Benzoate X Receptors α and β Are Pharmacologically Distinct and Do Not Function as Xenobiotic Receptors

The Xenopus benzoate nuclear hormone receptors, BXRα and BXRβ, share 82% identity within their ligand-binding domains and are classified as members of the NR1I2 subfamily that includes the mammalian steroid and xenobiotic receptor, SXR/PXR. Although alkyl benzoates have been identified as endogenous ligands, the exact role of the benzoate receptors in amphibian physiology has not been established. In this report, we show that BXRα and BXRβ are pharmacologically distinct from each other: BXRα is more promiscuous than BXRβ with respect to both ligand specificity and co-activator recruitment. BXRα can be transactivated by a number of benzoate derivatives including 4-amino-butylbenzoate (4-ABB), 4-hydroxy-butylbenzoate (4-HBB), 3-hydroxy ethyl benzoate (3-HEB), and benzyl benzoate, but only 4-HBB acts as an agonist for both receptors. Furthermore, BXRα-specific agonists such as 4-ABB, chlorpyrifos, and trifluralin act as antagonists on BXRβ. BXRs are widely distributed in adult tissues but do not show any enrichment in liver and intestine, major sites of SXR/PXR expression that are critical in xenobiotic metabolism. Neither BXR shows the broad specificity toward steroids or xenobiotics exhibited by SXR/PXR. Therefore, we conclude that the BXRs are pharmacologically distinct from each other and unlikely to serve as xenobiotic sensors.

Nuclear receptors are ligand-modulated transcription factors that respond to steroids, retinoids, and thyroid hormones to control development and body physiology. Orphan nuclear receptors possess apparent DNA and ligand-binding domains but lack identified ligands (1)(2)(3). Each orphan has the potential to regulate a distinct signaling pathway. The promise of orphan receptors is that the identification of novel and perhaps unsuspected classes of ligands may offer insight into potentially new principles of development and physiology. In recent years, a number of orphan receptors have been adopted or matched with physiological ligands (3,4). Consequently, new insights into cholesterol and bile acid metabolism and transport (4) have been gained.
Previously, we identified a Xenopus orphan nuclear receptor that represented a distinct branch of the nuclear receptor superfamily and named it benzoate "X" receptor (BXR) 1 (5) (also known as xONR1 (6)). The name reflects its activation by alkyl esters of amino and hydroxyl benzoic acids, one of which is found endogenously in the Xenopus embryo (5). The identification of BXR as a receptor for benzoate ligands illustrates the potential of uncovering previously unsuspected signaling pathways through orphan receptor characterization. Recently, a second Xenopus BXR cDNA was described (7). This BXR shares only 88% nucleic acid sequence identity and 83% amino acid sequence identity (see Fig. 1 and Ref. 7) with the BXR we characterized previously (5). Therefore, the two receptors have been designated as BXR␣ and BXR␤ (7). BXR is most closely related to the human steroid and xenobiotic receptor, SXR (8) (also known as pregnane X receptor, PXR (9), and pregnane activated receptor, PAR (10)). SXR and its rodent ortholog PXR function as xenobiotic sensors in the liver and intestine. They mediate the breakdown and elimination of steroids, drugs, and xenobiotic compounds by activating the expression of degradative cytochrome P450 enzymes and members of the ABC family of organic molecule transporters. BXR and SXR/PXR have been assigned to the NR1I2 family by the Nuclear Receptor Nomenclature Committee (11). This indicates that these receptors are orthologous, i.e. the same gene from different species. During our characterization of SXR, we noted that none of the compounds that activated SXR was able to activate BXR␣ (8). 2 This led us to question whether BXR and SXR are functionally equivalent, i.e. do BXRs function as xenobiotic sensors? As described below, we found that neither BXR␣ nor BXR␤ is activated by the types of xenobiotic compounds that activate SXR/PXR. In addition, BXRs are ubiquitously expressed at varying levels in different tissues rather than showing the high level expression only in the liver and intestine characteristic of SXR/PXR. We infer that BXRs are unlikely to be functioning as xenobiotic sensors and are therefore functionally distinct from their mammalian relatives. Lastly, we show that BXR␤ is activated only by 4-hydroxyl benzoates as compared with BXR␣, which can also be activated by other related compounds including amino benzoates, benzyl benzoate, chlorpyrifos, and trifluralin. Interestingly, several of the BXR␣-selective activators function as antagonists for BXR␤, thus suggesting that the two BXRs are also pharmacologically distinct from each other.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-COS-7 cells were cultured in phenol red-free DMEM supplemented with 10% FBS. For transient transfection experiments, COS-7 cells were seeded into 96-well plates at a density of 5000 cells/well. 4 -5 h after seeding, the cells were transfected with CMX-GAL4-xBXR␣ (5), CMX-GAL4-xBXR␤, or CMX-GAL4 (control) together with tk(MH100)4-luc reporter (12) and CMX-␤-galactosidase transfection control plasmids using standard calcium phosphate precipitation methodology. 22-24 h after transfection, the cells were washed twice with phosphate-buffered saline supplemented with 1 mM MgCl 2 or DMEM-ITLB (DMEM containing 5 g/ml insulin, 5 g/ml holo-transferrin, 5 g/ml selenium, 0.5% defined lipid mix (Invitrogen), 0.12% w/v delipidated bovine serum albumin (Sigma)) (13). Ligands were next added in DMEM-ITLB, and the cells were incubated for an additional 24 -48 h. Ligands were typically purchased from Sigma, ChemService (West Chester, PA), or Roche Molecular Biochemicals and made freshly from powder in Me 2 SO as 0.1 M stocks, diluted in Me 2 SO to appropriate concentrations and added to media with vigorous vortex mixing. The cells were incubated with ligands for 24 -48 h and then lysed in situ. Extracts were prepared and assayed for ␤-galactosidase and luciferase activity as described (8). Reporter gene activity was normalized to the ␤-galactosidase transfection controls, and the results were expressed as normalized relative luciferase units per OD of ␤-galactosidase per minute to facilitate comparisons between plates. Fold induction was calculated relative to solvent controls. Each data point represents the average of triplicate experiments Ϯ S.E. and was replicated in independent experiments.
Similarly, GAL4 coactivator plasmids were generated by cloning the receptor interaction domains of human TIF2 (GenBank™ accession number NM006540, amino acids 563-790), human F-SRC-1 (Gen-Bank™ accession number U59302, amino acids 600 -800), or human ACTR (GenBank™ accession number AF036892, amino acids 600 -788) into pCMX-GAL4. The GAL4-PBP construct was a gift from B. Forman (City of Hope Medical Center). To construct Herpesvirus VP16 activation domain fusion proteins, full-length BXR␣ and BXR␤ were PCRamplified and ligated in-frame into pCMX-VP16 vector. All constructs were sequenced to verify that no errors were introduced in the PCR.

RESULTS
Comparative Expression of BXR␣ and BXR␤-Recently, a second Xenopus BXR cDNA was described (7). This BXR shares only 88% nucleic acid sequence identity and 83% amino acid sequence identity ( Fig. 1A and Ref. 7) with the BXR we characterized previously (5). This is more than would be expected for the divergence between two duplicated genes in the pseudotetraploid X. laevis genome (17). Accordingly, the new receptor was called BXR␤ (7). To gain insight into the possible target tissues for BXR action, we examined the expression patterns of BXR␣ and BXR␤ in adult frogs by Northern blot and quantitative real time RT-PCR analysis. Both genes encode ubiquitously expressed single transcripts of ϳ3.2 kb (data not shown) and are found at very high levels in the ovary (Fig.  1B). BXR␣ is expressed at high levels in the brain and skin with moderate levels in the testis, stomach, and intestines and lower levels in the lung, kidney, liver, and heart (Fig. 1B). BXR␤ is expressed at comparable levels to BXR␣ in the intestines, lung and kidney with slightly higher levels in the testis and heart and lower levels in the liver, skin and brain (Fig. 1B). It is notable that these ubiquitous expression patterns for BXR differ considerably from those of its putative human ortholog, the steroid and xenobiotic receptor SXR (8). SXR functions as a xenobiotic sensor and is expressed primarily in the liver and intestine, where it modulates the levels of cytochrome p450 enzymes and ABC family transporters (8,9). Since BXRs are ubiquitously expressed, they do not show the tissue distribution expected for a xenobiotic sensor.
BXR␣ and BXR␤ Are Pharmacologically Distinct-Our previous work showed that alkyl esters of amino and hydroxyl benzoic acids specifically bind to and activate BXR␣ (5). Therefore, it was surprising that BXR␤ was reported to be activated only very weakly by 4-amino butyl benzoate (7), which strongly activates BXR␣ (5). We tested the activation profiles of these receptors to determine whether the two BXRs might exhibit different ligand specificity. BXR␣ and BXR␤ were transiently transfected into COS-7 cells, and then a panel of benzoates and related compounds was tested for their ability to activate transcription of a luciferase reporter gene (Fig. 2). Activation of BXR␣ paralleled our published results ( Fig. 2A) in that both hydroxyl and amino benzoates were robust activators. In addition, we observed that benzyl benzoate, trifluralin, and chlorpyrifos also activated BXR␣ ( Fig. 2A). In contrast, only 4-hydroxyl butyl benzoate was able to activate BXR␤ (Fig. 2B). It is particularly notable that 3-hydroxyl ethyl benzoate, which was identified as an endogenous embryonic activator of BXR␣, could activate BXR␣ but was inactive on BXR␤ (Fig. 2, A and  B). BXR␤ was considerably more active in response to ligand than was BXR␣ (Fig. 2, A and B). A similar trend was also observed in dose-response experiments (Fig. 2, C and D). BXR␣ was strongly activated by 10 M 4-ABB, 4-HBB, chlorpyrifos, and trifluralin (Fig. 2C). BXR␤ was only activated by 4-HBB among the many compounds tested (Fig. 2, B and D). The activation of BXR␤ was very robust with 50 M of 4-HBB yielding between 50-and 100-fold activation of the reporter gene (Fig. 2, B and D).
Co-activator Recruitment by BXRs-To confirm that BXR␣ and BXR␤ show distinct pharmacological responses to ligand, we conducted co-activator recruitment studies to determine the ability and preferences of the various ligands to support the formation of specific active transcriptional complexes. This mammalian two-hybrid assay utilized VP16-BXR␣ and VP16-BXR␤ together with fusions between the GAL4-DNA-binding domain and the receptor-interacting domain of the nuclear hormone receptor co-activators, TIF-2, ACTR, SRC-1, and PBP (18) to investigate the ability of BXR␣ agonists to promote productive transcriptional interactions.
VP16-BXR␣ was able to interact with all four co-activators in the presence of agonistic ligands (Fig. 3A). VP16-BXR␣ showed the strongest interaction with SRC-1 irrespective of the ligand tested (4-HBB, 12-fold; 4-ABB, 8-fold; chlorpyrifos and trifluralin 6-fold), although, a rank order of co-activator response could be identified (SRC-1 Ͼ TIF-2 Ͼ PBP ϭ ACTR) (Fig. 3A). This is notably different from the rank order of potency of SXR for the same coactivators (SRC1 Ͼ PBP Ͼ TIF-2/GRIP Ͼ ACTR) (19). Overall, the rank order of potency of compounds in the co-activator recruitment assay paralleled their potency in the activation assays (Fig. 2).
Co-activator interaction with BXR␤ was much more restricted in response to agonist (Fig. 3B). Consistent with the dose-response experiments (Fig. 2D), only 4-HBB possessed any strong ability to promote co-activator interaction. The response observed with SRC-1 was approximately equivalent to that seen with BXR␣ (10-versus 12-fold). The level of interaction with PBP was weaker (4-fold) although comparable with that seen with BXR␣ (Fig. 3A). The activation responses with ACTR and TIF-2 were poor. The rank order was SRC-1 Ͼ PBP Ͼ ACTR ϭ TIF-2. The data suggest that SRC-1 is a strong co-activator for both BXR␣ and BXR␤. BXR␣ is promiscuous both in its choice of ligand and in its choice of co-activator, whereas BXR␤ is not promiscuous for either (Fig. 3).
BXR␣-specific Activators Are Antagonists for BXR␤-The observation that most BXR␣ activators could not activate BXR␤ leads to two possible inferences. One is that the compounds specifically bind only to BXR␣. In this case, we would not expect to discern any effect of these compounds on BXR␤ activation. Alternatively, the compounds might bind to both receptors but only activate BXR␣. In this scenario, the BXR␣ activators could act as antagonists for BXR␤. Accordingly, we conducted antagonism experiments using the BXR␤ agonist 4-HBB and BXR␣-specific agonists. Two types of experiments were employed.
First, we tested the ability of BXR␣-specific agonists to directly interfere with BXR␤-mediated activation. COS-7 cells were transfected with GAL4-BXR␤, and inhibitory dose-response curves were derived. Cells were treated with increasing doses of 4-ABB, chlorpyrifos, or trifluralin in the presence of a fixed series of agonist 4-HBB concentrations in the range of 1-50 M. Transcriptional activation of BXR␤ by 4-HBB alone gave a derived mean EC 50 of 33 Ϯ 3.2 M (n ϭ 4) ( Table I). Titration with either chlorpyrifos or trifluralin resulted in a dose-dependent inhibition of 4-HBB-mediated transcriptional activation of the reporter gene (Fig. 4A). In contrast, 4-ABB showed a small additive effect up to 10 M and was inhibitory at 100 M under these conditions (Fig. 4A). Data were fitted by non-linear regression analysis, and inhibitory constants (K i ) for BXR␤-mediated transcriptional activation were calculated using the Cheng-Prusoff equation (20) and shown in Table I. Chlorpyrifos yielded a K i of 0.5 Ϯ 0.2 M (n ϭ 13), trifluralin yielded a K i of 5.2 Ϯ 0.8 M (n ϭ 8), and 4-ABB yielded a K i of Ͼ82 Ϯ 9 M (n ϭ 15). By comparison with the EC 50 value of 33 M for 4-HBB on BXR␤, the data suggest that both chlorpyrifos and trifluralin can act as potent competitive antagonists of BXR␤ activation, whereas 4-ABB demonstrates weaker antagonism on this receptor. 3-HEB did not antagonize the activation of BXR␤ (data not shown).
We next tested the ability of the BXR␣-specific agonists to   (20). Values represent the mean Ϯ S.E. calculated from the indicated number of inhibition curves (n). b EC 50 for BXR␤ activation by 4-HBB was determined from nonlinear regression analysis of dose-response curves.
interfere with 4-HBB-mediated co-activator recruitment in the mammalian two-hybrid assay as described above. COS-7 cells were transfected with VP16-BXR␤, GAL-SRC1, and CMX-␤galactosidase and treated with 50 M 4-HBB in combination with a dose series of 4-ABB, chlorpyrifos, trifluralin, or solvent controls. Fig. 4B shows that 4-ABB, chlorpyrifos, and trifluralin were each able to impair co-activator recruitment by 4-HBB. Therefore, we conclude that these BXR␣-selective activators are able to act as antagonists for BXR␤, supporting the contention that BXR␣ and BXR␤ are pharmacologically distinct from each other.
BXRs Are Not Xenobiotic Sensors-The very different tissue distributions of BXRs and SXR/PXRs led us to suspect that these receptors might be functionally different. To test this hypothesis, we examined a panel of compounds for their ability to activate BXR␣, BXR␤, and human SXR. We found that neither BXR␣ nor BXR␤ was activated by any of the classic SXR/PXR activators (e.g. rifampicin, pregnenolone-16 ␣-carbonitrile, phenobarbitals, or clotrimazole) (Fig. 5). The only xenobiotic compounds that activated BXR␣, chlorpyrifos and trifluralin, have chemical structures similar to benzoates. We also evaluated other known SXR activators including steroids and bile acids for their ability to activate BXR. None of these SXR activators were able to activate BXRs (data not shown). We also note that two compounds reported previously to activate BXR␤, dexamethasone and methylprednisolone (7), do not activate either BXR in our experiments. The reason for this discrepancy is unknown at present but may relate either to the different experimental systems used or to the very low levels of activation originally observed (ϳ2-fold) (7).

DISCUSSION
Despite a flood of genomic and expressed sequence tag sequence information in recent years, there is still little information available that suggests a function for BXR or confirms its presence in animals other than X. laevis. The recent identification of BXR␤ (7) and its activation by 4-HBB (Figs. 2 and 3) suggests that both BXRs are receptors for endogenous hydroxyl benzoates or closely related compounds. This narrows the range of compounds expected to activate both BXR␣ and BXR␤ to a relatively small group and suggests that a detailed focus on these structures will lead to the elucidation of endogenous ligands. It is particularly interesting that BXR␣ and BXR␤ are pharmacologically distinct and that several potent BXR␣ activators are BXR␤ antagonists. This means that treating early embryos with compounds such as 4-ABB would simultaneously activate BXR␣ while inhibiting the activity of BXR␤. This could explain why such treatments do not have obvious adverse effects on the embryos at subtoxic doses. 2 To sort out the biology of the two receptor subtypes, it will be necessary to perform targeted loss-of-function experiments in early embryos using, for example, morpholino antisense oligonucleotides (21,22) coupled with phenotypic rescue experiments.
The results presented in Fig. 2 show that both BXRs are robustly activated only by 4-hydroxyl benzoates, whereas BXR␣ is promiscuously activated by other benzoates and related compounds. This is notable because one of the endogenous benzoates found in Xenopus embryos, 3-HEB, is a BXR␣Ϫselective activator (5) (Fig. 2). Although three endogenous benzoates remain to be identified in Xenopus embryos, the total concentration of benzoate BXR␣ activators is 10 M in the blastula stage embryo (5), which is in the range required for receptor activation. In accord with our results, Moore et al. (23) recently showed that BXR␤ is robustly activated by 4-hydroxyl benzoates and noted, without comment, that BXR␣ is activated by amino benzoates, whereas BXR␤ is not. In contrast to our findings, Nishikawa et al. (7) report only a 1.5-fold activation of BXR␤ by 4-HBB. The reason for this difference is currently unknown but could result from different cell lines utilized in the different laboratories (COS-7 versus HeLa). Alternatively, since bona fide target genes for BXR are unknown, the reporter construct used (7) might not be an effective target for BXR␤ in vivo. We note that BXR␣ shows a strong preference for the sequence AGTTCAnnnnAGTTCA (5) as compared with the AG-GTCAnnnnAGGTCA used by Nishikawa et al. (7) (where n equals any nucleotide).
BXRs are most closely related to the mammalian SXR/PXR gene family. Indeed, BXR was used as a probe to isolate human SXR (8). BXR and SXR/PXR have been assigned to the NR1I2 family by the Nuclear Receptor Nomenclature Committee (11), indicating that these receptors are orthologous, i.e. they represent the same gene from different species. During our original characterization of BXR (5) and SXR (8), we noted that there was little overlap in the set of compounds that activated each receptor. This raised an important issue about the relationship between BXR and SXR/PXR because the latter are known to function as xenobiotic sensors. It is well known that human SXR and mouse PXR are pharmacologically distinct in that each has species-specific activators as well as compounds that activate across species (8,9,24; also reviewed in Refs. 1 and 25). Therefore, one possibility is that BXRs are xenobiotic sensors but that the spectrum of compounds that activates them is species-specific.
We conducted extensive ligand screening experiments and showed that neither BXR is activated by the broad spectrum of steroids and xenobiotic compounds that are known to activate SXR (Fig. 5). Moore and et al. (23) have also tested a large number of steroids and xenobiotic compounds for BXR activation. In accord with our results, they show that BXRs are not activated by most steroids or xenobiotics (23), which argues that BXRs do not function as broad specificity xenobiotic sensors. The only point of disagreement between our datasets is that they show 2.6-fold induction of BXR␣ by rifampicin, whereas it does not activate either BXR in our experiments. However, considering that activation of BXR␣ by bona fide ligands ranges from 180-to 1200-fold, it is difficult to ascertain whether the 2.6-fold induction by rifampicin represents liganddependent transactivation. We infer from our data and those of Moore et al. (23) that BXRs are unlikely to function as xenobiotic sensors.
SXR and PXR expression are enriched in the liver and intestine, where they activate the expression of cytochrome P450 genes and ABC family organic transporters to detoxify and eliminate xenobiotic compounds (25). On the other hand, BXRs are ubiquitously expressed with no particular enrichment in liver or intestine. No known bona fide BXR targets have been identified yet although it is believed that the Pit-1 gene contains a high affinity BXR target element (5). Taken together with their lack of activation by the types of xenobiotic compounds that activate SXR/PXR, one is forced to conclude that these receptors are functionally distinct. It is highly unusual for orthologous nuclear receptors to exhibit such different expression patterns, hence one wonders whether the evolutionary pressure operating on members of this gene family is the same in different species. One possibility is that the BXRs are not really orthologous to SXR/PXR but rather represent a distinct family of receptors with an entirely different biology. For this model to be correct, one would expect to find BXR relatives in other vertebrates and SXR/PXR homologs in Xenopus. We have not found sequences more closely related to BXR than SXR/ PXR in the draft human and mouse genome sequences. Moreover, no sequences more closely related to SXR/PXR than BXR have appeared in the more than 200,000 Xenopus expressed sequence tags identified to date. We have not identified it in numerous screening experiments (data not shown). Thus, if BXR is not orthologous to SXR, it might be a receptor restricted to amphibians or lower vertebrates. This would make it the first such nuclear receptor identified.
An alternative possibility is that BXR is orthologous to SXR in an evolutionary sense but has diverged functionally. This could be due to the different constraints on poikilothermic, aquatic animals versus homoeothermic mammals. Xenopus is a carnivorous frog that likely has a very different diet from humans and rodents. Moore et al. (23) recently identified SXR homologs from zebrafish, pig, dog, and monkey. They showed that the zebrafish gene was activated by several steroids and xenobiotic compounds (23). Interestingly, the BXRs are about equally similar to the zebrafish (46% identity) and the mammalian sequences (ϳ50% identity). Therefore, it appears that the functional divergence of BXRs from the SXR/PXR group of genes cannot be explained by the diet or aquatic nature of the organisms. Therefore, it will be very interesting to identify and characterize BXR and SXR homologs from other amphibians, reptiles, fishes, and lower chordates to determine at which point in evolution BXR localization and function diverged from SXR. Indeed, the question of whether BXR and SXR are true orthologs must await the identification of similar receptors from related species of fish and amphibians.