Identification of a retinoid/chicken ovalbumin upstream promoter transcription factor response element in the human retinoid X receptor gamma2 gene promoter.

To investigate the mechanisms involved in the transcriptional control of retinoid X receptor (RXR) gene expression, the 5′-flanking region of the human RXRγ2 isoform was characterized. An imperfect hexamer repeat (γ retinoid X response element; γRXRE) with a single nucleotide spacer (GGTTGAaAGGTCA) was identified immediately upstream of the RXRγ2 gene transcription start site. Cotransfection studies in CV-1 cells with expression vectors for the retinoid receptors RXRα and retinoic acid receptor β (RARβ) demonstrated that the γRXRE confers retinoid-mediated transcriptional activation with preferential activation by RXR in the presence of its cognate ligand, 9-cis-retinoic acid (RA). Electrophoretic mobility shift assays demonstrated that RXR homodimer binding to γRXRE is markedly enhanced by 9-cis-RA, whereas RAR·;RXR heterodimer binding is ligand-independent. DNA binding studies and cell cotransfection experiments also demonstrated that the nuclear receptor, chicken ovalbumin upstream promoter transcription factor (COUP-TF), repressed transcription via the γRXRE. Cotransfection experiments revealed that COUP-TF and RXRα compete at the γRXRE to modulate transcription bidirectionally over a wide range. These results demonstrate that the human RXRγ2 gene promoter contains a novel imperfect repeat element capable of mediating RXR-dependent transcriptional autoactivation and COUP-TF-dependent repression.

nuclear receptor superfamily in that they may trans-activate not only as 9-cis-retinoic acid (9-cis-RA) activated homodimers but also as obligate heterodimeric partners for retinoic acid receptor (RAR), thyroid hormone receptor (TR), vitamin D receptor, peroxisome proliferator-activated receptor, and several "orphan" receptors (Refs. 1-6; reviewed in Ref. 7). RXR has thus been described as a "master regulator" of a subset of nuclear receptor signaling pathways.
RXR␣ and RXR␤ exhibit ubiquitous expression patterns during murine development and in adult tissues (8,9). In contrast, RXR␥ expression is restricted both in fetal and adult tissues (8 -10). The mouse RXR␥ gene has two known mRNA isoforms (RXR␥1 and RXR␥2), produced via alternative exon splicing and differential promoter utilization (10). The RXR␥ isoforms exhibit a distinct tissue-restricted expression pattern; RXR␥1 is enriched in neural tissue, whereas RXR␥2 is cardiac enriched (9,10). Both transcripts are relatively abundant in skeletal muscle (9,10). During embryologic development, RXR␥ transcripts are expressed in distinct temporal patterns (8,9,11). The RXR␥ gene is therefore unique among the RXR gene family members in that its expression is spatially and temporally restricted, suggesting the possibility that the function of this nuclear receptor is distinct from the other RXRs. Little is known about the mechanisms involved in the control of RXR gene expression or if cross-signaling occurs between members of the RXR gene family.
As an initial step in the investigation of the transcriptional regulatory mechanisms involved in the restricted pattern of expression of RXR␥ isoforms, we have cloned the 5Ј-flanking region of the human RXR␥2 gene. In this report, we describe a novel autoregulatory retinoid X response element (␥RXRE) located within this promoter and present evidence that this element is capable of conferring transcriptional activation via retinoid pathways and transcriptional repression via the orphan receptor COUP-TF.
representing a fragment colinear with bp 451-669 relative to the published mouse RXR␥1 sequence (10), was generated by primers rx4 (5Ј-ATCAggatccCTTCTGCCATGGGTCCACCCTCA-3Ј) and rx5 (5Ј-TCTGggatccTCCCCACAGATGGCACAGATGTG-3Ј). hRX45 was used as a probe to screen a human genomic EMBL3 phage library (Clontech). Three genomic clones were isolated, two of which were characterized (G1SH10 and G2SH1). Restriction endonuclease and Southern blot analysis confirmed that these genomic clones were overlapping and contained sequences recognized by the hRX45 probe. Additional Southern blot analysis using a polymerase chain reaction-generated mouse RXR␥ exon 1b probe (corresponding to the 5Ј-untranslated region of the RXR␥2 cDNA sequence (10)) identified overlapping genomic DNA restriction fragments containing the putative human homolog of exon 1b. Primer extension analysis using total human heart RNA and 32 P-endlabeled rx20 (5Ј-AATCTGCCCATGCGATCCAGAGTC-3Ј) and rx15 (5Ј-GCCTTTTTTCCAGTGTCATC-3Ј) primers from within the putative human exon 1b mapped the transcription start site to 322 bp upstream of the ATG start codon located in exon 3. To compare the genomic sequence with cDNA sequence, 5Ј-RACE was performed using a human heart 5Ј-RACE-Ready cDNA library (Clontech) and primers from within the putative human exons 3 (rx5) and 1b (rx16, 5Ј-ATCGggatc-cCATGGGCAGATTATTCC-3Ј, rx20, and rx15). Comparison of the DNA sequence of G1SH10 and G2SH1 confirmed (i) there was 100% nucleotide identity in the 5Ј-untranslated region region between the 5Ј-RACE clones and genomic clones G1SH10 and G2SH1, (ii) the human RXR␥ exon 1b was spliced to human exon 3 in a manner identical to that found in mice (10), and (iii) no intervening sequences existed between exon 1b and the transcription start site.
Reporter Plasmids and Eukaryotic Expression Vectors-hRXR␥2.luc.Ϫ1140 was constructed by cloning a BamHI-PstI fragment of the hRXR␥2 5Ј-flanking region from Ϫ1140 bp upstream to ϩ71 bp downstream of the transcription start site into the luciferase reporter plasmid pGL2 basic (Promega). The mutated human RXR␥2 promoter reporter plasmid (hRXR␥2.luc.Ϫ1140.m1) was constructed by replacing the native sequence from Ϫ121 to ϩ71 (an NdeI-PstI fragment) with a polymerase chain reaction-generated fragment containing point mutations in the ␥RXRE (GCTTGAAACGTCA; underlined nucleotides represent substitution mutations, compared with Fig. 1). The construction of pTKLuc has been described (14). ␥RXRE.TKluc and ␥RXREm2.TKluc were each constructed by ligating two copies of the corresponding double-stranded oligonucleotide fragments in sense orientation into the BamHI site of pTKLuc (5Ј-gatccTGGGGTTGAAAGGTCAGATGGAtc-3Ј for ␥RXRE.TKluc and 5Ј-gatccTGGGCTTGAAACGTCAGATGGAtc-3Ј for ␥RXREm2.TKluc (underlined nucleotides represent mutations)). Dideoxy DNA sequencing confirmed the location and orientation of the inserts. The construction of eukaryotic expression vectors for use in cell culture transfection experiments (pCDMRXR␣, pCDMRAR␤, and pC-DMCOUP) have been described (14) and were the generous gifts of Drs. Tod Gulick and David Moore (Harvard University). The murine expression vectors pSG5RXR␣, -␤, and -␥ were generously provided by Dr. Pierre Chambon, Institut de Genetique et de Biologie Moleculaire et Cellulaire, Strasbourg, France.
Mammalian Cell Transfections-Simian CV-1 cells were employed for all transfection experiments. Cells were maintained and transient cotransfections performed as described (15). In brief, transient transfections were performed by the calcium phosphate coprecipitation method in 12-well tissue culture plates (Falcon) with 4 g of reporter construct and 1 g of receptor expression plasmid (as indicated in the figure legends) or an equivalent amount of pCDM without insert (pCDM(Ϫ)). Cells were harvested 48 h after transfection. One g of a Rous sarcoma virus ␤-galactosidase expression vector (RSV␤gal) was included to correct for transfection efficiency with the exception of experiments involving pCDMCOUP. pCDMCOUP was noted to exert a modest repressive effect on the transcriptional activity of the thymidine kinase (TK) promoter. Accordingly, activities were adjusted to the effect of COUP-TF on the TK promoter based on the results of parallel experiments using pTKLuc without insert. In experiments involving the addition of ligand, 9-cis-RA or all-trans-RA was added 48 h prior to harvest, and vehicle was added at the same concentration to control wells. Luciferase activity was measured using the standard luciferin-ATP assay (16), and ␤-galactosidase activity was measured using the Galacto-Light chemiluminescence assay (Tropix) in an Analytical Luminescence Monolight 2010 luminometer.
Electrophoretic Mobility Shift Assays (EMSAs)-EMSAs were performed as described (14,17). The pT 7 lac-RXR␣ and pT 7 lac-myc-COUP-TF bacterial expression vectors (14) were generously provided by Dr. Tod Gulick. The nuclear receptors were overproduced in bacterial cells and partially purified as described previously (14). Antibody su-pershift experiments were performed with monoclonal antibodies to RXR (4RX1D12, directed against the D or E domain of all three murine RXRs; a generous gift of Dr. Pierre Chambon) and a monoclonal antibody directed against an epitope in the c-Myc protein (9E10, Oncogene Science).

Identification of a Retinoid-responsive Element in the Human
RXR␥2 Gene Promoter-A human EMBL phage genomic library was screened with a partial human RXR␥ cDNA probe (hRX45; see "Materials and Methods") encoding a portion of human RXR␥ gene exon 2 and all of exon 3 (nucleotides 451-669 relative to the published mouse RXR␥1 cDNA sequence (10)). The human RXR␥ gene exon 3 was identified by the high degree of cross-species nucleotide identity with the murine exon 3 (Ͼ90%) (10). To determine whether the 5Ј-flanking region of the human RXR␥2 gene was contained within either of two genomic clones, Southern blot analysis was performed with a polymerase chain reaction-generated murine RXR␥2specific cDNA probe containing only the 5Ј-untranslated region sequence encoded by the murine exon 1b (10). A single 4.0kilobase pair BamHI restriction fragment was identified with this exon 1b probe and DNA sequence analysis defined a 250-bp region with over 70% nucleotide identity with the murine RXR␥ exon 1b sequence. 5Ј-RACE clones from a human heart library confirmed that, as in mouse, the human exon 1b sequence is spliced to exon 3 (10). Comparison of the genomic DNA sequence with that of multiple 5Ј-RACE clones revealed 100% nucleotide identity, confirming that no intervening sequences existed between exon 1b and the transcription start site. The 5Ј-RACE sequence data and primer extension analysis with human heart total RNA using two different antisense oligonucleotides from within the human exon 1b sequence (data not shown) localized the transcription start site to 322 bp upstream of the start codon. Of note, in addition to a high degree of identity, the human and mouse exon 1b nucleotide sequences are nearly colinear, diverging by no more than 4 consecutive nucleotides over the entire length of both sequences (data not shown). These results confirmed that the 4-kilobase pair BamHI genomic fragment contained 1.14 kilobase pairs of RXR␥2 gene 5Ј-flanking sequence, the 250 bp of 5Ј-untranslated region sequence encoded by the human homolog of murine RXR␥ exon 1b, and approximately 2.5 kilobase pairs of downstream sequence (Fig. 1).
Analysis of the DNA sequence of the RXR␥2 5Ј-flanking region revealed a putative TATA sequence (TATATTA) at bp Ϫ16 (relative to the transcription start site, ϩ1), numerous potential E boxes (18), and several putative CArG sites (19) (Fig. 1), consistent with the muscle-and cardiac enriched expression of RXR␥2. An imperfect repeat sequence located at Ϫ100 to Ϫ86 conformed to the binding consensus for class II and class III nuclear receptors (Fig. 1). This sequence contains two potential hexamer binding sites separated by a single nucleotide and thus conforms to the direct repeat-1 (DR-1) group of elements known to confer transcriptional regulation by retinoid receptors (20).
To test the possibility that the putative nuclear receptor response element was retinoid-responsive and to characterize the transcriptional activity of the RXR␥2 gene 5Ј-flanking region from Ϫ1140 to ϩ71, transient cell transfection studies were performed with this DNA fragment fused to a luciferase reporter (hRXR␥2.luc.Ϫ1140). A series of cotransfection studies was performed in simian CV-1 cells with eukaryotic expression vectors for human RXR␣ (pCDMRXR␣) and human RAR␤ (pCDMRAR␤) in the presence and absence of the retinoid ligands 9-cis-RA and all-trans-RA. As shown in Fig. 2A, the transcriptional activity of hRXR␥2.luc.Ϫ1140 was minimally increased in the presence of 9-cis-RA or RXR␣ alone but was induced 7-13-fold upon the addition of both 9-cis-RA and RXR␣, indicating that this promoter fragment was activated by RXR␣ in a ligand-dependent manner. In contrast, hRXR␥2.luc.Ϫ1140 transcription was only minimally activated by all-trans-RA or 9-cis-RA in the presence of RXR␣ and RAR␤ ( Fig. 2A). These results suggest that the RXR␥2 promoter is preferentially activated by RXR homodimers rather than RXR⅐RAR heterodimers.
To localize the region of retinoid responsiveness and to determine whether the imperfect repeat sequence located at Ϫ100 bp was indeed an RXR-responsive element, cotransfections were repeated with a 5Ј-deletion series of hRXR␥2.luc constructs. The results of these experiments (data not shown) revealed that the sequences conferring 9-cis-RA-mediated response resided primarily within the fragment flanked by NdeI (Ϫ121 bp) and PstI (ϩ71 bp) sites (see Fig. 1), which contained the putative RXR response element. Cotransfection studies were repeated with a mutated hRXR␥2.luc.Ϫ1140 construct containing cytidine substitutions for the invariant second position guanine within each hexameric half-site of the imperfect repeat sequence (hRXR␥2.luc.Ϫ1140.m1; Fig. 2B). The 9-cis-RA/RXR␣-mediated activation of hRXR␥2.luc.Ϫ1140.m1 was markedly lower (Ͼ75%) than that of hRXR␥2.luc.Ϫ1140, confirming that the imperfect repeat conferred the majority of the 9-cis-RA/RXR-mediated response (Fig. 2B). This retinoid-responsive element is here referred to as the ␥RXRE.
To test whether the ␥RXRE could confer retinoid responsiveness to a heterologous promoter and to examine its transcriptional regulatory properties further, including its potential to interact with other class II and class III nuclear receptors, two copies of the ␥RXRE were cloned upstream of the herpes simplex TK promoter fused to a luciferase reporter (␥RXRE.TKluc). Cotransfection studies showed that ␥RXRE.TKluc was activated 8 -10-fold by 9-cis-RA in the presence of RXR␣ (Fig.  3A). Significant RXR-mediated activation of ␥RXRE.TKluc occurred only in the presence of its ligand, 9-cis-RA, as was observed with the homologous promoter (hRXR␥2.luc.Ϫ1140). When point mutations identical to those present in hRXR␥2.luc.Ϫ1140.m1 were introduced into both copies of the ␥RXRE in the context of TKluc (␥RXREm2.TKluc), 9-cis-RAmediated responsiveness was abolished (Fig. 3A). The cotransfection experiments were repeated with expression vectors for murine RXR␣, -␤, and -␥ to determine whether the ␥RXRE was capable of conferring 9-cis-RA-mediated transcriptional activation via all known RXRs. All three RXRs mediated 9-cis-RAdependent activation to a similar level (data not shown).
Previous studies have demonstrated that RXR⅐RAR heterodimers may confer transcriptional activation via DR-1 elements in the presence of either 9-cis-RA or all-trans-RA (21,22). In contrast to RXR homodimer-mediated activation, the ligand-mediated activation of RXR⅐RAR heterodimers on a DR-1 element occurs mainly or solely via RAR (23,24). Accordingly, RXR⅐RAR heterodimers may function as transcriptional inhibitors of RXR homodimer activation on DR-1 elements. In fact, the transfection studies shown above ( Fig. 2A) revealed that, in the context of the homologous promoter, RXR␣-mediated activation of ␥RXRE was reduced by the presence of RAR␤. To explore the activation of ␥RXRE in the context of a heterologous promoter, cotransfection studies were performed with ␥RXRE.TKluc and pCDMRXR␣ and/or pCDMRAR␤ in the presence and absence of either 9-cis-RA (a potential ligand for either RXR or RAR) or all-trans-RA (an RAR ligand) (Fig. 3B). Cotransfection of pCDMRAR␤ alone or pCDMRAR␤ plus pCD-MRXR␣ in the absence of ligand did not significantly alter the transcriptional activity of ␥RXRE.TKluc. Activation of ␥RXRE.TKluc by either all-trans RA or 9-cis-RA in the presence of both RXR␣ and RAR␤ was lower than the induction obtained with RXR␣ alone in the presence of 9-cis-RA (mean of 4 -5-fold versus 8-fold, respectively). These results and the retinoid-mediated activation studies of the homologous promoter ( Fig. 2A) indicate that the ␥RXRE is preferentially activated by RXR and its cognate ligand 9-cis-RA. The retinoid-mediated transcriptional regulatory properties of the ␥RXRE is similar to that of other DR-1 elements such as the cellular retinol-binding protein II gene RXRE (23,25) in which cotransfection of RAR␤ blunts 9-cis-RA mediated transactivation by RXR␣. Additional experiments demonstrated that several other known RXR partners, including TR␣1, TR␤1, or peroxisome proliferator-activated receptor ␣, had no effect on ␥RXRE.TKluc transcriptional activity in the presence of appropriate hormone ligands (thyroid hormone) or peroxisome proliferator-activated receptor activators (fatty acids or clofibrate) with or without RXR (data not shown).

The ␥RXRE Is Bound by RXR␣ Homodimers in a Ligand-dependent Manner and by RXR⅐RAR Heterodimers in a Ligandindependent
Manner-To characterize the interaction of RXR with the ␥RXRE, EMSAs were performed with a 32 P-radiolabeled ␥RXRE oligonucleotide probe and bacterially overexpressed, partially purified RXR␣. As shown in Fig. 4A, RXR␣ homodimers bound the ␥RXRE as a single complex with an affinity that was significantly increased by the addition of 9-cis-RA (Fig. 4A, lane 3) compared with vehicle (Fig. 4A, lane  2). The specificity of the RXR homodimer-␥RXRE interaction was demonstrated by competition studies showing complete inhibition of formation of the complex by the addition of a 100-fold molar excess of unlabeled ␥RXRE but no reduction in complex formation with an equivalent molar amount of unlabeled, unrelated double-stranded oligonucleotide (Fig. 4A;  lanes 3-5). In addition, the specific complex was "supershifted" with anti-RXR antisera (Fig. 4A; lanes 6 and 7). Finally, when a mutated ␥RXRE probe, containing the same point mutations previously shown to abolish functional activity, was incubated with RXR␣ and 9-cis RA, no complex formed (Fig. 4A, lanes 8  and 9). These data confirm that RXR␣ homodimers can interact directly and specifically with the ␥RXRE and that ligand increases binding affinity.
To characterize RXR⅐RAR heterodimer binding to the ␥RXRE, EMSA was performed with bacterially overexpressed RAR␤ and RXR␣ (Fig. 4B). A minimal complex was observed when either RAR or RXR alone was added to ␥RXRE probe in the absence of ligand. In contrast, incubation of the probe with both receptors resulted in a marked increase in complex formation. Competition experiments confirmed that this interaction was specific ( Fig. 4B; lanes 4 -6). Accordingly, RXR⅐RAR heterodimers bind the ␥RXRE in a cooperative manner. In contrast to the interaction of ␥RXRE with RXR homodimers, the RXR⅐RAR-␥RXRE interaction was not influenced by the addition of 9-cis-RA or all-trans-RA (data not shown).
RXR␣ and the Orphan Receptor COUP-TF Compete on the ␥RXRE-A significant body of evidence indicates that RXR and the known orphan nuclear receptor COUP-TF often compete at a single DR-1-type element (26 -31). To examine the potential binding of COUP-TF to the ␥RXRE, EMSAs were performed using COUP-TF tagged with an NH 2 -terminal Myc peptide overproduced in bacteria (COUP-TFMyc). COUP-TFMyc formed a specific complex with the ␥RXRE, as demonstrated by competition studies (Fig. 5). "Supershift" experiments with an anti-Myc antibody provided additional evidence for the specificity of the COUP-TF-␥RXRE interaction. Thus, COUP-TF binds the ␥RXRE with high affinity.
Cotransfection mixing experiments were performed with ␥RXRE.TKluc, pCDMRXR␣, and pCDMCOUP to determine whether these transcription factors could compete at the ␥RXRE to modulate transcription (Fig. 6). For these experiments, increasing amounts of pCDMCOUP were transfected into CV-1 cells with a fixed amount of pCDMRXR␣ in the presence of 9-cis-RA. Because parallel experiments with pTK-Luc alone demonstrated a modest repressive effect of COUP-TF on TK transcription, all data presented for ␥RXRE.TKluc have been corrected for the effect on the TK promoter. COUP-TF blunted the 9-cis-RA-mediated RXR activation via the ␥RXRE in a dose-dependent fashion (Fig. 6). With the highest amounts of pCDMCOUP transfected, transcription of ␥RXRE.TKluc was repressed below basal levels, indicating that, in addition to competing with RXR, at higher levels COUP-TF actively represses transcription via the ␥RXRE, a property shown for most known COUP-TF response elements (14, 26 -37). The transcriptional activity of ␥RXRE.TKluc varied over 50-fold in these cotransfection experiments. These results, together with the binding studies, indicate that COUP-TF modulates retinoid-mediated activation of ␥RXRE and suggest a mechanism whereby transcriptional activity can be modulated over a wide range. DISCUSSION This report demonstrates that the human RXR␥2 gene promoter contains a RXR response element, a mechanism for the regulation of RXR␥2 gene expression by retinoid-mediated pathways. The presence of autoregulatory elements within the promoters of genes encoding other nuclear receptor isoforms, including RAR␤2, RAR␣2, RAR␥2, and TR␤1, suggests that the expression of a subset of nuclear receptor genes are controlled by this mechanism (38 -43).
The transcriptional regulatory properties of ␥RXRE are similar to those of previously reported DR-1-type retinoid response elements (21, 23, 25, 29 -31, 44 -46). A comparison of the ␥RXRE sequence with the relatively few known natural DR-1 RXREs is shown in Fig. 7. Although the 5Ј-half-site sequence of the ␥RXRE is novel compared with other known elements, it conforms to the known consensus (PuG(G/T)TNA) for binding class II or class III nuclear receptors (reviewed in Ref. 7). Furthermore, the ␥RXRE sequence, including the extended heptamer of the 5Ј-half-site (GGGTTGA) resembles the RXR␣ binding site (GGGGTCAaAGGTCA) and the high affinity RXR␥ consensus (RGRNCAaAGGTCA) determined by nonbiased random oligonucleotide selection (47)(48)(49). Interestingly, comparison of the sequences shown in Fig. 7 reveals that the 5Ј-half-site sequence often diverges from the idealized class II/III sequence (AGGTCA). To our knowledge, the role, if any, of such sequence differences in dictating transactivation proper- ties of RXR isoforms has not been established.
We show here that the interaction of ␥RXRE with RXR homodimers but not RXR⅐RAR heterodimers is induced by the ligand 9-cis-RA, a unique property. Although others have shown ligand-dependent binding of RXR homodimers to DR-1 elements with receptor produced in reticulocyte lysate expression systems, to our knowledge this is the first example of ligand-induced binding with receptor protein produced in bacteria. In fact, several other groups have shown that RXR homodimer (produced in bacterial expression systems) binding to perfect DR-1 elements is ligand-independent (25,26,48). We have also shown that binding of RXR␣ (produced in our bacterial expression system) to an idealized DR-1 (AGGTCAaAG-GTCA) is not dependent on or induced by 9-cis-RA. 2 Taken together, these results suggest that the unique sequence of the ␥RXRE dictates the relative affinity by which RXR homodimers bind this element and thus require ligand for this interaction.
Our results also demonstrate that the orphan receptor COUP-TF competes with RXR homodimers on the ␥RXRE to repress transcription. This finding is consistent with the known role of COUP-TF as a negative modulator of RXRmediated transcriptional regulatory pathways through competition for DNA binding. This competitive interaction has been demonstrated for a variety of natural and synthetic retinoidresponsive elements, including DR-1 (26, 27, 29 -31) and DR-5 elements (27,33) as well as complex retinoid-responsive elements (14,27,28,33,36,50). Our data also indicate that, in addition to interference with RXR homodimer binding to the ␥RXRE, COUP-TF represses transcription via this element as it does on the majority of other COUP-TF response elements.
A major unanswered question in nuclear receptor biology involves the specific biological roles of multiple RXR and RAR isoforms generated by differential promoter utilization and/or alternative splicing. The RXR␥2 isoform is an excellent focus for the study of the function of nuclear receptor isoforms because of its tissue-and developmental stage-restricted expression pattern (8 -10). The recent characterization of mice homozygous for targeted ablation of retinoid receptors demonstrates the importance of retinoids in murine cardiac development (51)(52)(53)(54)(55). In addition, recent studies by us and others suggest that retinoids play a role in the control of postnatal cardiac energy metabolism (56) and antagonize the cardiac hypertrophy program (57). The known cardiac enriched expression of RXR␥2 and our identification of the ␥RXRE raises the intriguing possibility that the cardiac specific effects of retinoids occur via retinoid signaling pathways that converge on this gene. Given that RXR␣ and RXR␤ are expressed prior to RXR␥ in the developing heart and somites, it follows that the RXR␥2 gene promoter could be a downstream target during embryologic development. Cotransfection studies performed in our laboratory indicate that the ␥RXRE is activated similarly by RXR␣, RXR␤, or RXR␥ (data not shown), suggesting that the human RXR␥2 promoter is a potential target for any of the known RXRs.