Key words
Echinacea purpurea
- Asteraceae - alkamides - adipocytes - glucose uptake - PPAR
γ partial agonist
Abbreviations
aP2:
adipocyte protein 2
C/EBPα
:
CCAAT/enhancer-binding protein α
CS:
calf serum
DCM:
dichloromethane
DI:
dexamethasone and insulin
DMEM:
Dulbeccoʼs modified Eagleʼs medium
FCS:
fetal calf serum
Glut:
glucose transporter
GST:
glutathionine-S-transferase
GU:
glucose uptake
LBD:
ligand binding domain
MDI:
1-methyl-3-isobutylxanthine, dexamethasone and insulin
PBS:
phosphate-buffered saline
PPAR:
peroxisome proliferator-activated receptor
qPCR:
real-time quantitative polymerase chain reaction
Rosi:
rosiglitazone
SCD1:
stearoyl-coenzyme A desaturase 1
T2D:
type 2 diabetes
TFA:
trifluoroacetic acid
TR-FRET:
time-resolved fluorescence resonance energy transfer
TZD:
thiazolidinedione
Introduction
Echinacea purpurea (L.) Moench. (Asteraceae) is commonly used in traditional medicine for the treatment
and prevention of upper respiratory tract infections and the common cold. Commercial
medicinal preparations of E. purpurea have been shown to possess immunomodulatory and anti-inflammatory effects which explain
the traditional use of the plant [1], [2]. Furthermore, we have previously shown that lipophilic extracts of E. purpurea activate PPARγ and enhance insulin-stimulated GU in adipocytes in vitro in a dose-dependent manner [3], [4] and thus may have an effect on insulin resistance and T2D.
Adipose tissue plays a key role in energy homeostasis as a metabolic and endocrine
organ. Adipocytes are targets for drugs reducing obesity or ameliorating obesity-associated
metabolic disorders such as T2D [5]. TZDs have been used as insulin sensitizers in the management of T2D. TZDs regulate
the expression of genes involved in insulin signaling and glucose and lipid metabolism
in mature adipocytes but also promote adipocyte differentiation by acting as PPARγ full agonists setting into motion gene cascades promoting differentiation and accumulation
of triglycerides in adipocytes accentuating adiposity [6]. Furthermore, the potent TZDs cause undesirable side effects, including edema, fluid
retention, weight gain, cardiac failure, and hepatotoxicity. Due to the observed hepatotoxicity,
two TZD drugs were recently withdrawn from the market and the remaining one, pioglitazone,
has been reassessed in light of increased risk of bladder cancer [7]. Limitations of the TZDs have driven research towards new types of drugs, which
are able to increase insulin sensitivity acting as
PPARγ partial agonists, and therefore possibly ameliorate obesity-related insulin resistance
without causing the side effects observed in relation to many, if not all, full PPARγ agonists.
A number of natural plant-derived compounds have been shown to improve insulin sensitivity
and may thus serve as starting point for discovery and design of novel drugs and herbal
medicine to treat and prevent metabolic disorders with fewer side effects compared
to synthetic drugs [8]. The bioactive metabolites considered responsible for the pharmacological activities
of E. purpurea comprise polysaccharides, caffeic acid derivatives and alkamides [1], [4], [9], [10], [11]. From the florets of E. purpurea, a C16-alkamide has been isolated that significantly activated PPARγ and stimulated GU in adipocytes in vitro
[4]. The roots of E. purpurea have a relative high content of alkamides [12], [13], [14], of which some are known for their immunomodulatory and anti-inflammatory effects
[1], [2]. However, very little is known about these bioactive compounds in relation to their
effects on GU and PPARγ. In the present study, it is shown that a crude DCM extract of E. purpurea roots was able to enhance GU in 3T3-L1 adipocytes and to activate PPARγ. Hence, we decided to identify the metabolites responsible for the potential antidiabetic
effect of this lipophilic extract using a bioassay-guided fractionation approach.
Basal and insulin-dependent GU in adipocytes were used to assess the bioactivity of
fractions and isolated compounds. In addition, we determined the PPARγ activating properties of active fractions and isolated metabolites in order to determine
how their effect on GU in adipocytes correlated with their PPARγ activating properties.
Results and Discussion
We have previously screened 133 plant extracts in different assays including PPARγ transactivation, adipocyte differentiation, and insulin-dependent GU [3]. In this former investigation, the crude DCM root extract of E. purpurea was found to activate PPARγ and to increase insulin-dependent GU, while it did not promote adipogenesis. The
present study stays in agreement with these data ([Fig. 1]; Fig. 1S, Supporting Information). The results revealed that basal GU in cells treated with
the DCM extract was increased approximately 2-fold compared to control and was enhanced
in the presence of 3 and 10 nM of insulin, suggesting that the extract increased insulin
sensitivity in 3T3-L1 adipocytes ([Fig. 1 A]). Furthermore, the DCM extract was shown to activate PPARγ in transfected cell cultures in a dose-dependent manner ([Fig. 1 B]), acting as a weak PPARγ agonist compared to the full agonist Rosi, and did not inhibit MDI-induced adipocyte
differentiation or promote adipocyte differentiation of DI-treated 3T3-L1
preadipocytes (Fig. 1S).
Fig. 1 Effects of dichloromethane extract of Echinacea purpurea roots and active fractions A and D on: A Glucose uptake in adipocytes at 100 µg/mL with insulin concentrations of 0, 3, and
10 nM. B PPARγ transactivation at 1, 10, and 100 µg/mL. DMSO (vehicle) was set to 1 and the results
normalized to this, while Rosi (1 µM) was the positive control. In the PPARγ transactivation assay, Rosi was 37-fold relative to DMSO. All values are expressed
as mean ± SD of three independent experiments in triplicates; * p < 0.001 indicates
significance relative to DMSO. (Color figure available online only.)
A bioassay-guided approach was used to identify active metabolites from the crude
DCM extract with interesting properties in relation to GU. Initial separation of the
extract by flash chromatography resulted in nine fractions (A–I), of which fractions
A and D were found to increase insulin-dependent GU in the presence of 3 nM and 10 nM
insulin ([Fig. 1 A]). The observed effect of the active fractions A and D on GU could be due to the
presence of natural products with PPARγ activity, and they were therefore tested for this activity. Fractions A and D were
found to activate PPARγ acting as weak agonists ([Fig. 1 B]). However, the lack of correlation between insulin-dependent GU and PPARγ activity ([Fig. 1]) suggests that the effect on GU uptake, at least partially, may be due to compounds
acting as partial agonists. In addition, fractions A and D did not promote adipogenesis
using the DI protocol (Fig. 1S, Supporting Information).
HPLC-PDA-APCI-MS/MS analysis revealed that fraction A contained fatty acids with α-linolenic acid being a major constituent, while fraction D contained alkamides. α-Linolenic acid is a well-known PPARγ agonist [4], [15] and has previously been shown to improve glucose tolerance, insulin sensitivity,
dyslipidemia, hypertension, and ventricular stiffness in rats [16], [17]; thus fraction A was not investigated further. Separation of fraction D by semi-preparative
HPLC yielded four alkamides ([Fig. 2 A]), of which two are new isomeric C12-alkamides (1, 2). The known alkamides dodeca-2E,4E,8Z-trienoic acid isobutylamide (3) and dodeca-2E,4E-dienoic acid isobutylamide (4) have previously been isolated from the roots of E. purpurea
[12], [14], [18].
Fig. 2 A Chemical structures of alkamides (compounds 1–4) isolated from an active fraction (fraction D) of a dichloromethane root extract
of Echinacea purpurea. Compounds 1 and 2 were isolated as an inseparable mixture (named compounds 1/2). B Predicted binding modes of compounds 1/2 shown as a 3D representation. C 2D representation of the chemical features of the interaction pattern derived from
the docking pose for compound 1. The interaction pattern is the same for compound 2. Chemical features in the 2D representation are: red arrow = hydrogen-bond donor;
blue line = hydrophobic contacts. (Color figure available online only.)
Compounds 1 and 2 were obtained as an inseparable 1 : 1 mixture (compounds 1/2) that gave a quasi-molecular precursor ion at m/z 262 [M + H]+. The molecular formula C17H27NO (5 degrees of unsaturation) was assigned based on HR-ESI-MS data. The MS/MS fragmentation
pattern of the quasi-molecular precursor ion of compounds 1/2 showed characteristic product ions for 2-methylbutylamides ([Fig. 3]) [13], [14], [19], [20], [21]. The presence of a 2-methylbutylamine moiety was confirmed by the 1H NMR signals at δ 0.90 (d, J = 7 Hz, H3-5′), 0.90 (t, J = 7 Hz, H3-4′), 1.17 (m, H-3′b), 1.41 (m, H-3′a), 1.58 (m, H-2′), 3.16 (m, H-1′b), 3.28 (m,
H-1′a), and 5.43 (brs, NH) [12], and was further substantiated by the 1H–1H COSY spectrum ([Table 1]). The UV spectrum showed absorption maxima at 235 and
261 nm indicating a 2,4-diene moiety and an additional diene chromophore in the structure
of 1 and 2
[12], [21]. The relative intensities of the fragments at m/z 149 (100 %, base peak) and m/z 147 (24 %) of the MS/MS fragmentation pattern were in agreement with compounds 1/2 being 2,4-diene alkamides [13], [14], [20]. The presence of four carbon-carbon double bonds in 1 and 2 was evident from the 1H NMR showing signals from 2 × 8 olefinic protons ([Table 1]). The low-field shift of the methine proton at C-3 (δ 7.18) confirmed that one of the double bonds was attached next to the amide moiety
[12], [22]. The positions of the other double bonds were determined from the 1H–1H COSY spectrum thus establishing the four double bonds in positions 2, 4, 8, and
10, respectively ([Table 1]). The double bonds in positions 2 and 4 were assigned an E configuration on the basis of a coupling constant of 15 Hz, and the double bond in
position 8 was assigned a Z configuration due to a coupling constant of 11 Hz [12]. Similarly, the double bond in position 10 was assigned as being E or Z ([Table 1]). Hence, the chemical structure of compounds 1 and 2 was established as dodeca-2E,4E,8Z,10E-tetraenoic acid 2-methylbutylamide and dodeca-2E,4E,8Z,10Z-tetraenoic acid 2-methylbutylamide, respectively ([Fig. 2 A]).
Fig. 3 Proposed fragmentation pathway of the protonated molecular ion (MH+) of dodeca-2E,4E,8Z,10E-tetraenoic acid 2-methylbutylamide (1) as determined by MS/MS, including chemical structures of some of the most characteristic
fragments. The base peak fragment at m/z 149 (not shown) is probably formed when a double bond of the original 2,4-diene shifts
to the 3-position, with a subsequent gain of two hydrogens [13], [14], [20]. The fragmentation pathway of the 2E,4E,8Z,10Z-isomer (2) is the same as illustrated above.
Table 1 1H NMR and 1H–1H COSY spectral data (CDCl3, 400 MHz) of dodeca-2E,4E,8Z,10E-tetraenoic acid 2-methylbutylamide (1) and dodeca-2E,4E,8Z,10Z-tetraenoic acid 2-methylbutylamide (2).
H
|
Compound 1
|
Compound 2
|
1H−1H COSY correlations
|
δ
1H (ppm) (J in Hz)
|
δ
1H (ppm) (J in Hz)
|
2
|
5.76 d (15 Hz)
|
5.76 d (15 Hz)
|
H-3
|
3
|
7.18 dd (15; 11 Hz)
|
7.18 dd (15; 11 Hz)
|
H-2, H-4
|
4
|
6.17 dd (15; 11 Hz)
|
6.17 dd (15; 11 Hz)
|
H-3, H-5
|
5
|
6.08 dt (15; 7 Hz)
|
6.08 dt (15; 7 Hz)
|
H-4, H2-6
|
6
|
2.28 m
|
2.28 m
|
H-5, H2-7
|
7
|
2.28 m
|
2.28 m
|
H2-6, H-8
|
8
|
5.25 m
|
5.36 m
|
H2-7, H-9
|
9
|
5.97 dd (11; 11 Hz)
|
6.30 dd (11; 11 Hz)
|
H-8, H-10
|
10
|
6.25 dd (15; 11 Hz)
|
6.32 dd (11; 11 Hz)
|
H-9, H-11
|
11
|
5.69 dq (15; 7 Hz)
|
5.55 dq (11; 7 Hz)
|
H-10, H3-12
|
12
|
1.78 brd (7 Hz)
|
1.75 brd (7 Hz)
|
H-11
|
NH
|
5.43 brs
|
5.43 brs
|
H2-1′
|
1′a
|
3.28 m
|
3.28 m
|
H-2′, NH
|
1′b
|
3.16 m
|
3.16 m
|
H-2′, NH
|
2′
|
1.58 m
|
1.58 m
|
H2-1′, H2-3′, H3-5′
|
3′a
|
1.41 m
|
1.41 m
|
H-2′, H3-4′
|
3′b
|
1.17 m
|
1.17 m
|
H-2′, H3-4′
|
4′
|
0.90 t (7 Hz)
|
0.90 t (7 Hz)
|
H2-3′
|
5′
|
0.90 d (7 Hz)
|
0.90 d (7 Hz)
|
H-2′
|
Compounds 3 and 4 showed a weak increase of basal GU and insulin-dependet GU at 3 nM insulin in mature
3T3-L1 cells at 30 µM, whereas no increase of insulin-dependent GU was observed at
10, 30 and 100 nM insulin (Fig. 2S, Supporting Information) despite their ability to weakly activate PPARγ, which is, for compound 4, in accordance with previous investigations [4]. Dodeca-2E,4E,8Z,10E/Z-tetraenoic acid isobutylamides are among the major alkamides in E. purpurea but have previously been shown not to activate PPARγ
[4]. This may also explain why these alkamides were detected in the non-active fraction
F according to the GU bioassay (data not shown). Compounds 1/2 at 30 µM significantly increased basal GU 6.3-fold compared to the vehicle and 3-fold
compared to Rosi ([Fig. 4 A]). A significant stimulation of insulin-dependent GU in the presence of 3, 10, 30
and 100 nM insulin was also observed in cells treated with 30 µM of compounds 1/2 ([Fig. 4 A]). Compounds 1/2 also dose-dependently increased GU in the presence of 10 nM insulin ([Fig. 4 B]). The effects of compounds 1/2 on GU resemble those of hexadeca-2E,9Z,12Z,14E-tetraenoic acid isobutylamide, previously isolated from the flowers of E. purpurea
[4]. Similar to hexadeca-2E,9Z,12Z,14E-tetraenoic acid isobutylamide, compounds 1/2 increased transactivation of PPARγ significantly at a concentration of 30 µM compared to vehicle but still weak relative
to Rosi ([Fig. 5 A]). Furthermore, compounds 1/2 and Rosi both competed with a fluorescent pan PPAR agonist in a ligand binding assay
indicating a common binding site in the PPARγ LBD ([Fig. 5 B]) [23]. These results indicate that compounds 1/2 may function as PPARγ partial agonists. The docking mode and interactions of compounds 1/2 with the LBD ([Fig. 2 B, C]) support the assumption that compounds 1/2 are PPARγ partial agonists. The PPARγ LBD contains a large Y-shaped ligand binding cavity consisting of an entrance (arm
III) that branches off into two binding pockets (arm I and II). Arm I is the only
substantially polar cavity of the PPARγ LBD, whereas arms II and III are mainly
hydrophobic [24], [25]. The predicted binding modes of compounds 1/2 with the PPARγ LBD included one hydrogen bond with Ser342 and several hydrophobic contacts with
Ala292, Ile281, Ile296, Ile326, Ile341, Leu333, and Met348 from arms II and III of
the LBD of PPARγ ([Fig. 2 C]). These contacts are typical for PPARγ partial agonists [24]. The fact that dodeca-2E,4E,8Z,10E/Z-tetraenoic acid isobutylamides do not seem to activate PPARγ suggests that the hydrophobic contacts to Ile341 and Met348 are important for compounds
1/2 being PPARγ partial agonists. In addition, no hydrogen bond interaction between 1/2 and residues Ser289, His323, His449, and/or Tyr473 from arm I of the LBD of PPARγ was predicted, which are typical for PPARγ full agonists. It has been a general notion that the binding mode of PPARγ partial agonists affects the recruitment of co-activators differently from that of
full agonists and decreases the transactivation
activity of PPARγ resulting in fewer side effects, but still maintains an insulin sensitizing effect
[26]. In the present study, the weak activation of PPARγ by compounds 1/2, combined with the results of the competitive PPARγ binding assay and the docking mode of 1/2, suggests that these compounds are acting as PPARγ partial agonists and that the 2-methylbutylamide moiety is important for their activity.
Fig. 4 Effect of compounds 1/2 on insulin-dependent glucose uptake: A At 30 µM of compounds 1/2 with insulin concentrations of 0, 3, 10, 30 and 100 nM. B Dose-dependent effect of compounds 1/2 at 10 nM insulin. DMSO (vehicle) was set to 1 and the results normalized to this,
while Rosi (1 µM) was the positive control. All values are expressed as mean ± SD
of three independent experiments in triplicates; * p < 0.001 indicates significance
relative to DMSO. (Color figure available online only.)
Fig. 5 Effect of compounds 1/2 on peroxisome proliferator-activated receptor γ: A Transactivation of PPARγ at a concentration of 30 µM. DMSO (vehicle) was set to 1 and the results normalized
to this. Rosi (1 µM) was the positive control. All values are expressed as mean ± SD
of three independent experiments in triplicates; * p < 0.001 indicates significance
relative to DMSO. B Effect of Rosi (0.001–3 µM) and compounds 1/2 (0.1, 0.3, 1, 3, 10, and 30 µM) in LanthaScreen™ TR-FRET competitive PPARγ binding
assay in the agonist mode. Relative fluorescence units (RFU) are expressed as mean ± SD
of three independent experiments in triplicates. An unrestrained sigmoidal (one-binding
site) dose-response curve was fitted to each data set by linear regression. (Color
figure available online only.)
The alkamides are structurally related to natural endogenous and/or dietary PPARγ ligands such as fatty acids [27], [28], prostanoids [28], [29], phospholipids [30], and polyacetylenes [31]. The PPARγ activating properties of these ligands depend on the length of their aliphatic chain
and functional groups such as carboxylate/carboxyl, α,β-unsaturated ketone, and/or hydroxyl, and thus on their interactions with the LBD.
For example, the PPARγ activating properties of some endogenous prostanoids and oxidized polyunsaturated
fatty acids seem to be related to their ability to bind covalently to Cys285 of PPARγ LBD through a Michael addition reaction with an α,β-unsaturated ketone [32], [33], [34]. Polyacetylenes of the falcarindiol-type containing hydroxyl groups at one or two
positions along their C17-aliphatic chain were also shown to display PPARγ
partial agonism, which was confirmed by molecular docking studies suggesting a binding
to the LBD in a manner similar to that observed for compounds 1/2 and other PPARγ partial agonists [31]. However, the large cavity of the PPARγ LBD, and thus the broad specificity of this receptor, makes the prediction of PPARγ activity of ligands such as alkamides and other fatty acid derivatives difficult
based on their chemical structure alone, in accordance with the results of the present
study.
Increased expression of adipose cell differentiation markers is associated with insulin
sensitivity [35], which may to some extent explain the insulin-dependent and basal GU data obtained
in this study. The critical window for ligand-dependent induction of adipocyte differentiation
of 3T3-L1 preadipocytes is days 0 to 4 after induction of differentiation [36]. To investigate possible mechanisms by which compounds 1/2 might affect the early stages of adipocyte differentiation, we examined the expression
of genes involved in adipogenesis (PPARγ, C/EBPα, and aP2, also known as fatty acid binding protein 4), glucose transport (Glut1 and
Glut4), lipogenesis (SCD1), and adipokine (adiponectin) ([Table 2]) in mature 3T3-L1 cells treated with 30 µM of compounds 1/2 during the first 4 days of differentiation (days 0–4) using the DI and MDI protocols.
Compounds 1/2 increased the expression of PPARγ, aP2, C/EBPα, adiponectin and Glut1 in the MDI differentiation medium ([Fig. 6 A]). However, when
the cAMP elevating agent 1-methyl-3-isobutylxanthine was omitted in the cocktail (DI
differentiation medium), the expression of several genes was significantly downregulated
([Fig. 6 B]). These results suggest that compounds 1/2 require elevated cAMP levels to increase the expression of key markers of adipogenesis.
SCD1 expression was significantly downregulated by treatment with 30 µM of compounds
1/2 in both the MDI and DI treatment ([Fig. 6]). SCD1 is involved in lipogenesis, and therefore a downregulation of this gene may
result in decreased lipogenesis and a small size of lipid droplets [37].
Fig. 6 Effect of compounds 1/2 in a concentration of 30 µM on gene expression of peroxisome proliferator-activated
receptor γ, adipocyte protein 2, CCAAT/enhancer-binding protein α, stearoyl-coenzyme A desaturase 1, adiponectin, glucose transporters 1 and 4 in differentiating
3T3-L1 cells in the: A MDI protocol. B DI protocol. DMSO (vehicle) was set to 1 and the results normalized to this, while
Rosi (1 µM) was the positive control. All values are expressed as mean ± SD of three
independent experiments in triplicates; * p < 0.001 indicates significance relative
to DMSO. (Color figure available online only.)
Table 2 Primer sequences used for analysis of gene expression.
Genes
|
5′-prime (forward)
|
3′-prime (reverse)
|
Adipogenesis process related
|
|
|
C/EBPα
|
CAAGAACAGCAACGAGTACCG
|
GTCACTGGTCAACTCCAGCAC
|
aP2
|
CTGGGCGTGGAATTCGAT
|
GCTCTTCACCTTCCTGTCGTCT
|
PPARγ
|
ACAGCAAATCTCTGTTTTATGC
|
TGCTGGAGAAATCAACTGTGG
|
Adipokine gene
|
|
|
Adiponectin
|
GATGGCAGAGATGGCACTCC
|
CTTGCCAGTGCTGCCGTCAT
|
Glucose transporter
|
|
|
Glut1
|
TGTATCCTGTTGCCCTTCTGC
|
CGACCCTCTTCTTTCATCTCCT
|
Glut4
|
CAGAAGGTGATTGAACAGAGC
|
CCCTGATGTTAGCCCTGAG
|
Lipogenesis related
|
|
|
SCD1
|
ACACCTGCCTCTTCGGGATT
|
TGATGGCCAGAGCGCTG
|
Reference gene
|
|
|
TATA box-binding protein
|
ACCCTTCACCAATGACTCCTATG
|
ATGATGACTGCAGCAAATCGC
|
PPARγ agonists are known to increase glucose transport and insulin sensitivity by regulating
expression of several genes involved in glucose metabolism [38]. GU in adipocytes is correlated with the number of glucose transporters [39]. Glut1 is expressed in both preadipocytes and mature adipocytes and is responsible
for basal glucose transport. By contrast, Glut4 is expressed only in mature adipocytes
and supports insulin-dependent GU [40], [41]. In this study, compounds 1/2 increased gene expression of Glut1, but not Glut4, when the MDI differentiation medium
was used ([Fig. 6 A]), whereas compounds 1/2 decreased the expression of the Glut4 transporter and did not significantly affect
the expression of the Glut1 transporter in the DI differentiation medium ([Fig. 6 B]). This could indicate that compounds 1/2 in the context of MDI-induced differentiation enhanced basal GU by upregulation of
Glut1 gene expression, whereas the observed
insulin-dependent GU requiring Glut4 was unrelated to an increased expression of this
transporter.
Adiponectin also plays an important role in mediating GU in adipocytes and was significantly
upregulated by addition of compounds 1/2 in the MDI protocol ([Fig. 6 A]); hence, this suggests that the adipocytes might be insulin sensitive [42]. This finding fits with the stimulatory effects seen for the insulin-stimulated
GU in adipocytes shown in [Fig. 4].
We have demonstrated that a lipophilic extract of the roots of E. purpurea had a significant effect on basal and insulin-dependent GU and was able to activate
PPARγ in transfected cell cultures. The bioassay-guided fractionation approach led to the
isolation of an inseparable mixture of two new C12-isomeric alkamides (compounds 1/2) that significantly enhanced basal and insulin-dependent GU in adipocytes and exhibited
the characteristics of PPARγ partial agonists.
Materials and Methods
Plant material
Plants of Echinacea purpurea (L.) Moench were propagated from seeds purchased from Rieger Hoffmann GmbH. The plantlets
were raised in a greenhouse and transplanted in a sandy loam soil (Aarslev, Denmark;
coordinates: 55.3° N, 10.5° E) in early June 2007. The roots were harvested in August
2010, washed and air-dried and stored at − 25 °C until use. E. purpurea plants were authenticated by one of the authors (Dr. Kai Grevsen), and a voucher
specimen (EP2010) is deposited at the Department of Food Science, Aarhus University,
Denmark.
Apparatus, materials, chemicals, and reagents
Semi-preparative HPLC was performed on a Dionex UltiMate 3000 binary semipreparative
LC system (Thermo Scientific), equipped with a DAD (DAD-3000 RS) and a Foxy Jr. fraction
collector (Teledyne ISCO Inc.). Separations were performed on a Develosil ODS-HG-5
RP-C18 column (5 µm; 250 × 20 mm i. d., Nomura Chemical Co.). UV and MS data were
collected from an HPLC-PDA-MS/MS system, LTQ XL (Linear Quadrupole 2D Ion Trap, Thermo
Scientific) mass spectrometer attached to a PDA detector. HR-ESI-MS was recorded on
a 10 205 micrOTOF-Q II focus high resolution mass spectrometer (Bruker Daltonics).
NMR spectra were recorded on a Bruker 400 MHz spectrometer. Flash chromatography was
performed in silica gel 60, particle size 0.063–0.2 mm, Merck and TLC in silica gel
60 aluminium sheets; 0.2 mm, 20 × 20 cm, Merck. MeCN, EtOAc, DCM, n-hexane, isopropanol, TFA, DMSO, dexamethasone, 1-methyl-3-isobutylxanthine, insulin,
Oil Red O, and Rosi (HPLC purity ≥ 98 %) were purchased from Sigma-Aldrich; DMEM,
FCS, CS, PBS, penicillin/streptomycin, and TRIzol reagent from Invitrogen.
Plant extraction and isolation
Fresh E. purpurea roots (3 kg) were homogenized and extracted twice with DCM (10 L) for 24 h in the
dark at 5 °C with periodical shaking. The extracts were combined, filtered, and dried
under vacuum (30 °C) to give 9.3 g crude DCM extract (yield (w/w): 0.31 %).
Initial separation of the DCM extract was done on a flash column (100 mm i. d., 300 g
silica gel) conditioned with n-hexane. The DCM extract (8 g) was re-dissolved in DCM (20 mL), applied to the column
and eluted using the following solvent gradient: 100 % n-hexane (600 mL), 10–90 % EtOAc in n-hexane in 10 % steps (600 mL each), and 100 % MeCN (1000 mL). Seventy fractions (100 mL
each) were collected and combined into nine fractions (A–I) based on TLC: A (1.6 g),
B (0.3 g), C (0.1 g), D (0.2 g), E (0.1 g), F (0.4 g), G (0.2 g), H (0.5 g), and I
(2.9 g). TLC plates were developed using n-hexane-EtOAc (60 : 40) as eluent and were inspected by UV light followed by visualization
with vanillin reagent. Testing of the fractions A–I in the GU bioassay revealed two
active fractions, A and D ([Fig. 1 A]), which were further tested in the PPARγ transactivation bioassay ([Fig. 1 B]). Non-active fractions according to the GU bioassay (fraction B, C, E–I, data not
shown) were not tested for PPARγ activity.
The active fractions A and D were analyzed by HPLC-PDA-APCI-MS/MS according to the
method described by Thomsen et al. [14]. Fraction A was found to contain fatty acids with α-linolenic acid as a major constituent. Fraction D contained alkamides and was re-dissolved
at a concentration of 10 mg/mL in MeCN and separated by semi-preparative HPLC. Mobile
phase: (A) water containing 500 ppm TFA and (B) 100 % MeCN containing 500 ppm TFA.
Solvent gradient: 0 min, 25 % B; 55 min, 100 % B; 70 min, 100 % B; 85 min, 25 % B;
column temperature 25 °C; flow rate was 5 mL/min; injection volume was 2 mL; detection
wavelengths were λ = 210, 254, and 280 nm. Separation of fraction D resulted in the isolation of dodeca-2E,4E,8Z,10E-tetraenoic acid 2-methylbutylamide (1) and dodeca-2E,4E,8Z,10Z-tetraenoic acid 2-methylbutylamide (2) as an 1 : 1 inseparable mixture (5.9 mg, HPLC purity ≥ 96 %; tR = 36.9 min) named compounds 1/2, and the known metabolites dodeca-2E,4E,8Z-trienoic acid isobutylamide (3; 22.3 mg, HPLC purity
≥ 96 %; tR = 38.4 min) and dodeca-2E,4E-dienoic acid isobutylamide (4; 17.5 mg, HPLC purity ≥ 98 %; tR = 44.7 min). Compounds 1/2 were obtained as colourless oil; UV λ
max (nm): 235, 261. 1H NMR (CDCl3, 400 MHz), see [Table 1] and Supporting Information. HPLC-MS/MS (APCI; positive mode): m/z 262 [M + H]+ (quasi-molecular precursor ion) gave the following product ions (intensities in %
relative to base peak in parenthesis): m/z 234 (4) [M + H – C2H4]+, 206 (3) [M + H – C4H8]+, 192 (6) [M + H – C5H10]+, 175 (28) [M + H – C5H13N]+, 149 (100), 147 (24) [M + H – C6H13NO]+, 133 (13), 121 (13) [M + H – C8H15NO]+, 107 (18), 93 (15), 91 (6) [see also Supporting Information]. HR-ESI-MS m/z 262.2176 [M(C17H27NO) + H]+ (calcd. 262.2171) [Supporting Information]. The structures of compounds
3 and 4 were established based on comparison of NMR and MS data with literature reports [12], [14], [18].
Adipocyte cell cultures
3T3-L1 preadipocytes were cultured in DMEM with 10 % CS supplemented with 1 % penicillin/streptomycin
at 37 °C in humidified 95 % air and 5 % CO2. At day two post-confluence (designated day 0), the cells were induced to differentiate
with 500 µM 1-methyl-3-isobutylxanthine, 1 µM dexamethasone, and 167 nM insulin (MDI
protocol), or with 1 µM dexamethasone and 167 nM insulin (DI protocol). For both protocols,
the medium was replaced at day 2 with DMEM supplemented with 10 % FCS, 1 % penicillin/streptomycin,
and 167 nM insulin and thereafter every second day with DMEM supplemented with 10 %
FCS and 1 % penicillin/streptomycin. Extract, fractions, and pure compounds were dissolved
in 0.1 % (v/v) DMSO and thereafter added to the medium to a final concentration as
indicated. Vehicle cells were treated with 0.1 % (v/v) DMSO equal to the DMSO concentration
in the medium. Mature adipocytes were treated with the compounds from days 0–8 or
days 0–4 as described.
Glucose uptake in adipocytes
Mature 3T3-L1 cells were seeded in 96-well plates and differentiated according to
the MDI protocol till day 8. At day 8, the cells were fed with medium supplemented
with DMSO, Rosi (1 µM), and crude extract, fractions, or pure compounds. GU bioassay
was performed 48 h later according to the method described by Christensen et al. [4]. GU was determined in eight parallel wells at each insulin concentration ([Fig. 1 A] and [Fig. 4]; Fig. 2S, Supporting Information).
Peroxisome proliferator-activated receptor γ transactivation bioassay
PPARγ transactivation of vehicle [0.1 % (v/v) DMSO], positive control (1 µM Rosi), crude
extract, fractions, or pure compounds dissolved in DMSO ([Fig. 5]) was performed on a mouse embryonic fibroblast cell line trypsinized at 70 % confluence
as described previously [4].
Competitive peroxisome proliferator-activated receptor γ binding assay
Competitive binding was performed using TR-FRET (LanthaScreen, Invitrogen) on Wallac
EnVision (PerkinElmer). A terbium-labeled anti-GST antibody was used to label purified
GST-tagged human PPARγ LBD. Energy transfer from terbium to the tracer, a fluorescent pan PPAR agonist,
enabled read-out of each test compoundʼs ability to displace the tracer. Relative
fluorescence values from dose-response curves for compounds 1/2 and Rosi (positive control) were analysed using GraphPad Prism software ([Fig. 5]). An unrestrained sigmoidal (one-binding site) dose-response curve was fitted to
each data set by linear regression.
Oil Red O staining
Mature 3T3-L1 cells were differentiated according to DI or MDI protocol and treated
every second day with the designated medium and DMSO, Rosi (1 µM), crude extract,
fractions, or pure compounds until day 8. At day 10, cells were washed in PBS, fixed
with 4 % paraformaldehyde for 1 h and washed with double-distilled water. Cells were
incubated with Oil Red O solution (8.57 mM Oil Red O in isopropanol) mixed 3 : 2 with
water for 30 min and washed twice in water.
Molecular docking studies
Docking studies were performed using GOLD version 5.1 [43] and default parameters (GoldScore, 100 % search efficiency). Protein Data Bank entry
2Q5S (human PPARγ) was selected as protein template. The active site was determined by selecting all
residues within a radius of 6 Å of the co-crystalized ligand for 2Q5S. Docking was
performed with a hydrogen bond to Ser342 set to constraint since this interaction
has been reported to be essential for binding of PPARγ partial agonists [23]. After docking, all compounds were minimized using LigandScoutʼs general purpose
MMFF94 implementation [44]. The best docking poses for the ligands were selected by developing a 3D pharmacophore
with LigandScout [45], [46], [47].
Real-time quantitative polymerase chain reaction
Total RNA was isolated on day 10 from maturing (days 0–4) 3T3-L1 adipocytes using
TRIzol reagent. 500 ng of RNA were reverse transcribed with RevertAid (Fermentas)
according to manufacturerʼs instructions. qPCR analysis was performed using the Mx3000P
qPCR system (Agilent Technologies) and SYBRGreen JumpStart Taq ReadyMix (Sigma-Aldrich).
Primers were purchased from Tag Copenhagen A/S. Data was analyzed using the ΔΔCt method,
and gene expression was normalized to TATA box binding protein (reference gene). Primer
sequences used for analysis of gene expression are shown in [Table 2].
Statistical analysis
All data were analyzed by SAS statistical programming software (version 9.2; SAS Institute
Inc.) and OriginPro software (OriginPro 8.0, OriginLab Corporation). Data were expressed
as the mean ± standard deviation (SD) of three independent experiments in triplicates.
The identification of significances between different groups was carried out with
Studentʼs t-test. A p value < 0.05 was considered statistically significant.
Supporting information
Adipocyte differentiation of DI protocol-treated 3T3-L1 preadipocytes with DMSO, 100 µg/mL
DCM extract, fraction A, fraction D, 30 µM compounds 1/2, and 1 µM Rosi, respectively (Fig. 1S), the effect of compounds 3 and 4 at a 30 µM concentration on insulin-dependent GU (Fig. 2S), and the HR-ESI-MS, MS/MS, 1H NMR and 1H–1H COSY spectra of compounds 1/2 are available as Supporting Information.
Acknowledgements
This research was supported by the Danish Council for Strategic Research (Project
No. 09–063086).