Retinoid X receptor alpha represses GATA-4-mediated transcription via a retinoid-dependent interaction with the cardiac-enriched repressor FOG-2.

Dietary vitamin A and its derivatives, retinoids, regulate cardiac growth and development. To delineate mechanisms involved in retinoid-mediated control of cardiac gene expression, the regulatory effects of the retinoid X receptor alpha (RXR alpha) on atrial naturietic factor (ANF) gene transcription was investigated. The transcriptional activity of an ANF promoter-reporter in rat neonatal ventricular myocytes was repressed by RXR alpha in the presence of 9-cis-RA and by the constitutively active mutant RXR alpha F318A indicating that liganded RXR confers the regulatory effect. The RXR alpha-mediated repression mapped to the proximal 147 bp of the rat ANF promoter, a region lacking a consensus retinoid response element but containing several known cardiogenic cis elements including a well characterized GATA response element. Glutathione S-transferase "pull-down" assays revealed that RXR alpha interacts directly with GATA-4, in a ligand-independent manner, via the DNA binding domain of RXR alpha and the second zinc finger of GATA-4. Liganded RXR alpha repressed the activity of a heterologous promoter-reporter construct containing GATA-response element recognition sites in cardiac myocytes but not in several other cell types, suggesting that additional cardiac-enriched factors participate in the repression complex. Co-transfection of liganded RXR alpha and the known cardiac-enriched GATA-4 repressor, FOG-2, resulted in additive repression of GATA-4 activity in ventricular myocytes. In addition, RXR alpha was found to bind FOG-2, in a 9-cis-RA-dependent manner. These data reveal a novel mechanism by which retinoids regulate cardiogenic gene expression through direct interaction with GATA-4 and its co-repressor, FOG-2.

Retinoids are compounds, derived from vitamin A, which exert pleiotropic effects on cellular differentiation, morphogenesis, and metabolism. The effects of retinoids on cardiac growth and development are well recognized. Offspring of rodents and birds fed a vitamin A-deficient diet display congenital defects involving many organ systems including the cardiovascular system (1,2). Vitamin A replacement at different stages of embryonic development in these animals actually alters the severity and type of heart defects seen at birth indicating that retinoid signaling pathways play key roles in a variety of developmental programs (1). Similarly, retinaldehyde dehydrogenase-deficient mice die in utero at embryonic day 8 secondary to poor maturation of the ventricular myocardium, an effect that can be rescued by supplementation of maternal vitamin A (3,4). In addition, retinoids can exert teratogenic effects. Human offspring born of mothers who ingested isotretinoin, a vitamin A analog, during pregnancy exhibited a high incidence of complex cardiac malformations including transposition of the great vessels, tetralogy of Fallot, hypoplastic aortic arch, and ventricular septal defects (5). Moreover, in the developing chick, local application of high concentrations of retinoic acid disrupts the migration of the pre-cardiac mesoderm, an effect that is dose-dependent and developmental stage-specific (6).
Retinoids bind to ligand-activated nuclear receptor transcription factors called retinoid receptors. The retinoid receptors comprise a subfamily within the nuclear receptor superfamily including the retinoic acid receptors (RARs) 1 and the 9-cis-retinoic acid or retinoid X receptors (RXRs). RXRs participate in nuclear receptor regulatory pathways in two distinct manners. First, RXRs serve as heterodimeric partners for other class II nuclear receptors such as thyroid receptor, vitamin D receptor, peroxisome proliferator-activated receptors (PPARs), or RARs. Heterodimeric complexes containing RXR in the presence of ligand, recognize and bind cognate DNA elements within the regulatory regions of target genes. Alternatively, RXR can form homodimers in the presence its own ligand, 9-cis-retinoic acid (9-cis-RA). Thus, RXR has been termed a "master regulator" of retinoid signaling pathways. Interestingly, the transcriptional regulatory effects mediated by retinoids is influenced at multiple levels including specificity of the ligand-receptor interaction, temporal and spatial variations in receptor expression, selection of heterodimeric partners (some of which are permissive for liganded RXR), and the levels of specific co-activators and co-represssor molecules within target tissues. More recently, RXRs and other members of the nuclear receptor superfamily have been shown to interact with nonnuclear receptor transcription factors providing yet another level of complexity in the retinoid signaling pathway (7)(8)(9). Accordingly, it is not surprising that the observed effects of retinoids on cardiac morphogenesis involves multiple developmental stages and cell types.
The cardiac phenotype of RXR␣-null mice provides additional evidence for the importance of retinoid signaling in the development, growth, and maturation of the heart. The ventricular myocardium does not develop normally in RXR␣-null mice leading to heart failure at embryonic day 15 (10,11). The incidence of conotruncal defects in RXR␣Ϫ/Ϫ mice is greater than 50% (12). Furthermore, RXR␣ heterozygous animals display a wide array of defects involving the conotruncal region, atrioventricular canal, and ventricular myocardium. Taken together, these data indicate that retinoid signaling via RXR␣ serves a critical role in cardiac growth and development programs. However, the molecular mechanisms by which RXR exerts regulatory effects within the complex process of cardiac morphogenesis remains unclear.
As an initial step in the characterization of mechanisms involved in retinoid-mediated cardiac gene regulation, we sought to use an experimental system in which retinoid signaling was isolated from the myriad of signals concurrent during development. To this end, we focused on the observations of others that retinoids inhibit cardiac myocyte hypertrophic growth in vitro. In response to a variety of mechanical or hormonal stimuli, the terminally differentiated cardiomyocyte undergoes hypertrophic growth, a physiological response characterized by an increase in cell size, an increase in sarcomeric organization, and activation of the fetal expression pattern of contractile and metabolic genes (reviewed in Ref. 13). The cellular regulatory pathways involved in the induction of cardiac hypertrophy and development of heart failure are currently being unraveled. Atrial natriuretic factor (ANF) gene expression is increased during cardiac hypertrophic growth and has been used as a consistent marker of the hypertrophy program. Several groups have shown that RXR␣, in the presence of 9-cis-RA, represses the hypertrophic growth of cardiomyocytes in culture (14,15). We hypothesized that delineation of cis-elements and trans-acting factors mediating the effects of 9-cis-RA on ANF gene transcription would provide information about the mechanisms involved in the regulation of cardiogenic gene transcription by retinoids. In this report, we demonstrate that liganded RXR␣ regulates ANF gene transcription through a mechanism uncharacteristic of previously described retinoid signaling pathways. Our results indicate that RXR␣ represses ANF promoter activity in cardiomyocytes by interacting directly with the transcription factor GATA-4 and its co-repressor FOG-2. These results identify a mechanism whereby retinoid signaling modulates gene expression in the heart and unveils a new permutation in the retinoid signaling pathway.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfections-Rat neonatal cardiomyocyte cultures were prepared as described (16). Briefly, hearts were removed from 1-day-old Sprague-Dawley rats, the atria and great vessels were trimmed off, and tissue was finely minced followed by sequential digestion with 0.5 mg/ml collagenase (WAKO, Richmond, VA). Ventricular cardiomyocytes were separated from fibroblasts by differential plating and were cultured in gelatin-coated 12-well tissue culture plates (0.4 hearts/well) in media containing Dulbecco's modified Eagle's medium, 10% horse serum, 5% fetal calf serum, 100 M bromodeoxyuridine, penicillin, streptomycin, and L-glutamine. Cardiomyocytes were transfected with promoter-reporter constructs and transcription factors using the calcium-phosphate precipitation method (17) with 4 g of reporter and 0.5 g of receptor per well. After 24 h, media was changed to serum-free media containing 10 g/ml transferrin, 10 g/ml insulin, and 100 M bromodeoxyuridine, with the addition of phenylephrine (5 M) and retinoid ligand daily for 3 days. CV-1 cells (African green monkey kidney fibroblasts, obtained from ATCC, Manassas, VA) were maintained in minimal essential medium with 10% fetal bovine serum. Transfections of CV-1 cells were also performed using the calcium-phosphate precipitation method in minimal essential medium supplemented with fetal bovine serum that was pre-stripped using activated charcoal, to remove retinoid ligand normally present in serum. 9-cis-Retinoic acid (9-cis-RA) and TTNPB ((E)-4-(2-(5,5,8,8-tetramethyl-5,6,7,8-tetrahydro-2-naphthalenyl)-1-propenyl)benzoic acid) were obtained from Sigma. Luciferase was assayed in cardiomyocytes 48 -72 h after transfection. Cells were washed with phosphate-buffered saline and lysed using Promega reporter lysis buffer. Luciferase activity in the cell lysates was assayed using BD Pharmingen substrates A and B according to the manufacturer's instructions in a analytical luminescence luminometer. Growth hormone was assayed from supernatants 72 h after transfection of cardiocytes or 48 h after transfection of CV-1s, using a kit from Nichols Institute Diagnostics (San Juan Capistrano, CA). All reporter activities were corrected for the activity of an SV40-␤-galactosidase plasmid.
GST Pull-downs-GST fusion proteins were produced using pGEX vectors (Amersham Biosciences) in DE3 bacteria according to the manufacturer's instructions with isopropyl-1-thio-␤-D-galactopyranoside induction of bacterial protein expression with some modifications: insoluble proteins were solubilized in 7 M urea/phosphate-buffered saline followed by dialysis. Proteins were verified by Coomassie staining of SDS-PAGE gels and by Western blot using an anti-GST antibody. Bacterial lysate containing 2 g of the GST fusion protein or control GST protein were bound to 15 l of glutathione-Sepharose in bead binding buffer (20 mM Tris, pH 7.4, 60 mM sodium chloride, 1 mM dithiothreitol, 15% glycerol, 0.1% Nonidet P-40 (Calbiochem)) at room temperature and then washed in bead binding buffer. Five l of in vitro translated protein (TNT T7 Quick-coupled kit according to the manufacturer's instructions) was incubated for 1 h at room temperature, washed with bead binding buffer, and resolved on SDS-PAGE, then visualized by autoradiography. In binding studies involving retinoid ligand, ligand was included in the wash buffers as well as the protein incubation period.
Statistical Analyses-Significant differences (p Ͻ 0.05) were determined using unpaired, two-tailed Student t tests.

RXR␣ Represses Basal and ␣ 1 -Adrenergic Agonist-mediated Activation of the ANF Promoter via Elements within the Proximal Promoter Region-
The traditional mechanism of retinoidmediated transcriptional control occurs via the action of liganded retinoid receptors bound to DNA recognition sequences within regulatory regions of target genes. Analysis of the rat ANF promoter sequence failed to reveal retinoid receptor recognition sequences (AGGTnA). Thus, we hypothesized that the known repression of ANF gene transcription by retinoids occurs through an indirect mechanism such as the modulation of trans-acting factors by RXR. The cis-acting regulatory elements necessary for RXR-mediated repression of ANF gene transcription were localized by assaying a 5Ј deletion series of the rat ANF promoter fused to a luciferase reporter (ANF.Luc) in rat neonatal cardiomyocytes. The experiments were performed in the absence and presence of the ␣ 1 -adrenergic agonist, phenylephrine, to allow evaluation of effects on basal activity and hypertrophic response. Cotransfection of an expression plasmid for RXR␣ markedly repressed the activity of Ϫ633ANF.luc in a 9-cis-RA-dependent manner in the presence or absence of phenylephrine (Fig. 1A). Similar RXR␣-dependent repression was observed for the Ϫ316ANF.luc and Ϫ147ANF.luc constructs (Fig. 1A). Given that the basal activity of the smallest construct (Ϫ107ANF.luc) was very low it was not possible to determine with certainty whether it localized the cis-acting element mediating the repression by retinoids.
These results localized the cis-acting elements involved in the RXR-mediated repression of ANF gene transcription to a region within 147 nucleotides of the transcription start site. Importantly, this region contains response elements for several transcription factors known to be critical for the transcriptional control of ANF and other cardiac genes; GATA-4, a zinc finger transcription factor important in cardiogenesis and the cardiac hypertrophic program (19 -21); serum response factor (SRF), a transcription factor critical for basal and ␣ 1 -adrenergic agonistinduced activation of ANF gene expression (22,23); and Nkx2.5, a cardiac-enriched homeobox transcription factor necessary for proper cardiac development (24) and a known coactivator of both GATA-4 and SRF (25)(26)(27)(28)(29)(30).
To evaluate the role of the proximal GATA-response element (GATA-RE) in the retinoid-mediated repression of the ANF gene promoter, the site was mutated in the context of Ϫ316ANF.luc (Ϫ316⌬GATA.ANF.luc) and Ϫ147ANF.luc (Ϫ147⌬GATA.ANF.luc) for use in cotransfection assays. The basal activity of both Ϫ316⌬GATA.ANF.luc and Ϫ147⌬GATA.ANF.luc, although still active, was significantly diminished compared with the corresponding wild-type constructs (Fig. 1B). The retinoid-mediated repression observed with Ϫ316⌬GATA.ANF.luc was significantly less than that of Ϫ316ANF.luc but was not completely obliterated (30% compared with 94%, Fig. 1B). Similarly, the repression of the Ϫ147⌬GATARE ANF.luc construct was also significantly less than that of the wild-type construct (42 compared with 88%, Fig. 1B). Taken together, these data indicate that the GATAresponse element is necessary for the full repression conferred by retinoids in cardiomyocytes, but also suggest that other elements may play a role. Furthermore, the importance of the interaction between adjacent elements within the native promoter cannot be underestimated, especially in light of the known cooperativity between GATA, Nkx2.5, and SRF. Retinoid-mediated Repression of the ANF Promoter Is Liganddependent and RXR-specific-RXR is capable of heterodimerizing with other nuclear receptors including members of the RAR, vitamin D receptor, and PPAR families. To define the receptor specificity of the RXR␣-mediated repression of the ANF promoter, the requirement of its activation domain and the effects of RXR homodimerization were analyzed using two mutants, mRXR␣F318A and RXR⌬AF-2. RXR␣F318A is a constitutively active mutant (31). The conformational change induced by the phenylalanine to alanine substitution allows the RXR␣F318A mutant to exist in the dimer but not the tetramer form and to possess constitutive trans-activation properties in the absence of exogenous ligand (32,33). RXR⌬AF-2, which lacks the COOH-terminal 19 amino acids, is unable to recruit co-activator molecules (31,34). We confirmed the predicted activity of the RXR␣ mutants in CV-1 cells using a known RXR-response element (RXR-RE) derived from the cellular retinoid-binding protein II gene ( Fig. 2A). CV-1 cells were used for these experiments because they are largely devoid of endogenous retinoid receptor activity. As expected, wild-type RXR␣ conferred marked activation of the RXR-RE only in the presence of the ligand 9-cis-RA ( Fig. 2A). In contrast, the mRXR␣F318A mutant activated the RXR-RE in the absence of ligand. As expected, RXR⌬AF-2 was inactive with the RXR-RE. We then assayed the effect of these mutants on ANF promoterreporter activity in ventricular cardiomyocytes (Fig. 2B). ANF promoter activity was profoundly repressed with the F318A mutant in the absence of ligand. In contrast, the RXR⌬AF-2 mutant did not confer repression. These latter results indicated that as predicted by the 9-cis-RA dependence, an intact AF-2 domain is required for the repressive effects of RXR␣ on the ANF promoter. Moreover, these data demonstrate that the repressive effect is mediated largely by RXR␣ rather than one of its nuclear receptor partners.
The RXR specificity of the repressive response was further evaluated by dose-response studies using RXR-and RAR-specific ligands in the cardiomyocyte culture system. The RXR ligand, 9-cis-RA, repressed Ϫ633ANF.luc in a dose-dependent manner only in the presence of RXR␣ (Fig. 2C). Addition of RAR␣ had no affect on the 9-cis-RA response (Fig. 2B). Specifically, 9-cis-RA did not mediate a repressive response in the presence of RAR␣ alone, and the effects observed with addition of both RXR␣ and RAR␣ were similar to that of RXR␣ alone. Moreover, a dose-response analysis using the RAR-specific ligand TTNPB revealed that at no concentration did liganded RAR alone confer repression (Fig. 2D). In the presence of RXR␣, however, repression was observed at a concentration 100-fold greater than the effective 9-cis-RA concentration (Fig.  2D). Similar studies to analyze "permissive" RXR heterodimers demonstrated that PPAR␣/RXR␣ heterodimers, at high concentrations of the PPAR␣ ligand ETYA (10 Ϫ5 M), conferred only 40% repression of ANF promoter activity in cardiomyocytes (data not shown). Furthermore, VDR/RXR␣ heterodimers demonstrated a ligand dose-dependent repressive effect on ANF promoter activity, with 30 and 60% repression at 10 Ϫ7 and 10 Ϫ6 M vitamin D, respectively (data not shown). Accordingly, the effects of ligands specific for permissive heterodimeric partners of RXR did not approach the potency seen with RXR ligands. Taken together these results indicate that liganded RXR is the dominant receptor conferring repression of the ANF promoter. Fig. 1 demonstrated that the GATA-response element present in the proximal region of the ANF gene promoter was involved in the retinoid-mediated repression. However, a role for SRF or Nkx2.5 was not excluded. To explore this further, the effect of RXR␣ on GATA-4, SRF, and Nkx2.5 was explored using independent reporters containing response elements for each of the factors. RXR␣/9-cis-RA had no affect in cardiomyocytes on the activity of reporters containing heterologous Nkx2.5 or SRFresponse elements, respectively (data not shown). In striking contrast, the activity of a GATA-4 responsive construct (GATA-RE 6 .GH) was markedly repressed by RXR␣/9-cis-RA. Specifically, the high level activity of GATA-RE 6 .GH in cardiomyocytes was abolished by RXR␣ in a 9-cis-RA-dependent manner (Fig. 3A). Interestingly, the RXR␣-mediated repression of GATA-RE 6 .GH was not observed in the noncardiomyocyte cell line CV-1, even in the presence of overexpressed GATA-4 (Fig.  3B). Furthermore, addition of Nkx-2.5, SRF, or the known co-repressing RXR interacting partner SHP, alone or in combination, did not recapitulate the RXR␣-mediated repression in CV-1 cells (data not shown). These results together with the data shown in Fig. 1B demonstrate that the GATA-response element is sufficient for retinoid-mediated repression and necessary for the full repressive effect on the ANF gene promoter.

RXR␣ Antagonizes GATA-4 in a Ligand-dependent, Cardiomyocyte-specific Manner-The data shown in
RXR␣ Interacts Directly with GATA-4 in a Ligand-independent Manner-The data shown above suggested that liganded RXR␣ inhibits GATA-4 activity through a cardiomyocyte-specific mechanism, possibly via the effects of a cell-specific repressor recruited by RXR␣. Alternatively, GATA-4 expression could be down-regulated by a retinoid signaling pathway. However, the amount of cardiomyocyte GATA-4 protein detected by Western blot analysis was not affected by overexpression of RXR␣ in the presence of 9-cis-RA (data not shown). To determine whether RXR interacts directly with GATA-4, GST pulldown assays were performed. GATA-4 bound to GST-RXR␣ in an interaction that, surprisingly, did not require ligand (Fig. 4). Indeed, addition of 9-cis-RA had no effect on the binding. The specificity of this interaction was confirmed by the lack of binding with GST alone (Fig. 4). The interaction of RXR␣ with PPAR␣ is also shown as a control. These data demonstrate a direct interaction between RXR␣ and GATA-4 but does not explain the precise role of ligand in the RXR-mediated repression of the ANF-promoter.
To further characterize the specificity of the interaction between RXR␣ and GATA-4 and to map the relevant domains within RXR␣, a series of RXR␣ truncation and point mutants were used in GST pull-down studies. The results demonstrated that the C domain of RXR, which contains the DNA binding site, is critical for the interaction with GATA-4 (Fig. 5A). Specifically, an 35 S-labeled RXR␣ amino-terminal deletion mutant protein containing the C domain (aa 132-467), bound GST-GATA-4 while an amino-terminal deletion mutant lacking the C domain (203-467) did not bind (Fig. 5A). An RXR␣ fragment containing only the C domain (132-210) also bound GATA-4 (Fig. 5A). In addition, the RXR⌬AF-2 mutant (1-449) also bound GATA-4. Of note, this latter mutant did not confer repression on the ANF promoter providing additional evidence that the RXR␣/GATA interaction alone is not sufficient to confer the observed repression. Collectively, these data suggest that GATA-4 interacts with the DNA binding domain of RXR␣ and that neither the AB domain nor the DEF domains are necessary for this interaction. Furthermore, this interaction is likely to occur on a GATA-RE in the absence of direct DNA binding of RXR␣ because a critical DNA interaction site would likely be obscured by the interaction with GATA-4.
To map the RXR␣ interaction regions within GATA-4, GST pull-down assays were performed using a series of GATA-4 truncation and point mutants. A GATA-4 carboxyl-terminal deletion mutant lacking the ZF1 and ZF2 regions did not bind RXR (Fig. 5B). A truncation mutant containing only the DNA binding domain of GATA-4-(204 -300) bound GST-RXR, but to a lesser extent than the wild-type GATA-4 protein. A robust interaction was noted between the GATA-4 mutant containing the COOH-terminal half of the protein including ZF2 but not ZF1 (aa 244 -440). Mutation of ZF2 significantly decreased but did not abolish the interaction, indicating that in addition to ZF2, another region in the COOH-terminal half of GATA-4 likely contributes to the interaction with RXR. These results indicate that RXR␣ requires the ZF2 region of GATA-4 for binding. In addition, a domain(s) in the region carboxylterminal to ZF2 may also participate in the RXR␣/GATA-4 interaction.

Liganded RXR␣ Recruits FOG-2: A Mechanism Whereby Retinoids Repress GATA-4 Transactivation in Cardiomyocytes-
Given the likely participation of a cardiac-specific co-repressor molecule in the cell-type specific modulation of GATA-4 by RXR␣, we assayed the effect of the known cardiac-specific, GATA-4 co-repressor, FOG-2 on RXR␣-mediated repression of GATA-4. Transient transfection assays were performed in cardiomyocytes to assess the effect of FOG-2 and RXR on ANF promoter activity. As expected the addition of RXR␣, in the presence of 9-cis-RA, resulted in repression of Ϫ633ANF.luc (Fig. 6A). The addition of FOG-2, alone, confirmed its known repressive effect on GATA-4-dependent transcription. The combination of RXR␣ and FOG-2, in the presence of 9-cis-RA, resulted in a significantly greater degree of repression than either alone (p ϭ 0.002; Fig. 6A). The cooperativity between RXR␣ and FOG-2 was further evaluated using titrations of both retinoid ligand and FOG-2. We first determined the minimal amount of exogenous RXR␣ necessary to achieve maximal repression by titrating the amount of RXR␣ expression vector. Full repression of Ϫ633ANF.luc was noted with the addition of 350 ng (per well) of pSG5.mRXR␣ at 10 Ϫ8 or 10 Ϫ7 M 9-cis-RA (Fig. 6B). Utilizing these same conditions, we assayed the effect of adding increasing amounts of FOG-2 on ANF promoter activity at two different ligand concentrations. As expected, FOG-2 mediated a dose-dependent repression of Ϫ633ANF.Luc activity in the absence of ligand (Fig. 6C, open bars), consistent with the known role of FOG-2 as a GATA repressor. Addition of retinoid ligand and increasing amounts of FOG-2 expression vector incrementally repressed the activity of Ϫ633ANF.Luc amounts at the lower ligand concentration of 10 Ϫ8 M. Thus, the ligand-dependent repression conferred by RXR␣ is further enhanced by FOG-2, in a dose-dependent manner.
The functional data shown in Fig. 6 strongly suggested that liganded RXR␣ mediates repression of GATA-4 by cooperating directly or indirectly with FOG-2. To distinguish between these possibilities, GST pull-down assays were performed to determine whether RXR␣ and FOG-2 interact. The known ligandenhanced interaction between the coactivator SRC-1 and RXR␣ was included as a positive control for the effect of ligand. As expected, addition of 9-cis-RA or substitution with the constitutively active RXR␣F318A resulted in a stronger RXR␣/SRC-1 interaction than observed with RXR␣ alone (Fig. 7). Pull-down experiments with 35 S-FOG-2 and RXR␣ demonstrated that FOG-2 bound GST-RXR␣ only in the presence of 9-cis-RA. In addition, FOG-2 bound the constitutively active form of RXR␣ (RXR␣F318A) in the absence of ligand. These results reveal a strong, ligand-dependent direct interaction between RXR␣ and FOG-2 and strongly suggest that RXR␣ recruits the co-repressor FOG-2 to mediate deactivation of GATA-4.

DISCUSSION
The importance of retinoid signaling in various stages of cardiac growth and development has been recognized for more than 50 years, yet the specific molecular targets within the heart are largely unknown. We sought to characterize a retinoid signaling/transcription factor interface relevant to cardiogenic gene regulatory programs. To this end, we focused on the transcriptional regulatory region of the ANF gene that has been well characterized. ANF gene expression has been established as a reliable marker of cardiac hypertrophy in rodents and humans (see review by Levin (35)). In this report, we provide evidence that retinoid signaling via RXR␣ converges on GATA-4-mediated transcriptional control of the ANF gene in cardiomyocytes. These results extend the findings of others that RXR antagonizes the cardiac gene hypertrophic growth response.
In previously published studies (36,37), both RXR and vitamin D receptor ligands were shown to suppress basal and endothelin-1-activated expression of the ANF gene in rat atrial myocytes. In addition, retinoid or vitamin D ligands were shown to inhibit ET-1-induced cardiomyocyte hypertrophy (15). In these studies, high concentrations of all-trans-retinoic acid (i.e. enough to allow substantial isomerization of all-trans-RA to 9-cis-RA) were necessary, leading to the conclusion that liganded RXR rather than RAR is the dominant retinoid receptor conferring this repression. Our results are consistent with this conclusion. Additional information was gleaned by the studies of Zhou and colleagues (14) who investigated the effects of retinoids on ANF gene expression and hypertrophic growth on rat neonatal ventricular cardiomyocytes in culture. Hypertrophic growth and ANF expression were inhibited by treatment with all-trans-retinoic acid. Furthermore, a dominant negative retinoic acid receptor mutant (hRAR␤403) abrogated the repression. These latter results suggested that RXR/RAR heterodimers conferred the repressive effect. However, the hRAR␤403 mutant likely alters most if not all RXR signaling pathways because of its potential to heterodimerize and "squelch" RXR. Our data demonstrate a key role for liganded RXR in this pathway. The constitutively active RXR␣ mutant RXR␣F318A confers repression of the ANF promoter in the absence of ligand. Moreover, we found that the RXR ligand 9-cis-RA is significantly more effective than the RAR-selective ligand TTNPB in mediating the repressive effect. In fact, TT-NPB only repressed at concentrations known to also activate RXR (38,39). In addition, repression was not seen with the RXR␣⌬AF-2 mutant, which should maintain trans-repressive properties if the heterodimeric partner was the operative factor (40). Taken together we conclude that liganded RXR␣ is critical for the observed repressive affect on ANF gene expression. These data do not eliminate the possibility that RXR␣ heterodimeric partners such as vitamin D or RAR participate in this pathway to regulate other genes involved in the hypertrophic growth program.
Analysis of the ANF promoter region revealed that it does not contain cognate retinoid receptor response element DNA recognition sites suggesting that RXR␣ alters ANF gene transcription through an indirect mechanism. Our results are consistent with this conclusion. We found that RXR␣ represses ANF gene transcription via direct interaction with GATA-4. Through its DNA binding domain, RXR␣ interacts with the zinc finger region (ZF2) of GATA-4 as well as additional site(s) GST pull-down assays were carried out as above using truncation or point mutants of RXR␣ and GATA-4. A, representative autoradiographs of 35 S-labeled proteins resolved on SDS-PAGE. B, wild-type or mutant GATA-4 proteins were synthesized as above and interactions with GST control protein-bound Sepharose or GST-RXR␣ fusion protein were assayed; representative autoradiographs are presented. Numbers below each autoradiograph represent quantitation (% total input) of bound GATA-4 proteins, using 10% of the input (left) as an internal standard.
within the COOH terminus of the protein. There is significant precedence for interaction of GATA factors with other transcription factors. In cardiac tissues, GATA-4 interacts with Nkx2.5, NFAT, and MEF-2 through ZF2 (25,41,42), and with FOG-2 through ZF1 (43,44). p300 requires other sites in the NH 2 -terminal and carboxyl-terminal regions of GATA for its interaction (45). Interestingly, GATA-4 interacts with RXR␣ in a manner similar to the GATA-SRF interaction, involving a portion of the COOH terminus of GATA-4 in addition to ZF2 (46). This suggests that competition among activating and repressing factors for binding to GATA-4 could modulate its activity. It should also be noted that our results with the mutated ANF promoter suggest that other elements participate in the RXR␣-mediated repression. Nkx2.5 and SRF are candidates for this role although our results suggest that they cannot confer this repression alone.
Several lines of evidence shown here indicate that the repression of GATA-4 conferred by RXR␣ involves an additional corepressor. First, regulation of GATA-4 by RXR␣ occurred in a cardiac myocyte-restricted manner. Second, although the RXR␣-mediated repression of the ANF promoter and a GATA-RE reporter required 9-cis-RA, the RXR␣/GATA-4 interaction is ligand independent. Third, the RXR␣/GATA-4 inter-action does not require the AF-2 region of RXR␣, yet the observed repression effect required an intact AF-2. These results led us to explore the possibility that FOG-2, the known cardiacenriched repressor of GATA-4 was recruited by RXR␣. Our data demonstrate that FOG-2 interacts with RXR␣ in a liganddependent manner. Interestingly, the interaction domains on GATA-4 are distinct for RXR␣ and FOG-2 and, thus, it is possible that GATA-4 can bind both proteins simultaneously. We propose that RXR␣ serves as a high affinity recruiter of FOG-2 to the GATA-4 interface. The ligand dependence of this interaction provides a mechanism for dynamic modulation of the repressive effect in a cell-and developmental stage-specific manner. In addition, these data suggest a novel signaling pathway through which retinoids modulate key cardiac transcription factors.
The precise mechanism by which FOG-2 mediates repression has not been completely characterized. Indeed, depending upon the cell context and the target, FOG-2 is capable of activating or repressing GATA function (44,47). Recently, FOG-2 was shown to interact with the orphan nuclear receptor COUP-TF (48). Given that COUP-TF confers repression to most of its targets via recruitment of co-repressors containing histone deacetylase activity, it is possible that COUP-TF participates in the RXR⅐FOG-2 complex to deactivate GATA-4. In addition, given that RXR binds GATA-4 via the same region (ZF2) as Nkx2.5 and SRF, it is possible that the RXR␣⅐FOG-2 complex displaces these coactivating transcription factors.
The results of previous studies have shown that retinoids antagonize the hypertrophic growth of cardiac myocytes (14,15). The results presented here focused on ANF gene expression, a single, albeit representative component of the cardiac hypertrophic growth program. The link between the RXR␣⅐ATA-4⅐FOG-2 complex and the broad hypertrophic program is suggested but not definitively established by this work. However, it is tempting to speculate that the RXR␣-mediated mechanism described here serves to block multiple pathways involved in the hypertrophic response. It will also be of interest to determine whether additional nuclear receptor or non-nuclear receptor RXR interacting proteins are involved in the proposed retinoid-mediated effects on the cardiac hypertrophic growth program. GST pull-down assays were performed using GST control, GST-RXR, and GST-RXR F318A proteins with in vitro translated, 35 Slabeled SRC-1 or FOG-2, in the absence or presence of 9-cis-RA (10 Ϫ7 M). SRC-1 is shown as a control for the effect of ligand.