A Novel Multifunctional Motif in the Amino-terminal A/B Domain of T3Rα Modulates DNA Binding and Receptor Dimerization*

We reported previously that deletion of the 50-amino acid NH2-terminal A/B domain of the chicken (c) or rat thyroid hormone (T3) receptor-α (T3Rα) decreased the T3-dependent stimulation of genes regulated by native thyroid hormone response elements (TREs). This requirement of the NH2-terminal A/B domain for transcriptional activation was mapped to amino acids 21–30 of cT3Rα. Expression of transcription factor IIB (TFIIB) in cells was shown to enhance T3-dependent transcriptional activation by cT3Rα, and this enhancement by TFIIB was dependent on the same 10-amino acid sequence. In vitro binding studies indicated that cT3Rα interacts efficiently with TFIIB, and this interaction requires amino acids 23KRKRK27 in the A/B domain. In this study we document the functional importance of these five basic residues in transcriptional activation by cT3Rα, further supporting the biological significance of these residues and their interaction with TFIIB. Interestingly, we also find that the same amino acids also affect DNA binding and dimerization of cT3Rα. Gel mobility shift assays reveal that a cT3Rα mutant that has all five basic amino acids changed from23KRKRK27 to 23TITIT27binds to a palindromic TRE predominantly as a homodimer, whereas cT3Rα with the wild-type23KRKRK27 sequence binds predominantly as a monomer. This results from both a marked decrease in the ability of the cT3Rα mutant to bind as a monomer and from an enhanced ability to dimerize as reflected by an increase in DNA-bound T3R-retinoic X receptor heterodimers. These effects of23KRKRK27 on DNA binding, dimerization, transcriptional activation, and the association of T3Rα with TFIIB support the notion that this basic amino acid motif may influence the overall structure and function of T3Rα and, thus, play a role in determining the distinct context-dependent transactivation potentials of the individual T3R isoforms.

Steroid, retinoid, and thyroid hormone nuclear receptors are ligand-dependent transcription factors that couple extracellular signals directly to transcriptional responses. These receptors activate or repress transcription of target genes by binding to specific DNA sequences referred to as hormone response elements (HREs) 1 (1). The nuclear receptor superfamily can be divided into the steroid hormone receptor family and the thyroid hormone/retinoid receptor family (1,2), which includes receptors that mediate the effects of thyroid hormone [L-triiodothyronine (T 3 ) (the T 3 Rs), all-trans-retinoic acid (the RARs), 9-cis-retinoic acid (the RARs and RXRs), and 1,25-dihydroxyvitamin D 3 as well as several orphan receptors (e.g. COUP-TF, c-erbA␣2) whose ligand(s), if any, are unknown (3)(4)(5).
The T 3 Rs are encoded by two distinct but closely related genes (␣ and ␤) which, in humans (h), map to chromosomes 17 and 3, respectively (6). Each gene expresses several alternatively spliced isoforms. The T 3 R␣ gene in the rat (r) and man expresses the T 3 -binding isoform T 3 R␣1 along with c-erbA␣2, which does not bind T 3 because of alternative splicing at the COOH terminus (3,7,8). The closely related chicken (c) ␣ gene expresses only cT 3 R␣, which is more than 90% similar at the amino acid level to rT 3 R␣1 and hT 3 R␣1 (6,9,10). The T 3 R␤ gene expresses T 3 R␤1 and T 3 R␤2 that differ only in their NH 2 -terminal A/B regions, which are distinct from the A/B region of T 3 R␣1 (3,11). Except for the A/B domains, the T 3 R␣ and T 3 R␤ receptors are more than 90% similar at the amino acid level. Thus, three T 3 Rs are expressed which differ primarily in the A/B domain, suggesting that this region may play a role in mediating different effects of these receptors.
One of the central issues in understanding the actions of the T 3 Rs and other nuclear receptors is elucidation of the details by which target genes are recognized. The T 3 Rs and certain other members of thyroid hormone/retinoid receptor family bind to their HREs as monomers, homodimers (12)(13)(14)(15)(16), or as heterodimers with the RXRs (17)(18)(19)(20)(21)(22)(23)(24). In particular, the T 3 Rs bind to and activate transcription from a wide variety of response elements organized as direct repeats (DR), inverted repeats (IR), or everted repeats (ER) of the optimized AGGTCA hexanucleotide half-site (25)(26)(27) and from native half-site motifs that diverge from the AGGTCA core binding sequence (28).
Recognition of specific base pairs within the half-site core binding motif is mediated by the highly conserved DNA binding domain (DBD), which defines the nuclear receptor superfamily. This highly conserved DBD contains 66 -68 amino acids that are organized into two zinc finger structures that include 9 perfectly conserved cysteines followed by a carboxyl-terminal extension (29,30). A helix in the carboxyl-terminal extension, with its extensive minor groove contacts, effectively extends the contact surface of the DBD beyond the consensus 6-base pair half-site (31). The ability of nuclear hormone receptors to distinguish among specific HREs is conferred by 3 amino acids at the base of the first zinc finger in the DBD (the P box) (32). This region is organized into an ␣-helix that penetrates the major groove and recognizes the specific nucleotide sequence of the HRE.
Amino acids at the base of the second zinc finger (the D box) are thought to provide a dimerization interface for proteinprotein interactions on certain HREs (32). Structural studies indicate that the DBDs of certain thyroid hormone/retinoid family members form a cooperative dimerization interface, when bound to DRs but not to IRs and ERs (31). For IRs and ERs, binding of homodimers or heterodimers is thought to result from a dimerization interface located within the ligand binding domain. In addition to homodimer and heterodimer binding, the T 3 Rs also bind IRs, DRs, and other DNA configurations as monomers (14). In the absence of RXR, T 3 R␣ binds more efficiently as monomers to these elements, and T 3 R␤ isoforms bind more efficiently as homodimers (33).
We reported previously that a 10-amino acid sequence within the A/B domain of cT 3 R␣ or rT 3 R␣1 was essential for liganddependent activation of native HREs and for interaction of T 3 R␣ with TFIIB (34). Interestingly, deletion of the 50-amino acid A/B domain of cT 3 R␣ markedly reduced monomer binding and increased homodimer binding of the receptor, suggesting that the A/B domain of T 3 R␣ imposes preferential monomer binding of the receptor (34). In this study we show that the same 5 basic amino acids 23 KRKRK 27 which are necessary for efficient binding to TFIIB are required for transcriptional activity of cT 3 R␣. These same amino acids are also responsible for imposing preferential monomer binding and influencing the efficiency of heterodimer formation with RXR. To our knowledge this is the first identification of specific NH 2 -terminal residues involved in the differential binding of T 3 R isoforms to DNA.

EXPERIMENTAL PROCEDURES
Plasmids-⌬MTV-TREp-CAT (14,25), ⌬MTV-TRE-GH-CAT (25), and ⌬MTV-TRE-Mal-CAT (35) have been described previously. These CAT reporter genes contain a single copy of each TRE cloned into the HindIII site at Ϫ88 of ⌬MTV-CAT, a mouse mammary tumor viral LTR-CAT reporter that lacks glucocorticoid response elements (25,32). The TREp (also known as the TRE-IR) is an inverted repeat of optimized AGGTCA half-sites (AGGTCA TGACCT). TRE-GH and TRE-Mal are from the rat growth hormone (13) and rat malic enzyme genes (36), respectively, and contain direct and inverted repeats. The TRE 1/2 is the same as the TREp except that it contains a single G to C change in one of the half-sites (AGGTCA to ACGTCA).
Cell Transfections and CAT Assays-HeLa cells were transfected by electroporation (14,34). After transfection, cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 15 mM Hepes, 0.1 mg/ml pyruvate, 50 g/ml streptomycin sulfate, 50 g/ml penicillin (DHAP medium) containing 10% (v/v) hormonedepleted calf serum with and without 1 M T 3 for 48 h (as indicated) before harvesting for CAT assay. CAT expression was measured by determining the extent of [ 14 C]chloramphenicol (50 mCi/mmol; NEN Life Science Products) acetylation as described previously (41,42). Protein concentration was determined (43), and amounts of cell protein assayed for CAT activity were varied to maintain the percentage of [ 14 C]chloramphenicol acetylated in the linear range (Ͻ30%) during a 16-h incubation. CAT activity values were normalized to represent the percent of [ 14 C]chloramphenicol acetylated by a specific amount of cell protein in 16 h at 37°C. All experiments were performed in duplicate, and the results represent the means of at least three independent transfections. Within each transfection, duplicate samples varied less than 10%.
Gel Mobility Shift Assays-cT 3 R␣, cT 3 R␣(21-408), cT 3 R␣(21-408, 7/8), cT 3 R␣(51-408), and the various cT 3 R␣ NH 2 -terminal amino acid mutants were expressed by in vitro translation in reticulocyte lysates (24). A fraction of each translation received L-[ 35 S]cysteine in the incubation, and the amount of each 35 S-synthesized protein was analyzed by SDS-gel electrophoresis and quantitated using a Molecular Dynamics PhosphorImager with ImageQuant software. The number of fmol of each cT 3 R␣ protein synthesized in the lysate was estimated by the relative amounts of the individual 35 S-proteins and quantitation of reticulocyte lysate-synthesized cT 3 R␣ by binding with L-[ 125 I]T 3 (24). Reticulocyte lysate-translated cT 3 R␣ proteins (about 10 fmol/l of lysate) were incubated with 5 fmol (30,000 dpm) of the 32 P-labeled TREp or the TRE 1/2 element. The 30-l incubation mixture contained 25 mM Tris (pH 7.8), 0.5 mM EDTA, 75 mM KCl, 1 mM dithiothreitol, 0.2 g of poly(dI-dC), 0.05% Triton X-100 (v/v), 30 g of ovalbumin, 0.3 M ZnCl 2 , 0.2 g of RNase A, and 10% glycerol (v/v) and either 0.5, 1.5, 3.0, or 4.5 l of reticulocyte lysate (24). Control reticulocyte lysate was added to the 0.5-, 1.5-, or 3.0-l receptor preparations to adjust the final amount of lysate in each sample to 4.5 l. Unless indicated otherwise, a comparison of the different cT 3 R␣ proteins used the same number of fmol of receptor. Samples were analyzed by electrophoresis at 4°C for 50 min in nondenaturing 6% polyacrylamide gels (acrylamide:bisacrylamide, 37.5:1) in buffer containing 10 mM Tris, 7.5 mM acetic acid, 40 M EDTA (pH 7.8) (12,14). The gels were then dried and autoradiographed. The assignment of receptor monomers and homodimers is based on the mobility of purified cT 3 R␣ (14). Certain gel shift studies were quantitated using a PhosphorImager as indicated.
Influence of the NH 2 Terminus of cT 3 R␣ on the Formation of cT 3 R␣⅐RXR Heterodimers in Solution in the Absence of DNA-The binding of 35 S-cT 3 R␣ and mutants to GST and GST-RXR in vitro were carried out as described previously (34). L-[ 35 S]Cysteine-labeled wildtype cT 3 R␣, cT 3 R␣(21-408), and the cT 3 R␣(21-408, 7/8) NH 2 -terminal mutant were prepared using TNT reticulocyte lysates (Promega). 25,000 dpm of 35 S-labeled protein was incubated with 25 ng of GST-RXR or 10 ng of GST protein immobilized on glutathione-agarose beads in 300 l of Buffer A for 1 h at 4°C on an Orbitron rotator. Buffer A consists of 50 mM KCl, 25 mM Hepes (pH 7.9), 6% glycerol, 5 mM EDTA, 5 mM MgCl 2 , 1 mM dithiothreitol, and 0.05% Triton X-100 (34). The amount of GST protein used (10 ng) was equivalent to the amount of GST in the GST-RXR fusion protein (25 ng). Beads were collected by centrifugation at 4°C for 5 min at ϳ500 ϫ g and washed three times with 1 ml of Buffer A. The bound proteins were eluted with SDS-gel loading buffer and analyzed by SDS-gel electrophoresis followed by autoradiography. 35 S-Labeled wild-type or mutant proteins in the bind-  (14). Panel A, no binding to the TREp was detected using 0.5, 1.5, or 4.5 l of control lysate (lanes 1-3, respectively). No homodimer binding was detected with 0.5, 1.5, or 4.5 l of cT 3 R␣-expressing reticulocyte lysate (lanes 4 -6, respectively). Monomer binding was detected using 1.5 l of cT 3 R␣(21-408, 7/8)-and cT 3 R␣(51-408)-expressing lysates (lanes 8 and 11, respectively), and both monomer and homodimer binding were detected using 4.5 l of the same lysates (lanes 9 and 12, respectively). Panel B, no homodimer binding was detected with 1.5, 3.0, or 4.5 l of reticulocyte lysate expressing cT 3 R␣(21-408) (lanes 1-3, respectively). Monomer binding was detected using 1.5 l of lysate expressing cT 3 R␣(21-408, 7/8) (lane 4), and both monomer and homodimer binding were detected using 3.0 and 4.5 l of the same lysate (lanes 5 and 6, respectively). ing assays were analyzed by electrophoresis and autoradiography to ensure that equal amounts of input radioactivity of the labeled protein were used.

Residues 23 KRKRK 27 of the NH 2 -terminal A/B Domain of cT 3 R␣ Are Necessary for Maximal Transcriptional Activation of
Native TREs-Our previous studies aimed at elucidating the functional role of the NH 2 -terminal A/B region of cT 3 R␣ indicated that amino acids 21-30 of cT 3 R␣, which are also conserved in the NH 2 -terminal A/B domains of rT 3 R␣1 and hT 3 R␣1, are essential for both transcriptional activation and for efficient binding to TFIIB (34). More detailed in vitro binding studies revealed that residues 23 KRKRK 27 centered within amino acids 21-30 are required for efficient binding of cT 3 R␣ with TFIIB (34). This suggests that the functional activity that we originally mapped to amino acids 21-30 may depend solely on these 5 basic residues. To test this possibility we compared the functional activities of cT 3 R␣(21-408) and cT 3 R␣(21-408, 7/8) in which 23 KRKRK 27 was changed to 23 TITIT 27 . In transfection experiments with a reporter gene regulated by a single idealized TRE organized as an inverted repeat (TREp) of the optimized AGGTCA half-site (⌬MTV-TREp-CAT), cT 3 R␣(21-408) and cT 3 R␣ (21-408, 7/8) showed similar activity (Fig. 1A). In contrast, with ⌬MTV-TRE-GH-CAT or ⌬MTV-TRE-Mal-CAT, which contain native TREs, cT 3 R␣(21-408) was much more active (Fig. 1, B and C). These results are similar to our previous findings that showed that the functional effect of the NH 2 terminus is much more prominent on reporters containing lower affinity native TREs compared with reporters containing the idealized higher affinity TREp (34). Finally, coexpression of TFIIB, although enhancing the activity of both receptors, results in a much higher level of T 3 -dependent stimulation by cT 3 R␣(21-408) (Fig. 1C). The ability of TFIIB to enhance the activity of cT 3 R␣(21-408, 7/8) partially is not altogether unexpected and is similar to our previous results with cT 3 R␣(51-408). This most likely stems from the low affinity of amino acids 119 -154 of the cT 3 R␣ D region for TFIIB (34).
Residues 23 KRKRK 27 Affect DNA Binding of cT 3 R␣ as Monomers and Homodimers-The NH 2 terminus of T 3 R isoforms may confer cell type and promoter specificity not just through divergent, isotype-distinct, interactions with other transcription factors (34,44) but also through conferring distinct DNA binding properties to receptor isotypes (45,46). For example, the DNA binding properties of v-erbA differ from those of c-erbA (cT 3 R␣) and more closely resemble those of the RARs. The structural basis behind this difference in DNA recognition by v-erbA results from one or more changes within the v-erbA NH 2 -terminal domain (45,47).
T 3 Inhibits Both Monomer and Homodimer DNA Binding of cT 3 R␣ NH 2 -terminal Mutants-Ligand binding to DNA-bound T 3 R is accompanied by a conformational change of the receptor (48) which may alter the T 3 R-DNA interaction (49). For example, T 3 inhibits homodimer formation in vitro on various direct and everted, but not inverted, TREs (49,50). In addition, T 3 increases the electrophoretic mobility of T 3 R monomers and homodimers and of T 3 R-RXR heterodimers on inverted repeat TREs without changing their apparent stability (14,24).
Given the involvement of the NH 2 -terminal domain and its 5 basic amino acids in cT 3 R␣-DNA binding, T 3 -induced conformational changes of cT 3 R␣ may, depending on the physical integrity or structure of the A/B domain, result in distinct receptor DNA binding affinities. To test this possibility, gel mobility shift studies were performed with the TREp using wild-type cT 3 R␣(1-408), cT 3 R␣(51-408), cT 3 R␣(21-408), and cT 3 R␣(21-408, 7/8) in the absence or presence of T 3 . As shown in Fig. 3, cT 3 R␣(1-408) and cT 3 R␣(21-408) in the absence of T 3 bind the TREp predominantly as monomers (lanes 1 and 3). As expected, T 3 increased the electrophoretic mobility of these complexes without altering their apparent stability (lanes 2 and 4). Thus, deletion of the first 20 amino acids from the A/B domain does not affect receptor DNA binding in either the absence or presence of T 3 . In the absence of T 3 , cT 3 R␣(21-408, 7/8) and cT 3 R␣(51-408) bound as homodimers and somewhat less efficiently as monomers (lanes 5 and 7). Surprisingly (lanes 6 and 8), T 3 almost completely eliminates monomer and homodimer DNA binding of cT 3 R␣(21-408, 7/8) and cT 3 R␣(51-408).

NH 2 -terminal Mutants of cT 3 R␣ Bind Poorly to a TRE 1/2 -
The preferential binding of cT 3 R␣(21-408, 7/8) or cT 3 R␣(51-408) to the TREp as a homodimer compared with the predominant monomeric binding of cT 3 R␣(21-408) or wild-type cT 3 R␣ may result from an increased potential to homodimerize and/or a decreased potential to bind DNA as a monomer. This latter instance might result in an increased amount of this mutant receptor available to bind as a homodimer. To assess effects of the 23 KRKRK 27 sequence in the NH 2 terminus on monomer binding, we compared the binding of cT 3 R␣(21-408) and cT 3 R␣(21-408, 7/8) with a TRE 1/2 in the absence and presence of T 3 (Fig. 4). The TRE 1/2 is the same as the TREp except that it contains a single G to C change in one of the half-sites (AGGTCA to ACGTCA). Whereas cT 3 R␣(21-408) binds to the  (lanes 1 and 2), cT 3 R␣(21-408, 7/8) does not bind to this element without or with T 3 (lanes 3 and 4). This suggests that NH 2 -terminal amino acids 23 KRKRK 27 impose preferential monomer DNA binding on cT 3 R␣.
The DNA Binding Properties of cT 3 23 KRKRK 27 -To assess whether all or some of the basic residues within 23 KRKRK 27 are necessary for preferential monomer binding of wild-type cT 3 R␣, we performed gel mobility shift assays with NH 2 -terminal mutants containing different numbers and combinations of basic and neutral amino acid residues. As shown in Fig. 5A, wild-type cT 3 R␣ and cT 3 R␣(21-408) bind the TREp predominantly as monomers (lanes 1 and 3), whereas cT 3 R␣(51-408) and cT 3 R␣(21-408, 7/8) bind predominantly as homodimers (lanes 2 and 4). Interestingly, mutant cT 3 R␣(21-408, 9/10), which has only the first 3 basic amino acid residues changed ( 23 TITRK 27 ), binds the TREp very poorly as a monomer and not at all as homodimer (lane 5). In contrast, mutant cT 3 R␣(21-408, 11/12, lane 6), which has the last 3 basic amino acid residues changed ( 23 KRTIT 27 ), binds the TREp almost as efficiently as cT 3 R␣(21-408, 7/8) or cT 3 R␣(51-408) with slightly more monomer than homodimer. Finally, mutant cT 3 R␣(21-408, 13/14) (lane 7), which has only the middle basic amino acid preserved ( 23 TIKIT 27 ), binds in a way similar that of cT 3 R␣(21-408, 11/12) but less efficiently. These data indicate that the affinity and mode of binding of these mutants to the TREp are not related directly to the number of the basic amino acid residues within amino acids 21-30. However, all 5 basic amino acid residues 23 KRKRK 27 are necessary for predominant high affinity monomer binding to TREp.
The effect of T 3 on the binding of the various NH 2 -terminal mutants to the TREp and the TRE 1/2 is shown in Fig. 5, B and C. The expected pattern of DNA binding of wild-type cT 3 R␣, cT 3 R␣(51-408), cT 3 R␣(21-408), and cT 3 R␣(21-408, 7/8) in the absence or presence of T 3 is shown in lanes 1-8 of Fig. 5, B and C. cT 3 R␣(21-408, 9/10) in the absence of T 3 binds poorly as a monomer and not at all as a homodimer, and this is reduced further by T 3 (lanes 9 and 10 in Fig. 5, B and C). Homodimer binding of cT 3 R␣(21-408, 11/12) and cT 3 R␣(21-408, 13/14) to the TREp is abolished by T 3 (lanes 11-14 in Fig. 5B), and monomer binding is reduced markedly (lanes 11-14 in Fig. 5, B and C).
Amino acids 23 KRKRK 27 Affect the Efficiency of T 3 R⅐RXR Heterodimer Formation-Although amino acids 23 KRKRK 27 influence monomer binding of receptor to DNA, these residues may also affect the ability of the receptor to dimerize. We examined this possibility by assessing the ability of cT 3 R␣(21-408) or cT 3 R␣(21-408, 7/8) to bind as heterodimers with RXR on the TREp (Fig. 6). Incubations were performed either in the absence or the presence of T 3 and/or 9-cis-RA. As shown previously, cT 3 R␣(21-408) alone binds to the TREp predominantly as a monomer, and the electrophoretic mobility of this complex is increased by T 3 (Fig. 6, lanes 1 and 2). In contrast, cT 3 R␣(21-408, 7/8) binds the TREp predominantly as a homodimer, and this complex is abolished by T 3 (lanes 7 and 8). Interestingly, in the absence of T 3 cT 3 R␣(21-408) forms heterodimeric complexes with RXR less efficiently than cT 3 R␣(21-408, 7/8) (lanes 3 and 9), and T 3 increases the electrophoretic mobility of both complexes (lanes 4 and 10). 9-cis-RA alone does not affect either complex ( lanes 5 and 11), whereas the combination of T 3 and 9-cis-RA affects the mobility of both complexes as with T 3 alone (lanes 6 and 12). Similar results were also found using a 32 P-DRϩ4 containing AGGTCA half-sites instead of the TREp (data not shown). Importantly, these results indicate that amino acids 23 KRKRK 27 affect both monomer DNA binding and dimerization.
This increase in bound heterodimers of the cT 3 R␣ mutant may reflect enhanced heterodimer formation and/or the binding of preformed heterodimer to DNA or a decrease in the "off rate" of prebound complexes. The stability of heterodimeric complexes between RXR and cT 3 R␣(21-408) or cT 3 R␣(21-408, 7/8) on the labeled TREp was examined by incubating preformed complexes for various times with a 1,000-fold excess of unlabeled TREp (Fig. 7A). The results shown in lanes 2-5 and 7-10 of Fig. 7A were quantitated using a PhosphorImager (Fig.  7B). Only ϳ20% of the total TREp-bound cT 3  prebound to the TREp does not differ significantly (Fig. 7B).
The results of Fig. 7B imply that the observed difference in the efficiency of heterodimer binding to the TREp results from an intrinsic difference in the formation rate and/or DNA binding of these complexes. To provide evidence for enhanced formation, we studied the binding of 35 S-cT 3 R␣(21-408) and 35 S-cT 3 R␣(21-408, 7/8) to GST and GST-RXR in the absence of DNA (Fig. 8). GST alone did not bind to either of the 35 S-cT 3 R␣ proteins. GST-RXR bound the NH 2 -terminal mutant more efficiently than 35 S-cT 3 R␣(21-408), and this difference is similar to the relative increase in binding of cT 3 R␣(21-408, 7/8)⅐RXR heterodimers to the TREp as shown in Fig. 7. 3 R␣-A conserved sequence of 10 amino acids from NH 2terminal A/B domain of cT 3 R␣, rT 3 R␣1, and hT 3 R␣1 is essential both for transcriptional activation and for efficient binding to TFIIB (34). Within this sequence, 5 basic amino acid residues, 23 KRKRK 27 , proved absolutely necessary for efficient receptor binding to TFIIB (34). A sequence similar to the 23 KRKRK 27 sequence found in T 3 R␣ is not found in the A/B domains of T 3 R␤1 or T 3 R␤2. Indeed, we have found that removal of the A/B domain of rat T 3 R␤1 does not alter its ability to interact with TFIIB or to activate the rat TRE-GH or TRE-Mal in ⌬MTV-CAT (data not shown). However, all nuclear receptors contain a conserved basic sequence that follows their DBD, which appears to play a role in TFIIB binding (34).

Amino Acids 23 KRKRK 27 of the A/B Domain Are Required for Maximal Transcriptional Activation of Native TREs by cT
In this study we show that the functional role of the A/B domain in ligand-dependent transcriptional activation by cT 3 R␣ requires the 5 basic amino acid residues 23 KRKRK 27 . This was determined by directly comparing the activities of cT 3 R␣(21-408) and cT 3 R␣(21-408, 7/8) in which all 5 basic residues ( 23 KRKRK 27 ) were changed to 23 TITIT 27 . Like cT 3 R␣(51-408), which lacks the entire NH 2 -terminal A/B domain, cT 3 R␣(21-408, 7/8) was only slightly less active than cT 3 R␣(21-408) in transient transfection experiments with a reporter gene containing a single idealized TRE organized as an inverted repeat (⌬MTV-TREp-CAT; 34 and this study, Fig.  1A). However, in transfection experiments where native TREs from the rat growth hormone or malic enzyme gene promoters were used, cT 3 R␣(21-408, 7/8) was much less active in the presence or absence of TFIIB (Fig. 1, B and C). Indeed, the requirement for amino acids 21-30 of the A/B domain for optimal transcriptional activation of native TREs appears to depend solely on amino acids 23 KRKRK 27 . cT 3 R␣(21-408, 13/14), where 23 KRKRK 27 was changed to 23 TIKIT 27 , was somewhat more active than cT 3 R␣(21-408, 7/8) but much less active than wild-type cT 3 R␣(1-408). The other 23 KRKRK 27 mutants were closer in activity to cT 3 R␣(1-408) (not shown) even though their affinity for TFIIB is reduced (34). This probably reflects an integrative effect of the lower affinity of the mutants for TFIIB (34) and, as we show in this study, their increased ability to bind to response elements as homodimers and as heterodimers with RXR.
Amino Acids 23 KRKRK 27 Affect DNA Binding of cT 3 R␣ as Monomers and Homodimers-Previous studies from our laboratory indicated that the NH 2 -terminal region of cT 3 R␣ might affect DNA binding of receptor to the TREp and the native TRE from the rat growth hormone gene promoter by altering the extent of monomer binding and dimerization (34). Hollenberg et al. (51) also found that the NH 2 terminus of hT 3 R␣1 reduced the ability of the receptor to dimerize. Our current study documents that the basic amino acid sequence 23 25,000 dpm of 35 S-labeled T 3 R␣(21-408) and cT 3 R␣(21-408, 7/8) were incubated with glutathione-agarose bound GST-RXR or GST in 300 l of Buffer A (34) for 1 h at 4°C as described under "Experimental Procedures." Beads were collected by centrifugation at 4°C for 5 min at ϳ500 ϫ g and washed three times with 1 ml of Buffer A. The bound proteins were eluted with SDS gel loading buffer and analyzed by SDS-gel electrophoresis followed by autoradiography. 35 S-Labeled wild-type or mutant proteins in the binding assays were analyzed by electrophoresis and autoradiography to ensure that equal amounts of input radioactivity of the labeled proteins were used. addition to its role in transcriptional activation and TFIIB binding. First, both cT 3 R␣(21-408, 7/8), in which amino acids 23 KRKRK 27 were changed to 23 TITIT 27 , and cT 3 R␣  bind to the TREp more efficiently as homodimers than cT 3 R␣ ( Fig. 2A). More direct evidence for the involvement of these residues in the DNA binding of receptor is provided by the finding that cT 3 R␣(21-408) binds predominantly as a monomer even at high concentrations, whereas cT 3 R␣(21-408, 7/8) binds efficiently as a homodimer even at low concentrations ( Fig. 2A). We obtained results in gel shift studies using a DRϩ4 containing AGGTCA half-sites similar to those with the TREp (not shown). Interestingly, although the binding of cT 3 R␣ and cT 3 R␣(21-408) to the TREp is not affected significantly by T 3 , the binding of cT 3 R␣(21-408, 7/8) and cT 3 R␣(51-408) to this element is inhibited strongly (Fig. 3). Hence, residues 23 KRKRK 27 may either stabilize a "proper" conformation of the DBD, or they may ensure optimal positioning of the DBD with respect to the TREp half-site. These alternative mechanisms may occur either through a direct interaction between amino acids 23 KRKRK 27 and the DBD or through interaction of these basic residues with some other region of the receptor such as the ligand binding domain, which may affect the structure of the DBD. Alternatively, residues 23 KRKRK 27 may affect both the integrity and positioning of the DBD and in addition may contact DNA directly. Irrespective of the actual mechanism by which amino acids 23 KRKRK 27 affect receptor DNA binding, they seem absolutely necessary in the presence of T 3 for the receptor to bind to the TREp (Fig. 3).
The predominant DNA binding of cT 3 R␣(21-408, 7/8) as a homodimer may reflect decreased monomer binding potential, increased homodimerization potential, or a combination of both. DNA binding studies using the TRE 1/2 to preclude homodimer binding suggest that amino acids 23 KRKRK 27 affect the ability of receptor to bind DNA as a monomer (Fig. 4). Although cT 3 R␣(21-408, 7/8) does not bind significantly to the TRE 1/2 , it does bind to the TREp as a monomer (e.g. compare Fig. 4, lane 3, Fig. 5C, lane 7 with Fig. 2A, lane 8, Fig. 2B, lane  4, Fig. 3, lane 5). One possible explanation for this finding would be that to bind to DNA cT 3 R␣(21-408, 7/8) has to make an initial contact exclusively as a homodimer. Once this homodimer-DNA contact is established one cT 3 R␣(21-408, 7/8) molecule could dissociate, leaving a relatively unstable monomer⅐DNA complex behind.
DNA Binding of cT 3 23 KRKRK 27 Sequence-The basic residues 23 KRKRK 27 are not equally important for preferential monomer binding of wildtype ␣-receptor. For example, cT 3 R␣(21-408, 9/10), which has amino acids 23 TITRK 27 , binds to the TREp very poorly and only as a monomer (Fig. 5). In contrast, cT 3 R␣(21-408, 11/12), which has amino acids 23 KRTIT 27 , binds to the TREp in a way similar to that of cT 3 R␣(51-408) with slightly more monomer than homodimer. Finally, cT 3 R␣(21-408, 13/14), which has amino acids 23 TIKIT 27 , binds to the TREp in a way similar to that of cT 3 R␣(21-408, 11/12) but less efficiently. Hence, the affinity and the mode of binding of these mutants to the TREp are not related directly to the number of the basic amino acid residues within the sequence 23-27, and therefore these residues do not contribute equally to the receptor DNA binding. However, all 5 basic amino acid residues 23 KRKRK 27 are necessary for predominant high affinity monomer binding to TREp. In contrast, the affinity of cT 3 R␣ for TFIIB does correlate directly with the number of these basic amino acids (data not shown).

R␣ as Monomers and Homodimers Is Influenced by Different Basic Amino Acids within the
The NH 2 -terminal 23 KRKRK 27 Sequence Influences the Extent of cT 3 R␣⅐RXR Heterodimer Formation-In addition to the effect of residues 23 KRKRK 27 on the binding of receptor monomers to DNA, these residues also influence the intrinsic dimerization potential of T 3 R␣. Fig. 6 indicates that cT 3 R␣(21-408, 7/8) binds as a heterodimer with RXR to the TREp element much more efficiently than cT 3 R␣(21-408). Thus, ϳ90% of the total TREp-bound cT 3 R␣(21-408, 7/8) is bound as a heterodimer with RXR, whereas only ϳ20% cT 3 R␣(21-408) participates in such complexes (Fig. 7A, lanes 2 and 7). Both complexes showed similar dissociation rates in the presence of a 1,000-fold excess of unlabeled TREp (Fig. 7A, lanes 2-5 and  7-10, and Fig. 7B), suggesting that the increased amount of heterodimers found with the cT 3 R␣ mutant results from the more efficient formation and/or binding cT 3 R␣(21-408, 7/8)⅐RXR heterodimers to DNA. GST binding studies (Fig. 8) suggest that this increase results from the more efficient formation of heterodimers in solution, which then bind to DNA. The increased dimerization potential of the cT 3 R␣(21-408, 7/8) on the TREp does not result in enhanced activation of the ⌬MTV-TREp-CAT reporter compared with cT 3 R␣(1-408) or cT 3 R␣(21-408) (Fig. 1A). This discrepancy may reflect the weaker interaction of cT 3 R␣(21-408, 7/8) with TFIIB, which could result in a number of transcriptionally active cT 3 R␣(21-408, 7/8)⅐RXR⅐TFIIB complexes on the optimized element similar to those with cT 3 R␣(21-408) or wild-type cT 3 R␣(1-408).

Influence of the NH 2 -terminal A/B Domain on DNA Binding of Other Members of the Thyroid Hormone/Retinoid Receptor
Subfamily-Several additional reports have provided evidence for involvement of the NH 2 termini of related members of the nuclear hormone receptor family in DNA binding. Two NH 2terminal amino acids of v-erbA, His 12 and Cys 32 (which correspond to Arg 24 and Tyr 44 of cT 3 R␣) have been shown, in conjunction with amino acid changes in the zinc finger domain, to contribute to a restricted half-site DNA binding specificity (45,47,52). RXR␣ and RXR␥, but not RXR␤, have been suggested to activate transcription by forming tetrameric complexes on DNA elements consisting of four reiterated weak half-sites (53). These isoform-specific DNA binding properties mapped to the NH 2 -terminal A/B domains. Replacing the RXR␤ A/B domain with that of RXR␥ resulted in both tetramer binding to DNA and transcriptional activation by the chimeric protein.
That the NH 2 -terminal domain and the zinc finger region of nuclear hormone receptors may functionally cooperate was also shown by a study of the DNA binding properties of the RORs (46,54). Thus, the differential DNA binding activities of ROR␣1, ROR␣2, and ROR␣3 depend on their distinct NH 2 termini, which, when fused to heterologous nuclear hormone receptors (e.g. hT 3 R␤1), may impose novel DNA binding specificities. Finally, T 3 R␣ and T 3 R␤ bind differently to the same DNA element (33). Whereas T 3 R␣ binds predominantly as a monomer to the TREp (14), T 3 R␤1 binds predominantly as a homodimer, which is thought to result in part from its increased ability to dimerize (33,51,55). Importantly, T 3 R␤1 and T 3 R␣ show no homology in their NH 2 -terminal A/B domains. In particular, T 3 R␤1 does not contain a sequence in its A/B domain similar to the T 3 R␣-specific KRKRK, which may account for the finding that mutant cT 3 R␣(21-408, 7/8) exhibits certain DNA binding properties more like T 3 R␤1 than wild-type cT 3 R␣.
In conclusion, the influence of the NH 2 terminus of T 3 R␣ on transcriptional activation, dimerization, and DNA and TFIIB binding supports the idea that this domain may play a role in the selective regulation of specific genes and impart distinct context-dependent transactivation potential to the individual receptor isoforms. The three-dimensional structural studies of the T 3 Rs performed thus far in either the absence (56) or in the presence of DNA (31) have not included the NH 2 -terminal A/B domain. The marked effect of the KRKRK residues of T 3 R␣ on both DNA binding and transactivation suggests important effects of this sequence on receptor structure. Although this interaction may involve the DBD, our finding that mutation of the KRKRK residues alters the effect of ligand on DNA binding also suggests an interaction between the NH 2 terminus and the ligand binding domain (Figs. 3 and 5). Thus, structural studies that analyze homo-or heterodimerization or other DNA binding properties of the T 3 Rs need to include receptor moieties containing the NH 2 -terminal A/B domain. Future studies that focus on the similarities and differences in the function of the NH 2 -terminal regions of the T 3 Rs and the related retinoid receptor isoforms should provide important insights into the roles and mechanisms of action of these diverse receptor forms in gene regulation and development.