Detailed analyses of thyroid response elements (TREs) have led to the identification of a canonical TRE half-site sequence (2,16, 85) (Figure 3d-4). The TRE half-site is generally considered to be a hexamer (AGGTCA), but TR binding is optimal with a more extended binding site (86,87). Specifically, the sequence TAAGGTCA is optimal for TR binding and T3-responsiveness. However, inspection of TREs from many different target genes reveals there is a relatively low degree of sequence conservation among these elements. This finding suggests the possibility that naturally-occurring TREs may have diverged from an ideal consensus element during evolution as a means to modulate the degree of TH responsiveness.

TR interactions with DNA are quite different from those observed with steroid receptors, which bind to palindromic DNA sequences as homodimers. Although TR also can bind to certain TREs as a homodimer, it binds preferentially to most TREs as a heterodimer with the retinoid X receptors (RXRs) (2,16,31,32). The TR-RXR heterodimer binds to half-sites that are arranged in several different configurations. These include palindromic arrangements (head-to-head), direct repeats (head-to-tail), and inverted repeats (tail-to-tail). Most naturally occurring TREs are direct repeats (Figure 3d-4), typically separated by four nucleotides. The ability of TR dimers to bind to TRE’s in different configurations suggests a flexible protein structure, or the possibility that distinct protein surfaces are involved in the formation of dimers (32, 86-89). Taken together, the specificity and affinity for the TR-RXR heterodimer is primarily determined by sequences within the half-site, the length of the spacer region between the half sites, and the sequence context within the spacer region.

Although TR can interact with a wide variety of other nuclear receptors and transcriptional adaptor proteins (see below), the RXR proteins (α , β , and ) represent its most important heterodimeric partners (90). The RXR proteins enhance TR binding to DNA and reduce the rate of receptor dissociation from DNA (91) . RXR binds to the 5’ sequence and TR binds to the 3’ sequence of TREs in which half-sites are arranged as direct repeats (93, 94). The DNA-binding domains interact with the major grooves of the half-sites on the same face of the DNA (39). The carboxy-terminal end of the TR DNA-binding domain forms an α -helical structure that interacts with the spacer region in the DNA minor groove between the TRE half-sites. Although protein-protein contacts between the RXR and TR DNA-binding domains are important for dimerization, the major sub-regions involved in dimerization reside in the carboxy-terminii of the receptors (31,32). The dimerization surface of the TR appears to involve residues that lie along the surfaces of helices 10 and 11. T3 binding enhances the formation of TR-RXR heterodimers (95). On the other hand, T3 dissociates TR-TR homodimers (96). These findings raise the possibility that T3 binding might induce disruption of TR homodimers and induce the formation of TR-RXR heterodimers. The RXRs bind a stereoisomer of all trans retinoic acid, 9-cis retinoic acid (97,98). which variably alters transcriptional activity depending on the nature of the TH responsive gene (2, 16,99). Additional studies are required to clarify the functional roles of RXRs and their ligands in TH action.

After binding to DNA, TR alters transcriptional activity by interacting directly or indirectly with a complex array of transcriptional cofactors. These proteins include corepressors (CoRs), coactivators (CoAs), integrators like CREB-binding protein (CBP), and general transcription factors (GTFs) (reviewed in 2, 16,31,32,99, 100). Many of these factors have been identified by protein-protein interaction assays such as the yeast two-hybrid and glutathione-s-transferase pull down assays.

In the absence of TH, TR represses basal transcription in proportion to the amount of receptor and the affinity of receptor binding sites in positively-regulated target genes (101). This phenomenon also is referred to as transcriptional silencing (102, 103). (Figure 3d-5, below). The addition of TH reverses basal repression and increases transcriptional activation above basal levels seen in the absence of receptor. Our understanding of the molecular mechanism for basal repression of transcription by unliganded receptor was advanced significantly by the discovery of a family of repressor proteins that bind selectively to unliganded TRs and RARs. This corepressor family includes silencing mediator for retinoid and TH receptors (SMRT) and nuclear receptor corepressor (NCoR) (32, 33, 104,105). These corepressors are 270 kD proteins that contain three transferable repression domains and two carboxy-terminal α -helical interaction domains. They are able to mediate basal repression by TR and RAR, as well as orphan members of the nuclear hormone receptor family such as rev-erbAα and chicken ovalbumin upstream transcription factor (COUP-TF). They have little or no interaction with steroid hormone receptors and therefore do not mediate basal repression by these receptors. Another protein, small ubiquitous nuclear co-repressor (SUN-CoR) enhances basal repression by TR and rev-erbA (65). This 16kD protein may form part of a co-repressor complex as it interacts with NCoR.

Within the interaction domains of NCoR and SMRT are consensus LXXI/HIXXXI/L sequences which resemble the LXXLL sequences that enable co-activators to interact with nuclear hormone receptors (40-43)(see below). Interestingly, these motifs allow both corepressors and co-activators to interact with similar amino acid residues on helices 3, 5, and 6 which are part of the ligand-binding pocket of TR. Differences in the length and specific sequences of the co-repressor and co-activator interaction sites coupled with the conformational changes in the LBD upon ligand binding, determine whether corepressor or coactivator binds to TR (32, 42).

Recently, it has been shown that corepressors can form a complex with other repressors such as Sin 3 and histone deacetylases that are mammalian homologs of well-characterized yeast transcriptional repressors RPD1 and RPD3 (2,16,32,99,100). Thus, local histone deacetylation likely plays a critical role in the basal repression by unliganded TR/corepressor complex by maintaining local chromatin structure in a state that decreases basal transcription. Upon T3 binding, TR undergoes a conformational change that dissociates CoRs and recruits an array of coactivators (CoAs). Thus, hormone binding relieves repression and stimulates transcription by altering receptor binding to distinct classes of cofactors. Additionally, DNA-methylation may play a role in basal repression as methyl-CpG-binding proteins can associate with a co-repressor complex containing Sin3 and histone deacetylases (106,107). This repression was relieved by the deacetylase inhibitor, trichostatin A. These findings suggest that two repression processes, DNA methylation and histone deacetylation, may be linked via methyl-CpG-binding proteins.

The fact that TR alters the level of gene transcription in both the absence and presence of T3 has important implications for TH physiology. At low hormone concentrations, such as hypothyroidism, the unliganded receptor is predicted to repress expression rather than exist as an inactive, passive receptor. In some respects, this model is borne out by targeted inactivation of the TRα and TRβ genes. The phenotype of these double knockout mice are, for the most part, much less pronounced than the clinical features of congenital hypothyroidism (108,109). Thus, basal repression of transcription may explain why absence of receptor has less deleterious effects than absence of hormone (82,108,109).