Structural Basis for Small Molecule NDB (N-Benzyl-N-(3-(tert-butyl)-4-hydroxyphenyl)-2,6-dichloro-4-(dimethylamino) Benzamide) as a Selective Antagonist of Farnesoid X Receptor α (FXRα) in Stabilizing the Homodimerization of the Receptor*

Background: The pharmacological mechanism for FXR antagonist is still unclear. Results: Crystal structure of hFXRα ligand binding domain and NDB as an antagonist of hFXRα was determined. Conclusion: NDB promotes the formation of FXR homodimerization, inhibits FXR/RXR heterodimer, and decreases gluconeogenic genes of db/db mice. Significance: The hFXRα-LBD·NDB structure may help understand the antagonistic mechanism of FXR. Farnesoid X receptor α (FXRα) as a bile acid sensor plays potent roles in multiple metabolic processes, and its antagonist has recently revealed special interests in the treatment of metabolic disorders, although the underlying mechanisms still remain unclear. Here, we identified that the small molecule N-benzyl-N-(3-(tert-butyl)-4-hydroxyphenyl)-2,6-dichloro-4-(dimethylamino) benzamide (NDB) functioned as a selective antagonist of human FXRα (hFXRα), and the crystal structure of hFXRα ligand binding domain (hFXRα-LBD) in complex with NDB was analyzed. It was unexpectedly discovered that NDB induced rearrangements of helix 11 (H11) and helix 12 (H12, AF-2) by forming a homodimer of hFXRα-LBD, totally different from the active conformation in monomer state, and the binding details were further supported by the mutation analysis. Moreover, functional studies demonstrated that NDB effectively antagonized the GW4064-stimulated FXR/RXR interaction and FXRα target gene expression in primary mouse hepatocytes, including the small heterodimer partner (SHP) and bile-salt export pump (BSEP); meanwhile, administration of NDB to db/db mice efficiently decreased the gene expressions of phosphoenolpyruvate carboxykinase (PEPCK), glucose 6-phosphatase (G6-pase), small heterodimer partner, and BSEP. It is expected that our first analyzed crystal structure of hFXRα-LBD·NDB will help expound the antagonistic mechanism of the receptor, and NDB may find its potential as a lead compound in anti-diabetes research.

Farnesoid X receptor ␣ (FXR␣) 4 is a bile acid-activated nuclear receptor responsible for the regulation of the specific target genes involved in a broad spectrum of biological processes including bile acids, glucose and lipid homeostasis, liver regeneration, bacterial growth, and tumor growth (1). FXR␣ highly expresses in the liver, intestine, kidney, and adrenals (2,3), and its modulator has revealed promising therapeutic potential against metabolic disorders. In most cases FXR␣ forms a heterodimer with retinoid X receptor ␣ (RXR␣) to modulate the expression of its target genes. Primary bile acids such as chenodeoxycholic acid (CDCA) and cholic acid (CA) have been identified as the endogenous ligands of this receptor (4). Bile acids as signaling molecules activate not only FXR␣ but also a cluster of receptors, such as the pregnane X receptor, vitamin D receptor, the constitutive androstane receptor, and the bile acid receptor TGR5 (5), and are toxic to cells at high concentrations. Thus, discovery of specific non-bile acid modulator of FXR␣ is expected to be valuable in the regulation of the diverse metabolic pathways for treating the related metabolic disorders.
In recent years, FXR␣ agonists have received much attention for the beneficial effects of FXR␣ activation in lowering plasma lipogenesis, repressing very low density lipoprotein production, increasing plasma triglyceride clearance, improving insulin sensitivity, and promoting the storage of glycogen (6). To date increasing numbers of FXR␣ agonists have been discovered, such as GW4064, fexaramine, 6a-ethyl-CDCA, MFA-1, XL335, and ivermectin (7)(8)(9)(10)(11)(12). However, it was determined that FXR␣ activation by full agonist may also cause undesired side effects in animals. For example, long term treatment of GW4064 even accelerated obesity and exacerbated diabetes in diet-induced obesity mice through reduction of bile acid pool (13). In addition, agonists such as bile acids, GW4064, and XL335 reduced high density lipoprotein cholesterol (14,15). By contrast, FXR␣ blockage may find its beneficial effect on the improvement of hypercholesterolemia. For example, the natural product guggulsterone extracted from the guggul tree as an FXR␣ antagonist FXR␣-dependently decreased hepatic cholesterol in mice fed with a high cholesterol diet (16), and the nonsteroidal FXR␣ antagonist 12u also functioned in lowering hepatic cholesterol in the high cholesterol diet-induced C57BL/6 mice (17). Moreover, the results from FXR␣-deficiency mice confirmed that FXR␣ suppression showed potential effects on the improvement of obesity and diabetes (18 -20). Additionally, AMP-activated protein kinase (AMPK) was recently identified as a co-regulator of FXR␣, and AMPK activator metformin as an anti-diabetic agent showed potently antagonistic effect on FXR␣ (21). It is thus believed that discovery of selective antagonist of FXR␣ will render special interests in anti-metabolic disorders research.
In structure, similar to most of the other nuclear receptors (1), FXR␣ is divided into several major domains: an N-terminal region containing a ligand-independent transactivation function domain 1 (AF1), a highly conserved DNA binding domain (DBD), and a hinge region linking DBD to the C-terminal LBD that also contains a strong transactivation function domain 2. To date, dozens of the structures of FXR␣-LBD in complex with different agonists have been determined (7,10,12,22). The results indicated that in the active conformation of FXR␣, FXR␣-LBD folds into a canonical three-layer helical sandwich that embeds a hydrophobic pocket for ligand binding, and function domain 2 interacts with the coactivator, including steroid receptor coactivator-1 (SRC-1), SRC-2, SRC-3, and peroxisome proliferator-activated receptor ␥ coactivator-1␣ (PGC-1␣). However, no data have as yet been obtained concerning the molecular mechanism of FXR␣ antagonism due to the lack of the structural information of either unliganded FXR␣ (apoLBD) or antagonist-bound FXR␣.
In the current work we determined the small molecule Fig. 1A) as a selective antagonist of human FXR␣ (hFXR␣), and the crystal structure of hFXR␣-LBD in complex with NDB was successfully analyzed. It was discovered that the binding of NDB caused an unusual rearrangement of helix 11 (H11) and H12 (AF-2) of the receptor by forming a homodimer of hFXR␣-LBD. The binding details between two hFXR␣-LBD monomers were also supported by the mutation analysis and co-repressor binding assay. Moreover, animal-based assays have highlighted the potential of NDB in the prevention of glucose and lipid metabolism dysfunction. It is expected that our first analyzed crystal structure of hFXR␣-LBD binding to its antagonist NDB may help to largely expound the antagonistic mechanism of the receptor, and NDB will find its potential as a lead compound in antidiabetes research.
Animal Experiments-Male C57BL/6J db/db mice at 8 weeks of age were fed with normal diet and intraperitoneally injected with either vehicle or NDB (24 mg/kg) once a day for 4 weeks. The animals were then killed after fasting overnight, and mice livers were collected and frozen in liquid nitrogen for real-time PCR.
Cell Cultures-The human hepatoma HepG2 and HEK293T cells were obtained from ATCC (Manassas, VA). Primary mouse hepatocytes were isolated from male C57/BL6 mice in 8 -12 weeks of age (Shanghai SLAC Laboratory Animal Co. Ltd) using a two-step collagenase perfusion and low speed centrifugation (23). Cells were washed and suspended in Williams' E medium supplemented with 10% FBS, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C in a 5% humidified CO 2 incubator and allowed to attach for 4 -6 h. The unattached cells were then removed followed by medium changing to fresh DMEM containing 10% FBS, 100 units/ml penicillin, and 100 g/ml streptomycin with the addition of indicated concentrations of compounds.
Protein Preparations-Residues 244 -476 of hFXR␣-LBD were subcloned into Nde1/XhoI sites of pET15b to generate N-terminal His-hFXR␣-LBD protein expression vector. His-hFXR␣-LBD for AlphaScreen assay was purified with a nickel-nitrilotriacetic acid column followed by a Superdex 75 column according to the published approach (24). To obtain enough stable protein for crystallization, the plasmid of mutant hFXR␣-LBD-C436E/C470E was constructed according to the previously published method (24). The mutant protein was purified by the method similar to that for the wild type His-hFXR␣-LBD protein, with the tag removed by incubation with thrombin overnight at 4°C.
Surface Plasmon Resonance (SPR) Technology-based Assay-Binding of NDB to hFXR␣-LBD was investigated by using an SPR technology-based Biacore T200 instrument (GE Healthcare). Purified hFXR␣-LBD protein was immobilized on Series S sensor chip CM5 by the standard primary amine coupling reaction followed by injection of different concentrations of NDB to the chip.
All experiments were carried out at 25°C with HBS-EP (10 mM Hepes, 150 mM NaCl, 3 mM EDTA, and 0.005% surfactant P20, pH 7.5) as the running buffer at a constant flow of 30 l/min. The equilibrium dissociation constant (K D ) was achieved by fitting the data using 1:1 binding model based on the Biacore T200 Evaluation Software Version 1.0.
Fluorescence Quenching Analysis-Fluorescence quenching analysis was determined according to our published approach (24). All fluorescent spectra were recorded on F-2500 fluorescence spectrophotometer (Hitachi) equipped with 1.0-cm quartz cells. 1 M hFXR␣-LBD was mixed with different concentrations of NDB for 30 min. The sample was then placed in a quartz cell and kept for 5 min in the dark before measurement. All final spectra were corrected by deducting the buffer contribution, and the obtained results were the average of three parallel measurements.
AlphaScreen-based Protein-Peptide Interaction Assay-The effect of NDB on the interaction of co-factor peptide with hFXR␣-LBD was investigated by an AlphaScreen technologybased assay. Briefly, the experiment was conducted with 100 nM His-hFXR␣-LBD and ϳ30 -100 nM biotinylated co-factor peptide in the buffer containing 25 mM Hepes, 100 mM NaCl, and 0.1% BSA, pH 7.4. The mixture was incubated in the dark at room temperature followed by a fluorescence measurement on an Envision microplate analyzer (PerkinElmer Life Sciences). The binding signals were detected with NDB combined with GW4064 (0.5 M) or CDCA (50 M) and with GW4064 (0.5 M) or CDCA (50 M) alone as a positive control and DMSO as a negative control. All values were reported as the mean Ϯ S.D. of triple measurements, and ZЈ-factor was used to validate the assay.
Luciferase Reporter Assay-HepG2 cells plated to 40 -50% confluence in 48-well plate in DMEM were cultured overnight. Transient transfection was conducted by Lipofectamine 2000 (Life Technologies) according to the instructions of the manufacturer. In the transactivation system, 100 ng of the plasmids of pcDNA3.1-RXR␣, pcDNA3.1-FXR␣, pGL3-FXRE-Luc, and 50 ng of pRL-SV40 were transfected into the cells. After transfection for 6 h, the medium was changed into DMEM and incubated with different concentrations of NDB overnight. In the assay of the selectivity of NDB over the other tested nuclear receptors, the corresponding plasmids were transfected into HepG2 cells and incubated with the indicated compounds for 24 h. The firefly and renilla luciferase activities were assayed with Dual-Luciferase Reporter assay system (Promega). The firefly luminescence was normalized based on the Renilla luminescence signal, and the ratio of treatment over control is expressed as -fold activation.
RNA Isolation and Quantitative Real-time PCR-The total RNA in primary mice hepatocytes and the livers of db/db mice were isolated with TRIzol reagent (Invitrogen), and 1 g of total RNA was reverse-transcribed into cDNA using PrimeScript RT reagent kit. mRNA levels of BSEP, SHP, phosphoenolpyruvate carboxykinase (PEPCK), glucose 6-phosphatase (G6-pase), and GAPDH were quantified by quantitative real-time-PCR using specific primers. mRNA levels were all normalized to GAPDH mRNA. The primers of genes were listed in Table 1. All values were reported as the mean Ϯ S.D. of triple measurements of each cDNA sample, and the thermal cycling conditions were 95°C for 10 min followed by 43 cycles of 30 s at 95°C, 30 s at 55°C, and 30 s at 72°C.
Crystallization and Data Analysis-For preparation of hFXR␣-LBD⅐NDB complex, a 5-fold molar excess of NDB was added into hFXR␣-LBD and concentrated to 5 mg/ml. hFXR␣-LBD⅐NDB crystals were grown at 4°C in hanging drops containing 1 l of protein-ligand solutions and 1 l of well buffer (22% polyethylene glycol 2000MME, 0.2 M NaCl, 0.1 M MES, pH 5.9). Crystals grew up within 1 week. Diffraction data were collected at BL17U of Shanghai Synchrotron Radiation Facility (China) and integrated with HKL2000. The structure was determined by molecular replacement methods (MOLREP) with CCP4 using the structure of FXR␣-LBD⅐XL335 complex (PBD code 3FLI) as the initial model. Phasing and refinement were carried out with Refmac5 in CCP4. Model building was manually performed with COOT. The statistics of the structure and data sets are summarized in Table 2. Molecular images were made with PyMOL. Atomic coordinates and structure factors of hFXR␣-LBD⅐NDB have been deposited to Protein Data Bank under accession code 4OIV.
Mutation Analysis-The mutants (S457A and W458F) of FXR␣-LBD and full-length hFXR␣ were prepared with the fast mutagenesis kit (Transgene Biotech) using WT-FXR␣-LBD and WT-FXR␣ as templates. All mutants were validated by sequencing.
Circular Dichroism Spectroscopy-Secondary structural changes in the mutants of FXR-LBD were monitored using a thermostated Applied Photophysics Chirascan spectrophotometer (Applied Photophysics). A high quality quartz cell with 1-mm path length was used. Protein samples were prepared in the solution of 50 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 2 mM KH 2 PO 4 . Far-UV CD spectra were collected from 200 to 280 nm. Experimental data were the average values of three measurements and corrected by subtracting the blank obtained under the same conditions in the absence of protein.
Time-resolved Fluorescence Resonance Energy Transfer (TR-FRET) Assay-TR-FRET assay was performed according to the published approach (25). HEK293T cells were plated in a 6-well plate and cultured overnight in DMEM. Transfection was performed by Lipofectamine 2000 according to the instructions of

Results
NDB Has a Binding Affinity to hFXR␣-In the discovery of FXR␣ ligands from our in-house laboratory compound library, SPR technology-based compound binding affinity assay was at first carried out, and the small molecule active compound NDB (Fig. 1A) was thus determined. To quantitatively evaluate the binding feature of NDB to hFXR␣-LBD, the association (k on ) and dissociation (k off ) rate constants and the dissociation equilibrium constant (K D ) were determined by fitting the sensorgrams with a 1:1 binding fitting model. As demonstrated in Fig.  1B, NDB bound to hFXR␣-LBD with a k on of 3.63Eϩ4 M Ϫ1 S Ϫ1 , k off of 0.301s Ϫ1 , and K D of 8.29 M. Given the two contained tryptophan residues of Trp-454 and Trp-469 in hFXR-LBD, an intrinsic fluorescence-quenching based assay was also performed to further confirm the binding of NDB to hFXR␣-LBD (24). In the assay, different concentrations of NDB were incubated with hFXR␣-LBD (1 M), and intrinsic fluorescence spectra were detected. As shown in Fig. 1C, hFXR␣-LBD displayed the maximal fluorescence at 337 nm, and treatment of NDB resulted in fluorescence quenching of hFXR␣-LBD in a dosedependent manner, which indicated NDB binding to hFXR␣-LBD. As shown by the dashed line, NDB exhibited no fluorescence absorption at this wavelength.
NDB Is a Selective Antagonist of hFXR␣-As NDB has been determined to be the ligand of hFXR␣, we next explored the potential agonistic or antagonistic ability of NDB against hFXR␣ by an AlphaScreen-based assay. As shown in Fig. 2A, in the presence of a synthetic agonist GW4064 (GW, 0.5 M) or a natural agonist CDCA (50 M) of hFXR␣, the peptide containing the coactivator SRC-1 LXXLL binding motif was specifically recruited to hFXR␣-LBD, and NDB (25 M) suppressed the corresponding agonistic effect induced by GW4064 or CDCA. NDB alone exhibited no agonistic activity on hFXR␣ coactivator recruitment (data not shown). Similarly, the results shown in Fig. 2, B-D, demonstrated that NDB also exhibited antagonistic activities against GW4064 or CDCA in the activation of hFXR␣-LBD⅐SRC-2, hFXR␣-LBD⅐SRC-3, and hFXR␣-LBD⅐PGC-1␣ interactions. As a representative, shown in Therefore, all of the above results indicated that NDB functioned as an antagonist of hFXR␣. Next, a luciferase report assay was further carried out to validate the antagonistic activity of NDB against hFXR␣ in HepG2 cells. Given that binding of FXR␣/RXR␣ heterodimer to FXR␣ response element (FXRE) is required for activation of the downstream gene transcription, we examined the effect of NDB on the agonist-induced transactivation of FXR␣/RXR␣ in the cells. It was observed that NDB antagonized either the CDCA (50 M)or GW4064 (50 nM)activated transactivation activity with IC 50 at 4.5 or 2.6 M (Fig.  3, A and B). These results thereby confirmed the antagonistic activity of NDB against hFXR␣. In the investigation of the selectivity of NDB against hFXR␣, we examined the transactivation activities of NDB toward a group of nuclear receptors, including liver X receptor ␣/␤ (LXR␣/␤), Mineralocorticoid receptor (MR), RXR␣, peroxisome proliferatoractivated receptor ␥ (PPAR␥), progesterone receptor B (PR-B), and estrogen receptor ␣/␤ (ER␣/␤). As indicated in Fig. 3C, NDB had no effects on the transactivation activities of these tested nuclear receptors, thereby implying that NDB is a selective antagonist of hFXR␣. NDB Modulated FXR␣ Target Genes SHP and BSEP-As NDB has been determined as a selective antagonist of FXR␣, we next evaluated the potential regulation of NDB against the key FXR␣ target genes of SHP and BSEP in mouse primary hepatocytes by quantitative RT-PCR. In the assay the primary hepato-cytes were treated with different concentrations of NDB in combination with GW4064 (GW, 10 M). As shown in Fig. 4, A and B, NDB could effectively reverse the GW4064-induced stimulation of either BSEP or SHP mRNA expression. This result thus revealed the antagonistic effects of NDB on FXR␣ target genes.  Crystal Structure of hFXR␣-LBD⅐NDB Revealed the Homodimeric Mode of the Receptor-The crystal structure of hFXR␣-LBD⅐NDB complex was determined at a resolution of 1.7 Å in the P4 1 2 1 2 space group by using FXR␣-LBD⅐XL335 (PBD code 3FLI) structure as the model for molecular replacement. The statistics of data collection and structure refinement were summarized in Table 2. The overall structure of hFXR␣-LBD⅐NDB complex was shown in Fig. 5, A and B. As indicated, the complex was in a homodimeric mode formed by two hFXR␣-LBD monomers in the asymmetric unit, with either monomer consisting of an hFXR␣-LBD molecule and one NDB (Fig. 5, A and  B). The buried surface area formed by the two monomers was calculatedat1818.7Å 2 (calculatedwithPDBePISA).Eithermonomer contained 12 helices and 1 ␤ strand (denoted as S1 and S1Ј) (Fig. 5B), which is different from the active conformation that adopts a 12 ␣-helix bundle (7,10). It is noted that the backbones of Asn-461 and Arg-459 in S1 formed two hydrogen bonds with the corresponding backbones in S1Ј from the other monomer (Asn-461Ј and Arg-459Ј, Fig. 5C). At the joint of the antiparallel ␤ sheet, residue Ser-457 (Ser-457Ј) formed H-bonds with both His-463Ј (His-463) of the backbone and the side chain Asp-462Ј (Asp-462) of the adjacent monomer, and Trp-458 (Trp-458Ј) formed an H-bond with Glu-286Ј (Glu-286) in H3Ј (H3) of the other monomer (Fig. 5D). In addition, hydrophobic interac-  NDB Binding Pocket Analysis-In the structure NDB was completely enclosed within hFXR␣-LBD (Fig. 6A), and 2F o Ϫ F c map in the ligand binding pocket (LBP) of hFXR␣-LBD⅐NDB structure revealed a clear electron density of NDB (Fig. 6A). As indicated in Fig. 6B, NDB bound to the LBP of hFXR␣-LBD by interacting with a cluster of hydrophobic residues, including Trp-458, Thr-292, Met-456, Leu-455, Ala-452, Phe-333, Ala-295, Ile-366, Met-332, Met-369, Leu-291, Ser-336, His-298, and Tyr-373, and forming one hydrogen bond through Tyr-365 of the backbone in H8 and two bridged-water molecules.
Superposition of the structures of hFXR␣-LBD⅐NDB and hFXR␣-LBD⅐GW4064 (PDB code 3DCT) revealed that both NDB and GW4064 occupied a similar binding pocket (Fig. 6C), although NDB took a smaller size in the LBP compared with GW4064 for the shorter molecule of NDB than GW4064 (Fig.  6D). Such a similar LBP occupation for NDB and GW4064 might suggest that the antagonistic effect of NDB possibly results from the competitive inhibition of NDB against the agonists, including GW4064. In comparison with the hFXR␣-LBD⅐GW4064 structure (Fig. 6C), the hydrophobic interactions of NDB within the LBP made residues of Leu-455Ј and Met-456Ј closer to the compound thus causing severe conformation changes for H11Ј and H12Ј, with new formations of ␤ strand (S1Ј) and two small helixes (H112Ј and H122Ј) (Fig. 6C), whereas the new formed H112Ј at the end of H11Ј moved closely to and sealed the LBP (Fig. 6A). In the structure, interaction of S1Ј in one monomer with S1 in the other forms an antiparallel ␤ sheet as a bridge to send the small helix His-122Ј to the co-factor binding domain of the other monomer (Fig.  5B).
Mutation Assay and Co-repressor Binding Analysis Supported the Dimerization of hFXR␣-LBD⅐NDB Complex-As determined in the crystal structure analysis, residues Arg-459, Asn-461, Ser-457, and Trp-458 played potent roles in the homodimerization of hFXR␣-LBD⅐NDB complex (Fig. 5, C and  D). Thus, a mutation assay was next performed trying to validate the dimerization interface. In the assay, considering that residues Asn-461 and Arg-459 are within the backbone of one monomer (Fig. 5C), only mutagenesis of S457A and W458F was designed. First, the structure changes of hFXR␣-LBD mutants were detected by CD (Fig. 7A). We found that the structure of the mutant Trp-458 made a rearrangement nevertheless mainly adopting ␣-helix conformation similar to those of WT-FXR␣-LBD and the mutant Ser-457. The results indicated that Trp-458 played an important role in hFXR␣-LBD conformation. Next, we also identified the transcriptional activities of FXR mutants induced by GW4064 alone or GW4064 with NDB using luciferase reporter assay. As shown in Fig. 7B, GW4064 increased the transcriptional activation of either mutant, and the capability of NDB in antagonizing the GW4064-induced transactivation activity against the mutant S457A or W458F was obviously reduced (from 100% to 93 and 100% to 77%, respectively) (Fig. 7C). Collectively, these results thereby supported that the residue of Ser-457 actually was important in hFXR␣-LBD dimerization, and residue Trp-458 was significant for both hFXR␣-LBD conformation and dimerization. As demonstrated in Fig. 7D, His-122Ј of one monomer was recruited by the groove formed by H3 and H5 of the other monomer, involving a group of H-bonds formed by Leu-469Ј of His-122Ј with Lys-307 of H3, Glu-470Ј of His-122Ј with His-317 of H5, and Glu-471Ј of His-122Ј with Arg-459 of S1. These data may imply the inactivity of the homodimer in accommodation of the cofactors. Actually, the results in Fig. 2, A-D, have already indicated that NDB inhibited co-activator binding to hFXR␣-LBD. As also shown in the AlphaScreen assay results (Fig. 7E), the co-repressor NCoR-2 was not recruited by hFXR␣-LBD, whereas as a positive control, ivermectin as an FXR␣ ligand was able to induce hFXR␣-LBD to recruit NCoR-2 (11). In addition, the classical antagonist guggulsterone also showed similar effect to NDB. Therefore, all the above-mentioned mutation assay and co-repressor binding analyses have supported the dimeric mode of hFXR␣-LBD⅐NDB complex. Taken together, our results thereby suggested that NDB exerted antagonistic activity against hFXR␣-LBD by inducing the formation of the auto-repressed conformation of the receptor. As indicated in Fig. 8A, GW4064 as an FXR␣ agonist promotes the binding of FXR␣/RXR␣ heterodimer to FXRE in the regulatory regions of target genes for initiation of FXR␣ target gene expression. Binding of NDB to FXR␣ causes FXR␣ homodimer formation leading to the release of co-activators and furthers the repression of the target genes. To confirm the model, we performed the TR-FRET assay, attempting to identify the effect of NDB on the disruption of FXR␣/RXR␣ interaction. As shown in Fig. 8B, FXR␣ and RXR␣ could form a heterodimer compared with FXR␣ or RXR␣ alone in the cell lysate. In the presence of GW4064 (5 M), the interaction of FXR␣ with RXR␣ was greatly increased, whereas NDB suppressed the interaction induced by GW4064 in a dose-dependent manner. Therefore, these results further confirmed the mechanism for NDB antagonizing FXR␣ function.
NDB Modulated the Key Genes Related to Metabolic Processes in the Liver of db/db Mice-Given that FXR␣ antagonists have revealed their great potential in the treatment of metabolic diseases, we next investigated whether NDB as a selective FXR␣ antagonist could improve metabolic dysfunctions in vivo. In the assay, db/db mice were treated with vehicle or 24 mg/kg NDB for 4 weeks. After the animals were killed, we detected the levels of the metabolism-related genes in the liver of the mice, including the gluconeogenesis-related genes phosphoenolpyruvate carboxykinase and glucose 6-phosphatase and the bile acid-related genes SHP and BSEP. As shown in Fig. 9, A-D, administration of NDB evidently depressed these four genes in vivo. Therefore, our results suggested that FXR␣ antagonist NDB was effective in regulating FXR␣ activity and exhibited potential in improving metabolic dysfunctions in vivo. FIGURE 6. Investigation of NDB binding details to hFXR␣-LBD. A, the ribbon diagram shows the structure of NDB in complex with monomer with indication of helixes H11Ј, H112Ј, S1Ј, and His-122Ј. The atoms of NDB are colored in yellow (carbon atoms), red (oxygen atoms), blue (nitrogen atoms), and green (chlorine atoms). The electron density map for NDB is shown as a blue mesh contoured at 1. B, Ligplots showed the related interaction residues within hFXR␣-LBD⅐NDB complex based on the crystal structure. The residues were generated using LigPlot program. Hydrophobic interaction amino acids are shown in red, whereas the hydrogen interaction amino acid (Tyr-365) interacting strongly with the two water molecules (Hoh11 and Hoh24) is shown in black sticks. H-bonds are indicated with green dashed lines. C, superposition of hFXR␣-LBD⅐NDB (hFXR␣-LBD in gray, NDB in yellow) with hFXR␣-LBD⅐GW4064 (PDB code 3DCT) (FXR␣-LBD in violet, co-activator peptide in pink, and GW4064 in magenta) illustrating the conformation difference between agonistic and antagonistic state of hFXR␣-LBD. Both the agonistic and antagonistic ligands occupied the same binding site in LBP. D, superposition of NDB (yellow) and GW4064 (magenta).

Discussion
In the current work we determined the small molecule NDB as a selective antagonist of hFXR␣ and successfully analyzed the crystal structure of hFXR␣-LBD⅐NDB complex. Based on the structure, a new potential antagonistic mechanism of FXR␣ has been suggested. Previously, the reported results demonstrated that FXR␣ antagonists such as guggulsterone (26) and compound 1 (27) stabilized the complex of co-repressor NCoR with FXR␣ on the promoters, thus failing to recruit co-activators to exert agonistic effects. However, NDB exerted its distinct antagonistic activity against FXR␣ compared with these two reported FXR antagonists. It was found that NDB at first competitively occupied the same LBP as the agonists, causing the hydrophobic part of H11 bended to the LBP to stabilize this inactive conformation of LBD, which seems to be common in other nuclear receptors such as RXR␣-LBD and photoreceptor cell-specific nuclear receptor gene (PNRLBD) (28), and H12 protruded and occupied the co-activators binding site to inhibit their bindings (Fig. 2,  A-E). Next, we found that NDB rendered no effect on the interaction of co-repressor NCoR-2 with hFXR␣-LBD (Fig.  7E), which demonstrated that NDB possessed the distinct antagonistic activity compared with guggulsterone and com-pound 1. Based on the crystal structure, we inferred that NDB antagonized FXR␣ function by inducing FXR␣ into homodimeric mode and further decreasing the conformation of FXR␣/RXR␣ heterodimer (Fig. 8A). Next, we attempted to assay the effect of NDB on FXR homodimer using TR-FRET. Unfortunately, we failed to find the FXR homodimer induced by NDB. We tentatively suggested that the transformation between monomer and homodimer of FXR is dynamic and the conformation of homodimer is physiologically weak, although we could observe the inhibitory effect of NDB on FXR dimerization by mutation and reporter assay. In addition it is suggested that the dimeric inactive state of hFXR␣-LBD might apply to the other FXR species, as the amino acid sequence in dimer interface is highly conserved in different species (Fig. 10).
Structural comparison of hFXR␣-LBD⅐GW4064 with hFXR␣-LBD⅐NDB has revealed the potency of either H11 or H12 in the dynamic equilibrium between the agonistic and antagonistic conformation of hFXR␣-LBD. As demonstrated in hFXR␣-LBD⅐NDB structure, upon NDB binding the extra space of LBP was fully filled by the hydrophobic end of H11, leading to the LBP inaccessible for other ligand binding, whereas the new formed H112 would be assumed to stabilize NDB binding and lock NDB in the LBP. Therefore, His-122 seems to play an important role in stabilizing the antagonistic conformation of the receptor.
As also demonstrated from hFXR␣-LBD⅐NDB structure, the co-factor binding site was occupied by His-122 itself as seen in passive antagonist conformation. Such passive antagonist con-  formation may not be unique to NDB binding to FXR␣. For example, in the domain C of GR-LBD, H12 also occupied the co-factor binding site thus preventing the binding of GR interacting co-repressor and co-activators (29). In addition, the example of specific passive antagonist of androgen receptor to treat advance prostate cancer revealed that a passive antagonist might be superior to an active antagonist in some cases (29,30). Therefore, such a passive antagonist might bring a new thought for FXR␣-targeted agent discovery.
Currently, several reports have been published regarding other nuclear receptors forming homodimers. For example, vitamin D receptor keeps stable homodimers after binding to the coactivators and increases the ligand-induced vitamin D receptor transcription on different transaction elements (31). Thyroid hormone receptor and orphan nuclear receptor RevErb can form homodimers by interacting with N-CoR via their specific amino acid residues, consequently mediating the repression functions (32). Moreover, RXR␣-LBD exhibits a tetrameric conformation, whereas residues Phe-437 and Phe-438 of helix 11 play potently in the tetramer stabilization of SMRT-induced RXR rearrangement (33). Although the nuclear receptors adopt the homodimer conformation, their mechanisms and functions might be totally different.
Previously, the study on FXR␣ response to a fasting-refeeding high carbohydrate diet showed that FXR␣ deletion displayed a more accelerated response through repression on gluconeogenic genes, such as glucose 6-phosphatase and phosphoenolpyruvate carboxykinase. In our work we detected that NDB could also regulate the glucose metabolism in vivo by reducing the expression of gluconeogenic genes including phosphoenolpyruvate carboxykinase and glucose 6-phosphatase (20). Our results have provided additional understanding of pharmacological functions for FXR␣ antagonist as an antidiabetes agent.
In conclusion, we discovered a new selective FXR␣ antagonist NDB that is effective in modulating transcription of FXR␣ downstream genes. The analyzed crystal structure of hFXR␣-LBD in complex with NDB revealed the homodimeric mode of hFXR␣-LBD⅐NDB, and the potential antagonistic mechanism of NDB against the receptor was proposed. It is expected that our results may help better understand the mechanism of FXR␣ antagonism, thereby providing a new framework for developing novel FXR␣-targeted therapeutic agents to treat metabolic disease.