Carbon-13 Non Magnetic Resonance (NMR) of Steroids

Carbon-13 Non Magnetic Resonance (NMR) of Steroids Carbon-13 nuclear magnetic resonance (CMR) spectra show a marked sensitivity to such important features of chemical structure as carbon hybridization, the electronegativity of heteroatoms, branching and steric crowding. Potentially, CMR is an extremely rich source of structural data in organic chemistry, capable of rivaling or even surpassing proton magnetic resonance. In the research proposed here, we intend to develop this potential in the field of steroid chemistry. The factors which determine the CMR spectra of steroids are only modestly well understood. We have begun, and propose here to continue, a systematic study of families of closely-related steroids (keto- and hydroxy-androstanes and cholestanes) with the conviction that only through such a systematic study can the basic factors governing the CMR spectra of steroids be brought to light. We intend to quantify those factors as predictive rules which relate spectra to structures and to develop computerized methods for using t hose rules to extract structural information from the CMR spectra of unknown steroids. We also propose to develop several chemical methods (derivatization procedures) for augmenting the information-content of such spectra. Within the last twenty years, the phenomenon of nuclear magnetic resonance†2 (NMR) has evolved from little more than a laboratory curiosity to one of the most powerful analytical tools in chemistry. The experiment itself consists of observing, in an applied magnetic field, the resonance frequencies (in the radio-frequency range) of magnetic nuclei in a liquid chemical sample. The analysis of NMR spectra yields chemical shifts and coupling constants which reflect, respectively, the chemical environments of and the bonding or spatial relationships between atoms whose nuclei are magnetic. Because protons are magnetic, interacting particularly strongly with electromagnetic fields, and because they are present in virtually all organic compounds, proton NMR (PMR) has found broad usefulness in organic chemistry. The literature on PMR spectroscopy is huge, and frequently it is found that PMR spectra yield chemical information which would be difficult, if not impossible, to obtain by an y other method. The determination of structure and conformation, 394 the analysis of mixtures, 435 the study of rate processes6 and the elucidation of reaction mechanisms 7 have all been aided substantially by PMR techniques. Other nuclei frequently observed via NMR are F-19, P-31 and C-13. The first two are not common inorganic compounds and are thus used for more specialized studies. Carbon, by definition, occurs in all organic molecules, but only about 1isotope C-13. This, together with the fact that C-13 nuclei are almost two orders of magnitude less sensitive than protons to the NMP experiment, has hampered the widespread use of C-13 NMR (CMR) as an analytical tool for organic chemists. However, recent instrumental 8 advances such as pulsed Fourier transform techniques and 9-l 1 noise-modulated proton decoupling have made it possible to obtain natural-abundance CMR spectra of even large molecules (e.g., steroids) or low-concentration (ca. 0.05 M) samples within a reasonably short time (0.5 10 hr.). The research to date 12 indicates that C-13 chemical shifts (which constitute the primary data usually collected in the CMR experiment) cover a broad range (ca. 200 p.p.m.) relative to H-l shifts (ca. 10 – P.P.m.1, and are highly sensitive to hybridization, the electronegativity of substituents, branching, and steric crowding. Thus CMR spectroscopy is pote ntially a rich and highly useful source of structural data. As further advances in instrumental design take place, CMR spectroscopy will become an increasingly available and informative tool in organic chemistry. OBJECTIVES AND SIGNIFICANCE The research proposed herein is directed toward understanding the factors which determine C-13 chemical shifts in steroids and toward developing computer-based methods whereby a chemist can obtain structural information from spectra of unknown steroids. This class of compounds was chosen for two reasons. First, the steroid skeleton is more or less rigid, providing a relatively controlled framework within which to study the effects of steric hindrance and other geometrical factors upon C-13 shifts. Second, a great fraction of steroid and natural products chemistry involves the identification or verification of steroid structures, and thus we expect our results to be of significant practical utility. At the current level of understanding of C-13 chemical shifts, it is not now possible to predict the CMR spectrum of a given steroid with much certainty, although by referring to simple model systems, one can often rationalize the signs and general magnitudes of the spectral changes which take place when the nature and position of substituents are altered. It is now a challenging problem simply to assign the spectrum of a known steroid, that is to identify which observed peaks belong to which carbons. The first definitive study of the CMR of steroids was presented only four years ago by Roberts et al. 13 – In that report, the assigned spectra of nearly thirty assorted steroids are presented, with the assignment task accomplished ’I using specific single- frequency and off-resonance proton decoupling, hydroxyl acetylation effects on chemical shifts, deuteration, and substituent influences in analogous -6- compounds.† Since then, several other authors 14 have reported research on the CMR of steroids, but only recently has the systematic study 15,16 of families of closely related steroids begun. We have reported 16 the assigned spectra of fourteen keto-substituted androstanes and cholestanes, where the keto group occupies every possible position around the skeleton. We are currently collecting data on a similar series of hydroxy-substituted steroids, and work is also in 17 progress on the series of steroids containing one endocyclic double-bond. The initial stage or our proposed research is to complete the hydroxyl series and to obtain the CMR spectra of several bifunctional (e.g., dihydroxy and keto-hydroxy) steroids. With such a collection of systematic data available, we will be able to study the influence upon C-13 shifts of these three types of functionality, alone and in combination, which are by far the most common types occur-ing in natural steroids. From these data, we expect to be able to extract rules which will allow the accurate prediction of CNR spectra of steroids containing these groups. Zffrcher’*-20 has derived an extremely useful set of rules relating skeletal substitutions in steroids to changes in the PMR shifts of protons in angular methyl groups. The C-13 rules we seek will relate not only to angular methyl groups but to all carbons in the skeleton, and will thus provide a great deal more information than the Zircher rules. These CNK rules will also form the base for our proposed work in the computerized interpretation of CNR data (vide infra). – As the second portion of our work, we propose a study of reversible derivatization procedures which will aid in the assignment of the spectra of known steroids, and in the analysis of the spectra of unknown -7- steroids. Roberts and co-workers 13 have found that acetylation of a hydroxyl group in a steroid produces characteristic changes, due primarily to steric effects, in the shifts of carbons close to that group. They have used this effect in assigning such shifts. We propose to study the effects of other hydroxyl-group derivatives, specifically, the 2,4,6- trimethylbenzene (benzoates themselves, in our hands, have not shown any advantages over acetates), trifluoroacetate and the trimethylsilyl ether. These derivatives have different steric and electronic properties than do acetates, and should thus produce different patterns of spectral change, providing a convenient means of augmenting the information-content of ordinary CMR spectra. We also propose to observe the C-13 shift changes which take place upon cyclic ketalization of carbonyl groups, and we expect that such changes will be useful in determining the local environment of keto groups on steroids. It has been found 21,22 that the presence of a paramagnetic complexing agent (e.g., a †lanthanide shift† reagent) causes large changes in C-13 chemical shifts of alcohols and ketones. These changes can be related to the geometry of the complex, which reflects the geometry of the Alcohol or ketone itself. We intend to explore the use of such shift-reagents in assisting the interpretation of steroidal CHR spectra. Of particular interest will be the difference between JA- and p-hydroxy steroids: It is expected that the grossly different steric environments of axial vs. equatorial hydroxyl groups will have a pronounced effect upon the geometry of the complex, and thus, very different lanthanide shift patterns should result. If so, the effect should provide a convenient means for distinguishing the stereochemistry of sterols. -8- A third facet of the proposed research involves the development of computerized techniques for automatically extracting structural information from CMR spectra. This represents a logical growth of our Heuristic DENDRAL project, 23-28 an eight-year joint effort between our laboratories and the Departments of Genetics and Computer Science. The purpose of the project is to develop applications of heuristic programming (†artificial intelligence†) to problems in chemical inference, with the bulk of the effort directed toward the computerized interpretation of mas s-spectroscopic (MS) data. In the early DENDRAL research, 24,25 only saturated, acyclic, monofunctional compounds were treated, but we have recently reported the successful identification of the structures cf estrogenic steroids 25 (and mixtures thereof 27 ) via the computerized interpretation of MS data. As the complexity of compound classes has increased, we have felt a growing need for sources of structural data other than MS. CMR data show a sensitivity to stereochemistry and substituent placement which complements, rather than duplicates, MS-derived information, and thus CMR is the ideal candidate. We have demonstrated 28 the feasibility of using CMR data in automated structure analysis. Using a detailed and accurate set of predictive rules 29 for saturated, acyclic amines, we have constructed a computer program which can †reason out† the structure of such an amine, starting from its empirical formula and CMR spectrum. A similar effort is proposed for the steroids (at least, those containing endocyclic double bonds, carbonyl groups and hydroxyl substituents) in which structural information would be inferred from CMR data using accurate predictive rules. This information could then be integrated with the results obtained from derivatization or special CMR techniques, 9- and (if necessary) from MS analysis to yield possible structures. Not only would such a system have substantial utility, but it would represent an important advance in the †state of the art† in both CMR spectroscopy and chemical information-processing. A. CMR Spectra of Steroids We plan to complete the series of sterols by synthesizing [emailprotected], [emailprotected], 74-, 9+, lbc-, 14ti-, 16x- and 176- androstanols or cholestanols, whose CMR spectra (pulsed Fourier-transform spectra, obtained at 25 Mhz. using noise modulated proton decoupling) will be recorded and assigned. We have worked out likely synthetic pathways for the preparation of these using commonly accepted procedures and starting from compounds available in our laboratories. In order to test the extent of additivity relationships and of various interactions of substituents, we shall similarly synthesize and record the spectra of two or three dozen dihydroxy and keto-hydroxy androstanes and cholestanes. The candidates chosen will depend upon the results of the analysis of the monofunctional steroids. Using statistical procedures similar to those of Dalling and Grant, 30 and of Lindeman and Adams, 31 we shall attempt to correlate s/structural variables with chemical shifts, the goal being the derivation of an accurate set of substituent parameters for steroids. In assessing the effects of steric crowding and skeletal distortion, we plan to utilize a computerized, classical-mechanical model of the molecular structure, such as the Westheimer-type models recently reviewed by Schleyer. B. Derivatization We propose to analyze the changes in C-13 shifts which take place when the hydroxyl group in several of the above androstanols and cholestanols is derivatized to the 2,4,6-trimethylbenzene, trifluoroacetate and trimethylsilyl ether. We propose similarly to investigate the effects of ketalizing (with ethylene glycol) several androstenone and cholestanones. We propose to analyze the effects of lanthanide shift reagents (in varying concentrations) upon the CMR spectra of several of the hydroxy- and keto-steroids, with particular emphasis upon pairs of sterols which differ only in the orientation of the hydroxyl group. These investigations will be directed toward the development of a repertoire of non-destructive, chemical methods for increasing the ’information available from CMR spectra. C. Computerized Interpretation of CMR data There are three phases to our proposed research in this area, all of which will make use of the heuristic programming techniques developed in our DENDRAL project. First, we intend to develop a program to assist in the assignment of spectra obtained in; arts A and B, using currently available techniques (i.e., using rules for acyclic systems together with analogies from appropriate model systems). The purpose here is twofold: On one hand, such a program will hasten a time-consuming procedure (in our work, the assignment of spectra requires about as much time as the preparation of samples and the recording of spectra, combined), while on the other hand, it will provide a context within which to develop techniques applicable to the more difficult problem of structure identification. Specifically, we will need methods for express; ing CMR rules as efficient computer code, and for deciding whether a good, unambiguous fit occurs between predicted and observed data. Secondly, we intend to u tilize the rules derived in part A, together with derivatization information from B, to write what is called a †planning† program in the DENDRAL terminology. Such a program is designed to examine the spectrum of an unknown and, referring to a set of heuristics, to attempt to verify the presence or absence of specific structural features in the unknown. Whereas the predictive rules allow one to predict a spectrum from a given structure, the heuristics represent transformations of the rules which allow one to infer structural information from a given spectrum. The primary challenge in constructing the planning program will be the design of heuristics which are as informative as possible, yet which run efficiently. This program will be a useful analytic tool in itself and will be used in the third phase of our proposed research. This third phase will involve merging the planning program with the existing DENDRAL system, which analyzes MS data for steroids. Modifications will be made to the structure generation program, which can construct all possible sets of acyclic substituents from a given set of atoms and attach those substituents in all possible ways to a given cyclic skeleton. The structure generator now makes use of IISplanner information, constructing only those steroids which are consistent with it. We shall modify the algorithm to make use of the output from both the MS and the CMR planners, and shall extend the algorithm to consider questions of stereochemistry, which are currently ignored. We believe that the augmented DENDRAL s ystem will have the capacity to identify, unambiguously, the structures of a wide variety of steroids using information from just these two spectroscopic sources. The programs will be written in the LISP language, and will thus be compatible with the rest of the DENDRAL system. Computer time on the PDP-10 will be provided through the NIH-funded SUMEX facility at Stanford, and we request no support in this proposal for computer facilities. Programs developed in our proposed research will be available to the scientific community over the ARPA computer network.