Discovery of a Potent Thiazolidine Free Fatty Acid Receptor 2 Agonist with Favorable Pharmacokinetic Properties

: Free fatty acid receptor 2 (FFA2/GPR43) is a receptor for short-chain fatty acids reported to be involved in regulation of metabolism, appetite, fat accumulation, and in ﬂ ammatory responses and is a potential target for treatment of various in ﬂ ammatory and metabolic diseases. By bioisos-teric replacement of the central pyrrolidine core of a previously disclosed FFA2 agonist with a synthetically more tractable thiazolidine, we were able to rapidly synthesize and screen analogues modi ﬁ ed at both the 2-and 3-positions on the thiazolidine core. Herein, we report SAR exploration of thiazolidine FFA2 agonists and the identi ﬁ cation of 31 (TUG-1375), a compound with signi ﬁ cantly increased potency (7-fold in a cAMP assay) and reduced lipophilicity (50-fold reduced clog P ) relative to


■ INTRODUCTION
−5 FFA2 is expressed in a variety of tissues including immune cells, 3 adipocytes, 6 pancreatic β-cells, 2 and enteroendocrine L-cells. 7olonic fermentation of dietary fiber produces large amounts of SCFAs, and results indicate that an interplay between FFA2 and dietary fiber via colonic fermentation to produce SCFAs is implicated in promoting a healthy composition of microorganisms in the gut. 8Fiber-rich diets and SCFAs can counteract colitis, 9,10 and exacerbated inflammation in colitis models has been reported for FFA2-knockout mice. 11owever, others have found that FFA2-knockout mice exhibited reduced inflammatory responses in colitis models, 12 and it is currently unclear if antagonists or agonists are preferable for the treatment of intestinal inflammation.So far, the antagonist GLPG0974 (1, Chart 1) is the only FFA2 modulator that has undergone clinical trials but unfortunately failed to meet end points against ulcerative colitis. 13ietary fiber is also known to exert positive effects in diabetic subjects. 14Acetate has been shown to potentiate glucose-stimulated insulin secretion from β-cells in an FFA2dependent manner, 15 and SCFAs were able to promote secretion of the incretin glucagon-like peptide-1 (GLP-1) from colonic cultures through FFA2 16 and both GLP-1 and the appetite regulating peptide YY (PYY) in vivo, 17 indicating that activation of FFA2 could have beneficial effects on type 2 diabetes (T2D).Interestingly and in contrast, loss of FFA2 and FFA3 was found to improve insulin secretion and glucose tolerance. 18Moreover, SCFA-mediated activation of FFA2 has been found to suppress adipose insulin signaling, leading to reduced fat accumulation in adipose tissue, and to promote GLP-1 secretion in the gut. 19−25 Several FFA2 agonists have been disclosed, including an allosteric agonist series from Amgen represented by 2 (known as AMG7703 or 4-CMTB) and an orthosteric agonist series originating from Euroscreen represented by 3. 20,26−30 A recent study found that agonist 4, also originating from Euroscreen, 31 caused only marginal GLP-1 secretion but led to significant PYY mucosal responses, inhibited insulin-promoted fat accumulation and intestinal functions, and suppressed food intake, supporting FFA2 as a possible antiobesity target. 21he FFA2 agonists disclosed hitherto have relatively high lipophilicity, which is associated with several undesired properties, 33−35 and only moderate potency (e.g., see Chart 1).Furthermore, the allosteric agonist 2 has solubility issues and has been reported to give results that differ from orthosteric FFA2 agonists, 15 and the orthosteric agonist 3 was found to have stability issues. 36We therefore wished to explore 4, an FFA2 agonist reported to have properties useful for in vivo studies. 21Initially, we sought to explore the SAR around the pyrrolidine scaffold of 4; however, the synthetic route to obtain this class of FFA2 agonists is rather lengthy and precludes efficient exploration of parts of the molecule.We therefore opted for replacing the central pyrrolidine by a thiazolidine.This replacement strategy has previously been employed for HIV protease inhibitors. 37Herein, we report SAR investigations of substituted thiazolidine FFA2 agonists and the discovery of an FFA2 agonist with improved potency, reduced lipophilicity, and favorable physicochemical and pharmacokinetic properties.
■ RESULTS AND DISCUSSION Synthesis.Pyrrolidine ligands were synthesized in seven linear steps from L-pyroglutamic acid (5, Scheme 1) essentially as previously described by Euroscreen, 31 with the modifications that the ring opening by aryllithium was allowed to reach room temperature and that the amide coupling with pyrrolidine 6 was performed with bis(tetramethylene)fluoroformamidinium hexafluorophosphate (BTFFH). 36hiazolidine ligands were synthesized in three steps by condensation of (R)-Cys-OMe hydrochloride with benzaldehydes via 7a−j (Scheme 2).In contrast to the pyrrolidine route, this strategy allowed rapid access also to variations on the 2-position of the thiazolidine scaffold in a single step from readily available starting materials 38 and formation of the desired (2R,4R)-thiazolidine amide as the major product in the next step. 39ondensation of (R)-Cys-OMe hydrochloride with aldehydes produced mixtures of C(2) epimeric methyl (4R)-2arylthiazolidine-4-carboxylates (7a−j). 38To obtain the (2R,4R)-configuration, 7a−j were treated with Et 3 N and subsequently reacted with the relevant acid chlorides to give the desired (2R,4R)-products. 39Finally, demethylation using LiI in EtOAc afforded the corresponding (2R,4R)-Nacylthiazolidine-4-carboxylic acids. 39 second set of analogues aimed at exploring the N-benzoyl part of the ligands was synthesized from central intermediates 8a,b by Suzuki−Miyaura or Sonogashira cross-coupling followed by demethylation, affording the desired biarylic compounds in low to good yields (Scheme 3).
The crystal structure of 14 confirms unambiguously the (2R,4R)-configuration of thiazolidine substituents (Figure 1).A relatively short intramolecular hydrogen bond of 2.598(2) Å is found between the carboxylic acid and pyridine group.The crystal packing (Supporting Information, Figure S1) shows no significant intermolecular interactions stronger than van de Waals forces.The intramolecular hydrogen bond is a likely reason for why crystals suitable for structural information were Chart 1. FFA2 Modulators a a clogP values are calculated using ChemBioDraw v16.0.1.4.pIC 50 for 1 and pEC 50 for 2 and 3 are previously reported. 28,30,32heme 1. Synthesis of Pyrrolidines 4 and 37 from L-Pyroglutamic Acid a Scheme 2. Synthesis of (2R,4R)-N-Acylthiazolidine-4carboxylic Acids from (R)-Cys-OMe a obtained for this compound but for none of the others.At physiological pH with the carboxylic acid ionized, the pyridyl hydrogen bond will be absent and the overall conformation is expected to be different.It is therefore not possible to draw conclusions regarding the configuration around the amide in the active conformation of 14.In fact, rather than the (Z)configuration apparent in the crystal structure of 14, computer minimizations indicate a general preference for the (E)configuration around the amide in the compound series.
Biological Testing and SAR Analysis.Compounds were tested on the human (h)FFA2 receptor using both a bioluminescence resonance energy transfer (BRET)-based βarrestin-2 interaction assay and a cAMP inhibition assay.The former assay is an upstream signaling assay measuring BRET arising from agonist-mediated recruitment of β-arrestin-2 to FFA2, and the latter assay evaluates inhibition of forskolininduced production of cAMP facilitated by coupling of FFA2 to Gα i/o proteins. 40he FFA2 agonist activity of 4 was confirmed in both assays.Compound 4 has previously been reported with an EC 50 of 81 nM. 21We obtained somewhat lower potencies of 1.2 μM in the β-arrestin-2 assay and 0.53 μM in the cAMP assay.We were pleased to find that the thiazolidine analog 9 had only moderately decreased potency in both the BRET and the cAMP assay (ΔpEC 50 = −0.67 and −0.31, respectively) relative to pyrrolidine 4 (Table 1).The easy access to substituted thiazolidines enabled a rapid screen of variations in the 2-position of the thiazolidine scaffold for 2-methoxy-1,1′biphenyl derivatives (Table 1).
Compared to 2-chlorophenyl (9), 3-chlorophenyl (10) or 4chlorophenyl (11) led to a decrease in potencies in the βarrestin-2 assay, whereas 11 retained potency in the cAMP assay.Thus, the 3-position appeared to be disfavored and 4position substitution may be a route to G-protein-biased ligands.Substituting the 2-chloro by bromo (12) or hydrogen (13) retained potency in both assays.
Installation of 2-pyridyl ( 14) or 2-furyl (15) onto the thiazolidine scaffold eroded potency in the β-arrestin-2 assay compared to the phenyl derivatives but mostly preserved potency in the cAMP assay, again suggesting a route to Gprotein-biased agonists.Since the goal was both to increase potency and decrease lipophilicity, values for ligand-lipophilicity efficiency (LLE) 33 based on calculated lipophilicity of the neutral compounds (clogP) were also used to guide the optimization.The hydrophilic pyridyl led to a marked improvement in LLE for 14 compared to all previous compounds including 4 but insufficient potency.The introduction of a tert-butyl group in the 2-position of the thiazolidine scaffold resulted in a compound ( 16) that only caused 50% receptor activation at 100 μM.Replacing sulfur with oxygen on 16 led to a completely inactive compound (17); hence, the oxazolidine scaffold was not explored further.
We next directed attention to the N-acyl substituent.As evident from 18 (Table 2), the biphenyl system is important for activity, as removing the terminal ring leads to complete loss of activity.Removing only the 2′-methoxy (19) also resulted in a pronounced loss of potency in both assays, as did moving the terminal phenyl to the 3-position (20).The 2′fluoro-3′-methoxy (21) and the 4′-methoxy (22) analogues both led to decreased potency in both assays relative to the 2′methoxy analogue 9. Compared to 19, installation of either acetyl (23) or nitrile (24) in the 4-position markedly boosted   potency in both β-arresin-2 and cAMP assays, indicating that certain hydrogen bond accepting groups are favored in this region.
Extension of the biphenyl 19 by insertion of acetylene between the rings (28) caused a minor loss of potency in the cAMP assay but no changes in the β-arrestin-2 assay.In a parallel manner, alkyne extension of 2-methoxyphenyl (9) into 29 reduced potency in the β-arrestin-2 assay but preserved activity in the cAMP assay.
Overall, 4-biphenyl compounds with the terminal phenyl ring carrying lipophilic (25, 26) or moderately hydrophilic (9)  substituents in the ortho-position showed higher potencies than a nonsubstituted phenyl analogue such as 19 or 20, indicating that a nonplanar biphenyl conformation is favored for optimal binding of this compound class.
With this and the encouraging results from 23 and 24 in mind, we wished to explore biaryl systems with orthosubstituted polar heterocycles as the terminal ring.Satisfyingly, installation of 1-methyl-1H-pyrazolyl produced a significantly improved compound (30), whereas introduction of 3,5dimethylisoxazolyl (31) appeared as the most potent compound of the series with a pEC 50 of 7.11 in the cAMP assay and an LLE increased by 2.3 relative to the initial pyrrolidine 4. Replacement of the methyl groups of 31 by phenyl and cyclopropyl (32) resulted in a 40-fold lower potency in the β-arrestin-2 assay, indicating a limit to the acceptable steric bulk in this region.
Having identified 4-(3,5-dimethylisoxazolyl)benzoyl as a preferred N-substituent, we directed the attention back to the 2-substituent of the thiazolidine scaffold to screen for improved substituents in a second iteration (Table 3).This effort showed again that the 2-chloro (31) was clearly favored  over a naked phenyl (33).The same was true for 2-ethynyl (34) and 2-methoxy (35).Furthermore, increasing the polarity by installation of 2-pyridyl (36) resulted in a weakly potent agonist in the β-arrestin-2 assay, whereas 36 was devoid of any activity when evaluated in the cAMP assay.Since pyridyl analog 14 was a reasonably potent compound in both βarrestin-2 and cAMP assays, the observation that 36 lost activity is likely explained by this compound being too polar to effectively reach the lipophilic binding site of FFA2.
As the initial idea was to use the thiazolidine as a bioisostere for pyrrolidine in the optimization of the compound series, the corresponding pyrrolidine 37 of thiazolidine 31 was synthesized and tested (Table 3).Although pyrrolidine 37 displayed appreciable FFA2 activity, its potency was lower than that of 31.
The occasionally diverging results in the cAMP and βarrestin-2 assays illustrate that potencies in the functional assays do not directly reflect receptor affinity.To better understand the relationship between affinity and functional activity, we proceeded with further examination of 31 and selected analogs in an FFA2 competition binding assay using a radiolabeled version of antagonist 1 (Table 4).It has previously been confirmed that 1 and radiotracer [ 3 H]-1 are orthosteric ligands and fully displaceable by propionate and the synthetic orthosteric agonist 3. 32,41 Pyrrolidines 4 and 37 and thiazolidines 9, 13, 19, and 31 were also all confirmed to fully displace [ 3 H]-1.Parent pyrrolidine 4 exhibited a moderate binding affinity, and the corresponding thiazolidine (9) was found to be a slightly stronger binder, in contrast to the slightly lower potency observed in both functional assays.Upon comparing 9 with the analogue lacking the 2-chloro substituent of the 2-aryl (13) or lacking the 2′-methoxy (19) substituent at the biphenyl system, it becomes clear that both these features contribute significantly to the affinity (ΔpK i = 0.55 and 0.46, respectively).The most potent thiazolidine (31) was indeed found to also exhibit the highest binding affinity within this selection.The corresponding pyrrolidine (37) displayed a significantly weaker binding affinity when compared to 31 (ΔpK i = −0.29),which correlated well with the observed differences in functional potency between these two compounds.Hence, for isoxazoles 31 and 37, the thiazolidine central scaffold boosted both affinity and receptor activation compared to the pyrrolidine.Replacement of the 2methoxyphenyl ( 9) by the more polar 3,5-dimethylisoxazole (31) not only gave rise to the more potent thiazolidine-based agonist (31) but also reduced lipophilicity resulting in a marked improvement in LLE over 4 (ΔLLE = 2.27), and 31 indeed exhibits the highest LLE of all compounds in the study.
Molecular Modeling.To better understand the interaction of 31 with FFA2, the compound was docked in a previously established homology model of the receptor. 32The lowest energy pose of 31 is depicted in Figure 2. Compound 31 was found to effectively engage Arg180 and Arg255 via ionic and hydrogen bonding interactions, an observation that is in agreement with previous reports that both residues are known to be vital for receptor binding and activation. 30,41imultaneously, the carboxylate carbonyl also interacted with the conserved Tyr238 through hydrogen bonding.The 2chlorophenyl moiety situated cis to the carboxylate was observed to interact with Phe89 via displaced π−π interactions, where the chloro substituent appears to orient the phenyl for  optimal contact.Furthermore, the dimethylisoxazole ring formed π−cation interactions with Lys65 and edge-to-face π−π interactions with Trp75 (Supporting Information, Figure S2).The shown pose also reflects higher energy poses.No alternative low-energy pose was identified.
Overall, groups of 31 found in the SAR analysis to be crucial for binding, such as 2-chlorophenyl and dimethylisoxazole, were confirmed by docking to contribute to the ligand receptor interaction, further supporting the proposed binding pose of 31 in hFFA2.Notably, both the carboxylate and the amide carbonyl of 31 were oriented toward the two arginine residues and Arg255 interacted directly with both groups through hydrogen bonds (Figure 2).
Characterization of 31.Since 31 showed the highest potency and LLE based on calculated lipophilicity in both functional assays and the binding assay, it was decided to characterize its properties in further detail.Thus, 31 was tested and confirmed to be active on the murine FFA2 orthologue (pEC 50 = 6.44 ± 0.13 in the cAMP assay).The compound was found to be inactive up to 100 μM concentration at the closely related SCFA receptor FFA3 (formerly GPR41) and evoked no significant agonist or antagonist response in the long-chain fatty acid receptors FFA1 (GPR40) and FFA4 (GPR120) or in nuclear receptors PPARα, PPARγ, PPARδ, LXRα, and LXRβ at 10 μM (see the Supporting Information).
Previous FFA2 ligands are generally, apart from SCFAs, lipophilic compounds.In contrast, experimental lipophilicity of 31 (Table 5) was found to be in the lower part of the optimal range, 34 giving an LLE of 6.2 based on the cAMP assay and measured lipophilicity.The compound also showed high aqueous solubility and chemical stability (Table 5).Recovery after incubation with PBS at pH 7.4 or with fasted-state simulated gastric or intestinal fluid (FaSSGF/FaSSIF) was close to quantitative, indicating excellent chemical stability.Investigation of metabolic stability of 31 in mouse liver microsomes also resulted in a high recovery.
Hepatocyte stability is a better approximation to the in vivo properties than liver microsomes since mitochondrial metabolism and conjugation processes also take place.Incubation of 31 with primary mouse hepatocytes gave a recovery of 72% after 2 h and low intrinsic clearance (<8 (μL/min)/10 6 cells), indicating high metabolic stability in mouse hepatocytes (Supporting Information, Figure S3 and Table S6).
The pharmacokinetic profile of 31 was investigated in mice to establish its properties as a tool for in vivo studies, revealing a satisfactory half-life of 2 h and a moderate to low clearance of 20.3 mL min −1 kg −1 in mice (Table 6).Oral dosing of 10 mg/ kg resulted in a maximal plasma concentration of 6.5 μM after 15 min and a bioavailability of 32%.Furthermore, ip administration of 31 at 5 mg/kg in mice and analysis of plasma concentrations indicated an AUC 0−∞ of 60 500 (ng/ mL)•min a half-life of 82 min and thus a favorable profile relative to 4 (AUC 0−∞ ≈ 25 000 (ng/mL)•min, t 1/2 ≈ 50 min). 21gonist 3 has been shown to affect inflammatory responses in human neutrophils via FFA2 by induction of intracellular calcium and superoxide production. 42Propionate is known to exert beneficial anti-inflammatory effects, 9 and the elimination of FFA2 from mice has previously been shown to cause increased inflammation in tissue models. 11Although the FFA2 antagonist 1 failed to meet the end point in clinical trials with ulcerative colitis patients, the compound was found to modulate neutrophil recruitment, 43 and the therapeutic potential of FFA2 modulation in inflammatory disease is still not clear, especially since neutrophil recruitment may have both beneficial and detrimental effects. 44Hence, 31 and propionate were examined for their ability to promote human neutrophil migration, as it is already established that antagonists, such as 1, can inhibit chemotaxis in vitro and that acetate can induce chemotaxis in mouse neutrophils. 45,46ompound 31 at 1 μM significantly induced migration of human neutrophils at a level comparable to propionate (C3) at a physiologically relevant concentration, 47 an effect that was more than doubled at 10 μM (Figure 3).
SCFAs have been shown to inhibit lipolysis in adipocytes mediated through FFA2 via the Gα i pathway. 48In this study we demonstrate that thiazolidine 31, in particular when evaluated in cAMP assays, is a potent agonist of hFFA2 with  a Administered at 10 mg/kg po and 5 mg/kg iv in Balb/c mice.

Journal of Medicinal Chemistry
Article retained activity at mFFA2.Therefore, 31 was evaluated for its ability to inhibit isoproterenol-induced lipolysis in murine adipocytes, a process known to be Gα i dependent. 48Even though 31 is slightly less potent on murine FFA2, 31 was, as predicted, 50-fold more potent than propionate in inhibiting lipolysis in isoproterenol-stimulated adipocytes (Figure 4).

■ CONCLUSION
Substitution of the central pyrrolidine of a published FFA2 agonist by a synthetically more tractable thiazolidine and rapid screening of substituents at this scaffold led to the discovery of the potent and selective FFA2 agonist 31, a compound with lipophilicity in the lower part of the optimal range and therefore also high LLE.The compound furthermore has excellent solubility, high chemical, microsomal, and hepatocyte stability, and favorable pharmacokinetic properties.Compound 31 was able to induce migration of human neutrophils and to inhibit lipolysis in murine adipocytes.Collectively, these properties render 31 an interesting tool compound for further exploration of FFA2 as a potential therapeutic target for treatment of inflammatory and metabolic diseases.

■ EXPERIMENTAL SECTION
Synthesis.All commercial starting materials and solvents were used without further purification unless otherwise stated.THF was freshly distilled from sodium/benzophenone. MeOH was freshly distilled from Mg. DCM was freshly distilled and stored over 4 Å sieves.TLC was performed on TLC silica gel 60 F254 plates (Merck) and visualized at 254 nm or by staining with ninhydrin, KMnO 4 , or FeCl 3 stains.Petroleum ether (PE) refers to alkanes with bp 60−80 °C.Water used in reactions was demineralized, and water used for freeze-drying was filtered Milli-Q.Purification by flash chromatography was carried out using silica gel 60 (0.040−0.063 mm, Merck).Test compounds suspended in Milli-Q water (1 mL/10 mg, 3 drops MeCN) were lyophilized on a Heto Drywinner freeze-dryer. 1 H and 13 C NMR spectra were recorded at 400 and 101 MHz, respectively, on a Bruker Avance III 400 at 300 K. Spectra were calibrated relative to residual solvent peaks: 1 H NMR (CDCl 3 ), 7.26 ppm; 13 C NMR (CDCl 3 ), 77.16 ppm; 1 H NMR (DMSO-d 6 ), 2.50 ppm; 13 C NMR (DMSO-d 6 ), 39.52 ppm; 1 H NMR (CD 3 OD), 3.31 ppm; 13 C NMR (CD 3 OD), 49.00 ppm; 1 H NMR (CD 2 Cl 2 ), 5.32 ppm; 13 C NMR (CD 2 Cl 2 ), 53.84 ppm.Rotamer chemical shift values in 1 H NMR are marked by " * " and have been assigned where possible.Rotamer peaks in 1 H NMR are labeled by an integral equal to that of its main partner. 1 H and 13 C spectra of 31 were obtained at 600 and 151 MHz, respectively, on an Agilent NMRS 600 at the indicated temperature (Supporting Information).High-resolution mass spectra (HRMS) were recorded on a Bruker micrOTOF-Q II (ESI).Specific optical rotation was recorded on Anton Paar MCP 100 polarimeter.Purity was determined by HPLC and confirmed by inspection of NMR spectra ( 1 H and 13 C NMR).HPLC analysis was performed using a Dionex 120 C18 column (5 μm, 4.6 mm × 150 mm); flow, 1 mL/ min; 10% MeCN in water (0−1 min), 10−100% MeCN in water (1− 10 min), 100% MeCN (11−15 min), with both solvents containing 0.1% formic acid as modifier; UV detection at 254 nm.All test compounds were of ≥ 95% purity.None of the test compounds contain substructures associated with pan-assay interfering activities (PAINS) 49 by inspection or by screening for PAINS or aggregators at http://zinc15.docking.org/patterns/home.
Molecular Modeling.A homology model of hFFA2 in complex with 1 was obtained as described previously. 32The model was based on a crystal structure of hFFA1 (PDB code 4PHU). 54As 31 fully displaces [ 3 H]-1 at increasing concentrations in a radioligand binding assay (Table 4), the compound was docked into the binding site defined by 1. Prior to docking, 31 was prepared (LigPrep, version 2.7, Schrodinger, LLC) using the OPLS-2005 force field. 55Ionization states were generated using Epik at pH 7.0 ± 2.0 (Epik, version 2.5, Schrodinger, LLC).Induced-fit docking was performed using the IFD 2006 protocol (Glide version 5.9, Schrodinger, LLC; Prime version 3.2, Schrodinger, LLC).Ligand conformational sampling was executed using default settings; initial Glide docking was performed using standard settings; the maximum number of poses of 31 was restricted to 20.Redocking was executed for 31-hFFA2 complexes within 30 kcal/mol of the lowest energy ligand−protein complex.Residues were refined within 5 Å of bound 31.
In Vitro Assays.β-Arrestin-2 Interaction Assay.HEK293T cells were maintained in Dulbecco's modification of Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 2 mM L- glutamine, and 1× penicillin/streptomycin mixture (Sigma) at 37 °C and 5% CO 2 .Cells were grown to 60% density and co-transfected at a 4:1 ratio with plasmids encoding an eYFP-tagged form of the receptor construct of interest and a β-arrestin-2 fused to Renilla luciferase using a polyethyleneimine-based transfection protocol. 56,57Cells were transferred into white 96-well plates at 24 h post-transfection and incubated for 24 h at 37 °C.Immediately prior to conducting the assay, cells were washed, and the culture medium was replaced with Hanks' balanced salt solution.To quantify β-arrestin-2 recruitment to the receptor induced by FFA2 agonists, the Renilla luciferase substrate coelenterazine h (Nanolight Tech, Pinetop, CA) was added to a final concentration of 2.5 μM, and cells were incubated for 5 min at 37 °C.Next, varying concentrations of agonist were added, and cells were incubated for an additional 10 min at 37 °C.BRET resulting from receptor-β-arrestin-2 interaction was assessed by measuring the ratio of luminescence at 535 and 475 nm using a PHERAstar FS plate reader fitted with the BRET1 optic module (BMG Labtech, Aylesbury, U.K.).pEC 50 values were determined with at least three independent replicates.cAMP Assay.All cAMP experiments were performed using Flp-In T-REx 293 cells modified to express receptors of interest in an inducible manner, which were maintained in DMEM without sodium supplemented with 10% fetal bovine serum, 1× penicillin/ streptomycin mixture, 5 μg/mL blasticidin, and 200 μg/mL hygromycin B. 58 Experiments were carried out using a homogeneous time-resolved FRET-based detection kit (Cis-Bio Bioassays, Codolet, France) in accordance with the manufacturer's protocol.Cells were plated at 2000 cells/well in low-volume white 384-well plates.The ability of compounds to inhibit 1 μM forskolin-induced cAMP production was quantified following a 30 min incubation of agonist compounds with cells, which were induced to express receptors of interest by a 16 h treatment with 100 ng/mL doxycycline.pEC 50 values were determined with at least three independent replicates.
Radioligand Binding Assay.Radioligand competition binding experiments were performed as established previously. 41The radioligand [ 3 H]-1 at approximately K d concentration was incubated with varying concentrations of unlabeled compounds and 5 μg of purified membranes isolated from Flp-In T-REx cells induced to express the receptor construct of interest.Nonspecific binding of the radioligand was determined in the presence of 10 μM CATPB. 30After a 2 h incubation at 25 °C, bound [ 3 H]-1 and free [ 3 H]-1 were separated by rapid vacuum filtration and radioactivity was quantified by liquid scintillation spectrometry.To determine the affinity of unlabeled ligands in terms of K i values, competition binding curves were fit to an inverse three-parameter sigmoidal one-site K i value fit with radioligand affinity and concentration as constraints.K i values were determined with n = 6 independent experiments for 31 and with n = 3 for all other compounds.
Isolation of Human Neutrophils and Migration Assay.Neutrophils were isolated from human whole blood as described previously. 59After isolation, neutrophils were immediately resuspended in RPMI 1640 containing 0.5% fatty acid-free bovine serum albumin.Test compounds were prepared at the indicated concentrations in the same buffer and added at the bottom of a 96well plate (Sigma-Aldrich).Inserts were then mounted to the plate, and neutrophils were added (3000 cells/μL).Cells were incubated at 37 °C for 1.5 h, and migrated cells were then collected and ATP content was assessed using ATPlite luminescence assay system (PerkinElmer) according to the manufacturer's instructions.Each experiment was performed with n = 4−6 independent replicates.
Derivation of Primary Mouse Adipocytes and Lipolysis Assay.Epididymal fat was collected from male mice, and isolation of preadipocytes was obtained as described previously. 59Animals were cared for in accordance with national guidelines on animal experimentation.Preadipocytes were subsequently differentiated in DMEM medium (DMEM, 10% fetal bovine serum, 4 mM glutamine, 10 mM HEPES, 10 μg/mL insulin, 25 μg/mL sodium ascorbate, 10 μM rosiglitazone) for 8 days.After this, mature adipocytes were challenged with test compounds and glycerol production was quantified as previously reported. 59Each experiment was performed with n = 3−4 independent replicates.Chemical Stability.A 10 mM stock solution of 31 in DMSO was diluted in phosphate buffer (10 mM, pH = 7.4) obtaining 1.2 mL of 50 μM compound solution in phosphate buffer.The sample was shaken (650 rpm) at 37 °C, and samples were taken out at different time intervals during the 3-week experiment.The study was performed in duplicate, each parallel and with double injection.The stability was calculated based on peak area of the 0 point sample.
Stability in Simulated Gastric and Intestinal Fluids.FaSSIF and FaSSGF were prepared in accordance with the manufacturer's procedure.A 50 μM solution of 31 was prepared from FaSSIF/ FaSSGF and a 10 mM stock solution of 31 in DMSO, and the samples were incubated in a thermomixer (37 °C, 650 rpm) for 0 and 120 min.Afterward, the samples were quenched with acetonitrile (+0.5% HCOOH), centrifuged for 10 min at 10 000 rpm and the supernatant was analyzed by HPLC.The stability was calculated based on peak area of the 0 point sample.The study was performed in duplicate.Average values are reported.
Microsomal Stability.Microsomal stability was studied in mouse liver microsomes (0.5 mg/mL) at a final test compound concentration of 1 μM, essentially as previously described. 60In brief, prewarmed (37 °C) 0.1 M PBS 7.4 was added to 10 mM NADPH in PBS and test compound 31 (1 mM in DMSO).The samples were incubated for 5 min at 37 °C before addition of newly thawn microsomes.The samples were mixed by gentle vortexing and incubated for 1 h at 37 °C, 300 rpm in a thermomixer.Samples were quenched by addition of ice-cold MeOH/MeCN (1:1) and centrifuged for 5 min at 10 000g.The supernantant was transferred to HPLC vials and stored in the freezer until analysis by HPLC.The metabolic stability was calculated based on the 0 min sample.The experiment was performed in triplicate.

■ ASSOCIATED CONTENT
* S Supporting Information The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jmedchem.8b00855.
Synthetic procedures, collection and crystallographic data for 14, and high-and low-temperature 1 H and 13 C NMR spectra of 31 (PDF) Coordinates for crystal structure of 14 (PDB) Molecular strings formula and some data (CSV)

Scheme 3 .a
Scheme 3. Synthesis of Biaryl Analogues via Coupling Reactions a

Figure 2 .
Figure 2. Lowest energy pose of 31 (gray) docked into hFFA2 (gold) with strongly interacting residues shown (cyan) and hydrogen bonds highlighted (yellow stipples).(A) Receptor as ribbon viewed from the angle of helix 4 and 5. (B) Same pose with receptor as surface viewed from the angle of helix 3 and 4.

Table 2 .
SAR Investigation of the N-Acyl Substituent

Table 4 .
Binding Affinities of Compounds for FFA2 a Binding affinities of agonists determined using [ 3 H]-1 in radioligand binding assays.