Retinoid X receptor (RXR) agonist-induced antagonism of farnesoid X receptor (FXR) activity due to absence of coactivator recruitment and decreased DNA binding.

The bile salt export pump (BSEP) plays an integral role in lipid homeostasis by regulating the canalicular excretion of bile acids. Induction of BSEP gene expression is mediated by the farnesoid X receptor (FXR), which binds as a heterodimer with the retinoid X receptor (RXR) to the FXR response element (FXRE) located upstream of the BSEP gene. RXR ligands mimic several partner ligands and show additive effects upon coadministration. Using real-time quantitative PCR and cotransfection reporter assays, we demonstrate that the RXR agonist LG100268 antagonizes induction of BSEP expression mediated by endogenous and synthetic FXR ligands, CDCA and GW4064, respectively. Moreover, this antagonism is a general feature of RXR agonists and is attributed to a decrease in binding of FXR/RXR heterodimers to the BSEP-FXRE coupled with the inability of RXR agonists to recruit coactivators to FXR/RXR. Our data suggest that FXR/RXR is a conditionally permissive heterodimer and is the first example of RXR ligand-mediated antagonism of FXR activity. Because FXR agonists lower triglyceride levels, our results suggest a novel role for RXR-mediated antagonism of FXR activity in the development of hypertriglyceridemia observed with RXR agonists in rodents and humans.

ligands also activate the PPAR/RXR and liver X receptor/RXR heterodimers, (4,6) referred to as permissive heterodimers. In contrast, RXR agonists fail to activate the thyroid hormone receptor/RXR and RAR/RXR heterodimers in transfection assays (11,12) and are therefore known as nonpermissive heterodimers. Although RXR ligands alone do not activate RAR/RXR, the combination of RAR and RXR ligands results in activity greater than that seen with RAR ligand alone (11). Synergistic activation is also observed with PPAR and RXR ligands (4,12). Therefore, the pleiotropic in vivo activity of RXR agonists should be evaluated in the presence of the partner's ligand.
Bile acids serve as natural ligands for FXR (7,13,14), which control many aspects of bile acid synthesis and transport. The FXR/RXR heterodimer binds to farnesoid X receptor response elements (FXREs) or bile acid response elements found in the promoter of FXR-responsive genes such as the bile-salt export pump (BSEP), which controls the excretion of bile acids from hepatocytes. The BSEP gene promoter is stimulated by bile acidbound FXR, and transactivation by FXR/RXR is lost when this FXRE is mutated (15). More recently, we and others have shown that CDCA induces BSEP mRNA in HepG2 cells (16,17).
Because RXR is the heterodimeric partner of FXR, we tested whether LG100268, a high affinity, selective RXR agonist (rexinoid) (Refs. 10, 12, and data not shown), would also induce BSEP expression. We demonstrate that RXR agonists alone do not induce BSEP expression, but antagonize ligand-bound FXR-induced expression of BSEP. Furthermore, RXR agonists fail to recruit certain coactivator peptides to FXR/RXR heterodimers. Hence, we propose that FXR/RXR heterodimers function as conditionally permissive heterodimers.
Expression Profiling by Real-time Quantitative PCR-Total RNA was prepared from cultured HepG2 cells (ATCC) using the RNeasy purification system according to the manufacturer's instructions (Qiagen). cDNA synthesis was performed using the Advantage RT-PCR system according to the manufacturer's instructions (Clontech). Real-time quantitative PCR was performed essentially as described (16,19). Briefly, 1 g of total RNA was DNaseI-treated and reverse transcribed using random hexamers and Moloney murine leukemia virus reverse transcriptase. Twenty five ng of each cDNA was added to the Taqman Universal PCR Master Mix along with 900 nM each primer and 200 nM probe according to the manufacturer's instructions (Applied Biosystems). Real-time fluorescence monitoring was performed with the Applied Biosystem 7700. Relative expression levels of the various transcripts were determined essentially as described (19). Values were normalized to 18S rRNA and expressed relative to Me 2 SO (taken as 1) and represent the average of three determinants carried out in duplicate Ϯ S.D. Probes for BSEP were modified at the 5Ј end with 6-carboxyfluorescein (6-FAM) and at the 3Ј end with 6-carboxytetramethylrhodamine (TAMRA) (Biosearch Technologies, Novato, CA); the 18S rRNA probe was modified at the 5Ј end with VIC and the 3Ј end with TAMRA. Primer and probe sequences were designed using Primer Express (Applied Biosystems) and are as follows: BSEP primer, 5Ј-GGTGAGAAAAGAGAGGTTGAAAGG, 5Ј-CCACA-CGAATCCAGTAAAGAATCC; BSEP probe, 5Ј-CCAACGCTGGGCGA-ACACAAGATTT; 18S rRNA primer, 5Ј-CGGCTACCACATCCAAGG-AA, 5Ј-GCTGGAATTACCGCGGCT; 18S rRNA probe, 5Ј-TGCTGGC-ACCAGACTTGCCCTC.
Plasmids and Transfection Assays-Cloning of the 1.5-kb human BSEP promoter by PCR has been described previously (15). Briefly, the forward primer (ccaagcttCCCTTTGACTGCTGTCAATAAC) and the reverse primer (ccaagcttGAGGAAGCCAGAGGAAATAATGG) were used to amplify the 1.5 kb (Ϫ1445 to ϩ86) human BSEP promoter and cloned into pGL3-Basic Vector (Promega) to give BSEP-LUC. The reporter plasmid BSEP-3xFXRE-pTA-LUC was constructed by subcloning three copies for the FXRE identified in the BSEP promoter (15) into the pTA-luc vector (Clontech). Complementary oligonucleotides containing the FXRE sequences were synthesized and cloned into the BglII site upstream of the TATA box using flanking BamHI and BglII sequences engineered in the oligonucleotides.
The full-length open reading frame of the human FXR was amplified by PCR using the published sequence (accession number U68233) and cloned into pCDNA3.1/TOPO (Invitrogen).
HepG2 cells were transfected in 24-well plates with FuGENE reagent according to the manufacturer's instruction (Roche Molecular Biochemicals). For each well, transfection mixes contained 0.25 g of FXRE-pTA-luc report plasmid, 0.05 g of human FXR, 0.05 g of human RXR␣ expression vectors, and 0.25 g of pSV ␤-galactodidase control plasmid (Promega). Twenty-four h later, FXR and RXR ligands or vehicle (Me 2 SO, 0.1% final concentration) were added to cells at the concentration indicated in the figures. After 24 h, cells were lysed and analyzed for luciferase and ␤-galactosidase activities as recommended by the manufacturer (Promega, Madison, WI and Stratagene, La Jolla, CA, respectively).
Fluorescence Resonance Energy Transfer (FRET) assays-Human FXR (amino acids 222-472) or PPAR␣ (amino acids 196 -468) with an N-terminal His tag and RXR (amino acids 225-462, untagged) were expressed in Escherichia coli. The cells were disrupted by sonication and FXR and RXR or PPAR␣ and RXR pastes mixed and the heterodimers formed purified by nickel column chromatography. Amino acid analysis indicated the complexes were a 1:1 heterodimer of FXR/RXR or PPAR␣/RXR. The FRET assay (20) was performed by incubating 7.5 nM FXR/RXR with 0.84 nM Europium-labeled anti-His antibody, 0.055 M streptavidin-conjugated allophycocyanin, and 100 nM biotinylated-SRC1 or CBP peptide (SynPEP) in 50 mM Tris (pH 8.0), 50 M KCl, 0.15 mg/ml bovine serum albumin, 10 M EDTA, and 1 mM dithiothretiol, in the presence of ligands (as indicated in the figure legends) for 4 h at room temperature. Data are expressed as the ratio of the emission intensity at 665 nm to that at 620 nm. The FRET method for PPAR␣ and the sequences for the coactivator peptides have been published (21).
Electromobility-shift Assays-In vitro-translated FXR and RXR were made using the TNT Quick Coupled transcription/translation system (Promega). Electromobility-shift assay was performed as described (22) with in vitro-translated FXR, RXR, or the combination (0.3 l of each per reaction) as indicated. Total lysate was kept constant by the addition of unprogrammed reticulocyte lysate. Ligands were added from 100ϫ stock solutions in Me 2 SO and incubated for 10 min on ice. 32 Plabeled probe was then added and incubated at room temperature for 20 min and the complex resolved on 6% polyacrylamide gels. The sequence of the oligonucleotide used is 5Ј-GATCCTGCCCTTAGGGACATTGAT-CCTTAGGCAAAA. The intensities of the protein-DNA complexes were quantitated using a PhosphorImager (Molecular Dynamics).

RESULTS
LG100268 Antagonizes BSEP Expression by FXR Ligands-Because RXR is the heterodimeric partner of FXR, we investigated whether the selective RXR agonist, LG100268, (10,12) would induce BSEP mRNA. HepG2 cells were treated with CDCA, GW4064, and LG100268 alone or in combination, and BSEP mRNA levels were analyzed by real-time quantitative PCR. Robust induction of BSEP mRNA was observed in HepG2 cells by CDCA, a natural bile acid ligand and GW4064, a synthetic, high affinity ligand for FXR (30-fold and 80-fold, respectively) (23) (Fig. 1A). Treatment with LG100268 induced BSEP mRNA very poorly (2-fold), but surprisingly, strongly inhibited CDCA-and GW4064-mediated induction by 80% (p Ͻ 0.0005) and 83% (p Ͻ 0.00005), respectively. Time course analysis of LG100268-mediated antagonism of FXR activity shows a significant effect (p Ͻ 0.05) as soon as 8 h after treatment, suggesting an immediate transcriptional effect (Fig. 1B).
To determine whether this antagonism is a general property of RXR ligands we tested 9-cis retinoic acid, a natural ligand for RXR (8). In addition, to address ligand selectivity of this effect, we utilized another synthetic ligand, LG100641 (LG641), which although structurally similar to LG268, binds very poorly to RXR (18). As with LG268, very poor induction of BSEP mRNA (less than 2-fold) was observed with 9-cis retinoic acid and LG641 (Fig. 1C). However, both LG268 and 9-cis retinoic acid antagonized GW4064-mediated induction. (82 and 65%, respectively) whereas LG641 inhibited weakly (32%) (Fig.  1C). This is the first example of RXR ligand-mediated antagonism of FXR activity.
RXR-mediated Antagonism Occurs via the BSEP-FXRE-To determine whether RXR agonists also antagonized FXR activation of the natural BSEP promoter, we cloned a 1445-base pair fragment of the BSEP promoter into a luciferase reporter. HepG2 cells were transfected with this reporter and treated with different ligands (Fig. 2A). In the presence of GW4064. we observed 5-fold induction from the BSEP promoter, but no effect with LG100268, 9-cis retinoic acid, or LG641. Coadministration of LG100268 or 9-cis retinoic acid with GW4064 antagonized FXR-mediated activation by 67% and 63%, respectively (p Ͻ 0.05, Fig. 2A) whereas LG641 antagonized less than 10%. Hence, RXR agonist-mediated antagonism of FXR-induced BSEP promoter activity is observed with endogenous levels of FXR/RXR heterodimers corroborating our real-time PCR data (Fig. 1).
An inverted repeat (IR-1) element has been identified in the BSEP promoter and shown to be a functional FXRE (15). In a cotransfection assay with expression plasmids for FXR, RXR␣, and a reporter containing three tandem copies of the FXRE from the BSEP gene promoter, CDCA and GW4064 induced FXR/RXR-mediated transcription from the BSEP-FXRE (9-fold and 28-fold, respectively) (Fig. 2B). In contrast, LG100268 weakly activated FXR/RXR (2-fold) and antagonized activation by CDCA and GW4064.
LG100268 and 9-cis retinoic acid antagonized FXR-mediated activation in a dose-dependent manner with an IC 50 of 6 nM and 300 nM, respectively (Fig. 2C). FXR-mediated induction was also antagonized by LGD1069 (9), another synthetic high affinity RXR ligand (data not shown). No antagonism was observed with LG641 (Fig. 2C). These data indicate that antagonism by RXR ligands is mediated via FXR/ RXR through the FXR response element in the BSEP promoter and is not observed with ligands that do not bind to RXR.
Increased Amounts of RXR Fail to Abrogate Antagonism by LG100268 -We next determined whether LG100268-mediated antagonism was caused by limiting amounts of RXR in HepG2 cells. Transient transfections were performed by increasing the ratio of RXR␣ to FXR expression plasmid from 1 to 5. This caused a slight increase in the signal observed with both ligands compared with GW4064 alone (Fig. 2D). However, even a 5-fold increase in RXR plasmid did not overcome LG100268-mediated antagonism.

RXR Agonists Do Not Recruit Coactivators to FXR/RXR Heterodimers-
To explain the lack of transcriptional activity of FXR/ RXR with LG100268, we determined whether LG100268liganded FXR/RXR recruits coactivators. In a FRET assay, GW4064 induced recruitment of a peptide identical to the receptor-interacting domain of SRC-1 and CBP (Fig. 3A) to the heterodimer but not LG100268. This implies that RXR agonists do not recruit these coactivators to the FXR/RXR heterodimer. To rationalize the mechanism of LG100268-mediated antagonism of FXR activity, we speculated that LG100268 may function as an FXR ligand, displacing GW4064 from the FXR ligand-binding pocket, thereby decreasing coactivator recruitment and activity of FXR. To test this, we coadministered increasing concentrations of LG100268 with a fixed concentration of GW4064 in the presence of FXR/RXR. LG100268 did not alter the signal observed with GW4064 (Fig. 3B). This suggests that LG100268, by binding to RXR, does not induce an allosteric conformation that reduces coactivator recruitment to FXR/RXR. A similar result was also observed with FXR alone (data not shown), indicating that LG100268 is not a ligand for FXR.
These results are in stark contrast to those observed with PPAR␣/RXR (Fig. 3C) where LG100268 induced SRC-1 recruit-ment similar to that seen with GW9578 (24), a potent and selective PPAR␣ agonist. Furthermore, a decrease in the EC 50 and increase in the maximum signal was observed when LG100268 was added in the presence of a fixed concentration of the PPAR␣ agonist. These results suggest that the RXR ligandbinding and coactivator recruitment properties of FXR/RXR are distinct from that of PPAR␣/RXR, a well characterized permissive heterodimer.
LG100268 Modulates Binding of FXR/RXR Heterodimers to the BSEP-FXRE-Previous reports have demonstrated that FXR/RXR heterodimers bind to the BSEP-FXRE in vitro (15). To determine whether LG100268 modulated FXR/RXR binding to DNA, we performed an electromobility-shift assay with in vitro-translated FXR and RXR protein and a radiolabeled oligonucleotide containing the FXRE sequence from the BSEP gene. No protein-DNA complex was observed with FXR or RXR alone in the presence or absence of their respective ligands, (Fig. 4A, lanes 1-4). However, a distinct complex was seen with the FXR and RXR in the same reaction (lane 5). This DNAprotein complex formation is enhanced in the presence of GW4064 (lane 6) but not with LG100268 (lane 7). When both ligands are present, LG100268 decreased the ability of GW4064 to enhance binding to the FXRE by 47% (p Ͻ 0.05,   Figs. 4A, lane 8 and 4B). Both LG268 and 9-cis retinoic acid decreased GW4064-bound FXR/RXR complexes to the BSEPresponse element (Fig. 5) in a dose-dependent manner. This effect was much reduced with LG641. Therefore, RXR ligands decrease the binding of FXR ligand-bound FXR/RXR heterodimers to the BSEP-FXRE. regulators in cholesterol/bile acid metabolism. In this report, we have demonstrated RXR ligand-mediated antagonism in a gene-and response element-specific manner. This antagonism in not observed with related compounds that do not bind to RXR. In contrast to our data involving antagonism of BSEP gene expression via the BSEP-FXRE, LG100268-mediated activation has been observed with the FXRE from the ultraspiracle (usp) gene (7,25) which was additive to an FXR agonist. Whether this translates to induction of usp mRNA is not known. These contrasting results may be due in part to sequence-specific differences between the BSEP and usp FXRE. Hence, the same ligand and receptor may induce one gene while repressing another, a phenomenon reported with another RXR partner, the vitamin D receptor. RXR ligands have been shown to activate the 25-hydroxyvitamin D 3 -24-hydroxylase promoter and to synergize with 1,25-dihydroxyvitamin D 3 (26), whereas RXR ligands inhibit 1,25-dihydroxyvitamin D 3dependent induction of the rat osteocalcin gene (27). We now demonstrate the complexity underlying RXR-permissiveness extends to the FXR/RXR heterodimer which, to date, has been described as a permissive heterodimer (28). Our data provide evidence that this may not be true in all cases and we therefore describe FXR/RXR as a conditionally permissive heterodimer.
One possible mechanism of RXR-mediated antagonism is by induction of the small heterodimerization partner (29 -31), which could dimerize with FXR forming an inactive transcription complex. However, the small heterodimerization partner was not induced in HepG2 cells with LG100268 (data not shown). Our results could also be reasonably explained by RXR agonist-induced formation of RXR homodimers (32), which do not bind to IR-1 FXREs but precludes formation of FXR/RXR heterodimers, as has been postulated for vitamin D receptor (33). Although possible, this may not be the predominant explanation because a similar magnitude of LG100268-mediated antagonism of FXR was observed with 5-fold excess of transfected RXR (Fig. 2D). Our data are also consistent with LG100268-induced dimerization of RXRs with other heterodimeric partners like PPARs in vivo, thus reducing the availability of RXR for dimerization with FXR. Reduced BSEP expression by antagonizing FXR activity is consistent with the phenotype of FXR knockout mice, which have significantly lower BSEP expression in the liver compared with wild-type controls (34) and therefore indicates a role for RXR and its ligand in bile acid transport and homeostasis.
The FXR agonist CDCA decreases triglyceride levels in humans (35), whereas GW4064 decreases triglyceride levels in rats (23). Conversely, RXR agonists increase triglyceride levels in mice and humans (36,37) which is postulated to occur by a mechanism involving RXR-mediated activation of liver X receptor and induction of SREBP1-c (38,39). Our data suggest that LG100268-mediated antagonism of FXR may also contribute to the hypertriglyceridemia induced by RXR agonists.
This report is the first example of RXR-mediated antagonism of FXR activity by decreasing DNA binding coupled with the inability of RXR agonists to recruit coactivators to the FXR/ RXR heterodimer. FXR/RXR thus falls into a distinct category as a conditionally permissive heterodimer.