Thyroid Hormone Receptor Dimerization Is Required for Dominant Negative Inhibition by Mutations That Cause Thyroid Hormone Resistance*

The syndrome of thyroid hormone resistance (THR) is caused by multiple distinct mutations of the ligand- binding domain of the thyroid hormone p receptor. Although the mutant receptors are transcriptionally inactive, they inhibit normal receptor function in a dominant negative manner to cause hormone resistance. Because most of the naturally occurring muta- tions are clustered within two areas that lie on either side of a putative dimerization region, we hypothesized that receptor dimerization was important for dominant negative inhibition. In gel mobility shift assays, two THR mutants (G345R and P453H) formed homodimers as well as heterodimers with the the retinoic acid X receptor a. In contrast, an artificial mutation (L428R) in one of the hydrophobic heptad repeats of the putative receptor dimerization domain impaired heterodimerization with retoinoic acid X receptor a without altering the formation of homodimers. Double mutants containing either of the THR mutations along with the dimer- ization mutation formed homodimers but not heterodimers, reflecting the properties of the dimerization mutant alone. In transient expression assays using positively (TRETKLuc) or negatively (TSHaLuc) regu- lated reporter genes, the dominant negative activity of the THR mutants was eliminated by the addition of the dimerization mutation. These results support a mech- anism for dominant negative activity by THR mutants in which functionally inactive heterodimers bind to DNA to inhibit access by normal receptors.

produce a higher steady state level of thyroid hormones. Thyroid hormone resistance occurs to a variable degree in other target organs, but is more difficult to quantify than the alterations in the thyroid axis (2). Additional features of the syndrome can include goiter, mental retardation, attention deficit disorders, language abnormalities, tachycardia, and delayed skeletal maturation (3). In almost all cases, the syndrome is inherited in an autosomal dominant manner (3,4).
Thyroid hormone action is mediated by binding to nuclear receptors that activate or repress the transcription of specific target genes. There are two thyroid hormone receptor genes, a and P, as well as splicing variants of each, resulting in a variety of receptor isoforms (5). There is now extensive evidence that THR is caused by mutations in the @ form of the thyroid hormone receptor (TRP) (3,4). More than 20 distinct mutations have now been described in the ligand-binding domain of TRP. In most, but not all cases, the mutations greatly decrease thyroid hormone binding affinity (3,4,6). In all cases that have been studied, the mutant receptors lack transcriptional activation or repression in transient expression assays (7-10). Thus, the mutations appear to inactivate the receptor either by reducing thyroid hormone binding and/ or altering critical domains involved in transcriptional control.
Although it is plausible that a mutation in one of the P receptor alleles could result in the partial resistance that is seen in THR, evidence derived from an unusual kindred with an autosomal recessive form of THR suggests that inactivation of a single allele is not sufficient to cause the syndrome (11). In this family, there is a large deletion that includes both DNA-binding and ligand-binding domains of the P receptor (12). Although individuals homozygous for this P receptor deletion exhibited hormone resistance, the heterozygotes were normal. This observation is consistent with the concept that the more common autosomal dominant form of THR is caused by mutant receptor inhibition of normal receptor function in a dominant negative manner (7,8). Dominant negative activity of THR mutants has now been documented in transient expression assays, and it occurs with respect to genes that are stimulated or repressed by thyroid hormone (7,8,10).
Three potential mechanisms have been proposed to explain the dominant negative activity of receptor mutants: 1) competition of the mutant receptor and normal receptor for the interactions with target DNA sequences; 2) functionally inactive heterodimer formation between normal and mutant receptors; and 3) titration of limiting transcriptional cofactors (8). In previous studies, we provided evidence that competition at the target DNA-binding site is the major mechanism for the dominant negative activity of THR mutants (13). This conclusion was based upon the observation that addition of DNA-binding mutations into the THR mutants abolished their dominant negative activities. Recently, thyroid hormone receptors have been shown to bind to DNA as homodimers (14)(15)(16) or as heterodimers with accessory proteins such as members of retinoid X receptor (RXR) family (17-21). Dimerization appears to be important for stabilizing receptor-DNA complexes (15,22) and could therefore play an important role in receptor action. The dimerization domain of the thyroid hormone receptor has been proposed to consist of a series of repeated hydrophobic motifs in the carboxyl terminus that overlap with the ligand-binding domain (23). It is notable that the receptor mutations that occur in THR are clustered in two discrete domains (centered at amino acids 345 and 453) that are separated by about 100 amino acids of the putative dimerization region (24,25). This finding raises the possibility that dimerization may be required for the mutant receptors to exert dominant negative activity. In this report, we examined the role of thyroid hormone receptor homodimerization and heterodimerization in the dominant negative activity of THR mutants.

Plasmid Construction and Receptor Mutagenesis-The TRETK
luciferase gene (Luc) contains two copies of a palindromic thyroid hormone-responsive element (TRE) upstream of the thymidine kinase promoter (26)  T3 Binding Assays"T3 binding assays were performed as described previously (13) using a filter binding method. Receptor affinity constants (&) were calculated using Scatchard plot analyses. The results of at least three individual experiments were used to determine mean f S.D.
Cell Culture and Transient Expression Assays-JEG-3 cells were grown in Optimem (BRL-GIBCO) containing 2% fetal bovine serum, penicillin (100 units/ml), and streptomycin (100 pglrnl). The cells were trypsinized 18 h prior to transfection and plated into medium containing 2% fetal bovine serum depleted of Ts by treatment with activated charcoal. Triplicate plates of cells were transfected with 5 pg of reporter plasmid (TRETKLuc or TSHa Luc) together with 0.2-2 pg of wild type or mutant receptor expression vector with addition of RSVCAT plasmid (31) as necessary to maintain the total amount of DNA constant. Following a 16-h exposure to a calcium phosphate-DNA precipitate, fresh serum-stripped media was added with or without 5 nM T3. After an additional 48 h, the cells were harvested, and luciferase activity was measured using a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego, CA) (32).

Structure of Receptor Mutants and Analyses of T3 Binding-
A Gly to Arg substitution at amino acid 345 (G345R) (33) and a Pro to His substitution at amino acid 453 (P453H) (34) have been reported in separate kindreds with THR. An artificial mutation was created in the ninth heptad repeat of the putative dimerization domain (23). The hydrophobic residue, Leu4% was substituted with a positively charged amino acid, Arg (L428R). Scatchard analyses were used to determine the TS binding affinities of the in vitro translated products of each of the receptor mutants. As described previously, the P453H mutant exhibited about 10-fold reduced affinity (2.73 f 1.5 X lo9 M-') (13). The affinities of G345R and L428R mutants were below detection (normal, 3.09 & 2.9 X 10'' M-').
Analyses of Mutant Receptor Binding to DNA-Gel mobility shift assays were used to analyze mutant receptor binding to a radiolabeled palindromic (Pal) TRE (26)  and fl wild type thyroid hormone receptors using the palindromic and inverted palindromic TREs. In vitro translated a1 and ,!? wild type receptors were preincubated with or without hRXRcu. The receptor complexes bound to the 32P-labeled palindromic (Pal) or inverted palindromic (Lap) TREs were analyzed by non-denaturing gel electrophoresis. The receptor-DNA complexes are indicated as arrows (1, monomer; 2, homodimer; 3, heterodimer).
Pal, the faster migrating complex (band 1 ) was predominant, whereas bands 1 and 2 were of similar intensity when bound to TRE-Lap. Coincubation of T R a l with RXRa formed a much slower migrating band (Fig. 1, arrow 3). Using a different DNA sequence that only contains a single TRE half-site, only band 1 was seen with T R a l (data not shown). Based upon these results and similar published findings (14,21,36), the protein-DNA complexes are concluded to represent monomer (band 1 ), homodimer (band Z), and heterodimer (band 3). In uitro translated TRB predominantly formed homodimer (band 2), which was very faint with Pal, but prominent with Lap. The addition of RXRa formed a strong heterodimer complex with both TREs. The receptor-DNA complexes with TRB migrated more slowly than those with TRal, likely reflecting the larger size of TRP.
The binding properties of the mutant TRB were also studied using the palindromic TRE sequence (TRE-Pal) (Fig. 2). The in uitro translated products of / 3 wild type receptor formed a faint receptor-DNA complex with the TRE-Pal, consistent with homodimer (Fig. 2, arrowhead). The addition of TRBspecific antibody resulted in a supershift of TRB-containing complexes, forming two slowly migrating bands at the top of the gel. Addition of TRB together with in vitro translated RXRa resulted in prominent complex that migrated more slowly than the receptor homodimer, consistent with the formation of a TRB-RXRa heterodimer (Fig. 2, arrow). The binding of the G345R and P453H mutant receptors to the palindromic TRE was similar to the wild type receptor in the absence or presence of RXRa. In the absence of RXRa, binding by the L428R mutant was similar to wild type receptor and addition of &specific antibody caused a supershift. However, the L428R mutant formed very little heterodimeric complex with RXRa. The double mutants containing the THR mutations and the dimerization mutant (G345R/L428R, P453H/L428R) displayed binding activity that was similar to that of the L428R mutant alone. Addition of 20 nM TI did not affect the pattern or intensity of receptor complexes bound to TRE-Pal (data not shown).
Receptor binding was also studied using the inverted palindrome of the TRE (TRE-Lap), as this element facilitates receptor binding as a homodimer (Fig. 3). The wild type receptor readily formed a homodimer complex with TRE-Lap (Fig. 3, arrowhead). Addition of RXRa resulted in formation of a heterodimer complex (Fig. 3, arrow) with a mobility that was indistinguishable from that bound to the TRE-Pal (Fig.  3, right lanes). Addition of 20 nM T a decreased the amount of homodimer binding to the TRE-Lap, but it did not change the amount of heterodimer complex (36). The two THR mutants also formed homodimer and heterodimer complexes. However, T3 did not alter homodimer binding substantially, consistent with the reduced affinities of these mutants for thyroid hormone. The L428R mutant and the double mutants formed homodimer complexes that were similar to wild type receptor, but the amount of heterodimeric complex with RXRa was greatly reduced. T3 did not affect the protein-DNA complexes of these mutants. These results indicate that the THR mutants bind to TRE-Pal or TRE-Lap as homodimers and as heterodimers with RXRa. The L428R mutant and the double mutants (G345R/L428R, P453H/L428R) also bind to these DNA sequences as a homodimers, but fail to effectively form the heterodimeric complex. Thus, the L428R mutation appears to selectively alter receptor heterodimerization.

Functional Properties and Dominant Negative Actiuity of Mutant Receptors in Transient Expression
Assays-To assess the transcriptional activity of the wild type and mutant receptors, each receptor DNA was transfected into JEG-3 choriocarcinoma cells along with a positively (TRETKLuc) or negatively (TSHaLuc) regulated reporter gene. Receptor activity was analyzed by measuring luciferase activity after treatment with 5 nM T3 (Fig. 4). In the absence of transfected receptor, TRETKLuc activity was stimulated to 1.3-fold and TSHaLuc was repressed to 74% of basal activity by treatment with T3.
These effects of T3 likely reflect the presence of a small amount of endogenous thyroid hormone receptor in JEG-3 cells. Transfection of wild type TRB conferred a marked increase in T3-mediated activation and repression of the reporter genes. TRETKLuc activity was stimulated by 24.7-fold and TSHaLuc was repressed to 15%. As shown previously (8), the G345R mutant did not mediate activation of TRE-TKLuc or repression of TSHaLuc (0.4-fold and 76%, respectively). The P453H mutant exhibited partial T3 responsiveness (7.1-fold in TRETKLuc and repression of TSHaLuc to 40%). consistent with its partial binding to T3 (8). The mutation in the dimerization region (L428R) and the double mutations with THR mutants (G345R/L428R, P453H/ L428R) did not activate TRETKLuc or repress TSHaLuc.
Having established the functional capabilities of the individual receptor mutants, their dominant negative activities were determined in the presence of wild type receptor (Fig.  5). Although the THR mutants partially block the activity of wild type receptor at a 1:1 ratio (data not shown), a 1:lO ratio of wild type to mutant receptors was used to more clearly illustrate the dominant negative activity of the mutant receptors. T3-dependent activation of TRETKLuc by wild type receptor (25-fold) was inhibited by cotransfection with G345R and P453H mutants (1-and &fold, respectively). In contrast, cotransfection of the dimerization mutant (L428R) did not inhibit wild type receptor function (18-fold), indicating that it is not a dominant negative mutant receptor. Insertion of the dimerization mutation into the THR mutants (G345R/ L428R and P453H/L428R) prevented dominant negative inhibition of the THR mutants (22-and 20-fold, respectively). Similarly, repression of TSHaLuc by the wild type receptor (15%) was blocked by cotransfection of G345R (67%) and to a lesser degree by P453H (42%). The L428R mutant and the two double mutants (G345R/L428R and P453H/L428R) did not inhibit wild type receptor function (23, 18, and 11%. respectively). These results suggest that the dimerization region is critical for the dominant negative activity of the THR mutants for positive and negatively regulated reporter genes.

DISCUSSION
The regulation of thyroid hormone action through its nuclear receptors is much more complex than predicted from physiologic studies that occurred before the receptors were cloned. In addition to the presence of multiple thyroid hormone receptor genes and protein isoforms (5), it has been demonstrated recently that the thyroid receptors can form homodimers as well as heterodimers with a variety of cellular proteins referred to thyroid hormone receptor auxiliary proteins (TRAPS) (15,(17)(18)(19)(37)(38)(39). There is also a high degree of variability in the primary sequence and arrangement of thyroid hormone response elements (TREs) (40). We have been interested in studying the naturally occurring mutations in the TRP gene as a probe of thyroid hormone action. The most striking feature of these mutations is their ability to function in a dominant negative manner to block the activity of wild type thyroid hormone receptors (7,8). The dominant negative activity of the mutant @ receptors is manifest with respect to wild type TRa and TRB proteins. In addition, the mutant receptors block thyroid hormone regulation of positively and negatively regulated genes (8).
Because a large number of mutations that cause THR have now been described, it is possible to begin to derive structurefunction implications based upon the locations of the mutations (Fig. 6). In contrast to mutations in the androgen (41) and vitamin D (42) receptors, it is notable that none of the TRD mutations that cause THR occur in the DNA-binding domain of the receptor. This observation likely reflects a requirement for DNA binding for the dominant negative effect of the mutant thyroid hormone receptors. Previous experiments demonstrating that addition of a DNA-binding domain mutation to the THR mutants are consistent with this hypothesis (13). Although the naturally occurring THR mutants appear to be restricted to the ligand-binding domain of the receptor, there are several reasons to suggest that alteration of thyroid hormone binding is not the only feature common to these mutations. First, some THR mutations have little effect on thyroid hormone binding (24). Second, thyroid hormone binding is eliminated by mutations in a number of regions in the carboxyl terminus other than the domain affected by naturally occurring THR mutations? Third, inspection of the locations of the THR mutations suggests that there is a predisposition for two domains centered around amino acids 345 and 453 (Fig. 6) (24, 25). These features suggest that in addition to alterations in thyroid hormone binding, the locations of the THR mutations have been se-T. Nagaya, unpublished data.

TRETK Luc TSHa Luc
Fold Stimulation by T3 % Relative Activity (+/-T3) lected on the basis of some other aspect of receptor function that is required for dominant negative activity.
The observation that the THR mutations are located on either side of a putative dimerization domain leads to an intriguing hypothesis that mutations that block dimerization would not be capable of dominant negative activity and therefore would not result in a dominant form of THR. The dimerization domain of the thyroid receptor has not been characterized in detail. However, it has been proposed that a series of heptad repeats between amino acids 334 and 428 create a hydrophobic interface that is important for proteinprotein interactions (23). The importance of the putative dimerization region was first demonstrated in transient expression assays in which a series of carboxyl-terminal fragments containing this region were shown to inhibit thyroid receptor-mediated transactivation (43). Because these carboxyl-terminal receptor fragments did not include the DNAbinding domain of the receptor, their inhibitory activity was proposed to occur on the basis of forming functionally inactive dimers. Subsequent experiments using gel mobility shift assays have confirmed the ability of the dimerization domain fragments to block the formation of homodimers (15,44).
Although the dimerization interface appears to involve an extensive region of the carboxyl terminus, the role of the last (the 9th) heptad repeat has been documented most thoroughly (18, 20, 21). For example, deletion of this region prevents formation of homodimers and heterodimers. In addition, the TRa2 splicing variant, which disrupts the 9th heptad repeat, does not form heterodimers effectively (45). The presence of a proline for serine substitution in the 9th heptad region of truncated mutant (entire deletion of amino terminus and DNA binding region) of c-erbA has also been shown to impair dominant negative activity by reducing dimerization (46). For these reasons, we created a mutation in one of the hydrophobic residues (Leu4%) of the 9th heptad repeat to test the requirement for dimerization for dominant negative activity of the THR mutants. Two THR mutants were selected to represent examples of the mutation clusters that are proximal (G345R) or distal (P453H) to the dimerization domain.
In the studies of mutant receptor binding to DNA, both types of THR mutants (G345R, P453H) bound to TRE sequences as homodimers or as heterodimers with RXRa (Figs. 2 and 3). The L428R mutant bound to TREs as a homodimer, but failed to form the strong heterodimer complex with RXRa. This finding was unexpected and suggests that the receptor sequences involved in homodimerization and heterodimerization may be different.
In the transient expression assays, the dimerization domain mutant (L428R) did not inhibit wild type receptor function, even though it was unable to bind thyroid hormone. This observation supports the view that the absence of thyroid hormone binding is not sufficient for dominant negative activity of mutant receptors. The observation that addition of the dimerization domain mutant to the THR mutations (G345R/L428R and P453H/L428R) eliminated their dominant negative activity strongly suggests that dimerization is required for this property. Based upon the DNA binding studies of the dimerization mutant, it is likely that heterodimer formation between the mutant receptor and auxiliary proteins is more important than homodimerization for dominant negative activity by the THR mutants.
In a previous study in which additional mutations were introduced into the DNA-binding domain of THR mutants (13), we suggested that the major mechanism of dominant negative inhibition by the THR mutants involves competition of the mutant receptors for DNA target sites. The current results are consistent with this model and extend it to include the concept that dimerization, probably with an accessory protein such as RXR, is required for high affinity competition for the DNA target sites (Fig. 7). This concept is also supported by the fact that the naturally occurring THR mutations have not been found in the DNA-binding domain or in the dimerization region. This model should be qualified, however, by the fact that these experiments do not define the nature of the thyroid receptor complexes in transfected cells. Moreover, the diversity of the TRE sequences in various genes leaves open the possibility that dominant negative activity of specific mutants could vary for different genes or in a tissuespecific manner.