Biochimica et Biophysica Acta 1830 (2013) 4813–4819

Contents lists available at SciVerse ScienceDirect

Biochimica et Biophysica Acta

j ourna l homepage: www.e lsev ie r .com/ locate /bbagen

Honokiol: A non-adipogenic PPARγ agonist from nature☆

Atanas G. Atanasov a, Jian N. Wang b, Shi P. Gu b, Jing Bu b, Matthias P. Kramer a, Lisa Baumgartner c,Nanang Fakhrudin a,1, Angela Ladurner a, Clemens Malainer a, Anna Vuorinen d, Stefan M. Noha d,Stefan Schwaiger c, Judith M. Rollinger c, Daniela Schuster d, Hermann Stuppner c,Verena M. Dirsch a, Elke H. Heiss a,⁎a Department of Pharmacognosy, University of Vienna, Althanstrasse 14, 1090 Vienna, Austriab Xi Yuan Hospital, China Academy of Chinese Medical Sciences, Beijing 100093, Chinac Institute of Pharmacy/Pharmacognosy, Center of Molecular Biosciences, University of Innsbruck, Innrain 80-82, A-6020 Innsbruck, Austriad Institute of Pharmacy/Pharmaceutical Chemistry, Center of Molecular Biosciences, University of Innsbruck, Innrain 80-82, A-6020 Innsbruck, Austria

Abbreviations: AMPK, AMP-activated kinase; ANOVAmerican type culture collection; BADGE, bisphenol Amorphogenic protein 4; BSA, bovine serum albumin; Ccellulose; DMEM, Dulbecco's modified Eagle's mediumEC50, effective concentration 50%; EGFP, enhanced grfetal calf serum; GST, glutathione-S-transferase; HPchromatography; IBMX, 3-isobutyl-1-methylxanthine;rapamycin; NBS, newborn bovine serum; NF-κB, nuclear faic resonance; PBS, phosphate buffered saline; PPARγ, pereceptor gamma; LBD, ligand-binding domain; MEF, mouNADPH-dependent oxidase; RXR, retinoic X receptor; SESPF, specific pathogen free; SREBP, sterol regulatory elemetional Chinesemedicine; TLC, thin layer chromatography;cence resonance energy transfer☆ This is an open-access article distributed under the tAttribution License, which permits unrestricted use, disany medium, provided the original author and source a⁎ Corresponding author. Tel.: +43 1 4277 55993; fax

E-mail address: elke.heiss@univie.ac.at (E.H. Heiss).1 Present address: Department of Pharmaceutical B

Gadjah Mada University, Sekip Utara, 55281 Yogyakarta

0304-4165/$ – see front matter © 2013 The Authors. Puhttp://dx.doi.org/10.1016/j.bbagen.2013.06.021

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 5 March 2013Received in revised form 3 June 2013Accepted 17 June 2013Available online 27 June 2013

Keywords:Peroxisome proliferator-activated receptorNatural productMetabolic disease

Background: Peroxisome proliferator-activated receptor gamma (PPARγ) agonists are clinically used to coun-teract hyperglycemia. However, so far experienced unwanted side effects, such as weight gain, promote thesearch for new PPARγ activators.Methods: We used a combination of in silico, in vitro, cell-based and in vivo models to identify and validatenatural products as promising leads for partial novel PPARγ agonists.Results: The natural product honokiol from the traditional Chinese herbal drug Magnolia bark was in silicopredicted to bind into the PPARγ ligand binding pocket as dimer. Honokiol indeed directly bound to purifiedPPARγ ligand-binding domain (LBD) and acted as partial agonist in a PPARγ-mediated luciferase reporterassay. Honokiol was then directly compared to the clinically used full agonist pioglitazone with regard to

stimulation of glucose uptake in adipocytes as well as adipogenic differentiation in 3T3-L1 pre-adipocytesand mouse embryonic fibroblasts. While honokiol stimulated basal glucose uptake to a similar extent aspioglitazone, it did not induce adipogenesis in contrast to pioglitazone. In diabetic KKAy mice oral applicationof honokiol prevented hyperglycemia and suppressed weight gain.Conclusion: We identified honokiol as a partial non-adipogenic PPARγ agonist in vitro which preventedhyperglycemia and weight gain in vivo.General significance: This observed activity profile suggests honokiol as promising new pharmaceutical leador dietary supplement to combat metabolic disease, and provides a molecular explanation for the use ofMagnolia in traditional medicine.

© 2013 The Authors. Published by Elsevier B.V. All rights reserved.

A, analysis of variance; ATCC,diglycidyl ether; BMP, bone

MCNa, sodium carboxymethyl; DMSO, dimethyl sulfoxide;een fluorescent protein; FCS,LC, high-performance liquidmTOR, mammalian target ofctor κB; NMR, nuclearmagnet-roxisome proliferator-activatedse embryonic fibroblasts; NOX,M, standard error of the mean;nt binding protein; TCM, tradi-TR-FRET, time-resolvedfluores-

erms of the Creative Commonstribution, and reproduction inre credited.: +43 1 4277 55969.

iology, Faculty of Pharmacy,, Indonesia.

blished by Elsevier B.V. All rights re

1. Introduction

Sedentary lifestyle with low physical activity and high caloric in-take promotes obesity, the metabolic syndrome, and type 2 diabetes,which pose a major risk for the individual's quality of life and a bur-den to the health care systems of industrialized societies. Peroxisomeproliferator activated receptor gamma (PPARγ) agonists are clinicallyused to combat hyperglycemia common to these pathological condi-tions and to alleviate related comorbidities [1–3]. Generally, PPARsare nuclear receptors and ligand-dependent transcription factorswhich control lipid and glucose metabolism [4–7]. Upon ligand bind-ing, PPARs form heterodimers with the retinoid X receptor (RXR), an-other nuclear receptor, and bind to response elements located in thepromoter region of their target genes [8]. After recruitment of nuclearreceptor coactivators [9,10] further chromatin rearrangement tran-scription is initiated [11]. From the three known subtypes of PPAR(α, β/δ, and γ), PPARγ is the best studied. It is expressed in adiposetissue, lung, large intestine, kidney, liver, heart, and macrophages

served.

4814 A.G. Atanasov et al. / Biochimica et Biophysica Acta 1830 (2013) 4813–4819

[12]. The well-established important role in the regulation of glucoseand lipid metabolism renders PPARγ a valid pharmacological targetfor combating metabolic diseases [13,14]. Currently available fullPPARγ agonists represented by thiazolidinediones (e.g. pioglitazone)are clinically effective, but they have serious side- and off-targeteffects (e.g. weight gain or edema formation), urging the retrieval ofnew PPARγ agonists. Ligands acting as partial agonists inducesubmaximal receptor activation and have been demonstrated toretain beneficial anti-diabetic properties with reduced side effects.Despite lower activation of the receptor, they are still able to inducePPARγ target genes responsible for the anti-hyperglycemic and insu-lin sensitizing, but not for the unwanted PPARγ actions [15,16]. Thereason for this is not entirely understood. However, it is conceivablethat partial agonists may elicit a different conformation of the receptor–ligand complex that leads—by altered recruitment of transcriptionalco-activators and repressors—to a more restricted set of expressed targetgenes compared to full agonists.

Using a computer-aided approach we have recently identifiedneolignans as a novel class of partial agonists occupying the PPARγligand-binding domain as dimers [17]. The hereby generated and opti-mized in silico tools now enabled us to identify and further characterizethe neolignan honokiol, a major bioactive constituent of the traditionalChinese herbal drug Magnolia bark, as novel non-adipogenic par-tial PPARγ activator. The previously identified neolignans exhibitPPARγ-dependent adipogenic properties similar to pioglitazone [17].In contrast, we show in this work that the newly identified PPARγpartial agonist honokiol does not trigger adipogenesis in two in vitrocell systems and prevents weight gain in the murine KKAy in vivodiabetes model, while retaining anti-hyperglycemic activity in vitroand in vivo.

2. Materials and methods

2.1. Chemicals, cell culture reagents, and plasmids

Fetal calf serum (FCS), newborn bovine serum (NBS) andDulbecco's modified Eagle's medium (DMEM) were from Lonza(Basel, Switzerland). Pioglitazone used for all experiments except invivo tests was from Molekula Ltd. (Shaftesbury, UK). All otherchemicals were from Sigma-Aldrich (Vienna, Austria). The test com-pounds were dissolved in dimethyl sulfoxide (DMSO), divided into al-iquots and kept frozen until use. In all experiments, DMSO was usedas solvent control. For in vitro and cell-based assays, the final concen-tration of DMSO was kept ≤0.2%. The PPAR luciferase reporter plas-mid (tk-PPREx3-luc) was a gift from Prof. Ronald M. Evans (HowardHughes Medical Institute, La Jolla, CA) [18], the plasmid encoding en-hanced green fluorescent protein (pEGFP-N1) was from Clontech(Mountain View, CA), and the plasmid encoding human PPARγ(pSG5-PL-hPPAR-γ1) was a gift from Prof. Walter Wahli and Prof. Be-atrice Desvergne (Center for Integrative Genomics, University of Lau-sanne, Switzerland) [19].

2.2. Isolation of honokiol

880 g powdered bark of Magnolia officinalis (Plantasia, Oberndorf,Austria; lot: 710786) was exhaustively macerated withdichloromethane (8.0 l, 12 times) at room temperature, yielding96.8 g crude extract. 80.0 g extract was separated by flash silica gelcolumn chromatography (400 g silica gel 60, 40–63 μm, Merck,VWR, Darmstadt, Germany; 41 × 3.5 cm) using a petroleum ether–acetone gradient with an increasing amount of acetone. The eluatewas collected in portions of 20 ml and analyzed by TLC. Comparableportions were combined to 18 fractions (A1–A18). Fraction A8(5.761 g) was further separated by Sephadex LH-20-column chroma-tography (CC) using methanol as mobile phase. Fractions were mon-itored by TLC and combined according to their composition to 15

fractions (B1–B15). Fraction B8 (810 mg) was further purified byHSCCC with a mixture of PE–EtOAc–MeOH–H20 1 + 0.5 + 1 + 0.5(v/v/v/v) using the upper layer as mobile phase. HSCCC was set to thetail to headmode with 800 rpm and a flow rate of 1 ml/min. The eluatewas collected in portions of 5 ml and combined to 18 fractions (C1–C18). Fraction C11 (eluted at 495–580 ml; 249 mg) was further puri-fied by Sephadex LH-20-CC using a dichloromethane–acetone mixture(85 + 15, v/v) as mobile phase to yield 217.54 mg of honokiol in a pu-rity (HPLC) of N98%. Identity of the obtained compound was provenby comparison of 1- and 2D NMR-spectra with literature [20] andby LC–MS analysis. The obtained NMR- and LC–MS-data are provid-ed as online supplement (supplemental Table S1 and supplementalFigs. S1–S4). The isolated honokiol was used for all in vitrobioassays.

2.3. PPARγ luciferase reporter gene transactivation

HEK-293 cells (ATCC, Manassas, VA) were grown in DMEMsupplemented with 2 mM glutamine, 100 U/ml benzylpenicillin,100 μg/ml streptomycin, and 10% FBS. The cells were seeded in10-cm dishes at a density of 6 × 106 cells/dish for 18 h, and thentransfected by the calcium phosphate precipitation method with4 μg of PPARγ expression plasmid, 4 μg of reporter plasmid(tk-PPREx3-luc), and 2 μg of pEGFP-N1 as internal control. Six hourslater, the cells were reseeded in 96-well plates (5 × 104 cells/well)in DMEM without phenol red with 5% charcoal-stripped FBS, gluta-mine and antibiotics. The cells were treated as indicated and incubat-ed for 18 h. After cell lysis, the luminescence of the firefly luciferaseand the fluorescence of EGFP were quantified on a GeniosPro platereader (Tecan, Grödig, Austria). The luminescence signals were nor-malized to the EGFP-derived fluorescence, to account for differencesin cell number or transfection efficiency. Neither pioglitazone norhonokiol interfered with the luciferase assay background determinedupon transfection of the cells with tk-PPREx3-luc in the absence ofPPARγ.

2.4. PPARγ competitive ligand binding

The LanthaScreen® time-resolved fluorescence resonance energytransfer (TR-FRET) PPARγ competitive binding assay (Invitrogen,Lofer, Austria) was performed using the manufacturer's protocol.The test compounds dissolved in DMSO or solvent vehicle (DMSO)alone were incubated together with the human PPARγ LBD taggedwith GST, terbium-labeled anti-GST antibody and fluorescently la-beled PPAR ligand (Fluormone Pan-PPAR Green). In this assay, thefluorescently labeled ligand is binding to the human PPARγ LBD,which brings it in close spatial proximity to the terbium-labeledanti-GST antibody. Excitation of the terbium at 340 nm results in en-ergy transfer (FRET) and partial excitation of the fluorescent PPAR li-gand, followed by emission at 520 nm. Test-compounds binding tothe human PPARγ LBD are competing with the fluorescently labeledligand and displacing it, resulting in a decrease of the FRET signal.The signal obtained at 520 nm is normalized to the signal obtainedfrom the terbium emission at 495 nm; therefore, the decrease in the520 nm/495 nm ratio is used as a measure for the ability of the testedcompounds to bind to the human PPARγ LBD. Neither pioglitazonenor honokiol interfered with the background 520 nm/495 nm fluo-rescence in the absence of PPARγ LBD.

2.5. Induction of adipogenic differentiation

3T3-L1 preadipocytes were subcultivated in DMEM (containing2 mM glutamine, 100 U/ml benzylpenicillin, 100 μg/ml streptomycin)plus 10% NBS. For differentiation experiments they were grown tillconfluency in 12-well plates. Two days later cells were treated withDMEM/10% FCS (negative control), DMEM/10% FCS/1 μg/ml insulin

4815A.G. Atanasov et al. / Biochimica et Biophysica Acta 1830 (2013) 4813–4819

(INS, basal 3T3 differentiation medium), or INSmedium + PPARγ ago-nists (1, 3 and 10 μM). An additional set of cells exposed to thehighest concentration of the respective agonist was pretreatedwith the PPARγ antagonist BADGE (50 μM) for 30 min in order to as-sess the PPARγ-dependency of an observable lipid accumulation. Allwells contained 0.2% DMSO. Medium was renewed every other day.After eight days, lipid accumulation (asmeasure of adipogenic differ-entiation) was determined by OilRed O staining. Solubilized dye wasquantified at 550 nm. Experiments were performed at least threetimes.

Immortalized mouse embryonic fibroblasts (MEF) (kind gift ofProf. Thomas Kensler, University of Pittsburgh) were differentiatedinto adipocytes as described [21] with minor modifications. Cellswere grown in DMEM/10% FCS till confluency (12-well plate). Twodays later, cells were treated with DMEM/10% FCS (negative control)or basal MEF diff (differentiation) cocktail (1 μg/ml insulin/200 μMisobutylmethylxanthine (IBMX)/500 nM dexamethasone/25 ng/mlbone morphogenic protein (BMP) 4) in the presence and absenceof PPARγ agonists (1–10 μM). An additional set of cells exposed tothe highest concentration of the respective agonist was pretreatedwith the PPARγ antagonist BADGE (50 μM) for 30 min in order to as-sess the PPARγ-dependency of a potential lipid accumulation. Allwells contained 0.2% DMSO. After three days, the medium waschanged to DMEM/10% FCS for control cells and to DMEM/10% FCS/1 μg/ml ± PPARγ agonists (1–10 μM) for differentiating cells, re-spectively. After an overall differentiation time of nine days, lipid ac-cumulation was determined by OilRed O staining (see supplementalFig. S6). Solubilized dyewas quantified at 550 nm. Experiments wereperformed at least three times.

2.6. OilRed O staining

Cells in plates were fixed with 10% formaldehyde for at least 1 h,washed with 60% isopropanol and dried, before filtered OilRed O solu-tion (2% OilRed O in 60% isopropanol) was added for 10 min. Afterwashing off the excessive dye with water, photos were taken, andbound dye was solubilized with equal volumes of 100% isopropanolper well and immediately photometrically quantified at 550 nm in aspectrophotometer (Tecan, Grödig, Austria).

2.7. 2-Deoxy-D-(1H3)-glucose uptake assay in adipocytes

Superconfluent 3T3-L1 preadipocytes were differentiated intomature adipocytes for 10 days (2 days in DMEM (containing 2 mMglu-tamine and antibiotics) supplemented with 10% FCS, 500 μM IBMX,1 μg/ml insulin and 500 nM dexamethasone followed by 2 days inDMEM/10% FCS/1 μg/ml insulin and 6 days in DMEM/10% FCS), andincubated with the test compounds or solvent vehicle for 24 h at37 °C. Before the experiment, cells were incubated for 1 h in KRHbuffer (50 mM HEPES, 136 mM NaCl, 23.5 mM KCl, 1.25 mM MgSO4,1.25 mM CaCl2 and 0.1% BSA). After an additional KRH buffer change,cells were allowed to rest for 10 min, then glucose uptake was initiatedby addition of 2-deoxy-D-glucose spikedwith 2-deoxy-D-(1H3)-glucose(final concentrations of 0.1 mM and 0.45 μCi/ml). After 15 min thereaction was stopped by three rapid washes with ice-cold PBS. Theglucose uptake rate was determined by liquid scintillation counting(PerkinElmer, Waltham, MA, USA) of cell lysates (0.05 N NaOH in PBS)and normalized to protein content assessed by the Bradford proteinassay and uptake time (to obtain incorporated mol glucose per mg pro-tein and minute).

2.8. Animal experiments

The KKAy mice represent a model for obesity and type 2 diabetes,with spontaneous development of hyperglycemia, hyperinsulinemia,glucose intolerance, and obesity [22,23]. KKAy mice (female, SPF

grade, qualification license No. SCXK (Beijing) 2009-0004, certificationNo. 0237501, utilization license No. SYXK (Military) 2007-030), andC57B1/6J mice were all supplied by Beijing Hua Fu Kang Bio-technologyCompany Ltd. All animal tests were conducted with full compliancewith local, national, and ethical principles and authority regulations.The KKAy mice were raised in a SPF experimental room, and servedstandard high fat diet (composed from sucrose, pork lard, egg yolkpowder, and ordinary feed in ratio of 10:10:10:7; from Beijing HuaFu Kang Bio-technology Company Ltd.) and freely available drinkingwater. Age-mated C57B1/6J mice were served ordinary feed andunrestricted drinking water. When the blood glucose of KKAy micereached ≥13.9 mM (confirming the successful establishment of thediabetic model) the animals were separated into four groups (10 ani-mals per group): (1.) “No treatment” group received through oralgavage administration sterilized water (0.1 ml/10 g body weight) onceper day for 35 days; (2.) “Vehicle” group received alcoholic (0.5%)CMCNa (5%) solution (0.1 ml/10 g body weight); (3.) “Pioglitazone”group received through oral gavage pioglitazonehydrochloride solution(Santa Cruz Biotechnology, Heidelberg, Germany; 1 mg/ml alcoholic(0.5%)) CMCNa (5%) solution (0.1 ml/10 g body weight), once per day(10 mg/kg/d) for 35 days; (4.) “Honokiol” group received throughoral gavage honokiol solution (Carbosynth, Compton, UK; 10 mg/ml al-coholic (0.5%)) CMCNa (5%) solution (0.1 ml/10 g body weight) onceper day (the 100 mg/kg/d treatment dosewas chosen because that con-centration demonstrated to have a clear effect in preliminary experi-ments), for 35 days. The C57B1/6J control group received through oralgavage administration sterilized water (0.1 ml/10 g body weight),once per day for 35 days. ACCU-CHEK advantage blood glucose detec-tion system fromRochewas used to determine the blood glucose levels.Glucose tolerance test was performed after 35 days of the indicatedtreatments following a 3 h of fasting. The blood glucose was measuredbefore (0 min) or 60 min and 120 min after administration of 2.5 g/kgglucose. Area under curve (AUC) was estimated using the formulaAUC = 1 / 2(0 h + 1 h) ∗ 0.5 + 1 / 2(1 h + 2 h) ∗ 1.5. To determineinsulin levels (right after the 35 day treatment period) the rat/mouse In-sulin ELISA Kit fromMillipore was used according to the manufacturers'instructions.

No differences between the treatment groups were detected interms of routine activity of the animals, hair condition, food intakeand water intake.

2.9. Molecular docking

Docking experiments of honokiol into the hPPARγ ligand bindingdomain (Protein Data Bank, www.pdb.org, PDB ID: 2VSR [24]) wereperformed using the CDOCKER module of Discovery Studio 2.5(www.accelrys.com) with default settings. For the first ligand, thebinding site was defined as a 9 Å sphere, centered on the guanidinecarbon of Arg288. The binding site for the second ligand was definedas an 8 Å sphere, centered on the sulfur atom of Cys285. Predictedbinding poses were visualized and ligand–protein interactions ana-lyzed with LigandScout 3.03 (www.inteligand.com).

2.10. Statistical methods and data analysis

Statistical analyseswere donewith the GraphPad Prism software ver-sion 4.03. One-way or two way analysis of variance (ANOVA) withBonferroni post hoc test was used to determine statistical significance.Nonlinear regression (sigmoidal dose response) was used to calculatethe EC50 values and maximal fold activation. Ki values of the investigatedcompounds in the receptor binding assay were calculated with theCheng–Prusoff equation: (Ki) = IC50 / (1 + L / KD)), where IC50 is theconcentration of the competitor that produces 50%displacement of the li-gand, L is the concentration of Fluormone™ Pan-PPAR Green used in theassay (5 nM), and KD is the binding constant (2.8 nM) of Fluormone™Pan-PPAR Green to PPARγ-LBD [25].

4816 A.G. Atanasov et al. / Biochimica et Biophysica Acta 1830 (2013) 4813–4819

3. Results

3.1. Identification of honokiol as a novel partial PPARγ agonist

We aimed to characterize novel partial PPARγ ligands, which canreduce blood sugar levels without inducing weight gain. Using thepharmacophore model collection previously developed [26], wehave recently identified several neolignans as representatives of anovel class of partial PPARγ agonists binding to the receptor LBDas dimers [17]. In the present study another structurally similarneolignan, the natural product honokiol (Fig. 1A), was predictedby molecular docking to also bind to PPARγ in a dimeric mode(Fig. 1B). In order to confirm that honokiol indeed binds to PPARγwe performed a competitive FRET-based ligand binding assay withthe purified human PPARγ LBD. Honokiol directly bound to the puri-fied human PPARγ LBD in vitro (Fig. 2A; Ki = 22.86 μM), althoughweaker than pioglitazone (Figs. 1A and 2A; Ki = 0.69 μM), whichserved as a reference full agonist for all conducted experiments.Honokiol also bound to the LBDs of PPARα and PPARβ/δ but only atconsiderably (three to four-fold) higher concentrations than required

Fig. 1. Chemical structures (A) of pioglitazone and honokiol, and honokiol modeled in the PPtacts, red and green arrows hydrogen bonds between honokiol and Cys285 and Ser289, res

Fig. 2. PPARγ receptor binding and luciferase reporter transactivation of honokiol. (A) PPARγDMSO and then mixed with a buffer solution containing the human PPARγ LBD tagged with1 h of incubation, the ability of the test compounds to bind to the PPARγ LBD and thus dispratio 520 nm/495 nm upon excitation at 340 nm. Each data point represents the mean ±duplicate. (B) Influence of honokiol on human PPARγ-mediated luciferase reporter gefull-length human PPARγ, a reporter plasmid containing PPAR-response element coupled toconcentrations of honokiol and pioglitazone for 18 h. Luciferase activity was normalized to twith the negative control (DMSO vehicle treatment). The data shown are means ± SEM of

for PPARγ binding (data not shown). In a next step we examinedhonokiol for its transactivation potential using a PPARγ-driven lucif-erase reporter gene assay. Honokiol hereby displayed partial PPARγagonism. It activated the receptor with an EC50 value of 3.9 μM andreached maximally two-fold activation compared to 12-fold maximalactivation with an EC50 value of 0.3 μM by the full agonist pioglitazonein the same system (Fig. 2B).

3.2. Honokiol stimulates basal glucose uptake in 3T3 adipocytes

In order to assess PPARγ agonism in a functional cell model with en-dogenous levels of PPARγ, we investigated the potential anti-hyperglycemic activity of honokiol and pioglitazone in mature 3T3-L1adipocytes. We incubated the cells with different non-toxic concentra-tions of pioglitazone or honokiol (1–10 μM, supplemental Fig. S5) for24 h and determined the basal cellular glucose uptake rate. As seen inFig. 3, both compounds enhanced the cellular glucose uptake in adipo-cytes in a concentration dependentmanner and also to a comparable ex-tent. At a concentration of 10 μM both compounds approximately

ARγ ligand-binding pocket (B). Yellow spheres mark hydrophobic protein–ligand con-pectively.

binding of honokiol. Serial dilutions of the honokiol and pioglitazone were prepared inGST, a terbium-labeled anti-GST antibody, and fluorescently labeled PPAR agonist. Afterlace the fluorescently labeled ligand was estimated from the decrease of the emissionstandard error of the mean (SEM) from three independent experiments performed inne transactivation. HEK-293 cells, transiently transfected with a plasmid encodinga luciferase reporter, and EGFP as internal control, were stimulated with the indicated

he EGFP-derived fluorescence, and the result was expressed as fold induction comparedthree independent experiments each performed in quadruplicate.

Fig. 3. Basal glucose uptake rate upon treatment with honokiol in differentiated adipo-cytes. Differentiated 3T3-L1 adipocytes were treated with DMSO (solvent control; 0 μM)or different concentrations of pioglitazone (Pio) and honokiol (Hon) (1–10 μM) for 24 hbefore the basal cellular glucose uptake rate was determined as described in detail inthe “Materials and methods” section. The bar graph depicts compiled results of three in-dependent experiments (means ± SEM, *p b 0.05, statistically different to DMSO controlgroup, ANOVA, Bonferroni post test).

4817A.G. Atanasov et al. / Biochimica et Biophysica Acta 1830 (2013) 4813–4819

doubled the amount of ingested glucose compared to vehicle-treatedcells.

3.3. Honokiol does not induce adipogenic differentiation in vitro

We further assessed the adipogenic potential of honokiol andpioglitazone. For this, we chose two different cellular adipogenesismodels [27]. In both 3T3-L1 preadipocytes (Fig. 4A) and immortal-ized mouse embryonic fibroblasts (Fig. 4B) pioglitazone significantlyenhanced adipogenic differentiation in a concentration-dependentmanner. The induction of differentiation by pioglitazone wasPPARγ-dependent and not an off-target result of pioglitazone treat-ment as it was blunted by co-treatment with the known PPARγ an-tagonist BADGE. Honokiol, even at the highest test concentration(10 μM), did not lead to a markedly enhanced formation of intracel-lular lipid droplets above basal levels (see also supplemental Fig. S6).

3.4. Honokiol prevents in vivo exacerbation of hyperglycemia and weightgain in the KKAy murine diabetes model

So far we have shown that honokiol binds to PPARγ and is a partial(eliciting submaximal activation) and apparently selective (hypergly-cemic but not adipogenic) PPARγ modulator in vitro. We went on to

A B

Fig. 4. Adipogenic potential of honokiol. 3T3-L1 preadipocytes (A) and mouse embryonic fibtiate into adipocytes and accumulate lipid in a PPARγ-dependent way as described in detailtiation medium, Pio: pioglitazone (0–10 μM), Hon: honokiol (1–10 μM), Diff Cocktail: bamorphogenic protein 4). BADGE (50 μM) was used as PPARγ antagonist. After staining theFig. S6). Bound dye was solubilized and quantified spectrophotometrically at 550 nm. The biments (means ± SEM, *p b 0.05, statistically different to “INS” and “Diff cocktail”, respecti

challenge these promising findings with the in vivo situation. Forthis, we chose diabetic KKAy mice which are characterized by gradu-ally worsening obesity and hyperglycemia. We administeredhonokiol (100 mg/kg), pioglitazone (10 mg/kg), or vehicle for35 days to these mice. Whereas both pioglitazone and honokiol suc-cessfully prevented exacerbation of hyperglycemia (Fig. 5A), onlyhonokiol also significantly suppressed weight gain (Fig. 5B). Further-more, honokiol significantly (p b 0.001) improved the glucose toler-ance and the insulin levels of treated animals (Table 1). Theseresults fit well with the in vitro experiments outlining honokiol asan efficient non-adipogenic blood glucose lowering agent, althoughfurther in vivo experiments, particularly with lower administrationdoses, are required to better estimate its therapeutic potential.

4. Discussion

We report in this study the identification and characterization ofthe natural product honokiol as a novel non-adipogenic partialPPARγ agonist. Interestingly, honokiol is one of the main bioactiveconstituents (1–5% content) of the traditional Chinese herbal drugMagnolia bark (Cortex Magnoliae, Hou Po), used to treat metabolicdisease among others [28], and may thus contribute to the beneficialproperties of this herbal remedy via partial activation of PPARγ.

Honokiol is a representative of the chemical class of neolignans,which we recently identified as novel partial PPARγ agonists bindingto the receptor LBD as dimers [17]. Shortly after our findings,magnolol, one of our investigated neolignan-type PPARγ agonists,was shown to also bind to RXRα [29]. The co-crystal structures withPPARγ and RXRα revealed a unique mode of dual PPARγ/RXRagonism by magnolol [29]. In addition to the partial PPARγ agonismthat we demonstrate here (Fig. 2A,B), honokiol has recently alsobeen shown to directly bind to RXR [30]. Honokiol therefore conformsto the dual PPARγ/RXR binding mode demonstrated for the structur-ally similar magnolol [29]. However, whereas all previously describedneolignans, including the dual PPARγ/RXR agonist magnolol, inducedPPARγ-dependent lipid accumulation in 3T3-L1 adipocytes similar topioglitazone [17], honokiol did not (Fig. 4) and therefore emerges assuperior novel PPARγ modulator. Interestingly, in contrast to ourfindings Choi et al. [31] reached the conclusion that honokiol didnot directly bind to PPARγ and induced adipogenic differentiation in3T3-L1 preadipocytes. One reason for this discrepancy could be thedifferent ranges of used honokiol concentrations. Whereas Choi etal. tested honokiol up to 30 μM, we have also investigated the effect

roblasts (B) were grown in 12-well plates to superconfluency and induced to differen-in the “Materials and methods” section (INS: insulin-containing 3T3-L1 basal differen-sal MEF differentiation cocktail containing insulin, IBMX, dexamethasone and bonecells with the lipophilic dye OilRed O the plates were photographed (see supplementalar graphs depict compiled results of four (3T3-L1) or three (MEF) independent exper-vely; ANOVA, Bonferroni post-test).

Fig. 5. Effect of honokiol on the blood glucose and body weight of diabetic KKAy mice. Blood glucose (A) and body weight (B) of KKAy mice administered orally with honokiol(100 mg/kg/day), pioglitazone (10 mg/kg/day) or vehicle for 35 days is presented (box and whisker plots representing median, upper and lower quartiles, and lowest and highestdetected values. ***p b 0.001, **p b 0.01, *p b 0.05 (n = 10, one-way ANOVA/Bonferroni)).

4818 A.G. Atanasov et al. / Biochimica et Biophysica Acta 1830 (2013) 4813–4819

of the compound at several fold higher concentrations (Fig. 1A). An-other relevant consideration is that those authors based the exclusionof honokiol as direct PPARγ-binder on a negative result from aTRAP220/DRIP-2 coactivator peptide recruitment assay [31]. An alter-native explanation for that observation could be that the binding ofhonokiol to PPARγ LBD induces a receptor–ligand complex with aconformation that is not optimal for recruitment of the TRAP220/DRIP-2 coactivator peptide in particular. In support of the latter inter-pretation, our receptor-binding assay (Fig. 2A) unbiasedly shows thathonokiol is able to displace fluorescently labeled PPARγ agonist fromthe LBD and thus directly binds to the ligand binding pocket of PPARγ.It is well known that different ligands induce receptor–ligand com-plexes with different conformations, leading to the recruitment ofdistinct sets of coactivators to PPAR-γ and expression of distinctsets of genes [32,33]. In line, another partial PPARγ agonist,MBX-102/JNJ39659100, was previously demonstrated to be unableto trigger TRAP220 recruitment to PPARγwhile retaining antidiabeticproperties without inducing weight gain [34]. To carefully investigatea potential pro-adipogenic activity of honokiol we used two differentin vitro adipogenesis models. Whereas the full agonist pioglitazoneinduced a potent PPARγ-dependent adipogenic differentiation inboth cellular models, honokiol did not display adipogenic properties(Fig. 4A and B). Moreover, the favorable activity profile observed invitro was further confirmed in vivo in the diabetic KKAy mousemodel in which honokiol not only prevented rising hyperglycemiasimilar to pioglitazone (Fig. 5A) but also suppressed weight gain(Fig. 5B) and clearly negatively influenced adipogenesis in our

Table 1Effect of honokiol or pioglitazone treatment on the glucose tolerance of KKAy mice. Glucose tindicated treatments following 3 h of fasting. The blood glucose was measured before (0 musing the formula AUC = 1 / 2(0 h + 1 h) ∗ 0.5 + 1 / 2(1 h + 2 h) ∗ 1.5. The shown datKKAy: vehicle group (ANOVA/Bonferroni).

Blood glucose (mM)

0 min 60 min

C57B1/6J 6.11 ± 1.03*** 6.83 ± 1.02***KKAy: no treatment 29.57 ± 4.29n.s. 30.6 ± 3.02n.s.

KKAy: vehicle 27.25 ± 3.25 29.4 ± 2.71KKAy: pioglitazone 19.54 ± 6.43*** 19.96 ± 5.30***KKAy: honokiol 15.59 ± 2.71*** 22.94 ± 5.51⁎⁎

experiments. At the current stage, however, it cannot be excluded thatproteins other than PPARγ contribute to the beneficial in vivo effectsof honokiol. Further support for PPARγ activation by honokiol in vivomay be provided in a recent work from Diano et al. who show thathonokiol administration mimics the modulatory effect of rosiglitazoneon the activation of hypothalamic neuronal circuits involved in bodyweight control [35]. However, this observation was then interpretedas a consequence of the ROS scavenging properties of honokiol, andthe possibility that the compound could act as a direct activator ofPPARγ was not considered by the authors.

Overall, the here identified partial PPARγ activation pattern, incombination with suppression of hyperglycemia and weight gain,makes honokiol an interesting natural product with a good poten-tial to be further explored as pharmaceutical lead or dietarysupplement.

Conflict of interest statement

The authors have no conflict of interest concerning this work.

Acknowledgements

The expert technical assistance by Elisabeth Geiger, JudithBenedics, Mia Gössinger, Daniel Schachner and Hortenzia Beres isgratefully acknowledged. This study was supported by grants fromthe Austrian Science Fund (FWF) [NFN S10704, S10702/10711,S10703, S10701 and P23317-B11], the Austrian Federal Ministries of

olerance test and blood insulin level determination were performed after 35 days of thein) or 60 min and 120 min after administration of 2.5 g/kg glucose. AUC was estimateda represent mean ± SD. ***p b 0.001, **p b 0.01, n.s. not significant, compared to the

AUC Blood insulin (ng/ml)

120 min

6.16 ± 1.08*** 12.98 ± 1.60*** 0.51 ± 0.15***26.38 ± 4.49n.s. 57.80 ± 5.94n.s. 33.52 ± 3.34n.s.

26.93 ± 2.20 56.42 ± 3.03 29.43 ± 6.4416.64 ± 4.72*** 37.34 ± 10.20*** 24.39 ± 8.65n.s.

12.63 ± 3.64⁎⁎⁎ 36.32 ± 7.41 ⁎⁎⁎ 15.23 ± 7.71***

4819A.G. Atanasov et al. / Biochimica et Biophysica Acta 1830 (2013) 4813–4819

Science and Research as well as Health [TCM and Age-Related Dis-eases] and National Natural Science Foundation of China (NSFC)[81274051].

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.bbagen.2013.06.021.

References

[1] F. Lalloyer, B. Staels, Fibrates, glitazones, and peroxisome proliferator-activatedreceptors, Arterioscler. Thromb. Vasc. Biol. 30 (2010) 894–899.

[2] B. Staels, J.C. Fruchart, Therapeutic roles of peroxisome proliferator-activated re-ceptor agonists, Diabetes 54 (2005) 2460–2470.

[3] D. Patsouris, M. Muller, S. Kersten, Peroxisome proliferator activated receptor li-gands for the treatment of insulin resistance, Curr. Opin. Investig. Drugs 5(2004) 1045–1050.

[4] B. Desvergne, L. Michalik, W. Wahli, Transcriptional regulation of metabolism,Physiol. Rev. 86 (2006) 465–514.

[5] R.M. Evans, G.D. Barish, Y.X. Wang, PPARs and the complex journey to obesity,Nat. Med. 10 (2004) 355–361.

[6] M.K. Hansen, T.M. Connolly, Nuclear receptors as drug targets in obesity,dyslipidemia and atherosclerosis, Curr. Opin. Investig. Drugs 9 (2008) 247–255.

[7] K. Schoonjans, G. Martin, B. Staels, J. Auwerx, Peroxisome proliferator-activated re-ceptors, orphans with ligands and functions, Curr. Opin. Lipidol. 8 (1997) 159–166.

[8] O. Bardot, T.C. Aldridge, N. Latruffe, S. Green, PPAR–RXR heterodimer activates aperoxisome proliferator response element upstream of the bifunctional enzymegene, Biochem. Biophys. Res. Commun. 192 (1993) 37–45.

[9] K. Ge, M. Guermah, C.X. Yuan, M. Ito, A.E. Wallberg, B.M. Spiegelman, R.G. Roeder,Transcription coactivator TRAP220 is required for PPAR gamma 2-stimulatedadipogenesis, Nature 417 (2002) 563–567.

[10] T.P. Yao, G. Ku, N. Zhou, R. Scully, D.M. Livingston, The nuclear hormone receptorcoactivator SRC-1 is a specific target of p300, Proc. Natl. Acad. Sci. U. S. A. 93(1996) 10626–10631.

[11] M. Ricote, C.K. Glass, PPARs and molecular mechanisms of transrepression,Biochim. Biophys. Acta 1771 (2007) 926–935.

[12] B.D. Abbott, Review of the expression of peroxisome proliferator-activated recep-tors alpha (PPAR alpha), beta (PPAR beta), and gamma (PPAR gamma) in rodentand human development, Reprod. Toxicol. 27 (2009) 246–257.

[13] L. Michalik, J. Auwerx, J.P. Berger, V.K. Chatterjee, C.K. Glass, F.J. Gonzalez, P.A.Grimaldi, T. Kadowaki, M.A. Lazar, S. O'Rahilly, C.N. Palmer, J. Plutzky, J.K. Reddy,B.M. Spiegelman, B. Staels, W. Wahli, International Union of Pharmacology. LXI.Peroxisome proliferator-activated receptors, Pharmacol. Rev. 58 (2006) 726–741.

[14] R.K. Semple, V.K. Chatterjee, S. O'Rahilly, PPAR gamma and human metabolic dis-ease, J. Clin. Invest. 116 (2006) 581–589.

[15] K. Bhalla, B.J. Hwang, J.H. Choi, R. Dewi, L. Ou, J. McLenithan, W. Twadell, E.Pozharskiy, J. Stock, G. Girnun, N-Acetyl Farnesyl cysteine is a novel class ofPPAR{gamma} ligand with partial and full agonist activity in vitro and in vivo, J.Biol. Chem. 286 (2011) 41626–41635.

[16] R. Agrawal, P. Jain, S.N. Dikshit, Balaglitazone: a second generation peroxisomeproliferator-activated receptor (PPAR) gamma (gamma) agonist, Mini-Rev. Med. Chem.12 (2012) 87–97.

[17] N. Fakhrudin, A. Ladurner, A.G. Atanasov, E.H. Heiss, L. Baumgartner, P. Markt, D.Schuster, E.P. Ellmerer, G. Wolber, J.M. Rollinger, H. Stuppner, V.M. Dirsch,Computer-aided discovery, validation, and mechanistic characterization of novelneolignan activators of peroxisome proliferator-activated receptor gamma, Mol.Pharmacol. 77 (2010) 559–566.

[18] S.A. Kliewer, K. Umesono, D.J. Noonan, R.A. Heyman, R.M. Evans, Convergence of9-cis retinoic acid and peroxisome proliferator signalling pathways throughheterodimer formation of their receptors, Nature 358 (1992) 771–774.

[19] L. Peyrin-Biroulet, J. Beisner, G. Wang, S. Nuding, S.T. Oommen, D. Kelly, E.Parmentier-Decrucq, R. Dessein, E. Merour, P. Chavatte, T. Grandjean, A.Bressenot, P. Desreumaux, J.F. Colombel, B. Desvergne, E.F. Stange, J. Wehkamp,M. Chamaillard, Peroxisome proliferator-activated receptor gamma activation isrequired for maintenance of innate antimicrobial immunity in the colon, Proc.Natl. Acad. Sci. U. S. A. 107 (2010) 8772–8777.

[20] Y. Shoji, N. Takashi, K. Akihide, N. Toshihiro, N. Itsuo, Isolation and characteriza-tion of phenolic compounds from magnoliae cortex produced in China, Chem.Pharm. Bull. 39 (1991) 2024–2036.

[21] S. Shin, N. Wakabayashi, V. Misra, S. Biswal, G.H. Lee, E.S. Agoston, M. Yamamoto,T.W. Kensler, NRF2 modulates aryl hydrocarbon receptor signaling: influence onadipogenesis, Mol. Cell. Biol. 27 (2007) 7188–7197.

[22] H. Iwatsuka, A. Shino, Z. Suzuoki, General survey of diabetic features of yellow KKmice, Endocrinol. Jpn. 17 (1970) 23–35.

[23] S.S. Braithwaite, B. Palazuk, J.R. Colca, C.W. Edwards III, C. Hofmann, Reduced ex-pression of hexokinase II in insulin-resistant diabetes, Diabetes 44 (1995) 43–48.

[24] T. Itoh, L. Fairall, K. Amin, Y. Inaba, A. Szanto, B.L. Balint, L. Nagy, K. Yamamoto,J.W. Schwabe, Structural basis for the activation of PPARgamma by oxidizedfatty acids, Nat. Struct. Mol. Biol. 15 (2008) 924–931.

[25] Y. Cheng, W.H. Prusoff, Relationship between the inhibition constant (K1) and theconcentration of inhibitor which causes 50 per cent inhibition (I50) of an enzy-matic reaction, Biochem. Pharmacol. 22 (1973) 3099–3108.

[26] P. Markt, D. Schuster, J. Kirchmair, C. Laggner, T. Langer, Pharmacophore model-ing and parallel screening for PPAR ligands, J. Comput. Aided Mol. Des. 21(2007) 575–590.

[27] E.D. Rosen, O.A. MacDougald, Adipocyte differentiation from the inside out, Nat.Rev. Mol. Cell Biol. 7 (2006) 885–896.

[28] G. Banos, I. Perez-Torres, M. El Hafidi, Medicinal agents in the metabolic syn-drome, Cardiovasc. Hematol. Agents Med. Chem. 6 (2008) 237–252.

[29] H. Zhang, X. Xu, L. Chen, J. Chen, L. Hu, H. Jiang, X. Shen, Molecular determinantsof magnolol targeting both RXRalpha and PPARgamma, PLoS One 6 (2011)e28253.

[30] H. Kotani, H. Tanabe, H. Mizukami, M. Makishima, M. Inoue, Identification of anaturally occurring rexinoid, honokiol, that activates the retinoid X receptor, J.Nat. Prod. 73 (2010) 1332–1336.

[31] S.S. Choi, B.Y. Cha, K. Iida, M. Sato, Y.S. Lee, T. Teruya, T. Yonezawa, K. Nagai, J.T.Woo, Honokiol enhances adipocyte differentiation by potentiating insulin signal-ing in 3T3-L1 preadipocytes, J. Nat. Med. 65 (2011) 424–430.

[32] P. Misra, R. Chakrabarti, R.K. Vikramadithyan, G. Bolusu, S. Juluri, J. Hiriyan, C.Gershome, A. Rajjak, P. Kashireddy, S. Yu, S. Surapureddi, C. Qi, Y.J. Zhu, M.S.Rao, J.K. Reddy, R. Ramanujam, PAT5A: a partial agonist of peroxisomeproliferator-activated receptor gamma is a potent antidiabetic thiazolidinedioneyet weakly adipogenic, J. Pharmacol. Exp. Ther. 306 (2003) 763–771.

[33] W. Liu, F. Lau, K. Liu, H.B. Wood, G. Zhou, Y. Chen, Y. Li, T.E. Akiyama, G. Castriota,M. Einstein, C. Wang, M.E. McCann, T.W. Doebber, M. Wu, C.H. Chang, L.McNamara, B. McKeever, R.T. Mosley, J.P. Berger, P.T. Meinke, Benzimidazolones:a new class of selective peroxisome proliferator-activated receptor gamma(PPARgamma) modulators, J. Med. Chem. 54 (2011) 8541–8554.

[34] F.M. Gregoire, F. Zhang, H.J. Clarke, T.A. Gustafson, D.D. Sears, S. Favelyukis, J.Lenhard, D. Rentzeperis, L.E. Clemens, Y. Mu, B.E. Lavan, MBX-102/JNJ39659100, anovel peroxisome proliferator-activated receptor–ligand with weak transactivationactivity retains antidiabetic properties in the absence of weight gain and edema,Mol. Endocrinol. 23 (2009) 975–988.

[35] S. Diano, Z.W. Liu, J.K. Jeong, M.O. Dietrich, H.B. Ruan, E. Kim, S. Suyama, K.Kelly, E. Gyengesi, J.L. Arbiser, D.D. Belsham, D.A. Sarruf, M.W. Schwartz,A.M. Bennett, M. Shanabrough, C.V. Mobbs, X. Yang, X.B. Gao, T.L. Horvath,Peroxisome proliferation-associated control of reactive oxygen species setsmelanocortin tone and feeding in diet-induced obesity, Nat. Med. 17 (2011)1121–1127.