Triiodothyronine (T3) Differentially Affects T3-Receptor/Retinoic Acid Receptor and T3-Receptor/Retinoid X Receptor Heterodimer Binding to DNA*

Thyroid hormone receptor (TR) heterodimerizes with retinoic acid receptor (RAR), retinoid X receptor (RXR), and triiodothyronine receptor auxiliary protein (TRAP) on natural and synthetic hormone re- sponse elements. Recently we showed that triiodothyronine (Ts) decreased TR bomodimer, but not TR/ TRAP heterodimer, binding to several thyroid hormone response elements (TREs). The effect of ligand on TR/RAR and TR/RXR heterodimer binding to DNA is not known. In this study, we showed that TR formed heterodimers with RAR and RXR on a retinoic acid (RA) response element and two TREs. Surprisingly, T3, but not RA, decreased TR/RAR heterodimer binding to DNA. In contrast, Ts, all-trans-RA, or 9-cis-RA did not affect TR/RXR binding to DNA. This finding suggests that TR/RXR heterodimer is a stable receptor complex that remains bound to response elements in the presence of ligand and therefore may be a receptor complex involved in T3-regulated transcription.


Paul M. Yen$., Akira Sugawara, and William W. Chin
From the Division of Genetics. DeDartment of Medicine. Brigham and Women's Hospital,Howard Hughes Medical Institute,and TREs (14). This finding suggests that TR/TRAP heterodimer is a stable receptor complex occupying TREs in the presence of T3. In this paper we studied the formation of TR/RAR and TR/RXR heterodimers on a retinoic acid response element (RARE), DR5 (8,9), and two TREs, DR4 and F2 (8,9,14,24), using an electrophoretic mobility shift assay (EMSA). Since both RARs and RXRs are expressed in a wide variety of tissues (25,26), it is possible that TR heterodimers containing these proteins may exist in many cells. Additionally, we examined the effect of T3 and retinoic acid (RA) on these heterodimers. Surprisingly, T3, but not RA, caused dissociation of TR/RAR heterodimers from these response elements. However, neither T3, all-trans-RA, or 9-cis-RA decreased TR/ RXR heterodimers binding to DNA, suggesting that TR/RXR heterodimers, like TR/TRAP heterodimers, are stable receptor complexes in the presence of T3. As such, they may play an important role in T3-regulated gene transcription.

MATERIALS AND METHODS
Preparation of in Vitro Translated Receptors-Previously described cDNA clones of rat TRa-1 (TRa-l), human TRP-1 (TRP-l), mouse retinoid X receptor P  in pBS (kindly provided by Dr. K. Ozato, National Institutes of Health), and human RARP in pGEM 1 (kindly provided by Dr. R. M. Evans, Salk Institute, La Jolla, CA) were used in these experiments (3,(27)(28)(29). Each cDNA was linearized with the appropriate restriction endonuclease and used as a template for RNA synthesis with T, RNA polymerase. Human TRP-1 truncation mutants missing the last 15,35, and 121 carboxylterminal amino acid residues (-15, -35, and -121) were generated from the TRp-1 cDNA plasmid with BsmI, BgZII, and PvuII restriction endonucleases, respectively before RNA synthesis. Unlabeled and [35S]methionine-labeled receptors then were produced from rabbit reticulocyte lysates according to the manufacturer's instructions (Bethesda Research Laborat,ories). Unprogrammed reticulocyte lysate also was incubated under the same conditions. The amount of translated protein was quantitated by trichloroacetic acid precipitation (15, 30) and by SDS-PAGE analysis of [35S]methionine-labeled receptors which showed proteins of the expected molecular weights.
Preparation of Nuclear Extracts-Nuclear extract from the rat pituitary lactotropic cell line, 235-1, was prepared and stored as previously described (15). These cells do not bind T , so presumably possess at most a minimal amount of endogenous TRs. Extracts then were dialyzed against 20 mM HEPES (pH 7.3), 5 mM 2-mercaptoethanol, 50 mM NaCl, 2 mM EGTA, 10% (v/v) glycerol, and 0.1 mM phenylmethylsulfonyl fluoride and centrifuged 10,000 X g for 15 min. Aliquots were stored frozen at -70 "C until used in DNA-binding assays.
Design and Preparation of Labeled DNA Probes-Double-stranded oligonucleotides containing TREs either from the chicken lysozyme gene (F2) [-2344 to -2326 (24)] or that are direct repeats of two half-sites (AGGTCA) separated by a gap of four (DR4) or five (DR5) nucleotides (8,9) in the context of the F2 flanking sequences were used in our experiments. The oligonucleotides were end-labeled with [y-32P]ATP by T4 polynucleotide kinase. The labeled probes then were purified on a 9% polyacrylamide gel under nondenaturing con-ditions. The full-length probe was cut from the gel and extracted in 400 pl of 0.5 M ammonium acetate and 1 mM EDTA a t 37 "C for 4 h. After microcentrifugation, labeled probe was extracted from the supernatant by two ethanol precipitations. The pellet was washed with 70% ethanol, resuspended in T E buffer (10 mM Tris, 1 mM EDTA (pH 8.0)), and stored at -20 "C until used in DNA-binding assays.
DNA-binding AssaylEMSA-Unlabeled receptor preparations or unprogrammed reticulocyte lysate (0.5-6.0 pl) and 10,000 cpm oligonucleotide probe were incubated in 20 pl of 25 mM HEPES (pH 7.5), 5 mM MgC12, 4 mM EDTA, 2 mM dithiothreitol, 110 mM NaCI, 5 pg/ ml bovine serum albumin, and 0.8 pg of sheared salmon sperm DNA (Sigma) for 30 min at room temperature, as previously described (14). Different concentrations of 3,5,3'-~-triiodothyronine (T3), all-trans-RA (Sigma), or 9-cis-RA (gift of A. A. Levin, Hoffmann-LaRoche) also were added to the samples in certain experiments. To study TR/ TRAP heterodimerization, 1.0 pg of 235-1 cell nuclear extract was added to each sample. To study TR/RAR and TR/RXR heterodimerization, samples were preincubated a t 30 "C for 10 min before adding oligonucleotide probe. After incubation, samples then were subjected to electrophoresis on 4% polyacrylamide gels in 0.5 X TBE buffer (45 mM Tris borate and 1 mM EDTA) for 75 min at 4 "C. Proteins on the gel were fixed by incubation in 7% acetic acid and 10% isopropanol for 15 min at room temperature. The gels were dried under vacuum at 80 "C for 1 h and autoradiographed with an intensifying screen. Film exposure to the gels ranged from 12 h to 1 week. In some experiments, the density of relevant bands on fluorographs was quantitated with a laser densitometer (Molecular Dynamics, Sunnyvale, CA).

RESULTS
We used EMSA to examine formation of TR/RAR heterodimers on the synthetic RARE, DR5, which previously was shown to mediate RA-regulated transcription (8,9). TRa bound as a monomer, and TRP bound as a homodimer, on this response element (Fig. 1, lanes 1 and 7). Retinoic acid receptor p (RAR) also bound as a homodimer on DR5 (Fig. 1, lane 6, and Fig. 2A, lanes 5-10). When increasing amounts of TRa or TRB were added to RAR, complexes with intermediate mobility were formed representing TRa/RAR or TRPIRAR heterodimers (Fig. 1). Formation of these heterodimers was favored over homodimers, and enhanced overall TR binding t o DR5. The presence of TR in these new complexes was demonstrated by supershifting the heterodimers with TRaor TRP-specific antibodies (data not shown). We next studied the effect of M TS or M RA on TRa/RAR heterodimer binding to DR5. These concentrations were used since we previously showed that T3 causes TR homodimer dissociation from several TREs at an EC60 of lo-' M (14), and the K d of RA for RAR is <lo-* M (31), so that essentially all of the RARs should be ligand-bound at M RA. Surprisingly, T3, but not RA, decreased TRa/RAR heterodimers from DR5 ( Fig. 2A, lanes [7][8][9][10]. RA had no effect on RAR homodimer binding to DR5 (Fig. 2A, lane 6). Addition of T3 increased slightly the mobility of the TRa monomer band, which is similar to previous observations for TRa binding to several TREs (14).
We next examined TRa, RAR, and TRa/RAR heterodimer binding to the synthetic TRE, DR4 (8,9). TRa bound mostly as a monomer to DR4, whereas RAR was unable to bind DR4 (Fig. 2B). As in the case of DR5, T3, but not RA, decreased heterodimer binding to DNA. Although not seen on this exposure, a faint homodimer band in TRa-containing samples (Fig. 2B, lane 1 ) disappeared after T3 addition similar to what we reported for several other natural and synthetic TREs (14). TRP also formed heterodimers with RAR on DR4 and DR5. Similar effects of T3 and RA on TRP/RAR heterodimer binding to these response elements were observed (data not shown).

T3 Decreases TRIRAR Heterodimer Binding
to DNA dimerized with RAR on this TRE, suggesting that the two T R isoforms can have different abilities to dimerize with RAR depending on the nucleotide sequence of the TRE (Fig. 3, lane 5, and Fig. 5, lane 3). Additionally, TRa formed relatively less heterodimer with RAR on F2 than on either DR4 or DR5, suggesting that different nucleotide sequences and/or orientation of half-sites in TREs determine the amount of TR/ RAR heterodimer bound to DNA. Similar to what we observed for DR5 and DR4, T3, but not RA, decreased TRaIRAR heterodimer binding to F2 (Fig. 3, lanes 5-8). When our studies of TR/RAR binding are taken together, they suggest that one particular member of the TR/RAR heterodimer (TR) needs to bind its cognate ligand for the heterodimer to decrease binding to DNA. In order to determine whether the T3-mediated decrease in DNA binding of TR/RAR heterodimers was due to dissociation of TR/RAR heterodimer from DNA, we examined the effect of T3 on TR homodimer and TR/RAR heterodimer prebound to DNA. We observed loss of TR homodimer binding to F2 when T3 was added either simultaneously with receptor and probe or after TR homodimer was prebound to probe (Fig. 4, lanes 2-4). We also observed similar results when we examined the effect of T3 on TR/RAR heterodimer prebound to DR4 (Fig. 4, lanes 6-8). These results suggest that T3 can induce dissociation of TR homodimers and TR/ RAR heterodimers from DNA. However, the rates of dissociation from DNA by these two dimers may be different since T3 caused complete dissociation of TR homodimer from F2
t i .

AB: A A B
FIG. 3. T3 and RA effects on TRa/RAR heterodimer binding to F2. In vitro translated RAR (4 pl) was mixed with TRa (4 pl) and labeled probe in the presence or absence of TI M) or RA (10"j M ) and analyzed by EMSA. Antibody supershifts were done as described in Fig (lanes 4 and 8). Samples then were analyzed by EMSA. For abbreviations, see Fig. 2 legend. by 5 min (data not shown), whereas 14.3% of TR/RAR heterodimer remained bound after 15 min.
We next examined TRIRXRP (TR/RXR) heterodimer formation on DR4, DR5, and F2. RXRB (RXR) alone did not bind to these response elements (data not shown); in contrast, TRPIRXR heterodimers bound to these DNA sequences (Fig.  5). RXR enhanced TRP binding to DR4 and DR5 more than an equivalent amount of RAR (Fig. 5). When nuclear extract was added to the DNA-binding reactions, TRP/TRAP heterodimer as well as TRP homodimer bound to each of the DNA sequences. It was not possible to compare directly the relative amount of TRPIRXR and TRPITRAP heterodimer formation since we could not quantitate the amount of TRAP in 235-1 nuclear extract. The TR/RXR heterodimer band migrated slower than the TR/RAR heterodimer but faster than the TR/TRAP heterodimer bands. Neither M T3 nor M RA affected the overall binding of TR/RXR to these response elements, although the bands containing T3-bound TR/RXR migrated slightly faster than bands containing TR/RXR in the presence or absence of RA (Fig. 6 ) . We also observed no effect of the putative RXR ligand 9-cis-retinoic acid on TR/ RXR heterodimer binding to DNA (data not shown). These findings suggest that TR/RXR, like TR/TRAP (14), is a stable complex bound to some TREs and RAREs in the presence of T3. Last, we examined TR subregions that may be important for homo-and heterodimerization. We made carboxyl-terminal truncation mutants of TRP in which the last 15, 35, and 121 amino acids were deleted. Deletion of the last 15 and 35 amino acids markedly decreased homo-and heterodimerization, suggesting that the distal carboxyl-terminal region of TRB may be important for dimerization (Fig. 7, A and B). T s did not affect mutant TR homodimer binding to F2 since the truncation mutants do not bind TI (Fig. 7A, lunes 2-4 and 6-8). Mutant -121 probably weakly heterodimerized with RAR but not RXR (Fig. 7B, lunes 4 and 8).

DISCUSSION
We have demonstrated the formation of TR/RAR heterodimers on two TREs (DR4 and F2) and a RARE (DR5). These heterodimers enhanced overall TR binding to these . . . . + + + +

FIG. 7.
Homo-and heterodimerization of TRB truncation mutants. Carboxyl-terminal truncation mutants of in vitro translated TRj3 were generated as described under "Materials and Methods." The mutant TRj3s were incubated with labeled F2 probe f either RAR or RXR and analyzed by EMSA. The mutant TRj3 homodimer, -15, migrated with slightly slower mobility, whereas mutant TRj3 homodimers, -35 and -121, migrated with faster mobility than native TRj3. A, homodimerization of truncation mutants in the presence or absence of T3 (10" M). To each reaction, 5 pl of in uitro translated receptor were added. B, heterodimerization of truncation mutants with RAR or RXR. To each reaction, in uitro translated TRs (2 pl) and in uitro translated RAR (4 pl) or RXR (2 pl) were added. Truncation mutants are designated by the number of deleted amino acids (-15, -35, and -122); j3, TRj3; remainder of abbreviations are as in Fig. 2. response elements in the absence of ligand. Interestingly, these heterodimers dissociated from DNA in the presence of T3 but not RA. In this regard, TR/RAR heterodimers behave similarly to TRa and TRP homodimers as well as TRa/TRP dimers which all dissociate from DNA in the presence of Ts (14). Recently, we observed that TRa and TRP form dimers with Mf-1 (32, 33), a mutant human TRP derived from a patient with generalized resistance to thyroid hormone in which glycine is substituted by arginine at amino acid position 345, and only minimally binds T3. These dimers dissociated from TREs at T3 concentrations similar to those that cause TR homodimer dissociation from DNA (33). As in the case of TR/RAR heterodimers, these findings would suggest that, for a certain subset of TR heterodimers, only one molecule of T3 needs to bind to the dimer before it dissociates from DNA. RA did not affect RAR homodimer binding to DR5 or TR/ RAR heterodimer binding to these response elements. The effects of T3 and RA on homodimer and TR/RAR heterodimer binding to response elements, then, seem to be fundamentally different.
Cotransfection studies have shown that both RARs and TRs mediate ligand-regulated transcription with TRE,.I as a response element (34, 35). However, unliganded TR blocked RA-regulated transcription with TRE,.I as response element (34,35). This repression was unaffected by RA addition, but T3 addition by itself, or in combination with RA, caused full transcriptional activation. Our results showed that T3 decreases TR/RAR heterodimers binding to two TREs and a RARE. Inasmuch as unliganded TRs repress basal transcription of T3-regulated genes (34, 36), it is possible that unliganded TR/RAR heterodimers also may play a role in basal repression of transcription since they bind some TREs better than TR homodimers (18). We speculate that addition of T3 would relieve this repression by causing dissociation of TR/ RAR heterodimer. Liganded TR complexes that can bind to DNA, such as TR/RXR or TR/TRAP heterodimers, could then bind to the response element and mediate transcriptional activation. Similar repression of RAREs such as DR5 or TREPaI by TR/RAR heterodimers may also occur. In support of these possibilities, recent data suggest there is a tramacting transcriptional inhibitor that binds to the carboxylterminal region of TRs which is released after T3 addition (37). Perhaps RAR, when heterodimerized with unliganded TR, could be such an inhibitor.
Recently, RXRa and RXRP have been shown to heterodimerize with TRs and enhance overall binding to TRh.1. We now have observed similar heterodimerization with TR and enhancement of overall TR binding by RXRP using hormone response elements containing half-sites arranged as direct repeats (DR4 and DR5) and an inverted palindrome (F2). Additionally, when T3 or RA was added, the heterodimers remained bound to DNA, suggesting that TR/RXR was a stable complex bound to response elements in the presence of ligand. These findings are similar to what we observed for TR/TRAP heterodimers (14). The issue of whether TRAP may be one of the known RXRs or a member of its family remains unresolved. While RXRa and RXRP display the properties of TRAP by enhancing TR binding to TREs (19-23), it is not clear that all TRAPS are RXRs. Cross-linking experiments suggest that TRAP in 235-1 rat pituitary cells is approximately 10 kD larger than the predicted molecular of the known murine, rat, and human RXRs (19,20,23,29,38)?
We recently compared the tissue distribution. of TRAP by performing EMSA of different tissue nuclear extracts and found that it had a different tissue distribution than reported * P. M. Yen, A. Sugawara, and W. W. Chin, unpublished results.
for the mRNA of known RXR isoforms (38).* EMSA experiments using anti-RXR antibodies to supershift TR/TRAP heterodimers would help determine whether TRAP is a member of the RXR family.
In cotransfection experiments, cells transfected with TR and RXR had greater T3-regulated transactivation than cells transfected with TR alone (20-22). These findings suggest that TR/RXR heterodimers can enhance T3-regulated transcription. When T3 and RA were added to cells transfected with TR and RXR, there was additional enhancement of transcription (22), suggesting the possibility of a synergistic dual ligand signaling system for TR/RXR heterodimers. Recently, 9-cis-RA has been shown to be an endogenous ligand for RXRa (38,39). We observed that g-cis-RA, like all-truns-RA, does not affect TR/RXR heterodimer binding to DNA. It would be important to know whether TRs can function as RXR-auxiliary proteins and enhance RXR binding to response elements of RXR-regulated genes such as cellular retinol acid-binding protein type I1 or apolipoprotein A1 (40,

41).
Last, we examined the ability of TRP truncation mutants to homodimerize and heterodimerize with RAR or RXR. We found that a 15-amino acid deletion of the carboxyl-terminal region markedly decreased both homo-and heterodimerization. Although we cannot rule out the possibility the deletion caused tertiary structure changes in the receptor that disrupted dimerization, our data suggest that this region, in addition to the previously described "heptad repeat" region, may be involved in dimerization (42, 43). Although no major differences in RAR and RXR interactions with truncation mutants were observed, it is possible that TR may contact RAR and RXR in similar regions in the unliganded state, whereas TR may no longer be able to interact with RAR in the liganded state due to conformational changes in TR. The increase in TRa monomer binding to DNA with concomitant decrease in TR/RAR heterodimer binding supports this possibility (Figs. 2-4). It is interesting to note that c-erbA a-2, an alternative splicing variant of TRa-1 which does not contain the highly conserved carboxyl-terminal 40 amino acids of the TRs, does not appear to homodimerize or heterodimerize with nuclear proteins.' Additionally, several point mutations in this region of TRP have been described in patients with generalized resistance to thyroid hormone (44-46), suggesting that a possible mechanism for T3 resistance in these patients may be poor dimerization by mutant TRs. We currently are investigating this possibility.
On the basis of our studies, we propose that there are two classes of TR heterodimers that bind to response elements. The first class contains TR/RAR heterodimers, which behave similarly to TR homodimers and TRaITRP dimers by dissociating from DNA in the presence of Ts. The second class contains TR/RXR heterodimers and TR/TRAP heterodi-mer8 which remain bound to DNA in the presence of T3 and thus may play a role in T3-regulated transcription. 14.