A Dominant-negative Peroxisome Proliferator-activated Receptor γ (PPARγ) Mutant Is a Constitutive Repressor and Inhibits PPARγ-mediated Adipogenesis

The nuclear receptor peroxisome proliferator-activated receptor γ (PPARγ) promotes adipocyte differentiation, exerts atherogenic and anti-inflammatory effects in monocyte/macrophages, and is believed to mediate the insulin-sensitizing action of antidiabetic thiazolidinedione ligands. As no complete PPARγ antagonists have been described hitherto, we have constructed a dominant-negative mutant receptor to inhibit wild-type PPARγ action. Highly conserved hydrophobic and charged residues (Leu468 and Glu471) in helix 12 of the ligand-binding domain were mutated to alanine. This compound PPARγ mutant retains ligand and DNA binding, but exhibits markedly reduced transactivation due to impaired coactivator (cAMP-response element-binding protein-binding protein and steroid receptor coactivator-1) recruitment. Unexpectedly, the mutant receptor silences basal gene transcription, recruits corepressors (the silencing mediator of retinoid and thyroid receptors and the nuclear corepressor) more avidly than wild-type PPARγ, and exhibits delayed ligand-dependent corepressor release. It is a powerful dominant-negative inhibitor of cotransfected wild-type receptor action. Furthermore, when expressed in primary human preadipocytes using a recombinant adenovirus, this PPARγ mutant blocks thiazolidinedione-induced differentiation, providing direct evidence that PPARγ mediates adipogenesis. Our observations suggest that, as in other mutant nuclear receptor contexts (acute promyelocytic leukemia, resistance to thyroid hormone), dominant-negative inhibition by PPARγ is linked to aberrant corepressor interaction. Adenoviral expression of this mutant receptor is a valuable means to antagonize PPARγ signaling.

first characterized as a transcription factor that regulates adipocyte-specific gene expression (1) and induces preadipocyte differentiation (2), but is now recognized to have a central role in other biological processes. PPAR␥ mediates inhibition of inflammatory cytokine production (interleukin-6 and tumor necrosis factor ␣) from monocytes (3), and receptor activation by oxidized low density lipoprotein-derived ligands promotes macrophage foam cell formation (4). PPAR␥ activation promotes colonic neoplasia (5), but inhibits the growth of breast cancer cells (6). Thiazolidinediones (TZDs), a novel class of antidiabetic agent that act as insulin sensitizers in vivo, bind PPAR␥ with high affinity (7), and prostaglandin J2 (8) and fatty acids have been proposed to be natural ligands. PPAR␥ regulates target gene transcription as a heterodimer with the retinoid X receptor, and this heterodimeric complex has been shown to be activated synergistically by TZDs and RXR-specific ligands (9). However, no complete synthetic or natural PPAR␥ antagonists have been described hitherto. We have therefore generated a dominant-negative PPAR␥ mutant to inhibit wildtype receptor action.
In keeping with other members of the nuclear receptor superfamily, PPAR␥ exhibits a modular structure consisting of a central DNA-binding domain, an amino-terminal activation domain, and a carboxyl-terminal ligand-binding domain (LBD) that encompasses a strong ligand-dependent transactivation (AF-2) function. The extreme C terminus of the PPAR␥ LBD forms an amphipathic ␣-helix that can also be delineated in a number of other nuclear receptors. There is striking conservation of hydrophobic (leucine) and negatively charged (glutamic acid) residues within this motif, and mutational analyses in the estrogen (10), thyroid (11,12), and retinoic acid (13) receptors have shown that they are critical for ligand-dependent transactivation and the recruitment of nuclear receptor coactivators (14). Resistance to Thyroid Hormone is associated with diverse thyroid hormone ␤ (TR␤) receptor mutations that inhibit the action of their wild-type counterparts in a dominant-negative manner (15). We have previously described a natural mutation of the conserved hydrophobic residue (Leu 454 ) in the AF-2 domain of TR␤ that exhibits strong dominant-negative activity and is associated with marked refractoriness to thyroid hormone action in vivo (16).
Here, we describe the mutation of homologous hydrophobic and charged residues (L468A and E471A) in PPAR␥. The compound mutant receptor exhibits impaired transcriptional activation and coactivator recruitment. In addition, it silences basal transcription by recruitment of corepressors and is a potent dominant-negative inhibitor of wild-type PPAR␥ action. Adenoviral expression of this mutant receptor in human preadipocytes inhibits thiazolidinedione-induced target gene transcription and cellular differentiation, providing direct evidence for the role of PPAR␥ in adipogenesis.
Hormone and DNA Binding Assays-Hormone binding assays were performed using bacterially expressed GST-PPAR␥ LBD fusion proteins and the PPAR␥-specific radioligand 125 I-SB236636 in a modification of a previously described filter binding assay (23). DNA binding was assessed as described previously (15) using in vitro translated WT PPAR␥1, L468A/E471A PPAR␥1, and human RXR␣ and oligonucleotide duplexes encoding the acyl-CoA oxidase PPARE (24).
Transfection Assays-Calcium phosphate-mediated transient transfection was performed in 24-well cultures of 293EBNA and JEG-3 cells. Each well was cotransfected with 50 -100 ng of receptor expression vector, 500 ng of reporter vector, and 100 ng of the internal control plasmid Bos-␤-gal. Cells were harvested and assayed as described previously (15).
Immunoprecipitation and Western Blot Analysis-293T cells were grown to 60% confluence in 10-cm dishes, transfected with 7.5 g of FLAG-tagged WT PPAR␥ or FLAG-tagged L468A/E471A PPAR␥, and cultured in Dulbecco's modified Eagle's medium and 10% fetal bovine serum with or without 1 M BRL49653. The following day, cells were lysed in ice-cold buffer (50 mM Tris-HCl, 0.15 M NaCl, 5 mM EDTA, 0.5% Nonidet P-40, Roche Molecular Biochemicals protease inhibitor mixture, 250 mM Na 3 PO 4 , and 0.1 mM okadaic acid (pH 8.0)) with or without 10 M BRL49653. Following centrifugation at 12,000 ϫ g for 10 min at 4°C, the supernatants were precleared with protein A beads and incubated with goat polyclonal anti-SMRT antibody (N-20, Santa Cruz Biotechnology). For Western blot analysis, detection was performed with anti-SMRT and mouse monoclonal anti-FLAG (Research Diagnostics, Inc.) antibodies.
Preadipocyte Culture-Preadipocytes (isolated from human breast adipose tissue by collagenase digestion) were cultured in serum-containing medium, and differentiation was induced using serum-free medium Ϯ 1 M BRL49653 as described previously (25). b-DNA assays (26) for adipocyte P2 (aP2) expression were performed at 24 h, whereas morphologic assessment and glycerol-3-phosphate dehydrogenase activity determination (25) were performed on day 10.

RESULTS
The transcriptional activity of WT and mutant L468A/E471A PPAR␥ was assayed by cotransfection of receptor expression vectors together with a reporter gene (PPARETKLUC) containing three copies of the PPARE from the acyl-CoA oxidase gene linked to the thymidine kinase promoter and luciferase (Fig. 1). Cells transfected with the WT receptor exhibited a strong liganddependent transcriptional response following exposure to increasing concentrations of the thiazolidinedione BRL49653. In comparison, the mutant receptor showed negligible transcriptional activity even at the highest concentration of ligand. However, ligand binding assays with bacterially expressed WT or mutant GST-PPAR␥ LBD fusion proteins and the radiolabeled thiazolidinedione 125 I-SB236636 (27) indicated that the  mutant receptor retained significant ligand binding (WT K d ϭ 45 Ϯ 12 nM; L468A/E471A K d ϭ 200 Ϯ 60 nM), suggesting that this did not account for its transcriptional inactivity. Likewise, DNA binding assays performed using WT or mutant PPAR␥, retinoid X receptor, and radiolabeled PPARE showed comparable formation of heterodimeric complexes (data not shown).
By analogy with the effect of homologous mutations in TR␤ (16), we hypothesized that the interaction of the L468A/E471A PPAR␥ mutant with transcriptional coactivator proteins might be altered. In a protein-protein interaction assay using bacterially expressed GST-PPAR␥ LBD fusion proteins, the WT receptor showed strong ligand-dependent recruitment of 35 Slabeled CBP and SRC-1 proteins (Fig. 2). In contrast, liganddependent coactivator recruitment by the L468A/E471A PPAR␥ mutant was abolished.
To assess dominant-negative inhibition by the L468A/E471A PPAR␥ mutant, cells were transfected with the WT receptor plus an equal amount of mutant receptor and increasing concentrations of BRL49653 (Fig. 1). In the presence of the L468A/ E471A mutant, reporter gene activation was markedly attenuated (ϳ50% of the WT response) at all ligand concentrations, whereas the transcriptional response to WT plus further WT receptor was unchanged (data not shown).
Cells transfected with empty expression vector (pcDNA3) showed a small but significant response, reflecting transcriptional activation mediated by low levels of endogenous PPAR␥ (unpublished Western blotting data not shown) in 293 cells (Fig.  1). In comparison, cells transfected with the L468A/E471A mutant exhibited even lower transcriptional activity, presumably reflecting dominant-negative inhibition of endogenous WT receptor (Fig. 1).
Other members of the nuclear receptor family (e.g. TR and RAR) are able to silence basal gene transcription in the absence of ligand by binding corepressor proteins such as NCoR (18) and SMRT (19). Furthermore, corepressor recruitment has been shown to be essential for dominant-negative inhibition by natural TR␤ mutants (28). We therefore examined the properties of the unliganded L468A/E471A PPAR␥ mutant. In comparison to empty pcDNA3 vector, the WT receptor exhibited moderate (ϳ5-fold) constitutive basal activation, whereas the PPAR␥ mutant showed striking silencing of basal transcription (pcDNA3 ϭ 1.0; L468/E471 ϭ 0.25) (Fig. 1, inset). To further substantiate that this silencing is PPAR␥-mediated, cells were transfected with vectors expressing either the Gal4 DBD alone or linked to the L468A/E471A PPAR␥ mutant LBD. Again, marked repression of basal transcription was observed (Gal4 ϭ 1.0; Gal4-PPAR␥ mutant ϭ 0.15) (Fig. 3a), suggesting that PPAR␥ might interact with corepressors in vivo.
The ligand-binding domain of TR mediates interaction with NCoR or SMRT when unliganded, and the addition of T 3 promotes corepressor dissociation and coactivator recruitment (18, 293T cells, cultured in 10-cm plates, were transfected with 7.5 g of expression vectors encoding SMRT and FLAG epitope-tagged WT or L468A/E471A PPAR␥. Whole cell lysates were treated with polyclonal anti-SMRT antibody in the absence or presence of 10 M BRL49653 as shown, and precipitates were Western-blotted and probed with anti-FLAG antibody. The dual band corresponds to phosphorylated and unphosphorylated forms of PPAR␥. Cell lysates were also Westernblotted and probed with anti-FLAG and anti-SMRT antibodies to verify comparable transfection efficiencies of PPAR␥ and SMRT, respectively. IP, immunoprecipitate. 19). When cotransfected with the Gal4-PPAR␥ mutant, the unliganded TR LBD relieved transcriptional silencing by the mutant receptor, and this effect was reversed by the addition of T 3 (Fig. 3a). In contrast, coexpression of a mutant (P214R) TR LBD with impaired corepressor binding (29) did not affect basal repression by the Gal4-PPAR␥ mutant (Fig. 3a). Cotransfection of a mutant (L454A) TR LBD that exhibits impaired hormonedependent corepressor release (30) also relieved silencing by the Gal-PPAR␥ mutant, but this persisted following the addition of T 3 (Fig. 3a).
To specifically address the role of individual corepressors in mediating silencing by the Gal4-PPAR␥ mutant, we examined the effects of mSiah2, a novel protein that has been shown to target the corepressor NCoR for proteasomal degradation (22). Cotransfected mSiah2 significantly impaired transcriptional silencing by the Gal4-PPAR␥ mutant (Fig. 3b). Furthermore, coexpression of NCoR was able both to enhance basal repression by the mutant receptor and to restore silencing in the presence of mSiah2 in a dose-dependent manner (Fig. 3b).
To examine the association of WT and mutant PPAR␥ with corepressor in vivo, cells were cotransfected with an SMRT expression vector together with FLAG epitope-tagged fulllength WT or L468A/E471A PPAR␥ in the absence and presence of BRL49653. Following immunoprecipitation with an SMRT antibody, PPAR␥ complexed with corepressor was quantitated by Western blotting (Fig. 4). In the absence of ligand, both WT and mutant PPAR␥ bound SMRT, with greater quantitative binding by the mutant receptor. The addition of 10 M BRL49653 ligand resulted in complete dissociation of SMRT from the WT receptor, whereas the L468A/E471A mutant retained significant corepressor binding, suggesting that liganddependent release of corepressor from the PPAR␥ mutant might be impaired. We tested this hypothesis in a mammalian two-hybrid assay using the Gal4-NCoR (residues 2276 -2454) or Gal4-SMRT (residues 982-1448) fusion together with the VP16 construct with WT or mutant PPAR␥ LBD. In the ab- sence of ligand, both WT and mutant PPAR␥ were recruited comparably to Gal4-SMRT, whereas interaction of the L468A/ E471A mutant with the Gal4-NCoR fusion was slightly enhanced. In contrast to a dose-dependent dissociation of the WT receptor from corepressors following the addition of ligand, the release of the PPAR␥ mutant from both SMRT and NCoR was markedly impaired (Fig. 3c).
To introduce the L468A/E471A PPAR␥ mutant into primary cells and tissues, we have constructed a recombinant adenovirus expressing the mutant receptor as well as GFP. In the first instance, we tested the ability of PPAR␥ mutant adenovirus (Ad␥ m ) to inhibit the action of transfected nuclear receptors. Mutant adenoviral infection of cells blocked ligand-dependent transactivation by both human PPAR␥1 and PPAR␥2 isoforms, whereas receptor-mediated activation was unaffected in cells infected with control adenovirus expressing GFP alone (AdGFP) (Fig. 5, a and b). Furthermore, the mutant receptor adenovirus only partially inhibited PPAR␣-mediated signaling (Fig. 5c) and was unable to block ligand-dependent transactivation by human RAR␣ (Fig. 5d), human RXR␣ (Fig. 5e), or human TR␤ (Fig. 5f).
PPAR␥ plays a central role in murine preadipocyte differentiation (2). We have shown previously that thiazolidinediones promote the differentiation of cultured human preadipocytes (25) and therefore tested the effect of the PPAR␥ mutant adenovirus on this process. Following infection with mutant receptor adenovirus, TZD-induced differentiation of cells into lipidladen adipocytes was markedly inhibited compared with cells infected with GFP adenovirus (Fig. 6c). The degree of differentiation was assessed quantitatively by measurement of glycerol-3-phosphate dehydrogenase enzyme activity (25) and aP2 mRNA accumulation (26) normalized to cell number as described previously. Thiazolidinedione induction of aP2 mRNA (Fig. 6a) and glycerol-3-phosphate dehydrogenase activity (Fig.  6b) was significantly reduced following infection with PPAR␥ mutant adenovirus compared with uninfected or GFP virus-infected cells. Thus, the L468A/E471A PPAR␥ mutant is capable of blocking responses mediated by endogenous wild-type PPAR␥.

DISCUSSION
The carboxyl terminus of a number of nuclear receptors, including PPAR␥, contains a C-terminal amphipathic ␣-helix that is required for ligand-dependent transcriptional activation (AF-2) function (10,11,13). We have mutated conserved hydrophobic (Leu 468 ) and negatively charged (Glu 471 ) residues in the putative AF-2 domain of PPAR␥ to alanine. Functional studies indicate that transcriptional activation by this compound mutant is severely impaired even in the presence of saturating concentrations of thiazolidinedione ligand sufficient to overcome its modestly reduced ligand binding affinity (Fig.  1). Protein-protein interaction assays indicate negligible liganddependent recruitment of CBP and SRC-1 coactivators by the PPAR␥ mutant, accounting for its functional impairment. The crystal structure of the PPAR␥ LBD in complex with the interaction domain of the coactivator SRC-1 (31) reveals that Leu 468 and Glu 471 are both situated at the receptor coactivator interface. The side chain of Glu 471 is oriented such that it makes several hydrogen bonds with the backbone amino groups of the coactivator helix, with the negative charge of the carboxyl complementing the positive end of the helix dipole. Leu 468 is situated at the bottom of the hydrophobic cleft into which the coactivator helix binds and as a result is completely buried between hydrophobic residues on helix 3 and others on the coactivator helix. It is clear that the impairment of coactivator binding seen in functional studies with this mutant is entirely consistent with the role of these residues in the structure of the receptor-coactivator complex.
The L468A/E471A PPAR␥ mutant was also able to inhibit the action of its WT counterpart in a dominant-negative manner (Fig. 1). Mutations in the C-terminal AF-2 domain of other FIG. 6. a, Ad␥ m impairs BRL49653-mediated induction of aP2 mRNA in human preadipocytes. Preadipocytes were grown to confluence in 96-well plates, infected with no virus (nil) or with 0.8 ϫ 10 7 plaqueforming units/well AdGFP or Ad␥ m , and harvested at 24 h. aP2 values were standardized against the glyceraldehyde-3phosphate dehydrogenase (GAPDH) housekeeping gene. Results represent the combined data from four independent experiments. b and c, Ad␥ m inhibits BRL49653-induced preadipocyte differentiation. b, human preadipocytes from breast tissue were infected with no virus or with 9.6 ϫ 10 7 plaque-forming units/well AdGFP or Ad␥ m and cultured in differentiation medium Ϯ 0.1 M BRL49653. At 10 days, cells were harvested, and glycerol-3phosphate dehydrogenase activity/g of protein was determined. Results represent the means Ϯ S.E. of four independent experiments. c, phase-contrast photographs (magnification ϫ 200) of oil red O-stained preadipocytes infected with AdGFP or Ad␥ m as shown and grown for 10 days in differentiation medium supplemented with 0.1 M BRL49653. nuclear receptors also generate mutant proteins with strong dominant-negative activity; for example, this region is deleted in the oncogene v-erbA, a potent inhibitor of TR and RAR action (32). We (16) and others (33) have described powerful dominant-negative amphipathic ␣-helix TR␤ mutants in the syndrome of resistance to thyroid hormone.
A subset of nuclear receptors including TR and RAR have been shown to repress basal transcription in the absence of ligand by recruitment of corepressor proteins such as NCoR (18) and SMRT (19). However, the role of corepressors in PPAR␥ signaling remains unclear. Whereas PPAR␥ can interact weakly with NCoR and SMRT in vitro, the WT receptor exhibits negligible transcriptional repression in vivo (24) (Fig.  1, inset), although mitogen-activated protein kinase-dependent phosphorylation has been shown to inhibit AF-2 function via SMRT recruitment (34). In contrast, our observations indicate that the L468A/E471A PPAR␥ mutant is a potent transcriptional repressor. Repression is exhibited by both the full-length mutant receptor (Fig. 1) as well as a Gal4-PPAR␥ LBD fusion (Fig. 3a), indicating that its silencing function is independent of N-terminal domain phosphorylation. Coexpression of the unliganded TR␤ LBD attenuates repression by the PPAR␥ mutant, suggesting that this function is mediated by shared cellular factors (35), and a TR␤ LBD mutant (P214R) that is defective for corepressor binding fails to inhibit repression. Evidence that NCoR partly mediates silencing by the PPAR␥ mutant is provided by the observation that coexpression of mSiah2, which targets NCoR for proteasomal degradation (22), also attenuates repression (Fig. 3b). Co-immunoprecipitation experiments (Fig. 4) demonstrate that the L468A/E471A PPAR␥ mutant interacts with SMRT in vivo, suggesting that this corepressor may also mediate transcriptional silencing. In addition to enhanced corepressor binding, the L468A/E471A PPAR␥ mutant also exhibits impaired ligand-dependent corepressor release (Fig. 3c), indicating a role for helix 12 of PPAR␥ in corepressor dissociation as has been documented with other nuclear receptors (30,33,36). Our observation that the dominant-negative PPAR␥ mutant is a powerful repressor is consonant with the properties of other nuclear receptors, including TR␤ mutants in Resistance to Thyroid Hormone (28), PML-RAR in acute promyelocytic leukemia (37), and v-erbA (38). Furthermore, corepressor recruitment has been shown to be required for dominant-negative inhibition (28). Our finding that the PPAR␥ mutant is a strong repressor raises the question as to why WT PPAR␥ appears to lack silencing activity. The crystal structure of the apo-PPAR␥ LBD indicates that it is possible for helix 12 to adopt the same conformation as the liganded receptorcoactivator complex. It is therefore likely that, in the absence of ligand, PPAR␥ is still able to recruit coactivator, albeit less efficiently than the holoreceptor. Abolition of such coactivator binding in the L468A/E471A PPAR␥ mutant facilitates corepressor binding, unmasking transcriptional repression.
Finally, we have used a recombinant adenovirus expressing the PPAR␥ mutant to block endogenous wild-type receptor action. It is well established that PPAR␥ is a key mediator of adipogenesis. Using a recombinant PPAR␥ mutant adenovirus that selectively inhibits thiazolidinedione-dependent PPAR␥ activation but not other nuclear receptor signaling pathways (Fig. 5), we have been able to inhibit human preadipocyte differentiation and induction of aP2, a PPAR␥ target gene, in response to thiazolidinediones (Fig. 6). These results provide compelling evidence that TZDs act directly via PPAR␥ to promote human preadipocyte differentiation. Our results also validate the utility of the PPAR␥ mutant in investigating receptor action in vivo. In contrast to a chemical antagonist, the dominant-negative mutant receptor can be used to selectively in-hibit thiazolidinedione-dependent PPAR␥ action in particular tissues. For example, the relative importance of insulin-sensitizing effects of TZDs in adipocytes versus skeletal muscle could be investigated by generating transgenic mice with the L468A/ E471A PPAR␥ mutant cDNA linked to tissue-specific promoters to target mutant receptor expression.