Transcriptional activation by peroxisome proliferator-activated receptor gamma is inhibited by phosphorylation at a consensus mitogen-activated protein kinase site.

The nuclear receptor peroxisome proliferator-activated receptor γ (PPARγ) regulates transcription in response to prostanoid and thiazolidinedione ligands and promotes adipocyte differentiation. The amino-terminal A/B domain of this receptor contains a consensus mitogen-activated protein kinase site in a region common to PPARγ1 and -γ2 isoforms. The A/B domain of human PPARγ1 was phosphorylated in vivo, and this was abolished either by mutation of serine 84 to alanine (S84A) or coexpression of a phosphoprotein phosphatase. In vitro, this domain was phosphorylated by ERK2 and JNK, and this was markedly reduced in the S84A mutant. A wild type Gal4-PPARγ(A/B) chimera exhibited weak constitutive transcriptional activity. Remarkably, this was significantly enhanced in the S84A mutant fusion. Ligand-dependent activation by full-length mouse PPARγ2 was also augmented by mutation of the homologous serine in the A/B domain to alanine. The nonphosphorylatable form of PPARγ was also more adipogenic. Thus, phosphorylation of a mitogen-activated protein kinase site in the A/B region of PPARγ inhibits both ligand-independent and ligand-dependent transactivation functions. This observation provides a potential mechanism whereby transcriptional activation by PPARγ may be modulated by growth factor or cytokine-stimulated signal transduction pathways involved in adipogenesis.

Adipocyte differentiation is a complex process regulated by extracellular hormone and cytokine stimulation (1). Cultured preadipocyte cell lines differentiate into lipid-laden adipocytes following exposure to insulin, glucocorticoid, and inducers of intracellular cyclic AMP (2). Conversely, epidermal growth factor (EGF) 1 and transforming growth factor ␣ act via the EGF receptor to inhibit both primary and preadipocyte cell line differentiation (3,4). Tumor necrosis factor ␣ (TNF␣) is also a potent inhibitor of differentiation. In addition, this cytokine promotes lipolysis and down-regulates adipocyte-specific gene expression in mature adipocytes (5,6).
Adipocyte differentiation is driven by the coordinate expression of a range of transcription factors, including C/EBP␣, -␤, and -␦ (7) and ADD1 (8), which lead to the expression of adipocyte-specific genes. In addition, the peroxisome proliferatoractivated receptor ␥ (PPAR␥) has been shown to be selectively expressed in adipocytes (9,10) and to modulate adipocytespecific gene expression (10). Alternative splicing generates two receptor isoforms such that mouse PPAR␥2 has a 28residue extension at its amino terminus compared with human PPAR␥1. PPAR␥2 mRNA is highly expressed in murine adipocyte cell lines (10), whereas both receptor isoforms are abundant in freshly isolated mouse adipocytes (11). The early induction of PPAR␥ mRNA expression during adipogenesis, combined with the ability of retrovirally overexpressed PPAR␥ to induce lipid accumulation and the expression of adipocytespecific genes (12), suggests that this receptor plays an important role in preadipocyte differentiation. This notion is strengthened by the observation that specific high affinity ligands for PPAR␥ including thiazolidinediones (which act as insulin sensitizers in vivo), as well as eicosanoids, promote the differentiation of murine preadipocyte cell lines (13)(14)(15).
PPAR␥ is an orphan member of the nuclear receptor family. These receptors share a conserved domain structure and modulate gene transcription through two transcription activation mechanisms: a hormone-dependent transcription activation function (AF-2) is located in the carboxyl-terminal hormonebinding domain, whereas the amino-terminal A/B domain contains a ligand-independent activation function (AF-1). Recently, the AF-1 activity of the estrogen receptor (ER), another member of this family, has been shown to be modulated following phosphorylation by mitogen-activated protein (MAP) kinase (16,17). Three MAP kinase pathways have been identified in mammalian cells. Members of the extracellular signal-regulated kinases, ERK1 and ERK2, are activated predominantly by growth factor stimulation via a Ras-dependent signal transduction cascade (18). In contrast, activity of Jun NH 2 -terminal kinase (JNK, also known as SAPK) and p38 kinase is increased by exposure of cells to environmental stress or to cytokines including TNF␣ (19,20). In turn, activated MAP kinases have been shown to regulate the activity of specific transcription factors including Elk-1, ATF-2, and c-Jun by phosphorylation of serine or threonine residues in the appropriate context (21).
Interestingly, it has been reported that PPAR␥1 and PPAR␥2 are similarly adipogenic but that a truncated receptor in which the amino-terminal domain of PPAR␥2 is deleted is a more potent inducer of adipocyte differentiation (12). We noted that NH 2 -terminal domain of PPAR␥ contains a consensus MAP kinase site in a region conserved between PPAR␥1 and PPAR␥2 isoforms. Furthermore, we and others (11,22) have observed that PPAR␥ proteins migrate on immunoblots as closely spaced doublets, a pattern suggestive of phosphorylation. In this report we show that this putative MAP kinase site is phosphorylated in vivo and also in vitro by ERK2 and JNK. Furthermore, we demonstrate that phosphorylation significantly inhibits both ligand-independent and ligand-dependent transcriptional activation by PPAR␥. Finally, we show that mutation of this MAP kinase site in PPAR␥ increases its adipogenic activity. These findings provide a potential pathway by which extracellular hormones and cytokines might regulate adipocyte differentiation by phosphorylation-dependent modulation of PPAR␥ activity.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-The GAL4 UAS-TKLUC luciferase reporter contains two copies of the GAL4 17-mer binding site (23) and the BOS␤gal and CMV␤-gal reference plasmids have been described previously (23,24). The GAL4-hPPAR␥ (A/B) wild type and S84A mutant expression vectors contain residues 1-108 of human PPAR␥1 cloned into the EcoRI site of pSG424 (25). These domains were also cloned into pGEX4T for expression of glutathione S-transferase (GST) fusion proteins. pCMVflag-p38, which contains the Flag epitope between codons 1 and 2 of p38, was obtained from J. Han (Dept. of Immunology, Scripps Research Institute, San Diego, CA). RSV-PP1 is an expression vector encoding protein phosphatase 1␣ (26). pSG-CL100 is an expression vector encoding a dual specificity MAP kinase phosphatase (27). pSPORT-mPPAR␥2 encodes full-length mouse PPAR␥2 (10), and the S112A point mutation was introduced into it by polymerase chain reaction-directed mutagenesis. Both PPAR␥2 and PPAR␥2-SA cDNAs were then subcloned into the SalI site of pCMX. All mutations and ligation junctions were confirmed by sequencing. The acyl-CoA ϫ 3-TK-LUC construct contains three copies of the acyl-CoA oxidase PPAR response element (5Ј-GATCTGGACCAGGA-CAAAGGTCACGTTCA) in pTK-luciferase.
Cell Culture and Transfection Studies-For functional studies, JEG-3 cells were cultured in DMEM containing 10% fetal bovine serum and transferred to DMEM plus charcoal-stripped fetal calf serum immediately prior to transfection. Cells were transfected with luciferase reporter, receptor expression vector, ␤-gal expression vector, and phosphatase expression vector where indicated using the calcium phosphate precipitation method. 5 M BRL49653 (in Me 2 SO) or vehicle control was added 16 h after transfection. Cells were lysed 24 h later, and luciferase and ␤-gal was measured as described (23,28). Luciferase values were normalized to ␤-gal activity and fold activation was calculated.
In Vivo Phospholabeling, Immunoprecipitation, and Western Blotting-JEG-3 cells, plated on 10-cm plates, were transfected with 30 g of GAL4-hPPAR␥ plus 10 g of PP1 expression vector as indicated. 40 h later cells were incubated in phosphate-free DMEM for 30 min, followed by a 4-h incubation with 1 mCi/ml [ 32 P]orthophosphate (DuPont NEN). Cells were lysed by a 30-min incubation at 4°C in RIPA buffer (1% Nonidet P-40, 1% deoxycholate, 0.1% SDS, 0.15 M NaCl, 0.01 M NaPO 4 , pH 7.2, 2 mM EDTA, 50 mM NaF, 0.2 mM sodium vanadate, 1 g/ml leupeptin, and 1 g/ml aprotinin). Extracts were cleared by centrifuging at 26,000 ϫ g for 30 min at 4°C.Samples were precleared with 10 l of whole rabbit serum (Cappel) in a 50% slurry of protein A-agarose (Life Technologies, Inc.). The resulting supernatant was incubated with 10 l of GAL4 DBD rabbit polyclonal antiserum (Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C. Immune complexes were then precipitated with a 50% slurry of protein A-agarose and washed five times with RIPA buffer. The immunoprecipitate was analyzed by 12% SDS-PAGE and autoradiography.
Whole cell extracts were prepared from JEG-3 cells transfected with wild type or mutant mPPAR␥2 as described above and analyzed by SDS-PAGE and Western blotting followed by chemiluminescent detection (Amersham Corp.) with anti-PPAR␥ IgG at a dilution of 1:1000.
Protein Kinase Assays-Phosphorylation of 2 g of GST fusion proteins by ERK2 (New England Biolabs, Bishop's Stortford, UK) or JNK (Calbiochem-Novabiochem, Nottingham, UK) were carried out as recommended by the manufacturers. p38 kinase assays were performed with 2 g of substrate at 30°C for 30 min in kinase buffer containing 18 mM magnesium acetate, 90 M ATP, and 2 Ci of [␥-32 P]ATP. Myelin basic protein (Life Technologies, Inc.) was used as a control substrate for ERK2, and a GST fusion protein containing a truncated form of activating transcription factor-2 (GST-ATF2, residues 1-96, Santa Cruz Biotechnology, Wembley, UK) provided positive controls for JNK and p38. 32 P incorporation was determined following SDS-PAGE fractionation. Gels were stained with Coomassie Blue to check equal loading of GST fusion proteins and autoradiographed.
Retroviral Infection and Adipocyte Differentiation of 3T3-L1 Cells-3T3-L1 cells were made to ectopically express PPAR␥2 or PPAR␥2-S112A using retroviral gene transduction as described previously (22). Infected 3T3-L1 cells were selected in G418 and grown to confluence in DMEM containing 10% iron-enriched fetal bovine serum. 2 days after reaching confluency, cells were treated with BRL49653 dissolved in Me 2 SO or Me 2 SO alone. After 7 days, cells were washed three times with phosphate-buffered saline, fixed by 10% formalin in phosphate buffer for 1 h at room temperature, washed once again with phosphatebuffered saline, then stained with 60% filtered Oil Red O stock solution (0.5 g of Oil Red O (Sigma) in 100 of ml isopropanol) for 15 min, washed four times with water, and photographed.

RESULTS
Sequence alignment of the amino-terminal A/B region that is common to PPAR␥1 and PPAR␥2 isoforms indicated that a consensus MAP kinase site is conserved between species (Fig.  1A). To determine whether the NH 2 -terminal region of PPAR␥ is phosphorylated in vivo, an expression vector encoding the the A/B domain of human PPAR␥1 (residues 1-108) fused to the heterologous DBD of GAL4 (residues 1-147) was transfected into JEG-3 cells. Orthophosphate labeling followed by immunoprecipitation with an antibody directed against GAL4 revealed that the GAL4 PPAR␥(A/B) fusion protein was highly phosphorylated, whereas Gal4 DBD alone was not (Fig. 1B,  lanes 2 and 3). In comparison, phosphorylation of transfected GAL4 PPAR␥(A/B S84A), in which a putative phosphoserine at position 84 has been mutated to alanine (S84A) was markedly reduced (Fig. 1B, lane 4), although this fusion is expressed and even more transcriptionally active than wild type Gal4-PPAR␥ (A/B) (see below). Furthermore, cotransfection of PP1, an activated form of a serine phosphatase, led to near complete loss of phosphorylation of the wild type PPAR␥ construct (Fig. 1B,  compare lanes 3 and 6).
Several MAP kinase cascades have been defined, each with a slightly different substrate specificity and activation pathway (21). In view of this, we tested the ability of three kinases, ERK2, JNK, and p38, to phosphorylate the isolated A/B domains of wild type and mutant hPPAR␥1 expressed as GST fusion proteins in E. coli. Purified recombinant ERK2 was able to phosphorylate the wild type GST-PPAR␥1 fusion protein ( Fig. 2A). This phosphorylation was abolished by mutation of the serine at position 84 in hPPAR␥1, which corresponds to the MAP kinase site to alanine (S84A) under conditions in which myelin basic protein was a good substrate for this kinase. Similarly, purified recombinant JNK was also able to phosphorylate wild type GST-PPAR␥ (A/B) in addition to a control GST-ATF2 fusion protein (Fig. 2B). Mutation of serine 84 to alanine markedly reduced but did not abolish phosphorylation by JNK, suggesting weak phosphorylation of a second kinase site in vitro, which is not phosphorylated in vivo (Fig. 1B). Similar experiments have been performed with p38 kinase and phosphorylation of wild type or mutant GST PPAR␥ was not detected under conditions in which GST-ATF2 was a good substrate for this enzyme (data not shown).
These results suggested that phosphorylation of a MAP kinase site by either ERK2 or JNK might regulate the ligandindependent transcriptional activity (AF1) of hPPAR␥. Transient transfection assays in JEG-3 cells showed that the A/B domain of hPPAR␥ acts as a weak, ligand-independent transcriptional activator when fused to the DBD of GAL4 (Fig. 3).
Mutation of serine 84 to alanine, which abolished phosphorylation of the GAL4-PPAR␥ fusion protein (see above and Fig.  1B), markedly enhanced the AF1 transactivation function of the PPAR␥ A/B domain. Similar results were obtained following transfection of wild type and mutant GAL4-PPAR␥ fusions in COS-7 cells (data not shown). Furthermore, coexpression of CL100, the human homologue of murine MAP kinase phosphatase (MKP-1), also augmented the transcriptional activity of wild type Gal4-PPAR␥ A/B (Fig. 3) but had no effect on the In addition to the constitutive AF1 function in the aminoterminal domain, PPAR␥ also contains a ligand-dependent transcription activation function (AF2). We therefore investigated whether phosphorylation of the MAP kinase site in the A/B domain could influence the ligand-dependent AF2 activity of full-length PPAR␥. Because PPAR␥ is conserved between species and the residues encompassing the amino-terminal MAP kinase site are identical in PPAR␥1 and PPAR␥2 isoforms (Fig. 1A), murine PPAR␥2, which is induced specifically during adipocyte differentiation, was used in these studies. Transfection of wild type mPPAR␥2 expression vector together with a reporter construct containing three copies of the PPAR-response element from the acyl-CoA oxidase gene and the thiazolidenedione ligand BRL49653 was associated with significant transcriptional activation (Fig. 4). Mutation of the homologous serine within the putative MAP kinase site in mPPAR␥2 to alanine (S112A) significantly increased ligand-dependent transactivation to approximately double that of the wild type receptor. These observations indicate that in addition to inhibiting the inherent AF1 activity of PPAR␥, phosphorylation of the MAP kinase site within the A/B domain also attenuates ligand-induced transcription activation by this receptor.
We next tested whether the enhanced transactivation by PPAR␥2-S112A resulted in increased adipogenic activity. Wild type PPAR␥2 or PPAR␥2-S112A were ectopically expressed in 3T3-L1 preadipocytes using a retroviral expression strategy that we have previously described (22). In the absence of adipogenic stimulation, confluent preadipocytes differentiate into adipocytes at a very low frequency (Ͻ1%). Ectopic expression of PPAR␥2 resulted in ϳ10% adipose conversion, as shown by Oil Red O staining of accumulated intracellular lipid (Fig. 5). This effect is presumably mediated by low levels of an endogenous PPAR␥2 ligand, although the possibility that this was a ligandindependent effect of PPAR␥ overexpression cannot be discounted. Remarkably, ectopic expression of PPAR␥2-S112A at levels similar to those of the ectopically expressed wild type PPAR␥2 (data not shown) induced adipocyte differentiation much more dramatically. In fact, Fig. 5 shows that the degree of adipocyte differentiation due to ectopic PPAR␥2-S112A expression was comparable with that achieved by cells expressing wild type PPAR␥ and treated with the ligand BRL49653 (50 nM). At this concentration BRL49653 alone is a relatively weak adipogenic stimulus for control cells (13,15,30), resulting in ϳ10% adipose conversion. These results show clearly that mutation of the MAP kinase consensus site produces a more adipogenic form of PPAR␥2, consistent with the increased transcriptional activity of this mutant receptor. DISCUSSION Our studies indicate that a consensus MAP kinase site located within the conserved amino-terminal A/B domains of PPAR␥1 and PPAR␥2 regulates the transcriptional activity of these adipogenic nuclear receptor isoforms. Phosphorylation of this site reduces the inherent transcriptional activity of the AF1 transactivation domain. In addition, ligand-dependent transactivation by the full-length receptor is also inhibited. This observation contrasts with previous studies of the ER showing that MAP kinase-mediated phosphorylation of the amino-terminal A/B domain enhanced the transcriptional activity of this receptor (16,17) and indicates that constitutive activity of the NH 2 -terminal domains of nuclear receptors can be modulated in response to other signal transduction pathways. Furthermore, the divergent transcriptional effects of phosphorylation in PPAR␥ versus ER suggest that the different sequences around the consensus MAP kinase site in each A/B domain may also influence the way AF1 activity is modulated.
Although the A/B domain sequence motif that we have identified represents a consensus MAP kinase site, we have demonstrated that this residue is amenable to phosphorylation by both ERK2 and JNK. Interestingly, both TNF␣ and EGF, which are potent inhibitors of adipocyte differentiation (3,4,5,6) are known to activate these pathways. In addition to enhancing JNK activity, TNF␣ stimulation has also been reported to activate MAP kinase (31,32), and EGF stimulation enhances MAP kinase activity in a variety of cell types (18). The present work suggests that activation of either pathway could phosphorylate and hence inhibit PPAR␥ activity, thus contributing to the anti-adipogenic effects of these agents.
It is also possible that other signal transduction pathways may be involved. For example, it is interesting to note that adipocyte differentiation is promoted by culture in conditions that raise intracellular cAMP (33,34), and elevated cAMP levels have been reported to inhibit ERK activity in rat adipocytes (35,36). In some cell types, cAMP-dependent stimulation of protein kinase A leads to direct inhibition of Raf-1, which in turn inhibits EGF-stimulated MAP kinase activity (37). Given the convergence of these pathways, it is tempting to speculate that altered PPAR␥ phosphorylation might contribute to the abnormal adipogenesis seen in mice harboring a mutant protein kinase A regulatory subunit (38).
Phosphorylation has been shown to regulate the activity of transcription factors by a number of different mechanisms (21). Because equal levels of mutant and wild type PPAR␥2 were detected in transfected cells (Fig. 4), phosphorylation does not appear to alter PPAR␥ protein stability, as has been shown for the transcription factor Fos (39). Furthermore, because the inhibitory effect of NH 2 -terminal PPAR␥ phosphorylation is transferable to the heterologous DNA binding domain of GAL4, we also consider it unlikely that phosphorylation alters the ability of PPAR␥ to bind to DNA. A third possibility is that phosphorylation inhibits the activity of PPAR␥ by altering receptor interaction with other transcription intermediary proteins. For example, an interaction between the A/B domain of thyroid hormone receptors TR␤2 and TR␣1 and TFIIB, which might influence transcription by altering preinitiation complex formation or stability, has recently been described (40,41). This raises the possibility of analogous interactions between basal transcription factors and the A/B region of PPAR␥ that are phosphorylation-sensitive. Alternatively, phosphorylation at this site might influence PPAR␥ interaction with specific coactivator or corepressor proteins, in the same way that phosphorylation of c-Jun enhances recruitment of the CREB-binding coactivator protein (42).
Our finding of increased adipogenicity of a PPAR␥2 point mutant that is not phosphorylated by ERK and related MAP kinases in its A/B domain strongly suggests that NH 2 -terminal phosphorylation inhibits the adipogenic activity of the wild type receptor. This also provides a molecular basis for an earlier observation that retroviral expression of NH 2 -terminally truncated PPAR␥2 induces murine preadipocyte differentiation more strongly than its wild type counterpart (12). The precise role of PPAR␥ phosphorylation in regulation of adipo-cyte differentiation remains to be elucidated. However, the ability to modulate the activity of this adipogenic transcription factor by altering its phosphorylation state might provide a novel strategy for development of anti-adipogenic therapeutic agents to treat obesity.