Functional characterization of a mutant thyroid hormone receptor in Xenopus laevis.

Thyroid hormone plays a causative role during frog metamorphosis, and its effect is mediated by thyroid hormone receptors (TRs). To investigate the function of Xenopus TRs, we have recently developed a thyroid hormone dependent in vivo transcription system by introducing TRs and RXRs (9-cis-retinoic acid receptors) into Xenopus oocytes. Interestingly, using this system, we have found that the TRαB cloned previously is defective in transcriptional activation compared with TRαA. In vitro DNA binding experiments show that TRαB·RXR heterodimers have drastically reduced affinity for a thyroid hormone response element. Site-directed mutagenesis shows that two of the seven amino acid residues that differ between TRαA and TRαB are responsible for the defect in TRαB function. These two residues affect the DNA binding by both TR·RXR heterodimers and TR homodimers. In contrast, heterodimer formation with RXRs is not affected as demonstrated by coimmunoprecipitation and dominant-transcriptional inhibition experiments. By cDNA and genomic DNA sequence analysis, we have demonstrated that the residues, which affect TRαB function when mutated, are identical between the wild type TRαB and TRαA. Thus, our experiments have discovered the first amphibian TR mutant. The DNA binding and transcription activation functions of the mutant are discussed in relation to the recently published TR crystal structure.

Thyroid hormone plays a causative role during frog metamorphosis, and its effect is mediated by thyroid hormone receptors (TRs). To investigate the function of Xenopus TRs, we have recently developed a thyroid hormone dependent in vivo transcription system by introducing TRs and RXRs (9-cis-retinoic acid receptors) into Xenopus oocytes. Interestingly, using this system, we have found that the TR␣B cloned previously is defective in transcriptional activation compared with TR␣A. In vitro DNA binding experiments show that TR␣B⅐RXR heterodimers have drastically reduced affinity for a thyroid hormone response element. Site-directed mutagenesis shows that two of the seven amino acid residues that differ between TR␣A and TR␣B are responsible for the defect in TR␣B function. These two residues affect the DNA binding by both TR⅐RXR heterodimers and TR homodimers. In contrast, heterodimer formation with RXRs is not affected as demonstrated by coimmunoprecipitation and dominant-transcriptional inhibition experiments. By cDNA and genomic DNA sequence analysis, we have demonstrated that the residues, which affect TR␣B function when mutated, are identical between the wild type TR␣B and TR␣A. Thus, our experiments have discovered the first amphibian TR mutant. The DNA binding and transcription activation functions of the mutant are discussed in relation to the recently published TR crystal structure.
Thyroid hormone receptors (TRs) 1 belong to the superfamily of steroid hormone receptors (1)(2)(3)(4), which also include receptors for 9-cis-retinoic acid (RXRs), and glucocorticoid receptor, etc. Members of this family are transcription factors whose activity is regulated by their cognate hormones. They have similar structural domains. The DNA binding domain, located in the NH 2 -terminal half of the receptor, shares considerable homology among different receptors and recognize specific hormone response elements, e.g. TREs (thyroid hormone response elements) for thyroid hormone receptors. The hormone binding domain, on the other hand, is unique for each hormone receptor and is located in the carboxyl half of the protein. Similarly, the other domains are very divergent in sequence among different receptors.
While TR can bind to TREs weakly as homodimers and monomers, it binds TREs with much higher affinity as heterodimers with some other members of the superfamily, expecially RXRs (5)(6)(7)(8)(9). Furthermore, TR⅐RXR heterodimers can mediate specific gene regulation by thyroid hormone in tissue culture cells. In addition, RXRs are expressed in most, if not all, tissues. These results suggest that TR⅐RXR heterodimers are likely to be the active complexes in vivo to mediate the effects of thyroid hormone.
We are studying the regulation and function of TRs during amphibian metamorphosis. This postembryonic process is entirely controlled by thyroid hormone and systematically transforms different tissues of a tadpole (10,11). Thus, blocking the synthesis of endogenous thyroid hormone inhibits the process, whereas the addition of thyroid hormone to the rearing water of premetamorphic tadpoles induces precocious metamorphosis. Furthermore, the response to thyroid hormone is organ autonomous (10 -13), suggesting that thyroid hormone acts on individual organs directly through its receptors. Indeed, four TR (TR␣A, TR␣B, TR␤A, and TR␤B) genes have been cloned in Xenopus laevis (14,15) and found to be expressed in all metamorphosing tissues of a tadpole (16 -21). More importantly, by analyzing the mRNA levels for both TR and RXR genes in different organs during development, we have observed a coordinated regulation of TR and RXR genes (21). Both TR and RXR genes are temporally regulated in an organ-specific manner. High levels of their mRNAs are present in a given organ when it undergoes metamorphosis, while low levels are generally observed in the same organ before or after metamorphosis. These results provide strong evidence that TR⅐RXR heterodimers mediate the causative effects of thyroid hormone during metamorphosis.
To study the function of these amphibian TRs, we have recently developed an in vivo thyroid hormone-responsive transcription system (21,22). This is achieved by taking advantage of the fact that Xenopus oocytes have only very low levels of TRs and RXRs that are incapable of activating a TRE-containing promoter (20 -22). In addition, it is easy to introduce exogenous receptors by injecting their mRNAs into the oocyte cytoplasm. Thus, when a single-stranded reporter plasmid containing a promoter with a TRE is injected into the oocyte nucleus, the DNA is replicated and concurrently assembled into chromatin in the presence of these receptors, thus allowing functional characterization of the receptors in a chromatin environment (23). Using such a system, we have found that both Xenopus TR␣A and TR␤A can activate efficiently a TREcontaining promoter as heterodimers with either Xenopus RXR␣ and RXR␥ (22). We report here that to our surprise, Xenopus TR␣B functions poorly as an activator. Furthermore, * 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  the TR␣B⅐RXR heterodimer has a drastically reduced affinity for a TRE. We demonstrate that the defect in TR␣B function is the result of mutations in the D domain located between the DNA and hormone binding domains in the previously reported TR␣B sequence, which was apparently derived from a mutant TR␣B gene. Our experiments discovered that two amino acid residues in the D domain are critical for DNA binding, but not for heterodimerization or transcriptional activation.

MATERIALS AND METHODS
Construction of the Mutants-The coding regions of Xenopus TR␣A and TR␣B (15) were cloned into the pSP64 vector (Promega) through a series of restriction and ligation. Mutants I-IV were made by switching domains between TR␣A and TR␣B cloned into the pSP64 vector. To construct mutant I, a DNA fragment was restricted out of TR␣A with EcoRI (site present in the vector) and NsiI (in the insert) and cloned into similarly restricted TR␣B vector. Mutant II was constructed by inserting the NheI (in the vector)/AflII (in the insert) fragment of TR␣A plasmid into TR␣B plasmid. Mutant III contained the BclI/AflII fragment of TR␣B cloned into the TR␣A plasmid, and mutant IV contained the NsiI/AflII fragment of TR␣B cloned into the TR␣A plasmid. Other mutants (V-XI) were constructed through PCR mutagenesis. First round of PCR was performed with two sets of primers: one (forward or reverse) containing the mutations and the other located at the very 3Ј (reverse primer Y: 5Ј-CGCTACCCGGGTCAAACTTCCTGGTCCTCA-AAGA-3Ј) or 5Ј (forward primer X: 5Ј-AGAGCTGGATCCATGGACCA-GAATCTCAGCGGG-3Ј) end of the coding region of TR␣A gene. Primers containing mutations were as follows: for mutant V, forward primer E: 5Ј-TGGCAATGGATCTTGACCTGGAA-3Ј and reverse primer F: 5Ј-TC-CAGGTCAAGATCCATTGCCA-3Ј; for mutant VI, forward primer G: 5Ј-GGATGATAGCAAGCGGGTAGC-3Ј and reverse primer H: 5Ј-GC-TACCCGCTTGCTATCATCC-3Ј; for mutant VIII, forward primer K: 5Ј-CGAGTGCGGCGGCGGAAGGA-3Ј and reverse primer L: 5Ј-TCCT-TCCGCCGCCGCACTCG-3Ј. In all above PCRs, TR␣A was used as a template. For mutant IX, mutant III was used as a template and primers used were I and J. For mutant X, mutant V served as a template and primers used were also I and J. Finally, for mutant XI, mutant III was used as a template and PCR was performed with E and F primers. Aliquots of above PCRs were mixed together and used as a template in the second round of PCR, performed with primers X and Y. DNA fragments containing mutations were restricted with NsiI/AflII (mutants V, VIII, X, and XI) or KpnI/AflII (mutants VI and IX) and cloned into TR␣A/pSP64 restricted with the same enzymes. Mutant VII was prepared by restricting mutant IX with BclI/BamHI enzymes, and DNA containing the mutations of interest was cloned into similarly cut pSp64-TR␣A vector. All mutants were verified by restriction analysis and sequencing.
Gel Mobility Shift Assay-Identical amounts of each TR␣ mutant (usually 1 l of the oocyte protein extract was diluted 20 times with the protein isolation buffer, and 2 l of this diluted extract was used/ sample) were mixed with 500 ng of dI-dC and 15 l of the binding buffer A containing 20 mM Tris, pH 7.5, 5 mM MgCl 2 , 100 mM KCl, 5 mM DTT, 0.1% Triton X-100, 10% glycerol, and a mixture of proteinase inhibitors (for DNA binding by homodimers) or binding buffer B containing 20 mM Tris, pH 7.5, 50 mM KCl, 2 mM DTT, 0.1% Nonidet P-40, 6% glycerol, and a mixture of proteinase inhibitors (for DNA binding by homo-and heterodimers). After 15-20 min of incubation on ice, 1 ng of synthetic, end-labeled double-stranded oligonucleotide bearing the TRE from Xenopus TR␤A gene (xTRE) (24) and 20 ϫ nonspecific (mutated TRE or mTRE; Ref. 24) or 20 ϫ specific competitor (xTRE) were added, where indicated. Samples were incubated at room temperature for an additional 20 min and analyzed on 5% native gel (at 150 V, for 140 min; Refs. 21 and 24).
Co-immunoprecipitation-This was done similarly as described (42). Oocytes were injected with mRNAs as described before, and incubated overnight in the presence of [ 35 S]methionine. Protein extracts were made as before, except that the isolation buffer was supplemented with 0.1% Triton X-100. Extracts were divided into two aliquots. One was used for Western blot analysis with TR and RXR antibodies to determine the levels of overexpressed proteins. The second aliquot (usually equivalent of 5 oocytes) was adjusted to 200 l with the protein isolation buffer, and 5 l of preimmune serum or serum containing the anti-TR antibody were added. Samples were incubated for 1 h at 4°C with delicate shaking, then 20 l of protein A-Sepharose was added and the incubation was continued for another hour. Samples were washed three times (10 min each wash) with the protein isolation buffer and three times (10 min each) in the same buffer in which the 70 mM KCl was replaced by 0.5 M NaCl. After removal of the last wash, 50 l of 1 ϫ sample buffer (37.5 mM Tris, pH 8.0, 2.5% ␤-mercaptoethanol, 1% SDS, 7.5% glycerol, 0.005% bromphenol blue) was added to the beads and the mixture was boiled for 5 min. The mixture was spun down, and 10 -20 l of the supernatant was analyzed on a 10% SDS-PAGE gel. The gel was dried and autoradiographed to detect the 35 S-labeled proteins.
Transcription Activation-This was done similarly as described by Wong et al. (22). Mutant TR␣ and/or RXR␣ mRNAs were co-injected into oocyte cytoplasm (0.125 ng of each mRNA/oocyte). After 2-3 h of incubation at room temperature, single-stranded plasmid DNA containing the thyroid hormone (T 3 )-responsive promoter of Xenopus TR␤A gene (22,24) was injected into the oocyte nucleus (5 ng/oocyte). Oocytes were incubated overnight at 18°C with or without 100 nM T 3 . After the incubation, the oocytes were homogenized in 0.25 M Tris, pH 7.5 (10 l/oocyte), and divided into two aliquots.
One aliquot (10 -12 oocytes equivalent) was mixed with 500 l of RNAzol B (Biotecx Laboratories) and 50 l of chloroform. The RNA was isolated according to manufacturer's recommendations and resuspended in RNase-free H 2 O (3.5 l/oocyte). To 7 l of the above RNA, 1 l of 32  To determine the amount of promoter DNA injected into the oocytes, the remaining half of the oocyte homogenate was mixed with equal volume of 30 mM EDTA, 20 mM Tris-HCl, pH 7.5, 1% SDS. Samples were treated with proteinase K (800 g/ml) for 1 h at 37°C, phenol/ chloroform-extracted twice, and precipitated with ethanol. Pellets were resuspended in 100 l of TE (10 mM Tris, pH 7.6, 1 mM EDTA) and treated with RNase A for 1 h at 37°C. The DNA was re-precipitated with ethanol and suspended in H 2 O (4 l/oocyte). NaOH was added to 0.1 M to denature the DNA. The DNA was slot-blotted onto a nitrocellulose membrane and fixed by UV irradiation (1200 mJ). Hybridization was done overnight at 42°C with a probe made of the CAT promoter vector as described (24).
Cloning and Sequence Analysis of Wild Type TR␣B cDNA in X. laevis-Total RNA isolated from stage 62 (39) tadpoles treated for 24 h with thyroid hormone (40) was reverse-transcribed using a primer that is common to both TR␣A and TR␣B gene and located near the carboxyl end of the coding region. A 5Ј-region encoding amino acids 62-186 and 3Ј-region encoding amino acids 353-414 (15) were PCR-amplified using primers common to both TR␣A and TR␣B. The 5Ј-region (amino acids 70 -186) and 3Ј-region (amino acids 353-402) were then cloned into pBluescript KS(Ϫ) (Stratagene). Five clones of the 3Ј-region were randomly isolated and sequenced, two of which were derived from TR␣A gene and three of which were from TR␣B gene, based on cDNA sequence comparison with the published sequences. For the 5Ј-region, the clones were hybridized with primers specific to either TR␣A or TR␣B (located at amino acids 95-100 as described previously; Ref. 41). Four TR␣B and three TR␣A clones were thus identified and sequenced.
Cloning and Sequencing of Genomic TR␣ Clones-A genomic library of homozygous X. laevis DNA (41) was screened with a 370-base pair cDNA probe of TR␣A (amino acids 62-186), a region that is highly conserved between TR␣A and TR␣B (15). The hybridization was performed overnight at 42°C in the buffer containing 50% formamide, 5 ϫ Denhardt's solution, 5 ϫ SSPE, 0.2% SDS, and 100 g/ml denatured salmon sperm DNA. Filters were washed three times (5 min each wash) at room temperature, in the buffer containing 2 ϫ SSC and 0.2% SDS and subsequently 2 ϫ 20 min at 65°C, in the buffer containing 0.25 ϫ SSC and 0.2% SDS. From 1,000,000 plaques screened, 5 positive clones were isolated.
Three clones were sequenced directly by using purified DNA and the Cycle Sequencing Kit (Pharmacia Biotech Inc.; also see Ref. 15). Briefly, phage DNA (2 g) was treated with RNase, alkali-denatured, and ethanol-precipitated. 5 ng of the 32 P-labeled antisense primer 5Ј-TGGTACTCCTGTGAGCTT-3Ј (amino acids 186 -181) were used in each sequencing reaction (94°C for 5 min, 55°C for 1.5 min, 72°C for 1.5 min: 1 cycle; followed by 25 cycles of 94°C for 1 min, 55°C for 1.5 min, 72°C for 1.5 min). The sequencing products were analyzed on a 5% sequencing gel. Two of the clones were found to encode TR␣A and the third one TR␣B. The identities were also confirmed by hybridization with a primer specific to either TR␣A or TR␣B (located at amino acids 95-100) as described previously (41).
PCR Analysis of TR␣A and TR␣B Expression in Tadpoles-The receptor cDNA prepared above with RNA from stage 62 tadpoles was PCR-amplified with two primers, one at amino acids 62-67 and the other at amino acids 186 -181, both of which are common to TR␣A and TR␣B (15). The amplified products were analyzed by Southern blot hybridization (41), using known amounts of TR␣A and TR␣B cDNA clones as the control and two oligonucleotide probes specific to each receptors (located at amino acids 95-100; Ref. 15).

RESULTS
To study the transcriptional properties of Xenopus TRs, especially as heterodimers with RXRs, we have recently established a thyroid hormone (T 3 )-responsive in vivo transcriptional system by introducing exogenous TRs and RXRs into Xenopus oocytes through microinjection of their mRNAs into the cytoplasm (21,22). As a reporter, we used the T 3 -responsive promoter of Xenopus TR␤A gene fused to a CAT gene fragment (22,24). Using such a system, we found that TR␣B is a much weaker transcription activator than TR␣A (see below). The experiments below were directed at determining the cause underlying the defects.
Two Residues in the D Domain of Xenopus TR␣B Cause Severe Reduction in TRE Binding Affinity-The Xenopus TR␣A and TR␣B are highly homologous to each other (15). The fulllength proteins differ only at seven amino acid positions (Fig.  1A). Five of the amino acid changes, three of which alter the charge of the protein, are located in the D domain between the DNA and hormone binding domains. Another is found in the NH 2 -terminal part of the protein, and the seventh one is localized to the hormone binding domain of the receptor. This last amino acid difference, an arginine in TR␣A versus a cysteine in TR␣B, is likely to be responsible for the 3-fold reduction in T 3 -binding affinity of TR␣B (about 0.084 nM, K d ) compared with TR␣A, TR␤A, or TR␤B (about 0.025 nM, K d ). 2 To investigate which amino acid substitution is responsible for the defect in transcriptional activation by TR␣B, the corresponding seven residues in TR␣A were replaced either individually or in different combinations with those from TR␣B (Fig. 1B). Particular focus was on the D domain, where the changes in charged residues are likely to affect DNA binding due to their proximity to the DNA binding domain. Thus, all combinational mutants of the double substitutions involving the charged residues were constructed for DNA binding and transcriptional activation studies.
To study the DNA binding properties of mutant TRs, we overproduced functional TRs and RXRs by microinjecting their mRNAs into the cytoplasm of the Xenopus oocytes. This method was chosen over the protein expression in Escherichia coli because TRs produced in the latter case were largely insoluble (24) and lacked potential post-translational modifications present in eucaryotic cells.
When equal amounts of the mRNAs for TR␣A, TR␣B, and the TR mutants were injected into Xenopus oocytes, essentially identical amounts of the different receptors were synthesized after the overnight incubation ( Fig. 2A). Whole oocyte protein extracts containing these TRs were then mixed with similarly prepared extracts containing Xenopus RXR␣ (21) and used to study their DNA binding properties. A double-stranded TREcontaining oligonucleotide derived from the Xenopus TR␤A promoter (xTRE; Ref. 24) was end-labeled and used with the oocyte extracts in the gel mobility shift assay. As expected, the TR␣A⅐RXR heterodimer formed a strong complex with xTRE ( Fig. 2B). In contrast, TR␣B extract showed only a very weak binding. Among the TR mutants, all of those involving substitutions other than V133D and/or N149D (e.g. mutants I, II, III, VI, and VIII) showed wild type DNA binding affinity (Fig. 2B). In stark contrast, all mutants containing either the V133D or/and N149D changes produced little TR⅐RXR⅐TRE complex, indicating that the introductions of a negative charge at either one of these two positions interferes with TRE binding by TR⅐RXR heterodimers. Interestingly, the Arg 3 Cys substitution had no effect on DNA binding, even though it resulted in a loss of a positive residue that is present in other TRs (15). In addition, the mutant IV exhibited slightly but reproducibly stronger binding to the TRE than TR␣B. This mutant differs from TR␣B only at positions 47 and 392, both of which do not affect DNA binding when mutated individually (mutants I and II). Thus, in addition to charges, other factors can also influence DNA binding by TRs.
Three types of RXRs exist in Xenopus as in other species (25,26). To test whether any mutations above affect RXR subtypespecific interaction with TRs, we compared TRE binding as heterodimers with Xenopus RXR␣ and RXR␥. Identical results were observed when either RXR␣ or RXR␥ was used in the gel mobility shift assay with the TRs (Fig. 2C and data not shown), indicating that these mutations do not affect any RXR subtype- specific contacts with TRs, if they exist.
To test the DNA binding specificity of the receptors, competition was performed with the wild type TRE (xTRE) itself or with the mutated TRE (mTRE), the mutations in which abolished both the response of the promoter to T 3 and binding to TR⅐RXR heterodimers (22,24). The binding by all TRs, including those that binds only very weakly, was found to be specific. Thus, the binding was efficiently competed out by xTRE itself but not by mTRE (Fig. 3A). In addition, competition was also done with two other known TREs: the TREp, a palindromic TRE consisting of two inverted repeats of the sequence AGGTCA, and TREgh, the TRE found in the human growth hormone gene, which is neither a palindromic TRE nor a TRE of two direct repeats of AGGTCA as in xTRE (27). Again, the complexes formed by xTRE and heterodimers of RXR␣ and all TR mutants showed similar competition profiles (Fig. 3B). The xTRE was found to be the strongest competitor and TREgh the weakest, similar to those observed for mammalian TRs and  idue in TR␣ (mutants TR␣IV, TR␣V,  TR␣VII, TR␣IX, TR␣X, TR␣XI, and  TR␣B) results in a TR with defective DNA binding. A, all TRs were translated with similar efficiencies in Xenopus oocyte. Equal amounts of TR mRNAs were injected into oocytes, and Western blot analysis with an anti-TR antibody was performed on the oocyte extract. CON-TROL, uninjected oocyte, indicating the lack of detectable endogenous TRs. B, gel mobility shift assay of the DNA binding properties of wild type and mutant receptors. The oocyte protein extracts prepared as in A were mixed with equal amounts of similarly prepared extract containing Xenopus RXR␣. The mixture was added to a labeled TRE oligonucleotide (xTRE) derived from the Xenopus TR␤A gene promoter (24) and the resulting TR⅐RXR⅐TRE complex was resolved on a native gel. The asterisk indicates a nonspecific complex that was formed even with the control extract lacking any introduced receptors. C, similar TRE binding affinity was observed for the TRs as heterodimers with either RXR␣ or RXR␥. Experiments were performed as in B.
FIG. 3. All TRs have identical sequence specificities even though their affinities for the xTRE differ. A, the complexes formed by RXR␣ and indicated TRs with labeled xTRE could be competed out by the unlabeled xTRE itself but not by the unlabeled mutated TRE (mTRE) (24). CONTROL, DNA binding with the oocyte extract without receptors and in the absence of competitors. B, both the strong and weak DNA binding TRs have identical sequence preferences. DNA binding was performed as in A but with different competitors as indicated. The xTRE was the strong competitor, TREp was a weaker one, and TREgh was the weakest for all TRs, as observed in mammals.
Heterodimer Formation with RXR␣ Is Unaffected in TR␣B and TR␣ Mutants-The defect in DNA binding observed above for some TR mutants could be either due to reduced ability to form heterodimers with RXRs or a decrease in TRs' intrinsic affinity for the TRE or both. To test these possibilities, gel mobility shift assay was performed under reduced stringency conditions using only extracts containing TRs. The result demonstrated that TR␣A could bind to xTRE in the absence of any RXR (Fig. 4). However, as reported for mammalian TRs (6 -8), the binding by TR␣A was much weaker than the TR␣A⅐RXR heterodimer (compare the signal ratio of the complex to the free TRE in Fig. 4 with that in Fig. 3B). The mobility of the complex suggest that it was formed by a TR␣A homodimer. Like the corresponding heterodimer, the TR␣B formed a homodimeric complex with the xTRE with the drastically reduced affinity. Similarly, all of the mutant receptors that were defective in binding as heterodimers with RXR were weak in binding as homodimers (TR␣IV, V, VII, IX, X, and XI). On the other hand, all the mutants that could bind xTRE strongly as heterodimers formed the homodimeric complex with xTRE as efficiently as TR␣A (TR␣I, II, III, VI, and VIII) (Fig. 4). These results suggest that the failure of TR␣B and some of the mutants to bind xTRE efficiently is due to a reduction in their intrinsic ability to bind DNA, but not the capacity to form heterodimeric complexes with RXRs.
To directly investigate the TR⅐RXR heterodimer formation by the TRs, mRNAs for TRs and RXR␣ were coinjected together with [ 35 S]methionine into oocytes. After overnight incubation, protein extracts were prepared and immunoprecipitated with either anti-TR or anti-RXR antibody. SDS-PAGE gel analysis of the immunoprecipitates showed that anti-TR antibody precipitated not only TRs, but also RXR␣ (Fig. 5A). For unknown reasons, different TRs were precipitated with varying efficiencies. However, the relative ratios of precipitated RXR␣ to TRs were similar for TR␣A, TR␣B, and all TR mutants, and similar amounts of TR and RXR were present in all samples (Fig. 5A,  Western blot panel). As a control for immunoprecipitation specificity, protein extracts from oocytes preinjected with no mRNA, mRNA for RXR␣, TR␣A, or TR␣B individually, were similarly precipitated with the anti TR antibody. The results showed that both TR␣A and TR␣B were immunoprecipitated (Fig. 5B, lanes 3 and 4). However, no signal was detected when FIG. 4. The relative binding affinities for xTRE by TR␣A, TR␣B, and the mutants alone are the same as their heterodimers with RXR␣, except that the absolute affinities are much weaker than those by the corresponding heterodimers. The binding assay was done as in Fig. 2A for heterodimers except that the RXR␣ extract was omitted. Note that the pattern of complex formation by different receptors was the same as in Fig. 2A except the intensities of the complexes were much weaker (compare the signal of the specific complex relative to the nonspecific complex labeled by an asterisk and that of the free DNA). The mobilities of the complexes formed by TRs alone were comparable with those when RXR␣ was also present, suggesting that they were formed by TR homodimers.

FIG. 5. All TR mutants can form heterodimers with RXR␣ with equal efficiencies.
A, the mRNAs for RXR␣ and various TRs were coinjected into the cytoplasm of oocytes and the oocytes were incubated in the presence of [ 35 S]methionine. After overnight incubation, the protein extracts were prepared and either immunoprecipitated (IP) with an anti-TR antibody followed by analysis on a SDS-protein gel (upper panel) or analyzed by Western blots with the RXR␣ or TR antibody (lower panel). Note that Western blots indicated that equal amounts of TR and RXR were present in all samples except for the control oocyte without mRNA injection. While for an unknown reason, different TRs were not immunoprecipitated with equal efficiency, the relative ratios of TR to RXR␣ in different samples were about the same, indicating equal formation of TR⅐RXR␣ heterodimers by different TRs. Note also that the 35 S label in the receptors was too light to produce a detectable signal during the short exposure time (5-10 s) used for the Western blot analysis. B, RXR␣ alone cannot be immunoprecipitated with the TR antibody. Oocytes were injected with no mRNA (CON-TROL), RXR␣, TR␣A, or TR␣B mRNA, and incubated in the presence of [ 35 S]methionine. Oocyte extracts were precipitated with the TR antibody and analyzed by gel electrophoresis, showing that both TR␣A and TR␣B but not RXR␣ (upper panel) were precipitated by the antibody. However, Western blot clearly showed that RXR␣ was efficiently translated in the oocytes (lower panel).
oocytes injected with no mRNA or only RXR␣ mRNA (Fig. 5B,  lanes 1 and 2), demonstrating the specificity of the antibody. These results thus clearly indicate that TR␣A, TR␣B, and all TR mutants formed heterodimers with RXR␣ with equal efficiency. This conclusion was also supported by co-immunoprecipitation with an anti-RXR␣ antibody (data not shown).
TRs That Are Defective in DNA Binding Are Weak Transcription Activators in Vivo-To compare the transcriptional activity of TR␣A, TR␣B, and the TR mutants, they were introduced into oocytes together with RXR␣ by injecting the corresponding mRNAs into the cytoplasm. After 2-3 h of incubation to allow the synthesis of the receptors, a single-stranded plasmid DNA containing the T 3 -responsive promoter of TR␤A gene (24) was microinjected into the nucleus and the oocyte was incubated overnight to allow the conversion of the single-stranded DNA into the double-stranded and chromatized form, and the subsequent transcription from the resulting template in the presence or absence of T 3 . The RNA transcribed from the TR␤A promoter was analyzed by primer extension (22). As we have shown recently, little transcription from the TR␤A promoter was detected in oocytes both in the absence and in the presence of T 3 in control oocytes without injected TR and RXR mRNA (Fig. 6A, lanes 1 and 2; Ref. 21). TR␣A alone slightly activated the promoter in the presence of T 3 (lanes 3 and 4) and this activation was drastically enhanced when RXR␣ was present (compare lanes 10 and 4). In contrast, no effect on the TR␤A promoter was observed in the presence or absence of T 3 when either TR␣B (Fig. 6A, lanes 5 and 6) or RXR␣ (Fig. 6A, lanes 7  and 8) was individually present. While co-introduction of TR␣B and RXR␣ was able to activate the promoter in the presence of T 3 , this activation was much weaker compared with that for TR␣A and RXR␣ (Fig. 6A, compare lanes 12 and 10). Among the mutant receptors, those that could bind the xTRE as efficiently as TR␣A (TR␣I, II, III, VI, and VIII) activated the promoter to a similar extent as TR␣A (Fig. 6A, lanes 25-34). All the other mutant TRs, which were defective in DNA binding, were found to be weak activators of the promoter (Fig. 6A,  lanes 13-24).
The relative transcriptional activity of the TRs did not correlate perfectly with their binding affinity for xTRE in vitro. For example, the mutant TR␣VII, involving a single amino acid substitution compared with TR␣A (N149D), showed no detectable TRE-binding in the gel mobility shift assay. It, however, could activate the promoter to as much as half of the level FIG. 6. TRs that are defective in DNA binding are also weak transcription activators. A, oocytes were injected with TRs and/or RXR␣ mRNAs as indicated. Following 2-3 h of incubation, the single-stranded, TRE-containing promoter plasmid was injected into the nuclei of the oocytes. Half of the oocytes were incubated in the absence (Ϫ) and the other half in the presence (ϩ) of T 3 . The transcription (TR␤ RNA) from the plasmid was assayed using the primer extension method. An endogenous oocyte RNA (Internal control) was also detected, which served as an control for the primer extension. The plasmid DNA was also recovered from the same oocytes and found to be equal among different samples by slot-blot hybridization. The relative TR␤ RNA levels were determined by normalizing the primer extension signals over the DNA recovery signals. Note that the defective DNA binders (TR␣B, TR␣IV, TR␣V, TR␣VII, TR␣IX, TR␣X, and TR␣XI) were all weaker activators compared with the rest and that TRs alone were not as effective in transcription activation as when RXR␣ was also present. The transcription signals in absence of T 3 were too weak to be visible in the plot of relative TR␤ RNA levels. B, when the same TR mRNAs used in A were injected into oocytes and assayed for TR levels, equal amounts of various TRs were produced. achieved by TR␣A (Fig. 6A, compare lanes 18 and 10). Similarly, the mutant TR␣X, containing a double substitution compared with TR␣A (V133D and N149D), showed moderate TRE binding among the weak DNA binding mutants but was the worst transcription activator.
As a control for the quantity of the RNA and efficiency of primer extension, the primer used for the primer extension assay was found to be able to hybridize to an endogenous oocyte RNA to generate equal amounts of a product in all samples (labeled as Internal control in Fig. 6A). In addition, when the DNA from the same oocytes used for RNA analysis was recovered and analyzed by Southern blot hybridization, identical amounts of the promoter plasmid were detected in all samples (Fig. 6A). Finally, when same amounts of TR mRNAs were injected into oocytes, the amounts of the receptors produced were found to be similar for all TRs (Fig. 6B), confirming that the observed differences in transcriptional activity were not due to any differences in the levels of TR or promoter DNA injected into the oocytes. As all mutant TRs were able to form heterodimers with RXR␣, they might be expected to dominantly interfere with wild type TR function. To test this, TR␣A and RXR␣ were introduced into oocytes together with increasing amounts of TR␣X, the worst activator among all mutants. In the absence of T 3 , little transcription was detected independent of whether TR␣A, RXR, and/or TR␣X was present or not (Fig. 7). However, in the presence of T 3 , TR␣A⅐RXR␣ heterodimer activated the transcription efficiently (Fig. 7, compare lane 8 to lanes 2 and  7). The presence of high levels of TR␣X (Ն4 times that of TR␣A) inhibited this activation to a level comparable with that when only TR␣X and RXR␣ were present. The result indicates that the mutant TR could compete efficiently with TR␣A for heterodimer formation with RXR␣, consistent with the coimmunoprecipitation data above.
The Previously Reported Xenopus TR␣B Was Derived from a Mutant TR Gene-The above results clearly demonstrate that the Xenopus TR␣B is a defective receptor and could interfere with the activity of TR␣A when copresent. TR␣ has been shown to be expressed during Xenopus metamorphosis (9 -14). However, due to the overall high level of sequence homology between TR␣A and TR␣B, these earlier studies could not differentiate the two genes. To test whether both TR␣A and TR␣B are expressed during metamorphosis, the receptor cDNA was amplified from metamorphosing tadpoles by PCR using primers common to both TR␣A and ␣B. The PCR products were analyzed by Southern blot hybridization using oligonucleotide probes specific to either TR␣A or ␣B. The results showed the TR␣A and TR␣B were expressed at similar levels at the climax of metamorphosis (data not shown).
It is puzzling why Xenopus expresses both a functional and a dysfunctional TR at a time when TR is critical for development. This prompted us to ask whether the previously cloned TR␣B was derived from the wild type gene or not. In addition, the possibility of a PCR artifact exists as the original clone was obtained by PCR (15). Therefore, total RNA from metamorphosing tadpoles was reverse-transcribed with a primer that is common to both TR␣A and ␣B and located near the end of the coding region. Two regions of the cDNAs were PCR-amplified, again using primers conserved in the two genes, and cloned. The 5Ј region encompassed amino acids 70 -186, including the D domain between the DNA and hormone binding domains (15). The 3Ј region covered amino acid 353-402, where an Arg 3 Cys substitution (amino acid 392, Fig. 1) in TR␣B seemed to cause a 3-fold reduction in T 3 affinity. 2 Five clones of the 3Ј region were randomly picked for sequence analysis. Three belonged to TR␣B and two to TR␣A based on sequence polymorphism other than at amino acid 392. All three TR␣B clones had an Arg residue at amino acid 392 (data not shown), i.e. as in TR␣A gene or a genomic clones of TR␣B gene (15). Thus, the wild type TR␣B sequence is identical to TR␣A in the hormone binding domain. The previously reported TR␣B cDNA sequence (15) was either derived from a mutant TR␣B gene or due to a PCR error at amino acid 392.
The clones of the 5Ј region were identified by hybridization with primers located at amino acids 95-100 that were specific to TR␣A or TR␣B by making use of the polymorphic differences between the two genes. Four TR␣B and three TR␣A clones were isolated and sequenced. The sequences of new cDNA clones confirmed the polymorphic sequence differences at amino acids 95-100 between the two genes (data not shown). In the D domain, sequences of the new TR␣B cDNAs had a Ser at amino acid 137 and Val at amino acid 151 (Fig. 8A), in agreement with the published TR␣B cDNA sequence (15). The new TR␣A clones had identical D domain sequences as previously reported except at amino acid 137, where an Ser was present in the new clones compared with a Gly in the original sequence. It is unclear whether this is due to polymorphism or a PCR artifact. However, as shown above, the TR with either an Ser or Gly at this position functioned identically.
Interestingly, at positions 133, 149, and 165 of the D domain, the amino acid residues in the new TR␣B clones were found to be identical to those in TR␣A. In the previously published Xenopus TR␣B sequence, these positions differed in sequence from those in TR␣A such that they resulted in a loss of a positive charge (position 165) and a gain of two negative charges (positions 133 and 149) compared with TR␣A (15). While the loss of the positive charge had no effect on receptor function, the introduction of either negative residue or both drastically reduced the affinity for a TRE and produced weak transcription activators as demonstrated above. Currently, it cannot be ruled out that PCR artifacts might be responsible for the previously published TR␣B sequence. However, it is extremely unlikely to generate three mutations within a region of less than 100 base pairs, considering the high fidelity of the Taq polymerase used during the PCR cloning (ϳ2 ϫ 10 Ϫ4 after a 10 6 -fold amplification; Ref. 29) and the fact that the nearby sequences in the DNA and hormone binding domains are the same in the previously published TR␣A and TR␣B sequences FIG. 7. A TR that is a weak activator can inhibit the function of wild type TR. Oocytes were injected without or with TR␣ and RXR␣ mRNAs together with indicated amounts of the mRNA for TR␣X, the weakest activator. The reporter promoter plasmid was then injected into the oocytes. After incubation in the absence (Ϫ) or presence (ϩ) of T 3 , the transcripts (upper panel) and promoter DNA were isolated and analyzed as in Fig. 6. Note that TR␣A⅐RXR␣ activated efficiently the transcription from the promoter and this activation was inhibited by excess TR␣X to the level obtained when only TR␣X and RXR␣ were present (Fig. 6). (15). Thus, the results suggest that the previously published TR␣B sequence was derived from a mutant TR␣B gene in X. laevis.
While several independent new TR␣B cDNA clones confirmed the new TR␣B sequence, these clones were obtained by PCR just like the previously published TR␣B sequence (15). To rule out any possibility of a PCR artifact, genomic DNA clones for both TR␣A and TR␣B were isolated and sequenced. The results confirmed the findings from our cDNA sequence analysis (Fig. 8B). They further revealed an intron/exon boundary at the NH 2 -terminal end of the D domain that is also conserved in the TR␤ genes (41). DISCUSSION We have provided evidence here that the previously cloned Xenopus TR␣B gene is a mutant TR gene. Three residues in the D domain are altered in this mutant. Two of these mutations introduce negative charges into the receptor and are responsible for the defect in DNA binding. In addition, our combinational mutagenesis experiments suggest that in addition to primary sequence, secondary, and tertiary structures are important for both DNA binding and transcriptional activation.
Structure-Function Correlations in DNA Binding-Each of the three mutated residues in the mutant TR␣B results in the loss of a positive charge (at position 165, Fig. 1) or gain of a negative charge in the D domain (at position 133 or 149). These mutations could be expected to affect charge-charge interaction between the receptor and DNA due to their proximity to the DNA binding domain. Interestingly, the Arg 3 Cys substitution at position 165, a position occupied by a positive amino acid in all other TRs, has no effect on DNA binding. The other two, both involving a neutral residue to aspartic acid substitution, cause severe reduction in DNA binding affinity.
FIG. 8. The previously reported Xenopus TR␣B sequence was derived from a mutant TR␣B gene. A, TR␣A and TR␣B cDNAs were PCR-amplified, cloned and sequenced. The regions of the D domain where the differences exist between the previously reported TR␣A and TR␣B were shown for one representative cDNA clone each for TR␣A and TR␣B. The numbers refer to the amino acid positions as previously reported (15). The asterisks and double asterisks in TR␣B indicate the residues that were different between the previous reported TR␣A and TR␣B. The asterisks mark the residues which were found to be different in new cDNA clones compared with those of Yaoita et al. (15) and the double asterisks mark those residues in the new TR␣B cDNA clone that are the same as those in Yaoita et al. (15), i.e. remain to be different from TR␣A. Thus three amino acid residues were found to be mutated in the previously reported TR␣B sequence in a 32-amino acid region. B, sequences of genomic TR␣A and TR␣B clones confirm the new TR␣ cDNA sequences. genomic clones were directly sequenced with an antisense primer. For a direct comparison with the cDNA sequences in A, the sequencing gel was presented upside down. Note that the genomic clones encode identical amino acids as the corresponding cDNA clones and that an intron/exon boundary is present at Leu 132 .
The importance of D domain in DNA binding has also been demonstrated in two previous studies. In one case, Zechel et al. (30) reported that a deletion of a region of chicken TR␣ encompassing the D domain strongly reduced its DNA binding ability. In the second case, Uppaluri and Towle (31) recently generated a library of mutant receptors for rat TR␤1 by PCR. Characterization of these mutants revealed a number of mutations in the D domain that affected DNA binding.
The structure of the D domain was recently reported in a crystal of a complex formed between a TR⅐RXR heterodimer and a TRE essentially identical to the xTRE used in our study. Numerous direct phosphate contacts formed mostly by the basic residues and water-mediated hydrogen bonds to the bases and phosphates in the D domain (positions 132-152) (32). In particular, the residues at positions equivalent to Val 133 and Asn 149 of Xenopus TR␣, and/or their flanking residues make direct and/or water-mediated contacts with the negatively charged phosphate backbone. Assuming that TR structure is conserved, it would not be surprising that the introduction of a negative charge into either position or both inhibits DNA binding. On the other hand, the residue Arg 165 is located within a disordered region in the crystal and this may not be important for DNA binding as we have observed here.
However, these simple amino acid-DNA interactions cannot explain the DNA binding defects for all mutants. For example, a single amino acid substitution at position 149 (Asn 3 Asp, mutant TR␣VII) or at position 133 (Val 3 Asp, mutant TR␣V) causes severe reduction in DNA binding ability (mutant TR␣VII is the weakest DNA binder among all mutants). Double mutation of these two residues produces receptors with higher DNA binding affinities than those with only a single substitution. Similarly, the mutant TR␣IV, which has all the substitutions in the D domain, but lacks two other substitutions (one in amino terminus and the other in the carboxyl terminus) present in the original TR␣B mutant (Fig. 1A), binds DNA with a higher affinity than the original TR␣B mutant, even though the substitution of the residue at the amino or carboxyl terminus alone does not influence DNA binding. These results suggest that the overall structure of the TR is an important aspect in DNA binding. These other amino acid residues, nearby or further away, are likely to play critical roles in secondary and/or tertiary structure formation, thus influencing receptor-DNA interaction.
Effect of D Domain on Transcriptional Activation-Our in vivo transcription assay clearly indicates that all mutant receptors can activate a TRE-containing promoter to varying degrees. This seems to contrast with the fact that some receptors have no detectable binding to a TRE in vitro. It is possible that these mutant receptors do have weak binding affinities for the TRE, but the complexes formed is too unstable to be detected by the gel mobility shift assay. Such weak, transient interaction with a TRE may be sufficient to activate the promoter in vivo. In this regard, it is interesting to note that essentially all D domain mutants generated by Uppaluri and Towle (31) that failed to bind DNA in vitro could nonetheless activate to varying degrees a TRE-containing promoter in cultured mammalian cells. On the other hand, all mutants that are defective in DNA binding are weak activators in oocytes, supporting a general correlation between DNA binding in vitro and transcriptional activity in vivo.
In addition to activating transcription in the presence of T 3 , TR⅐RXR heterodimers can also repress basal transcription in the absence of the ligand in oocytes (22). Unfortunately, under our in vivo transcription conditions, the basal transcription in the absence of the receptor is too low to be useful to study the transcriptional repressor activity of the unliganded receptors.
In addition to influencing DNA binding, the D domain also plays a role in transcription activation by TRs. For example, while TR␣VII, which contains a single substitution (V149D), is the weakest DNA binder, it is the strongest transcription activator among the mutants defective in DNA binding. The presence of additional mutations in the D domain in addition to this V149D substitution, e.g. in TR␣IX and TR␣X, further inhibits the transcription activation by the TR, even though these additional mutations actually enhance DNA binding. Similarly, several other mutants (TR␣V, TR␣XI, and the original TR␣B mutant) can bind DNA better than TR␣VII but are weaker activators in vivo. Thus, these other mutations in the D domain apparently affect transcription activation in a manner independent of their effects on DNA binding. Such an interpretation is also in agreement with two previous observations. Uppaluri and Towle (31) reported several mutants of the D domain generated by random PCR mutagenesis that had reduced transcription activation activities, but could bind TREs in vitro. Similarly, Lee and Mahdavi (38) identified by sitedirected mutagenesis two basic regions (amino acids 142-144 and 196 -198) that are important for transcription activation but not for DNA binding. The first region is flanked by the two amino acid residues that we have shown here to be important for DNA binding. Thus, the D domain of thyroid hormone receptors consists of intermixed subdomains critical for either DNA binding and/or transcription activation. The influence of the two functions of the receptor by the D domain is likely interrelated through structural regulation of the authentic DNA binding and activation domains.