Farnesoid X Receptor Modulation by Systematic Ligands of Translocator Protein (18KDA)

The synthesis of bile acids is the major biological mechanism for cholesterol removal in the human body. Strict regulation of both cholesterol and bile acid levels is necessary to maintain a healthy balance and to prevent health problems. Bile acids are natural ligands for famesoid x receptor (FXR), a nuclear receptor that controls gene expression for multiple proteins involved in maintenance of bile acid homeostasis. Many endogenous and exogenous chemical ligands have been found to activate FXR; chenodeoxycholic acid (CDCA) is the most well characterized endogenous ligand. This study identifies a synthetic indole-acetamide, FGIN-1-27, as a new FXR agonist. FGIN-1-27 is already a known ligand of the translocator protein 18 kDa (TSPO), a mitochondrial cholesterol transporter. FXR regulates target gene transcription through binding to special inverted repeat-I (IR-1) consensus DNA elements. Ligand binding to FXR was measured by inserting an IR-1 sequence upstream of a firefly luciferase detector gene that increased transcription of luciferase pr:oportional to ligand binding in a human hepatoma cell line (HuH-7). Results show that FGIN-1-27 is a partial agonist of FXR that activates FXR alone at 10 μM, but decreases activation from CDCA at 100 μM when cotreated. Two other well-known ligands of TSPO, FGIN-1-43 and PKl 1195 were investigated also for their effects on FXR mediated transcription. Both compounds acted as antagonists, decreasing the activity of CDCA (100 μM) while showing no activation of FXR alone at 1 O μM treatment. Agonist ligand binding to FXR increases the expression of the target gene, bile salt export pump (BSEP), and another nuclear receptor, small heterodimer partner (SHP). Through real time RT-PCR DNA amplification of both genes, we found FGIN-1-27 treatment in HuH-7 cells and primary human hepatocytes increased both BSEP and SHP gene expression. Additionally, expression of cholesterol 7a-hydroxylase (CYP7 Al), an enzyme involved in bile acid synthesis, is negatively regulated by FXR; we show that FGIN-1-27 decreased the expression ofCYP7Al. In addition to in vitro studies, we investigated in silica molecular modeling of the binding of these TSPO ligands to FXR and demonstrated that these synthetic compounds fit into the ligand-binding pocket of FXR with favorable energy measurements. We identified key amino acids involved in agonist ligand binding in silica, and through mutation assays we confirmed that H447 is the major amino acid responsible for FXR interaction with an agonist ligand. Taken together, FGIN-1-27 binding to and modulating two of the proteins involved in bile acid synthesis indicates there is overlap in the role of TSPO and FXR. FGIN-1-27 and related indole-acetamides may be potential therapeutic drugs beneficial to populations with enzyme deficiencies that cause high cholesterol levels. Further investigation of the role of mitochondria in bile acid synthesis will lead to a better understanding of the regulation of cholesterol and bile acid homeostasis.

Agonist ligand binding to FXR increases the expression of the target gene, bile salt export pump (BSEP), and another nuclear receptor, small heterodimer partner (SHP).
Through real time RT-PCR DNA amplification of both genes, we found FGIN-1-27 treatment in HuH-7 cells and primary human hepatocytes increased both BSEP and SHP gene expression. Additionally, expression of cholesterol 7a-hydroxylase (CYP7 Al), an enzyme involved in bile acid synthesis, is negatively regulated by FXR; we show that FGIN-1-27 decreased the expression ofCYP7Al.
In addition to in vitro studies, we investigated in silica molecular modeling of the binding of these TSPO ligands to FXR and demonstrated that these synthetic compounds fit into the ligand-binding pocket of FXR with favorable energy measurements. We identified key amino acids involved in agonist ligand binding in silica, and through mutation assays we confirmed that H447 is the major amino acid responsible for FXR interaction with an agonist ligand.
Taken together, FGIN-1-27 binding to and modulating two of the proteins involved in bile acid synthesis indicates there is overlap in the role of  and related indole-acetamides may be potential therapeutic drugs beneficial to populations with enzyme deficiencies that cause high cholesterol levels. Further investigation of the role of mitochondria in bile acid synthesis will lead to a better understanding of the regulation of cholesterol and bile acid homeostasis.   Cholesterol homeostasis must be maintained i~ the brain as in other tissues, but cholesterol cannot readily cross the blood brain barrier. To overcome this problem, CYP46Al produces 24S-hydroxycholesterol, an oxysterol that can cross the blood brain barrier, and be further converted into bile acids in the liver via the oxysterol 7ahydroxylase (CYP39Al) (20). The CYP27Al-initiated pathway forms predominantly 27-hydroxycholesterol, which is 7a-hydroxylated by CYP7Bl (21). CYP27Al is involved also in both pathways further downstream in ring modifications to oxidize and cleave the sterol side chain (1,19). The alternative pathway produces solely chenodeoxycholic acid (CDCA), while the classic pathway produces both CDCA and cholic acid (CA) (8). The relative abundance of CA versus CDCA is ultimately regulated by sterol 12a-hydroxylase (CYP8Bl) (7). Figure 1  concentrations accumulating in plasma, and accumulation of cholesterol in the liver with limited bile acid synthesis or excretion (22)(23)(24). Studies have shown that 40-60% of the Caucasian North American population are carriers of an A to C substitution polymorphism in the CYP7Al promoter region producing a high LDL-C phenotype with a recessive CYP7 Al -/-mutation that is more prevalent in men (23,25,26). Some individuals with this substitution have been shown to be resistant to cholesterollowering 3-hydroxy-3methylglutaryl-coenzyme A (HMG-CoA) reductase inhibitors and experienced premature gallstones from bile acid accumulation (23). It is hypothesized that the inhibition of cholesterol synthesis by HMG-CoA reductase inhibitors would not significantly decrease total cytosolic cholesterol concentrations, therefore the increase in LDL receptor expression normally resulting from inhibition 4 of cholesterol synthesis, thus decreasing LDL-C in blood, would be limiting with little effect on lowering LDL-C concentrations (23).
Lower than normal CA concentration is often an indicator of CYP7 Al deficiency as a result of the classic pathway being absent with compensation by the acidic pathway that produces only CDCA (22); typically the ratio of CA to CDCA is 2:1 (7,22). In the event of this compensation, CYP27Al activity doubles preventing complete deficiency of CYP7 Al from being lethal (23) The acidic pathway is limited not by the initial enzyme CYP27A1 itself, but by the delivery of cholesterol to CYP27 Al in the mitochondria (29). The transport of cholesterol into the mitochondria occurs through the translocator protein 18 kDa (TSPO), which is located predominantly on the outer mitochondrial membrane (OMM) (30,31) in cells of the adrenal glands, lung, heart, liver, and multiple other tissues (32). TSPO possesses five membrane-spanning domains that can form 5 multimeric protein polymers able to bind endogenous ligands that facilitate cholesterol binding (31,33). The polymer formation is facilitated by reactive oxygen species most likely produced from the CYP enzyme activity inside the mitochondria (31 ). TSPO, however, does not act alone and requires the assistance of steroidogenic acute regulatory (STAR) protein (34) (Fig. 2). Through a complex pathway, cholesterol binds to STAR in the cytoplasm for transport to the mitochondria (35) (37)(38)(39). Both endogenous and exogenous ligands increase 27-hydroxycholesterol production, identifying the availability of cholesterol to CYP27Al as the rate-limiting step in the alternative pathway (40).

Role of nuclear receptors
In high concentrations, bile acids can be toxic, so the potential toxicity is regulated by negative feedback ( 41 ). Bile acid synthesis is reduced in the presence of high bile acid concentrations, and conversely, low concentrations result in bile acid synthesis activation to increase the bile acid pool (1). Bile acid feedback is regulated by nuclear receptors that directly control target genes by activating or repressing transcriptional activities. Typically, nuclear receptors have a DNA binding domain that recognizes specific DNA sequences (hormone response elements) through a zinc finger region, and a ligand-binding domain (LBD). The response elements are comprised of halfsites at least 6 base pairs long (typically AGGTCA) (42,43). Nuclear receptors bind to e elements as either a homo/heterodimer to sequences of direct (DR) (-7-7), res pons everted (ER) (~-7) or inverted (IR) (-7~) repeats spaced by 1-5 nucleotides, or as a monomer, binding only to a half site ( 44 ). Helix 12 of the LBD is a ligand dependent activation function-2 (AF-2) domain, which upon agonist ligand binding to the receptor will recruit a coactivator protein with acetyltransferase activity ( 43).
Acetylation of residues on histone proteins causes relaxation of the chromatin structure so the transcriptional machinery can gain access to the DNA to increase gene transcription ( 45).
The endogenous bile acid receptor, famesoid x receptor (FXR; NR1H4), a member of the nuclear receptor superfamily ( 46), is a good potential target for pharmacological therapy to regulate bile acid concentrations, and thus cholesterol concentrations. FXR forms an exclusive heterodimer with retinoid x receptor a (RXRa; NR2Bl) (47) (the heterodimer formation occurs independent of ligand and DNA binding, but it is necessary for FXR receptiveness to bile acid ligand binding (47,48)). Multiple studies show that the primary and secondary bile acids: CDCA, lithocholic acid, and deoxycholic acid, are endogenous ligands of FXR, which in tum regulate bile acid homeostasis through transcriptional effects on specific genes ( 49-51 ). The most potent endogenous ligand of FXR is CDCA (50,51) and CDCA binding to FXR recruits the steroid receptor coactivator-1 (SRC-1) to the LBD ( 46,50,51 ). Along with bile acids, potent exogenous ligands of FXR have been made, including the potent synthetic agonists GW4064 (52), fexaramine (53) AGN29 and AGN31 (54). Additionally, guggulsterone, a compound isolated from the guggul tree traditionally used m Ayurvedic medicine, is a natural antagonist of FXR (55).
Agonist ligands of FXR play a major role in feedback regulation of bile acid synthesis.
In most cases, agonist activation of FXR should lower cholesterol up-take by diminishing the liver bile acid pool through increased efflux and through inhibition of CYP7Al activity, thereby inhibiting cholesterol absorption in the intestine (67). 8 r thl . s is not always the case since FXR agonists would show little effect in Howeve, people lacking functional CYP7Al enzymes. Also, CYP27Al is not a rate-limiting me so bile acids and their intermediates have less effect on this pathway in enzy , comparison to CYP7Al; agonist effects of FXR through bile acids typically do not regulate transcriptional activity of CYP27A1 directly ( 68,69). Individuals with poor synthesis via the classic pathway, therefore, may therapeutically benefit from relevant and useful targets of TSPO as a modulator of the alternative pathway.

TSPO and FXR interplay
It is possible that many of the known ligands of TSPO could additionally regulate the cholesterol turnover rate by acting upon other receptors. Since both mitochondrial and nuclear receptor signaling pathways are involved in maintenance of bile acid homeostasis, this study was designed to investigate the interplay between TSPO and FXR by demonstrating TSPO ligands modulate FXR activity also. PKl 1195, one of the most well known and widely used ligands of TSPO, is known to increase the cholesterol binding rate to the protein (31,70,71). Similarly, a series of 2-aryl-3indoleacetamides (named FGIN-1), designed by Romeo et al. (72), selectively bind to TSPO (73,74). The aim of this study was to investigate binding of PKl 1195 and FGIN-1 compounds to modulate FXR target genes and to provide evidence of binding pocket interactions with these compounds. We have found that FGIN-1-27 is a partial agonist of FXR that activates downstream transcription of FXR target genes, as demonstrated in both optimized luciferase assays and measurements of endogenous gene expression in liver cells. We show that FGIN-1-43 is a selective antagonist, able 9 to block FXR agonist activation of FGIN-1-27 but is less inhibitory of CDCA activation of FXR. PKl 1195, on the other hand, is a non-selective FXR antagonist.
Through in vitro transcriptional and mRNA expression studies and in silica molecular modeling studies we show that each of these compounds binds directly to the FXR LBD. Activation of both TSPO and FXR with one compound is favorable for dual maintenance of bile acid and cholesterol homeostasis. rat ios (µL reagent: µg DNA) were used for each type of transfection: reagent HD (3 ·1) Lipofectamine2000 (3.5:1), or PEI (4:1). Transfections were fuGene · ' perfonned in serum-free media for 6-24 hours. In addition to the reporters, 1.25-2.5 µg of emerald green fluorescent protein (pEGFP-Cl) (Clontech, Mountain View, CA) was transfected into the cells to monitor transfection efficiency. Transfected HuH-7 cells were then trypsinized and re-seeded into a 96-well plate and treated for 24 hours. The 60 GABA-ergic compounds (Table 1) were screened using single-well treatments.

Dual-Glo Luciferase Reporter kit (Promega) on a GloMax 96 Microplate
Luminometer (Promega). The luminescence from the firefly luciferase was normalized to the Renilla luciferase luminescence to control for transfection efficiencies and for well-to-well variation in cell numbers. The ratios of the measurements were calculated and reported as mean fold change relative to DMSO (control) ± standard error of the mean (SEM) (when greater than two replicates were performed).

FXR Mutants
Point mutations of single amino acids of FXR, generously provided by Dr. Ruitang Deng, were formed using a QuikChange site-directed mutagenesis kit (78,79). The mutation sequences are listed in Table 2. ZR-75-1 cells were seeded into 6-well plates and transiently transfected with 1 µg p(FXRE)4-TK-luc, 100 ng pRL-CMV, 200 ng pEGFP-Cl, and 500 ng of 3 .1 ( + ), 3 .1-FXR, or hFXR mutant per well. The cells were trypsinized and each well of a 6-well was re-seeded into a portion of a 96-well plate, treated for 24 hours and read on the Microplate Luminometer, as described in section 2.4.1.

Real-time RT-PCR
HuH-7 cells were seeded into 12-well plates and each well was treated with a different compound for 24 hours. Similarly, the human hepatocytes were obtained in 12-well plates and treated after 4-6 days of routine maintenance. Following the protocol from lnvitrogen/GIBCO, total RNA was harvested from the cells using TRizol reagent. The RNA was reverse transcribed into cDNA using a High Capacity cDNA Reverse 15 . t' n kit (Applied Biosystems, Carlsbad, CA). cDNA was subjected to generranscnp 10 'fi mplification of FXR target genes using SYBR Green PCR Master Mix spec1 1c a (Applied Biosystems). Reactions were made 50 µL at a time with 25 µL of 2x SYBR Green PCR master mix, 21 µL of nuclease free water, 1 µL each of the 10 µM forward and reverse primers, and 2 µL of cDNA (4 ng/µL). The actin, BSEP, CYP7Al, FXR, HNF4a, and LRH-1 primers were from Eurofins MWG Operon (Huntsville, AL), and RXRa primers were from Integrated DNA Technologies (Coralville, IA) (see sequences in Table 3). Each 50 µL reaction was split into 2 wells in a 96-well plate to provide technical replicates. RT-PCR SYBR Green amplification was performed using a 7500 Real-Time PCR System from Applied Biosystems with thermocycling as follows: 2 minutes at 50°C, 10 minutes at 95°C, followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. All values were normalized to actin, due to its ubiquitous and constant expression in all cells, and were reported as fold change relative to mean DMSO (control)± SEM. Applied Biosystems v2.0 SDS software was used for analysis; auto threshold and auto baseline settings were used to ensure that the optimum analysis settings were used for all reactions. Threshold values were set in the exponential phase of the change in normalized reporter dye fluorescence (~Rn).
The baseline of each sample was set to eliminate background noise from the measurements.

Molecular modeling
To further investigate the docking of compounds in the LBD of FXR, in silica molecular modeling with Scripps Research Institute's (La Jolla, CA) AutoDock v4.2 fi ed on CDCA, PKl 1195 and several FGIN-1 compounds. The crystal was per orm f FXR was obtained from the Protein Data Bank (PDB.org) (PDB ID code structure o 3 BEJ) ( 8 0). The structure had 237 residues (amino acids 235-472) of the LBD crystallized with Merck FXR agonist # 1 (MF A-1) that occupied the pocket in an active conformation, with a small fragment of SRC-1 (residues 676-700) that was bound to the AF-2 domain (helix 12) of FXR. The coordinates for residues 235-243 and 472 of FXR were missing and not accounted for.
Using Discovery Studio Visualizer v2.5 (Accelrys, San Diego, CA) the protein crystal structure was inspected and cleaned of any misplaced or misinterpreted atoms. Polar hydrogens were added to each amino acid to complete the valance of each atom (nonpolar hydrogens were implied). The valence of charged amino acids was adjusted so the overall charge on each residue was an integer, e.g. the guanidine group on arginine was given two hydrogens per nitrogen so a neutral charge resulted. Since the protein was only a portion of FXR, the end residues were adjusted to mimic the N-terminal and C-terminal for the purposes of docking. The ·N-terminal nitrogen of Glu244 was allowed two hydrogens to make a + 1 charge and the C-terminal Val4 71 was given a hydroxyl group to complete the carboxylic acid to make a -1 charge. These changes allowed the polar hydrogens and gasteiger charges to be added to the protein without errors in AutoDock Tools v4.2 (ADT). Polar hydrogens were added to the 3D coordinates of the ligand chemical structures generated from SMILES strings in Discovery Studio. One of the main advantages of AutoDock was that full ligand flexibility was possible for docking to static or partially flexible macromolecules (81 ). om num er value were disregarded ( calculat10n: # GA runs I # of 18 _ ndom value). Therefore, clusters that contained very few conformations clusters -ra 11 .kely formed by random chance. Average binding energies for each cluster were more ~ ed at random were calculated and graphed as histograms. The conformations not 1orm with the lowest binding energies were evaluated for possible hydrogen bonding using Discovery Studio's hydrogen bond monitor default parameters. Additionally, van der Waals interactions were measured using the intermolecular neighbor monitor in Discovery Studio. The distance between atoms in each residue and each ligand were further analyzed. The van der Waals interactions were calculated by addition of the radii of atoms in the ligand and surrounding amino acid atoms with the following radii values: carbon-1.87 A, nitrogen-1.50 A, oxygen-1. 40

3.RESULTS
Since many chemical compounds bind to both GABA receptors and TSPO, we screened a GABA-ergic chemical library (compounds are listed in Table 1), which contained specific TSPO ligands (F series), to examine the overlapping involvement of TSPO ligands binding also to FXR. In this screening assay, agonist ligand binding to FXR drove the expression of an IR-1 regulated luciferase reporter gene. In the library screen (Fig. 3A), one compound, F2 (FGIN-1-27), increased luciferase expression considerably compared to the control. This increase was similar to that observed with CDCA treatment. Conversely, two compounds structurally similar to FGIN-1-27 that are TSPO ligands also, FGIN-1-43 (F3) and PK11195 (FlO), did not increase luciferase expression. Since the FGIN-1 compounds were previously shown to be a selective ligand of TSPO (73,74) and PKl 1195 is a known ligand (31,70,71), these compounds were still included in further experiments for comparison. The chemical structures of the FGIN-1 compound contain the same 2-aryl-indole-3acetamide backbone with varying halide substitutions and hydrocarbon tail lengths.
PK11195 has similar aryl ring structures to the FGIN-1 compounds with chlorine on the phenyl ring. These compounds are quite structurally different from CDCA as seen in Figure 3B (the carbons of CDCA are numbered to correspond with the text). 20 3 . . . . d' tor that a compound is a ligand of a nuclear receptor is their ability to One m 1ca .
activators. In the presences of an agonist ligand, the histone recruit co acetyltransferase, SRC-1, is recruited to the LBD of FXR (82,83). To further validate that these compounds bind to FXR, a mammalian two-hybrid assay was used to demonstrate coactivator recruitment to a ligand-activated FXR (Fig. 4). CDCA at 10 µM and 100 µM significantly increased luciferase expression compared to the control, signifying coactivator recruitment. Similarly, FGIN-1-27 at 5 µM and 10 µM also significantly recruited SRC-1 to the LBD. FGIN-1-27 displayed maximum agonist activity at 5 µM -10 µM. Interestingly, PKl 1195 significantly decreased luciferase expression compared to the control, indicative of decreased basal SRC-1 recruitment.

HuH-7 hepatoma cell line expresses endogenous FXR
Human hepatocytes are the ideal cell type for in vitro studies of liver pathology and physiology studies, but cannot be easily obtained in large numbers because of limited availability of healthy donors. Therefore, we examined two cell lines as alternatives to primary hepatocytes. Our results showed that the HuH-7 cell line expressed similar amounts of endogenous FXR mRNA compared to human hepatocyte case HH1498 (Fig. 5A). ZR-75-1 breast cancer cell line, on the other hand, expressed very little FXR compared to HH1498 . ZR-75-1 was chosen as a good cell line to use for FXR over-expression studies that required limited endogenous FXR interference. All cell lines tested expressed similar amounts of RXRa, with no significant difference compared to HH1498. HuH-7 cells expressed very little endogenous SHP but did not 21 lack LRH-1 or HNF4a expression (Fig. SB). In HuH-7 cells, both CDCA and FGIN-1-27 significantly increased IR-1 driven luciferase expression in the presence and Of exogenous FXR (Fig. SC), while FGIN-1-43 did not modulate luciferase absence . ·ty There was a significant increase in luciferase expression when ZR-7 S-1 cells acttv1 .
were treated with CDCA and FGIN-1-27 in the presence of exogenous FXR (Fig. SD).
No change occurred from treatments in the ZR-7S-l cells in the absence of exogenous FXR.

Changes in BSEP and SHP expression
Another way to measure transcriptional effects of a compound on a nuclear receptor is to look at regulation of specific downstream target genes. FXR ligands directly increase transcription of downstream genes, including BSEP and SHP (84). In our studies, FGIN-1-27 increased mRNA expression of both BSEP and SHP in human hepatocytes (cases HH1486 and HH1498) and HuH-7 cells similar to levels seen with COCA treatment, while FGIN-1-43 did not increase basal expression( Fig. 7 and 8).
PKll 195 did not repress CDCA or FGIN-1-27 induced BSEP mRNA expression in HuH-7 cells or in HH1498 . Conversely, FGIN-1-43 and PKl 1195 increased SHP mRNA expression when combined with CDCA in HH1498 (Fig. 8) CYP7Al expression in human hepatocytes (Fig. 9). FGIN-1-43, on the other hand, does not alter basal expression of CYP7 Al . Unfortunately, CYP7 Al expression studies in HuH-7 cells yielded inconsistent results (data not shown), most likely due to the low expression of SHP (Fig. 5B). In another study, we also showed that PKl 1195 significantly repressed CYP7Al expression in HH1498, but to a lesser extant than the effects seen by COCA and FGIN-1-27 (Fig. 9).

TSPO ligands bind to the LBD of FXR
To further explore the capacity of PKl 1195 and 4 FGIN-1 compounds to fit into the LBD of FXR, in silico molecular modeling was performed. Docking results verified that these TSPO ligands could fit into an active conformation of FXR. The crystallized structure of human FXR (PDB ID code 3BEJ) (80) was used as the template to study the binding properties of COCA, PKl 1195, and 4 FGIN-1 compounds. This template was among 9 crystallized structures of FXR in the PDB, all of which had a ligand and coactivator bound, except for 1 OSH, which lacked a coactivator. Although this template has been used for other docking experiments (78), 1 OSH also lacked a significant portion of helix 3, and therefore was not an ideal candidate. Two of the structures were isolated from rat FXR, differing from human FXR by only 12 amino  (Fig. 1 lA). Conformation a oriented the carboxyl group on C24 of (con10 COCA near T288 to form hydrogen bonds (hydroxyl oxygen 2.39 A, hydroxyl hydrogen 1.88 A, carboxyl oxygen 1.77 A). In a flipped orientation (conformation b), the COCA carboxyl group hydrogen bonded to Arg331 (2.16 A) and the oxygen of C7 hydrogen bonded to H447 (2.20 A). Not surprisingly, CDCA in conformation a fit an orientation similar to MFA-1 (Fig. 10), which is a CDCA analog with an additional phenyl ring at C21 and a carboxyl group at C3 (80). Results show CDCA in both conformations was able to form van der Waal interactions with 13 residues on helices 3, 5, 10/11 and 12. oriented the rings perpendicular to the pocket. Similar to CDCA, FGIN-1-27 was most likely held in place in the LBD by van der Waals interactions with 16-18 residues on helices 3, 5, 6, 7, 10/11, and 12.

Docking of FGIN-1-20
Analogs of FGIN-1-27 were also docked into FXR to examine the significance of fluorine and the hydrocarbon tails to the binding properties. When the hydrocarbon tails were shortened from hexyls to propyls in FGIN-1-20 ( Fig. 11 C), the number of torsions decreased by 6, resulting in fewer overall conformational clusters.

Docking of PKJ 1195
The most well known TSPO ligand, PKl 1195, was also docked into the LBD of FXR.
Since PKl 1195 only had 5 possible flexible torsions, fewer GA runs were required to reach optimal refinement. Out of 600 GA runs ADT formed 9 conformational clusters ' for PKl 1195 with free energy of binding ranging from -9 .94 kcal/mol to -8.81 kcaVmol. Of these 9 clusters, 8 contained fewer than the random number value of conformations (600/9 = 66.7) so they were considered background binding. The 1 ster contained 496 conformations represented by an average binding energy largest cu of-9 . 64 ± 0.005 kcal/mol. Figure 1 lF shows that PKl 1195 adopted only one possible ~ nnation within the pocket with no possibility or opportunity to form any con10 hydrogen bonds, but was close enough to form hydrophobic interactions with 17 amino acids on helices 3, 5, 7, 10111, and 12.

7. Mutational studies of FXR
Since the molecular modeling studies were based on a fixed crystal structure, as opposed to a fully flexible molecule in a biological environment, point mutations formed in FXR were necessary to validate the studies. Only CDCA and FGIN-1-27 were examined due to their capacity to hydrogen bond. (Though FGIN-1-20 was also able to form a hydrogen bond, this compound is not commercially available). Figure   12A shows the conformations of CDCA and FGIN-1-27 explained above. The docking studies demonstrated that these compounds formed hydrogen bonds with Thr288 (helix 3), Arg331(helix5), Tyr369 (helix 7) or His447 (helix 10/11), so single amino acid mutations of these residues were fo~med (Table 2). Ser332 is the only other residue in the pocket that could form hydrogen bonds with side chain atoms so a mutation was also created for this residue as another potential key residue. As a control, ZR-75-1 cells were evaluated with and without exogenous FXR to compare to changes caused by mutated residues. In each mutation, the basal activation of FXR was decreased compared to the wild-type FXR transfection. Both CDCA and FGIN-1-27 hydrogen bonded to T288 in silico; when threonine (T288L) was mutated, activation by CDCA decreased to basal while FGIN-1-27 decreased by about half.

30
Only COCA in conformation b hydrogen bonded to R331, so as expected, the arginine mutation (R331L) prevented CDCA-induced expression, whereas FGIN-1-27 was not affected. Conformation b of FGIN-1-27 hydrogen bonded to tyrosine 369, but Y369L mutation resulted in a gain of function for FGIN-1-27. As expected, the Y369 mutation showed no change in CDCA. As predicted for CDCA in conformation b, the histidine mutation (H447F) prevented luciferase expression. The H447F mutation decreased activity also in response to FGIN-1-27. Even though no hydrogen bonding was seen with Ser332, the mutation (S332F) decreased luciferase activity also with both COCA and FGIN-1-27 treatments.

DISCUSSION
FXR is involved in multiple aspects of the maintenance of bile acid homeostasis acting as a mediator between bile acid synthesis and efflux from the liver (8). The alternative pathway of bile acid synthesis requires TSPO for the trans location of cholesterol into the mitochondria (29). Since both FXR and TSPO are important for maintaining bile acid homeostasis, it is not improbable that a single compound modulates the activity of each of these proteins. We first looked at a GABA-ergic chemical library in order to identify chemical treatments, already known as TSPO ligands, that increase transcription of an FXR-regulated luciferase reporter. We identified one compound, FGIN-1-27, that activated FXR to a level similar to CDCA (Fig. 3A). FGIN-1-27 was one of the specific TSPO ligands in the F series of compounds in the chemical library; therefore, we chose two other structurally similar TSPO ligands  PKl 1195) to investigate further.
Our studies show that FGIN-1-27 is a partial agonist of FXR (Fig. 6). We show through in vitro luciferase reporter gene assays that treatment of FGIN-1-27 activates FXR-mediated transcription but decreases FXR activation by CDCA when CDCA concentration is not limiting (100 µM) . Even though the FXR ligand-binding pocket preferentially binds amphipathic, non-planar bile acids that allow polar entities to form hydrogen bonds with amino acid residues (86), FGIN-1-27 fits into this pocket also.
Based on the size of the pocket and of the individual molecules, it is unlikely that both COCA and FGIN-1-27 bind to the pocket at the same time. According to in silica I odeling results, FGIN-1-27 fits into the pocket with low binding energies, inolecu arm bl to those of CDCA ( Fig. 1 lA and 1 lB). Both CDCA and FGIN-1-27 compara e bond to the same amino acid (T288) in silica, and show no FXR-activation hydrogen when H44 7 is mutated into phenylalanine (Fig. 12), which shows that both compounds bind similarly inside the pocket and compete for binding positions. Because CDCA bas more atoms that can form hydrogen bonds, compared to FGIN-1-27, CDCA can fit into the binding pocket in more than one favorable position with more favorable agonist binding. This is evident when CDCA (100 µM In this binding study, CDCA fits into the FXR ligand-binding pocket, with 87% of the possible conformations oriented so the C-24 carboxylate group hydrogen bonds with Thr288 ( Fig. l lA and 12A). The ?a-hydroxyl group of CDCA in the remaining 13% of the conformations hydrogen bonds with H447 and the C-24 oxygen binds to Arg3 3 t. In accordance with modeling studies performed by other groups, CDCA 2Th.
is calculation was based on a 1.5 kg adult liver 33 . .t elf similarly by positioning the 3a-hydroxyl group near His44 7, allowing the onents is C-24 carboxylate group to hydrogen bond to Arg331 (86), or it is oriented so the C-24 carboxylate oxygens hydrogen bond to Leu348 (78). The results of this study position COCA (conformation a) in an orientation more similar to that typically adopted by steroid hormones, and MFA-1, with the steroid rings rotated so that the 3a-hydroxyl group is near Arg331 (Fig. llA).
It is possible that there is more than one functional orientation of CDCA in the LBD that can cause agonist ligand effects from hydrogen bonding to more than one residue.
Our point mutation studies of FXR show that Thr288, Arg331 and His44 7 were the key residues responsible for the agonist effects of CDCA (Fig. 12B), all of which Conned hydrogen bonds in silica. This suggests that the ?a-hydroxyl and C-24 oxygens are key attributes for CDCA. Our studies are in agreement with other studies that have shown the 3a-hydroxyl group, present on all bile acids, is not necessary for FXR activation (86) and an oxygen in either a carboxyl group or an alcohol on C-24 is responsible for the enhanced ligand potency (88). It is most likely a combination of the hydrophobicity of CDCA and available oxygens to form hydrogen bonds that confers agonist-binding properties, and allows CDCA to bind in more than one conformation.
Although the FXR ligand-binding pocket evolved to recogmze non-planar amphipathic bile acids (86), FXR is able to bind compounds also with planar components, such as the FGIN-1 compounds. All four FGIN-1 compounds subjected to in silica modeling fit into four main conformations, a, b, c and d (Fig. 1 lB-1 lE). in conformation c could hydrogen bond to His44 7. This would correlate with the mutation of His447 diminishing FXR-activation by FGIN-1-27 (Fig. 12B). Upon entry into the pocket, it is conceivable that the FGIN-1 compounds would favor conformations c or d and never bend into an "L" shaped conformation. When Arg331 t d l ·nto Ieucine little change is seen in FGIN-1-27 activity compared with is muta e ' wild-type FXR (Fig. 12B). If FGIN-1-27 was in conformation d, fluorine could · ably hydrogen bond to the arginine, however, since Arg331 seems to have no conce1v significant interaction with FGIN-1-27, conformation c with the fluorophenyl group near His447 is more probable. Although the mutation of Ser332 (S332F) eliminates FXR-activation by treatments of both CDCA and FGIN-1-27, the addition of the bulky phenylalanine residue is most likely large enough to block the entrance to the binding pocket preventing any ligand entry.
To examine whether the hydrocarbon tails were factors in the binding of the FGIN-1 compounds, FGIN-1-20 was subjected to evaluation (Fig. llC). The original study with the FGIN-1 compounds (72) found the binding affinity to TSPO increased with increasing alkyl chain lengths, up to 6 carbons. In this study, however, decreasing the number of carbons on the hydrophobic tails minimally changes the binding orientations in conformations a and b, with little difference in binding energies.
However, the longer hydrophobic carbon tails in the other FGIN-1 compounds that fold alongside the indole backbone in conformations c and d create hydrophobic interactions favorable for ligand binding. The binding orientation changes when 3 carbons from each tail are removed; FGIN-1-20 does not fit into conformation d.
Additionally, the oxygen on FGIN-1-20 in conformation c forms a hydrogen bond with His447. This orientation is unlikely to occur in structures with long hydrocarbon tails, such as FGIN-1-27, because the tails would cause steric hindrance. We did, however, see complete elimination of the agonist effects of FGIN-1-27 with the 36 8447 F mutation of FXR, but this correlates with earlier speculations that the FGIN-1 d s may favor conformation c upon entry into the pocket with fluorine of compoun FGIN-l-Z? hydrogen bonding to His44 7.
The addition of chlorine does not prevent FGIN-1-43 from fitting into the binding pocket of FXR; instead, it only forms unfavorable binding energies (Fig. 1 lE). For this compound, conformation c is favored most often. It is possible that the sheer size of chlorine prevents FGIN-1-43 from binding efficiently. Interestingly, FGIN-1-43 is better at antagonizing the agonist effects of FGIN-1-27 on FXR more so than with COCA ( Fig. 6). This suggests that FXR can recognize the core indole-acetarnide structure without discrimination, but FGIN-1-27 has a higher binding affinity. When both FGIN-1 compounds are present, both will go into the pocket, but FGIN-1-27 will bind more favorably than FGIN-1-43. Similarly, because FGIN-1-43 does not bind to any residues specifically, CDCA will bind more efficiently so FGIN-1-43 will be displaced easier in the presence of CDCA. This idea also explains the antagonist effects of PKl 1195 on both CDCA and FGIN-r-27. According to the modeling results, PKl 1195 fits only into one orientation with an inhibition constant very similar to CDCA (Fig. 11 F). Even though PKl 1195 does not interact with any residues specifically, it is oriented so access to His44 7 is blocked. When PKll 195 is in the presence of an agonist ligand, PKl 1195 could compete for occupancy of the pocket of FXR and prevent CDCA or FGIN-1 -27 from binding (Fig. 6). 37 th S tudies we had to take into account the possibility of the FGIN-1 compounds In ese ' binding to TSPO on the mitochondria to indirectly increase the synthesis of CDCA, and in tum, activate FXR. In fact, FGIN-1-43 is found to be a more potent TSPO ligand than FGIN-1-27 (72). Ifthere were any downstream effects present, FGIN-1-43 would show equal, if not more, activation of FXR than FGIN-1-27. In all of the results, FGIN-1-43 never activates FXR, thus demonstrating FXR activation by FGIN-1-27 is not a result of TSPO ligand binding.
We also show that FGIN-1-27 is as efficient as CDCA in recruiting the coactivator, SRC-1, to FXR LBD (Fig. 4). At 10 µM, FGIN-1-27 binding causes greater fold increase in luciferase expression than did CDCA treatment at 10 µM. The decrease in capability of PKl 1195 to recruit SRC-1 correlates with this compound being an antagonist. However, PKl 1195 does not antagonize BSEP and SHP expression as expected ( Fig. 7 and 8 HH 1498 , where SHP is not limiting, PKl 1195 treatment causes a significant increase in SHP mRNA expression, but PKI 1195 does not inhibit CYP7 Al expression at the same level as CDCA or FGIN-1-27 (Fig. 9). Our data shows PKl 1195 has little · t effect on FXR alone, therefore, this increase in SHP expression and lack of agoms full CYP7Al repression could be due to PK11195 binding to LRH-1. If this were the e PKl 1195 could prevent SHP from binding to LRH-1, thus eliminating the cas' repressor function of SHP. Further studies of PKll 195 binding to LRH-1 to cause direct transcriptional effects are necessary to validate this theory.
Jn these experiments, FGIN-1-27 proves to be at least equally as potent a ligand of FXR as CDCA at 10 µM , despite inevitable variations between individual hepatocyte cases. FGIN-1-27 increases both BSEP and SHP expression significantly, while repressing CYP7Al expression, as expected ( Fig. 7-9). FGIN-1-43 and PKl 1195 treatments, however, rarely cause any differences in gene expression from FXR activation. With all data considered, we conclude that FGIN-1-43 is a selective antagonist, competing only with the ligand with similar binding affinities to itself  and PKl 1195 is a non-selective antagonist.
In addition to showing that these TSPO ligands modulate FXR, we also investigated a cell line with non-limiting endogenous FXR and RXR as an alternative to human hepatocytes (Fig. 5). Although primary human hepatocytes are the best in vitro representation of human liver, they vary among individuals and are expensive and difficult to acquire and maintain. Therefore, HuH-7 cells are beneficial for FXR 39 I I Sl ·nce they can be cultured in large numbers and passaged repeatedly. studies Unfortunately, HuH-7 lack SHP, so studies involving CYP7 Al repression through the SHP pathway may be difficult. Additionally, we show that the breast carcinoma cell . ZR 75-1 has very little endogenous FXR with non-limiting endogenous RXR. bne, -' Therefore, this cell line is ideal for mutational studies to avoid the interference of endogenous FXR.
In summary, targeting the rate-limiting step in the alternative pathway would be beneficial for upregulating this pathway. Correspondingly, TSPO ligands are known to increase cholesterol uptake into the mitochondria, which has been proven to be the rate-limiting step for the alternative pathway (29). Although controversial, some studies have shown that bile acids do not regulate CYP27A1 expression the same as CYP7Al (68,69), which is not surprising since increasing CYP27Al expression does not affect bile acid synthesis rates (29). Therefore, upregulating the alternative pathway apart from bile acid activation would be beneficial in people possessing faulty genes for CYP7 Al because the alternative pathway is heavily relied upon.
However, in healthy populations an upregulation would not be necessary since the alternative pathway contributes little to the overall synthesis (23).
We have shown that FGIN-1-27 increases FXR transcriptional activity to increase BSEP and SHP expression. Also, FGIN-1-27 increases the rate of cholesterol entering the mitochondria by binding to TSPO (72). As demonstrated in Figure 13, targeting both TSPO and FXR with one compound would increase the bile acid synthesis rate of l ative pathway while regulating homeostasis in the liver by controlling the the a tern . feedback through FXR. This would occur by 1) FGIN-1-27 binding to TSPO, negative facilitating the transport of cholesterol into the mitochondria where 2) CYP27 Al would initiate the production of COCA. Furthermore, 3) FGIN-1-27 binds to FXR to 4 ) increase BSEP expression that would increase the efflux of bile from the liver. As bile acids are removed from the liver, the bile acid pool would decrease and trigger more synthesis of COCA, thus lowering the cholesterol pool.
Future studies will be necessary to investigate the changes in the production of bile acid intermediates following FGIN-1-27 treatment. Multiple other genes involved in bile/lipid homeostasis are activated by FXR, including phospholipid transport protein, intestinal bile acid binding protein, and multidrug resistant protein 2 (MRP2) (84). The 1 . CYP enzymes involved in the two pathways of bile acid synthesis.
~ initiates t~e classic/neutral pathway to produce_ ~~olic acid (_CA) ~~d h odeoxycholic acid (COCA). CYP27Al and CYP46Al mittate alternative/acidic c ~~way forming oxysterols that must undergo ?a-hydroxylation before becoming roCA. The carbons on cholesterol are numbered to correspond with the names of h intermediate formed by each CYP enzyme shown (boxed); changes to each ::cture are illustrated in red. Note: other non-CYP enzymes are also involved in these pathways. Figure 2: The role of STAR and TSPO in cholesterol transport into the mitochondria. Intracellular cholesterol binds to STAR, and through a complex pathway involving other proteins not shown, cho.lesterol _is !ranspo:t:ed to the OMM. Cholesterol is then transferred to TSPO where ligand bmdmg facilitates cholesterol uptake into the IMM whe~e CYP27 Al resi~es. ~holesterol_ transport i~ the ratelimiting step in the alternative pathway for bile acid synthesis. COCA will then be synthesized once in the liver.  : Coactivator recruitment to FXR in mammalian two-hybrid assay. COS-1 cells were transfected with 5 µg of pFR-luc, 1.5 µg pM-SRC-1, 1.5 µg VP16-FXR and 0.5 µg pRL-CMV and treated for 24h. Luciferase activity of each treatment is reported as fold change relative to OMSO (control) represented by a solid line at 1. •denotes significance compared to control, p::::; 0.05, (n=8).   Human hepatocytes (HH1486 and HH1498) and HuH-7 cells were treated for 24h at a final concentration of 10 µM. Total RNA was reverse transcribed and cDNA was subjected to actin and SHP gene-specific amplification with SYBR green PCR. SHP expression was normalized to actin and treatments are relative to DMSO (control) in each experiment. In HH1498 and HuH-7, CDCA and FGIN-1-27 were cotreated with FGIN-1-43 or PKll 195. The CDCA and FGIN-1-27 control treatment values are represented by dotted lines. * denotes significance compared to control, p :S 0.05. t denotes significance compared to respective constant treatment, represented by dotted line, p :S 0.05, HH1486 (n=3), HH1498 (n=3), and HuH-7 (n=2). Figure 9: Expression of CYP7 Al mRNA in human hepatocytes. Human hepatocytes (HH1486 and HH1498) were treated for 24h at a final concentration of 10 µM. Total RNA was reverse transcribed and cDNA was subjected to actin and CYP7A1 gene-specific amplification with SYBR green PCR. CYP7A1 expression was normalized to actin and treatments are relative to DMSO (control) in each experiment. The schematic illustrates the direct repression of CYP7 Al gene expression by SHP via indirect FXR ligand activation. * denotes significance compared to control, p :s 0.05, (n=3).
F. ure 1 O: Crystallized structure of FXR LBD with MF A-1 in the binding pocket %BID code 3BEJ). Each _a-helix in the LBD is labeled (1-1.2) and v~ries by color ~or simpler visual representation. The small fragment of SRC-1 is shown m green. figure 11: Molecular modeling of CDCA and TSPO ligands in the LBD of FXR. Each compound formed various conformations, grouped into clusters, based upon orientation of each atom deviating by 2.0 A RMS. On the graphs, each bar represents the number of conformations in a cluster with shared mean free energy of binding ± SEM. The brackets, labeled a-d, represent the binding energy range for each conformation based upon core ring orientation. Within each bracket, the hydrocarbon tails vary in position while the rings maintain the same conformation. The lowest energy conformers representing the four bracketed conformation types are shown with key amino acid residues highlighted: T = Thr288, R = Arg331, Y = Tyr369, and H = His447. Dotted black lines represent hydrogen bonds. Each table shows the frequency and mean free energy of binding ± SEM for each bracketed conformation. A) CDCA fit into two conformations from 600 GA runs forming hydrogen bonds with T288 in conformation a and with R331 and H447 in conformation b. B) FGIN-1-27 formed four main conformations from 800 GA runs forming hydrogen bonds with T288 in conformation a and with Y369 in conformation b. C) FGIN-1-20 fit into three conformations from 800 GA runs forming hydrogen bonds with T288 in conformation a and H447 in conformation c. D) Four conformations of FGIN-1-51 were formed from 800 GA runs. E) FGIN-1-43 found four conformations from 800 GA runs. F) Only one conformation resulted for PKl 1195 from 600 GA runs.