Selective targeting of PPARγ by the natural product chelerythrine with a unique binding mode and improved antidiabetic potency

Considering the relationship between PPARγ and insulin resistance, we performed a high-throughput AlphaScreen™ assay, which determines the efficacy of small molecules in influencing binding affinity of PPARγ with coactivator peptides. Results from the AlphaScreen™ revealed that a type of QBA-chelerythrine weakly stimulated the binding affinity of PPARγ LBD and its coactivators. Notably, the chemical structure of chelerythrine exhibits a distinct scaffold from that of rosiglitazone (), which indicates that it may exhibit different activities. To confirm the binding potency of chelerythrine to PPARγ, we used a Lantha TR-FRET competitive binding assay to compare the half-maximum inhibitory concentration (IC50) of chelerythrine and rosiglitazone. The IC50 of chelerythrine is 566 nM (), which is ten-fold greater than that of rosiglitazone but less than that of another TZD, pioglitazone. Furthermore, in the transactivity assay, we used a GAL-4 driven reporter to determine the specificity and transactivity of chelerythrine. In various highly conserved NR LBDs, chelerythrine specifically activated PPARγ with a low transactivity distinct from rosiglitazone (and). Interestingly, as Supplementary Figure 2 indicated, the 10 μM of chelerythrine did not induce any adipogenesis, which was similar with the vehicle control, while 10 μM of rosiglitazone induced obvious differentiated adipocytes, demonstrating that the adipogenic activity of chelerythrine is much weaker than rosiglitazone (). Collectively, these data indicate that chelerythrine is a bona fide PPARγ selective modulator with high binding potency but weak classical agonism. ²⁶ Fig. 1a Fig. 1b Fig. 1c Supplementary Figure 1 Supplementary Figure 2

One characteristic of PPARγ ligands is that they regulate different target gene groups by regulating the recruitment and release of different cofactors. When binding to PPARγ LBD, agonists induce coactivator recruitment and corepressor dissociation, whereas antagonists behave in the opposite way. These cofactors bind to the nuclear receptor complex, modify the basic chromatin structure and sequentially modulate transcriptional machinery to regulate the target genes. ²⁷

We then used the AlphaScreen™ assay to determine the ability of chelerythrine to modify the cofactors profiling of PPARγ. Asdemonstrates, chelerythrine recruited the coactivator peptides from SRC1, SRC2, SRC3 and PGC-1α to PPARγ more weakly than rosiglitazone. Interestingly, with respect to its corepressor activity, chelerythrine slowed the dissociation of NCoR-2, a corepressor peptide from NCoR, compared to rosiglitazone treatment (). Furthermore, we used a gradient concentration of ligands to demonstrate that chelerythrine is substantially less effective in recruiting SRC family coactivators and dissociating the NCoR-2 corepressor (). Our study indicates that chelerythrine is a PPARγ selective modulator with a cofactor profile that differs from rosiglitazone, which leads to its low transcriptional activity and partial agonism. Fig. 2a Fig. 2a Fig. 2b–e

To further understand the molecular mechanism underlying chelerythrine interaction with PPARγ, we solved the crystal structure of PPARγ complexed with chelerythrine at a resolution of 1.98 Å (and). The structure reveals that chelerythrine bound PPARγ LBD adopts a classical three-layer helical sandwich structure, which is globally the same as almost all NR LBDs. In the electron-density map presented in, chelerythrine clearly binds in the PPARγ LBD. Although structural alignment of PPARγ/chelerythrine and PPARγ/rosiglitazone reveals the whole structure of chelerythrine-bound PPARγ is conserved relative to that of rosiglitazone(), the binding of chelerythrine gives rise to several changes in the PPARγ ligand-binding pocket (LBP). As a selective modulator with weaker transactivity, chelerythrine exhibits a nearly flat scaffold that is structurally distinct from rosiglitazone (). Both ligands docking in a similar binding site in the PPARγ LBP, whereas the two small molecules orient in a perpendicular manner (). Fig. 3a Supplementary Table 1 ^{28 29} Fig. 3b ³⁰ Fig. 3c,d Fig. 4a Fig. 4c,d

The interactions that ligands anchor in the LBP vary between PPARγ/chelerythrine and PPARγ/rosiglitazone. Chelerythrine is docked in the LBP mainly through reversible hydrophobic interactions and water-mediated contacts. There are two indirect hydrogen bond interactions through HO molecules (). First, the carbonyl oxygen of the amide from Ile326 makes a hydrogen bond contact with methoxy in chelerythrine through HO molecule. On the other side, the nitrogen or carbonyl oxygen of an amide from Leu353, Phe360 and Met364 make additional hydrogen bond contacting to the oxygen of heterocyclic chelerythrine through HO molecule. 2 2 2 Fig. 4b

The cofactor profiling assay in our study indicated that chelerythrine induces conformation changes in the PPARγ AF-2 surface to create a new cofactor binding pattern. Notably, AF-2 also correlates with the degree of ligand transactivation. Therefore, we further dissect the conformation changes induced by chelerythrine in the AF-2 surface to explain high-affinity PPARγ ligands with differential agonism. As shown in, rosiglitazone forms a hydrogen bond with Tyr473 and His449 through its nitrogen and oxygen atom in TZD group, which further allows helix 12 to dock against helix 3 and helix 11. However, this strong correlation is absent in the PPARγ/chelerythrine complex (), which is the direct evidence of the partial agonism of chelerythrine. Within the PPARγ LBP, chelerythrine lies parallel to helix 3 (), which is in a perpendicular orientation from rosiglitazone. Together, helix 7 and helix 11 also shift outwardly in response to the perpendicular orientation of chelerythrine relative to rosiglitazone. As a result, the hydrophobic benzene ring side chain of Phe363 shifts externally from its rosiglitazone-bound conformation, driving the side chain of Leu452 outward even as the major conformation of helix 11 is conserved as in PPARγ/rosiglitazone (). Collectively, these results suggest that the selective modulator chelerythrine binds PPARγ to induce dynamic changes in the AF-2 surface, leading to differential cofactor profiling and partial agonism. Fig. 4g Fig. 4h Fig. 4c,d Fig. 4e,f

As PPARγ is a significant drug target for metabolic syndrome associated with obesity-induced insulin resistance and hyperglycemia, we used thediabetic mouse model and their littermates to determine whether chelerythrine has the ability to improve glucose homeostasis in these mice. Treatingmice for 14 days with 3 mg/kg rosiglitazone or 3 mg/kg chelerythrine, we determined that rosiglitazone increased body weight significantly by approximatedly 30% compared to vehicle control with the equal food intake (), whereas the body weight of mice treated with chelerythrine kept stable (), demonstrating that as a selective modulator, chelerythrine did not cause the weight gain side effects that rosiglitazone did. In consist with this weight control advantages, the liver and epididymal fat pads weight of chelerythrine treatedmice were both obviously lower than that of vehicle and rosiglitazone treated mice (), which is also supported by that the expression of lipogenesis related genes Srebp-1c, Fasn, Scd1, etc. in liver and epi-WAT are significantly downregulated after chelerythrine treatment compared to rosiglitazone (). Of note, we have measured the activity of the serum alanine aminotransferase (ALT) of chelerythrine-treated mice, and chelerythrine did not cause liver toxicity in diabetic mice (). Moreover, the genes expression in mice liver, such as the downregulation of the gluconeogenesis enzyme G6Pase, elucidated the molecular mechanism of the glucose metabolism regulated by chelerythrine (). Meanwhile, fasted metabolic parameters such as serum glucose, insulin and cholesterol were all downregulated by both chelerythrine and rosiglitazone treatment (and). As gold markers in clinical diagnostic, hemoglobin A1c (Hb1Ac) and glycosylated serum protein (GSP) were both downregulated after chelerythrine and rosiglitazone treatment (). Our study also indicated that the treatment of chelerythrine downregulated CD36 and IL-1β as well as inflammation related genes like IFNγ and TNFα, which are of great benefits for the treatment of diabetes (). ^{31 32} Fig. 5a Fig. 5b Supplementary Figure 3a & 3b Fig. 5i,j Supplementary Figure 3d Fig. 5i Fig. 5c,d Supplementary Figure 3c Fig. 5e,f Supplementary Figure 4 db/db db/db db/db

We further performed a glucose tolerance test (GTT) and insulin tolerance test (ITT) to determine the insulin sensitivity. As shown in, chelerythrine and rosiglitazone both improved the glucose tolerance and insulin sensitivity efficaciously in diabetic mice, which is even as senstive as their littermate controls. Furthermore, the effect of chelerythrine in improving insulin sensitivity were consistent in another diabetic mouse model, KKAy, which indicated the same potency of chelerythrine in mediating glucose homeostasis (). Collectively, these data indicate that chelerythrine is as effective as rosiglitazone in improving insulin sensitivity and homeostasis of glucose but without any side effects of weight gain. Fig. 5g,h Supplementary Figure 5

The selective PPARγ modulator chelerythrine exhibited effects in maintaining glucose metabolism, although its transcriptional activity is considerably low. It was reported that inhibition of CDK5-mediated phosphorylation of PPARγ is related to insulin sensitivity. We used theandCDK5 kinase assay to determine whether the insulin sensitizer chelerythrine retains the ability to inhibit CDK5-mediated phosphorylation of PPARγ S273. As shown in, chelerythrine blocked the phosphorylation of PPARγ serine 273 by CDK5 in a concentration-dependent manner. More importantly, chelerythrine is more effective in the inhibition of PPARγ phosphorylation than rosiglitazone despite its weak transcriptional activity. Consistent with the data, both rosiglitazone and chelerythrine caused a similar reduction in PPARγ phosphorylation at Ser273 in mice white adipose tissues (). ¹⁷ Fig. 6a,b Fig. 6c,d in vitro in vivo in vitro in vitro

CDK5-mediated phosphorylation of PPARγ is associated with obesity, and dysregulated the expression of a subset of PPARγ-regulated genes, including the adiponectin() and adipsin, which were significant in PPARγ-mediated insulin sensitivity. As the data shown in the analyses of the gene expression of adipose tissue in obese mice (), adiponectin and adipsin were significantly upregulated by chelerythrine treatment, which was helpful in improvement of fat oxidation and insulin resistance. Adipoq ^{33 34} Fig. 5j

Human PPARγ LBD (residues 206–477) was expressed as an N-terminal 6xHis fusion protein from the expression vector pET24a (Novagen, Germany). BL21(DE3) cells transformed with the expression plasmid were grown in LB broth at 25 °C to an OD600 of approximately 1.0 and induced with 0.1 mmol/l isopropyl 1-thio-β-D-galactopyranoside (IPTG) at 16 °C. Cells were harvested and sonicated in 200 ml of extract buffer (20 mmol/l Tris pH 8.0, 150 mmol/l NaCl, 10% glycerol, and 25 mmol/l imidazole) per 6 liters of cells. The lysate was centrifuged at 20,000 rpm for 30 min, and the supernatant was loaded on a 5-ml NiSO-loaded HiTrap HP column (GE Healthcare, PA, USA). The column was washed with extract buffer, and the protein was eluted with a gradient of 25–500 mmol/l imidazole. The PPARγ LBD was further purified with a SP-Sepharose column (GE Healthcare, PA, USA). To prepare the protein-ligand complex, 5-fold excess of the chelerythrine (Sigma, USA) and 2-fold excess of steroid receptor coactivator 1 (SRC1) peptide (AQQKSLLQQLLTE) were added to the purified protein, followed by filter concentration to 10 mg/ml. 4

The crystals of the PPARγ/chelerythrine complex were grown at room temperature in hanging drops containing 1.0 μl of the above protein-ligand solutions and 1.0 μl of well buffer containing 0.2 mol/l sodium thiocyanate and 20% w/v polyethylene glycol 3350. The crystals were directly flash frozen in liquid nitrogen for data collection. The observed reflections were reduced, merged and scaled with DENZO and SCALEPACK in the HKL2000 package. The structures were determined by molecular replacement in the CCP4 suite (). Manual model building was carried out with Coot, followed by REFMAC refinement in the CCP4 suite. http://www.ccp4.ac.uk↗ ^{36 37}

The binding of various coregulator peptide motifs to PPARγ LBD in response to ligands were determined by AlphaScreen™ assays using a hexahistidine detection kit from Perkin-Elmer as described. The experiments were conducted with approximately 20–40 nmol/l receptor LBD and 20 nmol/l biotinylated coregulator peptides in the presence of 5 μg/ml donor and acceptor beads in a buffer containing 50 mmol/l MOPS, 50 mmol/l NaF, 0.05 mmol/l CHAPS, and 0.1 mg/ml bovine serum albumin, all adjusted to a pH of 7.4. ²⁶

The peptides with an N-terminal biotinylation are listed below.

SRC1, SPSSHSSLTERHKILHRLLQEGSP;

SRC2, QEPVSPKKKENALLRYLLDKDDTKD;

SRC3, PDAASKHKQLSELLRGGSG;

PGC-1α, AEEPSLLKKLLLAPA;

NCoR-1, QVPRTHRLITLADHICQIITQDFAR;

NCoR-2, GHSFADPASNLGLEDIIRKALMGSF.

293T cells(ATCC, USA) were maintained in DMEM containing 10% fetal bovine serum (FBS) and were transiently transfected using Lipofectamine 2000 reagent (Invitrogen, USA). All mutant PPARγ plasmids were created using the Quick-Change site-directed mutagenesis kit (Stratagene, USA). Twenty-four-well plates were plated 24 hours prior to transfection (5 × 10cells per well). For the Gal4-driven reporter assays, the cells were transfected with 200 ng of Gal4-LBDs of various nuclear receptors and 200 ng of pG5Luc reporter (Promega, USA). Ligands were added five hours after transfection. Cells were harvested 24 h later for the luciferase assays. Luciferase activities were analyzed as the the instruction of CheckMate™ Mammalian Two-Hybrid System (Promega, USA). 4

TheCDK kinase assay was performed as previously described. Briefly, 1μg purified His-tagged PPARγ LBD (residues 206–477) was incubated with 50 ng active CDK5/p25 (Invitrogen, USA) in assay buffer (25 mmol/l Tris pH7.5, 10 mmol/l MgCl, 5 mmol/l β-glycerophosphate, 0.1 mmol/l NaVO, 2 mmol/l DTT) (Cell Signaling Technology, USA) containing 100 μmol/l ATP in a 50 μl reaction for 30 min at room temperature. PPARγ ligands were preincubated with PPARγ LBD protein for 30 min before the assay was performed. Phosphorylation of the PPARγ LBD was analyzed by western blotting with an anti-CDK substrate antibody (Cell Signaling Technology, USA). in vitro ¹⁸ 2 3 4

White adipose tissues from mice treated with compounds were homogenized in RIPA buffer (50 mmol/l Tris pH7.5, 150 mmol/l NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS with protease and phosphatase inhibitors). For western blotting, a rabbit polyclonal phospho-specific antibody against PPARγ Ser273 was produced by AbMax Biotechnology Co., Ltd, China, with a synthetic phosphopeptide previously described. Total tissue lysates were analyzed with anti-PPARγ antibody (Santa cruz). ¹⁸

Animal experiments were performed according to procedures approved by the Institutional Animal Use and Care Committee of Xiamen University. 8–10 week-old male mice (+,and KKAy mice from Hua Fukang, Beijing, China) were acclimatized for 7 days under standard conditions before experiments. Mice were fed a high-fat diet (60% kcal fat, D12492, Research Diets Inc, USA) and intraperitoneally (i.p.) injected once daily with vehicle (40% of 2-hydroxypropyl-β-cyclodextrin, HBC) (Sigma, USA), 3 mg/kg of chelerythrine or 3 mg/kg of rosiglitazone for 14 days. Mice were euthanized after 6 h of fasting, and serum samples were collected to measure the metabolic parameters. db/ db/db

Mice treated with drugs were fasted for 6 h with free access to water. For the glucose tolerance test (GTT), 1 g/kg of glucose was i.p. injected into the mice, and blood glucose was measured with the Accu-Check® Performa (Roche Applied Science, Germany) at 0, 30, 60, 90, and 120 min. For the insulin tolerance test (ITT), 1 units/kg of recombinant human insulin (Novolin 30R, Novo Nordisk, Denmark) was i.p. injected into the mice, and blood glucose was measured at 0, 30, 60, 90, and 120 min after insulin injection. Serum glucose was determined using the Glucose Oxidase Method (APPLYGEN, China), and serum insulin levels were determined by ELISA using an ultra-sensitive mouse insulin kit (Crystal Chem, USA). The cholesterol, Hb1Ac, GSP and ALT levels were assayed using the kits from Nanjing Jiancheng Bioengineering Institute (Nanjing, China).

Total RNA was isolated from liver and epididymal fat pads using the Tissue RNA kit (Omega Bio-Tek, GA, USA). The RNA was reverse-transcribed using the TAKARA reverse transcription kit. Quantitative PCR reactions were performed with SYBR Green fluorescent dye using a CFXTM96 real-time system (BIO-RAD, USA). Relative mRNA expression was determined by the ΔΔ-Ct method normalized to actin levels. The sequences of the primers are listed in. Supplementary Table 2