Differential Effects of Xenoestrogens on Coactivator Recruitment by Estrogen Receptor (ER) α and ERβ*

It has been proposed that tissue-specific estrogenic and/or antiestrogenic actions of certain xenoestrogens may be associated with alterations in the tertiary structure of estrogen receptor (ER) α and/or ERβ following ligand binding; changes which are sensed by cellular factors (coactivators) required for normal gene expression. However, it is still unclear whether xenoestrogens affect the normal behavior of ERα and/or ERβ subsequent to receptor binding. In view of the wide range of structural forms now recognized to mimic the actions of the natural estrogens, we have assessed the ability of ERα and ERβ to recruit TIF2 and SRC-1a in the presence of 17β-estradiol, genistein, diethylstilbestrol, 4-tert-octylphenol, 2′,3′,4′,5′-tetrachlorobiphenyl-ol, and bisphenol A. We show that ligand-dependent differences exist in the ability of ERα and ERβ to bind coactivator proteinsin vitro, despite the similarity in binding affinity of the various ligands for both ER subtypes. The enhanced ability of ERβ (over ERα) to recruit coactivators in the presence of xenoestrogens was consistent with a greater ability of ERβ to potentiate reporter gene activity in transiently transfected HeLa cells expressing SRC-1e and TIF2. We conclude that ligand-dependent differences in the ability of ERα and ERβ to recruit coactivator proteins may contribute to the complex tissue-dependent agonistic/antagonistic responses observed with certain xenoestrogens.

It has been proposed that tissue-specific estrogenic and/or antiestrogenic actions of certain xenoestrogens may be associated with alterations in the tertiary structure of estrogen receptor (ER) ␣ and/or ER␤ following ligand binding; changes which are sensed by cellular factors (coactivators) required for normal gene expression. However, it is still unclear whether xenoestrogens affect the normal behavior of ER␣ and/or ER␤ subsequent to receptor binding. In view of the wide range of structural forms now recognized to mimic the actions of the natural estrogens, we have assessed the ability of ER␣ and ER␤ to recruit TIF2 and SRC-1a in the presence of 17␤-estradiol, genistein, diethylstilbestrol, 4-tert-octylphenol, 2,3,4,5-tetrachlorobiphenyl-ol, and bisphenol A. We show that ligand-dependent differences exist in the ability of ER␣ and ER␤ to bind coactivator proteins in vitro, despite the similarity in binding affinity of the various ligands for both ER subtypes. The enhanced ability of ER␤ (over ER␣) to recruit coactivators in the presence of xenoestrogens was consistent with a greater ability of ER␤ to potentiate reporter gene activity in transiently transfected HeLa cells expressing SRC-1e and TIF2. We conclude that ligand-dependent differences in the ability of ER␣ and ER␤ to recruit coactivator proteins may contribute to the complex tissue-dependent agonistic/antagonistic responses observed with certain xenoestrogens.
One of the greatest challenges in understanding the mechanisms of estrogen action has been to determine how different estrogen receptor (ER) 1 ligands (steroidal estrogens, antiestrogens, xenoestrogens) produce such diverse biological effects. The recent discovery of a second subtype of the estrogen receptor, named estrogen receptor-␤ (ER␤) to distinguish it from the classical ER (now renamed ER␣), adds another level of complexity to the mechanism of estrogen action and has opened new possibilities by which estrogens might exert tissue-and cell-specific effects (1). Indeed, it has been shown that ⌭R␣ and ER␤ differ in terms of their ability to activate gene expression from either the consensus estrogen response element (ERE) from the VTG gene or the divergent ERE from the luteinizing hormone ␤ gene in transiently transfected Cos-1 cells (2). Moreover, ⌭R␣ and ER␤ activate and inhibit, respectively, transcription from an AP1 enhancer site when complexed to 17␤estradiol (E2), whereas ER␤ was a transcriptional activator on AP1 sites when complexed to antiestrogens (3). Studies in rodents have revealed that the distribution and relative levels of ⌭R␣ and ER␤ expression differ among tissues. For example, ⌭R␣ is predominantly expressed in the pituitary, uterus, ovary (oviduct and germinal epithelium), mammary gland, testis, epididymis, and kidney, whereas ER␤ is the predominant form in regions of the hypothalamus, ovary (granulosa cells), prostate gland, lung, and bladder (4 -8). The coexpression of ⌭R␣ and ER␤ in certain tissues and cells, and the ability of ⌭R␣ and ER␤ to form heterodimers and bind to a consensus EREs in vitro, suggests that alternative estrogen signaling pathways may exist in cells expressing both ER isoforms (9,10), although this remains to be proven in vivo.
Estrogen receptors activate transcription of target genes via two activation functions; AF-1 (in the N-terminal domain) is ligand-independent and is regulated by phosphorylation in response to growth factors (11), whereas AF-2 is closely associated with the LBD and depends on ligand binding for its transcriptional activity. The activities of AF-1 and AF-2 of the ER vary depending upon the responsive promoter (12) and cell type, and in some cases both are required for full transcriptional activation of target genes (13,14). Although the amino acid homology within the LBD of rat ER␣ and ER␤ does not exceed 55%, a number of residues required for ligand binding and for the formation of the hydrophobic pocket are highly conserved between the two receptor isoforms. Therefore, the reported similarity in the relative binding affinities of ER␣ and ER␤ for a range of natural and synthetic estrogens may have been anticipated (15,16). However, ER␣ and ER␤ contain a region of relatively low amino acid homology within the LBD, a region that has been shown to be accessible to proteolytic attack (17), and is therefore likely to contain residues on an exposed surface of the receptor, which may be involved in subsequent receptor-protein interactions. The ligand-dependent transactivation domain AF-2 is of particular interest because it has provided a mechanistic explanation for the observed functional differences between agonists and antagonists of the ER. It is now believed that the primary role of 17␤estradiol binding to the ER is to induce a conformational change in the tertiary structure of the ER, which subsequently affects the alignment of a highly conserved amphipathic ␣-helix (helix 12) within AF-2 (18). Correct alignment of helix 12 (in the presence of 17␤-estradiol), exposes residues that interact with other proteins (known as transcriptional intermediary factors or coactivators) necessary for the formation of a stable pre-initiation complex (19), whereas it is misaligned with the estrogen antagonist raloxifen (20). Together, these findings suggest that ER␣ and/or ER␤ may display different profiles for coactivator protein recruitment depending on the structure of the ligand.
It was recently shown (using a range of short peptide sequences directed toward the surface of the ER) that a range of natural and synthetic estrogens/anti-estrogens induced distinct conformational changes in the tertiary structure of ER␣ and/or ER␤ (35). Thus, it has been proposed that the tissuespecific estrogenic and/or antiestrogenic actions of certain xenoestrogens may be associated with distinct changes in the tertiary structure of ER␣ and/or ER␤ following ligand binding; changes that are sensed by cellular factors required for normal gene expression. Indeed, ER␣ was recently reported to display ligand-dependent selectivity for coactivator recruitment in a two-hybrid system (expressing chimeric ER␣ and coactivator fusion proteins) in yeast (36).
In view of the wide range of chemical structures now known to mimic the actions of the natural estrogens (37), GST pulldown assays were used to assess the ability of both ER␣ and ER␤ to recruit transcriptional intermediary factor-2 (TIF2) and steroid receptor coactivator-1a (SRC-1a) in vitro, with 17␤estradiol, genistein (Gen, a phytoestrogen), diethylstilbestrol (DES, a stilbene), 4-tert-octylphenol (OP, an alkylphenol), 2Ј,3Ј,4Ј,5Ј-tetrachlorobiphenyl-ol (PCB-OH, a PCB metabolite), and bisphenol A (Bis-A, a biphenolic compound). We show that ER␣ and ER␤ differ in terms of their ability to recruit SRC-1a and TIF2 with the various xenoestrogens in vitro, despite the two receptors having relatively similar binding affinities for these compounds. The enhanced ability of ER␤ to recruit coactivators in vitro, in the presence of xenoestrogens, was also consistent with the greater capacity of ER␤ to potentiate reporter gene expression in transiently transfected HeLa cells carrying expression plasmids for SRC-1e and TIF2 relative to ER␣.
Coupled in Vitro Transcription and Translation-All in vitro translations were performed using a TNT ® coupled rabbit reticulocyte lysate system, with the appropriate RNA polymerase (T7), according to the manufacturer's instructions (Promega). Radiolabeled proteins were synthesized by substituting 1 mM methionine with [ 35 S]methionine (Amersham Pharmacia Biotech) in the reaction mixture. Radiolabeled products from the in vitro translations were analyzed by 8% ( 35 S-TIF2) or 10% ( 35 S-ER␣ and 35 S-ER␤) SDS-PAGE, respectively.
Ligand Binding-Ligand binding by ER␣ and ER␤ was assessed using [2,4,6, H]estradiol (Amersham Pharmacia Biotech). A working stock of 3 ϫ 10 Ϫ8 M [2,4,6,7-3 H]estradiol was prepared in ethanol. ER␣ and ER␤ proteins were synthesized in vitro from the pSG5-expression vectors using the TNT ® T7 Quick Coupled Transcription/Translation System (Promega), and reaction mixtures containing the translated receptors were snap-frozen in 45-l aliquots at Ϫ70°C until required. to each reaction tube. The tubes were mixed briefly, incubated on ice for 5 min, and centrifuged for 5 min at 4°C to pellet the charcoal. 80 l of supernatant was removed and added directly to ␤-vials (Pony Vial, Packard ® ) containing scintillant (Liquiscint, National Diagnostics). Bound radioactivity was measured using a TRI-CARB ® 2000CA liquid scintillation analyzer (Packard ® ). Specific binding in the presence of competitor was expressed as a percentage of the maximum binding (calculated by subtracting the nonspecific binding from the total binding). Receptor binding affinity (RBA) was calculated as the ratio of concentrations of E2 or competitor required to reduce the specific radioligand binding by 50% (RBA value for E2 was arbitrarily set at 100 for both receptors).
GST Pull-down Assays-Two types of GST pull-down assays were used to assess the ability of ER␣ and ER␤ to recruit SRC-1a and TIF2 following binding to a range of xenoestrogens. The first employed GST-AF2␣ or GST-AF2␤ fusion proteins with in vitro translated 35 S-TIF2, and the second employed a GST-SRC1a fusion protein with in vitro translated 35 S-ER␣ or 35 S-ER␤ (see Fig. 1).
Glutathione-Sepharose beads (Amersham Pharmacia Biotech) were pre-washed in NETN containing protease inhibitors (NETNϩP) and 0.5% powdered milk (to minimize nonspecific binding), and were resuspended in one volume of NETNϩP. Fusion proteins (GST-AF2␣/␤ or GST-SRC-1a) contained within the bacterial lysates were purified onto the pre-washed glutathione-Sepharose beads (25 l of beads/ml of lysate) by incubation for 1.5 h at 4°C on a rotary mixer. The beads (loaded with fusion proteins) were collected by centrifugation, washed four times with NETNϩP, and resuspended in one volume of NETNϩP prior to use. 50 l of beads, containing GST fusion proteins or GST alone (negative control), were incubated overnight at 4°C in Eppendorf tubes containing 845 l of NETNϩP, 80 l of Me 2 SO, and 15 l of the in vitro translated 35 S-labeled receptor or coactivator (Fig. 1), in the presence of 10 l of vehicle or test chemical (in Me 2 SO). The beads were then washed four times with NETN (all supernatant removed in final wash), dried in a Speed Vac for 30 min, resuspended in 61 l of 2ϫ protein loading buffer (reducing), and boiled for 4 min to release all the bound proteins from the beads. A 50-l aliquot of the protein loading buffer was then removed from the tubes; half of this sample was separated by SDS-PAGE, and the remaining half was added directly to ␤-vials (Pony Vial, Packard ® ) containing scintillant (Liquiscint, National Diagnostics), and the radioactivity was counted using a TRI-CARB ® 2000CA liquid scintillation analyzer (Packard ® ). SDS-PAGE gels were fixed and dried, and the 35 S-labeled proteins were visualized by fluorography. The relative recruitment ability (RRA) was calculated using Equation 1.
The RRA value for E2 was arbitrarily set at 100 for both receptors.
Cell Transfection Assay-HeLa cells were plated in 96-well microtiter plates in phenol red-free medium, and were incubated overnight at 37°C to reach approximately 30% confluence. Cells were transfected using the calcium phosphate coprecipitation method, as described previously (18). The transfected DNA included a reporter plasmid (p(Gal4) 5 .TK.GL3; 100 ng/well), an internal control plasmid (pRLCMV; 0.5 ng/well), human ER expression vectors (pGal4-AF2␣/␤; 10 ng/well), the coactivator expression vectors (pSG5-SRC1e or pSG5-TIF2; 10 ng/ well), and pMT2 to a total of 120 ng/well. DNA precipitates were prepared in 500-l volumes, to which 10 l was added to the appropriate wells. Transfections were allowed to occur overnight, after which time the cells were washed three times with fresh medium and were maintained with no hormone (Me 2 SO), 17␤-estradiol (1 nM), or various concentrations of xenoestrogen as required, using 100 l of medium/ well (Me 2 SO, 0.1% final concentration). After 24 h, 50 l of medium was removed from each well and was replaced with LucLite substrate according to the manufacturer's instructions (Packard). The cells were left to lyse at room temperature for 10 min, and the well contents (100 l containing LucLite reagent, medium, and lysed cells) were transferred to a white microtiter plate where the extracts were assayed for luciferase activity. Transfection efficiency was assessed 10 min after the addition of 25 l of Renlite reagent (1 mg/ml coelenterazine stock substrate in Me 2 SO, diluted 100-fold in 0.5 M HEPES, pH 7.8, 20 mM EDTA) to each well. The EDTA within the Renlite reagent chelates the divalent cations required for the firefly luciferase activity and quenches the light emission, allowing the Renilla luciferase activity from pRL-CMV to be determined. All treatments were carried out in duplicate (controls; pGL3.basic, p(Gal4) 5 .TK.GL3 alone, and p(Gal4) 5 .TK.GL3 plus pSG5-SRC1e or pSG5-TIF2), or quadruplicate (i.e. pGal4-AF2␣/␤ plus p(Gal4) 5 .TK.GL3 in the presence or absence of pSG5-SRC1e or pSG5-TIF2), and experiments were repeated for consistency. Mean values from a representative experiment are presented.

Specificity of ER␣ and ER␤ to a Range of Xenoestrogens-
The binding affinities of E2, DES, Gen, PCB-OH, OP, and Bis-A for ⌭R␣ and ER␤ were assessed by their ability to compete with [2,4,6,7-3 H]17␤-estradiol for binding to in vitro translated receptors over a 100,000-fold concentration range (Fig. 2). All of the chemicals tested were able to compete with tritiated 17␤estradiol for binding to both ER␣ and ER␤ in a dose-dependent manner. RBA for ⌭R␣ and ER␤ were 53 and 150 for DES, 0.7 and 15 for Gen, 1.6 and 3.0 for PCB-OH, 0.013 and 0.25 for OP, and 0.073 and 0.75 for Bis-A, respectively (Table I). ER␤ always had a greater RBA for all the xenoestrogens tested compared with ⌭R␣. The largest differences in RBA were seen with Gen and OP, which were approximately 20-fold higher with ER␤ than ⌭R␣ (Table I).
⌭R␣ and ER␤ Differ in Their Ability to Recruit Coactivator Proteins following Xenoestrogen Binding-The ability of either ER␣ or ER␤ fusion proteins to bind TIF2 and SRC-1a with different concentrations of ligand was assessed using two GST pull-down assay systems. Thus, the ability of GST-LBD to bind [ 35 S]methionine-labeled TIF2 or SRC-1a was examined, and conversely the ability of GST-SRC-1 (550 -770) to bind [ 35 S]methionine-labeled ER␣ or ER␤ was tested. Fig. 3 shows a typical autorad from a GST pull-down assay, where the dose-dependent recruitment of 35 S-TIF2 by GST-AF2␣ with E2 and DES is demonstrated. All of the chemicals tested enabled ⌭R␣ and ER␤ to bind TIF2 and SRC-1a in a dose-dependent manner with one exception; ER␣ was not able to bind either TIF2 or SRC-1a in the presence of PCB-OH over the concentration The ability of radiolabeled TIF2 to interact with the hER was investigated using a GST fusion protein containing the LBD and AF-2 of either hER␣ or hER␤ (A). In contrast, the ability of radiolabeled ER␣ or ER␤ (entire protein) to interact with SRC-1a was investigated using a GST-SRC-1a fusion protein (B). The GST fusion proteins associate with the beads, and the ligands interact with the LBD of the ER. Ligand-induced conformational changes in the tertiary structure of the estrogen receptor are associated with differences in the receptor's ability to interact with coactivator proteins (TIF2 or SRC-1a). range tested (Figs. 4 and 5, respectively). In contrast, ER␤ was able to bind TIF2 (albeit submaximally; Ͻ25% of the maximal inducible response) and SRC-1a with PCB-OH (Figs. 4B and 5B). The RRA of ⌭R␣ and ER␤ for TIF2 were 11 and 60 for DES, 0.005 and 60 for Gen, 0.0002 and 0.2 for OP, and Ͻ0.0001 and 0.05 for Bis-A, respectively (Table II). The RRA of ⌭R␣ and ER␤ for SRC-1a were 5 and 50 for DES, 0.06 and 2 for Gen, 0 and 0.05 for PCB-OH, 0.002 and 0.002 for OP, and 0.0003 and 0.0002 for Bis-A, respectively (Table II). The 20-fold greater affinity of ER␤ for Gen (Table I; Fig. 2) resulted in a 12,000and 33-fold greater ability to bind TIF2 and SRC-1a, respectively, relative to ⌭R␣ (Table II; Figs. 4 and 5). In contrast, the abilities of ER␣ and ER␤ to bind SRC-1a with octylphenol and bisphenol A were similar (Table II), despite the greater binding affinities of these compounds for ER␤ (Table I).
Mammalian Cell Transfection Assay-The consequence of differences in the ability of ⌭R␣ and ER␤ to bind coactivator proteins (as shown in the GST pull-down assays) on gene expression was assessed in transiently transfected HeLa cells, using expression vectors encoding the DNA-binding domain of Gal4 fused to the AF-2 domains of ⌭R␣ or ER␤. These chimeric receptors recognize and stimulate transcription from the p(Gal4) 5 .TK.GL3 reporter gene construct in the presence of ligand. As the magnitude of the transcriptional response was found to vary between AF2␣ and AF2␤ (i.e. pGal4-AF2␣ produced a response that was approximately 2-3 times greater than that of pGal4-AF2␤), the changes in the transcriptional response were expressed relative to the maximal inducible response observed with E2 (10 Ϫ8 M) in the absence of excess coactivator (response arbitrarily set at 100%). Differences in the base-line response for ER␣/␤ in the absence of hormone (NH) with excess coactivator required that potentiation (or -fold increases) of reporter gene activity were calculated from their respective base line in each case. The base-line response for Gal4-AF2␤ in the absence of hormone was not resolved above the background level of the assay. This is evident from the raw data, where the NH response of ER␣ without coactivator (14.0 Ϯ 4.1 luciferase units) occurred above the background level (4.4 Ϯ 1.8 luciferase units), whereas the NH response of ER␤ without coactivator (4.4 Ϯ 2.1 luciferase units) did not (data not shown). Therefore, the calculated percentage of response of ER␤ in the absence of hormone and coactivator was overestimated, and the resultant -fold increase in reporter gene activity (calculated using this baseline value) with ER␤ was underestimated. Consequently, we were unable to determine the true increase in reporter gene activity with ER␤ in this instance, and therefore potentiation of reporter gene activity in the absence of excess coactivator was not presented.
Figs. 6 -8 show the potentiation (expressed as a -fold increase above the respective base line) of reporter gene activity by ER␣ and ER␤ in transiently transfected HeLa cells carrying expression plasmids for SRC-1e and TIF2 exposed to xenoestrogens and E2. In all cases, potentiation of reporter gene activity was greater with ER␤ than ER␣. This is consistent with the results from the pull-down assay, in which the RRA values of ER␤ for SRC-1a and TIF2 with each xenoestrogens always exceeded those of ER␣ (Table II). The magnitude of reporter gene potentiation was also associated with the RRA values calculated in the pull-down assays (Table II). For example, in the presence of ER␤, genistein (10 Ϫ6 M) was able to potentiate reporter gene activity to levels around 72% and 86% of those produced by E2 (10 Ϫ8 M) with SRC-1e (RRA of 33) and TIF2 (RRA of 60), respectively. In contrast, potentiation of reporter gene expression by PCB-OH and Bis-A (which had lower RRA

FIG. 3. Results of GST pull-down assays showing the liganddependent interaction between hER␣ and coactivator in vitro.
In vitro translated 35 S-labeled TIF2 proteins were incubated with glutathione-Sepharose beads carrying GST-AF2␣ fusion proteins with various concentrations of E2 and DES (10 Ϫ5 M to 10 Ϫ9 M). 1/10 represents one-tenth of the amount of 35 S-TIF2 protein used in each pull-down assay incubation, ϩ indicates the maximal possible response in the assay (i.e. 100% TIF2 recruitment) obtained using 10 Ϫ5 M E2, and Ϫ denotes the response with Me 2 SO (carrier solvent) alone. After washing, the proteins were eluted from the beads and were separated by 7% SDS-PAGE. Gels were fixed and dried, and the labeled values) did not generally exceed 50% of the potentiation of E2 (10 Ϫ8 M) for both ER␣ and ER␤ even at higher concentrations. DISCUSSION The primary purpose of 17␤-estradiol binding to the ER is to induce a conformational change in the tertiary structure of the ER, such that AF-2 is in a position to mediate the assembly of the basal transcription machinery following the recruitment of coactivators or transcription initiation factors (18). However, the actual conformational change in the tertiary structure of the ER induced by xenoestrogens may differ from that of 17␤estradiol (35) due to differences in the steric and electrostatic properties of the various ligands. In this study, we show that binding of a range of natural and synthetic xenoestrogens to ER␣ and ER␤ alters their abilities (to different extents) to recruit coactivator proteins in vitro, and this may in turn affect their abilities to potentiate the expression of a reporter gene in transiently transfected HeLa cells. In short, ligand-dependent differences in the ability of ER␣ and ER␤ to recruit coactivator proteins may also contribute to the complex tissue-dependent responses observed with certain xenoestrogens.
Receptor-binding assays and GST pull-down assays were used to compare the affinities of xenoestrogens for both ER subtypes with their subsequent abilities to recruit SRC-1a and TIF2 in vitro. We found that all of the xenoestrogens tested were able to displace tritiated 17␤-estradiol from ER␣/␤ in a dose-dependent manner. The binding affinity of genistein for ER␤ was 20-fold higher than ER␣, which was consistent with previous findings (16). Although it is not possible to compare the recruitment profiles of the two GST pull-down assays di-rectly (because the two systems are not analogous; see Fig. 1), it was interesting to note that the RBA of the various xenoestrogens for ER␣/␤ were not always consistent with their subsequent ability to recruit coactivator proteins. This finding may explain previous reports that there is not always a direct correlation between binding affinity and transcriptional potency with certain ER ligands (40). With most of the xenoestrogens tested, ER␣ and ER␤ were able to recruit TIF2 and SRC-1a in a dose-dependent manner. However, ER␣ was unable to recruit TIF2 and SRC-1a, and ER␤ displayed submaximal recruitment of TIF2, with PCB-OH. In contrast, ER␤ was able to recruit SRC-1a fully (and in a dose-dependent manner) with PCB-OH. The observed differences in the ability of ER␣/␤ to recruit TIF2 and SRC-1a with PCB-OH were not anticipated (Table II), given the similar RBA of this compound for both ER subtypes ( Table I). The 20-fold selective affinity of genistein for ER␤ (Table I) resulted in a 12,000-and 33-fold greater ability of ER␤ to recruit SRC-1a and TIF2, respectively, compared with ER␣ (Table II). In contrast, the ability of ER␣ and ER␤ to recruit SRC-1a with octylphenol and bisphenol A were similar, despite the higher binding affinities of these two compounds for ER␤. In general, ER␤ had a greater RRA for SRC-1a and TIF2 than ER␣ with all the xenoestrogens tested. However, unlike their RBAs (which only varied by up to 1 order of magnitude between ER␣ and ER␤), their RRAs differed by as much as 4 orders of magnitude (Tables I and II). Thus, receptor binding affinities of xenoestrogens for ER␣ and ER␤ may not accurately predict the receptors' subsequent abilities to recruit different coactivator proteins. Certain coactivators appear to be differentially expressed among tissues, suggesting that they may be involved in the regulation of tissue-selective gene expression. As an example, the coactivator SRC-3 is abundant in the mammary gland and uterus (30), and has a higher affinity for ER␣ (relative to ER␤), which is the predominant ER form in these tissues. Moreover, it was recently shown that the ability of ER␤ to stimulate ERE-TK-Luc reporter gene expression in transiently transfected cells in the presence of 17␤-estradiol was dependent on the cell line used (38). Together, these findings suggest that (promoter context aside) the behavior of the ER subtypes within any given cell or tissue will be influenced by the type and level of the accessory proteins present. Thus, in the context of the whole tissue, the results from the GST pull-down assays may imply that genistein would be most effective in cells containing ER␤, and where TIF2 is the predominant coactivator present. In contrast, PCB-OH may be an agonist or antagonist, respectively, in cells containing ER␤ or ER␣ where SRC-1a is the main coactivator. In addition, bisphenol A is likely to be most effective in cells containing ER␤ where TIF2 is the main coactivator, but may be equally effective in cells containing either ER␣ or ER␤ when SRC-1a is predominant. In other words, these results indicate that liganddependent differences in the ability of the ER to recruit coactivators may alter, in part, the receptors' ability to potentiate gene expression in whole cells.
Transiently transfected HeLa cells were used to assess whether the differences in coactivator recruitment observed in the GST pull-down assays were also consistent with the ability of ER␣ and ER␤ to transactivate an estrogen-responsive reporter gene construct with excess coactivator (SRC-1e or TIF2). The HeLa cell system brings together all the different components of ER action (ligand binding, receptor dimerization, DNA binding, coactivator recruitment, and gene transcription), and the activity of the receptor is measured by the strength of the luciferase (reporter gene) response. Within the context of the whole cell, it is impossible to conceal the receptor from the influence of other endogenous coactivators, corepressors, or receptor-associated proteins contained within the cells, or to isolate the relative contribution of AF-1 and AF-2 on the transcriptional response. This issue was addressed, in part, using the Gal4 system, in which the N-terminal part of the ER is removed, thus restricting any activation of the reporter gene to the ligand-dependent activation function (AF-2) of the ER alone. These chimeric receptors recognize and stimulate expression of the firefly luciferase reporter gene construct (containing Gal4 DNA-responsive elements upstream of a TK promoter) in the presence of ligand. In these experiments, the coactivator SRC-1e was chosen because it was previously shown to be a more effective potentiator of gene expression compared with SRC-1a (27).
Interpretation of the cell line data was complicated by a detection-limit artifact in which the level of ER␤ expression (in the absence of hormone and excess coactivator) was not resolved above the background level of the assay. Therefore, it was not possible to compare the behavior of ER␣ and ER␤ in the presence and absence of excess coactivator. Potentiation of reporter gene activity in the presence of PCB-OH was consistently higher with ER␤ than ER␣ in cells expressing SRC-1e. However, PCB-OH was able to potentiate reporter gene activity in the absence of excess coactivator with ER␣, indicating that HeLa cells must contain other endogenous coactivators which can function in this case (Fig. 6). Genistein was more effective with ER␤ than ER␣ with both SRC-1e and TIF2 (RRA ratio ϭ 33 and 12,000, respectively) as predicted by the pull-down assays. Moreover, bisphenol A potentiated ER␣ and ER␤ similarly with SRC-1e (RRA ratio ϭ 0.67), but the response was greater with ER␤ plus TIF2 (RRA ratio Ͼ 500). In all cases the magnitude of the reporter gene response was greater with ER␣, whereas the relative increase in reporter gene activity was greater with ER␤. This suggests that factors other than the binding affinity of the ligand for the receptor and the ability of the receptor to recruit coactivators may also affect reporter gene activity. Nevertheless, given the increase in complexity between the binding assays and the whole cell system, there is a remarkable consistency in the results. Within the group of chemicals tested, we did not come across a ligand with ER␣selective coactivator recruitment, and therefore we were not able to compare this type of response profile in the transfection  6. Potentiation of reporter gene activity by ER␣-AF2 and ER␤-AF2 with E2 and PCB-OH in the presence of SRC-1e. Transiently transfected HeLa cells, carrying expression vectors for either Gal4-AF2␣ or Gal4-AF2␤, were assessed for their ability to stimulate reporter gene expression (p(Gal4) 5

assays.
It was interesting to note that the two chemicals that showed the biggest differences between ER␣ and ER␤ coactivator recruitment were the isoflavone phytoestrogen (genistein) and a hydroxylated PCB metabolite (2Ј,3Ј,4Ј,5Ј-tetrachlorobiphenylol). The mixed agonistic-antagonistic effects of these chemical groups on estrogen-mediated processes in mammals and mammalian cells are well established (41)(42)(43). Many flavanoids have now been shown to competitively bind to ER␣ (15,44) and induce reporter gene activity in transiently transfected MCF-7 cells and yeast containing E2-responsive reporter constructs (45,46). However, the same flavanoids are inactive in hormonedependent cell proliferation assays (MCF-7), and may inhibit both the proliferative activity of E2 in co-treated MCF-7 cells and E2-induced gains in uterine weight in immature rats (45,47). One possible explanation for this disparity is that different ligand-induced conformational changes in the receptor may enable ER␣ (the predominant ER form in the mammary gland and uterus) to activate gene expression on certain promoters, but not on others. Therefore, the much reported promotercontext specific action of ER action may be a consequence of the type of coactivators present, and the conformational change of the ER induced by the ligand. The fact that genistein had a significantly higher binding affinity and relative recruitment ability for SRC-1a and TIF2 with ER␤ versus ER␣ was intriguing, given the reported high expression of ER␤ in the secretory epithelial cells of the prostate (15), and the putative role these compounds play in preventing prostate cancer (48).
In utero and lactational exposure to PCBs (or commercial mixtures called "Arochlors") is associated with persistent neurobehavioral, reproductive, and endocrine alterations (reviewed in Ref. 49), which are species-, age-, and congenerspecific. Metabolism of PCBs by humans and rodents results in the formation of hydroxylated PCBs, many of which have been shown to be estrogenic in MCF-7 cells and transiently transfected HeLa cells (50). However, few, if any, studies have investigated the estrogenic and/or antiestrogenic activity of individual PCB congeners on the pituitary-hypothalamic axis and uterus in vivo, and therefore it is currently difficult to speculate whether the complex tissue-dependent effects of Arochlors are mediated by one or more specific congeners, which have selective ER␣/ER␤ agonistic-antagonistic effects. However, the fact that ER␣ is the predominant form expressed in the stromal and epithelial cells of the endometrium (uterus), and ER␤ is expressed in high amounts in the paraventricular and supraoptic nucleus of the hypothalamus (4) implies that the predominant ER isoform present may determine, in part, the type of tissue response observed following PCB exposure, i.e. the heterogeneity of estrogen receptor distribution, and the predominant types of coactivators present, may contribute to the diversity of tissue responses to estrogenic chemicals. However, the true significance of these findings may only become apparent when the functional roles of ER␣ and ER␤ with the various receptorinteracting proteins are known. These roles may become clearer when dominant negative versions of receptor-interactive proteins are analyzed, or when knockout animals are generated.