Structural investigations of metabolites from four marine invertebrates

Each of the projects described in this thesis entails either the three-dimensional or covalent structure determination of metabolites from marine invertebrates. The first chapter reviews literature reports of three-dimensional structures of conotoxins. As more structural work in completed for contoxins, additional biological insights are revealed about the binding sites for each of the toxins. A section summarizing the biological implications of the three-dimensional contoxin structures has been included for each pharmacological class of peptides. The second chapter discusses the solution NMR structure of the recently described conotoxin known as psi-PIIIE. Though this peptide contains the same disulfide bonding framework as the µ-conotoxins, its biological activity is unique as it targets the acetylcholine gated ion channel rather than the voltage gated sodium channel. Furthermore, inhibition of the acetylcholine gated ion channel by psi-PIIIE occurs by a different mechanism than the psi-conotoxins, demonstrating that the psi-conotoxin represent a novel pharmacological class of conotoxins. Determination of the three-dimensional structure of psi-PIIIE represents an important initial step in understanding what structural components of the peptides are required for this unique activity. Chapter 3 and 4 describe the isolation and structure determination of several nucleic acid derivatives from the ascidian Didemnum voeltzkowi and the sponge Amphimedon viridis. The covalent structure of nucleic acids is difficult to determine using only NMR spectroscopy, and mass spectroscopic and synthetic techniques were also employed in the complete structure determination of these metabolites. Chapter 5 discusses the isolation of a novel blue pigment from the ascidian Eudistoma olivaceum and includes evidence supporting a structural proposal for the metabolite. Purification of the compound required extensive investigations of the solubility and stability of the compound, which in turn yielded several important clues to the structure of the molecule. The results of chemical degradation experiments on the natural product and numerous spectroscopic techniques are also described, providing considerable evidence for the proposed structure.

STRUCTURAL INVESTIGATIONS OF METABOLITES FROM FOUR MARINE INVERTEBRATES by Scott S. Mitchell A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Medicinal Chemistry The University of Utah August 1997 THE UNIVERSITY OF UTAH GRADUATE SCHOOL SUPERVISORY COMMITTEE APPROVAL of a dissertation submitted by Scott S. Mitchell This dissertation has been read by each member of the following supervisory committee and by majority vote has been found to be satisfactory. Chaif.'·> Chris M. Ireland .. _---) Louis R. Barrows Darrell R. Davis Baldomero M. Olivera C. Dale Poulter THE UNIVERSITY OF UTAH GRADUATE SCHOOL FINAL READING APPROVAL To the Graduate Council of the University of Utah: I have read the dissertation of Scott Sherman Mi tche 11 in its [mal form and have found that (1) its format, citations, and bibliographic style are consistent and acceptable; (2) its illustrative materials including figures, tables, and charts are in place; and (3) the final manuscript is satisfactory to the supervisory committee and is ready for submission to The Graduate School. ., j / II } June 13th, 1997 Date i! \ I /~ \ / L/ "-.-\...".,/ Chris M. Ireland Chair, Supervisory Committee Aj ro~r:or the Major Department l'Uii)I~// ,Glenn D. Prestwich Chair!Dean Approved for the Graduate Council t Ann W. Hail Dean of The Graduate School Copyright © Scott Sherman Mitchell 1997 All Rights Reserved ABSTRACT Each of the projects described in this thesis entails either the three-dimensional or covalent structure determination of metabolites from marine invertebrates. The first chapter reviews literature reports of three-dimensional structures of conotoxins. As more structural work is completed for conotoxins, additional biological insights are revealed about the binding sites for each of the toxins. A section summarizing the biological implications of the three-dimensional conotoxin structures has been included for each pharmacological class of peptides. The second chapter discusses the solution NMR structure of the recently described conotoxin known as 'V-PI I IE. Though this peptide contains the same disulfide bonding framework as the ~-conotoxins, its biological activity is unique as it targets the acetylcholine gated ion channel rather than the voltage gated sodium channel. Furthennore, inhibition of the acetylcholine gated ion channel by 'V-PIllE occurs by a different mechanisn1 than the a-conotoxins, demonstrating that the 'V conotoxin represent a novel pharmacological class of conotoxins. Determination of the three-dimensional structure of 'V-PI I IE represents an important initial step in understanding what structural components of the peptides are required for this unique activity. Chapters 3 and 4 describe the isolation and structure determination of several nucleic acid derivatives from the ascidian Didemnum voeltzkowi and the sponge Amphimedon viridis. The covalent structure of nucleic acids is difficult to determine using only NMR spectroscopy, and mass spectroscopic and synthetic techniques were also employed in the complete structure determination of these metabolites. Chapter 5 discusses the isolation of a novel blue pigment from the ascidian Eudistoma olivaceum and includes evidence supporting a structural proposal for the metabolite. Purification of the compound required extensive investigations of the solubility and stability of the conlpound, which in tum yielded several important clues to the structure of the molecule. The results of chemical degradation experiments on the natural product and numerous spectroscopic techniques are also described, providing considerable evidence for the proposed structure. v TABLE OF CONTENTS ABSTRACT ................................................................................. o. 0 ••••• iv LIST OF FIGURES. 00.0 ••• 000000 •• 0 ••••••••••••••••••• 0 •• 0.0 ••• 00 •••••• 0 0.00 ••• 0. o ••••• 0 0 •••••••• viii LIST OF TABLES o .••• 00 •• 000.0000000 ••••• 0 •••••••••••••••••••• 0 ••••• 0 o. 0 ••• 0.0. o ••• 0 ••• 00 •••• 0 0 ••• xi LIST OF ABBREVIATIONS ... 0 •• 0 •• 0 •• 00.0 ••••• 000 •••••••••• 000000.00 •••• o. 0 •••• 0 •• 0 ••• 0 ••• o •• o.xii ACKNOWLEDGMENTS .. o.o ••••••• 0 0 •••••••••••••••••••••••••••••••••• 0 •••••••••••••••••• 0 ••••• xiv 1. INTRODUCTION AND BACKGROUND ............................................. 0 ••• 1 1.1 Chemical Diversity from Marine Sources .................................................... 1 1.2 Conotoxin Structural Diversity .................... 0 •••••••••••••••••••••••• 0 •••••••••••••••••• 2 1.3 Three-dimensional Structures of Conotoxins ......................... 0 ••••••••••••••••••••• 7 1.4 Perspectives ................. 0 •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 23 1 .5 References ...................................................................................... 26 2. THE THREE-DIMENSIONAL STRUCTURE OF CONOTOXIN tJ'-PIIIE .......... 28 2. 1 Background and Rationale .................................................................... 28 2.2 Results ........................................................... 0 •••••••••••••••••••••••••••••• 29 2.3 Discussion ...................................................................................... 50 2.4 Experimental ................................................................................... 53 2.5 References ................................ , ..................................................... 57 3. THE CHEMISTRY OF THE ASCIDIAN DIDEMNUM VOELTZKOWI ............. 59 3. 1 Pyrrolopyrimidine Natural Products from Marine Organisms ........................... 59 3.2 Natural Products from Didemnum voeltzkowi .............................. ............... 60 3.3 Isolation and Structural Elucidation of Tubercidin Analogs .............................. 61 3.4 Experimental ................................................................................... 65 3.5 References ...................................................................................... 67 4. ISOLATION OF 1,3-DIMETHYLISOGUANINE FROM THE BERMUDIAN SPONGE AMPHIMEDON VIRIDIS ..................... 68 4.1 Background and Rationale ................................................................... 68 4.2 Isolation and Structure Elucidation .......................................................... 69 4.3 Evaluation of Nucleosides from Amphimedon viridis .................................... 71 4.4 Synthesis of Nucleoside Analogs ............................................................ 74 4.5 Experin1ental ................................................................................... 75 4.6 References ...................................................................................... 78 5. ISOLATION OF A NOVEL BLUE PIGMENT FROM EUDISTOMA OLIV ACEUM .............................................................. 79 5.1 ~-Carboline Metabolites from Eudistoma sp. Ascidians ................................ 79 5.2 Project Overview ............................................................................. 82 5.3 Isolation of Novel Blue Pigment. .......................................................... 83 5.4 Structure Proposals for the Blue Pigment ................................................. 87 5.5 Biological Activity ........................................................................... 92 5.6 Experimental .................................................................................. 94 5.7 References .................................................................................... 97 Appendices A. NMR SPECTRA OF CONOTOXIN 'V-PIIIE .......................................... 98 B. FINAL IRMA REFINED RESTRAINT SET USED FOR SIMULATED ANNEALING OF CONOTOXIN U021 ................................................. 106 C. R-FACTOR DEFINITIONS ............................................................... 115 vii LIST OF FIGURES 1 . 1 Major disulfide bonding frameworks of conotoxins ................................... ,.4 1.2 Backbone ribbon with heavy atoms for sidechains for the a-PnlA crystal structure (left), the a-GI crystal structure (center), and the a-GI NMR structure (right), . , 10 1 .3 A comparison of superimposed backbone atoms of the /.1-GIIIA native and /.1- GIIIA R 13A mutant NMR structures .... , .... , ..... ' ............... , ., ......... , ........ 14 1 .4 Superimposed backbone structures of the Hill model of /.1-GlIIB .............. , .... 15 1 .5 Superimposed backbone structures of NMR structure of (0- conotoxins (top) with a ribbon backbone view of one representative structure (bottom), for (O-MVIIC (left), Australian model of co-GVIA, and California model of GVIA ...... " ....... 20 2.1 The primary sequences of /.1-GIIIA, /.1-GIIIB and 'V-PIllE aligned using the conserved disulfide bonding pattern ................................................ , .... 2g 2.2 The fingerprint region of the 400 ms WETNOESY spectrum with sequentially connected protons connected by sol id lines ................................ , ............ 30 2.3 Sequential assignment of fingerprint NOE crosspeaks for residues 16-26 for conotoxin 'V-PIllE .............................................................. " ......... 30 2.4 Summary of the observed NOE crosspeaks which allowed sequential assignment of amino acid spin systems ............................................................... 31 2.5 Residues H 1-C4 with arrows showing key NOE allowing sequential assignment.31 2.6 A plot of the Karplus equation for <\> angles. The regions denoted by the shaded bar correspond to <\> angles not allowed based on steric considerations ............. 32 2.7 Newman projections of the three most stable rotomers about the Ca- Cp bond. Expected coupling constant and NOE patterns are shown for each conformation. 33 2.8 A model of residue 02 build with ideal peptide geometry. The labeled NOE peaks reflect the intensity patterns expected based on the model, and were used for stereospecific p and b assignments ............................. '" ...................... 34 2.9 An overview of the molecular modeling scheme used for conotoxin 'V-PInE ..... 37 2.10 The three-dimensional disulfide folding pattern of conotoxin 'V- PInE ............. 40 2. 11 The scaling factors used for force constants during simulated annealing molecular dynamics calculations ................................................................... .442 2. 1 2 A. An average per residue plot of the RMSD for backbone atoms (dark bars) and all atoms (shaded bars) for final 17 structures of conotoxin 'V- PIllE. B. The per residue plot of the angle order parameter Sq for $ angles ............................. 43 2. 1 3 A $/'V Ramachandran plot including angles from the 17 converged structures of 'V­PIllE. Squares signify nonglycine residues, triangles signify glycine residues. The numbers on the symbols indicate the structure from which the data was calculated .................................................................................... 44 2.14 Stereoviews of the backbone atoms of the 17 converges structures of conotoxin 'V-PIlIE overlaid upon the lowest energy structure after simulated annealing ........ 45 2. 1 5 The backbone atoms of the lowest energy structure for conotoxin 'V-PIUE highlighting the proposed hydrogen bond between C4 and Rll ..................... 47 2. 16 The trans proline bond between Y13 and 014 ......................................... 48 2. 1 7 Ten structures superimposed on the lowest energy structure from modeling runs including Xss angle restraints. The lowest energy structure from a simulated annealing calculation which did not use Xss angle restraints is also included ...... 50 2.18 A backbone ribbon diagram of'V-PUIE, with charged residues shown as stick drawings. The discoidal shape and distribution of positive residues on one face of the molecule are evident in this display .................................................. 51 2.19 Superimposed backbone diagrams for Jl-GIIIA, Jl-GIIIB, and 'V-PIllE ........... 52 2.20 The primary structures of conotoxin 'V-PIllE aligned with Jl- GIlIA and ll-GIIIB. Brackets indicate the positions of ~ turn, I-I designates extended geometry, and V designates the y turn in 'V-PIlIE.minimal number of amino acids .......... " ....... 53 3.1 MS fragment assignments in thermospray LC-MS .................................... 63 3.2 The 5 Hz HI '_ H2' coupling constant distin~uishes between the anomers of the sugar, demonstrated as follows. (A) The H NMR proton spectrum of 3, highlighting the HI ,_ H2' coupling constants. (B) Energy minimized models of the a and ~ anomers of 3, looking down the C2' -C 1 I bond .............................. 64 4.1 The retro Diels-Alder reaction yielding the neutral methylcyanamide and observed mlz 122 ion in the EI mass spectrum .................................................... 71 4 . 2 The HPLC trace of nucleoside fractions from A. viridis (Top) and the resul ting ESI( -) mass spectra identifying inosine (upper left), thymidine (upper right), guanosine (lower left), and 1,3-dimethylisoguanine (lower right) in the chromatograph ............................................................................. 73 4 .3 The Vorbruggen coupling reaction applied to theophy Hine ........................... 74 4.4 Methylation on N3 of compound 16 using CH3I. ..................................... 75 5.1 Comparison of IH NMR spectra before (A) and after (B) anion exchange ix chromatography collected using 300 mg compound. Peaks at 3.3 and 4.9 ppm result from CD30D ........................................................................ 86 5.2 The 13C NMR spectrum of 15 mg of blue pigment in 250 mL 750/0 CDCl/25% CD30D ............................................................................. · ......... 87 5 . 3 A section of the aromatic region of an HMQC spectrum, with boxes surrounding several suggested duplicate signals arising from conformational or chemical equilibrium .................................................................................. 88 5.4 IH NMR spectra collected at _300 and 260 C ........................................... 89 5 . 5 Summary of proposed structures for fragment ions observed in and APCI+ daughter ion scan .......................................................................... 91 5 . 6 The structure of the proposed blue pigment, highlighting the portions of the structure likely derived from tryptophan (bold bonds), proline (dashed bonds), and tyrosine (light bonds) ...................................................................... 92 5 . 7 Data from an HCT 116 cytotoxicity assay at several concetrations of blue pigment. The absorbance at 540 nm measures the amount of MTT which has been metabolized by living cells, but the UV absorbance of the blue pigment interferes with this measurement. .................................................................... 93 5.8 The catalytic inhibition of topoisomerase I determined at various concentrations of blue pigment. ............................................................................... 94 x LIST OF TABLES 1.1 Overview of published conotoxin NMR structures ........................................ 6 1 .2 Amino acid sequences of peptides in the cystine knot motif ............................ 19 2.1 IH Resonance assignments for contoxin \jJ-PIIIE ...................................... 31 2.2 Coupling constants and NOE patterns allowing stereospecific assignments for contoxin \jJ-PIIIE ....................................................... 33 2 . 3 Data reflecting exchange rates of amide protons .......................................... 46 3.1 I Hand \3C Assignments fo compounds 5 and 3 ...................... " ................. 65 APCI CCC CID CVFF CH) CSa~ DEAE DG DMF DMSO DQF-COSY EI-MS LIST OF ABBREVIATIONS atmospheric pressure chemical ionization countercurrent chromatography collision induced dissociation consistent-valence forcefield methyliodide cysteine stabilized a~ motif diethylaminoethyl distance geometry N,N-dimethylformamide dimethylsulfoxide double quantum filtered correlation spectroscopy electron impact mass spectroscopy ESI electrospray ionization F ABMS fast atom bombardment mass spectroscopy GMQFCOSY gradient multiple quantum filtered correlation spectroscopy HCT human colon tumor HMBC heteronuclear multiple bond correlation HMDS 1,1,1,3,3 ,3-hexamethy ldisilazane HMQC heteronuclear multiple quantum correlation HPLC high performance liquid chromatography HR IC50 IRMA K2C03 LC-MS MD MS MIT NOESY o PECOSY PPM RMA RMD RMSD TFA lLC TOCSY TMSCI TMSOTf UV high resolution 50% inhibitory concentration iterative restrained matrix approach potassium carbonate liquid chromatography mass spectroscopy molecular dynamics mass spectrometry 3-[4,5-dimethylthiazoy-2-yl]-2,5-diphenyltetrazolium bromide nuclear Overhauser enhancement spectroscopy trans-4-hydroxyproline primitive exclusive correlation experiment parts per million restrained matrix approach restrained molecular dynamics root mean squared deviation angular order paranleter longitudinal relaxation time correlation time trifluoroacetic acid thin layer chromatography total correlation spectroscopy trimethylsilyl chloride trimethy lsily lmethy I trifluoromethanesulfonate ultraviolet xiii ACKNOWLEDGMENTS I would like to thank Chris Ireland for his guidance, support, and friendship throughout my studies at the University of Utah. Lou Barrows is responsible for the biological activity data which guided many of natural product projects described in this thesis, and for encouragement and advice. Toto Olivera and his research group provided the peptide sample used in the structural studies of 'II-PI I IE, as well insights regarding the biology of conotoxins. Darrell Davis and Jay Olsen deserve credit for management of NMR facilities, and for advice on many topics. Collection of marine organisms would not be possible without assistance from collaborators, including Giselle Concepcion in the Philippines, Henry Trapido-Rosenthal in Bermuda, and Bill Baker in Florida. Piotr Lassota performed additional biological activity testing on many of the compounds described in this thesis. I would like to thank Elliot Rachlin, Andy Whitehill, and Steve Pomerantz for acquiring mass spectral data. The faculty and members of the Medicinal Chemistry department have all contributed their knowledge and experience throughout my studies, and provided a pleasant atmosphere for scientific development. Finally, I thank all of the members of the Ireland research group, past and present, for their guidance, assistance, and friendship_ 1. INTRODUCTION AND BACKGROUND 1.1 Chemical Diversity from Marine Sources Natural products drug discovery is a rapidly evolving field, in large part because of the enormous chemical diversity that has arisen throughout the marine environment. The oceans have proven to be an prodigious source of novel natural products, providing many examples of novel and biologically active classes of compounds. Novel chemical functionalities discovered through marine natural products chemistry have been used by medicinal chemists in the design of new drugs, and aided in the discovery and development of new pharmacological targets used to direct drug development. The fundamental driving force for these advances is the discovery of new biologically relevant molecular configurations; molecules that have shapes, sizes, and functionalities to allow them to undergo a specific biochemical interaction with a receptor or enzyme target. The unifying theme of work described in this thesis is the description of novel three-dimensional and covalent structures of marine natural products. The natural world is a sensible place to search for these biologically important compounds. Despite the variety of life, all biological organisms have fundamental biochemical mechanisms in common, hence molecules synthesized by one organism to fulfill a specific biochemical role may have unique interactions with an enzymatic receptor of very distantly related organisms. The result for chemists is that while the natural world has created an unparalleled diversity of chemicals, the requirement that each of these compounds performs a specific biochemical function for the host organism increases the likelihood that the compounds might playa role in other biological systems. The conotoxins represent a rapidly expanding area of marine natural products, and serve as an excellent example of a class of organisms that has used the development of chemical diversity to its advantage. 1.2 Conotoxin Structural Diversity The conotoxin peptides are emerging as a premier example of a source of molecular diversity from natural sources. The biology and ecology of the predatory marine snails have led to the development of a peptide library, encouraging both structural diversity, and at the same time, a high degree of biological relevance. Structural diversity has resulted when snail species diverge, and adapt the components of their venom to fill a new ecological niche. A high degree of biological relevance is assured for these peptides, as their biological targets are presumably the receptors of the nervous system, assuring selection pressure for the continued production of only biologically active peptides. The conotoxins are small, conformationally-constrained peptides found in the venoms of the ca. 500 species of carnivorous marine snails belonging to the genus Conus. Each of these peptides is a highly specific ligand that patently affects receptor or ion channel function. In general, the cone snails use venom as the primary tool for paralyzing prey; a Conus venom can be extremely complex; many contain well over 50 different peptides. Furthermore, each different Conus species has an entirely different complement of peptides in its venoms. The conotoxins are nucleic acid-encoded, and hyper mutation of conotoxin­encoding genes occurs as Conus species diverge. The amount of sequence divergence between conotoxins when two Conus species are compared can be remarkable. It has been suggested that the cone snails have employed the equivalent of a combinatorial peptide library strategy to quickly evolve new peptide sequences. In the course of their evolutionary history, Conus species divergence has apparently resulted in considerable specialization both with regard to prey type, and prey capture strategies. It seems likely that the individual Conus venoms reflect this specialization. 3 The conotoxin peptide library suggests unifying themes for molecular diversity from natural sources: the generation of diversity based on general structural motifs, and specialization of binding affinity to receptor sUbtypes to avoid the production of redundant or competing toxins. Among the diverse molecular forms of Conus peptides, there is a conserved structural element: the disulfide-bonded framework that confers conformational rigidity to these peptides. There are three major conserved disulfide-bonded motifs (shown in Figure 1.1). One of these, the O-type framework, is shared by several different pharmacological families of conotoxins including the Ol-, 0- and J.l-O-conotoxins (examples of three Ol-conotoxins are shown). The conserved disulfide framework helps to assure that mutations in the variable loops will give rise to a stable three dimensional structure. A peptide with a rigid structure is more likely to have strong binding affinity to a receptor, because the entropy of restricting a flexible peptide to a single conformation need not be overcome. Hence, in part because of the development of stable structural motifs, the cone snails have been able to successfully develop thousands of peptide analogs. It has been suggested that the unprecedented degree of binding specificity of some conotoxins for one receptor subtype may be a result of evolutionary pressure to avoid the development of redundancy or antagonism within the family of peptide venoms. By targeting receptor subtypes, each novel peptide develops specific interactions with a different target. That there are only limited structural motifs is not a deterrent to this process, as the overall similar size and shape of two related peptides suggests that they should easily be able to adapt to the similar size and shapes of the related receptors. The stable conotoxin framework can be exploited to present a different three-dimensional array Conotoxin Examples: a-conotoxin GI a-conotoxin S I a-conotoxin Pn IA I II I C---C---CC---C---C I I Examples: Ol-conotoxin GVIA Ol-conotoxin MVIIC Ol-conotoxin MVI IA I I I I CC---C---C---CC I I Example: f.1-conotoxin GIIIA f.1-conotoxin GIIIB \fI-conotoxin PII IE Primary Sequence EOCN-PAOGRHY--SC* GRCCH-PAOGKr\lY--SC* GCCSLPI:?CAAl\l:NPDYC* O-type CKSCX3SSCSGrSYNCC-RSCNOYTKRCY* CKGK~SGSOGRRGK-C* CKGKGAKCSRlMYICCIGSC--RSGKC* M-type RICCTCOKKCKDRQ-CKCQRCCA* RICCTCORKCKDRR-CKO:M:KCCA* Hcx::ccLYGK-CRRYCX3CSSASCCQR* Figure 1.1 Major disulfide bonding frameworks of conotoxins. 4 of amino acid side chain functionalities, in a manner which maximizes potential positive interactions with the new receptor sUbtype. Thus, although the peptides found in Conus venoms comprise many thousands of different molecular forms, there are unifying structural themes. The three arrangements of cysteine residues shown in Figure 1.1 are found in ca. 80% of all Conus peptides sequenced so far; however, alternative Cys frameworks are found. Examples are the conantokins, conopressins, conodipines and aA-conotoxins. 5 By understanding the three-dimensional structural motifs of the conotoxins, an abundance of information regarding their physiological targets can be inferred. In effect, a conotoxin can serve as a framework for mapping amino acid residues of the target protein in the immediate vicinity of the bound ligand, thereby allowing exploration of the structure around the ligand binding site of the target receptor. Furthermore, because several Conus species target the same ligand binding site, peptides with significantly divergent sequences can be identified. A wealth of structure/function information is then made available, which theoretically becomes available for sophisticated drug design. NMR structures have now been published for at least one member of each of the structural classes of conotoxins (Table 1.1). The successful crystallization of two different a-conotoxins provides even higher resolution structures of conotoxins, though it is important to note that neither NMR or crystallography are likely to yield an accurate picture of most amino acid sidechains. Nonetheless, as additional structures are solved, comparison of the similarities and differences in related structures will be instrumental in describing key elements which give rise to the extraordinary binding specificities of the conotoxins. 6 Table 1.1 Overview of published conotoxin NMR structures Conotoxin Year Distance Molecular Back Software Author Published Geometry D~namics Calculation Packa~e a-Gl 1989 * DSPACE Pardi a-G1 1989 * DADAS Kobayashi a-S1 1993 * DSPACE Christensen Jl-GIIIA 1991 * DISMAN Ott Jl-GIIIA 1991 * * X-PLOR Lancelin Jl-GIIIA 1992 * * X-PLOR Lancelin co-GVIA 1993 * * * DSPACE Pallaghy co-GVIA 1993 * * VEMBED Davis GROMOS co-GVIA 1993 * * X-PLOR Pardi co-GVIA 1993 * * DIANA Sevilla GROMOS co-MVIIC 1995 * X-PLOR Japanese co-MVIIC 1995 * * VEMBED Farr-Jones SANDERS co-MVIIA 1995 * X-PLOR Japanese Il-GIIIB 1997 * X-PLOR Hill 7 1.3 Three .. dimensional Structures of Conotoxins 1.3.1 The a, .. conotoxins. The a,-conotoxins are a family of Conus peptides that inhibit nicotinic acetylcholine receptors. Among the major groups of Conus peptides, a,-conotoxins are the smallest; most are 1 17 amino acids, with two disulfide bonds. Consequently, for structure determination using multidimensional! H NMR spectroscopy, the relatively small size of the a,-conotoxins creates two problems. The estimated molecular correlation time of these small molecules is below one nanosecond, and thus it is close to the time scale where theoretical NOE intensity is near zero.! In addition, these molecules lack the tertiary structure necessary to generate long range distance constraints vital to the determination of high resolution structure. These inherent molecular characteristics have created difficulties for obtaining well-converged structures for this family of toxins. Nonetheless, two NMR structures of a,-conotoxin Gr have been published, a preliminary structure for a,-conotoxin Sr has been completed and further refinement is currently underway. The first NMR structures of any conotoxins published were investigations of a,­Gr published by groups in Colorad02 and Japan.3 The Colorado structure is probably the most biologically relevant since the NMR spectra were collected in an aqueous medium. Forty-nine distance constraints were generated using 200 and 300 ms NOESY spectra collected in H20ID20 at 5° C. No dihedral angle restraints were solved from 3JH coupling constants, and stereospecific assignments for prochiral atoms were not analyzed using Wagner's method.4 .5 Instead, distance geometry structures were manually examined and the proton which best satisfied the NOE data was used in the restraint. The modeling scheme employed distance geometry using an algorithm included in the software package DSPACE followed by constrained energy minimization of distance geometry structures. Energy minimization was performed using AMBER molecular mechanics software. Eleven structures were generated by distance geometry and 10 of these were energy-minimized in AMBER. The final 10 structures had a backbone rms deviation of 1.53 A and 1.62 A for all heavy atoms. 8 All 10 structures show two tight turns. The first turn is from N4 to C7 while the second turn is between Q8 and Yll. In most of the structures, the N4 carbonyl is generally within hydrogen bonding distance of the C7 amide proton indicating the first tum is a regular ~ turn. The second turn can be described as a ~ turn or two y turns. A previous model for the structure of a-G I was proposed by Gray6 using Chou and Fasman rules.7 This model predicted two ~ turns stabilized by the two disulfide bonds. The positions of the turns in this structure are in general agreement with this model. The Gray model also predicted the presence of a salt bridge between N 1 and R9. The salt bridge is not present in the Colorado structure as the two residues reside on opposite faces of the molecule. One model for competitive antagonists, primarily based on curare, requires the presence of two acetylcholine mimicking groups 11 A apart.8 Each group is composed of a cationic center separated by approximately 5 A from an electronegative group. By analogy, the two cationic groups in a-GI would be predicted to be the N-terminus and the guanidino side chain of R9. The average distance between the positively charged groups in the NMR structure is 15.5 A. Though this is larger than predicted by the alkaloid model, it has been proposed that small rotations of the torsion angles of the R9 side chain, perhaps facilitated by toxin binding, could easily move the groups closer together. 2 The Japanese structure of a-GI was determined using NMR data collected in DMSO.3 The structures were generated using a distance geometry algorithm included in the software package DADAS. The algorithm calculates a total error violation function which includes terms for restraint violations and repulsive core radii violations. The error function is minimized by randomly varying dihedral angles. One hundred random 9 structures were minimized and the 10 with the lowest total error function were used for the structural analysis. The rms deviation calculated for the backbone atoms among all conformers was 2.32 A whereas the rms deviation for all atoms was 3.54 A. The authors categorized the 10 structures into two groups of conformers that differed in the orientation of the tyrosine sidechain. Although the NMR data imply the presence of regular tight turns in the structure, the authors do not discuss the tertiary structure of their final model other than to state that the backbone folding pattern appears quite consistent with the Gray model, and could fit the proposed alkaloid and peptide toxin binding sites. Minor disagreements were considered to be due to the effects of DMSO. Two reports of crystal structures for conotoxin were reported in 1996, describing the structure of conotoxin a-GI at 1.2 A resolution, and a-PnIA at 1.1 A. The crystal structure of a-G I is generally in agreement with the NMR structure published by Pardi, with a backbone RMSD between an energy minimized NMR structure and the crystal structure of 1.6 A. The overall assignments of secondary structure are compatible with the NMR structure, and comparisons of the two structures are shown in Figure 1.2. Neither the structural effects of crystallization nor distortions due to lack of NMR data can be quantified, so the general agreement of the crystal and NMR structures for a-G I may serve as a benchmark for the overall degree of precision for conotoxin structures. Conotoxin a-PnIA and a-PnIB differ from other a-conotoxins in that they have been shown to inhibit neuronal nicotinic acetylcholine receptors (rather than skeletal muscle receptors). Consistent with this, the primary sequence of a-PnIA and a-PnIB differs from other a-conotoxins in that the length of the loops defined by the disulfide bridges are increased from three and five residues in most a-conotoxins, to four and seven residues, for a-PnIA and a-PnIB (Figure 1.1). The crystal structure of a-PnIA shows two well­defined secondary structural features, one type I ~ tum between residues 2-5, 10 .' ..... .1-..,.".. '. · C"".,.·..../. ·-.·..'.--.(\.A ... ·.. .". ,.•.· .·,··, . ,h • ",,,..-,\ . f'~ ~ Figure 1.2 Backbone ribbon with heavy atoms for sidechains for the a-PnlA crystal structure (left), the a-Ol crystal structure (center), and the a-Ol NMR structure (right). followed by two helical turns from residues 5-12. For several of the helical residues, the main chain hydrogen bonding pattern shows both 310 helix characteristics as well as a helix interactions, though the authors note many of the hydrogen bonds appear to be weak. A comparison of the structures for a-O I and a-PnlA reveals significant structural differences. The additional amino acids inserted into the disulfide loops of a-PnlA appear to strongly affect the ability of the peptide to adopt a helical conformation. Potential structure-activity relationships are intriguing, as antagonists of muscle acetylcholine receptors generally incorporate two positive charges, whereas a-PnlA contains only the N­terminal amino group. Since the structures of a-Ol and a-PnlA are significantly different, the structural requirements for receptor subtype specificity are not immediate]y evident. However, further structure-activity analysis using point mutation and other biochemical techniques should allow a more complete description of which structural characteristics determined by the a-conotoxin models are critical for biological activity. 1 I A preliminary structure for a-conotoxin Sr has also been reported.9 The a-Sr sequence differs from a-O r by three amino acids. a-S r blocks fish acetylcholine channels specifically whereas a-Or blocks acetylcholine receptors in both fish and mammals. A set of 50 structures was generated using distance geometry combined with energy minimization using BIOSYM software. Fifty distance constraints were included, and all but two dihedral angle constraints were added for stereo specifically assigned ~ protons. The backbone of the resulting structure can be overlaid convincingly with the backbone of the Colorado a-O r structure. However, the two structures do not agree with respect to placement of the cationic centers required by the curare model. The overall structure of the acetylcholine receptor from the electric organ of torpedo has been studied to a greater extent than almost any other ion channel. Electron microscopy and other techniques have shown that the muscle acetylcholine channel consists of five subunits embedded in the membrane in a circular fashion. The pentamer consists of two a subunits and one ~, yand 0 subunit. NUITlerOUS attempts have been made to localize toxin binding sites to specific subunits on the channel. Chimeras of the yand 0 subunits have been made to localize the a-conotoxin binding site to specific ayand ao subunit interfaces.10 These approaches have been successful in narrowing down binding site determinants for specific a-conotoxins to the level of individual amino acids. It should be noted that the majority of these studies have not directly employed a-Or. Interpretation of these results in terms of the a-Or structure is therefore an uncertain approximation. Future refinement of biochemical techniques and NMR methods may result in a model which could help exploit the potential of a­conotoxins as specific probes for acety lcholine receptors. 1.3.2 The J.l-conotoxins. The J.1-conotoxins are a family of Conus peptides which block voltage-gated Na channels. Three J.1-conotoxins from Conus geographus venom have been purified, characterized and chemically synthesized. The J.1-conotoxins 12 characterized so far are all specific for the skeletal muscle sUbtype of Na channels, and are 22 amino acids in length with three disulfide bonds (Figure 1.1). They target Site I on voltage-gated Na channels, and competitive binding with the guanidinium alkaloids tetrodotoxin and saxitoxin has been described. The only J.l-conotoxin NMR structures completed to date are of J.l-GI IIA. In 1990 a German group published the first structure of J.l-G I I IA.II In 1991 a consortium of Japanese groups published a series of papers that described the native structure of J.l­GIIIAI2 and a J.l-GIIIA analog with an arginine 13 alanine (RI3A) point mutation. 13 ,14 The initial J.l-G I I IA structure by the Japanese groups was built in is extended conformation using the QUANT AlCHARMm software package. In order to achieve complete sampling of conformational space in the initial structure set, random initial velocities were assigned to atoms according to a Maxwell distribution at 1000 K.15 Only distance constraints including pseudoatom corrections I were used during this first step These structures were further refined during a second stage of dynamics and simulated annealing,16 in which dihedral angle constraints and covalent disulfide bonds were introduced. Distance restraints were generated by measuring the intensity of cross peaks in a 250 ms NOESY experiment Sixty-two sequential and 24 long range constraints were generated. The ~ protons for seven residues were stereospecifically assigned and 11 angles were assigned from coupling constant measurements. The final structures contain a successive pair of tight turns from D2 to T5 and from T5 to K8. The C terminal region is comprised of a loop from D 12 to C 16 followed by a small right handed helix containing one turn up to Q 18. This is followed by a final loop which places the C-terminus in an almost opposite direction to the N-terminus. All of the positively charged residues from I to 11 reside on one face of the molecule and the three positively charged residues from 13 to 19 on the opposite face. 13 The structure-acti vity relationships for each of the residues of Il-G I I IA were examined by synthesizing mutants in which each of the residues was sequentially replaced by alanine. 14 It was determined that the basic residues of Il-G I I IA were particularly important. Arginine 13 was found to be extremely sensitive to mutation since even lysine substitution decreased the toxicity of the molecule significantly. Based on these findings, the researchers solved the NMR structure of the R 13A mutant to examine the effect of this mutation on the overall structure of the molecule. To facilitate comparison of native and mutant structures, the native structure was recalculated using the same parameters as the mutant structure. A comparison of diagonal plot representations for the NOEs from the native and mutant structures shows that only a few subtle changes occur in the spectra. It is therefore expected that the final calculated structures be almost the same except for areas directly associated with the mutated residue. Initial structures were generated by randomizing the <I> and \}1 angles for extended structures. All ro angles were constrained to trans geometry except for 07 which was set to cis geometry based on the previous structure calculations. The final structures were generated using the simulated annealing protocol YASApI6 ,17 included in X-PLOR. For the final run, 200 structures were generated for both molecules and the 10 best were used in the structural analysis. The backbone rms deviation for residues 2-21 for the native peptide was reported as 0.48 A and for the R13A mutant as 0.45 A. Values for all heavy atoms in residues 2-21 were 1.18 and 1.10 A. Inagaki's initialll-GIIIA structure showed significant disorder between residues K II-R 13. This was not reproduced in the second study and demonstrates that the specific modeling protocol used can have significant effects on the final calculated structures. This illustrates that it is important to compare the final structures with the NMR data to ensure the behavior of the predicted model is reflected in the experimental data. 14 A comparison of the backbone structure of the native peptide and the R 13A mutant shows that they are qualitatively the same, indicating that in this case the mutation has not changed the overall shape of the toxin. Figure 1.3 shows the best fit superposition of 10 structures of native Jl-conotoxin GIIIA (A) and the R(13)A mutant (B). This is strong evidence that the guanidinium group of arginine 13 has a specific interaction with the receptor. The German structure of Jl-GIIIA was derived using the program DISMAN for distance geometry calculations. lI Approximately 120 distance constraints were generated by measuring NOE buildup curves from a series of spectra with various mixing times. An additional 700 "non-NOE" constraints were added by restraining atoms that did not show NOE cross-peaks to be five angstroms or greater apart. The final structure appears to be similar to the structure calculated by the Japanese groups, but no parameters were reported that would allow a more quantitative analysis. Since this structure is the result of only distance geometry calculations it is possible that it does not represent the 10westenergy conformation of the molecule. G lilA Native GIlIA R13A Mutant Figure 1.3 A comparison of superimposed backbone atoms of the Jl-G I I IA native and Jl-GIIIA R13A mutant NMR structures 15 An NMR structure for ~-conotoxin GIIIB was published in 1996 by a group consisting of Hill, Alewood, and Craik, describing an overall fold consistent with previous models of ~-GIIIA. The similarity in the three-dimensional structures of ~-GlIIA and Jl-GIIIB is not surprising, as the peptides differ by only four amino acids. The X-PLOR program was used to perfonn simulated annealing calculations, with psuedo-NOE restraints replacing disulfide bonds. Final minimization with disulfide bonds replaced was perfonned using the CHARMm program. Trends in the J3C chemical shifts and NOE patterns for hydroxyproline residues were examined to identify amino acids in the trans confonnation. These methods suggest the 06-07 peptide bond may have cis geometry, but this was not experimentally verified, and models submitted to the Brookhaven Protein Data Bank have all trans geometry. Hill suggests the secondary structure of ~-G I I IB contains a section of anti-parallel ~ sheet followed by a C-tenninal helix; a more complete discussion of the secondary structure of ~-GIIIA, ~-GIIIB, and 'V-PIllE is included in Chapter 4. The final 20 structures that contained lowest energies are shown in Figure 1.4. The ~-conotoxins have been shown to compete for the same binding site as the alkaloid toxins tetrodotoxin and saxitoxin. 1 8 In light of this, the requirement for a specific arginine residue for activity in the ~-conotoxins is not surprising. The guanidinium Figure 1.4 Superimposed backbone structures of the Hill model of Jl-GlIIB. 16 binding site has been characterized in considerable detail. Though it is clear that the Jl­conotoxins interact with the receptor in different ways than alkaloid toxins, it is intriguing to speculate about potential interactions between the Jl-G I I IA structure and the models for the guanidinium binding site in the sodium channel. Though the structures of tetrodotoxin and saxitoxin are not related, the specific functional groups believed to be involved in binding, a guanidinium and two hydroxyl groups show remarkable similarity. The functional groups important for toxin binding have been investigated extensively and are reviewed by Kao. 19 The shared binding site for the toxins is most likely located in the outside mouth of the ion penneation pathway. Studies using carboxyl modifying reagents have shown that one or more negatively charged carboxyl groups on the channel protein play an essential role for toxin binding. Mutation studies using the a subunit of the sodium channel have also shown which amino acids in the sodium channel are critical for guanidinium toxin binding. 20,21,22 This work culminated in the prediction that four highly homologous peptides of the a subunit fonn the guanidinium toxin binding site. 22 . 23 A group at the University of Chicago used the Chou and Fasman rules to predict that the four peptides would most likely have ~ sheet topology and contain four ~ turns. Molecular modeling was then performed on the peptides using the predicted Chou and Fasman rules as guidelines. The resulting model contains several properties that are consistent with both toxin mutation and receptor mutation experiments. At the heart of the Chicago receptor model are four carboxyl residues that mutation studies have shown are required for toxin binding. The four negatively-charged residues are predicted to complex the positively charged guanidinium ion and hence block further ions from passing through the entrance to the channel. In the Chicago model the four negatively-charged groups are in a position where they may all interact with the same cation. In this model the two guanidinium groups of saxitoxin stabilize binding to a greater 17 extent than the single guanidinium group in tetrodotoxin. Saxitoxin is predicted to interact with all of the four subunits whereas tetrodotoxin binds only to subunits I and II. The Chicago group used the sodium channel vestibule model to investigate interactions between Jl-G I I IA and the proposed binding site. 18 A key observation is that a single E758Q mutation of adult rat skeletal muscle decreases the J.l-GIIIA affinity by 48- fold. The mutation also affects the binding of tetrodotoxin which is additional evidence for J.l-GIIIA binding in the same outer vestibule region. When the J.l-GIIIA structure was positioned in the binding site model it was observed that there were two potential orientations of the R13 guanidinium group. The position that was considered most favorable was one in which the guanidinium ion (as well as the rest of the molecule) is rotated 1800 from the analogous position in tetrodotoxin and saxitoxin. It was observed that the different alignment of the J.l-GIIIA guanidinium ion could explain the different effects of sodium channel mutations for J.l-conotoxins and tetrodotoxin. 1.3.3. The oo-conotoxins. The oo-conotoxins are a family of Conus peptides which block voltage-gated calcium channels. These are among the larger conotoxins, typically 25-30 amino acids with three disulfide bonds (Table I). Different (O-conotoxins can have different specificity for Ca channel subtypes. Thus, oo-conotoxin GVIA is highly specific for als-containing Ca channel complexes, whereas (O-conotoxin MVIIC has a higher affinity for alA-containing Ca channel subtypes. Four separate groups have published proposed structures for oo-conotoxin GVIA based on NMR data. This provides an opportunity to compare results from the variety of methodologies used for determining the different structures. The structures of 00- conotoxins MVIIA and MVIIC have also recently been published allowing some structure/activity analysis. The oo-GVIA structure published by an Australian group is significant because it utilizes many of the computational techniques available today.24 Structures were first 18 embedded into a metric matrix using the distance geometry algorithm included in DSPACE software. When embedded structures were submitted to simulated annealing protocols, the rms deviations and Ramachandran plots showed little change. Thus, the simulated annealing step was skipped in the final calculations. The criteria for embedded structures to be used for further calculations included the degree of convergence as well as consistency with the NMR data. The back calculation protocol included in the program BKCALC was used to assign additional NOE cross-peaks and evaluate the accuracy of the structure by comparing the predicted spectra with experimental data. When peaks were observed in the theoretical NOE spectrum and not in the 400 ms NOESY a lower bound restraint of 3-4 A was entered. The 12 final refined structures were then subjected to restrained energy minimization using the Discover program. The initial constraint set for ffi-GVIA generated from a 300 or 400 ms NOESY spectrum consisted of 175 distance restraints. After back calculation refinement, 35 additional restraints were added. The refined structures are described as a small antiparallel triple-stranded ~ sheet. Each strand in the sheet is very short and limited to two hydrogen bonds. The extent of structure is limited by turns involving residues 3 to 6, 9 to 12, 15 to 18, and 21 to 24. Tum one is a type II tum, tum two a type I tum, tum three a variant of a type III turn, and the fourth tum is not well resolved in the NMR structure. The Australian group performed a search of known three-dimensional peptide structures and found that the ro-GVIA structure and several other peptides (see Table 1.2) contain a common motif, describes as an inhibitor cystine knot motif. 25 The common features of the structural motif include the cystine bond topology and a triple-stranded ~ sheet. Many peptides are stabilized by mUltiple disulfide bonds, but those included in this discussion also contain the same three-dimensional arrangement of the cystine bonds. Topologically only the kalata peptide forms a true knot structure since its cyclic backbone prevents continuous deformation into unknotted structures. The other peptides discussed 19 here are pseudolinks, though the disulfide bonds make continuous deformation impossible. The other molecules containing the common structural elements are shown in Table 1.2. The biological targets of the peptides in the cystine inhibitor knot family are diverse. One general similarity is that all men1bers of the family appear to act as antagonists for their binding sites. It has been hypothesized that "the primary role of the inhibitor cystine knot motif is to provide a compact and stable framework for the presentation of active residues for a specific binding interaction.,,25 Justification for this statement is that throughout nature there are a limited number of ways to stabilize a and B structure. An extension of this is that there are a limited number of stable structural motifs, which in tum have arisen independently numerous times in nature. NMR structures of ro-GVIA have also been published by groups from Spain, California, and Colorado. Details for the modeling protocols used for these structures have been included in Table 1.1. All of the published reports for ro-GVIA have found the same triple stranded B sheet structure. However, the coordinates for some of these structures were not deposited in a public database such as the Brookhaven Protein Databank,26,27 thus it is difficult to compare the structures in a more detailed manner. The California structure is the best converged of all of the ro-GVIA structures published.28 The total backbone rms deviation of 0.58 A clearly defines the backbone angles for all residues. A comparison of the Australian and California backbone structures are presented in Figure 1.5. The modeling strategy consisted of distance geometry combined with two thorough molecular dynamics steps. The first dynamics calculation did Table 1.2 Amino acid sequences of peptides in the cystine knot motif. Molecule Sequence Receptor GVIA C KSOGSS C SOTSYN C C RS C NOYTKR C Y Ca Chan. Kalata NGLPVC GET C VGGT C NTPG C T C SWPV C TR Unkown CMTI-I RVC PRLIME C KKDSD C LAB C V C LEHGY C G Trypsin EETI-IT GC PRILMR C KTHDD C AG C V C GPNGF C G Protease CPI EQHADPIC NKP C CSGAWFC QA C WNSART C G* Protease AgaIVB EDNCIAEDYGKCTWGGTKC C RGRP CRCSMIGNTCE C T* Ca Chan. * Additional C-terminal amino acids present 20 Figure 1.5 Superimposed backbone structures of NMR structure of 0)­conotoxins (top) with a ribbon backbone view of one representative structure (bottom), for ro-MVIIC (left), Australian model of ro-GVIA, and California model of GVIA. not include the disulfide bonds to allow the structure enough freedom to find the global minimum. No back calculation steps were performed and the structures were not deposited in a public database. The O)-GVIA structures published by the Colorad029 and Spanish30 groups are quite similar. Both groups performed distance geometry and simulated annealing calculations. The final backbone by rms deviation reported the Colorado group for the 20 best converged structures was 1.70 A. The 0.82 A backbone rms deviation reported by the Spanish group included only eight structures in the calculation. It is worth noting that both structures showed little convergence between residues 9 to 16. It is not clear whether this reflects dynamic motion or paucity of NOE data for this portion of the molecule, since no data were collected that reflects the flexibility of the molecule. 21 Two NMR structures of the synthetic ro-conotoxin MVI IC, a brief communication by a Japanese group,31 and more detailed report by a California group, have recently been published. The latter report will be discussed in detail as it employs more rigorous modeling protocol and a complete description of the fmal structure. A standard modeling approach of generating structures via distance geometry and refining them using simulated annealing was employed.32 The iterative relaxation matrix protocol included in the software MARDIGRAS was used to generate distance restraints from the NMR data. Dihedral angle constraints and stereospecific assignments were included when NMR data satisfied the criteria described by Wagner. 4.5 The structures were refined using a back calculation protocol. The quality of the structures was judged by using the software CaRMA to compare the simulated NOESY spectra with experimental data. Non-NOE restraints were generated from this analysis and included in subsequent structural calculations. The non-NOE restraints were quite conservative since protons which did not show experimental NOE correlations were restrained to be only 3.5 A or greater apart. The best 15 structures (see Figure 1.5) show a backbone rms deviation of 0.84 A while the rms deviation for all heavy atoms was 1.18 A. The angular order parameter was also used to describe the final structures. The angular order parameter measures the homogeneity of <I>, 'II, and X angles. This calculation found three main areas of backbone disorder: residues C1 to K4, C15 to G 18, and G21 to R22. The overall structure shows the same triple-stranded ~ sheet topology as ro-GVIA. However, the overall solvent accessibility for the model was greater. This assessment stems from comparison of the previous ro-GVIA structure by the same California group which contained ten slowly exchanging amide protons versus three for ro-MVIIC. Regular ~ turns are found in roughly equivalent positions as in ro-GVIA. .' 22 A structure has recently been published for conotoxin ro-MVIIA by a different Japanese group (Kohno et al., 1995). NOE spectra were collected using mixing times of 100,200,300 ms to give 251 final distance constraints. The X-PLOR simulated annealing protocol was the only modeling procedure used in the calculation of the structure. One hundred structures with randomized coordinates were annealed which resulted in 13 converged structures. The final structure did show good agreement with the experimental data as no NOE violations greater than 0.5 A were reported. No further experiments to analyze the quality of the structure by back calculation or biochemical methods were performed. As expected, the ro-MVIIA structure shows the same overall triple-stranded ~ sheet topology as described for ro-GVIA and ro-MVIIC. The ~ strand regions for (t)-MVIIA are found between residues A6 to C8, S9 to R21, and K24 to C25. The structure also contains four ~ turns centered about residues K4G5, R10L11, C16C17, and G22K23. The overall rms deviation for the 13 best structures was 1.85 A and the backbone for only the backbone atoms was 0.68 A. The biochemistry of ro-conotoxins has been an area of active research due in part to their wide range of potentially useful medicinal properties. Some structure activity experiments have been completed which are relevant to the published NMR structures. ro­Conotoxins GVIA and MVIIA interact specifically with alB-containing N-type calcium channel subtypes.33 Whereas, ro-MVIIC binds to a broader subset of calcium channel sUbtypes. Despite their different targeting specificity, MVIIA shows greater sequence homology to MVIIC than to GVIA (see Figure 1.1). A series of ro-GVIA analogs have been synthesized in which each non cystine or hydroxyproline residue was replaced with alanine.34 The ability of each analog to interact with N-type calcium channels was then probed using several experimental methods. These experiments found that K2 and the hydroxyl group of Y13 were critical for the activity of ro-GVIA. The NMR structures of 23 ro-GVIA showed that K2 and Y 13 were both on the same side of the molecule suggesting that they are both directly involved in binding to the receptor site on the calcium channel. Similar experiments with ro-MVIIA demonstrated that Y13, K2, RIO and R21 are important in the binding of this toxin.33 The NMR structures for ro-GVIA, ro-MVI lA, and ro-MVI IC have demonstrated that all three have similar structures. In light of this, the similar binding properties of ro-GVIA and ro-MVI IA might be expected, but the altered specificity of ro-MVIIC for alA-containing calcium channel SUbtypes unexplained. This situation suggests that other residues of ro-MVIIC are not only responsible for impeding the binding of ro-MVIIC to alB-containing calcium channels but also for specifically binding to the alA -containing channels. If this explanation is correct then further experiments with the ro-conotoxin family could reveal information about Ca channels with unprecedented detail. The NMR structures completed for ro-conotoxins will certainly be invaluable tools for further experimental design in this area. 1.4 Perspectives The work on conotoxins reviewed above clearly demonstrates that NMR spectroscopy provides feasible technology for determining the structures of these peptides, given their relatively constrained conformation. At least one example of the three major structural classes of conotoxins has been analyzed to date. The technology involved in peptide and protein NMR structures is improving at a rapid pace. Greater confidence is warranted for recent conotoxin structures due to improvements in NMR sensitivity and new and more rigorous molecular modeling protocols. Several of the reports discussed in this review demonstrate the importance of back calculation and simulated annealing methods in refining minor errors in initial structures. The ro-conotoxins provide a specific example of how small refinements in an NMR model may be of critical importance in making the model relevant to biological activity. The ro-conotoxin family shows significant differences in receptor specificity 24 despite highly homologous structures. From this we can infer that subtle differences in three-dimensional structure can have dramatic effects on biological activity and specificity. This illustrates the importance of using both computational and biochemical methods to rigorously test an NMR model to uncover possible minor inconsistencies. For all three major structural classes of conotoxins, significant insights have also been gained by mutagenesis of their cloned receptor targets. Thus, mutations in the nicotinic acetylcholine receptor that are sensitive to a-conotoxins have revealed which amino acid residues in the receptors are responsible for the strong selectivity between the aid and the alg interface sites. Several laboratories have mutagenized voltage-gated sodium channels to determine requirements for f.L-conotoxin high affinity binding. The importance of specific glutamate residues in the specific binding of co-conotoxin GVIA to the alB subunit of the Ca channel complex was revealed by mutational analysis. At the same time, a large number of sequence analogs of the toxins have been chemically synthesized. Thus, structure/function studies of conotoxins and their receptor targets have the conceptual advantage that both ligand and receptor can be readily mutated, since both are encoded by nucleic acids. However, the ligands are small enough that mutagenesis through direct synthesis can also be carried out. The structural work which we have reviewed above is clearly key to further structure/function studies, since detailed structural information is a prerequisite to any mechanistic interpretation of how particular mutations in either the ligand or the receptor target might affect ligand/receptor interactions. Mutations in the ligand that knock out activity could in principle be compensated for by mutations in the receptor target. A rigorous test for the veracity of structural assignments and structure/function information is to be able to predict which mutation in the receptor target might compensate for a ligand mutation which strongly compromised interaction with the unmutated site. It is this level of predictive experimental verification to which we should aspire. For the long run, another 25 desirable goal is to directly determine by NMR spectroscopy the structure of ligand/receptor complexes, particularly for receptors or ion channel targets that are metnbrane-bound. Present biophysical techniques are insufficient to give an independent structure of membrane-bound proteins on a routine basis. The development of model systems that will allow the analysis of ligand/receptor target complexes by NMR, wherein a conotoxin is the ligand, should be an attractive avenue of future research. 26 1.5 References 1. Wtitrich, K. NMR of Proteins and Nucleic Acids, 1986 Wiley &Sons, New York. 2. Pardi,.A.; Galdes, A.; Florance, J.; Maniconte, D. Biochemistry 198928,5494- 5501. 3. Kobayashi, Y.; Ohkubo, T.; Kyogoku, Y.; Nishiuchi, Y.; Sakakibara, S. Biochemistry 28, 1989, 4853-4860. 4. Wagner, G.; Braun, W.L.; Havel, T.F.; Schaumann, T.; Go, N.; Wtitrich, K. 1. Mol. BioI. 1987,196,611-639. 5. Hyberts, S.; Marki, W.; Wagner, G. Eur. 1. Biochem 1987, 164, 625-635. 6. Gray, W.R.; Luque, A.; Olivera, B.M. 1. BioI. Chem 1981, 256,4734. 7. Chou, P.Y.; Fasman, G.D. Annu. Rev. Biochem. 1978,47,251-176. 8. Stenlake, J.B. in Burger's Medicinal Chemistry, p. 431, 1981, Wiley, New York. 9. Christensen, Dale NMR Spectroscopic Investigation of Peptide Structure and Enzyme Mechanism 1993 Ph.D. Thesis, University of Utah. 10. Sine, S.M.; Kreienkamp, N.B.; Maeda, R.; Taylor, P. Neuron 1995, 15, 205-211. 11. Ott, K.; Becker, B.; Gordon, R.D.; Rtiterjans, H. FEBS Lett. 1991,278, 160-165. 12. Lancelin, J.; Kohda, D.; Tate, S.; Yanagawa, Y.; Abe, T.,; Satake, M.; Inagaki, F. Biochemistry 1991, 30, 6908-6916. 13. Wakamatsu, K.; Kohda, D.; Hatanaka, H.; Lancelin, J.; Ishida, Y.; Oya, M.; Nakamura, H.; Inagaki, F.; Sato, K. Biochemistry 31, 1992, 12577-12584. 14. Sato, K.; Ishida, Y.; Wakamatsu, K.; Kato, R; Honda, H.; Ohizumi, Y.; Nakamura, H.; Ohya, M.; Lancelin, H.; Kohda, D.; Inagaki, F. 1. Bioi. Chem. 1991, 266, 16989-16991. 15. Nilges, M.; Gronenborn, A.M.; Brunger, A.T.; Clore, G.M. Protein Eng (PRl), 1988, 27-38. 16. Nilges, M.; Clore, G.M.; Gronenborn, A.M. FEBS Lett. 1988,229, 317-324. 17. Brunger, A.T., (1990). XPLOR software manual version 2.1. Yale University, New Haven, CT. 18. Dudley, S.; Todt, H.; Lipkind, G.; Fozzard, H. Biophys.l. 1995, 69, 1657-1665. 19. Kao, C. Y. Ann. N.Y. Acad. Sci. 1986,479, 52-67. 20. Guy, H.R.; Conti, R. Trends Neurol. Sci. 1990,13, 201-206. 21. Noda, M.S.; Suzuki, H.; Numa., S.; Stiihmer, W. FEBS Lett. 1989,259, 213- 216. 22. Terlau, H.; Heinemann, S. H.; Stiihmer, W.; Pusch, M.; Conti, F.; Imoto, K.; Numa, S. FEBS Lett. 1991, 293, 93-96. 23. Lipkind,G.; Fozzard, H. Biophys. 1. 1994, 66, 695-700. 27 24. Pallaghy, P. K.; Duggan. B. M.; Pennington, M. W.; Norton., R. S. 1. Mol. Biol. 1993, 234, 405-420. 25. Pallaghy, P. K.; Nielsen, K. J.; Craik, D. J.; Norton., R. S. Protein Science 1994,3, 1833-1839. 26. Abola, E.E.; Bernstein, F.C.; Bryant, S.H.; Koetzle, T.F.; Wong, J. Protein Data Bank in Crystallographic Databases - information content, Software Systems, Scientific Applications, 1977, eds. Allen, F.H., Bergerhoff, G., and Sievers, R., Data Commission of the Int'l Union of Crystallography, Bonn/Cambridge/Chester, 1987,pg 107-132. 27. Bernstein, F.C.; Koetzle, T.F.; Williams, G.J.; Meyer, E.F.; Brice, M.D.; Rodgers., J.R.; Kennard, 0.; Simanouchi, T.; Tasumi, M. J. Mol. Biol. 1977,112, 535-542. 28. Davis, J. H.; Bradley, E. K.; Miljanich, G. P.; Nadasdi, L.; Ramachandran, J.; Basus, V. J. Biochemistry 1993, 32, 7396-7405. 29. Skalicky, J. J.; Metzler, W.J.; Ciesla, D.J.; Galdes, A.; Pardi, A. Protein Science 1993,2., 1591-1603. 30. Sevilla, P.; Bruix, M.; Santoro, H.; Gago, F.; GarcIa, A. G; Rico, M. Biochem. Biophys Res. Com. 1993,192, 1238-1244. 31. Nemoto, N.; Kubo, S.; Yoshida, T.; Chino, N.; Kimura, T.; Sakakibara, S.; Kyogoku, Y.; Kobayashi, Y. Biochem. Biophys. Res. Com. 1995,2, 695-700. 32. Farr-Jones, S.; Miljanich, G. P.; Nadasdi, L.; Ramachandran, J.; Basus, V. J. 1. Mol. Bioi. 1995,248, 106-124. 33. Nadasdi, L.; Tamashiro, D.; Chung, D.; Tarczy-Hornoch, K.; Adriaenssens, P.; Ramachandran, J. Biochemistry 1995,34, 8076-8081. 34. Kim, J.L; Takahashi, M.; Ogura, A.; Kohno, T.; Kudo, Y.; Sato, K. 1. Bioi. Chern. 1994, 269, 23876-23878. 2. THE THREE-DIMENSIONAL STRUCTURE OF CONOTOXIN ,¥-PIIIE 2.1 Background and Rationale Conotoxin 'II-PIllE is a recently described peptide toxin from the venomous snail Conus purpurascens which has been shown to act as an antagonist of acetylcholine gated ion channels.35 'II-PIllE contains the same disulfide bonding pattern as the J.l-conotoxins (Figure 2.1), though it shares no other sequence homology with this pharmacological class of peptides. Three-dimensional structures and analysis of structure activity relationships have recently been published for the J.l-conotoxins GIIIA36.37.38 and GIl 39 describing a compact peptide structure built around a cage of disulfide bonded sulfur atoms. Conotoxin '¥-PIIIE has very little sequence homology with other peptides, apart from the disulfide bonding pattern, hence the unique primary structure and pharmacology of the peptide make it an attractive target for NMR structural studies. The acetylcholine receptor is a ligand gated ion channel that is involved in propagating an ion potential across the nerve synapse. When the postsynaptic receptor binds acetylcholine, it triggers the opening of a trans-membrane ion channel which allows Na+ and K+ ions to diffuse, respectively, in and out of the cell. Although a high resolution structure of the protein complex is not available, the receptor has been studied in detail, in RD CTOOKKC-KDRQCKOQR CA RD CTOORKC-KDRRCKOMKCCA HOO C LYGKCRRYOGCSSASCCQR Figure 2.1 The primary sequences of J.l-GIIIA, J.l-GIIIB and 'II-PIllE aligned using the conserved disulfide bonding pattern. 29 large part by characterizing binding sites of natural toxins. The a-conotoxins bind competitively with acetylcholine to the receptor, however conotoxin 'V-PIllE does not show competitive binding with either compound, and has been suggested to directly occlude the ion channel. The three-dimensional structure of conotoxin 'V-PIllE is of interest for two reasons: it will help characterize the acetylcholine receptor ion channel, and by comparing the differences in the structure of the f.l-conotoxins, allow insight into the structural differences that allow related disulfide bonded peptides to target two different receptors. 2.2 Results 2.2.1 Spectral assignment. DQF-COSY and TOCSY NMR experiments were used to assign individual spin-systems for each amino acid, which were then sequentially linked by key NOESY correlations (Figure 2.2, Figure 2.3) using the method of Wutrich et a1.40 Application of these methods to the majority of the sequence was straightforward (Table 2.1, Figure 2.4), except in the case of the N -tenninal three amino acids. The amino terminal proton of HI was not observed and amino acids 02 or 03 contain no amide protons; the residues were unambiguously assigned using the following the NOE crosspeaks (Figure 2.5). The AMX HI spin system was assigned based on NOE cross peaks from the aromatic HI 0 proton to the a and P protons, while the connectivity between HI and 02 was determined by characteristic HI a proton cross peaks to the 0 protons of 02, which is standard for a trans proline residue. The connectivity between 02 and 03 residues was demonstrated by NOE cross peaks from the a to 0 protons of 02 and 03, respectively. The a and amide protons of C4 show crosspeaks to the 03 P protons, allowing the complete sequence assignment for all amino acids of the peptide. 1'1 (ppm) 3.' 3.6 3.8 '.0 '.2 ,., '.6 '.8 5.0 5.2 5.' 5.6 1'1 (ppm) 3.' 3.6 3.8 '.0 '.2 ,., '.6 '.8 5.0 5.2 5.' 5.6 ~ ~ 8 ~ cO ~ C (1 ~O ~ ~~. 4 o~ 9.' 9.0 8.6 8.2 7.8 7.' 1'2 (ppm) Figure 2.2 The fingerprint region of the 400 ms WETNOESY spectrum with sequentially connected protons connected by solid lines I I~ ~ V i?J ~ cG "':!< e 4P 0 • 9.' 9.0 8.6 8.2 7.8 7.' 1'2 (ppm) Figure 2.3 Sequential assignment of fingerprint NOE crosspeaks for residues 16-26 for conotoxin 'II-PIllE 30 31 Table 2.1 IH Resonance assignments for conotoxin 'V-PIllE Amino HN a ~1'~2 'Y Other Protons Acid HI 02 03 C4 C5 L6 Y7 G8 K9 C10 R11 R12 Y13 014 G15 C16 S17 S18 A19 S20 C21 C22 Q23 R24 4.62 3.46, 2.64 07.06, e 8.70 4.97 2.49, 1.86 4.24 4.73 2.51, 1.92 4.69 9.05 4.72 3.06,2.85 8.08 5.39 2.95, 2.82 8.23 4.27 1.5, 1.12 0.91 9.45 4.33 3.32,3.03 07.05, e 6.74 8.65 4.17, 3.60 7.70 4.50 1.80, 1.35 1.4 o 1.65, e 3.0, <I> 7.59 8.52 4.68 2.93, 2.86 9.03 4.03 1.52, 1.14 1.66 06.17, e 6.69 7.97 4.01 1.79, 1.39 1.48 o 3.11, e 7.21 7.16 4.66 2.78, 2.14 07.02, e 6.72 4.55 2.32, 2.16 4.74 04.0,03.84 8.77 4.47, 3.80 8.12 4.95 3.51, 3.16 8.35 3.98 3.94, 3.91 7.70 4.45 3.98, 3.82 7.80 4.31 1.54 9.31 4.31 4.08, 3.94 7.40 4.71 3.37, 3.04 7.78 4.68 3.20,3.07 7.60 4.29 2.13,2.03 2.38 8.34 4.29 1.84, 1.76 1.63 HOOCCLYGKCRRYOGCSSASCCQR aNH * * * *- PNH * * *-* -- NHNH - - Figure 2.4 Summary of the observed NOE crosspeaks which allowed sequential assignment of amino acid spin systems. f\ H 0 H H6~ CD OH 2 Figure 2.5 Residues H1-C4 with arrows showing key NOE allowing sequential assignment. 32 2.2.2 Angle and stereospecific assignment restraints. Measurements of NH-a coupling constants from the proton spectrum were used to generate 10 <t> angle restraints, using the following criteria. Protons with coupling constants of less than 5 Hz (C4,C10,S17, A19, and C21) were assigned angle restraints of _90° to -40°, while protons with coupling constants of greater than 8 Hz (C5, K9, R11, S18, and C22) were restrained to angles of _900 to -180°. These angle restraint ranges can be derived directly from the Karplus equation, using coefficients detennined experimentally by Pardi et al.,41 as shown in Figure 2.6. Amino acids with NH-a coupling constants of greater than 8.0 Hz are only satisfied by <t> angles of -1600 to _800 , while amino acids with coupling constants of less than 5 Hz are satisfied by angles of _900 to _400 • The other potential Karplus equation solutions for coupling constants lower than 5.0 Hz, from 900 to 1800 , is not allowed based on steric constraints of the peptide backbone. Stereospecific ~ proton assignments for C5, R12, R13, Y13, C16, S18, and Q23 were made by analyzing ex to ~ coupling constants, as well as amide to ~ and ex to ~ NOE patterns, according to the method of Hyberts et aL42 This method relies on the observation of specific patterns of ex to ~ proton coupling constants and NOE intensities which occur if a sidechain exists in one of the three preferred rotamers about the ex-~ carbon bond (Figure 2.7). Table 2.2 shows coupling constant and NOE summaries for spin systems which 10~------------------------~ 9 8 7 6 5 4 3 2 0~~~~~~~~~·~~4i~"ijI~#«~@~ -180 -130 -80 ·30 20 70 120 170 Figure 2.6 A plot of the Karplus equation for <t> angles. The regions denoted by the shaded bar correspond to <t> angles not allowed based on steric considerations. 33 HU~ HUxR b HU:@ R I H1 H1 I H2 H2 I R NH NH NH ·60 180 60 Expected 3Jaf31 13 3.5 3.5 3Ja132 3.5 13 3.5 N0Ea!31 M S S NOEa132 S M S NOENHf31 M M W NOENHf32 W M M Figure 2.7 Newman projections of the three most stable rotomers about the Ca- C~ bond. Expected coupling constant and NOE patterns are shown for each conformation. Table 2.2 Coupling constants and NOE patterns allowing stereospecific assignments for conotoxin 'II-PIllE. Jnh-a J a-~l Ja-~2 NOEnh_~l NOEnh_~2 NOEa_~1 NOEa_~2 Xl HI 6.0 Hz 3.0 Hz 02 03 7.2 10.6 C4 2.0 5.9 C5 10.3 10.6 1.0 S M W S -60 L6 7.0 Y7 6.4 5.6 12.6 W W S M -60 G8 K9 8.0 CIO 5.0 R11 8.4 10.0 5.5 R12 6.5 4.2 8.0 M S S W -60 Y13 5.5 5.5 10.7 W S S M -60 014 9.6 9.8 G15 C16 6.7 2.0 9.0 W S S M -60 S17 1.0 S18 8.0 4.5 3.8 S S M OL A19 5.0 S20 5.2 C21 0.5 C22 9.6 M S S OL Q23 6.5 5.0 lOA W S S W -60 R24 6.7 34 were stereospecifically assigned in conotoxin 'V-PI I IE. Stereospecific assignments for ~ and B protons of the hydroxyproline rings were assigned by analyzing the NOE intensity patterns between ~ to (l and ~ to 'Y protons or 'Y to B protons. The conformational rigidity of the five membered proline ring fixes the relative orientation of these atoms to one another, and the relative distances of between atoms are distinct because of the stereochemistry at both the (l and 'Y carbons. An example of the interproton distances and the resulting NOE patterns is shown in Figure 2.8 for residue Hy Ha F1 ppm J 4.60 ~v ~ 464~ 4.681 4.72) .§ 4.761 1 4.80 ~ il 2.0 1.9 1.8 F2ppm Figure 2.8 A model of residue 02 build with ideal peptide geometry. The labeled NOE peaks reflect the intensity patterns expected based on the model, and were used for stereospecific ~ and B assignments. 35 02. For each hydroxyproline residue, the NOE cross peak between the a proton and the ProR P proton was more intense than the ProS proton, whereas the opposite pattern was observed for the 'Y to P protons. Similarly, the 0 proton with the more intense 0 to 'Y NOE crosspeak was assigned ProS stereochemistry. The NOE intensity patterns were similar for each of the three hydroxyproline residues of conotoxin 'V-PI I IE, lending credibility to this method. 2.2.3 Distance restraints and the RMA calculation. Initial molecular modeling experiments used distance constraints generated by categorizing NOE crosspeak volumes as strong, medium, or weak and assigning distance constraints of 1.5-2.5A, 2.5- 3.5 A, and 3.5-5.5 A based on these classifications. Portions of initial structures generated using the resulting constraint set were not well converged, and showed significant restraint violations. Further analysis of these structures demonstrated that restraint violations were not evenly distributed throughout the structures, but instead were often localized in regions of the molecule that were not well converged. Removing the violated restraints from the data set and reannealing the structures generally gave improved convergence around the affected residues. Data for each of the approximately 30 violated restraints (violations of greater than 1.0 A) were reexamined to assure that both assignments and crosspeak integrals were correct, resulting in four crosspeak reassignments. Nevertheless, generating distance restraints by classification of cross peak volumes as strong, medium, or weak distance approximations yielded structures in which 10% of the constraints were incompatible with the three-dimensional covalent requirements of the peptide. The conversion of NOE integrals to approximate distance bounds is a common method of restraint generation when applying NOE spectroscopy to structural studies. Potential reasons this method was unsuccessful in these studies include the quality of the NOE data, or that the resulting approximate distance ranges are not sufficient to restrain the model to one conformation in molecular modeling experiments. The first possibility was investigated by collecting NOE data using gradient H20 suppression techniques to improve 36 the signal to noise ratio of the peptide, and potentially observe new NOE crosspeaks that were entirely suppressed by the pre saturation of the H20 frequency. The resulting data contained no previously unobserved NOE correlations, and the increase in the signal to noise ratio was not large enough to justify investment of considerable resources necessary to recollect and reevaluate an entirely new data set. By using more precise approximations of the distance restraints derived from the NOE correlations, the second possibility was also evaluated. The approximations in the generation of distance constraints from NOE data are necessary because indirect magnetization transfer via spin diffusion competes with the NOE cross relaxation pathway, so the NOESY crosspeak volume does not directly reflect the distance between the two protons involved. However, the effect of the spin diffusion pathway can be predicted for given structural model, and when this factor is correctly accounted for, allows calculation of more precise distance constraints from NOESY data. Subsequent molecular modeling attempts used restraint sets generated using the RMA process, described as follows (Figure 2.9). In this method, experimental NOEs are combined with calculated NOE values based on a molecular model. Back-transformation of this mixed NOE matrix gives a relaxation matrix that provides a better estimation of the cross-relaxation rates than can be obtained directly from the NOE cross peaks. In practice, initial NOE volumes are derived directly from volume NOE values calculated by the FELIX software package. These volumes were used as input to the IRMA module of the Biosym molecular modeling software package. The correlation time (tC> was estimated by repeating the RMA calculation until the theoretical NOE buildup curve for ~ protons were in close agreement to experimental buildup curves. The measured tc for the ro-conotoxin GVIA was used as an initial starting point for this process, as the peptides are of similar size.43 Final values used were 0.6 s for tc and 1.6 s for T) leakage value. The initial IRMA calculation was performed using a linear peptide structure built with standard peptide geometry. These restraints were used in an initial distance geometry calculation, but were Linear Structure .RMA Dista,nce Geometry 500'tv1D RMA ~ +1000 0 Simulated Annealing Final Structures Figure 2.9 An overview of the molecular modeling scheme used for conotoxin 'V-PIllE. 37 modified by removing the lower distance bounds to compensate for errors due to the poor approximation of the staring model to the true structure. Ten structures were generated in this DG calculation, and this family of structures was used to refine the distance constraints using the RMA algorithm. Five iterations of 500° MD and RMA were used to determine the final refined set of distance constraints which were used in the final calculation of 50 structures. 2.2.4 Molecular modeling procedures. The molecular modeling experiments employed in determining the three-dimensional structure of conotoxin 'V­PI I IE included restrained energy minimizations, distance geometry, and molecular dynamics techniques. A brief description of how each of these processes was used, as well as an example of the results from each process, will be presented here. The evaluation of the energy of a structure is a central process to all of the molecular modeling experiments. The calculation of the potential energy for a configuration of atoms was calculated using the CVFF forcefield (Consistent-valence forcefield), which includes 38 terms to describe energy deformation of bond lengths, bond angles, torsion angles and out of plane stretching terms. The off-diagonal energy terms are also included in the forcefield, which represents couplings between deformations of internal molecular coordinates, but these terms can become unstable when the structure is far from an energy minimum, an are generally not included in calculation by the Biosym software. Finally, the forcefield includes van der Waals interactions (using a Lennard-Jones function) and a coulombic representation of electrostatic interactions. In the CVFF forcefield, hydrogen bonds arise naturally due to the van der Waals electrostatic terms. Since solvent atoms were excluded from calculations in the modeling of conotoxin 'V-PIllE, the coulombic terms were turned off, so electrostatic interactions would not dominate the energy calculation. Energy minimization algorithms used in molecular modeling of conotoxin 'V-PI I IE function by iteratively evaluating the CVFF forcefield potential for a given conformation of the molecule, and then searching for a lower energy configuration. The two minimization algorithms used, steepest decents and conjugate gradient methods, differ in the method by which they search for lower energy conformations. The steepest decents method employed by Biosym software searches conformational space by locating a trial point on the potential energy surface, and if the trial point has a lower energy, adjusts the direction of the next search to match the gradient of the current location on the energy surface. This constant reevaluation of the direction of the search towards a lower energy gradient is computationally efficient when the molecule is far from the energy minimum, but converges slowly on the minimum when the energy gradient is small. The conjugate gradient algorithm chooses the next search direction by calculating the direction which is orthogonal to all previous directions, and hence avoids retracing conformational space. This method is effective at finding the minimum energy conformation when the structure needs only minor adjustments, but can become unstable when the structure is far from the energy minimum. The steepest decents and conjugate gradient minimization algorithms are generally used in combination with one another, using steepest decents to bring the structure close to the energy minimum, and conjugate gradients to remove more subtle energy violations. 39 An inherent property of steepest decents and conjugate minimizations is that they only proceed down an energy gradient, so they are unable to overcome potential energy barriers, often resulting in structures representing local energy minima. To overcome this limitation, both distance geometry and simulated annealing molecular dynamics calculations were employed, and the resulting structures, having already thoroughly sampled conformations space, were then subjected to extensive energy minimization. The distance geometry algorithm converts distance restraints, supplied as NOE bounds calculated by the IRMA algorithm, into atomic coordinates which satisfy both the experimental data and the covalent geometry of the peptide model. As previously discussed, the NOE restraints are supplied as approximate distance ranges, so the distance geometry calculation computes and ensemble of structures which samples many of the possible conformations consistent with the data. The calculation takes place in three steps: bounds smoothing, embedding, and optimization. Bounds smoothing analyzes the structure to determine the minimum and maximum distances that are allowed by the covalent structure of the peptide. In addition to supplying additional distance restraints, this step also serves as a means of catching experimental errors, as incorrectly assigned NOE restraints may prove inconsistent with any possible conformation of the molecule, causing bounds smoothing to fail. The embedding process makes a random in the range of values of distances solved by bounds smoothing, and attempts to find atomic coordinates which best fit this molecular conformation. It is computationally efficient to allow the embedded coordinates to have fairly high potential energy values, as the resulting structures are further refined using the optimization process, which in the case of conotoxin '1'- PI I IE, involved use of a modified simulated annealing calculation. In modeling conotoxin \jI-PI I IE, distance geometry calculations were essential in solving the three-dimensional alignment of the disulfide bonds. Many studies of 40 conotoxins allow the disulfide bonds to pass through one another by replacing the covalent disulfide bond with a distance constraint. The CVFF forcefield as employed by Biosym software will not allow this procedure, as the calculation fails due to the unfilled valencies on the sulfur atoms. In an attempt to avoid this problem, a complete modeling cycle was performed with hydrogens capping the unfilled valencies of sulfur atoms. The resulting structures had very strained torsion angles throughout the disulfide bonds as they attempted to minimize steric hindrance of the additional protons. The ultimate solution to this problem involved performing the distance geometry calculation in four dimensions, which permits covalent bonds to pass through one another, allowing the structures to sample the lowest energy conformation of the disulfide atoms. Comparison of the resulting structures to the published structures of conotoxin J..1-GIIIA and J..1-GIIIB demonstrates that this procedure has found the same three-dimensional arrangement of disulfide atoms. To further explore the utility of the distance geometry calculation in solving the disulfide arrangement, a distance geometry calculation was performed with no NOE restraints applied. Although the resulting structures showed no structural convergence, all 10 structures contained the three-dimensional arrangement of disulfides shown in Figure 2.10. This suggest that the covalent geometry of conotoxin \if-PIllE, as analyzed by the bounds smoothing algorithm, can be satisfied only by one three-dimensional arrangement of the disulfide bonds. Figure 2.10 The three-dimensional disulfide folding pattern of conotoxin \If­PIllE. 41 The simulated annealing calculation is molecular dynamics technique that allows a model to cross high energy barriers and explore new folding patterns. Molecular dynamics calculations solve the classical equations of motion for the system of atoms comprising the molecule as it undergoes conformational and momentum changes and samples new geometries. As in the energy minimization process, the calculation attempts to minimize the value of a target function. In the simulated annealing calculation, the target function is separated into various different components. Weighting the various energy components differently as the simulation progresses controls the degree of conformational flexibility of the model, a process analogous to changing the effective temperature of the model. The simulated annealing calculation is performed in the following stages; preparation, sampling, folding, cooling, and minimization. The preparation phase involves loading the starting molecular coordinates, in the case of \j1-PIIIE generally resulting from distance geometry, and performing restrained energy minimization. The minimization is necessary so that the initial step of the molecular dynamics calculation can properly assign kinetic energy values to each of the individual atoms in the molecule. The sampling phase involves scaling back the weighting factors of the target function until the potential energy of the system is about equal to the kinetic energy at 1000 K. This stage of the calculation allows the molecule to cross high energy barriers and sample new folding patterns. The folding phase of the calculations involves first scaling up the weighting factor of the NOE distance constraints, then the covalent terms. When all energy terms have been scaled up to their full values, the model should satisfy all covalent and NOE restraints. The cooling phase drops the effective temperature to 300 K, followed by extensive restrained energy minimization of the resulting structures. This protocol is based upon that described by Nilges et al. using NMR data of the peptide crambin, which demonstrated that the final NMR structures were in agreement with crystal structure coordinates.59 Figure 2.11 shows a graphic representation of the scaling factors for the force constants used in simulated annealing calculations for conotoxin \j1-PI I IE . 10 ~ 0 t5 uc.t.I 0.1 Q) 'ii 0.01 0 CIJ 0.001 0.0001 0 0 o C\I Time (ps) o 1.0 o co --Covalent ,-,-,--Nonbond Figure 2.11 The scaling factors used for force constants during simulated annealing molecular dynamics calculations 2.2.5 Structure evaluation. The R-factor values calculated by the RMA 42 process give a quantitative assessment of the agreement between the experimental data and the resulting mode1.44 R factors have been defined in several different ways, but two of the most commonly reported values are the ensemble R value, the normalized standard deviation between the theoretical and experimental NOE intensities, and the R1I6 parameter, which represents a distance like quantity that is more closely related to distance restraint energies. A definition for the ensemble R factor and the ensemble R1I6 factor are given in equations 1 and 2: ~ .. W .. ( 't"m )IA:~lC ( rm) - A~~P (,rm)\ R = .£...I.},m l} lJ I} [1] ~ .. W··('t"m)A~~P(rm) '£"'I.}.m I} IJ W(tm ) is a weighting function proportional to the mixing time, and A(tm) represents the NOE crosspeak intensity. R-factors were observed to steadily decrease with refinement of 43 the model, with ensemble R factors of 0.98 for the initial linear peptide model to final values of 0.54 for the refined structure ensemble. Values for the parameter RlI6 decreased from 0.11 for the linear structure to 0.008 for the final set of 17 structures. Additional definitions for R factors, along with representative values for conotoxin 'V-PIllE, are included in Appendix C. RMSD values and the angular order parameter Sa were used to analyze the degree of convergence of the calculated structures.45 ,46 Figure 2.12 compares Sa for <p angles and the RMSD values per residue for backbone atoms as well the complete side chain. The order parameter Sa is related to the standard deviation of dihedral angles for a group of structures, and is calculated using equation 3. For a completely random angle distribution, Sa=O, while for an exactly defined angle Sa= 1. There are two main areas of conformational flexibility present in the structures, one centered about L6-Y7 as well as the C terminal two residues. Neither L6 or Y7 show long-range NOEs to other portions of the molecule, so it ·~IIIIIIIIIIIIIIIIIIIIIII 5 10 15 20 B Figure 2.12 A. An average per residue plot of the RMSD for backbone atoms (dark bars) and all atoms (shaded bars) for final 17 structures of conotoxin 'V­PIllE. B. The per residue plot of the angle order parameter Sq for <p angles. 44 is unclear whether the observed flexibility in the region arises from a paucity of NMR data or freedom of motion in solution. The decrease in the angular order parameter about residue 014 and G 15 is not reflected in the RMSD values, as the conformational flexibility of the glycine residue does not displace sidechain atoms to add to the overall RMSD for the residue. Figure 2.13 displays the a Ramachandran plot for the <I> and 'l' angles of the 17 best converged structures of conotoxin 'l'-P I I I E. All of the dihedral angles fall in allowed regions according of the plot, demonstrating that both torsion angle and steric contact issues have been resolved successfully by the molecular modeling algorithms employed. Phi (degrees) Plot statistics Figure 2.13 A <1>/'l' Ramachandran plot including angles from the 17 converged structures of 'l'-PI I IE. Squares signify nonglycine residues, triangles signify glycine residues. The numbers on the symbols indicate the structure from which the data was calculated. 45 2.2.6 Description of the structure. The structure of conotoxin 'V -P I I IE contains no a helix or ~ sheet regions, but consists of a series of ~ turns. A comparison of the positions of the ~ turns in J.L-GIIIA and 'V-PIllE reveals that each of the turns in conotoxin 'V-PIllE occurs in the analogous positions relative to the conserved cystine residues. Of particular interest, both pairs of adjacent cystine residues, C4-C5 and C21- C22, are the central residues of ~ turns; this structural feature is also observed in the two pairs of adjacent cystines in both the J.L-GIIIA and J.L-GIIIB structures. Other structural features include ~ turns between residues L6-K9, YI3-CI6, CI6-AI9, and AI9-C22. A Y tum appears between R 11 and Y 13, as well as a long distance hydrogen bond between the carbony I of C4 and the amide residue of R 11. Figure 2.14 displays a stereoview of the final 17 structures of conotoxin 'V-PIllE. The presence of hydrogen bonds often results in reduced accessibility of the amide proton to bulk solvent, which is indicated by slower exchange rates for hydrogen bonded amide protons. The relative exchange rates of amide protons are reflected in IH Tl values, the dependence of amide chemical shift on temperature (L\o/ ~), and slow amide exchange rates with D20. Data for these experiments are shown in Table 2.3. None of these N N c c Figure 2.14 Stereoviews of the backbone atoms of the 17 converges structures of conotoxin 'V-PI I IE overlaid upon the lowest energy structure after simulated annealing. 46 Table 2.3 Data reflecting exchange rates of amide protons. NH 0 IHT 1 f,.O/f,.T Slow Tum 2Qm s J2J2blK Ex. C4 9.05 1.02 7.3 C5 8.08 1.05 6.4 L6 8.23 0.98 6.1 * 03-L6 Y7 9.45 0.84 9.1 G8 8.65 0.69 6.5 K9 7.70 1.04 4.0 * L6-K9 C10 8.51 0.91 6.6 R11 9.03 1.20 6.7 * R12 7.97 0.78 5.0 Y13 7.16 1.21 3.8 * R11-Y13 G15 8.77 0.85 5.1 C16 8.12 1.07 2.4 * Y13-C16 S17 8.35 0.82 6.7 S18 7.70 1.03 3.3 A19 7.80 1.04 3.1 * C16-A19 S20 9.31 0.79 8.7 C21 7.40 1.20 1.9 C22 7.78 1.26 2.7 * A19-C22 Q23 7.60 0.46 1.8 R24 8.34 0.82 8.4 47 experiments can be considered direct evidence for hydrogen bonds, but a consistent trend between experiments, including slow D20 exchange rates, long T1, and small Ao/MC values strongly supports the involvement of the amide proton in a hydrogen bond. Analysis of these data for conotoxin 'V-PIllE predicts seven protons are involved in hydrogen bonds, each of which can be explained by the secondary structure of conotoxin 'V-PIllE. Amide protons of residues L6, K9, C16, A19, and C22 all appear to be hydrogen bonded to the carbonyl of the residue three amino acids earlier in the peptide sequence, with regular ~ tum geometries. The amide proton of Y13 is within hydrogen bonding distance of the carbonyl of R11, with residues R11, R 12, and Y 13 comprising a classical y tum. The amide proton of Rl1 appears to be involved in a hydrogen bond with the carbonyl ofC4 (Figure 2.15), an interaction that may stabilize the chain reversal centered about residues L6 and Y7. Figure 2.15 The backbone atoms of the lowest energy structure for conotoxin 'V-PIllE highlighting the proposed hydrogen bond between C4 and Rll. 48 All of the three hydroxyproline residues present in conotoxin \jf-PIIIE are in the trans conformation in the models. The trans proline bond of 014 is shown in Figure 2.16. Observed NOE patterns, with crosspeaks from the a proton of the previous amino acid to the 0 protons hydroxyproline residues and crosspeaks from the ~ protons of the hydroxyproline to the NH or a proton of the next amino acid in the sequence, support this finding. Several NMR structures of other conotoxins contain hydroxyproline residues in the cis conformation; however, as the positions of the hydroxyprolines are not conserved in the sequences of the conotoxins, conservation of the conformation of the amide bonds is not expected. Of the three disulfide bonds present in the sequence, only the sidechains of C4 and C16 appear to be well resolved in the final 17 models of conotoxin Y-PIIIE. This is consistent with the amount of spectral data available from NOESY and PE-COSY experiments for each of the cystine residues. Identical chemical shifts for ~ protons were observed for cystine residues C4, C5, and C10, complicating NOESY crosspeak assignments and coupling constant measurements. Crosspeaks across the disulfide linkage were assignable only for the C4-C 16, and sufficient information for stereospecific ~ proton assignments was available for only C16. Regions of flexibility throughout the model may be related to the lack of well-resolved spectral information needed to restrain 014 014 HA Figure 2.16 The trans proline bond between Y13 and 014. 49 key covalent disulfide linkages to specific conformations. The conformation of disulfide bonds in high resolution crystal structures has been analyzed by Richardson et aI., showing that the disulfide geometry generally conforms to either a left or right handed spiral, depending on the dihedral angle Xss (also known as X3)' In all cases analyzed Xss angle was either 900 (right-handed spiral) or -900 (left-handed spiral),47 and well-defined disulfide bonds in previous conotoxin NMR structures are consistent with this observation. To investigate the role of disulfide confoffilational flexibility on the overall structure of conotoxin 'V-PIllE, X3 dihedral angles between the two ~ carbons of disulfide bonded cysteine residues were restrained to the range of -850 to -950 , and the final simulated annealing refinement was repeated for the family of 50 DG structures. The resulting structures had considerably improved backbone rmsd values, averaging 0.65 A over residues 1-23 for the 10 best converged structures. The overall backbone geometry is essentially the same as structures generated without the disulfide dihedral angle constraints, with an rmsd value of 0.83 A for all backbone atoms between the lowest energy structures for both simulated annealing calculations. The decreased backbone flexibility is especially evident for residues L6-G8 (Figure 2.17), which appears as a well-defined ~ tum in the disulfide constrained structures. Arbitrary assignment of disulfide dihedral angles is not a reasonable procedure for NMR structure detennination; however these results demonstrate the importance of acquiring adequate spectral information to effectively restrain these key covalent linkages in the conotoxins. Conformational flexibility in the disulfide bonds appears to propagate throughout the backbone structure during molecular modeling experiments, suggesting that the overall accuracy of the resulting model may depend on correctly defining the conformation of disulfide bonds. N L6 c Figure 2.17 Ten structures superimposed on the lowest energy structure from modeling runs including Xss angle restraints. The lowest energy structure from a simulated annealing calculation which did not use Xss angle restraints is also included. 2.3 Discussion The positions of positive charges are likely to be relevant to structure activity relationships for conotoxin 'II-PIllE, as the toxin targets acetylcholine gated cationic 50 channel. As with conotoxins J.L-GIIIA and J.L-GIIIB, 'II-PIllE appears to have a flat, discoidal overall shape, with the majority of the positively charged residues congregated on one face of the molecule (Figure 2.18). It is reasonable to propose that the side of the structure containing the positively charged residues directly interacts with the toxin receptor. Side chain chemical shift values for K9, R12, and R24 are near the averaged values for residues which are fully exposed to solvent. The chemical shift of the guanidinium group of R 11 is 6.69 ppm, which is shifted upfield from the equivalent groups of R12 and R24 by 0.4 ppm. The NMR structure of conotoxin 'II-PIllE shows R 11 is well defined throughout the sidechain, and is less exposed to solvent than the other cationic groups of the peptide, which may be related to the hydrogen bond which appears between the amide proton of R11 and the carbonyl of C4. 2.18 A backbone ribbon diagram of lV-PI I IE, with charged residues shown as stick drawings. The discoidal shape and distribution of positive residues on one face of the molecule are evident in this display. 51 A comparison of the structure of conotoxin lV-PIllE and NMR structures of J.1- conotoxins J.1-GIIIA and J.1-GIIIB, which have the same cystine disulfide bonding pattern, reveals considerable similarities in the three-dimensional structures of the peptides (Figure. 2.19) Although J.1-GIIIA and J.1-GIIIB have extensive sequence homology, conotoxin lV-PIllE shares only the disulfide bonding pattern with the J.1-conotoxins. This suggests that the disulfide bonding pattern alone is able to direct the secondary structural characteristics of the peptide, creating an amino acid backbone scaffold, upon which various amino acid sidechains may be added without dramatically altering the overall structure of the peptide. The differences in the receptor specificities for the J.1-conotoxins and lV-PI I IE demonstrate that Conus geographus and Conus purpurascens have used the same structural backbone to target two different receptors, and suggest that the conotoxins should serve as excellent building blocks for rational drug design. Figure 2.19 Superimposed backbone diagrams for J.1-G I I lA, J.1-G I I IB, and 'V-PIllE. 52 Our results suggest that structural features common to conotoxins J.1-G I I lA, J.1- GIIIB, and 'V-PIllE define a structural motif unique to very small peptides, containing a cystine knot, which is stabilized only by the presence of conserved ~ turns, which create chain reversals necessary to satisfy the covalent restrictions of the cystine disulfide bonding pattern. Specifically, analysis of the secondary structure for the aligned sequences (Figure 2.20) of the peptides shows that turns occur in the same position in each peptide, relative to the cystine residues. The similarity of J.1-GIIIA, J.1-GIIIB, and 'V-PIllE, despite the lack of sequence conservation apart from the cystine knot, suggests that this is a stable structural motif which has been utilized by Conus purpurascens and Conus geographus to target different biological receptors. Hill et al. have suggested that the structures of J.1-GIIIA and J.1-GIIIB contain the core structural motifs of the cysteine stabilized a~ motif (CSa~), which includes a helical region stabilized by disulfide bonds to cystine residues of a ~ strand.48 Although the secondary structures of J.1-GIIIA, J.1-GIIIB, and 'V-PIllE are very similar, the secondary structure of 'V-PIllE is best described as a series of turns, rather than in terms of a helices or ~ sheets. 53 1111 f.---i 111111 R D ~ TOO K K m- K D R Q rnK 0 Q R ~C A RD CTOORKC-KDRRCKOMKCCA HOO C-LYGKCRRYOGCSSASCCQR I ILJ VL.JL.JLJ Figure 2.20 The primary structures of conotoxin 'V-PIllE aligned with Jl­GIIIA and Jl-GIIIB. Brackets indicate the positions of ~ tum, I-I designates extended geometry, and V designates the y tum in 'V-PI I IE.minimal number of amino acids The Jl-GIIIB structure described the C terminal 10 amino acids as a distorted helix, based upon six slowly exchanging protons in this portion of the sequence. In conotoxin 'V-PIllE, there are three slowly exchanging amide protons in the analogous portion of the sequence, each of which is involved in three successive ~ turns. Although conotoxin 'V-PIllE, Jl-GIIIA, Jl-GIIIB, and the members of the CSa~ motif all have the same disulfide bonding motif, conotoxins Jl-GIIIA and 'V-PIllE appear to satisfy these covalent restraints by containing two chain reversals, but without the a helical or ~ sheet secondary structure which is common to all members of the CSa~ motif. The purpose of the disulfide bonding pattern in the CSa~ motif is to maintain a well-defined a helix and two ~ strands in the in the same orientation with one another. The motif present in the conotoxins contains only the necessary to satisfy the covalent restraints of the disulfide bonds, and cannot accommodate the well-defined secondary structure present in the CSa~ motif. 2.4 Experimental 2.4.1 Sample. The peptide was synthesized using fmoc solid phase peptide synthesis techniques. Cysteine residues were oxidized with glutathione as previously described.49 The resulting isomers were separated using HPLC and tested for both biological activity and HPLC co-migration with native compound. The biological activity of the resulting material was found to be equivalent with the natural product. 54 2.4.2 NMR spectroscopy. All spectra were recorded on a Varian Unity 500 MHz spectrometer equipped with a triple channel waveform generator. The sample was prepared by dissolving 6 mg of the peptide in 475 JlL 90% H20IlO% D20 and adjusting the pH to 3.5 using trifluoroacetic acid, or alternatively in 100% D20, to yield a 4.3 mM solution. All experiments were performed at 4.00 C to decrease exchange rates of amide protons and lower the correlation time of the peptide. NOESY,50 DQF-COSY,51 PE­COSY, 52and TOCSy53 NMR spectra were collected in the phase sensitive mode using the hypercomplex method to achieve fl quadrature detection. 54 Solvent suppression was achieved using pre saturation using a 1.2 s gaussian shaped pulse centered on the H20 signal during preaquisition delay. Data were collected using time domain data points of 4096 (t2) by 512 (t) for all experiments. NOESY experiments were performed using mixing times of lOOms, 200 ms and 400 ms in H20ID20 solutions, and with mixing times of 200 and 400 ms in 100% D20. NOESY spectra were transformed with frequency domain time points of 4096 (f) by 2048 (h) using the FELIX software package. The data were multiplied by a shifted sine bell squared function, and the first data point in t2 was linear predicted using the subsequent four data points. Cross peak volumes were calculated by integration of all data points in a rectangular region which was manually defined for each peak. 2.4.3 Structural calculations. All structural calculations were performed using software included in the Insight II molecular modeling software package (Biosym Inc) on a Silicon Graphics Indig02 computer. An initial linear peptide structure was built using standard amino acid geometries with covalent disulfide bonds according to the experimentally determined covalent structure. Starting distance geometry structures were generated by metric-matrix embedding with random trial distances consistent with restraints from a smoothed bounds matrix as determined by the EMBED algorithm.55 These structures were optimized using the simulated annealing and minimization protocol included in the DGII module of the InsightII software. Ten structures were generated in a given calculation, and resulting structures were superimposed on the structure with the lowest overall error function value. 55 The family of 10 structures generated by DG was subjected to five iterations of RMA56 ,57 using 5000 MD for 10 ps, followed by 10 ps at 3000 and 100 and 1500 iterations of steepest and conjugate minimization, respectively. The RMA algorithm was applied to the refined structures and the process repeated until R-factors no longer improved and no restraint violations greater than 0.5 A were present in all 10 of the structures. The resulting constraint set was used to generate a family of 50 DG structures which were subjected to a simulated annealing protocol described below. 58,59 The target function used for the restrained molecular dynamics (RMD) calculation is comprised of terms displayed in Equation 4: The final 50 distance geometry structures were initially subjected to 500 iterations of quartic and conjugate minimization, with all force constants reduced to 0.001 kcaVmoVA2. The RMD calculation is initiated with this weak forcefield, and NOE force constants are scaled up to 100% of their value and covalent force constants are scaled to 15% of their values over 30 ps. The covalent and chiral terms were then increased to full value over an additional 10 ps of molecular dynamics. Finally, nonbond interactions and force constants for dihedral distance constraints were then scaled to 0.25 over 10 ps. The system was then cooled to 300 K according to a geometric progression over the final 10 ps of molecular dynamics. The resulting structures were then thoroughly minimized with all force constants at full value and the Lennard-Jones form of the nonbond forcefield using both steepest and conjugate minimization algorithms. The family of 17 structures from the simulated annealing experiments which best satisfied the NOE data was then subjected to the RMA. The relaxation matrix predicted using the IRMA algorithm was viewed as a theoretical NOE spectrum and manually compared to experimental data for structures in a procedure known as back calculation. Analysis of R-factor values computed over the ensemble of the 17 best structures also demonstrated the predicted RMA data agrees well with experimental NOESY data. 56 2.5 References 35. Olivera B.M. Personal communication. 36. Ott, K.; Becker, S.; Gordon, R.D.; Rilterjans, H. FEBS Lett. 1991, 278, 160. 37. Wakanlatsu, K.; Kohda, D; Hatanaka, H.; Lancelin, J.; Ishida, Y.; Oya, M.; Nakamura, H.; Inagaki, F.; Sato, K Biochemistry 1992,31, 12577. 38. Lancelin, J.; Kohda, D.; Tate, S.; Yanagawa, Y.; Abe T.; Satake, M.; Inagaki, F. Biochemistry 1991, 30, 6908. 39. Hill, J.M.; Alewood, P.F.; Craik, D.J. Biochemistry 1996,35, 8824. 57 40. Wiltrich, K. NMR of Proteins and Nucleic Acids, 1986, Wiley-Interscience, New York. 41. Pardi, A.; Milleter, M.; Wiltrich, K. J. Mol. BioI. 1984, 180, 741-751. 42. Hyberts, S.G.; Marki, W.; Wagner, G. Eur. 1. Biochem. 1987, 164, 625. 43. Davis, J.H.; Bradley, E.K.; Miljanich, G.P.; Nadasdi, L.; Ranlachandran, J.; Basus, V.J. Biochemistry 1993, 32, 7396. 44. Gonzalez, C.; Rullmann, J.A.C.; Bonvin, A.M.J.J.; Boelens, R.; Kaptein, R. 1. Magn. Res. 1991, 91, 659. 45. Hyberts, S.G.; Golberg, M.S.; Havel, T.S.; Wagner, G. Protein Sci. 1992, 1, 736. 46. Pallaghy, P.K.; Duggan, B.M.; Pennington, M.W.; Norton, R.S. 1. Mol. BioI. 1993, 234, 405. 47. Richardson, J. Adv. Protein Chem. 198134, 167-330. 48. Cornet, B.; Bonmatin, J.; Hetru, C.; Hoffmann, J.A.; Ptak, M.; Vovelle, F. Structure 1995, 3, 435. 49. Hopkins, C.; Grilley, M.; Miller, C.; Shon, K.; Cruz, L.J.; Dykert, J.; Rivier, J.; Yoshikami, D.; Olivera, B.M. 1. BioI. Chem. 1995,270, 22361-22367. 50. Jeener, J.; Meier, B.H.; Bachman, P.; Ernst, R.R. 1. Chem. Phys. 1979, 71, 4546-4553. 51. Rance, M.; S0rensen, O.W.; Bodenhausen, G.; Wagner, G.; Ernst,R.R.; Wiltrich, K. Biochem. Biophys. Res. Commun. 1983, 117, 479-485. 52. Milller, L. 1. Magn. Res. 1987, 72, 191. 53. Braunschweiler, L.; Ernst, R.R. 1. Magn. Res. 1983,53, 521. 54. States, D.J.; Haberkorn, R.A.; Ruben, D 1. Magn. Res. 1982, 78, 186. 55. Crippen G.M.; Havel, T.M. Distance Geometry and Molecular Conformation, 1988, Research Studies Press, Taunton, England. 56. Boelens, R.; Koning, T.M.G.; Kaptein, R. J.Mol Struct. 1988, 173, 299. 57. Boelens, R.; Koning, T.M.G.; van der Marel, G.A.; van Boom, J.H.; Kaptein, R. J. Magn. Res. 1989, 82, 290. 58. Clore, G.M.; Nilges, M.; Sukumaran, D.K.; Brunger, A.T.; Karplus, M.; Gronenbom, A.M. EMBO J. 1986, 5, 2729. 59. Nilges, M.; Clore, G.M.; Gronenbom, A.M. FEBS Lett 1988, 239, 129. 58 3. THE CHEMISTRY OF THE ASCIDIAN DIDEMNUM VOELTZKOWI 3.1 Pyrrolopyrimidine Natural Products from Marine Organisms Marine organisms have proven to be rich source of biologically active nucleosides, and one class of these compounds, the pyrrolo-[2,3-d]pyrimidines, has been isolated from a wide variety of marine sources. Tubercidin (1) and toyocamycin (2)were both initially isolated from Streptomyces tubercidicus and Streptomyces toyocaensis, respectively,6o,61 but they have also been reported from several blue green algae.62 The investigation of the ascidian Didemnum voeltzkowi collected at Apo Reef, Philippines, yielded S;-deoxytubercidin (3), S"-deoxy-3-bromotubercidin (4), and two isomers of S"-deoxy-3-iodotubercidin (5,6). Kazlauskas et al.,63 Davies et al.,64 and Phillis et al.65 described the isolation and biological activities of both S"-deoxy-3- iodotubercidin well as the 4-amino-3-bromopyrrolo-[2,3d]pyrimidine aglycone (7) from the red algae Hypnea valendiae and the sponge Echinodictym sp., respectively. Stewart et al. have reported that tubercidin or tubercidin analogs are often responsible for the cytotoxicity of cyanobacteria of the family Scytonemataceae.62 Zabriskie et al. have isolated S-(methoxycarboyl) tubercidin (8) and 2, along with the nucleobase aglycones (9,10), from the sponge Jaspis johnston; .66 The mycalisines (11,12) have been isolated from a Mycale sp. sponge, and contain both a pyrrolopyrimidine group and a modified sugar.67 R to N N I R2 7 R1=Br R2=H 60 1 R = H 2 R =CN 3 R =H ~ anomer 4 R =Br ~ anomer 5 R =1 ~ anomer 6 R =1 (X anomer 8 R1=C02Me R2 =Ribose 9 R1 =C02Me R2 =H MeO OH 12 1 0 R1=CN R2=H It is interesting to note the widespread occurrence of pyrrolo-pyrimidine nucleosides from various algal sources, as the ascidian Didemnum voeltzkowi is known to harbor a symbiotic micro-alga of the genus Prochloron. Although it would be reasonable to suggest that the tubercidin analogs isolated from Didemnum voeltzkowi are actually of algal origin, no experiments were performed to evaluated this hypothesis. 3.2 Natural Products from Didemnum voeltzkowi Lindquist and Sesin et al. have isolated didemenenones A and B (13, 14) from the Caribbean ascidian Trididemnum cf. cyanophorum and a Fijian collection of Didemnum voeltzkowi.68 The didemnenones were shown to have potent cytotoxicity in HCT 116 cell lines, while the crude extracts of Didemnum voeltzkowi from the Philippines collection showed no cytotoxicity in the same assay. In agreement with this finding, the didemnemones were not isolated from the Philippines collection, and purified samples of the tubercidin analogs isolated had little HCT 116 activity. OH 13 o ~ a\\ )l \ OH ; OH "'OH 14 3.3 Isolation and Structural Elucidation of Tubercidin Analogs 61 The frozen ascidian (400 g wet weight) was extracted with MeOH and the resulting crude extract partitioned according to a modified Kupchan fractionation protoco1.69 The CHCl3 soluble material and the aqueous MeOH were then subjected to C18 VLC (vacuum liquid chromatography) and HPLC. Compounds 4, 5, and 6 were purified from the CHCl3 soluble Kupchan fraction whereas 3 was purified from the aqueous MeOH soluble fraction. The nucleosides thymidine and uracil were also identified as components of the aqueous MeOH Kupchan fraction. The presence of pyrrolo[2,3-d]pyrimidine deoxyribose nucleosides was initially recognized by a fragment in the F ABMS at m/z 251 in all three tubercidin analogs. The molecular ion of the brominated compound showed the expected M +2 isotope pattern for a singly brominated molecule and the iodinated compound contained an M-126 ion suggesting halogen matrix exchange. The ribose sugar moiety was recognizable in 3 and 5 by characteristic anomeric proton doublets at 5.99 and 6.12 ppm respectively in the IH NMR spectra of the compounds. The absence of geminal proton coupling expected for 5" protons of nuc1eosides in the IH NMR of 3 and 5 suggested modification at this site. Doublets at 1.26 and 1.39 ppm which integrate to three protons in the IH NMR of 3 and 5 respectively were consistent with 5' deoxy sugars. The previous publication of 5 from the red algae Hypnea valendiae reported the purification of two isomers of the compound. Based on JH1 -H2 coupling constants Wells 62 proposed that the two compounds were anomers. We also have evidence for the presence of two chromatographically distinct isomers of 5 from Didemnum voeltzkowi. C 18 VLC of the CHC13 Kupchan fraction yielded a mixture of nucleosides which eluted using CH3CNI H20 (30:70). When this material was subjected to C18 HPLC using CH3CN/0.l M aqueous NH40Ac (20:80) two distinct peaks eluted at 16 min and 21 min. The fIrst peak appeared pure by NMR spectroscopy and contained one molecular ion at mlz 377 in the FABMS. The IH NMR spectrum of material isolated as the second peak showed slight doubling of all of the signals in the IH NMRspectrum and the F ABMS showed peaks at mlz 377 and two peaks of equal intensity at mlz 327 and 329. This suggested that the sample contained both a second isomer of 3-iodo-5"'-deoxytubercidin as well as a brominated analog. HPLC-MS was then employed to further investigate the structure of the brominated compound. Using a modifIed Buck gradieneo71 two major peaks were clearly resolved by HPLC. The thermospray mass spectrum of the fIrst HPLC peak contained ions at mlz 331, 329,251,215,213, and 134. The mlz 331 and 329 peaks corresponded to the molecular ion, and the 251 ion represented loss of a bromine atom with subsequent replacement by a proton. The 215 and 213 pair of ions arose from the loss of a deoxyribose sugar, demonstrating that the molecule is brominated somewhere on the nucleoside base. The mass spectrum of the second HPLC peak contained a molecular ion at mIz 377 representing 5 as well as ions corresponding to the same fragmentation pattern as the brominated compound. The IH NMR spectrum of the mixture contained singlets of different intensities at 7.78 and 7.71 ppm, assignable to C-2 of the pyrrolopyrimidine base. Further attempts to purify the compouds involved chromatography with TF A, which resulted in degradation of the compounds. We were unable to further characterize the brominated analog at this point as no additional sample was available. 5"'-deoxytubercidin was initially detected in the aqueous MeOH soluble Kupchan Br NH2 \~.--lN ;J H3ko~N N Ii't--t'li OH OH NH2 \~.--lN ;J H3C~O N N H H H H OH OH Figure 3.1 MS fragment assignments in thermos pray LC-MS. 63 fraction after it eluted from a C18 VLC column using MeOHlH20 (40/60). Further attempts to separate the compounds involved C18 HPLC with a solvent system of CH3CN/0.l M aqueous NH40Ac adjusted to pH 5 were successful, yielding pure compound 3 (10.2 mg). FABMS gave a molecular ion of mlz 251. IH coupling constants of 3.7 Hz for protons assigned to position 2 and 3 are expected values for the five membered aromatic ring. Chemical shifts for IH and 13C NMR data agree with expected values based on the halogenated analogs (Table 3.1). Based on the 5 Hz Hl'-H2' coupling constant, 5' -deoxytubercidin appears to contain the ~ anomer of the 5' -deoxyribose. Energy minimized molecular models are included in Figure 3.2, showing the difference in dihedral angles between the HI' and H2' protons for both the a. and ~ anomers. The larger dihedral HI '-H2' angle for the ~ anomer should give rise to a larger coupling 7 6 5 1 2 ppm A. B. Figure 3.2 The 5 Hz HI' -H2' coupling constant distinguishes between the anomers of the sugar, demonstrated as follows. (A) The IH NMR proton spectrum of 3, highlighting the HI '-H2" coupling constants. (B) Energy minimized models of the ex and ~ anomers of 3, looking down the C2' -C I ' bond. 64 Table 3.1. 1 Hand 13C Assignments of Compounds 1 and 3. 5 Compound 2 102.9 7.75, s 123.2 3 49.9 101.2 4 157.2 158.9 6 151.9 8.09, s 152.3 8 127.9 104.0 9 149.5 151.2 3 7.23, d (3.7) 6.64, d (3.7) 8.08, s l' 89.3 5.99, d (5.0) 89.7 6.12, d (4.6) 2' 77.9 4.37, dd (5.0,5.2) 75.8 4.42, dd (4.6,5.5) 3" 76.4 3.83, dd (5.3,5.3) 76.5 3.96, dd (5.5,5.5) 4' 81.3 3.87, dd (6.5,5.3) 80.8 4.06, dq (6.4,5.8) 5" 13.2 1.26, d (6.5) 19.2 1.39, d (6.4) a. Chemical shift values from spectra collected in CD30D b. Coupling constants reported in Hz constant between the protons, and literature precedence suggests HI '-H2" coupling constants larger than 3 Hz are generally indicative for P anomer of ribose sugars.72 3.4 Experimental 3.4.1 General procedures. IH and 13C NMR experiments were performed on a Varian Unity 500 MHz spectrometer. Spectra were referenced to residual undeuterated solvent peaks or solvent 13C solvent signals. High and low resolution 65 F ABMS mass measurements were performed on a Finnegan MAT 95 high-resolution gas chromatograph/mass spectrometer. The HPLC thermospray-MS was performed on an instrument of in-house design, the details of which have been previously published.73 The HPLC gradient used in the experiment is a slight modification of the gradient developed by Buck, using procedures that have been previously published.9 ,10 3.4.2 Purification and characterization of nucleosides. The ascidian was collected from isolated rocks scattered throughout shallow tide pools on Apo Reef, Philippines. A voucher specimen has been submitted to the University of the Philippines labeled as GC95-94-10. The frozen sample was ground using a blender and repeatedly extracted with MeOH. The dried extract was resuspended in 200 rr1L 90% MeOH 10% 66 H20. This solution was extracted with hexane, then 20 additional mL of H20 was added to the aqueous MeOH fraction. The resulting 30% aqueous MeOH fraction was then extracted with CHC13• The CHCl3 soluble material was then subjected to a C18 VLC column using 100 cm3 LiChroprep® RP18 in a 10 cm diameter column. Fractions were eluted using a step gradient using increasing amounts of CH3CN. All of the halogenated tubercidin analogs eluted in the CH3CNIH20 (70:30) wash. Reversed phase HPLC (Rainin Dynamax Microsorb 4.6 X 250 mm 5 11 column) of this fraction using CH3CN/0.1 M aqueous NH40Ac adjusted to pH 5 using acetic acid (15:85) yielded pure 5 (2.5 mg) and a mixture of the second anomer of 5 and compound 4 (2.3 mg). A step gradient C18 VLC column using MeOHlH20 solutions was performed on the aqueous MeOH Kupchan fraction. 5 "-deoxytubercidin eluted in the 60% MeOH 40% H20 wash. This material was further purified by C18 HPLC using 15% CH3CN 85% 0.1 M NH40Ac adjusted to pH 5 to yield compound 3 (10.2 mg). 3.5 References 60. Suzuki, S.; Marumo, S. 1. Antibiotics, Ser. A. 1960, 13, 360. 61. Nishimura, H.; Katagiri, K.; Satok K.; Mayama, M.; Shimaoka, N. 1. Antibiotics, Ser. A 1956, 9, 60-62. 62. Stewart, l.B.; Bornemann, V.; Chen, l.L.; Moore, R.E.; Caplan, F.R.; Karuso, l; Larsen, L.K.; Patterson, a.M.L. 1. Antibiotics 1988,41, 1048-1056. 63. Kazlauskas, R.; Murphy, P.T.; Wells, R.I.; Baird-Lambert, l.A.; Jamieson, D.O. Aust. 1. Chem. 1983, 36, 165-170. 64. Davies, L.P; Baird-Lambert, l.; Marwood, J.F. Biochem. Pharmacol. 1986,35, 3021-3029. 65. Phillis, l.W.; Smith-Barbour, M. Life Sci. 1993,53, 497-502. 66. Zabriskie, T.M.; Ireland, C.M. 1. Nat. Products 1989, 52, 1353-1357 67 67. Kato, Y.; Fusetani, N.; Matsunaga, S.; Hashimoto, K Tetrahedron Lett. 1985,26, 3484-3486. 68. Lindquist, N.; Fenical, W.; Sesin, D.F.; Ireland, C.M.; Van Duyne, G.D.; Forsyth, C.l.; Clardy, J. 1. Am. Chem. Soc. 1988,110, 1308. 69. Kupchan, S.M.; Britton, R.W.; Ziegler, M.F.; Siegel, C.W. 1. Org. Chem. 1973, 38, 178-179. 70. Pomerantz, S.C; McCloskey, l.A. Methods in Enzymol. 1990, 193, 796-824. 71. Buck, M.; Connick, M.; Ames, B.N. Anal. Biochem. 1983, 129, 1-13. 72 Wiitrich, K. in NMR of Proteins and Nucleic Acids 1986 John Wiley and Sons, Inc. 73. Edmonds, C.G.; Vestle, N.L.; McCloskey, J.A. Nucleic Acids Res. 1985, 13, 8197-8206. 4. ISOLATION OF 1,3-DIMETHYLISOGUANINE FROM THE BERMUDIAN SPONGE AMPHIMEDON VIRIDIS 4.1 Background and Rationale Marine sponges have proven to be an exceptionally rich source of modified nucleosides. The isolation of spongouridine and spongothymidine from Cryptotethia crypta 74and subsequent development of antiviral analogs demonstrate the potential medicinal importance of these compounds. More recently, several groups have reported the isolation of methylated guanine base analogs from sponges, including 7,9- dimethylguanine (herbipoline)/5 1,7,9-trimethylguanine,76 1,3,7-trimethylguanine77 , and 3,7-dimethylisoguanine.78 No physiological role for the myriad of methylated guanine analogs isolated from sponges is apparent. The chemistry of Amphimedon viridis from Bermuda was initially investigated due to high cytotoxicity levels in cell based cancer assays. Although most of this activity was traced to previously isolated compounds, 1,3-dimethylisoguanine (15) was observed