The Endocrine Disruptor Monoethyl-hexyl-phthalate Is a Selective Peroxisome Proliferator-activated Receptor γ Modulator That Promotes Adipogenesis*

The ability of pollutants to affect human health is a major concern, justified by the wide demonstration that reproductive functions are altered by endocrine disrupting chemicals. The definition of endocrine disruption is today extended to broader endocrine regulations, and includes activation of metabolic sensors, such as the peroxisome proliferator-activated receptors (PPARs). Toxicology approaches have demonstrated that phthalate plasticizers can directly influence PPAR activity. What is now missing is a detailed molecular understanding of the fundamental basis of endocrine disrupting chemical interference with PPAR signaling. We thus performed structural and functional analyses that demonstrate how monoethyl-hexyl-phthalate (MEHP) directly activates PPARγ and promotes adipogenesis, albeit to a lower extent than the full agonist rosiglitazone. Importantly, we demonstrate that MEHP induces a selective activation of different PPARγ target genes. Chromatin immunoprecipitation and fluorescence microscopy in living cells reveal that this selective activity correlates with the recruitment of a specific subset of PPARγ coregulators that includes Med1 and PGC-1α, but not p300 and SRC-1. These results highlight some key mechanisms in metabolic disruption but are also instrumental in the context of selective PPAR modulation, a promising field for new therapeutic development based on PPAR modulation.

The ability of pollutants to affect human health is a major concern, justified by the wide demonstration that reproductive functions are altered by endocrine disrupting chemicals. The definition of endocrine disruption is today extended to broader endocrine regulations, and includes activation of metabolic sensors, such as the peroxisome proliferator-activated receptors (PPARs). Toxicology approaches have demonstrated that phthalate plasticizers can directly influence PPAR activity. What is now missing is a detailed molecular understanding of the fundamental basis of endocrine disrupting chemical interference with PPAR signaling. We thus performed structural and functional analyses that demonstrate how monoethyl-hexyl-phthalate (MEHP) directly activates PPAR␥ and promotes adipogenesis, albeit to a lower extent than the full agonist rosiglitazone. Importantly, we demonstrate that MEHP induces a selective activation of different PPAR␥ target genes. Chromatin immunoprecipitation and fluorescence microscopy in living cells reveal that this selective activity correlates with the recruitment of a specific subset of PPAR␥ coregulators that includes Med1 and PGC-1␣, but not p300 and SRC-1. These results highlight some key mechanisms in metabolic disruption but are also instrumental in the context of selective PPAR modulation, a promising field for new therapeutic development based on PPAR modulation.
Exposure to endocrine disrupting chemicals can lead to detrimental effects in human and animal populations by promoting or inhibiting the synthesis, elimination, and action of hormones (1,2). At the molecular level, these compounds act by activating or inhibiting enzymatic activities of hormone biosynthesis and by targeting nuclear receptors (NRs), 4 a class of transcription factors that regulate gene expression programs in response to lipophilic hormones and mediators. In the past, research on endocrine disruption has focused on reproductive defects caused by interference with steroid signaling, largely linked to modulation of steroid NR activity (3). However, NRs constitute a large family of receptors regulating diverse physiological functions, out of which most members share the capacity to regulate gene expression in response to ligand binding. The concept of endocrine disruption is now being broadened to other receptors implicated in different aspects of homeostatic regulation (4).
Given their central role in metabolic regulations (5,6), peroxisome proliferator-activated receptors (PPARs) potentially constitute important targets for environmental factors. PPARs are lipid sensors that cooperate in different organs to adapt gene expression to a given metabolic status. PPAR␥ (NR1C3) controls fat storage in the adipose tissue by promoting the differentiation and the survival of adipocytes, but also plays major roles in the control of insulin sensitivity (7). At the molecular level, PPAR action relies on its constitutive association with the retinoid X receptor (NR2B) (8,9), which allows binding of the heterodimer to specific response elements located in target gene promoters. The unliganded receptors are engaged in large complexes of corepressors and coactivators that can promote repression and activation according to the promoter context, respectively (9 -11). To induce full transcriptional activation, ligand binding enhances the recruitment of coactivators acting through the remodeling of the structure and epigenetic marking of chromatin and by contacting the basal transcriptional machinery (12). PPAR activity can be pharmacologically modulated to treat major metabolic disorders (13). Thiazolidinediones are PPAR␥ activating drugs clinically used as insulin sensitizers but PPAR␥-targeted therapeutics suffer from important adverse effects (7). The hope of better pharmaceutical strategies arises from the identification of selective PPAR␥ modulators (SPPAR␥Ms) capable of uncoupling the beneficial actions of current PPAR␥ agonists from their side effects (13,14).
The large PPAR ligand binding pocket, which can accommodate a wide variety of ligands, raises the question of whether PPAR activity and PPAR-regulated pathways could be affected by exposure to endocrine disrupting chemicals. Actually, PPAR␣ was discovered as the receptor mediating hepatic peroxisome proliferation and carcinogenesis in rodents in response to a wide class of chemicals that include pesticides, industrial solvents, and plasticizers (15,16). We have focused the present study on the interference of phthalate esters with PPAR-regulated processes. Phthalates are widely used industrial chemicals that primarily serve as plasticizers to soften PVC but are also found in cosmetics, perfumes, and certain drugs as well as in industrial paints and solvents. Diethyl-hexyl-phthalate (DEHP) is among the most abundantly used phthalate esters with an annual worldwide production estimated around 2 million tons according to Swiss authorities (Federal Office of Public Health, www.bag.admin.ch/themen/chemikalien/00228/ 01378). DEHP is incorporated non-covalently into flexible plastics used for manufacturing a wide variety of daily products including medical devices and food packaging and its propensity to leach can expose humans to high concentrations of this compound (17). The biological effects of DEHP are hence of major concern but so far elusive. Upon ingestion, pancreatic lipases present in the intestine convert DEHP to its monoester equivalent monoethyl-hexyl-phthalate (MEHP), which is preferentially absorbed (18). In addition, MEHP can also be produced by plasmatic and hepatic lipases, which transform DEHP directly reaching the blood through absorption or medical contamination. This metabolite activates the three PPAR isotypes and mediates the action of DEHP on hepatic peroxisome proliferation via PPAR␣ (19 -21).
This study, aimed at going beyond the toxicology approach, focuses on the molecular mechanisms through which MEHP modulates PPAR␥ signaling. For that purpose, we use a combination of molecular and cellular assays to directly monitor the action of the receptor in living cells. We demonstrate that MEHP promotes PPAR␥-dependent adipogenesis, albeit to a lower extent than the full agonist rosiglitazone. MEHP induces the expression of a subset of PPAR␥ target genes required for adipogenesis when compared with rosiglitazone. Interestingly, the binding modes of MEHP and rosiglitazone to the ligand binding pocket of PPAR␥ seem similar. However, MEHP selectively modulates PPAR␥ activity according to the promoter context by promoting differential interactions with coregulators. Thus, in addition to understanding the molecular actions of MEHP beyond the toxicological observations, the combination of microscopy techniques and chromatin immunoprecipi-tation provides a strong molecular basis in living cells for the concept of selective PPAR␥ modulation.
Cell Culture and Transient Transfection Assays-COS7, C2C12, HeLa, and 3T3L1 cells from ATCC were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Invitrogen). Transient transfection assays were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. After transfection, cells were let to recover in medium supplemented with 10% fetal calf serum for 5 h and then grown in serum-free medium for 18 h in the presence of ligand or vehicle only. Unless otherwise stated, Wy14,643 (Cayman Chemical Co.), L-165041 (synthesized at custom in the laboratory), rosiglitazone (Sigma), MEHP (ICI), and GW9662 (synthesized at custom by Zydus Research Center, India) were used at respective final concentrations of 10 Ϫ5 , 5 ϫ 10 Ϫ6 , 10 Ϫ6 , 10 Ϫ4 , and 2 ϫ 10 Ϫ5 M.
Reporter Assay-PPAR activity was monitored on a (PPRE) 3luciferase reporter construct kindly provided by Dr. Evans (Salk Institute, San Diego, CA). Luciferase activity assays were performed with the Promega dual-reporter kit, according to the manufacturer's instructions. Renilla luciferase encoded by the normalization vector phRLTK (Promega) was used as an internal control for firefly luciferase normalization.
Adipocyte Differentiation-Two-day post-confluent 3T3L1 pre-adipocytes were induced to differentiate using two different protocols: a 10-day treatment with 10 g/ml insulin and PPAR␥ ligands (rosiglitazone or MEHP) or a 2-day treatment with a differentiation mixture (10 g/ml insulin, 1 M dexamethasone, 0.5 mM isobutylmethylxanthine) followed by an 8-day treatment with 10 g/ml insulin only. In both cases, the medium was changed every 48 h. After differentiation, cells were stained with Oil Red O for morphological analyses or RNA was extracted using phenol/chloroform (TRIzol, Invitrogen) and purified with Qiagen columns. For RNA extractions, four independent cultures were performed per condition. To determine triglyceride content, 3T3L1 cultures differentiated as described above in a 10-cm plate were lysed in 1500 l of phosphate-buffered saline, 0.1% Nonidet P-40 with a Dounce homogenizer and cytoplasmic extracts were collected. Relative triglyceride levels were evaluated by adding 100 l of cytoplasmic extract to 200 l of Infinity triglyceride reagent (Thermo-Electron Corporation) and measuring the absorbance at 500 nm. The levels of triglyceride were normalized to protein levels measured with the Bradford reagent (Bio-Rad) at 595 nm.
Quantitative Reverse Transcription-PCR-Reverse transcription was performed with random hexamers on 1 g of total RNA using the SuperScript first-strand synthesis system (Invitrogen) and the reaction was diluted 100 times for amplification. PCR were performed in triplicate in 384-well plates on an Applied Biosystems 7900HT cycler using commercial Taq-Man probes (Applied Biosystems). Results were normalized to 3 housekeeping genes (TATA-box binding protein, eukaryotic translation elongation factor ␣1, and ribosomal protein S9) and quantified using qBase (25).
Microarray Experiments-The mouse cDNA microarrays used in this study consisted of roughly 17,000 PCR products generated from cDNA clones and control DNAs spotted onto Nexterion AL slides (Schott). A complete description of the slides and their content can be obtained from the Lausanne DNA Array Facility (www.unil.ch/dafl). RNA quality was assessed using the RNA 6000 Nanochip assay (Agilent Technologies). A single round of amplification was performed with 3 g of total RNA using the MessageAmp aRNA Amplification Kit (Ambion) following the protocol provided. 5 g of amplified RNA was reverse transcribed with SuperScript II reverse transcriptase (Invitrogen) and random hexamers for 2 h at 42°C, in the presence of either Cy3-dCTP or Cy5-dCTP (GE Healthcare). RNA was hydrolyzed by heating at 65°C for 15 min in a basic environment. The solution was then neutralized and labeled cDNA was purified using a Qiagen MiniElute PCR Purification kit. The Cy3-and Cy5-labeled targets were combined, mixed with Cot 1 DNA (Invitrogen), polyadenylic acid (Sigma), and yeast tRNA (Sigma), and hybridized on custom glass microarrays at 64°C for 20 h in 3ϫ SSC, 0.4% SDS. Slides were then washed twice for 5 min in 2ϫ SSC, 0.1% SDS, twice for 1 min in 0.2ϫ SSC, once for 1 min in 0.1ϫ SSC, and once for 5 min in 0.1ϫ SSC, 0.1% Triton X-100. After drying, slides were scanned on a microarray scanner (Agilent Technologies) and the resulting TIFF images were analyzed using the GenePix Pro 6.0 software (Molecular Devices).
Statistical analysis was performed with the R software packages sma (26) and limma (27). Gene expression was quantified with the sma package using print tip group lowess normalization without background subtraction (26,28). The three treatments (V, rosiglitazone, and MEHP) were compared pairwise using a linear model. Differential expression was assessed by fitting the linear model for the effects of the two treatments considered and computing moderated t statistics and Ben-jamini and Hochberg false discovery rates with the limma package (27,29).
Fluorescence Correlation Spectroscopy (FCS) Experiments-FCS was performed on a LSM510 ConfoCor2 (Zeiss) as previously described (8). Briefly, intensity fluctuations of the fluorescence collected between 505 and 550 nm were detected using an avalanche photodiode and recorded at 5 spots in each nucleus, with a repetition of 10 measurements per spot. Using the Origin software, the autocorrelation curves were fitted to an anomalous diffusion model to derive diffusion times, which were subsequently converted to diffusion coefficients and averaged over at least 10 cells.
Fluorescence Resonance Energy Transfer (FRET) Experiments-Sensitized emission FRET was monitored at 37°C over at least 50 living cells using a TCS SP2 AOBS confocal microscope (Leica) as previously described (9,31). Briefly, fluorescence was recorded in three different settings: CFP ex , 405 nm, CFP em , 465-485 nm; YFP ex , 514 nm, YFP em , 525-545 nm; FRET ex , 405 nm, FRET em , 525-545 nm. Laser power and detector gain were adjusted in the different channels so that equimolar concentrations of CFP and YFP give equal intensities (equimolar concentrations of CFP and YFP were obtained by expressing a fusion protein of CFP and YFP spaced by 475 residues). Settings were kept unchanged for analysis of all samples. As the ratio of CFP spectral bleed-through into the FRET channel (SBT CFP ϭ I FRET /I CFP ) determined on cells expressing CFP alone has been previously observed to vary with CFP intensity, this variation was modeled using an exponential fit (31). In contrast, the ratio of YFP spectral bleed-through into the FRET channel (SBT YFP ϭ I FRET /I YFP ) determined on cells expressing YFP alone was a constant equal to the average ratio. FRET measured in co-expressing cells was then corrected for spectral bleed-throughs and normalized for expression levels according to the following formula (31), where I FRET , I CFP , and I YFP are the intensities measured with the FRET, CFP, and YFP settings in the presence of both the donor and acceptor, e, f, and g are the constants determined by the fitting of the CFP SBT ratio in the presence of CFP only, and b is the average YFP SBT ratio in the presence of YFP only. Structural Modeling-Missing parameters for MEHP, for use in conjunction with the CHARMM22 (32) all atoms molecular mechanics force field, were derived from the Merck Molecular Force Field (MMFF (33)), by taking the dihedral angle term as is, but only the quadratic part of the bond and angle energy terms. The partial charges and van der Waals parameters of the ligand atoms were taken from the MMFF. The ligand was modeled with all hydrogens.
To take account of a possible induced fit of the protein upon ligand complexation, the docking of MEHP to hPPAR␥ was realized based on two experimental x-ray structures of the hPPAR␥ ligand-binding domain (LBD) in complex with two different ligands. The first corresponds to the structure of the complex with the agonist AZ242 obtained at 2.35-Å resolution (PDB code 1I7I (34) in the Protein Data Bank (35)). The second structure corresponds to the complex of hPPAR␥ with an ␣-aryloxyphenylacetic acid partial agonist obtained at 2.5-Å resolution (code 1ZEO (36) in the PDB). The two structures differ in the conformation of residues Phe 282 (H3), Gln 286 (H3), and Phe 363 (loop between H6 and H7). In 1ZEO, these residues, together with residues Ile 281 (H3), Leu 356 (H6), Phe 360 (loop between H6 and H7), Leu 453 (H11), and Leu 465 (loop between H11 and H12), form a hydrophobic pocket where the benzisoxazol group fits. This pocket is closed in the 1I7I structure due to an alternate positioning of Phe 282 , Gln 286 , and Phe 363 . Both ligands of the template complexes have a carboxylate function, like MEHP. They were removed from the binding site prior to MEHP docking.
Four calculations were performed, corresponding to the docking of the R or S configurations of MEHP in the 1I7I or 1ZEO structures. The protein was held fixed in all cases, the protein flexibility being taken into account through the use of different experimental structures. The details of the calculations will be presented separately. 5 In brief, starting from a set of 250 randomly generated initial conformations, positions, and orientations of MEHP inside the known binding site of hPPAR␥, the coordinates of the ligand were refined using several operators, renewing 10% of the population at each generation. The thorough exploration of the accessible conformational space of the ligand relative to the protein surface was submitted during 400 generations to the evolutionary pressure of a scoring function that takes account of the solvent effect thanks to the GB-MV2 implicit solvent model (37). The lowest energy conformation was retained as the proposed binding mode. The accessible conformational space was defined as a 15-Å radius sphere centered on the center of mass of the AZ242 ligand. This region is large enough to explore possible pauses on the protein surface, out of the binding site. Actually, it has been observed that the pauses proposed by EADock for bad ligands are situated out of the known binding site of the targeted protein (data not shown). In all models of MEHP binding to PPAR␥, the energetically most favorable calculated binding mode is situated inside the known binding site of the protein, consistent with the ability of MEHP to transactivate PPAR␥.
Statistical Analyses-Data are represented as average Ϯ S.E. of the mean. Statistical significance was determined using an analysis of variance followed by Tukey's post hoc test.

MEHP Is a Partial PPAR␥ Agonist Whose Efficacy Varies
According to the Cell Type-MEHP was previously reported to activate PPARs in transactivation assays (19 -21). However, the levels of activation varied between studies, possibly because of the use of different cell types, different reporter systems, and receptors from different species. To characterize the action of this phthalate monoester on full-length PPARs, we assessed the ability of MEHP to induce the transcription of a reporter construct containing 3 PPREs upstream of the luciferase cDNA in the presence of the different PPAR isotypes. MEHP could activate mouse PPAR␣, PPAR␤, and PPAR␥1, albeit to a generally lower extent than the reference agonists (Fig. 1A). PPAR activation was detected with doses as low as 3.2 M and reached maximum levels at 100 M. Maximal activation of PPAR by MEHP was approximately half of that achieved with Wy14643 and rosiglitazone for PPAR␣ and PPAR␥, respectively, whereas the maximal levels of PPAR␤ activation achieved with MEHP and L165041 were comparable. We decided to focus this study on the activation of PPAR␥ by MEHP and therefore evaluated potential species and isoform differences. Both the mouse and human PPAR␥1 receptors where activated with similar affinity and efficacy in C2C12 cells (Fig. 1B). The additional 30 N-terminal residues of the PPAR␥2 isoform only mildly affected the ability of MEHP to activate the receptor by slightly increasing the maximal level of activation without affecting affinity (Fig.  1C). Surprisingly, the ability of MEHP to activate PPAR␥ was cell-type dependent (Fig. 1D). Whereas the affinity of the compound for PPAR␥ was comparable in the three cell lines tested (EC 50 around 30 M), the efficacy of MEHP relative to the reference agonist clearly differed between cell lines, the high dose reaching 80% of the rosiglitazone activation level in C2C12 cells, but only 60% in COS7 cells and 35% in HeLa cells. Altogether, these experiments demonstrate that MEHP is a partial pan-PPAR agonist with reduced affinity and efficacy compared with full agonists. To characterize whether differences in binding mode could potentially account for the differences in PPAR␥ activation by MEHP and rosiglitazone, we modeled the binding of MEHP to two reference structures (1I7I (34) and 1ZEO (36)) of the PPAR␥ LBD and compared it to the binding of rosiglitazone available from the crystal structure of the PPAR␥ LBD in complex with this agonist (38). The results obtained for the docking of MEHP in the 1I7I and 1ZEO structures of hPPAR␥ are very similar and both the R-and the S-enantiomer of MEHP could fit in the PPAR␥ LBD (data not shown). Although explored by the ligand during the docking process, the additional pocket of 1ZEO was not used in the proposed binding mode ("Experimental Procedures"). MEHP contacted Ser 289 , His 323 , His 367 , and Tyr 473 (Fig. 2A), a set of residues important for the stabilization of the interaction between rosiglitazone and the receptor (38) (Fig. 2B). Furthermore, the contact between the carboxylate function of the phtalic acid ring and Tyr 473 , a residue from helix 12 important for transcriptional activation, suggests that the activity of MEHP relies on the stabilization of helix 12. This observation supports the absence of activity of the DEHP parent compound (39) where this carboxylate is esterified by a bulky and hydrophobic chain. MEHP and rosiglitazone bind to the PPAR␥ LBD in similar configurations where only one side of the T-shaped binding pocket is occupied and where similar residues are contacted. Thus, the difference in affinity and in efficacy between MEHP and rosiglitazone most likely reflect subtle variations in the binding mode that may lead to less productive conformational changes upon MEHP binding. However, the full characterization of the differential changes in the three-dimensional structure of the LBD would require the crystallization of the PPAR␥ LBD in complex with MEHP.
Selective Recruitment of Coregulators by MEHP-Coregulator recruitment is the major functional determinant that follows conformational changes of the nuclear receptor LBD in response to ligand binding. We thus compared how MEHP and rosiglitazone modulate interactions with corepressors and coactivators, which assist PPAR␥ to regulate transcription. This experiment is classically assayed in vitro. However, many other factors in the nucleus of the living cell may also affect protein-protein interactions. FCS is a technique measuring the diffusion of fluorescent molecules at high temporal resolution that can be used in the living cell. Using this technique, we previously showed that the YFP-PPAR mobility in cells is reduced by ligand binding (8). Because this reduction in mobility is caused by the formation of large complexes through coactivator recruitment (9), we hypothesized that a selective recruitment of coactivators in response to MEHP compared with rosiglitazone binding could potentially translate into different patterns of YFP-PPAR␥ mobility. We thus compared YFP-PPAR␥ diffusion coefficients by FCS in COS7 cells in the presence and absence of rosiglitazone and MEHP. Diffusion coefficients were indeed reduced from 8 to 6.5 m 2 /s in Helixes contacting the ligand are colored as follows: H3, green; H5, orange; H11, red; and H12, blue. Hydrogen atoms were included in the modeling of the binding mode of MEHP but these atoms were removed from the representation for clarity because rosiglitazone hydrogens are not present in the structure from the data base. response to MEHP, whereas they were further reduced to 5 m 2 /s by rosiglitazone (Fig. 3A). The sizes of PPAR␥-coactivator complexes formed by MEHP binding, which are directly evaluated in living cells using this assay, are therefore smaller than those formed in response to rosiglitazone and reflect different complex compositions.
To identify the regulators recruited by rosiglitazone and MEHP, we analyzed their respective ability to trigger corepressor release and coactivator recruitment. Interactions with individual coregulators were assessed in the native context of a living cell using FRET imaging. The levels of energy transfer, which reflect direct interactions, were measured in living cells co-transfected with CFP-tagged PPAR␥ and large coregulator domains tagged with YFP. Transfection conditions were optimized to ensure low and equimolar expression levels and the analyses were restricted to cells without aberrant localization seen upon high levels of expression (8). The strong interaction between PPAR␥ and NCoR observed in the absence of ligand was totally abolished in the presence of rosiglitazone but MEHP had only a partial effect on this interaction (Fig. 3B). Using a similar assay, a basal interaction was detected between PPAR␥ and p300 and the Med1 subunit of the mediator complex (also called TRAP220, DRIP205, and PBP) in the absence of ligand (Fig. 3, C and D). Rosiglitazone significantly enhanced the interaction with p300 and Med1, whereas MEHP promoted the recruitment of Med1 but was totally inefficient in inducing the recruitment of p300. MEHP therefore selectively regulates interactions with coregulators in living cells both at the levels of transcriptional complex size and interactions with individual corepressors and coactivators.
The interaction with PGC1␣ could not be evaluated by FRET as this construct gave no FRET signal with PPAR␥, most probably because of inappropriate fluorophore spacing and orientation. We therefore confirmed our live cell results and extended them to other coregulators using an in vitro pull-down assay where GST-tagged coregulators were immobilized on beads and [ 35 S]mPPAR␥1 recruitment was evaluated (Fig. 4A). In this assay, the binding of SRC-1 to PPAR␥ was very weak and was not influenced by ligand binding but consistent with our FRET assays, MEHP induced only a partial release of NCoR, and a strong recruitment of Med1, but not of p300. The PPAR␥ coactivator 1␣ (PGC-1␣) strongly interacted with PPAR␥ in the absence of ligand and this interaction was modestly enhanced by rosiglitazone but strongly enhanced by MEHP. We also took advantage of this in vitro assay to assess species-and isoformspecific effects in coregulator recruitment in response to MEHP. The profile of coregulator recruitment to human PPAR␥1 and to the PPAR␥2 isoform in response to MEHP was very similar to that observed with mouse PPAR␥1 (Fig.  4B). Altogether, these experiments demonstrate that when compared with the action of the full agonist rosiglitazone, MEHP induces a selective recruitment of PPAR␥ coregulators with stronger interactions with PGC-1␣ but a lack of p300 recruitment.

Selective Recruitment of Coregulators by DNA-bound PPAR␥ in the Presence of MEHP-FCS
and FRET analyses cannot specifically assess the property of the complexes bound to DNA. We thus needed to establish whether the selective interactions in the nucleus of living cells also translate into selective recruitment of coregulators by DNA-bound PPAR␥-containing complexes. Using specific antibody against each of the coregulators, we performed chromatin immunoprecipitation and assessed PPAR␥ target promoter occupancy, in the presence of either rosiglitazone or MEHP (Fig. 5). The promoter of the ribosomal protein P0 was used as a reference gene, not regulated by PPAR␥, and precipitation with the unrelated hemagglutinin antibody served as a negative control. A control experiment using a PPAR␥ antibody could not be reliably realized, likely due to the poor quality of the various antibodies tested (data not shown). As previously described, in the absence of PPAR␥ ligand, NCoR was associated with the glycerol kinase (Gyk) and oxidized low density lipoprotein receptor 1 (Olr1) promoters, but not with the fatty acid-binding protein 4 (Fabp-4)/aP2 promoter, and this association disappeared with rosiglitazone treatment (10,40). Consistent with our FRET and pull-down observations, MEHP induced a partial clearance of NCoR from these two PPAR␥ target promoters. The recruitment of coactivators is also consistent with our observation in the living cells. Unlike rosiglitazone, MEHP did not promote recruitment of p300 or SRC-1 on target promoters. However, it induced a recruitment of Med1 to slightly lower levels than those achieved by rosiglitazone but a recruitment of PGC-1␣ at much higher levels. Thus, the global pattern of promoter-specific and ligand-dependent co-regulator recruitment strikingly correlates with the interaction assays.
MEHP Promotes Adipocyte Differentiation through PPAR␥-To investigate the action of MEHP on PPAR␥-regulated pathways at the cellular level, we studied the influence of MEHP on adipogenesis, a well characterized PPAR␥-regulated function. As expected, a 10-day treatment of 3T3L1 pre-adipocytes with the full PPAR␥ agonist rosiglitazone in the presence of insulin strongly induced adipogenesis in a dose-dependent manner, as evidenced by Oil Red O staining at low (Fig. 6A) and high magnification (Fig. 6B). Whereas insulin alone had only mild effects, the combination of insulin and MEHP also induced dose-dependent adipogenic effects, whereas the DEHP parent compound was inefficient in inducing adipocyte differentiation. The ability of MEHP to promote adipogenesis appeared at a dose of 10 M and was maximal at 100 M, with no further enhancement at higher doses (data not shown). However, the maximal effect induced by MEHP was lower than that induced by rosiglitazone. The actions of both ligands on adipocyte differentiation were confirmed by quantification of cellular triglyceride contents that were strongly induced by rosiglitazone but only partially induced by MEHP (Fig. 6C). Thus, although MEHP significantly induces adipocyte differentiation, this phthalate monoester has a reduced adipogenic potential compared with rosiglitazone, most likely related to the selective modulation of its activity through differential coregulator recruitment.
To address whether the actions of MEHP on adipogenesis require PPAR␥, we inhibited PPAR␥ either by antagonist treatment or by generating a 3T3L1 cell line stably expressing an anti-PPAR␥ siRNA through lentiviral infection. Treatment with an excess of the GW9662 antagonist partially but significantly inhibited the adipogenic actions of MEHP and rosiglitazone (Fig. 6D). Consistently, the effects of both compounds on adipocyte differentiation were significantly reduced in PPAR␥ siRNA cells, whereas cells infected with a virus containing the empty pLVTH vector underwent similar differentiation as WT cells (Fig. 6E).
MEHP Induces Selective Transcriptional Regulations during Adipocyte Differentiation-To characterize the differences in the adipogenic actions of MEHP and rosiglitazone and the underlying molecular pathways, we performed gene expression array analyses on glass slides spotted at custom with a collection of 17,000 mouse cDNAs. Expression levels from undifferentiated cells treated with insulin only and from cells differentiated with rosiglitazone or MEHP in the presence of insulin were compared using a linear model allowing direct comparisons between the three conditions. The vast majority of the genes regulated by MEHP were also regulated by rosiglitazone (Fig.  7A), again indicating that the adipogenic actions of MEHP are mediated by PPAR␥. Consistent with their ability to promote adipocyte differentiation, both compounds had strong effects on genes implicated in metabolism and cell cycle according to gene ontology analyses (Fig. 7B). The metabolic genes regulated by rosiglitazone and MEHP affected both catabolic and anabolic pathways (Table 1). Several genes implicated in glucose uptake, glycolysis, ␤-oxidation, citrate cycle, and oxidative phosphorylation were upregulated, likely providing the cells with high energetic levels to fulfill anabolic functions. The treatments with MEHP and rosiglitazone also induced genes required for the structure of lipid droplets (aP2 and Adrp) and the function of mature adipocytes (lipogenesis, triglyceride synthesis, and adipokines). Interestingly, we also observed an up-regulation of enzymes from the pentose phosphate pathway and from some intermediate steps of neoglucogenesis, which can potentially increase NADPH levels subsequently utilized for lipogenesis.
MEHP specifically regulated only a very low number of genes that could not be assigned to characterized adipocyte differentiation pathways. In contrast, rosiglitazone had a broader action than MEHP because around 30% of the genes were specifically regulated by this full agonist (Fig. 7A). Thus, MEHP potentially acts as a selective PPAR modulator regulating only a subset of PPAR␥ target genes compared with a full agonist. This target gene specificity could potentially translate into selective functional actions. However, genes specifically regulated by rosiglitazone were distributed within various pathways with no particular enrichment in a specific function ( Fig. 7B and Table 1). This observation precludes concluding on the respective selective biological functions of rosiglitazone and MEHP. The physiological selectivity, if any, may therefore rather arise from a differential regulation of rate-limiting enzymes or of crucial proteins rather than an entire pathway.
To validate the concept of a selective modulation of PPAR␥ targets by MEHP, we analyzed the expression of well described PPAR␥ target genes by quantitative reverse transcriptase-PCR ( Fig. 8 and Table 2). When differentiation was induced with rosiglitazone, most genes were up-regulated, from around 2-fold for PPAR␥ itself up to 400-fold for OLR1 (Fig. 8B), whereas the expression of some described PPAR␥ targets FIGURE 6. MEHP induces adipogenesis through PPAR␥. Two-day post-confluent 3T3L1 cells were treated for 10 days without insulin or with 10 g/ml insulin with vehicle (V; Me 2 SO, 1%), or the ligands at the indicated concentrations. After Oil Red O staining, the entire wells (A) or representative zones (B) were imaged. Triglyceride content was determined in cell lysates by colorimetric assay (500 nm) and normalized to the total quantity of protein measured by a Bradford assay (595 nm) (C). The same treatments were applied to wild type (WT) cells in the presence of the PPAR␥ antagonist GW9662 (D) and to 3T3L1 cells stably expressing an empty vector or a vector encoding an siRNA against PPAR␥ (E).
remained unaffected (Table 2). When MEHP was used as a PPAR␥ ligand, 8 of 10 genes were similarly induced (Ppar␥, Fabp-4/aP2, adiponectin (Adipoq), Cd36, acyl-CoA synthetase 1 (Acs-1), lipoprotein lipase (Lpl), C/ebp␣, and liver X receptor ␣ (Lxr␣)), whereas Gyk and Olr1 were significantly less induced, indicating that the efficacy of gene induction by MEHP depends on the promoter context. In addition, we confirmed new potential PPAR␥ targets from the microarray results by showing that expression of the lipogenic enzyme acetyl-CoA carboxylase ␤ (Acc␤) is induced both by MEHP and rosiglitazone, whereas that of acyl-CoA synthetase Bubblegum 1 (Acsbg1) is induced to higher levels with rosiglitazone than with MEHP.
In the experiment described above, PPAR ligands were added all along the differentiation process. Because the activity of PPAR␥ is intrinsically linked to adipocyte differentiation, it is therefore difficult to uncouple the direct induction of PPAR␥ target genes from their indirect activation via the global network of transcription factors controlling the adipogenic program. We hence performed the same gene expression analyses with 3T3L1 cells treated for 48 h with rosiglitazone or MEHP, either before differentiation or after differentiation with a classical adipogenic mixture, as described under "Experimental Procedures" (Fig. 8, A and C). Interestingly, the relative efficacy of MEHP compared with rosiglitazone varied between promoters and, for a given promoter, was different between the three differentiation status (non-differentiated, differentiated with a PPAR ligand, or differentiated with a classical mixture). When treatments were performed in differentiated adipocytes, target gene induction was globally reduced or alleviated (Fig. 8C), presumably because the expression of these genes was already high in differentiated cells. Nevertheless, genes such as Gyk, Olr1, Acsbg1, and Fabp-4 that could still be further induced, maintained an activation profile whose selectivity was similar to that obtained with treatments during differentiation. In contrast, Fabp-4 and Adipoq, which were strongly activated to equal levels both by MEHP and rosiglitazone in cells treated during differentiation, only responded to rosiglitazone in non-differentiated cells (Fig. 8A).
Altogether, these results demonstrate that in the same cellular context, MEHP exerts a selective action on different PPAR␥ target genes that varies according to the differentiation status of the cell. Two different classes of genes must be distinguished. The first group includes genes such as Gyk, Olr1, and Acsbg1 on which MEHP exerts a lower activity than rosiglitazone, independently of the differentiation status of the cell. A second group is exemplified by Fabp-4 and Adipoq, which equally respond to MEHP and rosiglitazone during differentiation but are principally induced by rosiglitazone in undifferentiated cells.

MEHP Induces Adipogenesis by Activating PPAR␥-
The activation of PPAR␥ by the phthalate monoester MEHP, a metabolite of the industrial pollutant DEHP, has been previously reported (20,21). However, these experiments, conducted in a toxicological perspective, did not address the molecular mechanisms of action as well as the physiological consequences of such an activation. We demonstrated herein that MEHP-dependent PPAR␥ activation promotes adipocyte differentiation, whereas DEHP has no effect. Several lines of evidence suggest that the adipogenic properties of MEHP are mediated by PPAR␥. First, MEHP concentrations required for minimal and maximal induction of adipogenesis parallel those required for PPAR␥ activation in transactivation assays. Second, the adipogenic actions of MEHP are reduced by knocking-down PPAR␥ or by inhibiting its activity with an antagonist, although the intricate link between PPAR␥ and adipogenesis does not totally rule out the possibility of a concomitant action of MEHP downstream of PPAR␥. Third, when analyzed on a genome-wide basis, the majority of the genes regulated by MEHP during adipogenesis are also regulated by the full PPAR␥ agonist rosiglitazone. Finally, MEHP can induce the expression of some PPAR␥ target genes important for adipocyte differentiation. Nevertheless, this compound differs from thiazolidinediones by reduced affinity and efficacy and by eliciting a partial adipogenic response compared with rosiglitazone.
MEHP Selectively Regulates PPAR␥ Activity during Adipogenesis-Modelization of MEHP within the PPAR␥ LBD revealed a configuration very similar to rosiglitazone. However, the conformational changes induced by both ligands or their ability to stabilize helix 12 in an active configuration are most likely different because MEHP promotes interactions with only a subset of PPAR␥ coregulators. Indeed, using three independent tools, i.e. in vitro, in living cells and on PPAR␥ target promoters, we demonstrated that MEHP only partially induces corepressor release and promotes the recruitment of the coactivators Med1 and PGC-1␣ but not of p300 and SRC-1. In addi- FIGURE 7. MEHP regulates only a subset of genes compared with rosiglitazone. RNA from 2-day post-confluent 3T3L1 cells treated for 10 days with 10 g/ml insulin in combination with vehicle (V; Me 2 SO, 1%), 1 M rosiglitazone, or 100 M MEHP was subjected to a custom microarray as described under "Experimental Procedures." Genes were considered to be significantly regulated when the p value was lower than 0.05 and the -fold change was above 2. The regulation by rosiglitazone and MEHP compared with the untreated control are summarized as Venn diagrams (A) and the significant gene ontology classes are represented for each condition (B).

TABLE 1 The metabolic actions of MEHP and rosiglitazone during 3T3L1 differentiation
The table represents the -fold induction of significantly regulated metabolic genes from the microarray analysis described in Fig. 3. Genes that are regulated by rosiglitazone only are in red and genes regulated by both ligands with a statistically higher effect with rosiglitazone are in green.
tion, other coregulators may be selectively recruited by MEHP and rosiglitazone. This is indeed reflected by the partial reduction of PPAR␥ mobility upon MEHP treatment in living cells where a full set of coactivators is present and where the reduction of PPAR mobility upon ligand binding is governed by the engagement in large complexes of coregulators (8,9).

TABLE 2 MEHP selectively induces PPAR␥ target genes in a differentiation-dependent manner
The results from Fig. 4 are presented as -fold inductions over vehicle treatment and extended to additional genes. 1 Cells were differentiated with the indicated ligands in the presence of insulin. 2 Cells were induced with a differentiation cocktail for 2 days (isobutylmethylxanthine/dexamethasone/insulin), differentiated for 8 days with insulin and then treated with the indicated ligands.
This selective recruitment of coregulators translates into a partial adipogenic program where only a subset of genes is regulated by MEHP compared with rosiglitazone. The partial release of NCoR in response to MEHP seems sufficient to alleviate the inhibitory effects of this corepressor on most PPAR␥ targets to promote adipogenesis (41). Indeed, genes important for adipocyte differentiation (Fabp-4, adiponectin, Acs, and Lpl among others) are fully induced. In contrast, other direct PPAR␥ targets such as Gyk, Olr1, and Acsbg1 are only partially activated by MEHP. Interestingly, Gyk and Olr1 are specifically silenced by corepressors during adipogenesis and require an exogenous PPAR␥ ligand for activation (10,40) (Fig. 5). Thus, the low levels of PPAR␥/NCoR interaction remaining upon MEHP binding could limit the maximal activation level of such genes. In contrast, genes such as Fabp-4 whose promoter is never associated with corepressors are insensitive to the limited action of MEHP on corepressor release, thereby potentially explaining why such genes are equally activated by MEHP and rosiglitazone. Adipoq is also activated by MEHP and rosiglitazone with equal efficacy. Amazingly, the persistence of NCoR on the Adipoq promoter in the presence of MEHP, as well as in the presence of rosiglitazone albeit to a lower level, has to be compensated by stronger levels of coactivator recruitment such as those observed on this promoter for PGC-1␣. The selective recruitment of coactivators may also play a role in the differential induction of target genes. Although we have not observed major promoter-specific differences in the association of the coactivators tested here, it is possible that subtle differences between promoters may explain differences in gene expression. In addition, the reduced adipogenic potential of MEHP compared with rosiglitazone could also be linked to a reduced efficacy in the recruitment of some coactivators. Indeed, both in live cell FRET assay and in chromatin immunoprecipitation experiments, MEHP induced only a partial recruitment of Med1, a coactivator required for efficient adipocyte differentiation (42).
Distinguishing Selective PPAR␥ Modulation from Partial PPAR␥ Activation-Using MEHP as a model, this study demonstrates the molecular basis of selective PPAR modulation and clearly establishes the difference between selective PPAR modulators and partial PPAR␥ agonists. SPPARMs are usually characterized by their molecular properties, i.e. the ability to induce selective coregulator recruitment (14). This is then related to a physiological output where a restricted PPAR␥ action is established. Between these two steps, the nature of the target gene regulation is, however, often underevaluated. Indeed, when compared with full agonists, SPPARMs differ from partial agonists as they promote selective gene regulation by differentially affecting target gene transcription in a gene-specific manner, with some genes induced to similar levels than those obtained with a full agonist, whereas others exhibit restricted activation. In contrast, partial agonists exhibit a global decrease in the activation of all target genes. Thus, a bona fide SPPARM should induce conformational changes different from full agonists that translate into selectivity in terms of coregulator interactions, target gene induction, and ultimately physiological effects. At the molecular level, MEHP fulfills these conditions as the dif-ferential recruitment of coregulators translates into a restricted profile of gene regulation.
Our results also point to the importance of the cellular context regarding coregulator equipment, illustrated by the differences of PPAR␥ transactivation by MEHP according to the cell line analyzed. In addition, gene expression analyses on 3T3L1 cells treated with PPAR␥ agonists before, during, or after differentiation, suggest that both the level and the selectivity of target gene induction rely on the expression of distinct sets of coregulators. Thus, PPAR␥ agonists may exert different effects on subpopulations of adipocytes and their progenitors in the adipose tissue, an aspect that should be taken into consideration in pharmacological strategies aimed at finding PPAR␥ modulators acting on adipocyte physiology.
The identification of SPPARMs that favor the beneficial pharmacological actions of thiazolidinediones over their adverse effects is a major challenge for pharmaceutical research on type 2 diabetes (13, 43), for which some candidate compounds are starting to emerge (44 -48). FK614 is a SPPARM causing impaired recruitment of CREB-binding protein but retaining beneficial effects on insulin sensitivity in hyperglycemic mouse models (46,49,50). The inability of MEHP to promote interactions between PPAR␥ and p300, a coactivator highly homologous to CREB-binding protein, and enhanced recruitment of PGC-1␣ would both potentially be compatible with a positive action of MEHP on insulin sensitivity (45,50). However, a metabolic study of phthalate exposure in animal models is required to understand how the molecular and cellular effects described herein potentially translate into the selective regulation of the physiological functions regulated by PPAR␥. Whereas mice and rats are the models most often used for this type of metabolic studies, the occurrence of sustained hepatic peroxisome proliferation mediated by PPAR␣, which does not occur in humans, is likely to affect the final metabolic phenotype of these animals, and alternate models might have to be considered.
MEHP Is a Potential Metabolic Disruptor-Our results on the adipogenic action of MEHP strongly argue that in addition to an action on hepatic carcinogenesis through PPAR␣-mediated peroxisome proliferation (51), the endocrine disrupting actions of DEHP through its MEHP metabolite should also be considered with respect to the development of obesity and associated metabolic disorders. We have observed that human PPAR␥ can be activated with similar affinities and efficacies than the mouse isotype, but we failed to detect the subtle differences in affinity described by other reports that may therefore be linked to the cellular context (20,21). In addition, MEHP induces the same pattern of coregulator recruitment with the mouse and human receptors, suggesting that the effects of MEHP on adipogenesis studied here in the context of the mouse receptor may translate into similar effects in humans if MEHP reaches the adipose tissue in sufficient concentration. Although the human exposure to DEHP is ubiquitous through daily products, plasmatic levels of both DEHP and MEHP generally remain low because of rapid urinary excretion (17). However, individuals requiring frequent blood transfusion or dialysis are subjected to repetitive acute exposures to high levels of DEHP because of the leaching of the compound from plastic bags and tubings in direct contact with biological fluids. Under such circumstances, the plasmatic levels of DEHP and MEHP can reach 50 M in humans (17), a concentration at which we already observed strong although not maximal induction of adipogenesis. Although the concentration of MEHP in adipocytes is difficult to assess, DEHP is lipophilic and can accumulate in adipose tissue (52,53). Despite the absence of data available to our knowledge on MEHP accumulation in the adipose tissue or on the adipose tissue expression of lipases that metabolize DEHP into MEHP, our data combined with these observations urge the need to consider the actions of DEHP and MEHP in the pathophysiology of this tissue. Together with reports showing that other pollutants including different phthalate esters and organotins can also target PPAR␥ and promote adipogenesis in cellular models (20,21,54,55) and in vivo (56), our study suggests that the metabolic functions of PPAR␥ can be targeted by a subclass of endocrine disruptors that we propose to define as metabolic disruptors.