The Differential Hormone-dependent Transcriptional Activation of Thyroid Hormone Receptor Isoforms Is Mediated by Interplay of Their Domains*

Human thyroid hormone nuclear receptor isoforms (TR (cid:97) 1 and TR (cid:98) 1) express differentially in a tissue-spe-cific and development-dependent manner. It is unclear whether these two isoforms have differential functions. We analyzed their interaction with a thyroid hormone response element with half-site binding motifs arranged in an everted repeat separated by six nucleotides (F2). Despite extensive sequence homologies, the two isoforms bound to F2 with different affinities and ratios of homodimer/monomer. Using F2-containing reporter gene, we found that the transcriptional activity of TR (cid:98) 1 was (cid:59) 6-fold higher than that of TR (cid:97) 1. The lower activity of TR (cid:97) 1 was not due to differences in expression of the two isoforms because similar nuclear localization pat- terns were observed. To understand the structural de-terminants responsible for these differences, we con- structed chimeric receptors in which hinge regions (domain D), hormone binding domains (domain E), and domains (D (cid:49) E) were sequentially interchanged and their activities were compared. Chimeric TRs contain- ing the domains D, E or (D (cid:49) E) of TR (cid:98) 1 showed in-creased propensities to form homodimers and mediated higher transactivation activities than TR (cid:97) 1. Thus, differential transactivation activities of TR isoforms are mediated by interplay of their domains and could serve as an important regulatory mechanism to achieve diver-

Human thyroid hormone nuclear receptor isoforms (TR␣1 and TR␤1) express differentially in a tissue-specific and development-dependent manner. It is unclear whether these two isoforms have differential functions. We analyzed their interaction with a thyroid hormone response element with half-site binding motifs arranged in an everted repeat separated by six nucleotides (F2). Despite extensive sequence homologies, the two isoforms bound to F2 with different affinities and ratios of homodimer/monomer. Using F2-containing reporter gene, we found that the transcriptional activity of TR␤1 was ϳ6-fold higher than that of TR␣1. The lower activity of TR␣1 was not due to differences in expression of the two isoforms because similar nuclear localization patterns were observed. To understand the structural determinants responsible for these differences, we constructed chimeric receptors in which hinge regions (domain D), hormone binding domains (domain E), and domains (D ؉ E) were sequentially interchanged and their activities were compared. Chimeric TRs containing the domains D, E or (D ؉ E) of TR␤1 showed increased propensities to form homodimers and mediated higher transactivation activities than TR␣1. Thus, differential transactivation activities of TR isoforms are mediated by interplay of their domains and could serve as an important regulatory mechanism to achieve diversity and specificity of pleiotropic T 3 effect.
Thyroid hormone receptors (TRs) 1 are the products of two genes, TR␣ and TR␤, located on chromosomes 17 and 3, respectively. Alternate splicings of their primary transcripts produce isoforms of the protein (␣1, ␣2, ␤1, and ␤2), which regulate the transcription of their target genes by binding to specific DNA sequences, known as thyroid hormone response elements (TREs). These contain repeats of a half-site binding motif with the sequence AGGTCA. Naturally occurring TREs can include these sequences as adjacent palindromic repeats, as direct repeats separated by 4 nucleotides, and as everted repeats sep-arated by 6 nucleotides (F2) (1,2). The sequences of TRs have been divided into four separate domains, A/B, C, D, and E. Domain C contains two zinc fingers and is involved in binding of the receptors to TREs. Domains D and E are structurally linked, in so far as part of domain D is required for the biological function of domain E, which is to bind thyroid hormones (3). Domains D and E are also involved in binding to co-repressors and dimerization, respectively (4). The crystal structures of TRE-bound domains C of TR␤1 and the retinoid X receptor (5) and of domains D/E complexed with a thyroid hormone agonist (6) have recently been solved. These structures give important information on interaction within domains but reveal nothing about the modes and roles of the interaction between domains in intact receptors, which may have important biological significance.
Comparison of the sequences between the human TR␣1 (w-TR␣1) and human TR␤1 (w-TR␤1) indicates that except domain A/B, there is extensive sequence homology between the two isoforms, specifically 88% in domain C, 71% in domain D, and 86% in domain E. Despite this high sequence homology, biochemical evidence suggests that they could have isoformspecific roles in mediating the action of thyroid hormones. TR␣ and ␤ genes are expressed at different stages during embryonic development (7,8) and during amphibian metamorphosis (9). Moreover, these two isoforms are expressed differentially in different tissues (8,10). More direct evidence to support the isoform-specific functional role of the TR␣1 and TR␤1 was provided by using gene transfer experiments. Strait et al. (11) showed that the gene encoding PCP-2 is regulated by TR␤1 but not by TR␣1. The 3,3Ј,5-triiodo-L-thyronine (T 3 )-dependent negative regulation of thyroid tropin releasing hormone promoter was shown to be mediated by TR␤1 but not by TR␣1 (12). Recently, using stably transfected neuronal cell line, Lebel et al. (13) showed that only cells that overexpress TR␤1, but not TR␣1, can respond to T 3 to exhibit morphological and functional characteristics indicative of neural differentiation.
At present, the molecular basis of isoform-specific gene regulation is not understood. It was suggested that the different homodimerization potentials of the two isoforms may underlie the functional differences. TR␤1 is known to bind to F2 and the TRE site on cardiac ␤-myosin heavy chain mainly as a homodimer, whereas TR␣1 forms homodimer poorly (14,15). These differences, however, are not eliminated by removal of A/B domains from the molecules (16) and consequently must arise from the remainder of the receptors. However, it is not clear that they are a consequence only of differences in sequence. They may also be caused by changes in interactions between domains in the intact receptors. Because of their marked effects on the properties of the receptors as transcrip-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ tion factors, we have investigated their origins by construction of a series of six chimeric receptors, in which domains A/B/C, D, and E from the two isoforms are joined in all possible combinations. We have measured their affinities for T 3 under identical conditions and their binding to an F2 TRE. We also determined the T 3 -dependent transcriptional activity of the wild type and chimeric receptors. We found that the domains C, D, and E are functionally linked, and the differential transcriptional activity of the two isoforms is mediated by interplay of their domains. Construction of Plasmids Encoding Chimeric TR␣1 and TR␤ 1-The T7-expression plasmids of the six chimeric receptors (see Fig. 2) were derived from the T7 expression plasmids of w-TR␣1 (pLC13) (17) and w-TR␤1 (pCJ3) (18). For cloning purpose, two restriction enzyme sites, NsiI and BamHI, were introduced into the boundaries between domains C and D and domains D and E of w-TR␣1 (nucleotide positions 412-414 (AAG/Lys to AAA/Lys) and 616 -621 (GGCAGC/Gly-Ser to GGATCC/ Gly-Ser)) to yield a new T7 expression plasmid of w-TR␣1, pCH␣. Only one restriction site, BamHI, was required to introduce into the boundary between domains D and E of w-TR␤1 (nucleotide positions 991-996 (GGCAGC/Gly-Ser to GGATCC/Gly-Ser)) because in pCJ3 (w-TR␤1) the NsiI already existed which yielded a new T7 expression plasmid of w-TR␤1, pCH␤. The introduction of these restriction sites was carried out by in vitro mutagenesis kit according to the manufacturer's instructions (Bio-Rad). The introduction of these two new restriction sites into TR␣1 and the BamHI site into TR␤1 did not change the amino acid sequences of TR isoform proteins. The six chimeric receptors were constructed by exchanging the domains between w-TR␣1 and w-TR␤1 using NsiI, BamHI, and the 3Ј EcoRI site immediately downstream of the termination codons of w-TR␣1 (nucleotide position 1306 for TR␣1 and 1672 for TR␤1 (pCJ3)) to yield T7 expression plasmids pCH1, pCH2, pCH3, pCH4, pCH5, and pCH6 for ␤␣␤, ␤␤␣, ␤␣␣, ␣␣␤, ␣␤␣, and ␣␤␤, respectively (see Fig. 2). The coding sequences for the six chimeric receptors were verified by restriction map analyses and direct DNA sequencing.
The mammalian expression plasmids of the TR␣1 and TR␤1 chimeric receptors were derived from the corresponding w-TR␣1 and w-TR␤1 expression plasmids, pCLC61 and pCLC51 (19), respectively. The expression of w-TR␣1 and w-TR␤1 is driven by cytomegalovirus promoter. To prepare the mammalian expression plasmids of chimeric receptor of TR␤1, pCLC51 was restricted by NotI followed by filling in with Klenow in the presence of deoxynucleotides. The coding sequence of the w-TR␤1 in pCLC51 was then released by treating the linearized and bluntended pCLC51 with HindIII, thereby providing the vector for ligation to the proper chimeric TR coding fragments. The wild type and chimeric TR␤1 coding fragments were derived from the above T7 expression plasmids (pCH␤, pCH1, pCH2, and pCH3) by treating the plasmids with EcoRI. After filling in, the fragments were released by treating with NdeI. An adaptor (HindIII/NdeI) was used in the final ligation of TR coding fragments to the vectors to yield plasmids pCDMCH␤, pCDMCH1, pCDMCH2, and pCDMCH3 for W-TR␤1, ␤␣␤, ␤␤␣ and ␤␣␣, respectively. The mammalian expression plasmids of w-TR␣1 and its chimeric receptors were prepared similarly except that the vector was derived from pCLC61. The resulting mammalian plasmids were pCDMCH␣, pCDMCH4, pCDMCH5, and pCDMCH6 for w-TR␣1, ␣␣␤, ␣␤␣, and ␣␤␤, respectively.
Electrophoresis Gel Mobility Assay (EMSA)-The probe, F2, was 32 Plabeled similarly as described (20). Briefly, two complementary oligonucleotides containing the F2 sequences as shown in Sequence 1 below, were annealed and the recess 3Ј-end filled with DNA polymerase (Klenow fragment) in the presence of [␣-32 P]dCTP. The labeled oligonucleotides were separated on a 12% polyacrylamide gel and purified by electroelution.
For EMSA, unlabeled TRs synthesized by in vitro transcription/ translation were used. The synthesized receptor proteins were quantified by measuring the intensity of the 35 S-labeled protein bands after SDS-polyacrylamide gel electrophoresis using PhosphorImager (Molecular Dynamics, CA). The 35 S-labeled protein was synthesized concurrently by using amino acid mixture minus methionine but with [ 35 S]methionine (4 l; 1190 Ci/mmol). Based on the quantitation of the labeled receptors, the amounts of the unlabeled receptors were calculated. For the determination of the binding constants of TRs to F2, equal amounts of the in vitro translated unlabeled receptors were incubated with increasing concentrations of the labeled probes (0.2-120 fmol) in the binding buffer (25 mM Hepes, pH 7.5, 5 mM MgCl 2 , 4 mM EDTA, 10 mM dithiothreitol, 0.11 M NaCl, and 0.4 g of single-stranded DNA). In some experiments, RXR␤ prepared as described by Meier et al. (20) was added. After incubation for 30 min at 25°C, the reaction mixture was loaded onto a 5% polyacrylamide gel and electrophoresed at 4°C for 2-3 h at a constant voltage of 250 V. The gel was dried and autoradiographed. The intensities of retarded bands and free probes were quantified by PhosphorImager. The binding data were analyzed based on the equations and considerations as described below.
Analysis of Binding Data from EMSA-Binding of glucocorticoid nuclear receptors to response elements with adjacent identical halfsites has been successfully analyzed using a simple two-site cooperative model, which ignored dimerization of free receptors in solution (21). Since dimerization of unbound TR's has never been detected, we interpret our results in a similar way (Equation 1). Receptor (R) can bind to either TRE half-site (D) to give monomeric complexes (DR and RD) or to both yielding a dimeric complex (RDR).
We assume that a receptor molecule can bind to either half-site on an empty TRE (D) with a binding constant K 1 and on a monomeric complex (DR or RD) with a binding constant K 2 . If K 2 ϭ s⅐K 2 , then s is the cooperativity parameter. Positive cooperativity implies s Ͼ 1, i.e. stronger binding of the second TR than the first to the TRE. The concentration of monomeric complexes and of dimeric complexes The total concentration of TRE, The concentrations of monomer, dimer, and TRE are measured on the gel from the known specific activities of the DNA probes. However, the recombinant TRs are produced in cell lysates. It was not possible to determine how much of the protein in each lysate was intact, competent TR, i.e.
[R] is unknown. In each experiment, in each lane, we can measure a ratio,

Substituting Equation 8 into 5 and 6 gives
For each combination of recombinant TR and TRE, values of K 1 and K 2 were estimated by fitting the measured concentrations [monomer] and [dimer] simultaneously to Equations 9 and 10 as functions of [D] 0 and r, with the constraints K 1 Ͼ 0, K 2 Ͼ 0. Analyses were performed using the PC-MLAB program (Civilized Software, Bethesda, MD). It must be pointed out that this procedure violates one basic assumption of least squares curve fitting, i.e. that experimental uncertainties in plotted data parallel the y axis (22). Here we have uncertainties along both axes. Together with the problems of the gel retardation method, which requires separation of reactants and products, perturbing the system from equilibrium, as discussed previously (16), could result in some uncertainty in K 1 and K 2 . Consequently, the values given in Table  I (3).
The binding data were analyzed by using Equation 11 based on direct competition between [ 125 I]T 3 and the unlabeled T 3 for a single site on the receptor. The concentration of radioactive complex is given by Equation 11: Transient Transfection Assay-CV1 cells (4 ϫ 10 5 cells/60-mm dish) were plated 24 h before transfection in Dulbecco's modified essential medium containing 10% fetal bovine serum. Cells were transfected with appropriate expression plasmids (0.2 g) for w-TR␤1 (pCDMCH␤), w-TR␣1 (pCDMCH␣), or the chimeric receptors (pCDMCH1, pCDMCH2, pCDMCH3, pCDMCH4, pCDMCH5, and pCDMCH6), TRE-containing TK-CAT reporter plasmid (0.2 g), and pCH110 (0.2 g; an expression plasmid for ␤-galactosidase) by using the lipofectamine transfection method according to the manufacturer's procedure (Life Technologies, Inc.). pBluescript SK II (ϩ) Strategene, La Jolla, CA) was used to bring the total DNA transfected to 3 g. After 6 h, the medium was replaced by fresh Dulbecco's modified essential medium containing 10% thyroid hormone-depleted serum. Fifteen hours before cells were harvested, T 3 (100 nM) was added to the appropriate dishes. After an additional 18 h, cells were lysed and chloramphenicol acetyltransferase (CAT) activity was determined as described previously (23,24). CAT activity was normalized by using equal amounts of lysate proteins.
Immunocytolocalization of TR Proteins by Immunofluorescence-Cultured CV-1 cells were transfected as described above. Two days later, cells were processed for immunofluorescence studies as described previously (25). Briefly, cells were fixed in 3.7% formaldehyde in phosphate-buffered saline for 5 min at 25°C. After washing, cells were incubated with monoclonal antibody C4 (10 g/ml; 26) in phosphatebuffered saline containing 0.1% saponin and 4 mg/ml normal goat globulin for 30 min at 25°C. After being washed with phosphatebuffered saline, cells were incubated with affinity-purified goat antimouse immunoglobulin conjugated with rhodamine (25 g/ml) for 30 min at 25°C. Cells were viewed and photographed using microscope equipped with rhodamine epifluorescence optics.
Western Blotting-Cell lysates (25 g) from transient transfection experiments as described above were loaded onto a 10% SDS gel. After electrophoresis, proteins were transferred onto a nitrocellulose membrane (PH79 membrane; Schleicher & Schuell). The membrane was gently shaken in 5% non-fat milk in TBS (25 mM Tris, pH 7.4, 150 mM NaCl) for 20 h and was subsequently washed three times with TBS. The membrane was incubated with monoclonal antibody C4 (1 g/ml) for 1 h. After washing, the membrane was incubated with affinity-purified rabbit anti-mouse immunoglobulin conjugated with horseradish peroxidase (1:1,000 dilution). TR protein bands were visualized by chemiluminescence using ECL kit (Amersham Life Sciences, Inc.).

Differential Interaction of TR␣1 and TR␤1 with F2-Previ-
ously it has been shown that TR␤1 binds to F2 mainly as a homodimer, whereas TR␣1 binds to F2 both as a homodimer and as a monomer (14,15). However, there was no quantitative comparison in the differential binding of F2 to the two isoforms. We therefore compared the binding affinities of F2 to the two isoforms. Fig. 1 shows the binding of TR␣1 and TR␤1 to F2 in a concentration-dependent manner. Consistent with previous observations (14,16), TR␤1 bound to F2 predominantly as a homodimer. Interestingly, when F2 concentration was higher than 15 fmol, weak binding of TR␤1 to F2 as a monomer was clearly detected (lanes [13][14][15][16]. However, TR␣1 bound to F2 differently from TR␤1. As shown in Fig. 1 The binding data shown in Fig. 1 were analyzed, and the K a values of homodimeric (K 2 ) and monomeric (K 1 ) binding for TR␤1 were found to be 400 and 0.1 ϫ 10 6 M Ϫ1 , respectively, indicating an increase of 4000-fold in the binding affinity when TR␤1 was bound to F2 as a homodimer (s ϭ 4000; Table I). Thus, binding of the first monomer of TR␤1 to F2 facilitated the binding of the second monomer. We designated "s" as the ratio of K 2 /K 1 to measure the extent of positive cooperativity in the binding of TR to TREs. The K a values of homodimeric and monomeric binding of TR␣1 to F2 were 300 and 3 ϫ 10 6 M Ϫ1 , respectively, which gave a substantially lower cooperativity (s ϭ 100; see Table I) than that for TR␤1.
Role of Domains in the DNA and T 3 Binding Activity of TR Isoforms-To identify the molecular basis of the differential interaction of TR␣1 and TR␤1 with F2, we interchanged the domains between the two isoforms and evaluated the effects of domain swapping on the F2 and T 3 binding activity. An examination of the sequences between the two isoforms indicates that there is no sequence homology in the A/B domain, whereas there is an 88, 71, and 86% homology in sequence in domains C, D and E, respectively (Fig. 2I). We have previously shown that the removal of domain A/B has no effect on the interaction of TR␤1 with TREs (16). Therefore, we grouped domain A/B together with domain C as a unit and constructed the chimeric receptors by swapping domains A/B/C, D, and E (Fig. 2II). The sequences encoding the chimeric receptors in the constructs were confirmed by restriction map analyses and DNA sequencing.
To assess the T 3 binding activity, we prepared the receptors by in vitro transcription/translation and carried out competitive T 3 binding assays. The displacement curves for w-TR␤1 and its chimeric receptors are shown in Fig. 3A and for w-TR␣1 and its chimeric receptors are shown in Fig. 3B. Binding data were analyzed, and the K d values are shown in Table II. The K d values for the binding of w-TR␤1 and w-TR␣1 to T 3 were 0.36 Ϯ 0.06 and 0.10 Ϯ 0.037 nM, respectively, indicating that w-TR␣1 bound to T 3 with an approximately 3-fold higher affinity than that of w-TR␤1. The 3-fold difference is very significant as indicated by the t test (p Ͻ 0.01). The difference in the binding affinity was not due to the different protein expression level by the in vitro transcription/translation. As shown in Fig. 3C, lane 2 shows the two translation products of w-TR␤1 initiated from the ATGs (Met-5 and -32) with the molecular weights of ϳ55,000 and ϳ52,000 (26, 27) that have the combined intensity similar to that of w-TR␣1 shown in lane 6 ( Fig. 3C). Similar binding experiments were carried out for the chimeric receptors, and as shown in Fig. 3, A and B, no significant differences were observed in the binding curves within the same subtype. The K d values for the chimeric receptors are virtually identical to those of the wild type receptors (Table II), indicating that the domain swapping between the two isoforms had no effect on the T 3 binding activity.
In contrast to the T 3 binding activity, domain swapping had a dramatic effect on the interaction of chimeric receptors with F2. Lanes 2-5 of Fig. 4 compare the binding of w-TR␤1 (␤␤␤; see Fig. 2) and its chimeric receptors to F2 by EMSA. Replacement of domains D or E of TR␤1 by that of TR␣1 had no significant effect on the binding of ␤␣␤ or ␤␤␣ receptor to F2 as a homodimer, but an increase in the formation of monomer was seen (lanes 3 versus 2; lanes 4 versus 2). However, when both domains D and E were swapped, a dramatic increase in the monomer formation was detected. The extent of monomer formation was similar to that seen for w-TR␣1 (␣␣␣, lane 6 of Fig.  4A versus lane 5). We further measured an F2 concentrationdependent binding to each chimeric receptor (␤␣␤, ␤␤␣, and ␤␣␣), similar to the experiments shown in Fig. 1, and determined their affinity constants. The binding data were ana-lyzed, and the K a values are shown in Table I. Swapping of domain D or E of TR␤1 by that of TR␣1 led to a 3-and 4-fold increase in the binding affinity of ␤␣␤ or ␤␤␣ to F2 as a monomer, respectively (K 1 ϭ 0.3 and 0.4 ϫ 10 6 M Ϫ1 , respectively, versus 0.1 ϫ 10 6 M Ϫ1 for w-TR␤1), but with little change in the binding affinity of these two chimeric receptors as a dimer (K 2 ϭ 500 ϫ 10 6 M Ϫ1 ). On the other hand, when both domains D and E were swapped, a dramatic 20-fold increase in monomer binding affinity (K 1 ϭ 2 ϫ 10 6 M Ϫ1 ) was detected. Thus, inclusion of domain D or E of TR␣1 facilitates the binding of TR as a monomer.
Lanes 7-9 of Fig. 4 show that replacement of either domain D or E alone or both domains D and E of TR␣1 by the corresponding regions of TR␤1 resulted in a similar reduction in the monomer formation ( lane 6 versus lanes 7-9). The ratios of monomer to homodimer were clearly reduced in ␣␣␤, ␣␤␣, and ␣␤␤. A more detailed analysis was carried out by determining the affinities in the binding of F2 to the chimeric TR␣s. Their K 1 , K 2, and s values are shown in Table I which indicate that there were only small changes in the values of positive cooperativity in ␣␣␤, ␣␤␣, and ␣␤␤ as compared with ␣␣␣ (s ϭ 50 -100).
RXRs have been shown to heterodimerize with TRs and modulate the activity of TRs (1,2). We therefore also examined the effect of domain swapping on the heterodimerization activity of the chimeric receptors. Similar to w-TR␤1 and w-TR␣1, all chimeric TRs were capable of forming dimers with RXR␤ on F2. No significant differences in the extent of formation of heterodimers were detected among the chimeric TRs (data not shown).
Role of Domains in the Differential Transactivation Activity of the Wild Type and Chimeric TRs-To assess the role of the domains in the transactivation activity of TRs, we constructed the mammalian expression vectors in which the expression of  the wild type and chimeric TRs was driven by the cytomegalovirus promoter. We co-transfected the TR expression plasmids with F2-containing reporter into CV1 cells. Fig. 5 shows that w-TR␤1 (␤␤␤) had a ϳ6-fold higher T 3 -dependent transactivation activity than w-TR␣1 (␣␣␣; bars 2 versus 6 of Fig. 5). The lower transactivation of w-TR␣1 was not due to the lower expression of w-TR␣1 proteins in CV1 cells. Using high titer monoclonal antibody C4 (26), we had concurrently carried out immunocytochemical localization of TRs in CV1 cells and Western blotting for quantitation of the expressed TRs. w-TRs and their chimeric TRs were similarly expressed in the nuclei (data not shown). Thus, the lower transactivation activity was not due to the inability of w-TR␣1 to be translocated into the nuclei. Furthermore, the Western blots show that w-TR␣1 (␣␣␣), surprisingly, was expressed ϳ2-fold higher than w-TR␤1 (lanes 6 versus lane 2 of Fig. 5), indicating that the lower transactivation activity was not due to the lower protein expression level of w-TR␣1.
To identify which domain in the TRs mediated the differential transactivation activity between these two isoforms, we further examined the transactivation activity of the chimeric TRs. On F2, swapping of domain D of w-TR␤1 by that of w-TR␣1 (bars 3 versus 2 of Fig. 5) reduced the transactivation activity of ␤␣␤ by 35%. As shown bars 4 and 5, the T 3 -dependent transactivation was reduced by ϳ60% when domain E alone or domains D ϩ E of w-TR␤1 was replaced by that of w-TR␣1. Swapping of domains E and D ϩ E in w-TR␣1 by the corresponding regions of TR␤1 led to a 1.8-(bar 8) and 2.5-fold (bar 9 of Fig. 5) increase in the transactivation in ␣␣␤ and ␣␤␤, respectively. The differences in the transactivation activity of the chimeric TRs were neither due to the differences in the ability of the chimeric TRs to be translocated into the nuclei because similar nuclear localization patterns were seen (data not shown) nor due to the TR expression levels because the transactivation activities shown in Fig. 5 were normalized against the amounts of proteins detected in Fig. 6. Taken together, these results indicate that domains D and E of TR␤1 had a propensity to mediate a higher transactivation activity, and those of TR␣1 mediated a lower transactivation activity.
To evaluate whether the differential transactivation activity of the two TR isoforms is mediated by TR/RXR heterodimer pathway, we co-transfected RXR␤ expression plasmid with F2-CAT reporter and w-TR␤1 or w-TR␣1 expression plasmids into CV1 cells. Consistent with the previous findings (24), the T 3dependent transactivation activity of w-TR␤1 mediated by F2  II Apparent affinity constants of the binding of w-TR␤1, w-TR␤1, and chimeric receptors to F2 Increasing concentrations of the 32 P-labeled F2 TRE were incubated with equal amounts of in vitro translated w-TR␣, w-TR␤1, or the chimeric receptor proteins. After EMSA, the intensities of the monomeric and hoomodimeric bands were quantified by PhosphoImager, k 1 and k 2 are binding constants for the association of a receptor molecule to a half/site on an empty TRE and on a monomeric complex. respectively. Cooperativity between the sites is measured by s(ϭk 2 /k 1 ). Their values were calculated according to the equations descdribed under "Experimental Procedures." was repressed ϳ60% by RXR␤. A similar extent of repression was also seen for the T 3 -dependent transactivation activity of w-TR␣1 by RXR␤ (data not shown). Therefore, the higher transactivation activity of w-TR␤1 was not due to the TR/RXR heterodimer pathway.

DISCUSSION
In an important study, Rastinejad et al. (5) recently determined the crystal structure of a heterodimeric complex of two proteins, made up from domains (C ϩ D) derived from the RXR and TR␤1, respectively, bound to a direct repeat TRE. The DNA in the complex is undistorted, with regular B-DNA geometry. Many amino acid side chains are involved in the TR␤1-DNA interactions. The specificity of binding is determined by side chains from the 11-residue long C domain "recognition helix," which starts at the third metal coordinating cysteine of the first zinc finger. These make direct contacts with the base pairs and backbone phosphates in the major groove of the half-site binding motif. Domain D contains a long ␣-helix, which makes extensive interactions with the minor groove between the half-site and upstream spacer sequence. The dimerization interface between RXR and TR␤1 lies across the minor groove of the spacer, involving mainly residues from the first zinc finger in TR␤1 and the second zinc finger in RXR. Of the many side chains identified as making DNA contacts, only one, K193R, is changed in TR␣1. Mutagenesis experiments indicate that at least one other conserved region of TRs, Leu-367-Leu-374, located in domain E, is involved in dimerization of intact receptors. The structure of the separate ligand binding domains D ϩ E of rat TR␣1 has recently been determined (6). The isolated protein is monomeric and gives no indication as to how this sequence, which forms "an extensive hydrophobic patch," participates in dimerization. The analogous sequence in human RXR-␣ does form a dimer interface in crystals of its isolated ligand binding domain (29). It has been suggested that this dimerization sequence from domain E has no selective pressure on response element recognition but only serves to stabilize these homodimer complexes (30), being active in all dimerization interfaces. Biochemical data, on the polarity of binding and the specificity for particular spacings in DNA response elements shown by heterodimers formed by various members of the steroid/thyroid hormone receptor family, were readily explained using the crystal structure solved by Rastinejad et al. (5). This indicates the generality of the binding mode which they detected and predicts that homodimers formed by TRs on the three different types of TREs will have distinct dimerization interfaces. Homodimers formed on F2 elements will be symmetrical, with a dimerization interface including the first zinc fingers of domain C and the D domain ␣-helices of both proteins. The situation is further complicated by the spatial arrangements of the binding sites. For DR4, the centers of the two binding motifs are on the same face of the DNA, one turn of the DNA helix apart (5). We can predict that for F2 they will fall a little further apart, on opposite faces of the DNA. Under most conditions, binding of TR␤1 to an F2 response element occurs as a dimer complex. Since the two half-sites involved are basically identical, the observed low levels of a 1:1 DNA/protein monomer complex indicates high positive cooperativity between the two half-sites (i.e. K 2 Ͼ Ͼ K 1 ). There are two extreme ways in which positive cooperativity between two intrinsically identical half-sites can be achieved (31). Relative to the empty TRE we can detect (i) stabilization of the dimer complex by a large positive free energy of interaction between both sites occupied by proteins, with no interaction between occupied and unoccupied sites, or (ii) destabilization of the monomer complex by a large negative free energy of interaction between occupied and unoccupied sites, with no interaction between two occupied sites. In general, less extreme situations, where both occupied-occupied and occupiedunoccupied interactions occur, must also be considered. The binding of TR␣1, TR␤1, and their chimeric receptors to F2 shows positive cooperativity (K 2 Ͼ Ͼ K 1 ). Table I clearly shows that the enhanced positive cooperativity of binding shown by TR␤1 over TR␣1 (i.e. its greater tendency to bind as a dimer) results mainly from the second of these causes. TR␤1 monomer complex is much less stable than the TR␣1 form, and this lower stability is relieved by formation of the dimer complex. As noted above, the sequences involved in DNA binding are essentially identical in TR␣1 and TR␤1. Consequently, the instability of the TR␤1 monomer complex must result from the overall structure of the receptor molecule and its effect on the binding interfaces. The data obtained with the chimeric receptors show that all proteins in which the DNA binding domain C is of the ␣ form show monomeric binding to F2, like TR␣1 (Table I, Fig.  4). However, exchange of only this domain to the ␤ form (␣␣␣ 7 ␤␣␣) is not sufficient to significantly enhance binding cooper- w-TR␤1 (pCDMCH␤), w-TR␣1(pCDMCH␣), or chimeric receptor expression plasmids (0.2 g; pCDMCH1, pCDMCH2, pCDMCH3, pCD-MCH4, pCDMCH5, or pCDMCH6) were co-transfected with the TRE-CAT reporter (0.2 g) and the ␤-galactosidase expression plasmid pCH110 (0.2 g) into CV-1 cells according to the methods described under "Experimental Procedures." Cell lysates were prepared, and the CAT activity was determined. The CAT activity was normalized to the protein concentration in the lysates. Data are expressed as mean Ϯ S.E. (n ϭ 6), each with duplicates.
FIG. 6. Expression of w-TR␤1, w-TR␣1, and their chimeric receptors in CV1 cells analyzed by Western blot. An aliquot of the lysates from cells transfected with plasmids as described in Fig. 5 were loaded to a 10% SDS-polyacrylamide gel and transferred to nitrocellulose PH79 membrane. The w-TR␤1, TR␣1, and chimeric receptor proteins were detected using C4 antibody and visualized by enhanced chemiluminescence.
ativity. In the case of F2, some destabilization of the monomer complex is detected when both domains C-D or C-E are exchanged, with a full effect with all three domains derived from ␤. In monomer complexes, these all fall on the occupied-unoccupied site interface, where the destabilization must occur. The physical origin of this destabilization interaction may only be revealed by determination of the structure of suitable complexes.
The hormone binding site of TRs is located in domain E. In earlier work, we showed that domain E, isolated from h-TR␤1 with part of domain D, can still bind T 3 but with reduced affinity (3). Addition of domains C and D restored the molecule's affinity to that of the intact receptor, indicating regulatory interactions between these domains. The results shown in Fig. 3 indicate that the chimeric receptors derived from TR␣1's domain C all have higher T 3 binding affinity than the chimeric receptors derived from domain C of TR␤1, at least in the absence of DNA (see Table II). Thus this aspect of the behavior of a chimeric receptor is determined exclusively by the origin of its domain C, the DNA binding domain, reinforcing the importance of interdomain interactions.
The results on the effect of domain swapping on the transcriptional activity of the TR isoforms revealed that despite a higher T 3 binding affinity, TR␣1 and its chimeric receptors had lower transcriptional activity (Fig. 5). This was unexpected, suggesting that the mode of DNA binding to TRs overrides the advantage gained from higher T 3 affinity. Inspection of the data further indicates that the extent of transcriptional activation by a receptor correlates better with the source of its domains than with its affinity for DNA. It is clear that domains D and E of TR␤1 tend to impart higher transcriptional activity than those of TR␣1. This in turn correlates well with the propensity of the chimeric receptors to form homodimers (see Fig. 4). The fact that w-TR␣1 and its chimeric receptors formed heterodimers with the RXR as well as w-TR␤1 and its chimeric receptors suggests that the higher transcriptional activity of TRs which contain the domains D and E of TR␤1 probably was not mediated by the heterodimer pathway. This notion is further supported by the findings that the transfected RXR repressed the T 3 -dependent transactivation activity of the two isoforms with similar extent. Therefore, this higher transcriptional activity of TR␤1 lies most likely in the interactions of domains D and E with domain C in homodimers. Genetic experiments have shown that the hormone-dependent transactivation activity depends on a short amphipathic ␣-helix at the extreme carboxyl terminus of domain E, which undergoes a large conformational change on hormone binding (6,26,30). The sequence is conserved in both isoforms but is probably located in different sequence contexts in relation to the DNA binding domain or in the context of the entire molecule. Thus, it may function with differing efficacies in the two different environments. Recently, several co-repressors and one co-activator for several members of the receptor superfamily including TRs have been reported (4,(32)(33)(34)(35). Their function has been proposed to act as bridging factors between the TRs and the basal transcriptional machinery (4). It is possible that the resultant tertiary structure of domains D and E of TR␤1 is less favorable to bind to a co-repressor. It is also possible that the structure of domains D and E of TR␤1 is such that the carboxyl-terminal ␣-helix is more easily accessible to a co-activator. Thus, domains D and E of TR␤1 may have a higher efficacy in transmitting the effects of conformational change of the carboxyl-terminal ␣-helix upon binding the hormone to regulate the interaction of the domain C with the target genes. These possibilities can only be distinguished when the x-ray crystallographic structures of ligand-bound intact TR␣1 and TR␤1 are solved and compared. Our present studies indicate that domains C, D, and E are functionally linked and the interplay of these domains underlines the differential transcriptional activity of the two isoforms.