A PPARγ ligand present in Actinidia fruit (Actinidia chrysantha) is identified as dilinolenoyl galactosyl glycerol

Activity-guided fractionation of Actinidia fruit species, including Kiwifruit, has identified DLGG (dilinolenoyl galactosyl glycerol) as a potent PPARγ (peroxisome-proliferator-activated receptor γ) ligand. DLGG is a type of MGDG (monogalactosyl diacylglycerol) and is present in all chloroplasts, and therefore all green fruits and vegetables. PPARγ is a ligand-activated transcription factor that regulates glucose metabolism and inflammation. An ethyl acetate extract of Actinidia chrysantha was fractionated by HPLC and the PPARγ-binding activity was detected by FP (fluorescence polarization). Linoleic and α-linolenic acids in A. chrysantha were readily detected as PPARγ ligands. Slow-binding PPARγ ligands were detected in several hydrophobic fractions. High-resolution MS identified DLGG as one of these ligands and confirmed that its binding is non-covalent. DLGG is a slow-binding PPARγ ligand with an IC50 of 1.64 μM, ±0.093 achieved after 45 min equilibration. DLGG is the first example of a form of DAG (diacylglycerol) that is a PPARγ ligand. In addition, DLGG is the first reported glycolipid ligand for PPARγ and also the first non-covalent, slow-binding PPARγ ligand.

Actinidia species were selected as a potential source of new PPARγ ligands not only because they are rich in α-linolenic acid, a well-known PPARγ ligand, but also because they contain a variety of polyunsaturated compounds such as lutein, β-carotene and various other carotenoids [13]. The Actinidia deliciosa species, better known as kiwifruit, is very common in the human diet yet several other Actinidia species are cultivated as a human food source or have been long-used in traditional remedies (e.g. Actinidia eriantha and Actinidia polygama) [14][15][16][17][18][19].
α-Linolenic acid is often a component of galactolipids such as MGDG (monogalactosyl diacylglycerols) and DGDG (digalactosyl diacylglycerols). MGDG and DGDG are abundant in the chloroplasts of plants and in the light harvesting plastids of cyanobacteria. In plants, their concentrations increase to maintain membrane fluidity as an adaptation to increasing light intensity [20]. In wheat exposed to strong light, α-linolenic acid represents approximately 75 % percent of total chloroplast MGDG [20]. The MGDG content of fruits and vegetable vary according to their chloroplast content; by wet weight, broccoli, kiwifruit and lemons contain approximately 350, 55 and 6 mg/kg, respectively [21].
In this study, we have used activity guided fractionation to identify a new PPARγ ligand in ethyl acetate extracts of Actinidia species, of which kiwifruit (A. deliciosa) is a member. Our interest has focused on DLGG (dilinolenoyl galactosyl glycerol) which is a specific type of MGDG in which the acyl groups are linolenic acid. PPARγ -binding activity was measured using a FP (fluorescence polarization) assay, and the active molecule was characterized using HPLC and MS.

Plant material and ethyl acetate extraction of Actinidia fruits
Actinidia species were grown in the Plant & Food Research orchard and the fruit harvested when ripe. Fruit were freezedried, pulverized, extracted for 1 h at room temperature (22 • C) in ethyl acetate (10 ml solvent: 1 g freeze-dried powder), filtered and freeze-dried again to remove solvent. The freeze-dried extract was then dissolved in DMSO at a ratio of 1 ml DMSO: 0.5 g original dry fruit equivalent). The Actinidia species used were: Actinidia chinensis, A. polygama, A. eriantha, Actinidia glaucophylla, A. chrysantha, A. deliciosa (unripe), A. deliciosa (ripe). Each extract was diluted serially in DMSO and assayed for PPARγ -binding activity using FP.

FP assay of PPARγ ligands
FP assays were performed on the Tecan Safire2 fluorescence microplate reader (Tecan) at 22 • C, in a volume of 20 μl in Nunc 384-well black, shallow microplates. The PPARγ (green) competitive binding assay (PolarScreen TM ) kit was supplied by Invitrogen Corporation. For measurement of FP, λ ex and λ em were set at 470 and 525 nm. Samples in DMSO were added to the preformed ligand:receptor complex such that the final DMSO concentration was 1 %. Polarization is expressed in mP (milli-polarization) units. The FP technique depends on the fact that small fluorescent ligand bound to a receptor has a slow rotation and therefore emits highly polarized light. As the fluorescent ligand is displaced from the PPARγ receptor by a non-fluorescent competitor, the polarization (mP) value decreases due to the more rapid rotation of the unbound, low molecular mass, fluorescent ligand.

Fraction preparation
Freeze-dried fruit of A. chrysantha were ground to a fine powder in a mortar and pestle and 50 g mixed with 400 ml of DCM (dichloromethane) and allowed to stand overnight at room temperature. The DCM was decanted, and the fruit residue extracted with a further 250 ml DCM. After decanting this both DCM extracts were combined, filtered through Whatman 3 filter paper and evaporated to dryness under vacuum at 50 • C using a rotary evaporator. The residue was dissolved in 15 ml DCM and stored at 4 • C until used. This extract was called AgDCM1. For fractionation by HPLC, 1 ml of AgDCM1 was evaporated to dryness under N 2 and the residue dissolved in 2 ml MeOH. The HPLC system was composed of a Waters 2690 Solvent Delivery System with a Waters 996 diode array detector connected to a Foxy Junior (Isco) fraction collector all controlled by Chromeleon Chromatography Management System V6.8 (Thermo Dionex). The separation column was a Gemini 5 μm C18 4.6×250 mm (Phenomenex) and a binary solvent system was used with solvent A, MilliQ water and solvent B, acetonitrile. The mobile gradient was as follows: 10 % A, 90 % B, 0-0.5 min; linear gradient to 100 % B, 0.5-15 min; composition held at 100 % B, 15-34 min; return to initial conditions, 34-35 min. The chromatogram at 205 nm was used for peak detection and peaks were automatically collected into a 1 ml 96-well plate between 2-34 min.

Semi-preparative HPLC
To isolate specific compounds the sample of AgDCM1 was fractionated using semi-preparative HPLC. The HPLC system was the same as that used above; however, the separation column was a Synergi Hydro 4 μm 10× 250 mm (Phenomenex) at a flow rate of 3.0 ml/min. The same solvents were used but the mobile gradient was as follows: 10 % A, 90 % B, 0-0.5 min; linear gradient to 100 % B, 0.5-15 min; composition held at 100 % B, 15-40 min; return to initial conditions, 40-41 min. Fractions were collected into 10 ml glass tubes and like fractions from individual injections were combined and evaporated to dryness.

LC-QTOF-HRMS
The LC-MS system was composed of a Dionex Ultimate ® 3000 Rapid Separation LC system and a micrOTOF QII mass spectrometer (Bruker Daltonics) and was operating in a positive mode with an ESI source. The LC system contained a SRD-3400 solvent rack/degasser, HPR-3400RS binary pump, WPS-3000RS thermostated autosampler and a TCC-3000RS thermostated column compartment. The analytical column was a

Survey of Actinidia species for PPARγ ligand activity
PPARγ -binding activity was detected in all extracts and the activity varied among the different Actinidia species ( Figure 1A). The strongest binding activity was observed with A. glaucophylla which has a PPARγ IC 50 of 807 + − 99. 6. Figure 1(B)

Fractionation of PPARγ -binding activity in A. chrysantha extract
A DCM extract of A. chrysantha was prepared and RP (reverse -phase)-HPLC used to isolate factions that were tested for PPARγ -binding activity. Figure 2 shows Based on the UV/Vis spectral properties (results not shown) compounds 5 and 6 appeared to be fatty acids and with the aid of authentic standards and comparison of HPLC retention times and UV spectral properties, compound 5 was identified as α-linolenic acid and compound 6 as a linoleic acid. Both α-linolenic acid and linoleic acid have previously been reported as PPARγ activators [22][23][24]. As shown in Figure 3, α-linolenic acid and linoleic acid bind to PPARγ immediately under the assay conditions used. In comparison compounds 11, 12 and 15 show time-dependent binding to PPARγ ( Figure 3C) as indicated by a reduction in mP over time. This behaviour suggests that these ligands might be binding covalently, whereas α-linolenic acid and linoleic acids bind non-covalently. The time-course binding curve for the covalent PPARγ ligand GW9662 at low concentrations is very       809.5013 gave an ion at m/z 277.2187 which is consistent with linolenic acid. These data suggest that compound 11 is a glycolipid and the mass of the ions are consistent with an MGDG containing two linolenic acid moieties. To confirm this identification, an authentic standard was purchased and compared with compound 11 using LC-ESI-MS. The results are shown in Figure 4 and the identity of compound 11 as DLGG is confirmed both by HPLC retention times and high resolution mass spectral data.

Direct binding to PPARγ using ESI-MS
Slow binding enzyme and receptor ligands are frequently covalent binders [25]. An example of a covalent binding PPARγ ligand (GW9662) is shown in Figure 3   behaviour of DLGG (compound 11) indicates that DLGG might bind covalently to PPARγ . ESI-HRMS of PPARγ was used to determine whether DLGG binds covalently to PPARγ and the results are shown in Table 1. The molecular mass of the PPARγ purchased from Invitrogen was shown to be 35919 Da with a further major component of 36096 Da. This is consistent with the amino acid composition of PPARγ . When IAF (5iodoacetamidofluorescein) was added to PPARγ , the molecular mass of PPARγ increased by 389 Da, consistent with covalent modification of cysteine in the binding site of PPARγ . Similarly, when the known covalent PPARγ binder, GW9662 [9], is added, the molecular mass of PPARγ increases by 241 Da indicating that covalent binding has occurred. However, when DLGG was added to PPARγ the molecular mass of PPARγ did not increase indicating that DLGG does not bind covalently to PPARγ .

DOG is not a PPARγ ligand
To determine whether the slow binding of DLGG could also be demonstrated with another DAG (diacylglycerol), DOG was assayed in the same manner. A dilution series of DLGG and DOG in DMSO was tested in the PPARγ FP assay. The result ( Figure 5) confirmed that DOG does not act as a PPARγ ligand.
Although the DOG was tested at the same concentrations as DLGG, only the top concentration is shown in Figure 5 because none of the DOG data showed evidence of binding and the DOG curves overlapped.
The formal possibility exists that the slow-binding phenomenon is due to the slow release of non-esterified fatty acids during the assay due to contamination of the assay reagents with lipase activity. Conceivably the recombinant, bacterially derived PPARγ protein might contain traces of lipase. The addition of pancreatic or Candida rugosa lipase to an emulsion of triolein leads to the time-and lipase-dependent release of oleic acid which displaces the FP assay ligand (results not shown) in a manner identical with the slow-binding curves of DLGG in Figures 3 and 5. However, Figure 5 confirms that no oleic acid is produced from DOG during the assay, indicating that the DLGG slow-binding phenomenon is not due to a contamination of the recombinant PPARγ protein with lipase. Thus the apparent slow binding of DLGG to PPARγ is not due to the accidental production of linolenic acid during the FP assay.

DISCUSSION
The finding that DLGG is a ligand for PPARγ is relevant for human health because DLGG and other MGDGs are abundant in the human diet. As a chloroplast component, MGDGs are present in all green (chlorophyll-containing) fruits and vegetables. Although we have isolated DLGG from Actinidia species, the highest concentrations of MGDG are in green vegetables, for example, parsley (Petroselinum crispum) is estimated to contain 184 mg MGDG/100 g wet weight, in comparison with 4.5-6.5 mg/100 g kiwifruit (A. chinensis) and 0.8 mg/100 g banana (Musa manzano) [21].
The time-course binding curve of DLGG is similar to that of GW9662, at low concentrations (results not shown). However, the mechanism underlying the slow binding of these two ligands appears to be quite different. The slow binding of GW9662 is due to the extremely low concentration of the compound in the assay. At low nanomolar concentrations of GW9662 the on-rate of binding is essentially diffusion controlled and GW9662 forms a covalent bond with PPARγ Cys285 [9]. By contrast, DLGG does not bind covalently to PPARγ (Table 1) and the slow binding of DLGG occurs in the low micromolar concentration range. The slow binding may indicate the ligand exists in a variety of conformations in the sample and only occasionally adopts a conformation which permits binding to the PPARγ . For example, if the assay concentration of DLGG is higher than its critical micelle concentration, the proportion of free DLGG in solution may be much lower than described in the Figure. Although the critical micelle concentration for DLGG is unknown to us, the critical micelle concentration of related plant lipid molecules is reported to be 13.1 μM for DGDG, 6.3 μM for sulphoquinovosyl DAG and 5.1 μM for DAG trimethylhomoserine [26]. These critical micelle concentration values are close to the DLGG concentrations which exhibit slow-binding to PPARγ and therefore suggest the true DLGG concentration in solution may be lower than the values reported in Figure 5. Another possible explanation for the slow binding of DLGG to PPARγ may be that the PPARγ conformation changes to accommodate DLGG in its binding site, i.e. DLGG may bind by an 'induced-fit' mechanism. Equilibration times of several hours have been observed when a peptide ligand binds to MHC class II molecules [27] by an induced-fit mechanism.
The physiological significance of DLGG-PPARγ interaction is unknown and it has not been confirmed in these experiments that DLGG can enter cells. Arguably the action of gut lipases might hydrolyse DLGG to release α-linolenic acid and glycerol before it is able to access gut epithelial cells. However, DLGG is reported to induce anti-inflammatory gene expression in chondrocytes in tissue culture [28]. In that study COX-2 (cyclooxygenase-2) is hypothesized to be the target of DLGG action. We would suggest that some anti-inflammatory activities of DLGG may stem from its binding to PPARγ .
DLGG is a galactose-containing analogue of DAG. DAG is a second messenger in PKC (protein kinase C) activation and PKC-mediated signal transduction. DAG recruits PKC to the internal surface of the plasma membrane and weakens the inhibitory pseudosubstrate domains control of the PKC catalytic domain [29]. Thus, the binding of DLGG, a glycosylated form of DAG, to PPARγ raises the possibility that PKC might compete with PPARγ for binding to limiting concentrations of DAG. DOG is a PKC activating co-factor [30] and is clearly not a PPARγ ligand ( Figure 5). However, it is conceivable that 1,2-dioleoylglycerol or 1,2-dilinolenoyl glycerol might show some affinity for PPARγ . Whether the DLGG binding to PPARγ is dependent on the presence of galactose remains to be determined.
The slow-binding of DLGG to PPARγ is interesting because of DLGG's abundance in the human diet. It is tempting to speculate that the slow-binding phenomenon may also occur with intrinsic PPARγ ligands and might even suggest some regulatory role for PPARγ in which lipophilic second messengers may be gradually sequestered. Further characterization of the additional PPARγ binding components present in chloroplast containing foods is warranted. Although we have not demonstrated the physiological effects of DLGG on cells in terms of PPARγ dependent gene-expression, this is the first demonstration of a non-covalent, slow-binding PPARγ ligand. In addition, this is the first evidence that PPARγ can accommodate either DAGtype or glycolipid ligands.