Half-site arrangement of hybrid glucocorticoid and thyroid hormone response elements specifies thyroid hormone receptor complex binding to DNA and transcriptional activity.

Thyroid hormone receptors bind to thyroid hormone response elements (TREs) as heterodimers with 3,5,3'-L-triiodothyronine (T3) receptor auxiliary protein (TRAP) and retinoid X receptors (RXRs). Currently, it is not known whether TR/TRAP or TR/RXR heterodimers need to bind to both TRE half-sites and whether there is a preferred orientation for TR/RXR heterodimer binding to TREs or transcriptional activation. Accordingly, we created a mutant TR alpha (TR-P box) by changing 3 amino acids in the P box region of the first zinc finger of the DNA-binding domain to that of the glucocorticoid receptor (GR), and we examined wild-type TR alpha and TR-P box complex binding to hybrid response elements containing TRE and glucocorticoid receptor element (GRE) half-sites arranged as a direct repeat with a four-nucleotide gap. TR-P box/RXR heterodimers selectively bound to the hybrid response elements in which GRE half-site was the downstream half-site, whereas TR alpha/RXR bound to hybrid response elements in which GREs were in either position. Additionally, TR/TRAP or TR/RXR heterodimer required two half-sites for binding to DNA, with strong binding to at least one of the half-sites. Last, co-transfection assays and methylation interference studies using the hybrid response elements suggest that the sequential arrangement of strong and weak half-sites in the TRE may be a critical determinant of TR/RXR heterodimer binding and transcriptional activation.


Thyroid hormone receptors bind to thyroid hormone response elements (TREs) as heterodimers with 3,5,3'-~triiodothyronine (T,) receptor auxiliary protein (TRAP)
and retinoid X receptors (RXRs). Currently, it is not known whether TWTRAP or TFURXR heterodimers need to bind to both TRE half-sites and whether there is a preferred orientation for TWRXR heterodimer binding to TREs or transcriptional activation. Accordingly, we created a mutant T R a (TR-P box) by changing 3 amino acids in the P box region of the first zinc finger of the DNA-binding domain to that of the glucocorticoid receptor (GR), and we examined wild-type TRa and TR-P box complex binding to hybrid response elements containing TRE and glucocorticoid receptor element (GRE) half-sites arranged as a direct repeat with a four-nucleotide gap.
TR-P bo-heterodimers selectively bound to the hybrid response element in which the GRE half-site was the downstream half-site, whereas T R d RXR bound to hybrid response elements in which GREs were in either position. Additionally, TWTRAP or TW RXR heterodimer required two half-sites for binding to DNA, with strong binding to at least one of the half-sites. Last, co-transfection assays and methylation interference studies using the hybrid response elements suggest that the sequential arrangement of strong and weak half-sites in the TRE may be a critical determinant of TWRXR heterodimer binding and transcriptional activation.
Thyroid hormone receptors (TRs)' are ligand-dependent transcription factors that belong to a large superfamily of nuclear hormone receptors (1,2). Unlike some members of this superfamily, such as the estrogen and glucocorticoid receptors (1,2) which bind to conserved hormone response elements (HREs) in the promoter region of target genes, TR binding to thyroid hormone response elements (TREs) appears to be a more complex process.
Second, we and others (7, [18][19][20][21] recently showed that T, itself can affect the nature of the receptor complexes that bind to TREs as T, decreases TR homodimer, but not TR/TRAP or TR/RXR heterodimer, binding to some TREs. On the basis of these data, and co-transfection experiments which showed that unliganded TRs repress basal transcription (22, 23), we have proposed that unliganded TR homodimers, when bound to TREs, may repress basal transcription of target genes which is relieved when ligand causes dissociation of TR homodimers from TREs. TR/TRAP and TR/RXR heterodimers remain bound to TREs in the presence of T,, further supporting their potential roles in mediating transcriptional activation above basal transcription levels.
Third, mutational analyses of TREs, and comparison of TREs from thyroid hormone (T,)-responsive genes, have suggested that TRs bind to a hexamer half-site sequence of AGGT(C/A)A (24). However, TRs bind to TREs which have considerable variation in their nucleotide sequences as well as their number, spacing, and arrangement of half sites (5,2628). In particular, TRs have been shown to bind TREs in which half-sites are arranged as palindromes (TRE,,s), direct repeats (DRs), or inverted palindromes (IPS). The optimal spacing for TR homoand heterodimer binding to DNA, and transcriptional activity, for TRE,,, DR, and IP appear to be 0, 4, and 6 nucleotides, respectively (TRE,,,O, DR4, and IP6) (5, 25, 28, 29). Additionally, the palindromic and inverted palindromic TRE half-site sequences are symmetric and thus would not be expected to specify a particular orientation of TlUTRAP or TFVRXR heterodimers bound to the TRE. On the other hand, the TRE half-sites in DR4 have a 5' to 3' polarity, so it is possible that the direct repeat motif may specify the orientation of TlUTRAP and TR/RXR heterodimer on the TRE.
Currently, it is not known whether TWTRAP needs to bind to both TRE half-sites to form heterodimers, and whether there is a preferred orientation for TWTRAP heterodimer binding to TREs or transcriptional activation. Recently, Zhou et al. (30) developed a n "oriented heterodimer" approach to study the onentation of the catalytic subunit of a bacterial transcription activator when bound to DNA. We now have used this approach to study TR/TRAP and TFVRXR heterodimer binding to TREs and transcriptional activation.
In these studies, we have created a mutant TRa (TR-P box) by changing 3 amino acids in the P box region of the first zinc finger of the DNA-binding domain to that of the glucocorticoid receptor (GR). This region previously has been shown to be important for TR, GR, and estrogen receptor recognition of their respective HREs (31-34), so TR-P box should have preferential binding to glucocorticoid response element (GRE) halfsites over TRE half-sites. We then compared the DNA binding of the TR-P box with wild-type TRa to hybrid response elements containing TRE and/or GRE half-sites arranged as direct repeats (GG, TG, GT, and 'M') (Table I). We also tested the abilities of these TRs to transactivate reporter plasmids containing these hybrid response elements. Our results suggest that two half-sites are required for TR/TRAP or TR/RXR heterodimer binding to DNA, but heterodimer can form when it binds well to only one of the half-sites. We also found that the sequential arrangement of strong (TRE) and weak (GRE) halfsites in the TRE may be a critical determinant for TWRXR heterodimer binding to the TRE and transcriptional activation.

MATERIALS AND METHODS
Preparation of in Vitro nunslated Receptors-Previously described cDNA clones of rTRa in pSG (351, mRXRp in pBS (13) (kindly provided by Dr. K. Ozato, National Institutes of Health, Bethesda, MD), were used in these experiments. TR-P box was created by changing the 3 amino acids of the P box on the first zinc finger of rTRa ( m C K G ) to that of the GR ( B C W ) by using a sense strand oligonucleotide (+209 to + 250) containing three point mutations specifying new codons: CCGCT-GTATCACTTGTGGGAGCTGCAAGGTC?TCTTTCGCCG, rTRa in pSG5, and the Clontech in vitro mutagenesis kit (Clontech, Palo Alto, CA). The mutations in the P box region of the TR-P box mutant cDNA were confirmed by DNA sequencing. Each cDNA was linearized with the appropriate restriction endonuclease and used as a template for RNA synthesis with T, or T, RNA polymerase. Unlabeled and [,%]methionine-labeled receptors then were produced from rabbit reticulocyte lysates according to the manufacturer's instructions (Life Technologies, Inc.). Unprogrammed reticulocyte lysate also was incubated under the same conditions. Translated [35S]methionine-labeled receptor protein was quantitated by trichloroacetic acid precipitation (8,  Preparation of Nuclear Extract-Nuclear extract from CV-1 cells were prepared and stored as described previously for 235-1 rat pituitary cells (8).
Design of Oligonucleotide Probes-Deoxyribonucleotides containing TRE or GRE half-sites arranged as direct repeats with a four nucleotide gap were used in our experiments: GG, GT, TG, and Tl' ( Table I). The half-sites were bounded by four nucleotides that were identical to the gap nucleotides. HindIII and XhoI sites were placed on either end to allow cloning into the PT109 reporter plasmid. GmT, TGm, GTm, and TmG were identical to their parent oligonucleotides except that either the GRE or TRE half-sites were replaced by 'ZTITTT (Gm or Tm).
The oligonucleotides were end-labeled with [y-32PlATP by T4 polynucleotide kinase. The labeled oligonucleotides were gel-purified and stored as described previously (7).
subjected to electrophoresis and autoradiography as described previ-Methylation Interference Studies-Methylation interference studies were performed similar to the method of Blanchard et al. (36). GT and TG oligonucleotides containing EcoRI and BamHI restriction sites on the 5' and 3' ends were cloned into the EcoRI and BamHI sites of pBluescript KS(+) plasmid. The plasmids were cleaved with XbaI and the coding strands end-labeled with [32P]dCTP by Klenow reaction, before excision with ClaI. The reverse procedure was performed to end-label the noncoding strands. End-labeled DNA fragments (100 ng) then were methylated with dimethyl sulfate as described previously (36). Preparative gel shifts using 5 pl of cell extracts of COS-1 cells transfected with rTRa in pMT2 (37, kind gift of Dr. S. K. b r a t h a n a s i s ,  labeled probe were performed and the heterodimer bands and free probe bands excised. The DNA was isolated, treated with 1 M piperidine for 30 min at 90 "C, purified, lyophilized, and suspended in formamide loading buffer, before being subjected to electrophoresis on 8% denaturing polyacrylamide gels (36). Gels were dried and subjected to autoradiography for 12 h.
Co-transfection Studies-GG, GT, TG, and Tl' oligonucleotides containing HindIII andXhoI restriction sites on 5' and 3' ends were cloned into the PT109 vector (which contains a herpes simplex thymidine kinase promoter coupled to luciferase coding sequence) (38). These reporter plasmids then were sequenced to ensure that only a single copy of the hybrid response element had been incorporated.
CV-1 cells were grown in Dulbecco's modified Eagle's medium, 10% fetal calf serum. The serum was stripped of T, by constant mixing with 5% (w/v) AG1-X8 resin (Bio-Rad) twice for 12 h at 4 "C before ultrafiltration. The cells were transfected with expression (0.25 pg) and reporter (5 pg) plasmids a s well as a Rous sarcoma virus p-galactosidase control plasmid (3 pg) unless otherwise indicated (39) in 6-cm plates using the calcium-phosphate precipitation method (40). Cells were grown for 48 h in the absence or presence of M T, before harvesting. Cell extracts were analyzed for both luciferase (41) and P-galactosidase (39) activity.

RESULTS AND DISCUSSION
We first analyzed TR-P box binding to hybrid response elements containing either TRE or GRE half-sites arranged as a direct repeat with a four nucleotide gap sequence (Table I). The nucleotide sequences that bounded the half-site sequences were identical to the gap sequences so that we could assess the orientation of TR complex binding to these response elements without potential confounding effects by different flanking sequences. TR-P box could not bind as a monomer or homodimer to any of the hybrid response elements (Fig. L4). It was unable to bind as a TR-P bo-heterodimer to GG or G T however, surprisingly, TR-P box/RXR heterodimer bound to TG and TT.
These results suggest that TR-P bo-heterodimer selectively bound to a response element in which the GRE half-site was the downstream half-site. Moreover, heterodimerization with RXR permitted TR-P box binding to TG and TT when it was unable to bind to these response elements alone. Additionally, TR-P bo-heterodimer unexpectedly bound to TT per- haps due to RXR interaction with the TRE half-site. Addition of M T3 did not affect the pattern of TR-P box binding to these response elements but did increase slightly the mobility of TR-P b0xRX.R heterodimer binding to TG and TT (data not shown).
We next analyzed TRa binding to these hybrid response elements (Fig. 1B ). TRa did not bind to GG but was able to bind to GT, TG, and 'I"" as a monomer. TRa also bound weakly to TT as a homodimer similar to our previous observations with DR4 (18). In contrast to TR-P bo-heterodimer binding to these response elements, TRoJRXR heterodimer bound to GT, TG, and TT with TRa/RXR heterodimer binding to TT > TG 2 GT. Additionally, we also examined TRcl/TRAP and TR-P b o x / " heterodimer formation on these response elements using CV-1 nuclear extract and found that they had a similar pattern as TRoJRXR and TR-P boxmXR heterodimer binding to these sequences (data not shown). Indeed, recent immunodepletion experiments in our laboratory using anti-RXR antibodies suggest the major TRAP in CV-1 cells is RXRP or a related protein (161.2 Our data showing TRa/RXR heterodimer binding to GT and TG, and TR-P bo-heterodimer binding to ! I T , suggested that the heterodimer could bind to elements containing at least one strong half-site. However, it is not known whether TFURXR heterodimer binding requires two half-sites or could occur on only one half-site with proteidprotein interactions alone stabilizing the heterodimer. Accordingly, we mutated either the GRE (TGm and GmT) or TRE (TmG and GTm) half-sites of TG   and GT and examined TRa and TR-P box complex binding to these response elements. TRa monomer bound to GmT and TGm as expected but could not form TRoJRXR heterodimers on these elements ( Fig. 2A). TRa could not bind as a monomer to TmG or GTm (Fig. 2B 1, demonstrating that TR monomer binds to the TRE, but not the GRE, half-site. Additionally, TRa/RXR heterodimer could not bind to TmG or GTm. TR-P box monomer or heterodimer could not bind to any of these mutated sequences (Fig. 2, A and B ) . Taken together with the previous findings (Fig. 11, these results suggest that two half-sites are required for TR/RXR heterodimer formation and that at least one of these half-sites must be a strong half-site for TFURXR heterodimer binding. We next examined the ability of TR-P box and TRa to mediate T,-dependent transcriptional activation in co-transfection experiments using reporter plasmids containing the hybrid response elements. TR-P box was unable t o stimulate transcription from any of the reporter plasmids in the presence M T, (data not shown). These results suggest that although TR-P box can form heterodimers on TG and ' I "J! , they are not transcriptionally active. In contrast, TRa was able to stimulate transcription from TG and T l ' , but was either unable, or minimally able, to stimulate transcription from GG and GT (Fig. 3). Thus, although TR/RXR heterodimer can bind to GT, TG, and TT, it is much more active when bound to the latter two elements.
In order to examine further the mechanism for the difference in T,-mediated transcriptional activation via TG and GT, we studied the methylation interference patterns of TRw'RXR heterodimer binding to these elements (Fig. 4). Both half-sites were occupied by TRdRXR heterodimer, further confirming that two half-sites are required for heterodimer binding. Additionally, greater methylation interference of guanine residues was observed in the upstream half-site of TG, whereas the opposite was observed for GT. These results suggest t h a t T R d RXR bound to these elements in different orientations andor distinctly different conformations.
In the foregoing studies, we have examined the DNA-binding and transcriptional activity of TR-P box and TRa on hybrid response elements containing GRE and TRE half-sites. TR-P box/RXR heterodimer was able to bind to TG but not GT, suggesting that the arrangement of half-sites in a direct repeat can dictate whether this heterodimer can bind to the hybrid response element. Additionally, these data suggest that TR-P box may bind to the downstream half-site of TG. Recently, while this work was in progress, two other groups used a similar approach to demonstrate that heterodimers containing P box mutants of TRP and RXRa prefer binding to the downstream and upstream half-sites, respectively (42, 43); however, the functional significance of either mutant or wild-type heterodimer binding to the hybrid response elements was not examined. At present, it is not known whether wild-type TFURXR has a similar orientation requirement when bound to direct repeats. However, the ability of TRa monomer to bind only to the TRE half-site, and the differences in TRdRXR heterodimer methylation interference patterns on TG and GT, suggest that TRw'RXR heterodimer may bind in different orientations on the hybrid response elements.
We also have shown that TRdRXR heterodimer formation on the hybrid response elements requires both half-sites since mutating even a weak half-site (GRE half-site) abolishes TRheterodimer binding to DNA. Methylation interference studies further confirmed that TRw'RXR was able to contact both TRE and GRE half-sites of a hybrid response element. However, TR/RXR heterodimer interaction with at least one strong halfsite (TRE half-site) is required for heterodimer binding to DNA since TR-can bind to GT and TG but not GG.
We also showed that formation of TWTRAP heterodimers on hybrid response elements may be necessary but not sufficient for T,-mediated transcriptional activation. TR-P bo-heterodimer bound to TG and l T but was transcriptionally inactive, suggesting that P box amino acid substitutions can affect transcriptional activity either directly or indirectly via conformational changes. In this respect, it recently has been shown that the second amino acid of the P box (glycine) may be important for T,-mediated transcription (44). Interestingly, a mutant TR in which the entire DNA-binding domain was replaced with the GR DNA-binding domain also was transcriptionally inactive on these response elements (32).2 TRoJRXR bound to GT, TG, and TT, but was transcriptionally active only on TG and 'IT, suggesting that TRdRXR bound to GT may not be in the optimal orientation or conformation for transcriptional activation.
Previously, it has been shown that nucleotide spacing and orientation of TRE half-sites, as well as flanking sequences, can influence TR binding and transcriptional activation (5, 25, 26,29,45). Our studies of TR-P box and TRa on the hybrid response elements strongly suggest that arrangement of upstream and downstream half-sites in the direct repeat motif can influence TFURXR and TWTRAP heterodimer binding to DNA. Such arrangement of half-sites can have functional consequences since the orientation andor conformation of liganded TR/TRAP heterodimer on a TRE may be critical determinants of its ability to transactivate on that TRE. We speculate that the nucleotide sequence degeneracy observed among natural TRE half-site sequences can result in the sequential arrangement of strong and weak half-site sequences, which, in turn, may play an important role in modulating T,-mediated transcriptional activation.