Identification of dual PPARα/γ agonists and their effects on lipid metabolism

Article abstract of DOI:10.1016/j.bmc.2015.11.013

The three peroxisome proliferator-activated receptor (PPAR) isoforms; PPARα, PPARγ and PPARδ, play central roles in lipid metabolism and glucose homeostasis. Dual PPARα/γ agonists, which stimulate both PPARα and PPARγ isoforms to similar extents, are gaining popularity as it is believed that they are able to ameliorate the unwanted side effects of selective PPARα and PPARγ agonists; and may also be used to treat dyslipidemia and type 2 diabetes mellitus simultaneously. In this study, virtual screening of natural product libraries, using both structure-based and ligand-based drug discovery approaches, identified ten potential dual PPARα/γ agonist lead compounds (9-13 and 16-20). In vitro assays confirmed these compounds to show no statistically significant toxicity to cells, with the exception of compound 12 which inhibited cell growth to 74.5% ± 3.5 and 54.1% ± 3.7 at 50 μM and 100 μM, respectively. In support of their potential as dual PPARα/γ agonists, all ten compounds upregulated the expression of cholesterol transporters ABCA1 and ABCG1 in THP-1 macrophages, with indoline derivative 16 producing the greatest elevation (2.3-fold; 3.3-fold, respectively). Furthermore, comparable to the activity of established PPARα and PPARγ agonists, compound 16 stimulated triacylglycerol accumulation during 3T3-L1 adipocyte differentiation as well as fatty acid β-oxidation in HuH7 hepatocytes.

Full text of DOI:10.1016/j.bmc.2015.11.013

Accepted Manuscript
Identification of dual PPARα/γ agonists and their effects on lipid metabolism
Quanqing Gao, Jacky Hanh, Linda Váradi, Rose Cairns, Helena Sjöström,
Vivian W.Y. Liao, Peta Wood, Seher Balaban, Jennifer Ai Ong, Hsuan-Yu
Jennifer Lin, Felcia Lai, Andrew Hoy, Thomas Grewal, Paul W. Groundwater,
David E. Hibbs
PII:
S0968-0896(15)30138-3
DOI:
http://dx.doi.org/10.1016/j.bmc.2015.11.013
BMC 12661
Reference:
Bioorganic & Medicinal Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
21 September 2015
9 November 2015
13 November 2015
Please cite this article as: Gao, Q., Hanh, J., Váradi, L., Cairns, R., Sjöström, H., Liao, V.W.Y., Wood, P., Balaban,
S., Ong, J.A., Lin, H.J., Lai, F., Hoy, A., Grewal, T., Groundwater, P.W., Hibbs, D.E., Identification of dual
PPARα/γ agonists and their effects on lipid metabolism, Bioorganic & Medicinal Chemistry (2015), doi: http://
dx.doi.org/10.1016/j.bmc.2015.11.013
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Identification of dual PPARα/γ agonists and their effects
on lipid metabolism
Quanqing Gaoa, †, Jacky Hanha, †, Linda Váradia, Rose Cairnsa, Helena Sjöströma, Vivian W. Y.
Liaoa, Peta Wooda, Seher Balabanb,c, Jennifer Ai Onga, Hsuan-Yu Jennifer Lina, Felcia Laia,
Andrew Hoyb,c,d, Thomas Grewala, Paul W. Groundwatera and David E. Hibbsa, *
a Faculty of Pharmacy, University of Sydney, NSW 2006 Australia
b Discipline of Physiology, School of Medical Sciences & Bosch Institute, University of Sydney, 2006,
Australia.
c Charles Perkins Centre, University of Sydney, Sydney NSW 2006, Australia
d Boden Institute of Obesity, Nutrition, Exercise & Eating Disorders, University of Sydney, Sydney,
NSW 2006, Australia
† These two authors contributed equally
* Corresponding author: David E. Hibbs; david.hibbs@sydney.edu.au
1

Abstract
The three peroxisome proliferator-activated receptor (PPAR) isoforms; PPARα, PPARγ and PPARδ,
play central roles in lipid metabolism and glucose homeostasis. Dual PPARα/γ agonists, which
stimulate both PPARα and PPARγ isoforms to similar extents, are gaining popularity as it is believed
that they are able to ameliorate the unwanted side effects of selective PPARα and PPARγ agonists; and
may also be used to treat dyslipidaemia and type 2 diabetes mellitus simultaneously. In this study,
virtual screening of natural product libraries, using both structure-based and ligand-based drug
discovery approaches, identified ten potential dual PPARα/γ agonist lead compounds (9 – 13 and 16 –
20). In vitro assays confirmed these compounds to show no statistically significant toxicity to cells,
with the exception of compound 12 which inhibited cell growth to 74.5%
3.5 and 54.1%
3.7 at
50 μM and 100 μM respectively. In support of their potential as dual PPARα/γ agonists, all ten
compounds upregulated the expression of cholesterol transporters ABCA1 and ABCG1 in THP-1
macrophages, with indoline derivative 16 producing the greatest elevation (2.3-fold; 3.3-fold,
respectively). Furthermore, comparable to the activity of established PPARα and PPARγ agonists,
compound 16 stimulated triacylglycerol accumulation during 3T3-L1 adipocyte differentiation as well
as fatty acid β-oxidation in HuH7 hepatocytes.
2

Graphical abstract
Keywords: Type 2 Diabetes, Dyslipidemia, Cholesterol and fatty acid metabolism, Pan PPAR agonists,
Virtual screening
3

Introduction
Type 2 diabetes mellitus (T2DM) is the most common form of diabetes mellitus  its clinical
presentation manifests as elevated blood glucose levels, due to insulin resistance and / or defects in
insulin secretion [1]. T2DM is a chronic metabolic disease which is usually associated with obesity,
dyslipidemia, and hypertension [2]. Patients with T2DM may suffer from loss of vision, strokes,
coronary artery disease, renal failure, and nerve damage [3]. The global prevalence of T2DM is
increasing dramatically, and it has been estimated that by 2030 it will affect over 360 million people [4].
Peroxisome proliferator-activated receptors (PPARs) belong to the nuclear receptor superfamily which
function as ligand-activated transcription factors and regulate the expression of specific genes [5].
These target genes are responsible for regulating adipogenesis, hepatic and adipose lipid metabolism,
reverse cholesterol transport, glucose homeostasis, insulin sensitivity, and vascular endothelial function
[6, 7]. The three PPAR isoforms, α, γ, and δ, are encoded by three distinct genes but share 80%
sequence similarities [8]. The isoforms have partially overlapping, but different, tissue distributions,
physiological functions, and pharmacological activators [9, 10]. Hence, the members of the PPAR
family are therapeutic targets for a specific cluster of risk factors collectively known as the metabolic
syndrome.
PPARα is predominantly expressed in tissues with high fatty acid oxidation, in particular the liver, but
also in macrophages, heart, muscle and kidneys [8, 11]. Agonists of PPARα, such as the fibrates, e.g.
fenofibrate 1, and WY-14643 2, (Figure 1), exploit the central role of PPARα in lipid and lipoprotein
synthesis, transport and metabolism, glucose homeostasis, and regulation of inflammation [12-14] to
manage dyslipidaemia. The main pharmacological benefit of PPARα agonists is a reduction in
4

cardiovascular diseases, predominantly through the elevation of high density lipoprotein cholesterol
(HDL-C) levels and the stimulation of fatty acid oxidation, thereby lowering triglyceride levels [15, 16].
These effects are primarily attributed to the ability of PPARα to promote the expression of ATP—
binding cassette transporters (ABCA1 and ABCG1), which increase cholesterol influx into
macrophages to drive reverse cholesterol transport [17-19]. In addition, PPARα upregulates several
hepatic enzymes that promote triacylglycerol degradation, which effectively reduces fatty acid
accumulation in liver and lowers triacylglycerol levels in circulating lipoproteins [20, 21].
Figure 1:
Chemical structures of selective PPARα agonists fenofibrate 1 and WY-14643 2,
selective PPARγ agonist rosiglitazone 3, and pan PPAR agonist bezafibrate 4.
PPARγ is mainly expressed in adipocytes where it plays a major role in adipogenesis and triglyceride
accumulation and storage [11-14, 22], however, it is also found in skeletal muscle, pancreatic β-cells,
macrophages, intestinal cells and the endothelium [11-13, 22]. PPARγ is also implicated in glucose
homeostasis, such as the enhancement of peripheral glucose uptake and utilisation [23]. The activation
of PPARγ improves insulin secretion and sensitivity [11-14, 22] and PPARγ agonists, such as the
5

thiazolidinediones, e.g rosiglitazone 3, (Figure 1), are of therapeutic importance in the treatment of
T2DM [14, 22].
Despite the clinical efficacy of selective PPARα and γ agonists in treating dyslipidemia [17, 24] and
T2DM [25-28] respectively, their clinical use is limited by their adverse effect profiles. In particular,
PPARγ agonists are associated with weight gain, fluid retention, oedema and myocardial infarctions
[14, 15, 25, 27, 29], while PPARα agonists are known to cause potentially fatal side effects, such as
rhabdomyolysis and myopathy [17, 24]. These adverse effects arise mainly from disparities in the
independent up-regulation of each isoform [14, 22]. Considering the beneficial effects of each PPAR
isoform, the concept of simultaneously activating both PPARα and γ subtypes with a single compound,
i.e. a dual PPARα/γ agonist, has emerged as a potential treatment for T2DM and metabolic syndromes.
It is believed that dual PPAR agonists which stimulate both PPARα and γ isoforms to a similar extent
would exhibit a more balanced biological profile and thus ameliorate the unwanted side effects [8, 14,
30].
In our previous study [31], we examined the dual PPARα and γ activity of five classes of flavonoids
(flavanones, flavones, isoflavones, chalcones, and their pyrazole derivatives) and identified four of the
isoflavones 5 - 8 (Figure 2) as the most active dual PPARα/γ agonists. Herein, we extended our study
on these four isoflavones 5 - 8, using ligand-based in silico screening technique with the aim of
identifying new isoflavone compounds with improved properties. We have also performed structure-
based drug discovery approaches to identify potential dual PPARα/γ agonistic lead compounds, which
are structurally different from fibrates and TZDs. The evaluation of potential lead compounds in cell-
based assays focused on the ability of the compounds to upregulate cholesterol transporters ABCA1
6

and ABCG1 in macrophages as well as PPARα/γ-dependent fatty acid metabolism in adipocytes and
hepatocytes.
Figure 2:
Chemical structures of the most active dual PPARα/γ isoflavone molecules 5 - 8
identified in previous study [31].
Methods and Materials
General
All starting materials for the chemical syntheses were commercially available (Sigma-Aldrich, Castle
Hill, NSW, Australia) and were used without further purification. All reactions were performed in
oven-dried glassware under a nitrogen atmosphere. Reactions were monitored by thin layer
chromatography on aluminum backed plates coated with silica gel (Merck, 60, F254) and visualised
under UV light at 254 nm. 1H and 13C NMR spectra were recorded at 400 MHz on a Varian Gemini 400
MR (Palo Alto, CA, USA). Coupling constants (J) are in Hertz (Hz), chemical shifts (δ) are expressed
in parts per million (ppm) and reported relative to residual solvent peaks. Melting points were obtained
on a Stuart Scientific SMP 10. Low resolution mass spectra were obtained on Thermo Scientific TSQ
Quantum Access Max LCMS/MS & TLX1 Turboflow Chromatography System. Infrared spectra were
recorded on Shimadzu FTIR-8400S.
7

Compounds 9, 10 and 13 were purchased from ChemBridge Online Chemical Store CA, USA
(www.hit2lead.com), while compounds 11, 12, and 17 were purchased from TimTec eChemStore, DE,
1
USA (www.echemstore.com). All compounds were characterised and checked for purity by H NMR
spectroscopy prior to biological testing. Fenofibrate, WY-14643, rosiglitazone and bezafibrate were
purchased from Sigma-Aldrich and were used without further purification (Castle Hill, NSW,
Australia). 2-Chloro-1-(indolin-1-yl)ethan-1-one 14 was synthesized according to reference [32].
Roswell Park Memorial Institute (RPMI) 1640 media, L-glutamine, penicillin, streptomycin and
phosphate-buffered saline (PBS) were purchased from Invitrogen, Life Technologies (Australia).
Dulbecco’s Modified Eagle Medium/F-12 (DMEM/F-12), low glucose DMEM, and trypsin were
purchased from Gibco, Life Technologies (Australia). Fetal calf serum (FCS), dimethyl sulfoxide
(DMSO), 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), geneticin, 2-
mercaptoethanol, 12-O-tetradecanoylphorbol-13-acetate and paraformaldehyde (PFA) were obtained
from Sigma-Aldrich (Australia). Mowiol and phorbol-12-myristate-13-acetate (PMA) were obtained
from Merck, Calbiochem (Australia). SDS-PAGE molecular weight markers were purchased from
Fermentas (Australia). Rabbit polyclonal antibodies against ABCA1 and ABCG1 were obtained from
Novus (Australia). Rabbit polyclonal anti-β-actin was purchased from BD Transduction Laboratories
(Australia). Mouse monoclonal antibody against Carnitine Palmitoyltransferase Iα (CPT1α) was
obtained from Abcam (Australia). Horseradish peroxidase labelled secondary antibodies were obtained
from Cell Signalling (Australia). Enhanced chemiluminescence reagents and [1-14C]-oleate were
purchased from Perkin-Elmer (Australia).
Molecular Modelling
Target protein preparation and receptor grid generation
8

The X-ray crystal structures of PPARα (1K7L) and PPARγ (1FM9) were retrieved from the RCSB
Protein Data Bank (http://www.rcsb.org/). These protein structures were prepared using protein
preparation and refinement protocols, directed by the Protein Preparation Wizard embedded in Maestro
v9.5 (Schrödinger, LLC, New York, USA). The process was carried out individually for each protein
by adding hydrogens, assigning bond orders, creating zero-order bonds to metals, generating disulfide
bonds, adding missing residues / closing loops via Prime v3.2, and deleting water molecules beyond 5
Å from heteroatoms. Next, the appropriate protein chains and the co-crystallised ligand were chosen,
the state penalties were calculated for the native ligand and the structure with the lowest state penalty
value was selected. The hydrogen bond network was optimised and the protein structure was
minimised to a RMSD (root mean square deviation) of 0.3 Å using the OPLS_2005 force field [33].
The Receptor Grid Generation tool in Glide v4.0 was used to characterise the binding site for docking
studies, using the structure of the co-crystallised ligand in each receptor as the centre of a 20 Å3
bounding box. A Coulomb-van der Waals scaling factor of 1.0 for receptor van der Waals radii was
applied to protein atoms with partial atomic charges less than 0.25 and rotation of receptor hydroxyl
and thiol groups were not allowed. For the ligands, a van der Waals scaling of 0.8 was applied [33].
Where multiple subunits existed, grid generation was only performed on a single subunit. For PPARα
(1K7L), chain A and its respective steroid receptor co-activator (chain B) were used.
Ligand preparation
The ligands from the natural product libraries (TimTec and SPECS, as available in the ZINC database,
University of California, San Francisco, www.zincdocking.org) were used directly without prior ligand
preparation, in the Glide high throughput virtual screening (HTVS) docking studies; ligand preparation
was only necessary for Glide extra precision (XP) docking studies. Ligand preparation involved
9

geometry minimisations with MacroModel v9.1 using the OPLS_2005 force field and water as the
solvent, followed by LigPrep v2.7 to generate all possible stereoisomers, tautomers, and ionisation
states at pH 7.0 ± 2.0. All generated structures were subjected to ConfGen v2.5 to generate alternative
low energy conformations using Standard ConfGen calculations in a Fast search strategy [33].
Structure-based drug discovery
Ligands from the TimTec and SPECS natural libraries were independently docked into the receptor
grids of PPARα and PPARγ with Glide v4.0 using the HTVS scoring function to estimate protein-
ligand binding energies. Hits ranked within the top 100 which overlapped in both receptor isoforms
were considered as dual PPARα/γ agonists and were subsequently prepared and re-docked into the two
receptors with Glide v4.0, using the XP scoring function to refine binding energy estimates [34, 35].
For all docking studies, ligands were docked with flexible states to allow for sampling of nitrogen
inversions and ring conformations, and ligands with more than 300 atoms or 50 rotatable bonds were
not docked.
The selection of potential lead compounds was based on the refined docking scores, simplicity of the
structure and ease of synthesis, commercial availability, and the suitability of physico-chemical
properties. Docking scores of < -10 kcal/mol were considered to indicate exceptional binding so hits
which demonstrated scores of < -10 kcal/mol were favoured as leads. In addition, hits which displayed
balanced docking scores (within 2 kcal/mol of each other) for both receptor isoforms were also
considered to be potential dual PPARα/γ agonists. Simple structures were preferred, as chirality is
known to play a crucial role in the activity of dual PPARα/γ agonists and chemically complex
compounds, with stereoisomers with potentially differing activities are undesirable [36, 37].
Compounds were also required to be easily prepared or commercially available so they could be readily
10

obtained for in vitro testing. Hits which satisfied these criteria were then evaluated for their physico-
chemical suitability as potential drugs using QikProp v3.5 and compounds that satisfied Lipinski’s
Rule of Five [38] were considered to be potential dual PPARα/γ agonist lead compounds and were
selected for in vitro studies.
Pharmacophore screening in ligand-based drug discovery
For the ligand-based approach to identify new lead compounds the chemical structures and biological
EC50 activities (half maximal effective concentration) of four known isoflavone dual PPARα/γ agonist
5 - 8 (Table 1) were obtained from a previous study [31] and used to generate a pharmacophore model
for dual PPARα/γ agonists using the Develop Common Pharmacophore Hypotheses tool in Phase v3.4
[39]. Briefly, this involved preparing the structures of the isoflavones to generate alternative ionisation
states and conformers at pH 7.0, followed by the generation of pharmacophore hypotheses (each with
five key functional groups). The pharmacophore model with the best survival score was used to screen
a Phase generated 3D database of the TimTec and SPECS natural libraries for matches using the
Advanced Pharmacophore Screening tool of Phase. The resultant hits were evaluated in a HTVS Glide
docking protocol. Hits with balanced HTVS scores for both PPARα and PPARγ isoforms were further
evaluated via XP protocols to identify potential dual PPARα/γ agonist lead compounds for in vitro
studies as described previously.
11

Table 1:
In vitro transactivation activity value (EC50) of the isoflavone derivatives identified from
our previous paper [31].
Transactivation EC50a (μM)
Compound
hPPARα
24.55
8.9
hPPARγ
18.86
26.94
15.38
22.29
5
6
7
8
23.1
33.13
aEC50 value is the concentration of compounds (μM) required for 50% fold activation of luciferase
activity.
Docking studies
Selected positive controls, fenofibrate 1, WY-14643 2, rosiglitazone 3 and bezafibrate 4, were built and
adjusted for chemical correctness in Maestro, minimised in MacroModel then, together with the
minimised isoflavone 5 - 8, derivatives prepared using LigPrep v2.7 to generate stereoisomers and to
assign protonation states at physiological pH. The conformers were generated for each ligand with
ConfGen v2.5 then docked into the ligand binding domain of the prepared PPAR structures. Glide with
extra precision (XP) was used to perform the docking study to estimate protein-ligand binding affinities.
Chemical Synthesis
N-{4-(Diethylamino)phenyl}-2-hydroxybenzamide 15
Salicylic acid (0.50 g, 3.62 mmol) was dissolved in excess thionyl chloride (10 mL) and the mixture
was heated at 110 °C for 3 hours to produce 2-hydroxybenzoyl chloride. A cold solution of 4-amino-
N,N-diethylaniline (0.6 mL, 3.62 mmol) in dichloromethane (10 mL) was then added dropwise into this
mixture, which was stirred overnight at room temperature. The volatiles were removed under vacuum
12

and the residues were taken up in water and extracted with DCM (3 × 10 mL). The combined organic
layers were then washed with NaHCO3 (15 mL) and brine (15 mL), and dried over anhydrous Na2SO4.
Filtration and reduction of the volatiles under vacuum gave the crude product which was purified by
column chromatography on silica, using EtOAc/hexane (30/70) as eluent, to give amide 15 as a dark
yellow powder (0.54 g, 50%). mp 101.8 – 104.7 °C. νmax (cm-1) 3329 (NH), 2970 (OH), 1519 (C=C).
1H NMR (400 MHz, CDCl3) δ 12.05 (1H, s, OH), 7.81 (1H, s, NH), 7.49 (1H, d, J = 8.4 Hz, CH-6),
7.39-7.43 (1H, m, CH-4), 7.35 (2H, d, J = 8.4 Hz, 2 × CHAr), 7.01 (1H, dd, J = 1.2 and 8.4 Hz, CH-3),
6.86-6.90 (1H, m, CH-5), 6.67 (2H, d, J = 8.4 Hz, 2 × CHAr), 3.35 (4H, q, J = 7.2 Hz, 2 × CH2), 1.15
13
(6H, t, J = 7.2 Hz, 2 × CH3). C NMR (400 MHz, CDCl3) δ 168.2 (C=O), 161.8 (quat., C-2), 145.9
(quat., -CArNEt2) 134.2 (CH, C-4), 125.3 (CH, C-6), 124.5 (quat., -CArNHC(O)), 123.7 (CH, 2 × CHAr),
118.8 (CH, C-3 or 5), 118.7 (CH, C-3 or 5), 114.7 (quat., C-1), 112.0 (CH, 2 × CHAr), 44.5 (2 × CH2),
12.5 (2 × CH3). MS (ESI) m/z 285.04 (100%) [M + H] +. Anal. Calcd for C17H20N2O2: C, 71.81; H,
7.09; N, 9.85. Found: C, 71.91; H, 7.15; N, 9.70.
N-{4-(Diethylamino)phenyl}-2-(2-{indolin-1-yl)-2-oxoethoxy}benzamide 16
2-Chloro-1-(indolin-1-yl)ethan-1-one 14 (0.25 g, 1.28 mmol), N-{4-(diethylamino)phenyl}-2-
hydroxybenzamide 15 (0.37 g, 1.28 mmol), potassium iodide (6.3 mg, 0.04 mmol) and potassium
carbonate (0.18 g, 1.28 mmol) were dissolved in acetone (50 mL) and the resulting mixture was heated
at 60 °C overnight. The reaction mixture was cooled down to room temperature and the volatiles were
removed under reduced pressure. The residues were taken up in water (30 mL) and extracted with
DCM (3 × 20 mL). The combined organic layers were then washed with NaHCO3 (20 mL), the brine
(20 mL), dried over anhydrous Na2SO4, filtered and concentrated. The product was recrystallized from
EtOAc to afford amide 16 as pale yellow crystals (0.3063 g, 54%). mp 197.7 – 200.7 °C. νmax (cm-1)
3317 (NH), 1670 (C=O), 1651 (C=O), 1516 (C=C). 1H NMR (400 MHz, CDCl3) δ 10.64 (1H, br, NH),
8.26 (2H, m, 2 × CHArindolin), 7.82 (2H, d, J = 7.2 Hz, 2 × CHAr), 7.39 (1H, t, J = 7.6 Hz), 7.17-7.25 (2H,
m), 7.11 (1H, t, J = 7.6 Hz, CH), 7.05 (1H, t, J = 7.6 Hz, CH), 6.90 (1H, d, J = 8.4 Hz), 6.70 (2H, d, J =
7.2 Hz), 4.84 (2H, s, -CH2OPh), 4.01 (2H, t, J = 8.0 Hz, CH2indolin), 3.33 (4H, q, J = 7.2 Hz, 2 × CH2, -
NCH2CH3), 3.24 (2H, t, J = 8.0 Hz, CH2indolin), 1.48 (6H, t, J = 7.2 Hz, 2 × CH3, -NCH2CH3) .13C NMR
(400 MHz, CDCl3) δ 164.6 (quat., C=O, -CH2C(O)), 162.5 (quat., C=O, benzamide), 155.3 (quat.),
144.8 (quat.), 142.3 (quat.), 132.7 (CH), 132.3 (CH), 130.6 (quat.), 128.2 (quat.), 127.8 (CH), 124.7
(CH), 124.5 (CH), 123.7 (quat.), 122.3 (CH), 122.2 (CH, 2 x CHAr), 117.14 (CH), 112.65 (CH, 2 ×
CHAr), 112.3 (CH), 66.85 (CH2, -CH2OPh), 46.2 (CH2, CH2indolin), 44.6 (2 × CH2, -NCH2CH3), 28.25
13

(CH2, CH2indolin), 12.6 (2 × CH3, -NCH2CH3). MS (ESI) m/z 444.06 (100%) [M + H] +. Anal. Calcd for
C27H29N3O3: C, 73.11; H, 6.59; N, 9.47. Found: C, 73.25; H, 6.62; N, 9.57.
Compounds 16 and 18 - 20 were characterised by single crystal X-ray diffraction, as described in the
supplementary information.
Cell culture
All cell lines were from the American Type Culture Collection (ATCC, Manassas, VA, USA). Human
embryonic kidney (HEK293) cells and hepatocarcinoma (HuH7) cells were cultured in DMEM/F-12,
THP-1 monocytes in RPMI-1640, supplemented with 10% FCS, L-glutamine (2 mM), penicillin (100
U/mL) and streptomycin (100 ug/mL). THP-1 monocytes were differentiated with 2 nM TPA for 24 h
before treatment with PPAR agonists. The mouse embryonic fibroblast 3T3-L1 cell line was
maintained in low glucose DMEM, 10% FCS, penicillin (100 U/mL) and streptomycin (100 ug/mL)
and then differentiated into adipocytes as described previously [40].
Cell proliferation assay
Cell proliferation was determined using a colorimetric (MTT) assay according to the manufacturer’s
instructions. HEK293 and THP-1 cells (2 × 103, in triplicate) were grown in media together with or
without 10-100 μM fenofibrate 1, WY14643 2, rosiglitazone 3, bezafibrate 4 and the potential dual
PPARα/γ agonists (9 – 13 and 16 – 20) for 48-72ꢀh. Cells incubated in 0.2% DMSO served as a
negative control. Cells were incubated with 20ꢀμl of MTT (5ꢀmg/ml) for 4 hours, the media was
removed and spectrophotometric absorbance of samples in DMSO was measured at 550 ꢀ nm
(Microplate Reader, Model 689, Bio-Rad, Hercules, CA, USA) with a reference filter of 655ꢀnm. All
experiments were performed in triplicate in at least three separate experiments.
14

Preparation of cellular extracts and Western blot analysis
Cells were incubated with 10-100 μM fenofibrate 1, WY14643 2, rosiglitazone 3, bezafibrate 4 and the
potential dual PPARα/γ agonists (9 – 13 and 16 – 20) for 24h. Then cells were harvested in lysis buffer
(20 mM Tris-HCl, pH 7.5, 2 mM EDTA, 100 mM NaCl, 5 mM MgCl2, 1% (v/v) Triton X-100, 5 mM
NaF, 10% (v/v) glycerol, 0.5% (v/v) 2-mercaptoethanol, 0.1 mM Na3VO4 and protease inhibitors) as
described [19]. After centrifugation at 13,000 rpm, the protein concentration of the cleared cell lysate
was determined. Cell lysates were separated by 12% SDS-PAGE gel and transferred to Immobilon-P
(Millipore). Proteins were detected using their specific primary antibodies, followed by HRP-
conjugated secondary antibodies and enhanced chemiluminescence detection. ImageJ software
(National institutes of Health, Bethesda, MD, USA) was utilised to quantify band intensities.
Oil Red O (ORO) staining
Differentiated 3T3-L1 adipocytes were grown on coverslips, incubated with and without rosiglitazone
3 (25 µM) or compound 16 (20 µM) overnight and fixed with 4% PFA. For ORO staining, cells were
washed with 60 % isopropanol, incubated with ORO (0.5% stock diluted 6/4 v/v in H2O) for 10 min,
then washed with water. Coverslips were mounted with Mowiol, and analysed the following day.
Images were taken using a Leica DMI300 B inverted fluorescent microscope [41].
Measurement of triacylglycerols
Differentiated 3T3-L1 adipocytes were incubated with and without rosiglitazone 3 (25 µM) or
compound 16 (20 µM) overnight and then lysed in ice-cold PBS. An aliquot was taken for protein
quantification by BCA (Thermo Fisher) while lipid extraction was performed on the remainder of the
lysate by the addition of 2:1 chloroform:methanol. The dried organic phase was re-suspended in 100%
15

ethanol and analysed using an enzymatic colorimetric method (GPO-PAP reagent, Roche Diagnostics)
[42, 43].
Fatty acid accumulation and oxidation
HuH7 cells were seeded in triplicate in 6 well plates and treated overnight with or without 20 µM WY-
14643 2 or compound 16. The following day, the medium was replaced with the assay medium,
containing 5 mM ‘cold’ oleate (prepared from stock fatty acid solutions of 100 mM oleate diluted in
100% ethanol), [1-14C]-oleate (0.5 μCi/ml) conjugated to 2% (w/v) fatty acid-free BSA in DMEM with
5 mM glucose and no antibiotics for 4 h. Mitochondrial oxidation was determined from 14CO2
production captured as previously described [42, 43]. Cells were harvested in 0.6 ml of ice-cold PBS to
14
determine C-oleate incorporation into intracellular triacylglycerol (esterification) and protein content.
Cellular lipids were extracted by the addition of 2:1 chloroform:methanol and separated by TLC using
14
heptane-isopropyl ether-acetic acid (60:40:3, v/v/v) as developing solvent and C activity in the TAG
band determined by scintillation counting [42, 43].
Statistical analysis
All the results are expressed as means
standard error of the mean (SEM). Student unpaired two-
tailed t-tests were performed on Microsoft Office Excel 2007 to identify statistical significance and p-
values less than 0.05 (p < 0.05) were considered to be statistically significant.
Results and Discussion
Structure-based screening results
Ligand Binding Domains (LBDs) of PPAR receptors
16

The ligand binding domains of PPARα and PPARγ was found to be very similar, comprising 12 helices
arranged in an antiparallel helix sandwich and 4 standard antiparallel β sheets [44]. The binding site is
a large Y-shaped cavity, composed of three parts; entrance (hydrophobic), arm I (predominantly polar),
and arm II (hydrophobic) [45]. PPAR ligands usually have a hydrophobic tail which interacts with the
entrance and arm II, and a polar head group which interacts with arm I [45]. Arm I includes helix 12
(the AF2 helix), which is known to be responsible for PPAR activation [46]. The polar head group
forms a network of hydrogen bonds with AF2 which generates a charge clamp, in turn reducing the
mobility of AF2 and regulating gene expression [46]. Besides these common features, there are also
some unique features for each receptor due to differences in LBD amino acid composition
(approximately 20%) [8]. PPARα is more hydrophobic than PPARγ [44].
Before the docking-based screening, the ligands bound in the crystal structures 1K7L (GW409544) and
1FM9 (farglitazar) were first docked into the corresponding receptors in order to examine the docking
accuracy. The published crystallographic binding pose of each ligand was successfully recreated,
indicating that the docking method was reliable. As a result of the HTVS protocol, six compounds (9 –
13 and 16) which were ranked in the top 100 hits for both receptor isoforms, were chosen for Glide XP
(extra precision) docking into the PPAR receptors. The structures of these potential dual PPARα/γ
agonists are illustrated in Figure 3, while the docking results are reported in Table 2.
17

Figure 3: Potential dual PPARα/γ lead compounds 9 – 13 and 16 - 20.
18

Ligand-based screening results
Surprisingly, in developing the pharmacophore hypothesis, PPARα and PPARγ were found to have the
same five hypotheses — ADRRR.1, AAARR.1, AADRR.1, AADRR.2, AARRR.1; A represents
hydrogen bond acceptor, D represents hydrogen bond donor, R represents aromatic ring. ADRRR.1
(Figure 4) was selected for virtual screening of the natural libraries, and another four compounds 17 -
20 were identified.
Figure 4:
Pharmacophore model of isoflavone pan agonists consisting of aromatic rings (from left
to right, A, B, and C), a hydrogen bond donor (blue) on ring A and a hydrogen bond acceptor (pink) on
ring B; A represents hydrogen bond acceptor, D represents hydrogen bond donor, R represents
aromatic ring. Isoflavone pan agonists used to generate the pharmacophore model are each in a
different colour.
19

Table 2: Docking scores and key interacting residues of compounds where docked into PPARα (1K7L)
and PPARγ (1FM9) receptors via extra precision (XP) docking.
Docking scorea
(kcal/mol)
Receptor
PPARα
(1K7L)
-8.4
Key interacting residuesb
Receptor
Compound
PPARγ
(1FM9)
-8.5
PPARα
(1K7L)
S280
PPARγ
(1FM9)
Fenofibrate 1
WY14643 2
F363, H449
-9.3
-11.1
S280, Y31,
Y464
H323, H449,
S289, Y473
Rosiglitazone 3 -9.8
-10.9
S280, Y314
H323, Y473
F363, L340
Bezafibrate 4
-10.55
-10.15
H440, S280,
Y314, Y464
-6.4
-7.5
-8.5
-8.2
-8.3
-10.0
-9.4
-9.3
H440
F360, F363,
H449, S289
5
F351, H440,
S280, Y314
H323, F363,
H449, S289
6
7
8
Y314
F363, H323,
H449, S289
E269, S280,
F363, H323,
20

H449, S289
F363, H449
F363, H449
-10.3
-10.3
-11.6
-9.2
S280, Y314
S280, Y314
H440, Y464
9
-9.8
10
11
-12.9
F282, F363,
H323, H449,
S289, Y473
-10.8
-12.7
H440, S280,
Y314, Y464
F363, H323,
H449, S289,
Y473
12
-8.65
-10.2
-10.2
-12.0
F363
13
16
C285, H449
-8.5
-7.9
-6.6
-8.6
S280
17
18
H440, Y464
F363, H449,
S289
-7.6
-7.6
-8.2
-8.2
H440, Y464
F363, H449
19
20
F273, F351,
H440
F363, H323,
H449
aDocking scores obtained from extra precision (XP) Glide docking as an estimate of the protein-ligand
binding energies. The more negative the score, the more favourable the protein-ligand binding
interaction.
bKey interacting residues were obtained from the 2D-Ligand-Interaction function of Maestro v9.3.
21

The docking scores obtained from Glide XP docking do not correlate directly with the number and type
of residue interactions. For example, compounds 1, 9, 10 and 19 all interact with PPARγ only through
π-π interactions with F363 and H449 but their PPARγ docking scores differ greatly (-8.5, -9.2, -9.8 and
-8.2 kcal/mol respectively) (Table 2). This difference can largely be attributed to the vigorous
calculation employed by Glide XP to score protein-ligand binding affinities. This calculation takes into
account residue independent factors, which may also favour ligand binding such as the displacement of
water molecules from the active site, desolvation effects and entropic effects as a result of ligand
binding. Therefore the Glide XP docking scores provide better estimates of in vitro efficacies and the
key interacting ligands should only be used as a general guide when developing structure-activity
relationships.
Synthesis
Compound 16 was synthesised following the strategy depicted in Scheme 1, involving a Williamson
ether coupling of phenol 15 and chloroacetylindoline 14. Compounds 18 and 19 were synthesised using
literature methods [47-50] and compound 20 was synthesised as previously described, with slight
modifications [51].
22

Scheme 1: Synthetic procedure for Compound 16. (i) SOCl2, 110 °C, 3h; then 4-diethylaminoaniline in
DCM; (ii) K2CO3/KI/acetone, 60 °C, overnight.
Cell proliferation assay
In order to address the cytotoxicity of all compounds, an MTT cell viability assay was performed in the
HEK293 cell line, with growth inhibitory effects investigated at 100 μM. Fenofibrate 1, WY-14643 2,
rosigiltazone 3 and bezafibrate 4 were used as controls, and DMSO (0.2 %) acted as vehicle control.
Compounds that led to reduced cell viability were further evaluated at 50 μM (Figure 5).
All compounds, with the exception of compound 12, exhibited no statistically significant effects on
HEK293 cell viability. Fenofibrate inhibited cell growth to 92.1%
1.4 and 53.6%
12.7 at 50 μM
and 100 μM, respectively, in agreement with previous studies which reported fenofibrate to have
greater impacts on cell growth than rosiglitazone, WY-14643 and bezafibrate [8, 22, 31, 52]. However,
the growth inhibition of fenofibrate, WY-14643, rosiglitazone and bezafibrate demonstrated in this
study (< 5% growth inhibition for bezafibrate, rosiglitazone and WY-14643 at 100 μM concentrations)
was less pronounced than has been previously reported, which may be attributed to differences in cell
lines and the assays employed [8, 22, 31, 52].
23

Figure 5: Effects of compounds on HEK293 cell viability. Cell viability is expressed as % cell growth
relative to DMSO 0.2% and all values are given as mean ± standard error of the mean (SEM) (n = 3), *
p < 0.05.
Biological evaluations
To address the potential of the compounds 9 – 13 and 16 – 20 as dual PPARα/γ agonists, a series of
biological evaluations were performed in macrophage, adipocyte, and hepatocyte cell models.
Transcriptional up-regulation of ABC transporters via nuclear receptors, including PPARα and PPARγ,
is well established [37, 59]. To determine if our compounds had PPAR agonist activity, we first
examined ABCA1 and ABCG1 expression in THP-1 macrophages. THP-1 monocytes were
differentiated with TPA for 24 h before treatment with 20 μM drugs for additional 24 h. Cell lysates
were prepared and ABCA1 and ABCG1 protein levels were determined by Western blot analysis. In
agreement with published data, rosiglitazone 3 and bezafibrate 4 induced ABCA1 and ABCG1
expression by 2-4 and upto 10-fold, respectively (Figure 6A). Similarly to the established pan-PPAR
and PPARγ agonists, compounds 13 and 16, but not 18 and 19, upregulated ABCA1 and ABCG1 by
approximately 3 to 7-fold in these experiments (Figure 6A). Compound 16 was selected for further
studies because of its’ pronounced ability to upregulate both ABCA1 and ABCG1.
24

Figure 6: Biological evaluations of compound 16. A. THP-1 monocytes were differentiated with 2 nM
TPA for 24 h before treatment with 20 μM of the compounds as indicated. Cell lysates were analysed
by Western blotting for ABCA1/G1 protein levels. ABCA1/G1 levels in each lysate were quantified
and normalised to actin. All values are given as mean ± SEM (n = 3). B. 3T3-L1 cells were
differentiated into adipocytes in the presence of rosiglitazone (25 μM) or compound 16 (20 μM) for 18
hours. Cells were fixed and neutral lipid accumulation was stained with ORO. C. 3T3-L1 adipocytes
were incubated as described in (B) and triacylglycerol concentrations in cell lysates were determined.
D. HuH7 cells were incubated overnight ± 20 μM WY-14643 2 and 20 μM compound 16. Cell lysates
were prepared and analysed by Western blotting for CPT1α levels, which were quantified and
normalised to actin. E. HuH7 hepatocytes were incubated with and without WY-14643 2 (20 μM) or
indoline 16 (20 μM) for 18 h before addition of 14C-oleate for 4h (see methods for details). The media
and cell lysates were collected and fatty acid oxidation (14CO2) in the media and cell-associated
radioactivity was determined. All values are given as mean ± SEM (n = 3).
25

We next examined PPARγ-dependent triacylglycerol accumulation during adipocyte differentiation [53,
54]. Therefore 3T3-L1 cells were differentiated into adipocytes, as described [40], in the presence of
rosiglitazone (25 μM) or, based on the findings above, compound 16 (20 μM) for 18 hours (Figure 6B-
C). Cells were fixed and neutral lipid accumulation was stained with ORO. Indeed, ORO staining was
substantially enhanced in the presence of rosiglitazone as well as compound 16, indicating increased
neutral lipid accumulation (Figure 6B). In support of this finding, triacylglycerol concentrations in cell
lysates from 3T3-L1 adipocytes were elevated, by approximately 20-30%, in the presence of the
PPARγ agonist rosiglitazone and compound 16 (Figure 6C), suggesting that the biological activity of
indoline 16 is comparable to established PPARγ agonists.
Next we analysed PPARα-dependent fatty acid metabolism in HuH7 hepatocytes. We first examined
expression levels of the PPARα-regulated gene CPT1α, which is involved in fatty acid oxidation [55,
56]. To compare the ability of fibrates and the indoline 16 to induce CPT1α expression, HuH7 cells
were incubated overnight ± 20 μM WY-14643 2 and 20 μM compound 16. CPT1α protein levels from
whole cell lysates were compared by Western blotting (Figure 6D) and shown to be upregulated more
than 2-fold in the presence of WY-14643 2 and compound 16. To further compare the activity of
indoline 16 to other PPARα agonists and enhance fatty acid oxidation, HuH7 hepatocytes were
incubated with and without WY-14643 2 (20 μM) or indoline 16 (20 μM) for 18 h. The medium was
14
then removed and a medium containing C-oleate was added and incubated for additional 4 h. Fatty
14
acid oxidation levels were measured by determining the amount of CO2 present in the media and in
cell lysates (Figure 6E). Indoline 16 substantially increased CO2 production, which was associated with
reduced incorporation of 14C-oleate into intracellular complex lipids, and is comparable to the effects of
WY-14643 2.
26

Conclusions
In summary, the combination of structure-based and ligand-based virtual screening of natural product
libraries identified ten potential dual PPARα/γ agonist lead compounds which are structurally different
to the current selective PPARα and PPARγ agonists. When tested next to established PPARα, PPARγ
and Pan-PPAR agonists, these lead compounds showed no statistically significant cell-toxicity, while
experiments performed in macrophage, adipocyte and hepatocyte cell models suggest that compound
16 has comparable biological activity in respect to cholesterol and fatty acid metabolism. Indeed,
compound 16 increased ABCA1 and ABCG1 expression in THP-1 macrophages (Figure 6A), elevated
CPT1α expression and fatty acid oxidation in HuH7 hepatocytes (Figure 6D-E), but also caused
increased lipid accumulation in 3T3-L1 adipocytes (Figure 6B). However, further investigations are
required in order to establish the effects of these compounds on cholesterol and fatty acid metabolism
in vivo. Additionally, any adverse metabolic side effects (which so far have limited the clinical use of
PPAR agonists) need to be determined to identify the therapeutic suitability of PPARα/γ agonists.
Acknowledgements
DEH, PWG and TG thank The University of Sydney, Australia (U1758, U7007), Faculty of Pharmacy,
The
University
of
Sydney,
Australia
and
NHMRC
for
funding.
27

References
[1] J.M. Chehade, A.D. Mooradian, A rational approach to drug therapy of type 2 diabetes mellitus,
Drugs 60 (2000) 95-113, http://dx.doi.org/10.2165/00003495-200060010-00006.
[2] P. Makadia, S.R. Shah, H. Pingali, P. Zaware, D. Patel, S. Pola, B. Thube, P. Priyadarshini, D.
Suthar, M. Shah, S. Giri, C. Trivedi, M. Jain, P. Patel, R. Bahekar, Effect of structurally constrained
oxime-ether linker on PPAR subtype selectivity: Discovery of a novel and potent series of PPAR-pan
agonists, Bioorg. Med. Chem. 19 (2011) 771-782, http://dx.doi.org/10.1016/j.bmc.2010.12.023.
[3] P.A. Carpino, B. Goodwin, Diabetes area participation analysis: a review of companies and targets
described in the 2008 - 2010 patent literature, Expert. Opin. Ther. Pat. 20 (2010) 1627-1651,
http://dx.doi.org/10.1517/13543776.2010.533171.
[4] S. Wild, G. Roglic, A. Green, R. Sicree, H. King, Global prevalence of diabetes: estimates for the
year 2000 and projections for 2030, Diabetes care 27 (2004) 1047-1053, http://dx.doi.org/
10.2337/diacare.27.5.1047.
[5] A. Rubenstrunk, R. Hanf, D.W. Hum, J.-C. Fruchart, B. Staels, Safety issues and prospects for
future generations of PPAR modulators, Biochim. Biophys. Acta. 1771 (2007) 1065-1081,
http://dx.doi.org/10.1016/j.bbalip.2007.02.003.
[6] B. Pourcet, J. Fruchart, B. Staels, C. Glineur, Selective PPAR modulators, dual and pan PPAR
agonists: multimodal drugs for the treatment of Type 2 diabetes and atherosclerosis, Expert. Opin.
Emerg. Drugs. 11 (2006) 379-379, http://dx.doi.org/10.1517/14728214.11.3.379.
[7] F. Chang, L.A. Jaber, H.D. Berlie, M.B. O'Connell, Evolution of peroxisome proliferator-activated
receptor agonists, Ann. Pharmacother. 41 (2007) 973-983, http://dx.doi.org/10.1345/aph.1K013.
[8] A. Matin, M.R. Doddareddy, N. Gavande, S. Nammi, P.W. Groundwater, R.H. Roubin, D.E. Hibbs,
The discovery of novel isoflavone pan peroxisome proliferator-activated receptor agonists, Bioorg.
Med. Chem. 21 (2013) 766, http://dx.doi.org/10.1016/j.bmc.2012.11.040.
28

[9] N. Wang, R. Yin, Y. Liu, G. Mao, F. Xi, Role of peroxisome proliferator-activated receptor-γ in
atherosclerosis: an update, Circ. J. 75 (2011) 528.
[10] B. Desvergne, W. Wahli, Peroxisome proliferator-activated receptors: nuclear control of
metabolism, Endocr. Rev. 20 (1999) 649-688, http://dx.doi.org/10.1210/er.20.5.649.
[11] H. Yki-Jarvinen, Thiazolidinediones, N. Engl. J. Med. 351 (2004) 1106-1118,
http://dx.doi.org/10.1056/NEJMra041001.
[12] L. Liu, Y. Ma, R.L. Wang, W.R. Xu, S.Q. Wang, K.C. Chou, Find novel dual-agonist drugs for
treating type 2 diabetes by means of cheminformatics, Drug. Des. Devel. Ther. 7 (2013) 279-288,
http://dx.doi.org/10.2147/DDDT.S42113.
[13] Y. Ma, S.Q. Wang, W.R. Xu, R.L. Wang, K.C. Chou, Design novel dual agonists for treating type-
2 diabetes by targeting peroxisome proliferator-activated receptors with core hopping approach, PLoS
One 7 (2012) e38546.
[14] A. Tenenbaum, E.Z. Fisman, Balanced pan-PPAR activator bezafibrate in combination with statin:
comprehensive lipids control and diabetes prevention, Cardiovasc. Diabetol. 11 (2012) 140,
http://dx.doi.org/10.1186/1475-2840-11-140.
[15] A. Rubenstrunk, R. Hanf, D.W. Hum, J.C. Fruchart, B. Staels, Safety issues and prospects for
future generations of PPAR modulators, Biochim. Biophys. Acta. 1771 (2007) 1065-1081,
http://dx.doi.org/10.1016/j.bbalip.2007.02.003.
[16] K. Kuwabara, K. Murakami, M. Todo, T. Aoki, T. Asaki, M. Murai, J. Yano, A novel selective
peroxisome proliferator-activated receptor alpha agonist, 2-Methyl-c-5-[4-[5-methyl-2-(4-
methylphenyl)-4-oxazolyl]butyl]-1,3-dioxane-r-2-carboxylic acid (NS-220), potently decreases plasma
triglyceride and glucose levels and modifies lipoprotein profiles in KK-Ay Mice, J. Pharmacol. Exp.
Ther. 309 (2004) 970-977.
29