Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Computer-Assisted Molecular Modeling of Benzodiazepine and Thyromimetic Inhibitors of the HepG2 lodothyronine Membrane Transporter Laura Kragie, Maureen Mary McCourt L. Forrester, Vivian Cody, and Biological Sciences, Faculty of Natural Sciences and Mathematics (L.K.) State University of New York at Buffalo Amherst, New York 14260 Molecular Biophysics (M.L.F., V.C.) and Electron Diffraction Departments (M.M.) Medical Foundation of Buffalo Buffalo. New York 14203 T3 cellular uptake is inhibited in the presence of benzodiarepines (BZs). The structure-activity relationship of BZ inhibition correlates strongly with halogen substitution of the nonfused phenyl ring and indicates that this ring is required for activity. A structure-activity series of thyromimetic (TH) inhibitors of the HepGP iodothyronine transporter further point out the critical importance of the amino group of the alanine side chain, its L-stereo configuration, and the size of the substituents of the inner and outer phenyl rings. A third series of compounds, reported to interact at related sites, were inactive as HepG2 iodothyronine transport inhibitors, and therefore the potent inhibitors were restricted to the BZ and TH compounds. Using both of these BZ and TH structure-activity series along with computer-assisted molecular modeling techniques, we determined which chemical structural components were important at the transporter interaction site. By superimposing structures from active chemicals, excluding residues from poor inhibitors, and incorporating molecular electropotential data, we developed a five-point model of BZ conformational similarity to the endogenous transporter ligand, L-TJ: the alkyl substitution at the Nl of the BZ ring seems to simulate the alanine side chain of TJ, and the electronegative halogen and oxygen atoms of substituents at M/R7/R2’/R4’ of BZ form a pyrimidal pharmacophore that seems to correspond with the 3-1/5-l/ 3’-l/4’-OH substituents of T3, respectively. These points, suggesting a tilted cross-bow formation, may be sites for ligand interaction with the iodothyronine transporter. (Molecular Endocrinology 8: 382-391, 1994) 0ea9-eae9/94/o3a2~391$03.00/0 Molecular Endocrimy Copyright 0 1994 by TIM Endocrine INTRODUCTION In this paper we compare the molecular conformations of thyroid hormones to benzodiazepines (BZs), a potential class of membrane iodothyronine transporter antagonists. Thyroid hormones have profound effects on growth, differentiation, maturation of tissues, and the turnover of substrates, vitamins, and hormones. In mammalian species, the active forms are Tq and its more potent deiodinated metabolite, T3 (1). Facilitated carrier-mediated transport allows the hormones to cross the cell membrane and then to interact with cytosolic, enzymatic, mitochondrial, microsomal, and nuclear binding sites (2-4). The iodothyronine membrane transport process has been studied in many tissues, and these data show that membrane transport of TJ is energy dependent, stereo-specific, and critical for cellular nuclear binding and metabolism of T3 (for review see Refs. 5-7). In a pharmacological survey of compounds that may interact with the membrane iodothyronine transporter, it was shown that compounds of the BZ class, like the thyromimetics (THs), are reversible inhibitors of the TJ membrane transporter (8, 9). Because the structureactivity relationship (SAR) describing BZ inhibition of TS uptake in the HepG2 cell line suggests a site that is different from the central and peripheral BZ receptor (BZR) sites, we hypothesized that the BZs and THs may directly interact at the TJ binding site on the membrane transporter. Using both of these SAM, in conjunction with molecular modeling strategies, we developed a model for the structural comparison of iodothyronine transport inhibitors and now propose a pharmacophore for ligand interaction with the TJ transport site. society 382 lodothyronine RESULTS Transporter Pharmacophore AND DISCUSSION in Table 1. This SAR of BZ inhibition reveals that the nonfused phenyl ring is necessary for activity in our series. The strongest correlation occurs at the 2’-position of that ring. Although there are no compounds in our series differing only at R2’, comparisons between similar compounds show that Cl substitutions have the greatest activity, followed by F and H. In addition, a halogen residue at the R4’ and an hydroxyl substitution at R3 increases inhibitory activity. An alkyl group at Rl or on the imidazole/triazole group of the 1,2-annelated BZ series enhances potency. For the R7 group, a Cl is preferable over an NO*. However dihalogenated compounds that are Cl substituted at R7 and R2’ are slightly less potent than the compound that is Cl substituted at R2’ and R4’ (Ro22 8349); monohalogenated compounds Cl substituted at R7 vs. R4’ (Ro5 5115 vs. diazepam) have equal potency. The carbonyl substituent at R2 of the 1,4-BZ series also enhances potency. In the resulting SAR, the importance of the halogensubstituted nonfused phenyl ring is highlighted; this SAR seems unique when compared to those series reported for other BZ sites and/or effects (1 O-l 2). In addition to the BZ SAR series, we studied a limited series of TH inhibitory compounds. TH compounds can act as substrates for the carrier protein and are transported into the ceils. These analogs have different SARs Figure 1 diagrams the structure and nomenclature of the prototypic classical 1,4-BZ and L-T~ compounds. Table 1 lists the substitutions at sites within the BZ molecules of our BZ SAR series. The BZ potencies for inhibition, illustrated by -Log IC& values, are also listed R3 I HO T3 i Fig. 1. Structure Table 1. Structural and Nomenclature Components BZ Triazolam Lormetazepam Prazepam (31) Lorazepam (32) Ro22 8349 Midazolam Delorazepam Ro5 4864 Temazepam (33) Flurazepam Ro5 5115 Oxazepam (34) Diazepam (35) PK 11195 Clonazepam (36) Alprazolam Nordiazepam Medazepam (37) Estazolam (38) Chlordiazepoxide (39) Flunitrazepam (40) Nitrazepam (41) Bromazepam (42) Ro5 3663 Flumazenil of BZ and L-T~ of BZR Ligands Rl fused methylated CH3 CHPcyclopropyl H CH3 fused methylated H CH3 CHz CH&H2N(CH&H& CH3 H CH3 H H fused methylated H fused fused triazolo ring imidazo R2 R3 ring 0 0 0 0 ring 0 0 0 0 0 0 0 b H OH H OH H H H H OH H H OH H H H H H H H H H H H H H 0 ring 0 H CH3 triazolo ring, without CH3 H NHCHJ 0 CH3 H 0 H 0 H 0 imidazo ring + COO(C2H5) triazolo Iv Cl Cl Cl Cl H Cl Cl Cl Cl Cl H Cl Cl H Non Cl Cl Cl Cl Cl Non NOz Br H F R2' Cl Cl H Cl Cl F Cl H H F H H H Cl Cl H H H H H F H Nat2’ no C-ring, R5=CH3 no C-ring, R4=CH3, R4' H H H H Cl H H Cl H H Cl H H H H H H H H H H H H R5=0 -Log 7.38 7.35 6.69 6.54 6.38 6.25 6.12 5.91 5.83 5.82 5.60 5.53 5.50 5.05 4.95 4.94 4.86 4.44 4.37 I&o +. SEM + 0.17 * 0.07 + 0.09 + 0.05 + 0.12 + 0.06 + 0.03 k 0.08 f 0.14 f 0.10 f 0.14 f 0.08 f 0.10 f 0.14 + 0.33 f 0.16 + 0.16 f 0.14 + 0.05 <4.3 <4.3 t4.3 <4.3 inactive up to inactive up to (N) (6) (5) (5) (4) (5) (4) (3) (8) (3) (4) (3) (3) (5) (5) (4) (3) (3) (3) (3) (3) (5) (4) (3) 10 /IM 10 PM Structurally analogous residues numbered according to the nomenclature of classic 1 ,+BZs, pictured in Fig. 1. I& values listed as negative log of molar concentrations. Portions of this table were previously published (7). Crystal structure references given in parentheses. ’ N-oxide at position 4. ’ CON(CH3)CH(CH3XC2HS) and PK 11195 is an isoquinoline carboxamide instead of a benzodiazepine. MOL END0.1994 Vol8 No. 3 384 potencies with regard to their abilities as intracellular THs, and therefore they may be considered as possible partial agonists/antagonists of L-T~ intracellular receptors. Structural substitutions, rank order, and -Log lCs0 values of compounds from this SAR are listed in table 2. The proposed endogenous ligand/substrate for the transporter, TJ, is the most potent of our series. Some of the Go values must be interpreted cautiously. Because of negative feedback from intracellular hormone concentrations, these nonzero experimental conditions generate IGo values that do not necessarily reflect the affinity of the protein for the analog (13). In addition, the potency reported for D-T~ (Sigma Chemicals, St. Louis, MO) could possibly be inflated due to its probable contamination with L-T,+ The bromine and chlorine substituted SKF compounds (SKF 94901, SKF 94424, and SKF 94918) may also be subject to less intracellular dehalogenation when compared to the iodinated compounds (SKF 93236, SKF 95050, and SKF 94690). The potency of 3,5dimethyl,3’-isopropyl-L-T3 (DIMIT) also could be inflated due to its probable lack of metabolism by native deiodinases. The decrease in potency of reverse TB (rT3) in the 15 PM (*T3) assay relative to the 40 PM (*T3) assay (Table 2) may indicate an inflated potency due to the probable metabolism of rTB into 3’,3-T2 by the 5’deiodinase. Indeed, rTB and 3’,3T2 have similar I& values. On the other hand, 3,5-T* might be less metabolized by the 5deiodinase relative to TI, and hence its potency too could potentially be inflated. Nevertheless, a rank order can be useful in defining Table SARs. The 3’-arylmethyl-substituted SKF compounds define the upper size limits of the 3’-substitution of the iodothyronines. The necessity for electronegative substitutions at the 3’,3,5 sites is not absolute as indicated by the moderate potency of DIMIT. Overall, the SAR rank order derived from this brief TH series suggests the importance of the halogen substituents at the 3’,5,3 sites, the size of the 3’ substituent, the L-configuration of the amino acid side chain carbon, a sterically restricted zone at the 5’ site, the presence of the side chain amino group, and the alkyl length of the alanine side chain. A number of non-BZ and non-TH compounds that are aromatic or known to interact with either BZRs or thyroid hormone systems (9) were also screened for inhibitory activity. For example, thyroid hormone disorders in rats are associated with changes in cardiac dihydropyridine binding site density (14, 15) and large amounts of phenytoin and bromosulfophthalein are reported to reduce TJ uptake in cells (for review see Ref. 9). Furthermore, higher concentrations of BZs can interact with neurotransmitter receptor systems (16, 17). Table 3 lists the compounds and their range of assayed concentrations. These drugs were all inactive and therefore demonstrate that, of the chemicals tested, high potency in inhibiting [‘251]T3 cellular uptake so far is limited to the THs and the BZs. Molecular Modeling The BZ structure is a nonplanar, seven-member, heterocyclic ring fused to a planar phenyl ring and a 2. Structural Components of THs TH -Log 3' 4' 5' 6' 5 3 I’& 15 PM *Tz L-Ts(43) L-3,5-Tz(44) L-DIMIT L-rT3(45) L-3,3'-T2 Triac" L-T, Tetrac (46) D-T; SKF L-94690 SKF L-94918 SKFL-93236 SKFL-94424 SKF L-94901 SKF L'-95050 3,5-T* proprionic acid (47) D-Tg L-3Sdiiodotyrosine L-thyronine D,L-thyronine H Isopropyl H I I I I CH*-pyridazinone CHa-pyridazinone CH&-OH benzyl CHP-pyridone CH2-pyridazinone CH,-pyridone H I OHHHI OHHHI OH H OH I OH H OHHHI OH I OH I OHIHI OHHHI OH H OHHHI OH H OH H OHHI OHHHI OH No outer H OH H OH ’ Significant contamination with L-T,. Crystal structure H H H CHB H H H H I I H Cl H H Br Br I I I CHB I I I I I I I Cl I Br Br H I L-alanine L-alanine L-alanine L-alanine L-alanine Acetic acid L-alanine Acetic acid o-alanine L-alanine L-alanine L-alanine L-alanine L-alanine L-alanine Proprionic HI I D-alanine ring I I H H H H L-alanine H H H H D,L-alanine references given in parentheses. H + SEM (N) Side chain 6.61 5.28 5.28 4.77 4.68 4.67 6.80(2) + 0.20(4) 6.20(2) 6.08 (2) 6.02 (2) 5.42 (2) + 0.28(2) 5.19 (2) + 0.28(2) 4.93 (2) k 0.17 (3) f 0.19 (3) + 0.15 (3) 4.60 (2) 4.36 (2) 3.76 (2) C4.3 5.67 + 0.18 (3) 4.99 f 0.03 (3) (2) L-alanine C4.3 C4.3 40 DM ‘Ts 6.76 f 0.07 (3) C4.3 (2) (2) (2) lodothyronine Table Transporter 3. Screened Compounds Drug BAY K 8644 Nisoldipine Diltiazem PN 200-l 10 Theophylline Adenosine Prazosin Phentolamine Benztropine Yohimbine 385 Pharmacophore Which Did Not Inhibit TJ Uptake Concentration range 1 nr+0.1 0.1 0.1 0.1 1opM 0.1-100 lo/.~M 0.5 0.5-10 l-10 1 IlM-1 Drug PM /lM PM fih4 Quercetin Rutin Glyburide Phenylephrine Ciproheptidine Diphenhydramine Chlorpheniramine Nomifensine Protoporphyrin IX Bromosulfophthalein Phenytoin /.tM PM PM PM PM nonfused planar phenyl group substituent. The nonfused phenyl ring is restricted in its freedom of rotation by the steric interactions with the R7 halogen substituent, resulting in an orthogonal position relative to the diazepine ring system. The typical 1,4diazepine conformation is a boat with the C3 atom as the bow point (Fig. 2). Although this ring has some conformational flexibility, it does have conformational stability due to ring annelation and the presence of the C=N bond. Given this conformational flexibility, we sought a minimum energy conformation and the compounds were minimized using the MOPAC AM1 molecular orbital method on SYBYL. Overall there is good agreement between the minimized structures and the initial crystallographic data. To explore the range of flexibility of rotation of the nonfused phenyl ring, the C-rings of a few of the minimized BZ structures were rotated 0’ to 180’ in 10” increments without reoptimization. The results of these calculations suggest that these molecules have a relatively flat potential surface, between torsion angles 90” and 140°, consistent with conformational flexibility within this range. T3 is an iodine-substituted tyrosine joined via an ether linkage to an iodine-substituted phenol. The ether linkage (with an angle of 120’) fixes the outer phenolic ring to an orthogonal position relative to the plane of the inner tyrosyl ring; minimal steric hindrance occurs when the outer phenolic ring is coplanar and the inner tyrosyl Fig. 2. Two Perpendicular Views Illustrating a Typical Conformation The 1,4-diazepine ring system has a boat conformation. BZ Concentration range 0.5 0.1-100 0.1 1OfiM lo/.~M 1orM lOjIM 0.5 0.1-10 0.01-10 0.01-10 /LM PM PM /.LM /.IM yM PM ring is perpendicular to the plane of the ether linkage. The conformation of the outer ring substituents can be distal (away) or a proximal (near). The alanine side chain lies perpendicular to the inner tyrosyl ring; therefore the alanine side chain and the outer phenyl ring can lie on the same (cisoid) side of the inner tyrosyl ring or the opposite (transoid) side (18). Figure 3 illustrates a prototypical L-T~ cisoid conformation in which the ether linkage fixes the outer phenolic ring to an orthogonal position relative to the plane of the inner tyrosyl ring, and the 3’-iodine is proximal relative to the alanine side chain. This is the conformation of TJ used for our molecular comparisons. Initially, conformational comparisons were made between triazolam, the most active BZ of the 1,2-annelated series, with the less active estazolam. Figure 4A highlights the regions that are unique to the active BZs, as determined by excluded volume analysis. Here, the R2’-Cl and the triazole-CH3 are the regions important for inhibitory potency. In Fig. 48, lormetazepam, one of the most active of the 1,4 diazepine analogs, is superimposed over the less active medazepam. Again, the analysis highlights the halogen at R2’ and the methyl group at Rl; it also points out the R2 carbonyl on Fig. 3. A Typical L-T3-cis Conformation Where the Ether Linkage Holds the Outer Phenolic Ring to an Orthogonal Position Relative to the Plane of the Inner Ring The 3’-I is shown in the proximal conformation; the distal conformation would result if the outer, phenyl ring were rotated 180’. The compound shown is uncharged, as would be the predominant species at pH 7-8. MOL 388 ENDO. Vol8 1994 active 1,2-annelated BZ, paired with a less active BZ, estazolam. Two stereo views of the superimposition of triazolam and estazolam, excluding their inactive VdW volumes and displaying their active VdW volumes. Note those regions which are unique to the active BZ. 8, Conformational comparisons of lormetazepam, the most active of the 1 ,Cdiazepine analogs, with the less active BZ, medazepam. Two stereo views of the superimposition of lormetazepam and medazepam where their inactive VdW volumes are excluded and the active VdW volumes are displayed. Note those regions which are unique to the active BZ. The hydrogens on the nonfused phenyl rings of lormetazepam and medazepam were undisplayed for clarity. C, Superimposition of triazolam, estazolam, lormetazepam, and medazepam. These stereo views show the union of the two active BZs, triazolam and lormetazepam, minus the union of the two less active of the BZ structures, estazolam and medazepam. The active residue volumes are highlighted. The hydrogens on the nonfused phenyl rings of lormetazepam and medazepam were undisplayed for clarity. lormetazepam and the buckling of the A-ring. Figure 4C combines both the 1,4- and the 1,2-BZ comparisons to show the baskets of Van der Waals (VdW) volumes used as our template for the initial fit of TB. To explore the outer ring region of THs, we compared TB, the most active TH, to the less active compound, No. 3 T4 (Fig. 5). Conversely, here the excluded volume analysis subtracted the active structure from the inactive compound, in order to illuminate regions of Tq that are not in common with TB. The 5’-substituent highlighted by this comparison may account for T4’s reduced potency; this suggests that there is a restricted area beyond the 5’-H of TS. We went on to compare the electronegative properties of BZs and THs. In Fig. 6A, L-T~ is compared with DIMIT, a compound with slightly less potency than LTB. The isopotential map is nearly identical to that of LT3. When we compared L-T~ to D-T~, we used their charged side chain conformations in order to highlight differences in the amino acid side chain. Figure 6B shows the isopotential maps of this pair of stereoisomers and illustrates the significant difference in the electronegative profile with the L- vs. o-configurations. In Fig. 6C we have compared lormetazepam to L-T~. Note that, except for the lack of an electronegative group at BZ’s R4’, the isopotential maps are similar. In Fig. 6D, we compare triazolam with L-T~ in order to determine the optimal alignment of the alanine side chain relative to the triazole group at Rl-R2. This comparison shows that the nitrogens in the triazole group are electronegative like the R2 carbonyl group in the 1,4-BZ series and that it is best aligned between the carboxyl group of the alanine side chain of TJ and one of its inner ring iodines (3-l). Therefore, when we ade our fit, we did not alter the rotation of the Cl2 bond from the minimized conformation. Due to the buckling of the C3 bow of the B-ring, -ere is an asymmetry to the BZ molecular structure. We displayed and compared BZ images of structures from databases composed of the inverted as well as the noninverted crystal structures of the BZs and compared their steric volumes and isopotential maps to both the transoid and cisoid conformers of L-T~. Figure 7 illustrates our best model for the superimposition of the cisoid, proximal conformation of L-T~ and the active BZs as represented by lormetazepam and triazolam. Based upon the superimposition of active structures, combined with exclusion of residues from poor inhibitors that are not in common with the active compounds, we suggest a model for BZ structural similarity to the cisoid conformation of L-T~. Specifically we point out five overlapping molecular substituents: 1) R2’ halogen residue occupies the same density as the proximal iodine on the outer phenolic ring of T3; 2) R4’ halogen residue occupies the same position as the T3’s outer 4’-OH; 3) R3-OH on the diazepine ring fits the density Fig. 5. Two Stereo Views of the Conformational of TB. the Most Active TH with the Less Active The VdW volume unique to Tq is displayed. Comparison Compound, T.I lodothyronine Transporter 387 Pharmacophore D Fig. 6. Isopotential Triazolam vs. Maps Comparing: A, DIMIT vs. L-T~; B, D-T~ structural homology. B, Same view as A but with the halogen and oxygen VdW volumes highlighted. The points for the RMS fit were (TH to lormetazepam): C4’ to C4’, Cl’ to Cl ‘, 5-l to C7, 3-l to R3-OH, alanine chiral carbon to Rl methyl. L-T~ (charged conformers); C, Lormetazepam vs. L-T~; and D, L-T~ ig. 7. A, Stereo views of the model for superimposition of LT3 and the active BZs, lormetazepam and triazolam, revealing their vs. ...:‘<.T ..,, The resulting RMS fit is 0.838 A. of the 3-l of T,‘s inner ring; 4) R7 halogen residue resides in the same position as the 5-l of T,‘s inner ring; and 5) Rl alkyl or triazole groups align along the amino acid side chain of TI with the more electronegative COOH oriented toward the R2 carbonyl (1,4-BZ) or tertiary amine (triazole/imidazole), and the Rl alkyl groups oriented toward T,‘s primary amine (aminoacid carbon). However, the cisoid distal (phenyl ring rotated 180’) conformation of L-T~ also can be fitted just as well to BZ structures with their nonfused ring rotated 180’. This conformation is slightly less energetically favorable for the BZs and slightly more favorable for the TB molecule. Our model must and does accommodate this alternative conformation. The mirror (inverted) image to the BZ molecular conformation we have displayed also fits the image of cisoid, proximal L-T~ that we have displayed in our figures. This fit requires flipping of the L-T~ molecule and instead aligning the opposite, symmetric inner ring iodine to the R7 or R3 groups of this inverted BZ molecule. In order to keep the isopotential maps consistent, the amino acid carbon of L-T~ must be rotated about its axis. This resultant five-point RMS fit is 1.3 A. We can also apply other examples from the SAR rank order to this model. The comparison of D-T~ vs. LT3 examines the effects of the stereoisomerism of charge/steric hindrance. Are the charge effects of the alanine side chain a necessity or just an enhancement when compared to the analogous BZ methyl group? According to the SAR, L-T* is quite active, but the deaminated compound, TP proprionic acid, has greatly MOL END0.1994 VoI8 No. 3 388 diminished activity. Triac, the deaminated and demethylated acid derivative of T3, shows a moderately diminished activity. Therefore the influence of NH3 on the chiral carbon is crucial, especially when there are only two other potential contact points in the pharmacophore. In our model, this amino group aligns with the methyl groups at Rl, while the carboxyl group aligns toward the BZ electronegative groups at R2 and R3. The CH,-cyclopropyl group on prazepam (see Fig. 8) is similar in length to the amino acid side chain of L-T~. and this similarity may explain prazepam’s potency. However, the large alkyl groups on flurazepam and PKll 195, compounds with weaker potencies may indicate the size limitations of the side chain receptor pocket. The 3’-arylmethyl substitutions of the SKF compounds, combined with the 3’-isopropyl group of DIMIT, delimit the size of the TH 3’pocket. DIMIT’s electrostatic potential map (see Fig. 6A) demonstrates that the electronic properties of two molecules can be similar even though the steric maps may not. This raises the question of the importance of TH’s central core of electronegativity for activity (perhaps more so than steric factors) at the 35 and 3’ sites. However, DIMIT might have better potency than other compounds due to its probable resistance to metabolism. Although the potencies of the SKF compounds are similar, there may be a slight enhancement with halogen substitution at the inner ring 35 positions, the order following I > Cl > Br. The presence of large electronegative groups in the outer ring region opposite the BZ’s R2’ group severely decreases potency, as demonstrated by L-T~, rT3, and SKF 905050 (see Fig. 5). These data suggest that the distal end of molecule requires a very precisely oriented and sized fit into the active transporter site. The compounds diazepam, Ro22 8349, Ro5 4864, and Ro5 5115 demonstrate the importance of the halogen sites of BZ in our pharmacophore criteria. Compounds Cl substituted at R7 and R2’ have slightly less potency than compounds Cl substituted at R2’ and Fig. 8. Stereo Views of the Model for Superimposition of L-T~ and the Active BZ, Prazepam. Revealing the Homology of the CH,-Cyclopropyl Substituent at Rl to the Alanine Side Chain on L-T~ Before fitting structures, the torsion angle of the nonfused phenyl ring of prazepam was rotated from -30” to -55” (these angles have equal heats of formation). Components fitted were (TH to BZ, respectively): 4’.OH to 4’-H, Cl’ to Cl ‘, 5-l to C7. C2 (TH inner ring) to C2 (diazepine ring), alanine CH, to CH? of Rl group, and alanine chiral carbon to base C of the Rl cyclopropyl structure. The resulting RMS fit is 1.02 A. R4’; compounds Cl substrtuted at R7 vs. R4’ have equal potency. Hence the 2’-halogen may be slrghtly more influential than the R7 halogen In the dihalogenated system, and therefore this position was allowed to be frt less precisely than the other proposed pharmacophore sites. CONCLUSION These SAR data and the molecular modelrng of crystallographrcally derived structures suggest a pharmacophore for the T3 site of the putatrve membrane iodothyronine transporter of HepG2 cells. To produce a compound wrth moderate (0.1-l PM) inhibitory activity, at least three of the following structural components are needed (BZ/TH. respectively): 1) R2’-halogen/outer ring iodine; 2) R4’-halogen/outer ring 4’-OH; 3) R3-OH/ inner ring 3-l; 4) R7-halogen/inner ring 5-1, 5) Rl alkyl group or tnazole group/amino acid side chain with Lconfiguration of the primary amine, and 6) no halogen at the 5’-outer ring positron of TH. Potentially good BZs for T3 uptake inhibitors may include modifications of the BZ triazolam such as an additional 4’-hydroxyl group, a halogen (Cl or I) or a hydroxyl at R3, and an amino group on the tnazole’s methyl. To improve on lormetazepam, an alanine side chain could be put at the Rl positton consistent with the L-configuration, as well as adding an OH at the 4’-position of the nonfused phenyl ring. BZs likely to be ineffective would have a nonfused phenyl rrng with a 3’- or 2’-halogen in addition to a 5’or 6’-halogen along with substitution of either a halogen or an OH at R3. These substitutions then would fix the nonfused ring into an orthogonal conformation and thrust the transord-positioned halogen into the restricted zone. As well, an alanine side chain In the Dconfiguration at positron Rl or R2 would render the BZ ineffective. The binding and TH pharmacophorefor the T3 nuclear receptor emphasize the importance of a precise fit of analogs to the receptor actrve site (19-22). Ion pairing to the receptor is suggested by the ionrc side chain on one end and a polar phenolic-OH on the other. The nuclear receptor. though, does not distinguish between the L- or o-configuration of the side chain. The 3,5 substrtuents define the size of the lipophilic pockets as well as the stenc constraints that hold the conformation of the aromatic rings; binding is enhanced by polanzable groups. A lipophilrc residue of limited size IS required at the 3’-site and must be capable of being positioned distal to the nonphenolrc inner ring; here an isopropyl group binds better than an Iodine. The similarities of our model to that of the T3 nuclear receptor pharmacophore include the restrictive zone at the 5’-position (although our SAR seems more restrictive) and increased activity with halogen substttution at 3’,3,5positions. Our differences Include the improvement of inhibitory activity with halogen substitutron at the 3’,3,5sites relative to substitution with alkyl groups and the lodothyronine Transporter Pharmacophore absolute requirement of L-configuration at the chiral amino acid carbon. We did not test the importance of the 4’OH site or ether linkage substitutions in our brief SAR series, but the outer ring 4’-OH of TS does overlap the 4’-halogen of BZ in our model of BZ/TH homology. As well, our 3’-I is fit to the BZs while in the proximal position, although the distal conformation can just as easily fit BZs with a rotated C-ring. And as the SAR and isopotential map comparisons show, DIMIT and triac are not the best ligands for the iodothyronine site, but they are high affinity ligands for the nuclear receptor (20). Finally, TJ binding in isolated rat hepatic nuclei show no competition from up to 10 @I triazolam and lormetazepam (8). Perhaps the nuclear receptor requires a more precise fit of halogens at the 35 and 3’ positions when compared to the transporter site. BZs are a class of pharmaceutical agents that are primarily used as sedative-hypnotics and anxiolytics. The BZRs for these central nervous system actions are contained in the r-aminobutyric acid receptor chloride ion channel complex (23, 24). At least three subtypes of BZRs have been described that exist in different brain regions and subserve different physiologic functions (25, 26). There is also a BZR that is not linked to the y-aminobutyric acid-BZ-Cl ion channel complex, termed the “peripheral BZ receptor” (10, 23, 24) although it is present in the central nervous system. Several models of the central BZR pharmacophore have been proposed. Common features of these models include two proton-accepting groups separated by approximately 3.0-3.5 A (11). The condensed aromatic ring is also a common feature for all high affinity ligands, but it may not be involved directly in the binding site (11, 12). For inverse agonists of the central BZR, Allen et al. (27) suggest an aromatic ring system constrained to one plane and containing two hydrogen binding sites; the antagonists have long substituents that access regions above or below this aromatic plane. Of the compounds with classic BZ structure, central BZRs show stereoselectivity, are not affected by the R-l alkyl groups, and compounds with 4’-substitutions have diminished affinity (28). The peripheral BZR is distinguished by such varied selective ligands as 4’-Cl diazepam, the isoquinoline carboxamide derivative PK 11195, and protoporphyrins, all of which are devoid of activity at the central BZR (10). Of the classical BZ compounds, the peripheral site prefers an alkyl group of three carbons or less at position 1, a carbonyl at position 2, a 4’-halogen or methoxy group, and halogens at R2’ and R7; substitution at R3 or R4 reduces affinity (28). Our pharmacophore model shares some characteristics from both the central and peripheral BZR pharmacophores. Our model for the iodothyronine transporter includes the planar aromatic ring system with electronegative substituents. However, it requires an additional halogen-substituted phenyl ring oriented perpendicular to the plane of the aromatic system, and our sites for the electronegative substitutions include both the central BZR-preferred (7, 2’) and the peripheral 369 BZR-preferred (4’) positions, as well as at the central/ peripheral BZR-preferred R2, and the peripheral BZRaversive R3 position (although we have not yet determined its stereoselectivity for this substitution). Like the peripheral site, we limit alkyl substitution at Rl to one to three carbons in our model. The alkyl substitution of BZ may simulate the alanine side chain of TJ in our model of fit. Electrostatically, the substitution at the BZ Nl may delocalize the alkyl electrons into the ring system and allow for greater nucleophilicity at substitutions in the aromatic ring system. As well, the necessity of the carbonyl at R2 allows for an amide linkage to Nl and prevents its protonation and the localization of the alkyl electrons (28). The highly electronegative halogen and oxygen atoms at R3/R7/ R2’/R4’ of BZ form a pyrimidal pharmacophore that seems to correspond with the 3-l/5-l/3’-l/4’-OH substitutions of TS. These five points, defining a tilted crossbow formation, may provide proton acceptor sites (such as protonated lysine residues) for interaction with the iodothyronine transporter. The highly sensitive stereospecificity of the TH chiral carbon may suggest hydrogen bonding and/or ion pairing to the active site at the side chain (BZ-Rl) portion of the molecule. MATERIALS AND METHODS Materials Tissue culture media and supplements were from Sigma Chemicals and GIBCO (Grand Island, NY). Our source for [‘251] TB, specific activity of 2200 Ci/mmol. was New England Nuclear, DuPont (Boston, MA). BZs were either purchased from Sigma (St. Louis, MO) or generously supplied from Drs. Peter Sorter (Hoffmann La Roche, Nutley, NJ) or David Triggle (State University of New York, Buffalo, NY). SKF compounds were a generous gift from Dr. Tony Underwood (Smith Kline Beecham, Welwyn, England). The source for the THs (highest purity available) L-T~, L-rT3, L-T~, L-T~, D-T~, and DIMIT was Henning/ Berlin (Berlin, Germany) and for D-T,, 3,5-~-T,, diiodotyrosine, D,L- and L-thyronine was Sigma Chemicals. 3,3,-~-T* was originally from Dr. E. C. Jorgensen. Our HepG2 cell line was a gift from Dr. B. Knowles. All other chemicals were from Sigma and Baker (Phillipsburg, NJ). Uptake Assay and lCso Values for Inhibition Our procedure to screen for inhibitors of labeled T3 uptake is as described (8,9). HepGP cells were grown in multiwell cluster plates in Dulbecco’s modified Eagle’s media low glucose with 10% fetal bovine serum. Cell suspensions were plated into 12to 24-well plates in equal volumes containing 1 04-1 O5 cells per well and grown to confluence (l-2 x 1 O6 cells/cm*) in 4-14 days. Before assay, wells were filled with Hank’s balanced salt solution plus the drugs or solvent control. To initiate the assay, [‘251]T3 was added to make a final concentration of 15 40 PM. At termination of the assay (30-60 min), wells were washed with a cold glycine buffer, hydrolyzed in NaOH. and then transferred to vials for counting in a r-counter. Each well is one of triplicates for each drug or solvent dose. Temperature-dependent uptake of [‘251]T3 was determined by subtraction of uptake at 4 C from the uptake at 22 C. The nonspecific binding/uptake was defined by uptake at 4 C incubation. I& values were derived from dose-response curves of individual experiments using a standard set of pharmacological statistical programs (29) implemented on an IBM personal computer. MOL 390 ENDO. Vol8 1994 Crystallographic Databases A search of the Cambridge crystallographic database (Quest 3D, Database 5.04, October 1992, Medical Foundation, Buffalo, NY) revealed crystal structures for 12 of the 25 BZs from our structure-activity series. Structures without available crystal data were created from modifications of the most homologous published BZ crystal structure using the SYBYL software package (Tripos Associates, St. Louis, MO). For example, triazolam was built from estazolam and lormetazepam from lorazepam. We used the thyroid analog crystal structures L-T~ and L-IT3 to build the remaining analogs of our TH series. All structures were energy minimized by using the MOPAC AM1 molecular orbital option in SYBYL on a Silicon Graphics work station (4D70GTB). We also explored the potential energy surface for rotation of the fused diazepine ring (A/B-rings) and the nonfused ring (C-ring). To do this, we started with the minimized structure and rotated the torsion angle between the fused A/B-rings and the nonfused C-ring in 10” increments form 0” to 180”. At each 10” increment a single point energy was calculated, without reoptimization of the structure. Least-Squares dispensable technical assistance. Dr. Kragie also thanks Drs. Jack Mendelson and Nancy Mello of the Alcohol and Drug Abuse Research Research Center, McLean Hospital, Harvard Medical School, for their current support. Received October 28, 1993. Revision received December 14,1993. Accepted December 22,1993. Address requests for reprints to: Dr. Laura Kragie, Biological Sciences, Cooke Hall, State University of New York at Buffalo, Buffalo, New York 14260-l 300. This work was supported by NIH Grants DK-01456 (to L.K.) and DK-41009 (to V.C.). Portions of this paper were presented at the 74th Annual Meeting of The Endocrine Society, San Antonio, TX, June 1992, and the 23rd National Medicinal Chemistry Symposium, Division of the American Chemical Society, Buffalo, NY, June 1992. REFERENCES Fit and VdW Volumes Because the fused benzene ring represents the only conformationally stable component in the BZ ring system, conformational comparisons between BZ molecules were made by a least squares fit of the fused benzene ring in each BZ (of the same stereoisomer) using the root mean square (RMS) fit option in SYBYL. The BZs comprised two structural groups, the classical 1,4-BZs (e.g. diazepam) and the 1,2-annelated BZs (e.g. triazolam). The most (A) and least (L) active compounds from each group were superimposed through the RMS fit option. The SYBYL option Compare Volumes subtracted their VdW volumes. For the BZ group, L was subtracted from A to highlight the features unique to the active compounds. The volume in common with the least active compound was subtracted to form the active volume displays (excluded volume analysis). A similar methodology was used for thyroid hormone analogs. These THs were RMS fit by superimposing the ether linkage and both phenyl rings. However, because our TH SAR data suggested that the 5’substituents of the inactive compounds reduced potency, we subtracted A from L to highlight groups unique to the inactive compound. These regions were then avoided in the template for T1 fit. The TS molecule was fit to the VdW surfaces in common to the most active BZ molecules (27). This produced our first approximation of T3/BZ fit. We then went on to RMS fit the following atoms (TH to BZ): Cl’ to Cl’; R4’ to R4’; chiral carbon of the alanine side chain to Rl methyl; 5-l to C7; and 3-l to R3 (see Fig. 1). lsopotential No. 3 Maps We created the isopotential maps by using the Isopotential Map option of the SYBYL software package. This option uses the MOPAC geometry and the charges that were created when the structures were energy minimized. It computes and displays the isopotentials at discrete points of the molecule. We calculated the negative isopotential maps of the minimum energy conformers. The level of the surface of -5.0 kcal was used for the comparisons of Fig. 6, A, C, and D. For the comparisons between the charged conformers of L-T~ and DT3 (Fig. 6B), a level of -17 kcal was chosen for display (30). Acknowledgments The authors thank Dr. Darrell Doyle for his sponsorship of Dr. Kragie and the financial support of Boots Pharmaceuticals (Lincolnshire, IL). Dr. David Smith provided assistance with the SYBYL software programs along with the Medical Foundation graphics department. Rich Smiehorowski provided in- 1. Wartofsky L. lngbar SH 1991 Diseases of the thyroid. In: Wilson JD, Braunwald E, lsselbacher KJ, (eds) Harrison’s Principles of Internal Medicine, ed 12. McGraw-Hill, New York, pp 1692-l 712 2. DeNayer P 1987 Thyroid hormone action at the cellular level. Horm Res 26:48-57 3. Fliegel L, Newton E, Burns K, Michalak M 1991 Molecular cloning of cDNA encoding a 55 KDa multifunctional thyroid hormone binding protein of skeletal muscle sarcoplasmic reticulum. J Biol Chem 26515496-l 5502 4. Ashizawa K, Kato H, McPhie P. Cheng S-Y 1990 Regulation of thyroid hormone binding to its cytosolic binding protein by L-a-alanine. Biochem Biophys Res Comm 167:587-592 5. Kragie L 1993 Neuropsychiatric implications of thyroid hormone and benzodiazepine interactions. Endocr Res 19:1-32 6. Kastellakis A, Valcan T 1989 Characterization of thyroid hormone transport in synaptosomes from rat brain. Mol Cell Endocrinol 67:231-241 7. Robbins J, Lakshmanan M 1992 The movement of thyroid hormones in the central nervous system. Acta Med Austriaca 19:21-25 8. Kragie L, Doyle D 1992 Benzodiazepines inhibit temperature-dependent L-[‘Z51]triiodothyronine accumulation into human liver, human neuroblast, and rat pituitary cell lines. Endocrinology 130:121 l-l 216 9. Kragie L 1992 Requisite structural characteristics for benzodiazepine inhibition of triiodothyronine uptake into a human liver cell line. Life Sci 51 :PL83-PL88 10. Verma A, Snyder SH 1989 Peripheral type benzodiazepine receptors. Annu Rev Pharmacol Toxicol 29:307-322 11. Villar HO, Davies MF, Loew GH, Maguire PA 1991 Molecular models for recognition and activation at the benzodiazepine receptor: a review. Life Sci 48:593-602 12. Haefely W, Kyburz E, Gerecke M, Mohler H 1985 Recent advances in the molecular pharmacology of benzodiazepine receptors and in the structure-activity relationships of their agonists and antagonists. In: Testa B (ed) Advances in Drug Research. Academic Press, New York, vol 14: 165-323 13. Osty J, Valensi P, Samsom M, Francon J, Blondeau J-P 1990 Transport of thyroid hormones by human erythrocytes: kinetic characterization in adults and newborns. J Clin Endocrinol Metab 71 :1589-l 595 14. Kragie L. Kwon Y, Smiehorowski R 1993 Rat cardiac calcium channels and their relationships with fi-adrenergic and muscarinic receptors in hypothyroidism. Endocr Res 19:57-71 15. Hawthorn M, Gengo P, Wei X-Y, Rutledge A, Moran J, lodothyronine 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. Transporter Pharmacophore Gallant S, Triggle D 1988 Effect of thyroid status on padrenoceptors and calcium channels in rat cardiac and vascular tissue. Naunyn-Smiedebergs Arch Pharmacol 3371539-544 Rampe D, Triggle DJ 1986 Benzodiazepine and calcium channel function. Trends Pharmacol Sci 7:461-464 Kenakin TP 1982 The potentiation of cardiac responses to adenosine by benzodiazepines. J Pharmacol Exp Ther 222:752-758 Cody V 1980 Thyroid hormone interactions: molecular conformation, protein binding, and hormone action. Endoer Rev 1 :140-l 66 Oppenheimer JH 1989 Tissue and cellular effects of thyroid hormones and their mechanism of action. In: Burrow G, Oppenheimer J, Volpe R (eds) Thyroid Function and Disease. WB Saunders, Philadelphia, pp 90-l 23 Jorgensen E 1976 Structure activity relationships of thyroxine analogs. Pharmacol Ther Part B Gen Syst Pharmacol2:661-682 Cheung EN-y 1985 Thyroid hormone action: determination of hormone receptor interaction using structural analogs and molecular modeling. Trends Pharmacol Sci 6:31-34 Cody V 1979 Thyroid hormones-receptor interactions: binding models from molecular conformation and binding affinity data. In: Olson E, Christoffersen R, (eds) Computer Assisted Drug Design. American Chemical Society, Washington, DC, pp 281-299 Drugan R, Holmes P 1991 Central and peripheral benzodiazepine receptors: involvement in an organism’s response to physical and psychological stress. Neurosci Biobehav Rev 15:277-298 Gavish M, Katz Y, Bar-Ami S, Weizman R 1992 Biochemical, physical and pathological aspects of the peripheral benzodiazepine receptor. J Neurochem 58:1589-l 601 Doble A, Martin IA 1992 Multiple benzodiazepine receptors: no reason for anxiety. Trends Pharmacol Sci 13:7681 Ruano D, Vizuete M, Cano J, Machado A, Vitorica J 1992 Heterogeneity in the allosteric interaction between the yaminobutyric acid (GABA) binding site and three different benzodiazepine binding sites of the GABAa/benzodiazepine receptor complex in the rat nervous system. J Neurochem 581485-493 Allen M, Tan Y-C, Trudell M, Narayanan K, Schindler L, Martin M, Schultz C, Hagen T, Koehler K, Codding P, Skolnick P, Cook J 1990 Synthetic and computer-assisted analyses of the pharmacophore for the benzodiazepine receptor inverse agonist site. J Med Chem 33:2343-2357 Wang J, Taniguchi T, Spector S 1984 Structural requirements for the binding of benzodiazepines to their peripheral-type sites. Mol Pharmacol 25:349-351 Tallarida R, Murray R 1986 Manual of Pharmacologic Calculations, ed 2. Springer Publications, New York Massa S, Artico M, Mai A, Corelli F, Botta M, Tafi A, Pantaleoni P, Giorgi R, Coppolino M, Gagnotto A, Skorupska M 1992 Pyrrolobenzodiazepines and related systems. 2. Synthesis of properties of isonoraptazepine derivatives J Med Chem 35:4533-4541 Brachtel G, Jansen M 1981 7-Chloro-2,3-dihydro-1 -cycle- 391 propylmethyl-5-phenyl-1,4-benzodiazepam: prazepam. Cryst Struct Comm 10:669-672 32 Bandoli G, Clemente D 1976 Crystal molecular and electronic structure of an antianxiety agent: 7-chloro-5-(2chlorophenyl)-l,3-dihydro-3-hydroxy-l,4benzodiazepine-2-one. J Chem Sci Perkin Trans 2:413-418 33 Galdecki Z, Giowka M 1980 7-Chloro-3-hydroxy-1 -methyl5-phenyl-l,3-dihydro-2H-l,4-benzodiazepin-2-one: temazepam: Acta Crystallogr Sect B Struct Sci 36:3044-3048 34 Gilli G. Bertolasi V. Sacerdoti M. Borea P 1978 7-Chloro1,3-dihydro-3-hydroxy-5-phenyl:2H-l,4-benzodiazepin-2one: oxazepam. Acta Crystallogr Sect B Struct Sci 34:2826-2829 35 Camerman L, Camerman N 1972 Diazepam: 7-chloro-1,3dihydro-l-methyl-5-phenyl-2H-1,4-benzodiazepin-2-one. J Am Chem Sot 94:268-272 36. Chananont P, Hamor T, Martin I 1979 5-(P-Chlorophenyl)1,3-dihydro-7-nitro-2H-l,4-benzodiazepin-2-one. Clonazepam. Cryst Struct Comm 8:393-396 37. Gilli G, Bertolasi V, Sacerdoti M 1978 7-Chloro-2,3-dihydro-1 -methyl-5-phenyl-1 H-l ,4benzodiazepine (MedazePam). Acta Crystallogr Sect B Struct Sci 34:3793-3795 38. Kamiya K, Wada Y, Nishikawa M 1973 8-Chloro-6-phenyl4H-S-triazolo(4, 3a)(1,4) benzodiazepine. Chem Pharm Bull (Tokyo) 21 :1520-l 522 39. Hernstadt C, Mootz D, Wunderlich H, Mohrle H 1979 7Chloro-2-methylamino-5-phenyl-3H-l,4benzodiazepine4-oxide hydrochloride. J Chem Sot Perkin Trans 2:735738 40. Butcher H, Hamor T, Martin I 1983 5-(2-Fluorophenyl)1,3-dihydro-1 -methyl-7-nitro-2H-1,4-benzodiazepin-2one: flunitrazepam. Acta Crystallogr Sect C Struct Commun 39:1469-l 472 41. Gilli G, Bertolasi V, Sacerdoti M, Borea P 1987 7-Nitro1,3-dihydro-5-phenyl-2H-1,4-benzodiazepin-2-one: nitrazepam. Acta Crystallc-gr Sect B Struct Sci 33:2664-2667 42. Butcher H, Hamor T, Martin I 1983 7-Bromo-1 .Sdihydro5-(2-pyridyl)-2H-1,4-benzodiazepon-2-one: bromazepam. Acta Crystallogr Sect C Cryst Struct Commun 39:14691472 43. Cody V 1974 Crystal structure and molecular conformation of the thyroid hormone distal 3,5,3’ triiodo-l-thvronine. J Am Chem Sot 96:6720-6725 44. Codv V. Duax WL. Norton DA 1972 Molecular conformatibn of the 3,5 diiodo-L-thyronine-n-methylacetamide complex (1:l). Acta Crystallogr Sect B Struct Sci 28:2244-2252 45. Okabe N, Fujiwara T, Yamagata Y, Tomita K-l 1982 The crystal structure of a major metabolite of thyroid hormone 3,3’,5’-triicdo-L-thyronine. Biochim Biophys Acta 717:179-l 81 46. Cody V, Hazel J, Langs DA, Duax WL 1977 Molecular structure of thyroxine analogues: crystal structure of 3,5,3’triiodothyroacetic and 3,5,3’,5’tetraiodothyroacetic acid-Ndiethanolamine (1 :l) complexes. J Med Chem 20: 1628 47. Cody V, Erman M, De Jarnette E 1977 Crystal structure of the active thyroxine analogue, 3,5 diiodothyropropionic acid (3-(4-hydroxyphenoxy-3,5-di-iodophenyl) propanoic acid. J Chem Res (M) 126:1444-l 465