A novel GPER antagonist protects against the formation of estrogen-induced cholesterol gallstones in female mice

Many clinical studies and epidemiological investigations have clearly demonstrated that women are twice as likely to develop cholesterol gallstones as men, and oral contraceptives and other estrogen therapies dramatically increase that risk. Further, animal studies have revealed that estrogen promotes cholesterol gallstone formation through the estrogen receptor (ER) α, but not ERβ, pathway. More importantly, some genetic and pathophysiological studies have found that G protein-coupled estrogen receptor (GPER) 1 is a new gallstone gene, Lith18, on chromosome 5 in mice and produces additional lithogenic actions, working independently of ERα, to markedly increase cholelithogenesis in female mice. Based on computational modeling of GPER, a novel series of GPER-selective antagonists were designed, synthesized, and subsequently assessed for their therapeutic effects via calcium mobilization, cAMP, and ERα and ERβ fluorescence polarization binding assays. From this series of compounds, one new compound, 2-cyclohexyl-4-isopropyl-N-(4-methoxybenzyl)aniline (CIMBA), exhibits superior antagonism and selectivity exclusively for GPER. Furthermore, CIMBA reduces the formation of 17β-estradiol-induced gallstones in a dose-dependent manner in ovariectomized mice fed a lithogenic diet for 8 weeks. At 32 μg/day/kg CIMBA, no gallstones are found, even in ovariectomized ERα (−/−) mice treated with 6 μg/day 17β-estradiol and fed the lithogenic diet for 8 weeks. In conclusion, CIMBA treatment protects against the formation of estrogen-induced cholesterol gallstones by inhibiting the GPER signaling pathway in female mice. CIMBA may thus be a new agent for effectively treating cholesterol gallstone disease in women.

found a GPER-selective agonist, G-1, that exhibits a binding constant of 11 nM with no significant binding to ER or ER at concentrations up to 1 µM (Fig. 1). Two GPERselective antagonists, G-15 and G-36, were then identified on the basis of the same dihydroquinoline scaffold (17,18). Although G-15 possessed a relatively strong binding constant of 20 nM to GPER1, further studies revealed significant binding to the classical ER and activation of EREs at concentrations above 100 nM (18). To address off-target binding and nonselectivity, Dennis et al. (18) modified the dihydroquinoline scaffold. As a result, G-36 was discovered. In this compound, the addition of a bulky, lipophilic group led to decreased binding to ER without significantly affecting activity at GPER. While decreased binding to the classical ER was reduced at concentrations above 1 M, the ERE activation was still weakly present (18). These results show that the G-series GPER antagonists may be limited as GPERselective antagonists due to the presence of off-target effects. Nevertheless, commercial availability of the G-series of ligands has made them valuable GPER chemical probes for research.
It is well known that the annual medical cost for treating gallstone disease in the United States exceeded $6 billion in 2004 and was even higher in 2019 (24). The economic burden of gallstone disease is exacerbated by the fact that laparoscopic cholecystectomy remains the standard treatment for symptomatic gallstones worldwide (25). In addition, clinical and epidemiological studies have clearly demonstrated that women are twice as likely to develop cholesterol gallstones as men, and oral contraceptives or other estrogen therapies significantly increase risk for gallstone formation (26,27). Although ER, but not ER, plays a key role in estrogen's lithogenic effects (28), new evidence has shown that GPER can produce additional lithogenic actions and promote gallstone formation in female mice independent from ER (29,30). More importantly, genetic analysis has found that Gper1 is a new gallstone gene, Lith18, on chromosome 5 in mice (31). All of these studies strongly suggest that GPER could play a critical role in the formation of estrogen-induced cholesterol gallstones (32).
To differentiate the GPER1-mediated effects from the ER-mediated events in estrogen-induced cholesterol gallstones, we have created three knockout mouse lines: GPER1 ( / ), ER (  (29). Our results have shown that compared with ovariectomized (OVX) wild-type mice, the deletion of either Er or Gper1 significantly reduced the prevalence of estrogen-induced gallstones but could not abolish it completely. Furthermore, no gallstones were found, even in OVX GPER1 ( / )/ER ( / ) mice treated with E 2 at 6 µg/day and fed the lithogenic diet for 8 weeks (29). These results clearly indicated that GPER plays a role that is independent of ER in the pathogenesis of cholesterol gallstone disease. Therefore, it is imperative to develop a new, potent GPER-selective antagonist that could prevent estrogen-induced gallstones in women.
Discovery of GPER ligands has relied on the identification of individual ligands via in silico molecular screening, and few synthetic GPER-modulating molecules have been reported (16)(17)(18)(19)(20)(21)(22)(23). There has been little effort to understand the important binding interactions of ligands within the binding pocket of GPER to aid in optimization of ligands. In this study, homology modeling was utilized to design and synthesize a series of GPER-selective antagonists (33,34). Using calcium mobilization and cAMP assays, we have established the first large series of compounds to probe the binding pocket of GPER and aid in the design of GPER-selective antagonists. Additionally, our animal studies have revealed that the lead compound discovered from this study can prevent the formation of E 2 -induced cholesterol gallstones in OVX mice by inhibiting the GPER signaling pathway and may provide an additional effective therapeutic option for gallstone disease in women and patients exposed to high levels of estrogen.

Chemistry
The compounds were synthesized according to the procedures shown in Fig. 2. Characteristics of the compounds were analyzed with NMR and high-resolution MS. 1 H and 13 C NMR spectra were performed in chloroform-D from Cambridge Isotope Laboratories (Andover, MA) on a 400 MHz Bruker AVANCE III (Bruker, Billerica, MA). Data were analyzed with TopSpin version 3.2. Experiments with high-resolution MS were carried out using a MaXis plus electrospray quadrupole TOF mass spectrometer. Ionization was done by electrospray with the samples infused into the instrument in 5 µM acetonitrile-water-formic acid (50:50:0.1; v/v/v) solutions at a flow rate of 3 µl/min. Nitrogen was used as a nebulizing, drying, and collision gas. HPLCgrade acetonitrile and water were purchased from Sigma-Aldrich (St. Louis, MO). CIMBA exhibited a purity of greater than 95%, Fig. 1. The endogenous ligand of GPER has been identified as E 2 . Other estrogenic compounds such as 17-estradiol, estrone, and estriol all exhibit weak binding at the GPER. In silico screening identified G-1 as a GPER-selective compound. The site-specific removal of the ethanone moiety (G-15) resulted in a change in pharmacological activity. G-15 exhibited offtarget binding at the classical nuclear ERs, ER and ER, and was further modified with an isopropyl moiety. The resulting compound, G-36, exhibited a similar efficacy to G-15 but did not show off-target binding. This study shows that CIMBA can prevent the formation of estrogen-induced gallstones in female mice. as did all synthesized compounds. High-resolution MS for CIMBA (C 23 H 32 NO + ) was calculated as m/z = 338.2478 and found to be m/z = 338.2471.

Cell line and culture
The human promyelocytic leukemia (HL-60) cells, purchased from the American Type Culture Collection (Manassas, VA), were cultivated in phenol red-free RPMI 1640 media containing 10% heat-inactivated FBS, 1% penicillin, and 1% GlutaMax. Three days prior to assays, the HL-60 cells were switched into the media containing 10% charcoal-stripped FBS. Cells were passaged every 3 days and maintained at a cell concentration below 1 × 10 6 to prevent differentiation. The cells were incubated at 37°C under 5% CO 2 .

Evaluation of GPER-associated calcium mobilization in HL-60 cells
The HL-60 cells were grown in 10% charcoal-stripped FBS, phenol red-free RPMI 72 h before the assays were performed. The cells were centrifuged and counted with a hemocytometer. The assay required 100,000 cells per well for a total number of cells on a 96-well plate assay of about 1 × 10 7 . HL-60 cells (1 × 10 7 ) were incubated in 50:1 HBSS-HEPES containing 5 M Indo-1 AM (Thermo Fisher Scientific, Waltham, MA) and 0.05% pluronic acid for 0.5 h at room temperature. Cells were spun down and washed with 50:1 HBSS-HEPES and resuspended in media. The resuspension was placed on ice for no longer than 5 min. Cells were loaded into the plate at 100,000 cells/well. For agonists, cells were immediately incubated for 15 min at 37°C. Following the 15 min of incubation, FlexStation 3 Multimode Plate Reader (Molecular Devices, Sunnyvale, CA) added the appropriate amount of agonist and was read for 150 s at 37°C. For antagonists, once the cells were seeded, the antagonist was added and incubated for 15 min at 37°C. After 15 min of equilibration, the experimentally determined EC 80 value of G-1 (3 M) was added by the FlexStation and read for 150 s at 37°C. Calcium mobilization was determined ratiometrically using  ex 350 nm and  em 405/490 nm, respectively.

Evaluation of cAMP response in HL-60 cells
Homogeneous time-resolved fluorescence (HTRF) components for cAMP were purchased from CisBio (Bedford, MA). HL-60 cells were grown in 10% charcoal-stripped FBS, phenol red-free RPMI media for 72 h prior to assays. Cells were centrifuged and counted with a hemocytometer. The total number of cells needed to complete the assay was determined based upon 8,000 cells/ well. The determined number of cells was diluted in 5:1 (5×) stimulation buffer containing 500 M 3-isobutyl-1-methylxanthine (Sigma Aldrich). To an HTRF 96-well low-volume white plate from CisBio, 5 l of cold cell suspension was added to each well, followed by 4 l of (2.5×) agonist or stimulation buffer (negative control and nonstimulated cells). Cells were covered with a clear plastic film and incubated at 37°C for 15 min in an incubator. After 15 min, 1 l of 10 M (10×) forskolin was added to each well. The plate was once again sealed and incubated at 37°C for 15 min. Subsequently, 5 l cAMP-d 2 (acceptor) was added to all wells, including the controls. Conversely, 5 l monoclonal anti-cAMP Eu3 + cryptate (donor) was added to only the wells with a test compound and nonstimulated cells. Following the addition, the plate was sealed, covered with aluminum foil, and incubated at room temperature for 30 min. After the allotted time cells were read using FlexStation3 with  ex 314 nm and dual-emission wavelengths for the acceptor and donor emission signals ( em donor = 620 nm;  em acceptor = 655 nm). A ratio of the acceptor to the donor was calculated and utilized in the determination of the EC 50 value of G-1. For an antagonism platform, the procedure was slightly altered. The EC 80 value of G-1 was redetermined on the basis of cAMP agonism results. To the plate, 2 l of (5×) antagonist were added, followed by the addition of 5 l of the cold cell suspension to all wells, with top and bottom wells being designated as G-1 control and forskolin control (20 M), respectively. Cells and antagonists were incubated for 15 min at 37°C. Then, 2 l of the (5×) EC 80 of G-1 was added to wells with the antagonist. For the G-1 control, 2 l of 5:1 stimulation buffer and 2 µl of the (5×) EC 80 G-1 were added. For the forskolin control, 4 l of 5:1 stimulation buffer and 1 l of (10×) forskolin were combined. For all other wells, following the addition of antagonists, cells, and agonists, Fig. 2. A: The proposed scaffold consists of two benzene rings connected via a secondary amine linker. One of these benzene rings is intended to interact with F208 (EL2) and H307 7.37 , and the other is meant to hold two hydrophobic moieties, R 1 and R 2 , to interact with the hydrophobic pocket. The secondary amine linker is anticipated to hydrogen bond to N310 7.40 , and varying substituents at R 3 have been included to explore the binding pocket in that region. B: Because all brominated anilines (isopropyl, methyl, and tert-butyl) are available commercially, they serve as a logical starting point for the synthesis. The brominated anilines undergo palladium-catalyzed Miyaura borylation with pinacolborane to produce borylated anilines (2a, 2b, and 2c). The reaction of 1-cyclohexenyl trifluoromethansulfonate successfully couples to the borylated anilines to form 3a, 3b, and 3c. Catalytic hydrogenation of the coupled cyclohexenyl ring is afforded quantitatively in 10% palladium-carbon and H 2 to give 4a, 4b, and 4c. The final step of the synthetic route involves the reductive amination between the prepared aniline (4a, 4b, and 4c) and various aldehydes (R 2 ) to yield 5-25. The aldehydes are chosen to study variation in size and electrostatic properties that may be favorable within the binding pocket of GPER.
1 l of (10×) forskolin were add for a final concentration of 1 M. After the addition of forskolin, cells were incubated at 37°C for 5 min before dyes and lysis buffer were added. Cells were allowed to equilibrate for 30 min at room temperature before being read at  ex 314 nm dual-emission wavelengths for the acceptor and donor emission signals ( em donor = 620 nm;  em acceptor 655 = nm). A ratio of the acceptor to donor was calculated and utilized to determine the IC 50 values of tested antagonists. HTRF values were converted to % cAMP using the manufacturer's suggested protocol.

ER and ER fluorescence polarization assay
The ER and ER PolarScreen Competitor Assays were purchased from Invitrogen (Carlsbad, CA). The concentrations of the ER and ER proteins varied between the two assays. A 4× sample of ER was prepared for a final concentration of 75 nM. Conversely, a 4× sample of ER was prepared for a final concentration of 23 nM within the assay. For both ER and ER, a 4× aliquot of fluoromone was prepared for a final assay concentration of 4.5 nM. Despite the differences in concentration of ER and ER, the binding assay protocol for the two different receptors remained the same. Drug dilutions were prepared for a 2× dilution in the ER-specific buffer that was provided within the assay kit. In addition to the drug dilutions, a 2× aliquot of fluoromone was prepared by adding 10 l of 4× with 10 l of ER-specific buffer. A 2× dilution of protein and fluorophore was achieved by adding equal portions of each to each other. To a black 384-well plate, 10 l of the compound was added to the wells, followed by 10 l of the 2× mixture of protein and fluoromone. Two controls were included in the plate design. In one control, the protein and compound were omitted so that only ER-specific buffer and fluoromone remained. A separate control contained the protein and fluoromone control. Once all of the components were added, the plate was covered with a film and aluminum foil. The plate was incubated for 2 h at room temperature before being read with  ex 485 nm and  em 535 nm, respectively.

Animals and diet
Although it has been found that inbred AKR/J mice are a gallstone-resistant strain, they are still susceptible to E 2 -induced cholesterol gallstone formation (28). Although AKR/J mice have an intact expression of the Gper1, Er, and Er genes, mRNA levels of Er in the liver are almost undetectable under normal physiological conditions. The hepatic expression of Er is 50-fold lower than that of Er, even under the stimulation of E 2 . In addition, we have established breeding colonies of ER ( +/ ) mice on an AKR/J genetic background in-house. ER ( +/ ) heterozygotes were also fertile and showed no obvious phenotypes in association with the disrupted Er genotype. A cross between heterozygous ER ( +/ ) mice produced the live birth of normal litter sizes of homozygous ER ( / ) mice. Mice were maintained in a temperature-controlled room (22 ± 1°C) with a 12 h day cycle (lights on between 6:00 AM and 6:00 PM) and were provided free access to water and normal mouse chow containing trace cholesterol (<0.02%) (Lab Rodent Diet, St. Louis, MO). To exclude possible interindividual differences in endogenous estrogen concentrations, all female mice, at the age of 4 weeks, were ovariectomized. At 8 weeks of age, these mice were implanted subcutaneously with pellets (Innovative Research of America, Sarasota, FL) releasing E 2 at 6 µg/day for 8 weeks. As reported (28), plasma estradiol concentrations were significantly increased to 81 ± 21 pg/ml in OVX mice treated with E 2 at 6 µg/day compared with wild-type mice (26 ± 11 pg/ml) receiving neither surgery nor E 2 treatment. To study the role of the GPER-selective antagonist CIMBA in the prevention of E 2 -induced cholesterol gallstones, the mice, at 8 weeks of age, were injected intraperitoneally with CIMBA at 0, 16, or 32 g/ day/kg, as well as fed the lithogenic diet containing 1% cholesterol, 15% butter fat, and 0.5% cholic acid for 8 weeks. All procedures were in accordance with current National Institutes of Health guidelines and were approved by the Institutional Animal Care and Use Committees of the Albert Einstein College of Medicine (Bronx, NY) and Saint Louis University (St. Louis, MO).

Gallstone studies
After 8 weeks of being fed the lithogenic diet, mice were fasted overnight but had free access to water. After anesthetization with pentobarbital, cholecystectomy was performed during laparotomy. The entire gallbladder bile was studied by polarizing light microscopy without a cover slip and then with a cover slip using phasecontrast optics for the presence of mucin gels, liquid crystals, cholesterol monohydrate crystals, sandy stones, and real gallstones according to previously established criteria (35). The images of cholesterol monohydrate crystals and gallstones were analyzed by a Carl Zeiss Imaging System with AxioVision Rel version 4.6 software (Carl Zeiss Microimaging GmbH, Göttingen, Germany). Under the circumstances, gallstones exhibit rounded contours and black centers from light scattering/absorption, and gallstones were counted at 100× or 200× magnification. The prevalence rate of gallstones is determined by the number of mice per group with evident stones. After microscopic studies, the gallbladder bile was collected, frozen, and stored at 20°C for lipid studies.

Biliary lipid analysis
Cholesterol, phospholipid, and bile salt concentrations in pooled gallbladder bile (n = 10 mice per group) were determined according to previously published methods (36). The cholesterol saturation index (CSI) of pooled gallbladder bile was calculated from critical tables (37) that were established for taurocholate, the predominant bile salts in the bile of mice on the lithogenic diet (38). The relative lipid composition of pooled gallbladder bile was plotted on condensed phase diagrams. For phase analysis, the phase limits of the micellar zones and the crystallization pathways were extrapolated from model bile systems developed for taurocholate at 37°C and at a total lipid concentration of 10 g/dl (39).

Quantitative real-time PCR assay
Total RNA was extracted from fresh liver tissues of mice (n = 4 per group) according to previously established methods (29). Primer Express software (Applied Biosystems, Foster City, CA) was used to design the primers on the basis of sequence data available from GenBank. Quantitative real-time PCR assays of the hepatic Gper1, Er, and Er genes were performed in triplicate according to previously established methods (29). The sequences of the primers for these genes have been reported (29). Relative mRNA levels were calculated using the threshold cycle of an unknown sample against a standard curve with known copy numbers. To obtain a normalized target value, the target amount was divided by the endogenous reference amount of mouse -actin as internal control.

Liver compound accumulation study
After liver samples were collected from mouse gallstone experiments, they were immediately frozen and stored at 80°C until analysis. Liver tissues were weighed and stored in Eppendorf tubes. Corresponding liver tissues were used for preparing standard curves in a tissue matrix. The appropriate volume of cold PBS was added to each tissue sample or standard to achieve a tissue concentration of 200 mg/ml. Stainless-steel beads (2-3 mm) were added to the tubes that were then placed in a bead beater for 1-2 min. Tissue samples and standards (100 µl) were then added to a 96-well plate. Standards (100 µl) were also added to a separate 96-well plate. To each tissue well, 400 l of cold acetonitrile containing 100 ng/ml of the internal standard enalapril was added. Plates were vortexed for 5 min at 4°C and then centrifuged at 3,200 rpm at 4°C for 10 min. The supernatant (400 µl) was transferred to a second 96-well plate, evaporated to dryness under nitrogen, reconstituted with 100 µl of 0.1% formic acid in wateracetonitrile (9:1; v/v), and vortexed for 5 min. The samples were then briefly centrifuged and submitted for LC/MS analysis. Compound concentrations were determined on a Sciex API-4000 LC/ MS system in positive electrospray mode. Analyses were eluted from an Amour C18 reverse-phase column (2.1 × 30 mm, 5 µm) using a 0.1% formic acid (aqueous) to 100% acetonitrile gradient mobile-phase system at a flow rate of 0.35 ml/min. Peak areas for the specific mass transitions were integrated using Analyst version 1.5.1. Peak area ratios of analyses to the internal standard were plotted against concentration with a 1/x-weighted linear regression to determine compound concentration.

Statistical methods
All data are expressed as means ± SDs. Statistically significant differences among groups of E 2 -treated OVX mice fed the lithogenic diet and administrated with various doses of CIMBA were assessed by Student's t-test, Chi-square tests, or the Mann-Whitney U test. The 50% excitatory (EC 50 ) and 50% inhibitory concentrations (IC 50 ) were determined by nonlinear regression analysis using GraphPad Prism version 5.02 (GraphPad, La Jolla, CA). All ANOVA analyses were also run using the same statistical software with Tukey's post hoc analysis for multiple comparisons. Statistical significance was defined as a two-tailed probability of P < 0.05.

Design and synthesis of the GPER-selective antagonists, as well as the inhibition of calcium mobilization and cAMP G i/o signaling, by GPER antagonists 5-25
Previous docking studies revealed the potential importance of hydrogen bonding to N310 7.40 (Ballesteros-Weinstein numbering) and E275 6.52 , - stacking with F208 (extracellular loop 2) and H307 7.37 , and various hydrophobic interactions with an extensive hydrophobic pocket (33). For this work, a scaffold was designed, as shown in Fig. 2A (34). The synthetic route that was utilized in the synthesis of the GPER-selective antagonists is summarized in Fig. 2B.
To assess the therapeutic efficacy of the synthesized GPERselective antagonists, the human promyelocytic leukemia (HL-60) cell line was utilized because it displays a high expression of GPER1 in addition to both ER and ER. Thus, using the HL-60 cells, off-target signaling was analyzed with a fluorescence polarization assay that is discussed later. Prior to running antagonism assays, compounds 5-25 were tested for GPER agonism in a similar fashion as G-1, and no compounds showed agonism below 10 M. Within the antagonism assay, nearly every compound exhibited an IC 50 value in the high nanomolar range ( Table 1). For comparison purposes, we determined that within the assay, the IC 50 values of G-15 and G-36 were 1,550 ± 170 nM and 1,350 ± 220 nM, respectively. The previously reported calcium mobilization IC 50 values of G-15 and G-36 varied from the determined values within this study. For G-15, the reported IC 50 value was 185 nM, whereas the G-36 exhibited slightly greater potency with an IC 50 value of 165 nM (18). Variation in the reported and observed IC 50 values for G-15 and G-36 may exist due to differences in the antagonism assay methodology between the study that reported these values and our study. For antagonism assays, Dennis et al. (18) utilized 200 nM G-1 to antagonize against G-15 and G-36 because there was a similar amount of calcium mobilization in SKBr3 cells for both E 2 and G-1 at this concentration. The standard method for performing an antagonism assay requires the EC 80/90 value of the agonist to create the necessary signal window for detecting an inhibitory response (40,41). While the EC 50 value of G-1 has been reported in the literature, based upon this methodology it is unclear whether 200 nM truly represents the EC 50/80 value of G-1. In the HL-60 cell line, the EC 80 value was determined (data not shown) in a dose-dependent manner and existed at approximately 3 M. Using 3 M as the EC 80 value of G-1, there was an approximate 10-fold magnitude difference in the IC 50  The results from the calcium mobilization data revealed several key interactions of antagonists with GPER. In this series of compounds (5-25), the R 1 substitution greatly affected GPER antagonism. Larger hydrophobic groups (isopropyl and tert-butyl) were well tolerated at the R 1 position, whereas the methyl-substituted compounds (12-18) exhibited the lowest antagonism activity. For the R 1 substitution, isopropyl derivatives (5-11) had lower IC 50 values than the tert-butyl derivatives (19)(20)(21)(22)(23)(24)(25). Overall, these findings suggest that small molecules may be limited in size due to restriction within the GPER binding pocket. Changes in activity observed by various R 2 substitutions were less straightforward. While monosubstitution was favored over 3,4disubstitution (11, 18, and 25), there was an insufficient number of compounds to verify whether the 3- or 4-position conferred favorable binding. The larger R 2 substituents such as naphthyl or biphenyl (5, 6, 12, 13, 19, and 20) had mixed tolerances based upon the R 1 substitution. Differences in these observations may exist due to the size of the GPER binding pocket or the existence of different binding modes. The 4-Cl derivatives (10, 17, and 24) were unaffected by the R 1 substitution, whereas the other small substituents were all affected by the R 1 substitution. The two compounds with the lowest IC 50 values, 8 (R 1 = isopropyl; R 2 = 4-OMe) and 21 (R 1 = tert-butyl; R 2 = 4-CH 3 ), had different substitution patterns and illustrated the potential for further optimization in future studies. These results were further verified in a secondary HTRF cAMP assay (Fig. 3A). In this assay, HTRF was inversely proportional to cellular levels of cAMP. Accordingly, it was observed that G-1 increased the HTRF signal, which indicated a decrease in the level of cAMP. Cellular response to G-1 was reversed with the treatment of pertussis-toxin (1 M), suggesting that G-1 signals through the G i/o pathway (Fig. 3A). Blocking GPER with an antagonist exhibited a decrease in the HTFR signal, which corresponded to an increase in cAMP levels. Results with G-36, 8, and 21 show that there was a return to basal levels of cAMP by blocking the activation of the receptor with the experimentally determined cAMP EC 80 value of GPER activation with the agonist G-1 (EC 80 = 3 M) (Fig. 3B). Presented data show % cAMP, which was calculated using the protocol suggested by the manufacturer.

Selectivity of 8 and 21 for GPER over the classical ER
Due to the existence of multiple targets for estrogenic compounds, the selectivity of 8 and 21 for GPER over the classical ER was established. Unlike E 2 , there was no appreciable binding for ER observed by G-1, G-36, or 8 at any tested concentrations (Fig. 4A). At the highest (10 M) concentration evaluated, 21 showed a low level of binding that was significantly different from G-1, G-36, and 8. For this reason, 21 may be limited in effectiveness as a GPER antagonist due to the potential of off-target effects. Studies of ER binding revealed that at high (10 M) concentrations, only G-1 and G-36 exhibited binding (Fig. 4B). In contrast to the G-series, both 8 and 21 showed a lack of binding at the 10 M concentration. Overall, binding studies revealed that both 8 and 21 do not exhibit significant binding to either ER or ER at concentrations below 10 M. Because 8 displayed greater selectivity at high concentrations compared with 21, we have established 8 as the lead compound within this series. Henceforth, the compound 8 [2-cyclohexyl-4-isopropyl-N-(4-methoxybenzyl)aniline] is referred to as CIMBA.

Prevention of estrogen-induced gallstones by CIMBA in OVX mice
Although inbred AKR/J mice have been found to be a gallstone-resistant strain, they are still susceptible to the formation of E 2 -induced cholesterol gallstones (28). Notably, the inbred AKR/J strain expresses Gper1, Er, and Er in the liver. To further explore whether GPER-selective antagonists play a key role in preventing the formation of estrogen-induced gallstones, CIMBA was first studied in OVX AKR/J mice fed a lithogenic diet and treated with exogenous E 2 at 6 g/day for 8 weeks. Figure 5A shows that without treatment of CIMBA (i.e., at 0 g/day/kg), 100% of E 2 -treated OVX mice developed gallstones in response to Fig. 3. A: An HTRF assay was used to determine cAMP accumulation. Forskolin was utilized as a positive control for elevated levels of cAMP. HTRF values for compounds were normalized to the average signal obtained from forskolin (20 M) to determine cAMP accumulation using the manufacturer's suggested protocols. The observed decrease in cAMP for 10 M G-1 was blocked with the use of 1 M pertussis-toxin (PTX), which inhibits the G i/o signaling pathway. B: Compounds 8 (CIMBA) and 21 were compared with a known GPER antagonist, G-36, and all three compounds exhibited the ability to antagonize the EC 80 (3 M) of the G-1-induced inhibition of adenylate cyclase. At 1 M, CIMBA exhibited a similar level of inhibition (P < 0.0001; one-way ANOVA and Tukey's multiple comparison) in response to the EC 80 of G-1. Compound 21 exhibited a less statistically significant inhibition (P < 0.001). A trend in the inhibition of adenylate cyclase was also observed at 100 nM in these three compounds, but only compound 21 exhibited significance (P < 0.01) compared with the EC 80 of G-1. All measurements were done in triplicate. being fed the lithogenic diet for 8 weeks. However, gallstone prevalence was significantly reduced from 80% to 40% in E 2 -treated OVX mice receiving CIMBA from 16 to 32 g/day/kg for 8 weeks. Figure 5B displays representative photomicrographs of amorphous mucin gel, liquid crystals, cholesterol monohydrate crystals, and gallstones in these mice, as observed by phase-contrast and polarizing light microscopy. As shown in Table 2, the highest mol% cholesterol and CSI values in pooled gallbladder bile were found in E 2 -treated OVX mice receiving no CIMBA. In contrast, the mol% of cholesterol in gallbladder bile was gradually reduced with an increase in doses of CIMBA. Fig. 4. A fluorescence polarization binding assay was utilized to assess the off-target binding of G-1, G-36, CIMBA, and compound 21 at ER and ER. Data were compared with the binding of E 2 , a known ligand for both receptors. A: At ER, no appreciable binding was observed for G-1, G-36, and CIMBA up to 10 µM. At the same concentration, compound 21 exhibited significant binding. B: At ER, no binding was observed for either CIMBA or compound 21 at 10 µM. At 10 µM, there was evidence of G-1 binding to ER, and G-36 seemed to have a greater affinity at 10 µM than CIMBA or compound 21. The relative lipid composition of pooled gallbladder bile (n = 10 per group) from E 2 -treated OVX mice injected intraperitoneally with various doses of CIMBA from 0 to 32 g/day/kg and fed the lithogenic diet for 8 weeks is plotted on a condensed phase diagram. Because of an 8 week feeding of the lithogenic diet, the relative lipid composition of pooled gallbladder bile from E 2 -treated OVX mice receiving no CIMBA (i.e., at 0 g/day/kg) is located in the central three-phase zone denoted region C, where at equilibrium, the bile is composed mainly of solid cholesterol crystals, liquid crystals, and saturated micelles. By treating mice with varying doses of CIMBA, the relative lipid composition of pooled gallbladder bile gradually shifts downward. These alterations show that gallstone prevalence is reduced in these mice treated with CIMBA in a dose-dependent fashion. The relative lipid composition of pooled gallbladder bile from E 2 -treated OVX mice fed the lithogenic diet for 8 weeks and treated with CIMBA at 0 (), 16 (), and 32 g/day/kg () is plotted on the condensed phase diagram. D: Effects of E 2 and CIMBA on the expression of Gper1, Er, and Er in the liver. The data are expressed relative to the mRNA levels of Gper1, Er, and Er in the liver of OVX AKR/J mice (n = 4 per group) receiving neither E 2 nor CIMBA, as well as fed the lithogenic diet for 8 weeks, and their relative expression levels are set at 1. Treatment of E 2 at 6 µg/day results in a significant increase in the mRNA levels of the liver Gper1 and Er, but not Er, genes in OVX mice. Notably, the expression of Er is approximately 50-fold less than the expression of Er in the mouse liver (28). Importantly, the expression of Gper1 in the liver is significantly reduced by the GPERselective antagonist, CIMBA, in a dose-dependent manner, even in OVX mice treated with E 2 .
Thus, CSI values of pooled gallbladder bile were dramatically decreased from 1.61 to 1.23 by CIMBA (Table 2), which is consistent with a dose-dependent reduction in gallstone prevalence in E 2 -treated OVX mice receiving various doses of CIMBA (Fig. 5A). Figure 5C shows that without CIMBA treatment, the relative lipid composition of pooled gallbladder bile from E 2 -treated OVX mice fed the lithogenic diet for 8 weeks is located in the central threephase area denoted as region C on a taurocholate-rich bile phase diagram, in which the bile is composed mainly of solid cholesterol monohydrate crystals, liquid crystals, and saturated micelles (39). With an increase in doses of CIMBA, the relative lipid composition of pooled gallbladder bile progressively shifts downward and to the left of the phase diagram. These alterations are caused by a dramatic reduction in cholesterol content, a relative decrease in phospholipid content, and a relative increase in bile salt content (Table 2).
In addition, utilizing pooled liver tissues harvested from the mice used in the gallstone studies described above, we found that CIMBA was absorbed into the liver at both doses as determined by HPLC/MS. When the doses of CIMBA treatment were augmented from 16 g/day/kg to 32 g/day/ kg for 8 weeks, the concentrations of CIMBA in the liver were increased from 5 ng/g liver tissue to 25 ng/g liver tissue in OVX mice. Taken together, these results are consistent with a dose-dependent reduction in gallstone prevalence in E 2 -treated OVX mice receiving CIMBA from 0 to 32 g/day/kg (Fig. 5A). Figure 5D exhibits the effect of E 2 and CIMBA on the expression of Gper1, Er, and Er in the liver. Compared with control OVX mice receiving neither E 2 nor CIMBA, the hepatic expression of Gper1 was significantly increased in OVX mice treated with E 2 at 6 g/day. However, the expression of Gper1 was significantly reduced by CIMBA in a dose-dependent manner. Of note, the expression of Er was significantly increased in three groups of OVX mice treated with E 2 regardless of whether the mice had received varying doses of CIMBA. As expected, Er expression was slightly increased in all mice because its expression was approximately 50-fold lower compared with Er in the mouse liver (28).
To investigate whether CIMBA protects against E 2 -induced gallstone formation through the GPER pathway, we studied OVX ER ( / ) mice fed the lithogenic diet and treated with exogenous E 2 at 6 g/day for 8 weeks. Consistent with our published results (30), at 8 weeks on the lithogenic diet, 30% of OVX ER ( / ) mice receiving no CIMBA (i.e., 0 g/day/kg) suffered from E 2 -induced gallstones (Fig. 6A). However, after 8 weeks of CIMBA treatment at 32 g/day/kg, no gallstones were detected in OVX ER ( / ) mice, and only 40% of these mice formed mucin gel, liquid crystals, and cholesterol monohydrate crystals (Fig. 6B). As shown in Table 2, the CSI value of pooled gallbladder bile in OVX ER ( / ) mice treated with CIMBA at 32 g/day/kg (CSI = 0.84) was markedly lower compared with the CSI value of pooled gallblader bile in OVX ER ( / ) mice receiving no CIMBA (CSI = 1.28). These results indicate that pooled gallbladder bile is unsaturated with cholesterol after 8 weeks of CIMBA treatment in OVX ER ( / ) mice.

DISCUSSION
Despite the implication of estrogen and GPER in a variety of health and disease states, there has been limited clinical success with currently available GPER ligands. Although the G-series (G-1, G-15, and G-36) has become the standard for studies pertaining to GPER, it may be limited by offtarget effects and weak solubility in water or oil. A particular disadvantage of the G-series is the presence of ERE activation at high concentrations (18). To improve the solubility and address promiscuous binding of the G-series, we have rationally and successfully synthesized a new series of the GPER-selective antagonists that provide the first evidence for key binding interactions within the binding pocket of GPER.
We previously explored binding interactions in a series of indole-thiazole derivatives that exhibited agonism at GPER (42). Molecular homology modeling found that the indole-thiazole derivatives exhibited - stacking similar to G-1 of the aromatic groups with F206 45.49 and F208 45.51 within the binding pocket of GPER (33,34,42). These results suggest that the tetrahydroquinoline is not required for GPER activity. Prior modifications of the G-series focused on site-selective alterations to the tetrahydroquinoline structure but have not focused on additional substitution patterns. To explore the binding modes and interactions that are important for antagonism, the indole-thiazole scaffold has been revised. In the revised scaffold, the amide bond was replaced with an amine, and the heteroatoms of the indole-thiazole were changed to substituted benzene ring systems. We hypothesized that removing the heteroatoms in the indole-thiazole would alter the pharmacological activity of the derivatives from agonists to antagonists. The results from this study validated this hypothesis. The structure of the G-series is similar in rigidity to estrogen and other steroids, which may contribute to weak solubility. Additionally, the lack of flexibility may reduce the ability of GPER ligands to effectively probe the binding pocket. To address the concerns of the solubility and probing ability, the scaffold used in this study included more rotatable bonds than the G-series.
Using the hypothesized scaffold, we developed a simple route to synthesize the initial set of compounds (5-25) with varying substituents. The results of calcium mobilization studies show that most compounds in the series can antagonize the EC 80 of the known GPER-selective agonist, G-1. Among these new series of compounds, CIMBA (8) and 21 display superior antagonism of the G-1 signaling activity in the HL-60 cell line. These two molecules have different substitutions at both the R 1 and R 2 positions of our scaffold, CIMBA (R 1 = isopropyl; R 2 = 4-OMe) and 21 (R 1 = tert-butyl; R 2 = 4-CH 3 ). At the R 1 substitution, it appears that GPER prefers bulkier hydrophobic groups, and varying the bulkiness of those groups influences the substitution pattern at the R 2 position. At the R 2 position, smaller electron donating groups are favored when R 1 is either an isopropyl or tert-butyl group. The electron donating nature of these two groups most likely contributes to the electronrich aromatic ring, which influences either aryl-aryl or aryl-cation interactions (43). Trends in efficacy and potency were validated by evaluating cAMP. Overall, the findings from the current studies provide some critical insights into the binding interactions that can be utilized in the development of new ligands for modulating the GPER activity.
The selectivity studies show that CIMBA and 21 specifically bind GPER over ER or ER below 10 M. While CIMBA does not bind to either ER or ER, 21 displays binding to ER at 10 M. These results are not surprising because the molecular modeling studies on GPER and the classical ER from our group and others are in agreement that estrogenic compounds bind within a similar binding pocket as each other (33,34,(44)(45)(46)(47). Due to the proposed similarity between the binding pockets for these three receptors, it is not surprising to observe off-target binding by some GPER ligands at high concentrations (16)(17)(18). In our binding studies, we found that G-36 exhibits an increasing trend in binding at ER. This is the first study to suggest potential off-target effects of G-36.
The results from this study show several key advantages of CIMBA over the G-series antagonists G-15 and G-36. One advantage of CIMBA is the removal of the tetrahydroquinoline moiety. This could reduce potential off-target binding because these compounds contain pan-assay interference moieties that generally lead to off-target effects, confusing structure-activity relationship of ligands, and poor downstream data (48,49). Moreover, removing the tetrahydroquinoline moiety allows for less molecular rigidity and a greater ability to probe the GPER binding pocket. An additional advantage of CIMBA over G-15 and G-36 is that CIMBA does not bind to either ER or ER at high concentrations, thereby preventing the induction of any ER-dependent ERE activation. Compared with the G-series antagonists, CIMBA also appears to be a more potent inhibitor of the calcium release induced by G-1. The increase in potency may be attributed to the increase in flexibility and the ability to probe the binding pocket of the receptor. Furthermore, CIMBA does not contain chiral centers that are present in G-1. The lack of chiral centers in CIMBA prevents a racemic mixture of both active and inactive stereoisomers, allowing for ease of synthesis and more accurate determination of potency. Last, CIMBA is more soluble than both G-15 and G-36 in 1% DMSO solutions. In a preliminary study, we found that both G-15 and G-36 dissolve sparingly in water, oil, and alcohols. To be appropriately dissolved, G-15 and G-36 must be dissolved in a high concentration of DMSO. The restriction in solvents has limited the use of G-15 and G-36 in animal studies due to the significant cytotoxicity of DMSO. In contrast, CIMBA in the ) mice (n = 10) treated with CIMBA at 32 g/day/kg, and only some mucin gels, liquid crystals, and cholesterol monohydrate crystals are detected in 40% of these mice. B: Representative photomicrographs of mucin gels, liquid crystals, cholesterol monohydrate crystals, sandy stones, and real gallstones as observed in the gallbladder bile of E 2treated OVX ER ( / ) mice (n = 10 per group) fed the lithogenic diet for 8 weeks and treated with CIMBA at 0 (top panel; bar = 200 µm) or 32 g/day/kg (bottom panel; bar = 25 µm) as observed by polarizing light microscopy. C: The relative lipid composition of pooled gallbladder bile from E 2 -treated OVX ER ( / ) mice fed the lithogenic diet and treated with CIMBA at 0 (♦) or 32 g/day/kg (◊) for 8 weeks is plotted on the condensed phase diagram. range of pharmacological dosages can be dissolved in alcohol first and then oil so that it is convenient to be used for animal experiments through intramuscular, subcutaneous, or intraperitoneal injection.
As found by clinical and epidemiological studies, the prevalence of cholesterol gallstone disease in women is twice that of men (24), and oral contraceptives or other estrogen therapies significantly increase that risk (26). Although the classical ER plays a critical role in estrogen-induced lithogenic effects, genetic studies have found that Gper1 is a new gallstone gene, Lith18, in mice (31) and is also involved in estrogen-dependent lithogenic pathways (29,50). In a mouse model of E 2 -induced gallstones, CIMBA reduces gallstone formation in a dose-dependent manner in OVX mice. Of note, at the highest (32 g/day/kg) concentration, CIMBA does not completely inhibit gallstone formation in OVX mice. One explanation for this observation is that even higher doses of CIMBA are needed to fully prevent E 2induced gallstone formation. However, the most likely explanation is that because CIMBA does not bind to the classical ERs, ER is still being activated by E 2 , thus leading to gallstone formation (29). Using OVX ER ( / ) mice, we further found that CIMBA protects against the formation of E 2 -induced gallstones by inhibiting the Gper1 activity in the liver. The current results are consistent with previous findings (27), supporting the novel concept that GPER is involved in E 2 -dependent lithogenic actions and works independently of ER. These results also support the notion that both GPER and ER can work through different pathways on hepatic cholesterol and bile salt metabolism, as well as biliary lipid secretion and cholesterol crystallization, to promote E 2 -induced cholesterol gallstones (29).
In summary, the discovery of the potent GPER-selective antagonist CIMBA may provide a novel and alternative strategy for preventing cholesterol gallstones in women and, particularly, in subjects who have to be exposed to high levels of estrogen. The response to CIMBA has established the precedence for GPER antagonists to reduce GPER mRNA levels in the liver. Furthermore, CIMBA and the other compounds from this study offer new pharmacological tools for evaluating the functions of GPER while exploring the ligand binding domain, which may greatly aid in the development of an orally administered, liver-specific GPER-selective antagonist for the prevention of cholesterol gallstone disease in a subgroup of women at high risk.