INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The actions of thyroid hormone, 3,5,3'-L-triiodothyronine (T₃), are mediated by ligand-inducible transcription factors that are members of the steroid/thyroid hormone receptor superfamily. Two T₃ receptor (TR) genes, TR (NR1A1) and TR(NR1A2), are conserved in vertebrates (32, 43), while two TR and two TR genes have arisen by gene duplication in Xenopus laevis (65). TR encodes three C-terminal variants in mammals: 1 (NR1A1a) binds T₃ and DNA and is a functional receptor, whereas 2 (NR1A1b) and 3 (NR1A1c) do not bind T₃ and are weak dominant negative antagonists in vitro, although their roles in vivo are unclear (33, 40, 50, 57). Recent studies have described a promoter in intron 7 of TR, which generates two truncated variants, 1 and 2, that are repressors in vitro but are of unknown physiological significance (9). In contrast, TR encodes two N-terminal variants, 1 (NR1A2a) and 2 (NR1A2b), which are transcribed from separate promoters (24, 28, 42, 61). The 1 N terminus is encoded by two exons that are replaced by a single exon in 2. These exons are alternatively spliced to six common exons that encode the DNA binding, ligand binding, and dimerization domains of the receptor (32). This arrangement is conserved in vertebrates, and the invariant splice site between the divergent N termini and the first of the common exons is known as the changing point (65). The changing point is retained in both Xenopus TR genes, although additional splicing in the 5' untranslated region (5'-UTR) results in many transcripts that are temporo-spatially restricted during development (51, 64). TR0 is expressed in chicken; it contains only 2 amino acids proximal to the changing point and is similar to a short TR in Xenopus, although a mammalian homologue has not been identified (18, 52, 65). T₃-regulated development in chicken also involves temporo-spatially regulated expression of TR but not of TR (18).

The actions of T₃ are diverse, providing important signals for nervous system, inner ear, muscle, heart, and skeletal development in mammals and for amphibian metamorphosis. T₃ is a key regulator of postnatal growth, when it is essential for endochondral bone formation, and is the major regulator of the basal metabolic rate during adulthood, when it also influences cholesterol homeostasis, myocardial contractility, and maintenance of bone mass. Accordingly, TR mRNAs are widely expressed, but there are differences in concentrations of the isoforms in individual tissues (32). In particular, TR2 is largely restricted to the anterior pituitary and hypothalamus (24), where it mediates feedback regulation of the hypothalamo-pituitary-thyroid axis (1). It has, however, been difficult to ascribe specific functions to other TR variants, which do not display such restricted patterns of expression. TR and TR each bind T₃ with high affinity and recognize identical thyroid hormone response element (TRE) DNA binding sites (22, 32), but some studies have suggested that  and  receptors may show preferential activation of certain target genes (14, 35, 53, 66). The emergence of synthetic ligands that display selective affinities for TR subtypes may help to clarify this issue (10).

Gene-targeting studies to delete either TR or TR (17, 19), creating / double-knockout mice (21), 1/ null mice (23), or mice lacking only the 1 (58) or 2 (1) isoforms, have shown considerable redundancy among the various TRs, indicating that loss of one variant can be overcome by the activities of other isoforms in many tissues. These studies, however, showed a discrete role for TR1 in the development of auditory function (1, 16); implicated TR in maintaining thyroid hormone production, development of the small intestine and skeleton (21, 45), and maturation of B-lymphocyte populations (2); and implicated 1 in control of basal heart rate and body temperature (25, 26, 58). Despite this, data from biochemical studies and studies with TR-null mice do not account fully for the diversity and specificity of T₃ action and have suggested an additional hormone-independent role for TRs (23).

In earlier studies to examine the action of T₃ in osteoblasts, we identified a new 7.0-kb TR mRNA (59). The conserved structure of TR in all vertebrates, and the lack of C-terminal variants in any species, led to the hypothesis that this transcript may encode a novel N-terminal isoform, and a strategy involving 5' rapid amplification of cDNA ends (5'-RACE) was adopted to test it, using an osteoblast cDNA library. These studies report the characterization of two new TR isoforms, which are generated by alternative mRNA splicing and are expressed in tissue-specific patterns which alter with changes in thyroid status. TR3 is a novel receptor, and TR3 is a potent dominant negative antagonist that binds T₃ with high affinity.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Cells and animals. Pituitary GH₃ and ROS 17/2.8, UMR 106, and ROS 25/1 osteosarcoma cells were cultured in Ham's F12 medium plus 5% fetal calf serum (FCS), and COS-7 cells were maintained in Dulbecco's modified Eagle's medium plus 5% FCS. Male Sprague-Dawley rats (6 weeks old) were treated for 6 weeks with saline or T₄ (50 µg/kg/day) and thyroidectomized rats were given saline to form euthyroid (n = 7), thyrotoxic (n = 8), and hypothyroid (n = 7) groups, respectively. All rats were fed a normal diet, and thyroidectomized animals received Ca²⁺ lactate supplements to drinking water. Studies were performed under licence in compliance with the Animals (Scientific Procedures) Act 1986 and were approved by the Imperial College School of Medicine Biological Services Unit ethical review. Plasma T₄ concentrations (euthyroid animals, 0.79 ± 0.09 ng/ml: hypothyroid animals 0.06 ± 0.05 ng/ml; thyrotoxic animals, 0.97 ± 0.20 ng/ml) were determined by an immunoradiometric assay (Euro/DPC Ltd., Caernarfon, Gwynedd, Wales), and plasma TSH concentrations (euthyroid animals, 2.7 ± 0.4 ng/ml; hypothyroid animals, 114 ± 22 ng/ml; thyrotoxic animals, 1.2 ± 0.2 ng/ml) were measured using reagents from the National Institute of Diabetes and Digestive and Kidney Diseases and the National Hormone and Pituitary Program (A. Parlow, Harbor University of California, Los Angeles Medical Center, Los Angeles, Calif.) as described previously (55).

5'-RACE and inverse RT-PCR. UMR106 and GH₃ adapter-ligated cDNA libraries were constructed from poly(A)⁺ mRNA (Marathon cDNA synthesis; Clontech), and 5'-RACE was performed to identify TR variants using primers BR2 and BR1 and Marathon adapter primers, AP1 and AP2. TR3 contained two exons (A and B), and TR3 contained only exon A (see Fig. 2). The 5' boundary of exon A was mapped by inverse reverse transcription-PCR (RT-PCR). cDNA was synthesized from ROS 25/1, UMR 106, and ROS 17/2.8 poly(A)⁺ RNA (4.5 µg) using 10 µM BR1 with 5× buffer (250 mM Tris [pH 8.3], 30 mM MgCl₂, 375 mM KCl), 1 nM (each) deoxynucleoside triphosphates (dNTPs) and Moloney murine leukemia virus MMLV reverse transcriptase (20 U) in 10 µl at 42°C for 1 h. Second-strand cDNA was synthesized with 10 nM dNTPs, 5× buffer (250 mM Tris [pH 7.8], 50 mM MgCl₂, 5 mM dithiothreitol DTT, 5 mM ATP, 25% polyethylene glycol 8000) and 20× enzymes (E. coli DNA polymerase I, 6 U/µl; DNA ligase, 1.2 U/µl; RNase H, 0.25 U/µl) in 80 µl at 16°C for 90 min (all reagents from Clontech). A 2-µl volume of T₄ DNA polymerase was added for 45 min at 16°C to blunt the cDNA, which was self-ligated, cut with PstI, amplified with AR1 and AF3, and sequenced.

Northern blotting. UMR 106 and GH₃ poly(A)⁺ blots were probed at 65°C in 50% formamide-5× SSPE (20× SSPE is 175.3 g of NaCl per liter, 27.6 g of NaH₂PO₄ · H₂O per liter, and 7.4 g EDTA per liter)-0.15 M Tris (pH 8.0)-1% sodium dodecyl sulfate-(SDS) 5× Denhardt's solution (50× is 10 g of Ficoll 400 per liter, 10 g of polyvinylpyrrolidone per liter, 10 g of bovine serum albumin per liter)-100 µg of salmon sperm DNA per ml with a TR1 riboprobe, transcribed from XhoI-linearized pBS62 (28) using T₇ RNA polymerase, and washed in 0.1× SSC (1× is 0.15 M NaCl plus 0.015 M sodium citrate)-0.1% SDS at 75°C for 1 h. Multiple tissue blots (Clontech) were hybridized at high stringency to 3- and 3-specific riboprobes (generated by primers EcoBF1 plus BR1 and EcoAF plus BR1, respectively), washed in 0.1× SSC-0.1% SDS at 65°C for 30 min, and autoradiographed for 14 days. Multiple tissue blots were also hybridized to specific TR1 (EcoRI-XbaI fragment excised from pBS62), TR2 (EcoRI-SacI fragment from pSG5hTR2), common TR1/2 (EcoRI-XbaI fragment from pBSmTR1), and full-length human actin cDNA probes in ExpressHyb (Clontech) hybridization buffer for 1 h at 68°C, washed in 0.1× SSC-0.1% SDS at 50°C for 1 h, exposed overnight, and analyzed with a Molecular Dynamics 445 SI PhosphorImager (Amersham-Pharmacia Biotech) using Molecular Dynamics ImageQuant v1.2 software.

RT-PCR. DNase I-digested RNA (8 µg) was reverse transcribed at 42°C for 30 min using primer BR4 (50 pmol/µl), random hexamers (0.016 unit of absorbance at 260 nm; Amersham-Pharmacia), RNase inhibitor (20 U, Gibco), dNTPs (0.8 mM each), avian myeloblastosis virus reverse transcriptase (10 U; NBL Gene Sciences), and RT buffer in 20 µl. A 5-µl volume of template, containing dNTPs, was amplified with 50 pmol of BR4 and forward primer (B1F1 for 1, B2F1 for 2, and AF1 for 3 and 3) per µl, 1.2 µl of Taq DNA polymerase/TaqStart antibody mix (Clontech), PCR buffer (Gibco), and 1.5 mM MgCl₂ in 25 µl. After denaturation at 94°C for 2 min, 25 cycles of denaturation at 94°C for 30 s, annealing at 60°C for 30 s, and extension at 72°C for 1 min were followed by a 5-min extension. A total of 25 cycles of nested PCR were performed with 1 µl of template, 0.2 mM (each) dNTPs, 50 pmol of BR3 and forward primer (B1F2, B2F2, or AF2) per µl, 1.2 µl of Taq polymerase/TaqStart mix, PCR buffer, and 1.5 mM MgCl₂ in 25 µl. Products were purified and sequenced. Semiquantitative RT-PCR was optimized to detect linear accumulation of TR3 and 3. A range of input RNA concentrations (2 to 16 µg) was tested over a range of PCR cycles (20 or 25 cycles for the initial PCR followed by 20, 23, 25, 27, 30, 35, and 40 nested cycles) in tissues that expressed TR3 (heart), 3 (spleen), or both mRNAs equally (kidney). The products were Southern blotted, probed with AR2, and the results showed linear accumulation of 3 and/or 3 cDNA in each tissue with 8 µg of RNA and 25 initial and nested PCR cycles (data not shown).

Southern blotting. Genomic DNA was prepared from normal rat liver, ROS 25/1, UMR 106, and ROS 17/2.8 cells, digested with EcoRI, HindIII, PstI, and BamHI, and Southern blotted. Duplicate filters were hybridized to an exon A probe, amplified using primers AF1 and BAR1 and a TR1-specific probe, obtained by EcoRI and XbaI digestion of pBS62, and washed (0.1× SSC, 0.1% SDS) at 65°C for 1 h.

Genomic library. A total of 1.2 × 10⁶ recombinant phages from a DASH rat genomic library were screened. Duplicate filters were hybridized to the exon A probe and washed to a final stringency of 3× SSC-0.1% SDS at 65°C for 15 min. Five positive plaques were replated and screened twice, and three independent identical clones (4, 12, and 16), containing inserts of approximately 16 kb, were isolated. Clone 4 was digested, Southern blotted, and hybridized to the exon A probe. A 4.5-kb EcoRI fragment was isolated, subcloned, and analyzed.

DNA constructs and in vitro transcription-translation. Full-length TR3 and 3 cDNAs were constructed using the XbaI site in TR1 (pBS62) (28) and inserted into pBSKS(+) (Stratagene). Then 5' deletions were made to optimize in vitro transcription-translation. Primers EcoBF1, EcoBF2, EcoTRBF1, or EcoTRBF2 were used with BR1 to create 3 constructs that lacked exon A or its entire 5' UTR and 3 constructs that lacked exon B or its 5' UTR. TR3 and 3 were subcloned into pCDM8 (Invitrogen) for transfections. Site-directed mutagenesis (QuikChange; Stratagene) was performed in 1, 3, and 3 to alter the 3 translation initiation codon to CTG. Both strands of all constructs were sequenced (Thermo-Sequenase I; Amersham) using an ABI 373A apparatus (Applied Biosystems). Transcription-translation (Promega) was performed using pBS62, pBS3, pBS3, pBShRXR (36), and pBSmTR1 (46) with T3 polymerase and pSG5hTR2 (31) with T7 polymerase. Products were separated on a Sephadex G-50 (Pharmacia) column in 20 mM Tris-Cl (pH 7.8)-50 mM NaCl.

Expression of TR and RXR fusion proteins. TR1, TR3, and TR3 were subcloned into pCAL-n (Stratagene) to express calmodulin binding-peptide fusion proteins in Escherichia coli BL21(DE3)pLysS. Cultures, containing 100 µg of ampicillin per ml plus 34 µg of chloramphenicol per ml, were induced at log phase by using 0.5 mM isopropyl--D-thiogalactopyranoside (IPTG) at 30°C for 3 h. Bacteria were resuspended in GTM 375 (15% glycerol, 25 mM Tris [pH 7.8], 375 mM KCl, 10 mM -mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride)-0.05% Triton X-100-2 mM CaCl₂ and sonicated. The sonicates were pelleted at 1,100 × g, and the supernatants were agitated for 90 min at 4°C with 1 ml of calmodulin resin in 1 ml of GTM 375-0.05% Triton X-100-2 mM CaCl₂. The resin was packed into a 0.5-ml column and washed with 10 ml of GTM 375-0.05% Triton X-100-2 mM CaCl₂ at 4°C. CAL-TR fusion protein was eluted with 2.5 ml of GTM 375-0.05% Triton X-100-2 mM EGTA at 4°C and stored at 80°C. RXR was expressed from pQEhRXR as described previously (67). TR fusion proteins were analyzed by Western blotting using a monoclonal antibody (MA1-215; Affinity Bioreagents Inc.) against amino acids 235 to 414 of TR1 (4) and enhanced chemiluminescence detection (Amersham-Pharmacia).

Gel shifts. DNA binding properties of TR3 and TR3 were determined using a synthetic DR+4 TRE (56). Receptors were incubated for 30 min at room temperature, followed by 10 min at 4°C, with ³²P-labeled DR+4 (15,000 cpm), 100 ng of poly(dl-dC) (Amersham-Pharmacia), and 5 µg of bovine serum albumin with or without unlabeled competitor oligonucleotide (DR+4, nonspecific -globin promoter sequence [12], or mutated DR+4 [MUT] containing single G-to-A mutations in each half-site) in a 30-µl reaction mixture containing final concentrations of 10% glycerol, 25 mM Tris (pH 7.8), 500 µM EDTA, 90 mM KCl, 10 mM -mercaptoethanol, and 0.05% Triton X-100. The reaction products were resolved by nondenaturing polyacrylamide gel electrophoresis (5% polyacrylamide) in low-ionic-strength mobility shift buffer (10 mM Tris-Cl, 7.5 mM glacial acetic acid, 40 µM EDTA), as described previously (60).

Ligand binding. The T₃ binding affinities of CAL-1, CAL-3, and CAL-3 proteins were determined by saturation and competition binding with minor adaptations to published methods (28, 42). The receptor was incubated with 0.01 to 5 nM [¹²⁵l]T₃ (2,200 Ci/mmol; NEN) with or without unlabeled T₃ (500 nM) to determine specific binding at each concentration in saturation binding studies. The receptor was incubated with 30,000 cpm of [¹²⁵l]T₃ in the presence of 0.01 to 3.2 nM unlabeled T₃ in competition studies, and 100 nM T₃ was used to determine nonspecific binding. Reactions were performed in 100 µl of GTME 400 (15% glycerol, 25 mM Tris [pH 7.8], 500 µM EDTA, 400 mM KCl, 10 mM -mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride) plus 0.0025% Triton X-100, and the mixtures were incubated for 30 min at room temperature and 4 h at 4°C. Bound T₃ was separated on a Sephadex G-25 (fine) column, equilibrated in GTME 400, eluted with 2 column volumes of GTME 400-0.0025% Triton X-100, and quantified by -scintillation counting. Data were analyzed by nonlinear regression, and binding constants were calculated using GraphPad Prism software.

Transfections. COS-7 cells were seeded in six-well plates (10⁵ cells/well) containing Dulbecco's modified Eagle's medium-5% charcoal-stripped FCS (CSS medium) (49) and transferred to serum-free medium for transfection with Lipofectamine PLUS (Gibco) as follows: 500 ng of luciferase reporter driven by a thymidine kinase (Tk) promoter controlled by TREs from the rat malic enzyme or -myosin heavy-chain genes (59) or by two copies of a palindromic TRE (54); 40 to 200 ng of TR plasmid (1, 3, or 3 in pCDM8 or 2 in pRSV) (47); and 100 ng of Renilla internal control reporter (Promega) and pCDM8 carrier DNA to a total of 1.5 µg of DNA per well. After 3 h, 1 ml of 10% CSS medium was added and the cells were incubated for 24 h. The contents of each transfected well were split into four in a 24-well plate containing 5% CSS medium with or without T₃ (10 nM) and incubated for 48 h. Reporter gene activities were determined using a dual luciferase assay (Promega), and luciferase was normalized to Renilla prior to calculation of T₃ induction ratios.

Nucleotide sequence accession numbers. Rat TR3 and TR3 mRNA and genomic sequences were submitted to DDBJ/EMBL/GenBank databases under accession numbers AF239914, AF239915, and AF239916.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

A novel 7.0-kb TR mRNA was identified previously in UMR 106 cells (59). This transcript was expressed with a 6.2-kb TR1 mRNA in osteoblastic cells but was found to be absent from GH₃ cells (Fig. 1A), which express separate 6.2-kb mRNAs encoding TR1 and TR2 (24). Southern blotting of normal liver and osteosarcoma cell DNA indicated that no major rearrangements of the TR genomic locus occurred in osteosarcoma cells that might generate an abnormal transcript (Fig. 1B). Alternative splicing of TR mRNA occurs only proximal to the changing point (32, 51, 64), and therefore, 5'-RACE was employed to isolate novel TR isoforms from UMR 106 and GH₃ cDNA libraries.

View larger version (47K): [in this window] [in a new window]   FIG. 1.   (A) Northern blot of pituitary GH3 and osteosarcoma UMR 106 cell poly(A)+ mRNA (5 µg, duplicate lanes) hybridized to a TR1 cRNA. A 6.2-kb mRNA is present in both cell types, and a 7.0-kb mRNA is present only in UMR 106 cells. (B) Southern blot of normal liver (N) and ROS 17/2.8, ROS 25/1, and UMR 106 osteosarcoma cell (lanes 17, 25, and U) genomic DNA, digested with EcoRI, HindIII, PstI, and BamHI and probed with TR1 (top) and exon A (bottom) probes.

Only TR and TR2 were isolated from the GH₃ library (data not shown). Three cDNA populations were identified in UMR 106 cells by sequencing of individual 5'-RACE clones; one population contained TR1, the other two were novel, and TR2 was not isolated. The shorter novel cDNA contained a 322-bp sequence (exon A) that joined the changing point (Fig. 2B). The longer cDNA contained a 653-bp sequence that included a new 315-bp sequence (exon B), which separated exon A from the changing point, and a further 16 proximal nucleotides that extended exon A to 338 bp (Fig. 2B). Sequencing revealed stop codons in all three reading frames of exon A, indicating that it is noncoding and forms a 5'-UTR. The 315-bp exon B consists of 245 bp of 5'-UTR that is continuous with exon A, followed by a 70-bp open reading frame (ORF) in continuity with the remainder of TR.

View larger version (39K): [in this window] [in a new window]   FIG. 2.   (A) The 4.5-kb EcoRI genomic fragment containing the 3 and 3 locus. Exons A (342 bp) and B (315 bp) are in continuity with the 101-bp first common TR exon. Noncoding sequence is shown as dotted lines, intron/exon boundaries are shown by vertical bars and the exon A 5' boundary is shown by four arrowheads representing a 20-bp region mapped by 5'-RACE and inverse RT-PCR. The changing-point splice site precedes the first common exon. Asterisks mark each end of the 4.5-kb fragment and are shown in panel D to indicate the location of this clone within the complete TR gene. (B) Novel cDNA clones obtained by 5'-RACE from a UMR 106 library. TR3 contains exons A and B; in TR3, exon B is skipped and exon A splices to the changing point. AP1 and AP2 are forward primers complementary to adapters used in library construction. Arrows show locations of the reverse primers used for 5'-RACE. Arrowheads below exon A indicate stop codons in each reading frame, and the arrow below exon B show an ORF in continuity with the rest of TR. The sequence at amino acid 103 in TR3 shows the next in-frame AUG codon within a Kozak consensus sequence and represents the 3 initiation codon. (C) TR3 and 3 predicted proteins of 390 amino acids (44.6 kDa) and 288 amino acids (32.8 kDa). 3 contains a 23-amino-acid N terminus encoded by exon B; 3 lacks a DNA binding domain and results from translation initiated at the downstream AUG codon. (D) Structural arrangement of the TR gene as deduced from published data from Xenopus (51, 65), human (3, 48), and mouse (18, 21, 61) TR genes and from data obtained in this study with rats. Exons are shown in the upper part of the diagram as solid lines and as shaded boxes beneath; introns are shown as dotted lines and below as thin continuous lines. In the lower part, promoter regions for 1, 2, and 3/3 are shown as thick solid lines, and the cross lines flanking the 2 region indicate that the relative positions of the 1 and 2 loci have not been determined in any species. Asterisks indicate the location of the 4.5-kb rat genomic fragment shown in panel A and cloned in this study. The common coding exons 3 to 8 lie 3' to the changing-point splice site, contain the DNA and ligand binding domains of TR, and are designated according to the published nomenclature for mouse TR1 (18, 21), which corresponds to exons 5 to 10 in human TR (3) and differs from that previously reported for mouse TR2 (61) and Xenopus (51, 65).

Thus, the longer cDNA containing exons A and B was designated TR3 and consisted of a 583-bp 5'-UTR with a 70-bp ORF spliced onto the six exons that encode DNA and ligand binding domains common to all TR isoforms (common exons 3 to 8 in Fig. 2D). Its 5'-UTR contains five overlapping ORFs encoding 5 to 60 amino acids, proximal to the ORF encoding TR3. The 70-bp in-frame ORF encodes a 23-amino-acid N terminus that replaces the A and B regions of 1 or 2. Database homology searches revealed no matches with known proteins or motifs. The predicted TR3 protein contains 390 amino acids with a molecular mass of 44.6 kDa (Fig. 2C).

The shorter cDNA that lacked exon B was designated TR3 and is predicted to direct translation from the next in-frame ATG codon, located at amino acid 174 of TR1 (28) (position 103 in 3) and situated within a consensus Kozak translation initiation sequence (30) (Fig. 2B and D). The 3 cDNA contained a 458-bp 5'-UTR derived from exon A and 136 bp of common TR sequence that precedes the next in-frame ATG codon described above (Fig. 2B and D). The 5'-UTR contained six overlapping ORFs encoding 4 to 60 amino acids, proximal to the predicted coding region. The predicted TR3 protein contains 288 amino acids with a molecular mass of 32.8 kDa (Fig. 2C). The 5' terminus of exon A was also determined independently, by inverse RT-PCR, in three populations of correctly spliced clones. The largest clone extended exon A by 4 nucleotides to 342 bp, and other populations contained mRNAs in which exon A was 338 or 325 bp long. These findings agree with the results obtained with RACE clones that contained exon A sequences of 322 and 338 bp and map the 5' mRNA boundary to a 20-bp region (Fig. 2A). Additional RT-PCR studies established that exons A and/or B were not spliced between TR1 or TR2-proximal exons and the changing point and vice versa (data not shown).

A 16-kB genomic clone was isolated, from which a 4.5-kb subclone was characterized and found to contain 886 bp of 5'-flanking sequence, exon A, exon B, the 101-bp common TR exon beginning at the changing point, and a 3' intron of at least 2.8 kb (Fig. 2A and D). The 16-kb clone did not contain TR1- or TR2-proximal exons, indicating that the 3/3 locus is in continuity with the common TR exon and located between the 1 and 2 loci and the changing point (Fig. 2D). Database searching revealed that the 5'-flanking region contains numerous possible binding sites for a variety of transcription factors, including Pit-1, Sp-1, Oct-1, and C-EBP, but lacks a TATA box upstream of the exon A transcription start site (data not shown). No sequence homologies to exon A and exon B 5'-UTRs were identified. However, the 70-bp ORF in exon B was 80 to 93% homologous to the 58 published nucleotides of the intron which precedes the changing point in the mouse TR gene (61). This 58-bp region begins with an ATG triplet and contains an in-frame ORF whose product has 70% amino acid identity to the 23-amino-acid N terminus of rat TR3. The homologous region in the human TR gene remains unpublished. The splice site between exons A and B consists of the exact metazoan 5' RNA splice site consensus sequence, A(G/G)UAAGU (8). These analyses support the finding that exon B is transcribed in TR3 mRNA but excised during 3 mRNA processing, and they indicate that the TR locus in this region is conserved between the rat and the mouse. Taken together with data that map the 5'-UTR of exon A to a 20-bp region, these findings suggest that TR3 and TR3 originate from a third TR promoter.

To determine whether exon B was retained in mature mRNA or whether its presence indicated the presence of unspliced RNA, an RT-PCR assay was designed using a reverse primer from the second common TR exon and a forward primer from exon A. The resulting products crossed splice sites between the two common DNA binding domain exons, at the changing point, and between exons A and B. The cDNAs generated in this assay were correctly spliced, indicating that 3 and 3 mRNAs are mature transcripts. Expression of 3 and 3 mRNAs differed between tissues; 3 alone was expressed in the spleen, only 3 was present in the heart and both were identified in the kidneys, suggesting that the relative expression of the transcripts may be tissue specific (Fig. 3).

View larger version (66K): [in this window] [in a new window]   FIG. 3.   RT-PCR using forward primers from exon A (common to 3 and 3) and TR reverse primers (from the second common TR exon) to show 3 (767 bp) and 3 (452 bp) mRNA expression in the spleen, heart, and kidneys. Genomic DNA, H2O, or RNA template lacking reverse transcriptase () negative controls confirm that expression is dependent on RT (+) and processing of the primary transcript.

Multiple-tissue Northern blots were hybridized to TR3 and TR3 probes to determine their tissue distribution. Both isoforms were expressed at low levels. TR3 was expressed widely as a 7.0-kb mRNA and was relatively abundant in the liver, kidneys, and lungs compared to other tissues; was present at lower levels in the skeletal muscle, heart, spleen, and brain; but was absent from the testes. A second, 4.0-kb transcript was expressed in skeletal muscle (Fig. 4A). A 7.0-kb 3 mRNA was expressed in the spleen and lungs and was present at very low levels in the brain but was undetectable in other tissues. Additional 3 mRNAs of 3.0 kb in the spleen, lungs, and brain and of 1.5 kb in the spleen were observed (Fig. 4B). Thus, both 3 and 3 are encoded by 7.0-kb mRNAs, which were not resolved by Northern blotting and are of identical size to the transcript first identified in UMR 106 cells. For comparison, multiple tissue blots were also hybridized to probes that detected each of the other TR and TR mRNA isoforms (Fig. 4D and E). The 6.2-kb TR1 mRNA was expressed predominantly in the kidney, liver, brain, and heart, was less prominent in skeletal muscle, and was just detectable in the lungs and spleen but was absent from the testes. The 6.2-kb TR2 mRNA was largely restricted to the brain, since expression was barely detectable in the lungs and heart and was absent from other tissues. TR1 and TR2 mRNAs were detected at the highest levels in the brain and at lower levels in the kidneys, skeletal muscle, lungs, and heart. TR1 and TR2 transcripts were barely detectable in the testes and liver and were absent from the spleen. Thus, the patterns of expression of the TR3 and TR3 mRNAs were different from those of the TR1, TR2, TR1, and TR2 isoforms and varied between tissues.

View larger version (83K): [in this window] [in a new window]   FIG. 4.   (A) Multiple-tissue Northern blot hybridized to a 3-specific exon B probe. A 7.0-kb mRNA is predominant in the kidneys, liver and lungs, expressed at lower levels in the skeletal muscle, spleen, brain, and heart, and absent from the testes. An additional 4.0-kb mRNA is restricted to skeletal muscle. (B) The same blot hybridized to a 3 probe at high stringency. A 7.0-kb mRNA predominates in the lungs and spleen and is present at low levels in the brain. A 3.0-kb mRNA is present in the same tissues, and a 1.5-kb transcript is evident in the spleen. (C) The same blot hybridized to a -actin cDNA to show that similar amounts of intact RNA are loaded in each lane. (D) Multiple-tissue Northern blot hybridized to a 1-specific probe. A 6.2-kb mRNA is predominant in the kidneys, liver, brain, and heart, expressed at lower levels in the skeletal muscle, just detectable in the lungs and spleen, and absent from the testes. (E) The same blot hybridized to 2- and 1/2-specific probes. A 6.2-kb 2 mRNA is clearly expressed in the brain and is just detectable in the lungs and heart but absent from other tissues. The 5.5- and 2.6-kb TR1 and TR2 transcripts were expressed at the highest levels in the brain and at lower levels in the kidneys, skeletal muscle, lungs, and heart, were barely detectable in the testes and liver, and were absent from the spleen. (F) Same blot as in panels D and E hybridized to a -actin cDNA.

In vitro transcription-translation of 3 and 3 cDNAs generated proteins migrating at 45 and 32.5 kDa (Fig. 5A), in agreement with calculated molecular masses of 44.6 and 32.8 kDa. Truncation of the 5'-UTR up to the exon A/B junction or initiation codon in 3 and to the changing point or putative initiation codon in 3 increased translation efficiency. Site-directed mutagenesis of the ATG at position 174 in TR1, position 103 in TR3, and position 1 in TR3 was performed to determine whether 3 was translated from the predicted initiation site and to examine whether the codon was used in 1 or 3 mRNAs. Transcription-translation of the mutants resulted in loss of the 32.5 kDa protein compared to the wild type in both 3 and 3 (Fig. 5B). Accordingly, translation of full-length TR3 was more efficient in the mutant than in the wild type. Thus, the TR3 mRNA encodes the predicted 288-amino-acid protein that lacks a DNA binding domain. Furthermore, the codon at position 103 was used to initiate translation of 3 from the TR3 mRNA, which thus encodes two proteins in vitro. Use of the codon in 1 was barely detectable, indicating that no significant translation occurs from this site in TR1 mRNA.

View larger version (64K): [in this window] [in a new window]   FIG. 5.   (A) [35S]methionine-labeled RXR and TR isoforms, with controls containing unprogrammed lysate incubated with T3 or T7 polymerase (RRL). 3 products were derived from full-length cDNA and constructs lacking exon A (TR3ExB) or the 5'-UTR (TR3ORF); 3 products were derived from full-length cDNA and constructs lacking exon A (TR3CP) or the 5'-UTR (TR3TR). The 32.5-kDa 3 product consists of two proteins, resolved after further electrophoresis. (B) Products from wild-type TR1, TR3, and TR3 cDNAs and constructs in which the AUG codon at position 174 in 1, 103 in 3, and 1 in 3 was mutated to CTG (TR1mut, TR3ORFmut, and TR3TRmut).

TR3, TR3, and TR1 fusion proteins were expressed in E. coli. Partially purified receptors were analyzed by Coomassie blue staining and Western blotting to confirm that full-length proteins were expressed (Fig. 6). T₃ binding affinities were determined in saturation and competition assays, and the binding constants derived from the two methods were in close agreement. Thus, 3 (K_D = 0.63 ± 0.13 nM) and 3 (K_D = 0.51 ± 0.09 nM) bound T₃ with high affinity, similar to 1 (K_D = 0.49 ± 0.10 nM) and in agreement with published data (28). E. coli extracts transformed with empty vector did not bind T₃.

View larger version (50K): [in this window] [in a new window]   FIG. 6.   (A) Coomassie blue-stained extracts from E. coli programmed with empty vector (pQE8 or pCAL) and receptor-expressing plasmids (RXR, TR1, TR3, and TR3). Expressed receptors were purified by Ni-nitrilotriacetate-agarose (pQE8 and RXR) or calmodulin (pCAL and TRs) column chromatography. Asterisks show receptor fusion proteins of the appropriate sizes. (B) Western blotting with a monoclonal antibody to amino acids 235 to 414 of TR, showing expression of full-length 1, 3, and 3 fusion proteins.

DNA binding was investigated by a gel shift assay using a DR+4 TRE (56). RXR, TR3, TR3, and TR1 failed to bind DNA when incubated alone with the element. Coincubation of RXR with 3 or 1 resulted in binding of 3-RXR or 1-RXR heterodimers, whereas coincubation of RXR with identical concentrations of 3 resulted in no binding (Fig. 7A). Addition of higher concentrations of 3 resulted in binding of 3-RXR complexes with greater mobility than 3 or 1 heterodimers (Fig. 7B and D and data not shown). 3, 3, and 1 heterodimer binding was sequence specific since complexes were competed by 50- to 100-fold excesses of unlabeled DR+4 but not by nonspecific competitor (Fig. 7B). Specificity was further confirmed by incubation of heterodimer complexes with up to a 150-fold excess of an unlabeled mutated DR+4 element (MUT; containing two AGATCA half-sites in place of the wild-type sequence AGGTCA), which failed to compete with the wild-type DR+4 TRE for heterodimer binding (Fig. 7C). In contrast, coincubation of preformed 3-RXR or 1-RXR heterodimers with increasing concentrations of 3 resulted in disruption of these complexes and formation of 3-RXR heterodimers, indicating that 3 competes efficiently with 3 or 1 for RXR. Reciprocal experiments demonstrated a less potent disruption of preformed 3/RXR complexes by increasing concentrations of 3 or 1 (Fig. 7D).

View larger version (68K): [in this window] [in a new window]   FIG. 7.   (A) Gel shift assay showing QE8 and pCAL extracts and overexpressed RXR, TR3, TR3, or TR1 (2 µl) incubated with 32P-labeled DR+4 in the first six lanes. The following lanes contain 2 µl of RXR with increasing concentrations (0.2, 1, and 2 µl) of TR3, TR3, or TR1. The migration position of the 1-RXR and 3-RXR heterodimers is shown on the right. (B) Competition of TR-RXR complexes with a 100-fold excess of unlabeled nonspecific oligonucleotide (NS) or increasing excess (10-, 50-, and 100-fold) of unlabeled DR+4. The lanes contain 2 µl of RXR coincubated with 2 µl of 3, 7.5 µl 3, or 2 µl of 1. The migration positions of 3-RXR and 1-RXR heterodimers are shown on the left and right, respectively, and the position of faster-migrating 3-RXR complexes in the middle lanes is shown on the right. (C) Competition of TR-RXR complexes with a 100-fold excess of unlabeled DR+4 or increasing excess (50-, 100-, and 150-fold) of unlabeled mutated DR+4 element (MUT). Lanes contain 1 µl of RXR coincubated with 2 µl of 3, 4 µl of 3, or 2 µl of 1. The migration positions of 3-RXR and 1-RXR heterodimers are shown on the left and right, respectively. 3-RXR complexes form only weakly with this reduced concentration of receptor compared to panels B and D. (D) Coincubation of increasing concentrations of 3 (1, 2, 5, and 10 µl) with preformed RXR-3 and RXR-1 heterodimers is shown in the first 10 lanes; the lanes contain 2 µl of RXR plus 2 µl of TR3 or TR1. The effect of increasing concentrations of TR3 or TR1 (2 and 5 µl) on preformed RXR-3 (2 µl of RXR plus 7.5 µl of 3) complexes is shown in the following lanes. The position of 3-RXR and 1-RXR heterodimers is shown on the right, and the position of faster-migrating 3-RXR complexes is also shown.

TR3, TR3, TR1, and TR2, a well characterized weak dominant negative TR isoform (47), were transfected with reporters driven by TREs from the -myosin heavy chain (MHC) or malic enzyme (ME) genes (59) or by two copies of a palindromic element (PAL) (54) into COS-7 cells, which lack significant concentrations of functional endogenous TRs (28). TR3 was a twofold more potent activator of the ME TRE than was TR1. This increased potency resulted from greater repression of basal gene expression by unliganded 3 compared with 1, as well as from increased T₃ activation (Fig. 8). On the MHC element, 3 was 1.5-fold more potent, because of increased gene activation rather than because of changes in basal expression. TR3 was a 2.4-fold more potent activator of the PAL TRE than was TR1, and this increased potency also resulted from increased gene activation rather than from differences in levels of basal expression. TR3 and TR2 did not activate any of the three elements in response to T₃, although 3 acted as a potent repressor of the ME TRE by markedly inhibiting both basal and T₃-activated expression (Fig. 8B and C). On the MHC and PAL TREs, 3 did not influence basal expression or T₃ activation.

View larger version (30K): [in this window] [in a new window]   FIG. 8.   COS-7 cells were transfected with TR (160 ng) or a luciferase reporter (500 ng) driven by the thymidine kinase promoter and controlled by either the ME or MHC gene TRE or a TRE containing two copies of PAL, an internal control Renilla reporter (100 ng) and pCDM8 carrier DNA (740 ng). (A) T3 induction of each TRE mediated by each receptor. Luciferase activity was normalized to Renilla to control for transfection efficiency, and the results are expressed as mean T3 induction ratio (with standard errors shown), calculated by dividing normalized luciferase activities following T3 treatment by basal values. (B) Induction of each TRE mediated by each receptor in the absence or presence of hormone. Luciferase activity was normalized to Renilla to control for transfection efficiency and the results shown as reporter gene activity in the absence () or presence (+) of T3 relative to the level of reporter gene activity under each condition in the absence of cotransfected receptor, which was normalized to a value of 1. Values below 1 in the absence of T3 indicate repression by unliganded receptor; values above 1 after addition of T3 indicate gene activation. (C) Complete data that were plotted in panels A and B, showing the values for mean basal (T3) and T3-induced (+T3) luciferase/Renilla ratios as well as T3 induction ratios (standard errors are given). Basal and T3-induced values are relative to reporter gene activity in the absence of cotransfected receptor, which was normalized to a value of 1.

In cotransfections, TR3 inhibited 3 and 1 induction of the ME TRE by up to 58%. Similarly, 3 inhibited 3 induction of the MHC TRE by 50% and inhibited 1 induction by up to 43%. Furthermore, 3 inhibited 3 induction of the PAL TRE by up to 44% and inhibited 1 induction by up to 80%. In marked contrast, low concentrations of 3 increased 3 induction of both the MHC and PAL TREs by more than twofold, but this effect was not seen on the ME TRE or for 1 induction of any of the three elements (Fig. 9A and B). TR2 did not inhibit or consistently influence 1 or 3 induction of any of the three elements (Fig. 9C and D).

View larger version (34K): [in this window] [in a new window]   FIG. 9.   Effect of cotransfecting a relative molar amount (0.125- to 1.5-fold) of TR3 (A and B) or TR2 (C and D) on TR3 (A and C) and TR1 (B and D)-mediated induction of ME, MHC, and PAL reporters. Cells were transfected with 3 or 1 (160 ng), 3 or 2 (0 to 240 ng), ME, MHC, or PAL luciferase reporter (500 ng), Renilla internal control (100 ng), and pCDM8 carrier DNA to a constant amount of 1.5 µg. Graphs show T3 induction mediated by each receptor on each TRE in the absence and presence of 3 or 2. Luciferase activities were normalized to Renilla, and the results are expressed as a percentage of the maximal T3 induction ratio, where mean induction in the absence of 3 or 2 was normalized to 100% (means and standard errors for three determinations are shown).

Since the activities of TR3 and TR3 differed and their mRNAs were expressed in tissue-specific patterns, semiquantitative RT-PCR was used to examine expression in euthyroid, hypothyroid, and thyrotoxic tissues. These studies confirmed that 3 and 3 mRNAs were expressed in tissue-specific ratios (Fig. 10). For example, in euthyroid animals only 3 was expressed in the liver and only 3 was present in the spleen, while both transcripts were expressed equally in the cerebral cortex but neither were present in the testes. Changes in the ratio of 3 to 3 occurred in some tissues after manipulation of the thyroid status. Both mRNAs were expressed equally in the euthyroid and hypothyroid cerebral cortex, but 3 predominated in the thyrotoxic cerebral cortex. In contrast, both transcripts were expressed equally in the euthyroid and thyrotoxic heart but 3 predominated in the hypothyroid heart. In the lungs, only 3 was detectable irrespective of the thyroid status. The concentrations of 3 and 3 mRNAs in the pituitary were also determined. TR3 alone was expressed in glands from each group, but the concentration was markedly increased in pituitaries from hypothyroid individuals. Although the assay was designed to determine relative expression within an individual sample and not to quantitate between samples, it is likely that this large difference reflects a true increased 3 expression in the hypothyroid pituitary since the method was optimized to measure linear product accumulation over a range of input RNA concentrations. These findings indicate that alternative splicing of 3 and 3 mRNA is tissue specific and show that it is influenced by thyroid status.

View larger version (45K): [in this window] [in a new window]   FIG. 10.   Southern blot of 3 and 3 products from rat tissues following manipulation of thyroid status, hybridized to an exon A internal probe. The 767-bp product is 3, and the lower, 452-bp band is 3. Expression in the pituitary is shown in the lower left panel using RNA from three to five glands of euthyroid (Eu), hypothyroid (Hypo), and thyrotoxic (Tox) rats. Lanes containing genomic DNA (Gen), water (H2O), and the absence or inclusion of reverse transcriptase (RT) () or (+) are shown. The upper panels show data from a representative experiment, and the graph represents mean values from three animals (with standard error shown). Graphed data are expressed as the relative ratio of 3 to 3; predominant expression of 3 is in the upward direction, and that of 3 is downward.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Two novel TR isoforms, TR3 and TR3, have been isolated. Exons encoding unique 1 or 2 N termini are replaced by two new exons in 3 or one in 3. Convergence of the four variants occurs at the common changing-point splice site (65). Thus, the existence of 3 and 3 cannot have been predicted with TR null mice, since gene targeting was directed at conserved exons common to all four variants (17, 21). TR1 and TR2 are transcribed from separate promoters (61), and data presented here suggest that 3 and 3 mRNAs are transcribed from a third promoter and arise by alternative mRNA splicing. TR3 mRNA encodes a 390-amino-acid, 44.6-kDa receptor that contains a unique 23-amino-acid N terminus, and TR3 mRNA encodes a 288-amino-acid, 32.8-kDa protein that lacks the DNA binding domain but retains the nuclear localization signal and ligand binding domain. 3 is a functional receptor, and 3 is a potent antagonist that may also potentiate the action of TR3, but not TR1, on some response elements under specific circumstances.

The patterns of expression of the 3, 3, 1, 2, 1, and 2 mRNAs differed in individual tissues and were distinct for each isoform, indicating that relative levels of expression of the various TR proteins could potentially differ between tissues. TR3 and TR3 were expressed primarily as 7.0-kb mRNAs, but shorter additional transcripts were seen in some tissues. The 5'-UTR and coding sequence of 3 is 1,445 bp long; in 3 and 1 the region is 1,130 and 1,386 bp, respectively, while the 3'-UTR is longer than 2.9 kb (28). Thus, the smaller 3 mRNA of 4.0 kb and the 3 mRNAs of 3.0 and 1.5 kb could contain the complete 5'-UTR and encode full-length proteins, which is likely since they hybridize to 5'-UTR probes. Such transcripts would include a short 3'-UTR resulting from differential mRNA processing, as described previously in human (48) and chicken (52) TR genes. Alternatively, smaller mRNAs could arise from events in which exon A or B skips some or all coding exons and splices to distal sites, resulting in untranslated transcripts that would also include short 3'-UTRs. This is less likely since the changing-point splice site contains a consensus sequence (8) and is used in all species (32). Furthermore, no splicing events that skip the changing point have been documented previously and none were identified in these studies.

In vitro transcription-translation and site-directed mutagenesis confirmed that 3 and 3 cDNAs encode the proteins predicted. These studies showed that TR3 is translated from 3 mRNA via the internal AUG at position 103, a site retained in TR and TR in all species and conserved in most nuclear receptors (37). Comigration of similar proteins resulting from translation of RXR, TR1, and TR2 cDNAs in these studies suggests that use of this codon may be a feature in other nuclear receptors. Previous studies have shown that internal AUG codons, including the conserved site used to translate 3 from 3 mRNA, initiate translation from chicken TR1 mRNA and that the use of these sites is influenced by the 5'-UTR. Thus, chicken TR1 mRNA encodes four proteins that are expressed in vivo and bind T₃ with high affinity (5, 6, 18).

The long 5'-UTRs of 3 and 3 are unusual and of likely functional significance. They contain multiple short ORFs, which occur in only 5 to 10% of eukaryotic mRNAs, particularly those encoding proteins that regulate cell growth and differentiation (29). Similar 5'-UTRs are present in Xenopus TR and TR and are proposed to regulate TR protein synthesis (65) from mRNAs that are temporo-spatially regulated during development (51, 64). 5'-UTR heterogeneity has been reported in mouse retinoic acid receptor (RAR) (27) and RAR2 (68) mRNAs. The RAR2 5'-UTR contains five overlapping ORFs that alter protein expression in a tissue-specific and developmentally regulated manner in transgenic mice (68). The human androgen receptor also contains a 577-bp 5'-UTR, which is necessary for induction of translation (41). These considerations suggest that translation of 3 and 3 proteins is likely to be tightly regulated. Expression of their mRNAs was also regulated (Fig. 4 and 10) and varied in relation to the levels of expression of 1, 2, 1, and 2 mRNAs in different tissues (Fig. 4). 3 and 3 mRNAs were expressed in all tissues studied except the testes, indicating that the proposed third promoter is transcribed widely. Additional, shorter transcripts were expressed in some tissues, and differing ratios of 3 and 3 mRNAs suggested that mRNA splicing is tissue specific. Alteration of 3 and 3 mRNA ratios following changes in thyroid status further suggests that splicing may be regulated by thyroid hormones.

DNA binding studies support the likelihood that regulated tissue-specific expression of 3 and 3 is functionally important. TR3, like TR1, bound specifically to DNA as a heterodimer with RXR, but 3-RXR heterodimers interacted poorly, presumably because these complexes contain a single DNA binding domain. RXR binds as a homodimer to elements organized as a DR+1 but does not bind DR+4 elements (22), as seen in Fig. 7A. It appears surprising, therefore, that 3-RXR complexes interact with the DR+4 element, albeit poorly, since TR3 lacks a DNA binding domain. Presumably, the formation of putative 3-RXR heterodimers results in a conformational change in RXR that enables a stable complex to bind weakly to the DR+4 element. This seems a likely probability since heterodimerization of RXR with other TR isoforms, which contain identical ligand binding and dimerization domains to 3, results in the formation of complexes that possess two DNA binding domains and bind to DR+4 elements with high affinity (Fig. 7A). Interestingly, preformed 1- or 3-RXR heterodimers were disrupted efficiently by lower concentrations of 3 than those required for DNA binding, suggesting that 3 competes for RXR in solution to limit its availability for heterodimerization with 1 or 3. Alternatively, 3 could interact with 1 or 3 to limit their access to RXR. This seems less likely since studies with chicken TR failed to identify interactions between truncated and full-length TRs (6). Thus, an equilibrium can be proposed in which the ratio of 3 to 3, relative to concentrations of other TR isoforms and RXR, influences which functional heterodimers coexist within T₃ target cells. Varying ratios of 3 to 3 mRNAs in different tissues indicate that these equilibria are tissue specific. In this model, small increases in the concentration of 3 are predicted to disrupt DNA binding of TR-RXR heterodimers and alter the balance of such equilibria, in keeping with its potent dominant negative activity. This model is likely to be applicable to signaling via other nuclear receptors that heterodimerize with and compete for RXR (11, 20, 34).

The mechanism of transcriptional activation by TRs and other nuclear receptors includes repression of basal gene expression by unliganded receptor, involving interaction with a corepressor complex containing histone deacetylase activity, followed by T₃ induction, involving displacement of the corepressor by the ligand and recruitment of coactivator proteins with intrinsic histone acetylase activity (39, 62). Similarly, TR3 induces reporter gene expression in two stages, involving repression by unliganded receptor followed by hormone activation, as described previously for TRs (7, 13). However, transactivation mediated by TR3 was TRE specific. T₃ activation of each of the three elements was more potently induced by 3 than by 1, but the magnitude of activation differed between elements and the differential potencies of 3 and 1 varied between elements (Fig. 8B and C). Repression of gene expression by unliganded receptor was greater with TR3 than with TR1 on the ME TRE but was similar for both receptors on the MHC and PAL elements. In contrast, 3 failed to activate transcription, although its actions were also TRE specific. 3 was a potent inhibitor of ME TRE gene expression in the absence and presence of T₃ (Fig. 8B and C), but expression of MHC and PAL was unaffected under either condition. The mechanisms responsible for these differences are unknown at this stage. The data, however, suggest the possibility that a specific factor whose activity is inhibited by both unliganded and liganded 3 is required for basal expression of the ME TRE in COS-7 cells but not for basal expression of the MHC and PAL TREs.

In cotransfections, 3 was a potent antagonist of both 1 and 3 at equimolar concentrations but 2 was inactive, in agreement with published data (47) and indicating that 3 is at least 10-fold more potent. Thus, 3 differs from 2 since it is a potent repressor and retains ligand binding activity whereas 2 displays weak activity, cannot bind T₃, and does not interact with corepressor proteins (54). Repression of 1-mediated induction of each TRE by 3 was proportional to the concentration of cotransfected 3 (Fig. 9B) and correlated with the degree to which the three elements were activated by T₃. Antagonism of 3 actions by 3 differed between the TREs and, in contrast, involved an initial potentiation of TR3 action on the MHC and PAL elements (Fig. 9A). On the MHC element, low concentrations of 3 increased 3-mediated induction of the element whereas addition of higher concentrations resulted in repression. A similar observation was seen for the PAL element but not for the ME TRE, on which 3-mediated activation was inhibited in a concentration-dependent fashion by 3. These data, together with the findings that 3 potently inhibits ME TRE gene expression but not MHC or PAL TRE gene expression, in the absence and presence of T₃ (Fig. 8B and C), indicate that additional unknown factors are likely to influence the actions of, and interactions between, TR3 and TR3. Thus, induction of MHC and PAL by 3, but not 1, is proposed to involve a cofactor that moderates its transactivation potency. In accordance with this model, low concentrations of cotransfected 3 are predicted to inhibit the activity of this moderating cofactor and result in the net potentiation of 3-mediated transactivation. At higher concentrations, 3 would inhibit the expression of MHC and PAL by acting as a dominant negative antagonist of 3. This hypothesis emphasizes the primarily inhibitory action of 3. The specificity of the model to TR3-induced activation of certain TREs may result from the unique 3 N-terminal domain. Thus, the 3 N terminus is predicted to interact specifically with the proposed moderating cofactor, which would therefore not interact with TR1. Interestingly the N terminus of 2, but not of 1, interacts with a subset of coactivators (44) and the SMRT corepressor (63) in the absence of T₃, supporting the view that the region confers specificity to TR isoforms (31).

These studies characterize two novel, alternatively spliced TR isoforms which are expressed widely. TR3 is a functional receptor, and TR3 is a potent dominant negative antagonist that binds hormone. Similar examples of truncated nuclear receptors that act as antagonists have been identified (15, 38), in addition to those described for chicken TR1 (6). Thus, a hypothesis can be proposed in which tissue-specific and hormone-regulated variation in the relative concentrations of TR3 and TR3 modulates target organ responsiveness to T₃. Such a mechanism defines a new level of specificity in the control of tissue thyroid status, which may be applicable to the actions of other nuclear receptors.