In vitro estrogenicity of polybrominated diphenyl ethers, hydroxylated PDBEs, and polybrominated bisphenol A compounds.

Polybrominated diphenyl ethers (PBDEs) are used in large quantities as additive flame retardants in plastics and textile materials. PBDEs are persistent compounds and have been detected in wildlife and in human adipose tissue and plasma samples. In this study, we investigated the (anti)estrogenic potencies of several PBDE congeners, three hydroxylated PBDEs (HO-PBDEs), and differently brominated bisphenol A compounds in three different cell line assays based on estrogen receptor (ER)-dependent luciferase reporter gene expression. In human T47D breast cancer cells stably transfected with an estrogen-responsive luciferase reporter gene construct (pEREtata-Luc), 11 PBDEs showed estrogenic potencies, with concentrations leading to 50% induction (EC(50)) varying from 2.5 to 7.3 microM. The luciferase induction of the most potent HO-PBDE [2-bromo-4-(2,4,6-tribromophenoxy)phenol] exceeded that of estradiol (E(2)), though at concentrations 50,000 times higher. As expected, brominated bisphenol A compounds with the lowest degree of bromination showed highest estrogenic potencies (EC(50) values of 0.5 microM for 3-monobromobisphenol A). In an ER alpha-specific, stably transfected human embryonic kidney cell line (293-ER alpha-Luc), the HO-PBDE 4-(2,4,6-tribromophenoxy)phenol was a highly potent estrogen with an EC(50) < 0.1 microM and a maximum 35- to 40-fold induction, which was similar to E(2). In an analogous ER beta-specific 293-ER betas-Luc cell line, the agonistic potency of the 4-(2,4,6-tribromophenoxy)phenol was much lower (maximum 50% induction compared to E(2)), but EC(50) values were comparable. These results indicate that several pure PBDE congeners, but especially HO-PBDEs and brominated bisphenol A-analogs, are agonists of both ER alpha and ER beta receptors, thus stimulating ER-mediated luciferase induction in vitro. These data also suggest that in vivo metabolism of PBDEs may produce more potent pseudoestrogens.

Although PCB concentrations in wildlife are still higher than PBDE concentrations, they are declining over the same time period.
The most sensitive end points of PBDE toxicity in vivo are effects on thyroid function, observed as induction of thyroid hyperplasia and alteration of thyroid hormone production [i.e., lowering of free and total thyroxine (T 4 ) concentrations] in rats and mice (11,12). Consistent with these findings is the recent observation that several pure PBDE congeners were able to displace T 4 from transthyretin (TTR; a plasma transport protein of thyroid hormones) in vitro, after metabolic conversion to hitherto unidentified metabolites (13). These phenomena have also been observed for other organohalogen compounds such as PCBs and their hydroxylated metabolites (14,15, and references therein).
Some studies have indicated that hydroxylated PBDEs (HO-PBDEs) are of potential environmental importance. In liver microsomes of rats, several PBDE congeners were biotransformed to metabolites (13). Örn and Klasson-Wehler (23) demonstrated that 2,2´,4,4´-tetraBDE (BDE-47) is biotransformed to HO-PBDEs in rats and mice. 3,5-Dibromo-2-(2,4-dibromophenoxy)phenol is a hydroxy-BDE that has been identified in blood plasma of Baltic salmon (24) at levels similar to those of the major PBDE congeners. Information on the endocrine activity of hydroxylated PBDEs is presently limited to the ability of several HO-PBDEs to bind competitively to the thyroid hormone receptor (25) and to TTR (13).
Studies showing that many industrial chemicals are weakly estrogenic compared to natural estrogens (26)(27)(28) have raised concern about their safety. For example, o,p´-DDT, bisphenol A, nonylphenol, and various phthalates possess estrogenic activity (27). The presumption is that these xenoestrogens may disrupt normal endocrine function, which can lead to reproductive failure and cancer of estrogen-sensitive tissues in humans and wildlife (29). Antiestrogenic activity by anthropogenic compounds has received less attention (30). Although the inhibition of hormone action and the resulting toxicological consequences have not been demonstrated conclusively, antiestrogenic action could critically affect sensitive reproductive and developmental processes as well (30). To date there have been no reports investigating the (anti)estrogenic activities of PBDEs and HO-PBDEs.
The aim of this study was to determine the (anti)estrogenicity of 17 PBDE congeners. We also examined three hydroxylated PBDEs that have halogen substitution patterns similar to those of thyroid hormones. The (anti)estrogenic activity of these compounds was tested in vitro, using an estrogen-responsive luciferase reporter cell line (T47D.Luc) (31). We compared the structure-activity relationships for (anti)estrogenicity of PBDE and HO-PBDE congeners with numerous other brominated flame retardants, such as differently brominated bisphenol A compounds. We also tested the most potent PBDEs and HO-PBDEs observed in T47D.Luc cells for estrogen receptor specificity using 293 human embryonic kidney cells stably transfected with recombinant human estrogen receptor (ERα or ERβs) cDNA and the luciferase reporter gene construct (32)(33)(34).
Cell culture. We used the human T47D breast cancer cell line stably transfected with an estrogen-responsive luciferase reporter gene construct (pEREtata-Luc) (31) to study the in vitro (anti)estrogenic activity of PBDEs and HO-PBDEs. The T47D.Luc cells were cultured in a 1:1 mixture of Dulbecco's Modified Eagle's (DMEM) medium and Ham's F12 (DF) medium (Gibco Brl, Life Technologies, Breda, The Netherlands) supplemented with sodium bicarbonate, nonessential amino acids, sodium pyruvate, and 7.5% fetal calf serum (heat inactivated) at 37°C and 7.5% CO 2 .
The preparation of the stably transfected 293-Luc cell lines (ERα and ERβs) has been described in detail elsewhere (32). Briefly, human 293 embryonal kidney (HEK) cells (ATCC, American Type Culture Collection, Rockville, MD, USA) were stably transfected with the pEREtata-Luc construct (31,32) cotransfected with an antibiotic resistance gene. This cell line was subsequently transfected with a recombinant    ER-CALUX assay. We performed the T47D.Luc-based assay as described previously (31). The cells were trypsinized, resuspended in assay medium, and seeded in 96-well plates (Packard, Meriden, CT, USA) at a density of 5,000 cells per well in 100 µL. The assay medium consisted of phenol redfree DF and fetal calf serum treated with 5% dextran-coated charcoal (DCC-FCS). DCC-FCS was prepared as described by Horwitz and McGuire (38). After 24 hr, when wells were approximately 50% confluent, the assay medium was renewed. After another 24 hr, the assay medium was replaced by incubation medium (for preparation, see below), containing DMSO or ethanol stock solutions of the test compounds or estradiol. Solvent concentrations did not exceed 0.1%. The incubation medium was removed after an incubation of 24 hr at 37°C in an atmosphere of 7.5% CO 2 . Cells were washed twice with 100 µL phosphate-buffered saline (PBS) and subsequently lysed in 30 µL low salt (LS) buffer containing 10 mM Tris (pH 7.8), 2 mM dithiothreitrol (DTT), and 2 mM 1,2-diaminocyclohexane-N,N,N´,N´tetraacetic acid. After 10 min of incubation on ice, the 96-well plates were frozen at -80°C for a minimum of 30 min and maximum of 1 day to lyse the cells. The plates were thawed on ice and shaken for 5 min at room temperature. We measured luciferase activity in a luminometer (Labsystems Luminoscan RS, Breda, The Netherlands) with automatic injection of 100 µL flash mix (pH 7.

293-ERα-and 293-ERβs-Luc assay.
The 293-ERα-and 293-ERβs-Luc-based assays were performed similarly to the ER-CALUX assay and have been described previously (32)(33)(34). Briefly, cells were trypsinized and resuspended in assay medium composed of phenol red-free DF containing 30 nM selenite, 10 µg/mL transferin, and 0.2% BSA supplemented with 5% DCC-FCS. The cells were seeded in 96-well plates at a density of 15,000 cells per well in 200 µL assay medium. After 48 hr the cells were 50-60% confluent, and the assay medium was replaced by incubation medium (i.e., containing a 1,000-fold dilution of test compounds) as described for the ER-CALUX assay. After an incubation of 24 hr at 37°C in an atmosphere of 7.5% CO 2 , the plates were transferred to ice and the medium was removed by suction. Luciferase production was assayed as described above for the ER-CALUX assay.
Exposure of cells. Before the T47D.Luc cell incubations, the PBDE and HO-PBDE stock solutions (prepared in DMSO) and the brominated bisphenol compounds (prepared in ethanol) were diluted 1,000-fold in assay medium in a 48-well plate (to obtain a solvent-concentration of 0.1% v/v) and thoroughly shaken, and 100 µL was added to the cells in 96-well plates. The nominal concentrations of the toxicants in the medium were 0.05, 0.1, 0.5, 1.0, and 5 µM, and for potent compounds concentrations of 2.5 and 10 µM were also included. For each experiment, we included a complete E 2 standard curve (1-100 pM, 7 different concentrations in total). In addition, we tested three calibration points (0, 10, and 30 pM E 2 ) on every 96-well plate within an experiment.
For the 293-ERα-and 293-ERβs-Luc assays, the DMSO stock solutions of the tested compounds were diluted 1,000-fold in the appropriate assay medium. The nominal concentrations of the toxicants exposed to the cells were 1.0, 5.0, and 10 µM. For each experiment a complete E 2 standard curve (0.001-10,000 pM in eight different concentrations) was included. For all three ER-CALUX assays, we tested every toxicant concentration in triplicate and repeated each assay at least twice.
Antiestrogenic effects. We tested the possible antiestrogenic effects of the compounds in the ER-CALUX assay at the same nominal concentrations as for the estrogenic activity screening. The T47D.Luc cells were coincubated with an E 2 concentration of 10 pM. This E 2 concentration was the approximate EC 50 for the induction of luciferase activity (31). The percentage (v/v) of DMSO present during these antiestrogenicity incubations was 0.2%. An antiestrogenic effect in this assay was defined by the capacity of a chemical to inhibit the luciferase activity induced by the approximate EC 50  [1] where I is the percent inhibition, and L test , L control , and L E2 are the average luciferase activity of three test wells, three control wells and six wells incubated with 30 pM of E 2 , respectively. Using Equation 1, a compound without antagonistic activity will show the same luciferase induction as 10 pM of E 2 , [i.e., 63.3 ± 7.5% (see "Results")]. On each plate a positive control of 10 nM of the competitive ER antagonist ICI 182,780 was included in triplicate. ICI 182,780 produces virtually total antagonism of E 2 -induced luciferase activity at this concentration [i.e., activity measured is equal to solvent control levels (31)].
Cytotoxicity. We measured possible cytotoxic effects of the tested compounds in the bioassays using 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide activity (39). To determine cytotoxic effects, we seeded cells and exposed them to the test compounds in the same manner as outlined in their corresponding assay procedures.
Dose-response curves and statistics. Possible dose-response relations were described by the sigmoidal function using SlideWrite Plus 4.0 (Advanced Graphics Software, Carlsbad, CA, USA), where y is the induction of luciferase activity compared to controls for estrogenic effects, or inhibition [I (%), Equation 1] for antiestrogenic effects, x is the logarithm of the dose, and a 1 is the maximum y-value. We tested the significance of the data fits using one-way analysis of variance at p < 0.05.

Cytotoxicity
In the concentration range of 0.01-10 µM, none of the incubations of the PBDEs or HO-PBDEs showed any significant effect on MTT activity relative to the solvent control (data not shown). Furthermore, no cytotoxic effect could be observed by microscopic examination in this concentration range.
The T 4 -like HO-BDE compound demonstrated no estrogenic effect up to 10 µM ( Figure 3). In contrast, the T 3 -like HO-BDE and T 2 -like HO-BDE showed the highest estrogenic potencies (EC 50 0.5 and 0.1 µM, respectively) among all compounds tested in this study (Table 1, Figure 3). The compound 4-phenoxy-phenol was included for comparison because it is structurally analogous to the hydroxylated PBDEs. The T 2 -like and T 3 -like HO-BDEs induced maximum luciferase activity at 0.5 µM and 1.0 µM respectively, and this maximum luciferase activity (160 ± 11 and 119 ± 22 %) exceeded that of the natural hormone E 2 ( Table 1).
Of the brominated bisphenols tested, MBBPA and diBBPA showed estrogenic activities comparable to the T 3 -and T 2 -like HO-BDE, with EC 50 values of 0.5 and 0.3 µM, respectively ( Figure 4, Table 1). The maximum luciferase activity of bisphenol A, MBBPA, and diBBPA exceeded the maximum activity induced by E 2 (Figure 4). Bisphenol A and 4-phenoxyphenol had the highest maximum luciferase activity of 199 ± 15% and 195 ± 17%, respectively, relative to the maximum of E 2 (set at 100%, Figure 4). Tetrabromobisphenol A (TBBPA) showed no estrogenic potency within the tested concentrations.

293-ERα-and 293-ERβs-293-Luc Cell Lines
As in previous findings (32)(33)(34), the luciferase activity for the 293-ERα-Luc assay was more sensitive and responsive to E 2 than the 293-ERβs-Luc assay (data not shown). In the present study, the 293-ERα-Luc assay had a 35-fold maximum induction relative to control, which was reached at about 100 pM E 2 . The lowest observed effect concentration (LOEC) and EC 50 for E 2 in the 293-ERα-Luc assay were 2.6 pM and 11.9 pM, respectively. In the 293-ERβs-Luc assay, a 16-fold maximum induction was attained at about 1,000 pM E 2 . The LOEC was 15.3 pM and the EC 50 was 90.2 pM for E 2 .

Discussion
In this study we investigated both the estrogenic and antiestrogenic activity of several PBDE congeners, three hydroxylated PBDEs, and some brominated bisphenol A compounds in vitro. To our knowledge, no studies have been performed on the agonistic or antagonistic activity of these compounds in vivo or in vitro at the level of the estrogen receptor. Of the 17 selected PBDEs, 11 congeners were able to exert estrogenic activities in T47D.Luc cells at LOECs as low as 0.05 µM, and EC 50 values ranging from 2.5 to 7.3 µM. In the same ER-CALUX assay, the organochlorine pesticides methoxychlor, endosulfan, and chlordane had a similar potency for luciferase induction, about 1.0 × 10 -6 times the potency of E 2 (31).
Though not very likely, hydroxylated PBDE metabolites formed in situ may have been involved in the estrogenic effects of the PBDEs in the T47D.Luc cells. T47D.Luc cells possess some metabolic capabilities such as cytochrome P450-mediated hydroxylation of estrogens and xenobiotics. P450 1A (42,43), P450 1B (43), and 17β-hydroxysteroid dehydrogenase (44) have been reported in T47D.Luc cells. However, few data are available on the metabolism of PBDEs, and in the only two studies known reporting PBDE metabolism, the major compound excreted was the parent PBDE. Örn and Klasson-Wehler (23) detected five hydroxylated PBDE metabolites (by GC/MS analyses) in feces and various tissues of rats and mice dosed orally with 2,2´,4,4´-tetraBDE (BDE-47), but the major compound excreted was BDE-47. Larsen et al. (45) reported a low (about 1% of the total given dose) biliary and urinary excretion of possible metabolites of 2,2´,4,4´,5-pentaBDE (BDE-99) in conventional and bile-duct cannulated rats.
The hydroxylated PBDE congeners tested in our study have structural resemblance with the thyroid hormones 3,5-diiodothyronine (T 2 ), 3,3´,5-triiodothyronine (T 3 ), and 3,3´,5,5´-tetraiodothyronine (thyroxine, T 4 ). These HO-PBDEs have been reported to bind to the human αand β-thyroid hormone receptor (THR) (25) and compete with the natural hormone T 4 for binding to a human thyroid hormone transport protein, transthyretin (13) in vitro. Several interactions between the thyroid hormone receptor-and estrogen receptor-mediated pathways have been reported, affecting testis development (46) and behavior (47). Since the structure of the hydroxy-phenyl ring in compounds interacting with the ER and THR (hydroxylated PCBs, hydroxylated PBDEs) is similar (with differences in halogen substitution), it is interesting to study the possible interaction of compounds with both pathways. The ranking of estrogenic potency in the T47D.Luc cells of the thyroid hormonelike HO-PBDEs was T 2 -like HO-BDE (EC 50 (31)] tested in the same ER-CALUX assay. Bisphenol A is one of several well-defined phenolic environmental estrogens that are known to elicit estrogen-mediated responses in vivo and in vitro such as the increased proliferation of MCF7 human breast cancer cells (48 -51). The ranking order for estrogenicity of the hydroxylated PBDEs (Figure 3) was the reverse order found for binding to the human αand β-thyroid hormone receptor (THR) (25) and human transthyretin (TTR) (13) in vitro. This comparison between ER and THR interactions emphasizes that nonbromination of the phenolic ring is necessary for optimum interaction with the ER, which was also found for HO-PCBs (40,41). Conversely, like the interaction of the natural, iodine-containing T 2 , T 3 , and T 4 thyroid hormones with THR and TTR, increasing bromination in adjacent positions on the HO-PBDEs increases THR and TTR binding affinity. The same is true for the brominated bisphenols. The ranking of estrogenic potency in the T47D.Luc cells of the brominated bisphenols was monoBBPA (EC 50 , 0.5 µM) d iBBPA (EC 50 , 0.3 µM) >> triBBPA (EC 50 > 10 µM) >>> TBBPA, and was also the reverse order found for interaction with human TTR in vitro (13). The addition of bromine atoms in the meta position of the aromatic ring (in diBBPA) had no significant effect on the estrogenic potency. This is in line with results published by Perez et al. (51), where the estrogenicity of 2,2-bis(4hydroxy-3-methylphenyl)propane (i.e., one methylgroup in the meta position of one aromatic ring) in a bioassay with MCF7 human breast cancer cells was not changed compared to bisphenol A. However, the introduction of two bromine atoms in the meta position of one aromatic ring drastically decreased the estrogenic potency (triBBPA, this study).
In contrast to the HO-PBDEs, the major HO-PCBs identified in human serum were mostly antiestrogenic but exhibited low to nondetectable estrogenic activities in several in vitro bioassays (48). At concentrations as high as 10 M, several 4-OH-substituted PCBs were not estrogenic toward binding of rat uterine ER. Furthermore, the same HO-PCBs did not induce the proliferation of MCF7 human breast cancer cells, or the luciferase activity of transiently transfected HeLa.Luc cells and MCF7 cells. Unlike the present HO-PBDEs, these HO-PCBs possessed tri-to tetrachlorine substitution on the phenolic ring. In this study, only three of the PBDEs [2,2´,4,4´,5,5´-hexaBDE (BDE-153), 2,3,4,4´,5,6-hexaBDE (BDE-166), and 2,3,3´,4,4´,5,6-hepta-BDE (BDE-190)] showed antiestrogenic activities with concentrations resulting in 50% inhibition (IC 50 values) ranging from 0.8 to 3.1 µM. These PBDEs are likely not metabolized in situ because the congeners are hexa-or heptabromine substituted, have two parabromines, and have no adjacent or ortho-meta brominated carbons. Since the T47D.Luc cells express a functional Ah receptor, it may be possible that the anti-estrogenicity of these PBDEs is Ah receptor-mediated, as is the case for 2,3,7,8-tetrachlorodibenzo-pdioxin (TCDD) and several other antiestrogens (52). BDE-153, -166, and -190 induced the highest maximal luciferase activity in an Ah receptor CALUX assay based on H4IIE.Luc cells, among the same set of 17 PBDEs (21).
The antiestrogenicity of Ah receptor ligands is directly correlated to their affinity for the Ah receptor and their CYP1A-inducing potency (52). As shown for TCDD-treated MCF7 cells (53), the result is enhanced estrogen catabolism, and lower availability of estrogen to the cell. This correlation between structure-antiestrogenicity-and structure-CYP1A-inducing potency has been shown for various halogenated aromatics such as TCDD and non-ortho PCBs in vivo and in vitro (55,56). The exact mechanism of antiestrogenicity is probably specific to species, cell type, and the estrogen-responsive gene. Other possible cellular mechanisms of Ah receptor-mediated antiestrogenicity of BDE-153, -166, and -190 may be that the Ah receptor decreases the binding of the ER to the estrogen-responsive element, or the Ah receptor could act as a repressor by inhibiting the binding of other transcription factors (ER) or the disruption of promotor function.
Interestingly, the HO-PBDEs induced luciferase to a higher maximum activity than the maximum induction generated by E 2 , though at higher concentrations. This has been shown for several other compounds mimicking the natural estrogen in reporter gene assays. Legler et al. (31) reported this phenomenon for the environmental estrogens genistein, nonylphenol, bisphenol A, o,p´-DDT, and methoxychlor in the same T47D.luc cells. Routledge and Sumpter (57) showed that genistein and 4-tert-octylphenol Articles • Estrogenicity of polybrominated diphenyl ethers Environmental Health Perspectives • VOLUME 109 | NUMBER 4 | April 2001