Molecular Interactions between Thyroid Hormone Analogs and the Rat Liver Nuclear Receptor PARTITIONING OF EQUILIBRIUM BINDING FREE ENERGY CHANGES INTO SUBSTITUENT GROUP INTERACTIONS*

Chromatin-localized non-histone proteins, which may mediate the actions of thyroid hormones, are present in rat liver nuclei. A study of the binding affinities of ~-3,6-[3’-’~~1]triiodothyronine ([‘261]T,) and a series of thyroid hormone analogs to the solubilized nuclear receptor has provided information about the structural and stereochemical nature and magnitude of interac- tions in the hormone receptor complex. The receptor binding affinities of 67 thyroid hormone anaIogs were determined in a competitive binding assay with I’*‘IJTS (& = 1.2 & 0.4 X loB ”l). These results were used to calculate the free energy of binding (AGO) for each analog in order to determine, by first order partitioning of free energies, the nature and magnitude of specific substituent interactions with the receptor. The binding of [’“IJT3 to the solubilized receptor is associated with a change in free energy. AGO = -12.4 (-t 0.2) kcal/mol. The individual substituent group con- tributions are: 1) The 4”hydroxyl participates in a donor hydrogen bond oriented toward the 6’ side of the outer ring and adds -1.2 (2 0.2) kcal/mol of binding free energy. 2) The 3’ substituent participates in direct hydrophobic and van der Waals bonding with a size limit at isopropyl. The 3’ substituent also enhances the strength of the 4’-hydroxyl interaction. The contribution

Chromatin-localized non-histone proteins, which may mediate the actions of thyroid hormones, are present in rat liver nuclei. A study of the binding affinities of ~-3,6-[3'-'~~1]triiodothyronine (['261]T,) and a series of thyroid hormone analogs to the solubilized nuclear receptor has provided information about the structural and stereochemical nature and magnitude of interactions in the hormone receptor complex.
The receptor binding affinities of 67 thyroid hormone anaIogs were determined in a competitive binding assay with I'*'IJTS (& = 1.2 & 0.4 X loB "l). These results were used to calculate the free energy of binding (AGO) for each analog in order to determine, by first order partitioning of free energies, the nature and magnitude of specific substituent interactions with the receptor. The binding of ['"IJT3 to the solubilized receptor is associated with a change in free energy. AGO = -12.4 (-t 0.2) kcal/mol. The individual substituent group contributions are: 1) The 4"hydroxyl participates in a donor hydrogen bond oriented toward the 6' side of the outer ring and adds -1.2 (2 0.2) kcal/mol of binding free energy. 2) The 3' substituent participates in direct hydrophobic and van der Waals bonding with a size limit at isopropyl. The 3' substituent also enhances the strength of the 4'-hydroxyl interaction. The contribution to the binding free energy of a 3"iodine in TI is -4.1 (& 0.4) kcal/mol. 3) The optimal 3,6 substituents are iodine atoms which can each contribute an average of -3.4 ( 2 0.7) kcal/mol. T h i s value contains the interactive effect on orientation of the outer ring, as well as the direct contribution to binding by the 3,S-iodine atoms and the aromatic rings. 4) The alanine side chain probably participates in an electrostatic attraction between the carboxylate anion and a positively charged amino acid residue in the receptor but due to the presence of the a-ammonio group adds a negligible -0. activities. They induce metamorphosis in amphibia (1) and play a fundamental role in regulating mammalian development and metabolism (2)(3)(4). The early work of Tata and associates resulted in the fmt proposal that thyroid hormones initiate their actions primarily by stimulation of transcription of genetic information at the nuclear level (5).
The presence of limited capacity binding sites for triiodothyronine' in the nuclei of both rat liver and kidney was reported first by Oppenheimer and co-workers in 1972 (6). In 1973, Oppenheimer presented a study of the binding affinities of a limited number of thyroid hormone analogs administered in vivo and found that if differences in metabolism and distribution were taken into account there was a good qualitative correlation between the analog binding affinity and its biological potency (7). Koerner et al. also demonstrated in vitro the presence of specific T S binding sites in isolated rat liver nuclei (8). In collaboration with Jorgensen, these workers used the binding to intact rat liver nuclei to study the affinities of an extensive series of thyroid hormone analogs (9). This study showed a high correlation between structural requirements for nuclear binding and hormonal activity. More recently, a quantitative correlation between in vitro receptor binding and in vivo rat antigoiter activity has been developed (10,11). These data have provided information on the structural requirements for thyroid hormone binding in vitro and the correlation of this with in vivo activity, which has supported further the physiological relevance of the nuclear binding sites.

Thyroid Hormone-Nuclear
Receptor Interactions chromatography (14,18,19). In order to define more clearly the nature of the molecular interactions between thyroid hormone and the nuclear receptor, we measured the binding affinities of an extensive series of thyroid hormone analogs to a solubilized nuclear protein preparation from rat liver. By utilizing the methods of Latham et al. (18), a preparation of solubilized nuclear protein was obtained which could be stored for over 6 months at liquid N2 temperature while retaining high affinity, saturable specific binding activity with [lZ5I]T3. Previous studies of in vivo and in vitro thyroid hormone activity have determined the basic structural features of the thyroxine molecule required for activity. We selected analogs for these binding studies to obtain information about the nature of the molecular interactions between the hormone and its receptor. The binding studies with the solubilized rat liver nuclear receptor have allows us to: (a) determine the strength and probable orientation of the 4"hydroxyl interaction with the receptor; ( b ) define the size limit of hydrophobic and dispersion force interactions of the 3,5, and 3' substituents with the receptor; and (c) define the probable type of electrostatic interaction between the alanine side chain and the receptor.  Green (20) and modified by Latham et al. (18). This hormone preparation usually contained 3% of unbound [1251]iodide ion. Due to the loss of [1251]iodide (1 to 3%/month) and the short radioactive half-life of 60 days for '"iodine, each preparation was used for no more than 2 months from the provider's assay date. In order to calculate the true molarity of a given radioactive sample of ['*'I]T3 from the measured counts per min, corrections were made to account for the amount of ['251]iodide present and for the loss of hormonally active molecules due to conversion of lz5I to I2"re by y emission (0.35 meV) and electron capture (see Appendix I, miniprint section). ' TI, To, and D -T~ were from Nutritional Biochemicals Co. (Cleveland, OH), triform was from Cyclo Chemicals (Los Angeles, CA), and triac and triprop were from Sigma Chemical Co. (St. Louis, MO). L-Thyronine, 3-iodo-~-thyronine, 3,3"diiodo-~-thyronine, 3,5-diiodo-3'methyl-L-thyronine, 3,5-diiodo-4-(4'-methoxy-3'-methylphenoxy)-~phenylalanine, 3,5-diiodo-3',5'-dimethyl-~-thyronine, 3,5-diiodo-4-(4'methoxy-3',5"dimethyl)-~-phenylalanine, 3,5-diethyl-3'-iodo-~~-thyronine, and 3,5-diethyl-T,5'-diiodo-~~-thyronine were from Dr. Paul Block, Jr., University of Toledo, Toledo, OH. Thyroxamine and 3,5diiodo-3'-fluoro-~~-thyronine were from Dr. Rosalind Pitt-Rivers, University College, London, England. 3,5,3'-Triiodothyrobutyric acid was from Dr. Benjamin Blank, Smith Kline and French Laboratories (Philadelphia, PA). The remaining compounds were prepared in our laboratory. All of the compounds were determined to be pure by migration as a single spot in thin layer chromatography on precoated silica gel plates with fluorescent indicators (250 pm coating on glass with preabsorbent loading zone (Whatman-Quantum, Inc., Clifton, NJ) using ch1oroform:methanol:concentrated ammonium hydroxide (2010:1).

Thyroid H0rmones-3,5-[3'-'~~I]Triiodo-~-thyronine
Preparation of Solubilized Nuclear Protein-Male Sprague-Dawley rats (350 g/rat, approximately 10 g of tissue/liver, supplied by Simonsen Laboratories, Gilroy, CA) were anesthetized lightly with chloroform and killed by cervical dislocation. The aorta was severed and blood was drained from the liver. The livers were excised and immediately stored in liquid nitrogen. The preparation of solubilized  (18). This extract was stored in 2-ml fractions at liquid nitrogen temperature in sealed plastic test tubes (A/S NUNC, Cole Scientific, 38 X 12.5 mm, No. 1076, with screw caps and Teflon sealing rings). The extract did not lose binding activity on freezing or after 6 months of storage. Endogenous triiodothyronine in the nuclear extract was reported by Latham to be less than 1 PM as measured by radioimmunoassay (18). Since the binding assays routinely used 0.1 nM hormone, 1 PM of endogenous cold triiodothyronine was considered to be insignificant. It was therefore unnecessary to remove the thyroid glands of the rats to lower endogenous hormone levels. Protein concentrations were determined by using the Lowry method (21).
Receptor Binding Assay-Two parallel sets of incubation tubes (12 X 75 mm) were prepared to determine: (a) the total amount of labeled hormone that is bound to the receptor and ( b ) the amount of "nonspecifically" bound radioactivity. The nonspecific component is defined as nonsaturable binding of [Iz5I]T3 to low affinity, high capacity binding sites and was detected as bound radioactivity in the presence of a 1000-fold excess of unlabeled triiodothyronine (5 X M).
A tracer dose of [IZ5I]T3 (0.4 nM, 100,000 cpm) was added to each tube. Incubation Buffer I (50 mM sodium phosphate (pH 7.6), 0.2 M ammonium sulfate, 1.0 mM dithiothreitol, 5% glycerol) was added to each tube to make the final volume (including nuclear extract) 0.5 ml.
To start the assay, 0.1 ml of nuclear extract (-200 PM of binding sites, 0.3 mg of protein) was added to the incubation tubes. The contents of the tubes were mixed rapidly and the tubes placed in a 25 r IoC constant temperature bath. After a 2-h incubation at 25"C, the tubes were cooled quickly to 4°C with an ice bath. This low temperature slows the dissociation rate constant ( k -I ) and allows for separation of bound and free counts without signifcant loss of bound counts in gel filtration medium. For 50 tubes, the bound and free hormone could be separated in 15 min. Since the dissociation rate constant ( k 1 ) at 4°C is 5.2 X rnin", there was no significant dissociation within 15 min. Bound and free hormone populations were separated by gel fdtration on a small Sephadex G-25 (medium) column (bed volume = 2.0 * 0.05 ml, equilibrated in Incubation Buffer I). The amount of Iz5I was then determined with a Searle Auto-Gamma spectrometer (efficiency = 74%). Specific binding (i.e. high affinity, limited capacity) was calculated by subtracting the counts per min for nonspecific binding from the counts per min for total binding. Free hormone was calculated by subtracting the counts per min for total binding from the counts per min for total labeled hormone added.
Competition Binding Assay-All competition studies were carried out at 25°C with nuclear extract (-200 PM binding sites) incubated for 2 h, in the presence of a constant tracer amount of ['251]T3 (onehalf of the K d at 25"C, ~0 . 4 m) and an increasing amount of unlabeled thyroid hormone analog. Nonspecific binding (usually 10% of total bound) was considered to be the amount of bound [1251]T3 following a parallel incubation which contained 103-fold excess unlabeled TB. The unlabeled analogs were weighed on a Cahn electrobalance to an accuracy of kO.01 mg and diluted to a stock concentration of 0.5 or 5 n m in 1-propanol with a few drops of 1 N HCl to help solubilization if needed. The concentration of I-propanol in the final incubation mixture never exceeded 1% which was shown to have no effect on binding activity.

Separation of Bound and Free [lZ5I]T3 by Gel Filtration-
The elution profie of the binding assay incubation mixture from Sephadex G-25 minicolumns is shown in Fig. 1. The triangles represent a normal incubation containing nuclear extract and [lZ5I]T3. The circles represent a parallel incubation which contained a 1000-fold excess of unlabeled T3. The fist peak to elute from the column was in the excluded volume (1.6 d) and represents macromolecule-bound [12511T3. The second peak at 2.5 ml represents ['251]iodide. The elution buffer must be changed to 0.25 N NaOH in order to elute the free counts per min between 5.5 ml and 8.5 ml . It can be seen that the addition of 1000-fold excess cold T3 displaces all but -10% of the macromolecule-bound [lZ5I]T3. This residual radioactivity represents nonspecific or nonsaturable binding and

Thyroid Hormone-Nuclear Receptor Interactions
was subtracted from the total bound [lZ5I]T3 to calculate the "specific" binding. Equilibrium Hormone and Analog Binding Affinity-An analysis of the binding of increasing concentrations of [1251]T3 to nuclear extract by the Scatchard method (22) indicates that the hormone binds to a single class of independent binding sites. The graph in Fig. 2 shows the average results of three separate sets of duplicate assays performed to determine the association constant K, = 1.5 X lo9 M" and the binding capacity Pr = 1.73 X 10"' M/incubation tube or 5.8 X 10"' M/mg of protein. The inset graph is a plot of specifically bound molarity versus free molarity, and it indicates that the binding reaction is saturable.
The   Table I  shown in Fig. 3. The binding affinities of 57 unlabeled analogs have been determined in this fashion and are presented in Table I. The free energy changes produced upon ligand binding are presented in the last column of Table I and can be used to determine the strength of substituent group interactions with the receptor.

Hydrophobic and Dispersion Force Interactions of 3' Sub-
stituents-The direct binding contributions of 3' substituents, separated from interactive effects on the 4'-hydroxyl group (see below), are determined by comparisons of the binding free energies of 3'-substituted 3,5-diiodo-4-phenoxyphenylalanines (Table 11). Nine substituents in three classes were studied: nitro, alkyl, and halogen. All 3' substituent groups showed negative values for their contributions to the free energy of binding (AAG), indicating an enhancement of binding relative to the unsubstituted molecule. Within the alkyl series this direct binding effect increased from -1.80 kcal/mol for methyl to -2.79 kcal/mol for isopropyl, then decreased for the more bulky t-butyl residue (-2.01 kcal/mol). Similarly, binding affinities increased with increasing size (24) and hydrophobic character (25) within the halogen series. The greater direct effect on binding affinity elicited by an iodine atom, as compared with an alkyl group of about the same size (Vu) and hydrophobic character (mz for ethyl or isopropyl), indicates that factors other than hydrophobic character are involved in the binding contributions of 3' substituents. The greater polarizability of a halogen atom relative to an alkyl group may account for this difference in binding. Previous studies involving binding to intact nuclei of rat liver (26) ascribed a size-limited hydrophobic binding role to the 3' substituent. The low contribution of the 3"nitro group to binding may be attributed to the weak hydrophobic character (~S Z = -0.28) of a nitro substituent.
The portion of the receptor that binds to the 3' substituent may be visualized as a size-limited region made up primarily    I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I   I   I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I   I   I  I  I  I  E t  Me  Me  i Pr  Et  I  I  I  I  I  H  H  I  H   R s   __   I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I   I  I  I   H  H   I  I  I  I   I   I  I  I  I  I  I

X Substituent
Group binding contribu-X AA G"  -deoxythyronine), using the data of Table I. of hydrophobic residues, with a polar residue component to account for the enhanced direct association of halogen atoms. Donor Hydrogen Bond Formation by the I'-Hydroxyl Group-The strength of the 4"OH interaction with the receptor was determined by subtracting the binding free energy of a 3"substituted 4'-deoxy-3,5-diiodothyronine from the binding free energy of a 3"substituted 4'-hydroxy-3,5-diiodothyronine and is presented in Table 111.

TABLE 111 Strength ofthe I'-hydroxyZ interaction with the soluble nuclear receptor in 3'-substituted 3,5-diiodothyronines
The addition of a 4'-hydroxyl to the unsubstituted outer ring increases the binding affinity by -1.24 kcal/mol. Within the series of 3'-alkyl thyronines, the enhancement in binding affiity due to addition of the 4"OH increases, reaching a maximum at isopropyl (-2.54 kcal/mol) then decreasing with the more bulky tertiary butyl group. The high binding affinities of these 3"alkyl thyronines, compounds in which the 4'hydroxyl group is essentially unionized at pH = 7.6, implies that the 4'-hydroxyl group acts as a proton donor in hydrogen bond formation with the receptor. A similar trend is seen in Table I11 for the 3'-halogen-substituted compounds. The enhancement in binding due to the addition of a 4'-hydroxyl increases with the size of the 3' group, reaching a maximum with bromine. For all 3"substituted compounds, except nitro, the strength of the 4'-hydroxyl interaction with the receptor is greater than for the 3'-unsubstituted compound. The positive correlation between size of the 3' substitution (except for iodine) and its enhancement of the strength of the 4'-hydroxyl interaction implies that one effect of the bulky 3' substituent is to orient the hydrogen atom of the 4'-hydroxyl away from the 3' side to a position cis to the 5' side, producing a more favorable donor hydrogen bond orientation relative to the receptor. The tertiary butyl substituent may interfere sterically with the 4'-hydroxyl receptor association. The very low BAG for a 3'-nitro substitution supports this conclusion since a nitro group forms a very strong intramolecular hydrogen bond with the 4"hydroxyl which would orient the 4'-hydroxyl away from the 5' side of the outer ring. Table I11 also indicates that for 3' substituents of similar size, the 3"halogen has a greater interactive effect on the 4'-hydroxyl than the 3'-alkyl. This is probably due to the inductive electron-withdrawing effect of the halogens which would tend to weaken the 0-H bond of the 4'-hydroxyl, making it a slightly better hydrogen bond donor. The weaker interactive effect of the 3'-iodine is compensated for by its strong direct effect in receptor binding ( Table 11).
Negative Influence of 5' Substitution-Previous studies have shown that for a given substituent, the 3'-monosubstituted compound has a higher in vivo (27,28) and in vitro (9) activity than the corresponding 3',5'-disubstituted compound. A similar relationship has been found for binding to the solubilized nuclear receptor as seen in Table IV. The change in the free energy (AAG) when an identical halogen or alkyl 5' substituent is added to a 3"substituted 3,5-diiodothyronine, is a direct measure of energy lost due to steric repulsion. Chlorine substitution in the 5' position of the 3'-chloro-3,5-diiodothyronine results in a negligible loss of binding free energy (+ 0.01 kcal/mol). However, as the size of the 5' substituent is increased, the loss due to the 5' substitution becomes greater except for 5'-bromine substitution. This is probably due to the greater polarizability of bromine which would cause less steric repulsion than the hard sphere of a 5'-methyl. Table IV also

Thyroid Hormone-Nuclear Receptor Interactions
shows that the decrease in binding affiiity due to 5"halogen substitution in 3,5-diiodo-3"isopropylthyronine is directly proportional to the size of the 5"halogen substituent. The magnitude of this effect is very close to that seen with 3',5'dihalogen substitution. This implies that the primary receptor binding association is with the 3'-isopropyl group, and that the smaller 5'-halogen substitutents interfere with binding in the 5' (proximal) position.

3,5 Substituent
Contributions to Receptor Binding-The binding free energy of 3'-iodo-3,5-dihydrothyronine, subtracted from the binding free energy of four 3'-iodo-3,5-disubstituted thyronines, is presented in Table V. This is a direct measure of the contributions of the 3,5 substituents in binding with the receptor. The data in Table V suggests that there are strict functional, conformational, and size factors for optimal 3,5 substitution. The strengths of the 3,5-dimethyl and 3,5-diethyl interactions with the receptor are drastically less than that of the 3,5-diiodo interactions, which may be due to the greater polarizability of halogen atoms. The 3,5 substitutions not only participate in hydrophobic and dipolar bonding, but they are also responsible for maintaining a perpendicular conformation of the outer ring. Dimethyl substituents are not bulky enough to constrain effectively the aromatic rings of the diphenyl ether nucleus in a mutually perpendicular conformation (29). Therefore, the interactive effect of 3,5-dimethyls on the 3"iodine is not as great as 3,5-diiodo substitution. Also, there appears to be a strict size limit for 3,5 disubstitution since the 3,5-diisopropyl interaction with the receptor is so low (-3.48 kcal/mol). The receptor binding regions for 3,5 substituents appear to be limited in size and comprised of primarily hydrophobic residues, with some polar character.
Electrostatic Interaction of the Alanine Side Chain-Based on the pK, values for tyrosine (pK,, = 2.20, pldz = 9.11) (30) the alanine side chain of 3,5,3'-triiodo-~-thyronine would be a zwitterion at pH = 7.6. Therefore, the potential for ionic bond formation exists for both the carboxylate and protonated amine portions of the amino acid. Because an analog without the alanine side chain was not available for this study, a direct measure of its contribution to the binding free energy can not be calculated. However, the propionic and acetic acid analogs, 3,5,3"triiodothyropropionic acid (triprop) and 3,5,3"triiodothyroacetic acid (triac), which contain no ammonium group, show AGO values -0.50 and -0.61 kcal/mol more negative than that of TY (-12.38 kcal/mol), respectively. The affinity of 3,5,3',5"tetraiodothyropropylamine (thyroxamine), which contains no carboxylate anion, was so low (< -6.4 kcal/mol) that it could not be measured within the solubility limits of  the compound. These results indicate that the most favorable ionic interaction of the alanine side chain is formation of electrostatic bonds between the carboxylate anion and some positively charged group on the receptor. However, the positively charged cy-NH3+ group of T3 contributes an unfavorable interaction with the receptor which results in the negligible contribution (-0.21 kcal/mol) of the alanine side chain to the overall binding free energy of T3. The relatively low affinities of the shorter side chain analog, 3,5,3'-triiodothyroformic acid (-10.91 kcal/mol), and of the longer side chain analog, 3,5,3'triiodothyrobutyric acid (-11.23 kcal/mol), indicate that the ionic binding site on the receptor has strict steric and size limitations, as has been previously seen for the rest of the molecule.

DISCUSSION
Equilibrium binding studies have been used to determine the strength and nature of the molecular interactions between thyroid hormone analogs and the solubilized rat liver nuclear receptor. If the binding free energies of specific group contributions are added (as seen in Fig. 4), then the sum of the calculated individual group contributions to the binding free energy of T3 (-12.4 (+ 1.8) kcal/mol) is not significantly different from the measured binding free energy for T s (-12.4 (+ 0.2) kcal/mol).
Using this method, the following features of the molecular interaction between thyroid hormone analogs and the solubilized rat liver nuclear receptor have been determined: 1) The 4'-hydroxyl participates in a donor hydrogen bond oriented toward the 5' side of the outer ring and adds -1.2 (& 0.2) kcal/ mol of binding free energy. 2) The 3' substituent participates directly in hydrophobic and van der Waals bonding to the receptor, with a size limit for alkyl substitution at isopropyl. The 3' substituent also strengthens the 4"hydroxyl interaction. The contribution of a 3"iodine is made up of a direct binding effect, -3.55 (-+ 0.12) kcal/mol, plus an indirect enhancement of the 4'-hydroxyl binding of -0.59 (+ 0.47) kcal/ mol. 3) The optimal 3,5 substituents are iodine atoms which can contribute an average of -3.4 (+ 0.7) kcal/mol/iodine. This value contains the interactive effect on orientation of the outer ring, as well as the direct contribution to binding by the 3,5-iodine atoms and the aromatic rings. Iodine atoms in the 3,s positions contribute significantly more to receptor binding than do alkyl groups of the same size and hydrophobic character. 4) The alanine side chain probably does participate in an electrostatic attraction between the carboxylate anion and a positively charged amino acid side chain in the receptor but due to the unfavorable interaction of the a-ammonio group adds a negligible -0.2 (i 0.8) kcal/mol to the binding free energy.
The magnitude of the binding affinity of T a or of T q to the nuclear receptor is the result of the summation of multiple individual group contributions from the entire hormone molecule. In the analogs studied to date, binding affinity to the nuclear receptor is directly related to the expression of hormonal activity in uiuo when factors of distribution and metabolism are considered. Therefore, it follows that all parts of the hormone molecule are involved both in binding and in the initiation of the event, possibly a conformational change in the receptor, which appears to regulate specific gene function and protein synthesis. The precise structural and stereochemical requirements for maximal hormonal activity suggest that the binding site is a deep cleft in which all portions of the TS molecule can interact with relatively inflexible binding domains in the protein.