Peroxisome Proliferator-Activated Receptor γ Is a Target for Halogenated Analogs of Bisphenol A

Background: The occurrence of halogenated analogs of the xenoestrogen bisphenol A (BPA) has been recently demonstrated both in environmental and human samples. These analogs include brominated [e.g., tetrabromobisphenol A (TBBPA)] and chlorinated [e.g., tetrachlorobisphenol A (TCBPA)] bisphenols, which are both flame retardants. Because of their structural homology with BPA, such chemicals are candidate endocrine disruptors. However, their possible target(s) within the nuclear hormone receptor superfamily has remained unknown. Objectives: We investigated whether BPA and its halogenated analogs could be ligands of estrogen receptors (ERs) and peroxisome proliferator–activated receptors (PPARs) and act as endocrine-disrupting chemicals. Methods: We studied the activity of compounds using reporter cell lines expressing ERs and PPARs. We measured the binding affinities to PPARγ by competitive binding assays with [3H]-rosiglitazone and investigated the impact of TBBPA and TCBPA on adipocyte differentiation using NIH3T3-L1 cells. Finally, we determined the binding mode of halogenated BPAs to PPARγ by X-ray crystallography. Results: We observed that TBBPA and TCBPA are human, zebrafish, and Xenopus PPARγ ligands and determined the mechanism by which these chemicals bind to and activate PPARγ. We also found evidence that activation of ERα, ERβ, and PPARγ depends on the degree of halogenation in BPA analogs. We observed that the bulkier brominated BPA analogs, the greater their capability to activate PPARγ and the weaker their estrogenic potential. Conclusions: Our results strongly suggest that polyhalogenated bisphenols could function as obesogens by acting as agonists to disrupt physiological functions regulated by human or animal PPARγ.

Bisphenols form a large family of chemi cals that are used mainly to produce poly carbonates and epoxy resins. By far, the most widely used bisphenol (> 3 million tons/year) is bisphenol A (BPA), which is used in the manufacture of items such as plastics, food can linings, dentistry seal ants, and thermal paper. BPA ( Figure 1) is a model xeno estrogen. Despite possessing only modest estrogenic activity compared with 17βestradiol (E2), over the last decade BPA has been shown to produce a range of adverse effects in laboratory animals, with major concerns regarding reproductive targets (Richter et al. 2007). More recently, it has been hypothesized that early exposure to BPA could play a role in the onset of obesity and other metabolic syndromes . In this regard, a large body of data about endocrinedisrupting chemicals (EDCs) under lines the importance of exposure dur ing early stages of development, which could result in reproductive defects in adult life (Newbold et al. 2009). Human exposure to BPA has been clearly demon strated (Calafat et al. 2008). However lowdose effects of BPA and the possible consequences of such exposure are controversial (Vandenberg et al. 2009;vom Saal and Hughes 2005).
Halogenated derivatives of BPA, which feature bromine or chlorine substituents on the phenolic rings, are used as flame retar dants. However, compared with BPA, little information is available regarding the poten tial endocrine disruption by these compounds. All brominated BPA analogs originate from tetra bromo bisphenol A (TBBPA), which is the mostproduced brominated flame retar dant (> 150,000 tons produced annually) (de Wit et al. 2010). TBBPA (Figure 1) is used to produce fireproof epoxy resins used in the manufacture of computer motherboards and other electronics; and it has been found in the environment (de Wit et al. 2010), in wild life (Darnerud 2003), and in human samples (Cariou et al. 2008;Shi et al. 2009). TBBPA is debrominated in the environment into lowerbrominated BPA analogs (monoBBPA, diBBPA, and triBBPA) (Arbeli et al. 2006). The closely related tetra chloro bisphenol A (TCBPA) (Figure 1) is also used as a flame retardant, but in much lower quantities than TBBPA (< 10 000 tons/year) (Chu et al. 2005), and its presence in environmental samples has been unequivocally demonstrated (Fukazawa et al. 2001). Given the low production level of TCBPA, its presence in the environment most likely originates from the spontaneous chlorina tion of BPA. Indeed, like many phenolic com pounds, BPA is readily chlorinated in aqueous media (Deborde et al. 2004).
The estrogenic activity of BPA exerted through binding to estrogen receptors (ERs) is likely involved in the onset of many of its adverse effects, and several studies in ani mal models have shown that such effects are observed after exposure to low doses (Vandenberg et al. 2009). Brominated BPA analogs are not as estrogenic as BPA, and the potency of brominatedBPAs as ER agonists decreases as the number of bromine atoms increases (Meerts et al. 2001). Conversely, the Background: The occurrence of halogenated analogs of the xeno estrogen bisphenol A (BPA) has been recently demonstrated both in environmental and human samples. These analogs include brominated [e.g., tetrabromobisphenol A (TBBPA)] and chlorinated [e.g., tetrachloro bisphenol A (TCBPA)] bisphenols, which are both flame retardants. Because of their structural homology with BPA, such chemicals are candidate endocrine disruptors. However, their possible target(s) within the nuclear hormone receptor superfamily has remained unknown. oBjectives: We investigated whether BPA and its halogenated analogs could be ligands of estrogen receptors (ERs) and peroxisome proliferator-activated receptors (PPARs) and act as endocrinedisrupting chemicals. Methods: We studied the activity of compounds using reporter cell lines expressing ERs and PPARs. We measured the binding affinities to PPARγ by competitive binding assays with [ 3 H]-rosiglitazone and investigated the impact of TBBPA and TCBPA on adipocyte differentiation using NIH3T3-L1 cells. Finally, we determined the binding mode of halogenated BPAs to PPARγ by X-ray crystallography. results: We observed that TBBPA and TCBPA are human, zebrafish, and Xenopus PPARγ ligands and determined the mechanism by which these chemicals bind to and activate PPARγ. We also found evidence that activation of ERα, ERβ, and PPARγ depends on the degree of halogenation in BPA analogs. We observed that the bulkier brominated BPA analogs, the greater their capability to activate PPARγ and the weaker their estrogenic potential. conclusions: Our results strongly suggest that polyhalogenated bisphenols could function as obesogens by acting as agonists to disrupt physiological functions regulated by human or animal PPARγ. volume 119 | number 9 | September 2011 • Environmental Health Perspectives estrogenic activity of chlorinated congeners could be similar to or higher than that of BPA (Mutou et al. 2006;Takemura et al. 2005). Similarly, both TBBPA and TCBPA interact with and disrupt thyroid hormone receptor signaling (Kitamura et al. 2002). Recently, Somm et al. (2009) showed that peri natal exposure to BPA altered early adipogenesis in the rat, which is mediated by peroxisome proliferator-activated receptor γ (PPARγ), a nuclear hormone receptor whose dysregulation is involved in the onset of diabetes and obesity (Swedenborg et al. 2009). This suggests that BPA and its derivatives may also interact with this receptor. In the present study, we exam ined the capacity of BPA and halogenated BPA derivatives to inter act with and perturb signaling by ERα, ERβ, PPARα, PPARδ, and PPARγ. We provide the first experimental evidence that flame retardants TBBPA and TCBPA are ligands and partial agonists of human PPARγ and also activate the corre sponding zebrafish and Xenopus receptors. Our findings indicate that these compounds should certainly be evaluated as EDCs with possible deleterious effects on humans and wildlife.
Reporter cell lines and stable gene expression assay. Generation of HGELN, HGELNERα, HGELNERβ, HGELNGALPPARα, HGELNGALPPARβ, and HGELNGAL PPARγ reporter cell lines was performed as pre viously described (Escande et al. 2006;le Maire et al. 2009). Briefly, reporter cells were seeded at a density of 20,000 cells/well in 96well white opaque tissue culture plates and main tained in phenolred-free Dulbecco's modified Eagle's medium (DMEM) supplemented with 5% dextrancoated, charcoaltreated fetal calf serum. Twentyfour hours later, culture medium was replaced with DMEM contain ing tested compounds. We performed assays in the absence of serum to avoid ligand capture by serum proteins. Sixteen hours after expo sure, we replaced media with media containing 0.3 mM luciferin. Luminescence was measured in intact living cells for 2 sec in a Microbeta Wallac luminometer (PerkinElmer).
NIH3T3-L1 differentiation. Twoday post confluent 3T3L1 pre adipocytes (a gift from L. Fajas; Institut de Génétique Moléculaire, Montpellier, France) were induced to dif ferentiate by 2day treatment with a differ entiation mixture (10 μg/mL insulin, 1 μM dexa methasone, 0.5 mM isobutyl methyl xanthine) followed by 8day treatment with 10 μg/mL insulin and PPARγ ligands. The medium was replaced every 48 hr. After differ entiation, cells were stained with Oil Red O for morphological analyses, or RNA was extracted from the cells using the RNeasy RNA isolation kit (Qiagen, Courtaboeuf, France). For RNA extractions, four independent cultures were performed per condition. Reverse transcrip tion was performed with random hexamers on 1 μg total RNA using SuperScript II reverse transcriptase (Invitrogen), and the reaction was diluted 100 times for amplification. Realtime polymerase chain reaction (PCR) quantifica tion was then performed using SYBR Green technology (LightCycler; Roche Diagnostics, Meylan, France). Results were normalized to two housekeeping genes (18S and 36B4) and quantified using qBase (Roche Diagnostics).
PPARγ expression and purification. DNA encoding the LBD of human PPARγ (amino acids Glu196Tyr477) was amplified by PCR and cloned into the expression vec tor pET15b. The plasmid PPARγ (Glu196 Tyr477)pET15b was transformed into Escherichia coli BL21(DE3) cells (Invitrogen). The PPARγ LBD was expressed and purified as previously described for retinoid X receptor (RXR)α LBD (Nahoum et al. 2007). Prior to crystallization trials, the purified PPARγ LBD was concentrated to 8.5 mg/mL in a buffer containing 20 mM TrisHCl, pH 8.5, 250 mM NaCl, 5 mM dithiothreitol, and 1 mM EDTA.
Crystallization. Crystals were obtained by vapor diffusion in hanging drops at 293 K. For crystals of unliganded (apo) PPARγ, 1 μL pro tein solution was mixed with 1 μL well solu tion containing 1 M tri sodium citrate, pH 7.5, 100 mM Hepes, pH 7.5, and 3% 1,2propane diol. Crystals appeared after 1 day and grew to about 200 μm within a few days. TCBPA was soaked into a PPARγ apocrystal by adding 0.5 μL TCBPA at a concentration of 1 mM suspended in well solution directly to the crys tal drop. The crystals were soaked for 4 days. For cocrystals of PPARγ in complex with TBBPA, 1 μL protein solution was mixed with 1 μL well solution containing 1 M tri sodium citrate, 100 mM HEPES, pH 7.5, 3.5% 1,2propanediol, and 0.2 mM TBBPA ligand, for a molar ratio of 1:2 of protein:ligand in the drop. Crystals appeared after 1 day and grew to about 200 μm within a few days. Crystals were transferred to a cryo protectant (well solution containing 20% glycerol and the correspond ing ligand at a concentration of 1 mM) and frozen in liquid nitrogen.
Crystallographic data collection, processing, and structure refinement. We collected diffrac tion data at the ID141 beamline at 2.55 Å and 2.70 Å resolution for TBBPAPPARγ and TCBPAPPARγ complexes, respectively, using an ADSC Quantum Q210 CCD detector at the European Synchrotron Radiation facility (ESRF, Grenoble, France). Diffraction data were processed using MOSFLM (Leslie 2006) and scaled with SCALA from the CCP4 pro gram suite (Collaborative Computational Project 1994). Structures were solved by using the previously reported structure 2ZVT (Waku et al. 2009) from which the ligand was omitted. Initial F o F c difference maps showed significant signals for the ligand, which could be fitted accurately into the electron density. The structures were modeled with COOT (Emsley and Cowtan 2004) and refined with phenix.refine from the PHENIX program suite (Afonine et al. 2005).

Results
Halogenated BPA derivatives activate human ERα, ERβ, and PPARγ. We monitored the agonistic potential of BPA and halo genated derivatives using stably transfected HGELN ERα, ERβ, PPARα, PPARδ, and PPARγ cell lines, allowing for a comparison of the effect of compounds on human ER and PPAR subtypes in a similar cellular context. All com pounds were first tested on the HGELN paren tal cell line containing only the reporter gene. We observed some toxicity at ligand concen trations of ≥ 10 μM but no unspecific modula tion of luciferase expression (data not shown). We then characterized the activity of BPA, TCBPA, and TBBPA on HGELNER cell lines. As shown in Figure 2A and B, despite a reduced affinity relative to E2, BPA exerted an almost full agonistic activity toward both ERα and ERβ. In contrast, TBBPA had little effect on either ER, whereas TCBPA partially activated both receptor subtypes with a slight preference for ERα. Similar experiments car ried out using HGELNPPAR cells demon strated that none of the compounds tested notably affected PPARα or PPARδ activity (data not shown). In contrast, both TBBPA and TCBPA were capable of partially activating PPARγ, despite being approximately 100fold less potent than the reference pharmaceutical compound rosiglitazone ( Figure 2C). The par ent compound BPA failed to activate PPARγ. The occurrence of lowerbrominated BPA analogs in the environment prompted us to meas ure their activity in HGELNER cell lines and HGELNPPARγ cells ( Figure 2D-F). Figure 2D shows that all brominated BPA con geners were partial ERα agonists with graded activities. BPA, monoBBPA, and diBBPA dis played the highest transactivation efficiency fol lowed by triBBPA, whereas TBBPA had almost no activity in the HGELNERα cells. These compounds were also tested in the HGELN ERβ cell line, providing a similar partial activ ity and ranking order of estrogenic potency ( Figure 2E). MonoBBPA and diBBPA, the brominated analogs charac terized by the highest estrogenic potency but the lowest molecular weight, exhibited a slight ERα selec tivity. Interestingly, when assayed in HGELN PPARγ cells, the halogenated compounds ranked in the reverse order, with triBBPA and TBBPA showing the highest potency to induce luciferase gene expression, followed by diBBPA and monoBBPA ( Figure 2F). Finally, we compared TBBPA and TCBPA with the wellknown environ mental PPARγ ligands MEHP (Feige et al. 2007), PFOS, and PFOA (Takacs and Abbott 2007) [see Supplemental Material, Figure 1 (http://dx.doi.org/10.1289/ ehp.1003328)]. As shown in Figure 2G, the halogenated BPAs triggered PPARγ activation at 10 to 100fold lower concentrations than the other candidate PPARγ disruptors. Binding activity of TBBPA and TCBPA to human PPARγ receptor. To further charac terize the inter action between human PPARγ and halogenated BPA derivatives, we per formed wholecell competitive binding assays using HGELNPPARγ cells. TBBPA and TCBPA competitively inhibited the binding of 10 -13 10 -12 10 -11 10 -10 10 -9 10 -8 10 -7 10 -6 10 -5 10 -12 10 -11 10 -10 10 -10 10 -9 10 -9 10 -8 10 -8 10 -7 10 -7 10 -6 10 -6 10 -5 10 -8 10 -7 10 -6 10 -5 10 -8 10 -7 10 -6 10 -5 10 -7 10 -6 10 -5 10 -4 10 -5 10 -8 10 -7 10 -6 10 -5

Log concentration (M)
Log concentration (M)  values for rosiglitazone, TBBPA, and TCBPA were 12.0 nM, 0.7 μM, and 6.0 μM, respec tively. Together with trans activa tion assays ( Figure 2C,F,G), these data demon strate that TBBPA and TCBPA bind to human PPARγ and activate the receptor at concentrations in the micro molar range. Halogenated BPAs promote adipocyte differentiation through PPARγ. Having shown that halogenated BPAs are PPARγ ligands, we investigated the action of TBBPA and TCBPA on endogenous genes by studying their ability to induce adipo genesis, a wellcharacterized PPARγregulated function. As we expected, treatment of 3T3L1 pre adipocytes with the full PPARγ agonist rosiglitazone strongly induced adipo genesis, as evidenced by Oil Red O stain ing, whereas the PPARγ antagonist CD5477 (le Maire et al. 2009) did not induce adipo cyte differentia tion ( Figure 4A). TCBPA and TBBPA at 10 μM also induced adipo genesis, whereas cotreatment with CD5477 inhibited the adipo genic action of TBBPA, indicating that halogenated BPAs mediate adipo genesis via PPARγ. Adipocyte differentiation by TBBPA and TCBPA was further confirmed by examining the endogenous expression of two PPARγ target genes, namely ApoA2/FABP4 (AP2) and PPARγ itself ( Figure 4B). Whereas PPARγ was expressed at similar levels upon treatment with halogenated BPAs or rosigli tazone, AP2 was expressed to a much lesser extent after treatment with either TBBPA or TCBPA compared with rosiglitazone. This dif ferential expression level of the two genes could reflect partial agonism of halogenated BPAs.

Log concentration (M) Log concentration (M)
TBBPA and TCBPA are activators of zebrafish and Xenopus PPARγ. Because the amino acid sequence of PPARγ differs between mammals and other species, we car ried out transient trans activation assays to examine the ability of TBBPA and TCBPA to act as agonists of zebrafish and Xenopus PPARγ ( Figure 5). These two animal species are often used as in vivo models to evaluate the impact of environ mental compounds on organisms (Fini et al. 2007;Legler et al. 2002). This experi ment confirmed that TBBPA, TCBPA, and MEHP are all activators of human PPARγ. We also found that TBBPA and TCBPA activated zebrafish PPARγ, whereas MEHP appeared to be a slightly weaker ligand, and rosiglitazone was completely inactive. In con trast, all compounds, including rosiglitazone, activated Xenopus PPARγ. Together, our data indicate that halogenated BPA can disrupt the activity of PPARγ from different species.
Structural analysis of the TBBPA-and TCBPA-PPARγ complexes. Finally, we solved the crystal structures of TBBPA and TCBPA bound to the PPARγ LBD to reveal the mech anism by which these compounds, which are structurally unrelated to known PPARγ ligands, bind to and activate this receptor [see Supplemental Material,  Figure 6A), the structures reveal the canonical tertiary fold of agonistbound nuclear hormone receptor LBDs (Bourguet et al. 2000). The TBBPA and TCBPA complex structures are indistin guishable, with a root mean square deviation (RMSD) value of 0.29 Å for superimposed alpha carbons (see Supplemental Material, Figure 2) and nearly identical to that of PPARγ in complex with the agonist rosiglita zone (Nolte et al. 1998), with an RMSD value of 0.62 Å ( Figure 6B). Omission of F o F c dif ference maps for the TBBPA and TCBPA structures revealed clear density for the ligands that could be positioned unambiguously into the PPARγ ligandbinding pocket (LBP) (see Supplemental Material, Figure 3B). PPARγ displays a large LBP that extends from the Cterminal helix H12 to the βsheet S1/S2 so that halogenated BPA occupies a small portion of the LBP ( Figure 6A). Whereas rosiglita zone occupies a region of the LBP spanning H11/H12 to the βsheet S1/S2, TBBPA and TCBPA occupy only the region between H3 and the βsheet S1/S2, with cycle A nestled between H3 and the βsheet (Figure 6B,C). In contrast with rosiglitazone, a consequence of the smaller size of halogenated BPAs is that they do not inter act directly with H12 ( Figure 6B,C). A close look at the LBP shows that the phenol groups of the BPA derivatives are involved in hydrogen bonds. The hydroxyl  group from cycle A ( Figure 6D) inter acts with the main chain nitrogen atom of Ser342 (βsheet S1/S2), whereas the second one (cycle B) is hydrogenbonded to Ser289 in H3 and Tyr473 from H12 through a watermediated hydrogen bond network ( Figure 6E). TBBPA and TCBPA contain four halogen atoms that contribute to ligand binding through van der Waals inter actions ( Figure 6D,E). Additional inter actions involving the ligand backbone were also observed ( Figure 6F). Interestingly, comparison of human, mouse, and zebrafish PPARγ sequences reveals several residue differences, which could explain the differential ligand specificity of the various spe cies (see Supplemental Material, Figure 3A). In particular, the replacement of human PPARγ Gly284 and Cys285 by serine and tyrosine res idues in zebrafish PPARγ provides a rationale for the weak binding affinity of rosiglitazone for this receptor compared with that observed for the human homolog (see Supplemental Material, Figure 3B). In contrast, the differ ent binding mode of halogenated compounds allows both human PPARγ and zebrafish PPARγ to accommodate TBBPA and TCBPA (see Supplemental Material, Figure 3B).

Discussion
Mounting data indicating the presence of halogenated bisphenols in environ mental and human samples clearly suggest that these com pounds should be considered an emerging class of contaminants whose cellular targets and effects require better under standing (Cariou et al. 2008;de Wit et al. 2010;Fernandez et al. 2007;Guerra et al. 2010). Brominated BPA analogs result mainly from the extensive use of TBBPA as a flame retardant, whereas a grow ing body of evidence suggests that chlorinated BPA analogs arise from the abiotic chlorination of BPA residues (Fukazawa et al. 2001). In the present study, we investigated the possible role of BPA and halogenated BPA derivatives as environmental ligands for ERs and PPARs. We used HeLa cells, which are charac terized by low intrinsic metabolic capabili ties, to examine ER and PPAR activities to minimize metabolic bio transformations of tested compounds, which potentially can lead to a mis interpretation of in vitro results. Toxicity of halogenated BPA derivatives toward HeLa cell lines was observed at con centrations > 10 μM, which is fully consis tent with previous reports showing that both TBBPA and TCBPA are more toxic than BPA (Nakagawa et al. 2007). For brominated ana logs, the ranking order of estrogenic potency in HELNERα and HELNERβ cell lines was monoBBPA > diBBPA > triBBPA, whereas TBBPA showed no estrogenic activity at all. In a similar study using T47D breast cancer cells, Meerts et al. (2001) reported simi lar findings. Interestingly, in HGELNPPARγ cells, the ability of BPA analogs to activate PPARγ was reversed, with TBBPA = triBBPA > diBBPA >>> monoBBPA. The activity of TCBPA was similar to that of TBBPA.
Full agonists of PPARγ (e.g., rosiglitazone) have been reported to fully activate their cog nate receptor by directly inter acting with and stabilizing helix H12, whereas compounds that do not directly contact H12 behave as weak or partial agonists by stabilizing other regions of the LBD, including H3 and the βsheet S1/S2 (Bruning et al. 2007;Kallenberger et al. 2003). Both the functional and structural studies reported here support this model of par tial PPARγ agonism by TBBPA and TCBPA. The rather weak affinity of halogenated BPAs for PPARγ compared with rosiglitazone can be explained by their smaller size and correspond ingly fewer direct atomic contacts with the pro tein. Notably, whereas rosig lita zone is engaged in five hydrogen bonds with the protein, only two hydrogen bonds are observed between TBBPA/TCBPA and PPARγ [see Supplemental Material, Figure 4 (Nolte et al. 1998) can readily accommodate the addition of bulky bromine or chlorine atoms, the much smaller LBP of the ERs cannot, thus providing an explanation for the differential pattern of inter actions of halogenated BPAs with the two receptor types. It is note worthy that some halo genated BPAs, including TCBPA and diBBPA, can interact with both ERs and PPARγ. This dual activity could increase the toxicity of these compounds compared with BPA or TBBPA, which are ER and PPARγselective ligands, respectively. The comparison of the adverse effects induced by the two types of compounds through in vivo experiments should provide information on whether the dual ER/PPAR halogenatedBPA ligands display a higher EDC potency on reproductive and metabolic func tions than more selective congeners. Until now, few environmental compounds (includ ing MEHP, PFOS, PFOA, and organotins) have been found to interfere significantly with PPARγ signaling (Feige et al. 2007;Grün and Blumberg 2006;Takacs and Abbott 2007). In this regard, we recently reported that organotins potently activate RXR/ PPARγ hetero dimers essentially through bind ing to the RXR subunit (le Maire et al. 2009). Conversely, the functional and structural data presented here demon strate that halogenated BPAs are capable of activating PPARγ via direct inter actions charac terized by binding affinities that are 10 to 100times higher than other proposed PPARγ disruptors. The discovery of a novel chemical class of PPARγ activators strengthens the hypothesis that environmental ligands could be involved in the disruption of energy balance in humans and wildlife. As with other EDCs, peri natal exposure could play a critical role. According to Cariou et al. (2008), significant levels of TBBPA can be found in human cord blood (200 pg/g fresh weight) and maternal milk (0.1-37.4 ng/g lipid weight), demon strating both pre natal and post natal exposure in a large fraction of the population. Furthermore, as other RXR (the main active form of PPARγ is the RXR/PPARγ heterodimer) and PPARγ activators are also present in the environ ment, additive (acting only through PPARγ) and syn ergistic (acting through both RXR and PPARγ) effects could occur and further increase the risk of metabolic diseases. In this regard, the "cocktail effect" resulting from a concomitant exposure to organotins and halogenatedBPAs could be particularly deleterious.