Identification of Gene-selective Modulators of the Bile Acid Receptor FXR*

BAR is a nuclear bile acid receptor (BAR) (FXR) receptor that regulates gene networks involved in cholesterol and bile acid homeostasis. We have identified two classes of synthetic compounds that differentially modulate BAR activity. The first class activates BAR target genes in the predicted fashion and is 25-fold more potent than endogenous bile acids. The second class, represented by AGN34, antagonizes BAR in transient reporter assays. Surprisingly, this compound acts in a gene-selective manner in vivo: it is an agonist on CYP7A1, an antagonist on IBABP , and is neutral on SHP. These findings indicate that synthetic BAR modulators can be developed to regulate transcription in a gene-specific fashion. Given the ability of BAR to regulate several lipid homeostatic pathways, the identification of gene-selective BAR modulators have important implications for the development of improved cholesterol lowering agents.

BAR is a nuclear bile acid receptor (BAR) (FXR) receptor that regulates gene networks involved in cholesterol and bile acid homeostasis. We have identified two classes of synthetic compounds that differentially modulate BAR activity. The first class activates BAR target genes in the predicted fashion and is 25-fold more potent than endogenous bile acids. The second class, represented by AGN34, antagonizes BAR in transient reporter assays. Surprisingly, this compound acts in a gene-selective manner in vivo: it is an agonist on CYP7A1, an antagonist on IBABP, and is neutral on SHP. These findings indicate that synthetic BAR modulators can be developed to regulate transcription in a gene-specific fashion. Given the ability of BAR to regulate several lipid homeostatic pathways, the identification of geneselective BAR modulators have important implications for the development of improved cholesterol lowering agents.
Cholesterol is a multifunctional molecule that is essential for a broad array of physiologic processes including membrane biogenesis, caveolae formation, and the distribution of embryonic signaling molecules. It is also as an essential precursor in the synthesis of transcriptionally active lipids including the steroid hormones and oxysterols (1). Although essential, cholesterol is highly insoluble and can form deposits that contribute to a variety of diseases including gallstones and heart disease (2,3). Indeed, excess circulating cholesterol is a major risk for atherosclerotic heart disease (3,4). This disease is responsible for nearly 500,000 deaths each year in the United States (42) and is the single largest cause of mortality in industrialized nations. It has been estimated that 10% of the U.S. population would benefit from cholesterol-lowering therapies (5). This has prompted an intense search for therapeutic agents that specifically modulate cholesterol homeostasis.
Cholesterol levels are controlled at a variety of levels including intestinal uptake, endogenous biosynthesis, transport, and elimination. The major pathway for cholesterol elimination is via hepatic conversion of cholesterol into water-soluble bile acids (6) and their subsequent secretion into the gastrointestinal tract. Approximately 95% of the secreted bile acids are recycled via intestinal uptake and are returned to the liver through the portal blood. The remaining 5% of bile acids are eliminated from the gut thereby forcing the liver to replenish these losses by converting as much as 0.5 g of cholesterol to bile acids each day (7). The liver therefore has an enormous capacity to metabolize cholesterol and therapies that target this process have the potential to eliminate cholesterol derived from a variety of sources including diet, synthesis, and atherosclerotic lesions (via the reverse cholesterol transport pathway).
Two metabolic pathways have been identified that convert cholesterol to bile acids (6). In humans, the classic pathway is responsible for at least 90% of all bile acid synthesis. The first and rate-limiting step in this pathway is catalyzed by CYP7A1, 1 a liver-specific cholesterol 7␣-hydroxylase. CYP7A1 transcription is strongly repressed by its bile acid end products (8). An important advance in understanding this feedback loop came with the identification of a member of the nuclear receptor superfamily (FXR, NR1H4, hereafter referred to as BAR) that suppresses CYP7A1 transcription in response to endogenous bile acids (9 -12). Two bile acid response elements (BAREs) have been identified in the CYP7A1 promoter. However, BAR is unable to bind directly to either element, suggesting an indirect role for BAR in the regulation of CYP7A1 (13). A mechanism has been proposed whereby BAR induces the negative transcriptional regulator SHP (small heterodimer partner), which in turn represses transcription factors that bind to the CYP7A1 BAREs (14,15). This mechanism for CYP7A1 repression was suggested based on experiments using transiently overexpressed SHP. Because SHP can repress (16,17) and/or activate (18) numerous nuclear receptors under these conditions, the SHP-induction model does not account for the specificity by which bile acids regulate gene transcription.
Although the mechanisms underlying transrepression by BAR is unclear, it is well known that BAR activates transcription by binding to specific response elements (19,20) as a heterodimer with the nuclear receptor RXR. Several genes have been identified whose transcription is activated by BAR including SHP, the ileal bile acid-binding protein (IBABP), and the hepatic bile salt export pump (BSEP, ABCB11) (21). These genes are critical for bile acid homeostasis. IBABP is an intra-cellular protein expressed in the distal ileum where the majority of bile acids are reabsorbed. It has been proposed that IBABP plays a role in transcellular shuttling and/or buffering the high and otherwise toxic levels of bile acids that pass through this tissue. BSEP is a canalicular ATP binding cassette transporter that is responsible for biliary secretion of bile acids. Indeed, inactivating mutations of this gene result in progressive familial intrahepatic cholestasis (type 2) and hepatic cirrhosis (22). Thus, in addition to regulating cholesterol degradation, BAR plays a more general role in coordinately regulating bile acid physiology.
BAR also controls other aspects of lipid homeostasis. For example, BAR agonists reduce triglyceride levels (23,24) and BAR-null mice have elevated triglycerides (12). This is potentially related to BAR-mediated regulation of apolipoprotein CII and/or phospholipid transfer protein (reviewed in Ref. 21). Regardless of the mechanism, it appears that BAR activation promotes reciprocal effects on cholesterol and triglyceride levels.
Given its critical role in repressing CYP7A1-mediated cholesterol degradation, BAR provides an attractive target for the development of novel cholesterol lowering agents. However, the effects of BAR on other target genes implies that a generalized antagonist would produce undesirable effects including elevations in triglycerides, a lowering of biliary bile acid transport, and/or cholestasis. Therefore, the most desirable therapeutic agents would be gene-selective modulators that selectively regulate a subset of BAR-specific genes. In a search for such compounds we have identified two classes of BAR modulators. The first class include agonists that are ϳ25-fold more potent than naturally occurring bile acids. These compounds activate BAR and produce the expected regulation pattern on endogenous target genes. Second, we have identified AGN34 as a gene-selective BAR modulator (BARM): it acts as an agonist on CYP7A1, an antagonist on IBABP, and is neutral on SHP. These data demonstrate that SHP induction is not required to repress CYP7A1. More significantly, we provide evidence for gene-selective BARMs, a finding with important implications for the treatment and prevention of atherosclerotic heart disease.
Transient Transfection Assay-CV-1 cells were grown and transiently transfected as described (9). The data presented were obtained with mouse BAR but human and rat BAR exhibited qualitatively similar results.
Northern Analysis-HepG2 cells were maintained in Eagle's minimal essential medium with 10% fetal bovine serum, 1 mM sodium pyruvate, 2 mM L-glutamine, and nonessential amino acids. Caco-2 cells were maintained in Dulbecco's modified Eagle's medium with 20% fetal bovine serum. Caco-2 cells were maintained for 20 days post-confluence to allow differentiation. One day prior to treatment, confluent HepG2 and differentiated Caco-2 cells were switched to phenol red-free media containing resin-charcoal-stripped fetal bovine serum and treated for an additional 24 h with the indicated compounds. Northern blots were prepared from poly(A) ؉ RNA and analyzed with the following probes: human CYP7A1 nucleotides 1617-2576 (accession number M93133), human SHP nucleotides 888 -1355 (accession number L76571), human IBABP coding sequence (accession number AI311734), and human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) nucleotides 105-331 (accession number NM_002046). The RNA levels were quantitated using a GS-800 calibrated densitometer and Quantity One software (Bio-Rad).
DNA-based Coactivator Recruitment Assay-This assay was performed as previously described (9). An optimized DR-1 was used for RXR homodimers and the mouse IBABP IR-1 response element (10) was used for BAR⅐RXR heterodimers. The data presented were obtained with rat BAR.
Competitive Ligand Binding Assay-RXR␣ proteins were produced using a baculovirus expression system and binding assays were performed as described previously (28).
Analysis of the crystal structure of retinoid-bound RAR indicates that the methyl group in the central isoprene unit (Fig.  1A, TTNPB, shaded area) interacts with critical residues in helix 5 of the RAR ligand binding domain (31). We hypothesized that placement of bulky residues in this position would create a steric hindrance to RAR binding. At the same time, compounds with a bulky moiety in this location might retain activity on BAR as this structure would more closely approximate the cyclohexane ring at the corresponding position of the BAR ligand chenodeoxycholic acid (CDCA) (Fig. 1A). To test this hypothesis, AGN29 and AGN31 were synthesized with trimethylsilyl and n-butyl groups in place of the methyl moiety on the central isoprene unit of TTNPB (Fig. 1A, shaded area). At concentrations of 5 M, AGN29 (91-fold) and AGN31 (85fold) were potent and efficacious activators of the BAR⅐RXR heterodimer ( Fig. 1B). By comparison, TTNPB was less efficacious in activating BAR (65-fold) and CDCA, an endogenous BAR ligand, required 20-fold higher concentrations (100 M) for optimal activity (Fig. 1B). As predicted, AGN29 and AGN31 were both dramatically less effective than TTNPB in activating RAR ( Fig. 1C) and had no effect on other nuclear receptors including AR, mouse, and human PXR, ER␣, CAR, LXR␣, PPAR␣, PPAR␥, PPAR␦, VDR, and T 3 R␤ (data not shown). Thus, unlike the parent ligand TTNPB, AGN29 and AGN31 exhibit significant selectivity for BAR.
RXR-specific ligands such as LG268 are known to activate the BAR⅐RXR heterodimer (Fig. 1B), however, we have previously established that this occurs via the RXR subunit and not via BAR (9). Because TTNPB derivatives were originally used to identify RXR-specific ligands (30), we tested the ability of AGN29 and AGN31 to activate RXR. Both compounds activated RXR at the same concentrations required to activate BAR⅐RXR heterodimers (5 M) and were inactive at lower doses (data not shown). Although AGN29 and AGN31 activate RXR, their action is distinct from that of LG268. For example, when compared with LG268, AGN29 and AGN31 preferentially activate BAR⅐RXR heterodimers whereas LG268 preferentially activates RXR (compare Fig. 1, B and D).
To further explore the molecular properties of AGN29 and AGN31 we examined their effects on coactivator recruitment in vitro. The primary function of nuclear receptor ligands is to induce a conformation change that facilitates recruitment of transcriptional coactivator proteins via the AF2 transactivation domain (32). Previous coactivator recruitment assays have demonstrated that bile acids and TTNPB bind directly to BAR (9,11). We used an electrophoretic mobility shift assay (9) to determine whether AGN29 and AGN31 also serve as BAR ligands. Thus, BAR⅐RXR heterodimers were formed on a 32 Plabeled BAR response element and incubated with a peptide containing the receptor interaction domains of the coactivator GRIP1. As expected, both AGN29 and AGN31 effectively recruited coactivator ( Fig. 2A, lanes 1-3).
LG268 also recruited coactivator ( Fig. 2A, lane 4) but was less effective than AGN29 and AGN31, a finding that parallels the transactivation data ( Fig. 1, B and D). To explore the subunit requirements of these compounds, we asked whether the BAR or RXR AF2 was required for these interactions. Mutation of the BAR AF2 resulted in loss of activity for both AGN29 and AGN31 whereas mutation of RXR AF2 had no effect ( Fig. 2A, lanes 5-7 and 9 -11). In contrast, recruitment by LG268 required the AF2 domain of RXR as well as BAR ( Fig. 2A, lanes 8 and 12). Thus, whereas AGN29 and AGN31 can activate an isolated RXR subunit (Fig. 1D), the results of Fig. 2A demonstrate that BAR is essential for the activity of AGN29 and AGN31 in the context of the BAR⅐RXR heterodimer.
To determine the relative potency of AGN29 and AGN31, transfected cells were treated with increasing doses of AGN29 and AGN31 (Fig. 2B). The half-maximal effective concentrations (EC 50 ) for AGN29 and AGN31 are ϳ2 M compared with Ͼ50 M for CDCA. Similarly, using the coactivator recruitment assay, AGN29 and AGN31 produced a dose-dependent increase in the GRIP1-containing complex (Fig. 2C). The concentration of AGN29 and AGN31 that promoted GRIP recruitment closely paralleled the potency of these compounds in transfection assays (Fig. 2B). These results indicate that AGN29 and AGN31 are not only BAR ligands but are Ͼ25-fold more potent than their endogenous counterparts.
Synthetic Agonists Regulate BAR Target Genes-BAR regulates transcription of specific target genes in the liver (CYP7A1, SHP) and intestine (IBABP, SHP) (reviewed in Ref. 21). This prompted us to test the effect of AGN29 and AGN31 on gene expression in HepG2 hepatoma cells and in differentiated intestinal Caco-2 cells. Both cell lines were treated for 24 h with optimal concentrations of AGN29, AGN31 (10 M), or CDCA (100 M). Like CDCA, AGN29 and AGN31 strongly induced SHP and IBABP expression in Caco-2 cells (Fig. 3A). Similarly, these compounds acted as agonists in HepG2 cells and produced the expected induction of SHP and repression of CYP7A1 (Fig. 3B). AGN29 and AGN31 had no effect on expression of BAR (data not shown) or the GAPDH internal control (Fig. 3). These results demonstrate that AGN29 and AGN31 are agonists on endogenous target genes.
Identification of a BAR Antagonist-We next explored the possibility that retinoid derivatives might serve as BAR antagonists. Indeed, using the transfection assay described above, AGN34 (Fig. 4A, 1 M) was identified as a compound that effectively blocked transcriptional activation by the endogenous CDCA ligand (50 M) (Fig. 4B). AGN34 had no effect on basal reporter activity (Fig. 4B). AGN34 also failed to antagonize the nuclear receptors AR, HNF4␣ CAR, LXR, mouse and human PXR, PPAR␣, PPAR␥, PPAR␦, and VDR and exhibited only minimal activity on ER␣, T 3 R␤, and RAR (Fig. 4C). These data demonstrate that AGN34 is a selective antagonist of BAR activity.
AGN34 Is a Direct and "Trans-antagonist"-Because AGN34 possesses a partial retinoid-like structure, we examined the effect of this compound on the retinoid receptors RAR and RXR. AGN34 was a weak antagonist of RAR and other receptors that utilize RXR as an obligate heterodimeric partner (Fig. 4C). However, AGN34 was an effective antagonist of RXR homodimers (Fig. 4C). This prompted us to compare the activity of AGN34 on BAR⅐RXR and RXR complexes. Dose-response analysis demonstrated that AGN34 antagonized both receptor complexes with a half-maximal inhibitory concentration (IC 50 ) of Ͻ10 nM (Fig. 5, A and C). Thus, the ability to antagonize RXR and BAR at similar doses suggest that AGN34 acts to trans- antagonize bile acid-activated BAR⅐RXR by binding to the RXR subunit.
We used an in vitro radioligand displacement assay to confirm that AGN34 acts directly on the RXR subunit. RXR was incubated with [20-methyl-3 H]9-cis-retinoic acid and binding was measured in the presence of increasing concentrations of unlabeled AGN34. We found that AGN34 associates with RXR with a binding constant of ϳ2 nM (Fig. 5B). The high in vitro affinity of AGN34 for RXR closely matches its IC 50 for both RXR and BAR⅐RXR heterodimers. These findings confirm that AGN34 can trans-antagonize BAR⅐RXR via the RXR subunit.
While AGN34 can function via the RXR subunit, this does not exclude the possibility that it could have additional activities via the BAR subunit. Indeed, examination of the doseresponse curve for BAR⅐RXR (Fig. 5A) indicates that AGN34mediated antagonism is somewhat biphasic. Antagonism is first seen at low doses corresponding to those associated with RXR binding (Յ10 nM). A plateau is seen at intermediate doses (100 -1000 nM) and then a second phase of antagonism is observed at high doses (Ͼ1 M). This pattern implies that AGN34 acts via the RXR subunit at low doses and via the BAR subunit at higher doses.
To further explore the activity of AGN34 on the BAR subunit, we examined the effect of low and high dose AGN34 on agonist-induced coactivator recruitment. Whereas agonist ligands recruit coactivators to nuclear receptors, antagonists occupy the ligand binding pocket but fail to recruit coactivators (33). When agonists and antagonists are mixed, the antagonist competes with the agonist for binding and promotes an apparent displacement of coactivator. As expected, the RXR-specific ligand LG268 induced coactivator recruitment to RXR homodimers (Fig. 5C, top panel, lanes 1 and 2). Addition of AGN34 resulted in a displacement of coactivator from RXR homodimers with an IC 50 of ϳ30 nM (Fig. 5C, top panel, lanes  3-10). In contrast, low doses of AGN34 (Ͻ30 nM) resulted in only a partial displacement of coactivator from BAR (Fig. 5C,  bottom panel, lanes 1-5). Full coactivator displacement and reappearance of the coactivator-free heterodimer required concentrations greater than 300 nM (Fig. 5C, bottom panel, lanes  6 -10). Similar results were seen using a mutated RXR with decreased ligand binding affinity (data not shown) suggesting that AGN34 functions via BAR at these high doses. This biphasic pattern of coactivator displacement in vitro further demonstrates that AGN34 antagonizes BAR⅐RXR via both subunits.
The ability to trans-antagonize BAR⅐RXR with RXR antagonists raises the possibility that previously characterized RXR antagonists such as LG754 (34) might generally serve as BAR⅐RXR antagonists. However, LG754 did not effectively antagonize BAR in transfection assays (Fig. 5D) and only displaced coactivator from agonist-occupied BAR⅐RXR heterodimers at very high concentrations (3 M) (Fig. 5E). Thus unlike previous RXR antagonists, AGN34 is distinct in its ability to trans-antagonize BAR⅐RXR.
A Gene-specific Bile Acid Receptor Modulator-The above data demonstrate that AGN34 is a potent and selective antagonist of BAR that functions via a unique form of trans-antagonism. This prompted us to explore the effect of this compound on a variety of endogenous BAR target genes. Differentiated Caco-2 and HepG2 cells were treated with 100 M CDCA alone or in combination with 1 M AGN34. In Caco-2 cells AGN34 dramatically inhibited bile acid-induced expression of IBABP (Fig. 6A), an effect that was observed with doses of AGN34 as low as 30 nM (data not shown). These results demonstrate that AGN34 is a potent antagonist of endogenous BAR target genes.
To further examine the activity of AGN34, we tested its effect on SHP expression. Although SHP is induced by BAR agonists in both cell lines, the antagonist AGN34 failed to inhibit bile acid-mediated induction of SHP (Fig. 6, A and B). Because AGN34 is a highly effective antagonist of IBABP expression, these findings demonstrate that AGN34 is a gene-specific antagonist. We next tested the effect of AGN34 on CYP7A1, a gene whose transcription is repressed by BAR agonists (Fig.  3B). One would expect that a pure BAR antagonist would relieve bile acid-mediated suppression of CYP7A1 expression. In contrast, AGN34 unexpectedly repressed CYP7A1 by itself and further repressed activity in concert with the CDCA agonist (Fig. 6B). Dose-response experiments with AGN34 further confirmed that AGN34 repressed CYP7A1 by itself and in combination with CDCA (Fig. 6C). The dose response in this assay is consistent with the dose response required to antagonize BAR activity in transient transfection and coactivator recruitment assays (Fig. 5, A and C). These results demonstrate that the activity of AGN34 is dramatically gene-selective: AGN34 acts as an antagonist on IBABP, has no effect on SHP, and is an agonist on CYP7A1. Thus, AGN34 provides the first example of a novel class of BAR ligands that we refer to as gene-selective BAR modulators (BARMs). Moreover, the ability of AGN34 to repress CYP7A1 without inducing SHP demonstrates that BAR can repress CYP7A1 through mechanisms that do not require the induction of SHP. is an RXR ligand. RXR was incubated with 10 nM [20-methyl-3 H]9-cisretinoic acid (9-cis-RA) and increasing concentrations of AGN34. Ligand-bound receptors were isolated with hydroxyapatite resin and the amount of bound radioactive ligand was determined using a Micro-Beta counter. Binding in the absence of AGN34 was set at 100%. C, AGN34 reverses coactivator recruitment. For RXR, coactivator displacement assays were performed by mixing in vitro translated RXR (0.6 l), 75 ng of purified GST-GRIP, and a 32 P-labeled DR-1 probe with agonist alone (25 nM LG268) or agonist with increasing concentrations of AGN34 as indicated (top panel). For BAR, assays were performed by mixing RXR (0.6 l) and BAR (0.6 l), 2 g of purified GST-GRIP, and a 32 P-labeled IBABP IR-1 probe with agonist alone (2 M AGN31) or agonist with increasing concentrations of AGN34 (bottom panel). The receptor dimer ( ‡) and receptorcoactivator complexes (ˆ) were separated by electrophoresis through nondenaturing polyacrylamide gels. D, AGN34 is a more effective antagonist than LG754. Transfections were performed as described in the legend to Fig. 1B and cells were treated with 50 M CDCA alone or in combination with 1 M AGN34 or 1 M LG754. The luciferase reporter activity was normalized to the ␤-galactosidase internal control. E, reversal of coactivator recruitment by LG754. Coactivator displacement assay was performed as in C, except with increasing concentrations of LG754. The receptor dimer ( ‡) and receptorcoactivator complex (ˆ) are indicated. and atherosclerosis. Indeed, adenoviral mediated overexpression of CYP7A1 is sufficient to reduce plasma low density lipoprotein concentrations by ϳ60 -75% (35). The most commonly used cholesterol-lowering drugs are the statin class of cholesterol synthesis inhibitors. These agents are nonspecific in that they inhibit the synthesis of both cholesterol and its biologically active precursors. In addition, inhibitors of cholesterol synthesis cannot eliminate pre-existing cholesterol that arises from dietary or other sources. Therefore, an urgent need exists for additional therapeutic strategies. Indeed, coronary arterial disease remains the leading cause of death in industrialized societies and 36-million Americans require cholesterol-lowering therapies (5).
It has been well established that CYP7A1 transcription is strongly repressed by its bile acid end products. Although bile acid-mediated repression is conserved in a variety of mammalian species, this pathway is particularly sensitive in humans (36). Drugs that antagonize bile acid-mediated repression of CYP7A1 would be particularly useful in stimulating cholesterol elimination in humans. A key advance in elucidating the molecular events underlying CYP7A1 repression was the demonstration that the nuclear receptor BAR suppresses CYP7A1 transcription in response to endogenous bile acids (9 -12). In principle, a BAR-specific antagonist would prevent CYP7A1 repression thereby facilitating further cholesterol catabolism. Whereas BAR has potent effects on CYP7A1, this receptor plays a broader role in regulating lipid homeostasis. For example, BAR activation lowers triglycerides (23,24) and stimulates expression of genes involved in biliary bile acid secretion. Thus, a generalized BAR antagonist has the potential to induce serious side effects including cholestasis and hypertriglyceridemia. An attractive means to bypass these effects would be the identification of antagonists whose activities are limited to a subset of target genes.
We describe two novel BAR agonists: AGN29 and AGN31. These compounds are 25-fold more potent than naturally occurring ligands and resulted in the expected activation (IB-ABP, SHP) or repression (CYP7A1) of BAR-target genes. AGN29 and AGN31 are derived from TTNPB, a synthetic retinoid that is a ligand for both BAR and RAR (11,29,30). In contrast, AGN29 and AGN31 are BAR-selective ligands in that they have weak activity on RAR and fail to activate other nuclear receptors. Structure-activity studies suggest that their weak activity on RAR results from the placement of bulky functional groups on the central isoprene unit. Indeed, analysis of the crystal structure of retinoid-bound RAR indicates that these bulky residues would clash with critical residues in helix 5 of the RAR ligand binding domain (RAR␥ Met 272 ) (31).
Interestingly, AGN29 and AGN31 are unique among BAR ligands in that they also activate RXR, the heterodimeric part-ner required for the formation of an active BAR complex. Thus, AGN29 and AGN31 function through both subunits of the BAR⅐RXR heterodimer. This prompted us to examine whether other retinoid-like compounds might antagonize the BAR complex. Using transient transfection and in vitro coactivator recruitment assays, we identified AGN34 as an extremely potent compound that antagonized BAR⅐RXR at concentrations as low as 10 nM. Antagonism at these low concentrations resulted from direct binding to the RXR subunit and subsequent reversal of agonist-induced coactivator recruitment (Fig. 5C). Because coactivator recruitment requires the AF2 transactivation domain of BAR ( Fig. 2A), the ability to displace coactivator while binding to RXR demonstrates that AGN34 is an allosteric inhibitor of BAR. Although RXR functions as an obligate partner for many receptors, the effect of AGN34 was specific to BAR (Fig. 4C) and other RXR antagonists were not effective at antagonizing BAR (Fig. 5D). This mode of specific "trans-antagonsim" via a partner receptor has not been previously described among the nuclear receptor superfamily.
Although AGN34 antagonizes BAR in transient reporter assays, this compound functions as a gene-selective modulator in vivo: it acts as an agonist on CYP7A1, an antagonist on IBABP, and is neutral on SHP. This divergent pattern of regulation is reminiscent of the activity of selective estrogen receptor (ER) modulators or SERMs (37). These compounds elicit an array of biological effects that are either estrogenic or antiestrogenic depending on the tissue. For example, SERMs such as tamoxifen and raloxifene are used for the prevention and treatment of breast cancer by virtue of their ability to antagonize ER in the breast. In contrast, tamoxifen and raloxifene act as ER agonists in other tissues. These unexpected activating properties have significant advantages as ER agonists have beneficial effects on bone density and plasma lipoprotein levels. By analogy to the SERMs, AGN34 represents a selective BARM. The ability to regulate BAR in a gene-specific fashion suggests that future compounds may be identified that selectively enhance cholesterol elimination without promoting negative effects such as hypertriglyceridemia or cholestasis.
Whereas the divergent activities of SERMs have been appreciated for over 2 decades the molecular mechanisms that underlie their action are only now being elucidated (reviewed in Ref. 38). ER regulates transcription either by binding to specific response elements or indirectly by tethering to other promoter-bound transcription factors. In both cases ER agonists recruit transcriptional coactivators to its targeted promoters. However, in breast cancer cells where corepressor proteins are more highly expressed, tamoxifen act as an antagonist by recruiting corepressors. In endometrial cells the same drug acts as an agonist by recruiting more highly expressed coactivators to ER-tethered promoters. Thus, the direction of SERM activity   FIG. 6. AGN34 is a selective BAR modulator. A and B, Northern blot analyses were performed as described in the legend to Fig. 3, except that Caco-2 (A) and HepG2 (B) cells were treated with 100 M CDCA and/or 1 M AGN34, as indicated. The relative change in expression of each gene, normalized to GAPDH, is indicated. RNA levels for each gene in untreated cells were given a value of one. C, dose response of AGN34 on CYP7A1 expression in HepG2 cells. Northern blot analyses were performed with HepG2 cells treated with ligands as indicated. The relative change in expression of CYP7A1 is also indicated with the RNA levels in untreated cells given a value of 1.
is determined in a combinatorial fashion by at least three factors: the conformation of the ligand-receptor complex, the promoter context, and the relative levels of expression of specific coactivator and corepressor proteins.
Like their SERM counterparts, the future development of BARMs will benefit from a more complete description of the mechanisms by which BAR regulates gene transcription. Direct binding sites for BAR have been identified in the promoters of certain target genes (e.g. IBABP and SHP) (21) but it remains unclear which coactivator proteins and/or transcription factors are utilized by BAR for in vivo regulation of these genes. In the case of transrepression, several negative BAREs have been identified in the CYP7A1 gene but BAR does not interact directly with these elements (13). Instead, it had been proposed that transrepression occurs by BAR-mediated stimulation of SHP, which in turn represses transcription at the negative BARE (14,15). Our findings demonstrate that alternative mechanisms must be utilized as CYP7A1 repression occurs without an induction of SHP. Indeed, very recent studies with SHP-null mice have reached a similar conclusion (39). By dissociating SHP induction from CYP7A1 repression, AGN34 provides a convenient tool for the future elucidation of this SHP-independent pathway of transrepression.
While this work was in progress, a plant-derivative known as guggulsterone was also identified as a BAR antagonist (40,41). However, this compound was 1000-fold less potent than AGN34 and was not specific: guggulsterone partially inhibited many nuclear receptors (40,41) and we find that it fully inhibited CAR and strongly activated PXR (data not shown). The activity of guggulsterone on the critical CYP7A1 target gene was not reported and would be difficult to assess as other receptor targets (PXR) for guggulsterone also regulate CYP7A1. Thus, guggulsterone is a low-affinity, low-specificity antagonist and it is unclear whether it possesses the geneselective activity that is characteristic of AGN34.