Different DNA Elements Can Modulate the Conformation of Thyroid Hormone Receptor Heterodimer and Its Transcriptional Activity*

Thyroid-hormone receptors (TRs) form heterodimers with retinoid-X receptors (RXRs) on thyroid-hormone-response elements (TREs). However, it is not known whether the formation of liganded TR/RXR heterodimer on a TRE alone is sufficient to dictate transcriptional activity. We designed several mutated DR4s (half-sites arranged as direct repeats with a nucleotide gap of 4) that bound TR/RXR heterodimers preferentially, and employed them to characterize functional and biochem- ical properties of the heterodimers on DNA. Although TR/RXR heterodimer binding was similar on some of the mutated DR4s, transient transfection assays showed that TR (cid:97) failed to support triiodothyronine (T 3 )-stimu- lated transcription on “inactive” DR4s but mediated basal repression on both “active” and inactive mutated DR4. T 3 binding assays showed that the mutated DR4s did not affect T 3 binding to the heterodimer. Finally, partial proteolysis studies revealed that binding of ac- tive DR4 elements and T 3 to the heterodimer synergis-tically enhanced heterodimerization-induced protease resistance of TR, but not RXR, in the heterodimer. These results suggest that: 1) liganded TR/RXR heterodimer binding to a DR4 is not sufficient for transcriptional activation of the target gene, and 2) DNA sequences in specific TREs may modify T 3 -mediated transcription by affecting the conformation of the liganded heterodimer. these bands were not observed These results suggest that the majority of TR and RXR can form heterodimers with each other on each DR4 element under the conditions used and that protease-resistant products from TR/RXR heterodimer can maintain the heterodimer formation and DNA binding activities. Further, active DR4 may possess a conformation of TR/RXR heterodimer that is different from that of the inactive DR4.

Thyroid-hormone receptors (TRs) form heterodimers with retinoid-X receptors (RXRs) on thyroid-hormoneresponse elements (TREs). However, it is not known whether the formation of liganded TR/RXR heterodimer on a TRE alone is sufficient to dictate transcriptional activity. We designed several mutated DR4s (half-sites arranged as direct repeats with a nucleotide gap of 4) that bound TR/RXR heterodimers preferentially, and employed them to characterize functional and biochemical properties of the heterodimers on DNA. Although TR/RXR heterodimer binding was similar on some of the mutated DR4s, transient transfection assays showed that TR␣ failed to support triiodothyronine (T 3 )-stimulated transcription on "inactive" DR4s but mediated basal repression on both "active" and inactive mutated DR4. T 3 binding assays showed that the mutated DR4s did not affect T 3 binding to the heterodimer. Finally, partial proteolysis studies revealed that binding of active DR4 elements and T 3 to the heterodimer synergistically enhanced heterodimerization-induced protease resistance of TR, but not RXR, in the heterodimer. These results suggest that: 1) liganded TR/RXR heterodimer binding to a DR4 is not sufficient for transcriptional activation of the target gene, and 2) DNA sequences in specific TREs may modify T 3 -mediated transcription by affecting the conformation of the liganded heterodimer.
Thyroid-hormone receptors (TRs) 1 belong to the nuclear receptor superfamily of ligand-inducible transcription factors (1) and are present as two major isoforms, TR␣ and TR␤ (2). They bind as monomers, homodimers, and heterodimers with nuclear proteins such as retinoid-X receptors (RXRs) to thyroidhormone-response elements (TREs) in the promoter region of target genes (2)(3)(4). RXRs are considered to be important heterodimer partners of TRs because RXR increases TR binding to TREs, augments TR-mediated in vitro transcription, and exists in almost all tissues (3)(4)(5)(6). Current evidence suggest that TR/ RXR heterodimers play a major role in mediating transcriptional activation on positively regulated target genes (7,8).
There are several parameters that influence the binding of TR/RXR heterodimer to TRE or change the transcriptional activity of TR. For example, the orientation of the two half-sites and the spacer size in a TRE (9 -12), the polarity of TR/RXR heterodimer (11, 13, 20 -23), the surrounding sequences and the variation in sequence of the half-sites in individual TREs (14,15,18,19), and addition of 9-cis-retinoic acid or heterodimerization with different RXR isoforms (16,17) appear to determine the binding property or transcriptional activity of TR.
Despite these findings, it is still not known whether the formation of TR/RXR heterodimer on a TRE alone is sufficient to dictate transcriptional activation or repression of basal transcription of a target gene. Further, it has not been demonstrated whether different response elements can induce distinct DNA-dependent structural changes within TR/RXR heterodimers, which may affect the magnitude of transcriptional activation.
In a previous study, we showed that different TR complexes, i.e. TR␣ monomer, homodimer, and TR␣/RXR␣ heterodimer, contact different regions of the same TRE (20). Based on these results, we designed several mutated DR4s (TRE half-sites arranged as a direct repeat with a 4-nucleotide spacer) that bind TR/RXR heterodimers preferentially. In order to characterize functional and biochemical properties of TR/RXR heterodimers on different mutated DR4 elements, we employed cotransfection and electrophoretic mobility shift assays (EMSAs), and identified functionally "active" and "inactive" mutated DR4 elements. Additionally, we compared the conformational change in TR/RXR heterodimer on active DR4 with that on inactive DR4, using a partial proteolytic digestion approach. We provide evidence that specific DNA sequences in DR4-TREs may modify triiodothyronine (T 3 )-mediated transcription by affecting the conformation of liganded TR/RXR heterodimer.
Plasmid Construction-The constructs containing rat TR␣ (pcDNAI/Amp-TR␣) was generated by inserting the EcoRI fragment from pBluescript rat TR ␣-1 into the EcoRI restriction site of the pcDNAI/Amp (Invitrogen, San Diego, CA) (25). Mouse RXR␤ (pcDNAI/Amp-RXR␤) was generated by inserting the EcoRI fragment from pBluescript mRXR␤ (kindly provided from Dr. R. Evans) (6). pcDNAI/Amp alone was also used in these experiments. Oligonucleotides containing DR4 and other mutated DR4s were cloned into the BamHI and HindIII restriction sites of the pT109 vector (which contains the herpes simplex thymidine kinase promoter coupled to a luciferase coding sequence) (26).
In Vitro Transcription/Translation-Unlabeled or [ 35 S]methioninelabeled receptor proteins were synthesized using a Promega transcription/translation kit (Promega, Madison, WI). Briefly, 2 g of circular plasmid DNA was transcribed and translated using T7 RNA polymerase in a 100-l reaction according to the manufacturer's procedure, except that 0.2 mM ZnCl 2 was added in the reaction buffer and incubated for 90 min at 30°C. Control reticulocyte lysate also was incubated without plasmid DNA under the same conditions. DNA Binding Assay/EMSA-Unlabeled receptor preparations (2-6 l) and 100,000 cpm oligonucleotide probe were incubated in 20 l of DNA-binding buffer (10 mM Tris-HCl (pH 7.5), 50 mM NaCl, 5% (v/v) glycerol, 1 mM EDTA, 1 mM dithiothreitol, and 1.5 g sheared salmon sperm DNA (Sigma)) for 30 min at room temperature as described previously (20). Different amounts of control reticulocyte lysate were added to some samples so that the total volume of the extracts was constant (6 l/lane). L-T 3 (Sigma) was added to some samples in certain experiments. After incubation, samples then were subjected to electrophoresis through a 5% polyacrylamide gel at 250 V in 0.5 ϫ TBE buffer (45 mM Tris borate and 1 mM EDTA) for 1.5 h at 4°C and analyzed by autoradiography as described previously (20).
To assess the effect of T 3 on TR/RXR heterodimer binding to DNA, a DNA binding reaction was performed as described by Demczuk et al. (27) with slight modification. [ 35 S]Methionine-labeled receptor (2 l), unlabeled receptor (2 l), and 10 ng of cold oligonucleotide probe were incubated in 20 l of blotting DNA-binding buffer (4% Ficoll, 80 mM KCl, 1 mM EDTA, 1.5 g of sheared salmon sperm DNA, 10 mM Hepes (pH 7.9)) in the presence of 10 Ϫ9 M unlabeled T 3 (27). To examine T 3 binding to receptor-DNA complexes, 10 Ϫ9 M L-[3Ј-125 I]T 3 (3000 mCi/g, Amersham Corp.) was added to gel shift reaction mixes containing unlabeled TR␣ and RXR␤ (2 l each) and 10 ng of cold oligonucleotide in siliconized tubes and incubated on ice for 1 h. Unlabeled T 3 was then added to 1 mM, and the incubation was continued for an additional 10 min. The protein, DNA, and T 3 complex in gel shift reactions were analyzed by 5% polyacrylamide gel in 0.25 ϫ TBE buffer for 2 h at 4°C. Labeled T 3 binding to receptor-DNA complexes was analyzed by autoradiography as described previously. To detect the labeled receptor signal, blotting of the gels was performed by the semidry-blotting method, which helped to reduce the background caused by free [ 35 S]methionine, using the Multiphor II NovaBlot electrophoretic transfer unit (Pharmacia Biotech Inc.) at a fixed current (mA) of 0.8 ϫ the gel surface area (cm 2 ) for 1.5 h. Semidry transfer to nitrocellulose (BA85, Schleicher & Schuell) was performed at 4°C in 48 mM Tris-HCl (pH 8.5), 39 mM glycine, 20% methanol. The membranes were analyzed by autoradiography at room temperature for 15 h. In experiments in which TR/RXR heterodimer dissociation rates were studied, receptor-containing samples were preincubated with labeled probe (100,000 cpm ϭϳ 0.5 ng) for 30 min at room temperature before addition of a 200-fold excess of unlabeled probe to the DNA-binding reactions. Samples then were subjected to electrophoresis as in the previous experiments after 0, 15, and 30 min of incubation at room temperature.
Cotransfection Studies-CV-1 cells were grown in Dulbecco's modified Eagle's medium plus 10% heat-inactivated (55°C, 1 h) fetal calf serum at 37°C. The serum was stripped of T 3 by constant mixing with 5% (w/v) AG1-X8 resin (Bio-Rad) twice for 12 h at 4°C before ultrafiltration. The cells were seeded at 10 5 cells per well of a six-well plate (Libron, Flow Laboratories, Inc. McLean, VA) 48 h prior to transfection. The cells in each well were transfected with various amounts of expression and reporter plasmids (1.6 g) as well as a RSV-␤-galactosidase control plasmid (1.0 g) by calcium phosphate coprecipitation as described previously (20). pcDNAI/Amp was added to some samples so that the total amount of expression plasmid was constant (3.0 g/well). Cells were grown for 24 h in the absence or presence of T 3 (10 Ϫ7 M) before harvesting. Cell extracts were analyzed for both luciferase and ␤-galactosidase activity (20).
Limited Proteolytic Digestion Analysis by SDS-PAGE-Protease di-gestion was carried out as described by Leng et al. (28) with slight modification. Trypsin (Sigma) was diluted in 10 mM Tris-HCl (pH 7.5). Fifty mg of T 3 were dissolved in 480 l of water and 20 l of 10 N NaOH, and 4.5 ml of ethanol (10 mg/ml stock) were added. T 3 stock was diluted to 1 mM by 0.005 N NaOH. Further dilution was performed with water. The carrier controls were prepared by the same procedure. Briefly, 2 l of labeled TR␣, 2 l of cold RXR␤ (or cold TR␣ and labeled RXR␤), 0.5 l of 10 ϫ DNA-binding buffer, and 0.5 l of T 3 or carrier were incubated in the presence of 250 ng of DNA element for 30 min at room temperature. A half l of trypsin was added to 5 l of receptor-DNA complexes, followed by a 15-min incubation at room temperature. The reaction was stopped by mixing with 5.5 l of 2 ϫ SDS loading buffer, heated for 5 min at 95°C, and subjected to electrophoresis (SDS-PAGE). Gels were then treated sequentially with 10% methanol, 10% acetic acid for 30 min and Amplify (Amersham Corp.) for 20 min. They were vacuum-dried at 80°C for 90 min. Signals were visualized by autoradiography at room temperature for 15 h. Limited Proteolytic Digestion Analysis by EMSA-Two l of unlabeled TR␣, 2 l of cold RXR␤, 0.5 l of 10 ϫ DNA-binding buffer, and 0.5 l of T 3 or carrier were incubated in the presence of 100,000 cpm of a oligonucleotide probe for 30 min at room temperature. A half l of trypsin was added to 5 l of receptor-DNA complexes, followed by a 15-min incubation at room temperature. To stop the reaction, the samples were put on ice and 15 l of 1 ϫ DNA-binding buffer were added to them. The samples were then immediately subjected to electrophoresis through a 5% polyacrylamide gel and analyzed as described in under "DNA Binding Assay/EMSA." ABCD Assay-The ratio of [ 35 S]methionine-labeled TR/RXR heterodimer bound and unbound to double strand oligonucleotides was examined using the avidin-biotin complex DNA binding (ABCD) assay, as described previously (29). Doublestranded oligonucleotides were filled-in with biotin-dUTP and Klenow, and biotin-dUTP homopolymer tails were added to them with terminal transferase (30). The binding reaction was performed as described under "Limited Proteolytic Digestion Analysis by SDS-PAGE," using 125 ng of biotinylated probe and 0.3 g/l sheared salmon sperm DNA. Receptor bound to DNA was separated from unbound receptor with streptavidin-agarose (Life Technologies, Inc.) which was pretreated with 1 mg/ml bovine serum albumin and 2 l of control reticulocyte lysate. Both unbound and bound receptors were analyzed in SDS-PAGE.
Quantification-EMSAs were repeated at least twice. Trypsin digestion studies were repeated more than five times. PhosphorImager and the ImageQuant software (Molecular Dynamics, Sunnyvale, CA) were used to quantify the radioactivity detected by the EMSA and SDS-PAGE. Transfection studies were repeated three times. The statistical significance of differences between luciferase activities in control samples containing pcDNAI/Amp alone and samples containing increasing amounts of TR␣ expression vector were evaluated in the presence or absence of T 3 , using the factorial ANOVA analysis (StatView TM II, Abacus Concepts Inc., Berkeley, CA) with p Ͻ 0.05.

RESULTS
Design of Oligonucleotide Probes-In a previous study, we identified the precise contact sites of various TR complexes on a synthetic DR4 (20). As shown in Fig. 1, in the absence of T 3 , TR monomer binds to the downstream half-site and the spacer of this particular DR4. TR homodimer binds to both half-sites and the spacer. In contrast, TR/RXR heterodimer also binds to both half-sites, but "prefers" to contact the downstream halfsite of DR4. Thus, the downstream half-site of this particular DR4 is a common binding site for all TR complexes. Based on these results, several mutated DR4s were designed that would bind TR/RXR heterodimers preferentially.
The first mutated DR4 is named UUU, because four base pairs of the spacer and downstream flanking sequences are changed to four base pairs of the upstream flanking sequence (U). UUU conserves two half-sites arranged as a direct repeat with a spacer of four nucleotides. Previously, TR/RXR heterodimer was reported to exhibit preferential contact with the downstream half-site of UUU, although TR does not bind to UUU as either monomer or homodimer (20).
It has been shown that two guanine residues in the primary half-site are important for TR homodimer binding (31). It was expected that a mutation of one of these guanine residues might produce differential binding of TR monomer, homodimer, and TR/RXR heterodimer. All residues in two halfsites were designated in numerical order (Fig. 1). M2 has a mutation of the no. 2 guanine site to adenine in the upstream half-site. M3 has a mutation of no. 3 guanine site. These M2 and M3 mutated DR4s have intact spacer and downstream half-site sequences. In contrast, M8 and M9 have mutations of no. 8 and no. 9 guanine sites in the downstream half-site, respectively.
TR␣/RXR␤ Heterodimer Can Bind to Mutated DR4 as Well as Wild Type DR4 -First, TR and TR/RXR heterodimer binding patterns on mutated DR4 elements were examined. Fig. 2 shows EMSAs using in vitro translated TR␣ and RXR␤. TR␣ monomer bound to DR4 (lane 2), with the complex migrating faster in the presence of T 3 (lane 3). TR␣ monomer also bound to M2 and M3 (lanes 11 and 15), but not to UUU, M8, or M9 (lanes 7, 19, and 23). Maximum amounts of TR␣ (6 l) did not show homodimer binding to DR4 or mutated DR4 elements (lanes 2, 7, 11, 15, 19, and 23). But a faint TR␣ homodimer band could be seen on DR4, but only on a long exposure (data not shown). In contrast, two complexes of TR␣/RXR␤ heterodimer bound to DR4, UUU, M2, and M3 with similar binding (lanes 5, 9, 13, and 17), but bound to M8 with moderate and M9 with less intensity (lanes 21 and 25). These bands likely reflect the presence of two in vitro translated TR␣ and RXR␤ products (Fig. 5, A and B), because these bands were competed by cold cognate oligonucleotides and were supershifted by both anti-TR␣ and anti-RXR␤ antibodies (20) (data not shown). No TR complexes were observed on Pit 1 probe as a control (data not shown). The binding of the heterodimer were not affected by the addition of 10 Ϫ7 M T 3 (data not shown) (Fig. 4).
These results confirm that the downstream half-site of this DR4 is important for TR monomer binding. These results also suggest that UUU, M8, and M9 are TR/RXR heterodimer selective elements. M2 and M3 are heterodimer as well as monomer binding elements.
T 3 -mediated Transcriptional Activation and T 3 -independent Basal Repression of DR4 and Mutated DR4 Elements-Func-tional properties of TR␣ on these mutated DR4 elements were examined by transient transfection experiments using CV-1 cells (Fig. 3). Previous immunodepletion experiments in our laboratory using anti-RXR antibodies suggest that the major TRAP (T 3 receptor auxiliary protein) in CV-1 cells is RXR␤ or a related protein (20). 2 If TR␣ alone is transfected to CV-1 cells, transactivation of exogenous TR␣/endogenous RXR heterodimer would be observed. Thus, it was reasoned that if TR/RXR heterodimer binding on DNA element alone was sufficient to dictate transcriptional activation in the presence of T 3 , enough binding of TR/RXR heterodimer might produce transcriptional activation on these mutated DR4 elements.
To test this hypothesis, increased amounts of TR␣ expression vector were cotransfected in CV-1 cells with a luciferase genecontaining reporter plasmid in which mutated DR4 elements were inserted in front of a TK promoter. Cells were then treated without (Fig. 3, filled bars) or with 10 Ϫ7 M T 3 (hatched bars) for 24 h and analyzed for luciferase activity. As shown in Fig. 3, increased amounts of unliganded-TR␣ mediated further repression of basal transcription on all mutated DR4 reporter genes, including heterodimer specific UUU. Surprisingly, TR␣ could activate transcription in response to T 3 on M2 (Fig. 3C) as well as DR4 (Fig. 3A), but failed to activate on UUU (Fig. 3B) and M3 (Fig. 3D). T 3 only derepressed basal transcription on UUU and M3. M8 and M9 gave results similar to those of UUU and M3 (data not shown).
Second, RXR␤ expression vector was cotransfected with TR␣ in order to increase the amount of RXR in CV-1 cells and increase further the formation of TR/RXR heterodimer on mutated DR4 elements. However, TR␣ and RXR␤ did not activate transcription in response to T 3 on either UUU, M3, M8, or M9 (data not shown). Considering these data, DR4 and M2 were designated active DR4 elements. UUU, M3, M8, and M9 were inactive DR4 elements. These results show that unliganded-TR heterodimer binding to mutated DR4s is correlated with and may be sufficient for basal repression of transcription. Inasmuch as TR/RXR heterodimer can bind to DR4, UUU, M2, and M3 with similar affinity, what is different between TR/RXR heterodimers on active DR4 and inactive DR4 elements?
T 3 Binding to TR/RXR Heterodimer Is Similar on Different DNA Elements-Kurokawa et al. (22) have shown that spacer size in a response element that consist of two direct repeat half-sites can influence the binding of the RXR-specific ligand to a RXR/retinoic acid receptor (RAR) heterodimer. It is possible that different spacer or half-site sequences may change T 3 binding to TR/RXR heterodimer on different elements. To test this hypothesis, we examined T 3 binding to TR/RXR heterodimer on different DNA elements by EMSA using cold TR, cold RXR, cold DNA, and trace amounts (10 Ϫ9 M) of [ 125 I]T 3 in the presence or absence of excess amount of unlabeled T 3 . As shown in Fig. 4B, labeled-T 3 binding to TR/RXR heterodimer/ DNA complex was similar among the different DNAs (lanes 2 ,  5, 8, and 11). As expected, addition of 10 Ϫ6 M unlabeled T 3 dissociated labeled T 3 from the heterodimer (lanes 3, 6, 9, and 12). To assess TR/RXR heterodimer binding to mutated DR4 elements in the presence of T 3 , EMSA was performed using in vitro translated cold TR␣, [ 35 S]methionine-labeled RXR␤, and equal amounts of each DNA probe in the presence of different concentration of unlabeled T 3 . As shown in Fig. 4A, two complexes of TR/RXR heterodimer bound to these four elements in similar binding in the presence of T 3 . Both bands were confirmed to contain both TR␣ and RXR␤ by antibody-supershift EMSA using anti-TR␣ and RXR␤ antibodies (20, data not shown). Because 10 Ϫ9 to 10 Ϫ6 M T 3 did not affect TR/RXR forma-2 P. M. Yen, personal communication. tion on different DNA elements, the T 3 binding shown in Fig. 4B is specific. Thus, T 3 binding to the heterodimer is not altered when the heterodimer binds to different DNA elements. Therefore, T 3 binding to TR heterodimer does not correlate with transcriptional activation by TR on the different mutated elements.
Binding to Active DNA Elements Increases Protease Resistance of TR␣/RXR␤ Heterodimer-Several groups have shown that T 3 induces conformational changes in TR␣, and that RXR enhances the ligand-dependent structural change of TR␤ (29,35,36). Some studies have employed partial proteolytic digestion as a useful method for analyzing conformational changes within TR and other proteins (36). It is suggested that a T 3induced conformational change may be an important step in the mediation of transcription of hormone-responsive genes.
Before carrying out conformational studies, it was important to confirm whether formation of the heterodimer was maximal, and whether proteolyzed TR/RXR heterodimer could retain DNA binding properties. First, to ensure that the maximal  6 g) and RSV-␤-galactosidase control plasmid (1 g) as described under "Materials and Methods." Additional pcDNAI/Amp control vector was added to some samples so that the total amount of plasmid cotransfected was constant for each sample. Cells were then treated without (filled bars) or with T 3 (hatched bars) for 24 h and analyzed for luciferase activity. Luciferase values were normalized by ␤-galactosidase activities and then calculated as fold-basal luciferase activity, with 1-fold basal activity defined as the luciferase activity with pcDNAI/Amp alone in the absence of T 3 . Each point represents the mean of three independent triplicate experiments with bars denoting the S.D. DNA elements are shown at the bottom of the graphs. The statistical significance of differences between luciferase activities in control samples containing pcDNAI/Amp alone and samples containing increasing amounts of TR␣ expression vector were evaluated in the presence or absence of T 3 , using the factorial ANOVA analysis (*p Ͻ 0.05; #p Ͻ 0.01).
binding was achieved, EMSAs were carried out with constant amounts of TR (2 l) and increasing amounts of RXR, or vice versa. No additional heterodimer-DNA complexes were formed on each mutated DR4 element beyond an equal volume of RXR and TR (data not shown). Under this condition, no TR monomers were observed. Second, the TR/RXR heterodimer dissociation rate from each mutated DR4 element was examined by preincubating both receptors and labeled probes in the absence or presence of T 3 for 30 min and then adding a 100-fold excess of unlabeled corresponding probe at time zero. As reported previously in the case of F2 element, more than 95% of the heterodimer on each mutated DR4 element still remained bound 30 min after the addition of unlabeled probes both in the absence or presence of T 3 (37) (data not shown).
Third, in order to obtain the ratio of bound and unbound TR/RXR heterodimer to each element, an ABCD assay was performed by incubating in vitro translated labeled TR with cold RXR and excess amounts of biotinylated probe (Fig. 5A), or labeled RXR with unlabeled TR and biotinylated probe (Fig.  5B). Bound receptors were then precipitated by avidin-agarose beads. After washing, both supernatant and pellets were analyzed by SDS-PAGE. In vitro translation of rat TR␣ produced two major proteins, 50 and 44 kDa (Fig. 5A). In vitro translation of mouse RXR␤ produced two major proteins, 65 and 48 kDa (Fig. 5B). Nonspecific binding was detected when receptors were incubated without any oligonucleotide probes. Because the nonspecific binding was the same as the signal using Pit 1 probe (lane 2 and data not shown), more than 90% of labeled TR or labeled RXR bound to each mutated DR4 probe (lanes 4, 6, 8, and 10). In contrast, the majority of receptors still remain in the supernatant using Pit 1 probe as a control (lane 1).
Lastly, the ability of TR/RXR heterodimer to bind to DNA after incubation with different amounts of trypsin was examined by EMSA (Fig. 5C). After the binding reaction with unlabeled receptors and labeled probe, different concentrations of trypsin were added to the reaction for a fixed time and then loaded to a polyacrylamide gel. Several faster migrated bands appeared with each probe, although T 3 did not change the binding and migration pattern significantly (data not shown). Surprisingly, larger amounts of the heterodimer bands were observed in the sample treated with 10 g/ml trypsin on active  1, 4, 7, and 10). To confirm the labeled receptor signal, electrophoresed samples were transferred to nitrocellulose membrane by the semidry blotting method, as described under "Materials and Methods." B, T 3 binding to TR/RXR heterodimer on different DNA elements. In vitro translated cold TR (2 l) and cold RXR (2 l) were incubated with cold DNA (10 ng), and trace amounts of  1, 4, 7, and 10).

FIG. 5.
Binding to active DNA elements alters protease sensitivity of the TR␣/RXR␤ heterodimer. Comparison of the ratio of bound and unbound TR/RXR heterodimer to various mutated DR4 elements. ABCD assays were performed by incubating in vitro translated [ 35 S]methionine-labeled TR with unlabeled RXR and excess amounts of biotinylated probe (A), or labeled RXR with cold TR and biotinylated probe (B). Bound receptors were then precipitated by avidin-agarose beads. After washing, both supernatants (unbound, U) and pellets (bound, B), as well as the molecular mass markers (M, numbers indicate kDa) were analyzed by SDS-PAGE, as described under "Materials and Methods." C, TR/RXR heterodimer binding to various mutated DR4 elements after trypsin digestion. In vitro translated TR and RXR were incubated with labeled mutated DR4 probes with carrier control or 10 Ϫ7 M T 3 (data not shown), then treated with indicated amounts of trypsin. Undigested controls were supplemented with an equal volume of 10 mM Tris-HCl (pH 7.5) (lanes 1, 5, 9, and 13). The products were analyzed by EMSA, as described under "Materials and Methods." Complex I was a protease-resistant complex. These representative data show the results from one of the three experiments. DR4 (DR4 and M2) (lanes 2 and 10) than on inactive DR4 (UUU and M3) (lanes 6 and 14). Complex I remained even after treatment of 80 g/ml trypsin (data not shown). But these bands were not observed in samples using control reticulocyte lysate, TR, or RXR alone (data not shown).
These results suggest that the majority of TR and RXR can form heterodimers with each other on each DR4 element under the conditions used and that protease-resistant products from TR/RXR heterodimer can maintain the heterodimer formation and DNA binding activities. Further, active DR4 may possess a conformation of TR/RXR heterodimer that is different from that of the inactive DR4.
Formation of TR␣/RXR␤ Heterodimer Induces a Conformational Change in TR␣-To examine the effects of T 3 and heterodimerization with RXR on possible structural changes of TR␣ in solution or on DNA, in vitro translated [ 35 S]methioninelabeled TR␣ was incubated with reticulocyte lysate or unlabeled RXR␤, with or without T 3 , in the presence of excess amounts of Pit 1 or DR4 probe, and then treated with increasing amounts of trypsin. The products were analyzed by SDS-PAGE (Fig. 6). Pit 1 oligonucleotides were used as nonspecific DNA controls that did not bind either TR or TR/RXR heterodimer by EMSA and ABCD assay, as described in the previous results.
Trypsin digestion produced 28 -30 kDa products which were increased by the addition of T 3 (lanes 3-5), indicating that T 3 increased proteolysis resistance of TR␣ in solution. Note again that in vitro translation of rat TR␣ produced two major proteins (Fig. 6, lane 1). Coincubation of TR␣ and RXR␤ produced a 46-kDa TR product after partial trypsin digestion (intermediate product), as indicated by asterisks (lane 7). This product was not observed when TR␣ alone was incubated with trypsin even in the presence of T 3 (lane 3). The addition of DR4 showed a shift to the right of the dose-response to partial trypsin digestion (an increased resistance to proteolysis, comparing intermediate products in lane 3 with those in lanes 7 and 11).
These results suggest that T 3 induces a conformational change in TR␣ in solution. These results also suggest that heterodimerization with RXR produces conformational change in TR, which is enhanced by binding of T 3 and DNA.

Binding to Active DNA Elements Increases Heterodimerization-induced Conformational Change of TR␣ in the TR␣/RXR␤
Heterodimer Complex-To analyze structural changes of TR in the TR/RXR heterodimer on DNA in depth, in vitro translated [ 35 S]methionine-labeled TR␣ and cold RXR␤ were incubated with different DNA elements, with or without T 3 , and then treated with increasing amounts of trypsin. The results for Pit 1, DR4, and UUU are shown in Fig. 7A, and those for Pit 1, M2, and M3 are depicted in Fig. 7B.
In the absence of T 3 , addition of mutated DR4s as well as DR4 showed increased protease resistance, by comparing the intermediate products in lane 3 with those in lanes 7 and 11. Incubation with T 3 increased protease resistance with each DNA element (contrast the upper panels with the lower panels). Comparing the effect of active DNA and inactive DNA, addition of active DNA, DR4, and M2 showed a shift to the right of the dose response to partial trypsin digestion (compare lanes 8, 9, and lanes 12, 13, significant alteration of T 3 induced protease resistance). The differences of TR trypsin sensitivity between active and inactive DR4 were statistically significant as shown by regression analyses (p Ͻ 0.05 to 0.01, Table I).
These results suggest that binding to DNA induces heterodimerization-mediated conformational change of TR␣ in TR/ RXR heterodimer complex both in the absence or presence of T 3 . Interestingly, active DR4 elements significantly alter this conformational change of TR␣ in TR/RXR heterodimer in the presence of T 3 .
Binding to DR4 Elements Produced Conformational Change of RXR␤ in TR␣/RXR␤ Heterodimer-Next, the effects of binding of the RXR␤ in TR␣/RXR␤ heterodimer to different DNAs on conformational change were examined (Fig. 8). The results for Pit 1, DR4, and UUU are shown in Fig. 8A, and those for Pit 1, M2, and M3 are in Fig. 8B. As shown in Fig. 5B, in vitro translation of mouse RXR␤ produced two major proteins (lane 1). Partial trypsin digestion produced a 46-kDa intermediate product (lanes 7, 8, 11, and 12). Although this product was already observed when RXR alone was incubated with trypsin, addition of TR increased protease resistance of RXR in TR/RXR heterodimer (data not shown). Addition of mutated DR4s as well as DR4 increased protease resistance (compare the intermediate product in lane 3 with those in lanes 7 and 11). In contrast to TR, incubation of T 3 did not increase protease resistance with each DNA element. But comparing the effect of active DR4 (DR4 and M2) and inactive DR4 (UUU and M3), active DR4 increased protease resistance better than inactive DR4 (compare lane 8 with lane 12). The differences of RXR trypsin sensitivity between active and inactive DR4 were statistically significant as determined by regression analyses (p Ͻ 0.05 to 0.01, Table II).
These results suggest that DNA binding to TR/RXR heterodimer may produce conformational changes in both TR and RXR in the TR/RXR⅐DNA complex. In particular, active DR4 elements, such as DR4 and M2, significantly alter heterodimerization-induced conformational changes of TR␣ in TR/RXR heterodimer in the presence of T 3 .

DISCUSSION
Two important processes involved in the T 3 -mediated activation of gene expression are T 3 binding to TR, and binding of the TR complex to TRE (2,4). However, although TR/RXR heterodimer appears to play an important role in this process (3, 4, 7, 8), the molecular basis of its action on a TRE is yet unclear. Here, we provide evidence that: 1) liganded TR/RXR heterodimer binding to DNA element itself is not sufficient for transcriptional activation of the target gene; 2) in contrast, unliganded TR/RXR heterodimer binding to DNA is sufficient for basal repression; and 3) DNA sequences in TREs may modify T 3 -mediated transcription by affecting the conformation of the liganded TR/RXR heterodimer.
First, we produced TR heterodimer-specific DR4 elements, by mutating the primary half-site, flanking and spacer sequences of DR4 (Fig. 1). Among these mutated DR4 elements, TR/RXR heterodimer showed similar binding and dissociation rates on UUU, M2, and M3 elements as compared to those on DR4 (Figs. 2 and 4A and data not shown). Surprisingly, T 3 could stimulate TR-mediated transcription on DR4 and M2 (active DR4s), but not on UUU and M3 (inactive DR4) in transient transfection studies (Fig. 3). In contrast, unliganded-TR␣ mediated repression of basal transcription on all mutated DR4 reporter genes. These data suggest that the TR/RXR heterodimer is correlated with and may be sufficient for basal repression. Also, TR homodimer is not necessary for basal repression, although TR homodimer may play a role in basal repression in the absence of T 3 and derepression in the pres-ence of T 3 (7). These data also support the idea that there are different pathways for TR-mediated transactivation and silencing (38).
One might argue that the TR monomer may also stimulate transcription on DR4 and M2 because TR monomer also binds to DR4 and M2 (Fig. 2). Indeed, several groups have reported transactivation by the TR monomer (15,39). However, it is unlikely that the TR monomer is involved in transcriptional activation on DR4 and M2. Previously, we identified TR monomer binding sites on DR4 which spans 10 nucleotides, including the spacer and the downstream half-site (20). Both M2 (active DR4) and M3 (inactive DR4) conserve this TR monomer contact sequence (Fig. 1). Further, the chemical DNA modification interference assay revealed that T 3 did not affect TR monomer contact sites on DR4, but caused weak contact of TR/RXR on no. 2 guanine site in the upstream half-site of DR4 (20). 3 On the other hand, a transfection study using TR␤, which does not bind to these mutated DR4 elements as a monomer, 3 (20,30, and 40 g/ml) (see Fig. 7). Comparison of regression slopes between active DR4 (DR4 and M2) and inactive DR4 (UUU and M3) in three to six experiments was performed by two sample t test. Significant differences were observed between active and inactive DR4 elements (p Ͻ 0.05 to 0.01). The regression equation (y ϭ bx ϩ a) was as follows: y; the amount of the intermediate products, showed similar results to those of TR␣. Therefore, the TR/RXR heterodimer may more likely be involved in transcriptional activation on DR4 and M2. Ligand plays an important role in nuclear receptor-mediated gene expression (1). Studies of steroid hormone receptors have shown that ligand is needed for receptor dissociation from hsp90, and that the formation of stable homodimer, which binds to the HRE, is important for function (40). It is also reported that antagonist/receptor complex can bind to HRE (41,42). These reports suggest that certain ligand-bound receptor-DNA complex may be critical for transcriptional activation. In the case of TR, "liganded" TR/RXR heterodimer binding to DNA element itself is not sufficient for transcriptional activation of the target gene, because T 3 binding to TR/RXR heterodimer is similar on active DR4 and inactive DR4 elements, according to the T 3 binding study (Fig. 4). We hypothesize that different response elements induce distinct DNA-dependent structural changes within TR/RXR heterodimers.
We performed limited proteolytic digestion studies employing SDS-PAGE and EMSA to examine the putative conformational changes of TR, induced by heterodimerization, T 3 , and DNA binding. First, the trypsin digestion EMSA study indi-cates that active DR4 elements allows a different trypsin access of the TR/RXR heterodimer compared to inactive DR4 elements (Fig. 5C). Although this method could not distinguish receptors in the TR/RXR heterodimer which caused the conformational change, the final trypsin-resistant products still retained heterodimerization and DNA binding activities according to EMSA. This suggests that trypsin cleavage site(s) of TR or RXR could be present in the N-terminal portion or the distal Cterminal portion of the receptors. The trypsin digestion study using SDS-PAGE further strengthens the finding of the different sensitivities of the DNA-dependent trypsin digestion pattern of TR/RXR heterodimer. It provides the following observations: 1) T 3 induces a conformational change in TR␣ in solution; 2) heterodimerization with RXR␤ produces a new trypsin-resistant intermediate product of TR␣ in TR␣/RXR␤ heterodimer, although the cleavage site(s) of this intermediate product is not known; 3) T 3 enhances conformational changes of TR␣, but not RXR␤, in the heterodimer-DNA complexes; 4) binding of DNA elements induces conformational changes of both TR␣ and RXR␤ in the heterodimer; and 5) T 3 significantly alters active DR4-mediated conformational change of TR in the heterodimer.   (20,30, and 40 g/ml) (see Fig. 8). Comparison of regression slopes between active DR4 (DR4 and M2) and inactive DR4 (UUU and M3) in three to six experiments was performed by two sample t test. Significant differences were observed between active and inactive DR4 elements (p Ͻ 0.05 to 0.01). A series of regression equation were calculated as described in Table I Recently, Kurokawa et al. reported that different spacer sizes in direct repeat type RAREs may alter the polarity of RAR/RXR heterodimer binding (22). In our study, the possibility that the putative different conformations of TR/RXR on active DR4 and inactive DR4 reflect the opposite polarity of TR and RXR on these two TREs is not completely excluded. However, this situation is unlikely inasmuch as all mutated DR4 have a conserved 4-nucleotide spacer, and perfect downstream half-site sequences.
It has already been shown that T 3 binding to TR induces conformational changes in the receptor, as measured by EMSA (8,9), antibody supershift EMSA (43), circular dichroism (34), or protease digestion assays (28,35,44,45). However, effects of heterodimerization with RXR and DNA binding on conformational changes in TR or RAR are still controversial. Keidel et al. (44) reported no effects of RXR and DNA on RAR conformation. Bendik and Pfahl (45) proposed similar ligand-induced conformational changes of TR in TR homodimer and TR/RXR heterodimer in the presence or absence of DNA. On the other hand, Leng et al. (28) reported that heterodimerization with RXR can enhance ligand-dependent conformational changes in TR or RAR/RXR heterodimer, but reported no major effects of DNA. A few groups have reported the effects of DNA on TR conformational change (34,46). However, there have been no reports about the effects of DNA on the conformation of the TR/RXR heterodimer, which may correlate with transcriptional activation. On the contrary, some groups have reported that TR or TR/RXR or TRAP produce DNA bending (47)(48)(49). Previously, we showed a different DNA-receptor contact pattern between TR homodimer and TR/RXR heterodimer (20). These observations support the idea that the liganded-TR/RXR⅐DNA complex possesses a distinct conformation to support transcriptional activation.
Recently, evidence of cellular factors that can interact with nuclear hormone receptors in a ligand-dependent manner has accumulated (50 -56). TR has been shown to bind directly to components of the basal transcriptional machinery such as TFIIB (57,58) and TBP (58). On the other hand, naturally occurring TREs have a wide variety of flanking, spacing and core half-site sequences. Rhodes et al. (59) demonstrated that TR␤ homodimer does not bind, but TR␤/RXR␣ heterodimer can bind, to the AGTTCA motif of DR4 elements in mouse Pit-1 gene. However, this element functions as a RARE and does not respond to T 3 . This observation and our results strongly support the idea that specific sequences of DNA elements can modify ligand-dependent conformational change in TR/RXR heterodimer, which then might permit the interaction with other cellular factors or component(s) of the basal transcriptional machinery, and transcriptional activation.
In conclusion, in addition to the orientation of the two halfsites and the spacer size in a TRE (9 -12), the polarity of TR/RXR heterodimer (11,13,(21)(22)(23), and DNA-mediated conformational change in the TR/RXR heterodimer can be important parameters that may determine the T 3 -mediated transcriptional activation.