Ligand-dependent synergy of thyroid hormone and retinoid X receptors.

The binding of thyroid hormone receptors to DNA is enhanced by heterodimerization with nuclear proteins. One such heterodimerization partner has recently been characterized as the retinoid X receptor. 9-cis-Retinoic acid has been identified as a natural ligand for retinoid X receptors, suggesting a potential receptor-mediated interaction between thyroid hormone and 9-cis-retinoic acid in the regulation of thyroid hormone-responsive genes. A transient cotransfection assay was used to test for such an interaction. When a complex thyroid hormone response element composed of both direct and inverted repeat hexamers was tested, these two ligands activated gene expression synergistically. In contrast, when the response element consisted only of directly repeated hexamers, unliganded retinoid X receptors enhanced thyroid hormone responsiveness, but 9-cis-retinoic acid induced no additional activation. The results suggest a unique mechanism to achieve differential suggest a unique mechanism to achieve differential thyroid hormone sensitivity of thyroid hormone-responsive genes within a cell. Genes with appropriate response elements will show amplification of the thyroid hormone response by 9-cis-retinoic acid in the presence of retinoid X receptors; other thyroid hormone-responsive genes will be influenced by retinoid X receptors, but not 9-cis-retinoic acid.

Retinoid X receptors (RXRs) also are members of the erbA superfamily. RXRs were discovered in the process of cloning transcription factors that regulate murine major histocompatibility class I genes (8), and also by low stringency hybridization of cDNA libraries with a retinoic acid receptor (RAR) DNA binding domain probe (9). The ligand for RXRs was not immediately apparent; hence, these were considered to be orphan receptors. However, pharmacologic doses of all-transretinoic acid (RA) were capable of activating RXRs, suggesting that the genuine ligand may be an unknown retinoid, dubbed retinoid X (9). Recently, 9-cis-RA has been identified as retinoid X (10, 11).
Several investigators have shown that RXRs can heterodimerize with TRs on TREs and that RXRs can enhance the ability of TRs to trans-activate target genes (12-17). Thus, RXRs can function as TRAPs. However, these initial studies identifying RXRs as TRAPs were performed before the actual RXR ligand was characterized. The discovery of 9-cis-RA suggests a potential interaction not simply between TRs and RXRs, but also indirectly between T3 and 9-cis-RA in the regulation of T3-responsive genes. We now describe a complex receptor-mediated interaction between these two ligands. In the presence of TR and RXR (but not RAR), T3 and 9-cis-RA synergistically* trans-activate a gene that contains a TRE with an inverted repeat motif. However, 9-cis-RA is without effect on a gene that has a simple direct repeat TRE. On this TRE, unliganded RXR is capable of enhancing T 3 induction of target gene expression.

MATERIALS AND METHODS
Cell Culture and Transfections-JEG-3 cells were grown in 90% Eagle's medium plus 10% Calf Supreme (GIBCO), and were transfected using calcium phosphate precipitation (18). Rat TRP in the expression vector pCDM has been described (18). Rat RXRp (12) and RARa (19) were expressed from the same vector. High level TR@ transfections utilized 3 pg of pCDMTR0, and low level transfections utilized 25 ng. Transfections also included 3 pg of pCDMRXRp or 3 pg of pCDMRARa as indicated. The vector pCDM was added where needed to achieve a total of 6 pg of pCDM-based plasmid/transfection. Three T3-responsive reporter plasmids were used at a dose of 4 pg/ transfection. The reporter plasmid pTK35BA contains two copies of a modified rat growth hormone gene TRE (including the direct and inverted repeats) 5' to a basal thymidine kinase promoter directing chloramphenicol acetyltransferase (CAT) expression (20). The plasmid pTK14AA contains two copies of a directly repeated element 5' to the thymidine kinase promoter but no inverted repeats (18, 21), and pTK39 contains only the inverted motif (TGAGGTCAT-GACCTCA). To control for transfection efficiency, each transfection also included 2 pg of pTKGH, in which the basal thymidine kinase promoter directs expression of human growth hormone (22). Cells in 60-mm Petri dishes were transfected in the presence of 10% charcoalstripped Calf Supreme supplemented with 100 nM dexamethasone, and then cultured for 2 days k T3 (0.2 or 100 nM) + 9-cis-RA (100 nM). This dose of 9-cis-RA was an ED,, in transfection studies using RXRa and a specific RXR response element (11). Cells were harvested for CAT assay and media for human growth hormone assay as described (18). Results are expressed as -fold CAT induction, defined as CAT/human growth hormone for cells cultured with ligand divided by CAT/human growth hormone for cells cultured without In this paper two ligands are considered to act synergistically if the response to both together is greater than the sum of their individual responses. ligand. Results are the mean & S.E. for at least 4 transfections/ condition.
Electrophoretic Mobility Shift Assay fEMSA)-[35S]Methioninelabeled and non-radiolabeled proteins (RXRP and TRP) were produced by translation in rabbit reticulocyte lysate. Protein DNA binding incubations included 15,000 cpm RXRD or TRP, 40 ng of rat growth hormone TRE or a palindromic TRE (23) and 1.4 pg of poly(dI.dC) in 35 pl of 20 mM HEPES, pH 7.8, 50 mM KCI, 1 mM dithiothreitol, 20% glycerol, 0.1% Nonidet P-40. To assess heterodimer formation, incubations included nonradiolabeled TRD or RXRP along with radiolabeled RXRP or TRP, respectively. In some experiments JEG cell extracts (7) were used in place of in uitro translated RXRP as a source of TRAP, which can function like RXRP. Incubations were at room temperature for 40 min prior to loading onto an 8% polyacrylamide gel at 4 "C. Gels were dried and analyzed by fluorography (AutofluorTM, National Diagnostics).

RESULTS AND DISCUSSION
T o evaluate the potential receptor-mediated interaction between T3 and 9-cis-RA, JEG-3 cells were transiently cotransfected with a TRP expression vector in the absence or presence of an RXRP expression vector. The reporter plasmid used, pTK35BA, contains a CAT gene driven by a TRE derived from the rat growth hormone 5'-flanking region (20). When cells were transfected with high levels of TRP but no RXRP, a physiologic dose of T 3 induced CAT expression a modest 4-fold ( Fig. la, top). 9-cis-RA by itself was without effect, but the combination of T 3 plus 9-cis-RA led to an 8fold induction of CAT, possibly due to interaction with endogenous RXRs. Synergy between T3 and 9-cis-RA was more clearly observed when RXRO was cotransfected with TRP. Under these conditions, induction by T 3 or 9-cis-RA alone was 4-6-fold, but the simultaneous addition of both ligands led to a 24-fold induction of CAT ( Fig. la, middle). It is important to note that, in the absence of 9-cis-RA, RXRP caused virtually no enhancement of the T3 induction of CAT ( Ligand-dependent synergy of TRj3 and RXRj3 in trans-activation through a TRE derived from the rat growth hormone promoter. a, JEG-3 cells were transfected with a high level (3 pg) of TRP alone, or that plus either RXRB or RARa. Cells were then cultured without ligands, with a physiologic dose (0.2 nM) of T3 alone, with 9-cis-RA alone (100 n M ) , or with T3 plus 9-cis-RA for 2 days prior to harvest. The combination of T3 plus 9-cis-RA synergistically induces CAT when TRP and RXRP are cotransfected (middle three bars), but not when TRP and RARa are cotransfected (bottom three bars). b, similar to above except transfection with a low level (25 ng) of TRP and cultured with a receptor saturating dose (100 nM) of T3.
Although this contrasts with the work of other investigators (12, 13, 16), we do find an effect of RXRP in the absence of 9-cis-RA under other circumstances (see below).
Retinoic acid receptors represent a second class of 9-cis-RA receptors (11) that, in addition, are activated by all-trans-RA. Both RXRs (12-16) and RARs (24) heterodimerize with TRs in uitro, and all of these receptors bind to DNA elements that contain the hexameric consensus AGGTCA (12,20,21,25). In a manner similar to that seen with RXRP, 9-cis-RA will induce CAT 6-fold in the presence of RARa and TRP (Fig.  la, bottom). However, in marked contrast to the situation with RXRP, the combination of RARa and TRP does not support further enhancement of gene expression with T3 plus 9-cis-RA. A trivial explanation for the lack of ligand-dependent synergy in the presence of RARa would be squelching of crucial transcription factors by excess RARa. However, RARa did not inhibit expression of the cotransfected internal control plasmid pTKGH (data not shown). RARa may compete with TRP for binding to the TRE, thus limiting the ability of T 3 to induce CAT with or without 9-cis-RA.
Similar results were obtained using low levels of transfected TR. In these experiments, a receptor-saturating dose of T3 was used so that the effects of T 3 alone could be measured reproducibly. In the presence of transfected TRP plus RXRP (Fig. lb, middle), CAT was induced 12-fold by T3, 5-fold by 9-cis-RA, and 37-fold by both ligands together. When RARa was substituted for RXRP, 9-cis-RA induced CAT almost 9fold, but no further enhancement was seen with T 3 plus 9cis-RA (Fig. lb, bottom). Once again, no enhancement of T3induced trans-activation was noted when RXRP was cotransfected with TRP in the absence of 9-cis-RA.
When vector was transfected in place of TRP, CAT induction with 100 nM T 3 alone was only %fold, and no synergy was observed with 9-cis-RA (data not shown). The low level of T3 induction is consistent with the low level of endogenous TRs in JEG cells (10 fmol/lOO pg DNA) (26). In addition, if the TRE was omitted from the reporter plasmid, CAT induction by T3 and 9-cis-RA did not exceed 2-fold, even in the presence of high levels of TRP and RXRP (data not shown).
Several mechanisms could potentially account for the synergy between T3 and 9-cis-RA on this TRE. 9-cis-RA could enhance binding of TR/RXR heterodimers to the TRE. This was tested directly using an electrophoretic mobility shift assay (EMSA), but 9-cis-RA had no effect on the magnitude of DNA binding or on the mobility of the protein DNA complex (Fig. 2). A second possibility would be for 9-cis-RA to increase the amount of RXR in the cell, or to induce another protein of similar function. To test this, extracts were prepared from cells that had been transfected with RXRP and then cultured with or without 9-cis-RA. Graded doses of these extracts were then incubated with TR and DNA, and heterodimer formation was assessed by EMSA (Fig. 3). Transfection with RXRP increased the potency of the extracts in this assay, as expected (data not shown). However, cells cultured with 9-cis-RA had no more heterodimerization activity than cells cultured without 9-cis-RA. It also might be possible for 9-cis-RA to increase the number of ligand-occupied TRs. This was excluded by finding no difference in nuclear [I2'I]T3 binding from cells that had been transfected with TRP plus RXRp and then incubated with or without 9-cis-RA (data not shown). In the absence of specific evidence to support any of the above mechanisms, it is likely that the ability of 9-cis-RA to synergize with T3 reflects a heterodimer conformation that is optimized for interaction with other transcription factors by the presence of both ligands.
The TRE used in the above studies is derived from the 5'- flanking region of the rat growth hormone gene (20). This TRE contains three nearly perfect copies of the consensus hexamer, the most 3' of which is inverted relative to the other two. Mutational analysis clearly indicates all three hexamers are required for full T 3 responsiveness of the rat growth hormone T R E (20). However, TREs can be constructed simply with directly repeated elements and no inverted repeats (18,21). The malic enzyme gene may contain a natural example of this class of T R E (27, 28). Therefore, it was of interest to examine the effects of T3, g-cis-RA, and their receptors on expression of a reporter gene driven by a TRE that contains only directly repeated hexamers.
Surprisingly, the results show only minimal effects of 9-cis-RA on T3-induced gene expression, but substantial effects of RXRP alone (Fig. 4). In the absence of cotransfected RXRP, T 3 induction of CAT was 17-fold. However, this rose to 56fold with cotransfected RXRP, using T3 as the only ligand. Although RXRB showed a clear effect in the absence of 9-cis-RA, no further enhancement of gene expression was seen when this ligand was added along with T3. A similar trend, although quantitatively less impressive, was seen when pTK14AA was transfected under conditions identical to those of Ligand-dependent and -independent effects of TRB and RXRB in trans-activation through a TRE containing only directly repeated elements (pTK14AA). JEG-3 cells were transfected with a high level of TRD with or without RXRP. Cells were then cultured without ligands, with T 3 alone, with 9-cis-RA alone, or with T3 plus 9-cis-RA for 2 days prior to harvest. The ligand doses (100 nM each) were identical to those in Fig. lb. RXRP enhances T 3 induction of CAT (compare first bar of upper section with first bar of lower section), but 9-cis-RA gives no further induction of CAT.
of cotransfected RXRB and 9-cis-RA. When RXRP and TRP were cotransfected, CAT induction was 6.2-fold with T3 as the only ligand and 6.4-fold with 9-cis-RA plus T3. Thus, in contrast to the rat growth hormone like TRE, this simple direct repeat TRE is not capable of showing ligand-dependent synergy between TRP and RXRP. Also in contrast to the rat growth hormone-like TRE, T3 induction of this TRE is enhanced by RXRB without 9-cis-RA. Perhaps on this direct repeat TRE the heterodimer conformation is optimized for interaction with other factors without g-cis-RA, and hence this ligand has no additional effect.
To confirm that the ligand-dependent synergy observed in Fig. 1 with the complex TRE did indeed depend on the inverted repeat motif, similar studies were performed with the reporter plasmid pTK39, in which the TRE contains only an inverted repeat. As predicted, synergy was observed in the presence of 3 pg of TRP and RXRP cDNAs; ligand induction of CAT was 7.3-fold for T3 alone, 3.1-fold for g-cis-RA, and 15.3-fold for both ligands together.
Thyroid hormone regulates a vast number of metabolic processes. The degree of T 3 response varies greatly from one gene to another, and some genes are highly T3-responsive in one tissue but not another, even though both may contain TRs (29, 30). The mechanisms underlying this diversity and specificity of T 3 action are not well understood, but the studies reported here suggest one possible explanation. Two TREs that are functionally identical in the presence of TRs may behave differently if 9-cis-RA and/or RXR also are present within the cell. The unoccupied RXR can heterodimerize with TRs and enhance the ability of T3 to induce expression from certain T3-responsive genes; other genes will show a synergistic enhancement of T 3 response only in cells that contain both RXR and 9-cis-RA. Given this, it will be important to determine the factors that regulate intracellular RXR and 9-cis-RA levels. If 9-cis-RA is produced enzymatically there may be important metabolic factors that regulate its production. (Precedent for enzymatic isomerization of retinoids is found in the visual system, where an isomerase catalyzes conversion of all-trans-to see Ref. 31.) The important discriminating effects of 9-cis-RA on T3 action may be specific for RXRs, since a second class of 9cis-RA receptors, the RARs, do not mediate similar effects when studied under identical conditions. These studies also may bear relevance to the syndrome of generalized resistance to thyroid hormone. Patients with this condition require supraphysiologic levels of T3 to remain clinically euthyroid. Most cases appear to be due to mutations that impair ligand binding of TRP (32, 33). Our studies would predict that mutations in an RXR gene might lead to a clinical syndrome Thyroid Hormone and 9-cis-Retinoic Acid Synergy 22013 with some features that overlap those of thyroid hormone resistance. Mutations that impair 9-cis-RA binding to an RXR might lead to diminished T3 induction of a subset of T3-responsive genes in those organs that contain RXR and 9-cis-RA. However, mutations of TRP or an RXR that impair heterodimerization might inhibit T3 induction of a wider array of genes. Clearly, the ultimate level of hormonal effect in uiuo is the result of complex interactions between multiple receptors and ligands which are just beginning to be understood.