Ligand-dependent Heterodimerization of Thyroid Hormone Receptor and Retinoid X Receptor*

The thyroid hormone receptors (TR) bind to cis-acting DNA elements as heterodimers with the retinoid X receptors (RXR). These heterodimers display distinct specificities in mediating the hormonal response to target gene transcription. We characterized the interaction between TR (cid:97) 1 and RXR (cid:97) via their ligand binding domains (LBDs) and the effect of ligands on the interaction using a yeast two-hybrid system. The DNA binding domain (BD) of yeast Gal4 fusion to the LBD of TR (cid:97) 1 had no transcriptional activity on its own, but when it was coexpressed with the activation domain (AD) of yeast Gal4 fusion to LBD of RXR (cid:97) conferred activation to a reporter gene harboring a Gal4 binding site, indicating that LBDs of TR (cid:97) 1 and RXR (cid:97) interact with each other in solution. Furthermore, T 3 and 9- cis -RA increased the reporter activity, and an additive effect was observed when both ligands were added, indicating that the TR (cid:97) 1 (cid:122) RXR (cid:97) heterodimerization is augmented by their respective ligands in vivo . Using an in vitro pull-down experiment, we confirmed the ligand-dependent interaction observed in the yeast system. Matrix-bound glu- tathione S -transferase-RXR (cid:97) specifically coprecipitated the 35 S-labeled TR (cid:97) 1 above the control, and associated 35 S-labeled TR (cid:97) 1 was increased by the addition of T 3 and 9- cis -RA. These results imply a

The thyroid hormone receptors (TR) bind to cis-acting DNA elements as heterodimers with the retinoid X receptors (RXR). These heterodimers display distinct specificities in mediating the hormonal response to target gene transcription. We characterized the interaction between TR␣1 and RXR␣ via their ligand binding domains (LBDs) and the effect of ligands on the interaction using a yeast two-hybrid system. The DNA binding domain (BD) of yeast Gal4 fusion to the LBD of TR␣1 had no transcriptional activity on its own, but when it was coexpressed with the activation domain (AD) of yeast Gal4 fusion to LBD of RXR␣ conferred activation to a reporter gene harboring a Gal4 binding site, indicating that LBDs of TR␣1 and RXR␣ interact with each other in solution. Furthermore, T 3 and 9-cis-RA increased the reporter activity, and an additive effect was observed when both ligands were added, indicating that the TR␣1⅐RXR␣ heterodimerization is augmented by their respective ligands in vivo. Using an in vitro pull-down experiment, we confirmed the ligand-dependent interaction observed in the yeast system. Matrix-bound glutathione S-transferase-RXR␣ specifically coprecipitated the 35 S-labeled TR␣1 above the control, and associated 35 S-labeled TR␣1 was increased by the addition of T 3 and 9-cis-RA. These results imply a complex, sensitive crosstalk in vivo among nuclear receptors and their respective ligands through distinct hormonal signaling pathways.
Effects of thyroid hormone on a wide variety of tissues are mediated via specific nuclear thyroid hormone receptors (TRs) 1 (1) which are the cellular homologs of v-erbA (2). Several isoforms of TRs and TR variants have been isolated (2)(3)(4)(5)(6) and shown to be the members of a gene superfamily that includes steroid receptors, retinoic acid receptors (RARs), vitamin D receptors, peroxisome proliferator-activated receptors (PPARs), and numbers of proteins with high homology but as yet unidentified ligands.
Based on sequence homology and functional analysis, the nuclear receptors exhibit a modular structure with functionally separable domains. Members of the superfamily are characterized by a highly conserved cysteine-rich DNA binding domain containing two zinc finger structures necessary for sequence-specific DNA interaction (7). The complex carboxyl-terminal region of the receptors contains ligand binding, receptor dimerization, and putative transcriptional activation functions (8).
The TRs, as well as other members of the superfamily, regulate transcription by binding to response elements containing two or more copies (often degenerate) of the consensus motif AGGTCA (9,10). Recently it was shown that RARs, TRs, vitamin D receptors, and PPARs form heterodimers with the RXRs on bipartite hormone response elements composed of nonsymmetrical head-to-tail tandem AGGTCA half-sites (11)(12)(13).
To date two distinct dimerization surfaces were proposed in the DNA binding domain (DBD) and ligand binding domain (LBD) of TR. The surface in the DNA binding domain conferred selective power in DNA-dependent dimer formation (14). In contrast to the interface within the DBDs, dimerization motifs in the LBDs permit the heterodimeric complex subsequently to interact with response elements. The carboxyl-terminal LBD is responsible for DNA-independent dimerization that in vitro allows performation of certain dimers in solution before DNA targeting. This dimerization function is believed to stabilize the complex and promote the recognition of DNA. Several heptad repeats in LBD are well conserved among the members of the erbA-related nuclear receptor family and have been proposed to form a hydrophobic surface that might act as a receptor dimerization interface (15,16), which is structurally similar to the leucine zipper dimerization domain found in Jun-Fos (17). The heterodimerization influences the recognition of DNA targets and confers specificity for a defined spacing between two directly repeating hexameric sequences. Deletion of heptad repeats abolished the trans-activation function of the receptor (18,19). Furthermore, the formation of heterodimers results in cross-talk among different ligands potentially to affect a range of physiological processes. The effect of ligand on the formation of TR⅐RXR heterodimers in solution has not been demonstrated clearly. To investigate the role of ligand in the process of receptor dimerization before binding to the target DNA, we have used the yeast two-hybrid system that allows the measurement of specific protein-protein interaction in solution. In the present study, we demonstrated that TR/RXR interaction occurred in solution before DNA targeting and was augmented by T 3 and 9-cis-RA additively both in vivo and in vitro.  1 The abbreviations used are: TR(s), thyroid hormone receptor(s); RAR(s), retinoic acid receptor(s); PPAR(s), peroxisome proliferator-activated receptor(s); RXR(s), retinoid X receptor(s); DBD, DNA binding domain; LBD, ligand binding domain; T 3 , 3,3Ј,5-triiodo-L-thyronine; 9-cis-RA, 9-cis-retinoic acid; PCR, polymerase chain reaction; AD, activation domain; CPRG, chlorophenol red-␤-D-galactopyranoside; GST, glutathione S-transferase; Triac, 3,3Ј,5-triiodothyroacetic acid; T 4 , strain is MATa, ura3-52, his3-200, lys2-801, ade2-101, trp1-901, leu2-3, 112, gal4 -542, gal80 -538, LYS2::GAL1-HIS3::URA3::(GAL4 17-mers)3-CYC1-lacZ (obtained from CLONTECH). Both yeast host strains carry a lacZ reporter gene under the control of GAL4 binding site. HF7C contains a second reporter gene (HIS3), also under the control of GAL4 response elements. Yeast strains were grown at 30°C in YPD medium (1% yeast extract, 2% Bacto-Peptone, 2% dextrose) or in synthetic selection medium with appropriate supplements.
␤-Galactosidase Assay-␤-Galactosidase activity, which was the product of the lacZ reporter gene, was used to indicate the interaction between the two hybrid proteins in vivo, which reconstitutes Gal4 function. Recombinant strains were grown overnight at 30°C in 5 ml of synthetic medium lacking leucine and tryptophan to select for the maintenance of plasmids. The yeast cultures were initially diluted to an A 600 nm of 0.05 and incubated in the presence of various concentrations of T 3 and/or 9-cis-RA at 30°C overnight in separate tubes. The A 600 nm was determined. 700 l of Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , 50 mM ␤-mercaptoethanol, pH 7.0) was added to 0.1 ml of culture. The cells in the reaction mixtures were permeabilized by adding 50 l of 0.1% SDS and 50 l of chloroform and vortex vigorously for 30 s. The reaction were started with the addition of 0.16 ml of chloroform red-␤-galactopyranoside (CPRG) (4 mg/ml) at 30°C and stopped by adding 0.5 ml of 3 mM ZnCl 2 , and the A 570 nm was read. ␤-Galactosidase activity was determined with the values at A 570 nm and A 600 nm using the equation U ϭ (1,000 where t is the time of reaction (min), and v is the volume of yeast culture used in the reaction mixture (ml) (21).
Expression of Recombinant Proteins-To express the fusion protein with glutathione S-transferase (GST), PCR-amplified full-length RXR␣ cDNA was inserted in-frame into BamHI and EcoRI cloning sites of the pGEX-2T vector (Pharmacia Biotech Inc.). The following oligonucleotides were used to amplify the full-length hRXR␣: 5Ј-agatctcatATGGA-CACCAAACATTTCCTG-3Ј (forward primer) and 5Ј-gaattcTAAGT-CATTTGGTGCGGC-3Ј (reverse primer). Overnight cultures of Escherichia coli JM109 carrying the recombinant GST-RXR␣ and GST control plasmid were diluted 100-fold, cultured for 5-6 h, and then induced with 0.1 mM isopropyl ␤-D-thiogalactopyranoside. After another 3 h, bacteria were collected and washed with phosphate-buffered saline. Pellets were suspended in phosphate-buffered saline containing 1% (v/v) Triton X-100 and sonicated. Debris was removed by centrifugation. The fusion protein or the GST control protein was bound to glutathione-Sepharose (Pharmacia) and washed extensively with phosphate-buffered saline containing 1% (v/v) Triton X-100. Matrix-bound proteins were used for interaction experiments. 35 S-Labeled TR␣1 protein was produced by in vitro translation (Promega TNT, Madison, WI) using a T7 expression vector containing full-length TR␣1 cDNA (pET TR␣1) (22).
Interaction Experiments-In vitro translated 35 S-labeled TR␣1 (1-2 l) was incubated for 20 min at room temperature with glutathione-Sepharose (10 l) preloaded with GST-RXR␣ fusion or GST control protein in 250 l of binding buffer (20 mM Tris-Cl, pH 7.8, 100 mM NaCl, 10% glycerol, 1 mM dithiothreitol, 1 mM EDTA, 1 mM phenylmethanesulfonyl fluoride, 1 mM leupeptin, 1 mM pepstatin) in the presence or absence of 10 Ϫ6 M T 3 and/or 9-cis-RA. After extensive washing with binding buffer, bound proteins were eluted in 25 l of Laemmli sample buffer, boiled for 10 min, and resolved by SDS-polyacrylamide gel electrophoresis (10%) followed by autoradiography. The results of the in vitro reactions and the amount of 35 S-labeled protein bound by GST or GST-RXR␣ were visualized and quantified using a PhosphoImager (Fuji BAS 1500).

Gal4BD-TR␣1(LBD) and Gal4AD-RXR␣(LBD) Fusion Proteins Interact in Yeast-
To investigate the ability of TR␣1 and RXR␣ to heterodimerize through their LBD, the interaction of Gal4BD-TR␣1(LBD) and Gal4AD-RXR␣(LBD) fusion proteins in yeasts was examined using the yeast two-hybrid system. The yeast reporter strain HF7c was cotransformed with pGBT9-TR␣1(LBD) or pGAD424-RXR␣(LBD) or both plasmids and selected with tryptophan and leucine dropout medium. Transformed colonies were streaked onto tryptophan and leucine dropout medium with or without histidine. In this assay, the formation of a complex between TR␣1(LBD) fused to Gal4BD and RXR␣(LBD) fused to Gal4AD confers histidine auxotrophy and ␤-galactosidase activity. As shown in Fig. 2, the yeasts expressing Gal4BD-TR␣1(LBD) and Gal4AD-RXR␣(LBD) are allowed to grow in the absence of histidine. The yeast cotransformed with pairs of Gal4BD-TR␣1(LBD) and Gal4AD, or Gal4BD and Gal4AD-RXR␣(LBD), did not permit growth in the absence of histidine, indicating that histidine auxotrophy is the result of the interaction between TR␣1(LBD) and RXR␣(LBD). Similar results were obtained in the reverse experiment using cotransformants expressing Gal4BD-RXR␣(LBD) and Gal4AD-TR␣1(LBD) (data not shown).
Then we tested the ability of the heterodimers between LBDs of TR and RXR as a transcriptional activator for Gal4 DBD. pRXR␣(LBD), just an expression vector for RXR␣(LBD), was cointroduced into SFY526 with pGBT9-TR␣1(LBD). No increase of ␤-galactosidase activity was observed in the yeasts transformed with Gal4BD-TR␣1(LBD) and RXR␣(LBD) expressing vector, and T 3 or 9-cis-RA did not show a significant increase of ␤-galactosidase activity (Fig. 3). It is concluded, therefore, that the heterodimers of LBDs of TR and RXR cannot function as a transcriptional activator in the yeasts.
When a combination of Gal4BD-TR␣1(LBD) and Gal4AD-RXR␣(LBD), or Gal4BD-RXR␣(LBD) and Gal4AD-TR␣1(LBD), were introduced into yeast strain SFY526, a significant increase in ␤-galactosidase activities was observed. Both T 3 and 9-cis-RA increased the ␤-galactosidase activity, and an additive effect was observed when both ligands were present. Dose dependence of ligands for activation of TR/RXR interaction was assessed. Yeast colonies were cultured overnight in the presence of various concentration of T 3 or 9-cis-RA. As shown in

FIG. 4. T 3 and 9-cis-RA augment the interaction between TR␣1(LBD) and RXR␣(LBD) in a dose-dependent manner.
A combination of pGBT9-TR␣1(LBD) (BD-TR␣) and pGAD424-RXR␣(LBD) (AD-RXR␣) was introduced into yeast SFY526 and selected with tryptophan and leucine dropout medium. Yeast colonies were cultured overnight in the presence of various concentrations of T 3 (E) or 9-cis-RA (q). ␤-Galactosidase activities of yeast cultures were determined using CPRG (see "Materials and Methods"). Values are the mean Ϯ S.D. for at least three independent experiments performed in duplicate.

FIG. 2.
Interaction between TR␣1(LBD) and RXR␣(LBD) in yeasts. The yeast reporter strain HF7c was cotransformed with pGBT9 (BD-) and pGAD424 (AD-), or pGBT9-TR␣(LBD) (BD-TR␣) and pGAD424 (AD-), or pGBT9 (BD-) and pGAD424-RXR␣(LBD) (AD-RXR␣), or pGBT9-TR␣(LBD) (BD-TR␣) and pGAD424-RXR␣(LBD) (AD-RXR␣) plasmids and selected with tryptophan and leucine dropout medium. A transformed colony from each plate was streaked on synthetic medium deficient in tryptophan and leucine (nonselective; left) or on the same medium lacking histidine (selective; right). The position and the expressed heterologous proteins of each strain are indicated in the diagram. Fig. 4, T 3 and 9-cis-RA increased the ␤-galactosidase activity in a dose-dependent manner. A half-maximal increase was observed at 10 Ϫ7 M for T 3 and 10 Ϫ6 M for 9-cis-RA. We examined further the effect of 9-cis-RA (10 Ϫ6 M) on the T 3 -dependent TR/RXR interaction. As shown in Fig. 5A, the presence of T 3 further activated the ␤-galactosidase activity without changing the sensitivity to 9-cis-RA. In Fig. 5B, the reverse experiments were performed to show the T 3 dose-response curve in the presence or absence of 10 Ϫ6 M T 3 . Similarly, the presence of 9-cis-RA further activated the ␤-galactosidase activity without changing the sensitivity to T 3 . All fusion proteins were expressed at approximately the same level in the transformed yeasts, as determined by ligand binding assay (data not shown). The above results were also verified by cotransforming the yeast host strain HF7c with the two hybrid vectors. In HF7c, the lacZ reporter gene is under the control of a promoter different from that used to control the lacZ gene in SFY526. These two promoter share only GAL4 response elements in common; the rest of the promoter sequences differ significantly.
We next tested the effect of T 3 analogs on activating the BD-TR␣1(LBD)⅐AD-RXR␣(LBD) complex. As shown in Fig. 6, the order of potency of each analog to induce ␤-galactosidase activity is consistent with previously determined affinity constant of binding of the ligands to the TR (23). Half-maximal activation was obtained at 10 Ϫ8 M for Triac, 10 Ϫ7 M for L-T 3 , D-T 3 , and 10 Ϫ6 M for L-T 4 (Triac Ͻ L-T 3 Ͻ D-T 3 Ͻ L-T 4 ). These compounds did not affect the growth property of the yeasts.
Effect of Ligand on the Interaction between the LBDs of TR␣1 and RXR␣ in Vitro-The GST-RXR␣ fusion protein was used to investigate the effects of ligands on the interaction between the LBD of RXR and TR in vitro. Complexes with GST-RXR␣ were retained on glutathione-Sepharose, the beads were washed and pulled down by centrifugation, and associated proteins were analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography. TR␣1 was labeled with [ 35 S]methionine by in vitro translation and incubated with GST (Fig. 7, upper panel,  second lane) or a GST-RXR fusion protein (third through sixth lanes) bound to glutathione-Sepharose beads, as indicated. 35 S-Labeled TR␣1 was specifically retained in the presence of GST-RXR␣ but did not bind to the GST control protein (second lane). Approximately 7.5% of the total input of TR␣1 was specifically bound to the GST-RXR fusion protein. Addition of T 3 or 9-cis-RA increased the amount of 35 S-labeled TR␣1 that associated with GST-RXR␣, and further increase was observed when both ligands were present.
Interaction Properties of TR␤1 and Their Mutant Receptors with RXR␣-We next investigated the ability of TR␤1(LBD) to interact with RXR␣(LBD). Similar to the results with TR␣1 and RXR␣, both ligand-independent (constitutive) and ligandinduced interactions were observed with TR␤1 and RXR␣ (Fig.  8). In addition, we also examined the heterodimerizing properties of specific mutant receptors, TR␤1(G345R) and TR␤1∧422, to compare ligand-dependent transcriptional activity with their heterodimerizing activity. TR␤1(G345R), which was isolated from a patient with resistance to thyroid hormone (24), has no detectable T 3 binding activity because of a glycine to arginine substitution at amino acid 345 in the hormone binding domain.

FIG. 6. Effect of T 3 analogs on the interaction between TR␣1(LBD) and RXR␣(LBD).
A combination of pGBT9-TR␣1(LBD) and pGAD424-RXR␣(LBD) was introduced into yeast SFY526 and selected with tryptophan and leucine dropout medium. Yeast colonies were cultured overnight in the presence of various concentrations of T 3 analogs (Triac (q), L-T 3 (f), D-T 3 (Ⅺ), L-T 4 (E)). ␤-Galactosidase activities were determined using CPRG ("Materials and Methods"). Values are the mean Ϯ S.D. for at least three independent experiments performed in duplicate.
TR␤1∧422 is an artificial mutant receptor with 4 amino acids insertion in the 9th heptad region, which is known to be essential for heterodimerization with RXR (18). The ligand bindingdefective mutant TR␤1(G345R) revealed the constitutive and 9-cis-RA-induced interaction but did not show the T 3 -induced interaction. Heterodimerization-defective mutant TR␤1∧422 failed to support both constitutive-and ligand-induced interaction in the yeasts (Fig. 8). DISCUSSION The TRs are believed to function as heterodimers with RXR and exert their physiological activities through binding to the thyroid hormone response elements in DNA (16). Dimerization of receptors through the specific interfaces that exist in the LBD has been demonstrated for the glucocorticoid receptor, estrogen receptor, and progesterone receptor (26 -28). Evidence for heterodimer formation between TR and RXR (20) and TR homodimer formation (19) has been suggested based on TR/ RXR interaction in cotransfection studies and gel mobility shift assay or cross-linking studies (11)(12)(13). To examine in detail the interaction between TR and RXR in solution and the effect of their ligands in vivo, we have established the yeast two-hybrid system as an in vivo approach to analyze the heterodimerization between the LBD of TR and RXR. In this report, we provide evidence that the TRs are able to heterodimerize with RXR in the absence of DNA, and T 3 and 9-cis-RA augment this interaction in vivo, supporting the physiological importance of the heterodimer formation in solution.
LBDs of TR␣1 and RXR␣ act as transcriptional regulators in mammalian cells when fused to the heterologous DBD of the transcription factors (29). Therefore, it is important to know whether Gal4BD-TR␣1(LBD) or Gal4BD-RXR␣(LBD) alone can activate the reporter gene in a ligand-dependent manner in the yeast system. If the ligand-induced activity of the reporter gene could be caused by activation of Gal4BD-TR␣1(LBD) or Gal4BD-RXR␣(LBD) by the ligands, the activities of the reporter gene do not exactly reflect the interaction of the two proteins. In Fig. 3, we showed that neither of Gal4BD-TR␣1(LBD) nor Gal4BD-RXR␣(LBD) alone activates the reporter gene, irrespective of the presence of ligands. Moreover, using pRXR␣(LBD), which lacks Gal4AD, we further analyzed the capability of the TR␣(LBD)⅐RXR␣(LBD) heterodimer to function as a ligand-dependent transcriptional activator in the yeasts. As shown in Fig. 3, a combination of Gal4BD-TR␣1 and RXR␣(LBD) did not activate the reporter gene even in the presence of ligands. These results indicate that neither TR␣1(LBD) alone nor a combination of TR␣1(LBD) and RXR␣(LBD) confers ligand-dependent activation to the reporter gene in the yeast system. Therefore, it is concluded that the activity of the reporter gene indicates the reconstitution of the Gal4 molecule, allowing the measurement of the strength of interaction between TR␣1(LBD) and RXR␣(LBD).
Because the yeast two-hybrid system is extremely sensitive and can detect very transient interactions, we utilized the in vitro pull-down experiment as a second more stringent assay to confirm the ligand-induced heterodimerization seen in the yeast system. As shown in Fig. 7, it is demonstrated that the dimerization can occur in solution and that it is augmented by ligands in a pull-down experiment using the bacterially expressed RXR␣ fused to GST.
It has been shown in mammalian systems that cotransfection of a Gal4-RXR expression vector and an expression vector FIG. 7. In vitro interaction of TR␣1 with RXR␣. 35 S-Labeled in vitro translated full-length TR␣1 was incubated with matrix-bound GST or GST-RXR␣ at 25°C for 20 min in the absence (Ϫ) or presence (ϩ) of 10 Ϫ6 M T 3 or/and 9-cis-RA. After washing of the beads, the proteins were eluted and separated on 10% polyacrylamide gel. Radioactivities were visualized and quantified by PhosphoImager (Fuji Film, BAS1500). The arrowhead shows the position of the TR␣1 protein of expected size. The bar chart was generated by PhosphoImage analyses using Fuji Mac BAS. Each bar represents the mean relative density (Ϯ S.E.) of each band. The values presented are the mean Ϯ S.D. for three independent determinations performed in duplicate. Results were analyzed by Student's t test, and the changes observed were statistically significant with a p value of less than 0.05. for the LBD of the TR can confer T 3 responsiveness to a promoter containing a Gal4 binding site (29). This experiment indicates that TR/RXR can form functional heterodimers in vivo in the absence of DBDs. Our results in the yeast system are consistent with the reported data in mammalian systems. In addition, augmentation of the TR/RXR interaction by T 3 analogs correlated with the biological potency and ligand affinity of thyroid hormone analogs (Triac Ͼ T 3 Ͼ T 4 ), suggesting a physiological importance of ligand-induced heterodimerization in solution.
In mammalian cells, the enhancement of T 3 -dependent reporter gene activation by TR⅐RXR heterodimers is well established, and the binding of TR to thyroid hormone response elements is much more efficient in the presence of RXR. Herein, we provide more direct evidence that TR and RXR form heterodimers before DNA targeting in vivo, and ligand binding clearly enhances receptor associations with each other. Previous studies have shown that ligand binding induced conformational change in TR in solution (30 -33). The structures of the LBD of RAR␥ (34) and TR␣ (33) crystallized in the presence of their respective ligands revealed that helix 12 was aligned over the ligand binding pocket, in contrast to its position in unliganded receptors, in which it protrudes away from the LBD. In liganded receptors, hydrophobic residues within helix 12 face toward the pocket, perhaps contacting the ligand, whereas the negatively charged residues are exposed on the protein surface. Thus, as suggested by Renaud et al. (34), realignment of helix 12 over the ligand binding pocket when the receptor binds ligand may generate a novel surface for interaction with the partner proteins or bridging proteins. Ligand binding seems to stabilize protein-protein interactions that lead to high affinity DNA binding and trans-activation. Although it is clear that ligand induces conformational change in the DNA-bound receptors, the precise molecular mechanism by which the TRs regulate transcription in response to the binding of ligands remain enigmatic. Ligand-dependent heterodimerization of TR and RXR is involved, at least in part, in the ligand-induced gene activation.
It has been suggested recently that the DBD of TR and RXR can bind to a thyroid hormone response element sequence as a heterodimer in vitro, and a dimer interface found in the DBD has selective power for recognition of specific DNA interaction (36). On the other hand, we demonstrated that DNA binding is not necessary for heterodimer formation via the surface of the LBD, suggesting that several domains cooperate for forming heterodimers on target DNA in vivo. These results are consistent with the two-step hypothesis for binding of heterodimers to DNA (37). In the first step, TR would form solution heterodimers with RXR through their LBDs. In the second step, the DBDs, by virtue of their proximity, would be able to bind a high affinity site in DNA. Once bound to DNA, the receptors are capable of modulating transcription.
Since RXR plays a central role in mediating many hormonal signals, including retinoids, thyroid hormone, vitamin D 3 , and peroxisome proliferator activators, ligand (T 3 )-induced heterodimerization leads to an influence on other nuclear receptor signaling by squelching out the common partner, RXR. Recently, Chu et al. (25) reported that TR inhibits PPAR signaling by squelching out RXR. They observed that T 3 enhances the inhibition of PPAR activity by TR. Our results explain this phenomenon well, because T 3 increases the number of TR⅐RXR heterodimers, resulting in a decrease in the number of active PPAR⅐RXR heterodimers.
Recently the apparent hormone dependence was observed in an estrogen receptor dimerization experiment (26). Wang et al. (35) studied extensively estrogen receptor homodimerization using a yeast two-hybrid system. A particularly useful feature of the yeast two-hybrid system is the ability to study the heterodimerization activity of the various receptor forms. It would be interesting to analyze solution interaction among nuclear receptors including orphan receptors. This approach may lead to an understanding the role of ligand in transcriptional activation by nuclear receptors.