Regulated Production of a Peroxisome Proliferator-activated Receptor-γ Ligand during an Early Phase of Adipocyte Differentiation in 3T3-L1 Adipocytes*

Peroxisome proliferator-activated receptor-γ (PPARγ) is a nuclear hormone receptor that is critical for adipogenesis and insulin sensitivity. Ligands for PPARγ include some polyunsaturated fatty acids and prostanoids and the synthetic high affinity antidiabetic agents thiazolidinediones. However, the identity of a biologically relevant endogenous PPARγ ligand is unknown, and limited insight exists into the factors that may regulate production of endogenous PPARγ ligands during adipocyte development. To address this question, we created a line of 3T3-L1 preadipocytes that carry a β-galactosidase-based PPARγ ligand-sensing vector system. In this system, induction of adipogenesis resulted in elevated β-galactosidase activity that signifies activation of PPARγ via its ligand-binding domain (LBD) and suggests generation and/or accumulation of a ligand moiety. The putative endogenous ligand appeared early in adipogenesis in response to increases in cAMP, accumulated in the medium, and dissipated later in adipogenesis. Organically extracted and high pressure liquid chromatography-fractionated conditioned media from differentiating cells, but not from mature adipocytes, were enriched in this activity. One or more components within the organic extract activated PPARγ through interaction with its LBD, induced lipid accumulation in 3T3-L1 cells as efficiently as the differentiation mixture, and competed for binding of rosiglitazone to the LBD of PPARγ. The active species appears to be different from other PPARγ ligands identified previously. Our findings suggest that a novel biologically relevant PPARγ ligand is transiently produced in 3T3-L1 cells during adipogenesis.

with the retinoid X receptor (RXR) (1) and binds to direct repeat 1-type motifs found in the promoter sites of target genes (2). PPAR␥ is a major modulator of several aspects of development and homeostasis. It is expressed in breast, colon, prostate, macrophages, and adipose tissue (3)(4)(5)(6) and has been shown to play a critical role in glucose and lipid metabolism (7), macrophage function (8), and adipogenesis (9).
As a member of the nuclear receptor superfamily, PPAR␥ is activated through ligand binding, which results in allosteric changes in receptor conformation, recruitment of coactivators, assembly of a transcriptional complex, and regulated transcription of target genes (10). Among known PPAR␥ agonists are the synthetic high affinity antidiabetic drugs thiazolidinediones (11) and a number of natural substances. These include the dehydration product of prostaglandin D 2 , 15-deoxy-⌬ 12,14 -prostaglandin J 2 (15-dPGJ 2 ) (12,13); derivatives of linoleate, 9-and 13-hydroxyoctadeca-9Z,11E-dienoic acids, both found in oxidized low density lipoprotein (14); certain polyunsaturated fatty acids (15,16); and oxidized alkyl phospholipids such as lysophosphatidic acid (17). Most of these natural ligands bind PPAR␥ with relatively low affinity compared with the affinity of most bona fide ligands for nuclear receptors. Additionally, they are weak agonists and exist at concentrations in vivo that may be below the levels needed for biological relevance. For these reasons, their physiological significance and role as true endogenous activators for PPAR␥ have been uncertain.
Adipogenesis is a complex process associated with coordinated changes in gene expression, cell morphology, and hormone sensitivity (18). Several transcription factors influence these processes, among which PPAR␥ and CCAAT enhancerbinding proteins (C/EBPs) have been extensively studied (19,20). Early gain-of-function studies showed that forced expression of PPAR␥ in cultured fibroblasts followed by treatment with exogenous ligands results in their efficient differentiation to mature adipocytes (9). Subsequent loss-of-function studies showed that PPAR␥ is required for adipogenesis in vivo and in vitro (21)(22)(23). Despite conclusive knowledge about the pivotal role that PPAR␥ plays in adipogenesis, surprisingly little is known about the production, regulation, and identity of endogenous ligands that activate PPAR␥ in specific cell types. It is possible, of course, that the receptor may respond solely to one or more nutritionally supplied lipid ligands and that no high affinity regulatory ligand is produced during adipogenesis.
Even so, it would be critical to know the net effective concentration of intracellular ligands under a variety of situations and in cells such as adipocytes, where PPAR␥ activation plays a major role. To address this fundamental question, we have developed an autoregulated inducible reporter system to monitor production of PPAR␥ ligand activity during adipogenesis in vitro. Here, we present evidence that 3T3-L1 cells produce and release into the medium a hydrophobic ligand(s) specific for PPAR␥. This activity is most abundant in the early stages of differentiation, is produced in response to a cAMP signal, and does not appear to be a previously described agonist. Our findings suggest that an as yet unidentified pathway for PPAR␥ ligand generation exists in 3T3-L1 cells. Activation of this pathway results in production of an endogenous PPAR␥ ligand that, along with the receptor itself, is likely to be a critical mediator of the adipogenic program.
In Vitro Interactions-GST protein purification and pull-down assays have been described previously (25,26). Briefly, for the pull-down assay, GST-PPAR␥ LBD was bound to glutathione-Sepharose beads and incubated with 35 S-labeled TIF2 RID for 20 min on ice. Then, 1 M rosiglitazone and increasing amounts (0.25-1 l) of organically extracted and concentrated control (C100ϫ) or conditioned (CM100ϫ) medium were added, and the reactions were incubated overnight at 4°C. The next day, the beads were washed, and the bound proteins were eluted with 15 mM glutathione, resolved by SDS-PAGE, and visualized by autoradiography. For the protease digestion assay, 2 l of in vitro translated and 35 S-labeled PPAR␥ was incubated with 2 l of in vitro translated RXR. The complex was allowed to dimerize for 10 min on ice before EtOH, 1 M rosiglitazone, or increasing amounts of C100ϫ or CM100ϫ were added. After a 30-min incubation, 250 ng of trypsin was added to the indicated tubes, and protein digestion was performed at room temperature for 7 min. Reactions were analyzed on a 15% SDS gel, after which the gel was dried and exposed to film. For the ligand binding competition assay, 100 g of GST-PPAR␥ fusion protein was bound to glutathione-Sepharose beads and incubated at 4°C in buffer containing 10 mM Tris (pH 8.0), 50 mM KCl, 10 mM dithiothreitol, and 1-500 nM [ 3 H]rosiglitazone (specific activity of 50 Ci/mmol; American Radiolabeled Chemicals, St. Louis, MO) in the presence or absence of unlabeled rosiglitazone, control medium, conditioned medium, or medium from mature adipocytes (day 8 medium). At the end of the incubation, the protein pellets were washed extensively with buffer containing 25 mM Tris (pH 7.5), 75 mM KCl, 15% glycerol, 0.05% Triton X-100, and 0.5 mM EDTA to separate bound from free radioactivity. Bound radioactivity was quantitated by liquid scintillation counting.
Cell Culture-3T3-L1 preadipocytes (American Type Culture Collection, Manassas, VA) were plated in 12-well dishes and cultured in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) containing 10% calf serum. Differentiation was induced as described previously (27). Briefly, at day 0 (2 days after confluence), the medium was changed to DMEM supplemented with 10% charcoal dextran-stripped FBS, 1 M dexamethasone, 0.5 mM MIX, and 5 g/ml insulin. 2 days later, the medium was replaced with DMEM supplemented with 10% charcoal dextran-stripped FBS and only 5 g/ml insulin for an additional 2 days and then changed to DMEM supplemented with 10% charcoal dextran-stripped FBS without insulin. The medium was replenished every 2 days. At day 10, the lipid content was determined by staining the cells with oil red O. The stable cell lines of 3T3-L1 preadipocytes expressing either MSV␥ or C-MSV were established by cotransfection of each vector with pWL-neo using a mammalian transfection kit from Stratagene. G418-resistant clones were selected and tested for ␤-galactosidase activity. Positive clones were propagated and maintained in DMEM containing 10% calf serum and 600 g/ml G418.
Transient Transfections and Reporter Assays-CV-1 cells were plated and maintained at 37°C in DMEM with 10% FBS (Hyclone Laboratories) and 100 units/ml penicillin/streptomycin (Invitrogen). For transient transfections, cells were plated in 24-well plates and maintained in DMEM with 10% charcoal dextran-stripped FBS for the duration of the experiment. All transfections were performed using LipofectAMINE 2000 reagent (Invitrogen) according to the manufacturer's instructions. In some sets of experiments, cells were transfected with pCMV-Gal4-PPAR␥ LBD, a pCMV-Gal4-␤Gal as a reporter, and pTK-Luc (Stratagene) as a control for normalization. For mammalian one-hybrid assays, the cells were transfected with pGal4-TK-Luc as a reporter, pCMV-Gal4-PPAR␥ LBD, and pCMV-␤Gal as a control. For mammalian two-hybrid assays, pVP16-TIF2 RID was added to the transfection mixture. The next morning, the medium was changed, and ligands (organically extracted and 3ϫ concentrated 3T3-L1 day 2 control medium (C3ϫ) or 3T3-L1 day 2 conditioned medium (CM3ϫ)) were added and incubated overnight. Cells were assayed for luciferase (Promega, Madison, MI) or ␤-galactosidase (Tropix Inc.) activity 24 h later according to the manufacturers directions. Similar results were obtained from at least four independent experiments. For some experiments, the relative ␤-galactosidase reporter activity was normalized to luciferase control activity, whereas for other experiments, the relative luciferase reporter activity was normalized to the ␤-galactosidase control activity.
Organic Phase Extraction-Conditioned media from 3T3-L1 cells collected at days 0, 2, and 8 after initiation of differentiation and day 2 control medium were acidified with 0.1 N HCl and mixed with ethyl acetate/acetone (1:1, v/v). The mixture was vortexed and centrifuged at 1000 ϫ g for 5 min. The upper phase containing molecules soluble in organic solvents was transferred to a new tube, washed twice with saturated NaCl, and evaporated in a SpeedVac. For transfection assays, evaporated extracts were reconstituted in DMEM containing 10% charcoal dextran-stripped FBS (3ϫ concentrated), and for in vitro interaction assays, extracts were reconstituted in EtOH (100ϫ concentrated).
Total RNA Extraction and Quantitative Reverse Transcription-PCR-RNA from day 4 differentiating 3T3-L1 cells was extracted using an RNeasy mini kit (QIAGEN Inc., Valencia, CA) and treated with DNase I (QIAGEN Inc.) according to the manufacturer's instructions. PPAR␥, aP2, lipoprotein lipase, and GLUT4 mRNAs were quantitatively measured using a Stratagene MX 4000 Multiplex quantitative PCR system and a Stratagene single step quantitative reverse transcription-PCR kit according to the manufacturer's instructions. Total RNA was diluted in the range of 0.1-100 ng and used to establish standard curves for each individual gene. The quantitation of each mRNA was further normalized by the corresponding cyclophilin mRNA measurement.
Prostaglandin J 2 Analysis-Organic extracts of conditioned media were subjected to micro-high performance thin layer chromatography for separation of lipid classes and confronted to the prostaglandin J 2 standard. Samples were evaporated to dryness under nitrogen and resuspended in an equal volume of chloroform/methanol (1:1, v/v). Aliquots were spotted on Whatman Silica Gel 60A plates and subjected to two consecutive developments in chloroform/ethanol/triethylamine/water (30:34:30:8, v/v/v/v) and hexane/ether (100:4.5, v/v). The positions and intensities of bands were determined by copper sulfate charring and spectrodensitometric analysis at 400 nm in a Shimadzu CS-9000U scanner in the reflectance mode.
HPLC Fractionation-Organically extracted conditioned medium was vacuum-dried using an ATR vacuum centrifuge (Appropriate Technical Resources, Inc., Laurel, MD), resuspended in chloroform/methanol (2:1, v/v), and microdispersed in a Branson sonication bath. 50-l aliquots were injected in a Waters HPLC using a Lichrosorb normal-phase column and a binary solvent system. Solvent A consisted of 100% chloroform, and solvent B was a mixture of chloroform/methanol/water (50:44.5:5.5). The separation method included 5 min of isocratic 100% solvent A and a linear gradient from 100% solvent A to 100% solvent B over 35 min, followed by 10 min of isocratic 100% solvent B and reconstitution of the original conditions. The flow rate was 1 ml/min, and fractions were collected on a time basis every 6 min.

Generation of a PPAR␥ Ligand-monitoring System in 3T3-L1
Cells-PPAR␥ is a ligand-regulated nuclear hormone receptor that is indispensable in the process of adipogenesis both in vivo and in vitro (23). It is therefore reasonable to hypothesize that a specific ligand is produced to activate PPAR␥ upon conversion of preadipocytes to mature lipid-containing adipocytes. To monitor the presence of such a ligand during adipogenesis, we developed and stably integrated ligand-inducible MSV␥ in 3T3-L1 cells. The vector carries the Gal4 DNA-binding domain fused to the PPAR␥ LBD, located upstream of the ␤-galactosidase reporter gene. Both the PPAR␥ fusion protein and the ␤-galactosidase reporter are under the control of five Gal4specific binding sites (UAS 5 ) and the hsp68 minimal promoter (MSV␥) (Fig. 1A). The combination of UAS 5 with the hsp68 minimal promoter and a receptor LBD has been successfully used previously in studies on reporter gene constructs both in vitro and in vivo (24). We also generated a second reporter construct containing just the ␤-galactosidase gene driven by the UAS 5 -hsp68 minimal promoter to use independently as a negative control vector (C-MSV) (Fig. 1A). The two vectors were transfected in CV-1 cells to test for their ability to respond to the synthetic PPAR␥ ligand rosiglitazone. As shown in Fig. 1B, the ligand-monitoring system vector MSV␥ responded well to 250 nM rosiglitazone in CV-1 cells, whereas the negative control vector C-MSV did not.
The two vectors were then stably transfected in 3T3-L1 preadipocytes, and two clones were selected for further experiments: the ligand-sensing effector clone 5B2 and the negative control clone UH12. Both cell lines showed normal lipid accumulation after induction of differentiation by standard protocols (Fig. 1C). Line 5B2 expressed PPAR␥ and C/EBP␣ at normal levels compared with the parental 3T3-L1 cells, and insulin-stimulated glucose uptake following induction of differentiation was also normal (data not shown). The data suggest that the 3T3-L1-based monitoring system line retains all the biological properties of adipocytes.
PPAR␥ Ligand Production during Adipogenesis-Next, we tested for increases in ␤-galactosidase activity after induction of differentiation of 5B2 and UH12 cells. Cells were maintained in 10% charcoal dextran-stripped FBS for this and all subsequent experiments. 2 days after confluence (day 0), cells were induced to differentiate using the standard adipogenic mixture of MIX, insulin, and dexamethasone. At the indicated time points, cells were lysed and assayed for ␤-galactosidase activity. As shown in Fig. 2A, no changes in ␤-galactosidase activity were observed in the negative control line UH12 up to 12 days after initiation of differentiation. On the other hand, a robust increase in activity was observed in line 5B2 as early as 24 h after induction (day 1). The activity remained high for another 24 h, gradually decreased to basal levels by day 4, and remained low up to day 12 after differentiation. DNA-binding domain-PPAR␥ LBD fusion driven by five Gal4-specific binding sites (UAS 5 (5xUAS)) linked to the hsp68 minimal promoter was generated. A reporter construct containing the bacterial ␤-galactosidase gene driven by UAS 5 linked to the hsp68 minimal promoter was also generated and either fused together with the effector plasmid to create the PPAR␥-monitoring system vector MSV␥ or used independently as a negative control (C-MSV). B, CV-1 cells were plated in 24-well plates and maintained in DMEM with charcoal dextran-stripped FBS. Cells were transfected with the two MSV plasmids and pTK-Luc as an internal control using LipofectAMINE 2000. The next day, the medium was changed, and cells were treated overnight with solvent (Ϫ) or 250 nM rosiglitazone (Rosi). Cells were then lysed and assayed for ␤-galactosidase and luciferase activities. Normalized activity is presented in arbitrary units. Results represent the average of at least three independent experiments. C, parental 3T3-L1 cells and lines 5B2 and UH12 were induced to differentiate following standard protocols. Cells were fixed and stained with oil red O at day 10 to visualize lipid content. An average area of staining is shown. If a PPAR␥ ligand required for adipogenesis is produced during a specific time window early in adipogenesis, then use of a highly specific PPAR␥ antagonist during the corresponding time frame should effectively block adipogenesis. Indeed, when 3T3-L1 cells were induced to differentiate in the presence of 30 M PD068235, subsequent lipid accumulation was completely blocked. As shown in Fig. 2B, the presence of the antagonist from days 0 to 2 during differentiation effectively blocked adipogenesis, whereas addition of the antagonists after day 2 had no effect. Collectively, these data suggest that generation and/or accumulation of a ligand activity takes place between days 0 and 2 after induction of adipogenesis, the same time frame during which activation of the ligand-sensing system, through the LBD of PPAR␥, was observed.
Next, we designed experiments to assess the potential contribution of molecules present within serum to the activation of the monitoring system vector. For this, 5B2 cells were incubated in serum-free medium containing rosiglitazone, differentiation mixture (DIM), or its different components and then assayed for ␤-galactosidase activity. As shown in Fig. 3A, the ␤-galactosidase activity increased in a serum-independent manner. Treatment with either rosiglitazone or DIM increased activity by 3.5-and 4-fold, respectively. Interestingly, one of the adipogenic mixture components, MIX, had an effect similar to that of the complete mixture because it increased ␤-galactosidase activity 4-fold. MIX is generally assumed to exert its proadipogenic effects by inhibiting phosphodiesterases that break down cAMP, but data exist to suggest that it may work though other mechanisms as well (28). We therefore treated cells with forskolin, which increases cAMP via the activation of adenylate cyclase, or with the non-hydrolyzable analog 8-bromo-cAMP. Forskolin and 8-bromo-cAMP both increased ␤-galactosidase activity to levels similar to those achieved by MIX (Fig. 3B). In addition, transient transfection experiments in CV-1 cells showed that neither the adipogenic mixture nor its components separately or in combination activated the PPAR␥ LBD under conditions in which robust activation was observed by rosiglitazone (data not shown). Of note, previous work has demonstrated a similar requirement of cAMP signaling pathways in the adipogenic process (29). It was shown that preadipocytes exposed to dexamethasone and insulin alone fail to differentiate and that addition of exogenous PPAR␥ ligands bypasses the defect and promotes adipogenesis in the absence of MIX (29). Collectively, these data suggest that production of the PPAR␥ ligand activity requires cAMP signaling pathways, is specific to preadipocytes and not other cell types, and is independent of pre-existing serum-derived activities.
Conditioned Medium from 3T3-L1 Cells Contains a Ligand Activity-Although no molecule has yet been demonstrated to serve as an endogenous PPAR␥ ligand in vivo, several potential biological activators have been identified. These include lipids such as prostaglandins, leukotrienes, hydroxyeicosatetraenoic acids, and polyunsaturated fatty acids (12)(13)(14)30). It has also been reported that ADD1/SREBP1 promotes the production and secretion of a PPAR␥ ligand activity in 3T3-L1 cells (31). We therefore tested the supernatants (conditioned media) of differentiating 3T3-L1 cells for the presence of ligand. 3T3-L1 cells plated on 10-cm dishes were induced to differentiate. 2 days later, the conditioned medium (hereafter called CM) was collected, extracted with organic solvents, and evaporated. Negative control medium from 3T3-L1 cells maintained in DMEM with charcoal dextran-stripped FBS, but without DIM, was also collected after 2 days of incubation and treated similarly. Both the evaporated control and conditioned media were reconstituted in DMEM with charcoal dextran-stripped FBS (3ϫ concentrated) before being testing on the 5B2 and UH12 lines. The two cell lines were incubated with control (C3ϫ) or conditioned (CM3ϫ) medium for 24 h, after which cells were lysed and assayed for ␤-galactosidase activity. Fig. 4A shows that incubation of the 5B2 cells with CM3ϫ resulted in an increase in ␤-galactosidase activity similar to that observed with DIM, whereas incubation with C3ϫ had no effect. In addition, incubation with organically extracted day 8 medium did not increase ␤-galactosidase activity (data not shown). The ␤-galactosidase activity of the control line UH12 remained unchanged. Next, CV-1 cells transiently transfected with C-MSV or MSV␥ were incubated with C3ϫ or CM3ϫ for 16 h. Again, CM3ϫ increased ␤-galactosidase activity by 3-fold, whereas C3ϫ had no effect. This was comparable with the level of activation seen with rosiglitazone, which induced ␤-galactosidase activity by 4-fold (Fig. 4B).
We then tested the concentrated media for their ability to promote lipid accumulation in 3T3-L1 preadipocytes in the presence or absence of the PPAR␥ antagonist PD068235. Cells were plated in 12-well dishes and, 2 days post-confluence, were incubated with the adipogenic mixture (DIM), rosiglitazone, C3ϫ, CM3ϫ, day 4 medium (3ϫ concentrated), day 8 medium (3ϫ concentrated), or differentiation medium (3ϫ concen- trated) as a negative control. The antagonist PD068235 was added as indicated. On day 2, the medium was changed, and insulin was added; on day 4, insulin was withdrawn. Cells were allowed to accumulate lipids for 6 additional days before they were fixed and stained with oil red O. As shown in Fig. 4C, CM3ϫ induced adipogenesis almost as effectively as the standard mixture or rosiglitazone. Cells incubated with control (C3ϫ), day 4 medium (3ϫ concentrated), day 8 medium (3ϫ concentrated), or differentiation medium (3ϫ concentrated) remained fibroblastic. Addition of the PPAR␥ antagonist completely blocked adipogenesis induced by DIM, CM3ϫ, or rosiglitazone.
We also tested the ability of CM3ϫ to induce PPAR␥ target genes involved in the process of adipogenesis. For this, 3T3-L1 preadipocytes were induced to differentiate with DIM, rosiglitazone, or CM3ϫ. Total RNA was collected 4 days later and analyzed for the presence of PPAR␥ and PPAR␥ target gene mRNAs. As shown in Fig. 4D, rosiglitazone strongly induced PPAR␥ mRNA, consistent with the hypothesis that activated PPAR␥ induces expression of C/EBP␣, and C/EBP␣ through a positive feedback loop maintains high expression of PPAR␥. Interestingly, CM3ϫ had an effect on PPAR␥ mRNA similar to that of rosiglitazone, both of which activated much more robustly that DIM. In addition, rosiglitazone and CM3ϫ activated aP2 and lipoprotein lipase to similar levels. These data indicate that an extractable hydrophobic ligand activity is produced during the early phase of adipocyte differentiation and dissipates later on, once the adipogenic program is fully established. This moiety activates the LBD of PPAR␥ in both 3T3-L1 and heterologous cell systems and induces adipogenesis of 3T3-L1 cells, a role consistent with that of a natural PPAR␥ ligand.

The Ligand Activity Is Distinct from That of Known PPAR␥
Ligands-Several naturally found ligands have been shown to activate PPAR␥ at high concentrations. These include components of oxidized low density lipoprotein (14), LPA (17), the prostaglandin D 2 dehydration product (15-dPGJ 2 ) (12, 13), and products of the 12/15-lipoxygenase-type enzymes (30). We therefore employed several strategies to examine our CM for the presence of previously known PPAR␥ ligands.
The enzyme cyclooxygenase-2 converts arachidonic acid to prostaglandins. One of its products is 15-dPGJ 2 , a PPAR␥ ligand that can promote adipocyte differentiation. Using TLC analysis, we did not find any significant amounts of prostaglandin J 2 in the CM or cell extracts at different time points during differentiation (Fig. 5A) (data not shown). This suggests that the ligand activity in the CM is not 15-dPGJ 2 .
The 12/15-lipoxygenase enzymes found in platelets, leukocytes, and macrophages produce PPAR␥ ligands from arachidonic and linoleic acids. It was recently reported that inhibitors of such enzymes block adipogenesis (32). More specifically, high concentrations of the 12-lipoxygenase inhibitor baicalein (20 -30 M) blocked lipid accumulation in 3T3-L1 cells. It was suggested that baicalein may act by inhibiting production of a putative PPAR␥ ligand (32). We tested the inhibitor for its ability to block activation of the ligand-sensing system in line 5B2 upon incubation with DIM. Not only was baicalein at 10, 20, and 40 M unable to inhibit induction of ␤-galactosidase activity (Fig. 5B), but it resulted in a modest enhancement of activation. This result, together with the reported absence of 12-lipoxygenase enzymes in 3T3-L1 cells (32), suggests that further examination is required to clarify the effect of lipoxygenase enzyme inhibitors on the adipogenic process.
Oleoyl-LPA was recently shown to activate PPAR␥ directly in RAW264.7 monocytic cells (17). This LPA species is also a potent paracrine mediator of adipocyte growth and function (33). We tested five different derivatives of LPA for their ability to activate the PPAR␥ ligand-monitoring system vector in 5B2 cells or to induce lipid accumulation in 3T3-L1 cells. We failed to see increases in ␤-galactosidase activity when 5B2 cells were incubated with the LPA derivatives for 24 h (Fig. 5C). In addition, the compounds did not promote lipid accumulation in confluent 3T3-L1 cells, whereas incubation with rosiglitazone during the same time frame resulted in significant lipid accumulation (data not shown). Our results suggest that, unlike the reported agonist-type function of LPA on PPAR␥ in RAW264.7 cells, a possible role of LPA as a mediator in adipogenesis of 3T3-L1 cells may be independent of PPAR␥.
Specificity of the Ligand Activity-To determine the specificity of the CM ligand activity among different members of the nuclear receptor superfamily, CV-1 cells were transfected with plasmids expressing various Gal4 DNA-binding domain-receptor LBD fusion proteins. The receptors tested included the PPAR␥ heterodimeric partner RXR␣, PPAR␣, estrogen receptor-␣, thyroid receptor-␤, and retinoic acid receptor-␣. Fig. 6 shows that the extracted and concentrated CM activated only PPAR␥ and not several other nuclear receptors under conditions in which robust activation of the receptors by their respective ligands was observed. This suggests that the ligand activity present in the CM is specific for PPAR␥.
Characterization of the PPAR␥ Ligand Activity-Since ligand-induced transactivation is achieved by the recruitment of coactivators to nuclear receptors, two sets of experiments were performed to determine the ability of the CM to recruit coactivators to the LBD of PPAR␥. Initially, a mammalian twohybrid assay was performed. For this, CV-1 cells were transfected with the pGal4-TIF2 RID and pVP16-PPAR␥ LBD expression plasmids in the presence of solvent, rosiglitazone, or the CM. Incubation of the cells with the CM resulted in enhanced interaction between the TIF2 RID and the PPAR␥ LBD to levels similar to those caused by rosiglitazone (Fig. 7A). Similar results were observed when the RID of the coactivator SRC-1 was used (data not shown). Next, we performed a coactivator-dependent receptor ligand assay. For this, the bacterially expressed PPAR␥ LBD was incubated with the in vitro translated and 35 S-labeled TIF2 RID in the presence of solvent, rosiglitazone, the control medium, or the CM. Again, incubation with the CM enhanced the interaction between the coactivator and the receptor as effectively as rosiglitazone (Fig. 7B).
We further assessed the direct interaction of the ligand moiety with the LBD of the receptor by testing the ability of the CM to protect the PPAR␥ LBD from digestion by proteases. As shown in Fig. 7C, incubation of 35 S-labeled PPAR␥ with trypsin  8 -10), whereas incubation with control medium failed to protect from degradation.
We then performed a "ligand depletion assay" using day 2 CM. The GST-PPAR␥ LBD fusion protein or equimolar amounts of control GST protein, both coupled to GST beads, were incubated with day 2 CM. The next day, the supernatant CM was separated from the beads, organically extracted, evaporated, and 3ϫ concentrated. If a specific PPAR␥ ligand activity is present, then only incubation with the PPAR␥ LBD should deplete the activity from the medium. This was indeed the case. As shown in Fig. 8A, GST-PPAR␥ LBD effectively depleted the activity from the CM. The resultant ligand-depleted CM failed to induce lipid accumulation in 3T3-L1 cells (Fig. 8A). In contrast, incubation of the CM with GST did not remove the ligand activity, and this CM promoted adipogenesis as effectively as DIM. Similarly, the ␤-galactosidase activity of 5B2 cells was increased after overnight incubation with the GST-depleted (but not PPAR␥ LBD-depleted) CM (Fig. 8B). Finally, control medium, day 2 medium, or day 8 medium was tested in a ligand binding competition assay. As shown in Fig.  8C, day 2 medium effectively competed for binding of [ 3 H]rosiglitazone to the LBD of PPAR␥, whereas control or day 8 medium did not. Collectively, these data demonstrate the presence of a bona fide PPAR␥ ligand in day 2 CM from 3T3-L1 cells.
Analysis of HPLC-fractionated Conditioned Media-We next attempted to purify the ligand activity further and to test for its presence in media from preadipocytes (day 0) and mature adipocytes (day 8). For this purpose, 12 ml of CM from day 0, 2, or 8 differentiating 3T3-L1 cells was organically extracted, evaporated, resuspended in 50 l of chloroform/methanol, and fractionated by HPLC. Fractions were collected and analyzed for their ability to induce adipogenesis in 3T3-L1 cells. Fig. 9A shows that fraction C derived from day 2 CM potently promoted lipid accumulation in 3T3-L1 cells. None of the remaining day 2 CM-derived fractions or any fractions from day 0 or 8 of differentiation were able to induce adipogenesis. Following this observation, we tested for the ability of the day 2 CMderived fractions to activate the LBD of PPAR␥ in CV-1 cells.

FIG. 7. The CM contains activities that interact with the PPAR␥ LBD.
A, CV-1 cells were cotransfected with the pVP16-PPAR␥ LBD expression vector and a vector expressing a Gal4 fusion of the coactivator TIF2 RID. After transfection, cells were incubated with solvent (Ϫ), 250 nM rosiglitazone (Rosi), or CM3ϫ (day 2 CM (CMD2)) and, 24 h later, were harvested for luciferase (Luc) and ␤-galactosidase assays. B, 35 S-labeled TIF2 was incubated overnight with bacterially expressed GST-PPAR␥ fusion protein in the presence of solvent (Ϫ), 1 M rosiglitazone (Rosi), or CM100ϫ. The next day, the reaction mixtures were washed, and bound proteins were eluted with glutathione, boiled, and subjected to SDS-PAGE and autoradiography. C, 35 S-labeled PPAR␥ was incubated with trypsin in the presence of solvent (Ϫ), rosiglitazone (R), or increasing amounts of C100ϫ or CM100ϫ (0.2, 0.6, and 0.8 l) for 7 min at room temperature. At the end of the incubation period, the reactions were subjected to SDS-PAGE, followed by autoradiography. The asterisks denote protected bands. In, input.
FIG. 8. The CM contains activities that bind to the PPAR␥ LBD. Bacterially expressed GST and GST-PPAR␥ proteins were incubated overnight with the CM. The next day, the supernatant was organically extracted, evaporated, and resuspended in medium with 10% charcoal dextran-stripped FBS. It was then tested for induction of adipogenesis (A) and activation of the PPAR␥ LBD in 5B2 cells (B). Control (C), day 2, and day 8 media were tested for their ability to compete for binding of [ 3 H]rosiglitazone (R) to the PPAR␥ LBD in a ligand binding competition assay (C).
Indeed, only fraction C activated Gal4-PPAR␥ LBD by 4-fold (Fig. 9B). These results show that a PPAR␥ ligand activity appeared early in adipogenesis and accumulated in the media. This activity was organically extractable and could be concentrated following HPLC fractionation. The medium from preadipocytes or, most interestingly, from mature adipocytes did not contain significant PPAR␥ ligand activities as detected by these assays. DISCUSSION Since its discovery almost 10 years ago (9), the role of PPAR␥ in cellular processes such as adipogenesis (34), insulin sensitivity (35), and glucose and lipid metabolism has been the subject of extensive investigation. Although the role of the receptor in the aforementioned cellular and metabolic processes (22, 36 -38) has been well established, the identity of its true biological ligands remains elusive. One hypothesis views PPAR␥ primarily as a sensor for nutritional fatty acids (39). Thus, a feed forward mechanism was proposed in which dietary fatty acids activate PPAR␥ in preadipocytic cells to promote adipogenesis, which, in turn, results in storage of the dietderived fat. In this model, which has parallels in other lipidsensing protein systems (39), unmodified nutritional free fatty acids are the endogenous ligands, and there is no requirement for active synthesis of a specific regulatory ligand for PPAR␥. However, free fatty acids bind to PPAR␥ with affinities in the range of 2-50 M, well above the published affinities of bona fide ligands for most nuclear hormone receptors. Although the effective concentration of free fatty acids in 3T3-L1 preadipocytes in close proximity to nuclear PPAR␥ is not currently known, we suspect that these concentrations are well below those likely to be required based on prior studies in cell-free systems.
3T3-L1 adipocytes have been the primary model for the study of adipogenesis, the process wherein fibroblastic cells differentiate into fat cells. Several studies have demonstrated the requirement for PPAR␥ in the initiation of the adipogenic program in vivo and in vitro (21)(22)(23). Since PPAR␥ gene expression requires the coordinate interaction of receptors ligands, and coactivators, it is surprising that studies examining production and/or accumulation of a putative ligand in these cells during adipogenesis are limited (29,31). In this study, we have shown that an endogenous PPAR␥ ligand is produced in a regulated fashion during the course of adipogenesis in 3T3-L1 cells. To accomplish this, we utilized an autoregulated inducible reporter system to show that ligand activity is generated in a time-and cAMP-dependent fashion during an early phase of the adipogenic process in 3T3-L1 cells. Activation of the PPAR␥ ligand-monitoring system by DIM in our stable cell line could be reproduced by MIX, forskolin, or 8-bromo-cAMP, an observation reinforced by a large number of in vitro studies demonstrating a pivotal role for cAMP in adipogenesis (20,40). For example, addition of plasma membrane-permeable cAMP analogs, forskolin, or MIX to cultured 3T3-L1 preadipocytes or primary rat preadipocytes enhances lipid accumulation and increases expression of several adipogenic markers such as aP2, stearoyl-CoA desaturase-1, glycerol-3-phosphate dehydrogenase, and lipoprotein lipase (41,42). Our data suggest that cAMP signaling pathways activated by MIX act on as yet unidentified enzymatic pathways that produce one or more specific endogenous PPAR␥ ligands.
The intracellular ligand activity detected early during differentiation by the ligand-sensing 3T3-L1 line was paralleled by the appearance of a ligand activity in the CM, which was extractable with organic solvents. This observation made possible the use of the organically extracted and concentrated CM in a large array of experiments. We have shown here that the concentrated CM activated the LBD of PPAR␥ in both the 3T3-L1-based system and a heterologous cell system, whereas it was unable to activate other nuclear receptors. More importantly, the concentrated CM was able to induce adipogenesis, similar to the ability of the established PPAR␥ ligands (43)(44)(45), a process that is completely reversed by the PPAR␥-specific antagonist PD068235 (46). Furthermore, the ligand present in the CM activated PPAR␥ directly rather than indirectly, as demonstrated by both coactivator recruitment and protease digestion assays. In addition, the ability of the PPAR␥ LBD to specifically deplete both the ligand-sensing and adipogenic activities from day 2 CM and to compete for rosiglitazone binding to the receptor demonstrates the existence of a bona fide agonist that binds to the PPAR␥ LBD. In addition, the activity present in the CM robustly increased the mRNA of PPAR␥ itself, possibly via a positive feedback loop involving C/EBP␣ (47), and of PPAR␥ target genes involved in fatty acid synthesis and storage.
Numerous reports have described a wide variety of structurally diverse molecules that the PPAR␥ LBD can accommodate in its large hydrophobic pocket (10,48). The prostaglandin 15-dPGJ 2 was the first molecule to be identified as a possible FIG. 9. HPLC fractions contain adipogenic and PPAR␥ ligandtype moieties. Day 0, 2, and 8 media from differentiating 3T3-L1 cells were organically extracted, concentrated, and fractionated by HPLC. A, the three sets of fractions collected (fractions A-H) were tested for their ability to induce adipogenesis in 3T3-L1 cells. Cells were treated with solvent (none), DIM, or the indicated fractions from days 0 to 2. Lipids were stained with oil red O at day 10. B, fractions A-H derived from day 2 medium were tested for their ability to activate the PPAR␥ LBD in CV-1 cells by transient transfection assay. Luc, luciferase. endogenous PPAR␥ ligand (12,13) and has been widely studied since. We were able to exclude 15-dPGJ 2 as the regulated adipogenic ligand in this system because of its almost complete absence in the CM or cells. Our results are in agreement with a recent report (49) that demonstrated only low picomolar levels of 15-dPGJ 2 in the medium of 3T3-L1 cells, amounts that do not increase during differentiation despite induction of cyclooxygenase-2. These observations, together with another report (50) that showed involvement of cyclooxygenase-2 enzymatic pathways solely in the clonal expansion state of the differentiation process, suggest that 15-dPGJ 2 is not the endogenous adipogenic PPAR␥ ligand. They also suggest that cyclooxygenase-2 enzymatic pathways may not be involved in production of this ligand in 3T3-L1 cells.
Madsen et al. (32) recently proposed that lipoxygenase enzymatic pathways generate PPAR␥ ligand(s) in 3T3-L1 cells by showing that baicalein, an inhibitor of 12-lipoxygenase, blocks lipid accumulation in 3T3-L1 cells without acting as a PPAR␥ antagonist. We did not see repression of the ligand-monitoring system when 5B2 cells where incubated with baicalein in the presence of DIM, suggesting that production of a PPAR␥ ligand may occur in the presence of this inhibitor. In addition, the lack of 12-lipoxygenase expression in 3T3-L1 cells (32) and of an adipocyte phenotype in the 12/15-or 5-lipoxygenase knockout mice (51,52) suggests that inhibition of adipogenesis by baicalein may take place via an alternative mechanism or via a novel yet unidentified enzymatic pathway.
LPA is a potent lipid mediator that controls growth and motility of preadipocytes (33). LPA also exists in monocytes present in atherosclerotic lesions and in aggregating platelets. A report by McIntyre et al. (17) proposed LPA to be a PPAR␥ agonist. The authors presented evidence for direct interaction between LPA and the LBD of the receptor and for stimulation of transfected PPAR␥ by medium derived from thrombin-activated platelets in RAW264.7 cells. Our experiments did not point to LPA as the regulated adipogenic ligand. None of the five LPA derivatives we tested activated the ligand-sensing system or promoted lipid accumulation. Hence, the action of LPA in preadipocytes may be different from its action in activated monocytes, possibly because specificity of receptor responses to cognate ligands depends on the coactivator milieu present in the relevant cell types.
Our finding that the ligand-monitoring system activity decreased to basal levels later in adipogenesis was not anticipated. Activity was high only for the first 48 h, a period that coincides with the up-regulation of the PPAR␥ mRNA and protein (9). This suggests that endogenous PPAR␥ ligand production initiates just before PPAR␥ protein levels begin to rise from low, but clearly detectable levels and dissipates once the adipogenic program is more fully established, even though PPAR␥ mRNA and protein levels remain high throughout the life of the mature adipocyte. This result is supported by the observation that the specific PPAR␥ antagonist PD068235 effectively blocked differentiation only when added at very early time points, when ligand-sensing activity was still maximal. The antagonist had no inhibitory effect if added at later time points, viz. after day 2, when the cells had already committed to differentiating and when the ligand-sensing activity was minimal. Another line of evidence that supports a decline in ligand production is illustrated in Fig. 8A. Only HPLC fraction C from day 2 CM contained adipogenic activity. The corresponding fraction C from day 0, when the cells were still fibroblastic and from later days when the cells were mature adipocytes, had no adipogenic activity. In agreement with this observation is the ability of fraction C from day 2 medium to activate the LBD of PPAR␥ in a heterologous system. It may be possible that acti-vation of PPAR␥ through its LBD by an endogenous ligand is not required once the adipogenic program is fully in place. It is well established that the level of the PPAR␥ protein peaks soon after differentiation and remains high in mature adipocytes. Along these lines, several investigators have demonstrated that exposure of mature adipocytes to exogenous PPAR␥ ligands results in reduction of both the mRNA and protein levels of the receptor (53)(54)(55). In other words, the continued presence of ligand causes PPAR␥ down-regulation, a phenomenon common among several hormone nuclear receptors (56,57). These findings, combined with data presented in this study, support a model in which the presence of the PPAR␥ ligand detected in our system may not be necessary for maintenance of the adipogenic program. If this is correct, one can hypothesize either that other transcription factors such as C/EBP␣ may complement the function of PPAR␥ or, alternatively, that ligandindependent activation functions in the N terminus of the receptor predominate and may contribute to maintaining differentiated function of mature adipocytes (58). Further investigation into the requirements for an endogenous PPAR␥ ligand in mature adipocytes is needed to clarify this question.
The biological importance of PPAR␥ has been well demonstrated through genetic gain-and loss-of-function experiments and through the proven efficacy of pharmacological PPAR␥ agonists, the thiazolidinediones, in the treatment of diabetes. In this context, it is disappointing that so little is known about the identity and levels of the cognate ligand(s) for this receptor in relevant tissues in normal physiology and disease states. It might be predicted that this knowledge, once obtained, would be highly relevant to our understanding of the biology of PPAR␥dependent pathways and their potential disregulation in diseases, including (but not limited to) those related to adipose function and glucose metabolism. The studies presented here provide compelling evidence that a hydrophobic PPAR␥ ligand is produced under a cAMP-dependent signal and is released into the medium early and transiently during adipogenesis in 3T3-L1 cells. Our studies with a PPAR␥ antagonist also strongly suggest that this ligand production is critical to the elaboration of the adipogenic program. Preliminary characterization of the ligand activity revealed that it interacts directly with the LBD of PPAR␥ and is capable of producing a conformation that promotes transcriptional activation. The partially purified activity promotes adipogenesis in a manner similar to that of known pharmacological and natural ligands. These studies provide important new information on the status of PPAR␥ ligand production during adipogenesis in 3T3-L1 cells that complements the large body of data on PPAR␥ itself and on PPAR␥-dependent transcriptional events in these cells. Current efforts to identify this ligand, if successful, will likely provide important new insights into the role of PPAR␥-dependent pathways in physiology and disease.