Discovery of Isoquinolinoquinazolinones as a Novel Class of Potent PPARγ Antagonists with Anti-adipogenic Effects

Conformational change in helix 12 can alter ligand-induced PPARγ activity; based on this reason, isoquinolinoquinazolinones, structural homologs of berberine, were designed and synthesized as PPARγ antagonists. Computational docking and mutational study indicated that isoquinolinoquinazolinones form hydrogen bonds with the Cys285 and Arg288 residues of PPARγ. Furthermore, SPR results demonstrated strong binding affinity of isoquinolinoquinazolinones towards PPARγ. Additionally, biological assays showed that this new series of PPARγ antagonists more strongly inhibit adipocyte differentiation and PPARγ2-induced transcriptional activity than GW9662.

rosiglitazone 1 contains a hydrogen bond between a nitrogen atom in rosiglitazone and the hydroxyl group of Tyr473, which lies in H12 (PDB: 2PRG) 13 . This interaction helps rosiglitazone stabilize conformational changes in PPARγ , particularly in the transcription cofactor-binding AF-2 region of H12 14 . In contrast, a PPARγ antagonist, GW9662 (in non-covalent complex with PPARγ , PDB: 3E00) does not have any interaction with H12 15 .
The LBD of nuclear receptors that contains the AF-2 region, is the primary site investigated for drug discovery. Our research group has succeeded in designing androgen receptor antagonists, nicotinamides, and demonstrated that the antagonist effect of these analogues is a result of their effect on the conformation of H12; agonists lock the conformation of H12 giving a closed conformation of ligand binding pocket (LBP), while antagonists give an open conformation of LBP 16 . On the basis of this principle, we investigated and synthesized isoquinolinoquinazolinones as a novel class of PPARγ antagonists. Compared with well-known PPARγ antagonists, such as GW9662, isoquinolinoquinazolinones which resemble berberine can be expected to possess more drug-like characteristics. Herein, we report a new series of PPARγ antagonists, which is much more potent than GW9662 according to biological evaluations.

Drug Design
We have previously reported the modification of protoberberines by altering the ring size or introducing a heteroatom into ring B [17][18][19][20][21] . In order to investigate a new series of PPARγ antagonists, we initially focused on 5-oxaprotoberberines, a class of berberine bioisosteres. The oxaprotoberberines affected adipogenesis; however, the activity was not better than berberine (Table 1, 10a-h). For an effective rational design strategy for PPARγ antagonists, molecular modeling was used to study the interaction between oxaprotoberberines and the GW9662 binding pocket of the PPARγ -GW9662-RXRα -retinoic acid-NCoA-2-DNA complex (PDB: 3E00) 15 .
Oxaprotoberberines, as shown in Fig. 1B, do not interact with H12, and the tetracyclic core is positioned in a hydrophobic region of the pocket. In addition, the oxygen atom in ring B is close to Cys285; this gave rise to an idea of exchanging the O atom with -NH, to generate isoquinolinoquinazolinones (Fig. 1C,D).
The binding mode between isoquinolinoquinazolinones and PPARγ indicates that the new amino group is located exactly where it was predicted, interacts through hydrogen-bonding to the Cys285 amino acid residue on helix 3 of PPARγ , and does not affect the conformation of H12 (Fig. 1C). These results provided an incentive to investigate the effects of isoquinolinoquinazolinones on PPARγ inhibition and adipocyte differentiation.

Results and Discussion
The isoquinolinoquinazolinones 8a-o were synthesized using a strategy that was similar to our previously reported synthesis method of oxaprotoberberine 20 . The synthesis of isoquinolinoquinazolinones 8 involved a cycloaddition reaction between lithiated toluamides and 2-aminobenzonitriles (Scheme S1). The o-toluamides 5 were deprotonated using n-BuLi and reacted with the 2-aminobenzonitriles 6 to give the 3-arylisoquinolones 7. Finally, intramolecular cyclization was performed using diiodomethane and K 2 CO 3 as base. Furthermore, 8o was reduced with lithium aluminium hydride to give the dihydro derivative 9 (Scheme S2).
The GW9662-mediated inhibition of PPARγ was confirmed using an Oil Red O staining assay to measure adipocyte differentiation 8 . To investigate whether the novel compounds could inhibit adipocyte differentiation, 3T3-L1 preadipocytes were incubated with differentiation medium (MDI; insulin, dexamethasone, and isobutyl methyl xanthine) in the presence of increasing concentrations of isoquinolinoquinazolinones. Adipogenesis was analyzed using Oil Red O staining after treatment with the compounds. Most of the compounds showed potential inhibitory activity towards adipocyte differentiation ( Fig. 2A and Table 1).
A brief structure and activity relationship (SAR) study was performed in the context of adipocyte differentiation inhibition. The different substituents in ring D affect the inhibitory activity of 5-oxaprotoberberines 10.
Compounds that contained a methyl group at both the C10 and C11 positions or a methoxy group at both the C11 and C12 positions (10c and 10d) were more active than those with a single methyl group at C11 or C12 (10a and 10b). Installation of a methylene dioxyl group across the C2 and C3 positions in ring A also increased the inhibitory activities (10e, 44.8% and 10f, 30.5%). In the case of isoquinolinoquinazolinones, introducing a single methyl group at C10 (8f and 8g) or a methyl group at both C10 and C11 (8h-j) decreased the activity. Introducing a chlorine substituent at C2 or C3 (8e, 8g, 8i, 8j) can increase the inhibitory activity. Among all the compounds tested, the isoquinolinoquinazoline compounds that had methoxy groups at both C10 and C11 and a chlorine substituent on ring A (8n and 8o) inhibited adipocyte differentiation with the greatest potency (71.5% and 82.4%, respectively). The mechanisms of action of these two compounds were further investigated.
Prior to testing the PPARγ antagonist potential of isoquinolinoquinazolinones, their cytotoxicity in 3T3-L1 cells (normal cells) was examined. At a concentration of 25 μ M, compounds 8n and 8o exhibited little or no cytotoxic effect (> 70% cell viability; Fig. 2B). These two compounds were then tested for PPARγ inhibitory activity. PPARγ alone or together with CCAAT-enhancer-binding protein α (C/EBPα ) regulates many adipocyte genes that are involved in creating and maintaining the adipocyte phenotype 22 . To further analyze the effect of isoquinolinoquinazolinones on adipogenesis, the expression of the adipocyte marker genes (C/EBPα and PPARγ) was analyzed. The differentiation of 3T3-L1 cells into adipocytes was induced by incubating in MDI induction medium for 2 days followed by incubation in differentiation medium for an additional 6 days. After MDI-induced adipocyte differentiation of 3T3-L1 cells, adipogenic transcriptional factors were analyzed. The transcriptional factors C/EBPα and PPARγ were inhibited by 8n and 8o in terms of both protein (Fig. 2C,D) and RNA (Fig. 2E,F) levels. In addition, the mRNA expression of the genes downstream of PPARγ including adiponectin and fatty acid synthase markers (FAS) were strongly reduced by treatment with 8o. The results indicated that isoquinolinoquinazolinones may suppress adipogenesis by affecting the PPARγ pathway.
To analyze whether isoquinolinoquinazolinones inhibit the transactivation of PPARγ , 3T3-L1 cells were cotransfected with full-length PPARγ 2 expression vector with a peroxisome proliferator response element (PPRE)-driven luciferase reporter gene (PPRE-Luc) that has been reported to respond to PPARγ 23,24 . The effectiveness of PPARγ 2 transfection was inhibited by every compound tested (berberine, GW9662, 8n and 8o) in a dose-dependent manner. Titration curves were generated using Graph Pad Prism ® , and the IC 50 value of each compound with respect to PPARγ 2-induced transcriptional activity was determined. The IC 50 of 8o was 2.43 μ M, while the IC 50 values of 8n, berberine, and GW9662 were 5.02 μ M, 6.63 μ M, and 4.47 μ M, respectively. Compound 8o showed the greatest inhibitory effect (lowest IC 50 value) on PPARγ 2 transcriptional activity (Fig. 3A). To examine isoquinolinoquinazolinone-mediated inhibition of ligand-induced PPARγ activity, the PPARγ agonist rosiglitazone was used. Luciferase expression was increased by up to 3-fold by rosiglitazone treatment. Treatment with a 5 μ M dose of 8o either alone or with rosiglitazone stimulation resulted in the inhibition of PPARγ 2 transfection by 82% and 88%, respectively. This was greater than berberine (33% and 40%) and GW9662 (46% and 48%) (Fig. 3B). The results showed that isoquinolinoquinazolinones were more potent inhibitors of PPARγ 2-induced transcriptional activity than GW9662 in both the absence and presence of the PPARγ agonist rosiglitazone.
A molecular modeling study was conducted to examine the hypothetical binding modes of PPARγ and isoquinolinoquinazolinones including 8o that exhibited potent inhibitory activities. The surflex-Dock program in Sybyl-X 2.1.1 was used to dock isoquinolinoquinazolinones into the ligand binding site of PPARγ following the removal of the ligand from the 3D crystal structure of PPARγ -GW9662-RXRα -retinoic acid-NCoA-2-DNA complex (PDB: 3E00). According to the docking model, isoquinolinoquinazolinone 8o occupied the hydrophobic region of PPARγ -LBP (Fig. 4). The amino group in ring B and methoxyl group in ring D formed hydrogen bonds with Cys285 and Arg288, respectively. Moreover, isoquinolinoquinazolinone 8o was located far away from H12. Furthermore, docking models showed that the isoquinolinoquinazolinones like 8l, which lacked significant inhibitory effect on adipocyte differentiation (39.1%), despite of subtle structural difference (absence of 10-OCH 3 ) than the potent counterpart like 8o (inhibitory effect on adipocyte differentiation: 82.9%) was basically due to unfavorable orientation of the compound in the LBP (Fig. S2).
A surface plasmon resonance (SPR) analysis demonstrated that 8o binds to PPARγ 2-LBD (Fig. S1). A Reichert SR7500 (Reichert, Depew, NY) biosensor was used to measure the binding affinity of rosiglitazone, GW9662, and 8o with PPARγ 2-LBD. PPARγ 2 protein was immobilized on the sensor chip, and the response (in resonance unit; RU) as a result of binding was continuously recorded and presented graphically as a function of time 25 . The association rate (k a also known as on rate (k on )), dissociation rate (k d also known as off rate (k off )), and equilibrium dissociation constant (K D ) were calculated using the Scrubber 2 program (Table 2). Rosiglitazone binds to PPARγ 2 quickly (k a : 24 M −1 s −1 ) and dissociates easily (k d : 0.0867 s −1 ) (Fig. S1A). On the other hand, GW9662 binds slowly to PPARγ 2 (k a : 15.09 M −1 s −1 ) but hardly dissociates (k d : 0.000059 s −1 ) from the protein confirming the irreversible covalent linkage of the ligand with PPARγ (Fig. S1B). Compound 8o binds and dissociates from PPARγ with intermediate rates (k a : 24 M −1 s −1 ; k d : 146.25 s −1 ) (Fig. S1C). The binding affinity of 8o to PPARγ is 4-folds more than rosiglitazone (K D : 146.25 μ M (8o), 581.88 μ M (rosiglitazone)) but is 37-folds less than GW9662 (K D : 3.91 μ M). The SPR result correlates well with the competitive assay of 8o and rosiglitazone.
To investigate the contribution site on the PPARγ 2 to the interaction with 8o, we generated mutant forms of the PPARγ 2 (Fig. 5). In the docking results, 8o showed the interaction with Cys285 and Arg288. Thus, we tested Cys-to-Gly (Cys285), Arg-to-Gly (Arg288) substitution mutants of PPARγ 2. Interestingly, the substitution of Cys285 to Gly abolished the inhibitory effect of both GW9662 and 8o on PPARγ 2 activity. However, the  substitution of Arg288 to Gly abolished the inhibitory effect of 8o not in GW9662. The results indicated 8o has interaction with both Cys285 and Arg288, which are in agreement with the docking study.

Conclusion
Based on the PPARγ antagonist behavior of berberine and the fact that the LBD of PPARγ contains a coactivator binding site, we designed a series of isoquinolinoquinazolinones. This novel class of PPARγ antagonists was synthesized in two steps and inhibited 3T3-L1 adipocyte differentiation by inhibiting transcription factor PPARγ . Compound 8o more effectively suppressed PPARγ transactivation than the existing PPARγ antagonist GW9662. The potent inhibitory effect of the isoquinolinoquinazolinones was well explained by their docking mode in the LBP of PPARγ . A molecular modeling study showed that, as expected, it was the presence of a quinazolinone -NH in ring B that enabled the isoquinolinoquinazolinones to have higher PPARγ inhibitory activities than GW9662. All of the biological assays, the SPR results and mutation study indicated that we have successfully designed and synthesized a new series of potent PPARγ antagonists. SAR data and docking models can provide helpful guidance in designing PPARγ antagonists. All of these findings suggest that isoquinolinoquinazolinones might exert multiple therapeutic effects and are potential treatments for obesity, type 2 diabetes, hyperlipidemia, and other metabolic syndromes.

Experimental Procedures
General Information and Instrumentation. Melting  . TLC was performed using plates coated with silica gel 60 F254 (Merck). Chemical reagents were purchased from Sigma-Aldrich and Tokyo Chemical Industry Co., Ltd. and were used without further purification. Solvents were distilled prior to use; THF was distilled from sodium/ benzophenone. All reactions were conducted under a nitrogen atmosphere in oven-dried glassware with magnetic stirring. The specifications of HPLC analysis are as follows: column, ACE C18-HL, 250 × 2.1 mm, flow rate, 0.2 mL/min; wavelength, 254 nm; mobile phase; acetonitrile:water (9:1, v/v). The purity of compounds was established by integration of the areas of all peaks detected and is reported for each final compound. All compounds tested in the biological assay showed more than 95% purity.
Cell culture and differentiation conditions. The mouse preadipocyte cell line 3T3-L1 was maintained at 37 °C in humidified air with 5% CO 2 . 3T3-L1 cells were cultured in Dulbecco's modified Eagle medium (DMEM; Life Technologies) supplemented with 10% bovine serum (BS; Gibco Invitrogen, Carlsbad, California, USA) as growth medium. For adipocyte differentiation, cells were cultured for 2 days to full confluence in a 24-well plate and the growth medium was then replaced (day 0) with DMEM supplemented with 10% fetal bovine serum (FBS, Gibco Invitrogen, Carlsbad, California, USA), 5 μ g/mL insulin (Sigma, St. Louis, Missouri, USA), 0.5 mM 3-isobutyl-1-methylxanthine (Sigma), and 1 μ M dexamethasone (Sigma). After 48 h, the differentiation medium was replaced (day 2) with DMEM + 10% FBS containing 5 μ g/mL insulin, and the cells were allowed to accumulate lipid droplets until experimental use.
Luciferase reporter assays. 3T3-L1 cells were transfected for indicated combinations of expression plasmids along with a luciferase reporter plasmid (PPRE-Luc). PPRE-Luc contains the consensus PPAR responsive element (PPRE). pCMV-β -gal was co-transfected for normalization of transfection efficiency. 24 h after transfection, cells were treated with indicated compounds for 12 h. Luciferase activities were measured using a luciferase reporter assay kit (Promega, Madison, WI, USA). Experiments were performed in triplicate and repeated at least three times.
Oil Red O Staining. The accumulation of lipids signifying the formation of adipocytes was observed by staining the differentiated cells with Oil Red O. Following differentiation, cells were washed twice with phosphate-buffered saline (PBS), fixed with 10% formalin for 60 min. Oil Red O stock solution (0.5%) was prepared in isopropanol and filtered in cellulose nitrate filters. Cells were stained with Oil Red O diluted 6:4 in water for 1 h at room temperature. Excess Oil Red O dye was washed off twice with distilled water and then dried. The stained lipid droplets within cells were visualized using an optical microscope and photographed with a digital camera at 100× magnification.
Molecular docking. The docking study was performed in Sybyl-X 2.1.1 (winnt_os5x) using the Surflex Dock program. The structure of PPARγ -GW9662-RXRα -retinoic acid-NCoA-2-DNA complex was downloaded from the Protein Data Bank (PDB: 3E00). Structures of RXRα , retinoic acid, NCoA-2, and DNA were removed. The ligand (GW9662) was extracted. Hydrogens were added and minimization was performed using the MMFF94s force field with MMFF94 charges, by using a conjugate gradient method, distance dependent dielectric constant and converging to 0.01 kcal/mol Å. Protomol, an idealized representation of a ligand that makes every potential interaction with the binding site, was generated on the basis of ligand mode. Oxaprotoberberines and isoquinolinoquinazolines were constructed in Sybyl; energy was minimized with MMFF94s force field and MMFF94 charges and stored in a Sybyl database. The compounds in the Sybyl database were docked into the binding site by Surflex Dock on the basis of the protomol constructed earlier.
Surface Plasmon Resonance analysis. Analysis of the interaction between immobilized PPARγ 2-LBD and rosiglitazone, GW9662 and 8o was performed using Reichert SR7500 Surface Plasmon Resonance dual channel instrument (Reichert, Depew, NY). Immobilization of the protein on the hydrophilic carboxymethylated dextran matrix of the sensor chip (Reichert) was carried out using a standard primary amine coupling reaction. Free carboxyl groups on the surface were modified by injecting a mixture of 0.1 M 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride and 0.05 M N-hydroxysuccinimide at a flow rate of 20 μ L/min to generate a reactive succinimide ester surface. Baseline equilibration was achieved by flushing the chip with PBS buffer for 1-2 h. All of the SPR data was collected at 25 °C with PBS as running buffer at a constant flow of 30 μ L/min. All the sensorgrams were processed by using automatic correction for nonspecific bulk refractive index effects. All the kinetic analyses of the binding to PPARγ 2-LBD were calculated using the Scrubber2 program (Bio-Logic Software).