Genetic Analysis of Adipogenesis through Peroxisome Proliferator-activated Receptor γ Isoforms*

Peroxisome proliferator-activated receptor (PPAR) γ is a nuclear receptor that is a key regulator of adipogenesis and is present in two isoforms generated by alternative splicing, PPARγ1 and PPARγ2. Studies of the ability of each isoform to stimulate fat differentiation have yielded ambiguous results, in part because PPARγ stimulates its own expression. We have thus undertaken a formal genetic analysis using PPARγ-null fibroblast cell lines to assess the specific role of each individual isoform in adipogenesis. We show here that both PPARγ1 and PPARγ2 have the intrinsic ability to stimulate robust adipogenesis. Adipose cells stimulated by either PPARγ1 or PPARγ2 express a similar gene profile and show similar responses to insulin. However, in response to low ligand concentrations, PPARγ2 shows a quantitatively greater ability to induce adipogenesis. Analyses involving coactivator binding and transcriptional assays indicate that PPARγ2 has an enhanced ability to bind components of the DRIP/TRAP complex, coactivators required for fat differentiation.

The peroxisome proliferator-activated receptor (PPAR) 1 ␥ is a member of the nuclear hormone receptor superfamily that is expressed at high levels in fat cells. It was independently identified as a transcriptional regulator of a fat cell-specific enhancer element (1) and as a novel member of the PPAR family (2). A key regulatory role for PPAR␥ in fat differentiation was demonstrated by gain of function experiments, which showed that ectopic expression and activation of this factor in fibroblasts or myoblasts promoted adipogenesis (3)(4)(5). More recently, it has been shown that PPAR␥ is necessary as well as sufficient for fat cell differentiation in cultured cells and mice (6 -8).
PPAR␥ activity is regulated by the binding of ligands and by the subsequent docking of coactivators. Several natural occurring ligands of relatively low affinity have been identified, including various fatty acids and prostaglandins (9,10); the biological significance of their function with PPAR␥ is largely unknown. Synthetic ligands such as the antidiabetic thiazoledinediones (pioglitazone, rosiglitazone, and troglitazone) and certain tyrosine analogs are effective agonists and bind with K d values between 5 and 750 nM (11,12).
The binding of agonist ligands to nuclear receptors promotes the docking of coactivator proteins such as members of the p160/SRC family, p300/CBP, and p300/CBP-associated factor, all of which exhibit histone-acyltransferase activity (13). In addition PPAR␥ can also dock the tissue-selective coactivators PGC-1␣ and -␤, although these docking events are not ligandgated (14,15). PPAR␥ also binds to PBP/DRIP205/TRAP220, initially identified as a PPAR␥-binding protein and subsequently characterized as a component of the mediator-like coactivator complex DRIP/TRAP/ARC (16 -18). This protein complex is essentially devoid of histone-acyltransferase activity and is believed to associate directly with the general transcription factors of the preinitiation complex. Recent studies illustrate that PBP/TRAP220/DRIP205 is absolutely required for PPAR␥-mediated adipogenesis (19).
PPAR␥ is present in two isoforms, PPAR␥1 and PPAR␥2, generated by alternative promoter usage; PPAR␥2 has an additional 30 N-terminal amino acids relative to PPAR␥1. The expression of PPAR␥2 is restricted mainly to fat, whereas PPAR␥1 is expressed in fat and many other tissues. Our earlier report indicated that both PPAR␥1 and PPAR␥2 could stimulate adipogenesis when introduced into fibroblastic cells (3). However, conclusions concerning the inherent adipogenic potential of these individual receptors are compromised by the fact that ectopic expression of PPAR␥1 turns on endogenous PPAR␥2 (3). In an attempt to circumvent this problem, one recent report has utilized engineered artificial transcriptional suppressors of the endogenous PPAR␥ gene in combination with ectopic expression of the individual isoforms. This study concluded that PPAR␥2 is adipogenic and PPAR␥1 is not (20). To avoid problems associated with these engineered suppressors and to investigate this key question in a more formal genetic way, we now make use of cells completely lacking a functional PPAR␥ gene to reassess the adipogenic action of PPAR␥1 and PPAR␥2. Our studies clearly illustrate that both PPAR␥ isoforms can drive the differentiation of fully functional fat cells. However, under limiting levels of a PPAR␥ ligand, PPAR␥2 has an enhanced ability to promote differentiation and to interact with PBP/DRIP205/TRAP220.
Cell Culture-Cells were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and penicillin/streptomycin in 5% CO 2 . For differentiation assays, cells were treated with 0.5 mM 3-isobutyl-1-methylxanthine, 5 g/ml insulin, and 1 M dexamethasone for 2 days, and, subsequently, cells were kept in maintenance medium consisting of Dulbecco's modified Eagle's medium with 10% fetal bovine serum supplemented with penicillin/streptomycin and insulin (5 g/ ml). Oil red O staining of differentiated adipocytes was performed as described previously (3).
Transient Transfections and Infections-95% confluent U2OS cells were transiently transfected with 25 ng of PPAR␥1 or PPAR␥2 retroviral constructs or empty vector, with 100 ng of DR1-luc, using Fu-GENE 6 (Roche Molecular Biochemicals) according to manufacturer's instructions. 24 h after transfection, cells were treated with various ligand concentrations, and 24 h later, the luciferase levels were measured. TRAP220Ϫ/Ϫ cells (19) were cotransfected with 25 ng of PPAR␥1 or PPAR␥2 and 100 ng of DR1-luc in the absence or presence of 400 ng of TRAP220. After 24 h, cells were treated with 10 nM rosiglitazone, and 48 h after treatment, cells were harvested for luciferase measurements. For retrovirus preparation, the Bosc23 packaging cell lines were transfected with PMCV retroviral PPAR␥1 and PPAR␥2 constructs or vector, according to the protocols described previously (14). PPAR␥ knockout cells were plated the day before infection and infected with the viral supernatant at 70% confluence. After an 8 -12-h incubation with virus and 8 M/ml Polybrene, cells were split 1:3 and selected with 2 M/ml puromycin for 1 week.
RNA and Protein Analysis-RNA was extracted from cultured cells using Trizol (Invitrogen), according to the manufacturer's instructions. 10 -15 g of RNA were run on an agarose formaldehyde gel and transferred on nylon membrane (ICN) by Northern blotting using 10ϫ SSC. Hybridizations were performed using Quickhyb (Stratagene) according to manufacturer's instructions. Whole cell extracts were obtained using phosphate-buffered saline supplemented with 1% Triton X-100, 0.5% deoxycholate, 10 mM sodium pyrophosphate, 2 mM sodium vanadate, 100 mM sodium fluoride, and 10 g/ml aprotinin and leupeptin. 100 -150 g of protein were used for Western blots using PPAR␥ antibody (E8) obtained from Santa Cruz Biotechnology.
Glucose and Fatty Acid Uptake-Cells were induced to differentiate with a differentiation mixture (as described under "Cell Culture"). At day 7 of differentiation, 90 -95% of cells accumulated lipid droplets. Glucose uptake was measured after 24 h of starvation in low-glucose medium supplemented with 0.5% bovine serum albumin. For glucose uptake assays, cells were treated for 8 min with 100 nM insulin, and deoxyglucose uptake was measured as described previously (21). For fatty acid uptake, radiolabeled oleic acid (PerkinElmer Life Sciences) was added to fully differentiated adipocytes at time 0. After 0, 3, or 6 h, cells were sonicated, and lipids were extracted from cell homogenates in chloroform/methanol. Incorporation of radiolabeled oleic acid was measured through scintillation counting.
Transcriptional Profiling-RNA collected from fully differentiated PPAR␥1 and PPAR␥2 cells was labeled with [␣-33 P]dCTP. Transcriptional profiling experiments were performed on custom DNA arrays composed of ϳ2,500 mouse cDNAs, half of which were from a mouse fat library (22). Duplicate filters per probe were used. The hybridization was done as described previously (22). The data generated were analyzed using the SOM clustering program.
Ligand Binding and Coactivator Recruitment Assays-Baculovirusinfected insect Sf21 cells were collected at day 2 after infection. GST PPAR␥ proteins were extracted using lysis buffer containing 400 mM KCl, 20 mM Tris, pH 8, 1 mM EDTA, 0.1% Triton X-100, and 1 mM dithiothreitol. After incubating the lysate with GST beads, PPAR␥ protein bound to the beads was quantified by Coomassie Blue staining. Ligand binding assays were performed with radiolabeled BRL, as described previously (23). The coactivator interaction assays were performed using 1 g of baculovirus PPAR␥ protein with 35 S-labeled in vitro-translated coactivators. Briefly, GST-PPAR␥ beads were incubated with 35 S-labeled in vitro-translated coactivators (Promega) at 25°C in the presence or absence of 10 nM rosiglitazone, according to a previously described protocol (14). After 1-2 h, beads were washed three times with 1 ml of binding buffer with or without 1 M rosiglitazone and loaded on an 8% SDS gel.

Adipogenic Functions of PPAR␥ Isoforms in PPAR␥-deficient
Cells-We investigated the adipogenic action of PPAR␥ isoforms utilizing PPAR␥-deficient cells. We infected immortalized 3T3 fibroblasts null for PPAR␥ (24) with an empty retroviral vector or with vectors expressing murine PPAR␥1 or PPAR␥2. Both RNA and protein for the two PPAR␥ isoforms were present in undifferentiated cells at virtually identical levels after drug selection for stable viral integration (Fig. 1A). These levels of PPAR␥ are approximately equal to those seen in differentiated adipocytes (data not shown). The cells were then stimulated to differentiate according to two different protocols. In the first set of experiments, strong induction stimuli were used: cells were treated for 2 days with a mixture commonly used to stimulate differentiation of 3T3-L1 cells, including dexamethasone, insulin, and isomethylbutylxanthine, plus the further addition of the PPAR␥ ligand rosiglitazone. At day 7 after induction, we observed extensive lipid accumulation in cell lines expressing either PPAR␥1 or PPAR␥2 (Fig. 1B). In addition to showing similar morphological changes, these cell lines also expressed similar levels of mRNA for molecular markers of white fat differentiation such as aP2 and CD36 (Fig.  1C). No induction of endogenous PPAR␥2 protein was observed in differentiated PPAR␥1 cells, confirming the total ablation of the PPAR␥ gene in these cells (data not shown). Similar effects on differentiation and gene expression were obtained when these cells were induced to differentiate with dexamethasone, insulin, and isomethylbutylxanthine in the absence of ligand (data not shown).
The development of insulin-sensitive glucose uptake is a very important part of the adipose differentiation program. The results shown in Fig. 2A indicate that insulin-stimulated glucose uptake is similar in cells induced to differentiate by either PPAR␥ 1 or PPAR␥2. We also measured the uptake of fatty acids in these cell lines and found that it is also similar in both cell lines (Fig. 2B).
Transcriptional Cascades Activated by PPAR␥1 and PPAR␥2 Isoforms-To investigate whether PPAR␥1 and PPAR␥2 can regulate similar sets of genes, we performed global transcriptional profiling using mRNA isolated from cells differentiated through PPAR␥1 and PPAR␥2 expression and activation. As shown in Table I, most of the genes associated with adipogenic differentiation or lipid metabolism are induced similarly by both isoforms, indicating that both PPAR␥1 and PPAR␥2 are competent to activate an adipogenic gene program. Fig. 2C illustrates that genes involved in the Krebs cycle, such as aconitase precursor and malate dehydrogenase, appear to be similarly induced in cells differentiated by the two different isoforms. However, a few genes appear to be regulated differently, at least quantitatively. For example, genes of the glycolytic pathway, such as phosphofructokinase and phosphoglycerate kinase, are induced to a greater extent in PPAR␥1 cells, whereas certain other genes associated with differentiation, such as adipsin, are induced at higher levels in PPAR␥2 cells.
Differentiation Capacity of PPAR␥ Isoforms under Higher Stringency-Whereas the data presented above illustrate clearly that both PPAR␥ isoforms can drive adipogenesis, it is reasonable to suppose that the extracellular influences may not be as consistently pro-adipogenic as the conditions that we have used. We therefore stimulated cells more gently, using only insulin and a titration of the PPAR␥ ligand levels. Fig. 3A shows that PPAR␥1 and PPAR␥2 drive cells to differentiate to differing extents when the ligand concentrations are submaxi-   3B); this difference is lost at 100 nM. Interestingly, more lipid accumulation and aP2 mRNA are also apparent in PPAR␥2 cells when no ligand is added.
To determine whether these differences in adipogenic action correlated with differences in transcriptional function of the two isoforms, their transcriptional activities were examined at low concentrations of rosiglitazone. As shown in Fig. 3C, PPAR␥2 can activate a multimerized PPAR response element linked to luciferase more effectively than PPAR␥1, both in the absence of rosiglitazone and at low doses of rosiglitazone. This difference, which is 2-3-fold at 10 nM rosiglitazone, is Ͻ30% at 1 M rosiglitazone.
Adipogenic Actions of PPAR␥ Isoforms Correlate with PBP/ DRIP205/TRAP220 Docking-The differences in adipogenic and transcriptional activity between PPAR␥1 and PPAR␥2 observed at low ligand concentrations could be due to differential affinity for agonists or could reflect other aspects of receptor function. To investigate the molecular basis of these differences, we first produced both PPAR␥ isoforms as full-length proteins in insect cells using baculovirus vectors. These were subsequently purified and used to investigate ligand binding capacity. As shown in Fig. 4A and B, PPAR␥1 and PPAR␥2 bind rosiglitazone and 15d-⌬ 12,14 PGJ 2 with essentially identical affinities.
Because nuclear receptors are known to activate transcription through docking of specific coactivator proteins, we assessed whether the quantitative differences in the ability of the two PPAR␥ isoforms to induce differentiation could reflect a difference in their ability to recruit known coactivators. We therefore performed in vitro interaction assays in the presence or absence of a low concentration of ligand using baculovirus purified full-length proteins and coactivators produced as in vitro-translated proteins in reticulocyte lysates. As shown in Fig. 5A, the interactions of the two PPAR␥ isoforms with both PGC-1␣ and SRC-1 are comparable in the presence or absence of rosiglitazone. In contrast, two members of the DRIP/TRAP complex, PBP/DRIP205/TRAP220 and DRIP150/TRAP170, appear to have a quantitative preference for PPAR␥2 in both the absence and presence of added ligand. PPAR␥ docks DRIP150/ TRAP170 independently of added ligand, consistent with the idea that this subunit of the DRIP/TRAP complex may bind to the N terminus of PPAR␥, as shown previously for the glucorticoid receptor (25).
To critically analyze the role of the DRIP/TRAP complex in the differential effects of PPAR␥1 and PPAR␥2, we utilized fibroblasts lacking the DRIP205/TRAP220 subunit of the DRIP/TRAP complex. Mice lacking DRIP205/TRAP220 die embryonically at day 11.0, and a DRIP205/TRAP220 Ϫ/Ϫ cell line was obtained by immortalizing cells isolated from embryonic day 10.0 littermate embryos (19). We utilized this cell line to analyze first whether the differences in transcriptional activity between PPAR␥1 and PPAR␥2 are dependent on TRAP220. As shown in Fig. 5B, in the absence of TRAP220, both PPAR␥1 and PPAR␥2 show an identical ability to transactivate the multimerized PPAR response element reporter gene in either the absence or presence of a low concentration (10 nM) of rosiglitazone. The addition of exogenous DRIP205/TRAP220 coactivator partially restores differential transcriptional activation observed between the two isoforms, with or without ligand addition. These data suggest a functional significance in the preferential interaction between DRIP205/TRAP220 and PPAR␥2. DISCUSSION A key question regarding the PPAR␥ isoforms is whether they can both stimulate fat cell differentiation in culture and in vivo. Whereas the latter cannot be answered with confidence, we provide definitive genetic evidence here that PPAR␥1 and PPAR␥2 can individually drive a full program of adipogenesis in culture. Both isoforms can stimulate lipid accumulation, the development of insulin-sensitive glucose uptake, and a pattern of gene expression that is very similar. Translating these results to an in vivo setting must be purely speculative. It is notable that a difference in relative adipogenic action of the two PPAR␥ isoforms becomes clear at lower ligand concentrations, with PPAR␥2 more adipogenic at 10 -50 nM rosiglitazone, approximating the K d of the receptor. If we assume that ligand levels are limiting in vivo, these results imply that the PPAR␥2 isoform is quantitatively more adipogenic. This might provide a biological rationale for the adipose-selective expression of the PPAR␥2 protein.
These data are quite different from those of Ren et al. (20), who conclude that the PPAR␥1 protein has no adipogenic action. Their experiments, like ours, used strong adipogenic conditions, including treating the cells with the mixture containing 3-isobutyl-1-methylxanthine, insulin, and dexamethasone. The suppression of expression of the endogenous PPAR␥ gene in 3T3-L1 cells was achieved by using an engineered repressor and subsequently reexpressing each individual PPAR␥ protein for comparison. It is possible that the effects observed by Ren et al. (20) could be due to a differential level of expression of the PPAR␥1 isoform relative to the PPAR␥2 isoform apparent in their protein data. Alternatively, it is possible that some other nonspecific effects of their artificial repressor blocked the adipogenic action of PPAR␥1 through unknown mechanisms.
In our study, the greater adipogenic activity of PPAR␥2 compared with PPAR␥1 correlates with better binding to PBP/DRIP205/TRAP220 and DRIP150/TRAP170 and, presumably, the entire DRIP/TRAP/ARC complex. Furthermore, the higher transcriptional activity of PPAR␥2 compared with PPAR␥1 seems to be dependent on the presence of the DRIP205/TRAP220 subunit. Because recent data have indicated an absolute genetic requirement for DRIP205/TRAP220 for the adipogenic action of PPAR␥, this may provide a mechanistic basis for the quantitatively greater adipogenic action of PPAR␥2.
At the present time, very little is known about the precise function of the N terminus of PPAR␥ or any nuclear receptor, for that matter. The present study, combined with earlier work showing that a substantial portion of adipogenic action of PPAR␥ is carried in the N terminus (26), suggests a real need to determine how this part of PPAR␥ functions in mechanistic detail.
FIG. 5. Recruitment of coactivators by PPAR␥ isoforms. A, comparison between the ability of the PPAR␥1 and PPAR␥2 isoforms to interact with different coactivators in the absence or presence of 10 nM rosiglitazone. Full-length PPAR␥ was expressed as GST fusion protein in insect Sf21 cells and immobilized on glutathione-coupled Sepharose beads. The coactivators SRC-1, PGC-1␣, DRIP150/TRAP170, and DRIP205/TRAP220 were in vitro-translated, labeled with [ 35 S]methionine, and incubated with GST-PPAR␥ isoforms, as described under "Experimental Procedures." B, transcriptional activity of PPAR␥1 and PPAR␥2 in DRIP205/TRAP220-null cells in the absence (Ϫ) or presence (ϩ) of 400 ng of PBP/DRIP205/TRAP220, after treatment with vehicle or 10 nM rosiglitazone.