Identification of a novel human constitutive androstane receptor (CAR) agonist and its use in the identification of CAR target genes.

The orphan nuclear constitutive androstane receptor (CAR) is proposed to play a central role in the response to xenochemical stress. Identification of CAR target genes in humans has been limited by the lack of a selective CAR agonist. We report the identification of 6-(4-chlorophenyl)imidazo[2,1-b][1,3]thiazole-5-carbaldehyde O-(3,4-dichlorobenzyl)oxime (CITCO) as a novel human CAR agonist with the following characteristics: (a) potent activity in an in vitro fluorescence-based CAR activation assay; (b) selectivity for CAR over other nuclear receptors, including the xenobiotic pregnane X receptor (PXR); (c) the ability to induce human CAR nuclear translocation; and (d) the ability to induce the prototypical CAR target gene CYP2B6 in primary human hepatocytes. Using primary cultures of human hepatocytes, the effects of CITCO on gene expression were compared with those of the PXR ligand rifampicin. The relative expression of a number of genes encoding proteins involved in various aspects of steroid and xenobiotic metabolism was analyzed. Notably, CAR and PXR activators differentially regulated the expression of several genes, demonstrating that these two nuclear receptors subserve overlapping but distinct biological functions in human hepatocytes.

The nuclear receptors CAR 1 (NR1I3) and PXR (NR1I2) play key roles in the response to chemical stress (1)(2)(3)(4). Both CAR and PXR have been shown to bind to a wide range of structurally unrelated ligands (5)(6)(7) and to regulate genes involved in the humoral response to both endobiotic and xenobiotic stress (2)(3)(4). Global gene expression profiling has shown that PXR and CAR regulate an overlapping set of genes that encode proteins involved in the detoxification of potentially harmful xenobiotics and endobiotics (8,9). For example, CAR-and PXRdependent signaling pathways converge on common response elements in the regulatory regions of a number of genes, notably members of the CYP3A and CYP2B subfamilies of xeno-biotic-inducible cytochromes P450 (10 -12). Current studies are aimed at broadening our understanding of the biology of these receptors and the genes that they regulate. An important goal is to delineate CAR-and PXR-specific target genes to define their distinct physiological roles.
The identification of target genes for each receptor is facilitated by the availability of potent and selective ligands. PXR has been shown to be activated by a structurally and chemically diverse set of ligands (13). Examples of human PXR activators include the xenobiotics rifampicin and SR12813 (14 -17), the endobiotics lithocholic acid (4,18) and 5␤-pregnane-3,20-dione (15), and the botanical hyperforin (19). In human studies, rifampicin has been shown to be a useful chemical tool to define PXR target genes in human hepatocytes (9). Expression studies using rifampicin have shown that PXR activates the expression of a battery of genes involved in the response to xenochemical and endobiotic stress, most notably those genes involved in oxidation (phase I enzymes), conjugation (phase II enzymes), and transport (9).
In contrast to PXR, a selective chemical tool has not been available to study the function of CAR in humans. The hepatomitogen TCPOBOP is a potent murine CAR ligand that has been used to delineate CAR target genes in mice, but does not activate human CAR (5,7,9). The barbiturate phenobarbital activates both human and mouse CARs; however, it does so though an indirect mechanism (7,20). Thus, although phenobarbital does not bind to the receptor (7), it causes CAR to be translocated from the cytoplasm to the nucleus (20 -22). Because CAR exhibits an intrinsically high transcriptional activity, nuclear localization of the receptor results in the activation of target gene expression in the absence of ligand binding (20,23). The induction of CAR translocation by phenobarbital can be blocked by the phosphatase inhibitor okadaic acid, suggesting that translocation involves a dephosphorylation event (21). Importantly, phenobarbital has been shown to induce large numbers of genes in a CAR-independent fashion, which may be due to its effects on the phosphorylation status of the cell (24); and moreover, in humans, phenobarbital also activates PXR, further complicating the interpretation of its effects (7). Similarly, the human CAR ligands 5␤-pregnane-3,20-dione (agonist) and clotrimazole (antagonist) are both effective activators of human PXR (7), which limits their utility in the identification of CAR target genes.
Given the species selectivity with respect to activators of human and mouse CARs and the unusual divergence of their respective ligand-binding domains (LBDs), it is possible that these two receptors perform different roles in mice and humans and in other species (25). The availability of potent and selec-tive CAR ligands for both the murine and human receptors will enable direct comparison of their species-specific roles.
Through a combination of in vitro and cell-based screening, we have identified an imidazothiazole derivative that is a selective human CAR agonist. This chemical tool has allowed us to unambiguously define CAR target genes. This compound should be a powerful tool in differentiating the role of human CAR and PXR. We have demonstrated the utility of this compound in human hepatocyte studies.
Fluorescence Resonance Energy Transfer (FRET) Ligand Sensing Assay-The FRET ligand sensing assay was performed by modification of a previously published procedure (26) and is described in Ref. 7. Polyhistidine-tagged human CAR LBD was purified from Escherichia coli as previously described (7).
Construction of a Green Fluorescent Protein (GFP)-CAR Expression Plasmid-Full-length human CAR cDNA (GenBank TM /EBI accession number Z30425) was amplified by PCR using primers with flanking EcoRI and BamHI sites and subsequently inserted into the EcoRI and BamHI sites of pEGFP-C1 (Clontech, Palo Alto, CA), producing pGFP-hCAR. The sequence of the human CAR cDNA was confirmed by sequence analysis.
Nuclear Translocation Assay-Primary rat hepatocytes were prepared by perfusion as previously described (28) and plated at a density of ϳ3 ϫ 10 6 cells/well of a six-well dish in Williams' E medium containing 10% fetal bovine serum, 100 nM dexamethasone, and 1% ITS-G (insulin/transferrin/selenium; Invitrogen). After overnight incubation, cells were transferred to 2 ml of Williams' E medium as described above, but without serum. For each well, 1.6 g of pGFP-hCAR expression plasmid in 100 l of Opti-MEM was mixed with 3 l of LipofectAMINE 2000 (Invitrogen) in 100 l of Opti-MEM according to the manufacturer's directions and subsequently added directly to the hepatocytes in serum-free medium. After a 4-h incubation, cells were incubated in Williams' E medium containing 10% serum and incubated overnight prior to the addition of compounds in fresh medium. The intracellular localization of the GFP-CAR fusion protein was determined by fluorescence microscopy ϳ4 h after the addition of compound.
Treatment of Primary Human Hepatocytes-Primary human hepatocytes were obtained from BioWhittaker, Inc. (Walkersville, MD) and plated at an approximate density of 3 ϫ 10 6 cells/well in a six-well plate. Cells were maintained in Williams' E medium supplemented with 100 nM dexamethasone, 2 mM L-glutamine, and 1% ITS-G. Cells were treated with vehicle (0.1% Me 2 SO), rifampicin (10 M), or CITCO (100 nM). Fresh compound and medium were added after 24 h, and cells were harvested after 48 h.
RNA Preparation and Expression Analysis-Total RNA from human hepatocyte cultures was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. Real-time quantitative (RTQ) PCR was performed using an ABI PRISM 7700 sequence detection system instrument and software (Applied Biosystems, Inc., Foster City, CA). RNA samples were prepared for RTQ-PCR as described (9). Gene-specific primers and probes were designed using Primer Express Version 2.0.0 (Applied Biosystems, Inc.) and synthesized by Keystone Laboratories (Camarillo, CA). All primers and probes were entered into the NCBI BLAST program to ensure specificity. -Fold induction values were calculated by subtracting the mean threshold cycle number for each treatment group from the mean threshold cycle number from the vehicle group and raising 2 to the power of this difference.

Identification of a Selective Human CAR Agonist
Several criteria were used to identify a potent and selective CAR agonist. These included (a) activity in an in vitro FRETbased assay, (b) Ͼ50-fold selectivity over PXR in a transient transfection assay, and (c) the ability to induce CAR translocation from the cytoplasm to the nucleus in primary hepatocytes. We then sought to use this chemical tool to identify genes that are regulated by CAR in primary cultures of human hepatocytes.
Identification of a CAR Agonist in a FRET-based Screen-A nuclear receptor-biased chemical library was screened in a CAR FRET-based assay using human CAR LBD and a peptide containing the second LXXLL motif of SRC-1 (steroid receptor coactivator-1; amino acids 676 -700). Compounds that induced increased interaction between these partners with potencies of Ͼ100 nM were chosen for further analysis. One of the compounds, the imidazothiazole derivative CITCO (Fig. 1A), displayed a half-maximal effective concentration (EC 50 ) of 49 nM in the CAR/SRC-1 FRET assay (Fig. 1B). CITCO was 50-fold more potent than the human CAR agonist 5␤-pregnane-3,20dione (EC 50 ϭ 3000 nM) (Fig. 1B). The human CAR antagonist clotrimazole (7) was also evaluated in this assay and demonstrated an IC 50 of 58 nM.
Selectivity for CAR Versus PXR of Ͼ50-fold-CAR agonists derived from the in vitro assay were tested for CAR selectivity. Using the CAR-and PXR-responsive XREM-CYP3A4-LUC reporter gene construct in CV-1 transient transfection assays, the selectivity of the compounds for human CAR over human PXR was assessed. The majority of compounds that were active
in CAR/SRC-1 FRET assays were Ͻ50-fold selective for CAR in comparison with human PXR (data not shown). CITCO was one of the few compounds that displayed Ͼ50-fold selectivity for CAR over PXR in the transient transfection assay (Fig. 2, A and  B). CITCO displayed calculated EC 50 values of 25 nM in the CAR transient transfection assay and ϳ3 M in the PXR transient transfection assay (Ͼ100-fold selectivity for CAR). We also carried out transient transfection studies to show that CITCO and the human CAR antagonist clotrimazole are competitive in their effects on CAR transcriptional activity ( Fig.  2A). In the presence of 1.5 M clotrimazole (an approximate EC 70 in this assay), the EC 50 of CITCO increased by Ͼ10-fold (EC 50 ϭ 304 nM), indicating competition between the two compounds for the receptor.
CAR Translocation Assay-Although CITCO was a potent activator of human CAR, the efficacy of this compound and other CAR activators was relatively weak in assays performed in immortalized cell lines, possibly because CAR is constitutively present in the nucleus in these assays. Thus, the results of the transfection assays do not accurately predict the overall efficacy of the compound in hepatocytes because inactive CAR is restricted to the cytoplasm in these cells. Thus, to assess the ability of CAR ligands to induce translocation of CAR from the cytoplasm to the nucleus, we developed a translocation assay in primary cultures of hepatocytes. The full-length human CAR coding region was fused in-frame to the GFP coding region and transfected into rat hepatocytes. The effects of various compounds on CAR translocation were visualized by fluorescence microscopy (Fig. 3). Although endogenous CAR has previously been shown to be predominantly localized in the cytoplasm in the absence of stimulation, GFP-CAR was present in both the cytoplasm and nucleus (Fig. 3, panels 1 and 2). This is likely due to the relatively high concentration of the GFP-CAR chimera expressed under these assay conditions. In addition to a widespread GFP-CAR distribution, control cells showed a somewhat crenulated pattern of fluorescence, indicating that CAR may be attached to a subcellular structure. Further stud- ies are needed to determine whether this is true for native CAR as well or whether this property is unique to the GFP-CAR chimera.
When hepatocytes expressing GFP-CAR were treated with phenobarbital, the GFP signal was localized predominantly in the nucleus (Fig. 3, panels 3 and 4), as expected. Notably, in cells treated with CITCO, the pattern of GFP localization was similar to that seen after phenobarbital treatment (Fig. 3, panels 5 and 6), indicating that CITCO causes efficient nuclear translocation of CAR in hepatocytes. In summary, CITCO fulfills our criteria for a useful human CAR chemical tool: it is a potent and selective human CAR ligand that activates the receptor in a transfection assay and promotes its translocation into the nucleus of hepatocytes.

Comparison of the Effects of CITCO and Rifampicin on Gene Expression in Primary Human Hepatocytes
In the absence of a selective human CAR agonist, it has been difficult to ascertain which genes are regulated by this receptor in human hepatocytes. We used CITCO to study the effects of CAR activation on gene expression in primary cultures of human hepatocytes. Hepatocytes derived from three separate donors were treated for 48 h with either 1 M CITCO or 10 M rifampicin, a selective PXR agonist. The comparative effects of the agonists on the expression of eight genes involved in a variety of aspects of xenobiotic metabolism were quantitated by RTQ-PCR (Table I). RNA in sufficient quantity was available from two donors (Donors 1 and 2) to further evaluate selected gene expression changes by Northern blot analysis (Fig. 4). The mRNAs evaluated included those encoding multiple cytochrome P450 enzymes (CYP2A6, CYP2B6, and CYP3A4); enzymes involved in supporting phase I metabolism (aldehyde dehydrogenase (ALDH1A4) and aminolevulinate synthase); and enzymes involved in phase II (conjugation) metabolism, including glutathione S-transferase A2 (GSTA2) and sulfotransferase (SULT1A1). The mRNA encoding the conjugation enzyme UDP-glucuronosyltransferase (UGT1A1) was examined exclusively by Northern analysis due to difficulties in generating functional RTQ-PCR primers for this mRNA. We also examined the effects of CITCO and rifampicin on the expression of the MDR1 (multidrug resistance-1) transporter mRNA.
Phase I Enzyme Genes-CAR regulates both Cyp2b10 and Cyp3a11 in mouse liver (8 -10, 12) and has also been implicated in the regulation of CYP2B6 and CYP3A4 in human hepatocytes (11,12,29). However, the role of CAR in the regulation of these genes has not been examined rigorously in human hepatocytes using a selective CAR chemical tool. Both CITCO and rifampicin induced CYP2B6 and CYP3A4 in human hepatocytes from Donors 1 and 2 as measured by RTQ-PCR (Table I).
Rifampicin was not effective at inducing CYP2B6 in Donor 3, whereas CITCO was not effective at inducing CYP3A4 in this donor. This likely reflects the relative efficacy of rifampicin and CITCO in induction of CYP3A4 and CYP2B6, respectively, coupled with higher basal levels of expression of these genes in Donor 3. When the effects of CITCO were evaluated by Northern blot analysis, CYP2B6 mRNA was found to be induced by 4.4-and 20-fold in Donors 1 and 2, respectively (Fig. 4). Rifampicin also induced CYP2B6 expression, although to a lesser extent than CITCO ( Fig. 4 and Table I). In all three donors, CYP2B6 was induced more efficiently by CITCO than by rifampicin, indicating that this gene is more responsive to CAR than to PXR.
In contrast to CYP2B6, CYP3A4 mRNA displayed a more robust response to rifampicin than to CITCO. When assessed by RTQ-PCR, rifampicin induced CYP3A4 mRNA by 15-fold in Donor 1 and by 78-fold in Donor 2, whereas CITCO induced CYP3A4 mRNA by 7.0-and 46-fold in the same donors. When evaluated by Northern analysis, rifampicin induced CYP3A4 expression by 30-and 5.9-fold, respectively, whereas CITCO induced CYP3A4 by 11-and 6.2-fold, respectively (Fig. 4). These data indicate that both PXR and CAR regulate CYP3A4 in human hepatocytes.
Although technically not a phase I enzyme, the CYP2A6 gene was also examined for PXR and CAR regulation. The mouse homolog (Cyp2a4) has previously been shown to be regulated by mouse CAR (8,9). Analysis of the three donors by RTQ-PCR showed that CYP2A6 mRNA was induced selectively by CITCO ( Table I). Analysis of Donors 1 and 2 by Northern blotting was consistent with these results (Fig. 4). In this analysis, CYP2A6 mRNA was induced by 5.4-and 9.6-fold in Donors 1 and 2, respectively, whereas rifampicin did not cause changes in the expression of this mRNA. Thus, CYP2A6 is selectively induced by CAR, but not by PXR.
Other genes involved in phase I metabolism were also evaluated by RTQ-PCR. Aminolevulinate synthase mRNA was induced by both CITCO and rifampicin in all of the donors as assessed by RTQ-PCR (Table I). Notably, ALDH1A4 showed a highly variable response to CITCO and rifampicin. Induction of ALDH1A4 varied from 1.8-fold (Donor 1) to 60-fold (Donor 3) in response to CITCO and from no induction (Donor 1) to a 160fold induction (Donor 3) in response to rifampicin. Thus, certain genes appear to display a high degree of inter-individual variability in terms of their response to selective CAR and PXR agonists.
Phase II Conjugation Genes-The effects of CITCO and rifampicin on the expression of the mRNAs encoding the conjugation enzymes GSTA2, SULT1A1, and UGT1A1 were also examined. Both CITCO and rifampicin were able to induce the expression of GSTA2 when assessed by RTQ-PCR and North-  Fig. 4 and Table I). When mRNAs from Donors 1 and 2 were evaluated by Northern analysis, CITCO induced GSTA2 mRNA by 3.4-fold in Donor 1 and by 19-fold in Donor 2 (Fig. 4). Rifampicin induced GSTA2 mRNA by 3.8-and 3.5-fold in these same donors (Fig. 4). SULT1A1 mRNA was robustly induced only by CITCO in one of the three donors. In Donor 1, induction by either compound was Ͻ2-fold as measured by Northern blotting or RTQ-PCR; and in Donor 3, no induction was seen by either compound by RTQ-PCR. In contrast, CITCO strongly induced SULT1A1 in Donor 2 (11-fold increase as measured by Northern blot analysis and 7.5-fold increase as measured by RTQ-PCR). The variation in donor response may again be attributable to interindividual heterogeneity in basal levels of gene expression. Northern blot analysis showed that Donor 1 had relatively high basal levels of expression of SULT1A1 (Fig. 4).
In Northern analysis, CITCO induced UGT1A1 mRNA by 1.8-and 3.7-fold in Donors 1 and 2, respectively (Fig. 4). In contrast, rifampicin did not induce expression of UGT1A1 significantly in Donor 2, but induced expression of UGT1A1 mRNA by 2.5-fold in Donor 1. Again, Donor 1 had a higher basal level of activity.
Transporter Expression-The multidrug resistance genes, including MDR1, function as broad-specificity transporters in the liver. We examined the response of MDR1 to CITCO and rifampicin by RTQ-PCR (Table I). MDR1 expression was induced by both CITCO and rifampicin in Donor 2, but no induction by either compound was seen in Donors 1 and 3. Thus, for MDR1, a significant inter-individual response is seen. In certain individuals, induction of MDR1 gene expression occurs in response to both PXR and CAR agonists.
Model of CITCO Binding in the Ligand-binding Pocket of CAR-Although no x-ray structure is available for CAR, x-ray structures have been done for the closely related receptors PXR (30) and VDR (31). The x-ray structure of PXR revealed a large and practically spherical ligand-binding pocket that can bind a wide range of lipophilic ligands, whereas VDR has a smaller pocket with polar side chains positioned to recognize specific ligands (31). In PXR, the pocket expansion is due primarily to a 50-60-residue insert between helixes 1 and 3. This helix 1-3 insert displaces helix 6, thereby opening the pocket. VDR also has a helix 1-3 insert; there is no evidence that it displaces helix 6. Although residues in the "core" of CAR LBD have greater identity to PXR (50%) than to VDR (40%), CAR lacks the helix 1-3 insert, and its helix 6 should have a geometry more similar to that in VDR than in PXR (6). Consequently, we chose to use VDR as the template in building a model for CAR. The MVP program (32) was used to build the model for CAR and to dock CITCO into the model. A number of different binding modes were obtained for CITCO, one of which is shown in Fig. 5. Asn 165 lies near the oxime linkage and might possibly donate a hydrogen bond to the oxime. This particular binding mode has the para-chlorophenyl ring directed downwards and the imidazothiazole group directed upwards, but the calculations also gave binding modes where the positions of these groups were interchanged, with the oxime linker still located near Asn 165 . The modeling and docking calculations are not accurate enough to distinguish among the possible binding orientations, but it is clear that the CAR binding pocket can accommodate CITCO. The model CAR pocket is smaller than that in PXR and somewhat more lipophilic than that in VDR, suggesting that CAR should be intermediate between VDR and PXR in terms of ligand promiscuity. DISCUSSION To date, little is known about the function of CAR in humans. Attempts to delineate CAR biology in man have been hindered by the significant overlap in the pharmacology of human CAR and PXR and the lack of a selective CAR activator. CITCO is a potent and, importantly, highly selective human CAR agonist that should prove to be a useful tool in dissecting the structure and function of this receptor.
Direct comparison of human CAR and PXR target genes is now possible, as is direct comparison of CAR target genes in mouse versus human cells. In mice, PXR and CAR differential gene expression studies have previously been carried out using the selective tools pregnenolone 16␣-carbonitrile (selective PXR agonist in mouse) and TCPOBOP (9). Gene expression studies with these compounds provide the framework to begin comparative analysis. For the majority of the genes that overlapped between the mouse and human studies, similar profiles were seen using PXR-and CAR-specific compounds. For example, similar to mice, both selective CAR and PXR agonists regulated CYP2B and CYP3A expression, consistent with previous studies suggesting that PXR and CAR cross-regulate these genes (10 -12, 22). Interestingly, CITCO had more robust effects on CYP2B expression, whereas rifampicin had more robust effects on CYP3A4, indicating that the receptors have different quantitative effects depending on the specific gene (7).
In our comparison of CITCO with the human PXR ligand rifampicin in three separate sets of human hepatocytes, we observed remarkable inter-donor heterogeneity. This observation is consistent with previous studies showing that cytochrome P450 expression is quite variable in primary human hepatocyte preparations (33-35). For example, inter-individual variations in CYP3A4 protein levels ranging up to 40-fold have been reported (36,37). Moreover, ethnic differences in CYP3A4-mediated drug metabolism have been reported (38), and it is estimated that ϳ90% of the inter-individual variability in CYP3A4 expression is due to genetic factors (39). In the case of the hepatocytes used in these studies, the inter-individual variability would be expected to be even higher due to the fact that the donors were typically undergoing drug therapy just prior to the harvest of the hepatocytes. The relatively high basal levels of multiple genes seen in Donor 1 versus Donor 2 might reflect genetic differences, differences in drug exposure, or both.
When comparing CAR target genes in mice versus humans, we found that, generally, genes regulated by only selective CAR ligands in mice were also regulated by the selective human CAR ligand. Notably, the phase II conjugation enzyme sulfotransferase gene (mouse SULTN/human SULT1A1) was more responsive to 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene in mouse cells and to CITCO in human cells. Also, the mouse Cyp2a4 and human CYP2A6 genes were typically more responsive to CAR compounds than to PXR compounds. These genes are particularly interesting because, in addition to hydroxylat-ing xenobiotics, the Cyp2a4 and CYP2A6 gene products hydroxylate a variety of steroid hormones, including androgens and estrogens (40). It is possible that CAR plays a different role from PXR in metabolizing endogenous steroids. Other similarities in gene response to CAR and PXR activators were also noted. For example, the phase II genes Gstm2 (mouse) and GSTA2 (human) and the transporter genes Mdr1a (mouse) and MDR1 (human) were induced by both PXR and CAR activators.
Differences between the human and mouse studies were also observed. For example, the UGT1A1 gene has been shown to be regulated by CAR in humans, and a CAR response element has been identified in its promoter (41). Interestingly, we found previously that the mouse CAR agonist 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene does not induce mouse UGT1A1 mRNA (9). In contrast and consistent with the findings of Sugatani et al. (41), we found that a human CAR agonist induces human UGT1A1 mRNA expression. These interspecies differences are intriguing, but, at this point, must be interpreted with caution because the mouse liver studies were carried out in an in vivo setting, whereas the human compounds were assessed in primary hepatocyte cultures. Despite dramatic differences in the amino acid sequences of their respective LBDs, we have demonstrated that human and mouse CARs regulate overlapping sets of genes. These studies have also begun to differentiate some of the distinct physiological pathways regulated by CAR versus PXR.
Finally, the majority of the compounds we identified in in vitro binding assays were not able to induce CYP2B6 mRNA in primary human hepatocytes (data not shown), suggesting that ligand binding alone is not sufficient to induce CAR translocation to the nucleus. Thus, CITCO, in conjunction with these compounds, should be a useful comparative tool to define the key determinants for ligand-induced CAR nuclear translocation. We have presented a model that supports direct binding of CITCO within the ligand-binding pocket of human CAR. Future crystallography studies of human CAR complexed with CITCO and CITCO analogs will further improve our understanding of the structural features required for CAR nuclear translocation and activation.