Thyroid Hormone Receptor Functions as Monomeric Ligand-induced Transcription Factor on Octameric Half-sites CONSEQUENCES ALSO FOR DIMERIZATION*

The thyroid hormone (3,5,3’-triiodothyronine) recep- tor (TsR) belongs to the nuclear receptor superfamily of ligand-inducible transcription factors. TsRs are known to bind as homodimers and heterodimers with retinoid X receptors (RXRs) to two hexameric half-sites in di- rectly repeated, palindromic, and inverted palindromic orientations. The binding of TsR monomers to indi- vidual half-sites was often reported, but no clear ligand-induced transactivational activity has been shown. Here, we analyzed interactions of TsR monomers with individual half-sites of the sequence NNAGGTCk We found that the two nucleotides 5’ of the AGGTCA core half-site strongly influence TSR binding and transcriptional activity: octameric half-sites of the consensus se- quence (T/C)(A/G)AGGTCA were bound by TsRs with the highest affinity. This suggests TsR functioning also as a monomeric transcription factor like the orphan nuclear receptors NGFI-B and FTZ-F1. Moreover, we observed that the function of TsR-RXR heterodimers on response elements composed of two half-sites in a directly repeated orientation spaced by 4 nucleotides is deter- mined in major parts by the S’-flanking sequence of the upstream half-site. Consequently, we noted that the af- finity of TsR homodimers

The thyroid hormone (3,5,3'-triiodothyronine) receptor (TsR) belongs to the nuclear receptor superfamily of ligand-inducible transcription factors. TsRs are known to bind as homodimers and heterodimers with retinoid X receptors (RXRs) to two hexameric half-sites in directly repeated, palindromic, and inverted palindromic orientations. The binding of TsR monomers to individual half-sites was often reported, but no clear ligandinduced transactivational activity has been shown. Here, we analyzed interactions of TsR monomers with individual half-sites of the sequence NNAGGTCk We found that the two nucleotides 5' of the AGGTCA core half-site strongly influence TSR binding and transcriptional activity: octameric half-sites of the consensus sequence (T/C)(A/G)AGGTCA were bound by TsRs with the highest affinity. This suggests TsR functioning also as a monomeric transcription factor like the orphan nuclear receptors NGFI-B and FTZ-F1. Moreover, we observed that the function of TsR-RXR heterodimers on response elements composed of two half-sites in a directly repeated orientation spaced by 4 nucleotides is determined in major parts by the S'-flanking sequence of the upstream half-site. Consequently, we noted that the affinity of TsR homodimers is influenced by both 6'-flanking sequences. Our findings suggest that the binding of dimerizing receptors like TsR and other nuclear receptors to their cognate response elements is determined not only by the half-site core sequence, orientation, and number of spacing nucleotides, but also by the nucleotide sequence preceding the half-sites.
Thyroid hormone (3,5,3'-triiodothyronine, T3)' regulates diverse aspects of cellular development and homeostasis by serving as a biological signal to control cell growth and differentiation in vertebrates (1,2). Its effects are primarily mediated by its nuclear receptors (T3Rs) which act as ligand-modulated transcription factors and belong to the nuclear hormone receptor superfamily (3-5).
T3Rs activate transcription through DNA sequences in the promoter of target genes, referred to as response elements. These T3 response elements (TREs) contain degenerated pairs of the consensus half-site AGGTCA, which can be arranged as direct repeats, palindromes, or inverted palindromes (6-9). It is generally accepted that TREs consist of hexameric core motifs, whereas it is less widely recognized that also the nucleotides flanking this core motif can have a major influence on the afiinity of T3R for its response elements. T3Rs regulate transcription in two ways. In the presence of T3, the receptor activates transcription from genes bearing TREs. In the absence of T3, the receptor still binds DNA but acts as a transcriptional repressor of genes bearing TREs. The binding of T3Rs to their response elements is enhanced by auxiliary proteins, of which major forms are identical to RxRs (10)(11)(12)(13)(14)(15). These receptors are abundantly expressed in adults (16) and are activated by their specific ligand 9-cis-retinoic acid (RA) (17, 18). RXRs can heterodimerize with T3Rs, thus influencing the selection of target sequences and modulating the binding of T3Rs to their cognate response elements (10)(11)(12)(13)(14)(15). The enhanced DNA binding and functional activity of these heterodimers seems to require two distinct hormones, suggesting a direct cross-talk between 9-cis-RA and T3, respectively (19, 20h2 Recent reports suggested that nuclear signalling by T3 is also transmitted through RXR-independent pathways (9, 21,22) most likely mediated by homodimers. T3R homodimers exhibit DNA binding specificities that are different only on inverted palindromes from those of the respective T3R-RXFt heterodimers (9h2 Classified by its interactions with response elements, which are determined essentially by the sequence of the DNA binding domain (P-box), T3R belongs to the same subgroup of the nuclear receptor family as RA receptors (RARs), FtXRs, vitamin D receptor (VDR), and peroxisome proliferator-activated receptors (PPARs) (23). All these receptors are known to function only in dimeric complexes, whereas for T3R in vitro binding of monomers to DNA has been reported (24)(25)(26)(27). These T3R monomers may function as precursors of dimeric complexes (21).
The studies described here were undertaken to characterize T,R as a transcription factor that is able to function as a monomer. We found that two 5'-flanking nucleotides preceding the AGGTCA core motif are important for efficient monomer binding. We show that on these octamer binding motifs T3Rs can transactivate in a ligand-induced manner. We also show that the selectivity to 5"flanking sequences influences the binding affinity of dimeric receptor complexes. for each element a pair of oligonucleotides was synthesized, purified, phosphorylated, and annealed to yield double-stranded DNAfragments. These fragments were fused to the tk promoter to drive the expression of the CAT reporter gene by subcloning into the XbaI site of pBLCAT2 (28). Chicken T3Ra and human RXRa cDNAs were subcloned into pSG5 (Stratagene). Cell Culture, E-ansfection, and CATAssays-The Drosophila cell line, SG3 (29), was grown at room temperature in Schneider's medium (Life Technologies, Inc.) supplemented with 15% fetal calf serum. For transfection, 5 x lo5 celldwell on a six-well plate were grown overnight in Schneider's medium supplemented with 5% charcoal-treated fetal calf serum. Liposomes were formed by incubating 1 pg of reporter plasmid, CAT assays were performed as described (30). The CAT activities were normalized to P-galactosidase activity, and induction factors were calculated as the ratio of CAT activity of ligand-stimulated cells to that of mock-induced control. Each condition was P .alyzed in triplicates and data are shown as mean with standard deviation.

DNA Constructs-For
In Vitro Panslation of Nuclear Receptors and DNA Binding Assays-Linearized cDNAs for T3Ra and RXRa were used for in uitro transcription as recommended by the supplier (Promega). Five pg of each RNA was mixed with 175 pl of rabbit reticulocyte lysate, 100 units of RNasin, and 20 p~ complete amino acid mixture (all from Promega) in a total volume of 250 pl and incubated at 30 "C for 180 min. The response element probes were prepared by double digestion of the respective plasmid DNAwith Hind111 and BamHI and purified and quantified by gel electrophoresis. All DNAtemplates were labeled by a fill-in reaction using [a-"P]dCTP and T7 DNA polymerase (Pharmacia). Five pl of in vitro translated receptors were preincubated for 10 min a t room temperature in a total volume of 20 pl of binding buffer (10 m~ Hepes (pH 7.9). 80 m~ KCI, 1 m M dithiothreitol, 0.2 pglpl poly(d1-dC) and 5% glycerol). About 1 ng of labeled probe (25,000 cpm) was added and the incubation was continued for 20 min. For Scatchard analysis 5 pl of in vitro translated T3R was incubated with different concentrations of probe ranging from 0.1 to 10 ng as described (26). Protein-DNA complexes were resolved on a 5% nondenaturing polyacrylamide gel (at room temperature) in 0.25 x TBE (22 m~ Tris, 22 m~ boric acid, 0.5 m~ EDTA (pH 8.3)). After exposing to a film to localize the free probe and protein-probe complexes, the respective bands were excised and counted directly in a scintillation counter. Extracts from SL.3 cells were prepared as described (31).

RESULTS
Orphan receptors that bind to DNA as monomers like, e.g. NGFI-B and FTZ-F1, show a preference for specific sequences in vitro binding affinity to T3R monomers. Thus, whereas a purine in the -1 position is preferred, a pyrimidine at the -2 position of the AGGTCA core motif is critical. As a control the same 16-monomer elements were incubated with in vitro translated RXRa (Fig. 1B). We observed only very little RXR monomer-DNA complexes on the elements preceded by the pairs A A , CA, and TA, suggesting that RXRs prefer an A residue in the -1 position of the AGGTCA half-site.
To quantify the binding affinity of T3R monomers, we performed Scatchard binding analysis on the complexes of the electrophoretic mobility shift assay. We plotted the ratio between T3R-bound and free DNA with respect to T3R-bound DNA. T3R monomers have dissociation constants (Kd) of 2.9 and 2.1 nM to the probes CG and TG, respectively (Fig. 2). These Kd values represent physiologically relevant affinities not much higher than those of FTZ-F1 or NGFI-B for their cognate DNAelements (32,331. In contrast, the probes AG and GG gave greater Kd values of 10.3 and 15.3 m, respectively. To measure the individual transcriptional activity of T3R and RXR on these response elements, we fused each of the 16 monomer response elements to the tk promoter driving the CAT gene. Since Drosophila cells are devoid of mammalian nuclear receptors by definition, we carried out our studies on T3R-and RXR-mediated transactivation in the Drosophila cell line SG3. However, we observed basal CAT activity on some of the response elements, even if only an empty expression vector was co-transfected with the reporter vector. We hypothesize that endogenous Drosophila nuclear receptors bind as monomers and activate the reporter gene. Therefore, we subtracted these basal values from the CAT activities measured for T3Ra or RXRa-driven expression. Based on these corrected CAT activities the -fold induction of ligand versus mock stimulation was determined for each response element (Fig. 3). The transactivation data closely parallel the in vitro DNA binding affinities of T3R for each of the 16 monomer response elements, suggesting a direct correlation between efficiency of monomer binding and transactivation activity. Accordingly, in SG3 cells expressing RXRs transiently none of the 16 monomeric response ele- ments provided substantial induction of CAT activity by 9 4 s -

RA.
We wished to exclude the possibility that the transactivating activity of T3R monomers that we observe in SL-3 cells is due to heterodimerization with an endogenous Drosophila receptor like, e.g. the RXR homologue ultraspiracle (usp) (341, which has been reported to form heterodimers with T3Rs (35). Therefore, binding of in vitro translated T3R to the monomeric half-site that provides strongest interaction, TGAGGTCA, was analyzed in the presence and absence of in vitro translated RXRa, usp, or SG3 cell extract (Fig. 4). We did not observe any enhanced T3R monomer-DNA complex formation nor could we detect any  T3   T3R RXR T3R T3R RXR T3R T3R RXR T3R  T3RRXR T3R T3R RXR  T3R  change in migration of the T3R.DNA complex brought about by demonstrate that T3R monomers are able to bind, and therefore higher order aggregations with auxiliary proteins like RXR, also to transactivate, in the absence of RXR or Drosophila usp, or other Drosophila proteins. Neither RXFt nor any Dro-"accessory factors." sophila proteins bound to the TG element. Interestingly, in the Having observed that T3Rs, and even RXFts, show preferpresence of T3 the mobility of the T3R.DNA complex was sig-ences for selected 5"flanking sequences when they bind to DNA nificantly increased, suggesting a ligand-induced conforma-as monomers (Fig. 11, we next investigated whether this might tional change of T3R monomers. Taken together, these data be also true for T3R homodimers and T3R-RXR heterodimers.

TSR Monomers
To this end, we have chosen the TRE of Moloney murine leukemia virus (36), which constitutes a direct repeat of two (G/ A)GGTC(A/C) half-sites spaced by four nucleotides (DR4). Its 5'-flanking nucleotide pairs are CA and TC, respectively, which we have shown to direct some T3R monomer binding affinity. On CAAGGTCA we also saw faint RXR monomer binding (Fig.  1). To gain more T3R and RXR specificity, we designed mutant Moloney murine leukemia virus TREs having the four possible combinations of TG and AA as 5'-flanking nucleotide pairs (see "Materials and Methods"). Analyzing the binding of T3R homodimers and T3R-FER heterodimers to the five different DR4s (Fig. 5.41, we observed the highest affinity of T3R homodimers to DfLlTGrPa, marked binding to the natural TRE f D R 4 c m ) and to DR&,u, but only low affinity to DR4and to D R~A A I A A .
In contrast, we observed highest affinity of T3R-RXR heterodimers to DR4-G, marked affinity to the Moloney murine leukemia virus TRE (DR4cm), and lower &hities to the other DWs. This analysis shows that although response elements, like DWhave a high affinity to heterodimers and only a low affinity to homodimers, other response elements like DR4TGnT. exhibit inverse specificity. We also performed CAT assays in SL-3 cells to monitor functional activity of ligand-induced T3R homodimers and T3R-RXR heterodimers on all 5 DR4s (Fig. 5B). The data supported the in vitro DNA binding analysis: highest -fold stimulation of T3R homodimers by T3 was observed on DR+GmG, whereas T3and 9-cis-RA-induced T3R-RXR heterodimers were found to be most active on the natural TRE and on DW-,.

DISCUSSION
Biological responses to T3 are mediated by its i n t~c e l l~a r receptor proteins. One of the advances in understanding the molecular mechanism by which nuclear receptors operate was provided by establishing the hexameric consensus half-site AG-GTCA as the core binding motif of response elements for T3Rs, RXRs, RARs, and PPARs (?,8). However, it has not been generally noted that 5"flanking nucleotides can determine the affinity of T3Rs for their T m . Here, we demonstrated that the octameric half-site TGAGGTCA is bound most strongly by T3R monomers. Comparing original and mutated half-sites of the rat growth hormone gene promoter, Kim et al. (37) reported that the TRE half-site sequence CTGAGGTMCG is bound by T3Ra with highest &lnity. The sequence fits the consensus motif (T/CX~G)AGGTCA we have identified. Moreover, we observed on this response element up to 10-fold induction of reporter gene activity in response to T3 stimulation. We showed that this transactivational activity is mediated by T3R monomers, apparently acting independent of any auxiliary protein.
Our finding suggests T,R as the first ligand-inducible monomeric transcription factor known. Its half-site specificity is modulated in major parts by the nucleotides preceding the AGGTCA half-site motif reminiscent of some orphan nuclear receptors like NGFI-B, F"I' %Fl , and Rev-ErbAa. Most classical nuclear receptors do not bind to DNA sequences that contain a single half-site, because the interaction of one monomer with one half-site does not provide sufficient free energy to stabilize the protein-~NA complex. In contrast, some orphan receptors are known to bind to their response elements exclusively as monomers. In this sense, TSRs are an exception, as they bind to their response elements both as monomers and as dimers. We recently identified a n orphan nuclear receptor, called RZR (381, which binds to certain response elements both as monomer and dimer. bind efficiently to DNA suggest a two-step mechanism for the recognition of response elements. Accordingly, T3R monomers screen the DNA for appropriate octameric half-sites. After having identified such a site a second receptor molecule, either T3R or R X R , binds in appropriate vicinity to facilitate dimerization. In this sense, T3R monomers may be interpreted as active precursors of dimeric complexes. Following the 3-4-5 rule (8), all DR4s, including the five we studied, should be equally good TREs. However, we observed that the affinity of homo-and heterodimeric T3R complexes to appropriate response elements composed of two hexameric consensus half-sites was modulated by 5'-flanking nucleotide pairs preceding the two half-sites. Highest atKnity of T3R-RXR heterodimers was observed only, if the 5'4anking sequence of the upstream half-site was optimal for RXRs, suggesting that in the heterodimeric complex RXR binds to the upstream half-site motif. This inte~retation is supported by two recent reports (39,40) that showed that in heterodimeric complexes with T3Rs, RXR always binds to the upstream half-site. Moreover, our observations indicate that the binding of T3R homodimers is even preferred over heterodimer binding, if both half-sites are preceded by nucleotide pairs, like TG, that are favored by T3R monomers. The discrimination of 5"flanking sequences as described here suggests a mechanism to distinguish T3 signaling pathways mediated by either T3R homodimers or TsR-RXR heterodimers. For VDRs we recently reported a different mechanism (31); VDR homodimers and VDR-RXR heterodimers discriminate their respective response elements by the number of spacing nucleotides between directly repeated half-site motifs. Taken together, not only the number of spacing nucleotides, but also the sequence of the spacer is important for response element recognition. Future investigations following the indications for the existence of RAR homodimers in vivo (9, 20, 41) will show whether the homo-and heterodimeric pathways of retinoid action are also defined at the level of one of these two mechanisms.