Identification and Mechanism of 10-Carbon Fatty Acid as Modulating Ligand of Peroxisome Proliferator-activated Receptors*

Background: Mechanism of action of medium chain fatty acid remains unknown. Results: Our results show that decanoic acid (C10) binds and activates PPARγ. Conclusion: Decanoic acid acts as a modulator of PPARγ and reduces blood glucose levels with no weight gain. Significance: This study could lead to design of better type 2 diabetes drugs. Peroxisome proliferator-activated receptors (PPARα, -β/δ, and -γ) are a subfamily of nuclear receptors that plays key roles in glucose and lipid metabolism. PPARγ is the molecular target of the thiazolidinedione class of antidiabetic drugs that has many side effects. PPARγ is also activated by long chain unsaturated or oxidized/nitrated fatty acids, but its relationship with the medium chain fatty acids remains unclear even though the medium chain triglyceride oils have been used to control weight gain and glycemic index. Here, we show that decanoic acid (DA), a 10-carbon fatty acid and a major component of medium chain triglyceride oils, is a direct ligand of PPARγ. DA binds and partially activates PPARγ without leading to adipogenesis. Crystal structure reveals that DA occupies a novel binding site and only partially stabilizes the AF-2 helix. DA also binds weakly to PPARα and PPARβ/δ. Treatments with DA and its triglyceride form improve glucose sensitivity and lipid profiles without weight gain in diabetic mice. Together, these results suggest that DA is a modulating ligand for PPARs, and the structure can aid in designing better and safer PPARγ-based drugs.

Metabolic syndrome, exemplified by type-2 diabetes and obesity, has become an epidemic of major proportions in the United States and around the world primarily because of the changing life styles and food habits (1,2). Management of the disease includes increased physical activity, dietary restrictions, and pharmacological intervention (3). Rosiglitazone and pioglitazone are members of the thiazolidinedione (TZD) 4 class of drugs used to treat type-2 diabetes. The action of these two drugs is mediated through binding and activation of peroxisome proliferator-activated receptor (PPAR)-␥ (4). However, TZDs have been associated with adverse effects, including weight gain, edema, and increased risk for cardiovascular diseases (5,6). These side effects have drastically reduced the clinical usage of TZDs in controlling type-2 diabetes. TZDs are full agonists of PPAR␥, and their strong PPAR␥ activation activities are considered to be part of the reasons for their side effects. Thus, there is an increasing focus on the development of partial agonists or modulators of PPAR␥ that retain their anti-diabetic properties but do not cause undesirable side effects (7,8).
PPAR␥ belongs to a subfamily of nuclear receptors that includes PPAR␣ and PPAR␤/␦. These receptors are key regulators of metabolic homeostasis in various tissues (9,10). PPAR␥ is expressed predominantly in adipose tissue where it is required for fatty acid synthesis and adipogenesis. PPAR␣ is enriched in the liver, and PPAR␤/␦ is expressed in many peripheral tissues (11,12). Both PPAR␣ and PPAR␤/␦ are mainly involved in fatty acid oxidation (13). PPARs form obligate heterodimers with retinoid X receptor (14), which then bind to specific DNA elements to regulate gene transcription in response to physiological signals (12). The exact identity of the endogenous PPAR␥ ligands remains unknown, but long chain unsaturated and oxidized/nitrated fatty acids (C16-C20) have been characterized as natural ligands of PPAR␥ (15)(16)(17)(18)(19)(20).
In addition to pharmacological intervention, dietary control of type-2 diabetes has received increasing attention. For example, foods containing fish oil enriched with -3 fatty acids have become a standard part of current dietary recommendations (21). In contrast, the benefits of foods enriched with other natural oils are greatly debated (22)(23)(24). Conventional oils such as long chain triglycerides (LCTs) consist of both saturated and unsaturated long chain fatty acids of 12 or more carbons. LCT diets have beneficial effects on serum triglyceride levels and serum lipid profiles but lead to adipogenesis and weight gain at the same time (23)(24)(25). Conversely, diets based on medium chain oils, which mainly consist of medium chain fatty acid (C6-C10) triglycerides, are known to alleviate diabetic conditions by increasing insulin sensitivity in adipose and muscle tissues (24). MCT oils have been used for treating cardiomyopathies and epilepsy and lipid absorption disorders (26,27). Furthermore, diets containing MCT oils are associated with a reduction of adipogenesis and weight gain as well as an improvement of serum triglycerides and lipid profile (28 -30).
Although the mechanism of how fish oils (-3 fatty acids) activate PPAR␥ is known, the mechanism of action for MCTs is less clear. In this study, we report that decanoic acid (DA, a C10 fatty acid), but not C6 or C8 fatty acids, is a direct ligand of PPAR␥, with its binding pocket drastically different from the binding pocket of TZDs or long chain fatty acids. Both cellbased and animal studies reveal that DA is a modulating ligand of PPAR␥ by improving glucose sensitivity and lipid profile without causing adipogenesis and weight gain. Identification of DA as a PPAR␥-modulating ligand establishes a solid framework for rationalizing the large amount of nutritional data on MCT oils in the literature (25, 26, 28 -31) and provides a mechanistic basis for designing the next generation of PPAR␥-based drugs that mimic the action of DA.

Chemicals
All the chemicals were obtained from Sigma, unless mentioned otherwise.

Competitive TR-FRET Binding Assay
The GST PPAR␥ LBD was labeled with a terbium-linked antibody and a fluorescent small molecule pan-PPAR ligand (Fluormone TM Pan-PPAR Green) is used as a tracer that is displaced by the ligand binding domain upon agonist binding. Excitation of the terbium at 340 nm results in FRET to the fluorescent tracer, with emissions detected at 520 and 495 nm. Upon ligand binding, the tracer is displaced from the PPAR␥ LBD and there is a loss of the FRET signal between the terbium label and the fluorescent tracer.

Coactivator Recruitment Assay
The experiments were performed with 20 nM receptor LBD and 20 nM biotinylated cofactor motif peptides in the presence of 5 g/ml donor and acceptor beads in a buffer containing 50 mM MOPS, 50 mM NaF, 0.05 mM CHAPS, and 0.1 mg/ml bovine serum albumin, all adjusted to a pH of 7.4. The reaction mixture was incubated for 90 min to equilibrate and read using a Wallac 2140 EnVision TM multilabel plate reader in a 384-well plate format. The intensity of the photon count is directly proportional to the binding efficiency of the protein to the peptide, and thus the relative binding affinities of the ligands were determined.

Transient Cotransfection Assays
COS-7 cells from ATCC were grown to 70% confluence in DMEM supplemented with 10% FBS and antibiotics. For assessing full-length PPAR receptors, COS-7 cells were transiently cotransfected with a plasmid containing the luciferase gene under the control of three tandem PPAR-response elements (100 ng) (PPAR-response element ϫ 3 TK-luciferase) and 50 ng of full-length human PPAR␥ plasmids using Lipofectamine 2000 (Invitrogen) along with the standard 5 ng of Renilla luciferase gene. 24 h after transfection, the cells were treated with 10 and 50 M concentrations of rosiglitazone, DA, and its triglyceride form (GT) for 24 h. Reporter luciferase assay kits from Promega were used to measure the luciferase activity, according to the manufacturer's instructions, with a luminometer (Envision, PerkinElmer Life Sciences). Luciferase activity was normalized by Renilla units. Each condition was performed with n Ն3 for each experiment. Luciferase reporter assays were also performed by transfecting 50 ng of the human lipoprotein lipase gene (Lpl) vector and treated with varying concentrations of GT. The cell culture media were supplemented with 0.1% BSA. Orlistat was used at 100 M concentration to inhibit Lpl. The siRNA of Lpl along with the scr siRNA was used to suppress the endogenous lipase gene and was treated with GT.

Adipocyte Differentiation Assay, RNA Isolation, and Real Time PCR Analysis
The adipocyte differentiation assay was performed with NIH 3T3-L1 preadipocytes obtained from ATCC. The 3T3-L1 preadipocytes were maintained in DMEM containing 10% FBS and antibiotics. For differentiation DMI (dexamethasone 1 M, 3-isobutyl-1-methylxanthine 0.5 mM, and insulin 167 nM) and 10 M of rosiglitazone were used as positive controls for the assay. DA was used at a 300 M concentration. All the treatments have insulin at 167 nM concentration. Media with DMSO were used as a negative control. Differentiation was induced by treating post-confluent cultures with media containing the respective ligands for 2 days. The media were changed every 2 days, and on day 12 cells were stained with Oil Red O to estimate the lipid accumulation. The images were analyzed using ImageJ (National Institutes of Health).
For real time PCR analysis, 1-2 g of total RNA was reversetranscribed using the SuperScript cDNA reverse transcription kit (Invitrogen). SYBR Green quantitative reactions using the SYBR Green PCR Master Mix (Applied Biosystems) were performed with the gene-specific primers (Table 3) using an ABI StepOne Plus machine. The relative expression of mRNA was determined after normalization to hypoxanthine-guanine phosphoribosyltransferase and GAPDH levels using the ⌬⌬Ct method.

Protein Purification, Crystallization, Data Collection, Structure Determination, and Refinement
The ligand binding domains of PPAR␣, PPAR␤/␦, and PPAR␥ (residues 206 -477 were cloned into a pRSET␣ vector with an N-terminal His 6 tag. Protein was expressed and purified as described before (32). The PPAR␥⅐PGC-1␣⅐decanoic acid complex was crystallized in the ratio of 1:1.2:20. Crystals were grown at 25°C in hanging drops containing 1.0 l of the above protein complex and 1.0 l of well solution containing 0.2 M sodium acetate (pH 6.2), 20% PEG 3350, 15% glycerol, and 1 mM concentration of the ligand. The crystals were directly frozen in liquid nitrogen for data collection. The PPAR␥⅐PGC-1␣/DA crystals formed in the P21 space group, with a ϭ 43.72 Å, b ϭ 54.34 Å, c ϭ 66.80 Å, ␣ ϭ ␥ ϭ 90°, ␤ ϭ 107.52°, and contain one molecule per crystallographic asymmetric unit. Full 360°data were collected from a single crystal using 1°oscillation by a MAR300 CCD detector at the 21-ID-D line of sector-21 at the Advanced Photon Source, Argonne National Laboratory. The observed reflections were reduced, merged, and scaled with the HKL2000 package (33). The structures were determined by a molecular replacement method using the crystal structure of PPAR␥ LBD (PDB code 3CS8) (32) as a template with the CCP4i program (34). Manual model building was carried out with Coot (CCP4i package), and structure refinement was performed with Refmac5 (35).

Animal Studies
Profiling the Effects of DA in db/db Mice by Subcutaneous Injection of DA-Six-week-old db/db male (BKS.Cg-m ϩ/ϩ Lepr db/J) mice from The Jackson Laboratory (Bar Harbor, ME) were obtained and acclimatized for at least 1 week before experiments. The mice were on ad libitum access to standard chow (10 kcal % fat; Research Diet) and water. Decanoic acid (250 mg/kg body weight daily), rosiglitazone (5 mg/kg body weight, daily), or PBS (500 l daily) were administered by subcutaneous injections. Blood was collected via retro-orbital bleed (200 l). The FreeStyle Freedom glucometer by Abbott was used for measurement.
Profiling the Effects of DA in db/db Mice by Including GT in Diet-Male db/db mice, 8 -10 weeks of age, were purchased from The Jackson Laboratory. All mice were maintained in a temperature-controlled facility with a 12-h light/dark cycle and were given free access to food and water. The mice were randomly divided into three groups of seven each and were given either a regular chow diet or a chow diet containing glyceryl tridecanoate (10 g/kg diet) or pioglitazone (100 mg/kg diet), respectively. The mice were treated for 5 weeks. Tail vein blood was used for glucose quantification with the Freestyle glucose meter (Abbott). After 3 weeks of treatment, intraperitoneal glucose tolerance tests (GTT) were performed by injecting a dose of 2 g of glucose/kg body weight into overnight fasted mice. The glucose levels were monitored at 0, 30, and 140 min after injection. Serum was collected and stored during the course of the study to carry out biochemical analysis. Total cholesterol and triglycerides were measured using kits from Wako Diagnostics (Richmond, VA).

Sample Preparation Procedure for Determination of the Decanoic Acid in Adipose Tissues
The DA in the adipose and liver tissue samples was determined by gas chromatography-mass spectrometry (GC-MS) as reported previously (36,37). Briefly, the wet adipose and liver tissues were blended into tissue homogenates after thawing to room temperature. Then 50 l of internal standard, deuterium substituted d 33 -heptadecanoic acid (0.8 mg/ml in hexane), was added into 100 l of the tissue homogenate sample or calibration standard containing decanoic acid in concentration of 0.1, 1, 10, 100, 1000, 10,000, or 100,000 nM in a matrix of phosphatebuffered saline containing 1 g/liter bovine serum albumin and 1 g/liter sodium azide. The mixture was further mixed with 2 ml of methanol/acetyl chloride (20:1 v/v) in a 10-ml glass tube closed with Teflon-lined caps. After sonication for 15 min at room temperature, the mixture in the glass tube was then incubated at 70°C for 90 min. After the tube was cooled to room temperature, 5 ml of 6% K 2 CO 3 was added to the mixture. Finally, 0.4 ml of hexane was added to the mixture and vortexed for 5 min. The upper organic layer containing the fatty acid methyl esters was transferred to the GC-MS sample vial for the GC-MS analysis.
Gas Chromatography-Mass Spectrometry Analysis-GC-MS analysis of decanoic acid was performed on an Agilent 6890N GC coupled to a 5975 inert XL mass selective detector (Agilent Technologies, Palo Alto, CA). An Omegawax-320 (Supelco, Bellefonte, PA) capillary column (30 m ϫ 0.32 mm inner diameter) was used for the separation of the methyl esters of the fatty acids. Purified helium gas (purity 99.9999%, Soxal, Singapore) was used as the carrier gas at a constant column flow rate of 3 ml/min. Standards and samples were injected using an Agilent 7683B series autosampler. The injection volume was 2.0 l, and the solvent delay was kept for 2 min. The initial oven temperature was set at 120 ºC, held for 3 min, followed by a ramp up to 180 ºC at 5 ºC/min, held for 2 min at 180 ºC, then followed by another ramp up to 230 ºC at 2 ºC/min, and a final hold for 10 min. The front inlet and mass selective detector transfer line temperatures were kept at 280 and 300 ºC, respectively. The mass selective detector was operated in selected ion monitoring mode with a dwell time of 100 ms for the ions m/z 143 and m/z 317 for methyl esters of decanoic acid and d 33 -heptadecanoic acid (internal standard), respectively. The decanoic acid concentration was quantified using a seven-point calibration curve of peak area ratio for decanoic acid to internal standard d 33heptadecanoic acid against the concentration.
Ser-273 Phosphorylation Analysis in Mouse Adipose Tissues-Western blotting was performed to analyze the Ser-273 phosphorylation in mouse adipose tissues according to Choi et al. (45).

RESULTS
Direct Binding of DA to PPAR␥ LBD-Because of the biological activity of MCT oils in improving glucose sensitivity and lipid profiles, we wanted to determine the relationship of medium chain fatty acids with PPAR␥. We first profiled fatty acids with a length between 6 and 22 carbons for their ability to promote PPAR␥ recruitment of the CBP-1 coactivator LXXLL motif by using an AlphaScreen biochemical assay. As shown in Fig. 1a, at 10 M concentrations, long chain saturated/unsaturated fatty acids with 12-22 carbons have varying abilities in promoting the binding of the CBP-1 LXXLL motif to PPAR␥. To our surprise, decanoic acid (C10) readily promotes recruitment of the LXXLL motif to PPAR␥, whereas C6 and C8 fatty acids were inactive, and the C12 fatty acid was less active, demonstrating a specific structure-activity relationship for the C10 fatty acid. DA at 10 M concentration also promotes binding of CBP-1 peptide to both PPAR␣ (1.8-fold) and PPAR␤/␦ (1.34fold), although to a lesser extent than PPAR␥ (2-fold) (Fig. 1b). However, DA at 50 M concentration promotes greater binding of CBP-1 peptide to PPAR␣ (3.6-fold), PPAR␤/␦ (5.2-fold), and PPAR␥ (3-fold). These results suggest that DA is also a ligand for PPAR␣ and PPAR␤/␦. We have also performed Alpha-Screen assay in the presence NCOR1 peptide that interacts well with PPAR␥ in the absence of ligands. In the presence of agonists such as rosiglitazone, the photon count is decreased, and we observe the same in the presence of DA but to a lesser extent (Fig. 1c), suggesting a weak agonist property of DA.
To further characterize the agonist/antagonist property of DA, we compared the ability of DA and rosiglitazone to promote PPAR␥ recruitment of various coactivator and corepressor motifs (Table 1). DA induced a strong interaction of PPAR␥ with various coactivator LXXLL motifs from the SRC1, CBP-1, TRAP220, and PGC-1␣ ID1 coactivator motifs but not with corepressor motifs from NCOR and SMRT, suggesting that DA is a PPAR␥ activator. The peptide binding profile of DA is similar to that of rosiglitazone (Fig. 1d), although generally DA has less ability than rosiglitazone to facilitate PPAR␥ recruitment of coactivator motifs. Interestingly, rosiglitazone enhanced binding of the SRC2-3 motif, whereas DA did not, further indicating the differential capacity of DA and rosiglitazone in promoting PPAR␥ recruitment of coactivators.
To determine the competitive binding affinity of fatty acids to PPAR␥, we used LanthaScreen TR-FRET technology from Invitrogen. The TR-FRET competitive binding assay was used to calculate the inhibition constants of octanoic acid (OA) and DA with rosiglitazone as a positive control. The inhibition con- stant k i of rosiglitazone is 53 nM. DA had a K i of 41.7 M although OA (C8 fatty acid) did not show any significant binding to the PPAR␥ LBD even at concentrations as high as 1 mM (Fig. 1e). The saturation levels obtained with DA were similar to that of rosiglitazone. The results show that DA is the only fatty acid among the medium chain group (C6 -C10) that binds to PPAR␥.
DA Activates PPAR␥ in Cells-To determine the cellular activity of DA, we transfected COS-7 cells with a PPAR␥ fulllength gene, a firefly luciferase reporter driven by a PPAR-response element, and a control reporter of Renilla luciferase. Cells were then treated with rosiglitazone, DA, and its triglyceride form, and the fold of activation was calculated against a DMSO control. At 10 and 50 M concentrations, rosiglitazone activated the PPAR␥ reporter by 5-and 7-fold, respectively, although DA increased the PPAR␥ reporter expression by 3.3and 4.3-fold, respectively (Fig. 2a). These results confirm our binding studies that DA binds to and activates PPAR␥ in a cellbased reporter system.
Interestingly, the triglyceride form of DA was also able to activate the PPAR␥ reporter to the same degree as DA in its free fatty acid form (Fig. 2b). Triglycerides in the body circulate as lipoproteins and can be broken down into individual fatty acids and glycerol components by lipoprotein lipases. To determine the role of lipases in PPAR␥ activation by the triglyceride form of DA, we inhibited lipase activity either by siRNA or by orlistat, which is a lipase inhibitor with broad specificity (38). As shown in Fig. 2b, lipase-specific siRNA, which efficiently knock down the LPL gene as demonstrated by quantitative PCR (Fig. 2c) reduced the DA triglyceride-induced PPAR␥ activation but had little effect on either the free fatty acid form of DA or rosiglitazone. In contrast, increased expression of lipase did not further enhance activation by the DA triglyceride, indicating the endogenous lipase is sufficient for the conversion of the triglyceride to free fatty acids for the full activation of PPAR␥. The lipase inhibitor orlistat had a similar effect as lipase-specific siRNA, which inhibited the activation of the DA triglyceride but had little effect on rosiglitazone or the free fatty acid form of DA. Together, these results indicate that DA, either in the fatty acid form or in the triglyceride form, was able to activate the PPAR␥ reporter in cell-based assays. However, the triglyceride form needs to be broken down into free fatty acids to activate the receptor.
DA Does Not Induce Adipogenesis-One of the major side effects of PPAR␥-based drugs is weight gain as PPAR␥ is the key activator of adipogenesis. To determine whether DA promotes adipogenesis, we treated mouse fibroblast 3T3-L1 cells with saturated concentrations of DA (300 M). Cells were also separately treated with rosiglitazone (10 M) or a DMI (dexamethasone 1 M, 3-isobutyl-1-methylxanthine 0.5 mM, and insulin 167 nM) mixture, a standard recipe for adipocyte differentiation. Cells were stained with Oil Red O that colors adipocytes in red. Adipocyte differentiation was observed after DMI treatment by day 7 (Fig. 3a) or after treatment with rosiglitazone by

Decanoic Acid Is a Modulator of PPAR␥
JANUARY 2, 2012 • VOLUME 287 • NUMBER 1 day 10 (Fig. 3c). In contrast, DA alone did not promote adipogenesis (Fig. 3e), which is similar to the negative control (Fig.  3f). Furthermore, when DA was added along with DMI and Rosi treatments, we observed a clear decrease in the total number of adipocytes (Fig. 3, b and d). This leads to the conclusion that DA does not activate adipogenesis despite activating PPAR␥, and furthermore, DA can inhibit adipogenesis activated by DMI and rosiglitazone. We have also performed adipogenesis assay with vehicle control, OA (300 M), oleic acid (OLA), along with Rosi (10 M) (Fig. 4, a--c and e). OA did not promote adipogenesis by itself, whereas OLA promotes adipocyte differentiation. When Rosi was added along with OA, we did not see reduction in the adipocytes suggesting that OA does not compete with Rosi for binding to PPAR␥ at 300 M concentration (Fig. 4d). We did not find any difference in number of adipocytes in the cells treated with OLA and Rosi (Fig. 4e). PPAR␥ activation is required for adipogenesis, which in turn up-regulates other adipogenic genes. To determine whether DA activates PPAR␥ in 3T3-L1 cells, we used quantitative PCR to measure the mRNA of endogenous PPAR␥-regulated genes (primers are listed in Table 2). Quantitative PCR was performed on DA, DMI, and rosiglitazone-treated cells after 2 and 8 days, and the results are represented with respect to the untreated cells on a log 2 scale. Treatment of the cells with DA leads to up-regulation of the Pparg gene in preadipocytes at day 2. DAinduced increase in Pparg mRNA levels was similar to that of rosiglitazone treatment but less than DMI-treated cells (Fig.   3g). DA treatment also led to increased expression of Pparg downstream genes that regulate fatty acid metabolism and adipogenesis such as Cebpa, Srebp1, Fabp4, and AdipoQ (Fig. 3g). At day 8 (Fig. 3h), the increased expression of Pparg target genes is similar between the DMI-and rosiglitazone-treated cells, but the activation of these Pparg target genes by DA is much less compared with DMI and rosiglitazone, especially the expression of Fabp4, Lep, AdipoQ, Cox7a1, and Pgc1a, whose expression is tightly connected with adipocyte differentiation (Fig. 3h). Together, these results indicate that DA is able to activate Ppar␥ target genes in preadipocytes, but to a lower level that is not sufficient to induce adipocyte differentiation.
Unique Binding Mode of DA to PPAR␥-To reveal the molecular basis of how DA modulates PPAR␥ activity, we determined the crystal structure of the PPAR␥ LBD in complex with DA and the PGC1␣ LXXLL motif at 1.52 Å resolution. The structure was deposited at the www.pdb.org with PDB code 3U9Q. Table 3 details the data collection and refinement statistics. The overall structure of the PPAR␥ LBD complex structure resembles the receptor bound with rosiglitazone and other long chain fatty acids (Fig. 5a). The C-terminal activation function-2 helix (AF-2 or helix 12) is packed closely against the main body of the LBD, forming a coactivator binding groove where the PGC1␣ LXXLL motif is docked. The binding mode of the PGC1␣ coactivator motif in the DA-bound structure is similar to its binding mode in the rosiglitazone-bound structure. A distinct feature of many ligand-bound PPAR␥ structures is the large ligand binding pocket (ϳ1500 Å 3 ) adjacent to the AF-2 helix, and this pocket is also present in the DA-bound structure. The major difference is found in the position of the C-terminal portion of helix 10 (Fig. 5d), where it forms a unique DA-binding pocket.
The most surprising aspect of the DA-bound structure is the unique DA-binding pocket that is dramatically different from the pocket occupied by rosiglitazone and long chain fatty acids. This unique binding mode of DA is well defined by the electron density map from the high resolution of the structure (Fig. 5f). The bound DA molecule is oriented into a narrow pocket formed by helices 3, 7, 10, and the AF-2 helix (Fig. 5a). In this orientation, DA occupies only a very small portion of the PPAR␥ pocket but leaves the vast volume of the pocket untouched. There is only one DA molecule in the PPAR␥ pocket. An alignment of DA, rosiglitazone (32) (PDB 3CS8), and nitrolinoleic acid (39) (LNO 2 ) (PDB 3CWD) ligands with respect to helix 3 reveals these differences (Fig. 5b). The acid headgroup of DA forms polar interactions with the residues His-323 (H5), His-449 (H10), and Tyr-473 (AF-2) along with Ser-289 (H3) that is not observed in case of TZDs (Fig. 5c). Although these hydrogen bonds are similar to those formed by the TZD group of rosiglitazone (32) or by the acidic group of long chain fatty acids, their geometry is less optimized as the hydrocarbon tail of DA is fitted into a narrow pocket formed by helices 3, 7, and 10 and the AF-2 helix (Fig. 5, g and h). Thus, the hydrocarbon tail of DA occupies a completely different pocket from that occupied by the tail of long chain fatty acids or by rosiglitazone. The tail of the DA is positioned in the hydrophobic pocket formed by the residues Ala-278, Ile-281, and Phe-282 of H3; Leu-353 of H6; Leu-356 of the loop after H6; Phe-360, Phe-363, and Met-364 of H7; Leu-453 of H10, and Leu-469 of AF-2. These snug interactions help to explain why short chain fatty acids (C6 and C8) do not bind to PPAR␥, and the long chain fatty acids (C12-C22) have to flip to the upper pocket, the same space that rosiglitazone occupies. The unique binding mode of DA in PPAR␥ may help to explain its distinct PPAR␥ activation property from that of TZDs and long chain    (32, 40 -42). The Phe-282 residue is displaced by DA to accommodate the tail of the fatty acid. The differences in the residue orientation in this region are documented in Fig. 5. Apart from the above-mentioned differences, the LBD structure is largely intact except for the H10. A side-on view of H10 shows a kink at the His-449 position in case of the rosiglitazone-bound structure when compared with the DA-bound structure (Fig. 5d). This is perhaps due to the greater pull by the hydrogen bond between His-449 and the TZD group of rosiglitazone. To determine the differential stabilization of the PPAR␥ LBD structure between DA and Rosi, we have performed hydrogen/ deuterium exchange followed by mass spectrometry (43)(44)(45). As shown in Fig. 5e, Rosi significantly reduced hydrogen/deuterium exchange in the structural elements surrounding its ligand binding pocket and the AF-2 helix, consistent with the fact that Rosi acts as a classical agonist of PPAR␥ by stabilizing the AF-2 helix. In contrast, DA had no effect on hydrogen/ deuterium exchange of PPAR␥ despite its binding to the receptor, further providing physical evidence that DA does not have the same ability as Rosi to stabilize the AF-2 helix, and therefore it has less ability to act as a full agonist.
Profiling DA Activity in Vivo-To determine the in vivo effects of DA, we used diabetic mice (BKS.Cg-m ϩ/ϩ Lepr db/J) as an animal model. DA was delivered to mice by two different methods. In one set of experiments, db/db mice were divided into three groups consisting of seven randomly sorted mice in each group. The control group was fed a normal chow diet, and other two groups were given a modified chow diet containing the triglyceride form of DA (glyceryl tridecanoate, 10 g/kg diet) or pioglitazone (100 mg/kg diet), respectively. GT was included  3U9Q). a, schematic representation of PPAR␥ LBD bound to DA. Helices are represented in red, sheets in yellow, and loops in green. DA is presented as blue dots. Helices are abbreviated as H. b, superimposition of DA (green), rosiglitazone (blue), and nitrolinoleic acid (LNO 2 ) (yellow) with respect to H3. Oxygen atoms are represented as red, nitrogen atoms as dark blue, and sulfur as orange. c, polar interactions of the TZD group of rosiglitazone (blue) and carboxyl headgroup of DA (green) with residues with Ser-289 from H3, His-323 from H5, His-449 from H10, and Tyr-473 from the AF-2 helix. d, side-on superimposition of the PPAR␥ H10 from the crystal structures bound to DA (red) and rosiglitazone (cyan) (PDB 3CS8) shows a kink at the His-449 position in case of the rosiglitazone (cyan)-bound structure. e, hydrogen/deuterium exchange (HDX) data plotted over the structures of PPAR␥ LBD bound with rosiglitazone (left, PDB 2PRG) and DA (right, 3U9Q). Percentage reduction in hydrogen/deuterium exchange relative to unliganded receptor is colored according to the key. f, electron density map (2F o Ϫ F c at 1) of DA in the ligand binding pocket of PPAR␥ and the residues that form polar interactions. g, stick representation of the PPAR␥ crystal structure bound to DA showing the displacement of F282. h, stick representation of Phe-282 residue of the PPAR␥ structure 2HFP bound to rosiglitazone.
in the diet as free fatty acids are not easily absorbed in the intestine. Blood glucose levels were measured every other week as well as body weight. In another set of experiments, mice were subcutaneously injected with 250 mg/kg DA on a daily basis or with 5 mg/kg of rosiglitazone and DMSO as positive and vehicle controls.
As shown in the Fig. 6a, mice fed pioglitazone and GT diets had a significant decrease in the fasted blood glucose levels when compared with the control chow diet. Injection of mouse with DA or rosiglitazone also showed significant decrease in the glucose levels with respect to the vehicle (Fig. 6b). One of the major problems associated with TZDs is the unwanted side effect of body weight gain. Indeed, both pioglitazone-and rosiglitazone-treated mice showed significant weight gain over the vehicle group after 2 and 4 weeks (Fig. 6, c and d). In contrast, DA-or GT-treated mice had no significant weight gain in comparison with control mice (Fig. 6, c and d).
Another hallmark of insulin resistance is the retention rate of excess glucose in the blood after external supply of glucose, which can be determined by a GTT. In this test, the GT-treated mice exhibited significantly better clearance of blood glucose when compared with the chow-fed mice but less than pioglitazone (Fig. 6e). GT-fed mice also showed a reduction in total serum cholesterol and triglycerides (Fig. 6, f and g). These results demonstrate that DA and its triglyceride form GT have pharmacological efficacy with respect to lowering blood glucose and lipids without causing additional weight gain as rosiglitazone and pioglitazone.
To determine the basis for the different pharmacological effects of DA and Rosi, we examined the expression profile by quantitative PCR on three major metabolic tissues as follows: white adipose tissue (WAT), brown adipose tissue (BAT), and liver (Fig. 7). In WAT and BAT, both DA and Rosi activated the Pparg-regulated gene fatty acid-binding protein Fabp4 and adiponectin AdipoQ, but the degree of activation by DA is much lower than that by Rosi (Fig. 7, a and b). In addition, rosiglitazone significantly up-regulated Pparg and its coactivator Pgc1a, although DA did not. Similarly in liver, rosiglitazone greatly activated Pparg and the liver fatty acid-binding protein Fabp1 (Fig. 7c). In contrast, DA only slightly elevated expression of Pparg and Ppar␣ but not Fabp1. Together, these data further support that DA is a partial agonist of PPAR␥ in vivo.
To further ascertain that the antidiabetic effects in DA are due to the activation of PPAR␥ in vivo, we have analyzed the adipose tissue concentration of DA. The total tissue lipid analysis was performed using gas chromatography-mass spectroscopy analysis with our established protocol (37). We observed a distinct DA peak in the GT-treated mice adipose tissue samples, whereas the control mice adipose tissues showed no peak (Fig. 7d). The mean concentration of DA in these tissues was found to be 56 (Ϯ 19.5) M. The DA concentration in control mice adipose tissue is 0.27 (Ϯ 0.3) M. The concentration of DA in the triglyceride-treated mice is close to the K d value of the DA binding to the PPAR␥. These results suggest that DA has accumulated in adipose tissues in physiologically relevant concentrations. Recently, Choi et al. (45) have reported that antidiabetic drugs inhibit obesity-linked phosphorylation of PPAR␥ by Cdk5 at Ser-273. The PPAR␥ Ser-273 phosphorylation was analyzed in four different adipose tissues from each treatment group. We did not observe any reduction in the overall phosphorylation in the GT-treated samples with respect to the control mice. However, reduction in Ser-273 phosphorylation was observed in case of pioglitazone-treated mice (Fig. 7e). However, despite no reduction of Ser-273 phosphorylation, we did observe a significant reduction in the blood glucose levels as well as an increase in the PPAR␥-dependent gene expression. Furthermore, pioglitazone-treated mice have larger hepatocytes and lipid droplets when compared with vehicle or GTtreated mice (Fig. 7, f-h). These results are consistent with our hypothesis that the mechanism of action of DA is different from the TZD class of PPAR␥ agonists.

DISCUSSION
In this study, we report that DA, a 10-carbon fatty acid, is a modulating ligand that binds and partially activates PPAR␥ as well as PPAR␣ and PPAR␤/␦. Our studies here are focused on PPAR␥ as it is the molecular target of the TZD class of antidiabetic drugs. PPAR␥ regulates glucose and lipid metabolism as well as adipogenesis (13,46). Although the physiological ligands of PPAR␥ largely remain unknown, long chain (C16 -C20) and polyunsaturated fatty acids have been known to bind and activate PPAR␥ (15,17). Oxidized and nitrated fatty acids have also been characterized as potent ligands of PPAR␥ (18,40,41). Identification of DA further expands the profile of natural fatty acids as PPAR␥ ligands. Based on our binding and structural data, PPAR␥ is able to bind fatty acids with 10 -22 carbons, some with varying degrees of unsaturation. These results further establish that PPAR␥ is a general receptor for fatty acids. PPAR␥ has the ability to sense a pool of fatty acid mixtures (either modified or unmodified), and the large PPAR␥ pocket size is a particular fit for this promiscuity of fatty acid recognition.
The expanded role of PPAR␥ as a sensor for medium chain fatty acids has important implications in fatty acid metabolism. There is extensive nutritional literature regarding the distinct biological activity of medium chain and long chain fatty acids and their corresponding triglycerides (23, 25, 28 -30, 47). Diets enriched with long chain triglycerides (LCT with Ͼ C12 fatty acids) and medium chain triglycerides (MCT predominantly with C6 -C10 fatty acids) are shown to have beneficial effects on serum triglyceride levels and serum lipid profiles (31). Nutrition studies reveal that LCT diets have side effects that include adipogenesis and weight gain (23,25). In contrast, MCT diets have been shown to reduce weight gain and control obesity (25). Currently several clinical trials are underway to study and confirm the effects of MCT oils on obesity, diabetes, and cardiovascular risks (clinical trial numbers NCT00207233 (48), NCT00529919 (31), and NCT00207272 (30)). The mechanism of action by MCT diets remains largely unknown. Based on our binding and cell-based assays, DA is able to bind and activate PPAR␥, but its activation level of PPAR␥ is too low to induce adipogenesis. Animal studies show that DA (or its triglyceride form) improves glucose and insulin sensitivity without weight gain. In this regard, DA is similar to a number of synthetic PPAR␥-modulating ligands, including GW0072 and others (7, 49 -51), which improve insulin and glucose sensitivity without FIGURE 6. Profiling of DA and its triglyceride form GT in db/db mice. In one set of experiments, male db/db mice were randomly divided into three groups of seven each and were given regular chow diet or modified chow diet containing GT (10 g/kg diet) and pioglitazone (pio) (100 mg/kg diet), respectively, for 4 weeks. In another set of experiments, mice were subcutaneously injected with vehicle DMSO or 5 mg/kg rosiglitazone or 250 mg/kg DA for 4 weeks. a, blood glucose levels of mice fed control chow diet and modified diet with GT and pioglitazone after 0, 2, and 4 weeks. b, blood glucose levels of mice injected with vehicle DMSO and treatments DA and Rosi 0, 2, and 4 weeks. c, body weight change in mice fed control chow diet and modified diet with GT and pioglitazone after 0, 2, and 4 weeks. d, body weight change in mice injected with vehicle DMSO and treatments DA and Rosi after 0, 2, and 4 weeks. e, GTT of different mice fed control chow diet and modified diet with GT and pioglitazone. f, total serum cholesterol after 4 week treatment. g, total serum triglycerides after 4 weeks of treatment. (Two-way analysis of variance was performed. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; n ϭ 7, Ϯ S.E.
inducing weight gain. Identification of DA as a modulating ligand of PPAR␥ thus partially explains the mechanism for the biological effect of MCT diets.
The crystal structure of PPAR␥ bound with DA reveals an unexpected mode of ligand recognition of DA by PPAR␥. In the structure, DA is positioned into a unique pocket space of PPAR␥ that is not occupied by other fatty acids or by rosiglitazone. Although the acidic head group of DA and other fatty acids form a similar interaction with the AF-2 helix, the hydro-carbon tail adopts a distinct conformation. The long chain hydrocarbon tail in most other fatty acid-bound PPAR␥ structures is oriented to wrap around Cys-285 of helix 3 (32,42,52). In contrast, the hydrocarbon tail of DA is oriented into a small pocket, which is too small for long chain fatty acids but is a perfect match for the 10-carbon length of DA. The difference in the binding mode of DA and long chain fatty acids may help to account for their distinct biological activity. Furthermore, the unique binding mode of DA thus provides a novel mode of FIGURE 7. Quantitative PCR of different tissues from db/db mice fed DA-containing chow diet. Mice were randomly divided into three groups of seven each, which were given regular chow diet, chow diet containing glyceryl tridecanoate (10 g/kg diet), or pioglitazone (Pio) (100 mg/kg diet), respectively, for 4 weeks. Results were quantitated by using ⌬⌬CT method and vehicle treatment as reference. Gapdh gene was used as an internal reference. a, relative mRNA expression of Pparg, Fabp4, AdipoQ, Pgc1a, and Pgc1a in WAT. b, relative mRNA expression of Pparg, Fabp4, AdipoQ, Pgc1a, and Pgc1a in BAT. c, relative mRNA expression of Ppar␥, Ppar␣, and Fabp1 in liver tissue (two-way analysis of variance was performed. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001) (n ϭ 5, Ϯ S.E.) d, total ion chromatogram for adipose tissue from GT-treated mice. After incubation of adipose tissue with methanol/acetyl chlorides (20:1 v/v) at 70 ºC for 90 min, all forms of DA were trans-esterified in the form of methyl ester. DA shows a distinct peak when compared with the internal standard. e, Ser-273 phosphorylation of PPAR␥ in mouse adipose tissues. Phosphorylation status of Ser-273 in mouse adipose tissues was analyzed using Ser-273-specific antibodies. Tissues from four mice from each treatment (vehicle, GT, and pioglitazone) were analyzed for Ser-273 phosphorylation as well as total PPAR␥. Cross-section of liver from mouse with different treatments. f, vehicle; g, glyceryl tridecanoate; h, pioglitazone (pio). ligand recognition by PPAR␥ and a structural basis of ligand design.
In summary, our data show that decanoic acid, a 10-carbon fatty acid, binds and partially activates PPAR␥. Activation of PPAR␥ by DA only induces expression of PPAR␥ target genes to a partial level that is not sufficient to induce adipogenesis. The crystal structure of DA bound to PPAR␥ reveals a new binding mode that is different from the binding mode of rosiglitazone and long chain fatty acids. Animal studies show that DA improves glucose sensitivity and lipid profile without weight gain in the diabetic db/db mouse model. However, because of the low affinity of DA, the decrease in glucose levels is less significant than TZD-treated mice. Hence, we propose that by using DA-bound structure as a scaffold, novel molecules can be designed that could have better affinity for PPAR␥ and with similar pharmacological properties.
Tissue lipid analysis shows significant accumulation of DA in the adipose tissue suggesting that DA accumulates in adipose tissue in physiologically relevant concentrations. However, it would be interesting to further analyze how DA affects liver, muscle, and cardiac tissues and their gene expression patterns.
Our in vivo quantitative PCR results of WAT, BAT, and liver tissues also suggests that DA acts as a modulator of PPAR␥ and increases the expression of PPAR␥-dependent genes without altering the PPAR␥ gene expression. Gene expression studies from liver also show an increase in the PPAR␥ and PPAR␣ levels. These results demonstrate that DA has a unique biological activity of a partial PPAR␥ agonist that is distinct from full PPAR␥ agonists like rosiglitazone. Given the side effects associated with full PPAR␥ agonists, the biological activity of DA and its unique binding mode in the PPAR␥ pocket may provide a mechanistic basis for designing better PPAR␥-based drugs that mimic the action and binding mode of DA.