Coordinate Functional Regulation between Microsomal Prostaglandin E Synthase-1 (mPGES-1) and Peroxisome Proliferator-activated Receptor γ (PPARγ) in the Conversion of White-to-brown Adipocytes*

Background: Microsomal prostaglandin E (PGE) synthase-1 (mPGES-1) is an inducible enzyme with unknown properties in adipose homeostasis. Results: mPGES-1 is necessary for pre-adipocyte differentiation into beige/brite adipocytes through functional interaction with peroxisome proliferator-activated receptor γ (PPARγ). Conclusion: A coordinate interaction between mPGES-1 and PPARγ is required for white-to-brown fat conversion. Significance: Increases in the number of beige cells in fat exerts beneficial metabolic actions. Peroxisome proliferator-activated receptor γ (PPARγ) is a ligand-activated nuclear receptor and a master regulator of adipogenesis. Microsomal prostaglandin E (PGE) synthase-1 (mPGES-1) is an inducible enzyme that couples with cyclooxygenase-2 for the biosynthesis of PGE2. In this study we demonstrate the existence of a coordinate functional interaction between PPARγ and mPGES-1 in controlling the process of pre-adipocyte differentiation in white adipose tissue (WAT). Adipocyte-specific PPARγ knock-out mice carrying an aP2 promoter-driven Cre recombinase transgene showed a blunted response to the adipogenic effects of a high fat diet. Pre-adipocytes from these knock-out mice showed loss of PPARγ and were resistant to rosiglitazone-induced WAT differentiation. In parallel, WAT from these mice showed increased expression of uncoupling protein 1, a mitochondrial enzyme that dissipates chemical energy as heat. Adipose tissue from mice lacking PPARγ also showed mPGES-1 up-regulation and increased PGE2 levels. In turn, PGE2 suppressed PPARγ expression and blocked rosiglitazone-induced pre-adipocyte differentiation toward white adipocytes while directly elevating uncoupling protein 1 expression and pre-adipocyte differentiation into mature beige/brite adipocytes. Consistently, pharmacological mPGES-1 inhibition directed pre-adipocyte differentiation toward white adipocytes while suppressing differentiation into beige/brite adipocytes. This browning effect was reproduced in knockdown experiments using a siRNA directed against mPGES-1. The effects of PGE2 on pre-adipocyte differentiation were not seen in mice lacking PPARγ in adipose tissue and were not mirrored by other eicosanoids (i.e. leukotriene B4). Taken together, these findings identify PGE2 as a key regulator of white-to-brown adipogenesis and suggest the existence of a coordinate regulation of adipogenesis between PPARγ and mPGES-1.

knockdown experiments using a siRNA directed against mPGES-1. The effects of PGE 2 on pre-adipocyte differentiation were not seen in mice lacking PPAR␥ in adipose tissue and were not mirrored by other eicosanoids (i.e. leukotriene B 4 ). Taken together, these findings identify PGE 2 as a key regulator of white-to-brown adipogenesis and suggest the existence of a coordinate regulation of adipogenesis between PPAR␥ and mPGES-1.
The formation of new adipocytes from precursor cells is a crucial aspect in controlling normal adipose tissue function (1,2). During the adipogenic process, adipocytes differentiated from mesenchymal stem cells give rise to two main types of adipose tissue: WAT, 3 characterized by the presence of adipocytes containing large unilocular lipid droplets, and BAT, composed by multiloculated adipocytes containing large numbers of mithocondria (1,2). WAT is the predominant fat type in human adults and is widely distributed through the body; its main function is to store excess energy as triglycerides (1,2). WAT is not only important for energy storage but also as an endocrine organ because it regulates whole body homeostasis by secreting adipokines (cytokines, chemokines, and growth factors) and biologically active lipid mediators (3,4). By contrast, BAT is located in discrete pockets and is specialized to generate heat by dissipating chemical energy, counteracting hypothermia, obesity, and diabetes (1,5). Apart from "classical" brown adipocytes that reside in BAT depots, another type of brown adipocytes, called beige or brite (brown-in-white) adipocytes, are sporadically found in WAT upon prolonged cold exposure or in response to ␤-adrenergic or thiazolidinedione exposure (1,5,6). Importantly, increases in the number of beige adipocytes in WAT are associated with protection against dietinduced obesity and metabolic diseases (7)(8)(9).
Adipocyte differentiation requires expression of peroxisome proliferator-activated receptor ␥ (PPAR␥), a member of the nuclear receptor superfamily that functions as a ligand-activated transcription factor regulating the genes implicated in adipogenesis (10). Although PPAR␥ functions as a master regulator of adipocyte differentiation and is absolutely necessary for both white and brown fat cell development (11,12), the mechanisms underlying terminal differentiation of pre-adipocytes into white or brown adipocytes up-stream and/or downstream from this transcription factor have not been completely elucidated.
Microsomal prostaglandin E (PGE) synthase-1 (mPGES-1) is an inducible enzyme that cooperates with cyclooxygenase-2 (COX-2) in the biosynthesis of PGE 2 , one of the most ubiquitous and biologically active endogenous lipid mediators (13). Several studies have linked this lipid mediator to the adipogenic program, although their results have yielded controversial views. For example, PGE 2 is recognized as a negative regulator of WAT lipolysis and adipocyte development (14,15), and preadipocytes stably transfected with either COX-1 or COX-2 show lower PPAR␥ expression (16). Consistent with this view, mice genetically deficient for mPGES-1 show basal elevations in PPAR␥ expression and transcriptional activity (17). However, more recent studies have shown that transgenic mice overexpressing COX-2 do not exhibit changes in PPAR␥ expression in adipose tissues (18), although PPAR␥ expression has been shown to be significantly reduced in adipose tissue from COX-2-deficient mice (19). The existence of this undefined role of the COX-2/mPGES-1/PGE 2 pathway in WAT physiology led us to investigate the functional interaction between PPAR␥ and mPGES-1 in the process of adipogenesis. Our results demonstrate that removing PPAR␥ specifically in fat cells triggers mPGES-1 expression and PGE 2 biosynthesis, uncovering an important role in the browning process of WAT pre-adipocytes.
Mice-Mice carrying the loxP-targeted PPAR␥ gene (PPAR␥ f/f ) were crossed with transgenic mice expressing Cre recombinase under the control of the aP2 promoter to yield adipocyte-specific PPAR␥ KO mice (⌬adip) (20). aP2 is an adipose-specific gene that is also expressed in undifferentiated adipogenic progenitors (21,22). The control group included wildtype (WT) mice carrying the loxP-targeted PPAR␥ gene. Mice were housed on wood-chip bedding cages at a ratio of 4 mice per cage with a humidity level of 50 -60% and a 12-h light-dark cycle and were given free access to food and water. Genomic DNA from mouse ear was isolated using the Omni-Pure Tissue Genomic DNA Purification System (Gene Link, Hawthorne, NY) following the manufacturer's protocol. Mice were genotyped by a multiplex PCR, which was performed in a total of 20 l containing 1 l of DNA, 2 l of 10ϫ reaction buffer (2.5 mM MgCl 2 ), 3.2 l of 1.25 mM dNTPs, 0.5 units of PRIME HotMaster TaqDNA Polymerase (5 PRIME, Hamburg, Germany), and 0.2-0.75 M 6-carboxyfluorescein and NED (2Ј-chloro-5Јfluoro-7Ј,8Ј-fused phenyl-1.4-dichloro-6-carboxyfluorescein) fluorescent-labeled primers (supplemental Table 1). PCR samples were amplified at 94°C (30 s), 60°C (30 s), and 71.5°C (90 s) for 40 cycles in a GeneAmp PCR System 9700, and PCR products were analyzed by capillary electrophoresis in a 3130 Genetic Analyzer (Applied Biosystems). Genotyping was also performed by conventional PCR in a total volume of 20 l containing 1 l of DNA, 2 l of 10ϫ reaction buffer (2.5 mM MgCl 2 ), 3.2 l of 1.25 mM dNTPs, 1 l of unlabeled primers, and 0.5 units of PRIME HotMaster TaqDNA polymerase. PCR samples were amplified as described above, and products were analyzed by electrophoresis in 2.5% LM Sieve-agarose gels and visualized by GelRed TM Nucleic Acid Gel Stain (Biotium, Hayward, CA) using a 100-bp DNA ladder marker (Invitrogen).
The mice used in the study were F5 or subsequent generations of age-matched littermates. To induce obesity, male WT and ⌬adip mice were fed a high fat diet (HFD) (60% Kcal from fat; Research Diets Inc., New Brunswick, Canada) for 12 weeks starting at week 6 of age. The chow group received a standard pelleted chow diet (13% kcal from fat) for 12 weeks. At the end of the study mice were euthanized under ketamine/xylazine (4:1) anesthesia via intraperitoneal injection (Ketolar, Parke Davis-Pfizer, Dublin, Ireland; Rompun, Bayer Leverkusen, Germany). Epididymal WAT (eWAT) and interscapular BAT (iBAT) were collected and snap-frozen in liquid nitrogen for further analysis. All animal studies were conducted in accordance with the criteria of the Investigation and Ethics Committee of Hospital Clinic and the European Community laws governing the use of experimental animals.
Analysis of Eicosanoids by Enzyme Immunoassay (EIA) and LC-Electrospray Ionization-MS/MS-LTB 4 , PGE 2 and 15d-PGJ 2 levels were extracted from eWAT samples (200 mg) from WT and ⌬adip mice. Each sample was individually homogenized in a Pellet pestle, Cordless Motor (Sigma) in 400 l of cold MeOH and kept at Ϫ80°C overnight. Subsequently, homogenates were centrifuged at 2000 rpm for 10 min at 4°C. Supernatants were collected and brought to a final volume of 10 ml with distilled water, transferred into syringes, acidified to pH 3.5, and loaded onto Sep-Pak C 18 Cartridges (Waters, Milford, MA). Materials were eluted with methyl formate, evaporated, and resuspended in EIA buffer. Confirmation of PGE 2 and 15d-PGJ 2 levels and analysis of 6-keto-PGF 1␣ , PGF 2␣ , PGD 2 , PGJ 2 , and thromboxane B 2 concentrations were performed by LCelectrospray ionization-MS/MS analysis. Briefly, 50 mg of frozen adipose tissue were extracted on solid phase columns before injection into an Agilent 1200 HPLC system coupled with an Agilent 6460 Triplequad mass spectrometer with electrospray ionization source. Analysis of lipid mediators was performed with MRM in negative mode.
Isolation of Pre-adipocytes from the eWAT Stromal Vascular Cell (SVC) Fraction-eWAT was excised, weighed, rinsed twice in cold carbogen-gassed Krebs-Ringer supplemented with 1% fatty acid-free BSA and 2 mM EDTA and centrifuged at 500 ϫ g for 5 min at 4°C to remove free erythrocytes and leukocytes. Tissue suspensions (300 -600 mg) were placed in 5 ml of digestion buffer containing Krebs-Ringer supplemented with 1% fatty acid-free BSA and 1 mg/ml collagenase A (Roche Applied Science) and incubated at 37°C for 30 min as described (23). Cell pellets corresponding to the SVC were incubated with erythrocyte lysis buffer (0.15 M NH 4 Cl, 10 mM KHCO 3 , and 0.1 mM EDTA) for 5 min and centrifuged at 500 ϫ g for 5 min. SVC with a predominant population of pre-adipocytes (ϳ72%) were further enriched using the magnetic labeling system Lineage Cell Depletion kit (Miltenyi Biotech, Auburn, CA), which depletes mature hematopoietic cells, such as monocytes/ macrophages, T cells, B cells, granulocytes, and erythrocytes and their committed precursors in the SVC fraction. Pre-adipocyte purity was tested by immunocytochemistry of Pref-1, which is expressed specifically in pre-adipocytes but not in mature adipocytes.
Pref-1 Immunocytochemistry-After cytospin at 500 ϫ g for 5 min, cells on coverslips were fixed with acetone for 10 min at Ϫ20°C followed by incubation with peroxidase blocking solution (S2023, DAKO, Glostrup, Denmark) for 15 min at room temperature to block endogenous peroxidase activity. After washing with DPBS, cells were incubated with blocking serum (Vectastin ABC Kit, Vector Laboratories, Burlingame, CA). Cells were then incubated overnight at 4°C with the primary rat anti-mouse Pref-1 antibody (1/250) followed by incubation for 30 min at room temperature with a biotinylated rabbit anti-rat IgG secondary antibody and incubation with ABC for 30 min at room temperature as described (25). Color was developed using the diaminobenzidine substrate (Roche Applied Science), and cells were counterstained with hematoxylin and mounted with aqueous solution. Cells were visualized at magnification ϫ200 in a Nikon Eclipse E600 microscope (Kawasaki, Japan).
Pre-adipocyte Differentiation into White Adipocytes-Freshly isolated pre-adipocytes were seeded onto 96-well plates (40,000 cells/well) with DMEM supplemented with 10% FBS and 10 ng/ml basic FGF (R&D Systems, Minneapolis, MN) and maintained in a 5% CO 2 atmosphere. Cells were allowed to grow to confluence for 2 days and then exposed to the white adipocyte differentiation mixture (25) (1 g/ml insulin, 0.25 g/ml dexamethasone, 0.

mPGES-1 Controls Pre-adipocyte Differentiation
with isopropyl alcohol for 30 min with shaking to elute the stain, and the optical density was measured at 500 nm in a FluoStar Optima microplate reader (BMG Labtech, Offenburg, Germany). For visualization, cells were counterstained with Gill's hematoxylin, washed with distilled water (4%), and mounted with aqueous solution. Cells were visualized at a magnification of ϫ200 in a Nikon Eclipse E600 microscope.
Differentiation of 3T3-L1 Adipocytes-Mouse 3T3-L1 cells were maintained in DMEM supplemented with 10% (v/v) FBS and 100 units/ml penicillin/streptomycin. Cells were maintained in a humidified atmosphere of 5% CO 2 at 37°C and differentiated into white or beige adipocytes as described before.
siRNA Transfection-Synthetic mPGES-1 siRNA and Universal Scrambled Negative Control (siCON) oligonucleotides were designed and synthesized by OriGene (Rockville, MD). 3T3-L1 cells were transfected at 70% confluence with 10 nM siRNA using Metafectene PRO (Biontex, Martinsried, Germany) at a 1:3 (w/v) in 24-well plates according to the manufacturer's instructions. To assess siRNA efficiency, RNA and protein samples were obtained after 48 h. Three different siRNAs targeting mPGES-1 were screened (data not shown), and the sequence of the most efficient was used. The effects of mPGES-1 siRNA on adipocyte differentiation were assessed 4 days after the addition of the differentiation mixture.
Western Blotting-Total protein from adipose tissue was extracted in lysis buffer containing 50 mM HEPES, 20 mM ␤-glycerophosphate, 2 mM EDTA, 1% Igepal, 10% glycerol, 1 mM MgCl 2, 1 mM CaCl 2 , and 150 mM NaCl, supplemented with protease inhibitor (Complete Mini) and phosphatase inhibitor (PhosSTOP) cocktails. For protein isolation from cell cultures, cells were scraped into ice-cold DPBS and resuspended in 150 l of lysis buffer. Homogenates from tissues were incubated on ice for 15 min and centrifuged at 16,000 ϫ g for 20 min at 4°C. Homogenates from cells were incubated on ice for 10 min and centrifuged at 1000 rpm for 2 min. Supernatants were collected, and COX-1, COX-2, and mPGES-1 protein expression was analyzed by Western blot. A total of 60 g of protein from tissues and 30 g of protein from cells were resuspended in SDS-containing Laemmli sample buffer, heated for 5 min at 95°C, and resolved on 8% (tissue) or 15% (cells) SDS-PAGE. Proteins were electroblotted for 60 -90 min at 100 V at 4°C onto polyvinyl difluoride membranes. Transfer was performed by the iBlot Dry Blotting System (Invitrogen) at 20 V in 7 min. The transfer was visualized by Ponceau S solution. Membranes were then soaked for 1 h at room temperature in TBS (20 mM Tris/HCl, pH 7.4, and 0.5 M NaCl) containing 0.1% (v/v) Tween 20 (0.1% TBS-T) and 5% (w/v) nonfat dry milk. Blots were washed 3 times for 5 min each with 0.1% TBS-T and subsequently incubated overnight at 4°C with primary mouse antimouse COX-1 (dilution 1:300), rabbit anti-mouse COX-2 (dilution 1:1000), and rabbit anti-mouse mPGES-1 (dilution 1:150) antibodies (Cayman Chemicals) and primary anti-mouse GAPDH (dilution 1:1000) (Abcam, Cambridge, UK) in 0.1% TBS-T containing 5% BSA. After washing the blots 3 times for 5 min each with 0.1% TBS-T, membranes were incubated for 1 h at room temperature with an HRP-linked donkey anti-rabbit secondary (dilution 1:2000) antibody (Biolegend, San Diego, CA) in 0.1% TBS-T (for COX-2 and GAPDH) and HRP-linked anti-mouse IgG (dilution 1:2000) (Cell Signaling, Danvers, MA) in 0.1% TBS-T for COX-1. Bands were visualized using the EZ-ECL chemiluminescence detection kit (Biological Industries) in a LAS 4000 imaging system (GE Healthcare) and quantified using Image GE ImageQuant TL analysis software.

RESULTS
Mice lacking PPAR␥ specifically in adipocytes were generated by introducing transgenic Cre recombinase under the control of the aP2 promoter into mice homozygous for the loxPflanked (floxed) allele of PPAR␥. The presence of respective cell-specific Cre recombinase transgenes was confirmed by DNA genotyping, which showed either a 210-bp band corresponding to WT mice carrying the loxP-targeted PPAR␥ gene or 210-and 524-bp bands corresponding to the presence of both floxed alleles and aP2-Cre recombinase (supplemental Fig.  1). Consistent with earlier studies showing that ϳ30% of cells in adipose tissue are nonfat cells (21,28), expression of PPAR␥ isoforms 1 and 2 was reduced in base-line conditions by 47-61% in eWAT from adipocyte-specific PPAR␥ knock-out (⌬adip) mice (Fig. 1A). Because PPAR␥ is a master regulator of adipocyte differentiation, we used the HFD model to induce obesity in WT and ⌬adip mice. As shown in Fig. 1B, ablation of PPAR␥ isoforms in ⌬adip mice under HFD feeding ranged between 63 and 65%. HFD feeding by itself produced a significant reduction in PPAR␥1 but not in PPAR␥2 in both WT and ⌬adip mice (supplemental Fig. 2). Compared with WT mice, ⌬adip mice showed a lesser body weight gain (Fig. 1C) and reduced eWAT weight (Fig. 1D) in response to HFD feeding. Under the chow diet, ⌬adip mice showed similar body weight gain and slightly reduced eWAT weight (Fig. 1, C and D).
PGs are key regulatory components of adipogenesis. The status of the PGE 2 biosynthetic pathway in adipose tissue in the absence of PPAR␥ has not been investigated. Consequently, we next assessed the expression of the key enzymes involved in adipose PGE 2 biosynthesis in WT and ⌬adip mice. COX-2 was found up-regulated at the protein level in adipose tissue from ⌬adip mice under HFD but not in chow diet ( Fig. 2A). No changes in COX-2 expression were detected at the mRNA level in both conditions (data not shown). In contrast, COX-1 was up-regulated in ⌬adip mice under the chow diet but not in HFD (Fig. 2B). Consistently, increased PGE 2 levels were detected in fat tissue from these KO mice either under chow or HFD conditions (Fig. 2C). Further evidence on the relationship between PGE 2 and PPAR␥ was obtained in ex vivo experiments in fat pads. First, fat explants from ⌬adip mice displayed a remarkable induction in COX-2 expression (Fig. 2D). Second, the addition of exogenous PGE 2 up-regulated COX-2 in fat explants from both WT and ⌬adip mice (Fig. 2E), suggesting the existence of a PPAR␥-independent positive regulatory feedback. Third, fat explants from ⌬adip mice also displayed mPGES-1 up-regula-FIGURE 1. Effects of targeted deletion of PPAR␥ in adipocytes. A, adipose mRNA expression for PPAR␥ isoforms 1 and 2 assessed by real-time PCR in WT (n ϭ 10) and aP2-Cre driven PPAR␥ KO (⌬adip) (n ϭ 9) mice under chow conditions. B, expression for PPAR␥ isoforms 1 and 2 in eWAT from WT (n ϭ 6) and ⌬adip (n ϭ 6) mice after 12 weeks of HFD (60% kcal from fat) feeding. C, body weight in WT (n ϭ 5) and ⌬adip (n ϭ 4) mice receiving a chow diet and in WT (n ϭ 13) and ⌬adip (n ϭ 11) mice during the obesity-induced model of HFD feeding for 12 weeks. D, eWAT weight in WT (n ϭ 18) and ⌬adip (n ϭ 15) mice under Chow and HFD conditions. E, representative photomicrographs of SVC fraction and pre-adipocyte cultures stained with Diff-quick or immunostained with Pref-1 specific antibody (ϫ200 magnification). F, PPAR␥ expression in pre-adipocytes from WT (n ϭ 3) and ⌬adip (n ϭ 3) mice. G, representative bright field images of pre-adipocytes from WT and ⌬adip mice incubated with rosiglitazone (Rosi, 1 M) or vehicle (Veh). Absorbance values from Oil Red-O staining of rosiglitazone-induced pre-adipocyte differentiation toward white adipocytes are shown on the right. H, mitochondrial UCP1 expression in eWAT and iBAT from WT (n ϭ 17) and ⌬adip (n ϭ 11) mice. I, UCP1 and PGC-1␣ expression (fold change versus eWAT for each condition and gene studied) in eWAT and iBAT from WT mice under Chow and HFD conditions. Results are expressed as the mean Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 versus WT mice. #, p Ͻ 0.05 versus Veh. a, p Ͻ 0.05 versus chow.

JOURNAL OF BIOLOGICAL CHEMISTRY 28235
ionization-MS/MS and found increased in ⌬adip mice was PGF 2␣ (supplemental Table 2).
To elucidate the ability of COX-derived products to regulate the adipogenic program, we next assessed the effects of PGE 2 on pre-adipocyte differentiation. Pre-adipocytes isolated from eWAT were primed with the differentiation-inducing mixture containing insulin, dexamethasone, isobutylmethylxanthine, and rosiglitazone in the presence of PGE 2 or vehicle, and the differentiation toward white adipocytes was monitored by Oil Red-O staining. As shown in Fig. 3A, PGE 2 significantly reduced rosiglitazone-induced white adipocyte differentiation. PGE 2 did not modify the Oil Red-O signal in the absence of rosiglitazone in the differentiation mixture (Fig. 3A, inset). Conversely, PGE 2 stimulated pre-adipocytes incubated with the brown adipocyte differentiation mixture containing insulin, dexamethasone, isobutylmethylxanthine, T3, and rosiglitazone

mPGES-1 Controls Pre-adipocyte Differentiation
to induce the development of beige/brite cells, as monitored by MitoTracker Red CMXRos (Fig. 3, B-D). Images from these experiments are shown at a lower magnification in supplemental Fig. 3A. Interestingly, a direct stimulatory effect of PGE 2 on the brown marker UCP1 was observed in fat pads from WT mice (Fig. 3E). In contrast, PGE 2 was unable to stimulate the browning of pre-adipocytes from ⌬adip mice lacking PPAR␥ (Fig. 3E). A suppression of both PPAR␥ 1 and 2 isoforms was observed in WT adipose tissue explants incubated with PGE 2 , suggesting a negative regulation of PPAR␥ by this eicosanoid (Fig. 3F).
Consistent with the stimulatory role of exogenous PGE 2 on pre-adipocyte differentiation toward beige/brite adipocytes, inhibition of endogenous PGE 2 synthesis was associated with pre-adipocyte commitment toward the WAT phenotype (Fig.  4A). The capability of inhibitors of PGE 2 synthesis to induce the  SEPTEMBER 27, 2013 • VOLUME 288 • NUMBER 39

mPGES-1 Controls Pre-adipocyte Differentiation
WAT program was more evident in the presence of rosiglitazone in the differentiation mixture (Fig. 4A). In these experiments, SC-560, a selective COX-1 inhibitor, produced similar effects to those of SC-58635, a selective COX-2 inhibitor (Fig.  4A). By contrast, in the absence of rosiglitazone, the benzo[g]indol-3-carboxylate compound, which is a potent inhibitor of mPGES-1 without significant activity on COX-1/COX-2 (29), exerted a more efficacious stimulatory effect on WAT development (Fig. 4A). Importantly, the mPGES-1 inhibitor significantly decreased pre-adipocyte differentiation into beige/brite cells induced by the differentiation mixture (Fig.  4B). Images from these experiments are shown at a lower magnification in supplemental Fig. 3B. To discard potential offtarget actions of the mPGES-1 inhibitor, we performed knockdown experiments using a siRNA directed against mPGES-1. Because primary pre-adipocytes are difficult to transfect, knockdown experiments were performed in the 3T3-L1 adipogenic cell line, the most studied cell line used for adipocyte differentiation (30). As shown in Fig. 4C, both mRNA and protein expression for mPGES-1 as measured by real-time PCR and Western blot, respectively, were significantly reduced 48 h after transfection. Interestingly, mPGES-1 gene silencing reduced the expression of markers of pre-adipocyte browning (i.e. UCP1, Cidea, and PGC-1␣) to a similar extent as that of pharmacological mPGES-1 inhibition (Fig. 4, D-G). Of note, genetic or pharmacological inhibition of mPGES-1 reduced the expression of the brown adipose determination factor PRDM16 (Fig. 4E). Consistent with findings described in Fig. 4B, pharmacological inhibition of either COX-1 or COX-2 did not translate into changes in adipocyte browning (Fig. 4H).
Consistent with previous publications, the mPGES-1 inhibitor benzo[g]indol-3-carboxylate blocked PGE 2 formation by ϳ55% (Fig. 5A) (29). In addition, confirming the existence of a negative regulatory loop between PGE 2 and PPAR␥, fat explants incubated with this mPGES-1 inhibitor displayed enhanced PPAR␥ expression (Fig. 5B). Given that changes in pre-adipocyte browning were only seen with the mPGES-1 inhibitor despite similar suppression of PGE 2 levels with selective COX-1 and COX-2 inhibitors, we next explored the pharmacological properties of this compound on other eicosanoid-generating pathways. In this regard, this benzo[g]indol-3-carboxylate compound has been shown to suppress leukotriene biosynthesis by inhibiting 5-LO (31). Accordingly, we assessed the effects of this compound on LTB 4 formation by adipose tissue. As shown in Fig. 5, C and D, this compound inhibited LTB 4 levels and 5-LO expression in a concentrationdependent manner in fat explants from WT mice. Noteworthy, ⌬adip mice exhibited increased adipose LTB 4 levels accompanied by reduced 5-LO expression (Fig. 5E). The presence of augmented LTB 4 levels in ⌬adip mice was likely related to a parallel increase in FLAP expression (Fig. 5E). When tested on ⌬adip mice, the benzo[g]indol-3-carboxylate compound also produced concentration-dependent inhibitory actions on LTB 4 levels and 5-LO expression (Fig. 5, F and G). The participation of LTB 4 in the browning process was excluded by demonstrating the lack of effect of this 5-LO product on the expression of UCP1, Cidea and PGC-1␣ in WAT pre-adipocytes (Fig. 5H).
Because in addition to adipogenesis, PPAR␥ also plays a role in regulating inflammatory response, we finally assessed the consequences of lacking PPAR␥ on TNF␣ expression in fat explants. As shown in Fig. 6A, ⌬adip mice exhibited increased expression of TNF␣ as compared with tissue explants from WT mice. This result was consistent with the observation that rosiglitazone was able to down-regulate the expression of this adipokine (Fig. 6B). Given that mPGES-1 products are well

mPGES-1 Controls Pre-adipocyte Differentiation
known lipid mediators participating in inflammatory response (13), we also assessed the consequences of mPGES-1 inhibition on adipose TNF␣ expression. In fat pads from both WT and ⌬adip mice, TNF␣ was down-regulated in a concentration-dependent manner by the pharmacological mPGES-1 inhibitor (Fig. 6C). This finding was reproduced in knockdown experiments using a siRNA directed against mPGES-1 (Fig. 6C, inset). Similar to TNF␣, ⌬adip mice exhibited increased expression of IL-6 and MCP-1 as compared with tissue explants from WT mice (Fig. 6, D and E). These two inflammatory adipokines were also blocked by different selective COX and 5-LO inhibitors (Fig. 6F).

DISCUSSION
PPAR␥ plays a unique regulatory role in adipose tissue homeostasis by exerting strict control of the adipogenic process. Here we present evidence that mice lacking PPAR␥ specifically in the adipose tissue are resistant to HFD-induced eWAT formation by mechanisms linked to PGE 2 biosynthesis. Indeed, our study demonstrates the induction of mPGES-1 expression and augmented PGE 2 levels in eWAT from aP2-Cre PPAR␥deficient mice. This mPGES-1-derived product is able to divert pre-adipocyte differentiation in WAT from white adipocytes to beige/brite mature adipocytes accompanied by up-regulation of UCP1. A proof of concept of the mPGES-1 role was obtained by inhibiting either the expression or the activity of this terminal enzyme. In particular, the addition of a selective pharmacological mPGES-1 inhibitor as well as a siRNA directed against mPGES-1 to adipogenic precursors resulted in the reduction of browning markers (i.e. UCP1, Cidea, and PGC-1␣) and browning determination factors (i.e. PRDM16). Together, these findings are relevant in terms of energy homeostasis because the engagement of beige/brite adipocytes and the induction of a thermogenic program in WAT depots are able to waste the surplus of energy through increased heat production, which ultimately exerts protection against obesity and obesity-related co-morbidities.
Previous studies indicate that beige/brite cells emerging in WAT depots come from a completely different cell lineage than those in the classical BAT depots (6). To date, the beige/brite cells in WAT depots have been described upon chronic cold exposure or in response to ␤-adrenergic or PPAR␥ agonists (1,5,6). Our study is the first investigation providing evidence that exposure of pre-adipocytes of WAT origin to the eicosanoid PGE 2 results in a browning effect during the adipocyte differentiation process. These findings add value to the previous observation that overexpression of COX-2, the first upstream enzyme in the PG biosynthetic cascade, was associated with de novo recruitment of brown adipocytes in WAT (18). Similarly, a previous study using COX-2-deficient mice reported that induction of UCP1 expression in WAT is dependent on COX activity (32). However, these studies left unanswered the relative role of PGs in beige fat development and whether WAT beige cells come from progenitor cells pushed to develop into beige/brown fat by PGs or whether white adipocytes could directly convert into beige cells. Moreover, these previous studies pose the limitation that in addition to PGE 2 , COX-2 activity gives rise to other PGs, including PGI 2 and specially PGD 2 . PGD 2 undergoes spontaneous dehydration to a number of derivatives including 15d-PGJ 2 , which is a potent ligand for the activation of PPAR␥, although its function in adipogenesis is still unproven (33). Our study obviates these limitations by providing evidence of the direct involvement of a COX-2 downstream terminal synthase responsible for the biosynthesis of PGE 2 (i.e. mPGES-1), which drives beige/brite fat development  SEPTEMBER 27, 2013 • VOLUME 288 • NUMBER 39 in WAT by direct stimulation of the conversion of pre-adipocytes into beige/brite adipocytes.

mPGES-1 Controls Pre-adipocyte Differentiation
An interesting aspect of our study is that we provide evidence of the existence of a coordinate functional regulation of brownin-white adipogenesis by mPGES-1 and PPAR␥. Indeed, our study presents a number of findings supporting a negative regulation between the nuclear receptor PPAR␥ and the PGE 2 biosynthetic pathway (Fig. 7). On one hand, mice deficient in PPAR␥ specifically in adipose tissue showed increased expression of COX-2 and mPGES-1 and augmented PGE 2 levels. We noted that elevations in COX-2 expression were more evident in ex vivo experiments (fat explants cultured in the absence of circulatory factors) than in vivo. This is probably related to the existence of complex mechanisms regulating COX-2 in vivo that may mask the effects of PPAR␥ deletion on adipose tissue COX-2 expression (34). Indeed, a number of circulating factors has been reported to exert opposite and counter-regulatory roles on adipose tissue COX-2 expression, including angiotensin II and glucocorticoids (down-regulating) and TNF␣ and IL-1␤ (up-regulating) (35,36). On the other hand, exogenous PGE 2 was able to suppress PPAR␥ expression, whereas opposite effects were seen after inhibition of endogenous PGE 2 biosynthesis in adipose tissue from WT mice. The existence of this negative regulatory loop is consistent with previous studies showing that pre-adipocytes stably transfected with either COX-1 or COX-2 had lower PPAR␥ expression (16) and that mice genetically deficient for mPGES-1 had basal elevations in PPAR␥ expression and transcriptional activity (17). Moreover, our results are suggestive of the existence of a positive feedback loop between PGE 2 and local COX-2 and mPGES-1 expression in adipose tissue (Fig. 7). This finding is consistent with a previously described positive feedback loop between PGE 2 and mPGES-1 in the porcine endometrium (37) and between PGE 2 and COX-2 in abdominal WAT (38).
Our findings indicate that the end result of the coordinate functional regulation between COX-2/mPGES-1/PGE 2 and the nuclear receptor PPAR␥ is the blockage of differentiation of pre-adipocytes into mature white adipocytes and the concomitant rise of beige/brite adipocytes. The mechanisms by which the COX-2/mPGES-1/PGE 2 axis and the nuclear receptor PPAR␥ interact during the process of adipogenesis are not completely delineated but might be related to the ability of PGE 2 to increase intracellular cAMP, which is a well known mediator of the induction of "brown fat-like" cells residing in the WAT (7,8). Another mechanism potentially implicated is the stabilization of several dominant transcriptional regulators of white-to-brown adipocyte development and function, including PRDM16 (9,11,38). In fact, our results demonstrate a reduction in PRDM16 in adipocytes incubated with an inhibitor of mPGES-1 activity or transfected with a siRNA that induce gene silencing of this terminal PG synthase. In any event, our findings indicate that this process requires the integrity of both systems because PGE 2 is not able to inhibit white adipocyte differentiation in the absence of a PPAR␥ agonist. Moreover, PGE 2 is able to directly induce the browning of WAT (i.e. UCP1) in WT mice but not in mice lacking PPAR␥ specifically in adipocytes. Finally, the existence of a loop between PGE 2 and PPAR␥ in the regulation of adipocyte browning appears not to be a generalized event to all fat depots. In this regard, ⌬adip mice showed more UCP1 expression in eWAT and less expression in iBAT than WT animals. Although the mechanisms underlying this observation are unknown, a remarkable reduction of UCP1 expression in interscapular, but not in perigonadal adipose tissue has been previously described in ⌬adip mice (39).
In addition to regulating adipogenesis, PGs are well known lipid mediators that reproduce the cardinal signs of inflammation (40). PGs are formed through the sequential actions of COXs and a variety of terminal PG synthases, which are expressed with some tissue specificity (41). Among the different PGs, PGE 2 is the most abundant and closely associated with inflammatory conditions and is synthesized by the concerted action of COXs and PGE synthases. In particular, mPGES-1 is preferentially linked with inducible COX-2, which is then coupled with inflammatory conditions in a variety of cells and tissues (13). On the other hand, PPAR␥ is a ligand-activated transcription factor implicated in the regulation of inflammatory responses and a nuclear factor assisting the resolution of inflammation (42). Therefore, it was pertinent in our study to assess the inflammatory status in adipose tissue from mice specifically lacking PPAR␥ in this tissue. Our findings showing increased proinflammatory adipokine expression (i.e. TNF-␣, IL-6, and MCP-1) in fat from PPAR␥ knock-out mice further support the previously described anti-inflammatory role of this nuclear receptor (42). Our findings also show that either pharmacological or genetic inhibition of mPGES-1 results in a reduction of TNF-␣ expression in adipose tissue from WT mice, a response that was not significantly different from that seen in PPAR␥ KO mice. This observation suggests that contrary to the adipogenic process, there is no coordinate functional regulation between mPGES-1 and PPAR␥ in the control of adipose tissue inflammation. Considering the duality of FIGURE 7. Schematic diagram of the proposed coordinated functional regulation of mPGES-1 and PPAR␥ in beige/brite adipogenesis. PPAR␥ deletion in WAT uncovers a negative regulation of COX-2 and mPGES-1 expression by this nuclear receptor. In addition, PGE 2 is able to down-regulate PPAR␥ expression in WAT. Furthermore, in this tissue the interaction between PGE 2 and PPAR␥ has the ability to divert pre-adipocyte differentiation from white into beige/brite adipocytes. Interestingly, PGE 2 exerts a PPAR␥-independent positive feedback loop with local COX-2 and mPGES-1 expression in adipose tissue. PGE 2 in WAT where it promotes the browning and heat dissipating function of adipocytes in parallel with a causative responsibility in adipose inflammation, further studies are needed to fully elucidate the overall role of the COX-2/mPGES-1/PGE 2 pathway on the insulin axis and the development or prevention of obesity-related co-morbidities.

mPGES-1 Controls Pre-adipocyte Differentiation
In conclusion, the present study demonstrates the existence of a functional interaction between PPAR␥ and mPGES-1 in the process of adipogenesis, especially in the formation of beige/ brite adipocytes from WAT pre-adipocytes. The PGE 2 -mediated actions on pre-adipocyte differentiation are not mirrored by other lipid mediators such as LTB 4 . Finally, our findings suggest that the coordinate functional interaction between PPAR␥ and mPGES-1 does not appear to contribute to the regulation of the unbalanced inflammatory tone associated with obesity.