Thyroxine-Thyroid Hormone Receptor Interactions*

Thyroid hormone (TH) actions are mediated by nuclear receptors (TRs α and β) that bind triiodothyronine (T3, 3,5,3′-triiodo-l-thyronine) with high affinity, and its precursor thyroxine (T4, 3,5,3′,5′-tetraiodo-l-thyronine) with lower affinity. T4 contains a bulky 5′ iodine group absent from T3. Because T3 is buried in the core of the ligand binding domain (LBD), we have predicted that TH analogues with 5′ substituents should fit poorly into the ligand binding pocket and perhaps behave as antagonists. We therefore examined how T4 affects TR activity and conformation. We obtained several lines of evidence (ligand dissociation kinetics, migration on hydrophobic interaction columns, and non-denaturing gels) that TR-T4 complexes adopt a conformation that differs from TR-T3 complexes in solution. Nonetheless, T4 behaves as an agonist in vitro (in effects on coregulator and DNA binding) and in cells, when conversion to T3 does not contribute to agonist activity. We determined x-ray crystal structures of the TRβ LBD in complex with T3 and T4 at 2.5-Å and 3.1-Å resolution. Comparison of the structures reveals that TRβ accommodates T4 through subtle alterations in the loop connecting helices 11 and 12 and amino acid side chains in the pocket, which, together, enlarge a niche that permits helix 12 to pack over the 5′ iodine and complete the coactivator binding surface. While T3 is the major active TH, our results suggest that T4 could activate nuclear TRs at appropriate concentrations. The ability of TR to adapt to the 5′ extension should be considered in TR ligand design.

Thyroid hormone (TH) 1 plays important regulatory roles in metabolism, homeostasis, and development by binding and altering the transcriptional regulatory properties of two related nuclear receptors (NRs), the thyroid hormone receptors (TRs) ␣ and ␤ (1,2). Most TH produced in the thyroid gland is secreted in the form of thyroxine (T 4 ; 3,5,3Ј,5Ј-tetraiodo-L-thyronine) (2,3). The thyroid gland also produces smaller amounts of triiodothyronine (T 3 ; 3,5,3Ј-triiodo-L-thyronine) and reverse T 3 (rT 3 ; 3,3Ј,5Ј-triiodo-L-thyronine), and 80% of T 4 is converted to T 3 and rT 3 in peripheral tissues by two selenium deiodinases, which are tissue-specific (4). Current beliefs are that T 3 is the dominant active form of TH; T 3 binds the TRs with an affinity about 20 -30 times higher than that of T 4 (5)(6)(7)(8)(9), and some studies suggest that T 3 is present at higher concentrations in the nucleus than T 4 (10,11). Nonetheless, the question of whether T 4 is simply a prohormone or an active TH species is not completely resolved. T 4 exerts rapid nongenomic effects at several loci distinct from TRs (12). Moreover, saturating levels of T 4 activate transcription of TH-responsive genes in cell culture (see for example Ref. 5). Whereas it is possible that at least some of this activity is due to T 3 generated from T 4 in the cell, these results suggest that T 4 may act as a TR agonist. Normal concentrations of plasma-free T 4 are about 4 -6-fold higher than those of T 3 (19 pmol/liter of T 4 versus 4.3 pmol/liter T 3 ) and intracellular T 4 and T 3 levels can differ because of variations in uptake and T 4 to T 3 conversion (3); thus, it is conceivable that intracellular T 4 in some context could occupy a significant fraction of nuclear TRs.
If T 4 does behave as an agonist, then it should bind to TR in a similar way to T 3 and induce conformational changes in the TR similar to those induced by T 3 (13,14). T 3 interacts with the TR ligand binding domain (LBD), located in the receptor C terminus. The x-ray crystal structure of TR␣ or TR␤ complexed with T 3 reveals that hormone is completely enclosed in a ligand binding pocket within the core of the LBD. It is thought that the enclosure is due to ligand-induced packing of the LBD C-terminal helix 12 (H12) against the LBD; a rearrangement that also disrupts the corepressor binding surface and completes the coactivator binding surface, leading to exchange of coregulators and influence on gene expression in vivo (15).
Unlike T 3 , T 4 possesses a bulky iodine substituent at the 5Ј-position of the first thyronine ring. X-ray crystal structures have been determined for TR-LBDs complexed with several different high affinity agonists, including T 3 , Dimit (3, 5-dimethyl-3Јisopropyl-L-thyronine), and the TR␤-specific ligands GC-1 (3,5 dimethyl-4-(4-hydroxy-3Ј-isopropylbenzyl)-phenoxy acetic acid), and KB141 (3,5-dichloro-4-[(4-hydroxy-3-isopropylphenoxy)phenyl] acetic acid) (16 -18). In each of these cases, the agonist contains a 5Ј hydrogen group that lies close to the inner surface of H12. We therefore predicted that compounds with bulky side groups would perturb the folding of H12 against the body of the LBD and exploited this feature to create TR antagonists based on the notion that 5Ј extensions would preclude appropriate H12 packing and coactivator binding (19 -24). For example, addition of a 5Ј isopropyl group, similar in size to an iodine group, to the agonist MIBRT (3,5-dibromo-4-(3Јisopropyl-4Ј-hydroxyphenoxy)benzoic acid) creates the TR antagonist DIBRT (3,5-dibromo-4-(3Ј,5Ј-diisopropyl-4Ј-hydroxyphenoxy)benzoic acid) (22). Thus, it is conceivable that T 4 , with a 5Ј iodine extension, could even behave as an antagonist in some settings. Improved understanding of the way that the TRs adapt to the 5Ј iodine group will be therefore important for understanding T 4 action and key principles of NR antagonist design.
In this study, we examine T 4 interactions with TR, the way that T 4 influences TR activity in vitro and in cells in culture and determined the x-ray crystal structure of TR in complex with T 4 . We find that the TR-T 4 complex is less stable than the TR-T 3 complex, and that T 4 -liganded TRs exhibit properties that are similar to unliganded TRs in solution. Nonetheless, T 4 behaves as an agonist in cell-free assays and transfected cells. The x-ray structure of the TR LBD-T 4 complex reveals that a previously undetected niche in the ligand binding pocket widens, relative to the size of the pocket observed in the TR␤-T 3 complex, to accommodate the 5Ј iodine, permitting H12 to pack against the LBD surface in the presence of the larger ligand. Thus, the enclosed TR hormone binding pocket accommodates T 4 without complete disruption of overall TR-LBD structure. These results suggest that T 4 will act largely as a TR agonist if present at high enough concentrations in the nucleus.

MATERIALS AND METHODS
Thyroid Hormone Binding Assay-Thyroid hormone binding and analog competition assays were performed as previously described (9). K d values were calculated by fitting saturation curves and competition data to the equations of Swillens (25) using the GraphPad Prism program (GraphPad Software, San Diego, CA).
Hydrophobic Interaction Chromatography of TR-TR-LBDs were expressed in E. coli and partially purified on phenyl-Toyopearl, TSK-DEAE, TSK-heparin, and TSK-phenyl columns without TH as described previously (9). For each analog tested, TR was incubated for 1 h with a 5-fold molar excess of the analog relative to the final TR concentration, ammonium sulfate concentration was adjusted to 0.7 M, and the sample was loaded and chromatographed on a 0.8 ϫ 7.5 cm TSK-phenyl column (Tosoh Biosep, Montgomeryville, PA) at 0.75 ml/min with a 60-min gradient from 0.7 M ammonium sulfate, no glycerol to no salt, 20% glycerol. TR levels were assessed by measuring absorption at 280 nM and, where appropriate, radiolabeled T 4 in complex with the TR was detected by scintillation.
Reporter Cells-The assay procedure, described previously (22), utilized Chinese hamster ovary cells stably expressing TR␣1 and TR␤1 containing a stably integrated reporter gene with a single TRE (DR-4) cloned into the position of the mouse mammary tumor virus promoter hormone response element, driving expression of alkaline phosphatase coding sequences.
Crystallization and Data Collection-The TR␤ LBD was purified for crystallization trials using cobalt affinity and hydrophobic interaction chromatography first without and then with hormone as previously described (18). Crystals of the T 3 complex were obtained by the hanging drop method, with a 10.5 mg/ml protein stock solution and mother liquor consisting of 100 mM sodium cacodylate and 900 mM sodium acetate, pH 7.2. Crystals were cryoprotected by immersion in sequential baths of 100 mM sodium cacodylate and 1.1 M sodium acetate, pH 7.2, with 3, 8, and 15% glycerol. Crystals were subjected to a final swipe through a bath with 25% glycerol before flash-freezing in liquid nitrogen.
Crystals of the T 4 complex were obtained similarly, with a crystallization mother liquor of 100 mM sodium cacodylate and 700 mM sodium acetate, pH 7.4. Use of extremely fresh protein and microseeding with Triac-hTR␤ LBD crystals (Ͻ 0.1% of the final crystal) were found to be essential to obtaining diffraction quality crystals. Crystals were then cryoprotected using sequential glycerol baths as for the T 3 complex, but with a mother liquor of 100 mM sodium cacodylate and 900 mM sodium acetate, pH 7.4. Crystals were analyzed at the Advanced Light Source synchrotron facility, beamline 5.02. For the T 3 complex, 135°of data were collected with 1.5°oscillations; for the T 4 complex, 73°of data were collected with 0.5°oscillations. Reflections were indexed in DENZO and scaled in SCALEPACK.
Structural Refinement of the T 3 Complex-A molecular replacement solution was found using EPMR, employing the wild-type TR␤/TRIAC structure with hormone omitted as a probe. The structure was then subjected to multiple rounds of simulated annealing, followed by positional and B-factor refinement in CNS. Occupancies were refined in CNS for arsenic atoms only. Refinement was then continued using REFMAC of the CCP4 suite. In the final stages of refinement, water molecules were added to the structure both manually and in ARP/ wARP. Refinement steps alternated with manual rebuilding steps in Quanta98 and O, guided by F o Ϫ F c and 2F o Ϫ F c maps calculated using FFT of the CCP4 suite, and a simulated annealing omit map calculated using CNS.
Structural Refinement of the T 4 Complex-The T 4 data set was subjected to molecular replacement in EPMR and simulated annealing and positional, B-factor, and occupancy refinement in CNS as for the T 3 complex. In light of the markedly higher resolution of the data for the T 3 complex, the refined T 3 structure was then least squares fitted in O to the model of the T 4 complex. A composite model consisting of protein from the fitted T 3 complex and ligand from the T 4 complex was created and used for subsequent refinement against the T 4 complex data.
To guard against model bias, this composite model was subjected to simulated annealing and positional refinement in CNS. However, this treatment raised R free markedly, as compared with R free of the composite after positional refinement only. This indicates that the true structure of the T 4 complex is close to the T 3 complex model used to create the composite. Subsequent refinement and model rebuilding were carried out as described above for the T 3 complex.

The TR-T 4 Complex Is Less Stable Than the TR-T 3 Complex and Adopts a Different Conformation in Solution-
We first investigated interactions of TRs with T 3 and T 4 (Fig. 1A). We previously determined that T 3 has an affinity for TRs (K d ϭ 0.06 nM) about 30-fold higher than T 4 (K d ϭ 2 nM) (6). T 4 also dissociates from TRs more rapidly than T 3 (Fig. 1, B and C). Whereas it took 8.4 and 6.2 h for half the T 3 to dissociate from in vitro translated preparations of TR␣ and TR␤ at 4°C, it only took about 0.15 h and 0.06 h for half of the T 4 to dissociate from TR␣ and TR␤, respectively. Thus, TRs form a complex with T 4 that is significantly less stable than the TR-T 3 complex.
We next examined elution of T 4 -liganded TRs from TSK-phenyl hydrophobic interaction columns (HIC), an assay that detects ligand-dependent conformational alterations in TR-LBDs and provides a crude index of hydrophobicity (9). In accordance with previous results (9), liganded TR␣-LBDs eluted ahead of unliganded TRs ( Fig. 2A). Moreover, TR preparations in complex with several agonists T 3 , Triac (3,3Ј,5Ј-triiodothyroacetic acid), 3Ј-IpT 2 (3Ј-isopropyl-3,5diiodo-L-thyronine), and Dimit eluted nearly together, with elution order paralleling relative affinities of ligand for TR (T 3 ϭ Triac Ͼ IpT 2 Ͼ Dimit). By contrast, the TR␣-T 4 complex eluted closer to unliganded TR. Similar results were also obtained with the TR␤ LBD; the TR␤-T 4 complex eluted from the column between the TR␤-T 3 complex and unliganded TR␤ (Fig. 2B). It is unlikely that the unusual elution profile of the TR-T 4 complex is related to the low affinity of T 4 for TRs; the TR␣-Dimit complex eluted at a similar position to other TR agonist complexes even though Dimit binds TR␣ with an affinity five times lower than T 4 (K d ϭ 9 nM for Dimit versus 2 nM for T 4 ). It is also unlikely that the unusual elution profile is related to rapid T 4 release during passage over HIC. Continuous dissociation of T 4 would lead to a broad curve and not a discrete symmetric peak as observed here, although the "shoulder" observed with the TR␤-T 4 complex may reflect T 4 dissociation (Fig. 2B). Nonetheless, we directly examined migration of [ 125 I]T 4 prebound to TR␣ on HIC (Fig. 2C). Radiolabeled T 4 migrated at the same position as the TR␣-T 4 complex, whereas free T 4 did not elute from the column in these timescales (not shown), confirming that TR remains bound to T 4 as it passes over HIC. Thus, the unique elution profile of the TR-T 4 complex reflects an unusual conformation that exposes more hydrophobic surface than TRs in complex with T 3 or other analogues.
Modulator Binding Properties of TR-T 4 Complexes Resemble Those of TR-T 3 Complexes-We next determined whether T 4 behaved as an agonist under cell-free conditions. Fig. 3A shows that T 4 and T 3 promoted equivalent levels of TR binding to bacterially expressed nuclear receptor interacting regions of the coactivators GRIP1 and TRAP220. T 4 and T 3 also showed comparable activity in promoting binding of radiolabeled fulllength GRIP1 and TRAP220 to bacterially expressed TR␤-LBD (not shown). Finally, both ligands promoted TR release from bacterially expressed preparations of the receptorinteracting region (C terminus) of the corepressor, N-CoR (Fig. 3B). Thus, T 4 and T 3 behave as agonists in cell-free conditions.
TR agonists promote near complete dissociation of TR dimers, but not RXR-TR heterodimers, from DNA response elements (TREs) containing half-sites aligned as inverted palindrome (F2/IP-6) or direct repeats (DR-4) (29,30). T 4 and T 3 both promoted TR␣ homodimer release from an F2/IP-6 element and enhanced interactions of TR monomers with the same element (Fig. 4). Both forms of TH only modestly reduced RXR-TR␣ heterodimer binding in the same conditions. Similar results were also obtained using DR-4, and with TR␤ and both TREs (not shown). Nonetheless, TR migration was slower in the presence of T 4 than T 3 (this was most evident for the monomer). Thus, T 4 resembles T 3 in terms of regulation of DNA binding activity, but TR-T 4 and TR-T 3 complexes exhibit different mobilities, underscoring the idea that TR-T 4 complexes adopt a different structure from TR-T 3 complexes in solution (see Fig. 2).
T 4 Behaves as a TR Agonist in Cell Culture-We next examined the behavior of T 4 in cell culture. T 4 elicited a similar maximal response to T 3 in cultured chinese hamster ovary cells that were stably transfected with a TH-regulated reporter gene and a vector that expresses either TR␣ (TRAF␣ cells, Fig. 5A) or TR␤ (TRAF␤ cells, Fig. 5B) (22,32). In both cases T 4 exhibited a potency that is about 10% that of T 3 . It is unlikely that T 4 to T 3 conversion accounts for the activity of administered T 4 in these conditions for several reasons. Treatment of TRAF cells with optimal doses of the deiodinase inhibitors iopanoic acid or propylthiouracil (PTU) did not alter the T 3 or T 4 dose response, and we did not detect significant T 4 to T 3 conversion assayed by high performance liquid chromatography of TRAF cell extracts (not shown). Moreover, spiking T 4 preparations with T 3 (to account for 10% of total TH on a molar basis) yielded a leftward shift in EC 50 relative to T 4 alone (Fig. 5C), suggesting that T 4 to T 3 conversion of 10% (and probably lower) can be detected in this system. Thus, T 4 behaves as a full agonist in cells in conditions in which conversion to T 3 is unlikely to contribute significantly to agonist activity.
The X-ray Crystal Structure of the TR␤-T 4 Complex Reveals That the LBD Adapts to the 5Ј-Iodine of T 4 and Completes the Coactivator Binding Surface-To determine how TRs accommodate T 4 within the ligand binding pocket and perceive the compound as an agonist despite the presence of the bulky 5Ј iodine group, we obtained structures of the TR␤-LBD in complex with T 3 and T 4 . Data collection and refinement statistics are presented in Table I. A ribbon diagram of the T 4 -liganded TR␤ structure (pink), superimposed over the T 3 -liganded TR␤ structure (cyan) is shown in Fig. 6. The complexes are similar in overall fold, both to each other and to previously determined structures of TR LBDs bound to agonists. Moreover, T 4 is completely buried within the core of the LBD, just like other agonists. The TR␤-T 3 structure has a pocket volume of 572 Å 3 , whereas the TR␤-T 4 structure exhibits a pocket volume of 607 Å 3 , as determined by GRASP (33). The TR␤-T 3 and TR␤-T 4 structures share the same space group and crystal contacts. Thus, differences in pocket volume between these two structures are probably attributable to the difference in the size of the ligand.
Direct comparisons of the TR-T 4 and TR-T 3 complexes (using "blinking" between aligned LBD structures in Insight II (Accelrys)) revealed concerted backbone shifts in four distinct regions. The first comprises H12, the H11-H12 loop and the wall of the ligand binding pocket and lies close to the 5Ј iodine group (Fig. 6). The other regions include: N-terminal residues 199 -212 (part of the DNA binding domain C-terminal helix (H0), which is included in this structure); H2 residues 234 -243, portions of the underlying ␤-sheet (residues 318 -321 and 325-339) and the loop between H2 and H3; and the N terminus of H3 (residues 248 -267). Each of these regions of TR usually exhibits poor electron density in crystals, suggesting that they correspond to mobile regions of the protein (16 -18). Thus, alterations in these regions are less likely to be significant for understanding ligand discrimination than those of H11-H12 region.
The H11-H12 loop (residues 445-453) is shifted by about 1 Å in the TR-T 4 structure relative to the TR-T 3 structure (Fig. 7), the C terminus of H11 (residues 437-444) is pulled inwards toward the pocket, accentuating a kink also present in the T 3 structure and other agonist-bound TR LBD structures, and the C-terminal end of H12 (residue 460) is pushed outwards in the presence of T 4 . Despite these alterations, residues that comprise the coactivator binding surface (on H12 and the upper part of H3 and H5) adopt a structure with backbone positions identical to those seen in the TR-T 3 structure, and side chain positions nearly identical. This is consistent with the finding that T 4 promotes coactivator binding in vitro, and displays agonist activity in vivo. Nonetheless, direct comparisons in Rasmol indicate that H12 (residues 452-460) has closer contacts with the main body of the LBD in the T 3 complex than in the T 4 complex. This suggests that H12 packs less tightly against the LBD. Moreover, the average B-factor for protein atoms was higher for TR-T 4 (54.05) than for TR-T 3 (49.61), and the TR-T 4 structure had lower resolution (3.1 Å) than the TR-T 3 structure (2.5 Å). Thus, the TR␤-T 4 complex exhibits a greater degree of disorder than the TR-T 3 structure.
The conformational alterations that occur within the hormone binding pocket near the T 4 5Ј iodine group are shown in detail in Figs. 8 and 9. Strikingly, the 5Ј iodine fits neatly into a small "niche" in the wall of the pocket (Fig. 8A). This feature is analogous to similar niches that accommodate the other iodine groups in the T 4 and T 3 structures and is comprised of two distinct parts: an upper region that consists of residues This precisely positioned niche permits TR to accommodate T 4 completely within the enclosed pocket despite the presence of the 5Ј iodine group. The niche is also present within the TR␤-T 3 structure, but it is smaller (Fig. 8B). Superimposition of the TR-T 4 and TR-T 3 complexes reveals this region of the pocket expands slightly in the presence of T 4 (Fig. 9, compare mesh surface, T 4 with solid surface, T 3 ). This expansion is a result of a number of amino acid side chain shifts. The largest involves Met 310 (on H6), which lies above the 5Ј iodine in the TR-T 4 complex. If one considers the receptor in the orientation seen in Fig. 9, a steric clash between Met 310 and the 5Ј-iodine shifts the entire ligand toward the "left" of the receptor relative to the position of ligand in the TR␤-T 3 complex (detailed in Table II). This repositioning accentuates further steric clashes between the 5Ј iodine group and side chains of two residues (Phe 455 , Phe 459 ) on H12 itself. In addition, the kink in H11 probably results from a steric clash between the 5Ј iodine and His 435 (see Table II for distances). These alterations enlarge the niche that accommodates the 5Ј iodine substituent and permit H12 to pack against the LBD and complete the coactivator binding surface.

FIG. 5. T 4 acts as an agonist in mammalian cell cultures.
A, graph shows alkaline phosphatase activity (the alkaline phosphatase gene is under control of a TH-regulated promoter) as a function TH concentration in chinese hamster ovary cells stably transfected with human TR␣ (TRAF-␣ cells). B, as in A, except that the experiment was performed in CHO cells stably transfected with TR␤ (TRAF-␤ cells). C, as in A, except that an additional treatment (T 4 preparations spiked with 10% T 3 ) was employed.  6. Structure of the TR␤-T 4 complex. Superimposed ribbon diagrams of TR␤-T 3 complex (cyan) and the TR␤-T 4 complex (pink) show that the overall fold is nearly identical. The coactivator binding surface H3, 4, 5, and 12 is labeled.

FIG. 7. Superimposed views of the backbone trace of the TR␤ H11-H12 region derived from TR␤-T 4 and TR␤-T 3 structures.
Backbone displacements in TR-T 4 (red) relative to TR-T 3 (blue). Helix 12 is highlighted. Displacements near the ligand binding pocket are confined to three areas: (a) C terminus of helix 12, (b) loop connecting helices 11 and 12, (c) disordered loop consisting of residues 250 -265 linking nearby helices 2 and 3. Despite these structural alterations, the center of helix 12 shows no backbone displacement and completes the coactivator binding surface.

DISCUSSION
The current studies address T 4 interactions with TRs. As developed in the Introduction, T 3 is thought to be the major active form of TH and binds TRs with ϳ30-fold the affinity of T 4 (6). Free circulating T 4 levels are, however, about four to six times that of T 3 and intracellular T 4 /T 3 ratios can vary, so it is conceivable that T 4 could occupy a significant fraction of nuclear TRs in some contexts. The lower affinity of T 4 for TRs is probably related to the bulky 5Ј-iodine moiety that, based on our previous structures of TR-LBDs in complex with T 3 and related agonists, should not fit readily into the hormone binding pocket (14, 16 -18). Indeed, placement of some bulky 5Ј extensions on high affinity TR agonists can even create antagonists (21). Thus, we asked how T 4 interacts with TR, whether it behaves as an agonist or antagonist, and how it can fit into the TR ligand binding pocket.
We initially examined properties of TR-T 4 complexes. We confirmed that T 4 bound to TR more weakly than T 3 , and further demonstrated that T 4 dissociates from TRs faster than T 3 , (Fig. 1). Moreover, the TR complex with T 4 is less compact than that with T 3 , as suggested by migration of TR-T 4 complexes closer to unliganded TRs than to TR-agonist complexes on HIC (Fig. 2, A and B) and in gel shift assays with DNA (Fig.  4). The unusual HIC elution profile is not a reflection of lower affinity of TR for T 4 , because TR complexes with Dimit (which lacks a 5Ј substituent yet only exhibits 20% of the affinity of T 4 for TR) elute at a similar position to TR-T 3 complexes (Fig. 2). It is also unlikely to reflect rapid dissociation of T 4 while on the column, because TRs in complex with radiolabeled T 4 also elute at a similar position to TRs in complex with unlabeled T 4 (Fig.  2C). Despite the less compact nature of the TR-T 4 complex, maximally effective doses of T 4 were as effective as those of T 3 in stimulating association of coactivators (GRIP1 and TRAP220; Fig. 3), release of corepressors (N-CoR and SMRT; Fig. 3) and dissociation of TR homodimers from DNA (Fig. 4). Moreover, maximally effective doses of T 4 were as effective as those of T 3 in stimulating activity of a TRE-regulated reporter in cultured cells that express TR␣ or TR␤, and did so in conditions in which it is unlikely that agonist activity of T 4 stems from contaminating T 3 or intracellular conversion of T 4 to T 3 . Thus, it is likely that T 4 and T 3 promote similar overall conformational rearrangements within the TR-LBD in these conditions and that H12 must fold into the active conformation in the presence of T 4 .
The crystal structure of the TR␤-T 4 complex supports the notion that T 4 induces a TR conformation similar to that observed with higher affinity agonists (Fig. 6). TR adopts this fold because, overall, T 4 fits tightly into the ligand binding pocket despite the presence of the 5Ј iodine group. The pocket accommodates the bulky 5Ј iodine via shifts in the position of several amino acid side chains in the pocket relative to their positions in the TR-T 3 complex. These changes enlarge a niche that lies close to the 5Ј-position of the first thyronine ring and closely matches the size and shape of the 5Ј iodine (Figs. 8 and 9). The requirements for these structural alterations for fitting of T 4 relative to T 3 likely explain the reduced affinity of the TRs for T 4 relative to T 3 . However, the niche permits H12 to fold over the bulky iodine group and complete the coactivator binding surface. Thus, the presence of an adaptable niche in the TR ligand binding pocket allows T 4 to behave as agonist, despite the 5Ј-extension.  Although H12 adopts the typical active conformation in the presence of T 4 , our crystal structures indicate that the H11-H12 loop is more mobile and more loosely packed against the LBD in the presence of T 4 than in the presence of T 3 (Figs. 6 and 7). These features suggest explanations for the observed differences between the behavior of TR-T 4 and TR-T 3 complexes. A tendency of H12 to oscillate between conformations that resemble liganded and unliganded states would reduce the efficiency of the capping of the pocket and allow T 4 to dissociate more readily. Loose packing of H12 would also expose more of the hydrophobic interior of the protein, explaining unusual mobilities of the TR-T 4 complex in HIC and gel shifts. While T 4 consistently behaves as a full agonist in our hands, it is conceivable that the loose packing of H12 induced by T 4 versus T 3 could leave the TR open to external influences that alter the response to the ligand. For example, in cells with high corepressor and/or low coactivator levels, H12 might be forced into the unliganded conformation and T 4 could display partial agonist, or even antagonist, activity. This issue will require further investigation.
We do not yet have a similar structure of TR␣ in complex with T 4 , but there are great overall similarities between the TR isoforms in terms of overall LBD fold (18), sequence and, as reported here, activity in the presence of T 4 and T 3 (Figs. 1, 2, and 5) suggesting that TR␣ will adapt to the 5Ј iodine extension in a similar way to TR␤. Interestingly, T 4 dissociates from TR␤ even more rapidly than from TR␣ (Fig. 1). In this regard, TR␤ tends to exhibit less rigidity in the vicinity of the H11-H12 region than equivalent structures of TR␣ in complex with the same ligand (as judged by temperature factors) (18). These differences in rigidity within the TR H11-H12 region may explain the increased dissociation rate of T 4 .
We previously proposed that TR ligands with bulky 5Ј side chains should perturb H12 and act as antagonists (reviewed in Ref. 21). This idea, the extension hypothesis (19,20), has been partly validated by our synthesis of novel TR antagonists based on these principles (22)(23)(24), and structures of other NRs in complex with antagonists (such as selective estrogen receptor modulators) (34,35). Nonetheless, T 4 acts as an agonist, just as we have learned that many other TR ligands with extensions that are even bulkier than the T 4 5Ј iodine group can behave as agonists (23,24). Thus, the nature of the extension and its relationship to the rest of the ligand is important for overall agonist/antagonist activity, and TR must accommodate larger ligands in ways that cannot be easily predicted from structures of TR ligands without "extensions." NR antagonists perturb H12 position in two ways, by directly interfering with H12 packing or occupying the pocket without inducing the structural changes required for the agonist configuration (reviewed in Ref. 21 and references therein). Our studies add to an emerging pattern, which suggests that NRs alter their conformations in a variety of ways to accommodate hormone analogs and allow them to act as agonists. The TR-T 4 crystal structure reported here reveals that the pocket can reorganize to accommodate the 5Ј iodine group, but with a resulting strain of the overall structure relative to T 3 . An extreme case of accommodation is for PXR, where the pocket expands to accommodate larger ligands and collapses to accommodate smaller ligands (36 -38). In this case, packing does not appear to result in the stability differences we have detected between T 4 -versus T 3 -liganded TRs. Finally, we recently showed that TR accommodates a ligand (GC-24) that binds TR␤ with about 40-fold the affinity of TR␣, and has a 3Ј phenyl extension and a 5Ј hydrogen (39), by opening up a hydrophobic patch on the inner surfaces of H3 and H11 that is not normally part of the pocket. It will be interesting to determine how TR accommodates ligands with even bulkier 5Ј extensions (22)(23)(24), or why ligands with particular 5Ј extensions, such as the DI-BRT isopropyl group and the NH 3 phenyl group, act as antagonists.
Our studies do not address the question of whether T 4 is a relevant species of TH in physiological settings. As indicated in the Introduction, many factors regulate relative T 4 /T 3 concentrations in the nucleus, making it difficult to gauge the extent to which intracellular T 4 participates in TR binding. Nonetheless, the observation that TR can reorganize to create a niche that precisely accommodates the T 4 5Ј iodine, coupled with the fact that the potency of T 4 is about 10% that of T 3 in cell culture and that free circulating T 4 concentrations are 4 -6-fold those of T 3 , raises the distinct possibility that T 4 could exhibit significant agonist activity in humans.