Introduction
The thyroid hormones (thyroxine or 3,3',5,5'-tetraiodo-L-thyronine [T4], and its metabolites) are synthesized and secreted from the thyroid gland and circulate throughout the body bound to transport proteins (transthyretin TTR; thyroxine-binding globulin TBG; serum albumin). They are also protein-bound when transported through the cell membrane and into the nucleus. Studies of thyroid hormones provide insight into the molecular events that control their biosynthesis, transport, and mechanism of action. Therefore, their molecular interactions with proteins are of paramount interest. Currently, structural data have elucidated thyroid hormone-protein interactions for the ligand-binding domain of nuclear thyroid hormone receptors, and the serum transport proteins, TTR and serum albumin. Structure-activity relationships among the metabolites of T4 and its deiodination products have shown that specific stereochemical features of the thyronine nucleus are crucial in determining protein binding specificity and binding affinity (Table) (1).

Thyroid hormone structure and stereochemical characteristics

Structural data for the thyroid hormones showed that the minimal steric interaction between the tyrosyl ring 3,5 iodine atoms and the phenolic ring 2',6' hydrogen atoms is maintained when one ring is coplanar with, and the other perpendicular to, the plane of the two ether bonds (2,3). These observations showed the hormone to have a skewed conformation and the concept of preferred, if not somewhat rigid, orientations of the molecule.Thus, the 3'-iodine of the hormone 3,3',5-triiodothyronine (T3) can exist in two positional isomers - distal and proximal - depending on whether the phenolic ring iodine atom is oriented away from, or near, the tyrosyl ring (4). Activity measurements of rigid analogues revealed that a distal T3 conformation was the more active analogue (1).

Crystallographic analysis of thyroid hormones showed that a skewed diphenyl ether conformation is observed in the structures of all 3,5-disubstituted hormone analogues (2-4). As mentioned, the bulk of the tyrosyl ring substituents forces the diphenyl ether to adopt a skewed conformation, whereas removal of one of these substituents releases this constraint, permitting an antiskewed conformation (Fig. 1), as observed in the crystal structure of 3,3',5'-triiodothyronine (rT3) (5).

Fig. 1. Two examples of the skewed and antiskewed diphenyl ether conformations of T4. In each case the molecule is viewed parallel and perpendicular to the tyrosyl ring plane (2).

T4 cannot adopt an antiskewed conformation because the bulk of the tyrosyl ring iodine atoms would be placed into the electron density of the phenolic ring. The conformational space available to the thyroid hormone side chain is defined by three torsion angles from which only specific subsets are predicted to be favored energetically, and most of these have been observed (2,4,6). As a result of these stereochemical features, the overall hormone conformation can be either cisoid or transoid, depending on whether the phenolic ring and side chain are on the same or opposite side of the plane of the tyrosyl ring (2). These features play a role in differentiating the binding affinity of hormone analogues for their various hormone-binding proteins. For example, T4 has the strongest binding affinity for thyroxine-binding globulin, whereas the metabolite tetraiodothyroacetic acid (T4Ac) has the strongest binding affinity for TTR (Table 1).

Protein binding interactions

TTR

Transthyretin is responsible for the serum transport of thyroid hormone and for the binding of retinol-binding protein that binds vitamin A. TTR is a 127-amino acid monomer that forms a homotetrameric structure that has two hormone-binding domains that are partially occupied under normal physiological conditions (7,8). Structure-activity data show that thyroid hormone analogues have different binding affinities for TTR, depending on their substituent patterns (Table 1). In addition, many pharmacologic agents and natural products, such as plant flavonoids, non-steroidal analgesic drugs, inotropic bipyridines and organohalogen environmental agents, are strong competitors for T4 binding to TTR with binding affinities much greater than T4 (9). Studies of T4 displacement from TTR revealed that a synthetic plant flavonoid, EMD 21388 (3-methyl-4',6-dihydroxy-3',5'-dibromo-flavone), is the strongest competitor for T4 binding to human TTR (10,11) and showed that this T4 antagonist alters the circulating total and percentage of free thyroid hormones and serum thyrotropin concentrations (12).

To understand the structure-activity correlations observed for thyromimetic ligand binding, a number of ligand complexes with human and rat TTR have been studied (13-28). Most crystals of human TTR ligand complexes are isomorphous with the orthorhombic lattice reported for the native structure (29,30) and has two independent monomers in the asymmetric crystal lattice. The presence of the crystallographic two-fold axis through the center of the binding domains requires that the ligand either possess two-fold symmetry, or occupy two statistically disordered positions within the binding binding sites. Because T4 does not have such symmetry, it must bind with a statistical disorder. On the other hand, crystals of rat TTR crystallize such that there is a unique tetramer in the asymmetric tetragonal crystal lattice and these data provided the first observation of the unique binding environment for thyroid hormone (31). When structures of human TTR-T4 complexes were crystallized in space groups that did not require the crystal symmetry, binding occurred with the ligand occupying multiple positions within the hormone binding domain (18). These data further revealed that there was significant flexibility in the binding preferences for thyroid hormones.

Ligand binding modes for TTR

Structural data for the human TTR-T4 complex (30) revealed that T4 binds in a "forward" mode, with its 4'-OH buried deep within the channel running through the tetrameric protein and has its amino acid side-chain near the channel entrance interacting with Lys-15 and Glu-54. Recently, the TTR-T4 complex crystal structure was determined as a co-crystallized hormone complex (17). These data showed that T4 binds deeper in the channel and displaces the bound water observed in the crystals soaked with T4 (30). Although the orientation is similar, the hormone is rotated such that it shares common binding sites for the 3- and 3'-iodine atoms. These data verify that T4 binding does not affect the main chain conformation significantly but results in local rearrangements of residue side-chains in the binding channel.

The orientation of the weak binding metabolite 3,3'-T2 differs significantly from that of T4 as it binds deeper in the channel than T4, and in this orientation, 3,3'-T2 occupies the binding domain in a completely different manner from T4 (15). The binding affinity of 3,3'-T2, which is 100-fold lower than that of T4, reflects the lack of the second pair of iodine atoms interacting in the channel. The thyroid hormone metabolite T4Ac (Table 1), a tight binder of TTR, shows multiple binding modes in structures with human and rat TTR (13,20). These data reveal the hormone metabolite binds with a mixed population of both "forward" and "reverse" orientations in one binding domain, and two, alternate "forward" binding positions in the second domain (Fig. 2).

FIG. 2. Position of T4Ac (red) in forward binding mode in domains A and B of rat TTR complex (20). Also shown is the reverse binding mode of T4Ac (green) in the same complex.

The added intermolecular interactions of T4Ac, as reflected in the multiple binding contacts that result from the partial occupancy of two different orientations, may be indicative of the tight binding affinity observed for this metabolite for TTR.

A comparison of the structures of TTR ligand complexes reveals differences in the channel geometry and in the diameter of the two hormone binding domains of TTR. The analysis of the ligand occupancy permits the identification of the primary site that binds ligand with higher affinity (22). The ligand-induced changes result in a lower binding affinity of ligand to the second binding domain. These data suggest a mechanism to explain the negative cooperativity observed in the binding of hormone to the second binding domain (7,8).

Structural results show that flavones bind to transthyretin in a manner different from T4. Data for the structure of the hTTR-EMD21388 complex revealed that bromoflavone binds deeper in the channel than T4; the bromine atoms occupy symmetric sites in a "forward" mode (that is positioned near the channel center at the tetramer interface) and in a "reverse" mode, with the bromophenolic ring near the channel entrance in TTR (13,21). In this binding, there is partial occupancy of each molecule that bind in opposite directions along the binding domain channel axis. A bromoaurone analogue binds in a similar manner (14). The observation of two alternative binding orientations for EMD 21388 may explain its greater binding affinity for TTR (11). Similar results have been observed for the organohalogens, tri- and penta-bromophenol that were observed to bind exclusively in the "reverse" mode (24).

TTR variants and amyloidic disease

Increased efforts have been made to understand the mechanisms underlying TTR tetramer stability and its relationship to the formation of amyloid fibrils that characterize amyloid diseases including familial amyloid polyneuropathy (FAP), senile systemic amyloidosis, and Alzheimer's disease (32). Recent data show that wild-type and variant TTR form amyloid fibrils that are the causative agents in familial amyloid polyneuropathy (FAP) and senile systemic amyloidosis diseases (33-44). To date, more than 70 single amino acid variants of the 127 residue monomer of TTR have been implicated in FAP disease (33); however, structural studies of TTR variants have failed to identify major differences that could explain their amyloidogenicity (35-38). Several hypotheses have been proposed based on monomeric or dimeric amyloidogenic intermediates to explain fibril formation from TTR monomers. One model proposes head-to-tail polymerization of monomeric intermediates (33,42). Another model is based on the formation of linear aggregates of TTR molecules, each linked by a pair of disulfide bonds involving Cys-10 (35,37,44). In this case, the intermediate is a dimer. A third model is based on data from two engineered amyloidogenic mutants and requires dimers that are associated by antiparallel organization of the F and H strands of the native protein. This model requires the destabilization of the tetramer prior to fibril formation (33,34). Yet another model invokes proteolytic cleavage as the initiation step in fibril formation (34,41).

Although the mechanism of tetramer stabilization is still unclear, it has been shown that ligand binding stabilizes tetramer formation in all variant species (33,39). Therefore, one means of intervention for disease treatment could involve binding of nonthyromimetic analogues that can stabilize the TTR tetramer and possibly delay the onset of fibril formation. To this end, numerous compounds have been screened (27,28,39). Structural data for the TTR complex with flufenemic acid show that the ligand mediates intersubunit hydrophobic interactions and hydrogen bonds that stabilize the normal tetrameric fold (39). Data for the TTR complex with another analgesic analogue, VCP-6, that has 6 times the affinity of T4 (45) show that the molecule forms strong hydrogen bond interactions of its 2-carboxylate with Lys-15 and with the 3,5-dichloro atoms in symmetric hydrophobic pockets near the channel center at the tetramer interface (25,26). These results suggest that the strategy of stabilization by strong competitors may prove fruitful. Comparisons of the environment near the 22 residues in the rat sequence that differ from those in human TTR also permit evaluation of their influence on hormone-binding interactions and tetramer assembly. There are small differences in the local environment of the changed residues that affect ligand binding interactions. Understanding the influence of these residue changes upon binding affinity and tetramer stability may help understand the significance of point mutations in human TTR that result in amyloid fibril formation.)

Human Serum Albumin

Human serum albumin (HSA) is the major protein component of blood plasma and serves as a transporter for T4 and other hydrophobic compounds, including fatty acids (46). Structural data show that HSA is a globular protein of 585 amino acids that is composed of three similar domains, each of which contains two subdomains (47). Structural data for the T4-HSA complex reveals that there are four T4 binding sites in HSA located in subdomains IIA, IIIA, and with two adjacent sites in IIIB (48). These data further reveal that the binding orientation of thyroxine differs in these sites with the hormone binding with a cisoid conformation in some sites and transoid in others. In addition, it was shown that myristate competes for the hormone binding sites, as the fatty acid binding sites partially overlap those of the hormone. Comparison of the crystal structures of HSA crystallized in the presence of both myristate and T4 reveals the presence of a fifth hormone binding site that is formed as a result of fatty acid-induced conformational change.

Naturally occurring mutants R218H or R218P of HSA can give rise to enhanced T4 binding affinity and are the cause of autosomal-dominant condition of familial dysalbuminemic hyperthyroxinemia (FDH) (46). Structural analyses were carried out on the R218H and R218P mutations in complex with T4 (48). The residue R218 is located in the center of a helix near the entrance to the hormone binding site in subdomain IIA that shows conformational flexibility caused by disorder in the local environment give rise to a mixture of protein conformational states that results in a partial occupancy of the hormone binding site. These mutant albumin structures reveal that T4 binds with 100% occupancy in contrast to the 50% occupancy observed for the wild-type protein. The replacement of arginine with small side chains, coupled with movement of surrounding residues, removes the steric strain present in the native protein complex.

Nuclear Receptor Binding

Biochemical data have shown that the site of action of T3 is the nuclear receptor which is a member of the large nuclear receptor superfamily that contains a central conserved DNA-binding domain, a less conserved carboxy terminal ligand binding domain, and an amino-terminal domain that varies in size and composition among family members (46). Structural data for the thyroid hormone ligand binding domain reveal a largely helical structure with a unique internal-binding mode for T3 that is completely buried in a hydrophobic core (49). Two different thyroid hormone receptor genes (TR and TR) have been identified that show differential hormone binding characteristics for these subtypes of receptor. The TR1 product represents a functional receptor that responds to thyroid hormone, whereas the TR2 isoform does not bind thyroid hormone but can antagonize thyroid hormone action. The TR1 and TR2 isoforms differ in their amino termini, but both can bind and respond to thyroid hormone (50).

Structural data have been reported for rat TR ligand binding domain in complex with the non-iodinated thryomimetic, 3,5-dimethyl-3'-isopropyl thyronine (DIMIT) (49), as well as human TR ligand-binding domain in complex with T3 acetic acid and GC-1 analogue (50). These results identify differences in the ligand-binding pocket between the a and b subtypes involving the single amino acid difference (Ser-277 in TR or Asn-331 in TR) that could be used in pharmaceutical design. More recent efforts have extended this concept to the design of thyromimetics with tissue-selective actions. For example, thyromimetics have been synthesized that are selective for TR(51) while other analogues elicit a T3 response from mutant forms of the thyroid receptor (TR R320C) (52). These analogues suggest that such "pharmacological rescue" agents have potential for the treatment of thyroid disease.

Acknowledgements
Supported in part by TW00225.

REFERENCES