patrick Gallivan In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy California Institute of Technology Pasadena, Cali~ornia 2000 (Defended May 22, 2000) 11 @ 2000 Justin Patrick Gallivan All Rights Reserved lll Acknowledgements O.K, admit it. This is likely the only section of the thesis that you are going to read. Oh sure, you might flip through the rest. You might even pause to admire the fancy graphics on page 35, but the fact of the matter is, you really want to see if your name made it in here. Thus, I apologize in advance for anyone that I might inadvertently leave out. First of all , I must thank my advisor, Dennis Dougherty. I don't think I could have asked for a better advisor. Dennis has provided a great environment in which to do science. He has given me the freedom to choose problems that I thought were interesting, and has been a constant source of advice and wisdom. But aside from being an excellent scientist, Dennis is quite simply a good person, and I have really enjoyed having the opportunity to work with him. Special thanks also go to Ellen Dougherty, who on many occasions welcomed the group into her home. Ellen realizes that beneath all of our academic bravado, we are just like elementary school kids who miss home. I would also like to thank Henry Lester for allowing me to spend time in his lab. Although the draw of computational chemistry eventually led me back to the second floor of Crellin, I sincerely appreciate the time I spent on the third floor of Kerckhoff. I also extend thanks to my committee. Peter Dervan has always offered sage advice. I will always remember a talk he gave about the importance of choosing good mentors - fortunately, I have had several. I would also like the thank Barbara Imperiali, Doug Rees , and John Bercaw, who have all taken time out of their busy schedules to attend committee meetings or write letters of recommendation on my behalf. The Dougherty group has been a great place to work over the past several years. Kraig Anderson and Seth Miller made 247 Crellin a very interesting place during my first few years. Kraig's staunch cynicism and Seth's unwavering optimism nicely IV balanced one another. Along with Kraig, Jon Forman and Mick Murray were always willing to go out for a couple of pints. 1 Sandro Mecozzi has to be one of the nicest guys in the world , and I appreciate having had the opportunity to work with him. Jeff Clites is one of the brightest people I have ever met and was always fun to talk with about any subject. I owe special thanks to many people who helped me get started on various projects. Scott Silverman is a great teacher and friend. If they gave a Nobel Prize in knowledge of "The Simpsons," he would win it every year. I wish him all the best as he begins his academic career. Mark Nowak taught me most of the molecular biology that I know and I greatly appreciate that. Anthony West is an incredibly smart guy and he got me started down the computational chemistry path - this thesis would probably be a lot different without his influence. Wenge Zhong and I collaborated on the work described in Chapter 2, and the combination of theory and experiment made the project quite fun. Pam England and I had our differences at times, but I am glad that we got to interact for so many years, and I am a better person for it. Of the "younger" students in the group, I would particularly like to thank Gabriel Brandt, Josh Maurer, and Don "Fancy Pants" Elmore (sorry, Don, it had to be done). Gabriel has always been willing to critically discuss almost any subject, scientific or otherwise, and he always offers an interesting opinion. Josh Maurer and Don Elmore are excellent people to bounce ideas off of, and reading the "Pint-Sized Punchlines" was always fun. 24 7 Crellin will be left in the very capable hands of Darren Beene, James Petersson, and Sarah Monahan. I have enjoyed having them around these last few months. I would also like to thank all of the other members of the Dougherty group that have made my time here special. Outside of the group, I owe thanks to the Eastman Kodak Corporation for providing me with several years of funding, and to my three mentors there: Eric Ginsburg, Rakesh Jain , and Frank Corns. I learned a lot from all of you and always enjoyed my visits to Rochester. Special thanks also go to Dian Buchness, Tom Dunn, Steve 1 It should be noted that along with the decline in the number of magnetism projects in the group came a decline in the amount of alcohol consumed by the group. Determining the significance of this is left as an exercise for the reader. V Gould, and Chris Smith. Aside from being very nice people, they have all gone above and beyond the call of duty to help me out in one way or another. I would also like to thank Cory Nelson and Dave Lynn for being great friends. The trips to Las Vegas and San Francisco, poker nights , McMurphy's, and running the basketball pool were all lots of fun. I had a great time hanging out with these guys, and their presence at Caltech is missed. I would also like to thank Adam Matzger, who has been a good friend and an excellent source of wisdom. I wish he and Anna the best of luck as they start their professional careers. Finally, I would like to thank my family. My wife, Martha, has been a constant source of support throughout my time here. Martha probably understands me better than I understand myself, and she always seems to know exactly the right thing to say. But most of all, Martha is always there to listen, and often that is the best thing. I would also like to thank my parents. They, too, have always been there for me and have supported me in everything that I have done. Mom, Dad, and Martha: I love you all. Vl Abstract Electrostatic interactions such as hydrogen bonds, salt bridges, and cation-1r interactions play a large role in structural biology. A major goal of this thesis is to build upon previous studies of the cation-1r interaction to further understand its role in biological systems. Put simply, we wish to understand how, when, and why Nature uses cation-1r interactions. We begin by highlighting a cation-1r interaction important in the binding of acetylcholine to the nicotinic acetylcholine receptor (nAChR). By combining ab initio calculations and molecular neurobiology, we provide compelling evidence that a cation-1r interaction is a major determinant of the recognition of acetylcholine by the nAChR. We then ask a broader question: To what extent does Nature use cation-1r interactions within protein structures? By surveying the protein databank, we demonstrate that energetically significant cation-1r interactions are quite common within protein structures. To explain why, we ask what advantages cation-1r interactions have over other noncovalent interactions commonly found in proteins. Using quantum mechanical calculations, we study the strengths of cation-1r interactions and salt bridges in both water and in a range of organic solvents. The results suggest that cation-1r interactions maintain their strength over a wide range of solvents, whereas the strength of a salt bridge is severely attenuated when it is placed in a high-dielectric solvent. We then turn our attention to a different type of electrostatic interaction - the interaction between water and hexafluorobenzene. We find that in the gas phase, water binds to hexafluorobenzene in a geometry in which the lone pairs of electrons located on the oxygen are directed towards the 1r-system of the aromatic. This surprising result is easily explained using electrostatics. In addition, we present computational studies of the triphenylene· · · perfluorotriphenylene "supramolecular synthon." Finally, we return to the nAChR. A challenge in the study of integral membrane proteins is determining their transmembrane topology. Here we present a potentially Vll general method for determining not only the transmembrane topology of a functional neuroreceptor expressed in a living cell, but also the surface accessibility of individual amino acids. Vlll Contents Acknowledgements iii Abstract VI 1 Introduction 1 1.1 Introduction . Bibliography 2 7 2 From Ab Initio Quantum Mechanics to Molecular Neurobiology: A Cation-1r Interaction in the Nicotinic Acetylcholine Receptor 10 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . 11 2.2 Characterizing the nAChR at the Molecular Level 12 2.2.1 Structural Data . . . 12 2.2.2 Sequence Landmarks 13 2.2.3 The Agonist Binding Site 14 2.3 Cation-1r Interactions and Structure-Function Studies 2.3.1 16 Site-Directed Mutagenesis in Molecular Neurobiology- Varying Structure . . . . . . . . . . . . . . . . . . . . . . . . 16 2.3.2 Electrophysiological Recording - Measuring Function 17 2.3.3 Correlating Structure and Function . . . . . . . . . . 18 2.4 Previous Studies of Possible Cation-1r Interactions in the nAChR . 19 2.5 The Search for a Cation-1r Binding Site within the nAChR 21 2.5.1 Theory . . . 21 2. 5. 2 Experiment 26 2.6 Summary . . . . . 31 2. 7 Appendix 1 - Details of Calculations 32 IX 2.8 Appendix 2 - Related Cation-1r Calculations 33 Bibliography 38 3 44 Cation-1r Interactions in Structural Biology 3.1 Introduction . 45 3.2 Methods 48 3.3 Results and Discussion 54 3.3.1 Cationic Amino Acids 59 3.3.2 Aromatic Amino Acids 64 3.3.3 Pairwise Interactions 67 3.4 Gallery of Cation-1r Interactions 69 3.5 Summary 73 3.6 Appendix 1 - CaPTURE Manual 73 3.6.1 Overview 73 3.6.2 Installation 74 3.6.3 Getting Started 75 3.6.4 Interpreting the Screen Output from CaPTURE 75 3.6.5 Adjusting the Parameters 79 3.6.6 Limitations of CaPTURE 82 3.7 . Appendix 2 - Description of CaPTURE Source 87 Bibliography 92 4 Computational Studies of Cation-1r Interactions and Salt Bridges 96 4.1 Introduction . 97 4.2 Results . . . . 99 4.2 .1 Ab Initio Calculations 4.2.2 Protein Studies • 104 4.3 Summary .. . . . . . 108 4.4 Appendix 1 - Methods 109 4.5 Appendix 2 - Sample Gaussian Input Files 113 99 X 4.5.1 CH 3 NH! ···Benzene Z-matrix 113 4.5.2 CH 3 NH! ···Acet ate Z-matrix 114 Bibliography 116 5 Weak Electrostatic Interactions Involving Perfluorinated Aromatic Rings 120 5.1 Introduction . 121 5.2 Results .. . . 124 5.2.1 On-Axis Geometries 124 5.2.2 Off-Axis Geometries 129 5.3 Summary . . . . . . . 130 5.4 Computational Methods 131 5.5 Appendix 1 - Studies of the Supramolecular Synthon Triphenylene· · • 5.6 Perfluorotriphenylene . . . . . . .. . . . . . .. .. .. . 132 Appendix 2 - Total Energies and Geometric Parameters 138 Bibliography 6 140 Site-Specific Incorporation of Biotinylated Amino Acids to Identify Surface-Exposed Residues in Integral Membrane Proteins 142 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 6.2 Previous Approaches to Determining Topology and Accessibility 145 6.2.1 Endogenous Glycosylation Sites and N-Glycosylation Scanning Mutagenesis. . . . . . . . .. 145 6.2.2 Cysteine Accessibility /SCAM 146 6.2.3 Epitope Protection 146 6.2.4 Tyrosine Iodination . 147 6.3 The Biocytin Approach . 147 6.4 Results and Discussion 150 6.5 Summary . . 158 6.6 Experimental 159 Xl 6.6.1 Synthesis of Biotinylated Amino Acids 6.6.2 General Method for Coupling of N-Protected Amino Acids to 159 dCA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 6.6.3 Site-Directed Mutagenesis of the AChR a67-76 TAG Mutants 163 6.6.4 In Vitro Transcription of mRNA 164 6.6.5 Oocyte Preparation and Injection 164 6.6.6 Electrophysiology . . . . . . . . . 164 6.6. 7 General Procedure for Radioactive Ligand Binding Experiments 165 Bibliography 166 7 Other Experiments 172 7.1 Amino Acid Synthesis . .. .. . . . . . . . . . . . . . . . 173 7.1.1 General Method for the Protection of Amino Acids 174 7.1.2 General Method for the Synthesis of Amino Acid Cyanomethyl Esters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1.3 7.2 7.3 7.4 175 General Method for Coupling of N-Protected Amino Acids to dCA . . . . . . . . . . . . 176 Synthesis of Natural Amino Acids 177 7.2.1 dCA-NPS-Phenylalanine . 177 7.2.2 dCA-4-PO-Phenylalanine. 178 7.2.3 dCA-NVOC-Glycine 180 7.2.4 dCA-NVOC-Alanine 181 7.2.5 dCA-NVOC-Proline 182 7.2.6 NVOC-Glutamine Cyanomethyl Ester 183 7.2.7 (NVOC)2-Lysine Cyanomethyl Ester 184 7.2.8 NVOC-Methionine Cyanomethyl Ester 186 Synthesis of Unnatural Amino Acids 187 7.3.1 dCA-NVOC-Allyl-Glycine . . 187 7.3.2 NVOC-,6-(2-Thienyl)-L-Alanine Cyanomethyl Ester 188 Synthesis of Photoreactive Amino Acids .. . .. . . . . . 190 Xll 7.4.1 alanine . . . . . . . . . . . . . . . . . 190 dCA-4-PO-p-Benzoyl-Phenylalanine . 196 Synthesis of Caged Amino Acids . . . . . . . 197 7.5.1 dCA-4-PO-Cysteine-S-o-Nitrobenzyl Thioether 197 7.5.2 a-4-PO-c-NVOC-Lysine Cyanomethyl Ester 199 7.5.3 dCA-4-PO-Serine-O-o-Nitroveratryl Ether . 200 7.5.4 dCA-4-PO-Threonine-O-o-Nitroveratryl Ether 201 Molecular Biology . . . . . . . . . . . . . . . . . . . . 205 7.6.1 General Procedure for Site-Directed Mutagenesis 205 7.6.2 General Procedure for In Vitro Transcription of mRNA 205 7.6.3 Mutants for Nonsense Suppression in the nAChR 206 7.6.4 Conventional Mutants in the nAChR 206 7.4.2 7.5 7.6 7.7 7.8 dCA-N-( 4-PO )-3-[p-[3-( trifluoromethyl)-diazirin-3-yl]phenyl]-DL- Oocyte Experiments . . . . . . . . . . . 207 7.7.1 Oocyte Preparation and Injection 207 7. 7. 2 Electrophysiology . . . . . . . . . 208 Results of Nonsense Suppression Experiments 209 7.8.1 Natural Amino Acids . . . 209 7.8.2 Biotinylated Amino Acids 210 7.8.3 Chemically Reactive Amino Acids . 211 7.8.4 Photochemically Reactive Amino Acids . 212 Bibliography 215 Xlll List of Figures 1.1 Schematic view of a cation-1r interaction . . . 2 1.2 The electrostatic potential surface of benzene 4 2.1 Structural features of the nAChR 13 2.2 Cutaway view of the nAChR . . . 14 2.3 Schematic view of the a subunit of the nAChR 15 2.4 Calculated HF /6-31G** electrostatic potential surfaces and Na+ binding energies of the aromatic side chains of phenylalanine, tyrosine, and tryptophan . . . . . . . . . . . . . . . . . . 20 2.5 Side chain structures of tryptophan analogs 22 2.6 The HF /6-31G** Na+ binding energies of fluorinated indole derivatives 24 2.7 Electrostatic potential surfaces of the side chains of Trp, 4,5,6,7-F4 Trp, 5-Br-Trp, and 5-CN-Trp 2.8 . . . . . . . . . . . . . . . . . . . . . . Plot of log[EC 50 / EC 50 (wild type)] vs. quantitative cation-1r binding ability of Trp and its fluorinated derivatives expressed at a149 . . . . 2.9 25 28 Plot of log[EC 50 /EC 50 (wild type)] vs. quantitative cation-1r binding ability of all Trp derivatives expressed at a149 . . . . . . . . . . . . . 29 2.10 Plot of log[EC 50 / EC 50 ( wild type)] vs. quantitative cation-1r binding ability of the 5-membered ring of Trp and its fluorinated derivatives expressed at a 149 . . . . . . . . . . . . . . . . . . . . . 30 2.11 Side chain structure of the tethered agonist Tyr-O-3Q . 31 2.12 Plot of HF /6-311G** vs. HF /6-31G** Na+ binding energies for a series of aromatics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 2.13 HF /6-31G* * electrostatic potential surfaces for all aromatic side chains 35 3.1 Limiting geometries for cation-1r interactions. . . . . . . . . . . . . . . 47 XlV 3.2 Plots of Ees vs. counterpoise-corrected HF 6-31G** binding energies for selected cation-1r interactions. . . . . . . . . . . . . . . . . . . . . 51 3.3 Scatter plot for all Lys/Phe pairs from all 323 single subunit proteins. 57 3.4 Plot of the number of cation-1r interactions vs. the sequence distance of the interacting pair. . . . . . . . . . . . . . . . . . . . . . . . . . . 59 3.5 An intrahelical cation-1r interaction in the vaccinia virus protein VP39 61 3.6 Distribution of Ees for cation-1r interactions involving arginine and lysine. 62 3. 7 The interplane angle of cation-1r interactions involving arginine. 63 3.8 Scatter plots for energetically significant cation-1r interactions 65 3.9 Views of all Lys-Trp cation-1r interactions from single subunit proteins. 66 3.10 Schematic view of the model used to calculate the preferred location for lysine phenylalanine pairs 69 3.11 Representative cation-1r interactions involving lysine . 70 3.12 Representative cation-Jr interactions involving arginine 71 3.13 A lysine involved in 4 cation-1r interactions 72 3.14 Flowchart of Capture.c . . . . . . . . . . . 88 4.1 Geometries for CH 3 NHt · · • benzene and CH 3 NHt · · · acetate .. 99 4.2 Interaction energies for CH 3 NHt ···benzene and CH 3 NHt ··· acetate in water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Interaction energies for CH 3 NHt ···benzene and CH 3 NHt ···acetate in various solvents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 100 102 The percentage of the total number of residues that have a given relative exposure . . . . . . . . . . . . . 106 4.5 Integration of the data in Figure 4.4 . 107 5.1 Experimentally-determined geometry of water· · · benzene .. 121 5.2 Electrostatic potential surfaces of benzene and water. 122 5.3 Electrostatic potential surfaces of hexafluorobenzene . 123 5.4 Predicted binding geometry of water··· hexafluorobenzene. 123 5.5 On-axis geometry of the water• • • hexafluorobenzene complex .. 124 xv 5.6 Schematic potential energy curves . . . . . . . . . . . . . . . . 128 5. 7 Off-axis geometry of the water- · · hexafluorobenzene complex. . 129 5.8 In-plane geometry of water··· hexafluorobenzene . . . . . . . . 131 5.9 Electrostatic potential surfaces of triphenylene and perfluorotriphenylene 133 5.10 Calculated binding energies of triphenylene· · • perfluorotriphenylene 134 5.11 Calculated binding energies of benzene··· hexafluorobenzene 136 6.1 Predicted topologies of the nAChR and GluR3 . 144 6.2 Structure of biocytin . . . . . . . . . . . . . . . 149 6.3 Average whole-cell currents for oocytes injected with nAChR a70TAG or a76TAG and tRNA . . . . . . . . . . . . . . . . . . . . . . . . . . 151 6.4 Whole-cell currents from biocytin suppresion at each of the 10 positions of the MIR 6.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 [125 I]-Streptavidin binding to oocytes injected with nAChR a70TAG or a76TAG and tRNA . . . . . . . . . . . . . . . . . . . . . . . . . . 154 6.6 Whole-cell binding of [125 I]-streptavidin and [125 I]-bungarotoxin to AChR a70 biocytin mutants . . . . . . . . . . . . . . . . 155 6.7 NMR structure of the MIR peptide (PDB:lTOR) 156 6.8 Biocytin analogs 157 7.1 ACh-induced currents for nAChR glycosylation mutants) 208 7.2 Natural amino acids incorporated into the nAChR. . . 209 7.3 Biotinylated amino acids incorporated into the nAChR 210 7.4 Amino acids with reactive side chains incorporated into the nAChR. 211 7.5 A water-soluble olefin metathesis catalyst. . . . . . . . . . . . . . . 212 7.6 Resting membrane potentials for oocytes treated with a ruthenium- . . . . . . . . . . . . . . . . . . based olefin metathesis catalyst . . . . . . . . . . . . . . . . . . . . . 213 7.7 Amino acids with photochemically reactive side chains incorporated into the nAChR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214 XVI List of Tables 2.1 Cation-n binding energies for the aromatic derivatives shown in Figure 2.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.2 EC 50 values for ACh for mutants at a86, a149, a184, 1 55, and b57 26 2.3 Cation-n binding abilities and EC 50 values for ACh for tryptophan derivatives incorporated into the nAChR . . . . . . . 27 2.4 Total energies of monomers and 6-ring Na+ complexes. 32 2.5 Total energies of the 5-ring Na+ complexes. 33 2.6 Calculated binding energies and distances for the Na+··· benzene com- . ..... plex at several levels of theory. . . . . . . . . . . . . . . . . . . . . . . 2. 7 36 Calculated binding energies for the NHt • · · benzene complex at differing symmetries and levels of theory. . . . . . . . . . . . . . . . . . . 37 2.8 Calculated binding energies for various Na+··• aromatic complexes 37 3.1 PDB ID Codes for single-subunit proteins. 55 3.2 PDB ID Codes for multi-subunit proteins. 56 3. 3 Interacting pair data . . . . . . . . . . . . 60 3.4 Frequency of cation-n interactions for individual residue types within the 593 protein dataset. 3.5 . . . . . . . . . . . . . . . . . . . . . . . . . Frequency of cation-7!' interactions for residue pairs within the 593 protein dataset. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 60 67 Interaction energies for CH 3 NHf · · · benzene and CH 3 NHf · · · acetate in a range of solvents . . . . . . . . . . . . . . . . . 101 4.2 Total energies for the CH 3 NHf ···acetate complex. 111 4.3 Total energies for the CH 3 NHf · · · benzene complex. 112 XVll 5.1 Results of ab initio calculations for the water· · · hexafluorobenzene complex shown in Figure 5.5 . . . . . . . . . . . . . . . . . . . . . . . 125 5.2 Results of ab initio calculations for the complex shown in Figure 5.7. 129 5.3 Geometries and total energies for H2 O at various levels of theory . 138 5.4 Geometries and total energies for C6 F 6 at various levels of theory 138 5.5 Total energies for the off-axis complex 138 5.6 Total energies for the off-axis complex 139 5. 7 Total energies of triphenylene and perfluorotriphenylene . 139 5.8 Total energies for the triphenylene· · · perfluorotriphenylene complex 139 6.1 Sequence analysis of the MIR from 34 nAChR a subunits in the SWISSPROT database. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 7.1 TAG mutants in the nAChR . 206 7.2 Conventional site-directed mutants in the nAChR. 207 7.3 Incorporation of tyrosine into the nAChR. 210 7.4 Biocytin incorporation into the nAChR. 211 7. 5 Mid-chain-biocytin incorporation into the nAChR. . 211 7.6 Allyl-glycine incorporation into the nAChR. . .. 212 7.7 Benzoyl-phenylalanine incorporation into the nAChR. 214 1 Chapter 1 Introduction 2 1.1 Introduction One of the greatest achievements of modern organic chemistry is a thorough understanding of the nature of covalent bonding. Successes ranging from modern ab initio electronic structure methods to the total synthesis of complex molecules all testify to our ability to understand and manipulate covalent bonding. What is less well understood, however, is the nature of noncovalent bonding. It is the collective behavior of weak , noncovalent interactions that ultimately determines the structure and function of biological systems. As such, a thorough knowledge of weak interactions, such as hydrogen bonds, salt bridges, and the hydrophobic effect, is necessary to understand biological systems at the molecular level. In addition to these well-studied interactions, the cation-7!' interaction, the electrostatically favorable interaction between a cation and a 7!'-system, has been shown to be an important noncovalent binding interaction of relevance to structural biology (Figure 1.1) [1]. ···········O Figure 1.1: Schematic view of a cation binding to an aromatic - a cation-7!' interaction . In the early 1980s, high-pressure mass spectrometry and ion cyclotron resonance studies demonstrated that cations bind to simple aromatics in the gas phase [2 , 3] . In pioneering work , Kebarle showed that the interaction energy between K+ and benzene in the gas phase (19 kcal/mol) is even stronger than the interaction between K+ and water (18 kcal/mol) [2]. Later experiments have shown that in clusters containing both benzene and water molecules bound to a single cation, the benzenes are more 3 tightly bound than the water molecules [4]. The interaction is not limited to alkali metals, and it is now appreciated that organic cations such as tetramethylammonium and acetylcholine bind to aromatics in the gas phase [3, 5]. The cation-1r interaction might have remained a gas phase curiosity were it not for work done by the Dougherty group in the 1980s and early 1990s. Using a series of water-soluble cyclophane hosts containing a hydrophobic cavity, e.g. 1, Dougherty and co-workers demonstrated that these molecules could bind not only hydrophobic guests, but also cationic guests in aqueous media [6- 11] . .0 C CO2· 1/ ~ ~o 9 0 ~ ~ o o ~/4 /o ~ ·O2C CO£ 1 This is significant because to bind a cationic guest, a substantial (>40 kcal/mol) desolvation penalty must be paid to transfer the cation from bulk water to the interior of the host. This suggests that the binding energy gained in forming a cation-1r interaction must be quite strong to overcome the desolvation penalty. A key result from the small molecule studies is that host molecules such as 1 bind the quaternary ammonium group of acetylcholine. This led to the prescient prediction by Dougherty and Stauffer that aromatic amino acids might be involved in the binding of acetylcholine in biological systems [9]. Indeed this is the case, and the x-ray structure of acetylcholinesterase reveals an "aromatic gorge" that presumably guides acetylcholine to the active site of the esterase [12]. Remarkably, the "anionic subsite" that contacts the quaternary ammonium group of acetylcholine is the aromatic sidechain of a tryptophan residue. This study clearly established a role for cation-1r interactions in biological systems. 4 In addition to the host-guest studies, numerous computational studies have been undertaken to understand the physical basis of the cation-1r interaction [13- 17]. It is now appreciated that the cation-1r interaction is a complex phenomenon, involving dispersion, polarization, and electrostatic interactions [l]. Interestingly, although the dispersion and polarization effects may represent a substantial contribution to the total binding energy in a cation-1r interaction [14, 17], these effects remain relatively constant across a related series of aromatics [15] . As such, variations in the cation-1r binding ability of aromatics correlate very well with variations in the electrostatic potential of the aromatic. Thus electrostatic potential maps of the aromatics, such as the one for benzene, shown in Figure 1.2, provide a useful guide to predicting the cation-1r binding ability of a given aromatic [16]. Figure 1.2: The calculated HF / 6-31G** electrostatic potential surface of benzene. Red regions represent areas of negative electrostatic potential and blue regions represent positive electrostatic potential. Scale: -25 to +25 kcaljmol. A major goal of this thesis is to build upon previous fundamental studies of the cation-1r interaction to further understand the role of cation-1r interactions in biological systems. Put simply, we wish to understand how, when, and perhaps more importantly, why Nature uses cation-1r interactions. We do this in several different contexts, ranging from ab initio quantum mechanical calculations of the cation-1r interaction both in the gas phase and in solution, to studies of the cation-1r interaction between acetylcholine and the nicotinic acetylcholine receptor expressed in a living cell. Chapter 2 hi ghlights a particular cation-1r interaction important in the binding 5 of acetylcholine to the nicotinic acetylcholine receptor. By combining high-level ab initio calculations and molecular neurobiology, we provide compelling evidence that a cation-n interaction is a major determinant of the recognition of acetylcholine by its natural receptor. In doing so, we provide some of the highest resolution structural information to date about the nicotinic acetylcholine receptor. In Chapter 2, we only consider potential cation-n interactions between a particular receptor-ligand pair. The work presented in Chapter 3 addresses a broader question. In particular, we ask to what extent Nature uses cation-1r interactions within protein structures. By surveying a large database of protein structures, we demonstrate that energetically significant cation-n interactions are quite common within protein structures. In fact, one quarter of all tryptophans in our dataset are involved in at least one cation-n interaction. Extensive statistics and the methodology used are presented. The results m Chapter 3 present clear evidence that cation-n interactions are common in protein structures. In Chapter 4, we attempt to discover why this is the case. In particular, we ask what advantages cation-n interactions have over other noncovalent interactions commonly found in protein structures. Using high-level quantum mechanical calculations, we study the strengths of cation-1r interactions and salt bridges in both water and in a range of organic solvents. The results suggest that cation-n interactions maintain their strength over a wide range of solvents, whereas the strength of a salt bridge is severely attenuated when it is placed in a high-dielectric solvent. In Chapter 5, we turn our attention to a different type of electrostatic interaction. In particular, we computationally explore the interaction between two very simple molecules, water and hexafluorobenzene. We find that in the gas phase, water binds to hexafluorobenzene with 2.1 kcal/mol of binding energy. Most surprisingly, the geometry is such that that the lone pairs of electrons located on the oxygen are directed towards the 1r-system of the aromatic. This result would not normally be predicted from our views of orbital interactions, but it is easily explained using electrostatics. In addition to studies of the water· · · hexafluorobenzene system, we present initial 6 computational studies of the triphenylene· · · perfluorotriphenylene "supramolecular synthon." In Chapter 6, we return our attention to the nicotinic acetylcholine receptor. A major challenge in the study of integral membrane proteins is determining the transmembrane topology of the protein. Here we present a potentially general method for determining not only the transmembrane topology of a functional neuroreceptor expressed in a living cell, but also the surface accessibility of individual amino acids. Finally, in Chapter 7 we present experimental details of our efforts to develop new unnatural amino acids to explore membrane protein structure and function. 7 Bibliography [1] J. C. Ma and D. A. Dougherty. The cation-1r interaction. Chem. Rev. , 97:13031324, 1997. [2] J. Sunner, K. Nishizawa, and P. Kebarle. Ion-solvent molecule interactions in the gas phase. The potassium ion and benzene. J. Phys. Chem. , 85: 1814- 1820, 1981. [3] C.A. Deakyne and M. Meot-Ner (Mautner). Unconventional ionic hydrogen bonds. 2. NH+ 1r. complexes of onium ions with olefins and benzene derivatives. J. Am. Chem. Soc., 107:474- 479 , 1985. [4] 0. M. Cabarcas, C. J. Weinheimer, and J. M. Lisy. Size selectivity by cation-1r interactions: Salvation of K+ and Na+ by benzene and water. J. Chem. Phys., 110:8429-8435 , 1999. [5] C.A. Deakyne and M. Meot-Ner. Ionic hydrogen bonds in bioenergetics. 4. Interaction energies of acetylcholine with aromatic and polar molecules. J. Am. Chem. Soc., 121(7):1546- 1557, 1999. [6] T.J . Shepodd, M.A. Petti, and D.A. Dougherty. Molecular recognition in aqueous media: Donor-acceptor and ion-dipole interactions produce tight binding for highly soluble guests. J. Am. Chem. Soc., 110:1983- 1985, 1988. [7] M. A. Petti, T. J. Shepodd, R. E. Barrans Jr. , and D. A. Dougherty. "Hy- drophobic" binding of water-soluble guests by high-symmetry, chiral hosts. An electron-rich receptor site with a general affinity for quaternary ammonium compounds and electron-deficient 1r systems. J. Am. Chem. Soc ., 110:6825- 6840, 1988. 8 [8] D.A. Stauffer, R.E. Barrans Jr. , and D.A. Dougherty. Concerning the thermodynamics of molecular recognition in aqueous and organic media. Evidence for significant heat capacity effects. J. Org. Chem., 55:2762- 2767, 1990. [9] D.A. Dougherty and D.A. Stauffer. Acetylcholine binding by a synthetic receptor. Implications for biological recognition. Science, 250: 1558-1560, 1990. [10] A. McCurdy, L. Jimenez, D. A. Stauffer, and D.A. Dougherty. Biomimetic catalysis of SN2 reactions through cation-1r interactions. The role of polarizability in catalysis. J. Am. Chem. Soc. , 114: 10314- 10321 , 1992. [11] P. C. Kearney, L. S. Mizoue, R. A. Kumpf, J. E. Forman, A. McCurdy, and D. A. Dougherty. Molecular recognition in aqueous media. New binding studies provide further insights into the cation-1r interaction and related phenomena. J. Am. Chem. Soc. , 115:9907- 9919, 1993. [12] J.L. Sussman, M. Harel, F. Frolow, C. Oefner, A. Goldman, L. Toker, and I. Silman. Atomic structure of acetylcholinesterase from Torpedo californica: A prototypic acetylcholine-binding protein. Science, 253:872- 879, 1991. [13] R.A . Kumpf and D.A. Dougherty. A mechanism for ion selectivity in potassium channels: Computational studies of cation-1r interactions. Science, 261: 17081710, 1993. [14] J . W . Caldwell and P. A. Kollman. Cation-1r interactions: Nonadditive effects are critical in their accurate representation. J. Am. Chem. Soc., 117:4177- 4178, 1995. [15] S. Mecozzi , A. P. West Jr. , and D. A. Dougherty. Cation-1r interactions in simple aromatics. Electrostatics provide a predictive tool. J. Am. Chem. Soc., 118:2307- 2308 , 1996. [16] S. Mecozzi, A. P. West Jr., and D. A. Dougherty. Cation-1r interactions in aromatics of biological and medicinal interest : Electrostatic potential surfaces as a useful qualitative guide. Proc. Natl. Acad. Sci. USA , 93:10566- 10571 , 1996. 9 [17] E. Cubero, F. J. Luque, and M. Orozco. Is polarization important in cation-7r interactions? Proc. Natl. Acad. Sci. USA, 95:5976-5980, 1998. 10 Chapter 2 From Ab Initio Quantum Mechanics to Molecular Neurobiology: A Cation-n Interaction in the Nicotinic Acetylcholine Receptor 11 2.1 Introduction The nicotinic acetylcholine receptor (nAChR) is the prototype ligand-gated ion channel [1- 3]. It is the longest known and most well studied member of a superfamily that includes receptors for serotonin, ,1-aminobutyric acid , and glycine. There are two main classes of nicotinic acetylcholine receptors, neuronal and muscle. Neuronal nicotinic acetylcholine receptors represent a large family of receptors that function in the central nervous system and play an important role in nicotine addiction. In contrast, the muscle-type nAChR, which is the primary focus of this chapter, functions at the neuromuscular junction. When an action potential reaches the end of a motor neuron, a pulse of acetylcholine (ACh) is released into the synaptic cleft. ACh diffuses across the synaptic cleft where it may bind to the nicotinic acetylcholine receptor. When two molecules of ACh bind to the receptor, a non-selective cation channel within the receptor opens, which allows Na+ and K+ to flow into the post-synaptic cell and causes the muscle to contract. Throughout this process, acetylcholinesterase, an enzyme located on the post-synaptic cell, rapidly hydrolyzes acetylcholine to produce choline and acetate, thereby terminating the chemical signal. The nicotinic acetylcholine receptor is a remarkable molecular machine. Upon binding two small molecules , the receptor converts a chemical signal into an electrical one. As chemists, we would like to understand both the structure and function of the nicotinic acetylcholine receptor at the chemical scale. In doing so, we would like to answer several fundamental questions: What does the receptor look like? How does ACh bind? How is this binding converted into a conformational change? The major focus of this chapter is to answer the question of how acetylcholine binds to the nicotinic acetylcholine receptor. The studies in this chapter were carried out in collaboration with Wenge Zhong, who performed all of the experimental work described. 12 2.2 Characterizing the nAChR at the Molecular Level Because the nicotinic acetylcholine receptor is an integral membrane protein, it has resisted structural characterization at the molecular level. Standard high-resolution structural methods, such as x-ray crystallography and nuclear magnetic resonance spectroscopy, are challenged to provide molecular scale information about integral membrane proteins. This is often due in part to the lack of sufficient quantities of material to perform structural studies. In the case of the nAChR, Nature has provided a plentiful source of protein that can be used for structural and biochemical studies. The electric organ of certain species of rays and eels, such as To rpedo califomica, contains a high density of nAChRs. As a result, much is known about this receptor from both classical biochemistry as well as groundbreaking electron microscopy studies performed in the laboratory of Nigel Unwin [4- 7]. Our present view of the receptor suggests that it is a large protein (MW 290 kD) comprising five homologous subunits. The subunits are thought to be arranged in a pentagonal array around a central pore that forms the ion channel. The embryonic form of the receptor has a subunit stoichiometry of a 2 , (3, '"Y, <5. The arrangement of these subunits around the pore is still the subj ect of debate, but the most commonly advocated arrangement is a clockwise arrangement of a , '"Y, a, <5, (3 when viewed from the extracellular side of the membrane as shown in Figure 2.1. 2.2.1 Structural Data The three-dimensional structure of the nicotinic acetylcholine receptor has been determined to 9 A resolution [5] (and 4.6 A in projection) [7] using electron microscopy. Although these experiments do not reveal the atomic positions of either the protein backbone or individual side chains, they do offer insight into the global layout of the receptor with respect to the lipid membrane. Several structural features are highlighted in Figure 2.1. The receptor is approximately 120 A long and 85 A across at 13 -I·- 85 A ---- 8 -1 120 A j Figure 2.1: Structural features of the nAChR taken from the work of Unwin [5], with the subunit arrangement taken from Hucho et al. [3] its widest point. The extracellular domain extends approximately 76 A above the lipid bilayer, and the agonist binding sites, which are believed to be located primarily on the a-subunits, are proposed to lie approximately 30 A above the lipid bilayer [5] . It has been proposed that the gate of the receptor is a hydrophobic plug consisting of a ring of highly conserved leucine residues located within the membrane [8]. These residues , which are proposed to be located in the second membrane-spanning region (known as M2) of the individual subunits, are believed to be located over 50 A away from t he putative agonist binding sites (Figure 2.2) . 2.2.2 Sequence Landmarks Although protein isolated from Torpedo has provided many details about the structure and fun ction of the nicotinic acetylcholine receptor, it serves as only a starting point. As physical organic chemists, we would like to perform structure-function studies on the receptor. These studies were enabled by the cloning of the subunits of the receptor by Numa [9- 12] and Claudio [13], and the subsequent heterologous expression of the receptor in Xenopus oocytes by Numa [14] . The sequences reveal that the subunits of the receptor are highly homologous. Hydropathy analysis predicts that each subunit contains a large ( ca. 200 amino acid) extracellular domain, followed by four membrane-spanning regions (designated Ml-M4, Figure 2.3). 14 agonist binding sites "gate" Figure 2.2: Cutaway view of the nAChR. The agonist binding sites are postulated to be located over 50 A away from the gate of the ion channel. In the extracellular domain of each subunit , there are two cysteine residues that form a 15-residue disulfide-loop. This signature disulfide cys-loop is a hallmark of the ligand-gated ion channel superfamily and is present in all nicotinic receptors, as well as the receptors for glycine, serotonin, and 1-aminobutyric acid (GABA) [2]. In addition to the cys-loop, there is a vicinal disulfide linkage that is unique to a -subunits of the nAChR that is known to be near the ACh binding site. Although binding of acetylcholine is localized to the extracellular domain, the gate, which controls the passage of ions, is believed to be located in the membrane. Extensive studies of the receptor suggest t hat the M2 region of each subunit lines the ion channel , and that one residue in particular, leucine 9', forms the gate of the channel [15 , 16], although this model is not universally accepted [17, 18]. 2.2.3 The Agonist Binding Site The agonist binding sites of the nicotinic acetylcholine receptor are believed to be located at the interfaces of the a and I subunits and the a and i5 subunits [19- 22]. Binding of acetylcholine to these sites is cooperative and leads to channel opening [23]. A fundamental goal of this work is to identify which residues bind ACh, 15 a subunit cxx::::c:x:) N N E ' I I L C:X:::CC:O Figure 2.3: Schematic view of the a subunit of the nAChR. The other subunits are highly homologous and are believed to share the same overall topology. The aromatic residues highlighted have been affinity-lab eled (see text). and through what forces. Although no direct high-resolution structural data exist for the nAChR, pioneering labeling studies carried out on the Torpedo nAChR have implicated several amino acids that may be important in the binding of ACh. Using a variety of labeling reagents , such as dimethylaminobenzene diazonium tetrafluoroborate (DDF) [24, 25], 4-(N-maleimido)-benzyltrimethylammonium (MBTA) [26], nicotine [22], d-tubocurarine [27], lophotoxin [28], and acetylcholine mustard [29], several residues that are likely involved in the binding of ACh have been identified, including aTrp86, a Tyr93, a Trp149 , a Tyr151 , aTyr190 , aCys192 , aCys 193, a Tyr198 , ')'Trp55 , and 6Trp57. Not surprisingly, all labeled residues within the a -subunits are highly conserved and some (aTrp86, a Tyr151 , and aTyr198) have been called structural canonical residues. What is at first glance surprising is that the majority of labeled residues are aromatic. Conventional wisdom might suggest that to bind a cationic agonist such as acetylcholine, there must be an anionic residue. However, no anionic residues were directly labeled. This large number of aromatic amino acids led 16 Dougherty and Stauffer to propose that cation-n interactions might be important in the binding of acetylcholine to the nicotinic acetylcholine receptor [30]. The remainder of this chapter describes our attempts to test this proposal. 2.3 Cation-1r Interactions and Structure-Function Studies One of the most time-tested methods of physical organic chemistry is the determination of a structure-function relationship. Put simply, how do changes in molecular structure effect the function of a system? To answer this question , we must be able to do three things. We must first be able to perform structural changes on the nicotinic receptor. Second, we must be able to measure the effect of these structural changes on the function of the receptor. The third requirement is that we must be able to correlate the results from the structure and function studies to some known molecular property, in this case, cation-n binding ability. 2.3.1 Site-Directed Mutagenesis in Molecular NeurobiologyVarying Structure Arguably the most powerful weapon in the arsenal of the molecular neurobiologist is site-directed mutagenesis - changing one amino acid to another. By altering the molecular structure of an amino acid side chain, one can attempt to correlate the structural change with a functional change. Although site-directed mutagenesis is quite versatile, it is limited by the fact that each amino acid may only be changed to one of the other nineteen translationally-incorporated amino acids. As physical organic chemists, we would like finer control over the amino acid substitutions. Fortunately, work from the laboratories of Schultz [3 1, 32], Chamberlain [33- 35], and Hecht [36] has provided us with a method to expand the genetic code and sitespecifically incorporate unnatural amino acids into proteins expressed in vitro using a technique known as nonsense codon suppression . In the nonsense codon suppres- 17 sion method , a mutant mRNA is produced that contains a stop codon at the site of interest . Normally, this codon would signal the termination of translation because there is no tRNA molecule with an anticodon that recognizes the stop codon. To circumvent this , a tRNA with an anticodon that recognizes the stop codon is produced in vitro. This "suppressor" tRNA is chemically aminoacylated in vitro with the desired amino acid. By combining the mutant mRNA and aminoacyl-tRNA in an appropriate translation system, proteins containing the desired amino acid at the site of interest can be produced. Nonsense codon suppression is a powerful technique and has been used to site-specifically incorporate both natural and unnatural amino acids into proteins expressed in vitro [31- 36] . Unfortunat ely, as originally developed, the nonsense codon suppression method is not generally applicable to problems in molecular neurobiology. This is largely due to the fact that the proteins of molecular neurobiology are often complex, heavilyprocessed , mult i-subunit integral membrane proteins. Although these proteins may be expressed in vitro , they may not fold or assemble correctly, and thus they cannot be assayed in a proper biological context. Ideally, one would like to assay the function of these proteins in a biologically relevant context - within a living cell. In the past several years, the Dougherty and Lester laboratories have adapted the nonsense codon suppression method for use in a living cell, specifically a X enopus oocyte [37- 39]. Today, it is possible to site-specifically incorporate unnatural amino acids into proteins expressed in vivo. vVe now have unprecedented structural control over the molecular properties of individual amino acid side chains. Over 60 natural and unnatural amino acids have been incorporated into ion channel proteins expressed in a living cell. 2.3.2 Electrophysiological Recording - Measuring Function Site-directed mutagenesis is a powerful method of making structural changes in a protein . However, we must also be able to assay the fun ctional effects of such changes. Because the nicotinic acetylcholine receptor is an ion channel, one way to assay its function is to measure the flow of ions through the channel. This can be done using the 18 two-electrode voltage clamp technique. In the two-electrode voltage clamp technique, a cell expressing the protein of interest is pierced with two electrodes. The first electrode (known as the voltage electrode) serves to monitor the membrane potential of the cell. The membrane potential can be set to an arbitrary value and 'clamped' (i.e., maintained at a constant value). The second electrode (known as the current electrode) delivers current to the cell such that the membrane potential remains at the clamped value. If the membrane potential deviates from the clamped value due to the opening or closing of ion channels, the current electrode must deliver current to compensate for this change to maintain the clamped voltage. The current delivered is equal in magnitude, but opposite in sign, to the current passed by the ion channels. Thus, by measuring t he total amount of current required to maintain a voltage clamp (the whole-cell current), one can determine the total amount of current passing through the open channels. The nicotinic acetylcholine receptor is a ligandgated ion channel. In the absence of an agonist, such as acetylcholine, the probability of the channel opening is quite low. Application of ACh increases the probability that a conformational change will occur that allows ions to flow across the membrane. The whole-cell current of Xenopus oocytes expressing the nAChR can thus be measured in response to varying concentrations of ACh. The concentration of agonist that elicits half of t he maximal response is known as t he EC 50 for t hat agonist. It is important to note that although EC 50 measures the sensitivity of a receptor for a given agonist, it is not a binding constant. Rather, EC 50 is a composite value that incorporates the effects of both agonist binding and channel gating. Nonetheless, small variations in EC 50 caused by mutations often correlate well with the free energy change of binding of the ligand [40]. Thus, by measuring the EC 50 of ACh for a series of mutant receptors, we can quantitatively evaluate receptor function. 2.3.3 Correlating Structure and Function It is now clear that we can vary t he structure of the nAChR by using nonsense codon suppression. In addition, we can assay the effects of the mutations by measuring the 19 sensitivity of the mutant receptors to ACh by recording the EC 50 . The remaining challenge is to correlate changes in the structure and function of the receptor to a change in the cation-1r binding ability of an individual amino acid side chain. Fortunately, Anthony West and Sandro Mecozzi convincingly demonstrated that variations in the cation-1r binding abilities of a structurally-related series of aromatics could be accounted for by variations in the electrostatic potential surfaces of the aromatics [41 , 42]. Thus, we can modify the electrostatic potential (and thus the cation-1r binding ability) of an amino acid via organic synthesis; we can evaluate the variations using ab initio quantum mechanical calculations; and we can measure the changes in EC 50 for ACh for receptors containing the mutant side chains. If the changes in EC 50 for a series of closely-related aromatics at a particular site in the receptor follow a consistent electrostatic trend, we will have evidence that the particular amino acid is involved in a cation-1r interaction with acetylcholine. 2.4 Previous Studies of Possible Cation-1r Interactions in the nAChR As noted in Section 2.2, several aromatic amino acids have been implicated in the binding of acetylcholine by the nAChR, including aTrp86 , a Tyr93 , a Trp149 , a Tyr1 51, aTrp184, a Tyr190 , a Tyr198, and the homologous 1 Trp55 and o"Trp57. Previous work from the Dougherty group explored the possibility t hat the tyrosines at positions a93 , a 190, and a 198 might bind ACh through a cation-1r interaction [37]. Using the in vivo nonsense suppression method, it was determined that the hydroxyl group of a Tyr93 functions as a hydrogen bond donor, and that a Tyr 198 serves as a steric place holder in the receptor. a Tyr190 was found to be very sensitive to subtle mutations, but no consistent structure-function relationship could be derived [37]. Perhaps surprisingly, none of the t hree tyrosines studied appeared to bind to ACh through a cation-1r interaction. In addition to the tyrosine residues, there are several conserved tryptophans that have been affinity labeled. It might be anticipated that a tryptophan would con- 20 -27.1 -26.9 -32.6 Figure 2.4: Calculated HF /6-31G** electrostatic potential surfaces and Na+ binding energies (see text) of the aromatic side chains of phenylalanine (benzene), tyrosine (phenol), and tryptophan (indole). Red represents a region of negative electrostatic potential, blue positive. (Scale: -25 to +25 kcal/mol) . stitute a cation-1r binding site because the indole side chain of tryptophan provides the most potent cation binding site among the aromatic amino acids (Figure 2.4) [42]. There are relatively few studies of the roles of the conserved tryptophans in the nAChR. Both White [43] and Changeux [44] have demonstrated that 1Trp55 and <5Trp57 are important binding determinants of both ACh and d-tubocurarine in the receptor from Torpedo. In addition, Changeux has shown that a Trp-+ Phe mutation at position a 148 in the a7 neuronal receptor (analogous to a 149 in the muscle receptor) increases the EC 50 for ACh nearly one hundred-fold [44] . Finally, in an NMR study of peptide fragments from the a -subunit, Fraenkel reported that the quaternary ammonium group of ACh contacts the indole side chain of a184 [45]. Although these studies suggest the possibility of cation-1r interactions between ACh and the tryptophans in the nAChR, they do not constitute proof that such an interaction exists. The remainder of this chapter describes our efforts to find a cation-1r binding site within the nicotinic acetylcholine receptor. 21 2.5 The Search for a Cation-1r Binding Site within the nAChR 2.5.1 Theory With several potential candidates for a cation-1r binding site in the nAChR, we set out to predict the cation-1r binding ability of a series of tryptophan derivatives that could be used as replacements in a nonsense codon suppression experiment. Previous studies by Anthony West and Sandro Mecozzi have shown that the cation-1r binding ability of a simple aromatic can be quantitatively determined by using high-level ab initio calculations [41 , 42]. To predict the cation-1r binding ability of an aromatic, the binding energy between the aromatic ring and a probe cation, such as Na+, is determined using an ab initio calculation. Earlier studies have suggested that Hartree-Fock calculations with a medium-sized basis set, such as 6-31G**, can accurately reproduce experimentally-determined binding energies of aromatics and non-polarizable cations [46]. In addition, incorporation of the effects of electron correlation using post-Hartree-Fock methods, such as M0ller-Plesset (MP) perturbation theory, does not significantly improve the agreement between experiment and theory in systems containing non-polarizable cations. Thus, we used Hartree-Fock theory with the 631G** basis set to calculate the Na+ binding energy for the series of indole analogs shown in Figure 2. 5. To perform the calculations, a Na+ ion was placed over the ring of interest, and the geometry of the complex was fully optimized using Gaussian 94 or Spartan. The binding energies were calculated using the supermolecule method and are not corrected for basis set superposition error (BSSE) [47], which is known to be small in these simple systems. The calculated binding energies of these systems are shown in Table 2.1. There are several items worthy of explanation in Table 2.1. Perhaps most surprising is the extraordinary cation-1r binding ability of the indole side chain of tryptophan. Of the 22 indole analogs shown in Table 2.1, only three bind cations more strongly 22 (()N H Trp H3C'O)I ~ ::,... N H (()N CH3 H2Nm~ ::,... I N H 5-Me-Trp N-Me-Trp 5-NHrTrp (X)N Qq 7-F-Trp \ F Or) Fml~ 4-F-Trp 5-F-Trp 6-F-Trp N H Fml~ F ::,... N H F H F~I~ F ::,... N F H 5,7-FrTrp 5,6-FrTrp 6,7-FrTrp 5,6,7-F3-Trp m FLO~ ,_ I Cll.Ll I ~ ,_ N N H 4,5,6,7-F4-Trp 7-aza-Trp 5-F-7-aza-Trp 5-Cl-7-aza-Trp NCO) I ~ ::,... N H Brm I ~ ()) ()) 5-CN-Trp 5-Br-Trp 1-Np-Ala 2-Np-Ala 00 0 (IQs co Bf-Ala Bt-Ala 2-Qu-Ala N H F w .~l ::,... F F F~I~ F ::,... F N H ::,... N N H H yQ F N N H ::,... F N H N H A F N H A N" Figure 2.5: Side chain structures of a series of tryptophan analogs. The derivatives are named as amino acids; thus indole is considered Trp. 23 Aromatic Trp 5-Me-Trp N-Me-Trp 5-NH2-Trp 4-F-Trp 5-F-Trp 6-F-Trp 7-F-Trp 5,7-F 2-Trp 5,6-F 2-Trp 6,7-FrTrp 5,6,7-F3-Trp 4,5,6,7-F 4 -Trp 7-aza-Trpt 5-F-7-aza-Trpt 5-Cl-7-aza-Trp r 5-CN-Trp 5-Br-Trp + 1-Np-Ala, 2-Np-Ala[::. Bf-Ala Bt-Ala 2-Qu-Ala Cation-1r binding of 6-ring (kcal/mol) 32.6 33.4 33.7 36.4 27.9 27.5 27.4 28.2 23.3 23.2 23.4 18.9 14.4 26.0 21.2 20.8 21.5 27.8 28.9 26.9 26.9 25.8 Cation-1r binding of 5-ring (kcal/mol) 29.0 n.d. n.d. n.d. n.d. 25.4 n.d. n.d. 21.9 n.d. n.d. 18.9 15.6 25.0 n.d. n.d. 19.1 n.d. n.a. 22.3 n.d. n.a. Table 2.1: Cation-1r binding energies for the aromatic derivatives shown in Figure 2.5. The cation-1r binding is defined as the negative of the binding energy (kcal/mol) of a Na+ ion to the aromatic ring determined from a HF /6-31G** calculation. In cases where the 5-membered ring of the aromatic might compete with the 6-membered ring for cation binding, the binding energy to the 5-membered ring is presented. t The true minima for the aza-Trp derivatives have the cation bound to the pyridine nitrogen. The values reported in the table in some cases had to be constrained to the ring axis. t The value for 5-Br-Trp is determined from an HF /6-311G** calculation (see text). 6 1-N p-Ala and 2-N p-Ala are listed together because they have the same side chain. The Np-Ala values were calculated by Sandro Mecozzi [41]. n.a. - not applicable, n.d. - not determined. 24 35 0 30 Ei ---cilu c >-. oO .... <!) 25 i:: i:.tJ oO i:: :ai:: i:i5 +"' 20 z 15 0 2 3 4 Number of Fluorines Figure 2.6: The calculated HF /6-31G** Na+ binding energies of a series of indole derivatives fluorinated on the 6-membered ring. Each fluorine reduces the Na+ binding energy by 4.5 kcal/ mol. (R = 0.99). than indole itself. The most potent of these, 5-NH 2 -indole, might be expected to easily oxidize and be less stable. Thus, it would appear that Nature has chosen a particularly strong, yet chemically stable, cation-binder in the indole side chain of tryptophan. Although it is difficult to strengthen the cation-n binding ability of the indole ring, it is quite easy to weaken it. Consistent with previous studies, electronwithdrawing substituents, such as fluorine and cyano groups, substantially weaken the cation-n binding ability of an aromatic ring [41]. In addition, increased fluorination directly correlates with decreased cation-n binding ability. As shown in Figure 2.6, each additional fluorine reduces the cation-n binding ability of the ring by a constant 4.5 kcal/ mol. The data in Table 2.1 clearly suggest a broad range of cation-n binding abilities for the various indole analogs. Thus, it should be possible to synthesize a series of tryptophan derivatives that have widely differing cation-n binding abilities. However, to establish whether a cation-n interaction is occurring, we need to show a correlation between the predicted cation-n binding ability of a given amino acid and the function 25 Figure 2.7: Calculated electrostatic potential surfaces of the side chains of Trp, 4,5,6, 7F 4 -Trp, 5-Br-Trp, and 5-CN-Trp. It is clearly seen that the introduction of electronwithdrawing groups, such as fluorine and cyano groups, lowers the electrostatic potential above the rings. In addition, this figure highlights the similar size of the bromine and cyano substituents. The surfaces are calculated at the HF /6-31G** level for Trp, 4,5,6,7-F 4 -Trp, and 5-CN-Trp, and at the HF /6-311G** level for 5-Br-Trp. Justification for the use of the 6-311G** basis set appears in an appendix to this chapter. (Scale: -25 to +25 kcaljmol). of a receptor containing that amino acid. Four entries in Table 2-1 are particularly useful in this regard: Trp, 4,5,6,7-F4 -Trp, 5-Br-Trp, and 5-CN-Trp. The calculated electrostatic potential surfaces of these side chains are shown in Figure 2. 7. Figure 2. 7 highlights several important comparisons between the various side chains. It is clearly seen, both qualitatively and quantitatively (Table 2.1), that the addition of four fluorines reduces the cation-rr binding ability of the indole side chain of tryptophan. Fluorine is generally regarded as a conservative steric substitute for a hydrogen atom. Thus, the substitution of four fluorines alters the electronic properties on the ring dramatically, while leaving the steric properties unperturbed. In a similar manner, 5-Br-Trp and 5-CN-Trp are nearly isosteric, yet their cation-rr 26 binding abilities are quite different. Thus, by mutating the conserved tryptophans in the nAChR to only these 3 Trp derivatives, we can screen the effects of varying electronic structure (and thus cation-n binding ability) , while keeping the steric profile of the side chains the same. If the properties of mutant receptors containing these derivatives are similar, it suggests that changes in cation-n binding ability of the side chains do not correlate with a change in receptor function . However, if mutant receptors show functional differences that correlate with the changes in cation-n binding ability of the side chains, there is evidence of a cation-n interaction, and additional tests can be performed using other analogs shown in Figure 2.5. 2.5.2 Experiment To determine whether one or more of the conserved tryptophans in the nicotinic acetylcholine receptor binds ACh through a cation-n interaction , Wenge Zhong synthesized several of the Trp analogs outlined in Figure 2.5. Using t he nonsense codon suppression method, he incorporated Trp and the three analogs , 4,5,6,7-F4 -Trp , 5Br-Trp, and 5-CN-Trp at several sites within the nAChR. The EC 50 for ACh for the wild-type Trp and each of the three Trp derivatives was measured at several sites, a86, a 149, a184, 155, and 557, and the results are reported in Table 2.2. Side chain Trp 5-Br-Trp 5-CN-Trp 4,5,6 ,7-F 4 -Trp a86 50 53 61 45 ECso (µM) a 149 a184 155/557 50 50 50 48 23 88 46 25 4750 2700 62 92 Table 2.2: EC 50 values for ACh for several Trp mutants at a86, a 149, a184, 1 55, and 557. There are only substantial variations at a149 , suggesting a possible cationn interaction between ACh and a Trp149. The mutations in the 5 and 1 subunits are expressed simultaneously, so each agonist binding site contains the mutant Trp. Because the EC 50 is so high for some mutations, it is difficult to measure accurately. To make these receptors more responsive to ACh, an additional mutation, Leu 9' -Ser was introduced and the values were extrapolated. For full details of this process, see [48]. The results in Table 2. 2 suggest that at three of the four Trp sites (a86, a 184, 27 Aromatic Trp 5-Me-Trp N-Me-Trp 5-NH2-Trp 4-F-Trp 5-F-Trp 6-F-Trp 5,7-F2-Trp 5,6,7-F3-Trp 4,5,6,7-F 4 -Trp 7-aza-Trp 5-CN-Trp 5-Br-Trp 1-Np-Ala 2-Np-Ala Bt-Ala Cation-n binding (kcal/mol) 32.6 33.4 33.7 36.4 27.9 27.5 27.4 23.3 18.9 14.4 26.0 21.5 27.8 28.9 28.9 26.9 ECso (µM) 50 49 95 280 56 200 48 550 1400 2700 130 4750 88 51 82 174 Table 2.3: Calculated cation-n binding abilities and experimentally-measured EC 50 values for ACh for various tryptophan derivatives incorporated into the nAChR. o:55/o:57), large variations in the cation-n binding ability of the amino acid side chain do not significantly alter the sensitivity of the receptor for ACh. Thus, there is little evidence to suggest a cation-n interaction at these sites. However, at o:149, there is a large change in the EC 50 for ACh that does correlate with the cation-n binding ability of the amino acid side chains. In particular, the EC 50 values for both 5-CNTrp and 4,5 ,6, 7-F4 -Trp are substantially higher than the values for either 5-Br-Trp or Trp. This is consistent with weaker ACh binding leading to a higher EC 50 . To further investigate the possibility of a cation-n interaction between the quaternary ammonium group of ACh and the side chain of o:Trpl49, Wenge Zhong synthesized several other Trp analogs, incorporated them into the nAChR, and measured the EC 50 for ACh. The results of these experiments are shown in Table 2.3. Although the results for 5-CN-Trp are highly suggestive of a cation-71' interaction between o:Trp149 and ACh, an even more compelling correlation is observed for the series of fluorinated Trp derivatives shown in Table 2.3. In Figure 2.8, the experimentally-determined log(EC 50 ) values for a series of fluorinated Trp derivatives 28 2.0 • ~ F N H 1.5 ~ ~ u"' 1.0 F ~o • N H FY'y\ • -i:.r.l YN F H 0 u"' i:.r.l c.O .3 • 0.5 ~ lvLJ ~ H t,J-N' H • 0.0 15.0 20.0 25 .0 30.0 35.0 Cation-rr binding Figure 2.8: Plot of log[EC 50 /EC 50 (wild type)] vs. quantitative cation-1r binding ability of Trp and its fluorinated derivatives expressed at a149 . Data are taken from Table 2.3 and fit the line y = 3.2 - 0.096x, R = 0.99. The log[EC 50 ] values are used to place the y-axis on an energy scale. is plotted against the calculated cation-1r binding abilities of the derivatives. Figure 2.8 incorporates two very different measurements. On one axis are the predicted cation-1r binding abilities for the various amino acid side chains calculated using ab initio quantum mechanics. On the other axis are experimental measurements performed on a molecular weight 290,000 protein expressed in a living cell. Nonetheless, a striking correlation is observed between these two very different measurements. It should again be stressed that the EC 50 for ACh is not a binding constant. Rather, it incorporates the effects of both binding and gating. In Figure 2.8, we do not assume that gating is unimportant, but rather that the changes in EC 50 are primarily due to changes in the binding affinity for ACh. To the extent that EC 50 is proportional to a binding constant, log[EC 50 ] is the appropriate quantity to correlate with a binding energy. The EC 50 change observed on going from Trp to 4,5,6,7-F 4 -Trp represents an energy difference of 2.4 kcal/mol of binding energy. Since it is unlikely that this mutation substantially alters the structure of the receptor, it is reasonable to assume that this energy change is directly related to the binding of ACh. The energy dif- 29 2.0 • • 1.5 • ~ ~ ~o • 1.0 u"' --u i:.Ll • 0 "' ~ 0.5 • bl) .3 •• •• ••• 0.0 15.0 20.0 25.0 30.0 • •• 35.0 Cation-rt binding Figure 2.9: Plot of log[EC 50 /EC 50 (wild type)] vs. quantitative cation-1r binding ability of all Trp derivatives expressed at a149. Data are taken from Table 2.3. The outliers correspond to 5-CN-'Irp and 5-NH 2 -Trp. The correlation coefficient for the line shown is 0.74. Removing the outliers improves the correlation to R = 0.90. ference of 2.4 kcal/mol is quite large for a noncovalent interaction and thus suggests that ACh and the 'Irp side chain are interacting directly. In addition to the data from the fluorinated tryptophans, a plot containing all of the data in Table 2.3 reveals a similar trend as shown in Figure 2.9. The outlying data points in Figure 2.9 correspond to 5-CN-'Irp and 5-NH 2-Trp. It is not clear why these particular amino acids are outliers, but there is a possibility that 5-NH 2-Trp is protonated or oxidized and thus leads to a higher than expected EC 50 . Although there is scatter in the data, the trend is clear that a decrease in cation-1r binding ability of an amino acid side chain correlates with an increase in EC 5o for ACh. Until this point , we have assumed that the ACh is in contact with the 6-membered ring of the 'Irp. This is a reasonable assumption given that it is a more potent cation1r binder than the 5-membered ring. However, there is a possibility that ACh is in fact binding to the 5-membered ring of the 'Irp. A plot of the EC 50 data taken from 30 2.0 • 1.5 ~ ~ o 1.0 • • u'° "'-l --u'° 0 "'-l ~ • 0.5 .3 • 0.0 -0.5 ' - - - - - - ~ - - - - - - ' - - - - - - - - - - - ' 15.0 20.0 25.0 30.0 Cation-it binding to 5-membered ring Figure 2.10: Plot of log[EC 50 / EC 50 ( wild type)] vs. quantitative cation-7!' binding ability of the 5-membered ring of Trp and its fluorinated derivatives expressed at cd49. Data are taken from Table 2.3 and fit the line y = 3.84 - 0.13x, R = 0.99. Table 2.3 vs. the calculated cation-7!' binding abilities of the 5-membered rings of Trp and its fluorinated derivatives (taken from Table 2.1) is shown in Figure 2.10. Indeed an excellent correlation is also observed when considering the 5-membered rings of Trp and its fluorinated derivatives. However , if the ACh were in contact with the 5-membered ring of t he Trp, it might be expected that phenylalanine, with a side chain aromatic t hat aligns with the 5-membered ring of the Trp, would make a good substitute for Trp. Experimentally, this is not the case. Changeux has observed that an analogous Trp--+ Phe mutation in the a 7 neuronal nAChR markedly increased the EC 50 for ACh [44], and we find that an a Trpl49--+Phe mutation increases the EC 50 for ACh approximately 100-fold. Thus it is likely that the quaternary ammonium group of ACh is in contact with the 6-membered ring of the Trp. However, one cannot exclude the possibility t hat the 6-ring of Trp positions the side chain such that ACh can bind to the 5-ring, whereas Phe, lacking a second place-holder ring, might not be able to bind as well. An amino acid such as 4,5,6,7-tetrahydro-Trp might be able to experimentally distinguish the two possibilities. 31 Figure 2.11: Side chain structure of the tethered agonist Tyr-0-3Q. Finally, inspired by the information derived from this study, Wenge Zhong incorporated into position a149 an amino acid with a quaternary ammonium group tethered to its side chain (Tyr-0-3Q, Figure 2.11). Incorporation of Tyr-0-3Q at position a149 produces a constitutively active receptor. In the absence of ACh, standing inward currents that can be inhibited by nAChR open-channel blockers are seen in Xenopus oocytes expressing the mutant channel. The side chain of Tyr-0-3Q positions the quaternary ammonium group such that it can mimic the binding of ACh to the 6-membered ring of aTrpl49. This result and others further suggest that ACh binds to aTrpl49 of the nAChR via a cation-71' interaction. 2.6 Summary The work presented in this chapter illustrates the power of combining theoretical and physical organic chemistry with modern neurobiological techniques to provide chemical-scale information of a complex molecular machine. By combining ab initio quantum mechanical calculations, organic synthesis, and molecular neurobiology, we provide compelling evidence that the quaternary ammonium group of acetylcholine binds to the nAChR through a cation-71' interaction with aTrp149. In doing so, we provide some of the highest-resolution structural information known about this receptor. In addition, we demonstrate the power of the nonsense codon suppression method to apply the principles of physical organic chemistry to complex biological systems. 32 Aromatic 6-31G** minimum energy of monomer (Hartrees) Trp 5-Me-Trp N-Me-Trp 5-NH 2-Trp 4-F-Trp 5-F-Trp 6-F-Trp 7-F-Trp 5,7-F 2-Trp 5,6-F2-Trp 6,7-F2-Trp 5,6,7-F3-Trp 4,5,6,7-F 4 -Trp 7-aza-Trp 5-F-7-aza-Trp 5-Cl-7-aza-Trp 5-CN-Trp 5-Br-Trp t Bf-Ala Bt-Ala 2-Qu-Ala - 361.4829178 -400.5221750 - 400.5146125 - 416.513102 - 460.3330449 - 460.3316404 - 460. 3327604 - 460. 3308288 - 559.1788382 - 559 .173604 7 - 559.1723013 - 658.0126328 - 756.8518351 - 377.4819407 - 476.3267973 -836.3784228 - 453.2199701 - 2933.331860 - 381.2991455 - 703. 9623685 - 399.3623064 6-31G** minimum energy of Na+ complex with 6-membered ring (Hartrees) - 523 .1941820 -562.2346708 - 562.2275997 - 578.2304357 - 622.0367912 - 622.0347153 - 622.0357012 - 622.0351006 - 720.8752736 - 720.8691313 - 720.8688588 - 819. 7020758 - 918.5340764 - 539.1826815 -638.0199006 - 998.0708002 - 614.9135390 - 3095. 036333 -543.0012739 - 865. 6646036 -561 .0627336 Table 2.4: Total energies of monomers and 6-ring Na+ complexes. t 5-Br-Trp is calculated using the 6-311G** basis set (see text). E(Na+) (HF /6-31G**) = -161.6592883 Hartrees. E(Na+) (HF / 6-311G**) = -161.6642329 Hartrees. 2. 7 Appendix 1 - Details of Calculations All calculations in this chapter were performed using Gaussian 94, Rev. D3. [49], or Spartan (versions 4 and 5) [50]. Na+ binding energies were determined by the supermolecule method in which the total energy of the complex was minimized, and minimum energies of the individual partners were subtracted. In nearly all cases, Hartree-Fock theory was used with the 6-31G** basis set. The total energies of the individual molecules and their Na+ complexes are given in Table 2.4 and Table 2.5. Because bromine is not defined in the 6-31G** basis set, the calculations for 5Br-Trp had to be performed using the 6-311G** basis set. The binding energy from 33 Aromatic Trp 5-F-Trp 5,7-F2-Trp 5,6 ,7-F3-Trp 4,5,6,7-F4 -Trp 7-aza-Trp 5-CN-Trp Bf-Ala 6-31G** minimum energy of Na+ complex with 5-membered ring (Hartrees) - 523.1884613 - 622.0314230 - 720.8730516 - 819.7020001 - 918.5359377 - 539.1810810 - 614.9096883 - 542. 9939364 Table 2.5: Total energies of the 5-ring Na+ complexes. the 6-311G** calculation was extrapolated back to a 6-31G** binding energy based on a linear correlation of Na+ binding energies calculated in both basis sets as shown in Figure 2.12. Clearly, Na+ binding energies calculated using the two basis sets correlate well, although binding energies calculated with the larger 6-311G** basis set are a constant 2.25 kcal/mol less than those calculated with the 6-31G** basis set. In addition to the total energies of the monomers, the electrostatic potential surfaces of each molecule were calculated at the HF /6-31G** level (HF /6-311G** for 5-Br-Trp). These are presented in Figure 2.13. All are displayed on an equal energetic scale, ranging from - 25 kcal/mol (red) to +25 kcal/mol (blue). 2.8 Appendix 2 - Related Cation-1r Calculations In addition to the calculations previously described in this chapter, several related calculations involving simple ions binding to aromatics were performed and are briefly presented in this Appendix. To evaluate the effects of both theoretical method and basis set size on the calculated Na+••• benzene binding energies, a series of calculations was performed (Table 2.6). All complexes were constrained to have C6 v symmetry and are fully geometry optimized. It is clearly seen that using Hartree-Fock theory, the binding energies 34 35 30 0 6 ---elu ~. ;,..,• 25 OllCJ ......... O.i C'") C ' [..Ll'-0 bl)--- c.r.. :a·2 :r:: '9 z 20 • +11) 15 Na+-Biniding Energy (kcal/mo!) HF/ 6-31 lG** Figure 2.12: Plot of HF /6-311G** vs. HF /6-31G** Na+ binding energies for a series of aromatics. The data fit the line y = 2.25 + l.Olx with a correlation coefficient of R = 0.998. The aromatics, left to right, are: 1,3,5-trifluorobenzene, 1,4difluorobenzene, 1,4-dichlorobenzene, chlorobenzene, fl.uorobenzene, 5-cyanoindole, benzene, 5-fluoroindole, and aniline. decrease with increasing basis set size. Although consideration of basis set superposition error might mitigate some of this effect, there is nonetheless a substantial range of energies. Interestingly, both HF and B3LYP theory using the 6-31G** basis set perform quite well with respect to predicting the experimentally determined binding energies. This may be due to fortuitous cancellation of errors [46]. In addition to the calculations involving Na+, a series of calculations involving ammonium ion (NHt) binding to benzene was performed. In addition to the ab initio calculations shown in Table 2.7, calculations using the AMl and PM3 semiempirical methods are presented. These methods perform poorly, and both severely underestimate the experimentally-determined binding energy of the complex. Similar to the calculations for the Na+··· benzene complex, most methods and basis sets underestimate the experimental binding energy. This is likely due to the fact that Hartree-Fock methods neglect the effects of electron correlation, and thus underestimate some of the van der Waals attraction between the partners. Density functional 35 Trp 5-Me-Trp N-Me-Trp 5-NHrTrp 4-F-Trp 5-F-Trp 6-F-Trp 7-F-Trp 5,7-FrTrp 5,6-FrTrp 6,7-FrTrp 5,6,7-FTTrp 4,5,6,7-FrTrp 7-aza-Trp 5-F-7-aza-Trp 5-Cl-7-aza-Trp 5-CN-Trp 5-Br-Trp Np-Ala Bf-Ala Bt-Ala 2-Qu-Ala Figure 2.13: Calculated HF /6-31G** electrostatic potential surfaces for all aromatic side chains. (Scale: - 25 to +25 kcaljmol). 36 Method Basis Set HF HF HF HF HF HF HF B3LYP BPW91 Experiment 3-21G* 3-21G** 6-31G* 6-31+G* 6-31G** 6-31+G** 6-311G** 6-31G** 6-31G* - Binding Energy (kcal/ mol) 34.6 34.2 27.0 23.5 27.1 23.5 24.2 28.4 25.9 28.0 Distance A 2.36 2.36 2.45 2.48 2.45 2.47 2.47 2.38 2.40 Table 2.6: Calculated-binding energies and distances for the Na+ ·· · benzene complex at several levels of theory. methods, which include the effects of electron correlation, fare somewhat better than the HF methods. Finally, Table 2.8 presents the binding energies of a series of Na+ ··· aromatic complexes . Again, the trend is observed that binding energies decrease with an increase in the size of the basis set. Interestingly, the results of the HF calculations at the 6-31G* and 6-31G** basis sets are nearly identical - most differences are due to round-off errors. This suggests that although the polarization functions on the hydrogen atoms represent an additional computational cost, they do not improve the accuracy of the calculations. This might be anticipated because the hydrogens are not directly involved in the interaction between the ring and the cation. However, in the case of the ammonium ion above, a similar trend is seen, even though the NHt hydrogens are pointing into the face of the ring. 37 Method AMI PM3 HF HF HF HF B3LYP B3LYP B3LYP BPW-91 BPW-91 BPW-91 Experiment Basis Set Symmetry C3v C3v 6-31G* 6-31G* 6-31G** 6-31G** 6-31G** 6-31G** MIDI!(6D) 6-31G* 6-31G* MIDI!(6D) C3v C2v C3v C2v C3v C2v C2v C3v C2v C2v -Binding Energy (kcal/mol) 9.2 6.8 14.1 15.3 14.1 15.3 15.7 17.5 18.4 14.5 16.6 17.5 19.3 Table 2. 7: Calculated - binding energies for the NHt · · · benzene complex at differing symmetries and levels of theory. Benzene Fluorobenzene 1,4-Difluorobenzene 1,3,5-Trifluorobenzene Chlorobenzene 1,4-Dichlorobenzene Aniline 3-21G** 34.2 28.1 22.1 16.7 28.7 23.7 40.0 Hartree-Fock 6-31G* 6-31G** 27.1 27.0 21.9 22.0 16.7 16.8 12.4 12.4 21.4 21.5 16.4 16.4 31.7 31.8 6-311G** 24.2 19.2 14.1 9.8 19.2 14.6 29.1 B3LYP 6-31G** 28.4 24.4 20.2 16.8 23.5 19.1 n.d. Table 2.8: Calculated - binding energies for various Na+•·· aromatic complexes. All energies are in kcal/mol. 38 Bibliography [1] A. Devillers-Thiery, J. L. Galzi, J. L. Eisele, S. Bertrand, D. Bertrand, and J.P. Changeux. Functional architecture of the nicotinic acetylcholine receptor: A prototype of ligand-gated ion channels. J. Membrane Biol. , 136:97- 112, 1993. [2] A. Karlin and M. H. Akabas. Toward a structural basis for the function of nicotinic acetylcholine receptors and their cousins. Neuron, 15:1231- 1244, 1995. [3] F . Hucho, V. I. Tsetlin, and J. Machold. The emerging three-dimensional structure of a receptor - the nicotinic acetylcholine receptor. Eur. J. Biochem., 239:529- 557, 1996. [4] E. Kubalek, S. Ralston, J. Lindstrom, and N. Unwin. Location of subunits within the acetylcholine receptor by electron image analysis of tubular crystals from Torpedo marmorata. J. Cell Biol. , 105:9- 18, 1987. [5] N. Unwin. Nicotinic acetylcholine receptor at 9 A resolution. J. Mal. Biol. , 229: 1101- 1124, 1993. [6] N. Unwin. Projection structure of the nicotinic acetylcholine receptor: Distinct conformations of the a subunits. J. Mal. Biol., 257:586- 596, 1996. [7] A. Miyazawa, Y. Fujiyoshi, M. Stowell, and N. Unwin. Nicotinic acetylcholine receptor at 4.6 A resolution: Transverse tunnels in the channel wall. J. Mal. Biol., 288:765- 786, 1999. [8] N. Unwin. Acetylcholine receptor channel imaged in the open state. Nature, 373:37- 43, 1995. [9] M. Noda, H . Takahashi, T . Tanabe, M. Toyosato, Y. Furutani, T. Hirose, M. Asai, S. Inayama, T. Miyata, and S. Numsa. Primary structure of a-subunit 39 precursor of Torpedo californica acetylcholine receptor deduced from cdna sequence. Nature, 299:793- 797, 1982. [10] M. Noda, H. Takahashi, T. Tanabe, M. Toyosato, S. Kikyotani, T. Hirose, M. Asai, H. Takashima, S. Inayama, T. Miyata, and S. Numa. Primary structures of {3 and 5 subunit precursors of Torpedo californica acetylcholine receptor deduced from cdna sequences. Nature, 301:251- 255, 1983. [11] M. Noda, H. Takahashi, T. Tanabe, M. Toyosato, S. Kikyotani , Y. Furutani, T. Hirose, H. Takashima, S. Inayama, T. Miyata, and S. Numa. Structural homology of Torpedo californica acetylcholine receptor subunits. Nature, 302:528532, 1983. [12] M. Noda, Y. Furutani, H. Takahasghi, M. Toyosato, T. Tanabe, S. Shimizu, S. Kikyotani , T. Kayano, T. Hirose, S. Inayama, and S. Numa. Cloning and sequence analysis of calf cdna and human genomic dna encoding a -subunit precursor of muscle acetylcholine receptor. Nature, 305:818-823, 1983. [13] T. Claudio, M. Ballivet, J. Patrick, and S. Heinemann. Nucleotide and deduced amino acid sequences of Torpedo californica acetylcholine receptor --y subunit. Proc. Natl. Acad. Sci. USA, 80:1111 , 1983. [14] M. Mishina, T. Kurosaki , T. Tobimatsu, Y. Morimoto, M. Noda, T. Yamamoto, M. Terao, J. Lindstrom, T. Takahashi, M. Kuno, and S. Numa. Expression of functional acetylcholine receptor from cloned cDNAs. Nature, 307:604-608, 1984. [15] P. Charnet, C. Labarca, R. J. Leonard, N. J. Vogelaar, L. Czyzyk, A. Gouin, N. Davidson, and H. A. Lester. An open-channel blocker interacts with adjacent turns of a-helices in the nicotinic acetylcholine receptor. Neuron, 2:87- 95, 1990. [16] M.H. Akabas, D.A. Stauffer, M. Xu, and A. Karlin. Acetylcholine receptor channel structure probed in cysteine-substitution mutants. Science, 258:307310, 1992. 40 [17] H. Zhang and A. Karlin. Identification of acetylcholine receptor channel-lining residues in the ml segment of the ,6-subunit. Biochemistry, 36: 15856- 15864, 1997. [18] H. Zhang and A. Karlin. Contribution of the ,6 subunit m2 segment to the ionconducting pathway of the acetylcholine receptor. Biochemistry, 37:7952- 7964, 1998. [19] P. Blount and J.P. Merlie. Neuron, 3:349- 357, 1989. [20] S. E. Pedersen and J. B. Cohen. d-tubocurarine binding-sites are located at a"/ and a- 8 subunit interfaces of the nicotinic acetylcholine-receptor. Proc. Natl. Acad. Sci. USA, 87:2785- 2789, 1990. [21] C. Czajkowski and A. Karlin. Agonist binding site of Torpedo electric tissue nicotinic acetylcholine receptor. J. Biol. Chem., 266:22603- 22612, 1991. [22] R. E. Middleton and J. B. Cohen. Mapping of the acetylcholine binding site of the nicotinic acetylcholine receptor: [3 H]nicotine as an agonist photoaffinity label. Biochemistry, 30:6987- 6997, 1991. [23] J.-1. Galzi and J.-P. Changeux. Neuronal nicotinic receptors: Molecular organization and regulations. Neuropharmacology, 34:563- 582 , 1995. [24] M. Dennis, J. Giraudat, F. Kotzyba-Hibert, M. Goeldner, C. Hirth, J.-Y. Chang, C. Lazure, M. Chretien, and J.-P. Changeux. Amino acids of the Torpedo marmorata acetylcholine receptor a subunit labeled by a photoaffinity ligand for the acetylcholine binding site. Biochemistry, 27:2346- 2357, 1988. [25] J.L. Galzi, F. Revah, D. Black, M. Goeldner, C. Hirth, and J.-P. Changeux. Identification of a novel amino acid a-tyrosine 93 within the cholinergic ligandsbinding sites of the acetylcholine receptor by photoaffinity labeling. Chem., 265:10430- 10437, 1990. J. Biol. 41 [26] P. N. Kao , A. J. Dwork, R. R. J. Kaldany, M. L. Silver, J. Wideman, S. Stein, and A. Karlin. Identification of the a subunit half-cystine specifically labeled by an affinity reagent for the acetylcholine receptor binding site. J. Biol. Chem., 259: 11662- 11665, 1984. [27] D. C. Chiara and J. B. Cohen. Identification of amino-acids contributing to high and low affinity. d-tubocurarine (dtc) sites on the Torpedo nicotinic acetylcholinereceptor (nAChR) subunits. FASEB J., 6:A106- A106, 1992. [28] S.N. Abramson, W. Fenical, and P. Taylor. Lophotoxins: Irreversible activesite-directed inhibitors of nicotinic acetylcholine receptors. Drug Develop. Res., 24:297- 312, 1991. [29] J.B. Cohen, S.D. Sharp, and Liu W.S. Structure of the agonist-binding site of the nicotinic acetylcholine receptor. J. Biol. Chem., 266:23354- 23364, 1991. [30] D.A. Dougherty and D.A. Stauffer. Acetylcholine binding by a synthetic receptor. Implications for biological recognition. Science, 250: 1558- 1560, 1990. [31] C. J. Noren, S. J. Anthony-Cahill, M.C. Griffith, and P.G. Schultz. A general method for site-specific incorporation of unnatural amino acids into proteins. Science, 244:182- 188, 1989. [32] J. Ellman, D. Mendel, S. Anthony-Cahill, C. J. Noren, and P. G. Schultz. Biosynthetic method for introducing unnatural amino-acids site-specifically into proteins. Meth. Enzym. , 202:301- 336, 1991. [33] J. D. Bain, C. G. Glabe, T. A. Dix, and A. R. Chamberlin. Biosynthetic sitespecific incorporation of a non-natural amino acid into a polypeptide. J. Am. Chem. Soc., 111:8013- 8014, 1989. [34] J. D. Bain, D. A. Wacker, E. E. Kuo , M. H. Lyttle, and A. R. Chamberlain. Preparation of chemically misacylated semisynthetic nonsense suppressor trnas employed in biosynthetic incorporation of non-natural residues into proteins. J. Org. Chem. , 56:4615- 4625, 1991. 42 [35] J. D. Bain, C. Switzer, A. R. Chamberlin, and S. A. Benner. Ribosome-mediated incorporation of a non-standard amino acid into a peptide through expansion of the genetic code. Nature, 356:537- 539, 1992. [36] T. G. Heckler, L-H. Chang, Y. Zama, T. Naka, M. S. Chorghade, and S. M. Hecht. T4 RNA ligase mediated preparation of novel "chemically misacylate tRNA-Phes. Biochemistry, 23:1468- 1473, 1984. [37] M. W. Nowak, P. C. Kearney, J. R. Sampson, M. E. Saks, C. G. Labarca, S. K. Silverman, W. Zhong, J. Thorson, J. N. Abelson, N. Davidson, P. G. Schultz, D. A. Dougherty, and H. A. Lester. Nicotinic receptor binding site probed with unnatural amino-acid incorporation in intact cells. Science, 268:439- 442, 1995. [38] M. E. Saks, J. R. Sampson, M. W. Nowak, P. C. Kearney, F. Du, J. N. Abelson, H. A. Lester, and D. A. Dougherty. An engineered tetrahymena tRNA Gln for in vivo incorporation of unnatural amino acids into proteins by nonsense suppression. J. Biol. Chem., 271:23169- 23175, 1996. [39] M. W. Nowak, J. P. Gallivan, S. K. Silverman, C. G. Labarca, D. A. Dougherty, and H. A. Lester. In vivo incorporation of unnatural amino acids into ion channels in a Xenopus oocyte expression system. Meth. Enzymol., 293:504-529, 1998. [40] M. D. Martin and A. Karlin. Functional effects on the acetylcholine receptor of mutiple mutations of 1Asp174 and 5Asp180. Biochemistry, 36:10742- 10750, 1997. [41] S. Mecozzi, A. P. West Jr., and D. A. Dougherty. Cation-n interactions in simple aromatics. Electrostatics provide a predictive tool. J. Am. Chem. Soc., 118:2307- 2308, 1996. [42] S. Mecozzi, A. P. West Jr., and D. A. Dougherty. Cation-n interactions in aromatics of biological and medicinal interest: Electrostatic potential surfaces as a useful qualitative guide. Proc. Natl. Acad. Sci. USA, 93:10566- 10571, 1996. 43 [43] M.E. O'Leary and M.M. White. Mutational analysis of ligand-induced activation of the Torpedo acetylcholine receptor. J. Biol. Chem., 267:8360-8365, 1992. [44] P.-J. Corringer, J.-L. Galzi, J.-L. Eisele, S. Bertrand, J.-P. Changeux, and D. Bertrand. Identification of a new component of the agonist binding site of the nicotinic a7 homooligomeric receptor. J. Biol. Chem., 270:11749- 11752, 1995. [45] Y. Fraenkel, J. M. Gershoni, and G. Navon. Acetylcholine interactions with tryptophan-184 of the a-subunit of the nicotinic acetylcholine receptor revealed by transferred Nuclear Overhauser Effect. FEES Lett., 291:225- 228, 1991. [46] J. C. Ma and D. A. Dougherty. The cation-1r interaction. Chem. Rev., 97:13031324, 1997. [47] S. F. Boys and F. Bernardi. Mol. Phys., 19:533- 566, 1970. [48] W. Zhong. Physical Organic Chemistry on the Nicotinic Acetylcholine Receptor. Ph.D. Thesis, California Institute of Technology, 1998. [49] M. J. Frisch, G. W. Trucks, H. B. Schlegel, P. M. W. Gill, B. G. Johnson, M.A. Robb, J. R. Cheeseman, T. A. Keith, G. A. Petersson, J. A. Montgomery, K. Raghavachari, M. A. Al-Laham, V. G. Zakrzewski, J. V. Ortiz, J. B. Foresman, C. Y. Peng, P. A. Ayala, M. W. Wong, J. L. Andres, E. S. Replogle, R. Gomperts, R. L. Martin, D. J. Fox, J. S. Binkley, D. J. Defrees, J. Baker, J. P. Stewart, M. Head-Gordon, C. Gonzalez, and J. A. Pople. Gaussian 94 (Revision D.3). Gaussian, Inc., 1995. [50] Spartan, version 5.0., 1995. 44 Chapter 3 Cation-1r Interactions in Structural Biology 45 3.1 Introduction The three-dimensional structure of a protein is uniquely determined by a delicate balance of weak interactions. Hydrogen bonds, salt bridges, and the hydrophobic effect all play important roles in folding a protein and establishing its final structure. In addition to these extensively studied interactions, the cation-7!" interaction [1- 3] is increasingly recognized as an important noncovalent binding interaction of relevance to structural biology. Theoretical and experimental studies have shown that cation-7!" interactions can be quite strong, both in the gas phase and in aqueous media. A wide array of biological structures has been shown to employ cation-7!" interactions, with perhaps the most compelling cases involving the binding of the quaternary ammonium group of acetylcholine by both acetylcholinesterase [4] and the nicotinic acetylcholine receptor [5] .1 These studies and others have emphasized the role of cation-n interactions in biological recognition , such as receptor-ligand and enzyme-substrate interactions. In this chapter, we attempt to answer a broader question - How does Nature use cation-n interactions within protein structures? In doing so, we present a detailed analysis of the extent and nature of cation-7!" interactions that are intrinsic to a protein's structure and are likely to contribute a protein's stability. We find that energetically significant cation-n interactions are common in proteins - a "typical" protein will contain several. In addition, we document some significant preferences for certain amino acid pairs as partners in a cation-7!" interaction. Important early work indicated a role for cation-7!" interactions in protein structures. Following work by Levitt and Perutz [6- 8] which suggested that aromatic rings could be involved in a hydrogen bonding-like interaction with amino groups, a pioneering study by Burley and Fetsko identified the "amino-aromatic" interaction [9], showing that amino groups tend to be positioned near aromatic rings within proteins. No distinction was made between neutral and charged NH groups in that study. It is now appreciated that the interaction of a cationic group with an aromatic - a cation-n interaction - is more favorable energetically by a full order of magnitude 1 This work is discussed in Chapter 2. 46 than an analogous interaction involving a neutral amine. For example, ammonium ion binds to benzene with a strength of 19 kcal/mol [10], whereas the binding interaction between ammonia and benzene is less than 2 kcal/mol [11]. As such, it is best to separate interactions involving the cation Lys and Arg from those involving the neutral, amide side chains of Asn and Gln. Subsequent studies by Thornton and co-workers [12- 17] modified the Burley and Petsko analysis, especially with regard to the amino-aromatic hydrogen bond. Although Thornton and co-workers observed the same preference for NH-containing groups to be located near aromatic amino acids , they suggest that it is not an amino-aromatic hydrogen bond that drives the formation of such structures. In particular, the side chains of Asn and Gln tend to stack parallel to aromatic side chains. This geometry precludes hydrogen bonding to the aromatic ring, but allows the amino group to satisfy its hydrogen bonding potential by donating a hydrogen bond to either solvent or other parts of the protein. Concerning cationic residues, explicit studies of Arg interacting with aromatic residues have been reported by Flocco and Mowbray [18], and by Thornton [14], and other efforts to search the PDB for cation-1r interactions between ligands and proteins have been reported [19, 20]. A recent study by Schultz has shown that an engineered cation-1r interaction can contribute to protein stability with a magnitude comparable to that of a hydrogen bond [21]. As is typical of such work, previous searches of the protein database relied primarily on geometric definitions of side chain interactions, focusing on when a cationic side chain displayed a certain distance/angle relationship to an aromatic side chain. Because of the substantially different geometries of Lys vs. Arg and Trp vs. Phe/Tyr (Figure 3.1), such comparisons can be problematic. In addition, not all cation- aromatic contacts represent energetically favorable, cation-1r interactions. Unlike ion pairs (e.g., Lys/ Asp), for which any close contact will be energetically favorable, the interaction between a cation and an aromatic can be attractive (over the face) or repulsive (on the edge). The electrostatic potential surfaces of the aromatics, which control such distinctions , can be complex,2 and so it is difficult to clearly distinguish 2 See: Chapter 2, Figure 2.4. 47 Parallel (Stacking) Perpendicular (T-Shaped) Arginine (Arg) sp 2 nitrogen Tryptophan (Trp) Lysine (Lys) w. ,,,, •'' H sp 3 nitrogen H H HO ~ Ca Tyrosine (Tyr) Figure 3.1: Limiting geometries for cation-1r interactions. Top: Stacked and T-shaped alignments for arginine. Bottom: Lysine interacting with Phe, with the c-carbon (left) or the NHt (right) making closest contact with the ring. These structures are not meant to indicate a preference for any particular pair or geometry. attractive from repulsive cation-aromatic contacts using geometrical criteria (r, 0, ¢) alone. For example, while the interaction of Lys with Phe can perhaps be adequately modeled considering only geometry - good interactions have the RNHt group "above" the phenyl ring - such arguments become much more challenging when the cation is Arg (T-shaped vs. stacked vs. intermediate) and the aromatic is Trp (6-ring vs. 5-ring). To circumvent this problem, and to put the diverse array of potential cation-1r interactions on a more nearly equal footing, we have chosen to use energybased, rather than geometry-based criteria in this study. Our goals in this study are two-fold. First, we wish to develop meaningful statistics for cation-1r interactions for structures within the Protein Data Bank (PDB) [22]. Second, we wish to develop a simple, unambiguous protocol for identifying cation-1r interactions which can be easily applied by other workers. Within a protein chain, cation-1r interactions can occur between the cationic side chains of either lysine (Lys, K) or arginine (Arg, R) and the aromatic side chains of phenylalanine (Phe, F), tyrosine (Tyr, Y), or tryptophan (Trp, W) . Because histidine 48 can participate in cation-1r interactions as either a cation (when the side chain is protonated), or as a 1r-system, and there can be some ambiguity on this issue, we do not consider histidine in this study. We assume Lys (pKa = 10.5) is always protonated in this study. There are several possible geometries for interactions involving arginine and lysine. With lysine, the obvious geometry places the NHt group over the ring. However the E-carbon of Lys ( CE) also carries a substantial positive charge, and putting it near an aromatic also constitutes a cation-1r interaction. Arginine has two limiting geometries, a perpendicular, or T-shaped geometry that places the NH groups of the side chain in the face of the aromatic, and a parallel geometry in which the guanidinium group is stacked upon the face of the aromatic (Figure 3.1). 3.2 Methods In this section we describe in some detail our strategy for identifying and ranking cation-1r interactions in proteins. To provide an energetic evaluation of all potential cation-1r interactions in a protein, we use a variant of the OPLS force field [23 , 24]. Only cation-1r interactions with binding interactions that rise above a certain energetic threshold are retained. The details of and justification for the protocol constitute the remainder of this section. Further discussion of this method appears in an appendix to this chapter. Our development of a simple, general protocol to identify cation-1r interactions within the PDB proceeded as follows: 1. Potential cation-1r interactions from a test dataset of 68 proteins were identified using only geometric criteria. 2. Ab initio quantum mechanical calculations were performed on the candidates from the test dataset to evaluate their binding energies. 3. A force field- based method was developed to reproduce t he trends in the ab init io data. 49 4. The force field- based method was then used to select energetically significant cation-1r interactions from a larger dataset of 593 proteins. To search the PDB, a computer program (CaPTURE - Cation-1r Trends Using Realistic Electrostatics) was developed to calculate the distance between the cationic group (the ammonium nitrogen (NZ) in Lys or the guanidinium carbon (CZ) in Arg) and the centers of all aromatic rings (the centroids of the 6-membered rings in Phe, Tyr and Trp, as well as the 5-membered ring of Trp). We chose 68 proteins from the PDB whose resolution is better than 2.0 A and which had been previously shown by Thornton's group to be non-homologous [16]. Using a 6.0 A distance cutoff and no other geometrical constraints, 359 potential cation-1r pairs were selected. Each pair was then reduced to a system that could be studied computationally. The cations Lys and Arg were represented as ammonium and guanidinium ions, and the aromatics, Phe, Tyr and Trp, were represented as benzene, phenol and indole, respectively. Using HF /6-31G** minimum energy structures for these fragments, we established the relative orientations of the partners using geometrical parameters from the PDB and converted each potential interaction to a Z-matrix format suitable for input to GAUSSIAN 94 [25]. The binding energy of each was determined using HF /6-31G** calculations with counterpoise corrections for basis set superposition error (BSSE). We appreciate that a higher level of theory (such as MP2) would provide a better estimate of the gas phase interaction energy for any pair [26, 27]. However, that is not our goal. We seek a simple criterion that will put all cation-aromatic contacts on a consistent scale, and it is clear that the trends from the level of theory applied here reproduce the trends of the higher levels. In addition, it must be remembered that even high resolution macromolecular structures have some structural ambiguity, and we feel that attempting very high level calculations on relatively "low level" geometries is not sensible. Our ultimate goal is to develop a simple, rapid procedure to identify cation-1r interactions, and a force field- based method is best suited to this. While standard forcefield-methods are challenged to quantitatively model the cation-1r interaction [28-30], 50 it is clear that force field- based methods can correctly reproduce trends in the binding energies. We therefore implemented within CaPTURE a subset of the OPLS force field [23] in which only electrostatic and van der Waals interactions are considered. We next calculated the OPLS binding energies for each of the 359 interactions described above, and compared them to the HF calculated binding energies. At first, the results were not encouraging, producing a poor correlation between the HF energies and the OPLS total energies (Etot, which equals the sum of an electrostatic term, Ees, and a van der Waals term, EvdW). Previous studies have shown that a large component of the cation-1r interaction is electrostatic, and that trends in the electrostatic component determine trends in cation-1r binding ability [31]. Thus, we attempted to correlate the electrostatic component of the OPLS binding energy (Ees) with the total ab initio binding energy. We find that these measurements correlate well for the different cation-1r pairs (Figure 3.2), and represent an enormous computational savings. The correlations are better when Lys is the cation than when Arg is the cation. This is possibly because the larger, more complex Arg side chain has several different binding modes available, and these produce variations in van der Waals interactions, which are now excluded in our force field calculation, but are included in the ab initio calculations. In addition, the HF structures were built under the assumption that the fJ carbon (CD) of Arg is in the same plane as the guanidinium unit. However, in some protein structures this is not so, and the OPLS calculations use the structure directly from the PDB file. This introduces some further scatter in the data. Our major concern, however, is with the trends in the data - we want to know when one cation-1r interaction is stronger than another. We feel this approach is more than adequate for such purposes. Because force field calculations are computationally inexpensive, we included the "methyl-groups" for the cationic side chains (Arg, CD; Lys, CE) on moving to the much larger dataset of proteins discussed below. Thus, Lys was now represented as methylammonium and Arg as methylguanidinium. This change was motivated by the fact that a significant fraction of the positive charge in Lys and Arg is associated with these carbons. Methyl groups were represented as united atoms because of 51 Lys-Phe 'a .. ... ,· ~ 0 I uP •2 ...... Lys-Trp .. ~ .4 Jj ..... bO .s ".s ·6 "' -8 .. : ~~~~~~~~~~~~~~ -12 -10 -8 -6 -4 -10 -2 ~~~~~~~~~~~~~~ -15 -10 Binding Energy HF/6-3JG•• (kcal/mo!) Arg-Phe -5 ... . .. .,.. , Binding Energy HF / 6-3JG•• (kcal/ mo!) 10 15 Arg-Trp ...... . . . .. -6 ~~~~~~~~~~~~~~ -8 4 4 ~ 0 0 Binding Energy HF/6-31G"' (kcal/mo!) . -10 ~~~~~~~~~~~~~~ -8 4 -4 -2 2 Binding Energy HF / 6-3JG•• (kcal / mol) Figure 3.2: Plots of E es vs. counterpoise-corrected, HF 6-31G** binding energies for selected cation-1r interactions. The data fit to the following lines: Lys-Phe y = 0.98 + 0.68x, R = 0.95, Lys-Trp y = 1.72 + 0.74x, R = 0.96, Arg-Phe y = 0.42 + 0.88x, R = 0.81, Arg-Trp y = 1.33 + 1.23x, R = 0.92. Three outliers are circled in the Arg-Phe plot. Inspection of these pairs reveals that they contain spurious close contacts (see text), which lead to erroneously high energies in the HF calculations, but not in the OPLS calculations, because no van der Waals term is included in the latter. If they are removed from the plot, the correlation coefficient improves to 0.85. Note that due to complications concerning the representation of Tyr (see text), such plots are not meaningful when Tyr is the aromatic. uncertainty in the position of the hydrogens. The inclusion of methyl groups on the aromatics does not significantly alter their cation-1r binding ability [32], and they were not added. The hydrogens of the aromatics were placed in idealized geometries and treated explicitly (not as united atoms) using the standard parameters in the OPLS force field. Tyrosine presents a special problem for this analysis. The position of the proton of the phenolic OH is typically not well-determined in macromolecular crystal structures, and its location can significantly affect cation binding ability through an interaction of the ion with the strong OH bond dipole. In principle, we could treat the OH of 52 tyrosine as a united atom, as we have treated other groups with uncertainty in the positions of the hydrogens. Unfortunately, this presents a new problem, as the partial negative charge on this atom strongly attracts cations, but not through a cation-n interaction. Since phenylalanine and tyrosine are nearly equivalent in their idealized cation-n binding ability [31, 32], 3 we have chosen to treat them with identical charge parameters. The phenolic oxygen of tyrosine retains its steric parameters, but has zero charge. To make the ring electrostatically equivalent to Phe, a dummy charge is placed in an equivalent location to that of the para-hydrogen of phenylalanine, and all other atomic charges are set equal to those of phenylalanine. Thus, in interactions of a cation with Tyr, we are considering only the cation-n interaction. In the actual protein, the cation-tyrosine interaction may be larger due to interactions with the oxygen (see below). Another advantage of excluding EvdW from our calculation is that many of the most favorable cation-n interactions we find in proteins produce spurious, repulsive EvdW terms. This results from the inherently low resolution of macromolecular crystal structures, which occasionally have unrealistic atom-atom contacts. Such pairs would be rejected from a search for cation-n interactions if Etot were the evaluation criterion. Since electrostatics are much less sensitive to close contacts ( 1/ r) than van der Waals repulsions (1/r 12 ), the effects of such small geometry errors are minimized when only Ees is considered. With this computationally efficient, energy-based criterion established, we could now examine a larger number of proteins (see below), and we could expand the interaction distance cutoff from 6 A to 10 A. Within this expanded search, we found some cation-n pairs involving tryptophan that gave a favorable value of Ees but had the cation quite far from the aromatic rings. Although these structures are certainly attractive in the gas phase, they are likely attenuated in the interior of a protein. Because of this, a simple, geometry-based criterion was added that would eliminate these long range structures. To do this, we asked whether a water-sized probe (a sphere 3 The calculated cation-7r binding energies and electrostatic potential surfaces of these residues can be found in Chapter 2. 53 2.8 A in diameter) could fit between the van der Waals surfaces of the cation and the aromatic at their point of closest approach. If the water molecule fits, the structure is rej ected, regardless of its electrostatic energy. In practice, this criterion eliminates a small number of distant interactions. Indeed, in some structures, inspection of the full protein rather than just the cation-1r pair we had selected for evaluation, revealed intervening atoms from the protein separating the "interacting" partners. We felt it was sensible to rej ect such structures. It remains to choose a threshold value for Ees , below which a pair is considered to experience a cation-1r interaction. The simplest model would assume that if Ees is less than zero, i.e., if the electrostatic interaction is favorable, then the pair should be counted as a cation-1r interaction. However, inspection of structures with Ees only slightly below zero shows that these interactions are much too weak to consider. They may be significant in the gas phase, but we consider them unlikely to contribute to protein stability. On the other hand, it is clear that if Ees ~ -2.0 kcal/ mol, then the pair is experiencing a significant cation-Jr interaction. Also, if Ees > -1.0 kcal/mol, no cation-1r interaction should be considered . For -2.0 < Ees ~ -1.0 kcal/ mol, the choice is not so clear. Some structures are clearly desirable, while in other cases the interacting partners are too far apart, even though there is no 2.8 A gap. In order to distinguish these , we consider the van der Waals term of the OPLS interaction energy as an indicator of whether the pair is interacting significantly. It is safe to include EvdW for these relatively weaker interactions , because they are generally not the closest contacts, and therefore not susceptible to spurious van der Waals interactions. We conclude that if EvdW is ~ -1.0 kcal/mol, the interaction is significant. We emphasize that these final refinements to the protocol are meant to produce the best possible selection criterion, but they do not have a major impact on the final list. None of the global conclusions presented below would be substantially altered if they were not included. To summarize, our protocol for selecting cation-1r interactions is as follows: All cation-1r pairs (K or R with F , Y, or W) within 10 A of each other are considered. If there is a gap large enough to insert a water molecule (2.8 A separation between 54 van der Waals surfaces at closest contact), the structure is rejected, and the residues are considered non-interacting. For the remaining interacting pairs, the OPLS electrostatic energy, Ees , is evaluated. If Ees ~ -2.0 kcal/mol, the pair is counted as a cation-1r interaction. If E es > -1.0 kcal/mol, the structure is rej ected. If -2.0 < Ees ~ -1.0 kcal/mol, the structure is retained only if EvdW is ~ -1.0 kcal/mol. It is worth remembering that the interaction energies we will discuss below are only the OPLS electrostatic energies. The actual interaction energy - the true magnitude of the cation-1r interaction - is larger by an amount equal to the van der Waals interaction energy. For most pairs considered, EvdW is comparable to Ees, and so the true cation-1r interaction energy is roughly twice as large as Ees. 3.3 Results and Discussion Using the above criteria, we scanned a larger dataset of representative protein crystal structures, taken from the "PDB Select" list of Hobohm and Sander [33, 34]. We initially divided the proteins into two sets. The first consisted of 323 proteins that exist as a single subunit (Table 3.1); the second consisted of 270 proteins that contained multiple subunits (Table 3.2) . In most analyses, however, no significant differences between these two sets were found, and thus the combined set of 593 proteins is considered unless otherwise noted. All proteins considered had resolutions better than 2.5 A. For evaluating possible cation-1r interactions, residues with fractional occupancies < 0.95 were rejected, and structures that contained residue insertions were not considered . For the combined sets, 230,504 residues were considered, producing 14,030 interacting pairs (structures within 10 A of each other and containing no "gap", Table 3.3) , and 2,994 significant cation-1r interactions (Table 3.4). Figure 3.3 gives a visual representation of our selection procedure. All Lys/Phe pairs from single subunit proteins within the 10 A cutoff are shown. They are divided into three categories: those that are rejected based on the gap criterion (open circles); those that are interacting (no gap) but have E es > -1.0 kcal/mol (blue circles); and 55 1191 1531 lA0P 1A17 lAlX 1A26 1A62 1A68 1A6Q 1A7J 1A8E 1A8I 1A9S lAA0 lAAC lABA lACC 1AD2 lADS 1AF7 1AH7 lAHO lAIE 1AI1 1AJ2 1AJ6 lAJJ lAK0 lAKl 1AK5 lAKO lAKZ 1A13 1A1O 1A1U 1A1Y 1AM3 lAMM lAMP lAMX 1AN8 lANF 1AO1 lAOP lAQB lARB lARV lASH lAT0 lAUA lAUK lAWD lAXN 1AY1 lAZO lBAl lBDO lBEO lBFD lBFG lBGC lBGP lBKF lBPl lBRT lBTN lBVl lBYB 1C52 lCEM lCEO lCEX lCFB lCFR lCHD 1C1C lCNV lCPO lCSH lCSN lCTJ lCYX lDAD lDAR lDHR lDHS lDHY lDUN lDXY lECA 1EC1 lEDE lEDG lEDT lERV lESC lEUU lEZM lFDR lFDS lFIT lFKX lFNA lFTS lFUA lFUS 1G3P lGAI lGCB lGKY lGND lGOH lGPC lGSA lGVP lHAl lHFC lHJP 1H1B lHMY lHOE lHTP lHXN lIDK lIDO lIGD lIHP lINP lIRK lISO lITG lIXH lJDW lJER UPC lKID lKIT lKNB lKPF lKTE lKUH lKVU 11AM 11BA 11BU 11C1 11IS lLIT 11KI 11M1 lMAI lMAZ lMBD 1MM1 lMRJ lMRP lMSC lMSI lMSK lMUP lMZM lNAR lNEU lNFN lNIF 1N1S lNOX lNPK lNXB lOIS lONC lOPD lOPR lOPY lOYC lPDA lPEA lPGS lPHE lPHP lPII 1P1C lPMI lPNE lPOA lPOC lPOT lPPN lPPT lPTA lPTQ lPTY lPUD lQNF lQUF 1R69 1RA9 lRCF lREC lRHS lRIE 1R1W lRMD lRMG 1RN1 lRPO lRRO lRSS lRYT lSBP lSFE lSFP lSKZ lSMD lSRA lSVB lTCA lTDF lTDX lTFE lTFR lTHV lTHX lTIB lTIF 1TM1 lTSP 1TU1 lTYS lUAE lUBI lUBY lUCH lUTG lUXY 1V39 lVCC lVHH lVID lVIF lVIN lVJS 1V1S lVNC lVSD lWAB lWBA lWER lWHI lWHO lXJO lXNB lYER lYGS lYTW lZIN lZXQ 2A0B 2ABK 2ACY 2AYH 2BAA 2BGU 2CAE 2CBA 2CTC 2CYP 2DKB 2DPG 2DRI 2END 2ENG 2ER1 2FDN 2FHA 2GDM 2HBG 2HFT 2HTS 2I1B 2I1K 21BD 21IV 2MCM 2OMF 2PGD 2PHY 2PIA Table 3.1: PDB ID Codes for single-subunit proteins. 2PII 2P1C 2POR 2PTH 2RN2 2SAK 2S11 2SN3 2SNS 2STV 2TCT 2TGI 2WEA 3B5C 3CHY 3C1A 3COX 3CYR 3GRS 31ZT 3N11 3NU1 3PTE 3SEB 3TSS 3VUB 4MT2 4XIS 5EAU 5ICB 5P21 5PTI 6CE1 7RSA 8ABP 56 lAlI 1A28 1A2P 1A2Y 1A34 1A6J 1A74 1A7T 1AA7 lABR lADO lAER lAFR lAFW lAGJ lAGQ lAIJ lAIK lAJS 1AK4 lALV 1AM7 lAMU 1AN9 1AO6 lAOC lAOH lAOJ lAOQ lAOR lAOZ lAPY lAQ0 1AQ6 lAQZ lASK lATL lATZ lAUl lAUR lAUV lAVM 1AW8 lAWC lAWS 1AX4 lAXI lAYM lAYY lAZS lBBP lBDM lBEB lBEN lBFT lBHM lBKR lBNC lBND lBOV lBTK lCEW lCFY lCHK lCHM lCKA lCKN lCNT lCPC lCSB lCSE lCYD lDEA lDEL lDFJ lDJX lDKZ lDOK lDOR lDOS lDUB lDUP 1E2A lECE lECP lEDM lEFV lETP lEXN lFIE lFJM lFMT lFTl lFTP lFUR lFVK lFWC lGAR lGCl lGDO lGGG lGIF lGNW lGOT lGPM lGTM lGTQ lGUQ lHAV lHCR lHGX lHJR lHSB lHSS lHTR lHUL lIDA lIGN lIHF lIIB lIPS lIPW lISU 11TB lJAC lJET lJFR lJHG lJLY lJMC usu lJXP lKIM lKNY lKPT lKVE lKWA lKZU lLAT lLGH lLKK lLMB 1LT5 lLTS lLUC lMKA lMLD lMNM lMOL lMPG lMSP lMUC lMUG lNBA lNBC lNCI lNOZ lNPl lNSG lNUL lNWP lONR lOSP lOTG lPBO lPCF lPDN lPHN lPNK lPOI lPRC lPRX lPYA lQTQ lREG lREQ lRGE lRHO lRLA lRVA lSFT lSLT lSMN lSMP lSMT lSPU lSTM lSVP lTAD lTAF 1TC3 lTDT lTHT lTII lTRK lTUP lTVX lUMU lUNK lURN lVCA lVDE lVDF lVHR lVMO lVPF lVPS lVWL lWDC lWPO lWSY 2PHL lXBR 2POL lXGS 2PSP lXIK 2RSL lXVA 2RSP lXYZ 2SCP lYAI 2SIC lYAS 2SPC lYCQ 2TMK lYCS 2TRC lYPD 2TYS lYRN 2VHB lYTB 2VIU lYTF 3DAA lYVE 3DAP 2ARC 3MDD 2BBK 3MIN 2BOP 3PCG 2CCY 3SDH 2CHS 3ULL 2DYN 4AAH 2FIV 4CPA 2HHM 4PGA 2HPD 4PGM 2KIN 5CSM 2MAS 5HPG 2MPR 6GSV 2MSB 7AHL 2MTA SATC 2NAC SRXN Table 3.2: PDB ID Codes for multi-subunit proteins. 57 Figure 3.3: Scatter plot for all Lys/Phe pairs from all 323 single subunit proteins. The circle denotes the location of the side chain N. Each side of the cube is 10 A. The phenyl ring plus the fJ carbon are denoted by black lines. The color code is as follows : red - a cation-7!' interaction; blue - an interacting pair; white - structures rejected by the gap criterion. In this projection view, a few open/blue circles are seen to lie over the ring, but they are too far away to have a favorable cation-7!' interaction. cation-7!' interactions (red circles) . Clearly our selection is "sensible" - the three classes form concentric rings about the centroid of the aromatic ring. It is worth noting here a significant advantage of energy-based selection criteria over geometry-based methods. Like other workers, our initial geometric screen involved a distance (r) from the center of positive charge (NZ of Lys or CZ of Arg) to the aromatic, but there is a danger in rejecting structures solely on the basis of r. For example, there are many structures in which the NZ of Lys will be more than 5 A from the aromatic, but the CE will be well within 5 A. Since the CE of Lys has a substantial positive charge, such a structure should be considered as a possible cation-7!' interaction, and our calculations reflect this. An example of such a structure can been found in Section 3.4. While we considered all pairs with r :S 10 A, in 99% of our accepted cation-7!' structures, r is :S 6 A when both NZ and CE of Lys or CZ 58 and CD of Arg are considered, while 88% have r :S 5 A. With 2,994 cation-n interactions in 230,504 residues, there is an average of 1 energetically significant cation-Jr interaction for every 77 residues in a protein. This number does not vary substantially with the length of the protein, although there is certainly some scatter. For example, the 126 amino acid mutant form of the protein human fibroblast growth factor (PDB: lBFG) contains 6 significant cation-Jr interactions, while the 323 amino acid protein penicillopepsin (PDB: 2WEA) is the largest single chain protein studied that contains no energetically significant cationn interactions, although this protein contains only 5 lysines and no arginines. The number of cation-n interactions per residue is the same whether single chain or multisubunit proteins are considered. For comparison, one might ask how common salt bridges are in proteins. The energetic significance of salt bridges continues to be an issue of debate - a debate that we will consider further in Chapter 4. Nevertheless, to get a sense of the relative frequency of the two types of interactions, we considered all ion pairs (Lys/ Asp, Lys/Glu, Arg/ Asp , Arg/Glu) that meet the "interacting pair" criterion, i.e., they have no gap. We find that salt bridges are quite common, with almost 27,000 occurring in our collection of proteins (compared to 14,030 cation-aromatic interacting pairs). There is no significant preference for one pair over another. The most common cation-n interaction is between neighboring residues in the sequence, with 7.3% of the interactions occurring between adjacent residues (Figure 3.4). Interactions between residues at the i and i+4 positions are the second most common, suggesting that cation-Jr interactions may commonly occur within a-helices, as in the structure of the vaccinia virus protein VP39 (PDB: 1V39) shown in Figure 3.5. Table 3.3 summarizes the likelihood that individual amino acids will be involved in an interacting pair, while Table 3.4 lists analogous data for cation-n interactions. Table 3.4 presents two sets of numbers: all cation-n interactions and those with Ees :S -4.0 kcal/mol. The latter set contains strong interactions that, by any criterion, would be considered to make a substantial contribution to protein stability. Several 59 200 200 Cl) 8 150 (i, i+4) :.0 u ro ~ 150 Ir~ 100 100 ] 0 ..... .....ro u 0 50 ~ (1) 's 50 ;:J z 0 0 10 20 30 ~~.iJ~jl\'1,.l~,11.r u~L 11-,t11,, ,L11 L.1,.,. 1,1L,. 1~ .. 1,,. 1 100 200 300 400 0 0 Sequence Distance Between Residues Figure 3.4: Plot of the number of cation-1r interactions vs. the sequence distance of the interacting pair. The inset highlights interactions between the i and i+4 residues that may be located on a-helices. Only single chain proteins were considered for this analysis. interesting trends concerning individual amino acids are evident. 3.3.1 Cationic Amino Acids With respect to the cationic amino acids, Table 3.3 indicates a striking preference for the side chain of arginine to be located near aromatic side chains in proteins. Over 70% of all Arg side chains are near an aromatic side chain, consistent with earlier studies of the arginine-aromatic interaction [12] . In Table 3.4, which considers only favorab le structures, we again note a strong preference for arginine over lysine that mimics the trends in Table 3.3. It is perhaps surprising that arginine is more likely to be found in a cation-1r 60 Total numbera 13446 10919 9162 8309 3412 Amino acid K R F y w Frequency (% )b 5.8 4.7 4.0 3.6 1.5 Interacting pairsc 5881 8149 4969 6615 2446 % interacting pairsd 43.7 74.6 54.2 79.6 71.7 Table 3.3: a) The total number of times a particular amino acid appears in the dataset of 593 proteins. b) Occurrence of a particular amino acid in the dataset expressed as a percentage. c) The number of times a particular amino acid occurs as a member of an interacting pair. d) The likelihood of a given amino acid to appear in an interacting pair expressed as a percentage of the number of times that it appears in the dataset. Amino acid K R F y w All cation-?T interactions Number Frequency (%)a 7.5 1006 18.2 1988 915 10.0 14.3 1187 892 26.1 Ees ::; -4.0 kcal/mol Number Frequency (% )a 2.1 281 4.0 432 2.2 199 2.4 197 317 9.3 Table 3.4: Frequency of cation-?T interactions for individual residue types within the 593 protein dataset. a) The likelihood that a particular amino acid type will be in a cation-?T interaction, expressed as a percentage of the total number of times that amino acid appears in the dataset. Preferences for arginine and tryptophan are clearly apparent . 61 Figure 3.5: An intrahelical cation-n interaction between Arg 79 and Tyr 83 of the vaccinia virus protein VP39 (PDB: 1V39). Plot generated with MOLSCRIPT [35]. interaction than lysine, since arginine is not expected to be innately better at binding to aromatics than lysine. Ab initio calculations by Jorgensen clearly indicate that in the gas phase, ammonium (model for Lys) interacts more strongly with aromatics than guanidinium (model for Arg) [36]. We find at the HF /6-31G** level, the binding energy for ammonium to benzene is -15.3 kcal/mol, substantially stronger than that of guanidinium binding to benzene in both the parallel (-4.1 kcal/mol) and T-shaped (-10.6 kcal/mol) geometries. We expect this pattern to hold up at higher levels of theory. To determine whether our statistics reflect an artifact of our electrostatic models for either Lys or Arg, we modeled each side chain as a point charge, placing a unit charge at either the location of the e-NHf group of lysine or the location of the guanidinium carbon of arginine. Since these positions essentially overlap in space and the charges are identical, we eliminate any potential bias in the charge model for the cation. We then recalculated Ees using this simplified model, and we still find that Arg is substantially more likely than Lys to be found in a significant cation-n interaction. This suggests that other, non-electrostatic effects may be contributing to the propensity of Arg to be involved in cation-n interactions. Since the side chain of arginine is larger and less well water-solvated than that of lysine, it is likely that Arg benefits from better van der Waals interactions with the aromatic ring. In addition, as suggested by Thornton et al. [14], the side chain of arginine may still donate several energetically significant hydrogen bonds while simultaneously binding to an aromatic ring (if it is stacked), whereas lysine would typically have to relinquish hydrogen 62 200.0 200.0 Arginine Lysine 150.0 150.0 ... § 100.0 ... <iJ <iJ .£l .£l ;::l z s 100.0 z 50.0 50.0 0.0 f-.,..-~,........,.-=f-LJ...l+'-+-'--l--'+'J..ld-1-'-+Lf.-Lµ+"+'-!,J..,......., -10.0 -8.0 -6.0 -4.0 -2.0 0.0 0.0 ____,..__p::1-j::bj:J.IJ...l-'.4-L~,4...t.µ.i.4±µ...i.-4-1-1---.----, -10.0 -8.0 -6.0 -4.0 -2.0 0.0 Ees (kcal/mol) E es (kcal/mol) Figure 3.6: Distribution of Ees for cation-n interactions involving arginine and lysine. bonds to water (or other parts of the protein) in order to bind to an aromatic. Thus, although arginine is more prevalent than lysine in cation-n interactions, we suggest that this does not reflect the intrinsic cation-n binding ability of arginine, but rather other factors as discussed above. Consistent with this view, the average strengths of a cation-n interaction involving either Lys or Arg are similar (Lys: -3.32 ± 1.50 kcal/ mol; Arg: -2.87 ± 1.42 kcal/mol) , with perhaps a slight preference for Lys. This is evident in Figure 3.6 which shows the distribution of Ees for cation-n interactions involving arginine and lysine. It is also interesting to note that the 12 most energetically significant interactions involve lysine as the cation, consistent with the ab initio calculations discussed above. The side chain of arginine may participate in cation-n interactions in two limiting geometries (Figure 3.1). The first is a stacked geometry that places the guanidinium group parallel to the plane of the aromatic ring. The second is a T-shaped geometry that places the guanidinium group perpendicular to the ring. Computational studies have suggested that in the gas phase, the perpendicular geometry is favored energetically, while in solution, the parallel geometry is preferred [36]. Previous studies of arginine-aromatic interactions within proteins by Thornton and co-workers [12, 14], as well as Flocco and Mowbray [18], have suggested that the parallel geometry is preferred in protein structures. Our findings largely agree with these studies, in that 63 -1.0 -2.0 -3.0 ,......._ 0 -4.0 <ii -5.0 Ei u ...:.:: '-' ,,, µ.i "' -6.0 -7.0 . :. . ( -8.0 _9 _0 -+-P_a~ra_l_le~l~~~~-~T~-_S~h_a......,_ed~~~~-~~P_a_r~a_ll~el 45 90 135 180 0 Interplane Angle Figure 3. 7: The interplane angle of cation-1r interactions involving argmme. Although the parallel geometry is preferred, T-shaped geometries represent some of the strongest interactions. there is a preference for parallel geometries, as shown in Figure 3.7. However, some of the strongest cation-1r interactions involve T-shaped geometries. Earlier work proposed that when Arg is in the T-shaped geometry, it always has one, not two, NH 2 groups in the face of the ring. While this is certainly a strong preference, we do find some structures with two NH 2 groups interacting with the aromatic (Figure 3.12). With respect to the orientation of the lysine side chain in cation-1r interactions, the c-carbon (CE) of lysine is 2.4 times more likely to be closer to the ring centroid than the nitrogen (NZ). This preference, which is contrary to expectations based only on electrostatics, disappears when only the strongest binding structures (Ees ~ -5.0 64 kcal/mol) are considered, although some very strong interactions have a close CEaromatic contact (Figure 3.11). These results suggest that while the .s-carbon carries substantial partial positive charge, binding to the nitrogen leads to the strongest cation-1r interactions. We note, however, that positioning the carbon closest to the ring may contribute favorable van der Waals binding, and exposing the ammonium group may lead to better interactions with solvent or hydrogen bonding groups in the protein. 3.3.2 Aromatic Amino Acids With respect to the aromatic amino acids, a marked preference for tyrosine and tryptophan to interact with the side chains oflysine or arginine is observed (Table 3.3). On average, 80% of all tyrosines and 72% of all tryptophans in the dataset interact with a cationic side chain, vs. 54 % for phenylalanine. To determine whether cation7r interactions explain these observations, the results in Table 3.3 can be compared to those in Table 3.4. Concerning the phenylalanine/tyrosine results, there is no strong bias when considering only the strongest cation-1r interactions (Ees :S -4.0 kcal/mol). When all cation-1r interactions are considered, however, a slight preference for Tyr is evident. Theory suggests that tyrosine and phenylalanine are equivalent in their cation-1r binding ability, and that the increased number of cation-1r interactions involving tyrosine must be due to other effects, such as the ability of the OH group of the tyrosine to act as a hydrogen bond donor. It has been shown that if the tyrosine OH is involved in a hydrogen bond, it substantially potentiates the cation-1r binding ability of the phenolic ring [32]. Also, the negative electrostatic potential on the oxygen could directly contribute to cation binding, as discussed above. In fact, the OH could accept a hydrogen bond from a cationic side chain, which might position these side chains more favorably so as to benefit from cation-1r binding. This is suggested in Figure 3.8, which shows the spatial distribution of cations engaged in cation-1r interactions. In the case of Tyr, there appears to be a bias for the cation to be closer to the OH, suggesting that in the proteins, this interaction is energetically 65 :~ / • .·~... :'i• . •, ••••• .... ~ ...a:>-aX) CX>-0:>0 :.... c:..z:..... ..... • / .. ~ Figure 3.8: Scatter plot for the energetically significant cation-7r interactions for Lys/Phe pairs (left) and Lys/Tyr pairs (right) from all 323 single subunit proteins. The small red circles denote the location of the side chain N of the lysines. Each side of the cube is 10 A. The aromatic rings are viewed edge on, wit h t he fJ carbons located on the right. The phenol oxygen is in red. Note the cluster of structures above and below the oxygen. significant. Perhaps the most surprising result is that 26% of all tryptophans in the dataset are involved in at least one energetically significant cation-7!" interaction. It has been postulated that tryptophan would be overrepresented at cation-7!" sites, because in the gas phase, indole binds cations more tightly than either benzene or phenol [1, 31 , 32]. This preference for indole has a strong electrostatic component, and since our energy calculation mimics this trend, it could be argued that it was inevitable that tryptophan would appear to be a better cation-7!" binder when using an energetic criterion. To address this concern, two limiting reasons why tryptophan might be more prevalent at cation-7r sites were considered. The first is that t he larger volume of tryptophan allows it to contact a greater number of cations relative to phenylalanine or tyrosine. To test this , we again consider the purely geometric criterion, counting all cation-aromatic interactions (Table 3.3). We find that cationic side chains show a marked preference to be near tyrosine relative to tryptophan and phenylalanine (Phe, 1.0 : Tyr, 1.47 : Trp, 1.32) . When compared to the ratios of Table 3.4, column 2 (Phe, 1.0 : Tyr, 1.43 : Trp, 2.61), it is clear t hat the Trp bias is not geometric in origin. 66 I~ .. ...• ••• .'....~.:,.. . . . ..... .. I/ . .. . .' .• • .. • -.•-:.-£••.'i,k'. . . ooe:a .. • • ••..J•4.., J..•• :. -•: .. .. ·- / ~ Figure 3.9: Views of all Lys-Trp cation-1r interactions from single subunit proteins. The indole nitrogen is shown in blue. Left: Top-down view, right: edge-on view. To determine whether the bias for Trp simply reflects a bias in our energy model, we reduced tryptophan's cation-1r binding ability 50% by halving the electrostatic energy of each interaction involving Trp. This reduction substantially penalizes tryptophan, making it a much less potent cation binder than phenylalanine. Nevertheless, using this different model for Trp, we still find that it is favored over phenylalanine by a ratio of 2.0: 1.0. We thus conclude that the substantial overrepresentation of Trp in our collection of cation-1r interactions does not result from a bias in our select ion criteria. Instead, the likelihood that an aromatic side chain will be involved in a cation-1r interaction directly reflects the innate cation-1r binding ability of the ring. Two views of the Lys-Trp interaction are shown in Figure 3.9. A clear bias for the six ring of Trp to be involved in cation-1r interactions is evident. This is in complete agreement with expectations based on the electrostatic potential surface of the Trp side chain shown in Chapter 2. Also contributing to this bias is a steric effect; the protein backbone is nearer the five-ring, and may interfere with cation binding. Such an effect is also evident in Figure 3.8, but it is much less pronounced than t he bias seen with Trp . 67 Amino acids KF KY KW RF RY RW All cation-1r interactions cation-1r (%) interacting pairsa 285 14.5 438 14.7 283 30.2 630 21.0 749 20.6 609 40.4 Ees ~ -4.0 kcal/mol cation-1r 75 84 122 124 113 195 (%) interacting pairsa 3.8 2.8 13.0 4.1 3.1 12.9 Table 3.5: Frequency of cation-1r interactions for residue pairs within the 593 protein dataset. a) The likelihood that a particular amino acid pair is in a cation-1r interaction expressed as a percentage of interacting pairs. Within the dataset, the total numbers of interacting pairs: KF 1962; KY 2981; KW 938; RF 3007; RY 3634; RW 1508. 3.3.3 Pairwise Interactions Table 3.5 considers the interactions between the six possible cation-1r pairs. When the side chain of lysine approaches that of either phenylalanine or tyrosine, there is an approximately 15% chance they will be involved in an energetically significant cation1r interaction. However, when lysine approaches tryptophan, the chances are nearly one in three that the residues will be involved in a cation-1r interaction. These trends also hold for interactions involving arginine. Again the preference for tryptophan stands out. An interesting question concerns the location of cation-1r interactions within protein structures. It is generally accepted that cationic residues prefer to be on the surface of a protein, whereas aromatic amino acids prefer to remain in the hydrophobic core. Since a cation-1r interaction within a protein contains both a cation and an aromatic, it is worth considering whether the interacting pairs prefer to be located on the surfaces of proteins, or in the cores. Traditional methods for determining residue surface accessibility rely on calculating the surface area of an amino acid that is in contact with water. Since the partners of a cation-1r interaction are necessarily in contact with one another, their water accessible surface is diminished, even though the interacting pair as a unit may be fully solvated. Thus, it is difficult to determine computationally whether a cation-1r interaction is on the surface of a protein using only surface accessibility. We have not visually inspected all 2,994 cation-1r interac- 68 tions, but examination of many structures suggests that cation-1r interactions tend to be on the surfaces of proteins, consistent with an earlier conclusion by Flocco and Mowbray [18]. We will revisit this subject in further detail in Chapter 4. Although the cationic side chains of arginine and lysine often experience favorable electrostatic interactions with the side chains of aromatic amino acids via the cation-1r interaction, the question remains whether Nature uses this advantage to orient these side chains in folded proteins. That is, when all interactions between a cationic and an aromatic side chain are considered, is there a tendency for geometries that are favorable in a cation-1r sense compared to just a random distribution? To answer this question, we consider the simplest system - lysine /phenylalanine - and ask whether the side chains of these residues are placed such that the lysine is in an electrostatically favorable location above the phenylalanine ring. To do this, we must define what a random distribution would look like. While others have based such a distinction on criteria such as interplane angle, this approach was somewhat ambiguous for our purposes. So, we sought a clear geometric approach. Our approach is best understood with reference to Figure 3.10. We first create a model for benzene that describes its size and shape, as well as the volume excluded by the benzene and atoms touching it. We also create a "shell" with the same shape of the benzene, but a uniformly larger size, and we consider only structures for which the Lys NHt lies within this shell. We then create a cylinder encompassing the excluded volume of benzene and the region above it. If a cation lies within this cylinder, it is considered to be in an electrostatically favorable region and so is expected to experience a cation-1r interaction. The cylindrical region above the ring occupies 32% of the total volume, suggesting that 32% of the Lys should be in this region. However, we find that 48% of the lysines lie in this cylinder. With a dataset of 1,716 structures, 4 this indicates a non-random distribution at a confidence level greater than 99.999%. Similar trends are observed for the other cation-1r pairs. Thus, proteins do position cations at non-random positions relative 4 This number differs from the value in Table 3.5 (1,962) because many legitimate cation-1r inter- actions have the N outside of the shell, but CE is inside the shell. 69 Shell Excluded Volume \ Cylinder Figure 3.10: Schematic view of model used to calculate the preferred location for lysine/phenylalanine pairs. The excluded volume region represents the volume of benzene plus the volume unavailable to atoms ofradius 1.7 A. The cylinder is tangent to the benzene (radius = 4.8 A obtained by adding the radius of benzene (1.4 A + 1.7 A) and the radius of a neighboring carbon atom (1.7 A)). The shell is obtained by adding a constant radius of 2.8 A - the diameter of a water molecule - to the excluded volume region. This view shows the only the top half of the excluded volume and has portions of the shell and cylinder removed for clarity. To determine whether there is a preference for cations to be located within the cylindrical region, we compare the number of lysine/phenylalanine structures located within the cylindrical region, to the total number of lysine/phenylalanine structures inside of the shell. to aromatics, so as to optimize cation-1r interactions, providing further support for the notion that cation-1r interactions contribute significantly to protein stability. Although the methodology differs, our results agree with the findings of Thornton et al. [17] who observe non-random angle preferences for cationic side chains interacting with aromatic side chains. 3.4 Gallery of Cation-1r Interactions Finally, we present a gallery of representative cation-1r interactions obtained in this study, along with their Ees values. 70 A. B. C. D. Figure 3.11: Representative cation-n interactions involving lysine. A. Interaction between Lys379A and Trp376A in a copper amine oxidase (PDB: lSPU). Ees for this interaction is -9.8 kcal/mol, making it the strongest in the dataset. B. Interaction between Lys408 and Trp356 in a sulfite reductase (PDB: lAOP). E es for this interaction is -9.4 kcal/mol, making it the third strongest in the dataset. C. Interaction between Lys133 and Trp140 in Staph. nuclease (PDB: 2SNS) . E es for this interaction is -7.3 kcal/mol. It is the strongest interaction in the dataset that places t he methylene of the lysine closer to the aromatic than the ammonium group. D. Interaction between Lys76B and Trp58B in transaldolase B. (PDB: lONR) . E es for this interaction is -4.4 kcal/mol. It 1,lso places the methylene of the lysine closer to the aromatic than the ammonium group. 71 A. B. C. Figure 3.12: Representative cation-1r interactions involving arginine. A. Interaction between Arg77 A and Trp211A in the oligopeptide binding protein OPPA (PDB: lJET). E es for this interaction is -8.4 kcal/mol, making it the strongest interaction involving arginine in the dataset. When E v dW is included, this is the strongest interaction overall. B. Interaction between Arg335 and Trp481 in a mutant of glucose6-phosphate dehydrogenase (PDB: 2DPG). E es for this interaction is -7.7 kcal/mol. It is an example of a strong stacked arginine interaction. C. Interaction between Arg1136 and Trpll 75 in the tyrosine kinase domain of the insulin receptor (PDB: lIRK). E es for this interaction is -6.8 kcal/mol. Two nitrogens face the aromatic ring in this T-shaped structure. Thornton has noted that this is an uncommon motif [14]. D. Interaction between Arg1253 and Trpl193 in the tyrosine kinase domain of the insulin receptor (PDB: lIRK) . E es for this interaction is -1.0 kcal/mol. It is the weakest interaction that we accept. E vdW for this interaction is -2.0 kcal/mol. 72 Trp52 Figure 3.13: A remarkable example of a lysine involved in 4 cation-n interactions taken from the protein glucoamylase (PDB: lGAI). The total Ees for this interaction is ~ -22 kcal/mol. 73 3.5 Summary By developing an energy-based criterion that puts all cation-1r interactions on an equal footing , we have been able to develop meaningful statistics for the frequency of occurrence of cation-1r interactions in proteins and to evaluate whether specific cation1r pairs are preferred. We find that cation-1r interactions are common - one favorable interaction can be expected for every 77 residues of protein length. While the weakest of the interactions considered here (Ees :::::: -1.0 kcal/mol) may make only a small contribution to the overall stability, it is clear that some of the more favorable pairs contribute at least as much to protein stability as the more conventional interactions such as hydrogen bonds and salt bridges. We find that Trp is the most likely of the aromatics to be involved in a cation-1r interaction, with a remarkable 26% of all Trps involved in energetically significant cation-1r interactions. This is consistent with theoretical arguments [1, 32] that predicted the Trp side chain would be especially well suited to cation-1r interactions. We also find that- Arg is more likely than Lys to be involved in an energetically significant cation-1r interaction. This is likely not due to an intrinsic aspect of the Arg cation-1r interaction, but more likely reflects the differing geometric features of the Arg and Lys side chains. These results make a compelling case that cation-1r interactions should be considered alongside the more conventional hydrogen bonds, salt bridges and hydrophobic effects in any analysis of protein structure. 3.6 Appendix 1 - CaPTURE Manual 3.6.1 Overview Cation-1r interactions are widely distributed throughout protein structures. CaPTURE (Cation-1r Trends Using Realistic Electrostatics) is a program that allows the user to search Protein Data Bank (PDB) files to find energetically significant cation1r interactions. Using user specified parameters, CaPTURE searches a PDB file for potential cation-1r interactions using geometric criteria, and evaluates them energet- 74 ically using a modified version of the OPLS force field. The user can specify both the geometric and energetic criteria that define a cation-1r interaction. CaPTURE supports both an interactive mode which leads the user through the process, and a batch mode that lends itself to quick, large-scale database searching. For a study presenting results obtained by CaPTURE, see: "Cation-Jr Interactions in Structural Biology," J.P. Gallivan and D.A. Dougherty, Proceedings of th e National Academy of Sciences, USA, 1999, 96, 9459. 3.6.2 Installation CaPTURE is distributed as C++ source code in the file capture. tar. CaPTURE has been compiled on SGI workstations running IRIX 6.2 and IRIX 6.5.5. In addition, CaPTURE has been run with minimal modifications under both Solaris and Linux. Although there is no reason to suspect that it won't compile under other syst ems, t he author does not the have the time or resources to explore this possibility. If you succeed in compiling CaPTURE on another system, congratulations!'- I would love to hear of your experiences, and with your permission, I can make your distribution available. To install on an SGI system, simply untar the distribution file and type make: tar -xvf capture.tar make There will likely be several warnings, but hopefully no fatal errors. The finished product is an executable named capture which is about 450K. CaPTURE also uses a configuration file t hat will be generated t he first time t he program is run, or anytime that t he program can't find it. 75 3.6.3 Getting Started CaPTURE supports two modes of operation, an interactive mode and a batch mode. The first time the program is run, it will enter the interactive mode and prompt the user to answer several questions. When the questions are answered, a configuration file (capture. cfg) is generated that can be used in the future. Once this file is present and is to the user's liking, the program can be run in batch mode, bypassing the Q&A. Running CaPTURE in Interactive Mode To run CaPTURE in the interactive mode, simply type 'capture'. The program will prompt you for a PDB filename. Once the filename is entered, CaPTURE will ask the user whether it should use default parameters or custom parameters to carry out its search. In most cases, the default parameters are acceptable. Descriptions of the parameters appear later in this document. The default parameters were used I in our published study: "Cation-1r Interactions in Structural Biology," J.P. Gallivan and D.A. Dougherty, Proceedings of the National Academy of Sciences, USA , 1999 , 96, 9459. Running CaPTURE in Batch Mode To run CaPTURE in batch mode, simply type ' capture pdbfilename' . Provided both the pdb file and the configuration file (capture. cf g) are present , the program will run without the need for further user intervention. 3.6.4 Interpreting the Screen Output from CaPTURE When the program starts, you should see the following: Reading the PDB file filename Reading parameters from the file capture . cfg. Initializing. At this point , there may be some warnings printed: 76 ******************* Amino Acid Warnings********************* WARNING! Atoms in the following amino acids have fractional occupancy < 0.95. This means that there is uncertainty in their positions. These amino acids will be excluded from the calculations. Atom 68 in residue THR 9 has an occupancy < 0.95 and is excluded. Atom 69 in residue THR 9 has an occupancy < 0.95 and is excluded. Atom 70 in residue THR 9 has an occupancy < 0.95 and is excluded. Atom 71 in residue THR 9 has an occupancy < 0.95 and is excluded. Atom 72 in residue THR 9 has an occupancy < 0.95 and is excluded. Atom 73 in residue THR 9 has an occupancy < 0.95 and is excluded. Atom 304 in residue LYS 38 has an occupancy < 0.95 and is excluded. Atom 305 in residue LYS 38 has an occupancy < 0.95 and is excluded. Atom 306 in residue LYS 38 has an occupancy < 0.95 and is excluded. Atom 307 in residue LYS 38 has an occupancy < 0.95 and is excluded. Atom 308 in residue LYS 38 has an occupancy < 0.95 and is excluded. ************************* End of Amino Acid Warnings******************* In particular, if atoms in the PDB file have fractional occupancies less than 0.95, they will be excluded from the calculations. This is because there is some uncertainty in their positions. This is usually not a serious problem, unless there are a very large number of warnings. In addition, many of these atoms may belong to residues that cannot be involved in a cation-1r interaction. You will then see something like the following: 77 ********** Report for the protein filename************** This protein has 108 amino acids of which: 3 (2.8%) are Lysine 3 (2.8%) are Arginine 4 (3.7%) are Phenylalanine 6 (5.6%) are Tyrosine 3 (2.8%) are Tryptophan This protein exists as a single chain The protein has an apparent molecular weight of 12151 D These are largely self-explanatory. You may also see something like: WARNING! These amino acids are missing atoms needed for the calculations: LYS 38 The warning in this case means that the program knows that the amino acid is there , but some of the atoms may have a fractional occupancy of < 0.95, or may be missing from the PDB file altogether. Amino acids listed here are only those potentially involved in cation-n interactions or salt bridges. Note that there may be amino acids in this list that are not listed in the amino acid warnings above (for example, a lysine that does not have its E-nitrogen in the PDB file may appear in this list, but not above). The following section summarizes the major findings. *********************** Cation-Pi Summary*********************** Number of ARG/PHE interacting pairs: 1 Number of energetically significant ARG/PHE cation-pi interactions: The following are the energetically significant ARG/PHE cation-pi interactions: 1 78 Cation AA # ARG 92 Chain Pi AA # PHE 21 Chain Number of ARG/TYR interacting pairs: E(es) (kcal/mol) -1.84 E(vdw) (kcal/mol) -1.07 0 Number of energetically significant ARG/TYR cation-pi interactions: Number of ARG/TRP interacting pairs: 0 Number of energetically significant ARG/TRP cation-pi interactions: Number of LYS/PHE interacting pairs: 0 1 Number of energetically significant LYS/TYR cation-pi interactions: Number of LYS/TRP interacting pairs: 0 0 Number of energetically significant LYS/PHE cation-pi interactions: Number of LYS/TYR interacting pairs: 0 0 0 Number of energetically significant LYS/TRP cation-pi interactions: 0 * There is a total of 1 energetically significant cation-pi interactions. * There are 0 cation-pi interactions with E(es) <= -2.0 kcal/mol * There are 1 cation-pi interactions with E(es) between -2.0 and -1 . 0 kcal/mol and with E(vdw) <= -1.0 kcal/mol * There is an average of 1 cation-pi interaction for every 108.0 residues. Successful Exit An interacting pair is a pair of amino acids in which the appropriate atoms of the side chains are arranged such t hat a water molecule cannot fit in between at the point of closest approach. Using the default parameters, an energetically significant cation-1r interaction is an interacting pair whose OPLS electrostatic energy, Ees, is either :S -2.0 kcal/mol OR whose Ees is :S -1.0 kcal/mol AND whose OPLS van der Waals energy, EvdW, is :S -1.0 kcal/mol. The criteria used for these calculations may be changed by the user. For examining a small number of proteins, the screen output provided by the default parameters is normally quite sufficient for discovering cation-1r interactions. 79 However, for analysis of large numbers of proteins, or for more details regarding the interactions, you may want to try to adjust the some of the parameters for tailoring your search . 3.6.5 Adjusting the Parameters The simplest way to change t he parameters is to run t he program from t he command line wit hout specifying a pdb filename. Simply type ' capture '. You will t hen be given t he option to change t he parameters. This section will describe each of t he options presented . The first questions involve energy cutoffs for determining cation-1r interactions. We use a 3-tiered system to determine whether an interaction is significant. Tier 1 : Interactions with E(es) below a primary cutoff (-2 . 0 kcal/mol) are automatically counted . Tier 2 : Interactions with E(es) below a secondary cutoff (-1.0 kcal/mol) are counted only if E(vdW) is less than a given energy (-1. 0 kcal/mol). Tier 3 : Interactions with E(es) above the secondary cutoff are rejected. What 0PLS electrostatic energy, E(es), in kcal/mol should be used for the primary cutoff (Tier 1)? (Default = -2 . 0) -> What DPLS electrostatic energy, E(es), in kcal/mol should be used for the secondary cutoff (Tier 2)? (Default = -1. 0) -> What 0PLS van der Waals energy, E(vdW), in kcal/mol should be used in conjunction with the secondary cutoff (Tier 2)? (Default = -1. 0) 80 -> The next question asks whether CaPTURE should consider salt bridges. In this context, a very liberal definition of a salt bridge is used. An interaction is considered a salt bridge when a cationic residue (Arg or Lys) forms an interacting pair with an anionic residue (Asp or Glu). As described above, an interacting pair is a pair of amino acids in which the appropriate atoms of the side chains are arranged such that a water molecule cannot fit in between at t he point of closest approach . Would you like to consider salt-bridges also? With this option, you will only see interactions determined by distance . Energy calculations are not performed . (Enter 1 for yes or 0 for no. Default= 0) -> The next question asks whether CaPTURE should write a PDB file showing only the two amino acids in a given interaction. Be forwarned t hat there may be many files written, which may take up a lot of space. Would you like a PDB file describing each interaction to be written? (Enter 1 for yes or 0 for no. Default= 0) -> The next question asks whether you would like to see t he advanced options. For most purposes, these need not be changed. However, the option to write '.stat' files (described below) found within the advanced options may be useful to some users. 81 Would you like to see advanced options? (Enter 1 for yes or 0 for no . Default= 0) -> The following questions all deal wit h t he geometrical definitions used in t he init ial screen for cation-?T interactions. Since CaPTURE uses energy-based criteria for determining cation-?T interactions, t hese would typically not be changed. Increasing t he distance beyond lOA, for example, would likely not result in finding more energetically significant cation-?T interactions, but would increase the calculation time by forcing t he program to calculate t he energies of interactions t hat are unlikely to be significant. In addit ion, t he angle preferences are of little use, unless you would like to use t he program for more specific searches. What is the minimum distance (in angstroms) between interacting pairs? (Default = 0. 0) -> What is the maximum distance (in angstroms) between interacting pairs? (Default= 10.0) -> What is the minimum theta angle (in degrees) between interacting pairs? (Default = 0. 0) -> What is the maximum theta angle (in degrees) between interacting pairs? (Default= 360 . 0) -> What is the minimum phi angle (in degrees) between interacting pairs? (Default = 0 . 0) -> What is the maximum phi angle (in degrees) between interacting pairs? (Default= 360.0) 82 -> The next question asks whether CaPTURE should write a file showing the OPLS energies for each interaction. Each file will contain the pairwise interaction energies for all atoms considered. Be forwarned that there may be many files written, which may take up a lot of space. This option is not very useful , except for debugging. Should I write an output file detailing the 0PLS energies for each interaction? These can take up a lot of space. (Enter 1 for yes or 0 for no. Default= 0) -> The final question asks whether CaPTURE should write an output file showing the statistics calculated by the program in a tabular form. The resulting '.stat' file (called pdbfilename.stat) can be processed easily using programs such as awk or perl to extract data, or by using a spreadsheet program to read the file. A description of the '.stat' files appears later in the document. Should I write an output file containing all of the statistics in tabular form? This may be hard to parse, but is useful for large scale searching. (Enter 1 for yes or 0 for no. Default= 0) -> 3.6.6 Limitations of CaPTURE As distributed, CaPTURE has a few limitations for the sizes of proteins that it can use. In normal use, these are unlikely to be a problem. CaPTURE can handle proteins 83 of up to 100,000 atoms, with up to 2,000 residues per chain, and up to 30 chains. Currently the maximum number of interactions is set at 5,000. All of these values can be changed by editing the file 'global. h ' and recompiling. CaPTURE cannot currently deal with non-protein units such as ligands, posttranslational modifications, and nucleic acids. All of these are ignored by the program. Because of the difficulty in determining the appropriate OPLS parameters for heteroatoms, there are no plans to introduce support for these. CaPTURE's PDB parser has recently been improved, but is certainly not perfect. If you have a PDB file that causes the program to crash, you may try to 'convert' it using the freely available program Babel. Simply convert the file from a PDB to a PDB. This will rewrite the file in a standard form that may be easier to read in. CaPTURE cannot handle inserted residues within PDB files. An example occurs in the protein crambin (PDB: lCBN), where residue 22 is a praline and residue 22B is a senne. CaPTURE will warn whether a protein has inserted residues and gracefully exit. Bugs Although CaPTURE has been extensively tested, there are bound to be some bugs in the program. If you spot one, please let me know by email (jpg@igor.caltech.edu) and I will take a look at it. CaPTURE has been tested on hundreds of files from the PDB and works reasonably well. As with any computational package, you should make sure that the results from the program pass the common sense test. If the program produces seemingly non-sensical results, please let me know and I will try to track down the problem. Notes The configuration file, capture. cf g, is created by the program the first time it is run. It contains all of the information that CaPTURE needs to do its job other than the name of the PDB file. You may edit this file , but it must NOT contain any blank lines or the program will hang. Additionally, there are options in the file that are not supported in the public distribution of CaPTURE (such as calculating solvent accessible surface areas). It is best that you not modify this file by hand. If you want 84 to modify it, you can run CaPTURE in the interactive mode and choose to change the parameters. This will update the configuration file. Contents of the '.stat' file For large-scale searching, it is often useful to have the output of the program written to a file in tabular form. This can be set under the advanced options setting (see above). When this option is set, CaPTURE will produce a '.stat' file, with the name 'pdbf ilename . stat ' where pdbfilename is that of the pdb file entered. The '.stat' file has four main sections: a remark section (REM); a sequence section (SEQ); a warning section (WARN); and an interaction section (INT) . Descriptions of these sections appear below. Remark Section (REM): This consists of a single line in the file (the first line) that begins with 'REM:'. This line contains the name of the original PDB file. Sequence Section (SEQ): This section is a single line beginning with the word SEQ. This line is tab-delimited into 'fields'. The first field contains the word SEQ. The second contains the filename. The following fields contain the number of residues in the protein, followed by the number of times each amino acid (limited to K,R,F ,Y,W,D, and E) occurs in the protein, as well as the percent composition of each amino acid. Warning Section (WARN): This section is a single line beginning with the word WARN, followed by the file name. It reports whether amino acids are missing atoms needed for the calculations. For example, sometimes only the backbone trace of a protein will be in the PDB file. Even though the protein may have 3 lysines in its sequence, there are no coordinates for the side chains and thus the program cannot determine whether they are in cationJr interactions. These amino acids also appear under the 'WARNING! These amino acids are missing atoms needed for the calculations: ' section in the screen output. Also on this line are the percentages of the amino acids that are missing atoms. For example, if a protein has 3 lysines in the sequence, and one is missing atoms, this number will be 1/3 = 33.3%. If the percentages are very high, this indi- 85 cates a PDB file that is missing a significant number of atoms, and care should be taken in interpreting the output of the program. Interaction Section (INT): This section is the most important of the '.stat' file. It contains information about each pairwise interaction considered. Each line consists of tab-delimited fields that are as follows: 1. The word INT. 2. A representation of the filename of the interaction. If you have specified the option of writing PDB files, this will be the filename of the PDB file for the interaction (if you replace the .z with a .pdb). 3. The type of cation. 4. The residue number of the cation. 5. The chain identifier of the cation (or a dash if the protein is a single chain). 6. The type of 1r or anion. 7. The residue number of the 1r or anion. 8. The chain identifier of the 1r or anion (or a dash if the protein is a single chain). 9. The distance, r, in A between the cationic center and either the centroid of the aromatic, or the carbonyl carbon of an anionic side chain. 10. The angle, 0, in degrees, between a vector from the ring centroid and the cationic center, and a vector between the ,8-carbon of the 1r-system and the centroid. This is undefined for an anionic partner and will appear as -1000.0. 11. The dihedral angle, ¢, in degrees, between a vector from the ring centroid and the cationic center, and a vector between the ,8-carbon of the 1r-system and a normal to the ring. This is undefined for an anionic partner and will appear as -1000.0. 86 12. The OPLS van der Waals energy, EvdW, (in kcal/ mol), for t he interaction. Note that this is not applicable for salt bridges. 13. The OPLS electrostatic energy, E es, (in kcal/ mol) , for the interaction. Note that this is not applicable for salt bridges. 14. The tot al OPLS energy (in kcal/ mol), for the interaction. This is the sum of the van der Waals and electrostatic energies (12 + 13). Note that this is not applicable for salt bridges. 15. The interplane angle between an arginine side chain and a ring. 90 degrees is perpendicular (T-shaped) , and O and 180 are parallel. 16. The 'gap' measurement. If this number is less than 0, no water molecule can fit between t he part ners at t heir closest approach point. 17. This is r as specified in 9, but it is calculated relative to t he 6-ring in the case of Trp. This makes plotting easier. 18. This is 0 as specified in 10, but it is calculated relative to t he 6-ring in t he case of Trp. This makes plotting easier. 19. This is¢ as specified in 11 , but it is calculated relat ive to the 6-ring in the case of Trp. This makes plotting easier. 20. This is the distance between either the c:-carbon of Lys or the b-carbon of Arg and t he cent roid of t he closest aromatic ring. This will appear as -1000 .00 for a salt bridge. 21. The total number of residues in the protein. Using standard UNIX tools such as grep, cut, and awk, you can easily write scripts to extract information from t hese files. Alternatively, they may be read into spreadsheets. 87 3. 7 Appendix 2 - Description of CaPTURE Source This section is meant to describe the CaPTURE source code so that others may be able to modify it. The various features of many of the functions will be briefly outlined. CaPTURE is written in C++ and was developed under IRIX 6.2 and IRIX 6.5.5. There are no graphics or !RIX-specific functions within the code. Thus , CaPTURE should be able to be compiled under other systems with minimal modifications. The main body of code is found in the source file Capture. c. A schematic of the files and functions called by Capture.c are shown in Figure 3.14. Descriptions of the major functions in each source file follow. Capture.c Capture. c is the heart of the program. Many of the structures and variables are initialized within Capture. c. Welcome.c Welcome. c introduces the user to the program. It first checks whether the user entered a PDB filename on the command line. If so, it checks for the existence of the configuration file capture. cfg. If capture. cfg exists, the program uses the parameters stored within the configuration file and runs without further user input. If either a PDB filename or the configuration file are missing, the program enters interactive mode , and prompts the user for more information. If there is no configuration file, the WriteConfig function (defined in Wri teConf ig. c) is called, and creates the file capture . cfg. ReadPDB.c ReadPDB. c reads information from the PDB file into the PDB structure. It also determines the total number of chains within the protein. ReadParams.c ReadParams . c assigns default values to many of the user-changeable parameters. It then reads information from the configuration file capture . cf g and adjusts the parameters accordingly. There are a few parameters in here, such as the options for 88 Capture.c Welcome.c WriteConfig.c ReadPDB.c t ReadParams.c t Tessel.c t Divide.c t AssignRadii.c ~ ~ FindFunctions.c AssembleProtein.c t Add DummyAtoms.c ________t ------' I '. __ _:1_d~1!!fd_r~tel'_!~c_ __ ; t AddOPLS.c t IMineBuriedCharges.c I -- ~------------. ~------------. WriteZMat.c •- *, WritePDB .c ' _ - ..-, ...... ...... ___ ScreenCandidates.c I t - - - - - - - - - - - - -· ~------------. - - - - - - - - - - - - -· - ...., CalculateOPLS.c • I - - - - - - - - - - - - _I '. __ ~a_}c_f-.!~x~!:!rJ~c:·: __ ; t CalcWeight.c I WriteScreenOutput.c I ________t ------' I '. ____ iy~i~e~1!rJ-: ____ ; t Cleanup.c Figure 3.14: Flowchart of Capture. c The header files (in ovals) are available to all functions, as are the functions found in FindFunctions. c. Files shown in solid boxes are always called. Those in dashed boxes are optional. The arrows show the direction of program flow. 89 Gaussian calculations, which can only be altered at compile time. Tessel.c and Divide.c Tessel. c and Divide. c contain functions used by the GEPOL algorithm for calculation of the solvent-accessible surface area of the protein. (See: Silla, E.; Tuii6n, I.; Pascual- Ahuir , J. L. J. Comput. Chem. 1991, 12, 1077- 1088.) They are small functions and are called regardless of whether the solvent-accessible surface is actually calculated. These should not be modified. AssignRadii.c and AddSolvent.c AssignRadii. c assigns an atomic radius to each atom for use in calculating the solvent-accessible surface area. It uses the values from: Rose, G. D.; Geselowitz, A. R.; Lesser, G. J.; Lee, R. H.; Zehfus, M. H. Science 1985, 229, 834- 838. If the solvent-accessible surface is to be calculated, the AddSol vent function adds the user-defined solvent radius to each atom. Surface.c and PrintSurface.c These files contain the code to determine the solvent-accessible surface of each atom. They are adapted from: Silla, E. ; Tuii6n, I.; Pascual- Ahuir, J. L. J. Comput. Chem. 1991, 12, 1077-1088. AssembleProtein.c Until now , all data are treated as individual atoms. This function parses the atoms and creates the Protein structure. A protein consists of chains, which consist of amino acids, which consist of a series of pointers to the atoms in the Atom structure. This function also determines whether the occupancy of the individual atoms is greater than a user-specified value. In addition, the relative accessibility of each amino acid can be calculated. AddDummy Atoms.c This function adds dummy atoms to each aromatic ring, including the centroids and surface normals used to calculate various geometric parameters. AddHydrogens.c This function adds hydrogen atoms to each aromatic ring (provided they are not there already) for use in calculating the OPLS energies. Since Arg and Lys are treated 90 as united atoms, hydrogens are not added. AddOPLS.c This function adds t he OPLS parameters, q, a , and E to the atoms using the Assign □ PLS function , defined within Add0PLS. c. The parameters are taken from [23, 24]. MineBuriedCharges.c This function is largely a relic and should probably be phased out. It was originally intended to determine whether an amino acid is buried by comparing its solventaccessible surface area to a user-defined parameter. It also calculates the total charge of a protein and the population of each type of amino acid. ScreenCandidates.c The ScreenCandidates function is the computational engine of the CaPTURE program. ScreenCandidates searches through the protein for potential cation-1r interactions, or salt bridges. When it finds a potential pair of residues , it checks to see if all of the necessary coordinates are present, and uses the functions defined in FindFunctions . c to calculate the distance, angle, and dihedral information. Using this information, it can call the functions Wri te*ZMatrix. c and Wri tePDB . c to write files describing the interactions. ScreenCandidates can also call the function Calculate0PLS (defined in Calculate0PLS. c) , to determine the OPLS energy of the interacting pair. If the interacting pair is between an Arg and an aromatic, the interplane angle (IPA) is calculated. In addition, ScreenCandidates performs various book-keeping activities, including counting the total number of each interaction, and creating the filenames for the .z and .PDB files. CalcMaxSurface.c If CaPTURE is set to calculate the solvent-accessible surface area, t his function calculates the maximum accessible area of a given residue (i) by calculating its area in the presence of the backbone coordinates of the (i - 1) and (i + 1) residues. CalcWeight.c This function calculates the molecular weight of the protein. 91 WriteCatalog.c This function writes t he ' . stat ' file which contains all of t he data generated by the program in t abular form , useful for large-scale database searching. WriteScreenOutput .c This function formats the most important results of t he program for display on t he screen. In many cases, t his is t he only out put a user will see or need. WriteSurf.c If t he solvent-accessible surface area is calculated, a file containing all of the results for each amino acid is printed. Cleanup.c This function clears the memory. Other files In addit ion t o t he ' . c' source files outlined here, t here are also two ' . h' header files: global. h and function. h. global. h contains t he definitions of all structures and global variables within the program. function. h contains t he prototypes for all of the functions available globally. 92 Bibliography [1] D. A. Dougherty. Cation-1r interactions in chemistry and biology. A new view of benzene, Phe, Tyr, and Trp. Science, 271:163- 168, 1996. [2] J. C. Ma and D. A. Dougherty. The cation-1r interaction. Chem. Rev., 97:13031324, 1997. [3] N. S. Scrutton and A. R. C. Raine. Cation-1r bonding and amino-aromatic interactions in the biomolecular recognition of substituted ammonium ligands. Biochem. J., 319:1- 8, 1996. [4] J.L. Sussman, M. Harel, F. Frolow, C. Oefner, A. Goldman, L. Toker, and I. Silman. Atomic structure of acetylcholinesterase from Torpedo califomica: A prototypic acetylcholine-binding protein. Science, 253:872-879, 1991. [5] W. Zhong, J. P. Gallivan, Y. Zhang, L. Li, H. A. Lester, and D. A. Dougherty. From ab initio quantum mechanics to molecular neurobiology: A cation-1r binding site in the nicotinic receptor. Proc. Natl. Acad. Sci. (USA), 95:12088-12093, 1998. [6] M. F. Perutz, G. Fermi, D. J. Abraham, C. Poyart, and E. Bursaux. Hemoglobin as a receptor of drugs and peptides: X-ray studies of the stereochemistry of binding. J. Am. Chem. Soc., 108:1064- 1078, 1986. [7] M. Levitt and M. F. Perutz. Aromatic rings act as hydrogen bond acceptors. J. Mal. Biol., 201:751- 754, 1988. [8] M. F. Perutz. The role of aromatic rings as hydrogen-bond acceptors in molecular recognition. Phil. Trans. R. Soc. A, 345:105- 112, 1993. [9] S.K. Burley and G.A. Petsko. Amino-aromatic interactions in proteins. FEES Lett., 203: 139- 143, 1986. 93 [10] C.A. Deakyne and M. Meot-Ner (Mautner). Unconventional ionic hydrogen bonds. 2. NH+ 1r. complexes of onium ions with olefins and benzene derivatives. J. Am. Chem. Soc., 107:474- 479, 1985. [11] D.A. Rodham, S. Suzuki, R.D. Suenram, F.J. Lovas, S. Dasgupta, W.A. Goddard III, and G.A. Blake. Hydrogen bonding in the benzene-ammonia dimer. Nature, 362:735- 737, 1993. [12] J. Singh and J. M. Thornton. Sirius. An automated method for the analysis of the preferred packing arrangements between protein groups. J. Mol. Biol., 211:595- 615, 1990. [13] J. B. 0. Mitchell, C. L. Nandi, S. Ali, I. K. McDonald, J. M. Thornton, S. L. Price, and J. Singh. Amino/aromatic interactions. Nature, 366:413, 1993. [14] J. B. 0. Mitchell, C. L. Nandi, I. K. McDonald, J. M. Thornton, and S. L. Price. Amino/ aromatic interactions in proteins: Is the evidence stacked against hydrogen bonding? J. Mol. Biol., 239:315- 331, 1994. [15] C. L. Nandi, J. Singh, and J. M. Thornton. Atomic environments of arginine side-chains in proteins. Protein Eng., 6:247- 259, 1993. [16] J. B. 0. Mitchell, R. A. Laskowski, and J. M. Thornton. Non-randomness in side-chain packing: The distribution of interplanar angles . Proteins, 29:370-380, 1997. [17] J. Singh and J. M. Thornton. Atlas of Protein Side-Chain Interactions, volume 1 and 2. IRL Press, Oxford, 1992. [18] M. M. Flocco and S. L. Mowbray. Planar stacking interactions of arginine and aromatic side-chains in proteins. J. Mol. Biol., 235:709- 717, 1994. [19] M. Hendlich. Database for protein-ligand complexes. Acta Cryst., D54:11781182, 1998. 94 [20] J. Wouters. Cation-1r (Na+-Trp) interactions in crystal structure of tetragonal lysozyme. Protein Sci., 7:2472- 2475, 1998. [21] A. Y. Ting, I. Shin, C. Lucero, and P. G. Schultz. Energetic analysis of an engineered cation-1r interaction in staphylococcal nuclease. J. Am. Chem. Soc., 120:7135- 7136, 1998. [22] E. E. Abola, F. C. Bernstein, S. H. Bryant, T. F. Koetzle, and J. Weng. Protein Data Bank. Data Commission of the International Union of Crystallography, Bonn/Cambridge/Chester, 1987. [23] W. L. Jorgensen. The OPLS potential functions for proteins. Energy minimizations for crystals of cyclic peptides and crambin. J. Am. Chem. Soc., 110:16571666, 1988. [24] W. L. Jorgensen, D. S. Maxwell, and J. Tirado-Rives. Development and testing of the OPLS all-atom force field on conformational energetics and properties of organic liquids. J. Am. Chem. Soc., 118:11225- 11236, 1996. [25] M. J. Frisch, G. W. Trucks, H. B. Schlegel, P. M. W. Gill, B. G. Johnson, M.A. Robb, J. R. Cheeseman, T. A. Keith, G. A. Petersson, J. A. Montgomery, K. Raghavachari , M. A. Al-Laham, V. G. Zakrzewski, J. V. Ortiz, J. B. Foresman, C. Y. Peng, P. A. Ayala, M. W. Wong, J. L. Andres, E. S. Replogle, R. Gomperts, R. L. Martin, D. J. Fox, J. S. Binkley, D. J. Defrees, J. Baker , J. P. Stewart, M. Head-Gordon, C. Gonzalez, and J. A. Pople. Gaussian 94 (Revision D.3). Gaussian, Inc., 1995. [26] K. S. Kim , J. Y. Lee, S. J. Lee, T-K. Ha, and D. H. Kim. On binding forces between aromatic rings and quaternary ammonium compounds. J. Am. Chem. Soc., 116:7399- 7400, 1994. [27] A. Pullman, G. Berthier, and R. Savinelli. Interaction of the tetramethylammonium ion with the cycles of aromatic amino acids beyond the SCF ab initio level. J. Am. Chem. Soc., 120:8553- 8554, 1998. 95 [28] R.A. Kumpf and D.A. Dougherty. A mechanism for ion selectivity in potassium channels: Computational studies of cation-1r interactions. Science, 261:17081710, 1993. [29] J. W. Caldwell and P. A. Kollman. Cation-1r interactions: Nonadditive effects are critical in their accurate representation. J. Am. Chem. Soc., 117:4177- 4178, 1995. [30] 0. Donini and D. F. Weaver. Development of modified force field for cation-amino acid interactions: Ab initio-derived empirical correction terms with comments on cation-1r interactions. J. Comput. Chem., 19:1515-1525, 1998. [31] S. Mecozzi, A. P. West Jr., and D. A. Dougherty. Cation-1r interactions in simple aromatics. Electrostatics provide a predictive tool. J. Am. Chem. Soc., 118:2307- 2308, 1996. [32] S. Mecozzi, A. P. West Jr., and D. A. Dougherty. Cation-1r interactions in aromatics of biological and medicinal interest: Electrostatic potential surfaces as a useful qualitative guide. Proc. Natl. Acad. Sci. USA, 93:10566- 10571, 1996. [33] U. Hobohm, M. Scharf, R. Schneider, and C. Sander. Selection of representative protein data sets. Protein Sci., 1:409- 417, 1992. [34] U. Hobohm and C. Sander. Enlarged representative set of protein structures. Protein Sci., 3:522- 524, 1994. [35] P. J. Kraulis. J. Appl. Cryst., 24:946- 950, 1991. [36] E. M. Duffy, P. J. Kowalczyk, and W. L. Jorgensen. Do denaturants interact with aromatic hydrocarbons in water? J. Am. Chem. Soc., 115:9271- 9275, 1993. 96 Chapter 4 Computational Studies of Cation-1r Interactions and Salt Bridges 97 4.1 Introduction Of the myriad noncovalent interactions that contribute to protein stability, few are both specific and strong when fully exposed to an aqueous medium. Hydrogen bond formation , for example, is opposed by competing interactions with water, and the hydrophobic effect is not specific for any particular pair of residues. Although salt bridges are specific and quite strong in non-polar media, there is debate over the extent to which water-exposed salt bridges contribute to protein stability [1- 6]. Many studies estimate that a water-exposed salt bridge contributes 1-2 kcal/mol to protein stability, while others suggest a salt bridge is worth very little energetically, and may even be destabilizing. There is, however , a relatively underappreciated noncovalent binding force that is potentially both specific and strong in an aqueous environment. Recent studies have shown that cation-n interactions can be quite strong in aqueous media [7]. In addition, evidence from the preceding chapters suggests that not only are cation-n interactions widely used in protein structures, but that they are frequently found on the surfaces of proteins, exposed to aqueous solvation. This naturally leads to the question: "Why?" In this chapter, we attempt to answer this question by comparing the attributes of cation-n interactions to those of other motifs commonly found in protein structures. In particular, we present a computational study of the relative strengths of salt bridges and cation-n interactions both in aqueous media and in a broad range of organic solvents. In doing so, we hope to define a unique role for cation-n interactions within biological systems. While there have been other computational studies of salt bridges [8-14] and of cation-n interactions [15- 20] in water, a wide range of theoretical models of differing complexities and reliabilities has been employed. Previous studies of cation-n interactions in water have relied primarily upon force field-based methods and statistical perturbation theory [15- 17]. Although traditional two-body force field-based methods are able to reproduce the trends in cation-n binding abilities, they are challenged to accurately reproduce the results of both experiment and high-level it ab initio calculations. Kollman and co-workers have attempted to improve two-body de- 98 scriptions of the cation-n interaction by adding "10-12" functions to the Amber force field [16]. Although these functions allow the force field method to reproduce the results of high-level ab initio calculations in the gas phase, they nonetheless ignore the effects of both polarization and charge-transfer, which are important in the accurate description of cation-n interactions [21, 22]. Similarly, a previously described hybrid quantum mechanical/molecular mechanical study [18] is compromised by the use of the AMl Hamiltonian, which is similarly poor in treating cation-n interactions. 1 Since ab initio calculations more accurately reproduce experimental results with regard to cation-n interactions [7], we anticipated that a modern implementation of aqueous salvation in a quantum mechanical calculation might be useful. Cramer and Truhlar recently introduced the SM5.42R/HF method which combines self-consistent reaction field calculations at the Hartree-Fock level with the generalized Born approximation plus surface tension terms to calculate free energies of salvation [23- 25]. The SM5.42R/HF method accurately predicts the free energies of salvation of hundreds of neutral molecules and ions, including the monomers used in this study, and has recently been used to explore solvent effects on bimolecular reactions in solution [26]. In addition, the calculations exhibit rapid convergence and proper dissociation behavior in supermolecule calculations, and they are very fast. Most importantly, the method treats the intermolecular interactions at the Hartree-Fock level, thus accounting for the effects of polarization and charge transfer [21 , 22]. For these reasons , we used the SM5.42R/HF method to compare the relative strengths of cation-n interactions and salt bridges at the same level of theory, so as to determine the potential significance of the less-studied cation-n interaction vs. the well-appreciated salt bridge. 1 See Chapter 2. 99 4.2 Results 4.2.1 Ab Initio Calculations As a model for the cation-1r interaction , we chose methylammonium ion binding to benzene (Figure 4.1), which mimics the interaction between lysine and phenylalanine in protein structures. For the salt bridge, we chose methylammonium ion binding to acetate, which represents the interaction between lysine and glutamate or aspartate (Figure 4.1). To calculate the interaction energies, we first optimized the geometries of the individual molecules in the gas phase using HF /6-31+G* calculations. Then, using fixed monomer geometries, we optimized the complexes in the gas phase at the HF /6-31+G* level. We find that gas phase binding energies for methylammonium to benzene and acetate are -12.5 kcal/ mol and -125.5 kcal/mol respectively. Not surprisingly, the ion-pair is much stronger than the cation-1r interaction in the gas phase. It should also be noted that the gas-phase geometries were restricted to exist as the ion pair. Allowing the proton to transfer to form a neutral hydrogen bonded geometry would further lower the energy [27]. Figure 4.1: Geometries for methylammonium· ··benzene (left) methylammonium· ··acetate (right). The distance,d, is reported in Figure 4.2. and To perform calculations in solution, we used the gas phase monomer geometries, altered the distances between the monomers ( d in Figure 4.1), and performed SM5.42R/HF /6-31 +G* supermolecule calculations to determine an "interaction en-: ergy," defined in Equation 4.1 Int eraction Energy =Eab(gas) - Ea(gas) - Eb(gas) + (4.1) 100 where Eab(gas) is the gas phase energy of the complex, Ea(gas) and Eb(gas) are the gas phase energies of monomers a and b, D.Gab(sol) is the free energy of salvation of the complex, and D.Ga(sol) and D.Gb(sol) are the free energies of salvation of the monomers a and b. The results of these calculations with water as the solvent are shown in Figure 4.2. 8.0 ---0 6.0 s 4.0 "@ ---u 0 >, cO ....<l) C - - - Methylammonium • • • Acetate --o-- Me thy lammoni um • • • Benzene 2.0 0.0 µ.l C 0 -2.0 u ....co -4.0 ·z .....<l) C -6.0 -8 .0 3 4 5 6 d (A) 7 8 9 10 Figure 4. 2: Interaction energies at the SM5.42R/ HF / 6-31+G* level for methylammonium· · ·benzene and methylammonium· · · acetate in water. The data used to generate this figure can be found in Tables 4.2 and 4.3 in the Methods section. Figure 4.2 clearly shows t hat both the salt bridge and the cation-1r interaction have well-defined minima in water at this level of theory. The salt bridge experiences a broad, shallow minimum of -2.2 kcal/ mol at a distance of 3.5 A. This result is compatible with other studies, and the overall interaction surface is smooth and wellbehaved, further supporting the reliability of the SM5.42R method as applied here. In contrast, the cation-1r interaction experiences a substantially deeper minimum, with -5.5 kcal/ mol of interaction energy at a distance of 3.75 A. We note that Hartree-Fock theory underestimates the magnitude of the methylammonium· · · benzene interaction energy in the gas phase. In part icular, the level of theory used here (HF / 6-31+G*) 101 gives a value of -12.5 kcal/mol, while the experimental 6.H is -18.8 kcal/mol [28]. As such, it may well be that the approach used here underestimates the cation-n well depth in water. While higher levels of theory perform better on the gas phase binding energy (e.g., the MP2/6-31+G* value is -17.3 kcal/mol), such methods are not presently compatible with the SM5.42R method. Since any change to a higher level of theory is expected to increase the magnitude of the cation-n interaction in water, our basic conclusion that the cation-7!' interaction is stronger than a typical salt bridge in water remains. What is perhaps most striking is that in aqueous solution, the magnitude of the salt bridge interaction is reduced by over fifty-fold relative to the gas phase, whereas the cation-n interaction is reduced by less than three-fold. This suggests that cation-n interactions will remain strong across a wide range of dielectric constants. To test this notion, we performed calculations analogous to those shown in Figure 4.2 in differing solvents. The results of calculations using carbon tetrachloride, ethyl acetate, ethanol, and acetonitrile as solvents are shown in Table 4.1. solvent none CCl4 CH3CO2CH2CH3 CH3CH2OH CH3CN H2O dielectric 1.0 2.23 5.99 24.85 37.5 78.0 cation-n interaction energy -12.5 -7.8 -6.2 -5.6 -5.6 -5.5 salt bridge interaction energy -125.5 -53.4 -19.7 -5.2 -3.8 -2.2 Table 4.1: Interaction energies (kcal/mol) at the SM5.42R/HF /6-31+G* level for methylammonium- ··benzene and methylammonium· · ·acetate in a range of solvents. As might be anticipated, both the cation-n interaction and the salt bridge show slightly deeper minima in acetonitrile than in water. In the non-polar solvents carbon tetrachloride and ethyl acetate, the results tend to approach the gas phase numbers, and the energy minimum of the salt bridge is substantially deeper than that of the cation-n interaction. Interestingly, the minima for both the cation-7!' interaction and the salt bridge are very similar in ethanol. A plot of the dielectric dependence of the 102 -20 0 E -40 el --1 ';: -60 ... o() <!) C i:l.l C 0 -80 ·.ou ... ro 2 -100 .E: --e-- Methylammonium •••Benzene --+- Methylammonium• • •Acetate -120 -140 ~~~~~~~~~~~~~~~~~~ 0.2 0.4 0.6 0.8 0 1 1/E Figure 4.3: Interaction energies at the SM5.42R/HF /6-31+G* level for methylammonium· ··benzene and methylammonium· ··acetate in water, acetonitrile, ethanol, carbon tetrachloride, and the gas phase, plotted against the dielectric constant of the medium. R > 0.99 for both lines. The intersection of the lines is at E = 22.7. binding energies is shown in Figure 4.3. Clearly variations in the dielectric constant of the medium are the overwhelming contributor to variations in the strength of the interactions. This is to be is expected because of the dependence of the Born salvation model as implemented here on the dielectric constant of the medium. Nonetheless, Figure 4.3 predicts that the strengths of the cation-1r interaction and the salt bridge are approximately equal in a solvent with a dielectric constant of 22. 7 - slightly less polar than ethanol, and slightly more polar than acetone (c=20.7). Given these results, it is perhaps tempting to suggest that the minima in Figure 4.2 and Table 4.1 may represent !:::..G for the association of the two partners.2 This would suggest that these interactions have association constants in the range observable by methods such as NMR. However, to obtain a better estimate of the association constant of the binding process, one should use Equation 4.2 to integrate the potential 2 Jorgensen has suggested that this interaction energy approximates 6.Gbind [29]. 103 of mean force (PMF) curves of Figure 4.2 [16]. (4.2) This approach has shortcomings because the curves above were derived using only a very small number of configurations, rather than by sampling the large number of configurations typically examined in a statistical perturbation study. Because only a limited number of configurations was sampled, any estimate of the association constant will be too favorable, and will thus represent an upper limit to the magnitude of the interaction. The calculated association constant for the cation-1r interaction between methylammonium and benzene in water is 401 M- 1 , while that for the salt bridge between methylammonium and acetate is 6 M- 1 . This suggests that the upper limit for 6.Gbind for the cation-1r interaction is approximately -3.6 kcal/mol, while that of the salt bridge is approximately -1.0 kcal/mol. We again stress that these values must be considered with caution because of the limited number of configurations and the symmetry constraints imposed. For example, in a study of a similar system using molecular mechanics methods, Kollman found that relaxing constraints similar to those employed here diminished the binding constant by a factor of 30 [16]. Nonetheless, we do note that these numbers are consistent with earlier computational studies that employed similar constraints. In particular, the value of 6.Gbind for the cation-1r interaction (-3.6 kcal/mol) agrees remarkably well with the value of -3.2 kcal/mol determined by Kollman and co-workers for a similarly constrained simulation [16]. In addition, the value for 6.Gbind for the salt bridge is in good agreement with experimental values ranging from -0.2 to -1.25 kcal/mol for surface exposed salt bridges determined by Fersht and co-workers [4]. This agreement with both experiment and theory further demonstrates the applicability of the SM5.42R/HF method for calculating weak interactions in solution. We propose the following simple explanation for why the formation of a cation-1r interaction may be favored over the formation of a salt bridge in water. To form an ion pair in water, two well-solvated ions must become desolvated to a considerable 104 extent. The binding energy of the complex may be very strong, but the desolvation penalty may be of comparable magnitude. In addition, although the ion pair complex is strongly polar, to some extent there is a neutralization of charge in bringing two opposite charges in close contact. This will diminish the long range Born solvation of the ion pair relative to the isolated ions. To form a cation-1r interaction, a cation must pay a similar penalty to partially desolvate. However, the aromatic partner is poorly solvated by water, and so the desolvation associated with cation binding is most likely beneficial. In addition, no "charge neutralization" occurs, and so the cationic complex can still reap the substantial benefits of long-range solvation by water. 4.2.2 Protein Studies The possibility that cation-1r interactions are strong and specific in aqueous media suggests that exposed cation-1r interactions may also contribute to protein structure and stability. In Chapter 3, we demonstrated that cation-1r interactions are quite common within proteins. In that chapter, it was suggested, though not proven, that cation-1r interactions tend to occur on the surfaces of proteins, rather than buried in the cores. To more rigorously explore this hypothesis in the context o{ the present chapter, we have determined the extent to which cation-1r interactions are exposed to water in protein structures. To do this, we implemented within the CaPTURE (Cation-1r Trends Using Realistic Electrostatics) program a variant of the GEPOL method [30] to calculate the solvent accessible surface area (SASA) of individual interactions. Traditionally, calculation of relative surface accessibility of amino acids has relied on comparing the SASA of an individual amino acid within the protein to the SASA of the same amino acid in a GXG peptide (where G is glycine and X is the amino acid). However, in a cation-1r interaction, since the interacting partners are in contact with one another, their surface exposed area is by definition less than that of the amino acid in a GXG peptide, even though the interacting pair as a unit may be highly exposed. To overcome this difficulty, we have chosen to calculate the 105 accessibility of a cation-n interaction considered as a unit within the protein, relative to the accessibility of the amino acids in their identical orientation, but not in the context of the protein. To do this, the solvent accessible surface area for each partner in the cation-n interaction is calculated. The total exposed area of the two partners is then compared to the maximum exposed area of the two partners. This is determined by "excising" the interaction from the rest of the protein. The coordinates of the amino acids involved in the cation-n interaction, as well as the backbone coordinates of the (i - 1) and (i + 1) residues of both the cation and the aromatic are retained, and the rest of the atoms in the protein are deleted. The solvent accessible surface area of the cation-n interaction in this minimal model is calculated. This represents the maximum possible surface area of the cation-n interaction taken as a unit. The ratio of t he exposed area of the cation-n interaction within the protein to the maximum area of the cation-n interaction considered by itself can be e:xpressed as a "percent exposed". Using this method, we can also determine the "percent exposed" value for each amino acid within the dataset. As a dataset of proteins, we chose the 593 non-homologous proteins evaluated in Chapter 3. The exposure of each amino acid, as well as 2878 energetically significant cation-n interactions, was calculated using the method outlined. The results are shown in Figure 4.4. The results in Figure 4.4 are largely as expected for both the cationic and aromatic amino acids. Phe, Tyr , and Trp tend to be buried within the proteins with over 40% of the aromatics completely buried. In addition, the aromatic amino acids are rarely highly exposed - only 20% have more than 20% of their surfaces exposed to water. Not surprisingly, the cationic amino acids Lys and Arg tend to be exposed - 70% of all cationic amino acids expose more than 20% of their surfaces to water, and much less than 10% of these residues are completely buried. It might be anticipated that cation-n interactions would resemble the average of the cationic and aromatic amino acid curves in Figure 4.4. However, this is not the case. Instead, cation-n interactions tend to behave as a hybrid between the cationic and aromatic amino acids. On the one hand , cation-n interactions are rarely completely buried within proteins. Only 11 % of 106 50 40 - + - Cation-re --o--- Aromatics (F /Y /W) ......ell - . - Cations (R/K) 0 30 --e-- Average of aromatics and cations E-< "-< 0 ~ (I) u tij 20 p... 10 0 0 20 40 60 80 100 Percent Exposure Figure 4.4: The percentage of the total number of residues (interactions) that have a given relative exposure. Abbreviations: F=phenylalanine, Y =tyrosine, W =tryptophan, R=arginine, K=lysine. cation-1r interactions are completely buried, compared to 7% for cationic amino acids and 42% for the aromatics. In this regard, they tend to behave more like cations than like aromatics. On the other hand , while cation-1r interactions are more often highly exposed than are aromatics, they are certainly less often exposed than are the cationic amino acids. This is perhaps not surprising as a cation-1r interaction is by definition more hydrophobic than a cationic amino acid. Nevertheless, the question of whether cation-1r interactions are "exposed" still remains. The choice of what degree of exposure constitutes "exposed" is somewhat arbitrary. Previous workers have offered varying criteria for exposure. In particular, Rost and Sander have suggested several models [31]. The first is a two-state model in which amino acids whose exposed area is greater than 16% are considered exposed, 107 100 80 -+-- Cation-n --o-Aromatics (F/Y /W) _ . _ Cations (R,K) 20 0 0 20 40 60 80 100 Percent Exposure Figure 4.5: Integration of the data in Figure 4.4 for cationic amino acids, aromatic amino acids, and cation-1r interactions. The vertical dotted line represents the 16% criterion in the two-state model. The shaded box represents the "intermediate" state of the three-state model (9% < exposure < 36%). while those less than 16% are considered buried. The second is a three-state model in which amino acids whose exposed area is< 9% are considered buried, those between 9 and 36% exposure are intermediate, and those > 36% are considered exposed. Figure 4.5 integrates the data for cationic amino acids, aromatic amino acids, and cation-1r interactions shown in Figure 4.4. The vertical dotted line represents the 16% criterion in the two-state model points to the right of this line are exposed, points to the left are buried. The shaded box represents the "intermediate" state of the three-state model (9% < exposure < 36%). Points to the left of the box are considered buried and those to the right are considered exposed in the three-state model. To determine the nature of a particular interaction, we define the PE50 as the percent exposure at which 50% of the amino 108 acids are more exposed than this point and 50% are less exposed. For a particular curve, PE 50 is the intersection between it and the horizontal dotted line in Figure 4.5. PE 50 is 3% for the aromatic amino acids, 20% for the cation-n interactions and 34% for the cationic amino acids. Using the two-state model suggests that the average aromatic amino acid is buried, and that the average cation-n interaction or cationic amino acid is exposed. Using the three-state model, aromatic amino acids are still considered buried, whereas both cation-n interactions and cationic amino acids are considered "intermediate", although the cationic amino acids have a higher PE 50 than the cation-n interactions. Thus, using two different sets of criteria for determining exposure, the cation-n interaction falls into the same broad classification as the cationic amino acids. Regardless of the particular (arbitrary) cutoffs, the trend is clear - cation-n interactions tend not to be buried, reflecting their cationic character, but they are not as exposed as cationic residues by themselves, reflecting their aromatic component. 4.3 Summary The results presented in this chapter represent the state-of-the-art in continuum calculations and compare, at a consistent level of theory, the strengths of salt bridges and cation-n interactions in water and in a range of organic solvents. The SM5.42R/HF method is well behaved for this problem, and , while future efforts may refine the present results , we believe the basic trend will survive. In particular, we suggest that in water, a cation-n interaction is more stabilizing than an analogous salt bridge. Not surprisingly, the strengths of the interactions are strongly dependent on the dielectric constant of the medium, as would be expected for interactions with large electrostatic components. Interestingly, the cation-n interaction maintains its strength across a range of solvents. Whereas the strength of the salt bridge is reduced over 50-fold on moving from the gas phase to water, the strength of a cation-n interaction is weakened by less than a factor of three. Thus, cation-n interactions can provide strong, specific interactions on the surfaces of proteins. Consistent with this prediction, we find that 109 cation-1r interactions are rarely buried within proteins in the Protein Data Bank, and more often than not, are exposed to water. These results suggest a novel mechanism to increase the stability of proteins. Most efforts at protein design have emphasized optimization of packing in the core of the protein [32]. Engineering solvent-exposed cation-1r interactions, especially at the interfaces between domains or subunits, may provide a complementary strategy. Perhaps most importantly, these results may suggest the raison d'etre for cation-1r interactions in protein structures. In contrast to interactions such as salt bridges and hydrogen bonds, which are severely attenuated in water, or the hydrophobic effect, which lacks specificity, cation-1r interactions are both strong and specific in water. As such, cation-1r interactions may play a unique role by providing a means to stabilize regions of proteins that are exposed to solvent. 4.4 Appendix 1 - Methods All calculations were performed using Gaussian 94, Rev. D3 [33], enhanced with the Minnesota Gaussian Salvation Module Version 98.2.5 (GSM) [23]. All solution phase calculations were performed with SCRF (self-consistent reaction field) model 2, using the option ISCRF=2. For the CH 3 NHt · · · benzene complex, the nitrogen was constrained to the C6 axis of benzene, the complex had Cs symmetry, and the monomer geometries were fixed at their gas phase minima. For the CH 3 NHt · · · acetate complex, the distance between the N and the carboxylate carbon was varied, keeping all other parameters fixed. The solution phase minimum energy geometry was a "doublyhydrogen bonded" structure that placed the carboxylate oxygens nearly coplanar with the ammonium hydrogens as shown in Figure 4.1. Consideration of other geometries gave comparable results. For solution phase calculations, supermolecule calculations with the SM5.42R/HF /6-31 +G* method produced an "interaction energy," defined in Equation 4.1. To validate the performance of the method, we compared the calculated ~G(sol) for the monomers methylammonium, benzene and acetate to the experimentally de- 110 termined values in water. In all cases, these agree well with experiment. For acetate, !J.G(sol) calcd. = -77 kcal/ mol, l:J.G(sol) expt. = -77 kcal/ mol; methylammonium !J.G(sol) calcd. = -76 kcal/ mol , /:J.G(sol) expt. = -73 kcal/ mol; benzene /:J.G(sol) calcd. = -0.6 kcal/ mol, !J.G( sol) expt. = -0.9 kcal/ mol [23, 34]. Anilinium, a cation comparable in size to the cation-1r complex of Figure 4.1, is also well treated by the method (!J.G(sol) calcd. = -68 kcal/mol, !J.G(sol) expt. = -68 kcal/mol). Since the method treats all the individual molecules properly, as well as a larger molecule that contains the relevant functional groups, it is reasonable to expect good performance for the supermolecules of Figure 4.1. The largest discrepancy between experiment and theory involves methylammonium ion, although it should be noted that the error in experimental values for ions are often several kcal/mol. In this study, methylammonium is used in calculations for both the cation-1r interaction and the salt bridge. Thus, to the extent that this value may differ from experiment, it will effect both calculations in the same way. Tables 4.2 and 4.3 contain the results used to generate Figure 4.2. Inspection of !J.G (sol) for both complexes is quite revealing. For the salt bridge, !J.G(sol) is contin- uosly reduced as the partners approach one another. For the cation-1r interaction, !J.G(sol) remains relatively constant - consistent with the cationic nature of the inter- action. Calculations on protein structures were performed using the CaPTURE program (Cation-1r Trends Using Realistic Electrostatics) 3 , modified to use the GEPOL algorithm [30] to determine the solvent accessible surface area of individual amino acids and pairs. A solvent radius of 1.4 A was used for the probe and van der Waals radii were taken from the work of Rose et al [35]. 3 CaPTURE is described in Chapter 3. 111 distance (A) 2.75 3.0 3.125 3.25 3.5 3.75 4.0 4.25 4.5 5.0 5.5 6.0 7.0 8.5 10.0 CH3NHf CH3CO 2 E(gas) E(sol) 6.G( sol) (hartrees) - 323.0130812 -323.0165081 - 323.0134842 - 323.0089197 - 322.9977698 - 322.9861580 - 322.9752304 - 322.9653184 - 322.9564439 - 322.9414938 - 322.9295546 - 322.9198377 - 322.9049581 - 322.8896950 - 322. 8792790 - 95.57415921 - 227.2495112 (hartrees) - 323.0519352 - 323.0670636 - 323. 0700798 - 323.0714869 - 323.0718733 - 323.0716323 - 323.0715380 - 323.0711991 - 323. 0708662 - 323. 0704304 - 323.0701273 - 323.0699184 - 323. 0695550 - 323. 0692843 - 323.0691178 - 95.69577654 - 227.3725958 (kcal/mol) - 24.381 - 31.724 - 35.514 - 39.262 - 46.501 - 53.636 - 60.434 - 66.441 - 71.801 - 80.909 - 88.211 - 94.177 - 103.286 - 112.694 - 119.126 - 76.316 - 77.237 Table 4.2: E(gas), E( sol), and 6.G(sol), for the CH3NHf ···acetate complex calculated at the SM5.42R/HF /6-31 +G* level. 6.G(sol) is the energy difference between E(gas) and E(sol)· 112 distance E(gas) E(sol) 6.G(sol) (A) (hartrees) - 326. 2974629 - 326.3040451 - 326.3049767 - 326.3050329 - 326.3037308 - 326.3016819 - 326. 2995494 - 326.2975917 - 326.2958886 - 326.2932258 -326.2913382 - 326.2899857 - 326.2882708 - 326.2869585 - 326.2863071 - 326.2853837 - 230. 7110925 (hartrees) - 326.3963187 - 326.4076982 - 326.4108134 - 326.4130617 - 326.4158784 - 326.4165876 - 326.4162292 - 326.4154505 - 326.4145117 - 326.4128132 -326.4115623 - 326.4107064 -326.4096779 - 326.4088708 - 326.4084967 - 326.4079465 - 230.712035 (kcal/mol) - 62.033 - 65.043 - 66.414 - 67.789 - 70.374 - 72.104 - 73.218 - 73.958 - 74.437 - 75.042 - 75.442 - 75.753 - 76.184 - 76.501 - 76.675 - 76.909 - 0.591 2.75 3.0 3.125 3.25 3.5 3.75 4.0 4.25 4.5 5.0 5.5 6.0 7.0 8.5 10.0 20.0 C6H6 Table 4.3: E(gas), E(sol), and 6.G(sol), for the CH3NHt · · · benzene complex calculated at the SM5.42R/ HF / 6-31 +G* level. 6.G (sol) is the energy difference between E(gas) and E(sol ) · 113 4.5 Appendix 2 - Sample Gaussian Input Files The fo llowing are sample input files for Gaussian 94 for t he CH3 NHt ·· ·benzene and CH3 NHt · ·· acetate complexes. Calculations at different distances were performed by allowing the d variable to change. 4.5.1 CH 3 NHf · ··Benzene Z-matrix $RunGauss %Mem=32MB #p rhf/6-31+g* scf=(direct,tight , novaracc) pop=min test CH3NH3+ benzene 6-31+g* gsm d=3.75 1, 1 X X 1 1. 0 X 2 1.0 1 90.0 C 1 r1 2 90.0 3 0.0 C 1 r1 2 90.0 3 60.0 C 1 r1 2 90.0 3 120 . 0 C 1 r1 2 90 .0 3 180 . 0 C 1 r1 2 90 . 0 3 240 . 0 C 1 r1 2 90.0 3 300 . 0 X 4 1.0 1 90.0 2 0.0 X 5 1.0 1 90.0 2 0.0 X 6 1.0 1 90 . 0 2 0 . 0 X 7 1.0 1 90.0 2 0.0 X 8 1.0 1 90.0 2 0.0 X 9 1.0 1 90 . 0 2 0 . 0 H 4 r2 10 ach 2 180 . 0 H 5 r2 11 ach 2 180 . 0 H 6 r2 12 ach 2 180 . 0 H 7 r2 13 ach 2 180 . 0 H 8 r2 14 ach 2 180 . 0 H 9 r2 15 ach 2 180 . 0 N 1 d 4 90 . 0 3 0.0 C 22 rnc 1 ang 4 180.0 H 23 rch1 22 cha1 1 0.0 H 22 rnh1 23 nha1 24 180 . 0 H 23 rch2 22 cha2 25 chd H 23 rch2 22 cha2 25 -chd H 22 rhn2 23 nha2 24 nhd H 22 rhn2 23 nha2 24 -nhd 114 d=3.75 ang=129 . 95970693 rnc=l.50695973 rchl=l.07800176 cha1=108.14397235 rnh1=1. 01126514 nha1=111 . 63851485 rch2=1.07807961 cha2=108 . 13096365 chd=59.99795508 rhn2=1. 01126634 nha2=111.62792752 nhd=59.99832444 rl=l . 38809164 r2=1.07566132 ach=90. $CM2 ISCRF=2 ICDS=8 IAQU=1 ICMD=8 $END 4.5.2 CH3NHf ·: ·Acetate Z-matrix $RunGauss %Mem=32MB #p rhf/6-31+g* scf=(direct,tight) pop=min test methyl-ammonium/acetate 6-31+g* gsm d=3.75 0 1 N C 1 rcn H 2 rch1 1 rha1 H 1 rnh1 2 rna1 3 180.0 H 2 rch2 1 rha2 4 nhdih H 2 rch2 1 rha2 4 -nhdih H 1 rnh2 2 rna2 3 chdih H 1 rnh2 2 rna2 3 -chdih C 1 d 4 ang1 3 dih1 0 9 rco1 1 ang2 3 dih2 C 9 rec 10 ccoa1 3 dih3 H 11 arh1 10 arha1 9 0 . 0 H 11 arh2 10 arha2 9 rhd H 11 arh2 10 arha2 9 -rhd 0 9 rco2 11 occa1 12 0.0 ang1=54.53975247 115 ang2=64.41786955 d=3.75 dih1=222.92249489 dih2=39.98401842 dih3=170.87183184 rcn=i . 50715703 rch1=1.07809806 rch2=1.07809841 rnh1=1 . 01122464 rnh2=1. 01122615 rha1=108.14559 rha2=108.14503852 rna1=111. 6358302 rna2=111.63544777 nhdih=59.99984654 chdih=60.00004857 rco1=1.23873918 rco2=1.23715177 rcc=i.54738322 ccoa1=114.91243311 arh1=1.08530293 arha1=140.30124156 arh2=1.08795034 arha2=93.43710349 rhd=126.35484655 occa1=116.33222371 $CM2 ISCRF=2 ICDS=8 IAQU=1 ICMD=8 $END 116 Bibliography [1] A. Matouschek, J. T. Kellis Jr., L. Serrano, M. Bycroft, and A. R. Fersht. Transient folding intermediates characterized by protein engineering. Nature, 346:440- 445, 1990. [2] L. R. Brown, A. De Marco, R. Richarz, G. Wagner, and K. Wiithrich. The influence of a single salt bridge on static and dynamic features of the globular solution conformation of the basic pancreatic trypsin inhibitor. J. Biochem., 88:87-95, 1978. [3] S. Dao-pin, U. Sauer, H. Nicholson, and B. W. Matthews. Contributions of engineered surface salt bridges to the stability of T4 lysozyme determined by directed mutagenesis. Biochemistry, 30:7142- 7153, 1991. [4] A. Horovitz, L. Serrano, B. Avron, M. Bycroft, and A. R. Fersht. Strength and co-operativity of contributions of surface salt bridges to protein stability. J. M ol. Biol. , 216:1031- 1044, 1990. [5] H.-J. Schneider, T. Schiestel, and P. Zimmermann. The incremental approach to noncovalent interactions: Coulomb and van der Waals effects in organic ion pairs. J. Am. Chem. Soc. , 114:7698- 7703, 1992. [6] J. P. Schneider, J. D. Lear, and W. F. DeGrado. A designed buried salt bridge in a heterodimeric coiled coil. J. Am. Chem. Soc., 119:5742- 5743, 1997. [7] J. C. Ma and D. A. Dougherty. The cation-71" interaction. Chem. Rev., 97:13031324, 1997. [8] K. T. No, K-Y. Nam, and H. A. Scheraga. Stability of like and oppositely charged organic ion pairs in aqueous solution. J. Am. Chem. Soc., 119:12917- 12922, 1997. 117 [9] S. Saigal and J. Pranata. Monte Carlo simulations of guanidinium acetate and methylammonium acetate ion pairs in water. Bioorg. Chem., 25:11-21, 1997. [10] X. Barril, C. Aleman, M. Orozco, and F. J. Luque. Salt bridge interactions: Stability of the ionic and neutral complexes in the gas phase, in solution, and in proteins. Proteins, 32:67- 79, 1998. [11] Z. S. Hendsch and B. Tidor. Do salt bridges stabilize proteins? A continuum electrostatic analysis. Protein Sci., 3:211- 226, 1994. [12] B. Honig and A. Nicholls. Classical electrostatics in biology and chemistry. Science, 268: 1144- 1149, 1995. [13] A. H. Elcock. The stability of salt bridges at high temperatures: Implications for hyperthermophilic proteins. J. Mal. Biol., 284:489- 502, 1998. [14] R. Luo, L. David, H. Hung, J. Devaney, and M. K. Gilson. Strength of solventexposed salt-bridges. J. Phys. Chem. B, 103:727-736, 1999. [15] E. M. Duffy, P. J. Kowalczyk, and W. L. Jorgensen. Do denaturants interact with aromatic hydrocarbons in water? J. Am. Chem. Soc., 115:9271- 9275, 1993. [16] C. Chipot, B. Maigret, D. A. Pearlman, and P. A. Kollman. Molecular dynamics potential of mean force calculations: A study of the toluene-ammonium 1r-cation interactions. J. Am. Chem. Soc., 118:2998-3005, 1996. [17] R.A. Kumpf and D.A. Dougherty. A mechanism for ion selectivity in potassium channels: Computational studies of cation-1r interactions. Science, 261:17081710, 1993. [18] J. Gao, L. W. Chou, and A. Auerbach. The nature of cation-1r binding: Interactions between tetramethylammonium ion and benzene in aqueous solution. Biophys. J., 65:43-47, 1993. [19] M. Gaberscek and J. Mavri. Phenol forms complexes with tetramethylammonium ions in aqueous solution? Chem. Phys. Lett., 308(5-6):421- 427, 1999. 118 [20] M. A. L. Eriksson, P.-Y. Morgantini, and P. A. Kollman. Binding of organic cations to a cyclophane host as studied with molecular dynamics simulations and free energy calculations. J. Phys. Chem. B, 103:4474-4480, 1999. [21] J. W. Caldwell and P. A. Kollman. Cation-1r interactions: Nonadditive effects are critical in their accurate representation. J. Am. Chem. Soc., 117:4177- 4178, 1995. [22] E. Cubero, F. J. Luque, and M. Orozco. Is polarization important in cation-1r interactions? Proc. Natl. Acad. Sci. USA, 95:5976- 5980, 1998. [23] J. Li, G. D. Hawkins, C. J. Cramer, and D. G. Truhlar. Universal reaction field model based on ab initio Hartree-Fock theory. Chem. Phys. Lett., 288:293-298, 1998. [24] C. J. Cramer and D. G. Truhlar. Implicit salvation models: Equilibria, structure, spectra, and dynamics. Chem. Rev., 99:2161-2200, 1999. [25] J. Li, T. Zhu, C. J. Cramer, and D. G. Truhlar. New class IV charge model for extracting accurate partial charges from wave functions. J. Phys. Chem. A , 102:1820- 1831, 1998. [26] Y.-Y. Chuang, M. L. Radhakrishnan , P. L. Fast, J. Cramer, and D. G. Truhlar. Direct dynamics for free radical kinetics in solution: Solvent effect on the rate constant for the reaction of methanol with atomic hydrogen. J. Phys. Chem. A , 103:4893- 4909 , 1999. [27] Y-J. Zheng and R. L. Ornstein. What happens to salt-bridges in nonaqueous environments: Insights from quantum mechanics calculations. J. Am. Chem. Soc ., 118:11237- 11243, 1996. [28] C.A. Deakyne and M. Meot-Ner (Mautner). Unconventional ionic hydrogen bonds. 2. NH+ 1r. complexes of onium ions with olefins and benzene derivatives. J. Am. Chem. Soc., 107:474- 479, 1985. 119 [29] M. L. Lamb and W. L. Jorgensen. Computational approaches to molecular recognition. Curr. Opin. Chem. Biol., 1:449- 457, 1997. [30] E. Silla, I. Tunon, and J. L. Pascual-Ahuir. GEPOL: An improved description of molecular surfaces II. Computing the molecular area and volume. J. Comput. Chem., 12:1077- 1088, 1991. [31] B. Rost and C. Sander. Conservation and prediction of solvent accessibility in protein families. Proteins, 20:216- 226, 1994. [32] B.I. Dahiyat and S.L. Mayo. De novo protein design: Fully automated sequence selection. Science, 278(5335):82- 87, 1997. [33] M. J. Frisch, G. W. Trucks, H. B. Schlegel, P. M. W. Gill, B. G. Johnson, M.A. Robb, J. R. Cheeseman, T. A. Keith, G. A. Petersson, J . A. Montgomery, K. Raghavachari, M. A. Al-Laham, V. G. Zakrzewski, J. V. Ortiz, J. B. Foresman, C. Y. Peng, P. A. Ayala, M. W. Wong, J. L. Andres, E. S. Replogle, R. Gomperts, R. L. Martin, D. J. Fox, J. S. Binkley, D. J. Defrees, J. Baker, J. P. Stewart, M. Head-Gordon, C. Gonzalez, and J. A. Pople. Gaussian 94 (Revision D.3). Gaussian, Inc., 1995. [34] W.C. Still, A. Tempczyk, RC. Hawley, and T. Hendrickson. Semianalytical treatment of salvation for molecular mechanics and dynamics. J. Am. Chem. Soc., 112:6127- 6129, 1990. [35] G. D. Rose, A. R. Geselowitz, G. J. Lesser, R. H. Lee, and M. H. Zehfus. Hydrophobicity of amino acid residues in globular proteins. Science, 229:834- 838, 1985. 120 Chapter 5 Weak Electrostatic Interactions Involving Perfluorinated Aromatic Rings 121 5.1 Introduction In the preceding chapters, we have emphasized the ability of aromatic rings to bind to cations via cation-1r interactions. The implication has been that benzene and other aromatics act as "pseudo-anions ," and thus bind cations t hrough electrostatic interactions. If this idea is correct, one might predict t hat "pseudo-anions ," such as benzene, might also bind "pseudo-cations ," or molecules that display regions of positive electrostatic potential. Indeed this is the case. A prototype system is water· • • benzene. Water binds to benzene with a calculated -1.8 kcal/mol of binding energy in the geometry shown in Figure 5.1 [1]. H H H~H H H Figure 5.1: Experimentally determined geometry of water··· benzene [1]. Using standard arguments , this interaction is somewhat surprising. Water is a polar molecule, whereas benzene is generally considered to be non-polar. Furthermore, in solution, water and benzene do not mix to an appreciable extent. So why, then, are these molecules attracted to one another in the gas phase? The answer lies in the electrostatic properties of water and benzene. Although benzene does not have a dipole moment , benzene has a large, permanent quadrupole moment, 1 such that there is substantial negative electrostatic potential above and below the plane of the ring , and a belt of positive potential around the edge [2 , 3]. This can be easily seen from the electrostatic potential surface of benzene, shown in Figure 5.2. Ionic and polar species interact with this moment as anticipated from purely electrostatic arguments. Thus, inspection of the electrostatic potential surface of water, also shown in Figure 5.2, suggests that the hydrogens of water, with their region 1 A quadrupole moment arises from moment. an arrangement of dipoles such that there is no net dipole 122 Figure 5.2: Calculated HF /6-31G** electrostatic potential surfaces of benzene (left) and water (right). Scale: -25 - +25 kcaljmol. of positive electrostatic potential (shown in blue), should bind to benzene's region of negative electrostatic potential (shown in red). Indeed this is what is observed both theoretically and experimentally [1]. Thus, simple inspection of the electrostatic potential surfaces allows one to predict the likely binding geometry of the two molecules. 2 It has long been recognized that hexafluorobenzene has a quadrupole moment that is comparable in magnitude but opposite in sign to that of benzene. 3 This effect manifests itself in the electrostatic potential surface for hexafluorobenzene shown in Figure 5.3. Relative to the electrostatic potential surface of benzene, positive regions become negative and negative positive. Inspection of the electrostatic potential surfaces of benzene (Figure 5.2) and hexafluorobenzene (Figure 5.3) suggests that benzene and hexafluorobenzene should interact favorably by binding through a face to face interaction. This stacking is observed both computationally [4] and experimentally [3], and aryl-perfluoroaryl interactions are now being employed as "supramolecular syn2 This is not to imply that non-electrostatic effects are not important contributors to the binding - they certainly are. However, the geometries in these simple systems can often be predicted by consideration of electrostatics alone. 3 The quadrupole moments of benzene and hexafluorobenzene are -29.0x 10- 4 o C m 2 and + 31.7x 10- 4 C m 2 , respectively [3]. ° 123 Figure 5.3: Calculated HF /6-31G** electrostatic potential surface of hexafluorobenzene. Scale: -25 - +25 kcaljmol. thons" [5- 7]. 4 Until this point, all interactions discussed have found experimental support. Although the water··• benzene interaction may be somewhat surprising at first, consideration of the electrostatic potential surfaces of the partners explains the behavior of the system. Taking electrostatic reasoning further, one might predict that water should also bind to hexafluorobenzene, but in the geometry shown in Figure 5.4, with the water dipole moment reversed relative to the water· · · benzene interaction. F F F*F F F Figure 5.4: Predicted binding geometry of water· ·· hexafluorobenzene. While quite reasonable from an electrostatic viewpoint, from another perspective this interaction is somewhat counterintuitive. It suggests that a n-electron system should be attracted to a set of lone pair electrons. One might expect significant electron-electron repulsions that are not present for the water· · · benzene interaction. As such, the water· • • hexafluorobenzene interaction pushes the limits of electrostatic 4 Calculations describing two of these systems are presented in an appendix to this chapter. 124 arguments in rationalizing noncovalent binding interactions to aromatics. 5.2 Results 5.2.1 On-Axis Geometries To investigate whether water can bind to hexafluorobenzene, we performed ab initio molecular orbital calculations on this system. We first considered the "on-axis" complex shown in Figure 5.5. I :r Figure 5.5: On-axis geometry of the water··· hexafluorobenzene complex. Using Hartree-Fock theory and the 6-31G** basis set, we find that water binds to hexafluorobenzene with -2.49 kcal/mol of binding energy at an oxygen- centroid distance of 3.22 A. Although these calculations indicate an interaction between the partners, there are two potential problems. The first is that the binding energies of complexes determined from supermolecule calculations tend to be overestimated due to basis set superposition error (BSSE). The second is that Hartree-Fock theory tends to underestimate the binding energies of neutral complexes because the effects of electron correlation are ignored. The effects of BSSE can be minimized by performing counterpoise (CP) corrections according to the method of Bernardi and Boys [8]. Furthermore, the effects of electron correlation can be accounted for in a post-HartreeFock method, such as M0ller-Plesset (MP) perturbation theory. The results of these calculations are shown in Table 5.1. 125 theory HF /6-31G** CP-HF /6-31G** MP2(full) /6-31G** CP-MP2(full) /6-31G** B3LYP /6-31G** r (A) 3.22 3.50 2.94 3.20 3.08 - binding energy (kcal/mol) 2.49 1.67 3.90 2.11 2.43 Table 5.1: Results of ab initio calculations for the water· · · hexafluorobenzene complex shown in Figure 5.5. See Table 5.5 in Section 5.6 for the total energies. Calculations on noncovalent systems can be challenging. Our goal here is not to present the highest level of theory imaginable for this system. Rather, we employ a competent level of theory that has been shown to handle the water· · · benzene system well [1]. While more advanced levels of theory may change some quantitative aspects of our results, we are confident that no qualitative conclusions will change. The highest level of theory used here involved MP2/6-31G** calculations with counterpoise (CP) corrections for basis set superposition error (BSSE). Similar to a previous study of benzene· • • hexafluorobenzene [4], we find that BSSE strongly affects the interaction distance between water and the aromatic (but no other geometrical parameters), and so this distance has been optimized with the CP correction. This level of theory is equivalent to that applied previously to water· · · benzene [1], making direct comparisons meaningful. The global minimum for water··· hexafluorobenzene (Figure 5.5) is as anticipated from electrostatic arguments, and the binding energy (-2.1 kcal/mol) is quite comparable to that for water··· benzene (-1.8 kcal/mol). There appear to be no adverse electron-electron repulsions, consistent with an earlier study of polar molecules binding to hexafluorobenzene [9]. Not surprisingly, the potential energy surface is quite flat in this region, and the binding energy in the eclipsed geometry shown differs by less than 1 cal/mol from the 15° and 30° staggered geometries at the CP-MP2(full)/6-31G** level. The magnitude of the binding energy in this system is substantial - 2.1 kcal/mol. It is perhaps most surprising given that one might expect substantial electronelectron repulsions, even though electrostatics predicts that water and hexafluoroben- 126 zene should bind. To estimate the fraction of the binding energy associated with electrostatics, we make the assumption that the total binding energy of the complex (Etot) can be represented as the sum of the electrostatic contribution (Ees) and the non-electrostatic contributions (Enon- es)- Thus, the total energy for the water··· hexafluorobenzene system can be represented as in Equation 5.1. Unfortunately, one cannot directly calculate E es and Enon-es using ab initio calculations. One can only calculate Etot · However, recall that the quadrupole moment of benzene is nearly equal in magnitude to that of hexafluorobenzene, but opposite in sign. Thus, the electrostatic contribution of water binding to benzene in the geometry of Figure 5.5 should be equal and opposite to that of water binding to hexafluorobenzene in the same geometry. (5.1) (5.2) By making the assumption that E es (c6 p 6 ... H 2 0) Equations 5.1 and 5.2, we find that Furthermore, it is reasonable to make the assumption that to first order (5.4) because the closest contacts are between the oxygen and the carbon atoms in both systems. Thus, substituting Equation 5.4 into Equation 5.3 and rearranging, we find that (5.5) Thus, by determining Etot (C 6 H6 . .. H 2 o) using an ab initio calculation, we can estimate 127 Enon- es (C 6 F 6 ... H 2 0) from Equation 5.5 and E es (C 6 F 6 ... H 2 0) using Equation 5.1. We thus calculated Etot (C 6 H6 ... H 2 o) at the CP-MP2/6-31G** level. Surprisingly, Etot (C 6 H 6 ... H 2 0) = +2.1 kcal/ mol - equal and opposite to Etot (C 6 F 6 ... H 2 0)· This im- plies t hat Enon-es for both systems is zero, and that the water-·· hexafluorobenzene interaction is completely dominated by electrostatics. If this were true, one would predict that the energy of water binding to 1,3,5-trifluorobenzene (a molecule with a near-zero quadrupole moment, and thus limited potential for electrostatic interaction) in the geometry of Figure 5.5 would be 0. Indeed this is the case - the CP-MP2/6-31G** binding energy is +0.009 kcal/mol. Although t hese results are consistent , they are at first glance surprising. Chemical intuition might suggest that Enon-es, which consists largely of van der Waals forces, would be favorable in the water-·· hexafluorobenzene system. However, further consideration of the interaction suggests why this is not the case. If we consider two point • charges of opposite sign in the gas phase, Coulomb's law predicts that they will experience the maximum attraction at the minimum distance. Using similar reasoning , water and hexafluorobenzene are attracted to one another , and become more so at smaller distances. Of course at small distances, the molecules will repel, because other non-electrostatic forces, including van der Waals forces, will become repulsive. This is illustrated in Figure 5.6 which shows three potential energy profiles. The red curve is the potential energy profile for an electrostatic interaction between oppositely charged partners ( - ~ dependence). The potential energy approaches O as r approaches oo, and it approaches -oo as r approaches 0. In this system, the electrostatic interaction is always favorable. The blue curve represents the potential energy profile for the Lennard-Jones "12-6" potential which is commonly used to represent van der Waals forces ( ) 2 - / 6 dependence). At very short distances, this force is quite repulsive. At a somewhat longer distance, there is a minimum, and as r approaches oo, the red curve approaches 0. The purple curve is the sum of the red and blue curves. It , too, has a minimum - one that is shifted to shorter distances than that of the minimum of the blue curve. Clearly, at short distances, the repulsion from the "12- 6" curve dominates the potential energy curve. However , because the electrostatics of the system 128 ro ~ CJ ~ i:.i:I ....--4 .m ...... ~ ~ ---, r CJ ...... 0 ~ Figure 5.6: Schematic potential energy curves. Red: electrostatic interaction between oppositely charged partners(-¼ dependence); blue: Lennard- Jones potential (ri 2 - r~ dependence); purple: sum of the red and blue curves. The vertical dotted lines correspond to the minima of the two curves. The horizontal dotted line shows that at the true minimum, the Lennard- Jones potential (blue curve) is not at a minimum. are always favorable, the minimum of the purple curve is found at a shorter distance than the minimum of the "12- 6" curve. In fact , at the distance of the true minimum (represented by a purple dot), the blue curve is nearly in positive territory. This is similar to the situation observed in our calculations. Depending on the relative magnitudes of the two curves, one can envision situations where the true minimum of the purple curve corresponds to a location of the blue curve where the potential energy is 2: 0. Similarly, in cases where the red curve is flat (such as the water·· · 1,3,5trifluorobenzene interaction), the two minima will correspond. Indeed, water binds to 1,3,5-trifluorobenzene with -0.3 kcaljmol of binding energy at a distance of 3.7 A - greater than the 3.2 A seen in the water··· hexafluorobenzene interaction. If we consider the 0.3 kcal/mol of binding energy to be equal to the non-electrostatic component of the water· · · hexafluorobenzene interaction, it suggests that electrostatics contribute 86% of the binding energy in the water· · · hexafluorobenzene system. This should serve as a cautionary tale to those who attempt to decompose binding energies into their components - for interactions dominated by favorable electrostatics, van der Waals interactions serve to keep the molecules apart, and thus may not always be favorable. 129 5.2.2 Off-Axis Geometries In addition to the on-axis geometries discussed in the previous section, we also considered other geometries for the water· · · hexafluorobenzene complex. Optimization starting from an electrostatically unfavorable geometry with an O- H bond located on the C6 axis and the hydrogen pointed toward the face of the ring led to a second minimum shown in Figure 5.7, in which the water has moved well off the C6 axis. This structure is calculated to be only slightly less stable than the on-axis geometry Figure 5. 7: Off-axis geometry of the water· · · hexafluorobenzene complex. shown in Figure 5.5 (-1.7 vs. -2.1 kcal/mol at the CP-MP2/6-31G** level). Results at several levels of theory are presented in Table 5.2. Rather than a symmetrical theory HF/6-31G** MP2(full) /6-31G** CP-MP2(full) /6-31G** B3LYP /6-31G** r (A) 3.96 3.77 4.05 3.95 0 </> 41.8 42.5 45.0 39.6 36.9 48.5 14.9 60.7 H- · -F (A) 2.61 2.33 2.93 2.30 - binding energy (kcal/mol) 2.95 5.32 1.69 4.32 Table 5.2: Results of ab initio calculations for the complex shown in Figure 5. 7. See Figure 5.7 for definitions of r, 0 and ¢. See Table 5.6 in Section 5.6 for the total energies. alignment of molecular moments (dipole/quadrupole), this geometry represents a favorable alignment of local O- H and C- F bond dipoles, and so presumably still has a substantial electrostatic component. To determine whether an analogous minimum exists for water·· · benzene, calculations were performed at the HF /6-31G** level. These efforts were not successful - all starting geometries led to structures similar to that previously reported for water· · · benzene. 130 Given the increasing interest in DFT methods, we also evaluated both the on-axis and off-axis structures using the Becke3LYP density functional and the 6-31G** basis set. Since no counterpoise corrections were applied, the most appropriate comparison is to the MP2 results. In an earlier study of the benzene· • • hexafluorobenzene interaction [4], DFT performed poorly, being comparable to simple Hartree-Fock calculations. However, for water··· hexafluorobenzene DFT produces good results for the global minimum, producing values that are in acceptable agreement with the MP2 results and are actually quite close to the full CP-MP2 results (Table 5.1). While CP-MP2 predicts a significant preference for the on-axis geometry, DFT finds the offaxis structure to be more stable. Both the DFT and the MP2 structures have a much larger value of ¢, the tilt angle of the water molecule, than the CP-MP2 structure. This tilting brings the water hydrogens very close to the fluorines , well inside van der Waals contact, indicative of a hydrogen bond. Dunitz has concluded that hydrogen bonds to fluorine are not energetically significant [10], but it appears that both DFT and MP2 consider such an interaction to be important. It is clear that BSSE will be most important when noncovalent contacts to fluorine are involved, because of the large demands fluorine puts on its basis set. Thus, it is perhaps not surprising that the CP corrections are most significant for the off-axis structure. 5 For these reasons we consider the CP-MP2 level to be the best for comparisons of the two structures. The results for the off-axis geometry suggest that a fully in-plane structure such as that shown in Figure 5.8 might be favorable. However, at the HF /6-31G** level such a structure is in fact a transition state separating two equivalent forms of the structure shown in Figure 5. 7. 5.3 Summary These results further illustrate the power of electrostatic reasoning in studies of noncovalent interactions involving aromatics. In the absence of such effects, one would never predict a stabilizing interaction for a geometry that points oxygen lone pairs 5 There are no close contacts to fluorine in the on-axis structure. 131 F FY)'F F~F H 'o H / F Figure 5.8: In-plane geometry of water··· hexafluorobenzene. HF /6-31G** calculations indicate that this is a transition state structure, separating two equivalent forms of the structure shown in Figure 5.7 directly into the face of a 1r system as observed here. As always, other effects related to polarizability and dispersion interactions certainly contribute to the binding energy, but all the structures discussed here can be accurately anticipated on the basis of electrostatics alone. These results further support the idea first put forth by Reisse [2] that aromatics such as benzene and hexafluorobenzene are correctly thought of as polar molecules, and clearly underscore the importance of electrostatic reasoning in evaluating novel intermolecular interactions. 6 5.4 Computational Methods All calculations were performed using Gaussian 94 Rev. D3 [12]. Geometry optimizations were first performed at the HF level using the 6-31G** basis set. Frequency calculations performed on fully optimized structures at the HF level produced no negative frequencies (except for the in-plane structure), indicating that the geometries were true minima. Using the Hartree-Fock frequency calculations, we estimate the zero-point energy (ZPE) correction to be +0.58 kcal/mol for the structure in Figure 5.5 and +0.89 kcal/ mol for the structure in Figure 5. 7 using the frequency scaling factor of 0.8992 suggested by Scott and Radom [13]. These results are not included in Tables 5.1 and 5.2. Counterpoise corrected optimizations (designated CP) were performed using the method of Bernardi and Boys [8]. First , the component molecules, water and hexafluorobenzene, were individually optimized at either the HF / 6-31G** level or the MP2 (full)/ 6-31 G ** level (the full indicates the "frozen-core" method was 6 Recently, Danten et al. reported a study containing similar findings to those described here [11]. 132 not used in the MP2 calculations - all electrons were considered). The internal coordinates of the optimized monomers were kept fixed and the complexes were optimized by scanning the potential energy surface at several points. CP corrections were performed at each point, and this corrected binding energy was minimized. The results of these calculations are designated CP-HF / 6-31G** or CP-MP2(full)/6-31G**. The on-axis complex had C 2v symmetry, whereas the off-axis complex had Cs symmetry. Tables detailing the geometries and total energies of several of the complexes can be found in Section 5.6. 5.5 Appendix 1 - Studies of the Supramolecular Synthon Triphenylene· · · Perfluorotriphenylene In recent years, the field of crystal engineering has seen a flourish of interest. The ability to rationally engineer solid-state structures poses great challenges. In particular, a thorough understanding of noncovalent interactions is necessary to enable one to design and manipulate solid-state architectures. Desiraju has advocated the "supramolecular synthon" approach, in which the goal of crystal engineering is to "recognize and design synthons that are robust enough to be exchanged from one network structure to another, which ensures generality and predictability " [14]. One such supramolecular synthon is the phenyl-perfluorophenyl system, in which the electrostatically favorable face-to-face stacking of fluorinated and non-fluorinated aromatic rings is employed to direct crystal packing. This system has been studied extensively, both theoretically [4] and experimentally [3, 5, 6, 15]. Recently, Grubbs and coworkers have extended the phenyl-perfluorophenyl supramolecular synthon to include fluorinated and non-fluorinated polycyclic aromatics [7]. To further explore such interactions, we performed ab initio calculations to investigate the energetics of the interaction between triphenylene and perfluorotriphenylene, and to compare them to the well-studied benzene··· hexafluorobenzene interaction. We first calculated the electrostatic potential surfaces of both triphenylene and perflu- 133 Figure 5.9: Calculated HF /6-31G** electrostatic potential surfaces of triphenylene (left) and perfluorotriphenylene (right). Scale: -25 - +25 kcal/mol. orotriphenylene at the HF /6-31G** level (Figure 5.9). For triphenylene, we calculated the geometry by optimizing the structure at the HF /6-31G** level in D 3 h symmetry. For perfluorotriphenylene, we used the coordinates from the crystal structure determined by Alex Dunn [16]. Several features are evident in Figure 5.9. The first is that qualitatively, the electrostatic potential surfaces for the fused ring systems are quite similar to those of the parent compounds shown in Figures 5.2 and 5.3. In addition, the propeller twist of perfluorotriphenylene is quite pronounced. To calculate the binding energy of the triphenylene· · · perfluorotriphenylene complex, we used the coordinates from Weck et al. [7]. Because the locations of the hydrogens were not determined in this structure, they needed to be modeled. This was done by taking the coordinates for the triphenylene ring by itself, adding the hydrogens, and optimizing their positions at the HF /6-31G** level while keeping the coordinates of the carbon atoms fixed. We then evaluated the energetics of various aspects of the complex as outlined in Figure 5.10. We first calculated the "distortion energy" of the partners in the interaction at the HF /6-31G** , B3LYP /6-31G**, and MP2/6-31G** levels. This measures the energetic cost of twisting the triphenylene and perfluorotriphenylene rings from their ground state structures to their structures in the complex. At all three levels of theory, the distortion cost is similar, and the combined distortion 134 Distortion Binding +2.3 kcal/mol +5.1 kcal/mol +3.8 kcal/ mol +4.0 kcal/mol +2.8 kca l/mo] +3.1 kcal/mol Total +5.8 (+12.1) kcal/mol +5.7 (+13.3) kca l/ mot -20.7 (-9.4) kcal/mol = Triphenylene = Perfluorotriphenylene t:.E HF/6-31G** t:.E B3LYP /6-31G** t:.E MP2/6-31G** Figure 5.10: Calculated binding energies of the triphenylene· · · perfluorotriphenylene complex. Numbers in parentheses are corrected for basis set superposition error. See text for a full description. energies for both partners range from + 6.3 kcal/mol to + 7.9 kcal/mol, depending on the method applied. We next calculated the binding energies of the complex at several levels of theory. This was done by calculating the total energy of the complex, and subtracting from it the total energies of the individual monomers calculated in the geometry of the complex. We first used simple supermolecule calculations and did not correct the binding energies for basis set superposition error (BSSE). The non-corrected binding energies are -0.5 kcal/mol at the HF /6-31G** level, -2.1 kcal/mol at the B3LYP /6-31G** level, and -27.6 kcal/mol at the MP2/6-31G** level. Clearly there are large differences between these levels of theory. Hartree-Fock theory and DFT both suggest very small binding energies, whereas the MP2 value is quite large. This trend is consistent with an earlier study of the benzene··· hexafluorobenzene complex [4]. When we performed counterpoise corrections to account for BSSE, the binding energies dropped significantly (these numbers are shown in parentheses in Figure 5.10). In fact, when the CP correction is included, both Hartree-Fock theory and DFT consider this interaction 135 to be repulsive. Finally, the total energies, including both distortion and binding are shown at the bottom of Figure 5.10. Including the effects of distortion, both Hartree-Fock theory and DFT consider the pairwise interaction to be strongly repulsive. The CP-MP2 value (-9.4 kcal/mol) seems more reasonable, especially compared to the -3. 7 kcal/mol of binding energy calculated for the benzene- · • hexafluorobenzene complex. It is quite surprising to see the wide range of binding energies at the different levels of theory. In the benzene- · · hexafluorobenzene study, it was noted that the equilibrium distance for the partners at the MP2 level was much shorter than that at the HF and DFT levels. Perhaps, then, the distance observed in the crystal structure is more representative of the distance at the MP2 mininum than the minima at the other levels. Unfortunately, due to the size of the system, we cannot easily calculate the equilibrium binding distance at the different levels of theory. However, we can attempt analogous calculations on the benzene-·· hexafluorobenzene complex to determine if similar effects are seen when using the coordinates from the crystal structure [15]. In the previous computational study, benzene and hexafluorobenzene were constrained to a C6 symmetry axis [4]. In the crystal structure, however, the rings are not perfectly stacked, but slipped. We thus calculated the binding energies for this system as shown in Figure 5 .11. In this case, the ground state structures are fully geometry optimized structures calculated at each level of theory. As evidenced by their small distortion energies, the DFT and MP2 monomer geometries are very similar to those observed in the crystal structure, whereas the Hartree-Fock monomer structures pay an energetic penalty of +6.3 kcal/mol to distort to the crystal structure geometry. In terms of binding energy, the trend mimics that seen in the triphenylene- · · perfluorotriphenylene case. MP2 theory suggests a strong binding energy (-7.2 kcal/mol) for the interaction, whereas both Hartree-Fock theory and DFT report much weaker interactions (-0.2 kcal/mol and -2.1 kcal/mol, respectively). Including the effects of the counterpoise corrections weakens the binding energies further, to +2.0 kcal/mol (HF), +1.7 kcal/mol (DFT), and -3.5 kcal/mol (MP2). Interestingly, the counterpoise-corrected MP2 energy is similar in magnitude to that observed for the C6 symmetric complex (-3.7 kcal/mol), while the geometries 136 Distortion Binding +2.8 kcal/mol +0.6 kcal/ mo! +1.1 kcal/mol -0.2 (+2.0) kcal/mo! -2.1 (+1.7) kcal/mo! -7.2 (-3.5) kcal/mol +3.7 kcal/mol +0.8 kcal/mol +0.5 kcal/mol • Total +6.3 (+8.5) kcal/mo! -0.7 (+3.1) kcal/mo) -5.6 (-1.9) kcal/ mol = Benzene = Hexafluorobenzene t!.E HF/6-31G** t.E B3LYP / 6-31 G ** t!.E MP2/6-31G** Figure 5.11 : Calculated binding energies of the benzene··· hexafluorobenzene complex. Numbers in parentheses are corrected for basis set superposit ion error. See text for a full description. are quite different. This may suggest that· the true CP-MP2 minimum corresponds to a slipped geometry, as observed in the benzene· · · hexafluorobenzene co-crystal. This would have implications for the use of this motif as a supramolecular synthon. 7 Further calculations may shed light on this. Sadly, the results in this appendix probably serve as a greater testament to the weaknesses of computational methods in describing noncovalent interactions, than to our understanding of them. Nonetheless, several trends emerge from these results. The first is that Hartree-Fock theory performs poorly in describing t he weak interactions between neutral molecules. This is not at all surprising given that HF theory ignores electron correlation. Perhaps somewhat more surprising is that density functional theory performs nearly as poorly as Hartree-Fock theory, despite its greater computational cost. M0ller-Plesset theory seems to perform better than both HF theory and DFT. At the very least, it considers the two interactions described here as attractive. In addition, the magnitudes of the interactions seem "sensible," but only 7 Many crystal structures of mixtures of fluorinated and non-fluorinated aryl rings exhibit the slipped-stacked geometry. See [17] for several examples. 137 experiment will confirm whether or not this is the case. Unfortunately, we are left with the dilemma that our "best" level of theory may not even accurately predict the results of an experiment, and it is prohibitively expensive computationally. 8 So where does this leave us? The major theme of this thesis is electrostatic interactions. Fortunately, electrostatics provide a predictive tool in these systems. Both of these interactions could be predicted solely by inspection of the electrostatic potential surfaces of the partners, just as in the cases of water··· benzene and water· · · hexafluorobenzene. As such, although these results may reveal the weaknesses of comparatively highlevel theoretical models, they undoubtedly emphasize the strengths of "electrostatic reasoning" in evaluating noncovalent interactions. 8 The CP-MP2 binding energy for the triphenylene· • • perfluorotriphenylene interaction required over 45 CPU-days to calculate using a Silicon Graphics 195 MHz RlO000 processor. 138 5.6 Appendix 2 Total Energies and Geometric Parameters Theory HF /6-31G** MP2(full) /6-31G** B3LYP / 6-31G** O- H bond length (A) 0.943 0.961 0.965 H- 0 - H angle (degrees) 105.97 103.88 103.74 Total energy (hartrees) - 76.0236150 - 76.2224487 - 76.4197397 Table 5.3: Geometries and total energies for H2 O at various levels of theory. Theory HF/ 6-31G** MP2(full) /6-31G** B3LYP /6-31G** C- C bond length (A) 1.378 1.391 1.394 C-F bond length (A) 1.314 1.340 1.336 Total energy (hartrees) - 823. 7535727 - 825.5613207 -827. 5969442 Table 5.4: Geometries and total energies for C6 F 6 at various levels of theory. Theory HF/ 6-31G** MP2(full) / 6-31G** CP-MP2(full) /6-31G** B3LYP /6-31G** Total energy -899. 7811587 - 901. 7899813 - 901.7893611 - 904.0205567 Table 5.5: Total energies (in hartrees) for the on-axis complex. For geometries, see Table 5.1. The energies of the monomers can be found in Tables 5.3 and 5.4. The counterpoise corrected monomer energies (in hartrees) at the CP-MP2(full)/6-31G** level are - 76.2240342 and - 825.5619603 for water and hexafluorob enzene, respectively. 139 Theory HF/6-31G** MP2(full) / 6-31G** CP-MP2(full) /6-31G** B3LYP /6-31G* * Total energy -899. 7818682 - 901. 792246 -901.7890015 - 904.023565 Table 5.6: Total energies (in hartrees) for the off-axis complex. For geometries, see Table 5.2. The energies of the monomers can be found in Tables 5.3 and 5.4. The counterpoise corrected monomer energies (in hartrees) at the CP-MP2 (full) / 631 G ** level are - 76.22406769 and -825.5623169 for water and hexafluorobenzene , respectively. Triphenylene Theory Ground state Complex HF /6-31G** - 688.681883 - 688.6780185 MP2/6-31G** - 691.1236012 - 691.1175665 B3LYP /6-31G* * - 693.2001571 - 693.1921083 Perfl uorotriphenylene Ground state Complex - 1874. 735386 - 1874. 7290964 - 1879.1682995 - 1879.163388 - 1883.8631578 - 1883.8586723 Table 5.7: Total energies (in hartrees) of triphenylene and perfluorotriphenylene. For triphenylene, the ground state corresponds to a calculated minimum energy structure at the given level of theory. For perfluorotriphenylene, the ground state structure is taken from the crystal structure coordinates [16] and the energy is a single-point energy calculated at the given level of theory. The structures in the complex are taken from the triphenylene· · · perfluorotriphenylene co-crystal [7]. The energies are single-point energies, calculated at the given level of theory. The difference in energies between the ground state and the complex is the distortion energy reported in Figure 5.10. Theory HF / 6-31G* * MP2/6-31G** B3LYP / 6-31G** Complex - 2563.4078617 - 2570.3249285 - 2577.0541744 Counterpoise Corrected Tri phenylene Perfl uorotri phenylene - 688.6825751 - 1874.7345637 - 1879.1734479 - 691.1255296 - 693.1966762 - 1883.8660902 Table 5.8: Total energies (in hartrees) for the triphenylene· · · perfluorotriphenylene complex and for the counterpoise corrected monomers. The energies for the counterpoise corrected monomers are determined by calculating a single-point energy for the monomer in the presence of the basis functions of the other monomer , in the geometry of the complex. The differences in these values are the counterpoise corrected energies shown in parentheses in Figure 5.10. 140 Bibliography [1] S. Suzuki, P.G. Green, R.E. Bumgarner, S. Dasgupta, W.A. Goddard III, and G.A. Blake. Benzene forms hydrogen bonds with water. Science, 257:942-945, 1992. [2] M. Luhmer, K. Bartik, A. Dejaegere, P. Bovy, and J. Reisse. The importance of quadrupolar interactions in molecular recognition processes involving a phenyl group. Bull. Soc. Chim. Fr., 131:603-606, 1994. [3] J. H. Williams. The molecular electric quadrupole moment and solid-state architecture. Acc. Chem. Res., 26:593- 598, 1993. [4] A. P. West Jr., S. Mecozzi, and D. A. Dougherty. Theoretical studies of the supramolecular synthon benzene··· hexafluorobenzene. J. Phys. Org. Chem., 10:347- 350, 1997. [5] G. W. Coates, A. R. Dunn, L. M. Henling, D. A. Dougherty, and R. H. Grubbs. Phenyl-perfluorophenyl stacking interactions: A new strategy for supramolecular construction. Angew. Chem. Int. Ed. Engl., 36:248- 251, 1997. [6] G.W. Coates, A.R. Dunn, L.M. Henling, J.W. Ziller, E.B. Labkovsky, and R.H. Grubbs. Phenyl-perfluorophenyl stacking interactions: Topochemical[2+2] photodimerization and photopolymerization of olefinic compounds. J. Am. Chem. Soc., 120(15):3641- 3649, 1998. [7] M. Weck, A.R. Dunn, K. Matsumoto, G.W. Coates, E.B. Labkovsky, and R.H. Grubbs. Influence of perfluoroarene-arene interactions on the phase behavior of liquid crystalline and polymeric materials. Angew. Chem. Int. Ed. Engl., 38(18):2741- 2745, 1999. [8] S. F. Boys and F. Bernardi. Mol. Phys., 19:533- 566, 1970. 141 [9] I. Alkorta, I. Rozas, and J. Elguero. An attractive interaction between the 1rcloud of C6 F 6 and electron-donor atoms. J. Org. Chem., 62:4687- 4691, 1997. [10] J. D. Dunitz and R. Taylor. Organic fluorine hardly ever accepts hydrogen bonds. Chem. Eur. J. , 3:89- 98, 1997. [11] Y. Danten, T. Tassaing, and M. Besnard. On the nature of the water- hexafluorobenzene interaction. J. Phys. Chem. A, 103(18):3530- 3534, 1999. [12] M. J. Frisch, G. W. Trucks, H. B. Schlegel, P. M. W. Gill, B. G. Johnson, M. A. Robb, J. R. Cheeseman, T. A. Keith, G. A. Petersson, J. A. Montgomery, K. Raghavachari, M. A. Al-Laham, V. G. Zakrzewski, J. V. Ortiz, J. B. Foresman, C. Y. Peng, P. A. Ayala, M. W. Wong, J. L. Andres, E. 8. Replogle , R. Gomperts, R. L. Martin, D. J. Fox, J. 8. Binkley, D. J. Defrees, J. Baker, J. P. Stewart, M. Head-Gordon, C. Gonzalez, and J. A. Pople. Gaussian 94 (Revision D.3). Gaussian, Inc., 1995. [13] A. P. Scott and L. Radom. Harmonic vibrational frequencies: An evaluation of Hartree-Fock, M0ller-Plesset, Quadratic Configuration Interaction, Density Functional Theory, and Semiempirical Scale Factors. J. Phys. Chem. , 100:1650216513, 1996. [14] G. R. Desiraju. Supramolecular synthons in crystal engineering - A new organic synthesis. Angew. Chem. Int. Ed. Engl. , 34:2311- 2327, 1995. [15] J. H. Williams, J. K. Cockcroft, and A. N. Fitch. Structure of the lowest temperature phase of the solid benzene-hexafluorobenzene adduct. Angew. Chem. Int. Ed. Engl., 31:1655- 1657, 1992. [16] A. R. Dunn and R. H. Grubbs. Unpublished results. 1999. [17] 8. A. Miller. Magnetic Interactions in Molecular Materials. Ph.D. Thesis, California Institute of Technology, 1999. 142 Chapter 6 Site-Specific Incorporation of Biotinylated Amino Acids to Identify Surface-Exposed Residues in Integral Membrane Proteins 143 6.1 Introduction In this chapter, we return our attention to the nicotinic acetylcholine receptor, and outline our attempts to develop a general method to determine both the transmembrane topology of the receptor, as well as the surface accessibility of individual residues. Determining the transmembrane topology of integral membrane proteins such as ion channels and neurotransmitter transporters is a necessary first step in evaluating their structure and function at the molecular level [l]. After topology has been established, a more demanding question is the relative exposure of individual residues to intracellular or extracellular media. Knowledge of transmembrane topology and exposure allows one to identify regions of a protein that may become glycosylated or form binding sites for external ligands, to identify intracellular regions that may be subject to phosphorylation, to assign potential binding sites for other intracellular secondary messengers such as G proteins, and to make predictions regarding the overall structure and function of an integral membrane protein. In this chapter, we address the "higher resolution" issue of surface exposure. We wish to determine which residues are clearly on the "outside" surface of a receptor. Of course, if a residue is determined to be exposed, the topology question is answered. Residue accessibility is clearly a related measure of protein structure, but, as discussed below, addresses a somewhat different issue. We consider here a relatively well understood membrane protein, the nicotinic acetylcholine receptor. 1 There is general agreement concerning the overall topology of this receptor, although final proof will require high resolution structural data. Each subunit is thought to have a large, extracellular N-terminal domain, followed by four transmembrane regions and a necessarily extracellular C-terminus (Figure 6.lA). However, little is known about the relative exposures of residues throughout the receptor. Many other membrane channels, transporters, and receptors have less well understood topologies. The first step (and, unfortunately, often the last step) in determining the topology is the generation of a hydropathy plot based on the protein 1 For an introduction to the nicotinic acetylcholine receptor, please see Chapter 2. 144 A.) NH 2 COOH B)~ Out COOH Figure 6.1: A.) Generally accepted topology of each subunit of the nicotinic acetylcholine receptor. This is also the original topology model (1989- 1995) proposed for the glutamate receptor, GluR3 [2]. B.) The revised topology for the glutamate receptor [3, 4] . sequence. Although hydropathy plots are useful in suggesting a possible topology for a membrane protein, they are by no means an effective substitute for experimental evidence. For instance, the primary amino acid sequence for the ionotropic glutamate receptor was first determined in 1989 and led to a topological model analogous to that for the nicotinic receptor - four transmembrane domains and both the N and C termini located extracellular [2]. Subsequent work showed that the putative second transmembrane domain most likely forms a reentrant loop, so that the N and C termini are on the opposite sides of the membrane (Figure 6.lB) [3, 4]. Such a drastic change certainly alters ones view of the structure and mode of action of the receptor. Other recent experiments on a variety of ion channels and transporters, including the potassium channel, ROMKl [5]; the GABA transporter, GAT-1 [6]; the glycine transporter, GLYTl [7]; and the Na+ /glucose cotransporter, SGLTl [8] have suggested inconsistencies in the transmembrane topologies predicted from hydropathy analysis. These studies underscore the importance of having definitive experimental approaches to determine transmembrane topology of membrane proteins. Although several methods are available, a new approach would be welcome if it overcomes 145 some of the inherent weaknesses of the existing methods. 6.2 Previous Approaches to Determining Topology and Accessibility A wide variety of techniques have been developed to address the topology issue [l]. We will briefly describe below some of the higher precision approaches that target specific regions of integral membrane proteins. 6.2.1 Endogenous Glycosylation Sites and N-Glycosylation Scanning Mutagenesis. One of the most common methods to determine transmembrane topology is to assay a particular region of a protein for endogenous N-linked glycosylation. Because N-linked glycosylation occurs, with few exceptions [9], on the luminal side of the endoplasmic reticulum, only regions destined to become extracellular are glycosylated. Thus, identifying a site that is glycosylated in the wild-type protein is good evidence that it is located extracellularly. Expanding upon this tactic, one can engineer Nlinked glycosylation sites into transmembrane proteins to determine topology. This approach has been used recently to assay the transmembrane topologies of many ion channels and transporters, including the GABA transporter, GAT-1 [6]; the glycine transporter, GLYTl [7]; CFTR [10]; and the Na+ /glucose cotransporter, SGLTl [8]. In all of these studies, the functional competence of the glycosylated mutants was often quite poor compared with that of the wild-type protein, and in many cases the modified proteins did not function at all. This is of particular concern because the introduction of a non-native glycosylation site may alter topology [4]. If a particular protein is found to be glycosylated in the experiment but does not function , it is often difficult to determine whether the protein fails to function because glycosylation interferes with the normal function of the protein, or because glycosylation has caused the protein to fold with an altered topology. 146 6.2.2 Cysteine Accessibility /SCAM Another common method for determining transmembrane topology is the introduction of cysteine residues into the protein of interest, followed by labeling with thiolreactive reagents. If the reagent is membrane impermeant, topology can be evaluated by comparing the results of intracellular vs. extracellular application [11, 12]. For example, cysteines have been modified with biotin reagents , and the reaction detected by streptavidin binding [13, 14]. The cysteine modification approach has an advantage over glycosylation methods in that a single cysteine mutant is often less perturbing to the overall protein structure than the introduction of a glycosylation site. The method has been generalized by Karlin to SCAM (substituted-cysteine accessibility method), allowing the "accessibility" of specific residues to be probed with relatively smaller thiol-reactive reagents [15 , 16]. Note that such accessibility is quite different from "exposure" as used here. SCAM has labeled residues that nominally lie in a transmembrane region, but that are nonetheless accessible because they lie along an ion channel or within an agonist binding site. It is in this regard that we distinguish accessibility vs. surface exposure. Potential disadvantages of these methods arise from the contributions of endogenous cysteine residues that may also be labeled in the assay. In addition, the introduction of new cysteines may lead to improper disulfide formation and incorrectly folded proteins. A possible solution to these problems is to mutate all of the endogenous cysteines to another amino acid, and to then introduce a unique cysteine. This has been successful in some proteins, but there are certainly cases where removal of all the endogenous cysteines will render the protein nonfunctional. Furthermore, for an in vivo assay, there may be complications arising from reactive cysteines found in endogenous cellular proteins that cannot be mutated away. 6.2.3 Epitope Protection In this assay, a fusion is created between the protein of interest and a well defined epitope [4, 17, 18]. The proteins are synthesized in vitro in the presence of microsomal 147 vesicles so that the expressed protein is inserted into the membrane. Digestion of the translation mixture with proteases cleaves extramicrosomal (cytoplasmic) regions, destroying the epitope if it is displayed in this region. If the epitope is located inside the vesicle, it is protected from protease digestion and can be immunoprecipitated or detected in a Western blot. There are several concerns raised by this assay. The assay is performed in a cell-free system; therefore the topology may not be identical to that produced in an intact cell [18]. The second concern is that the epitope introduced is often quite large (the prolactin epitope is 195 amino acids!) and consequently must often be introduced as a C-terminal fusion to a truncated and often non-functional protein [4]. Finally, there is always a concern that the introduction of a large epitope may cause the protein of interest to fold improperly. This potential problem, coupled with the fact that the proteins often cannot be assayed for function, makes the assay less attractive. 6.2.4 Tyrosine Iodination Hucho and co-workers have used [125 1]-labeling to probe surface exposure of tyrosine residues of the nicotinic receptor harvested from Torpedo californica [19]. This method has some attractive features, but, since all exposed tyrosines are labeled simultaneously, one must sequence the labeled protein to determine exposure. This is not feasible with proteins for which there is not a naturally occurring plentiful source. 6.3 The Biocytin Approach To address the potential shortcomings of the present methods, we sought to develop a new experimental approach to discover surface exposed residues. Necessarily, such information will bear strongly on issues of transmembrane topology and accessibility. Our design goals were as follows: • To determine the surface exposure/topology of a protein in vivo • To assay functional proteins with wild-type behavior 148 • To introduce a single, minimally perturbing mutation • To use a marker that is endogenous neither to the protein of interest nor to other cell surface proteins • To detect the marker using robust and readily available methods • To produce a graded series of reagents that allows conclusions about relative exposures As a first step, we have combined two commonly used techniques: the biotinstreptavidin interaction, and the in vivo nonsense suppression method. Biotin is a small molecule that binds with high affinity (Kd ~ 10- 15 M) to streptavidin, a 60 kD tetrameric protein. Biotin can be readily functionalized at its carboxyterminus while retaining its ability to bind tightly to streptavidin, rendering it an ideal probe [20- 22]. Biotinylated amino acids have previously been incorporated into proteins in vitro by supplementing the translation mixture with full-length lysyl-tRNA functionalized with biotin derivatives at the c-amino group of lysine, thus leading to random incorporation of biotinylated amino acids at various lysine sites in the protein [23]. Although this suggested that biosynthetic incorporation of biotinylated amino acids was possible, the remaining challenge was to introduce the biotin label site-specifically in the protein of interest in a living cell. To accomplish this we employed the nonsense suppression method for incorporation of unnatural amino acids into proteins [24-27] as adapted in our labs to the in vitro heterologous expression system of the Xenopus oocyte [28- 30]. In the in vivo nonsense suppression method, the oocyte is co-injected with 2 mutated RNA species: 1) mRNA synthesized in vitro from a mutant cDNA clone containing a stop codon, TAG, at the amino acid position of interest; and 2) a suppresser tRNA containing (a) the complementary anti-codon sequence (CUA) and (b) the desired unnatural amino acid chemically acylated to the 3' end. During translation by the oocyte's endogenous synthetic machinery, the unnatural amino acid is specifically incorporated at the appropriate position in the protein encoded by the 149 mRNA. Recent studies from these labs have shown that this method is broadly applicable, being compatible with many different proteins, a large number of sites in a given protein, and a wide range of amino acids [28, 30]. As a test amino acid we chose biocytin, a naturally occurring (but not translationally incorporated) biotinylated amino acid. Biocytin's relatively small size, similarity to the natural amino acid lysine, and commercial availability made it particularly attractive. Figure 6.2: Structure of biocytin To test the method, we chose the prototype ligand-gated ion channel, the nicotinic acetylcholine receptor (nAChR) , a 290 kD pentameric ion channel found at the neuromuscular junction [31- 35]. In particular, we chose to evaluate the main immunogenic region (MIR) of the a subunit. The MIR is composed of ten residues in the N-terminal region of the a subunit (residues 67-76) and binds many antibodies, including those associated with myasthenia gravis, a human autoimmune disease [36- 38]. Furthermore, this region has been shown by electron microscopy to be located near the synaptic end of the nAChR [39]. Because it is known to be topologically extracellular and accessible as a unit to large proteins, the MIR appeared to be an ideal location to test the biocytin methodology for evaluating the surface exposure of individual residues. Given the tight fit between biotin and streptavidin, we anticipated that only the most highly exposed positions of the MIR would form strong biocytin/streptavidin complexes. 150 6.4 Results and Discussion We first mutated the codons for each of the 10 residues of the main immunogenic region of the nAChR from mouse muscle (a67-76 - see Table 6.1 for sequence) to the stop codon TAG. We then co-injected Xenopus oocytes with a mixture of mutant mRNA (1.2 ng total/oocyte, 4:1:1:1 a:(3:,:b subunit stoichiometry) containing the TAG codon at a single position in the a subunit, and an aminoacyl-tRNA (25 ng/ oocyte) charged with the amino acid of interest. After 36-48 hours, oocytes were assayed electrophysiologically for response to 200 µM acetylcholine (ACh) by the two electrode voltage clamp technique [40]. The results from a successful biocytin incorporation are shown in Figure 6.3. In this experiment, biocytin incorporation was successful at both positions a70 and a76. As controls, suppression with a natural amino acid , tyrosine at a76, and experiments using uncharged tRNA (tRNA-dCA) [28- 30] were performed as well. Clearly, the biotinylated mutants functioned at levels comparable to those for mutants suppressed with a natural amino acid. In addition, the very small currents arising from the dCA control (uncharged tRNA) demonstrate that the currents observed in the biocytin experiments arise solely from receptors containing the biotinylated amino acid. At five of the ten positions in the MIR - a68, 70, 71, 75 and 76 - biocytin was incorporated efficiently into functional receptors with whole-cell currents ranging from several hundred to several thousand nA. At the other five positions - a67, 69, 72, 73 and 74 - whole-cell currents were generally less than 200 nA. In each experiment, the a70 biocytin mutant was expressed as a positive control so that whole-cell currents, which often vary substantially between oocyte batches, could be compared to an internal standard. A summary of these experiments is shown in Figure 6.4. Again as a control, uncharged tRNA was tested at all the sites. In all cases, the current evoked from 200 µM ACh was less than 20% of that obtained in the suppression experiment, and in nine out of ten cases, it was less than 5%. The electrophysiology experiments clearly established the presence of functional, biocytin-containing ACh receptors on the oocyte surface for 5 of the 10 positions in 151 10000 8000 16000 "E ~ '"" ::l U 4000 2000 0 _l____a------a70 dCA et.70 Bio Low et.70 Bio Hi et.76 Bio et.76 Tyr Figure 6.3: Average whole-cell currents in response to 200 µM ACh for oocytes injected with nAChR a70TAG or a76TAG (4:1:1:1, 1.2 ng) and uncharged tRNA (dCA) , or tRNA charged with biocytin (Bio) or tyrosine (Tyr). The two bars for the a70 biocytin are the averages of the currents for the five lowest expressing oocytes and the five highest expressing oocytes. Error bars are ± s.e.m. (n=5) the MIR. Sequence conservation analysis of 34 nAChR a subunits shows that biocytin produced functional receptors at five of the six least conserved positions in the MIR (Table 6.1). The fact that biocytin did not express well at the sixth, aGly73, may arise because a glycine (side chain molecular weight= 1) to biocytin (side chain molecular weight=298) mutant may epitomize the phrase "non-conservative mutation." Furthermore, the four most highly conserved residues, three of which are greater than 70% conserved across all nicotinic subunits, are the positions at which biocytin did not express. Since these four residues are highly conserved, they may play crucial structural or functional roles in the receptor, and even non-conservative conventional mutations may not be tolerated. We believe that the low currents observed from attempted biocytin incorporation at these positions indicate that the relatively large side chain of biocytin is not tolerated at these positions. 152 0.8 ..... i::: ~ ...;:::s 0.6 u "d Q.) -~ cil s... 0.4 z 0 0.2 Trp67 Asn68 Pro69 Asp70 Asp71 Tyr72 Gly73 Gly74 Val75 Lys76 Figure 6.4: Whole-cell currents in response to 200 µM ACh from attempted biocytin suppression experiments at each of the 10 positions of the MIR of the nAChR. Results are normalized to the whole-cell current of a70 biocytin expressed in the same experiment. Bars represent the mean of two experiments, each with five oocytes. With electrophysiological evidence for functional biotinylated receptors, we turned our attention to binding streptavidin to assay the surface exposure of particular side chains. To assay for streptavidin binding, batches of five oocytes whose whole-cell currents had previously been determined were incubated in a solution of 300 pM [ 125 I]-streptavidin for two hours, followed by extensive washing (see Experimental section). Individual oocytes were then measured in a ')'-counter. Of the five mutants that showed large whole-cell currents, only one, a70 biocytin, showed streptavidin binding above background levels. In addition, no streptavidin binding was observed with any of the five mutants that did not give large whole-cell currents, or to cells expressing non-biotinylated nAChRs. The results of a typical experiment are shown in Figure 6.5. In the case of a70 biocytin, the mutant that showed streptavidin binding, we also evaluated the binding of [125 I]-a-bungarotoxin - a snake toxin that is highly 153 Position Trp67* Asn68 Pro69 Asp70 Asp71 Tyr72* Gly73 Gly74* Val75 Lys76 Percent conservation in a subunits 97 56 79 26 53 91 47 82 62 44 ACh response with biocytin incorporation + + + + + Table 6.1: Sequence analysis of the MIR from 34 nAChR a subunits in the SWISSPROT database. Positions marked with a * are > 70% conserved across all nicotinic subunits (i.e. , not solely a subunits). specific for the nAChR. We found that the amounts of streptavidin and bungarotoxin binding were similar, suggesting that all nAChRs were completely labeled by both bungarotoxin and streptavidin. This also provides concrete biochemical evidence to support the electrophysiological evidence that the in vivo suppression method produces full length protein incorporating only the desired amino acid. If this were not the case, i.e. , if some nAChR was produced with a natural amino acid incorporated at the suppression site due to read through of the stop codon or reacylation of the tRNA, we would have expected the amount of bungarotoxin binding to exceed the amount of streptavidin binding. To explore whether the streptavidin binding sites and bungarotoxin binding sites overlap, we incubated oocytes expressing nAChR a70 biocytin with unlabeled streptavidin. We then applied [125 1]-a-bungarotoxin to these oocytes to determine if it could still bind. We also performed the reverse experiment, incubating oocytes expressing nAChR a70 biocytin with unlabeled a-bungarotoxin and attempting to bind [ 125 1]-streptavidin. The results of these experiments are shown in Figure 6.6. These results clearly demonstrate that binding of bungarotoxin does not strongly inhibit binding of streptavidin to the a70 biocytin mutant and vice versa. This is consistent with previous studies that demonstrated that both bungarotoxin and 154 300 250 .....;::s(l) 200 .s ~ I-< (l) 0.. 150 .....rJ) § 0 u 100 50 0 a70dCA a70 Bio Low a70 Bio Hi a76 Bio a76 Tyr Figure 6.5: [125 I]-Streptavidin binding to oocytes injected with nAChR a70TAG or a76TAG (4:1:1 :1, 1.2 ng) and uncharged tRNA (dCA), or tRNA charged with biocytin (Bio) or tyrosine (Tyr). The oocytes used in this experiment are identical to those whose whole-cell currents are shown in (Figure 6.3), thus demonstrating that binding of streptavidin scales with whole-cell current. The a76 Tyr mutation establishes that streptavidin does not bind to non-biotinylated nAChRs. Detector background has been subtracted in all cases. Error bars are ± s.e.m. (n=5) antibodies against the MIR are capable of binding simultaneously to the intact receptor [41]. Finally, we have measured the EC 50 value for ACh for the a70 biocytin mutant both before binding streptavidin (EC 50 ~ 39 µM) and after binding (EC 50 ~ 34 µM). These values are not significantly different from one another, nor are they significantly different from EC 50 of the wild-type receptor (49 µM). At first it may seem surprising that binding a large protein to t he receptor would not alter its pharmacological behavior. However, previous studies have shown that antibodies binding to the MIR do not substantially alter the physiology of the receptor [42]. To gain a clearer physical picture of the above results, we have considered them in light of a previously published structural model. Tsikaris et al. have used NMR 155 20 --<:::t 15 .....;::j -----Q) .5 :::8 10 1--< Q) 0... .....Cf.J i:: ;::j 0 u 5 0 Btx* Stp + Btx* Stp* Btx + Stp* Figure 6.6: Whole-cell binding of [125 1]-streptavidin and [125 1]-bungarotoxin to AChR a70 biocytin mutants. Btx*: binding [125 1]-bungarotoxin to untreated oocytes; Stp + Btx*: binding [125 1]-bungarotoxin to oocytes pretreated with streptavidin; Stp*: binding [125 1]-streptavidin to untreated oocytes; Btx + Stp*: binding [125 1]-streptavidin to oocytes pretreated with bungarotoxin. Within the error bars shown, the data demonstrate that [125 1]-bungarotoxin and [125 1]-streptavidin do not interfere with one another in binding to the a70 biocytin nAChR. Binding is expressed as counts per minute per µA of whole-cell current to account for different expression levels between groups of oocytes. Counts have been normalized to account for the differences in specific activity between the [125 1]-streptavidin and [125 1]-bungarotoxin. Error bars are ± s.e.m. (n=5) to determine the structure of a ten-residue peptide corresponding to the sequence of the MIR from the a subunit of the nAChR from Torpedo califomica [43] . This sequence differs from that of the mouse muscle MIR studied here at two sites Ala70 and Ile 75 . Starting with the coordinates for the Torpedo MIR taken from the Protein Data Bank [44], we altered the side chain identities of these two residues to the corresponding mouse muscle residues (Asp and Val, respectively) using standard geometries. This structure is shown in Figure 6.7. We have color-coded the structure as follows: purple - sites at which biocytin incorporation was unsuccessful; gray sites that incorporated biocytin but did not bind streptavidin; red - a70, the sole 156 Figure 6.7: NMR structure (CPK model) of MIR peptide (PDB:lTOR) with modifications described in text. Color scheme: purple - sites at which biocytin incorporation was unsuccessful; gray - sites that incorporated biocytin but did not bind streptavidin; red - a70, the sole streptavidin-binding site. The right view is derived from the left by a 180° rotation around the vertical (y) axis. streptavidin-binding site. The structure in Figure 6. 7 leads naturally to a rationalization of our results. The MIR peptide clearly has two sides - one that can tolerate biocytin, and one that cannot. Biocytin is larger than any of the 20 natural amino acids. Perhaps it will produce functional receptors only if there is a pre-existing (or easily created) open space for this larger side chain. We suggest that the "gray" side has this open space, and so presumably faces out away from the bulk of the protein. In contrast, the "purple" side cannot tolerate biocytin, and so apparently these side chains project down into the rest of the protein. The selectivity of the biocytin method is shown by the result that biocytin was incorporated at five sites, yet only one of these sites was competent to bind streptavidin under the conditions of these experiments. The crystal structure of the biotinstreptavidin complex reveals that the ligand is buried fairly deeply in the protein, with even the terminal carboxylate not fully exposed [45]. Only residues that position the biocytin in an unencumbered environment will lead to streptavidin binding. As shown in Figure 6. 7, the Asp70 side chain (red) points "up", away from , for example, the 157 Trp67 or Try72 side chains that are postulated to project down into the receptor. The other "biotin-accepting" side chains (gray) project approximately normal to the Asp70 side chain. Apparently, these sites (68, 71, 75, 76) point to a space that is large enough for biocytin but not for streptavidin. There is precedent for binding interactions with dissociation constants significantly weaker than 10- 15 M for substituted biotins [46], and perhaps higher streptavidin concentrations would have allowed detectable binding at the "gray" sites. However, such studies were not feasible due to non-specific binding of streptavidin to the oocytes at higher concentrations. Our results thus suggest a particular orientation for the MIR on the nAChR. Residue a70 points directly out away from the receptor. The side chains emanating from residues 67, 69, and 72 (73 and 74 are Gly) project down into the protein. The remaining residues, 68, 71, 75, and 76, project into open space, but perhaps toward some other structural feature of the receptor that prevents streptavidin binding. This analysis implies that a tether longer than is present in biocytin might allow streptavidin binding at the "gray" sites. We therefore prepared two longer-chain analogs of biocytin, A and B. Unfortunately, we were not able to incorporate the longer analog B, into functional receptors. We could incorporate biocytin-analog A at position a70 at levels that were readily detectable by electrophysiology (500-1000 nA) . This is a remarkably "unnatural" amino acid, suggesting considerable flexibility in the ribosomal machinery. However, this amount of expression was not sufficient for the less sensitive streptavidin assay, and so we were not able to make use of the longer chain analog. 0 H 0 H N ~ N ~ N~ ' ' 0 H [ H2N COOH B q HN NH d Figure 6.8: Biocytin analogs We conclude that the method of site-specific biocytin incorporation with streptavidin detection provides a potentially general way to identify highly surface exposed residues in complex, integral membrane proteins. It thus provides information that 158 goes beyond the "simpler" question of transmembrane topology. This technique will likely be applicable to other channels, receptors, and transporters. Our data also show that biocytin derivatives with longer tethers do not express well, so that the biocytin method alone cannot be simply generalized to a wide range of exposures. Alternative strategies can be envisioned, perhaps involving incorporation of unnatural residues that are rather small but that have specific reactivities allowing attachment of biotin moieties with longer chains [47- 49]. 6.5 Summary The in vivo nonsense suppression method for the incorporation of unnatural amino acids is a powerful tool for obtaining chemical-scale information about functional proteins expressed in X enopus oocytes. We have extended the method to incorporate a biotinylated amino acid site-specifically within the main immunogenic region (MIR) of the nicotinic acetylcholine receptor (nAChR), the prototypical ligand-gated ion channel. Using [125 I]-labeled streptavidin as a probe, we have identified an amino acid in the MIR that is highly exposed on the surface of the protein. We believe that this method represents an improvement over present approaches to determining surface exposure and transmembrane topology for a number of reasons. In particular, the method does not require purified protein; uses a minimally perturbing point mutation; uses a marker that is not endogenous to either the protein of interest or to other cell surface proteins; allows detection by a commercially available assay; and determines the topology /surface exposure in a functional, characterizable protein expressed in vivo. For these reasons, we consider the method to be a potentially general one for determining the surface exposure of specific residues of transmembrane proteins expressed in vivo. In a broader sense, these results further demonstrate the power of the in vivo nonsense suppression method to apply the tools of organic chemistry to the problems of molecular neuroscience. 159 6.6 Experimental Reagents were purchased from Aldrich unless otherwise noted. (3-[ 125 1]-iodotyrosyl54)-a-bungarotoxin (2000 mCi/mmol) and [125 1]-streptavidin (30 mCi/mg) were purchased from Amersham. Anhydrous THF was distilled from Na/benzophenone; anhydrous DMF was obtained from Aldrich. All chemical syntheses were performed under a positive pressure of argon unless noted otherwise. NMR spectra were recorded on a General Electric QE300 (300 MHz 1 H, 75 MHz 13 C), JEOL Delta 400 (100 MHz, 13 C), or a Bruker AM-500 (500 MHz 1 H, 125 MHz 13 C) spectrometer and are refer- enced to residual solvent protons. Mass spectrometry (FAB) was performed by the mass spectrometry facilities at the University of California, Riverside, or at Caltech. UV /Vis spectra were recorded on a Beckman Instruments DU-640 continuous-wave spectrometer. HPLC separations were performed on a Waters dual 510 pump liquid chromatograph system equipped with either a Waters 490E variable wavelength UV detector or a Waters 960 photo diode array detector. Analytical HPLC samples were separated using a Waters NovaPak C18 column (3.9 x 150 mm); semi-preparative samples were separated using a Whatman Magnum 9 column (9.4 x 500 mm, Partisil 10, ODS-3). Oligonucleotides for mutagenesis were synthesized by the Biopolymer Synthesis Facility at Caltech. PCR reactions were performed using a Perkin-Elmer TC-1 DNA Thermal Cycler. Automated sequencing of PCR products was performed by the DNA Sequencing Core Facility at Caltech. Ligation of aminoacyl-dCA to tRNA-THG73(-CA) was performed as described previously [29, 30]. 6.6.1 Synthesis of Biotinylated Amino Acids NVOC-Biocytin Cyanomethyl Ester Biocytin (50 mg, 0.13 mmol, Molecular Probes) and sodium carbonate (28 mg, 0.27 mmol) were dissolved with stirring in water (7.5 mL) and freshly distilled THF (5 mL). A solution of 6-nitroveratryloxycarbonyl chloride (NVOC-Cl) (37 mg, 0.13 mmol) in 5 mL THF was added dropwise [50, 51]. After three hours, solvents were removed in 160 vacuo and the remaining residue was dissolved in dry DMF (1.5 mL) and chloroacetonitrile (1.5 mL). Triethylamine (400 µL) was added and the mixture was stirred overnight. Volatiles were removed in vacuo and the remaining solid was purified by flash chromatography (silica gel, CH 2 Cb + MeOH 1-10%) to give 67 mg of NVOCbiocytin cyanomethyl ester as a pale yellow solid (77% yield for 2 steps): 1 H NMR (500 MHz, DMSO-c4;) c5 8.10 (d - br, 1=6.8 Hz, lH), 7.74 (t - br, J 5.0 Hz, lH), 7.70 (s, lH), 7.18 (s, lH), 6.39 (s, lH), 6.33 (s, lH), 5.38 (AB, J =15.0 Hz, lH), 5.34 (AB, J =15.0 Hz, lH), 5.01 (s, 2H), 4.27 (m, lH), 4.11 (m, lH), 3.92 (s, 3H), 3.86 (s, 3H), 3.13 (m, lH), 3.04 (m, lH), 2.99 ,(m, 2H), 2.80 (dd, J =10 Hz, 5.5 Hz, lH), 2.57-2.46 (m, 3H), 2.03 (t, 2H), 1.70-1.15 (m, 10H); 13 C NMR (75 MHz, DMSO-c4; c5 172.0, 171.6, 162.8, 155.8, 153.5, 147.8, 139.2, 127.8, 115.8, 110.3, 108.2, 62.8, 61.1, 59.3, 56.3, 56.1, 55.5, 53.6, 49.5, 39.9, 38.1, 35.3, 30.2, 28.7, 28.3, 28.1, 25.4, 22.8. a-NVOC-c-(6-(biotinoyl)amino )hexanoyl-L-lysine (NVOC-A) A solution of a-NVOC-lysine (42 mg, 0.11 mmol, 1.0 eq., prepared in two steps from a-Boe-lysine and NVOC-Cl, followed by removal of the Boe-group with TFA), 6((biotinoyl)amino)caproic acid N-hydroxysuccinimide ester (50 mg, 0.11 mmol, 1.0 eq., Molecular Probes) and Na2 CO 3 (47 mg, 0.44 mmol, 4.0 eq.) was stirred in dry DMF (5 mL) under argon. After 3 days, the mixture had become gel-like and a solution of lN NaHSO 4 (50 mL) was added, followed by CH 2 Cl 2 (50 mL). The mixture was stirred and a pale yellow precipitate formed. The solid was filtered, washed with CH 2 Cb, water, acetone, and dried to afford 42 mg of a pale yellow solid (53% yield): 1 H NMR (500 MHz, DMSO-c4;) c5 7.74 (d - br, J =8 Hz, lH), 7.66 (m, 2H), 7.64 (s, lH), 7.12 (s, lH), 6.39 (s, lH), 6.33 (s, lH), 5.33 (AB, J =15.lHz, lH), 5.25 (AB, J =15.lHz, lH), 4.23 (t, J =5.5 Hz, lH), 4.06 (m, lH), 3.85 (s, 3H), 3.80 (s, 3H), 3.02 (m, lH), 2.93 (m, 4 H), 2.82 (s, lH), 2.76 (m, lH), 2.73 (s, lH), 2.50 (d, J =12.4 Hz, lH), 1.94 (m, 4H), 1.63 (m, lH), 1.52 (m, 2H), 1.37 (m, 4H), 1.28 (m, 8H), 1.15 (m, 2H); 13 C NMR (100 MHz, DMSO-c4;) c5 173.8, 171.9, 171.8, 162.7, 155.8, 153.5, 147.6, 139.0, 138.1, 128.2, 110.0, 108.1, 62.4, 61.0, 59.2, 56.2, 56.1, 55.4, 53.8, 38.3, 38.1, 35.4, 35.2, 30.4, 28.9, 28.7, 28.2, 28.0, 26.1, 25.3, 25.0, 23.0. 161 a-NVO C-c-( 6-(biotinoy I) amino) hexanoy 1-L-lysine Cyanomethy I Ester To a solution of a-NVOC-c-(6-(biotinoyl)amino)hexanoyl-L-lysine (40 mg, 0.055 mmol, 1.0 eq.) in 2 mL dry DMF was added triethylamine (24 µL, 0.165 mmol, 3.0 eq.) and chloroacetonitrile (2 mL). After stirring for 9 hours under argon, volatiles were removed in vacuo and the remaining solid was partitioned between water and CH 2 Cb . A pale yellow solid remained and was combined with the organic fraction, concentrated and dried in vacuo to afford the cyanomethyl ester as a pale yellow solid (20 mg, 47%): 1 H NMR (500 MHz, DMSO-<k) c5 8.07 (d, J =7.5 Hz, lH) , 7.70 (m, 2H) , 7.18 (s, lH) , 6.39 (s, lH) , 6.33 (s, lH) , 5.38 (AB, J =14.8 Hz, lH) , 5.34 (AB, J =14.8 Hz, lH) , 5.01 (s, 2H) , 4.29 (t , J =5.6 Hz, lH) , 4.12 (m, 2H) , 3.92 (s, 3H) , 3.86 (s, 3H) , 3.08 (m, lH) , 2.97 (m, 4H) , 2.80 (s, lH) , 2.57-2.49 (m, 2H) , 2.01 (m, 4H) , 1.63 (m, lH) , 1.43 (m, 4H) , 1.34 (m, 8H), 1.17 (m, 2H). a-NVOC-c-( 6-( ( 6-( (biotinoy I) amino) hexanoy I) amino) hexanoy I)- L-lysine (NVOC-B) Fluorenylmethyl Ester. A mixture of 6-((6-((biotinoyl)amino)hexanoyl)amino)hexanoic acid, succinimidyl ester (50 mg , 0.088 mmol, 1 eq.) and a-NVOC-lysine fluorenylmethyl ester (75 mg, 0.13 mmol, 1.5 eq.) [52] was stirred in dry DMF (1 mL) with triethylamine (25 µL , 2 eq.). The reaction was stirred for four days after which time the volatiles were removed and the yellow residue was purified by column chromatography (silica, 4:1 CH 2 Cl 2 :MeOH) to afford 46 mg of the desired product (47% yield): 1 H NMR (500 MHz, DMSO-<k) c5 7.92-7.84 (m, 3H) , 7.69-7.55 (m, 6 H) , 7.36 (q, J =7.6 Hz, 2H), 7.25 (m, 2H) , 7.13 (s, lH) , 6.35 (s, lH) , 6.30 (s, lH), 5.34 (AB, J =13.5 Hz, lH) , 5.28 (AB, J =13.5 Hz, lH) , 4.45 (m, lH) , 4.38 (m, lH), 4.23 (m, lH) , 4.19 (t, J 5.0 Hz, lH) , 4.07 (m, lH) , 3.89 (q, J-7.0 Hz, lH) , 3.84 (s, 3H) , 3.80 (s, 3H) , 3.05 (m, lH) , 2.75 (m, lH) , 1.97 (m, 6H) , 1.65- 1.20 (m, 26H) ; 13 C NMR (125 MHz, DMSO-<k) c5 172.0, 171.6, 171.5, 162.4, 155.5, 153.2, 147.5, 144.0, 143.4, 140.5, 139.0, 127.5, 127.4, 126.8, 124.7, 119.8, 110.3, 108.0, 65.5 , 62 .2, 60.8, 59.0, 55.9, 55 .9, 55.1, 53.7, 46.1 , 40.3, 38.0, 37.8, 35.1 , 35.0, 29.9 , 28.7, 28.4, 27.9, 27.7, 25.9, 25.0, 24.8, 22.5. 162 a-NVOC-c:-(6-( (6-( (biotinoyl)amino )hexanoyl)amino )hexanoyl)-L-lysine Cyanomethyl Ester. a-NVOC-c:-( 6-( (6-( (biotinoyl)amino )hexanoyl)amino )hexanoyl)- L-lysine fluorenylmethyl ester (19 mg, 0.02 mmol, 1 eq.) was dissolved in dry DMF (1 mL). Piperidine (20 µL , 10 eq.) was added and t he mixture was stirred. After 2.5 hours , volatiles were removed in vacuo, hexane (15 mL) was added, and the suspension was filtered. The remaining solid was dissolved in aqueous DMF. The solution was lyophilized and the solid residue was dissolved in dry DMF (4 mL) , chloroacetonitrile (0.8 mL) and triethylamine (100 µL). After 24 hours , the mixture was separated using preparative HPLC with a linear gradient of 100% 10 mM aqueous acetic acid to 100% acetonitrile over one hour. The product-containing fractions were lyophilized to give 5 mg of a pale yellow solid (31 % yield for 2 steps): 1 H NMR (500 MHz , DMSO-c4;) r5 8.08 (d, J=7.5 Hz, lH), 7.95 (s, lH) , 7.18 (s, lH), 6.40 (s, lH) , 6.34 (s, lH) , 5.40 (AB, J =14.8 Hz, lH) , 5.35 (AB, J = 14.8 Hz, lH) , 5.01 (s, 2H) , 4.31 (m, lH) , 4.12 (m, 2H) , 3.92 (s, 3H) , 3.86 (s, 3H) , 3.07 (m, lH) , 2.98 (m, 6 H) , 2.88 (s, lH) , 2.80 (dd, J 7.3 , 5.2Hz, lH), 2.72 (s , lH) , 2.57 (d , J=12.4 Hz, lH), 1.75- 1. 52 (m , 4H) , 1.50-1.41 (m, 6H), 1.41-1.25 (m, SH) , 1.25-1.12 (m, 4H); 13 C NMR (125 MHz, DMSO-c4;) r5 171.8, 171.7, 171.3, 162.5, 162.2, 155.6 , 153.4, 148.6, 139.3, 127.5, 115.5, 110.5, 108.3, 66.9, 62.5, 61.0, 59 .1 , 56. 1, 56. 1, 55.2, 53.5 , 49.7, 49.3, 44.2 , 38.2, 37.9, 35.6 , 35.3 , 35. 1, 30.7, 28.8 , 28.5 , 28.0 , 27.9, 26.0, 25.2, 24.9, 24.8, 22.6. 6.6.2 General Method for Coupling of N-Protected Amino Acids to dCA. The N-protected amino acid (~ 30 µmol , 3 eq. ) was mixed with the tetra-n- butylammonium salt of the dCA dinucleotide (~ 10 µm ol, 1 eq.) in dry DMF (400 µL). The reactions were monitored by analytical HPLC with a linear gradient of 100% 25 mM aqueous ammonium acetate buffer (pH 4.5) to 100% acetonitrile over one hour. Following the disappearance of the dCA (5 min. to 3 hours), the reaction mixture was separated using semi-preparative HPLC with a linear gradient of 100% 25 163 mM aqueous ammonium acetate buffer (pH 4.5) to 100% acetonitrile. The fractions containing the aminoacyl dinucleotide were collected and lyophilized. The lyophilized solid was redissolved in 10 mM aqueous acetic acid/ acetonitrile and lyophilized a second time to remove salts. The products were characterized by UV-Vis and mass spectrometry. dCA-NVOC-Biocytin Prepared as above: FAB-MS [M-HJ- 1228, calcd for C4sH51N13O22P2S 1228; UV Amax (10 mM Aq. HOAc) 258 nm, 350 nm. dCA-0::-NVOC-c-( 6-(biotinoyl )amino )hexanoyl-L-lysine Prepared as above: FAB-MS [M-HJ-1342, calcd for Cs1H12N14O23P2S 1342; UV Amax (10 mM Aq. HOAc) 256 nm, 350 nm. dCA-0::-NVOC-c- (6-( (6-( (biotinoyl)amino )hexanoyl)amino )hexanoyl)-L- lysine Prepared as above: ESI-MS [M-Na]+ 1480, calcd for C57 HsoN 15 O2 4P2SNa 1480; UV Amax (10 mM Aq. HOAc) 253 nm, 350 nm. 6.6.3 Site-Directed Mutagenesis of the AChR a67-76 TAG Mutants The mouse muscle AChR 0::67-76 TAG mutants were prepared using a standard two step PCR-based cassette mutagenesis procedure using Pfu Polymerase (Stratagene). For each mutant, a trimmed BsmBI-MscI PCR-generated fragment was subcloned into the dephosphorylated BsmBI-MscI fragment of mouse muscle AChR a subunit in the pAMV-PA vector [30]. When possible, silent mutations were introduced to facilitate screening of positive colonies by restriction endonuclease digestion. We have found the Bfal recognition sequence CTAG to be particularly helpful for screening. 164 All mutants were verified by automated sequencing of the entire PCR cassette across both ligation sites. 6.6.4 In Vitro Transcription of mRNA mRNA was synthesized in vitro from Not! linearized plasmid DNA (1 µg/40 µL reaction) using the T7 MessageMachine kit from Ambion per the manufact urer's instructions. The products were purified by phenol:chloroform:isoamyl alcohol extraction and isopropanol precipitation. Purity of the mRN A was assayed by gel electrophoresis. 6.6.5 Oocyte Preparation and Injection Oocytes were removed from Xenopus laevis as described previously [40] and maintained at 18 °C in ND96 solution consisting of 96 mM NaCl, 2 mM KCl , 1 mM MgCb, 1.8 mM CaCb, 5 mM HEPES (pH 7.5), supplemented with sodium pyruvate (2.5 mM), gentamicin (50 µg/mL), theophylline (0.6 mM) and horse serum (5%). Prior to microinjection, the NVOC-aminoacyl-tRNA (1 µg/ µL in 1 mM NaOAc, pH 5.0) was deprotected by irradiating the sample for 5 minutes with a 1000W xenon arc lamp (Oriel) operating at 600W equipped with WG-335 and UG-11 filters (Schott). The deprotected aminoacyl-tRNA was mixed 1:1 with a water solution of the desired mRNA. Oocytes were injected with 50 nL of a mixture containing 25 ng aminoacyltRNA and 1.2 ng of total mRNA (4:1:1:1 a:{3:,:6 subunit stoichiometry). 6.6.6 Electrophysiology Two-electrode voltage clamp measurements were performed 36- 48 hours after inject ion using a GeneClamp 500 amplifier (Axon Instruments, Foster City, CA). Microelectrodes were filled with 3M KCl and had resistances of 0.5 to 2 MO. Oocytes were perfused with a nominally calcium-free bath solution containing 96 mM NaCl, 2 mM KCl, 1 mM MgCb, 5 mM HEPES (pH 7.5). Macroscopic ACh-induced currents were 165 recorded in response to application of 200 µM ACh at a holding potential of -80 mV. Whole-cell current measurements are reported as the mean ± s.e.m. 6.6. 7 General Procedure for Radioactive Ligand Binding Experiments Batches of five oocytes were incubated at room temperature for 15 min with gentle agitation in 400 µL calcium-free ND96 (see Electrophysiology section) containing 10 mg/ mL BSA (FractionV, Sigma). The radioactive ligand was then added to a final concentration of 300 pM for [125 I]-streptavidin (Amersham, 30 µCi/ µg) or 1 nM for [125 I]-a-bungarotoxin (Amersham, 2000 Ci/mmol). After two hours, the oocytes were transferred to an ice-cold solution of ND96 + 10 mg/mL BSA (1.5 mL) and agitated for five min. The oocytes were then washed extensively in fresh solutions of ice-cold ND96 + BSA and transferred individually to scintillation vials and counted in a Beckman 185000 ')'-counter. In all cases, a sample of the final wash solution was counted to ensure that unbound ligand had not been carried through the assay. In addition, five empty scintillation vials were counted in each experiment to determine the detector background which was then subtracted from subsequent measurements. In experiments involving the binding of both labeled and unlabeled ligand, the unlabeled ligand was added at high concentration (166 nM for streptavidin, 312 nM for bungarotoxin) to the initial 400 µL solution. After 20 min, the labeled ligand was added and the assay proceeded as described above. 166 Bibliography [1] S. H. White, editor. Membrane Protein Structure. Methods in Physiology. Oxford University Press, Inc, New York, 1994. [2] M. Hollmann, A. O'Shea-Greenfield, S. W. Rogers, and S. Heinemann. Cloning by functional expression of a member of the glutamate receptor family. Nature, 342:643- 648, 1989. [3] M. Hollmann, C. Maron, and S. Heinemann. N-glycosylation site tagging suggests a three transmembrane domain topology for the glutamate receptor GluRl. Neuron, 13:1331- 1343, 1994. [4] J. A. Bennett and R. Dingledine. Topology profile for a glutamate receptor: Three transmembrane domains and a channel-lining reentrant membrane loop. Neuron, 14:373-384, 1995. [5] R. A. Schwalbe, Z. Wang, L. Bianchi, and A. M. Brown. Novel sites of Nglycosylation in ROMKl reveal the putative pore-forming segment H5 as extracellular. J. Biol. Chem., 271:24201- 24206, 1996. [6] E. R. Bennett and B. I. Kanner. The membrane topology of GAT-1, a (Na+ + c1-)-coupled 1 -aminobutyric acid transporter from rat brain. J. Biol. Chem., 272:1203- 1210, 1997. [7] L. Olivares, C. Aragon, C. Gimenez, and F. Zafra. Analysis of the transmembrane topology of the glycine transporter GLYTl. J. Biol. Chem., 272:1211- 1217, 1997. [8] E. Turk, C. J. Kerner, M. P. Lostao, and E. M. Wright. Membrane topology of the human Na+ /glucose cotransporter SGLTl. J. Biol. Chem. , 271:1925- 1934, 1996. 167 [9] C. H. Pedemonte, G. Sachs, and J. H. Kaplan. An intrinsic membrane glycoprotein with cytosolically oriented N-linked sugars. Proc. Natl. Acad. Sci. USA , 87:9789- 9793, 1990. [10] X. B. Chang, Y. X. Hou, T. J. Jensen, and J. R. Riordan. Mapping of cystic fibrosis transmembrane conductance regulator membrane topology by glycosylation site insertion. J. Biol. Chem., 269:18572- 18575, 1994. [11] S. Frillingos and H. R. Kaback. Probing the conformation of the lactose permease of Escherichia coli by in situ site-directed sulfhydryl modification. Biochemistry, 35:3950- 3956, 1996. [12] T . W. Loo and D. M. Clarke. Membrane topology of a cysteine-less mutant of human P-glycoprotein. J. Biol. Chem., 270:843- 848, 1995. [13] J. Sun, J. Li, N. Carrasco, and H. R. Kaback. The last two cytoplasmic loops in the lactose permease of Escherichia coli comprise a discontinuous epitope for a monoclonal antibody. Biochemistry, 36:274- 280, 1997. [14] S. L. Slatin, X. Q. Qiu, K. S. Jakes, and A. Finkelstein. Identification of a translocated protein segment in a voltage-dependent channel. Nature, 371:158161, 1994. [15] M.H. Akabas, D.A. Stauffer, M. Xu, and A. Karlin. Acetylcholine receptor channel structure probed in cysteine-substitution mutants. Science, 258:307310, 1992. [16] J. A. Javitch, D. Fu, J. Chen, and A. Karlin. Mapping the binding-site crevice of the dopamine D2 receptor by the substituted-cysteine accessibility method. Neuron, 14:825- 831, 1995. [17] R. A. Chavez and Z. W . Hall. Expression of fusion proteins of the nicotinic acetylcholine receptor from mammalian muscle identifies the membrane-spanning regions in the a and 5 subunits. J. Cell Biol., 116:385- 393, 1992. 168 [18] R. Anand, L. Bason, M. S. Saedi, V. Gerzanich, X. Peng, and J . Lindstrom. Reporter epitopes: A novel approach to examine transmembrane topology of integral membrane proteins applied to the a subunit of the nicotinic acetylcholine receptor. Biochemistry, 32:9975- 9984, 1993. [19] M. Mund, C. Weise, P. Franke, and F. Hucho. Mapping of exposed surfaces of the nicotinic acetylcholine receptor by identification of iodinated tyrosine residues. J. Prat. Chem., 16:161- 170, 1997. [20] N. M. Green. Avidin. Advan. Protein Chem., 29:85- 133, 1975. [21] M. Wilchek and E. A. Bayer, editors. Avidin-Biotin Technology, volume 184 of Methods in Enzymology. Academic Press, Inc., San Diego, CA, 1990. [22] K. Hofmann, G. Titus, J. A. Montibeller, and F. M. Finn. Avidin binding of carboxyl-substituted biotin and analogues. Biochemistry, 21:978- 984, 1982. [23] T. Kurzchalia, M. Wiedmann, H. Breter, W. Zimmermann, E. Bauschke, and T. A. Rapoport. tRNA-Mediated labelling of proteins with biotin. A nonradioactive method for the detection of cell-free translation products. Eur. J. Biochem., 172:663- 668, 1988. [24] C. J. Noren, S. J. Anthony-Cahill, M.C. Griffith, and P.G. Schultz. A general method for site-specific incorporation of unnatural amino acids into proteins. Science, 244: 182- 188, 1989. [25] J. Ellman, D. Mendel, S. Anthony-Cahill, C. J. Noren, and P. G. Schultz. Biosynthetic method for introducing unnatural amino-acids site-specifically into proteins. Meth. Enzym. , 202:301- 336, 1991. [26] S. A. Robertson, J. A. Ellman, and P. G. Schultz. A general and efficient route for chemical aminoacylation of transfer RNAs. J. Am. Chem. Soc., 113:2722- 2729, 1991. 169 [27] V. W. Cornish, D. Mendel, and P. G. Schultz. Probing protein structure and function with an expanded genetic code. Angew. Chem. Int. Ed. Engl., 34:621633, 1995. [28] M. W. Nowak, P. C. Kearney, J. R. Sampson, M. E. Saks, C. G. Labarca, S. K. Silverman, W. Zhong, J. Thorson, J. N. Abelson, N. Davidson, P. G. Schultz, D. A. Dougherty, and H. A. Lester. Nicotinic receptor binding site probed with unnatural amino-acid incorporation in intact cells. Science, 268:439- 442, 1995. [29] M. E. Saks, J. R. Sampson, M. W. Nowak, P. C. Kearney, F. Du, J. N. Abelson, H. A. Lester, and D. A. Dougherty. An engineered tetrahymena tRNA Gin for in vivo incorporation of unnatural amino acids into proteins by nonsense suppression. J. Biol. Chem., 271:23169-23175, 1996. [30] M. W. Nowak, J. P. Gallivan, S. K. Silverman, C. G. Labarca, D. A. Dougherty, and H. A. Lester. In vivo incorporation of unnatural amino acids into ion channels in a Xenopus oocyte expression system. Meth. Enzymol., 293:504- 529, 1998. [31] A. Karlin. Structure of nicotinic acetylcholine receptors. Curr. Opin. Neurobiol., 3:299-309, 1993. [32] A. Devillers-Thiery, J. L. Galzi, J. L. Eisele, S. Bertrand, D. Bertrand, and J.P. Changeux. Functional architecture of the nicotinic acetylcholine receptor: A prototype of ligand-gated ion channels. J. Membrane Biol., 136:97- 112, 1993. [33] H. A. Lester. The permeation pathway of neurotransmitter-gated ion channels. Annu. Rev. Biophys. Biomol. Struct., 21:267-292, 1992. [34] R.M. Stroud, M.P. McCarthy, and M. Shuster. Nicotinic acetylcholine receptor superfamily of ligand-gated ion channels. Biochemistry, 29:11009-11023, 1990. [35] N. Unwin. Nicotinic acetylcholine receptor at 9 A resolution. 229: 1101- 1124, 1993. J. Mol. Biol. , 170 [36] E. Kubalek, S. Ralston, J. Lindstrom, and N. Unwin. Location of subunits within the acetylcholine receptor by electron image analysis of tubular crystals from Torpedo marmorata. J. Cell Biol., 105:9-18, 1987. [37] S. J. Tzartos, M. T. Cung, P. Demange, H. Loutrari, A. Mamalaki, M. Marraud, I. Papadouli, C. Sakarellos, and V. Tsikaris. The main immunogenic region (MIR) of the nicotinic acetylcholine receptor and the anti-MIR antibodies. Molecul. Neurobiol., 5:1- 29, 1991. [38] A. Mamalaki and S. J. Tzartos. Nicotinic acetylcholine receptor: Structure, function and main immunogenic region. Adv. Neuroimmunol., 4:339-354, 1994. [39] R. Beroukhim and N. Unwin. Three-dimensional location of the main immunogenic region of the acetylcholine receptor. Neuron, 15:323- 331, 1995. [40] M. Quick and H. A. Lester. Methods for Expression of Excitablitity Proteins in Xenopus Oocytes. Academic Press, San Diego, CA, 1994. [41] S. Verrall and Z. W. Hall. The N-terminal domains of acetylcholine receptor subunits contain recognition signals for the initial steps of receptor assembly. Cell, 68:23- 31, 1992. [42] S. Tamamizu, D. H. Butler, J. A. Lasalde, and M. G. McNamee. Effects of antibody binding on structural transitions of the nicotinic acetylcholine receptor. Biochemistry, 35: 11773- 11781, 1996. [43] V. Tsikaris, E. Detsikas, M. Sakarellos-Daitsiotis, C. Sakarellos, E. Vatzaki, S. Tzartos, M. Marraud, and M. T. Cung. Conformational requirements for molecular recognition of acetylcholine receptor main immunogenic region (MIR) analogues by monoclonal anti-MIR antibody: A two-dimensional nuclear magnetic resonance and molecular dynamics approach. Biopolymers, 33:1123- 1134, 1993. 171 [44] E. E. Abola, F. C. Bernstein, S. H. Bryant , T. F. Koetzle, and J. Weng. Protein Data Bank. Data Commission of the International Union of Crystallography, Bonn/Cambridge/Chester, 1987. [45] P. C. Weber, D. H. Ohlendorf, J. J. Wendoloski, and F. R. Salemme. Structural origins of high-affinity biotin binding to streptavidin. Science, 243:85- 88, 1989. [46] K. Fujita and J. Silver. Surprising lability of biotin-streptavidin bond during transcription of biotinylated DNA bound to paramagnetic streptavidin beads. BioTechniques, 14:608- 617, 1993. [47] V.W. Cornish, K.M. Hahn, and P.G. Schultz. Site-specific protein modification using a ketone handle. J. Am. Chem. Soc., 118(34):8150- 8151, 1996. [48] L.K. Mahal, K.J. Yarema, and C.R. Bertozzi. Engineering chemical reactivity on cell surfaces through oligosaccharide biosynthesis. Science, 276(5315):1125- 1128, 1997. [49] E. Saxon and C.R. Bertozzi. Cell surface engineering by a modified Staudinger reaction. Science, 287(5460):2007- 2010, 2000. [50] P. C. Kearney, M. W. Nowak, W. Zhong, S. K. Silverman, H. A. Lester, and D. A. Dougherty. Dose-response relations for unnatural amino acids at the agonist binding site of the nicotinic acetylcholine receptor: Tests with novel side chains and with several agonists. Mal. Pharmacol., 50:1401- 1412, 1996. [51] B. Amit, U. Zehavi, and A. Patchornik. Photosensitive protect- ing groups of amino sugars and their use m glycoside synthesis. 2Nitrobenzyloxycarbonylamino and 6-nitroveratryloxycarbonylamino derivatives. J. Org. Chem. , 39:192- 196, 1974. [52] S. K. Silverman. I. Conformational and Charge Effects on High-Spin Organic Polyradicals; II. Studies on the Chemical-Scale Origin of Ion Selectivity in Potassium Channels. Ph.D. Thesis, California Institute of Technology, 1998. 172 Chapter 7 Other Experiments 173 7.1 Amino Acid Synthesis In this section, the synthesis of a number of natural and unnatural amino acids is outlined. The syntheses generally occur in three stages: 1. The a-NH 2 group and other reactive side chain groups are protected 2. The protected amino acid is activated as the cyanomethyl ester 3. The activated ester is coupled to the dCA-dinucleotide General methods describing these steps are outlined below. For simple amino acids with unreactive side chains, only the a-NH 2 group needs to be protected. Protection of the a-NH 2 group increases the stability of the aminoacyldinucleotide and aminoacyl-tRNAs. Prior to use , the aminoacyl tRNAs must be deprotected. The conditions used to deprotect the amino acid must be mild, such that neither the amino acid, the tRNA, nor the oocytes are damaged. Originally, the o-nitroveratryloxycarbonyl (NVOC) protecting group was used becuase it could be removed under mild conditions using 350 nm light. However, in some applications, it is desirable to have an alternate protecting group for the a-amine. For example, it is possible to produce "caged" amino acids that have side chains that are quickly unmasked at a specified time. Photochemical decaging offers the ability to spatially and temporally reveal new functionality within a protein. However, because the 350 nm used to remove the NVOC a-amine protecing group would likely also remove the side chain caging group, new a-amine protecting groups had to be adapted for use that were compatible with both aminoacyl-tRNAs and the Xenopus oocyte. Two protecting groups were investigated as potential complements to the NVOC group. The 4-pentenoyl group (4-PO), originally developed by Fraser-Reid and co-workers is normally removed by a solution of 12 in aqueous THF [1]. To test whether this protecting group was compatible with the oocyte system, the N-4PO-phenylalanine was synthesized, coupled to the dCA dinucleotide, and ligated to tRNA-THG73(-CA). The protected aminoacyl-tRNA was treated with 1 volume of a saturated solution of 12 in 1 mM NH 4 OAc, pH 4.5, for five minutes, combined with 174 mRNA encoding the nAChR (1.1 ng total/oocyte, 4:1:1:1 ratio of aE180TAG, /3, ,, b subunits), and injected into X enopus oocytes. After 24 hours, whole-cell currents were robust, indicating that the protecting group had been sufficiently removed. In addition to the 4-PO group, the o-nitrophenyl sulfenyl (NPS) group was also investigated [2]. The NPS group is normally removed by reducing reagents, such as sodium thiosulfate. Thus, NFS-phenylalanine was synthesized, coupled to the dCA dinucleotide, and ligated to tRNA-THG73(-CA). The protected aminoacyltRNA (12.5 ng/oocyte) was treated with 20 mM Na2 S2 O3 in 1 mM NH 4 OAc, for 10 minutes, combined with mRNA encoding the nAChR (1.1 ng total/oocyte, 4:1:1:1 ratio of aE180TAG, /3, ,, b subunits), and injected into Xenopus oocytes. After 24 hours, whole-cell currents in response to 200 µM ACh were robust (> 1 µA), indicating that the protecting group had been removed. Since the interior of the oocyte is a reducing environment, it was believed that deprotection prior to injection might be unnecessary. Preliminary results indicate that this is the case. In cases where aminoacyl-tRNA stability is critical, this group may provide a means to keep the tRNA protected (and thus stable) for longer periods of time. However, NPS amino acids are somewhat less attractive synthetically than those protected with either the NVOC or the 4-PO group. 7 .1.1 General Method for the Protection of Amino Acids Scheme 7.1 outlines the syntheses of N-protected amino acids using the NVOC, 4-PO, and NPS protecting groups. For the NVOC and 4-PO groups, Schotten-Baumann conditions are used to install the protecting group. 1 The amino acid (1.0 eq.) and Na 2 CO 3 (2.0 eq.) are dissolved in a 1:1 mixture of water/dioxane. A solution of either NVOC-Cl (1.0 eq.) [3] or 4-PO-anhydride (1.1- 1.3 eq.) [4] in dioxane is added and the mixture is stirred until TLC indicates that the reaction is complete (usually 1-4 hours). Reactions involving 4-PO-anhydride may be sluggish and additional 4PO-anhydride may be added. When the reaction is judged to be complete by TLC, 1 For a procedure to install the NPS group, see Section 7.2.1. 175 CH 2 Cb is added, followed by an equal amount of 1 N NaHSO 4 and the aqueous layer is extracted three times with CH 2 Cb. The organic extracts are combined, dried over Na2 SO 4 , and concentrated in vacuo. The crude products are purified by flash chromatography on silica. In many cases, a mixture of hexane/ethyl acetate will elute the byproducts, and switching to ethyl acetate with 1% acetic acid will elute the protected amino acid. NVOC-Cl +~o~ 0 0 0 Na2C03 H20 /Dioxane R ~N~OH H 0 4-PO-anhydride NaOH NPS-Cl Scheme 7 .1: Protection of amino acids. 7.1.2 General Method for the Synthesis of Amino Acid Cyanomethyl Esters The protected amino acid (~ 0.5 mmol, 1.0 eq.) is dissolved in 2 mL dry DMF under argon (Scheme 7.2). Triethylamine (3 eq.) is added, followed by 2 mL chloroace- 176 tonitrile. The reactions are stirred under argon for 1-24 hours and the volatiles are removed in vacuo. The products are purified by flash chromatography. Useful solvent systems are CH2 Cb , or 1:1 hexane/ethyl acetate. A PG,N~O..___,...,CN H 0 Scheme 7.2: Cyanomethylation of protected amino acids. 7.1.3 General Method for Coupling of N-Protected Amino Acids to dCA This procedure is adapted from previously published procedures [5- 7]. The N- protected amino acid (~ 30 µmol , 3 eq.) is mixed with the tetra-n-butylammonium salt of the dCA dinucleotide (~ 10 µmol, 1 eq.) in dry DMF (400 µL) . The reactions are monitored by analytical HPLC with a linear gradient of 100% 25 mM aqueous ammonium acetate buffer (pH 4.5) to 100% acetonitrile over one hour. Following the disappearance of the dCA (5 min. to 3 hours) , the reaction mixture is separated using semi-preparative HPLC with a linear gradient of 100% 25 mM aqueous ammonium acetate buffer (pH 4.5) to 100% acetonitrile. The fractions containing the aminoacyl dinucleotide are collected and lyophilized. The lyophilized solid is then redissolved in 10 mM aqueous acetic acid/ acetonitrile and lyophilized a second time to remove salts. The products are characterized by UV-Vis and mass spectrometry. NH~ C -6 7,,0~ 0 I dC N~o -o-P-o A )__ PG-NH' .OCH2CN '](" + 0 protected amino acid cyanomethyl ester JVVVVVVVV' ~ N~ 9 Ni: -o-~=o f I N 0 N N_) H 0 0 ---........., A 7,,o~ H HO OH •2 n-Bu N+ 4 DMF H A l("'NH-PG 0 dCA-amino acid Scheme 7.3: Synthesis of dCA amino acids. 177 7.2 Synthesis of Natural Amino Acids 7.2.1 dCA-NPS-Phenylalanine NPS-Phenylalanine Phenylalanine (330 mg, 2.0 mmol) was dissolved in dioxane (20 mL) , water (5 mL) and 1.14 M NaOH (1.75 mL, 1 eq). Subsequently, 2-nitrophenylsulfenyl chloride (417 mg, 2.2 mmol, 1.1 eq) was added in small portions, turning the yellow mixture dark green. Addition of 1.93 mL of 1.14 M NaOH returned the mixture to its original yellow color. After 30 minutes, CH 2Cb (25 mL) was added, followed by NaHSO 4 (1 N, 25 mL). The mixture was extracted 3 times with CH 2Cb (25 mL). The combined organic layers were dried over Mg8O 4 , filtered and the solvent was evaporated to give a yellow solid. The crude product was purified by flash chromatography (silica, 50:50 hexane/ethyl acetate followed by 1% HOAc in ethyl acetate) to give the protected amino acid as a yellow solid (503 mg, 79% yield): 1 H NMR (300 MHz, CDCb) 6 11.35 (br s, 1 H, COOH), 8.21 (dd, J = 7.0, 1.2 Hz, 1 H, NPS-3-H), 7.44- 7.20 (m, 7 H) , 7.17 (td, J = 5.5, 1.5 Hz, 1 H, Phe-4-H) , 3.80 (m, 1 H, a-CH), 3.34 (dd, J =14.0, 4.0 Hz, 1 H, side chain CH), 3.00 (dd, J = 14.0, 4.0 Hz, 1 H, side chain CH), 3.20 (br s, 1 H) ; 13 C NMR (75 MHz , CDC1 3 ) b 178.9 COOH, 144.3 NPS-C2, 142.5 NPS-Cl, 136.4 Phe-Cl, 133.8 NPS-C5, 129.5 Phe-C2,C6, 128.8 Phe-C3 ,5, 127.3 NPS-C6, 125.6 Phe-C4, 124.7 NPS-C3, 124.4 NPS-C4, 65.5 a-CH, 39.6 side chain CH 2; HRMS er+ (NH 3 ) [M-H]+ 319.0757, calcd for C 15 H14 N2O 4 S 319.0753. 178 NPS-Phenyalanine Cyanomethyl Ester NFS-phenylalanine (100 mg, 0.31 mmol) was dissolved in 2 mL dry DMF under argon. Triethylamine (0.13 mL, 3 eq.) was added, followed by chloroacetonitrile (2 mL). After 5 hours, the volatiles were removed in vacuo to give a yellow oil. The product was purified by flash chromatography (silica, hexane, 80:20 hexane/ethyl acetate, 50:50 hexane/ethyl acetate) to give a yellow oil (104 mg, 93%): 1 H NMR (300 MHz, CDC13) 8 8.18 (dd, J = 7.0, 1.0 Hz, 1 H, NPS-3-H), 7.45- 7.20 (m, 7 H), 7.18 (td, J = 5.6, 1.2 Hz, 1 H, Phe-H-4), 4.82 (dd, J = 15.6, 12.0 Hz, 2 H, OCH2), 3.82 (m, 1 H, a-CH), 3.35-3.25 (overlapping peaks, 2 H, side chain CH, and NH), 3.03 (m, 1 h, side chain CH); 13 C NMR (75 MHz, CDC1 3) 8 171.8 COOR, 143.8 NPS-C2, 142.4 NPS-Cl, 135.7 Phe-Cl, 133.8 NPS-C5, 129.4 Phe-C2,C6, 128.8 Phe-C3,5, 127.3 NPS-C6, 125.6 Phe-C4, 124.8 NPS-C3, 124.4 NPS-C4, 113.80 CN, 65.4 a-CH, 48.9 OCH 2, 39.5 side chain CH 2; HRMS EI+ [M-H]+ 357.0779, calcd for C 17 H15 N3O4S 357.0783. d CA-NPS-Phenylalanine Prepared as above: FAB-MS [M-HJ- 935, calcd for C34H3sN10O15P2S 935; UV >-max (10 mM Aq. HOAc) 251 nm, 384 nm. 7.2.2 dCA-4-PO-Phenylalanine 4-P O-Pheny lalanine A solution of phenylalanine (100 mg, 0.6 mmol) and sodium carbonate (158 mg, 0.6 mmol) in 15 mL water and 7 mL dioxane was cooled in an ice bath. A solution of 179 4-pentenoic anhydride (130 mg, 0.72 mmol, 1.2 eq) in dioxane was added dropwise. After 24 hours, 50 mL 1 N NaHSO 4 was added, followed by 40 mL CH 2 Cb. The mixture was extracted three times with 40 mL portions of CH 2 Cb. The organic layers were combined, dried over MgSO 4 and concentrated in vacuo to give 175 mg of crude product that crystallized in an NMR tube. The remainder was used without further purification: 1 H NMR (300 MHz, CDCh) 5 10.36 (s, 1 H, COOH), 7.25 (m, 3 H, Ar-H), 7.16 (dd, J = 6.2, 1.7 Hz, 2 H, Ar-H), 6.50 (d J = 7.7 Hz, 1 H, NH), 5.90-5.65 (m, 1 H, CH 2 =CH), 5.12- 4.86 (m, 3 H, CH 2 =CH and a -CH), 3.16 (dd, J = 6.0, 5.5 Hz., 2 H, side chain CH 2 ), 2.47-2.26 (m, 4 H, CH 2 's in 4-PO group); 13 C NMR (75 MHz, CDCh) 5 178.4 COOH, 174.7 CONH, 136.5 CH 2 =CH, 135.7 Ar-Cl, 129.4 Ar-C2 ,6, 128.5 Ar-C3,5, 127.1 Ar-C4, 115.8 CH 2 =CH, 53.2 a-CH, 37.2 side chain CH 2 , 35.4, 29.4 CH 2 's in 4-PO group. 4-PO-Phenylalanine Cyanomethyl Ester To a solution of 4-pentenoyl phenylalanine (59 mg, 0.24 mmol) in 2 mL dry DMF was added triethylamine (100 µL, 0.72 mmol, 3 eq.) and 2 mL chloroacetonitrile. After 10 hours, the volatiles were removed in vacuo to afford a crude white residue that was purified by flash chromatography (silica, hexane--+3:1 hexane/ethyl acetate--+1:1 hexane/ethyl acetate) to afford 51 mg of a white solid (75% yield): 1 H NMR (300 MHz, CDC1 3 ) 5 7.36-7.25 (m, 3 H, Ar-H), 7.14 (dd, J = 6.2, 1.6 Hz, 2 H, Ar-H), 6.07 (d J = 7.5 Hz, 1 H, NH), 5.85- 5.68 (m, 1 H, CH 2=CH), 5.12-5.04 (m, 3 H, CH 2=CH and a -CH) , 4.91 (dd, J = 32.3, 15.7 Hz., 2 H, OCH2CN), 3.20- 3.06 (m, 2 H, side chain CH 2) 2.47-2.26 (m, 4 H, CH 2's in 4-PO group); 13 C NMR (75 MHz, CDC13) 5 178.4 COOH, 174.7 CONH, 136.5 CH2=CH, 135.7 Ar-Cl, 129.4 Ar-C2,6, 128.5 Ar-C3,5, 127.1 Ar-C4, 115.8 CH 2=CH, 53.2 a-CH , 37.2 side chain CH 2, 35.4, 29.4 CH 2's in 4-PO group. dCA-4-PO-Phenylalanine Prepared as above: FAB-MS [M-HJ- 864.3, calcd for C33 H41 NgO 15 P2 865.2. 180 7.2.3 dCA-NVOC-Glycine NVOC-G lycine Prepared in 85% yield as above starting with glycine (75 mg, 1 mmol), NVOC-Cl (276 mg, 1 mmol) and (212 mg, 2 mmol): 1H NMR (300 MHz, DMSO-d6) b 13.25- 12.25 (br s, 1 H, COOH), 7.85 (t, J = 6.2 Hz, 1 H, NH), 7.68 (s, 1 H, Ar-H), 7.17 (s, 1 H, Ar-H), 5.35 (s, 2 H, NVOC-CH 2), 3.92 (s, 3 H, CH 3O), 3.85 (s, 3 H, CH 3O), 3.69 (d, J = 6.2 Hz, 2 H, a-CH 2); 13 C NMR (75 MHz, DMSO-d6) b 172.1 COOH, 156.7 NHCO, 153.96 Ar-C, 148.08 Ar-C, 139.44 Ar-C, 128.65 Ar-C, 110.46 Ar-CH, 108.52 Ar-CH, 63.05 NVOC-CH 2, 56.65 CH 3O, 56.54 CH 3O, 42.60 a-CH 2; HRMS er+ (NH3) [M-NH 4]+ 332.1107, calcd for C12H1 4N2Os + NH3 332.1094. NVOC-Glycine Cyanomethyl Ester NVOC-Glycine (100 mg, 0.32 mmol) was dissolved in 2 mL dry DMF. Triethylamine (135 µL, 0.96 mmol, 3 eq.) was added, followed by 2 mL chloroacetonitrile. After 6 hours , the volatiles were removed in vacuo and the crude product was purified by flash chromatography (silica, 1:1 hexane/ethyl acetate) to give 112 mg of a pale yellow solid (quantitative yield): 1H NMR (300 MHz, DMSO-d6) b 8.07 (t, 1 H, J = 6.2 Hz, NH), 7.67 (s, 1 H, Ar-H), 7.16 (s, 1 H, Ar-H), 5.36 (s, 2 H, NVOC-CH 2), 3.99 (d, J =6.0 Hz, 2 H, a-CH 2), 3.91 (s, 3 H, CH 3O), 3.85 (s, 3 H, CH 3O); 13 C NMR (75 MHz, DMSO-d6) b 170.0 COOR, 156.6 NHCO, 153.9 Ar-C, 148.2 Ar-C, 139.6 Ar-C, 128.2 Ar-C, 116.3 CN, 110.7 Ar-CH, 108.5 Ar-CH, 63.3 NVOC-CH 2, 56.7 CH 3O, 56.5 CH 3O, 49.9 OCH 2CN, 42.3 a- CH 2; HRMS Er+ [M-H]+ 353.0852, calcd for C14H1sN3Os 353.0859. 181 dCA-NVOC-Glycine Prepared as above: FAB-MS [M-HJ- 932.2, calcd for C31 H38 N10 O 20 P 2 932.17. 7.2.4 dCA-NVOC-Alanine NVOC-Alanine The reaction was performed as above, and purified by flash chromatography (silica, CH 2Cb--+5:1 CH 2Cb/ethyl acetate--+1:1 CH 2Cb/ethyl acetate--+ethyl acetate --+1% acetic acid in ethyl acetate). The product was obtained as an orange/yellow solid in 47% yield: 1H NMR (300 MHz, CD 3OD) b 7.73 (s, 1 H, Ar-H) , 7.19 (s, 1 H, Ar-H), 5.52 (AB, J = 15.1 Hz , 1 H, NVOC-CH), 5.40 (AB, J = 15.1 Hz, 1 H, NVOC-CH) , 4.20 (m, 1 H, a -CH) , 3.97 (s, 3 H, CH 3O), 3.90 (s, 3 H, CH 3O), 1.42 (d, J = 7.5 Hz, 3 H, CH 3); 13 C NMR (75 MHz, CD 3OD) b 176.8 COOR, 158.4 CONH, 155.8 Ar-C , 149.9 Ar-C , 141.0 Ar-C , 130.3 Ar-C , 111.1 Ar-CH, 109.7 Ar-CH, 64.9 NVOCCH 2, 57.4, 57.2 CH 3O's, 51.3 a-CH, 18.2 CH 3; HRMS EI+ [M-H] + 328.0914, calcd for C13H16N2Os 328.0907. NVOC-Alanine Cyanomethyl Ester NVOC-alanine (202 mg, 0.62 mmol) was dissolved in 2 mL dry DMF. Triethylamine (250 µL , 1.8 mmol, 3 eq.) was added, followed by 2 mL chloroacetonitrile. After 1.5 hours, the volatiles were removed in vacuo and the crude product was purified by flash chromatography (silica, CH 2Cb--+95:5 CH 2Cb/ethyl acetate) to give 189 mg of pale yellow crystals (84% yield): 1H NMR (300 MHz, CDC1 3) b 7.68 (s, 1 H, Ar-H) , 6.98 (s, 1 H, Ar-H), 5.55 (m, 1 H, NH), 5.53 (AB, J = 15.1 Hz, 1 H, NVOC-CH) , 182 5.45 (AB, J = 15.1 Hz, 1 H, NVOC-CH), 4.79 (m, 2 H, OCH 2CN), 4.20 (m, 1 H, a-CH), 3.99 (s, 3 H, CH 3O), 3.93 (s, 3 H, CH 3O), 1.48 (d, J = 7.5 Hz, 3 H, CH 3); 13 C NMR (75 MHz, CDC1 3) fJ 171.8 COOR, 155.4 CONH, 153.8 Ar-C, 148.2 Ar-C, 139.6 Ar-C, 128.9 Ar-C, 114.1 CN, 109.9 Ar-CH, 108.2 Ar-CH, 64.1 NVOC-CH 2, 56.6, 56.5 CH3O's, 49.6 a-CH, 49.2 OCH 2CN, 17.9 CH 3; HRMS EI+ [M-H]+ 367.1026, calcd for C1sH17N3Os 367.1016. dCA-NVOC-Alanine Prepared as above: FAB-MS [M-HJ- 945, calcd for C32H 4oN 10 O 20 P 2 945; UV >-max (10 mM Aq. HOAc) 249 nm, 265 nm (sh), 350 nm. 7.2.5 dCA-NVOC-Proline NVOC-Proline The reaction was performed as above. The crude product was purified by flash chromatography (silica, 1:1 hexane/ethyl acetate--+1% acetic acid in ethyl acetate). Product obtained as a foamy yellow solid in 85% yield. The 1H NMR spectrum shows 2 poorly resolved conformational isomers: 1H NMR (300 MHz, CD 3OD) fJ 7.59 (s br, 1 H), (7.00,6.99*) (s, 1 H) , H2O obscures 5.8- 5.5 ppm, 5.27 (d, J = 23 Hz, 1 H) , 5.22 (d, J = 23 Hz, 1 H), (4.48,4.44)(dd, J = 8.0, 3.5 Hz, 1 H), 4.36 (dd, J = 8.0, 3.5 Hz, 1 H), (3.92,3.91 *) (s, 3 H) , (3.84,3.83*) (s, 3 H) , 3.65- 3.40 (m, 2 H), 2.42-1.90 (m - br); 13 C NMR (75 MHz, CD 3OD) fJ (175.9, 175.2*), (155.9, 155.6*), (155.3, 155.0*) , (149.2, 149.0*), (140.2, 139.9*), (110.3, 110.0*), (108.9, 108.8*), (65.3, 65.1 *), (60.5, 60.1 *), 57.25, (56.8 (2)), (56.7 (2)), (48.2, 47.8*), (31.9, 30.9*), (25.2, 24.3*). 183 NVOC-Proline Cyanomethyl Ester The reaction was performed as above. The crude product was purified by flash chromatography (silica, 1:1 hexane/ethyl acetate). Product obtained as a yellow solid in 80% yield. The 1H NMR spectrum shows 2 poorly resolved conformational isomers: 1H NMR (300 MHz, CDC1 3) 6 (7.71, 7.69*) (s, 1 H), (7.01, 6.98*) (s, 1 H), 5.77-5.38 (m, 2 H) , 4.93- 4.68 (m, 2 H) , 4.56- 4.42 (m, 1 H), (4.04, 4.02*) (s, 3 H) , (3.96, 3.95*) (s, 3 H), 3.70- 3.57 (m, 2 H), 2.38-2.34 (m, 1 H), 2.19- 1.90 (m, 3 H + EtOAc); 13 C NMR (75 MHz, CDCh) 5 (171.4, 171.1*) , 154.3, (153.7, 153.6*) , (148.1, 148.0*) , (139.6, 139.5*), (128.1, 127.6*), (114.1, 114.0*) , (110.8, 109.6*), 108.1, 102.0, (64.7, 64.3*), (58.8, 58.4), (56.7, 56.4), 49.0, (47.1, 46.3*), (30.9, 29.7*), (24.4, 23.5*). dCA-NVOC-Proline Prepared as above: FAB-MS [M-H]+ 973.3, calcd for C34 H42 N 10 O 20 P2 972.2. 7.2.6 NVOC-Glutamine Cyanomethyl Ester NVOC-Glutamine The reaction was conducted as above. After addition of CH 2Cb and 1 N NaHSO 4, a pale yellow precipitate formed. This was filtered to afford the product in 83% yield; FAB HRMS [M-H]+ 386.1182, calcd for C15H19N3Og 386.1200. 184 NVOC-Glutamine Cyanomethyl Ester NVOC-Glutamine (50.8 mg, 0.13 mmol) was dissolved in 2 mL dry DMF. Triethylamine (55 µL , 0.39 mmol, 3 eq.) was added, followed by 2 mL chloroacetonitrile. After three hours the volatiles were removed in vacuo and the residue was purified by flash chromatography (silica, hexane-+ 1:1 hexane/ethyl acetate-+ethyl acetate) to afford 24 mg of an off-white solid (43% yield): 1 H NMR (300 MHz, DMSO-d6 ) 5 8.15 (d, J = 8 Hz, 1 H, a -NH) , 7.65 (s, 1 H, Ar-H) , 7.18 (s, 1 H, Ar-H), 7.35 (s, 1 H, NH) , 6.80 (s, 1 H, NH), 5.38 (m, 2 H, NVOC-CH 2 ), 5.00 (s, 2 H, OCH 2 CN) , 4.15 (m, 1 H, a -CH) , 3.91 (s, 3 H, CH 3 O), 3.86 (s , 3 H, CH 3 O) , 2.18 (m, 2 H, 1-CH 2 ), 1.97- 1.79 (m, 2 H, 5-CH 2 ); 13 C NMR (75 MHz, DMSO-d6 ) 5 173.6 COOR, 171.9 CONH 2 , 156.2 CONH, 153.9 Ar-C , 148.2 Ar-C, 139.5 Ar-C , 128.2 Ar-C , 116.3 CN, 110.7 Ar-CH , 108.4 Ar-CH , 63.2 NVOC-CH 2 , 56.7, 56.6 CH 3 O's, 53.7 a-CH, 50.1 OCH 2 CN, 31.3 1-CH 2 , 26.7 5-CH 2 ; FAB HRMS M-Na+ 447.1123, calcd for C 17 H20N 4 O9Na 447.1128. 7.2.7 (NVOC)2-Lysine Cyanomethyl Ester 0-......,...--CN 0 (NVOC)2-Lysine (NVOC)2-Lysine was prepared as above using 250 mg (1.7 mmol) L-lysine, 943 mg (3.4 mmol , 2 eq.) NVOC-Cl and 360 mg (3.4 mmol, 2 eq.) sodium carbonate. Purified by flash chromatography (silica, 1:1 hexane/ethyl acetate-+ 1% acetic acid in ethyl acetate) to give 830 mg of a pale orange solid (78% yield): 1 H NMR (300 MHz, 185 DMSO-d6) b 7.81 (d, J = 8.1 Hz, 1 H, a -NH) , 7.67 (s, 1 H, Ar-H) , 7.65 (s, 1 H, Ar-H) , 7.43 (m, 1 H, c-NH), 7.14 (s, 1 H, Ar-H) , 7.15 (s, 1 H, Ar-H) , 5.42 (m, 4 H, NVOCCH2 's) , 3.91 (s, 3 H, CH 3O) , 3.89 (s, 3 H, CH 3O) , 3.86 (s, 3 H, CH 3O) , 3.85 (s, 3 H, CH 3O), 2.98 (m, 2 H, ,8-CH 2), 1.80- 1.25 (m, 6 H, side chain CH 2's); 13 C NMR (75 MHz, DMSO-d6) b 173. 7 COOH, 155.58, 155.67 CONH, 153.45, 153.26 Ar-C, 147.46, 147.59 Ar-C, 138.94, 139.27 Ar-C, 127.82, 128.11 Ar-C , 110.54, 109.99 Ar-CH , 108.11 (2) Ar-CH , 62.29 , 62.12 NVOC-CH 2, 56.12, 56.02 CH 3O's, 53.74 a-CH , 39.5 (within solvent peak) , 30.37, 28.82, 22.77 side chain CH 2's; FAB HRMS [M-H]+ 625.1964, calcd for C26H32N4O14 625.1993. (NVOC)2-Lysine Cyanomethyl Ester (NVOC)2-Lysine (103 mg, 0.16 mmol) was dissolved in dry DMF (1 mL). Triethylamine (67 µL , 0.48 mmol, 3 eq.) was added, followed by 1 mL chloroacetonitrile. After one hour the volatiles were removed in vacuo and the remaining yellow solid was purified by flash chromatography (silica, CH 2Cb --,5:1 CH 2Cl2/ethyl acetate--,ethyl acetate) to afford 106 mg of a pale yellow solid ( 100% yield): 1H NMR (300 MHz, CDC1 3) b 7.60 (s , 1 H, Ar-H) , 7.58 (s, 1 H, Ar-H) , 6.94 (2 s overlap, 2 H, Ar-H), 5.95 (s, 1 H, a-NH) , 5.38 (m, 3 H, NVOC-CH 2 and c-NH), 4.76 (m, 2 H, OCH 2CN) , 4.35 (m, 1 H, a-CH), 3.93, 3.91, 3.88 (2 overlapping) (s , 3 H, CH 3O), 3.17 (m, 2 H, ,8-CH 2), 1.9- 1.35 (m, 6 H, side chain CH 2's) ; 13 C NMR (75 MHz, CDCb) b 170.89 COOR, 156.03, 156.2 CONH, 153.27, 153.45 Ar-C, 147.9 (2) Ar-C , 139.57, 139.41 Ar-C , 127.58, 127.42 Ar-C , 113.80 CN , 110.26, 109.92 Ar-CH , 107.96 (2) Ar-CH , 63 .69, 63.31 NVOC-CH 2, 56.24 (2) , 56.11 (2) CH 3O's, 53.41 a-CH , 48.74 OCH2CN, 39.93, 30.77, 29.03 , 21.91 side chain CH 2's; FAB HRMS [M-H]+ 664.2079, calcd for C2sH33N5O14 664.2102. 186 7.2.8 NVOC-Methionine Cyanomethyl Ester NVOC-Methionine NVOC-Methionine was prepared in 98% yield, as above, starting with L-methionine (50 mg, 0.34 mmol): 1 H NMR (300 MHz, DMSO-d6 ) fJ 7.88 (d, J = 8.1 Hz, 1 H), 7.67 (s, 1 H) , 7.14 (s , 1 H), 5.34 (dd, J = 15.0 Hz, 2 H), 4.18 (m , 1 H) , 3.90 (s , 3 H), 3.84 (s, 3 H) , 2.50 (m, 2 H) , 2.05 (s, 3 H) , 1.98 (m, 2 H); 13 C NMR (75 MHz , DMSO-d6 ) fJ 173.7, 155.9, 153.5, 147.7, 138.0, 128.3, 109.1, 62.5, 56.2, 56.1, 52.8, 30.3, 29.8, 14.5. NVOC-Methionine Cyanomethyl Ester NVOC-Methionine cyanomethyl ester was prepared in 82% yield, as above, starting with NVOC-L-methionine (32 mg, 0.08 mmol): 1 H NMR (300 MHz, CDCh) fJ 7.75 (s, 1 H), 7.00 (s, 1 H) , 5.65 (d, J = 15.0 Hz , 1 H) , 5.53 (dd, J = 32.0 Hz, J = 15.1 Hz, 2 H) , 4.80 (dd, J = 32.0 Hz, J = 15.1 Hz, 2 H), 4.63 (m, 1 H), 4.01 (s, 3 H), 3.96 (s, 3 H), 2.60 (m, 2 H) , 2.30- 2.10 (m, 2 H), 2.12 (s, 3 H). 187 7.3 Synthesis of Unnatural Amino Acids 7.3.1 dCA-NVOC-Allyl-Glycine NVOC-Allyl-Glycine Allyl-glycine (100 mg, 0.87 mmol) and one equivalent of sodium carbonate (184 mg, 0.87 mmol) were dissolved in 15 mL water and 8 mL dioxane. A solution of one equivalent of NVOC-Cl (240 mg, 0.87 mmol) in 10 mL dioxane was added dropwise and the mixture was allowed to stir. After one hour, CH Cb (25 mL) was added, 2 followed by 1 N NaHSO 4 (25 mL). The aqueous layer was extracted three times with 25 mL CH Cl 2 The organic extracts were combined, dried over MgSO . 2 , 4 filtered and concentrated in vacuo to give a pale yellow solid. The crude product was purified by flash chromatography (silica, 50:50 hexane/ethyl acetate ----d% acetic acid in ethyl acetate) to give 303 mg of a pale yellow solid (94% yield): 1 H NMR (300 MHz, DMSOd6 ) <5 7.89 (d, 1 H, J = 8.2 Hz, NH), 7.68 (s, 1 H, Ar-H), 7.15 (s, 1 H, Ar-H), 5.78 (dt, J = 16.8, 9.9 , 6.8 Hz, 1 H, CH 2 =CH) , 5.38 (AB, J = 9.0 Hz , 1 H, NVOC-CH), 5.32 (AB, J = 9.0 Hz, 1 H, NVOC-CH) , 5.13- 5.02 (m, 2 H, CH 2 =CH), 4.03 (m, 1 H, a-CH) , 3.90 (s, 3 H, CH 3 O), 3.84 (s, 3 H, CH 3 O), 2.37 (m, 2 H, side chain CH 2 ); 13 C NMR (75 MHz, DMSO-d ) 6 <5 173.3 COOH, 155.8 CONH, 153.6 Ar-C, 147.7 Ar-C, 139.0 Ar-C, 134.3 CH =CH, 128.4 Ar-C, 117.8 CH =CH, 109.9 Ar-CH, 108.1 Ar-CH, 2 62.5 NVOC-CH , 2 2 56.3, 56.1 CH O's, 53.8 a-CH, 35.2 side chain CH 3 . 2 NVOC-Allyl-Glycine Cyanomethyl Ester NVOC-Allyl-glycine (100 mg, 0.27 mmol) was dissolved in 2 mL dry DMF. Triethylamine (113 µL , 0.81 mmol, 3 eq.) was added , followed by 2 mL chloroacetonitrile. 188 After 6 hours, the volatiles were removed in vacuo and the crude product was purified by flash chromatography (silica, hexane---+8:2 hexane/ethyl acetate---+1:1 hexane/ethyl acetate) to give 106 mg of a pale yellow oil (93% yield): 1 H NMR (300 MHz, CDC1 3 ) 8 7.20 (s, 1 H, Ar-H), 6.99 (s, 1 H, Ar- H), 5.71 (m, 1 H, CH 2=CH), 5.55 (AB, J = 15.0 Hz, 1 H, NVOC-CH), 5.47 (AB, J = 15.0 Hz, 1 H, NVOC-CH), 5.25- 5.19 (m, 3 H, CH 2=CH and NH) , 4.75 (dd, J = 20.7, 15.7 Hz, 2 H, OCH 2CN), 4.53 (m, 1 H, a-CH), 4.00 (s, 3 H, CH 3 O), 3.95 (s, 3 H, CH 3 O) , 2.61 (m, 2 H, side chain CH 2); 13 C NMR (75 MHz, CDC1 3 ) 8 170.6 COOR, 155.4 CONH, 153.7 Ar-C, 148.1 Ar-C, 139.5 Ar-C, 131.5 CH2=CH, 127.8 Ar-C, 120.4 CH 2=CH, 114.0 CN, 109.7 Ar-CH, 108.1 Ar-CH, 64.0 NVOC-CH 2, 56.5, 56.4 CH 3 O's, 53.1 a -CH, 49.1 OCH 2CN, 36.0 side chain CH 2. 7.3.2 NVOC-,6-(2-Thienyl)-L-Alanine Cyanomethyl Ester NVOC-,6-(2-Thienyl)-L-Alanine NVOC-,6-(2-Thienyl)-L-alanine was prepared as above using 35 mg (0.2 mmol) ,6(2-Thienyl)-L-alanine, 56 mg (0.2 mmol, 1 eq.) NVOC-Cl and 43 mg (0.4 mmol, 2 eq.) sodium carbonate. Purified by flash chromatography (silica, hexane---+7:3 hexane/ethyl acetate---+1:1 hexane/ethyl acetate---+1% acetic acid in ethyl acetate) to give 66 mg of pale yellow crystals (80% yield): 1 H NMR (300 MHz, DMSO-d6 ) 8 7.98 (d, J = 8.4 Hz, 1 H, a-NH), 7.66 (s, 1 H, Ar-H), 7.31 (t , J = 3.1 Hz, 1 H, thiophene-H) , 7.11 (s, 1 H, Ar-H) , 6.92- 6.90 (br d , 2 H, thiophene-H), 5.36 (AB, J = 13.0 Hz , 1 H, NVOC-CH), 5.30 (AB, J = 13.0 Hz, 1 H, NVOC-CH), 4.17 (m, 1 189 H, a-CH), 3.86 (s, 3 H, CH 3O), 3.82 (s, 3 H, CH 3O), 3.40-3.05 (m, 2 H, side chain CH ); 13 C NMR (75 MHz, DMSO-d ) 8 172.5 COOH, 155.6 CONH, 153.5 Ar-C, 147.7 2 6 Ar-C, 139.6 thiophene 2-C, 139.0 Ar-C, 128.0, 126.7, 126.1, 124.5 thiophene-C's and Ar-C, 110.0 Ar-CH, 108.2 Ar-CH, 62.3 NVOC-CH 2 , 56.16, 56.06 CH 3O's, 55.5 a-CH, 30.68 side chain CH 2 ; HRMS er+ (NH 3) [M-NH 4 ]+ 428.1145, calcd for C 17 H18 N2 O8S + NH 3 428.1128. NVOC-;3-(2-Thienyl)-L-Alanine Cyanomethyl Ester NVOC-;3-(2-Thienyl)-L-alanine (60 mg, 0.15 mmol) was dissolved in 2 mL dry DMF. Triethylamine (61 µL, 0.44 mmol, 3 eq.) was added, followed by 2 mL chloroacetonitrile. After 9 hours , the volatiles were removed in vacuo and the crude product was purified by flash chromatography (silica, hexane---+7:3 hexane/ethyl acetate---+1:1 hexane/ethyl acetate) to give 61 mg of pale yellow crystals (93% yield): 1H NMR (300 MHz, CDC1 3) 8 7.71 (s, 1 H, Ar-H), 7.21 (dd, J = 5.l, 0.75 Hz, 1 H, thiophene-H), 6.97 (overlapping s, 2 H, Ar-H, thiophene-H), 6.87 (d J = 3.5 Hz, 1 H, thiophene-H), 5.58 (AB, J = 6.5 Hz, 1 H, NVOC-CH), 5.52 (AB, J = 6.5 Hz, 1 H, NVOC-CH), 5.55 (m, 1 H, a-NH) , 4.87- 4.73 (m, 3 H, a-CH, CH 2 CN), 3.97 (s, 3 H, CH 3O), 3.95 (s, 3 H, CH 3O), 3.42 (d, J = 6.5 Hz, 2 H, side chain CH 2); 13 C NMR (75 MHz, CDCb) 8 169.7 COOR, 155.1 CONH, 153.6 Ar-C, 148.0 Ar-C, 139.5 Ar-C, 136.0 thiophene C2, 127.6, 127.3, 127.2, 125.3 thiophene-C's and Ar-C, 113.6 CN, 109.7 Ar-CH, 108.0 Ar-CH, 64.0 NVOC-CH 2 , 56.4, 56.3 CH 3O's, 54.5 a-CH, 49.1 CH 2 CN, 31.8 side chain CH 2 ; HRMS er+ (NH 3) [M-NH 4 ]+ 467.1225, calcd for C 18 H19 N3OsS + NH3 467.1237. 190 7.4 Synthesis of Photoreactive Amino Acids 7.4.1 dCA-N-(4-P0)-3-[p-[3-(trifluoromethyl)-diazirin-3yl]phenyl]-DL-alanine 0 ~N H 0 An overview of this synthesis is shown in Scheme 7.4. The various steps are adapted from several published procedures [8- 10]. 2,2,2-Trifluoro-1-( 4-tolyl)ethanone (Compound 1) Compound 1 was prepared with slight modification to the procedure of Platz et al. [10]. t-Butyllithium (100 mL, 1.7 Min pentane, 2.05 eq.) was cannulated into a dry three-neck 500 mL round-bottom flask at -78 °C. After 30 minutes, 4-bromotoluene (10.3 mL, 0.083 mol, 1.0 eq.) in 125 mL of dry THF was cannulated into the flask over a period of 15 minutes. After one hour , ethyl trifluoroacetate (19 mL , 0.16 mols, 1.9 eq.) was added dropwise in 0.5 mL portions. The mixture was allowed to warm in the CO 2 / acetone bath for 14 hours , after which time it was quenched with water (25 mL) , saturated ammonium chloride solution (15 mL) and an additional 100 mL water. The mixture was extracted four times with 125 mL portions of ether, the organics were combined, dried over Mg8O 4 , and concentrated in vacuo to give 20.5 grams of an orange liquid that was used without further purification. 1 H NMR of the crude product indicated a 54:46 ratio of desired product to starting material. 2,2,2-Trifluoro-1-(4-tolyl)ethanone Oxime (Compound 2) To the crude mixture of 2,2,2-trifluoro-1-(4-tolyl)ethanone in a 500 mL round-bottom flask was added hydroxylamine-HCl (9.38 g, 0.14 mol , 3 eq. based on the amount 191 N'OH 0 f} Br h' t-BuLi ffCF3 NH20H•HCI TsCI ffCF3 CF3C02Et Et3N 1 2 N=N ~ )l) CF3 4 3 NBS/AIBN 5 NaOEt, l)AcOH 2)NaOH 3) 6 ~ 0 0 ~,Na2C03 0 7 DMF Scheme 7.4: Synthesis of dCA-N-( 4-PO)-3-[p--[3-(trifluoromethyl)-diazirin-3- yl]phenyl]-DL-alanine. of starting material determined by NMR) and NaOH (5.52 g, 137 mmol, 3 eq.) in absolute ethanol (250 mL). After refluxing for 28 hours , the mixture was concentrated to a volume of 50 mL. Water (100 mL) was added and the mixture was brought to pH 1 with 3 M HCL The mixture was extracted four times with ether (100 mL portions) , the organic layers were combined, dried over MgSO 4 , and concentrated to give an orange oil. The crude product was purified by flash chromatography (silica, 192 CH2Cb eluent) to give white crystals conatminated with a small amount of an orange impurity that was removed by recrystallization from hexane to give a white solid (8.00 g) in 88% yield, 47% for 2 steps: 1H NMR (300 MHz, CDC1 3) 8 9.85 (s, 1 H, NOH), 7.47 (d, J = 8.l Hz, 2 H, Ar-H), 7.27 (d, J = 8.l Hz, 2 H, Ar-H), 2.37 (s, 3 H,CH 3); 13 C NMR (75 MHz, CDCb) 8 147.64, 147.21 - 2 peaks for C=N 2 bond coupling to CF3, 141.30 Ar-Cl , 129.37, 128.70 Ar-C2,C6 and Ar-C3 , C5, 122.83 Ar-C4, ( q - 127, 122.57, 118.94, 115.5, J = 275 Hz, CF 3), 21.49 CH 3. 2,2,2-Trifluoro-1-( 4-tolyl)ethanone O-(Toluenesulfonyl)oxime (Compound 3) To a cooled solution (0 °C) of 2,2 ,2-trifluoro-l-( 4-tolyl)ethanone oxime (7.85 g, 39 mmol, 1 eq.) in 150 mL toluene was added tosyl chloride (11.18 g, 58 mmol, 1.5 eq.) and triethylamine (8.7 mL , 1.6 eq.). After 15 minutes, the reaction was allowed to warm to room temperature. The volatiles were removed in vacuo, and the remaining residue was taken up in 100 mL ether. Water (100 mL) was added, and the mixture was extracted four times with 100 mL portions of ether. The organic layers were combined and washed with water (100 mL, followed by 50 mL), dried over MgSO 4 , and concentrated to give an off white solid. The crude product was recrystallized from hexane to give a white solid (12.19 g, 88% yield): 1H NMR (300 MHz, CDCb) 8 7.88 (d, J = 8.5 Hz, 2 H, Ts Ar-H) , 7.38 (d, J = 8.3 Hz, 2 H, Ar-H), 7.32 (2 overlapping d, 4 H, Ar-H), 2.46 (s, 3 H,CH 3), 2.39 (s, 3 H,CH 3); 13 C NMR (75 MHz, CDC1 3) 8 154.0 (d, C=N 2 bond coupling to CF 3), 146.11 Ts-Cl , 142.35 Ar-Cl , 131.21, 129.87, 129.49, 129.26, 128.46, 121.61 - Ar, Ts-C's, (d - 121.55, 117.86, other two peaks buried, J = 278 Hz, CF 3), 21.78, 21.56 CH 3's. 3-(4-Tolyl)-3-(trifluoromethyl)diaziridine (Compound 4) A 250 mL round-bottom flask charged with 2,2,2-trifluoro-1-(4-tolyl)ethanone O(toluenesulfonyl)oxime was fitted with an ammonia condenser. The flask and the condenser were cooled with CO2/acetone and anhydrous ammonia (150 mL) was condensed into the flask. The reaction was stirred at -78 °C for 9.5 hours, after which 193 time the ammonia was allowed to evaporate. The residue was dissolved in ether (100 mL) and 100 mL water was added. The mixture was extracted with ether (4x100 mL), the organic layers were combined, dried over Mg8O 4 , and concentrated to give 6. 76 grams of a pale yellow solid that was used without further purification: 1H NMR (300 MHz, CDCb) 5 7.47 (d, J = 8.0 Hz, 2 H, Ar-H) , 7.20 (d, J = 8.0 Hz, 2 H, Ar-H) , 2.77 (d, J = 8.6 Hz, 1 H), 2.36 (s, 3 H,CH 3), 2.22 (d, J = 8.6 Hz, 1 H). 13 C NMR (75 MHz, CDCb) 5 140.27 Ar-C4, 129.88 Ar-Cl, 129.42, 128.04 Ar-C2,6 and 3,5, (d - 125.49, 121.80, other two peaks buried, J = 275 Hz, CF 3), 58.16, 57.69 q, J = 35 Hz C-(NH)2, 21.28 CH 3. 3-(4-Tolyl)-3-(trifluoromethyl)diazirine (Compound 5) To a solution of 3-(4-tolyl)-3-(trifluoromethyl)diaziridine (6.75 g, 33 mmol) in 100 mL dry ether was added 15 grams of freshly prepared Ag 2 O. After 2 hours, another 22 grams of Ag 2 O was added and the mixture was stirred for an additional two hours. The reaction mixture was filtered through Celite and washed with ether. The ether was removed using a rotary evaporator without heating to give an off-white solid which was purified by flash chromatography (silica, 9:1 hexane/ethyl acetate---+ethyl acetate) to give 3.49 grams of a pale yellow liquid (51% yield, 59% based on recovered starting material): 1H NMR (300 MHz, CDC1 3) 5 7.13 (d, J = 8.2 Hz, 2 H, Ar-H), 7.05 (d, J = 8.2 Hz, 2 H, Ar-H), 2.36 (s, 3 H, CH 3); 13 C NMR (75 MHz, CDC1 3) 5 140.02 Ar-C4, 129.70 Ar-Cl , 126.56, 126.44 Ar-C2,6 and 3,5, 124.35, 120.71 - q, J = 274 Hz, CF3 , 28.86, 28.33 d, J = 40 Hz C<(N)2, 22.92 CH 3. 3-( 4-bromomethylphenyl)-3-(trifluoromethyl)diazirine (Compound 6) To an argon purged solution of 3-( 4-tolyl)-3-(trifluoromethyl)diazirine (3.33 g, 16.6 mmol) in 80 mL CC1 4 was added N-bromosuccinimide (2.95 g, 16.6 mmol). A reflux condenser was attached and the mixture was heated to 68 °C under an argon atmosphere. AIBN (100 mg, 0.6 mmol) was added and the mixture was allowed to reflux for 20 minutes. The mixture was cooled and the solid succinimide was filtered off. The filtrate was concentrated and purified by flash chromatography (silica, 20:1 194 petroleum ether:CH2 Cb) to give 2.40 grams of a colorless oil (52% yield): 1H NMR (300 MHz, CDCl3) o7.41 (d, J = 8.1 Hz, 2 H, Ar-H), 7.15 (d, J = 8.1 Hz, 2 H, ArH), 4.45 (s, 3 H,CH2 Br); 13 C NMR (75 MHz, CDC1 3) o 139.64 Ar-C4, 129.58 Ar-Cl, 127.17, 126.94 Ar-C2,6 and 3,5, 123.98, 120.35 - q, J = 274 Hz , CF 3, residual solvent obscures C<(N) 2 , 29.17 CH 2 Br. N-( diphenylmethylene )-3-[p-[3-( trifluoromethyl )-diazirin-3-yl] phenyl]-DLalanine ethyl ester) (Compound 7) Sodium spheres (79 mg, 3.4 mmol) were placed in a dry 100 mL round-bottom flask under argon. Ethanol (7.8 mL) was added, followed by a solution of N- (diphenylmethylene)glycine ethyl ester (Aldrich) (880 mg, 3.3 mmol in 3.3 mL THF). The mixture was stirred for 45 min at °C, after which time, a solution of 3-( 4bromomethylphenyl)-3-( trifluoromethyl)diazirine (810 mg, 2.9 mmol in 5.5 mL dry THF) was added. After 4 hours, ether (100 mL) was added, and the organic layer was washed with sat. NH 4 Cl (2 x 50 mL), and H2 O (50 mL), dried over Mg8O 4 , and concentrated to give a yellow oil. The product was purified by flash chromatography (silica, CH 2 Cb--+5% ether in CH 2 Cb) to afford 485 mg of a yellow oil (36% yield): 1H NMR (300 MHz, CDCb) 7.77 (d, J = 7.1 Hz, 1 H), 7.60 (d, J = 6.8 Hz, 1 H), o 7.50 (m, 1 H), 7.40 (m, 1 H), 7.35- 7.2 (m, 6 H), 7.05 (d, J = 8.2 Hz, 1 H), 7.00 (d, J = 8.2 Hz, 1 H) , 4.25 (m, 1 H) , 4.15 (m, 1 H), 3.35-3.15 (m, 2 H) , 1.20 (t, J = 7.1 Hz, 3 H) ; 13 C NMR (75 MHz, CD Cb) o 171.3, 171.0, 140.4, 139.2, 135.9, 132.4, 130.5, 130.1 , 128.8, 128.6, 128.3, 128.1 , 127.6, 127.0, 126.4, 122.5 (q, J = 275 Hz) , 67.0, 61.1 , 39.3, 53 .6 , 28.4 (q, J = 40 Hz), 14.2. N-( 4-PO )-3-[p-[3-( trifluoromethyl)-diazirin-3-yl]phenyl]-DL-alanine) ( Compound 8) To a solution of (7) in 28 mL THF was added 7 mL 10 M AcOH. The solution was stirred for 24 hours, and another 7 mL 10 M AcOH was added. Volatiles were removed in vacuo, water (10 mL) was added, and the mixture was extracted with ether (2 x 25 mL). Cone. NaOH was added to the aqueous fraction to adj ust the pH 195 to ~13. This mixture was stirred at RT for 18 hours and neutralized with AcOH. The mixture was lyophilized, and the remaining solid was dissolved in a minimum amount of water. Dioxane (~20 mL) was added, followed by a solution of 450 mg 4-PO-anhydride in 20 mL dioxane. After 7 hours , the reaction was purified using an ion-exchange column (Dowex 50x 4-400). The eluent was lyophilized and cone. HCl was added to acidify the mixture. The insoluble yellow solid was filtered and the remaining aqueous solution was extracted with CH 2 Cb (3 x 50 mL). The organic extracts were combined with the filtered solid, dried (Na2 8O 4 ), and concentrated to give a yellow oil. The product was purified by a flash column (silica, hexane---+1:1 hexane/ ethyl acetate---+ethyl acetate---+ ethyl acetate + 1% AcOH) to give 215 mg of a whitish solid (58% yield for 3 steps): 1 H NMR (300 MHz, CDC1 3 ) 6 11.01 (s - br), 7.22 (d, J = 8.3 Hz, 2 H), 7.10 (d, J = 8.0 Hz, 2 H) , 6.95 (d, J = 7.8 Hz, 1 H), 5.71 (m, 1 H) , 5.02- 4.89 (m, 3 H) , 4.11 (q, J = 7.2 Hz, 1 H), 3.24- 3.18 (m, 2 H) , 2.31 (m, 4 H); 13 C NMR (75 MHz, CDC1 3 ) fJ 177.3, 174.2, 138.1 , 136.5, 130.0, 128.0, 128.8 (q, J = 275 Hz) , 126.7, 116.0, 53.2, 37.1, 35.4, 29.6, 28.5 (q, J = 40 Hz). N-(4-P0)-3-[p-[3-(trifluoromethyl)-diazirin-3-yl]phenyl]-DL-alanine cyanomethyl ester) (Compound 9) Prepared in 91% yield as above, starting with N-(4-PO-)-3-[p-[3-(trifluoromethyl)diazirin-3-yl]phenyl]-DL-alanine: 1 H NMR (300 MHz, CDC1 3 ) 6 7.19 (d, J = 8.3 Hz, 2 H) , 7.14 (d, J = 8.3 Hz, 2 H), 6.15 (d, J = 7.6 Hz , 1 H) , 5.73 (m, 1 H) , 5.10- 4.90 (m, 2 H) , 4.80 (dd, J = 15.7 Hz, 2 H) , 3.18- 3.12 (m, 2 H) , 2.35- 2.24 (m, 4 H); 13 C NMR (75 MHz, CDC13) fJ 172.4, 170.2, 137.1, 136.5, 129.7, 128.4, 126.9, 122.0 (q, J = 275 Hz), 115.9, 113.8, 52.6, 49.1, 37.2, 35.3, 29.1, 29.0 (q, J = 40 Hz). dCA-N-( 4-P0-)-3-[p-[3-(trifluoromethyl)-diazirin-3-yl]phenyl]-DL-alanin 196 7.4.2 dCA-4-PO-p-Benzoyl-Phenylalanine 0 0 ~N H 0 4-PO-p-Benzoyl-Phenylalanine Cyanomethyl Ester L-p-Benzoyl-phenylalanine·AcOH·H 2 O (Bachem, 200 mg, 0.67 mmol) was mixed with Na 2 CO 3 (284 mg, 2.68 mmol) in 30 mL 1:1 water/dioxane (not all solid dissolved). 4-PO-anhydride (125 mg, 0.67 mmol) in 5 mL dioxane was added and the mixture was stirred for 7 hours, after which 2 additional drops of 4-PO-anhydride were added. The mixture was stirred overnight, and CH 2 Cb (25 mL) and 1 N NaHSO 4 (25 mL) were added, and the mixture was extracted with CH 2 Cl 2 (3 x 25 mL) and the combined organics were dried over Na 2 8O 4 . The volatiles were removed to give 600 mg of a pale yellow oil which was purified by flash chromatography (silica, ethyl acetate-----,ethyl acetate + 1% AcOH) to give 532 mg of a colorless oil. 1 H NMR indicated that there was still 4-PO-anhydride or acid present. 250 mg of the crude product was dissolved in DMF (2 mL) and chloroacetonitrile (2 mL) was added, followed by triethylamine (0.3 mL). After 39 hours (not optimized) the volatiles were removed and the compound was purified by flash chromatography (silica, 1:1 hexane/ethyl acetate) to give 131 mg of a clear oil (49% yield): 1 H NMR (300 MHz, CDC1 3 ) <5 7.76 (m, 4 H), 7.60 (t, J = 7.4 Hz , 1 H), 7.48 (t, J = 7.4 Hz, 2 H), 7.31 (d, J = 7.7 Hz, 2 H) , 6.65 (d, J = 6.8 Hz , 1 H), 5.85- 5.65 (m, 2 H), 5.05 (m, 2 H), 4.81 (dd, J = 15.0 Hz , J = 2.0 Hz, 2 H) , 3.22 (AB, J = 7.2, 2 H) , 2.32 (m - br, 4 H) (impurities: 4.08, 3.62); 13 C NMR (75 MHz, CDC13 ) <5 196.3, 172.5, 170.0, 140.3, 137.2, 136.4, 136.3, 132.4, 130.4, 129.8, 129.0, 128.2, 115.6, 113.8, 107.1 , 90.6, 69.9, 65.8, 62.4, 52.5, 48.9, 37.1, 34.9, 29.0. 197 dCA-N-4-PO-p-Benzoyl-Phenylalanine Prepared as above: FAB-MS [M-HJ- 968.4, calcd for C40 H45 N9 O 16 P 2 968.8. 7.5 Synthesis of Caged Amino Acids 7.5.1 dCA-4-PO-Cysteine-S-o-Nitrobenzyl Thioether Y'NO, o (s ~N~dCA H 0 An overview of this synthesis is shown in Scheme 7.5. 4-PO-Cysteine-S-o-Nitrobenzyl Thioether (Compound 10) Sodium spheres (49 mg, 4.1 mmol) were dissolved in 4 mL dry MeOH at O °C and the mixture was warmed to room temperature over 10 min. L-Cysteine (250 mg, 2.06 mmol) was added, followed after 5 min. by o-nitrobenzyl bromide (356 mg, 1.65 mmol). After 1 hour, the mixture had gelled. Methanol (~5 mL) was added, but the gel did not dissolve. Volatiles were removed on a rotovap, and the residue was dissolved in water (10 mL) and dioxane (10 mL). Na2 CO 3 was added (436 mg, 4.1 mmol), followed by 4-PO-anhydride (380 mg, 2.1 mmol). After 3 hours, an additional 50 mg 4-PO-anhydride was added. After an additional hour, CH 2 Cl 2 (25 mL) and 1 N NaHSO 4 (25 mL) were added, and the mixture was extracted with CH 2 Cb (3 x 25 mL) and the combined organics were dried over Na2 SO 4 . The volatiles were removed, and the product was purified by flash chromatography (silica, 1:1 hexane/ethyl acetate--+ethyl acetate+ 1% AcOH) to give 472 mg of a pale yellow solid (85% yield for 2 steps): 1 H NMR (300 MHz, CD 3 OD) b 7.93 (AB, J = 9.0 Hz, J = l.3 Hz) , 7.62- 7.40 (m, 3 H) , 5.90-5.75 (m, 1 H) , 4.97 (dd, J = 32.0 Hz, J = 198 1) NaOMe NO 2 ~o~ 0 0 _ ~ CICH2CN / Et3N r~ s ~ DMF NlOH H 0 10 ()N02 _) , r~s O ~ N l OH 0 () DMF CN _ ) ' N02 dCA•2 n-Bu4N+ o (s ~N~dCA H 0 11 Scheme 7.5: Synthesis dCA-4-PO-Cysteine-S-o-Nitrobenzyl Thioether. 1.8 Hz, 2 H) , 4.94 (dd, J = 32.0 Hz, J = 1.8 Hz, 2 H), 4.59 (dd, J = 4.8 Hz, 1 H) , 4.08 (m, 2 H) , 2.90- 2.75 (m, 2 H) , 2.34 (m, 4 H); 13 C NMR (75 MHz, CD 3OD) c5 175.4, 173.5, 149. 7, 138.1 , 134.9, 134.3, 133.3, 129. 7, 126.5, 116.0, 53.0, 36.0, 34.4, 34.3, 30.7. 4-PO-Cysteine-S-o-Nitrobenzyl Thioether Cyanomethyl Ester (Compound 11) Prepared in 85% yield, as above, starting with 4-PO-Cys-S-NB (360 mg, 1.06 mmol): 1H NMR (300 MHz , CDC1 ) c5 8.00 (d, J = 8.2 Hz , 1 H) , 7.60 (t, J = 7.7 Hz, 1 H) , 3 7.44 (m, 2 H) , 6.8 (m - br, 1 H) , 5.83 (m, 1 H) , 5.10- 4.95 (m, 2 H) , 4.80 (s, 2 H) , 4.11 (m, 2 H) , 2.94 (m, 2 H) , 2.36 (m - br, 4 H) ; 13 C NMR (75 MHz, CDCb) c5 172.7, 169.5, 148.3, 136.7, 133.4, 133.2, 132.1 , 128.7, 125.6, 115.7, 114.2, 51.6, 49.4, 35.0, 34.0, 33.4, 29.3. dCA-4-PO-Cysteine-S-o-Nitrobenzyl Thioether (dCA-Caged-Cys) Prepared as above: FAB-MS [M-H]+ 957.2, calcd for C34 H42N10O 17 P2S 956.2. 199 7.5.2 a-4-PO-c:--NVOC-Lysine Cyanomethyl Ester 0 ~N H 0-.....,_,...-CN 0 a-4-PO-c-BOC-Lysine E-BOC-Lysine (500 mg, 2.0 mmol) and Na2 CO 3 were stirred in 1:1 water/dioxane (30 mL). A solution of 4-PO-anhydride (407 mg, 2.23 mmol, 1.1 eq. in dioxane) was added and the mixture was stirred for 1 hour. CH 2 Cb (25 mL) and 1 N NaHSO 4 (25 mL) were added, and the mixture was extracted with CH 2 Cb (3 x 25 mL) and the combined organics were dried over Na2 8O 4 . ' The volatiles were removed to give 1.27 g of a pale yellow oil. The product was purified by flash chromatography (silica, pet. ether-1:1 pet. ether/ethyl acetate-ethyl acetate-ethyl acetate+ 1% AcOH) to give 600 mg of a colorless film (90% yield): 1 H NMR (300 MHz, CD 3 OD) b 5.82 (m, 1 H), 5.03 (d, J = 17.0 Hz, 1 H) , 4.67 (d, J = 10.0 Hz, 1 H), 4.38 (AB, J =5 .0 Hz, 1 H), 2.91 (t, J = 7.6 Hz , 2 H) , 2.34 (m - br, 4 H) , 1.95- 1.38 (m, 6 H); 13 C NMR (75 MHz, CD 3 OD) o 175.6, 175.3, 138.2, 115.8, 53.3, 40.4, 36.0, 31.9, 30.8, 27.9, 23.8. a-4-PO-Lysine To a solution of a-4-PO-c-BOC-lysine (600 mg, 1.83 mmol) and 1.25 g 1,4dimethoxybenzene in 10 mL CH 2 Cb was added trifluoroacetic acid (10 mL). After 30 minutes, volatiles were removed (CO 2 /acetone trap), toluene (15 mL) was added, and the volatiles were removed again. The remaining residue was triturated with ether (3 x 25 mL). Attempts to dissolve the mixture in MeOH and precipitate the product with ether failed, and the methanol/ ether mixture was removed to yield 400 mg of an oil (96% yield) which was used directly in the synthesis of a-4-PO-c-NVOC-lysine. 200 o:-4-PO-c-NVOC-Lysine Prepared in 58% yield as above from 100 mg o:-4-PO-lysine, 121 mg NVOC-Cl, and 93 mg Na2 CO 3 : 1 H NMR (300 MHz, CDCh) c5 9.25 (s - br), 7.67 (s, 1 H), 7.01 (s, 1 H), 6.84 (d, J = 8.4 Hz, 1 H) , 5.80 (m, 1 H), 5.60- 5.40 (m, 3 H) , 5.06 (m, 2 H) , 4.56 (m, 1 H) , 3.97 (s, 3 H) , 3.94 (s, 3 H), 3.19 (d - br, 2 H), 2.35 (s - br, 4 H), 1.95-1.35 (m, 6 H); 13 C NMR (75 MHz, CDCh) c5 176.2, 175.4, 173.7, 156.5, 153.6, 148.1, 139.7, 136.7, 128.0, 115.7, 110.3, 108.1, 63.6, 56.4 (2), 52.2, 40.5, 35.4, 31.4, 29.5, 29.3, 22.3. o:-4-PO-c-NVOC-Lysine Cyanomethyl Ester Prepared in 81 % yield as above, starting with 100 mg o:-4-PO-c-NVOC-lysine: 1 H NMR (300 MHz, CDC1 3 ) c5 7.69 (s, 1 H) , 7.01 (s, 1 H), 6.52 (d - br, 1 H) , 5.80 (m, 1 H), 5.48 (s, 2 H), 5.25 (t - br, 1 H) , 5.01 (dd, J = 18.0 Hz, J = 1.2 Hz, 2 H), 4.88 (dd, J = 15.7 Hz , J = 2.0 Hz, 2 H) , 4.61 (m, 1 H) , 3.98 (s, 3 H) , 3.96 (s, 3 H), 3.24 (m, 2 H) , 2.35 (m, 4 H), 1.85 (m, 2 H) , 1.58 (m, 2 H) , 1.43 (m, 2 H); 13 C NMR (75 MHz, CDC1 3) c5 172.7, 171.1 , 156.3, 153.6, 148.2, 139.8, 136.8, 127.9, 115.6, 114.2, 110.3, 108.2, 63.6, 56.5, 56.4, 51.8, 48.9, 40.2, 35.2, 30.9, 29.4, 29.3, 22.1. 7.5.3 dCA-4-PO-Serine-O-o-Nitroveratryl Ether 02NyYOCH3 O (O~OCH3 ~N~dCA H 0 o:-4-PO-Serine-O-o-Nitroveratryl Ether Prepared in 82% yield as above, starting from the Na+ salt of L-serine-O-NV [11]: 1 H NMR (300 MHz, CDCh) c5 7.62 (s, 1 H), 7.12 (s, 1 H) , 6.79 (d, J = 7.l Hz , 1 H) , 5.80 (m, 1 H) , 5.15- 4.78 (m, 5 H), 4.10 (dd, J = 9.0 Hz, J = 3.2 Hz, 1 H) , 3.82 (dd, J = 9.0 Hz, J = 3.2 Hz, 1 H), 3.94 (s, 3 H), 3.92 (s, 3 H) , 2.39 (m, 4 H); 13 C NMR 201 (75 MHz, CDCh) <5 173.8, 173.5, 153.8, 147.7, 139.1, 136.5, 129.6, 115.9, 109.6, 107.9, 70.8, 70.3, 56.4, 56.3, 52.9, 35.4, 29.4. a -4-PO-Serine-O-o-Nitroveratryl Ether Cyanomethyl Ester Prepared in 94% yield as above, starting from 90 mg a-4-PO-Ser-O-NV: 1 H NMR (300 MHz, CDC1 3 ) <5 7.63 (s, 1 H), 7.01 (s, 1 H) , 6.52 (d, J = 7.l Hz, 1 H), 5.85 (m, 1 H) , 5.12- 4.75 (m, 7 H), 4.11 (dd, J = 8.0 Hz, J = 3.2 Hz, 1 H), 3.83 (dd, J = 8.0 Hz, J = 3.2 Hz, 1 H), 4.00 (s, 3 H), 3.94 (s, 3 H), 2.40 (m, 4 H) ; 13 C NMR (75 MHz, CDC1 3) <5 172.4, 169.1, 153.7, 148.0, 139.5, 136.7, 128.7, 115.7, 113.9, 110.2, 108.0, 70.7, 70.6, 56.6, 56.4, 52.4, 49.2, 35.3, 29.3. dCA-4-PO-Serine-O-o-Nitroveratryl Ether Prepared as above: FAB-MS [M-HJ- 999.3, calcd for C36 H46 N10 0 20 P 2 1000.24. 7.5.4 dCA-4-PO-Threonine-O-o-Nitroveratryl Ether An overview of this synthesis is shown in Scheme 7.6. The procedure is adapted from [12]. a -N-Trityl-Threonine Methyl Ester (Compound 12) L-Threonine methyl ester- HCl (1.0 g, 5.9 mmol, 1.0 eq) was dissolved in 50 mL dry CH 2 Cb in a dry 100 mL round-bottom flask under argon at O°C. Triethylamine (1.19 g, 11.8 mmol, 2.0 eq) was added, followed by trityl chloride (1.64 g, 5.9 mmol, 1.0 eq). After 10 minutes at O°C, the mixture was allowed to warm to room temperature and was stirred for an additional hour. A solution of 10% citric acid (20 mL) was added, 202 H3C,,,(OH 1) TsCJ, pyridine Tr, NTTOMe H3~>t-H l)TFA Tr-NyOCH3 2) Na2C03, 0 0 12 13 ~0 ~ 0 0 1) NaOH/MeOH 02N'r(yOCH3 H3C,, o ~ 0 ·r - , OCH 3 ~NTTO-CN H 0 DMF 16 Scheme 7.6: Synthesis dCA-4-PO-Threonine-O-o-Nitroveratryl Ether. and the organic layer was extracted, washed with water, and dried over Na2 SO 4 with activated charcoal. The mixture was filtered and concentrated in vacuo to give 1.82 g of a foamy white solid (83% yield) : 1H NMR (300 MHz, CDCb) 8 7.48 (d, J = 7.4 Hz, 6 H), 7.30- 7.10 (m, 9 H), 3.81 (m, 1 H), 3.39 (d, J = 7.3 Hz, 1 H), 1.19 (d, J = 7.3 Hz, 3 H); 13 C NMR (75 MHz, CDC1 3 ) 8 173.6, 145.4, 128.9, 127.8, 126.6, 70.7, 69,7, 62.4, 51.6, 18.9. Compound 13 a -N-trityl-threonine methyl ester (1.74 g, 4.63 mmol) was dissolved in 10 mL dry pyridine in a dry 100 mL round-bottom flask under argon. In a separate dry 100 mL round-bottom flask under argon, tosyl chloride (2.65 g, 13.9 mmol, 3.0 eq) was dissolved in 10 mL dry pyridine. The amino acid solution was cooled to O °C, and the TsCl solution was cannulated in. The reaction was allowed to warm to room temperature, and was stirred for 20 hours. The pyridine was removed in vacuo, and EtOAc (25 mL) and water (25 mL) were added and the mixture was extracted. The aqueous layer was further extracted with EtOAc (2 x 25 mL), and the combined organic extracts were washed with 10% citric acid (3 x 10 mL) and water (2 x 10 mL). 203 The organic extracts were dried over Na28O 4 and concentrated in vacuo to give 2.66 g of an orange oil. 2.54 g of the oil was dissolved in dry THF (20 mL) under argon, triethylamine (937 mg, 2.0 eq) was added, and the mixture was refluxed overnight. The solvent was removed in vacuo, and the residue was dissolved in EtOAc and 10% citric acid and extracted with EtOAc (3 x 25 mL). The combined organic extracts were washed with 1 M NaHCO 3 and water, dried over Na28O 4 and concentrated in vacuo to give 1.66 g of a dark brown solid. The product was purified by flash chromatography (silica, 4:1 hexane/CH2Cb-----+l:l hexane/CH 2Cl2-----+l:3 hexane/CH 2Cl 2 to give 804 mg of a foamy white solid (yield: 49% for 2 steps. Repeated on larger scale, yield: 78%): 1 H NMR (300 MHz, CDCh) c5 7.52 (d, J = 7.2 Hz, 6 H), 7.27- 7.15 (m, 9 H), 3.69 (s, 3 H), 1.89 (d, J = 6.4 Hz, 1 H), 1.63 (m, 1 H), 1.36 (d, J = 5.5 Hz, 3 H); 13 C NMR (75 MHz, CDC13 ) c5 170.6, 143.9, 129.3, 127.6, 126.8, 75.0, 51. 7, 35.9, 34.8, 13.3. Compund 14 To a solution of 13 (400 mg, 1.1 mmol) in 5 mL 1:1 MeOH/CH 2Cl 2 at O°C was added TFA (5 mL). The reaction was stirred at O °C for 10 min and allowed to warm to RT. After 3 hours, an additional 5 mL TFA was added and the mixture was stirred overnight. The solvent was removed by forming an azeotrope with toluene. Water (20 mL) and dioxane (20 mL) were added, along with 4-PO-anhydride (261 mg, 1.43 mmol, 1.3 eq) and Na2CO 3 (466 mg, 4 eq). After 5 hours, CH 2Cb (25 mL) was added, and the mixture was extracted with CH 2Cl 2 (3 x 25 mL). The combined organic extracts were dried over Na28O 4 and concentrated in vacuo to give 660 mg of a colorless oil. The product was purified by flash chromatography (silica, 9:1 hexane/EtOAc-----+1:1 hexane/EtOAc-----+EtOAc to give 184 mg of a colorless oil (yield: 84% for 2 steps): 1 H NMR (300 MHz, CDC13 ) c5 5.82 (m, 1 H), 5.10- 4.97 (m, 2 H), 3.79 (s, 3 H), 3.20 (d, J = 6.4 Hz, 1 H), 2.81 (m, 1 H), 2.53 (m, 2H), 2.39 (m, 2 H), 1.37 (d, J = 5.6 Hz, 3 H); 13 C NMR (75 MHz, CDCh) c5 183.5, 167.8, 136.7, 115.7, 52.4, 38.7, 37.9, 35.6, 28.6, 13.1. 204 a-4-PO-Threonine-O-o-Nitroveratryl Ether Cyanomethyl Ester (Compound 16) Compound 14 (170 mg, 0.86 mmol, 1.0 eq) and o-nitroveratryl alcohol (NVOH, 550 mg, 2.58 mmol, 3.0 eq) were placed in dry 100 mL round-bottom flask under argon. Dry CDC1 3 (30 mL) was added (CDC1 3 was used because it is not stabilized with EtOH). Most but not all of the NVOH dissolved and the flask was cooled to 0 °C. 5 drops of BF 3 -OEt 2 were added and after 15 min, the mixture was warmed to RT and allowed to stir overnight, after which time, an additional 100 µL BF 3 -OEt 2 was added. After 6 additional hours, 5% NaHCO 3 was added, and the mixture was extracted with CH 2 Cl 2 (3 x 25 mL). The combined organic extracts were dried over Na 2 8O 4 and concentrated in vacuo to give 652 mg of a yellow solid. Attempts to separate 15 from NVOH by TLC failed, and the decision was made to hydrolyze the ester directly. The solid was dissolved in MeOH (30 mL) , and 1 N NaOH (15 mL) was added. After 6 hours, additional 1 N NaOH (5 mL) was added and the reaction was stirred overnight. The MeOH was removed by rotary evaporation, and the mixture was extracted with CH 2 Cb (1 x50 mL, 2 x 25 mL) to remove NVOH. CH 2 Cb (50 mL) and 1 N NaHSO 4 (50 mL) were added to the aqueous layer, and it was extracted with CH 2 Cb (3 x 25 mL). The combined organic extracts were dried over Na 2 8O 4 and concentrated in vacuo to give 107 mg of an orange-brown oil. The product was purified by flash chromatography (silica, 1:1 hexane/ethyl acetate--+ethy1 acetate--+ethy1 acetate + 1% AcO H). The product-containing fractions were decolorized with activated charcoal and concentrated to give 95 mg of a pale yellow oil. 1 H NMR indicated that the product was still not pure, and it was decided to continue to the cyanomethylation. 58 mg of the crude product and 50 µL of Et 3 N were dissolved in 2 mL dry DMF. Chloroacetonitrile (2 mL) was added and the mixture was stirred overnight. Volatiles were removed in vacuo and the product was purified by flash chromatography (silica, 1:1 hexane/ethyl acetate) to give 49 mg of a pale yellow oil. (Yield: 77% for final step, 21 % for 3 steps): 1 H NMR (300 MHz, CDC1 3 ) b 7.63 (s, 1 H) , 7.05 (s, 1 H) , 6.29 (d, J = 10.1 Hz, 1 H) , 5.85 (m, 1 H), 205 5.17- 4.65 (m, 6 H), 4.30 (m, 1 H), 4.02 (s, 3 H), 3.94 (s, 3 H), 2.42 (m, 4 H), 1.34 (d, J = 6.2 Hz, 3 H) ; 13 C NMR (75 MHz, CDC1 3 ) b 172.9, 169.6, 153.6, 148.0, 139.6, 136.7, 128.7, 115.8, 114.0, 110.8, 108.0, 75.6, 68.1, 56.6 , 56.5, 56.4, 49.1 , 35.5, 29.4, 16.0. dCA-4-PO-Threonine-O- o-Nitroveratryl Ether Prepared as above: FAB-MS [M-HJ- 1013.3 (also see 1254.7), calcd for C37 H48 N10 0 20 P 2 1014.25. 7.6 Molecular Biology 7.6.1 General Procedure for Site-Directed Mutagenesis In most instances, site-directed mutants were prepared using a standard two-step PCR-based cassette mutagenesis procedure using Pfu Polymerase (Stratagene). For each mutant, an enzymatically trimmed PCR-generated fragment was subcloned into the dephosphorylated fragment of the desired vector. In some cases, it was possible to synthesize the inserted mutant fragment without resorting to the PCR method. When possible, silent mutations were introduced to facilitate screening of positive colonies by restriction endonuclease digestion. We have found the Bfal recognition sequence CTAG to be particularly helpful for screening TAG mutants. All mutants were verified by automated sequencing of the entire PCR cassette across both ligation sites. 7.6.2 General Procedure for In Vitro Transcription ofmRNA mRNA was synthesized in vitro from linearized plasmid DNA (1 µg/40 µL reaction) using the T7 MessageMachine kit from Ambion per the manufacturer's instructions. The products were purified by phenol:chloroform:isoamyl alcohol extraction and isopropanol precipitation. Purity of the mRNA was assayed by gel electrophoresis. 206 7.6.3 Mutants for Nonsense Suppression in the nAChR The following is a list of mutants prepared for nonsense codon suppression in the nAChR from mouse muscle. All mutants are in the pAMV-PA vector [13]. aW67-TAG aN68-TAG aP69-TAG aD70-TAG aD71-TAG aY72-TAG aG73-TAG aG74-TAG aV75-TAG aK76-TAG aD89-TAG aF124-TAG 1 Yl17-TAG Table 7.1: TAG mutants in the nAChR In addition to these mutants, two other TAG mutants were prepared in other constructs. The first (a210GPI D70-TAG) was prepared in a truncated N-terminal fragment of the mouse nAChR a subunit fused with a sequence directing membrane anchorage via a glycosyl-phosphatidylinositol (GPI) linkage [14]. The second is the aW54-TAG mutant, prepared in the a7 neuronal nicotinic subunit from the mouse. This residue is homologous to the 1 55/857 mutants discussed in Chapter 2. 7.6.4 Conventional Mutants in the nAChR In addition to the mutants outlined above, a series of mutants was prepared to investigate the roles of putative glycosylation sites in the nAChR. Sequence alignnments of the ligand-gated ion channel (LGIC) superfamily reveal an interesting trend. In the main immunogenic region of the a subunit of the nAChR, there is a praline residue at position 69. In the aligned positions of the LGICs, this praline is often replaced with a consensus asparagine-linked glycosylation sequence (NX( ¥)). Both pralines and N-linked glycosylation are thought to induce f3 turns in protein structures [15, 16]. 207 aP69-----+A j)D68-----+ N, P69-----+ I 1 D68-----+N, P69-----+I bT72-----+D a7869-----+D aP69-----+ A, D70-----+S j)D68-----+ N, P69-----+ I, A 70-----+S ,D68-----+N, P69-----+I, A70-----+S 6171-----+P, T72-----+D a7V68-----+ P, 869-----+ D Table 7.2: Conventional site-directed mutants to investigate the role of pralines and glycosylation sites in the extracellular domain. Thus, several mutants were made in various subunits of both the muscle and a7 neuronal nAChR from the mouse to explore whether either a praline or a glycosylation site is vital to the function of the receptor, and if so, to determine if pralines and glycosylation sites are interchangable. The mutants are shown in Table 7.2. Although these mutants were not tested extensively, early experiments comparing the aP69-----+A mutant (which lacks the praline and cannot be glycosylated) and the aP69-----+A, D70-----+S mutant (which lacks the praline, but may be glycosylated) suggest that the aP69-----+ A, D70-----+S expresses better (as evidenced by larger whole-cell currents in response to ACh) than the aP69-----+A mutant (Figure 7.1). This may imply that a turn-promoting unit in this region of the receptor is necessary for receptor function. This is consistent with the NMR structure of the MIR peptide shown in Chapter 6. 7. 7 Oocyte Experiments 7. 7 .1 Oocyte Preparation and Injection Oocytes were removed from Xenopus laevis as described previously [17] and maintained at 18 °C in ND96 solution consisting of 96 mM NaCl, 2 mM KCl, 1 mM MgCb, 1.8 mM CaC1 2 , 5 mM HEPES (pH 7.5), supplemented with sodium pyruvate (2.5 mM), gentamicin (50 µg/mL), theophylline (0.6 mM) and horse serum (5%). Prior to microinjection, the NVOC-aminoacyl-tRNA (1 µg/ µL in 1 mM NaOAc, pH 5.0) 'was deprotected by irradiating the sample for five minutes with a 1000 W xenon arc lamp (Oriel) operating at 600 W equipped with WG-335 and UG-11 filters (Schott). 208 5000 4000 ,--._ ~ r:: 3000 ..., '-' r:: a) 1--4 1--4 ::l u 2000 1000 0 P69A P69A D70S Figure 7.1: Whole-cell currents in response to 10 µM ACh for oocytes expressing nAChR aP69-+A or aP69-+A, D70-+S. 4-PO-protected tRNAs were deprotected by treatment with 1 volume of a saturated solution of 12 in 1 mM NaOAc, pH 5.0. The deprotected aminoacyl-tRNA was mixed 1:1 with the desired mRNA. In general, for experiments with suppression in the a subunit of the nAChR, oocytes were injected with 50 nL of a mixture containing 25 ng aminoacyl-tRNA and 1.2 ng of total mRNA (4:1:1:1 a:/3:--y:6 subunit stoichiometry). 7. 7. 2 Electrophysiology Two-electrode voltage clamp measurements were performed 36- 48 hours after injection using a GeneClamp 500 amplifier (Axon Instruments, Foster City, CA). Microelectrodes were filled with 3M KCl and had resistances of 0.5 to 2 MO. Oocytes were perfused with a nominally calcium-free bath solution containing 96 mM NaCl, 2 mM KCl, 1 mM MgCh, 5 mM HEPES (pH 7.5). Macroscopic ACh-induced currents were recorded in response to application of 200 µM ACh at a holding potential of -80 mV. Whole-cell current measurements are reported as the mean ± s.e.m. 209 7.8 Results of Nonsense Suppression Experiments In this section, we outline the results of nonsense suppression experiments in the nicotinic acetylcholine receptor. For each experiment, a successful incorporation is one in which the whole-cell currents in response to 200 µM ACh for the mutant receptor are greater than than those obtained when an identical mRNA mixture and an uncharged tRNA (tRNA- dCA) are injected. In most cases, this "dCA control" produced very small currents, however, at aN68, and several sites in the --y subunit, the dCA currents were not negligible (>500- 1000 nA). 7.8.1 Natural Amino Acids The natural amino acids shown in Figure 7.2 were all incorporated into various sites within the nAChR. Phenylalanine (protected with both the NPS and 4-PO groups) was successfully incorporated at position aE180 in the nAChR, thus demonstrating the viability of these protecting groups. Valine was successfully incorporated at aG73 and aV75, but no currents were seen at aG74. Tryptophan was successfully incorporated at two sites, aW67 and aY72. Praline was incorporated at a natural praline site, aP69 - no other sites were tested. Tyrosine was incorporated at several sites shown in Table 7.3 and was often used as a positive control for expression when the natural amino acid was not available. H, N ; $ 0 Phe Yr(oH H Pro 0 Val Figure 7.2: Natural amino acids incorporated into the nAChR by nonsense suppression. 210 Successes aN68 aY72 aW67 Failures aD70 aK76 aP69 aD71 Table 7.3: Incorporation of tyrosine into the nAChR. 7.8.2 Biotinylated Amino Acids The biotinylated amino acids shown in Figure 7.3 were used in several suppression experiments. In general, biocytin 2 was efficiently incorporated into the nAChR (Table 7.4), with whole-cell currents that were comparable to natural amino acids at the same position (>5 µA). The mid-chain biocytin analog could also be incorporated at several sites in the receptor (Table 7.5). Although the currents were often well above background (500-1000 nA), they were not large enough to allow biochemical experiments (such as the binding of streptavidin) to be performed. Finally, we were unable to see expression of the long-chain biocytin analog at any of the sites tested (aD70, aK76, and aE180). Mid-Chain-Biocytin (A) Biocytin 0 H 0 HN~N~N~"·Q 0 H OH H2 N O HN)(NH 0 Long-Chain-Biocytin (B) Figure 7.3: Biotinylated amino acids incorporated into the nAChR by nonsense suppression 2 For full details, see Chapter 6 211 Successes Failures aN68 aK76 1 Vl32 1 Fl37 aW67 aG74 aD70 aE180 1 Tl33 a D71 ,8129 1 Yl34 aV75 ,8131 ,F135 aP69 aY72 aG73 Table 7.4: Biocytin incorporation into the nAChR. Successes Failures aN68 aD71 a D70 aP69 Table 7.5: Mid-chain-biocytin incorporation into the nAChR. 7.8.3 Chemically Reactive Amino Acids In this section, we outline the incorporation of amino acids with chemically reactive side chains. Such amino acids might be useful for post-translational modification of proteins and may complement methods for site-specific cell labeling, such as SCAM and the biotinylated amino acids described above. e I H,N,<,OH 0 Allyl-Gly H2N 0 Keto-Tyr 0 0 o~ H2N ~ 0 Acetyl-Phe Figure 7.4: Amino acids with chemically reactive side chains incorporated into the nAChR by nonsense suppression. The first amino acid is allyl-glycine, which contains a terminal olefin. Since terminal olefins are not common on the surfaces of cells, we reasoned that this amino acid might display unique reactivity. Water-soluble olefin metathesis catalysts, such as the one shown in Figure 7.5 can mediate the reaction shown in Scheme 7. 7 [18, 19]. By using an appropriately substituted catalyst, we can efficiently and selectively transfer a specific functionality to a protein containing a terminal olefin. To this end, we incorporated allyl-glycine at several sites in the nAChR. Whole-cell currents were uniformally large (often greater than 10 µA) at several sites shown in Table 7.6. Unfortunately, at high concentrations, the catalyst shown in Figure 7.5 is acutely toxic 212 to cells (Figure 7.6) , and this approach may be limited. Figure 7.5: A water-soluble olefin metathesis catalyst. ); + /,N~\ H 0 Scheme 7. 7: Olefin metathesis Successes Failures aD70 aK76 aE180 None Table 7.6: Allyl-glycine incorporation into the nAChR. In addition to allyl-glycine, two other amino acids with specific side chain reactivies were incorporated into the nAChR at position aD70. Both keto-tyrosine [20, 21] and 4-acetyl-phenylalanine [21] gave whole-cell currents in excess of 1 µA in response to 200 µM ACh. Since ketones are not prevalent on the cell surface, it may be possible to label these amino acids with functionalized hydrazines or hydroxylamines [20, 22]. 7.8.4 Photochemically Reactive Amino Acids Finally, we attempted to incorporate several amino acids with photo-reactive side chains into the nAChR (Figure 7.7). Benzoyl-phenylalanine contains a benzophenone moiety, which upon irradiation with UV light , forms a triplet excited state that can efficiently abstract hydrogen atoms from a variety of C- H bonds. The remaining carbon-centered radicals may recombine to form new carbon-carbon bonds. Thus, this amino acid may be used to cross-link separate subunits in a multi-subunit protein 213 35 30 f average membrane potential for control oocytes without catalyst 25 20 ;----------;- - - - - - . - - - - - - - 15 estimated lower limit for electrophysiology assay 10 5 0 Concentration of Catalyst (M) Figure 7.6: Resting membrane potentials for oocytes treated with the catalyst shown in Figure 7.5. Each point represents the average resting membrane potential of 5 oocytes treated for 15 minutes with the given concentration of catalyst in degassed ND96 , pH 6.0. Catalyst concentrations > 10 µM rapidly kill the cells. such as the nAChR, and allow us to determine which amino acids lie at subunit interfaces. The results of attempts to incorporate benzoyl-phenylalanine at several sites in the nAChR are shown in Table 7.7. Benzoyl-phe was successfully incorporated at several sites, including positions where the wild-type residue is either a polar or charged amino acid. However, in many cases, especially in the I subunit, where dCA currents were high, the whole-cell currents were not much larger than the dCA control. Nonetheless, this amino acid may be useful for photocross-linking studies when the appropriate biochemical assays to identify successful cross-linking (such as gel-shift assays, or mass-spectrometry) are perfected. In addition to benzoyl-phenylalanine, we also attempted to incoporate a diazirinecontaining amino acid (Figure 7. 7) into the nAChR. Although this amino acid has been incorporated into proteins via nonsense suppression [23] , we did not observe 214 N ale (S ~ H2N~:H N02 H2N~OH 0 0 Benzoyl-Phe Diazirine Caged-Cys Figure 7. 7: Amino acids with photochemically reactive side chains incorporated into the nAChR by nonsense suppression. Successes Failures aD70 ,S131 ,F135 ,1130 ,W139 aY72 1 Tl33 1 Fl37 ,V132 ,Q140 ,'8129 ,Y134 ,D138 1 Pl36 Table 7. 7: Benzoyl-phenylalanine incorporation into the nAChR. ACh-induced currents during attempts to incorporate this amino acid at a70. Although a70 is a permissive site, it is possible that this residue may be incompatible with a70, and further testing at other sites may be fruitful. Fortunately, benzoylphenylalanine is tolerated at a number of sites, and it may represent a better alternative to the diazirine-containing amino acid. Finally, we incorporated "caged-cysteine" at position a70 of the nAChR. In several experiments , caged-cysteine produced whole-cell currents of 1- 2 µA m response to 200 µM ACh. Caged-cysteine is attractive because it allows one to block all endogenous cysteines by reacting them with a maleimide-based reagent. Subsequent decaging of the cysteine side chain with 320 nm light will reveal a unique cysteine, which may react with a wide variety of commercially-available maleimide or methanethiosulfonate (MTS) reagents. Caged cysteine may eventually provide a complementary method for site-specific cell-labeling, which may be particularly useful for labeling cells with multiple fluorescent probes for fluorescence-resonance energy transfer experiments [24]. 215 Bibliography [1] J.S. Debenham, R. Madsen, C. Roberts, and B. Fraser-Reid. 2 new orthogonal amine-protecting groups that can be cleaved under mild or neutral conditions. J. Am. Chem. Soc., 117(11):3302- 3303, 1995. [2] L. Zervas, D. Borovas, and E. Gazis. J. Am. Chem. Soc., 85:3660, 1963. [3] B. Amit, U. Zehavi, and A. Patchornik. Photosensitive protect- mg groups of amino sugars and their use m glycoside synthesis. 2Nitrobenzyloxycarbonylamino and 6-nitroveratryloxycarbonylamino derivatives. J. Org. Chem., 39:192- 196, 1974. [4] U. Ellervik and G. Magnusson. 2-(Trimethylsilyl)ethyl glycosides. 9. Transformation into the corresponding 1-O-acyl sugars. Acta. Chem. Scand., 47(8):826- 828, 1993. [5] C. J. Noren, S. J. Anthony-Cahill, M.C. Griffith, and P.G. Schultz. A general method for site-specific incorporation of unnatural amino acids into proteins. Science, 244: 182- 188, 1989. [6] S. A. Robertson, J. A. Ellman, and P. G. Schultz. A general and efficient route for chemical aminoacylation of transfer RNAs. J. Am. Chem. Soc., 113:2722- 2729, 1991. [7] J. Ellman, D. Mendel, S. Anthony-Cahill, C. J. Noren, and P. G. Schultz. Biosynthetic method for introducing unnatural amino-acids site-specifically into proteins. Meth. Enzym., 202:301- 336, 1991. [8] M. Nassal. 4'-(l-azi-2,2,2-trifluoroethyl)phenylalanine, a photolabile carbenegenerating analog of phenylalanine. 1984. J. Am. Chem. Soc., 106(24):7540- 7545, 216 [9] L. B. Shih and H. Bayley. A carbene-yielding amino acid for incorporation into peptide photoaffinity reagents. Anal. Biochem., 144:132-41, 1985. [10] M. Platz, A. S. Admasu, S. Kwiatkowski, P. J. Crocker, N. Imai, and D.S. Watt. Photolysis of 3-aryl-3-(trifluoromethyl)diazirines: A caveat regarding their use in photoaffinity probes. Bioconjug. Chem., 2:337- 41, 1991. [11] S. K. Silverman. I. Conformational and Charge Effects on High-Spin Organic Polyradicals; II. Studies on the Chemical-Scale Origin of Ion Selectivity in Potassium Channels. Ph.D. Thesis, California Institute of Technology, 1998. [12] M.C. Pirrung and D.S. Nunn. Synthesis of photodeprotectable serine derivatives - caged serine. Bioorg. and Med. Chem. Lett., 2(12):1489- 1492, 1992. [13] M. W. Nowak, J. P. Gallivan, S. K. Silverman, C. G. Labarca, D. A. Dougherty, and H. A. Lester. In vivo incorporation of unnatural amino acids into ion channels in a Xenopus oocyte expression system. Meth. Enzymol., 293:504- 529, 1998. [14] A. P. West Jr. Theoretical Studies of Molecular Magnetism and Molecular Recognition, and Experimental Studies of the Extracellular Domain of the Nicotinic Receptor. Ph.D. Thesis, California Institute of Technology, 1998. [15] K.W. Rickert and B. Imperiali. Analysis of the conserved glycosylation site in the nicotinic acetylcholine receptor - potential roles in complex assembly. Chemistry E3 Biology, 2(11):751- 759, 1995. [16] B. Imperiali and K.W. Rickert. Conformational implications of asparagine-linked glycosylation. Proc. Natl. Acad. Sci., U.S.A., 92(1):97- 101, 1995. [17] M. Quick and H. A. Lester. Methods for Expression of Excitablitity Proteins in Xenopus Oocytes. Academic Press, San Diego, CA, 1994. [18] D.M. Lynn, S. Kanaoka, and R.H. Grubbs. Living ring-opening metathesis polymerization in aqueous media catalyzed by well-defined ruthenium carbene complexes. J. Am. Chem. Soc., 118(4):784- 790, 1996. 217 [19] D.M. Lynn, B. Mohr, and R.H. Grubbs. Living ring-opening metathesis polymerization in water. J. Am. Chem. Soc., 120(7):1627- 1628, 1998. [20] V.W. Cornish, K.M. Hahn, and P.G. Schultz. Site-specific protein modification using a ketone handle. J. Am. Chem. Soc., 118(34):8150- 8151, 1996. [21] W. Zhong. Physical Organic Chemistry on the Nicotinic Acetylcholine Receptor. Ph.D. Thesis, California Institute of Technology, 1998. [22] L.K. Mahal, K.J. Yarema, and C.R. Bertozzi. Engineering chemical reactivity on cell surfaces through oligosaccharide biosynthesis. Science, 276(5315):1125- 1128, 1997. [23] V.W. Cornish, D.R. Benson, C.A. Altenbach, K. Hideg, W.L. Hubbell, and P.G. Schultz. Site-specific incorporation of biophysical probes into proteins. Proc. Natl. Acad. Sci., U.S.A., 91(8):2910- 2914, 1994. [24] L.M. Mannuzzu, M.M. Moronne, and E.Y. Isacoff. Direct physical measure of conformational rearrangement underlying potassium channel gating. Science, 271(5246):213- 216, 1996.