Assessing environmental chemicals for estrogenicity using a combination of in vitro and in vivo assays.

Because of rampant concern that estrogenic chemicals in the environment may be adversely affecting the health of humans and wildlife, reliable methods for detecting and characterizing estrogenic chemicals are needed. It is important that general agreement be reached on which tests to use and that these tests then be applied to the testing of both man-made and naturally occurring chemicals. As a step toward developing a comprehensive approach to screening chemicals for estrogenic activity, three assays for detecting estrogenicity were conducted on 10 chemicals with known or suspected estrogenic activity. The assays were 1) competitive binding with the mouse uterine estrogen receptor, 2) transcriptional activation in HeLa cells transfected with plasmids containing an estrogen receptor and a response element, and 3) the uterotropic assay in mice. The chemicals studied were 17 beta-estradiol, diethylstilbestrol, tamoxifen, 4-hydroxytamoxifen, methoxychlor, the methoxychlor metabolite 2,2-bis(p-hydroxyphenyl)-1,1,1-trichloroethane (HPTE), endosulfan, nonylphenol, o,p'-DDT, and kepone. These studies were conducted to assess the utility of this three-assay combination in the routine screening of chemicals, or combinations of chemicals, for estrogenic activity. Results were consistent among the three assays with respect to what is known about the estrogenic activities of the chemicals tested and their requirements for metabolic activation. By providing information on three levels of hormonal activity (receptor binding, transcriptional activation, and an in vivo effect in an estrogen-responsive tissue), an informative profile of estrogenic activity is obtained with a reasonable investment of resources.

ifen, methoxychlor, the metho'ychlor metaboliee 2,2-bis(p-hydroxyphenyl)-1,1,1trichloroehane (HPTE), endosulfan, nonylphenol, o,p -DDT, and kepone. These studies were conducted to asses the utility ofthis thre-assay combination in the routine ofchemicals, or combinaions of chemicals, for Ie activity. Results were cnsnt among the dthre with repect to what is known about theeogei acivities of the chemicals tested d their for mtocavation. By infmatin on te levels of orTmonal actvity (receptor binding and an iX vivo ect in an responsie tissue), an inormative profile of estroenic activity obtaied with a resonable ivestment of resources. Key w e com bindipg endocrine disruptrs, estrogen receptor, tansection, uterotropic my. Envin Heal Perset 104: 1296-1300 (1996) Estrus is that period of the reproductive cycle in mammalian females in which ovulation occurs and the female is typically most receptive to mating. In rodents, the most commonly used laboratory model for studies of reproduction, estrus follows the surge of estrogen produced by the maturing ovarian follicles during proestrus. Thus, chemicals that induce estrus, or a biological response associated with estrus, are traditionally defined as estrogens (1). The capacity of a substance to induce such effects is termed estrogenicity.
With growing concern that estrogenic chemicals in the environment, either naturally occurring or man-made, may adversely affect the health of humans, domestic animals, and wildlife (2,3), the need for meaningful, standardized, and widely accepted methods for reliably detecting and characterizing estrogenic chemicals has gained importance. Many assays for estrogenicity have been proposed and several are in broad use. Descriptions of many of these are found in the proceedings of the conference on Estrogens in the Environment, III: Global Health Implications (4).
Because of the multiple biological effects of estrogens and the influence of absorption, metabolism, distribution, and excretion on the manifestation of their estrogenic activity, any single assay can provide only limited information on those effects. In vitro assays can provide valuable insights on mechanisms of action but are restricted in their capacity to mimic whole animal metabolism and distribution. In vivo assays permit the detection of effects resulting from multiple mechanisms but may give indications only of gross effects and reveal little about mechanisms of activity.
In an initial effort to assess the utility of using a combination of in vitro and in vivo assays to screen for estrogenicity, 10 known or alleged estrogenic chemicals were tested using assays that assess estrogenic activity at three different levels ofaction. The first assay, competitive binding with the estrogen receptor (ER), uses a cell-free system to determine the extent to which the test chemical binds to the ER, as reflected by its effect on the binding of 170-estradiol (5). Second, if a chemical is estrogenic, the consequence of its binding to the ER should be transcriptional activation of estrogen responsive genes. This effect has been determined using an assay that employs human HeLa cells transfected with an ER and an estrogen response element (ERE) linked to a chloramphenicol acetyl transferase (CAT) reporter gene. Transcriptional activation is determined by measuring the amount of CAT protein produced by the cells following treatment with the test chemical (6). The third assay was used to determine effects on an estrogen-responsive tissue in an intact animal. The weanling mouse uterotropic assay, one of the most well-established and widely used methods to detect estrogenicity, was employed to determine if exposure to the test chemical led to increased uterine wet weight (7).
Thus, 10 chemicals were assessed for evidence of estrogenicity at three levels of activity. The results indicate that this combination of tests may provide an efficient and effective approach to screening environmental chemicals for estrogenic activity. Articles * Three assays for estrogenicity  MgCI2, pH 7.6) and homogenized with a Polytron (Brinkmann Instruments, Westbury, NY) for 15 sec at speed setting 6.5 at a ratio of 50:1 (milligrams of tissue weight/per milliliter of buffer). The homogenate was filtered through 100-125 mm mesh Nitex filtering media and centrifuged at 1,000g for 10 min; the supernatant was then decanted and centrifuged at 45,000 rpm for 50 min. The 105,000g supernatant was used for cytosol receptor binding assays.
Aliquots of 100 pl cytosol were incubated with 5 nM [3H]E2 and increasing concentrations of unlabeled competitors (0.5 nM-5 pM). The mixtures were incubated at 40C for 18 hr, and then 250 pl of 60% HAP (hydroxyapatite) in TEGM buffer was added to each tube. Tubes were centrifuged at 1,000g for 10 min and the resulting HAP pellet was washed twice with 3 ml TEGM buffer then suspended in scintillation cocktail. The radioactivity was measured using a Beckman CS 9800 scintillation counter (Beckman, Fullerton, CA). The binding affinities were deter-mined using Ligand Competition Analysis Software by Lundon Software (Chagrin Falls, OH).
Transcriptional activation assay in ERtransfected HeLa cells. HeLa cells were used because of their estrogen responsiveness in the presence of ER and estrogens, allowing for detection of weak estrogens. In addition, they can be treated in serum-free media to ensure that the estrogen background is null. Testing with and without the ER proves that any response is through direct interactions with the ER. The mouse ER was used in these experiments for consistency among the three assays.
The estrogen responsive reporter vector, ERET81CAT, and the pRSV vector containing the mouse ER cDNA (without the neomycin resistance cassette) was constructed as previously described (9). HeLa cells, which do not contain ER, were cotransfected with both vectors or with only the ERET8 1 CAT vector to determine if observed activity was ER dependent. The cells were grown in DMEM/F12 medium (1:1) without phenol red (Sigma), supplemented with 5% fetal bovine serum and penicillin-streptomycin. Cells were electroporated and treated as previously described (6). During and after transfection, the cells were maintained in DMEM/F12 medium plus insulin-transferrin-sodium selenite (Sigma). All procedures using serum-free conditions were performed with Falcon plastics (Becton Dickinson, Franklin Lakes, Articles -Shelby et al.

Results and Discussion
Results of the competitive binding (Fig. 2), transcriptional activation (Fig. 3,4) (Fig. 5 O Kepone transfection assay (Fig. 3) and the receptor binding assay (Fig. 2), which do not employ an exogenous source of metabolic activation. These results show the inherent estrogenic activity of tamoxifen and demonstrate that this activity is enhanced by hydroxylation. Evidence of the in vivo estrogenic activity of methoxychlor and its requirement for metabolic conversion to HPTE, the active estrogen, is reviewed by Bulger and Kupfer (12,13). In contrast to tamoxifen, methoxychlor exhibits no intrinsic estrogenic activity, and conversion to HPTE is required for activity. No activity was seen with methoxychlor in either in vitro assay, whereas HPTE was clearly active in both assays (Fig. 2 of the test systems and to pro-Over the same dose range where effects e data against which to comwere seen with nonylphenol, o,p'-DDT Its of the other eight chemigave a greater response in the receptor bind-*oduced the effects expected, ing assay (Fig. 2). Although o,p'-DDT gave Lbstantially greater estrogenic a response in the transfection assay (Fig. 3), three assays than any of the it was less active than nonylphenol. o,p'iemicals.
DDT was uterotropic (Fig. 5), showing ,t cancer adjuvant therapeutic roughly the same level of effect in the same ifen and its metabolite, 4-dose range as HPTE and nonylphenol.
Articles * Shelby et al.
In the present studies, the reported estrogenic chemicals endosulfan (20) and kepone (13,(21)(22)(23)(24) were both negative in the receptor binding assay (Fig. 2) and the transcriptional activation assay (Fig. 3), indicating no activity in vitro. These kepone results are in contrast to reports of binding to the estrogen receptors of animals other than mice. Consistent with these negative results in vitro, endosulfan was negative in the uterotropic assay (Fig.  5). However, kepone gave a small doserelated increase in uterine weights over a dose range of 100-10,000 pg/kg (Fig. 5). This result agrees with earlier reports of similar effects in birds and rats (13) and mice (24). In the uterotropic assay, test doses for both endosulfan and kepone were limited to 10,000 pg/kg by toxicity. Further studies on the estrogenic potential of these two chemicals are underway.
As seen in Table 1, results were the same in the two in vitro assays for all 10 chemicals. Because these two assays measure different endpoints, they should not, at this time, be considered redundant. Testing of a larger, more diverse set of chemicals may reveal differences in responses. This is likely to occur with estrogen antagonists that bind the receptor but do not activate transcription. Further, the lack of an exogenous source of metabolic activation is a current limitation to the in vitro assays, for which there is only partial compensation by the use of the uterotropic assays in the three-test combination.
Finally, in reaching judgments on the biological significance or hazard associated with estrogenic activity identified in tests such as those reported here, it is important to keep in mind that the activity of some chemicals is observed only at levels of exposure that are orders of magnitude higher than those where effects are seen with estrogens such as estradiol and DES.
In summary, the combination of three assays employed in this study provides a rational and informative approach to assessing the estrogenicity of chemicals. Consistency of results among all three assays, as seen with the tamoxifens, HPTE, and o,p'-DDT, offers assurance that the chemical under study is truly estrogenic or, with consistent negative results as with endosulfan, lacks meaningful estrogenic activity. Inconsistent results may also be informative, providing clues as to mechanism of action of the test chemical, e.g., binding to the ER and failure to elicit transcription or uterotropic responses would suggest that a chemical inactivates the receptor. Activity in both the receptor binding and transcriptional activation tests and negative results in the uterotropic assay would suggest inactivation of the chemical in vivo or failure to distribute to the uterus. Likewise, failure to bind to the ER or induce transcription while leading to uterotrophy, as was seen with methoxychlor, indicates the requirement for metabolic activation or a mechanism independent of the estrogen receptor.
This three-test combination offers a systematic and mechanistically informative approach to assessing estrogenicity. It provides a useful profile of activity using a reasonable amount of resources and is compatible with the study of individual chemicals as well as the investigation of interactions among combinations of chemicals. Such an approach is needed if the presence of estrogenic agents in the environment is to be determined as a first step toward assessing the health hazards they may present to humans and other forms of life.