Conformational analysis of environmental agents: use of X-ray crystallographic data to determine molecular reactivity.

This paper explores the use of crystallographic techniques as an aid in understanding the molecular reactivities of a number of agents that are of concern to pharmacologists and toxicologists. The selected examples demonstrate the role of structural data in the determination of absolute configuration, configurational flexibility and active-site topology for a reactive species. For example, the role of absolute stereochemistry in understanding synthetic pyrethroid structure-activity relationships is shown from analysis of their crystal structures; conformational flexibility among DDT analogues, and the importance of conformational and electronic properties in phenylalkanoic acid herbicides are shown from systematic analysis of their crystal structures; and interpretation of active-site stereochemistry is made by study of computer modeling of enzyme inhibitors in the active sites of related protein crystal structures. Thus, the observed patterns in conformational flexibility and their resultant effects on substrate pharmacological profile can be interpreted in understanding the molecular level events that influence biological reactivity.


Introduction
Many strategies have been employed to understand the molecular reactivities of small molecule substrates acting at their target sites. The premise underlying the development of structure-activity relationships is that the correlations that exist between molecular structure (or molecular shape) and observed biological responses are the result of these substrates (hormones, drugs, metabolites, agonists, antagonists, etc.) interacting with their macromolecular target sites (receptors, enzymes, binding proteins, etc.). Thus, interaction with the target molecule not only depends upon substrate chemical constitution, including such properties as charge, polarity, and hydrophobic character, but also on the geometric disposition of key moieties. Therefore, a detailed knowledge of molecular structure is fundamental to the understanding of structure-activity relationships.
An X-ray crystal structure determination provides a detailed picture of the contents of a crystal lattice at the atomic level and will uniquely define the three-dimensional conformation of a molecule from its atomic positions. Provided one has a reasonably good quality crystal, these parameters can provide an accurate description of the intramolecular geometry (bond lengths, angles, and torsion angles to within ± 0.0003-0.005 A and + 0.1-0.4°), and also describe intermolecular in-*Medical Foundation of Buffalo Inc., 73 High St., Bufalo, NY 14203. teractions such as hydrogen bonding and charge-transfer between molecules. Moreover, the presence of anomalously scattering atoms in the structure (e.g., halogens) will permit the absolute configuration of all chiral centers in the molecule to be defined. Also, data from neutron diffraction experiments permit accurate location of hydrogen atoms in the structure. Finally, high resolution X-ray diffraction studies can describe the electron density distribution within the bonding atoms from charge-density studies.
Crystallographic observation of the substrate, target, and target-substrate complex can provide the most reliable description of these molecular interactions, as well as provide information concerning conformational preferences in flexible molecules. However, in general, little is known of the target site, and the stereochemical requirements of the substrate-target interactions are often poorly understood. For this reason, target-site topology has been deduced from comparisons of structural data (solid state, solution, or isolated molecule) for a series of compounds that bind to the target site and elicit varying degrees of activity.
However, to understand how substrate molecular conformation can influence observed reactivity, it is necessary to understand the factors that control conformational properties. While there are a number of terms contributing to the total lattice energy in organic crystals, only a few of these intermolecular forces are large enough to influence molecular conformation. The prin-cipal terms considered in calculating the energetics of conformational change include van der Waals forces, hydrogen bonding, charge-transfer, and electrostatic energies.
While the crystal structure of a substrate will provide information concerning its preferred conformation in a particular environment, it provides no information concerning the energy differences between conformations and the energy barriers that separate them. Therefore, by comparing the crystallographically observed conformation with that (1) in solution, (2) in the free state, (3) in different polymorphs, and (4) of different conformers in the same crystal, the relative influence of these intermolecular forces on molecular conformation can be investigated. Also, by analyzing the crystal structures of a number of closely related molecules, one can look for patterns in the consistency of their structural features, and credence can be given to trends which may not be statistically significant by themselves.
Thus, X-ray crystallographic techniques permit accurate identification of the chemical constituents, definition of the absolute stereochemistry, and determination of the range of conformational flexibility of substrate molecules. Also, patterns in conformational flexibility and their resultant effects on substrate pharmacological profile can be of importance in understanding the molecular level events that influence biological reactivity. This paper will explore the use of crystallographic techniques as an aid in understanding the molecular reactivity of a number of systems that are of concern to pharmacologists and toxicologists. In choosing these examples, the aim is not to provide an exhaustive review of the literature, but to illustrate, with selected examples, the role of crystallographic data in the determination of configuration, conformation, or activesite topology of a reactive species.

Structure-Activity Relationships Environmental Agents
Given in sufficient doses, most chemicals are toxic poisons. Even the use of chemicals for their intended purpose can lead to problems of contamination and disposal. Also, a growing awareness and concern over the use of chemical agents that are either deliberately (e.g., pesticides) or inadvertantly (e.g., pollutants) added to the environment have led to renewed efforts to understand their degradation and translocation in the environment, as well as their metabolism and effect on other species.
The release into the atmosphere of polycyclic aromatic hydrocarbons that result from incomplete combustion of fossil fuels and vegetation has caused concern since many of these environmental pollutants are carcinogenic. Also, the extensive use of halogenated aromatic compounds in many commercial products has made their disposal an environmental hazard because of their high chemical stability and lipophilicity which lead to  long-term bioaccumulation (1,2). Pesticides are those agents that are purposely, and often routinely, added to the environment for the purpose of killing some form of life; plant (herbicides) or animal (insecticides). Pyrethroid Insecticides. Synthetic pyrethroid insecticides are photostable analogs of the naturally occurring pyrethrin esters ( Fig. 1) and have low mammilian toxicity, short persistence in the environment, and elicit rapid insect paralysis (3). The most potent pyrethrine analogs can have eight possible isomers formed by combinations of configurations about the three asymmetric carbons in the structure (Fig. lb). Structure-function data show that the most active pyrethroid compounds are cyclopropane carboxylic acid esters with two methyl groups on C-2 and an unsaturated side chain on C-3 which is oriented cis to the ester group ( Fig. 1) in synthetic pyrethroids and trans in the natural esters (3). Substitution of these groups results in a great loss of activity. These characteristics suggest the importance of molecular shape in activity. X-ray crystallographic analyses were carried out to determine reliable structural information concerning absolute stereochemistry, molecular dimensions, and conformation (4)(5)(6)(7).
The absolute configuration of the most potent insecticide (Fig. 2a) in this series is (a.S) cis (1R,3R), as determined from its crystal structure (5). The structural data for three other analogs (5)(6)(7) show that although the molecule is flexible (Fig. 2), there are regions with conformational similarities. All the structures were resolved into their optical antiopods except that of Figure  2c, which crystallized as a racemic mixture and is shown as the ((xS) tranis (1S,3R) isomer.
C-1. The conformation of the diphenyl ether moiety in these structures is either twist or antiskewed (8,9), as defined by the torsion angles about the ether oxygen. The net effect of these different diphenyl ether conformations is to change the overall shape of the molecule. In the structure without the cyano substituent, the diphenyl ether conformation is twist and the outer phenyl ring folds back toward the gem dimethyl of the cyclopropyl ring (Fig. 2b,d). In the active isomer (Fig. 2a), the diphenyl ether conformation is twist and the molecule is more extended. The racemic structure ( Figure  2c) has the opposite chiral stereochemistry at C-1 from that of Figure 2a, and further has an antiskewed diphenyl ether conformation. This places the vinyl chlorine atoms toward the center of the inner ring. Preferred conformations computed by theoretical calculations (10) were carried out for these pyrethroids and show that the minimum energy conformation corresponds to that observed in the crystal structure. These calculations also indicate that there is restricted rotation of the vinyl group, resulting in a preferred conformation that is also observed in solution (10) and in the crystal structure (Fig. 2d). Although these calculations show no preferred conformation for rotation about the alpha carbon, the extended form is observed both in solution and in the crystal structure.
Therefore, these structural studies have delineated the absolute configuration about the critical chiral centers and have shed light on the conformational influence of active group substitution to the overall molecular shape. These data are also consistent with solution studies (10) showing that the proximity of the cyano and ester groups is important for activity.
Organochlorine Insecticides. The organochlorine insecticides include the chlorinated ethane derivatives of which DDT is the best known example. Other classes include the cyclodienes such as chlordane, aldrin, and endrin, which are chloromethano-bridged cyclic carbon compounds; the hexachlorocyclohexanes such as lindane; and the fused chlorocyclopentanes such as mirex and kepone (Fig. 3). While all of these classes act as neuropoisons, their mechanism or site of action is not yet clearly understood (1).
Toxicological studies indicate that DDT alters the transport of Na+ and K+ ions across the membranes of nerve axons by producing a repetitive discharge of the insect nervous system. Its mechanism of action is thought to involve a stereospecific fit into nerve membrane-pore channels and the formation of charge-transfer complexes with constituents in the nerve fibers (11)(12)(13). In this general receptor model, a constraint is placed upon the conformation of the two phenyl rings such that their symmetrical dipoles are sterically fixed relative to the trihalomethyl bridging group. This fixes the two phenyl rings in a "butterfly" conformation, e.g., with both rings perpendicular to the plane of the bridging atoms.
These structure-activity data show that several features of the DDT molecule are important to its insecticidal properties. For example, substitutions larger than chlorine on the bridge ethane C-1 atom produce inactive analogues; chlorination ofthe bridge head carbon results in a twisted diphenyl ring conformation and loss of activity; and an ethylene bridging unit causes loss in ac-tivity, presumably because the bridge angle opens to 1200 and changes the orientation of the p,p'-chloro groups (12). Therefore, the important conformational features for active DDT analogs include (1) the rotation of the phenyl rings about the bridge atom, (2) the nature of the bridge substituent, and (3) the nature and position of the phenyl ring substituents.
To investigate the molecular mechanism of DDT action, the crystal structures of several active DDT analogs were undertaken (13)(14)(15)(16)(17)(18)(19)(20)(21). As illustrated (Figs. 4 and 5), the overall conformation of the diphenyl rings is variable but similar to the proposed model. Since these molecules crystallize in centrosymmetric space groups, there are two mirror-image conformations possible. The absolute configuration has been determined for only one DDT analog; (-) o,p'-DDT (15), and all other structures have been made consistent with this enantiomorph for the purpose of comparison.
The proposed active DDT conformation has two phenyl rings open and facing each other (e.g., butterfly) with the bridging torsion parameters near 900 for both angles. These structural studies show ( Fig. 5) that, while there is a broad range of values observed for these parameters, they all fall within the same quadrant of conformational space. The most toxic analogs are clustered close together (e.g., Nos. 1-4), while those that are moderately active (e.g., Nos. 7,13-15) are more broadly grouped. This group also includes the active halogen-free analogs (Nos. 13,14). The inactive analogs (Nos. 8,9,11,12) are located at the extremes of this conformational space. As illustrated (Fig. 6), there is a large conformational difference between active 2,2-dichloromethyl analog (No. 10) and its inactive dichloroethylene analog (Nos. 8,9). The bridging angle in the ethylene structure is 1180 and 1090 in the methyl compound, and there is a difference in the volume occupied by these bridging groups that appears to be important for activity. A similar difference is noted for the inactive acid metabolite, p,p'-DDA (Nos. 11,12) which also occupies a smaller volume and has electronic properties similar to those of the ethylene compound. The importance of these properties is further reflected in the moderate activity of the propyl bridge analog (No. 13), which is halogen-free. These studies have defined the molecular size and shape necessary to elicit insecticidal activity and have shown that, as the aromatic character of the bridging
unit increases (e.g., Figs. 4c,d), its toxicity decreases. Furthermore, these data show that an isostructural unit of size similar to that of the trichlormethyl group can function to sterically fix the conformation ofthe diphenyl rings and that chloro substitution of the bridge is not necessary.
The chlorinated cyclodienes (Fig. 3) are wide-spectrum insecticides produced by a Diels-Alder reaction between the cyclic diene, hexachlorocyclopentadiene, and an olefin. The insecticidal activity of the resulting adducts depends on the nature and complexity of the dienophile and its substituents. Structure-activity data illustrate that the geometrical patterns of these compounds have an important bearing on their insecticidal properties (22). These data further suggest that the structural features required for optimal insecticidal activity include the presence of two electronegative centers positioned within a narrow range of distance and direction from each other and with a specific orientation with respect to the dichloromethano bridge of the cyclodiene compound. It has been shown that the cyclodiene epoxides are metabolically active; thus those compounds that do not form epoxides have little or no toxicity (22).
To understand the stereochemical nature of these insecticides, crystal structure analyses of the most active cyclodienes have been investigated (22)(23)(24)(25)(26)(27)(28). The most toxic compounds are a family of hexachlorodimethano- naphthalenes, e.g., aldrin (Fig. 3). While there are four different geometric configurations possible for these compounds, structural data show that only two isomers are observed: endo-exo aldrin (23) and endo-endo isodrin (24) and endrin (23) (Fig. 6).* The importance of molecular shape in relation to insecticidal activity is illustrated by the comparison of endo-exo-(O)exo dieldrin (25) with its endo-endo-(O)exo isomer, endrin (24) (Fig. 6), which has lower toxicity (22). This indicates that the spatial orientation of the electronegative oxygen in dieldrin is critical for toxicity. The overall shape of dieldrin is cylindrical with the chlorines describing the maximum circumference of the cylinder. Dieldrin is more elongated than its endo-endo isomer endrin (Fig. 7). The distance between the two electronegative centers (midpoint of the double bond to the midpoint of the C-C bond of the oxo ring) is also shown for these molecules (Fig. 6). These data suggest that the active site recognizes only a narrow range of values for these compounds.
Another family of active dienes, the hexachloromethanoindenes, e.g., heptachlor (Fig. 3), show increased toxicity with addition of a chlorine in the 1-position. The structure of the heptachlor (27) and its epoxide (28) have been carried out and show the epoxide is exo, an important configuration for the high toxicity of this class of compounds (22). Therefore, these structural studies have confirmed the stereochemical assignment of the active cyclodienes and show that the relative orientation of the electronegative groups is critical for their insecticidal activity.
Herbicides. The halogenated phenoxyalkanoic acids selectively kill broad-leafed plants by stimulating excessive and disorganized growth, thus acting as auxins or plant hormone herbicides (1). These compounds also *A feature of the Diels-Alder reaction involves the stereochemical orientation of the addends. These additions favor the orientation which has the diene double bonds close to the unsaturated substitutents of the dienophile and is designated endo (inside). Thus, the substituent is directed to the convex side of the resulting cyclohexene ring. Conversely, exo (outside) has the substituent directed to the concave side of the ring. demonstrate molecular engineering techniques by taking advantage of species metabolic differences to create species selective herbicides. The most potent herbicides in this class are 2,4-dichlorophenoxyacetic acid (2,4-D), 2,3,5-trichlorophenoxyacetic acid (2,4,5-T), and 4-chloro-2-methylphenoxyacetic acid (MCPA) (Fig. 8). The corresponding 2-methyl-2-phenoxypropionic acids are inactive as herbicides, but do possess antilipolytic and antihypocholesterolemic activity (1).
Structure-activity data indicate stringent stereospecific requirements for recognition by a receptor in order to elicit herbicidal activity (29,30). Phenoxyalkanoic acids with only phenyl ring substituents at either the 2,4, or 5 position (or in all of them) are active, whereas symmetrically substituted phenyl ring compounds are not. Also, those chlorophenyl analogs with a methyl at the acetic acid C-1 position are active, but not those with a dimethyl substitution. The influence of phenyl ring substitution and side chain substitution is reflected by the activity and inactivity of the pairs of compounds p-chlorophenylacetic acid (No. 5, Fig. 8) and phenylacetic acid (No. 1) and 2,4-D (No. 6) and its 2,2-dimethyl side chain analog (No. 7). These data also show that an ether link is essential for auxin activity and that when there is an asymmetric carbon in the side chain, the D-isomers are more potent than the L isomers (29,30). Based on these structure-activity relationships, a three-point contact receptor model was proposed which required the proper orientation of the acid side chain carboxyl group, the oahydrogen and phenyl ring with appropriate electronegative characteristics (29,30).
Crystallographic analyses of phenoxyalkanoic acids were carried out to determine the conformational properties of this class of compounds and to elucidate the specific stereochemical features that could explain their herbicidal activity (31,32). These data show striking conformational similarities among these structures and indicate that there is a distortion of the exo-C-1 acetic acid bond angles and also of the central atom of the acid function (31). In addition, the ether bond angle is near 1180, and the bridging bond distances reflect the influence of their aromatic and alkyl substitutions. As illustrated (Fig. 9), the conformation of the active analogs can be planar (2,4,5-T) or, as in 2,4,5-TP, the carboxy function is perpendicular to the plane of the rest of the molecule. The inactive analog 2,4,6-T has the sidechain perpendicular to the phenyl ring. However, in the inactive 2-methyl analog, the acid function is nonplanar, as observed for 2,4,5-TP. Conformational energy calculations have not been carried out, but free rotation about the ether-acetate bond is expected. Calculations of the charge-density distribution within these compounds could show differences that might indicate why symmetrically substituted phenyl ring compounds are inactive.
Polyhalogenated Aromatic Hydrocarbons. The halogenated aromatic hydrocarbons most prevalent in the environment are the polychlorinated (PCB) or brominated (PBB) biphenyls (Fig. 10) whose lipophilic properties lead to their accumulation in many tissues. However, one of the most toxic is 2,3,7,8-tetrachlorodibenzo-p-dioxin (dioxin), a trace by-product in the synthesis of 2,4,5-T (1,2). There are 75 possible chlorine substituted dioxin isomers and their toxicity depends upon both the number and position of the halogens. Since a number of these halogenated aromatic hydrocarbons are formed as trace contaminants in the synthesis of several commercial products, structure-activity relationships have been derived in order to understand thier mechanisms of bioactivity (33)(34)(35)(36)(37)(38)(39)(40).
A number of PCBs and dioxin analogs have been used as model compounds to investigate the specific structural requirements for cytosol receptor binding and for induction of aryl hydrocarbon hydroxylase activity (33)(34)(35)(36)(37). X-ray crystallographic analyses also have been carried out to elucidate their conformational properties (38)(39)(40)(41)(42). As illustrated (Fig. 10), many of these compounds have common structural features.
Toxicological studies have further demonstrated that the most toxic PCBs and PBBs are those that have halogens in both the meta and para positions of both rings, but no ortho substituents (40). Crystallographic analysis of these biphenyls show that the most stable conformation is nonplanar, with the dihedral angle between the phenyl rings of 470 (39) (Fig. 11). Analysis of these compounds using molecular orbital techniques (43) shows that the minimum energy conformation of structures without ortho substituents is 420, while it is 900 for ortho-substituted compounds, in agreement with the crystallographic results. However, activity data for the dioxin or naphthyl compounds indicate that a planar conformation is required for toxicity. Therefore, interpretation of these data suggests that the minimum requirements for toxicity include four halogens positioned laterally on the biphenyl rings, and a planar conformation.
However, more recent quantitative structure-activity data (37) indicate that molecular shape and polarizability are important factors in determining molecular reactivity and that hydrophobic, hydrogen bonding and electronic forces play a critical role in determining these properties. These data further suggest that halogen substitution in both phenyl rings is not a requirement for binding, nor is a planar structure. The molecular shape and thickness of these compounds will be affected by the conformation about the biphenyl bond and by the buttressing effects of adjacent halogens on the phenyl rings.
The results of these studies suggest that a model for  toxic structures requires four lateral chlorines arranged in a box of approximate area 3 x 1oA, or if brominated, three lateral bromines arranged in a triangle (40) (Fig.  10). This area defines the surface of the smallest envelope that a planar molecule may occupy and elicit its toxic response at its target site. Polycyclic Aromatic Hydrocarbons. Many of the polycyclic aromatic hydrocarbons (PAH) that are carcinogenic are derived from a benzo[a]anthracene skeleton (Fig. 12). Structure-activity data show that the carcinogenic properties of this large class of PAHs are highly stereospecific and depend upon their chemical structure (1,44,45). It was also noted that most carcinogenic PAHs contain an angular three-ringed phenanthracene unit and that the greatest chemical reactivity is associated with the 5-6 bond, called the K region (Fig.  12). Thus, the earliest structure-activity profiles related the electronic structure of these compounds to their carcinogenicity (45). Recent data now suggest that their carcinogenic action is related to their metabolism involving arene oxides, dihydrodiols and diol epoxide formation (Fig. 12b).
Although a large number of polycyclic aromatic hydrocarbon metabolites have been identified, especially those of benzo[a]pyrene (BP), the ultimate carcinogen is believed to be an epoxide dihydrodiol formed by the action of mixed function oxygenase enzymes. These enzymes can produce two diastereoisomeric diol epoxides, designed by the disposition of the epoxide function with respect to the hydroxy group. Thus, the syn isomer has the epoxide and hydroxy on the same side of the ring, while the anti has them on opposite sides. These data show that the mutagenic and carcinogenic properties of these isomers differ greatly. The anti isomer has been identified as the ultimate carcinogen (45), while the syn isomer is more mutagenic. Additionally, K region oxides and diol epoxides have been shown to bind covalently to nucleic acids (44)(45)(46)(47). Crystallographic studies of many of the more potent carcinogenic PAHs have been carried out and show important conformational differences when compared with their non carcinogenic parent hydrocarbons (46,4 7). Many ofthe PAHs are planar, whereas their methylated derivatives are nonplanar due to the steric interactions of the methyl hydrogens with those of the polycyclic ring hydrogens, particularly in the crowded Bay region of the molecule (Fig. 13). This is the case for the most potent carcinogenic PAHs, 7,12-dimethylanthracene (DMBA) and 5-methylchrysene.
The stabilities of these epoxy diol BP analogs have been computed using quantum mechanical methods (53) which show that the cis axial-axial diol epoxide is the more reactive species; however, no intramolecular hydrogen bond was indicated. More recent data using molecular shape analysis of polycyclic aromatic hydrocarbons indicate that the quantitative structure- activity relationships derived can be used to estimate the mutagenic potency without explicit consideration of metabolic activation pathways (54). Thus, these studies have led to definitive information concerning specific structural parameters of importance for the mutagenic and carcinogenic characteristics of these compounds. While these data have helped shape the concepts of carcinogenicity, there still remains much to be learned about how these stereospecific interactions signal their molecular reactivity.

Pharmacological Agents
In contrast to the toxic properties associated with environmental agents, pharmacological or therapeutic agents are biologically active compounds which, when applied appropriately, modify physiological systems for the benefit of the recipient. Since many pharmacologically important drugs act by inhibiting enzymes, kinetic studies of their inhibitory effects provide valuable information on their potency and mechanisms of action. Therefore, studies of drug (or hormone) action are concerned with understanding the relationships observed between drug structure and biological activity, and developing a receptor theory to account for these observations.
The receptor can be characterized by describing the properties of compounds that react with it, thus producing a "negative replica" of the receptor from the composite of the structure-activity information assembled. Compounds that combine with the receptor and produce a change in cellular properties are agonists, while those that combine with it, but do not elicit the given responses are antagonists. Understanding these (C) H- (8)  Although there are several examples of pharmacologically important agents that have been studied crystallographically, only dihydrofolate reductase inhibitors and thyroid hormone analogs will be examined in detail. Steroid hormone structures are described in another paper in this issue.
Dihydrofolate Reductase Inhibitors. The enzyme dihydrofolate reductase (DHFR), a necessary component for all cell growth, is strongly and specifically inhibited by certain substrate analogs with binding affinities so great that they are not readily displaced by the natural folic acid substrates (Fig. 14). It has been shown that the general requirement for tight binding of inhibitors to DHFR from any species is a 2,4-diaminopyrimidine ring structure (55). Also, the binding selectivity and specificity of inhibitors for DHFR from different species can vary significantly with only a single substituent change. Comparison of several lipophilic 5-substituted analogs show that those with a 5-adamantyl substituent are the most effective inhibitors of mammalian DHFR and possess significant antineoplastic activity (56). For example, there is a 500-fold increase in inhibitory potency within this series of 5-adamantyl-6-substituted antifolates with 6-H < propyl < methyl < ethyl. In order to investigate the conformational aspects of binding specificity to DHFR, crystallographic analyses were carried out on this series and their results compared with the structural data from other antifolate classes (57)(58)(59)(60)(61)(62)(63). These data reveal that the most striking feature of these adamantyl antifolates is their "bowed" shape ( Fig. 15), caused by the steric crowding of the 5-adamantyl and the 4-and 6-substituents. This interference forces the pyrimidine ring to be significantly distorted from planarity. These data further demonstrate the effects of Ni protonation of the pyrimidine ring geometry. Comparison of the endocyclic bonds at C2 (Fig. 16) indicates that protonation causes the N1-C2 bond to lengthen and the C2-N3 bond to decrease (58). To investigate the significance of these structural changes to the binding energies and basicity of these compounds, CNDO/2 molecular orbital calculations were carried out (64) and were able to reproduce the pyrimidine ring torsional distortions observed in their crystal structures, as well as the geometry changes upon protonation at Ni. The rotational energy barrier was also computed and showed that the minimum energy conformation was the same as observed in the crystal structure. However, these calculations were not able to accurately reproduce the exocylcic amino geometry and this discrepancy can affect the charge density distributions within the pyrimidine ring.
Pyrimidine antifolates with a 5-benzyl substituent, such as trimethoprim (TMP) (Fig. 14), have the greatest selectivity in their binding for bacterial DHFR. Structural studies of TMP (59,60) show that it can adopt several equi-energy conformations. Structure-activity data show that those TMP compounds with a functionalized 4'-substituent are more potent than TMP itself. Analysis of the 4'-isopropenyl and 4'-carbomethoxy derivatives (59) shows that these substituents are perpendicular to the benzyl plane and, unlike TMP, place density both above and below the ring plane.
Structural studies have also been carried out on the natural substrate folic acid (63) and show that the overall molecular shape is similar to that of the inhibitor methotrexate as observed in the protein structures of DHFR from bacterial sources (65). Quantum mechanical and dynamic energy calculations have also been carried out on the folate molecules (66,67), and these results show that there is considerable flexibility in the orientation of the side chains and that the low energy conformation corresponds to that found in the crystal structures.
As a result of these studies the geometries of these compounds have been accurately defined, as have their hydrogen bonding patterns and conformational and electronic properties. These observations may explain their antifolate characteristics and provide a means for the rational design of more selective antifolates.
Thyroid Hormone Structures. Thyroid hormones (Fig. 17) are secreted from the thyroid gland into the general circulation where they are bound to one of several serum proteins. These hormones are also the only halogenated molecules that are biologically active, and have a number of specific proteins and enzymes that transport and metabolize them. Therefore, it is of interest to speculate on the role that these thyroid proteins and enzymes may have in the concentration and transport of other halogenated species. As observed, many of the most toxic environmental agents are halogen-containing compounds. Therefore, the possibility exists that these compounds could be recognized by these thyroid hormone specific proteins.
Structural studies carried out on the thyroid hormones and their analogs (68) have shown that these molecules are flexible and that each of the hormonespecific binding proteins requires a different conformation for maximal binding affinity. The specific stereochemical properties of the thyroid hormones arise from the steric influence of the diortlho inner ring iodines upon the diphenyl ether conformation. Minimal steric interaction between the 3,5-iodines and the outer ring 2',6'hydrogens is maintained when the conformation is skewed (e.g., the two rings are mutually perpendicular and bisecting). Because of the restricted rotation about the ether bridge, the crystal structures of all diorthosubstituted thyroactive analogs have a skewed conformation (68), while that of 3',5',3-triiodothyronine, (rT:) (69) is antiskewed (Fig. 18).
Conformational analysis of the crystallographic data permits the delineation of characteristic patterns within these structures that can be correlated with their hormone action. These data show that there exist preferred orientations of hydrogen bonds and iodine intermolecular contacts that can indicate spatial preferences for critical functional groups within the binding environment of a hormone responsive protein or enzyme (9,70).
Of particular interest in the mechanism by which the thyroid hormones are deiodinated (Fig. 17). The most active form of the hormone, 3,5,3'-triiodothyronine (T;), is derived from the deiodination of thyroxine (T,) in peripheral tissues. The enzyme iodothyronine deiodinase produces either metabolically active (T:) or inactive (rT) products (Fig. 17). Structure activity studies of the inhibition of this enzyme (70-72) by a variety of compounds including radiographic contrast agents, dyes, and flavones indicate that the enzymes recognition process is under stereospecific control involving stringent requirements for maximal inhibition. Many of these inhibitors are of concern since their ingestion can interrupt normal thyroid function and lead to physiological problems. Such is the case with the radiographic contrast agents which have slow rates of metabolic clearance and have been shown to cause hypothyroidism by inhibition of the deiodination pathway (73). Flavones are ubiquitous natural products (Fig. 19) derived from plant pigments. Aurones are the most potent inhibitors of iodothyronine deiodinase (74). In an effort to identify structural features common to natural substrates and inhibitors of the deiodinase enzyme, three conformational models of the most potent aurone, aureusidin, were compared with the observed crystal structures of the thyroid hormones (Fig. 20) (75). The skewed conformation is not a permitted aureusidin conformation since it requires rotation about the bridging double bond. This, on the other hand, is the conformation observed for all hormonally active thyroid compounds (68). A planar structure is also an energetically favorable aurone conformation, but not one that can be adopted by the thyroid compounds. Finally, the anti skewed conformation has been observed in the crystal structure of rT3, a natural inhibitor of deiodinase, and is a likely aurone conformation. Thus, these studies suggest that the deiodinase enzyme can bind a substrate or inhibitor in either the skewed or antiskewed conformation.

Receptor Models
Since crystal structures of the protein-substrate complex have been determined for dihydrofolate reductase and prealbumin, many previously proposed hypotheses concerning the structural basis for binding and activity of the antifolate and thyroid hormones can be tested. The results of the crystal structure determination of DHFR from four species, complexed with the cofactor NADPH and with inhibitors (65,76), provide specific information concerning the backbone conformation, spatial arrangement of the side chains, and a general description of the inhibitor binding site.
Computer graphic modeling of the fit of various inhibitors in the active site of DHFRs has been undertaken. Lipophilic antifolate binding to chicken liver DHFR was modeled by positioning the diaminopyrimidine ring in the same orientation as observed structurally for the diaminopteridine ring of methotrexate (57). This orientation was also verified from crystallographic studies of the binary complex of chicken liver DHFR and DAMP (Matthews, personal communication). These observations showed that there is no steric crowding of the adamantyl ring in the binding cavity. However, when the structure of the 5-(1-naphthyl)diaminopyrimidine, a noncompetitive inhibitor (57), was modeled in the same manner, the outer portion of the naphthyl ring penetrates through the surface of the binding pocket and comes into close contact (< 1.3 A) with a phenylalanine residue (Fig. 21a). Even when the naphthyl ring is rotated, there is no satisfactory orientation for the ring. On the other hand, when the active 2-naphthyl analog is similarly modeled, there are no unfavorable interactions (Fig. 21b). More subtle features of the inhibitor-substrate interactions are observed when the binding of the 6-ethyl adamantyl structure is compared with that of the less active 6methyl analog. tion of the antifolate structure as observed from its crystal structure (58), while Figure 22b shows that a 1100 rotation of the ethyl sidechain places the outer carbon in an alternate binding pocket that appears to be more suitable. This observation would suggest that the ethyl analog is more active because of the added hydrophobic interactions with the enzyme and that there is not sufficient room to fit the propyl sidechain in this pocket, hence its lesser activity. Similar observations have been made for the binding interactions of the thyroid hormones with the serum protein prealbumin (77). Prealbumin is composed of four identical subunits that form a channel running through the length of its structure, forming the thyroxine binding site. These structural data also provide a detailed environment for the iodine atoms of the hormone as well as a quantitative description of the iodine interactions in the active site (Fig. 23a). Iodothyronine analogs that are potent inhibitors of iodothyronine deiodinase are also strong binders to thyroxine-binding prealbumin. Because of this similarity, the thyroxine binding site in the crystal structure of prealbumin was used as a model for the active site of iodothyronine deiodinase. As illustrated (Figs. 23b and c) both the planar and anti-skewed models for aureucidin are able to fit into the thyroxine binding site of prealbumin, suggesting a mode of inhibitor recognition. This model helps explain the iodothyronine analog inhibition data and suggests that more specific enzyme inhibitors can be designed.
Since the halogenated pollutants such as dioxin, DDT or the phenylalkanoic acid insecticides may bind to the thyroid hormone specific proteins in much the same manner as the thyroid hormones, models of their binding to the active site in prealbumin were also studied. As shown (Fig. 23), these compounds can fit into the thyroxine binding site and position at least one of their chlorines in an iodine binding pocket. These toxic agents can be accommodated in the binding site with their ring structures matched to either the inner or outer phenyl ring of thyroxine. While these models show that these pollutants can fit into the binding channel of prealbumin, the data for the deiodinase inhibitors suggest that not all of the inhibitor structure need overlap the thyroid hormone to elicit enzyme recognition. Therefore, these studies indicate the type of possible interactions that may be operative in the concentration of these pollutants by the thyroid hormone proteins.* Summary This paper has described the role of X-ray crystallographic data in delineating those structural characteristics that are relevant for enhanced active-site binding and function among a series of environmental and pharmacological agents. These examples show how structural data were used to define absolute configuration, molecular shape, conformational flexibility and active-site topology.
For example, structural analysis of the synthetic pyrethyroid insecticides has defined the absolute stereochemistry of the three chiral centers in the molecule and demonstrate the importance of conformation to the stereoselectivity of these compounds. The presence of a cis or trans configuration of the functional groups attached to the cyclopropane ring has a dramatic influence on the overall shape of the molecule with the (S)cis(1R,3R) configuration being required for maximal activity.
Within the series of DDT analogs, analysis of the diphenyl ring conformation indicates that there is a narrow range of values permitted for the most toxic compounds. These data further show that the overall shape of the molecule is important and that there is a steric size and aromatic character prerequisite for the bridging groups to elicit maximal toxicity.
Structural analysis of the phenylalkanoic acid herbicides indicates that factors other than molecular conformation dominate the structure-activity relationships since the active and inactive analogs can have the same conformation. These data, however, do suggest that the charge-density properties of these compounds are of importance to activity. Thus, the availability of accurate geometrical parameters for this series of compounds is a necessary starting point for these calculations.
Similarly, structural analysis of the lipophilic antifolates has shown that adamantyl substitution causes distortion of the pyrimidine ring planarity by steric crowding with adjacent ring substituents. This results in an overall "bowed" shape to the molecule and could explain the enhanced antineoplastic activity of this family of compounds compared with other antifolates with similar lipophilicities. Furthermore, the crystal structure determination of DHFR from several species has permitted observation of the active site and modeling of the binding of other antifolates within it. These studies have defined the conformational space available for binding and have delineated the important interactions of the inhibitor-enzyme binding complex and have further permitted the rational design of other inhibitors. Perhaps most importantly, this type of structural study now permits a more realistic interpretation of structure-activity relationships of systems where no specific structural data are yet available. For example, use of structural data from a related family of proteins to model the inhibitor-enzyme complex permits a more precise description of the types of interactions that are important. This approach is demonstrated in the interpretation of the structure-activity relationships among flavones as potent inhibitors of the enzyme iodothyronine deiodinase. By using the thyroxine binding site of prealbumin as a model, correlations were observed that could help explain the mode of action of these compounds. In a similar manner, this site was also used to explore the possible binding modes accessible to the halogenated environmental agents such as dioxin and DDT and in this way show a mode of entry of these species into tissues.
The future will see a greater emphasis on the input of specific and well-determined structural data, and as more protein structures and enzymes become available from solid state and solution, the "negative replica" of the receptor will develop into a clear "picture" of the active site mechanism of action of these important systems.
The author thanks Catherine DeVine, Danielle Smith and Yi-Yan Hong for their efforts in preparing the data for computer analysis and for the literature searches; Melda Tugac, Gloria DelBel, and Queen Bright for graphic illustrations, and Dr. William Duax for his critical review of this manuscript. Molecular volumes were computed by using a pseudoelectron density function computed on the MMS-X graphics display system, designed and fabricated by the Computer Systems Laboratory, Washington University, St. Louis. This research was supported in part by a grant from AM-15051.
*Recent workers have provided experimental evidence for their modeling predictions that chlorinated dibenzodioxins and dibenzofurans can be effective competitive binding ligands for thyroxine binding sites in prealbumin (78).