Thyroid hormone receptor variant alpha2. Role of the ninth heptad in dna binding, heterodimerization with retinoid X receptors, and dominant negative activity.

Thyroid hormone receptors bind DNA with highest affinity as heterodimers with retinoid X receptors, and such heterodimers generally are thought to be the biological mediators of thyroid hormone action. An alternative splice product of the thyroid hormone receptor a gene, thyroid hormone receptor variant a2, does not bind thyroid hormone and functions as a weak dominant negative inhibitor of thyroid hormone action. Thyroid hormone receptor variant a2 is missing one-half of the ninth heptad, a region of the bona fide receptor thought to be important for heterodimerization with retinoid X receptors. The role of the ninth heptad in heterodimerization has been evaluated further. Thyroid hormone receptor variant a2-retinoid X receptor heterodimers form on a subset of direct repeat response elements but not on palindromic or inverted palindromic elements. Restoration of the missing ninth heptad sequence is critical for restoring heterodimerization on the palindromic DNA, but either the ninth heptad amino acids or a stretch of alanines is equally able to restore heterodimerization on the inverted palindrome. Thus, the role of the ninth heptad in heterodimerization differs on direct repeat, palindromic, and inverted palindromic response elements, suggesting that the protein-protein interactions differ on each of these elements. The dominant negative activity of thyroid hormone receptor variant a2 requires DNA binding, but the relatively weak nature of the dominant negative activity is only partially explained by the weak DNA binding.

Thyroid hormone receptors bind DNA with highest affinity as heterodimers with retinoid X receptors, and such heterodimers generally are thought to be the biological mediators of thyroid hormone action. An alternative splice product of the thyroid hormone receptor ␣ gene, thyroid hormone receptor variant ␣2, does not bind thyroid hormone and functions as a weak dominant negative inhibitor of thyroid hormone action. Thyroid hormone receptor variant ␣2 is missing one-half of the ninth heptad, a region of the bona fide receptor thought to be important for heterodimerization with retinoid X receptors. The role of the ninth heptad in heterodimerization has been evaluated further. Thyroid hormone receptor variant ␣2-retinoid X receptor heterodimers form on a subset of direct repeat response elements but not on palindromic or inverted palindromic elements. Restoration of the missing ninth heptad sequence is critical for restoring heterodimerization on the palindromic DNA, but either the ninth heptad amino acids or a stretch of alanines is equally able to restore heterodimerization on the inverted palindrome. Thus, the role of the ninth heptad in heterodimerization differs on direct repeat, palindromic, and inverted palindromic response elements, suggesting that the protein-protein interactions differ on each of these elements. The dominant negative activity of thyroid hormone receptor variant ␣2 requires DNA binding, but the relatively weak nature of the dominant negative activity is only partially explained by the weak DNA binding.
The metabolic actions of thyroid hormone (3,5,3Ј-triiodothyronine (T 3 )) 1 are initiated by the binding of T 3 to nuclear thyroid hormone receptors (TRs), which are members of a large family of zinc finger transcription factors that includes the receptors for all known steroids, retinoids, and vitamin D (see Refs. 1-3 for recent reviews). Thyroid hormone receptors are transcribed by two genes, TR␣ and TR␤. TR␤ produces two functional TRs (TR␤1 and -␤2) that contain identical DNA and ligand binding domains but that possess unique amino-terminal domains (Fig. 1). TR␣ also produces two proteins (Fig. 1), but only one of them (TR␣1) is a functional TR. The alternative splice product TR variant ␣ 2 (TRv␣2) contains a unique 122amino acid sequence in place of the carboxyl-terminal portion of the TR␣1 ligand binding domain; hence, TRv␣2 is not capable of binding T 3 (or any other known ligand). TRv␣2 is not a functional TR but, rather, in transfection systems is an inhibitor of thyroid hormone action (4,5). TRv␣2 is widely expressed, and in some tissues such as brain its expression greatly exceeds that of the functional TRs, at least at the RNA level (6). Although the physiological role of TRv␣2 is not known, it is plausible that it serves to dampen T 3 regulation of gene expression, perhaps in a tissue-or gene-specific manner.
TRs are unusual in that they regulate gene expression from a wide variety of T 3 response elements (TREs). TREs usually contain two receptor binding sites (half-sites), which are related to the traditional 6-base pair sequence AGGTCA (7) or the higher affinity octamer TAAGGTCA (8) and which can be arranged either as a direct repeat (DR), palindrome (Pal), or inverted palindrome (IP) (9 -12). Whether there is physiological significance to these different half-site orientations is not known.
In general the functional TRs bind to TREs with highest affinity as heterodimers with retinoid X receptors (RXRs), and it is thought that the TR⅐RXR heterodimer is the biological mediator of T 3 action (13,14). RXRs are members of the steroid receptor superfamily, and they also serve as heterodimerization partners for other nuclear receptors, such as those for vitamin D and retinoic acid. It appears that multiple domains within the TR ligand binding domain may be involved in heterodimerization with RXRs (15)(16)(17)(18). One such domain, located at amino acids 368 -374 of TR␣1 (LMKVTDL), is known as the ninth heptad. Mutations in this domain can severely impair or abolish TR␣1⅐RXR heterodimerization (16,17). It is interesting that the alternative splice site for production of TRv␣2 is located in the middle of the ninth heptad, between amino acids 370 and 371 (Fig. 1). Since TRv␣2 is missing the second half of the ninth heptad, one would predict that it would be unable to heterodimerize with RXR. Indeed, this has been shown to be true on a palindromic or inverted palindromic TRE (19,20). Surprisingly, however, it has been reported that TRv␣2⅐RXR heterodimers do form on a DR TRE (20). This raises questions about the role of the ninth heptad in TR⅐RXR heterodimerization, as well as questions about the mechanism of the dominant negative activity of TRv␣2. For example, is DNA binding required for the dominant negative activity of TRv␣2? Some evidence in the literature supports this (20,21), but other data do not (22). If DNA binding is required, will the TRv␣2 domi-nant negative activity be confined to those response elements that support TRv␣2 heterodimerization with RXR? The goal of these studies was to further evaluate the role of the ninth heptad in TR⅐RXR heterodimerization, as well as the role of DNA binding in the dominant negative activity of TRv␣2. Furthermore, it is clear that the dominant negative activity of TRv␣2 is weak relative to that of mutant TRs that are naturally found in patients with resistance to thyroid hormone (RTH) (23,24); whether this weakness is due to lack of the full ninth heptad or poor DNA binding has been examined.

EXPERIMENTAL PROCEDURES
Thyroid Hormone and Retinoid X Receptors-Mouse TR␣1 (25), rat TRv␣2 (6), and mouse RXR␣ (14) cDNAs were transcribed from pBluescript plasmids and then translated in rabbit reticulocyte lysate (Promega) in the presence of [ 3 H]leucine or [ 35 S]methionine as described previously (25). Trichloroacetic acid precipitable-protein counts per minute were determined, and the products were analyzed by SDS-polyacrylamide gel electrophoresis. Receptors also were produced in Escherichia coli using the vector pMalc2 (New England Biolabs) as described previously (26,27). This vector produces a fusion protein of maltosebinding protein (MBP), followed by a cleavage site for factor Xa and the receptor of interest. The fusion proteins were purified by amylose affinity chromatography and cleaved with factor Xa.
Electrophoretic Mobility Shift Assay (EMSA)-The assay was performed as described previously (8,28). In brief, 4,000 cpm of [ 3 H]leucine-labeled receptors (or, where indicated, 5-10 ng of E. coliderived receptors) were incubated with 20,000 cpm of Klenow filled-in 32 P-labeled DNA for 40 min at room temperature. In some experiments reticulocyte lysate-translated receptors were incubated with calf intestinal alkaline phosphatase (or alkaline phosphatase that had been inactivated by heating at 100°C for 10 min) as described (29) prior to the EMSA incubation. The incubation mixtures were electrophoresed through 6% nondenaturing polyacrylamide gels and analyzed by autoradiography. The DNA sequences used are listed in Table I.
Site-directed Mutagenesis-Mutations were placed in TRv␣2 or TR␣1 using the polymerase chain reaction-based splice overlap extension method (30) or the Promega Altered Sites kit. Mutant products were sequenced to confirm the mutations and to exclude errors.
TR⅐RXR Heterodimerization in the Absence of DNA-The MBP-RXR␣ fusion protein was produced in E. coli and adsorbed onto a series of 1-ml amylose resin columns. MBP was similarly adsorbed as a control. The columns were washed with 20 ml of EMSA incubation buffer, and then 1,200,000 cpm of [ 35 S]methionine-labeled TRs were loaded onto the columns. After a 30-min incubation at 4°C, the columns were washed with another 20 ml of EMSA incubation buffer. Radiolabeled TRs that remained bound were eluted with 10 mM maltose and quantified by liquid scintillation counting. SDS-polyacrylamide gel electrophoresis of aliquots confirmed that the eluted TRs were of the appropriate sizes.
Transient Transfections-JEG-3 human choriocarcinoma cells were maintained and transfected as described previously (31). Receptors were expressed from CDM (32) or Rous sarcoma virus (33) plasmid. Transfections included (as indicated) 10 -100 ng of TR␣1 vector, 1-3 g of TRv␣2 vector, and 3 g of RXR vector. Empty vector was added to maintain 6 g of expression plasmid/60-mm Petri dish transfection. Transfections also included 4 g of a T 3 -responsive chloramphenicol acetyltransferase (CAT) reporter vector derived from pUTKAT3 (34). The CAT vectors had single copies of the TREs listed in Table I placed 5Ј to the basal thymidine kinase promoter. Transfection efficiency was determined using 1 g of pTKGH/Petri dish, in which the basal thymidine kinase promoter (without TRE) drives expression of human growth hormone (GH). CAT and GH were assayed as described previously (31). T 3

inductions were calculated as (CAT/GH for cells transfected with T 3 )/(CAT/GH for cells transfected without T 3 ).
Immunoprecipitation-JEG-3 cells were transfected in 100-mm Petri dishes with plasmid amounts scaled accordingly. Two days after transfection the cells were placed in methionine-free media with 10% dialyzed fetal bovine serum and [ 35 S]methionine, 0.15 mCi/ml (4 ml/Petri dish). The cells were cultured for 4 h. Cell harvesting and immunoprecipitation, using a rabbit antiserum directed against the unique carboxyl terminus of TRv␣2, were performed as described (29), except that incubation with the antibody was for 2 h, and the immunoprecipitates were isolated using protein A-agarose. Immunoprecipitates were analyzed by SDS-polyacrylamide gel electrophoresis and quantified on a Molecular Dynamics PhosphorImager.

Heterodimerization of TRv␣2 and RXR␣ on Direct Repeat
TREs-Since TRv␣2 lacks one-half of the ninth heptad, it was not expected to be able to heterodimerize with RXR on DNA. Although this expectation was confirmed for a palindromic or inverted palindromic TRE (19,20), it was unexpectedly shown that such heterodimers do form on a DR TRE (20). We tested TRv␣2⅐RXR heterodimer formation on four DR TREs (Table I), which differ only in whether the half-sites contain the traditional hexamer AGGTCA (7) or the extended octamer TR binding site TAAGGTCA (8). As shown in Fig. 2, TRv␣2⅐RXR heterodimers form on the TRE in which both half-sites are octamers (8DR; lane 6), but DNA binding is not detectable when both half-sites are hexamers (6DR; lane 24). Heterodimer formation is barely detectable when just the 5Ј-half-site is an octamer (86DR; Fig. 2, lane 12) and is somewhat more prominent when just the 3Ј-half-site is an octamer (68DR; Fig. 2, lane 18). Similar results were obtained using proteins produced in E. coli, except that binding of TRv␣2⅐RXR heterodimers to 8DR was as strong as the binding of TR␣1⅐RXR heterodimers, and TRv␣2 monomer binding to DNA was detectable (data not shown). The data indicate that TRv␣2⅐RXR heterodimers can form on DR TREs, but this requires at least one half-site to be an octamer (preferably the 3Ј-half-site), and strong binding requires both sites to be octamers.
Effect of the Ninth Heptad on TRv␣2⅐RXR Binding to DNA-The fact that TRv␣2⅐RXR heterodimers can form on some DR TREs but not on Pal or IP TREs suggests that the importance FIG. 1. Schematic diagram of thyroid hormone receptor proteins. The TR␤ gene encodes two proteins, TR␤1 and TR␤2, which differ in their amino-terminal regions but which have identical DNA and ligand binding domains. Two protein products of the TR␣ gene, TR␣1 and TRv␣2, also are shown. These two proteins are identical for their first 370 amino acids but then diverge completely. TRv␣2 does not bind T 3 . The DNA and ligand binding domains of the TR␤s and TR␣1 are ϳ85% identical, but their amino-terminal domains are unrelated. Numbers above each protein, amino acid positions. Not shown, TR␣ encodes a second splice variant (TRv␣3) that is similar to TRv␣2, except it uses a splice acceptor site 117 nucleotides (39 amino acids) downstream of the TRv␣2 site.

TABLE I DNA sequences of T 3 response elements used in EMSA and transfection experiments
The sequences shown are the top strands and do not include GATC overhangs at both ends. The sequences are named to denote whether the half-sites are octamers (TAAGGTCA) or hexamers (AGGTCA) and also to denote the relative orientation of the half-sites. Thus, 8DR contains two octamer half-sites as a direct repeat, 86DR contains a 5Ј-octamer half-site and a 3Ј-hexamer half-site as a direct repeat, etc. The AGGTCA hexamers are underlined.
of the ninth heptad for heterodimerization is not the same for all TREs. To address this further, several mutant TR proteins were produced, as shown schematically in Fig. 3. The full ninth heptad was restored in TRv␣2 by inserting TR␣1 amino acids 371-378 immediately following TRv␣2 amino acid 370. This construct is denoted ␣2ϩ9H. To control for the spacing effect of this insertion, the construct ␣2ϩAla was made in which a series of alanines was placed immediately after TRv␣2 amino acid 370. Finally, a substitution mutation was made in TR␣1, in which the ninth heptad amino acids 371-376 were replaced by TRv␣2 amino acids 371-376 (␣1Ϫ9H). These mutant proteins were produced in reticulocyte lysate and were then used in a series of EMSAs with TREs in which both half-sites are octamers but the orientations of the halfsites differ (8DR, 8Pal, and 8IP). The results are shown in Fig.  4. As expected, TR␣1 forms strong monomer and heterodimer bands on all three TREs (Fig. 4, lanes 5 and 6). TRv␣2 does not form monomer bands and forms a heterodimer only on 8DR (Fig. 4, lanes 7 and 8), suggesting that the full ninth heptad is not critical for heterodimerization on 8DR but is critical on 8Pal and 8IP. This is supported by ␣2ϩ9H, which forms a strong heterodimer band on all three TREs (Fig. 4, lanes 9 and 10). The need for the specific ninth heptad sequences for heterodimer formation on 8Pal is confirmed by the minimal heterodimerization of ␣2ϩAla (Fig. 4B, lanes 13 and 14). Most surprisingly, however, ␣2ϩAla heterodimerizes strongly on 8IP (Fig. 4C, lanes 13 and 14). This suggests that on 8IP the ninth heptad serves a critical role, but the effect is amino acid sequence independent; therefore, this region is presumably serving as a nonspecific spacer to separate inhibitory sequences in the unique TRv␣2 carboxyl terminus from other critical sequences in the protein. The varied importance of the ninth heptad in TRv␣2 heterodimerization was consistent with the findings using the TR␣1 mutant ␣1Ϫ9H, which heterodimerizes well on 8DR and 8IP but not on 8Pal (Fig. 4C, lanes 11 and   12). Results essentially identical to those of Fig. 4 were obtained using E. coli-derived proteins, except that TRv␣2 monomer binding to DNA also was detected (data not shown, but see Fig. 6 for E. coli-derived TRv␣2 monomer binding to 8IP).
It also is of interest that ␣1Ϫ9H monomers bind these TREs poorly relative to wild type TR␣1 (Fig. 4, lanes 11 versus 5). Furthermore, ␣2ϩ9H monomer binding to DNA is detectable (Fig. 4B, lane 9), whereas wild type TRv␣2 monomer binding is not (Fig. 4B, lane 7). The enhanced ␣2ϩ9H monomer binding to DNA is more dramatic when using E. coli-derived protein (data not shown, but see Fig. 6 for E. coli-derived ␣2ϩ9H monomer binding to 8IP). Since the ninth heptad is not thought to contact DNA directly, these results suggest an indirect effect of this region to enhance TR monomer-DNA interactions.
Because the ability of TRv␣2 and its mutant versions to heterodimerize with RXR differed on the different TREs, we wished to determine whether such heterodimers can form in the absence of DNA. To accomplish this, a MBP-RXR fusion protein was produced in E. coli and adsorbed to an amylose affinity column (MBP was adsorbed as a control). After extensive washing, reticulocyte lysate-translated, radiolabeled TR␣1, TRv␣2, ␣2ϩ9H, or ␣2ϩAla was applied to the columns. After further extensive washing, the bound radiolabeled proteins were eluted, and the radioactivity was determined. The results are presented in Fig. 5. TR␣1⅐RXR heterodimers formed easily under these conditions. TRv␣2⅐RXR heterodimers were detectable, but only at ϳ6% of the level seen for TR␣1⅐RXR. Heterodimerization was largely restored for ␣2ϩ9H but not for ␣2ϩAla. The importance of the ninth heptad was confirmed by ␣1Ϫ9H, which heterodimerized poorly relative to wild type TR␣1. Thus, heterodimerization in the absence of DNA most closely resembles that seen on the 8Pal TRE.
The Ability (or Lack of Ability) of TRv␣2 Proteins to Heterodimerize with RXR on 8IP Is Not Related to Their Phosphorylation State-It has been shown that TRv␣2 is phosphorylated in its unique carboxyl terminus when translated in reticulocyte lysate, and that this phosphorylation underlies the deficient TRv␣2 monomer-TRE binding (29). A series of EMSAs was performed to determine whether phosphorylation influences the ability of the TRv␣2 proteins to heterodimerize. Specifically, we wished to know whether the unexpected ability of reticulocyte lysate-translated ␣2ϩAla to heterodimerize with RXR on 8IP was due to moving the phosphorylated residues away from other critical regions of the TRv␣2 protein. The strategy, modeled after that of Katz et al. (29), was to produce TRv␣2 proteins in E. coli, in which case they are not phosphorylated, or to treat the reticulocyte lysate-translated proteins with alkaline phosphatase. If ␣2ϩAla promotes heterodimerization on 8IP by sterically separating the phosphorylated res- The sequences of TRv␣2 and TR␣1 amino acids 365-380 are shown, with the full ninth heptad underlined in TR␣1 as well as the partial ninth heptad in TRv␣2. An insertion was made in TRv␣2 after amino acid 370 to restore the full ninth heptad (underlined), denoted ␣2ϩ9H. To control for the spacing effect of this mutation, the insertion mutant ␣2ϩAla was made. A substitution mutation was made in TR␣1 (denoted ␣1Ϫ9H), replacing amino acids 371-376 by the equivalent residues in TRv␣2, thus creating a partial ninth heptad identical to that in TRv␣2.

FIG. 2. TR⅐RXR heterodimerization on direct repeat TREs.
The binding of TR⅐RXR heterodimers to DNA was assessed by EMSA on a series of DR TREs, which differ only in whether the half-sites are the traditional hexamer AGGTCA or the octamer TAAGGTCA. As indicated in Table I, 8DR contains two octamers, 86DR contains an octamer in the 5Ј-position and a hexamer in the 3Ј-position, etc. The TRs and RXR were translated in reticulocyte lysate. The lanes that lack TR contain mock-translated reticulocyte lysate. D, position of the TR⅐RXR heterodimer bands; M, TR monomer bands.
The Role of TRv␣2⅐RXR Heterodimerization and DNA Binding in the Dominant Negative Activity of TRv␣2-To assess the role of TRv␣2⅐RXR heterodimerization in the dominant negative activity of TRv␣2, we first compared this activity on 8DR, 86DR, 68DR, and 6DR by transient transfection (Fig. 7). The percentages of repression of T 3 induction by TRv␣2 on these response elements were: 8DR, 38 Ϯ 3.9%; 86DR, 30 Ϯ 7.6%; 68DR, 21 Ϯ 3.0%; and 6DR, 29 Ϯ 5.2%. These data roughly parallel the relative binding of TRv␣2⅐RXR to these TREs (Fig.  2) in the sense that binding to 8DR was the strongest. However, the difference in dominant negative activity between 8DR and the other elements was not statistically significant (Dunnett's test). Thus, the lack of a strict parallel between the EMSA and transfection data might suggest that heterodimerization and DNA binding is not critical for the dominant negative activity or that the EMSAs do not reflect the DNA binding within the cell.
To help address this question, a TRv␣2 protein was produced with a cysteine to alanine mutation within the first zinc finger (C56A). This mutation prevents formation of the zinc finger and abolishes DNA binding of TRv␣2⅐RXR heterodimers on 8DR (Fig. 8). In a transfection system, TRv␣2 C56A no longer possesses dominant negative activity on 8DR, 8Pal, or 8IP (Fig.  9). To confirm that this mutant protein is expressed, transfected cells were metabolically labeled with [ 35 S]methionine, and immunoprecipitation was performed using a TRv␣2-specific antibody. The results, quantified by PhosphorImager analysis, indicate that C56A is expressed at 65% of the level of TRv␣2 (Fig. 10). It is unlikely that this modest difference in expression could account for the complete loss of dominant MBP-RXR fusion protein was produced in E. coli and adsorbed onto amylose columns; MBP was adsorbed to amylose as a control for nonspecific binding. 35 S-labeled TRs were produced in reticulocyte lysate and loaded onto the amylose columns. After washing, the bound TRs were eluted and quantified. Nonspecific binding to the MBP column was subtracted to yield the specifically bound radioactivity; the former averaged 11 Ϯ 3% of the latter for all radiolabeled proteins. Bars, mean specific binding of two experiments, with dots representing each experiment. Results are normalized to the binding of TR␣1, which was 51% of input cpm. negative activity. This suggests that the dominant negative activity of TRv␣2 requires DNA binding.
To further evaluate this issue, the dominant negative activity of ␣2ϩ9H also was analyzed by transfection (Fig. 11). The data indicate that ␣2ϩ9H has powerful dominant negative activity on 8DR, 8Pal, and 8IP, much stronger than wild type TRv␣2 but similar to that reported for mutant TRs from patients with resistance to thyroid hormone (23,24). For 8Pal and 8IP, ␣2ϩ9H also restores heterodimerization and DNA binding (as analyzed by EMSA); hence, this restoration could explain the potent dominant negative activity. However, by EMSA wild type TRv␣2 and ␣2ϩ9H heterodimerize with RXR nearly equivalently on 8DR (Fig. 4A), even though only ␣2ϩ9H has potent dominant negative activity. This discrepancy could be explained either by the EMSA not reflecting DNA binding within the cell, by DNA binding not being a critical regulator of the dominant negative activity, or perhaps by there being critical regulators in addition to DNA binding.
To address this, the C56A mutation was placed in the context of ␣2ϩ9H and analyzed by transfection. Immunoprecipitation from metabolically labeled cells demonstrated that ␣2ϩ9H and its C56A mutant were expressed at equivalent levels (Fig. 10). The transfection results (Fig. 11) indicate that ␣2ϩ9H C56A loses ϳ50% of its dominant negative activity on 8DR and 8IP, but it remains a powerful inhibitor on 8Pal. This suggests two independent mechanisms for the dominant negative activity, one requiring DNA binding and one not. The simplest explanation for the dominant negative activity that requires DNA binding is competition between TR␣1⅐RXR and ␣2ϩ9H⅐RXR heterodimers for DNA. A straightforward explanation for the dominant negative activity that does not require DNA binding would be sequestration of RXR in solution. If this were the case, one would predict that cotransfection of RXR would partially alleviate the dominant negative activity of ␣2ϩ9H. This was tested and, surprisingly, found not to be the case (Fig. 11). Thus, although the non-DNA binding-dependent portion of the dominant negative activity is likely due to sequestration of some critical factor in solution, the factor probably is not RXR.

DISCUSSION
This study sought to evaluate several questions regarding thyroid hormone receptor heterodimerization with RXR and the dominant negative activity of TRv␣2. It is clear that TR⅐RXR heterodimers bind to many TREs with much higher affinity than TR monomers or homodimers (13,14), and this has led to the belief that the heterodimer is the biological mediator of T 3 action. A substantial body of mutagenesis data supports the notion that the TR ninth heptad is a critical contact site in the heterodimer with RXR (16,17), and this is further supported by the observation that the ninth heptad sequence is well conserved among nuclear receptors that interact with RXR (35). However, direct evidence of contact between the ninth heptad and RXR is lacking. Although this region was originally named as one of nine possible leucine zipper motifs that might form an interface for interaction with RXR (35), these heptad repeats are unlikely to form a traditional leucine Reticulocyte lysate-translated TRv␣2 or its C56A mutant was incubated with or without RXR in the presence of radiolabeled 8DR, and protein-DNA complexes were resolved by EMSA. D, TRv␣2⅐RXR heterodimer complex. Lanes 1-3 were run on a separate gel from lanes 4 -6, but all were from a single experiment. The data indicate that, as expected, disruption of the first zinc finger abolishes TRv␣2⅐RXR heterodimer binding to 8DR.
FIG. 6. Effect of TRv␣2 phosphorylation on binding to 8IP DNA. TRv␣2, ␣2ϩ9H, and ␣2ϩAla were produced in E. coli or reticulocyte lysate (R.L.). The reticulocyte lysate-translated material was, where indicated, pretreated with active alkaline phosphatase (CIP: ϩ) or boiled alkaline phosphatase (CIP: ⌬) as a control. These proteins were incubated without or with E. coli-derived RXR as well as radiolabeled 8IP DNA, and protein-DNA complexes were resolved by EMSA. HTD, position of the TR⅐RXR heterodimers; HMD, TR homodimers; M, TR monomers; ‫,ء‬ a band due to binding of a truncated TR that is common to all TR␣1 and TRv␣2 preparations in E. coli.

FIG. 7. Dominant negative activity of TRv␣2 on direct repeat
TREs. JEG-3 cells were transfected with a TR␣1 expression vector, a T 3 -responsive CAT reporter (8DR, 86DR, 68DR, or 6DR), pTKGH to control for transfection efficiency, and either TRv␣2 or empty vector. Cells were cultured for 2 days after transfection with or without T 3 . Fold T 3 induction is calculated as (CAT/GH for cells cultured with zipper structure (36), and a full understanding of the function of the ninth heptad remains to be developed. Crystallographic analysis of the TR␣1 monomer ligand binding domain (36) indicates that the ninth heptad is part of ␣ helix 11. This helix contains a number of hydrophobic residues that could form a surface suitable for protein-protein interactions. Helix 11 also contains two residues that directly contact the ligand.
Since TRv␣2 is missing a portion of the ninth heptad, it was expected to be unable to heterodimerize with RXR. This was confirmed using a palindromic or inverted palindromic TRE (19,20), but heterodimerization was unexpectedly observed on a direct repeat TRE (20). This suggests that the role of the ninth heptad in heterodimerization is dependent on the nature of the TRE. We have confirmed that TRv␣2⅐RXR heterodimers can form on a DR TRE, and we show that this requires the half-sites to comprise the high affinity octamer TAAGGTCA rather than the more traditional hexamer AGGTCA. The halfsites used in the original report (20) were TCAGGTCA, sug-gesting that the 5Ј-Thd is more critical than the following Ado in the octamer, and this has been confirmed. 2 Still, even with octamer half sites, TRv␣2⅐RXR heterodimers are not detectable on Pal or IP TREs. For DR TREs, it appears that having an octamer sequence as the 3Ј-half-site is more important than as the 5Ј-half-site. This fits with the polarity preference of TRs occupying the 3Ј-half-site when they heterodimerize with RXR on DR TREs (37,38).
To test whether restoration of the full ninth heptad would permit TRv␣2 to heterodimerize on 8Pal and 8IP, we made the construct ␣2ϩ9H. As anticipated, ␣2ϩ9H heterodimerizes with RXR on both of these TREs, suggesting that the full ninth heptad is critical for this process. However, restoring the full ninth heptad into TRv␣2 also adds separation between the first 370 amino acids of this protein and its unique 122-amino acid carboxyl terminus. To control for this, we made the construct ␣2ϩAla. As expected, ␣2ϩAla had only minimal ability to heterodimerize with RXR on 8Pal. Surprisingly, however, ␣2ϩAla and RXR heterodimerize strongly on 8IP. This suggests that the role of the ninth heptad in TR⅐RXR heterodimerization is different on each of the orientations of TREs: (a) on 8DR the full ninth heptad is not critical for heterodimerization; (b) on 8Pal the full ninth heptad sequence is critical; this situation most closely mimics heterodimerization in the absence of DNA; and (c) on 8IP the full ninth heptad also is critical, but only as a nonspecific spacer, since a series of alanines functions equally well. Presumably heterodimerization of wild type TRv␣2 on 8IP is sterically impaired by the unique TRv␣2 carboxyl terminus, and the extra alanines or ninth heptad serves to move the inhibitory domain out of the way. Although the exact location of this inhibitory domain is unknown, it is not simply the TRv␣2 2 M. Lazar, personal communication.  11. Dominant negative activity of TRv␣2, ␣2؉9H, and C56A mutant of ␣2؉9H. JEG-3 cells were transfected with a TR␣1 expression vector, a T 3 -responsive CAT reporter (8DR, 8Pal, or 8IP), pTKGH to control for transfection efficiency, and TRv␣2, ␣2ϩ9H, ␣2ϩ9H C56A, or empty vector. Some transfections with ␣2ϩ9H also included an RXR expression vector. Results are the mean Ϯ S.E. (bars) for four to six independent transfections per condition, except for ␣2ϩ9H with RXR on 8Pal and 8IP; n ϭ 2. The data demonstrate that ␣2ϩ9H is a substantially more potent inhibitor of T 3 induction than is TRv␣2 on all TREs tested. ␣2ϩ9H C56A loses ϳ50% of its dominant negative activity on 8DR and 8IP but retains nearly fully dominant negative activity on 8Pal. Cotransfection of RXR is not able to overcome the dominant negative activity of ␣2ϩ9H. phosphorylation sites, since the phosphorylation state of TRv␣2 and its mutants did not regulate heterodimerization on 8IP. However, TRv␣2 phosphorylation may have some influence on TRv␣2⅐RXR binding to 8DR, since (phosphorylated) reticulocyte lysate-translated TRv␣2 heterodimerized somewhat less well on this element than did (nonphosphorylated) E. coli-derived TRv␣2. This probably would represent an extension of the process whereby phosphorylation impairs TRv␣2 monomer binding to DNA (29).
The above conclusions suggest that the steric interactions between TR and RXR are not identical on DR, Pal, and IP TREs. Previous models of heterodimerization (1) postulated that a flexible hinge separates the DNA and ligand binding (heterodimerization) domains, such that the TR⅐RXR ligand binding domain contacts can be the same, even though the relative orientations of the DNA binding domains differ. Our data, however, are not consistent with this model. Even if the putative hinge does exist, it apparently is not fully flexible, since it is clear that the stereochemistry of heterodimerization differs on DR, Pal, and IP TREs. This implies that details of the TRE sequence, specifically the half-site orientations, actively influence the TR-RXR interaction. Thus, the TRE is not simply a passive recipient of a preformed protein heterodimer, but rather, it actively influences the heterodimer structure.
This suggests that certain TR mutations could impair heterodimerization selectively depending on the orientation of the TRE half-sites. This is of possible relevance to the syndrome of RTH, which is caused by mutations in the TR␤ ligand binding domain (see Ref. 39 for a recent review). The syndrome is generally inherited in an autosomal dominant manner, and the mutant TR is not neutral but, rather, has dominant negative activity on the wild type TR. It has been shown that mutant TR␤s in this syndrome must retain DNA binding and RXR binding activity for them to possess dominant negative activity (17,40). The clinical phenotype in this syndrome is variable and cannot be predicted simply by the decrease in T 3 binding affinity. If a mutation impaired heterodimerization on a subset of TREs, this could result in greater dominant negative activity on some TREs than on others and could help explain variations in phenotype.
In addition to its role in heterodimerization, the ninth heptad also appears to play a role in TR monomer binding to DNA. For example, when the ninth heptad of TR␣1 is mutated to become the partial ninth heptad of TRv␣2 (␣1Ϫ9H), there is a decrease in monomer binding to 8DR, 8Pal, and 8IP, even though heterodimer binding is significantly impaired only on 8Pal. Similarly, restoration of the full ninth heptad into TRv␣2 tends to enhance monomer binding to DNA, especially with E. coli-derived protein. In this regard, it is interesting to note that, due to their nonphosphorylated state, E. coli-derived TRv␣2 monomers bind DNA much better than do (phosphorylated) reticulocyte lysate-derived TRv␣2 monomers (29). This suggests a functional interaction between the ninth heptad and the phosphorylation state of TRv␣2. In any case, since the ninth heptad is not thought to contact DNA directly, it is presumed that this effect of the ninth heptad on TR monomer DNA binding is indirect.
The mechanism of the dominant negative activity of TRv␣2 has been contentious. A simple model, direct competition for DNA binding, is supported by some studies (20,21). However, other data suggest that the effect is independent of DNA binding (22). Related to this is the question of why TRv␣2 has such weak dominant negative activity relative to mutant TRs from patients with RTH. Using essentially all TREs except 8DR, TRv␣2 shows a very minimal ability to heterodimerize with RXR and, hence, a similarly minimal ability to bind to the DNA. This would help explain the weak dominant negative activity of TRv␣2 if the mechanism is direct competition for DNA binding (presumably heterodimerization and DNA binding do occur on TREs such as 8Pal, 8IP, and 6DR, but too weakly to detect by traditional EMSA).
The DNA binding-deficient mutant of TRv␣2 (C56A) sheds light on this issue, since this protein is devoid of dominant negative activity. This supports the simple model of direct competition for DNA binding. However, the dominant negative activity on 8DR is only marginally greater than that on other TREs, even though by EMSA the DNA binding of TRv␣2⅐RXR heterodimers appears to be substantially greater on 8DR. This quantitative discrepancy could be explained in a number of ways. It is plausible that DNA binding is necessary but not sufficient for the dominant negative activity of TRv␣2. Recently nuclear receptor corepressors have been identified, which can bind to unliganded TRs and cause them to repress basal transcription (41,42). A similar molecule might need to interact with DNA-bound TRv␣2 to permit dominant negative activity, and if this were the case, there may only be a weak correlation between the magnitude of TRv␣2⅐RXR binding to TREs by EMSA and the dominant negative activity in a transfection assay.
An alternative explanation for the above discrepancy between DNA binding and dominant negative activity on 8DR might be that the EMSA conditions do not accurately mimic DNA binding within the cell. Thus perhaps in vivo, binding of TRv␣2⅐RXR heterodimers to 8DR may be much weaker than binding of TR␣1⅐RXR, even though our standard EMSA conditions show similar degrees of DNA binding. To assess this we examined several properties of TRv␣2⅐RXR and TR␣1⅐RXR binding to 8DR by EMSA. First, we performed the EMSA incubations with increasing concentrations of KCl (50 -500 mM). Increasing salt had a parallel inhibitory effect on TRv␣2⅐RXR and TR␣1⅐RXR binding to 8DR and thus did not yield conditions that could substantially weaken TRv␣2 binding relative to TR␣1 (data not shown). Second, we examined the stability of TRv␣2⅐RXR and TR␣1⅐RXR complexes on 8DR by adding a 1000-fold excess of nonradiolabeled 8DR after the EMSA binding reactions reached equilibrium and then loading aliquots onto the EMSA gel at various time points. The TR␣1 and TRv␣2 heterodimer complexes were remarkably stable, with both showing only minimal loss of signal after 80 min of incubation with the competitor DNA (data not shown). In short, we were unable to find EMSA conditions that could demonstrate TRv␣2⅐RXR heterodimers to be substantially less stable or less likely to form on 8DR than TR␣1⅐RXR heterodimers.
The above discussion presumes that the lack of dominant negative activity of TRv␣2 C56A indicates that DNA binding is essential for dominant negative activity. However, we cannot totally exclude the possibility that the TRv␣2 C56A mutation disrupts some unknown protein-protein interaction, which accounts for the loss of dominant negative activity, rather than the disruption of TR-DNA binding.
The reason why the dominant negative activity of TRv␣2 is substantially weaker than that of RTH mutant TRs is complex. Restoring the full ninth heptad into TRv␣2 (␣2ϩ9H) restores highly potent dominant negative activity, indicating that the loss of the full ninth heptad accounts for the weak TRv␣2 dominant negative activity (all known RTH mutations occur outside the ninth heptad). It would be expected that the role of the ninth heptad in this situation is either to enhance DNA binding (by enhancing heterodimerization with RXR) and/or to sequester RXR in solution. The DNA binding mutant ␣2ϩ9H C56A retains ϳ50% of its dominant negative activity on 8DR and 8IP, indicating that enhanced DNA binding is only onehalf of the explanation on these TREs. We expected sequestration of RXR in solution to be the other half, in which case cotransfection of RXR should at least partially overcome the dominant negative activity of ␣2ϩ9H. Unexpectedly, this did not occur, suggesting that the full ninth heptad plays a role in sequestering a currently unknown non-RXR factor, which is important for full T 3 induction of gene expression on 8DR and 8IP. The results with 8Pal are somewhat different in that ␣2ϩ9H C56A retains near complete dominant negative activity. This indicates that enhanced DNA binding plays little role in the potency of ␣2ϩ9H as a dominant negative inhibitor on 8Pal. As with 8DR and 8IP, cotransfection of RXR did not relieve the dominant negative activity of ␣2ϩ9H on 8Pal, again suggesting that the mechanism involves sequestration of some unknown non-RXR factor. Clarification of this mechanism and identification of the postulated factor would be of importance in understanding T 3 regulation of gene expression, the dominant negative activity of TRv␣2, and the syndrome of RTH.