Introduction

The metabolic syndrome is a cluster of metabolic abnormalities, including type 2 diabetes, obesity, dyslipidaemia, insulin resistance, liver steatosis and hypertension [1]. Peroxisome proliferator-activated receptors (PPARs) belong to the nuclear receptor superfamily that function as ligand-inducible transcription factors and control the expression of multiple target genes [2, 3]. Pharmacological activation of each of three PPAR isoforms (α, β/δ and γ) is effective at correcting a specific aspect of the metabolic syndrome, such as insulin resistance, dyslipidaemia and hyperglycaemia [2]. Fibrates such as bezafibrate and fenofibrate, which are synthetic PPAR-α agonists, have been used to treat dyslipidaemia by lowering triacylglycerol (TG) levels and increasing HDL-cholesterol (HDL-C) levels [4]. Thiazolidinediones (TZDs), such as rosiglitazone (RSG) and pioglitazone (PGZ), which are classic selective PPAR-γ agonists, are used as insulin sensitisers to treat type 2 diabetes [5–8]. Recent studies have highlighted the metabolic functions of PPAR-β/δ activators in the treatment of obesity [9–14]. Thus, the activation of three PPAR subtypes simultaneously (pan-PPAR agonism) is a promising avenue for reversal of multiple metabolic syndrome disorders.

Evidence has suggested that pan-PPAR agonists could complement each other and circumvent the limitations of selective or dual PPAR agonism. However, there are still some limitations to developing pan-PPAR agonists. Bezafibrate, a PPAR-α agonist, has been found to activate all three PPAR isoforms [15, 16], but its glucose-lowering effect is not significant in clinical studies [17]. Most pan-PPAR agonists, including GW677954, DRL-11605, PLX-204, GW625019 and netoglitazone, have been discontinued because of safety problems [18–20]. Additionally, preclinical studies on pan-PPAR agonists LY465608 and BPR1H036 are not progressing smoothly [21–24].

Natural products have served as the source and inspiration for a large number of current drugs. Malaytea scurfpea fruit, the seed of Psoralea corylifolia Linn., has been used in traditional Chinese formulas to prevent and treat type 2 diabetes in clinical practice [25, 26]. In this study, bavachinin (BVC), a component of malaytea scurfpea fruit, is identified as a novel natural pan-PPAR agonist in vitro and in vivo. Interestingly, BVC does not antagonise, but synergises with TZDs and fibrates in metabolic syndrome phenotypes. This synergism is induced by unique binding modes of BVC with PPAR-γ or -α.

Methods

All experimental details are shown in the electronic supplementary material (ESM) Methods.

Isolation and purification of BVC

See ESM Methods for details.

Reporter gene assay

293T cells were transfected over a 6 h period with target expression plasmid, relevant reporter plasmid, pREP7 reporter and FuGENE-HD transfection reagent, and were subsequently treated with compounds for 24 h. Luciferase activity was detected using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA).

TR-FRET assay

LanthaScreen time-resolved fluorescence resonance energy transfer (TR-FRET) competitive binding or coactivator assays (Invitrogen, Carlsbad, CA, USA) use the ligand-binding domain (LBD) of human nuclear receptors tagged with glutathione S-transferase (GST), the terbium-labelled anti-GST antibody and the fluorescent small molecule or coactivator peptide.

3T3-L1 differentiation

3T3-L1 cells were purchased from Cell Bank of Shanghai Institutes of Biological Sciences, Chinese Academy of Sciences, China. The cells were induced and maintained in the presence of test compounds (day 0–day 6), and then stained with Oil Red O as previously reported [27]. To investigate the effects on the expression of PPAR target genes, 3T3-L1 cells exposed to compounds were harvested on days 0, 2 or 6.

Animal experiments

In accordance with the Guide for the Care and Use of Laboratory Animals, all animal studies were guided by the Animal Ethics Committee of Shanghai University of Traditional Chinese Medicine.

Thirty-three 9-week-old female db/db mice were randomly administered with vehicle, or 4 mg kg−1 day−1 RSG or 100 mg kg−1 day−1 BVC by gavage for 3 weeks (n = 11). Twenty-one 19-week-old female diet-induced obese (DIO) mice received RSG (100 mg/kg) or BVC (2 g/kg) milled in a high-fat diet (HFD) or regular HFD (Research Diets, New Brunswick, NJ, USA) for 3 weeks (n = 7). The age-matched, lean wild-type C57BL/6J mice were treated as normal control groups (n = 10).

To study the synergistic effects on glucose metabolism, 53 male db/db mice were randomly administered with vehicle, 2 mg kg−1 day−1 or 4 mg kg−1 day−1 RSG, 50 mg kg−1 day−1 or 100 mg kg−1 day−1 BVC, or 2 mg kg−1 day−1 RSG plus 50 mg kg−1 day−1 BVC by gavage for 4 weeks (n = 8–9). To study the synergistic effects on lipid regulation, groups of db/db mice were treated with one of the following by gavage for 4 weeks: (1) vehicle; (2) fenofibrate, 25 mg kg−1 day−1 or 100 mg kg−1 day−1; (3) BVC, 50 mg kg−1 day−1 or 100 mg kg−1 day−1; or (4) fenofibrate, 25 mg kg−1 day−1, plus BVC, 50 mg kg−1 day−1.

At the end of treatment, all overnight-fasted mice were anaesthetised. Blood samples were collected by cardiac puncture, and multiple organs and tissues were harvested and either fixed in formalin or snap frozen and stored at −80°C.

Metabolic studies

All mice were subjected to an overnight (8–12 h) fast followed by intraperitoneal glucose tolerance test (ipGTT) and non-fasting conditions followed by intraperitoneal insulin tolerance test (ipITT). Hepatic lipids were extracted as described previously [27], and TGs and total cholesterol (TCH) were detected using colorimetric kits. Values were calculated as nmol/l (TG and TCH) per mg wet tissue. Serum levels of insulin, adiponectin, leptin, NEFA, TG, TCH, HDL-C, LDL-cholesterol (LDL-C), alanine transaminase (ALT) and aspartate transaminase (AST) were analysed from collected fasted blood using commercial kits.

Histochemistry

Livers were fixed, processed and stained with haematoxylin and eosin (H&E) and Oil Red O as described [27].

Mouse cDNA genome-wide gene expression analysis

The details are shown in ESM Methods.

Quantitative real-time PCR

Total RNA was prepared using TRIzol reagent (TaKaRa, Dalian, China). The first strand cDNA was synthesised using a reverse transcription kit (Thermo, Waltham, MA, USA), and was amplified with SYBR Green QPCR Master Mix (Applied Biosystems, Waltham, MA, USA) using an ABI StepOnePlus sequence detector (Applied Biosystems).

Nuclear magnetic resonance

All 19F nuclear magnetic resonance (NMR) experiments were performed at 298 K on a Bruker 500 MHz AVANCE spectrometer equipped with 5 mm QCI-F cryoprobe without 1H decoupling, using standard pulse sequence ‘zg’ provided with Topspin 3.0.

Molecular docking

The crystal structure of PPAR-γ (PDB code 3K8S) was retrieved from the Research Collaboratory for Structural Bioinformatics (RCSB) protein data bank. Docking was performed using Glide 5.6 (Schrödinger, New York, NY, USA) with its SP mode.

Statistical analysis

All data are expressed as mean ± SEM. Statistical analysis was performed using SPSS 16.0 (SPSS, Chicago, IL, USA). Between-group differences were determined by one-way ANOVA with the least significant difference (LSD) test. All statistical tests are two-tailed. A value of p ≤ 0.05 was considered to be statistically significant. The IC50 and half maximal effective concentration (EC50) values were calculated using GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA). Experimenters were not blind to group assignment and outcome assessment, and no data were excluded from the experimental results.

Results

BVC is a natural pan-PPAR agonist with potent binding affinities

A reporter gene assay was used to determine the activities of BVC on PPAR-α, -β/δ or -γ. As shown in Fig. 1, BVC (Fig. 1a) dose-dependently induced the transcriptional activities of the mouse LBD (Fig. 1b–d) and full lengths (Fig. 1e−g) of the three PPAR isoforms. Notably, BVC showed stronger activities with PPAR-γ than with PPAR-α and PPAR-β/δ (EC50 = 0.74 μmol/l, 4.00 μmol/l and 8.07 μmol/l, respectively; Table 1).

Fig. 1
figure 1

BVC functions as a potent pan-PPAR agonist. (a) Structure of BVC. (b–d) BVC induced the transcriptional activities of LBDs of mouse PPAR-γ (b), -α (c) and -β/δ (d). ***p < 0.001 vs 0 μmol/l BVC. (e–g) Compounds induced the transcriptional activity of full-length mouse PPAR-γ (e), -α (f) and -β/δ (g). Blue, BVC; Red, RSG (e), WY14643 (f) and GW0742 (g). (h) BVC-induced 3T3-L1 pre-adipocyte differentiation. Cells were exposed to BVC or RSG at the beginning of pre-adipocyte induction (day 0). The adipocytes were stained with Oil Red O on day 6. Ins+Dex, 10 μg/ml insulin + 1 μmol/l dexamethasone. (i–k) Binding to LBDs of human PPAR-γ (i), -α (j) and -β/δ (k) in a TR-FRET competitive binding assay. Blue, BVC; red, RSG (i), WY14643 (j) and GW0742 (k); green, fenofibric acid (j)

Full size image
Table 1 Binding affinity constants, inhibitory concentrations and effective concentrations of tested compounds for three PPAR subtypes
Full size table

To examine the cellular effect of BVC on PPAR-γ, insulin, dexamethasone and RSG or BVC were used to induce differentiation of 3T3-L1 pre-adipocytes. BVC promoted 3T3-L1 adipocyte differentiation in a dose-dependent manner (Fig. 1h). 3T3-L1 differentiation was effectively activated by 1 μmol/l BVC, and almost fully by 10 μmol/l BVC. The results suggest that BVC activates PPAR-γ at a cellular level.

Competitive binding assays (Fig. 1i–k) showed that BVC had stronger binding to human PPAR-γ (K i = 223 nmol/l; Table 1) than to human PPAR-α and -β/δ (K i = 7.88 and 5.28 μmol/l; Table 1). Moreover, BVC has a fivefold weaker binding affinity than RSG for PPAR-γ, and has twofold stronger binding affinity than fenofibric acid and WY14643 for PPAR-α.

In contrast to the full PPAR-γ agonist RSG and PPAR-α agonist GW7647, BVC did not induce the recruitment of classic transcriptional cofactors, peroxisome proliferative activated receptor, gamma, coactivator 1 α (PGC1α) and mediator complex subunit 1 (TRAP220/DRIP) (ESM Fig. 1a–c). Using reporter gene and coactivator recruitment assays, BVC did not activate other nuclear receptors related to the metabolic syndrome, such as liver x receptor (LXR)-α, LXR-β/δ and farnesoid x receptor (FXR) (ESM Fig. 1d–g), suggesting the specificity of BVC for PPARs.

BVC treatment ameliorates diabetes and hyperlipidaemia in db/db and DIO mice

Next, we evaluated the glucose-lowering effects of BVC in leptin-receptor-deficient db/db mice and DIO mice. In groups of db/db mice with similar food intake (Fig. 2a), RSG greatly increased body weight by approximately 8 g within 3 weeks (p < 0.001), whereas BVC treatment did not significantly induce weight gain (Fig. 2b). BVC treatment significantly decreased fasting and non-fasting glucose levels (Fig. 2c, d) and enhanced glucose tolerance (17% decrease in glucose AUC vs vehicle) and insulin sensitivity (34% decrease in glucose AUC vs vehicle) during intraperitoneal glucose tolerance test (ipGTT and ipITT in db/db mice (Fig. 2e, f). Similar results were detected in DIO mice (ESM Fig. 2). Insulin resistance, one hallmark of type 2 diabetes, is usually accompanied by hyperinsulinaemia. In both models, insulin levels were notably reduced by BVC comparable with the effect of RSG (Fig. 2g). Additionally, BVC significantly upregulated levels of adiponectin (Fig. 2h), which also promotes insulin sensitivity and decreases glucose production in the liver [28]. Meanwhile, BVC, but not RSG, decreased serum leptin levels to further alleviate insulin resistance in DIO mice (ESM Fig. 2f). The above results demonstrated that BVC ameliorates insulin resistance and glucose metabolism in db/db and DIO mice.

Fig. 2
figure 2

BVC exhibits potent glucose-lowering effects in db/db mice. Mice were treated with vehicle or 100 mg kg−1 day−1 BVC or 4 mg kg−1 day−1 RSG for 3 weeks (n = 11). (a) Food intake. (b) Weight gain. (c) Fasting blood glucose. (d) Non-fasting blood glucose. (e) Glucose levels during ipGTT (inset, AUC, mmol/l × min) were measured after an 8 h fast. (f) Glucose concentrations during ipITT (inset, AUC, mmol/l × min) were determined in non-fasting conditions. (g) Fasting serum insulin, (h) adiponectin, (i) TG and (j) LDL-C levels were investigated in DIO and db/db mice after an 8 h fast. Orange, vehicle; red, RSG; blue, BVC. *p < 0.05, **p < 0.01 and ***p < 0.001 vs db/db vehicle mice. † p < 0.05, †† p < 0.01 and ††† p < 0.001 vs DIO HFD mice. CD, chow diet; d, day

Full size image

High serum TG and LDL-C and low HDL-C levels are characterised as major lipid risk factors for patients with the metabolic syndrome [29]. In both mouse models, BVC lowered TG levels (p < 0.05), while RSG did not show consistent TG-lowering effects (Fig. 2i). Additionally, BVC reduced LDL-C levels in both models, while RSG increased LDL-C levels in db/db mice and decreased levels in DIO mice (Fig. 2j). Moreover, BVC strongly lowered serum NEFA in db/db mice and increased the levels of HDL-C/LDL-C in both models (ESM Fig. 3). The results indicate that BVC is more effective in dyslipidaemia than RSG in DIO.

BVC improves hepatotoxicity in db/db and DIO mice

Non-alcoholic fatty liver, an additional feature of the metabolic syndrome [30], was observed in db/db and DIO mice (Fig. 3 and ESM Fig. 4). Oil Red O staining showed that BVC decreased lipid accumulation in liver compared with vehicle or RSG treatment (Fig. 3a), which was verified with hepatic lipid content quantification (Fig. 3b). Compared with RSG, BVC lowered liver TG content by approximately 50% and liver TCH contents by 40% in db/db mice. BVC suppressed serum levels of ALT and AST (Fig. 3c), which are commonly used to evaluate clinically hepatocellular injury, and it further ameliorated vacuolar degeneration in db/db livers (Fig. 3d). More significantly, similar results were observed in DIO mice (ESM Fig. 4). These findings indicated that BVC improves hepatotoxicity in db/db and DIO mice.

Fig. 3
figure 3

BVC alleviated hepatotoxicity in db/db mice after 3 weeks of treatment. (a) Oil Red O staining of liver sections (n = 9), (b) liver TG and TCH contents (n = 7–9), (c) fasting serum AST and ALT levels (n = 9–10) and (d) H&E staining of liver sections (n = 9). **p < 0.01 vs db/db mice exposed to vehicle. ‡ p < 0.05, ‡‡ p < 0.01 and ‡‡‡ p < 0.001 vs RSG control. Scale bars, 50 μm

Full size image

BVC regulates PPAR gene expression in vitro and in vivo

To reveal the mechanism underlying BVC-mediated improvement in the metabolic syndrome, liver gene expression profiles of db/db and DIO mice were analysed by cDNA genome-wide gene expression assay. Figure 4a shows the 138 overlapping differential genes of both mouse livers that were identified (fold change ≥2, p < 0.05). Gene ontology pathway and hierarchical cluster analysis further revealed the most enriched pathways regulated by BVC were related to adipocytokine signalling, PPAR signalling, type 2 diabetes, glycerophospholipid metabolism, mitochondrial fatty acid beta oxidation and butanoate metabolism (Fig. 4b). Furthermore, the above results were confirmed by quantitative PCR. BVC increased the expression of the PPAR-γ marker gene glucose transporter 4 (Glut4, also known as Slc2a4) to enhance transport and utilisation of glucose in white adipose tissue (Fig. 4c) and skeletal muscle (Fig. 4d). BVC also upregulated other classic PPAR target genes involved in the synthesis and transport of fatty acid and insulin sensitivity in white adipose tissue [31], such as Fabp4, Acox1, Fsp27 (also known as as well as Cidec), Adipoq and Adipor1 (Fig. 4c). Moreover, BVC significantly induced the expression of the PPAR-β/δ genes Pdk4 and Angptl4, but not Cpt1b, in skeletal muscle (Fig. 4d). Additionally, BVC enhanced transport and β/ω-oxidation of fatty acid by inducing the expression of Fabp4, Cd36, Lpl, Slc27a4, Fgf21, Acacb, Acadm and Acox1 but not Cpt1a, Acadl and Hmgcs2 (Fig. 4e and ESM Fig. 5c). Finally, the expression of inflammatory cytokines Tnf-α, Il-6 and Il-1β (also known as Tnf, Il6 and Il1b) in white adipose tissue concurrent with insulin resistance was decreased by BVC (Fig. 4f). In contrast to RSG, BVC increased Fgf21 mRNA expression to ameliorate insulin resistance and weight gain, but there was no difference in the expression of Acacb, Acadm, Cpt1a, Acadl and Hmgcs2 in liver or Cpt1b in skeletal muscle (ESM Fig. 5a, b). In addition, BVC enhanced Glut4 and Cyp4a10 expression at day 2 and day 6, as well as the expression of Cyp7a1 and Scd1 at day 6 of exposure period of 3T3-L1 adipocyte differentiation, which is similar to RSG (ESM Fig. 5d–g). These data suggest that BVC enhances glucose transport and utilisation, hepatic lipid turnover and fatty acid metabolism through PPAR networks, thereby improving insulin sensitivity, dyslipidaemia and fatty liver [32].

Fig. 4
figure 4

BVC selectively regulates the expression of PPAR target genes. (a) Overlapped differential genes in livers of db/db and DIO mice with greater than twofold change (n = 3). (b) The 11 most enriched pathways (p < 0.05). (c–e) BVC regulated PPAR target genes in white adipose tissue (c), skeletal muscle (d) and liver (e) in db/db mice (n = 6). (f) BVC decreased the expression of inflammatory cytokines in white adipose tissue (n = 6). Data are expressed as mean ± SEM. *p < 0.05 and **p < 0.01 vs mice exposed to vehicle. Ppar-γ is also known as Pparg; Ppar-α is also known as Ppara

Full size image

BVC synergises with RSG and WY14643 in glucose-lowering effect

Interestingly, in the reporter gene assay, BVC did not antagonise, but rather synergised with, synthetic PPAR-α and -γ agonists to induce PPAR-mediated transcription. In Fig. 5a, b, the combination of BVC plus RSG caused dose-dependent augmentation of the transactivation of LBDs and full-length PPAR-γ compared with 20 μmol/l RSG. Similar synergism was also observed when combining BVC with PGZ (ESM Fig. 6a, b). Consequently, the synergistic effects of BVC plus RSG on glucose metabolism were evaluated in db/db mice. With similar food intake (ESM Fig. 6c), the combination of 50 mg kg−1 day−1 BVC plus 2 mg kg−1 day−1 RSG significantly reduced body weight gain compared with 2 mg kg−1 day−1 RSG treatment (p < 0.05, ESM Fig. 6d). Compared with vehicle treatment, both BVC and RSG treatment significantly decreased fasting glucose concentrations in a dose-dependent manner (Fig. 5c). The combination of BVC plus RSG exhibited almost the same effects as 4 mg kg−1 day−1 RSG in lowering glucose levels, which was more effective than either treatment alone (Fig. 5c). Furthermore, BVC synergised with WY14643 (Fig. 5d, e) and fenofibric acid (ESM Fig. 6e, f) to induce the transcriptional activity of PPAR-α. However, no synergism between BVC with GW0742 in PPAR-β/δ-mediated transcription was detected (ESM Fig. 6g, h). Figure 5f showed that both BVC and fenofibrate dose-dependently decreased fasting TG levels in db/db mice. Treatment with 50 mg kg−1 day−1 BVC plus 25 mg kg−1 day−1 fenofibrate treatment exhibited similar effects to 100 mg kg−1 day−1 fenofibrate treatment in lowering TG levels, which was more effective than either treatment alone (p < 0.01). The above results indicate BVC synergises with synthetic PPAR-γ and -α agonists in decreasing glucose and lipid levels.

Fig. 5
figure 5

BVC synergises with PPAR-γ and -α agonists in vitro and in vivo. (a, b) The combination of BVC plus RSG induced the transcriptional activities of mouse PPAR-γ (n = 3). (c) The effects on fasting blood glucose after 4 weeks of treatment (n = 7–11). (d, e) The effects of BVC plus fibrates on the transcriptional activities of mouse PPAR-α (n = 3) and (f) fasting TG levels (n = 7–11) of db/db mice. *p < 0.05 and ***p < 0.001 vs negative or vehicle control. ‡ p <0.05, ‡‡ p <0.01 and ‡‡‡ p <0.001 vs positive control. §§ p < 0.01 and §§§ p < 0.001 vs BVC 50 mg kg−1 day−1. d, day

Full size image

Binding behaviour of BVC with PPAR-γ

To study the binding events of BVC at the cellular level, the reporter gene assay was performed using a PPAR covalent antagonist GW9662, which can bind to the canonical ligand-binding pocket (CLBP) of synthetic PPAR ligands, blocking the binding to PPAR-LBD. Figure 6a, b as well as ESM Fig. 7a, b show that 20 μmol/l GW9662 reduced the transcriptional activities of PPAR induced by RSG and WY14643, but did not block the stimulation of BVC. The results demonstrate that BVC binds to at least one other binding site in PPAR in addition to the CLBP.

Fig. 6
figure 6

Binding behaviour of BVC to PPAR-γ. (a, b) GW9662 blocked the transcriptional activities of PPAR-γ induced by RSG (a), but was not affected by BVC (b). Data are expressed as mean ± SEM. (c) The structure of F-BVC. (d) Titration of F-BVC resulted in two PPAR-γ LBD-bound 19F-NMR resonances at −115.7 ppm and −136.1 ppm. (e) The 19F-NMR resonance at −115.7 ppm was saturated at the ratio of 1:1 (ligand:PPAR-γ). (f) The disappearance of the 19F chemical shifts at −115.7 ppm indicated that F-BVC shared the same canonical binding site as RSG on PPAR-γ. The retaining signal at −136.3 ppm suggested that F-BVC could occupy the other binding site after RSG bound to the CLBP of PPAR-γ. (g) Docking one BVC (yellow sticks) into the canonical site and the other BVC (green sticks) into the alternative site, respectively. (h) Docking of BVC (green sticks) into the alternative site after RSG (yellow sticks) occupies the CLBP. PPAR-γ is shown with the disordered Ω loop between H3 and H2′in red dashed lines

Full size image

To further clarify the binding sites of BVC to PPAR-γ-LBD through 19F-NMR, 3′-F-BVC (F-BVC; Fig. 6c) was prepared and validated as a potent PPAR activator (ESM Fig. 7c). Stoichiometric addition of F-BVC to PPAR-γ revealed sequential saturation of two binding sites as shown by 19F chemical shifts at −115.7 ppm and −136.1 ppm (Fig. 6d). The 19F-NMR resonance at −115.7 ppm indicated F-BVC had a strong binding site in PPAR-γ, and it falls into a slow-exchange regime on the NMR time scale, resulting in the narrow line width of 19F signal in bound form. Integration of 19F-NMR signal at −115.7 ppm was plotted against titration molar ratio. As shown in Fig. 6e, the strong binding site was saturated at a ratio of 1:1 (ligand/PPAR-γ). In addition, the broader line width of the 19F-NMR resonance at −136.1 ppm of F-BVC indicated the other weaker binding site in PPAR-γ, which may involve an intermediate chemical exchange between bound and free species of F-BVC. To further understand the binding behaviour of BVC, F-BVC was added into the sample of RSG-saturated PPAR-γ at the molar ratio of 1:1.5 (Fig. 6f). Compared with the 19F resonance of F-BVC/PPAR-γ, the disappearance of the signal at −115.7 ppm indicated that F-BVC shared the same binding site (CLBP) in PPAR-γ as RSG, and RSG completely blocked BVC from binding to CLBP. Conversely, the retaining signal at −136.3 ppm confirmed that F-BVC occupied another site on PPAR-γ, but this binding could not be displaced by RSG.

Next, docking studies were performed to investigate the exact binding sites of BVC using the Glide program. Docking results further proved that BVC could bind to two binding sites in PPAR-γ. One is CLBP with the predicted docking score of −34.72 kJ/mol, and the other is a second binding site with a docking score of −28.03 kJ/mol, indicating BVC occupied the CLBP with higher binding affinity. The second binding site for BVC was consistent with the alternative binding site (ABS), which was composed of H2′, Ω loop, H3 and β sheets [33]. In addition, if one BVC was docked to CLBP first (Fig. 6g, yellow sticks), then a second BVC could be docked to the ABS with −25.94 kJ/mol (Fig. 6g, green sticks). If one RSG was docked to the CLBP first (Fig. 6h, yellow sticks), then BVC could bind to the ABS with −28.87 kJ/mol (Fig. 6h, green sticks). In addition, docking results showed that the 4′-hydroxy group formed a hydrogen bond with the guanidinium group of R288, and the 6-prenylated isoflavanone skeleton was surrounded by a hydrophobic subpocket composed of I281, F282, L353, L356, F360 and F363 (ESM Fig. 7d). In addition, the 7-methoxy group of BVC was surrounded by Y327, K367 and H449, where van der Waals forces play a dominant role in the interactions (ESM Fig. 7d). In conclusion, BVC could bind to CLBP and the alternative site in PPAR-γ, and RSG can block BVC binding to CLBP but not to the ABS.

Discussion

Matin et al have identified some synthetic flavanones, flavones and isoflavones as dual PPAR-α and -γ agonists [34], and further discovered some synthetic isoflavones that can activate three PPAR isoforms in vitro [35]. Here, for the first time, we identified BVC, a natural prenylated isoflavanone, as a potent pan-PPAR agonist. The prenylated group of flavones can produce an increased affinity for biological membranes and an improved interaction with proteins [36].

The in vitro functional analyses demonstrate that BVC binds to and activates three PPAR isoforms. Mouse studies conclude that BVC displays properties desired for a glucose-lowering pan-PPAR agonist. Previous studies have indicated that full agonism of PPAR-γ, as with RSG and PGZ, may be responsible for the side effects associated with glucose-lowering effects [37–39]. Therefore, BVC does not cause weight gain and hepatotoxicity because of its submaximal activation of PPAR-γ and additional activation of PPAR-α and -β/δ. The enhancement of fatty acid oxidation, with the decrease in uptake and synthesis of fatty acids and the associated increased lipid mobilisation in adipose tissue, hints at how PPAR-α activation leads to body weight gain and increased adiposity [40]. Activation of PPAR-β/δ has shown to increase the fatty-acid-burning capacity of skeletal muscle, which is accompanied by a redistribution of fatty acid flux adipose tissue toward skeletal muscle [41]. Thus, the sum of PPAR-α and PPAR-β/δ activation offsets fat accumulation and weight gain induced by PPAR-γ agonism. Also, the combination of PPAR-β/δ and PPAR-γ agonists has been shown to alleviate insulin resistance, regulate glucose metabolism and enhance the ability to exercise [11, 16, 42]. Therefore, BVC shows glucose-lowering effects without inducing weight gain and hepatotoxicity.

Interestingly, BVC synergises with synthetic PPAR-γ and -α agonists to improve carbohydrate and lipid metabolism, but not with PPAR-β/δ agonists. Reporter gene assays demonstrate that BVC synergises with synthetic agonists to induce transcriptional activities of PPAR-α and -γ. As a result, the synergistic effects are extensively confirmed in dose–response mouse studies. PPAR senses ligands through ligand-binding sites in the LBD [43, 44]. The PPAR-γ-LBD has shown two binding sites, CLBP and ABS; synthetic PPAR ligands are designed to bind to CLBP by mimicking endogenous ligands [33]. GW9662, a PPAR covalent antagonist, can compete for the CLBP of synthetic ligands to PPAR and block binding to CLBP. Reporter assays demonstrate that GW9662 blocks transcriptional activities of PPAR induced by RSG and WY14643, but does not antagonise BVC, which indicates that BVC has one additional binding site to the CLBP. Competitive binding and NMR studies demonstrate that BVC shares the CLBP with PPAR-α and -γ synthetic agonists. In addition, 19F-NMR and docking results further reveal that BVC occupies the ABS in addition to the CLBP, although it binds to the CLBP with higher affinity than it binds to the ABS. After the CLBP of PPAR-γ is saturated with RSG, BVC continues to occupy the ABS. This unique binding mode of BVC produces special synergistic effects as a result of induced differential conformational change. Although the LBDs of the three PPAR subtypes share 80% homology, the cavity of the PPAR-β/δ-LBD is markedly narrower than the LBDs of PPAR-α and -γ [45]. Thus, it is speculated that the PPAR-β/δ-LBD does not have enough room to accommodate two ligands, which could explain why BVC does not synergise with GW0742.

BVC binds to PPAR in a unique mode, resulting in a special conformational change, differential cofactor recruitment, gene expression and, finally, distinct biological responses. Agonist binding to PPAR-γ regulates activity by causing conformational changes to the transcription activation function-2 (AF-2) domain of the LBD [46]. Changes to AF-2 allow for coactivator recruitment, followed by transcriptional activation. These changes vary depending on the type of ligand that binds to the LBD. TZDs and fibrates induce direct stabilisation of AF-2 conformation through distinct interactions with helix 12 of the AF-2 domain, followed by the recruitment of coactivators, including PGC-1α and TRAP220 [33]. BVC does not recruit PGC-1α and TRAP220 during the transcriptional process—its unique binding mode results in indirect stabilisation of AF-2 mediated via an Ω loop and helix 3. Thus, BVC exhibits differential effects on the expression of PPAR target genes involved in fatty acid metabolism, cholesterol metabolism and glucose transport. As a result, BVC decreases TG and LDL-C levels in both mouse models, but RSG exhibits the discrepancies in lipid regulation, which are also probably linked to species-specific regulation of VLDL and LDL particle size and concentration, and apolipoprotein [47–49]. Compared with the only clinical pan-PPAR agonist bezafibrate, BVC exhibited a stronger glucose-lowering effect [17]. While another pan-PPAR agonist, LY465608, has been shown in preclinical studies to improve glucose and lipid disorders, it significantly increases body weight and causes severe hepatotoxicity [21, 23]. Although BVC exhibits broad-spectrum potency without inducing weight gain and hepatotoxicity, additional studies are required to assess the side-effect profiles of BVC compared with current PPAR agonists.

In conclusion, the synergistic effects occur because of the unique binding events of BVC to PPAR-α and -γ. Therefore, the combination of natural BVC plus TZDs or fibrates could help increase effectiveness with fewer doses, followed by controlled side effects. This discovery may provide a new promising therapeutic approach for the treatment of type 2 diabetes and other metabolic syndrome indications.