Overview of a workshop on screening methods for detecting potential (anti-) estrogenic/androgenic chemicals in wildlife

Authors

  • Gerald Ankley,

    Corresponding author
    • The current address of G. Ankley is U.S. Environmental Proctection Agency, National Health and Environmental Effects Research Laboratory, 6201 Congdon Boulevard, Duluth, MN 55804–1136, USA.
    Search for more papers by this author
  • Ellen Mihaich,

  • Ralph Stahl,

  • Donald Tillitt,

  • Theo Colborn,

  • Suzzanne McMaster,

  • Ron Miller,

  • John Bantle,

  • Pamela Campbell,

  • Nancy Denslow,

  • Richard Dickerson,

  • Leroy Folmar,

  • Michael Fry,

  • John Giesy,

  • L. Earl Gray,

  • Patrick Guiney,

  • Thomas Hutchinson,

  • Sean Kennedy,

  • Vincent Kramer,

  • Gerald LeBlanc,

  • Monte Mayes,

  • Alison Nimrod,

  • Reynaldo Patino,

  • Richard Peterson,

  • Richard Purdy,

  • Robert Ringer,

  • Peter Thomas,

  • Les Touart,

  • Glen Van Der Kraak,

  • Tim Zacharewski


  • This paper has been subjected to a U.S. EPA technical review; however, the views expressed are those of the authors and do not reflect U.S. EPA policy, or that of any of the organizations mentioned herein.

Abstract

The U.S. Congress has passed legislation requiring the U.S. Environmental Protection Agency (U.S. EPA) to develop, validate, and implement screening tests for identifying potential endocrine-disrupting chemicals within 3 years. To aid in the identification of methods suitable for this purpose, the U.S. EPA, the Chemical Manufacturers Association, and the World Wildlife Fund sponsored several workshops, including the present one, which dealt with wildlife species. This workshop was convened with 30 international scientists representing multiple disciplines in March 1997 in Kansas City, Missouri, USA. Participants at the meeting identified methods in terms of their ability to indicate (anti-) estrogenic/androgenic effects, particularly in the context of developmental and reproductive processes. Data derived from structure-activity relationship models and in vitro test systems, although useful in certain contexts, cannot at present replace in vivo tests as the sole basis for screening. A consensus was reached that existing mammalian test methods (e.g., with rats or mice) generally are suitable as screens for assessing potential (anti-) estrogenic/ androgenic effects in mammalian wildlife. However, due to factors such as among-class variation in receptor structure and endocrine function, it is uncertain if these mammalian assays would be of broad utility as screens for other classes of vertebrate wildlife. Existing full and partial life-cycle tests with some avian and fish species could successfully identify chemicals causing endocrine disruption; however, these long-term tests are not suitable for routine screening. However, a number of short-term tests with species from these two classes exist that could serve as effective screening tools for chemicals inducing (anti-) estrogenic/androgenic effects. Existing methods suitable for identifying chemicals with these mechanisms of action in reptiles and amphibians are limited, but in the future, tests with species from these classes may prove highly effective as screens. In the case of invertebrate species, too little is known at present about the biological role of estrogens and androgens in reproduction and development to recommend specific assays.

BACKGROUND

Endocrine-disrupting chemicals (EDCs) have been variously defined as “exogenous agents that interfere with the production, release, transport, metabolism, binding, action, or elimination of the natural hormones in the body responsible for the maintenance of homeostasis and the regulation of developmental processes” [1] or “an exogenous substance that causes adverse health effects in an intact organism, or its progeny, secondary to changes in endocrine function” [2]. Given these broad definitions, it is clear that there are many mechanisms through which effects could occur.

Endocrine disruption (or modulation [3]) has been demonstrated in controlled laboratory studies with mammalian, fish, reptilian, amphibian, avian, and invertebrate models following exposure to a variety of anthropogenic and natural chemicals. Of greater concern, of course, are incidences of endocrine disruption that occur in the environment. The degree to which human populations are being affected by EDCs is currently debated [4,5], however, good evidence exists that potentially adverse effects occur in some wildlife at the individual and population levels, caused by disruption of one or more endocrine systems [6]. For example, wild male fish with measurable levels of vitellogenin, a female-specific protein, were observed in different riverine environments in the United Kingdom and United States [7,8]. Female mosquitofish in streams dominated by pulp mill effluents were found to possess male-specific gonadopodia [9,10], and other fish species (white sucker) exposed to pulp mill effluents exhibited delayed sexual maturation, reduced gonadal growth, and altered steroidogenic capacity [11,12]. A chemical spill of organochlorine pesticides in Lake Apopka, Florida, USA, was proposed as responsible for morphologic abnormalities and decreased reproduction in the alligator population of the lake [13]. Imposex (development of male genitalia by females) and decreased reproductive success by some marine gastropod species was linked to exposure to tributyltin [14,15]. It should be noted that, although there are seemingly clear examples of adverse effects of EDCs in individual organisms, the extent to which broad population- level impacts in wildlife are occurring is uncertain [16].

Conferences and workshops held over the past few years [17], as well as the recent publication of a book on the topic [18], were instrumental in focusing political and public attention on EDCs. In response to these concerns, the Office of Research and Development of the U.S. Environmental Protection Agency (U.S. EPA) has identified EDC research as high priority [19], and initiated a number of planning exercises to address the issue [1,16]. Similarly, the chemical industry has developed a number of proposals for endocrine-related research in wildlife that currently are under consideration (R. Stahl, R. Miller, unpublished data). In addition, the U.S. EPA recently released a review concerning the risk of EDCs to human and ecological health [20]. In 1996, the U.S. Congress passed two pieces of legislation, the Food Quality Protection Act (PL 104–170) and the Safe Drinking Water Act (PL 104–182), both of which issued a mandate to the U.S. EPA to initiate a testing program designed to identify chemicals as EDCs. The U.S. EPA was ordered to develop screening tests by August 1998 and to implement a screening program by August 1999 that would evaluate substances found in drinking water and all pesticides for potential endocrine-mediated effects. The legislation focused specifically on that category of EDCs that has been the most publicly recognized, estrogen mimics. However, in recognizing the potential harm of other classes of EDCs, the U.S. EPA has extended the definition to include other modes of action, including antiestrogenic, androgenic, and antiandrogenic effects. Also, although the legislation primarily focused on potential human health consequences, the U.S. EPA has included wildlife health as an endpoint of concern. In this context, wildlife includes mammals, fish, birds, reptiles, amphibians, and invertebrates.

The Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC) is a multistakeholder advisory group to the U.S. EPA that is charged with developing recommendations as to the course of action to take to comply with the congressional mandate. The EDSTAC comprises individuals representing various groups affected by the EDC issue, including government, academia, industry, and environmental advocacy groups. In part to assist the EDSTAC group in the identification and evaluation of screening tools, the U.S. EPA, the Chemical Manufacturers Association, and the World Wildlife Fund cosponsored a series of workshops. The first of these was in July 1996, in Durham, North Carolina, USA, and addressed screening methods designed to detect or predict human health effects associated with chemicals with the potential to exhibit (anti-) estrogenic/androgenic activity [21]. The meeting described in this manuscript was held in March 1997 in Kansas City, Missouri, USA, and focused on screening methods to detect (anti-) estrogens/androgens in wildlife species. The wildlife workshop was deemed necessary because of concerns that mammalian data could not be accurately extrapolated to other vertebrates (or invertebrates) because of differences in mammalian versus nonmammalian endocrine systems. Finally, a third workshop focused on the effects of EDCs on thyroid- mediated processes in both humans and wildlife was held in Durham in June 1997. The report from that meeting is pending.

The wildlife screening methods workshop was organized by a steering committee consisting of representatives from governmental agencies (G. Ankley, S. McMaster, D. Tillitt), industry (R. Miller, E. Mihaich, R. Stahl), and environmental advocacy groups (T. Colborn). Participants at the workshop represented multiple affiliations and disciplines. The scientists were specifically charged with proposing and considering various endpoints in wildlife that could be fashioned into suitable screening tools for the identification of EDCs, defined as those chemicals with the potential to exhibit (anti-) estrogenic/androgenic activity, particularly from the standpoint of reproductive and developmental effects. The workshop was initiated with a discussion of potential confounding factors in extrapolation among species, specifically from laboratory mammals such as rats and mice to vertebrate wildlife. This was done in the context of the general screening recommendations for mammals made by Gray et al. [21]. Following this were presentations and discussions focused on five groups of wildlife: mammals, fish, birds, reptiles/amphibians, and invertebrates. Included in the discussions was consideration of evidence of endocrine disruption in the environment, identification of unique or relevant aspects of endocrinology and biology of each group of animals, evaluation of current testing protocols and their suitability for detecting EDCs, and proposed modifications or additions to these tests in the context of screening. We also considered, separately, the role of in vitro assays and structure-activity relationship (SAR) models in the screening process.

GENERAL CONSIDERATIONS FOR SCREENING

An ideal screening tool should be relatively rapid and cost- effective so that many chemicals can be tested. For example, 70,000 existing industrial organic chemicals (which does not include pharmaceuticals and pesticides) and 1,500 to 2,000 new industrial chemicals are submitted yearly in the United States for evaluation under the Toxic Substances Control Act, and more than 100,000 chemicals are listed on the European Inventory of Existing Chemical Substances [22,23]. A screening design should not consist of a single endpoint, but rather a suite of endpoints reflective of the mechanisms of concern (i.e., [anti-] estrogenicity/androgenicity). The endpoints should be relatively easy to measure to foster development of standardized screening protocols and interlaboratory validation. Finally, screening tools should be appropriate to hazard identification, and not necessarily reflective of effects characterization or dose-response analysis. That is, positive results in a screen with a particular chemical would result in secondary testing aimed at better characterization of specific effects and doses necessary to elicit effects. In this context, screening tools should be biased in favor of minimizing the occurrence of false negatives.

The general consensus from the workshop was that in vivo tests, rather than in vitro tests or SAR models, currently are the most defensible screening tools for wildlife. This conclusion arose from uncertainties related to differences in xenobiotic metabolism, bioavailability, and toxicokinetics in in vivo versus in vitro systems. It should be stressed that as in vitro assays and SAR models evolve, they should be considered as routine wildlife screening tools; however, at present their greatest utility lies in the ability to delineate specific mechanisms of action (e.g., receptor binding). We feel that short-term in vivo screens would be most useful if a suite of endpoints could be assessed. Such tests generally would be integrative (i.e., not necessarily unequivocally indicative of a specific mechanism of action), but capable of assessing overall function of the endocrine system. A critical consideration contributing to this latter point is the limited knowledge of basic endocrinology and physiology for many classes of animal.

An important issue in screening tests is the dose utilized. Because of resource considerations, testing that utilizes a limited number of doses would be most desirable. However, several participants at the workshop felt that given the unpredictable nature of dose-response curves for some EDCs (e.g., the U-shaped dose-response relationship observed with estrogenic compounds in recent mammalian studies [24]), tests with a limited number of doses might not be appropriate. However, even with U-shaped dose-response curves, the highest doses seem not to be completely devoid of effects, and these effects would trigger a positive response in a screen. Further testing could detect any unusual dose-response relationships. Thus, a screening paradigm based on a limited number of doses seems reasonable; however, the selection of these doses, for example in terms of a maximum tolerated dose (MTD), remains problematic. That is, a very high dose of a test chemical could elicit overt toxicity that might obfuscate more subtle effects on specific aspects of endocrine function. In exploring the possible value of an MTD approach, consideration of recent progress in other areas of wildlife toxicology, such as the utilization of MTDs in designing in vivo screens for detecting chemically induced genetic damage, might be useful. This approach was first utilized in mammalian genetic toxicology [25], and recently has been applied successfully to genotoxicology testing from an ecological perspective [26].

An important consideration in the identification of a suite of screening tools is the degree to which extrapolation of (anti-) estrogenic/androgenic effects among species is possible. Many aspects of endocrine function are conserved among species [27,28], and this similarity serves as a potentially useful basis for among-species extrapolation [16,29]. This is particularly true from the standpoint of organization of the neuroendocrine regulatory system and the hypothalamic-pituitary axis, to the synthesis of hormones in the gonads and their action at target tissues in terms of binding to appropriate receptor(s), and subsequent interaction with DNA and induction of protein synthesis. Moreover, several of the biochemical pathways involved in synthesis and action of hormones are similar among vertebrates. Similarities also exist regarding second messengers, hormone structure, gene sequences, regulatory elements, and gene interactions. Although there are some differences at the level of hypothalamic control of the endocrine system (e.g., factors that govern the onset of breeding cycles), the most divergence among vertebrate classes in terms of endocrine function occurs downstream from the hormone receptor at the levels of physiological and cellular responses [30]. For example, in male birds primary differentiation of the testes is influenced by 17-β estradiol, whereas in mammals differentiation of testes is under control of testosterone [31]. As another example, androgens control certain aspects of secondary sexual characteristics in males of all vertebrates, but these effects in terms of morphologic and behavioral endpoints can be quite species- or class-specific, for example, antler/horn formation in ungulutes, coloration in fish and birds, calling (singing) in frogs and birds, bone formation in frogs, dewlap formation in lizards, and so on.

Species or class differences with respect to metabolic activation/deactivation of EDCs also exist. For example, fish generally have lower phase I and II xenobiotic-metabolizing activity than mammals or birds [32,33]. If the chemical of concern was, for example, deactivated by oxidative metabolism, it might not produce a response in a mammalian system (negative), but might be active (positive) in a fish model. Similarly, among-species differences in toxicokinetics of potential EDCs could lead to uncertainty in extrapolation.

An important uncertainty in among-species (across-class) extrapolation of potential (anti-) estrogenic/androgenic effects arises from a lack of knowledge concerning basic processes such as ligand-receptor binding and interactions of the ligand- receptor complex with DNA. Although evidence exists of considerable structural homology, for example, of the estrogen receptor (ER) across species, it is not clear that this homology is adequate as a basis for quantitative extrapolation in terms of relative affinity of the receptor for endogenous or exogenous ligands. The rainbow trout ER recently has been reported to differ structurally and functionally somewhat from mammalian ER(s) [34,35]. However, the degree to which these structural differences might be translated into variations in ligand affinity and subsequent in vivo effects is unclear. At least some ligands that do not bind to the mammalian ER (e.g., atrazine), recently have been reported to bind (albeit weakly) to reptilian (alligator) ER [36]. Again, whether this would translate into significant differences in in vivo responses between mammals and reptiles is unknown, but these examples serve to illustrate the nature of extrapolation uncertainty that could exist even at subcellular levels. This type of uncertainty, of course, is not restricted to receptor binding, but includes processes related to events ranging from endogenous ligand synthesis/degradation to hormone transport and release.

In summary, because mechanisms involved in the synthesis, release, and actions of a given hormone (in this case estrogens and androgens) are similar in target cells of most vertebrate wildlife, the uncertainty in extrapolations across species may be acceptable for screening-level assessments at the molecular level. An important caveat to this conclusion is that little is known about comparative aspects of basic processes, such as receptor binding, among species. Extrapolation of effects at the cellular, tissue, and whole organism levels is more problematic because of differences in endocrine function among species, and metabolic and toxicokinetic variations. These variables may result in instances where chemical potency differs significantly among species or a chemical is active in one species but not in another. Thus, at present, a screening paradigm for wildlife should include in vivo assays with one or more model species representative of the animal class(es) of concern. Furthermore, it would be desirable to have model species for the various animal classes that are as well characterized in terms of basic endocrinology and physiology as, for example, rats or mice.

TOPICAL SUMMARIES

In the following sections, we cover in some detail the discussions that ensued in the various topical areas considered at the workshop, including recommendations of specific assays and endpoints suitable for screening for (anti-) estrogenic/androgenic effects in both the near and long term. We have attempted to include as much relevant discussion (and pertinent references) as possible; however, given the extreme breadth of the subject, this coverage should not be considered to be a comprehensive literature review.

IN VITRO TESTS AND STRUCTURE-ACTIVITY RELATIONSHIPS

Types of in vitro tests

Several types of in vitro tests have been used to identify potential (anti-) estrogenic/androgenic compounds in mammals and other taxa. These include receptor binding, cell proliferation, gene expression, and inhibition or stimulation of steroid hormone synthesis (for reviews, see [21,37]). In addition, a number of SAR models have been developed for screening based, in large part, on data from in vitro assays (for review, see [38]).

Receptor binding affinity. A classic method of identification of possible (anti-) estrogens/androgens is measurement of relative binding affinity to the ER or androgen receptor (AR) using competitive ligand binding techniques. Increasing concentrations of the chemical in question are included in incubation mixtures consisting of radiolabeled ligand (e.g., 17-β estradiol or dihydroxytestosterone) and a cytosolic or nuclear preparation from a tissue containing ER or AR (e.g., in rats and mice, the uterus or prostate). The free radiolabeled ligand is separated from the bound ligand and expressed as the percentage displaced by the competing chemical. The tissue source for receptor preparations is usually a target tissue of the hormone of interest. In oviparous animals, ER competitive binding experiments have been performed using hepatic or oviduct tissues, whereas brain, skin, and testicular tissue have been used to study xenobiotic interaction with the AR.

The advantages of ligand binding assays in identifying ER and AR ligands include widespread acceptance and use not only in mammalian, but nonmammalian species, including fish, reptiles, and birds [36,39,40]. The assay does not distinguish between agonists and antagonists, which might be considered an advantage in a primary screen designed to generate the fewest number of false negatives. As discussed above, a disadvantage associated with competitive ligand binding assays is uncertainty of extrapolation among species because of possible differences in receptor structure. This can be addressed to some extent by performing binding assays with tissues from a wide range of species. However, certain receptors may be difficult to obtain for testing from species of potential concern, for example, the bald eagle. A highly desirable research avenue, therefore, would be a detailed among-species comparison of receptor binding data to allow consideration in limiting the number of species for which such a screen would be used. For example, if it was established that ER and AR binding affinity did not vary widely among fish species, then a screen with only one model species could be employed.

Note that, for receptor binding assays to serve as an effective primary screen, a standard operating procedure would have to be accepted; this could be a difficult task given species- and situation-dependent diversity in tissue expression levels of receptors. For example, in fish ER and AR levels in various tissues are known to vary widely throughout a seasonal cycle [41–44] and ER and AR binding affinities can also vary seasonally in a variety of species [45,46]. In the future, the use of recombinant receptors from different species may eliminate some of these confounding factors, as well as provide a supply of receptor(s) without animal sacrifice [37].

Cellular proliferation. With the promotion of the E-Screen assay, cellular proliferation methods for detection of EDCs have generated interest. These assays measure proliferation in cell lines that are dependent upon hormones for stimulation of growth. For example, the E-Screen is based on the human breast cancer cell line MCF-7, which requires estrogens to proliferate [47]. Hence, estrogenic xenobiotics cause the cells to proliferate, whereas antiestrogens inhibit the proliferation response to 17-β estradiol. Although this assay is sensitive and can distinguish between ER agonists and antagonists, there are disadvantages. Due to the sensitivity of the assay to culture conditions (e.g., from one lot of serum to the next and between different batches of cells), there is interlaboratory variability in results [37]. In addition, cell lines responsive to other hormones (e.g., androgens) are not currently available. Because of these limitations, cell proliferation assays are not, at present, acceptable screens for wildlife.

Gene expression. Gene expression assays measure the induction of gene transcription following hormone receptor activation [37]. One technique used to assess gene expression is measurement of the mRNA of an endogenous gene product stimulated by estrogens or androgens. Reporter gene assays in eukaryotic cell lines or yeast also can be used to assess gene expression. In the reporter gene assay, the response element induces not the transcription of the endogenous product, but instead mRNA for an enzyme (the “reporter”). The endogenous machinery in the cell or yeast translates this product into the enzyme, such as luciferase, chloramphenicol acetyltransferase, β-galactosidase, or alkaline phosphatase. When substrate is added, the enzyme catalyzes a light-emitting, radioactive, or colorimetric reaction that indicates the amount of gene expression. Reporter gene assays constructed in a variety of systems including MCF-7, MVLN (an MCF-7 derivative), and yeast cells have been used by investigators to detect estrogenic activity of single chemicals and complex mixtures [21,48,49].

Gene expression assays in eukaryotic cell lines have an advantage over receptor binding assays of being able to distinguish between agonists and antagonists. Because of widespread use in the pharmaceutical industry for drug screening, protocols are relatively standardized and easily adapted to automated microtiter plate formats allowing large numbers of samples to be processed quickly.

Disadvantages of gene expression assays include the specialized equipment and training required for both mRNA detection and reporter gene endpoints. Also, most of the reporter gene assays are based on transient transfection for inserting plasmids; this is a labor-intensive procedure and introduces interassay variation into results. Although the lack of background endogenous hormones and receptors in yeast limits some confounding variables, results between yeast and mammalian cells are sometimes poorly correlated, which may in part be due to transport differences of xenobiotics across cell walls versus cell membranes, and to differences in receptor populations [37]. Finally, little work has been done with these types of assays in nonmammalian species, thus limiting their utility as screening tools in wildlife.

Steroidogenic stimulation/inhibition. Gray et al. [21] recommended the use of ex vivo assays to detect interference of xenobiotics with the oxidative synthetic enzymes of the steroidogenic pathway. Based upon their recommendations, laboratory animals would be exposed in vivo and excised gonads would then be cultured in vitro to assess relevant enzyme activities. Following addition of radiolabeled precursors to the incubation media, intermediates or final products in pathways of concern would be detected using simple methods such as thin-layer chromatography or radioimmunoassay (RIA). The assay could be simplified further through the use of in vitro contaminant exposures; however, this eliminates potential metabolic transformation that might occur in vivo. The advantages of these assays include their reproducibility and simplicity. They also enable investigation of both female (ovary) and male (testes) function.

These types of ex vivo and in vitro steroidogenesis assays have been used with nonmammalian species to identify the presence of potential EDCs. For example, McMaster et al. [50] reported a decline in the steroidogenic capacity of ovarian follicles from fish collected near pulp and paper mills. However, these techniques have received little or no attention with other animal classes, so uncertainty exists as to their current utility as screening tools.

Structure-activity relationships. Structure-activity relationship models offer the potential for rapidly screening large numbers of chemicals for potential biological activity associated with specific mechanisms of action. From the perspective of (anti-) estrogenic/androgenic effects, existing models are based largely upon data derived from in vitro assays, most commonly receptor binding assays [38]. As such, SAR models have the same basic strengths and limitations as in vitro assays. For example, models are based only on a limited number of mechanisms of action, which almost invariably cannot reflect all potential mechanisms of concern in vivo. However, a positive result for a given mechanism of action obtained either through SAR modeling or in vitro assays could serve as a reasonable basis for further in vivo testing. From the standpoint of screening for (anti-) estrogenic/androgenic effects, readily applied models for ligand binding to the ER and AR exist [51–55].

Advantages and disadvantages

In vitro models represent ideal test systems for studying the actual mechanisms of endocrine-modulating activity. Interactions of interest can be studied independent of the many possible elements that can confound the ability to identify actual mechanisms of action. In addition, in vitro models can facilitate obtaining reproducible empirical data in a time- and cost-effective manner. However, in vitro models have limitations as predictive tools in risk assessment. The models fail to account for the complexity of the whole animal and several important mechanisms inherent to in vivo systems. Cellular and cell-free systems lack the intact signaling pathway with regard to intercellular interactions and endocrine homeostasis. In addition, most in vitro models lack the ability to metabolically alter chemicals. As a result, extrapolation from in vitro to in vivo systems can lead to false negatives for compounds that are bioactivated, and overestimates of potency for compounds readily degraded in vivo. In vitro models also represent a totally different system with regard to toxicokinetics than in vivo systems, making estimates of potency difficult. Because current SAR models are based in large part upon data from in vitro systems, these models also are prone to the these same types of shortcomings.

Role of in vitro tests in screening

In summary, although in vitro tests and SAR models have some utility, particularly in terms of identifying specific mechanisms, such approaches cannot at present be the sole basis of a primary screening paradigm. We suggest, however, that in vitro tests ultimately could serve as one of three components of a screening effort consisting of short-term in vivo assays, an in vitro assay (or battery of assays), and SARs. Indications of (anti-) estrogenic/androgenic activity in both in vitro and SAR screens would be sufficient to identify a chemical for further testing, as would any activity detected in vivo. A further use of in vitro assays and/or SAR models might be for the prioritization of chemicals for more costly in vivo testing.

MAMMALIAN TESTS

Background

Current methods for assessing endocrine disruption in laboratory mammals (rat, mouse) are fairly well developed. The challenge, therefore, is to decide whether data from these models can be successfully extrapolated to mammalian wildlife. From a risk assessment perspective, toxicity data from studies with these laboratory mammals are used in decisions regarding human, and thus individual health, whereas risk assessment decisions regarding mammalian wildlife are concerned more with population-level effects [29]. However, from the standpoint of screening/hazard identification, this distinction is of less concern than for a full risk assessment.

Evidence exists for effects of EDCs on mammalian wildlife populations [6]. The most attention, by far, has been on carnivores, such as the Florida panther [56], river otter [57], and mink, particularly in the Great Lakes region [58–61]. This latter example is probably the best documented linkage of chemical exposure to impacts at the individual and population levels; however, the exact mechanism through which the putative chemicals of concern (polychlorinated biphenyls [PCBs]) are exerting effects is uncertain [6]. Although the issue has received little attention to date, possible effects of phytoestrogens on herbivorous wildlife also might be of concern in terms of (anti-) estrogenic/androgenic effects [62].

Aspects of endocrinology and biology

Although a high degree of similarity exists in endocrine function across mammalian species, there are some important differences. For example, relative steroid levels can differ markedly among mammalian species [63–65]. Confounding this is the fact that basic information is generally lacking on the comparative endocrinology and physiology of wildlife species of potential concern. Also, reproductive strategies can differ markedly; for example, a number of species including mustelids, bears, and bats have delayed implantation, which might influence how the effects of (anti-) estrogens/androgens are expressed.

Existing tests with mammals

Strengths and weaknesses of protocols with potential utility for identifying (anti-) estrogenic/androgenic compounds in tests with laboratory mammals were reviewed by Gray et al. [21]. The assays recommended by Gray et al. [21] as primary screens are in vivo, and based on early life-stage exposure followed by assessment of morphologic abnormalities, such as anogenital distance and nipple retention. Such assays have standardized protocols and have been validated with known EDCs. Gray et al. [21] also suggested that receptor binding, transcriptional, and ex vivo steroidogenesis assays have potential utility as screening tests. These latter assays could prove useful both for screening and elucidating specific mechanisms of action.

Virtually no existing tests with mammalian wildlife are suitable for screening for (anti-) estrogenic/androgenic effects. Protocols for partial life-cycle tests with mink likely would identify developmental and reproductive effects of these types of toxicants (e.g., [66]); however, these tests are relatively costly and can only be implemented by a small number of specialist laboratories.

Conclusions

Current methods for assessing endocrine-mediated effects in laboratory rodents (rat, mouse) are well developed, and enjoy wide application in many toxicology laboratories worldwide. For this reason, and given the fact that endocrine function is similar among mammals, we believe that the results of laboratory rodent screening studies for endocrine disruption can be extrapolated qualitatively, perhaps not as readily quantitatively, to the majority of mammalian wildlife species. Implicit in the acceptance of this extrapolation from laboratory rodents to wild mammals is the fact that uncertainties do exist and that further studies, perhaps in indigenous species, may be necessary to more clearly understand potential effects in wild populations, particularly from the standpoint of dose- response and effects characterizations.

FISH TESTS

Background

Teleosts (bony fish) are a very diverse taxon, comprising more than 20,000 species. This number does not include cartilaginous fish, such as sharks. The biology of only a small percentage of fish has been studied to any significant degree. Yet, more is known about the endocrinology and physiology of teleost fish than any other class of vertebrate wildlife. The endocrinology of fish is similar on the molecular level to that of mammals, but diverges significantly at higher levels of biological organization, making extrapolation from mammalian systems difficult. In addition, within the teleost taxon, many different reproductive strategies and several distinct factors influence sexual differentiation.

Various reports have been made of effects of EDCs in field populations of fish [6]. In one of the better publicized examples, Sumpter and coworkers reported that exposure of freshwater fish to certain municipal effluents in the United Kingdom could result in elevated levels of plasma vitellogenin (egg yolk protein), a response in oviparous organisms that is relatively specific to ER agonists [7,67]. Vitellogenin induction in fish exposed to effluents also has been recently reported in freshwater species in the United States [8,68] and marine species in Europe [69,70]. Bortone et al. [9] reported the masculinization of female mosquitofish (development of a gonopodium) exposed to pulp mill effluents. A Canadian research team also reported significant alterations in a number of endocrinologic parameters (e.g., gonadotropin and sex steroid concentrations) in fish exposed to pulp mill effluents [11,71,72].

Aspects of endocrinology and biology

In teleosts, as in other oviparous vertebrates, the reproductive cycles of the male and female are complex and highly coordinated, and rely upon the integrated activities of the pituitary, controlled by the hypothalamus and higher brain centers, and the gonads [73,74]. In common with other vertebrates, reproduction is regulated by gonadal steroids, under the direct control of two pituitary gonadotropins, GtH I and GtH II. These gonadotropins are analogous to the mammalian gonadotropins, follicle-stimulating hormone and luteinizing hormone [75]. These gonadotropins stimulate steroidogenesis at the gonads and maturation of the oocytes. Estradiol, synthesized primarily in the ovary, stimulates hepatic synthesis of vitellogenin. The female gonad also produces progestogens, which serve to induce final oocyte maturation. Membrane receptors for these steroids are found in the eggs, as well as receptors for vitellogenin. Apparently, no analogous receptors occur in mammals and, thus, these receptors may be unique sites of endocrine disruption in lower vertebrates. A nuclear progestogen receptor that regulates ovulation also has been identified in teleost ovaries [76]. This receptor has different steroid specificity than progesterone receptors in tetrapods, and has little or no affinity for a variety of organochlorines that bind to these receptors in birds and mammals [77]. Male fish also can differ in steroid profiles from mammals, with 11-ketotestosterone being a more potent androgen than testosterone in some species [78–81].

Several developmental strategies exist among fishes, often classified by the degree of nutrition provided by the female fish to the developing embryos. These include oviparity, viviparity, and ovoviparity, each of which could be differentially affected by xenobiotics. Also, several different patterns of sexual differentiation occur in fish. Some fish are naturally hermaphroditic; others are plastic, reversing sex in response to environmental parameters such as temperature, populational sex ratio, or age [82,83]. Although mechanisms of sex differentiation in fish are poorly understood, it is known that steroids can be influential in the sexual development of fish [84]. For example, androgen treatment at early critical lifestages usually induces offspring that are skewed to male, whereas estrogens produce predominantly female animals. However, notable exceptions to this occur, including paradoxical sex reversals where estrogens produce skewed sex ratios toward males, as well as intersex, in fathead minnows (R. Patino, unpublished data), and where nonaromatizable androgens produce monosex and functional females in channel catfish [85,86]. In addition, treatment of some species with sex steroids can result in hermaphroditic fish, with one ovary and one testis or with a gonad that has both ovarian and testicular tissue [87,88].

Sexual dimorphism in fish can be affected by exposure to estrogenic or androgenic substances. For example, males of certain mosquitofish species, which fertilize eggs internally, possess an altered anal fin (gonopodium). In these species, treatment of females with androgens results in alterations in the anal fin to a gonopodiumlike structure [89]. Hormonally influenced sexual dimorphism also occurs in fathead minnows, where sexually active males develop breeding tubercles, a process that can be diminished by treatment with 17-β estradiol [90]. Finally, an interesting and potentially useful (from the standpoint of testing for [anti-] estrogenic/androgenic effects) mutation in medaka has resulted in a strain that exhibits coloration differences reflective of genotypic sex [87,91].

Existing tests with fish

Both the Organization of Economic Cooperation and Development (OECD) and the U.S. EPA require in vivo tests with teleost fish for various regulatory programs, including short-term lethality, early life-stage, and partial and full life- cycle studies [92–98]. Commonly tested species include carp, guppy, rainbow trout, sheepshead minnow, fathead minnow, medaka, and zebrafish. Primary endpoints used in standard partial and life-cycle tests include mortality, behavior, growth, development, and reproduction [99–101]. Although several of these endpoints could reflect (anti-) estrogenic/androgenic effects, most are relatively nonspecific in terms of mechanism of action. Moreover, these long-term tests generally would be too resource-intensive to be considered as screening assays.

Potential assays and endpoints for detection of EDCs

Vitellogenin. Vitellogenin induction has received much attention as an endpoint potentially indicative of exposure to ER agonists. The endpoint is most sensitive in male fish where vitellogenin production is an abnormal process. Evidence exists that inappropriate vitellogenesis in males can result in kidney damage, and vittellogenesis has been proposed to divert metabolic resources from growth and spermatogenesis [102]. The induction of vitellogenin in male rainbow trout following estrogen treatment has been correlated with decreased testicular growth, which is generally accepted as a negative reproductive consequence [103]. Vitellogenin induction in female fathead minnows exposed to ethynylestradiol also has been shown to be associated with suppression of egg production [104].

Induction of vitellogenin has been observed in a variety of fish species (both male and female) exposed via different routes to a number of putative ER agonists including nonylphenol, methoxychlor, and chlordecone [103,105,106]. The preferred methods for vitellogenin detection involve analyzing blood samples with antibodies to vitellogenin using routine procedures such as RIA, enzyme-linked immunoabsorbent assays (ELISAs) or western blotting. Vitellogenin mRNA has also been used as a probe for vitellogenin induction in rainbow trout; however, although very sensitive, this technique requires specialized training and equipment [107]. Less sensitive methods for detecting vitellogenin also exist, such as measuring changes in alkaline-labile protein-bound phosphorus [108,109]. Vitellogenin induction also can be used as an endpoint in vitro in fish tissue systems [110,111]. In in vitro systems, in particular, vitellogenin induction may possibly be used as an indicator of both agonists and antagonists of the ER; in the latter case, activity might possibly be inferred as a function of decreased responsiveness to 17-β estradiol.

Limitations in vitellogenin induction assays include procurement of sufficient blood from small fish such as the fathead minnow or medaka; however, in these species it may be possible to monitor liver cytosolic fractions [106]. Also, because vitellogenin structure differs among species, antibodies developed for one fish may or may not cross-react with another, thus requiring species-specific antibody development [112–114].

Plasma steroid concentrations. In contrast to vitellogenin induction, alterations in plasma sex steroid concentrations can result from several different mechanisms of action, including direct effects on steroidogenic enzymes or indirect modifications associated with altered feedback loops. For example, an estrogen antagonist could interfere with the feedback inhibition of steroidal synthesis, leading to increased steroidogenesis in the ovaries. Decreases in plasma androgens have been shown to accompany exposure to 17-β estradiol in several fish species, presumably due to feedback inhibition of androgen synthesis [8,103,115]. Plasma steroid concentrations in fish are usually measured using RIA or ELISA with monoclonal antibodies to mammalian 17-β estradiol and testosterone. In some teleosts, 11-ketotestosterone may be a more proximal measure of androgenic activity than testosterone [78–80]. However, antisera to this hormone can be difficult to obtain. Similarly, antisera for the maturation-inducing progestogen steroids in female teleosts are limited in availability.

As with vitellogenin, steroids are more difficult to measure in small species than in large species because of limited blood volume; however, steroid measurements of extracts of whole body or gonadal tissue may be useful as alternative indicators of steroidgenic activity in small fish. Another concern regarding the use of steroid concentrations as an endpoint is that fluctuations can occur through the course of the reproductive cycle requiring that, for comparative purposes across populations, fish be sampled at exactly the same point in the cycle. In addition, sampling should occur at the same time of day to avoid the diurnal rhythms reported for some fish species [116,117]. Although these types of cyclical variations have been assessed in some species, comparative data are not available for many species of concern.

Receptor binding. The interaction of xenobiotics with hepatic ER has been assessed in several fish species including goldfish [118], spotted seatrout [40], and rainbow trout [34,105,119] Although qualitative differences in ligand binding of mammalian versus fish ER appear to be limited, too few data exist to recommend that results from mammalian studies would suffice for fish. Considerable uncertainty also is found in the identification of chemicals that interact with fish AR because relatively little work has been conducted with this receptor. Finally, the extent to which EDC screening should include those receptors unique to fish (and perhaps other oviparous animals), such as the membrane-bound progestogen, vitellogenin, and gonadotropin receptors in the gonads, oocytes, and sperm, is unclear. For example, Ghosh and Thomas [120] found that the estrogenic chemicals methoxychlor and chlordecone, as well as the nonestrogenic compound naphthalene, can bind to oocyte receptors and thus inhibit maturation of eggs.

Alteration in secondary sexual characteristics. Alterations in secondary sexual characteristics can serve as endpoints indicative of endocrine disruption. For example, exposure to pulp mill effluent resulted in gonadopodium formation in female killifish [9,121], a response that has been simulated in laboratory studies with androgens [89]. In fathead minnows, size of breeding tubercles and fatpads on males diminishes with exposure to 17-β estradiol [90]. Endpoints such as gonadopodium alteration, size of breeding tubercles, and color changes are relatively easy to identify and can serve as complementary assay endpoints because animals are not euthanatized; however, these measures have some degree of subjectivity.

Alteration in sexual differentiation. The hormonal environment of young fish affects their gonadal and phenotypic development. For example, Gray and Metcalfe [88] found that the degree of intersex increased in medaka exposed at an early life stage to the ER agonist nonylphenol. Although sexual differentiation can serve as a potential indicator of ER or AR agonists, various factors make this endpoint uncertain. For example, the differentiation mechanism in different fish species is uncertain in that outcomes are not always predictable; this is the case when channel catfish exposed to androgens become females [85,86] or when fathead minnows exposed to estrogens become males (R. Patino, unpublished data). Moreover, environmental factors such as water temperature can affect sex determination and ratios in fish [122]. Thus, if such a test endpoint were routine, the model species would have to be well characterized with respect to the regulation of sexual differentiation. Although apparent (phenotypic) sex ratios are relatively easy to assess in some species, histopathology is needed to detect more subtle changes in gonad structure (e.g., hermaphrodism), thus potentially increasing the cost of testing.

Gonadosomatic index (GSI). The ratio of gonad weight to body weight is a reasonable screening endpoint in terms of indicating (anti-) estrogenic/androgenic chemicals because it is easily measured and applicable to both sexes of oviparous fish species. The GSI is highly integrative in that it is not necessarily specific for any particular mechanism of action, but can reflect, for example, circulating hormone levels. Changes in GSI also can be indicative of likely reproduction success and, hence, population-level effects. However, as is true for steroid levels, such an endpoint should only be compared in fish that are at the same stage of gametogenesis.

Steroidogenesis. As is the case for mammalian systems [21], different researchers have assessed gonadal steroidogenesis in fish with both in vitro and ex vivo protocols. For example, gonads of wild fish collected from rivers receiving paper pulp mills were shown to have diminished steroid-producing capacity [50,72]. Similar experiments also were performed with fish following laboratory exposures to chemicals [123]. In addition, the steroidogenic capacity of isolated ovarian follicles of fish was shown to be significantly altered when exposed to xenobitics in vitro [124].

Final gamete maturation. This endpoint involves collecting eggs and sperm from exposed fish and evaluating gamete quality and maturation at the completion of vitellogenic growth and spermatogenesis. Like the GSI, final gamete maturation is a broadly integrative endpoint that reflects a variety of mechanisms affected by EDCs and is generally applicable to all fish species. Gamete size, germinal vesicle breakdown (GVBD), ovulation, hydration, and lipid and oil droplet formation all are relatively easy endpoints to assess using basic light microscopy. Sperm motility using image analysis is more difficult to measure and requires a reasonable level of expertise. In some species, this assay may require administration of gonadotropin-releasing hormone; however, such a requirement would bypass possible disruption in the neuroendocrine axis or in behavior, both of which regulate release of gonadotropin- releasing hormone in fish.

Germinal vesicle breakdown. In order to determine if a xenobiotic is capable of inhibiting maturation of oocytes, an in vitro assessment of GVBD (the final maturation of the eggs) can be performed in a culture system that includes the chemical of interest, oocytes removed from their follicles, and maturational gonadotropin [120]. Chemicals that inhibit GVBD probably interact with the membrane steroid receptors for maturation-inducing hormones. Several compounds could be tested simultaneously using eggs from a single ovary. The endpoint clearly could be of significant environmental relevance; however, from a screening perspective, the GVBD assay is not widely used or standardized.

Hypothalamic-pituitary function. A variety of chemicals have been shown to disrupt reproductive endocrine function in teleosts by altering gonadotropin secretion [72,125], which may be accompanied by changes in hypothalamic neurotransmitter metabolism [126]. However, gonadotropin RIAs have been developed for relatively few species and are unsuitable for gonadotropin measurement in most teleosts. The lack of suitable gonadotropin antisera, therefore, prevents the measurement of gonadotropin secretion in routine screening assays.

Test species

Because of the great diversity of fish, choosing just one species as a model is difficult. For screening purposes inclusion of species representative of multiple orders and environments (e.g., salmonids, cyprinids, percids) may be desirable. Several considerations must be made in choosing model species for EDC screening. For example, goldfish, some salmonids (e.g., rainbow trout, brook trout), Fundulus, and perciforms (e.g., croaker, yellow perch) represent species for which a reasonable knowledge base of reproductive endocrinology exists. For other species, such as fathead minnow, sheepshead minnow, zebrafish, and medaka, a large amount of acute and chronic toxicity data exists, as well as a history of use in regulatory programs [127–133]. Finally, some species (e.g., zebrafish and medaka) are attractive models in terms of an emerging understanding of the basis of the genetics underlying development and reproduction [134,135]. In the final analysis, decisions concerning choice of species should include consideration of the specific goals of the screening process, as well as capabilities of the laboratories involved in testing.

Route of exposure

The basic purpose of EDC screening tests in fish and other wildlife is to establish the plausibility of a chemical acting through a particular mechanism, that is, in terms of (anti-) estrogenic/androgenic responses, but not necessarily prediction of subsequent effects at the population or community levels, or dose-response relationships. In the environment, the principal route of exposure of fish to a given EDC depends upon a variety of factors, including water chemistry, the physical and chemical characteristics of the compound, and the biology of the species in question. In general, more hydrophobic chemicals would be expected to be accumulated by fish primarily from food, whereas less hydrophobic chemicals would be accumulated by absorption from water at the gills [136,137]. For chemicals of intermediate hydrophobicity, both food and water can contribute significantly to their uptake.

Most tests with fish deliver toxicants via the water or diet, as environmentally relevant routes of exposure. The simplest types of water exposures are static and static-renewal tests. Flow-through tests provide more realistic data than static systems because all conditions can be kept more or less constant during the exposure. Although these traditional routes of waterborne exposure can be employed to study EDCs, they have limitations that, in some cases, might be overcome through the use of alternative routes of exposure. For example, the amount of an EDC taken up by an egg or adult fish cannot always reliably be predicted during a waterborne exposure. Water exposures therefore typically require analytical support to relate effects observed with actual accumulated body burdens. Other problems sometimes associated with waterborne exposures include the following: the length of exposure potentially required to reach steady state for highly lipophilic chemicals, lack of significant absorption from the water, differential rates of absorption for different water exposure concentrations, and possible volatilization or adsorption of the chemical from/to the exposure container.

Because the purpose of EDC screening tests in fish is to identify mechanism-based responses and not to establish risk assessment models, other routes of exposure could provide advantages over the more conventional waterborne approach. In screening for EDC effects in fish, exposure concentration may be more important than exposure duration, and the route of exposure may not be as important as the actual body burden accumulated during exposure. For example, relatively large doses of 17-β estradiol are required for successful induction of vitellogenin in immature or male teleosts. This is due, at least in part, to the rapid clearance of 17-β estradiol by most fish species. Hence, investigators have utilized a variety of alternative methods to deliver 17-β estradiol, including silastic implants, coca butter pellets, and intraperitoneal injections [138].

The use of an exposure technique such as microinjection of eggs or intraperitoneal injection of adults allows a known dose of EDC to be administered, and avoids some of the problems associated with water-based exposures. Injection techniques also might prove useful for screening EDC effects in fish because they are rapid (i.e., overcoming the length of time needed to reach steady state), utilize small amounts of test materials, and reduce the risk of worker exposure through the use of a closed system [138–140]. In addition, injection techniques can be modified to screen complex mixtures of potential EDCs through the use of the fixed-ratio method of isobole analysis [141]. Zabel et al. [142] recently used this approach to analyze toxicity interaction patterns between pairs of polychlorinated dibenzo-p-dioxins, furans, and biphenyls in rainbow trout embryos.

Potential disadvantages of injection exposure techniques include the difficulty in selecting sublethal dose levels based upon available data from existing study designs [96–98]. Injection and similar dosing methods may also generate problems when, for example, the test substance is noxious and causes a cessation of feeding and/or important behavioral effects. Injection of such substances directly into fish may cause changes in endocrine function as a secondary effect of other toxicity mechanisms. These considerations need to be borne in mind, according to the nature of the test substance and the objectives of the screening exercise.

Conclusions

Despite the relatively large amount of research with teleosts regarding endocrinology and reproductive toxicology, experience is lacking with specific assays for screening for (anti-) estrogenic/androgenic activity, including standardization of such assays and uncertainty of outcomes associated with them. We suggest, therefore, that short-term in vivo tests that consider a suite of integrative endpoints could be used as screening tools. The endpoints suggested, although not always mechanism-specific, reflect the mechanisms of concern in their representation of endocrine status as a whole.

Specifically, we suggest that a 2- to 3-week in vivo protocol be used in which one or more species of reproductively mature fish are exposed via the water or injection, following which several endpoints are assessed, including GSI (with gross pathology of gonad), secondary sexual characteristics, final gamete maturation (ovulation/spermiation), plasma sex steroid concentrations, and induction of plasma vitellogenin. These endpoints were selected on the basis of their broad applicability and ease of routine measurement. Although such protocols would ideally employ reproductively mature fish for the most comprehensive set of endpoints, other well-defined life stages may be of use for a subset of endocrine-mediated endpoints (e.g., vitellogenin induction in juvenile fish [97]). Depending upon availability of resources, more comprehensive (and costly) endpoints could be added in this screen, including gonadal histology, in vivo responsiveness to gonadotropin-releasing hormone challenge, ex vivo steroidogenesis, fertilization success, and sexual differentiation in (F1) offspring. Note that the addition of these endpoints to existing partial and full lifecycle tests with various fish species would greatly enhance sensitivity of the assays to (anti-) estrogens/androgens.

In fish, in vitro assays at present could be considered as secondary tests because of the requirement for a reasonably sophisticated level of expertise. But, because in vitro assays provide information about potential mechanisms of action, as such assays become more routinely performed and validated, they could be included as part of the initial screening process. For example, steroid receptor competitive binding assays could prove to be valuable components of screening approaches. Other in vitro tests that in the future could be suitable for screening include steroidogenesis, vitellogenin production by hepatocytes, and the GVBD assay.

AVIAN TESTS

Background

Birds were one of the first classes of wildlife to draw attention to the potential dangers of anthropogenic chemicals in the environment, specifically organochlorine pesticides such as dichlorodiphenyltrichloroethane (DDT) [143–145]. Exposure to these pesticides resulted in decreased populations of some species, which was due in part to eggshell thinning and decreased survival of offspring. Later studies by Fry and Toone [146] suggested that exposure of gull embryos to DDT and its metabolites resulted in histologic signs of feminization. In one of the best characterized examples of the effects of environmental contaminants on avian species, in the Great Lakes region embryo mortality and terata in colonial fish-eating water birds has been linked to PCB exposure [147]. Bans on organochlorines such as DDT and PCBs have decreased their concentrations in the environment; however, at more contaminated sites, for example in the Great Lakes, a relatively high prevalence of deformities persists in some avian species [6].

Aspects of endocrinology and biology

Molecular aspects of avian endocrinology and the hormones involved in sexual development and reproductive processes are similar to known pathways in mammals. The central feedback mechanisms that control hormone production and release, as well as many of the associated target tissues (liver, gonads, and accessory organs) that respond to factors released from the pituitary, parallel those observed in other vertebrates. Sexual differentiation and reproduction are controlled through similar estrogenic/androgenic mechanisms, including receptors that act as ligand-activated transcription factors that regulate specific genes. A unique aspect of genetic sex determination in birds, as compared to mammals, is that the male is the default sex and is homozygous, whereas females are the heterozygous sex. During embryogenesis in birds, 17-β estradiol is required for differentiation of the female reproductive system [148], and male embryos exposed to natural estrogens or estrogenic contaminants (e.g., o,p′-DDT, methoxychlor) became feminized [149–153].

Divergence of avian steroid endocrinology from that of mammals and other classes of vertebrates can occur to a significant extent at higher levels of organization. For example, behavioral aspects of reproduction, such as nest building, parental care patterns, and migration are very different from other vertebrate models. Additionally, these endocrine-controlled behaviors can be quite distinct even among various avian wildlife species. Parental care of the young and nesting behaviors are known to be affected by environmental contaminants [146,154], and plumage in birds is also under control of steroid hormones and may be susceptible to potential EDCs.

Another important difference among birds with respect to reproduction has to do with feeding strategies. Newborn chicks are categorized as either altricial or precocial. Precocial birds hatch at a later stage, have open eyes, and are covered with down. These chicks can thermoregulate, have advanced muscular activity, and can self-feed, as opposed to altricial birds, which are dependent on their parents. Thus, successful production of offspring by altricial birds does not stop at game- togenesis, fertilization, and hatching, but has a strong parental behavioral component that is controlled by neuroendocrine factors.

Birds are oviparous organisms, and like most fish, amphibians, and reptiles, females produce vitellogenin in response to estrogen. Vitellogenin receptors are present in avian tissues (i.e., oviducts) and the vitellogenic response to ER agonists occurs in birds. In addition, exposure of chickens to estrogenic compounds has been reported to cause eggshell thinning through developmental effects on the oviduct and eggshell gland [155], suggesting this might be a potentially sensitive endpoint for (anti-) estrogenic/androgenic effects of chemicals in avian species.

Existing tests with birds

For pesticide registration, avian reproductive toxicity tests usually are required by both the U.S. EPA and the OECD [156]. At present, these tests are conducted with bobwhite quail and mallard ducks, but the OECD also permits the use of Japanese quail. Tests with the former two species require more than 20 weeks for completion, including a time period for acclimatization, and a pre-reproductive 8-week exposure, followed by 10 weeks of egg collection (with continuing exposure). Both males and females are exposed to the chemicals via feeding, and reproduction is assessed using pairs of animals. The collected eggs are examined for viability and hatching success (in incubators), and survival is monitored for 14 d following hatch.

The strengths and weaknesses of these avian reproductive tests have been discussed in detail elsewhere [157,158]. In summary, since implementation in 1982, the tests have generated a large amount of toxicity data for bobwhite quail and mallard ducks. The tests are excellent for chemicals that have a propensity to bioaccumulate or have a relatively high persistence in the environment in that they simulate a chronic exposure. However, the protocol is not as appropriate for relatively short-lived pesticides, and the length of the test decreases statistical power due to a lack of synchrony in egg production across paired animals. Also, the artificial incubation used removes parental behavior as a component of reproduction. Finally, except for assessment of eggshell thickness, the tests as currently performed may not serve as an indicator of (anti-) estrogenic/androgenic chemicals, in part because the protocol does not comprehensively evaluate the F1 progeny or second generation reproductive effects.

In a proposed modification of the standard avian reproductive toxicity test, Japanese quail breeding pairs are used as their own controls by collecting eggs prior to and during the feeding exposure to the test substance [158]. Through the use of this species, the entire test is shortened significantly, lasting only 10 weeks. Endpoints can include shell thickness of laid eggs, sex ratios of the progeny, gonad weights, sex accessory gland weights, and oviduct weights of hatched chicks. Histopathology also could be performed on reproductive tracts. All of these endpoints would enhance detection of potential EDCs. Despite these improvements, however, even a 10-week test likely would be considered too lengthy or costly to be used as a screening tool.

Potential assays and endpoints for detection of EDCs

In vivo assays with chickens. Induction of oviduct weight. This measure is analogous to the mammalian estrogenic bioassay of uterine weight [21], because estrogen acts upon the oviduct of birds to increase water retention and growth, usually in a dose-dependent fashion. In the assay, immature females could be exposed either through food or intramuscular injection daily for at least 7 d. Following exposure, the entire genital tract, from infundibulum to the cloaca, is weighed and compared to that of a positive control. Increases in oviduct weight in domesticated birds have been observed following exposure to some pesticides [159], phytoestrogens [160], and avian growth stimulants [161]. By measuring the inhibition of estrogen-induced oviduct weight increase, progesteronic activity also can be assessed [162].

Comb development. Testosterone is the primary androgen produced and secreted by avian testes. Testosterone influences several secondary sexual characteristics including vocalizations, plumage and bill color, and comb size. An increase in comb weight has been used as a sensitive avian bioassay for androgenic activity [163], and has been used to detect natural and synthetic androgenic compounds [164,165]. By measuring the inhibition of androgen-mediated comb weight increase, antiandrogens also can be identified [166,167]. For this assay, 2- to 3-d-old male chicks are treated with test chemicals in a carrier solvent directly applied to the lateral surface of the comb. Following at least 7 d of daily treatments, chicks are sacrificed and the comb is removed and weighed and compared to the comb from a positive androgen control.

In ovo exposure/embryonic development. Because hormonally active chemicals have been shown to affect sexual differentiation and organization of the reproductive tract of birds [146,149–152,168], in ovo exposure to contaminants could be used as the basis for a primary screen for detecting (anti-) estrogenic/androgenic compounds. In ovo exposure could be achieved through maternal transfer (e.g., after feeding laying females), or by direct injection of the eggs. Injection studies can provide results similar to the those from traditional feeding studies [169], and have the advantage of being less costly in terms of resources and time. Compounds that are fairly recalcitrant to metabolic processes are particularly amenable to the egg injection approach. Additionally, species that cannot be propagated in the laboratory may be tested with egg injections if their eggs are collected from an uncontaminated colony and incubation parameters are known [170]. Endpoints measured in hatchlings from eggs exposed via maternal transfer or injection could include sex ratios, weights of gonads, sex accessory organs, and oviducts; and, if warranted by resources and previous data, gonadal histology.

Vitellogenin induction and receptor binding. As is true for other classes of oviparous vertebrates, induction of vitellogenin in birds is indicative of exposure to ER agonists. For example, vitellogenin induction was reported in male Japanese quail exposed to organochlorine pesticides and phytoestrogens [171]. Also, in vitro primary cultures of avian hepatocytes have been used to examine effects of anthropogenic chemicals [172,173], and, therefore, an in vitro assay for vitellogenin induction conceivably could be developed for birds, as has been done for fish.

Receptor binding assays with avian ER or AR also can be used to screen for EDCs. Both receptors have been reasonably well characterized in birds relative to basic structure, distribution, and function [174–179]. However, less work has been done evaluating interactions of xenobiotics with avian ER or AR; one example of a study of this type was reported by Eroschenko and Palmiter [39], who evaluated binding of chlorderone to avian oviduct ER.

Overall, at present, neither vitellogenin induction nor xenobiotic binding to hormone receptors have been explored to the same extent in birds as these endpoints have been tested in other vertebrate species; consequently, recommendation of these assays as screening tests is premature.

Aromatase-mediated responses. Aromatase, a member of the cytochrome P450 family of enzymes, catalyzes the conversion of androgens to estrogens. In birds, aromatase plays a critical role in determining sexual characteristics, feathering traits, and reproductive behavior. For example, the gonads of chicken embryos that are less than 7 d old have both male and female characteristics. After this period the gonads develop into either ovaries or testes depending on the genotype of the embryo. Elbrecht and Smith [180] discovered that if 5-d-old chicken embryos are treated with an inhibitor of aromatase, the birds with the female genotype developed male gonads and were capable of spermatogenesis. These birds also had the physical appearance of male birds in that they had the male comb, hackles, and wattle. These and additional studies that confirmed and extended the original findings [181] show that aromatase plays a key role in sexual development in birds. Aromatase also plays a role in the henny feathering traits of domestic chickens. The henny feathering trait in males is caused by a mutation that causes an increase in aromatase activity in the skin [182]. Finally, aromatase also plays an important role in the copulatory and singing behavior in birds [183]. Therefore, studies of the effects of EDCs on aromatase- mediated responses appear to be warranted to determine if screening methods could be developed for this mechanism of action.

Conclusions

Proposed changes by the U.S. EPA and the OECD to existing avian reproductive toxicity tests, including substitution of Japanese quail, would increase the probability of detecting (anti-) estrogenic/androgenic chemicals. The sensitivity of these tests would be enhanced by evaluation in progeny of endpoints such as sex ratios, weights of gonads and oviducts, and histology of gonads. Also, inclusion of an altricial species, such as the zebra finch, for testing purposes might prove critical to detecting EDCs that act at a neuroendocrine level [184]. However, even with these improvements, the length of the existing avian reproductive toxicity tests means they are practical for use as screening tools.

Short-term in vivo assays of potential use for screening include 7-d exposures of immature male and female chicks to chemicals of concern, followed by assessment of comb development and oviduct weight in the two sexes, respectively. These endpoints previously proved successful as bioassays for hormonally active chemicals. Such tests should be standardized and further validated with known (anti-) estrogens/androgens.

In addition to the short-term chicken assays, an in ovoexposure followed by observation of sex ratios and weights of gonads, sex accessory glands, and oviducts of hatchlings of one or more avian species could prove to be a useful screen for EDCs. For these assays, egg injection might be the optimal route of exposure in terms of cost effectiveness and simplicity Such a screen would take longer to conduct than the in vivo chicken tests described above because of incubation and growth time for the chicks.

Future directions of EDC screening in avian species should focus on whether data for other classes of vertebrates can be successfully extrapolated to birds. Also, comparison of in vitro data from hepatocyte vitellogenin induction and/or receptor binding studies to responses in vivo is needed. Due to uncertainties concerning all potential points of endocrine disruption in birds, even the assays recommended above might not be totally adequate. For example, alterations in behavior could be a more important mechanism of endocrine disruption in birds than in some other species. Therefore, a definitive evaluation may require testing to evaluate, for example, courtship, singing, or parenting behavior.

REPTILE AND AMPHIBIAN TESTS

Background

One of the most publicized examples of possible EDC effects on a wildlife population is that of reptiles of Lake Apopka, Florida, USA. In 1980, Lake Apopka was contaminated by high concentrations of DDT, dichlorodiphenyldichloroethylene (DDE), and other compounds after a chemical spill and, shortly thereafter, a documented decline occurred in the alligator population [13,185]. In recent years, a variety of endocrine-related effects have been reported in both alligators and turtles from Lake Apopka. The majority of the male alligators in Lake Apopka appear to have been demasculinized, and females have been found to have a plasma 17-β estradiol to testosterone ratio twice that of normal animals. Although these effects are well documented, the underlying mechanism of action and chemicals(s) responsible for them remain unresolved [6]. Numerous laboratory studies also have been conducted in which the effects of hormonally active chemicals resulted in alterations in endocrine status and sexual differentiation in various reptiles [186,187].

Little information is available from the field concerning possible EDC effects on amphibians. Amphibian populations appear to be declining in North America and worldwide [188], and many factors have been theorized as causative of these declines, including habitat loss, increases in ultraviolet radiation, and toxic chemicals [189]. However, amphibian testing has been extremely limited, so assessment of the relative importance, for example, of chemical stressors on individuals or populations, is difficult [6]. Not unexpectedly, however, amphibians are susceptible to adverse effects of EDCs in laboratory studies. For example, Ramsdell et al. [190] found that exposure of Xenopus laevis tadpoles to nonylphenol, a putative ER agonist, resulted in vitellogenin induction and skewed sex ratios.

Aspects of endocrinology and biology

Although some vertebrates exhibit genetic sex determination, in which gender is fixed at conception, certain reptiles (e.g., crocodiles, turtles), as well as some amphibians, exhibit sex determination based on the temperature of embryo incubation, which is termed temperature-dependent sex determination (TSD). For example, in alligators, constant incubation at 30°C results in formation of female phenotypes in all hatchlings, whereas incubation at 33°C results in males. In contrast, red-eared slider turtles produce all females at warm temperatures (e.g., 31°C), whereas cooler temperatures produce males. Exogenous steroid administration has been shown to override temperature determination in developing alligators and turtles [186,191–193]. In alligators, 17-β estradiol treatment produces females when administered to eggs incubated at male-producing temperatures, whereas the ER antagonist tamoxifen causes sex reversal in embryos at male-producing temperatures [194]. Similar types of effects on sexual differentiation have been observed in amphibians. For example, exposure of X. laevis larvae to 17-β estradiol during a critical developmental period can cause genetic males to a exhibit a female phenotype [195].

Although reptiles with TSD appear to be potentially susceptible to certain classes of EDCs, prediction of exact effects from species to species is difficult. However, the sensitivity of sexual differentiation of reptiles to potential EDCs may allow development of useful screening tools. Although the elucidation of the genetically determined sex pathway in reptiles and amphibians is an area of active research, further work is needed to help clarify and contrast this pathway with the TSD pathways.

Existing tests with reptiles and amphibians

No standardized tests with reptiles and no statutory requirements for testing either reptiles or amphibians exist. One amphibian test has been standardized and validated in several laboratories [196–198]. The frog embryo teratogenesis assay (FETAX) is a 96-h whole-embryo developmental toxicity screen that involves exposure of X. laevis eggs with subsequent examination of embryonic development to assess teratogenic potential of chemicals [199,200]. Specific test endpoints are survival, growth, and malformation of embryos. Xenopus laevis is a good model because it is commercially available, easy to maintain, breeds year round, eggs are plentiful and transparent, and an extensive database exists on various aspects of its biology, including an emerging literature on molecular aspects of development.

Potential assays and endpoints for detection of EDCs

Life-cycle tests with alligators and turtles, which include the species most studied with regard to EDCs, are not practical because of the length of time it takes these animals to reach sexual maturity. In addition, sufficient numbers of suitable eggs can be difficult to obtain because they are collected from the wild, and are often contaminated with persistent organochlorines such as PCBs and DDT. Lizards, as well as frogs, mature faster than alligators and turtles and are easier to breed in the laboratory, thus providing a constant supply of contaminant-free eggs.

Modified FETAX assay for reproduction. The FETAX assay could be adapted to include exposure scenarios and endpoints appropriate for detection of (anti-) estrogens/androgens. Exposure protocols could include injection, feeding, or waterborne exposure of both males and females prior to mating. Endpoints in parents could include breeding behavior, fertilization rate, GSI, number or weight of released gametes, and stage of eggs in ovaries. In addition, more resource-intensive tests could examine parameters of sperm motility, velocity, and abnormality. In the offspring, sex ratios could be measured and vitellogenin production assessed. Antibodies are available that are specific for X. laevis vitellogenin [201]. However, a modified FETAX assay might not be suitable for screening because of the complexity and length of the test. Instead, this assay might be useful for secondary testing to confirm results from primary screen(s) and to examine dose-response relationships.

In ovo exposure/embryonic development. In ovo contaminant exposure can be acheived through maternal transfer or via injection/painting, but the latter routes might be advantageous because they eliminate the need to dose adults, thus reducing resource requirements for testing. Because of the sensitivity of sexual differentiation and gonadal development to endocrine disruption, as demonstrated both through egg painting and injection studies with reptile embryos [187,202], a protocol that exposes eggs and assesses sexual development could be a fairly straightforward in vivo screen for (anti-) estrogens/androgens. Amphibians could be assayed in a similar manner, because injection of X. laevis or other amphibian eggs is routine. Sensitive endpoints in the hatchlings would include embryonic mortality, hatching success, sex ratio, GSI, steroidogenic enzyme activities, and, if resources allow, gonadal histopathology. Plasma steroid concentrations are not recommended as an endpoint in newly hatched individuals because little difference in 17-β estradiol/testosterone ratios apparently occurs, even in sex-reversed hatchlings (L. Guillette, University of Florida, personal communication).

Receptor binding. Some basic research has been conducted characterizing reptilian hormone receptors [42,203,204], as well as a limited amount of investigation of the interaction of xenobiotics with these receptors. For example, Vonier et al. [36] demonstrated the affinity of known or suspected EDCs for the ER and progesterone receptor isolated from oviduct tissue of the alligator. Crews and colleagues have placed genes for the ER, AR, and progesterone receptor of whiptail lizards into expression vectors, creating in vitro receptor production [205,206]. A cell line augmented with these expression vectors would negate the need for animal sacrifice and greatly facilitate receptor binding assays. The ERs and ARs in X. laevis also have been cloned [207–210], and the same binding assays used in other vertebrates could be performed for amphibians. As with other classes, we recommend that comparative binding data be collected for reptiles and amphibians, and then used to decide whether results from a single species could be extrapolated to the rest of the class(es) and/ or whether binding affinities are similar enough to utilize data from other vertebrate classes.

Germinal vesicle breakdown. The hormonal events controlling the onset of meiosis have been well studied in X. laevis [211,212]. Xenobiotics that affect the breakdown of the oocyte nucleus (germinal vesicle) can easily be assayed either in vivo or even in vitro. These tests would be of short duration, and the results would be interpretable in the context of a large body of literature in this model system. At the present, however, no attempt has been made to standardize or validate a GVBD assay for amphibians.

Sex reversal in tadpoles. Sex reversal has been accomplished in X. laevis with the administration of 17-β estradiol to genetic male larvae at stages 51 through 54 [195]. Ramsdell et al. [190] were able to demonstrate this same response with exposure of X. laevis to nonylphenol. Several laboratories are presently working to develop this assay as a standard technique, and the test could be added to a suite of screening methods. However, at present, such tests are not ready for adoption into a screening program because they lack sufficient validation with known EDCs, and the mechanism of sexual differentiation in amphibians is not completely understood.

Vitellogenin. Palmer and Palmer [201] reported vitellogenin induction in adult male red-eared slider turtles by estrogenic compounds and environmental contaminants, including 17-β estradiol, diethylstilbestrol, and DDT. Males of other reptile or amphibian species also should respond to estrogenic compounds via vitellogenin induction. Typically, the animals are injected on consecutive days with test chemicals of concern and a positive control, blood samples are collected, and vitellogenin is measured with immunodetection techniques such as ELISA, RIA, or western blotting. However, little information exists at present concerning cross-reactivity of existing reptilian or amphibian antibodies for vitellogenin. Induction of vitellogenin in X. laevis also has been assessed by detection of mRNA for the protein [213]. In addition to in vivo monitoring, vitellogenin induction screens could be adapted to in vitro systems using established protocols, for example, based on teleost fish methods, but with primary cultures of reptilian or amphibian hepatocytes.

Conclusions

In terms of screening for chemicals with (anti-) estrogenic/ androgenic properties, at present information on the endocrine system of reptiles and amphibians is insufficient to make direct comparisons to other vertebrate classes. Thus, how predictive screens and tests from other species will be for these two animal classes is unclear.

At present, the currently available in vitro test methods with reptiles and amphibians do not appear to be adequate to predict potential (anti-) estrogenic/androgenic effects of xenobiotics. Thus, short-term in vivo methods should be used until sufficiently robust and validated in vitro methods are developed. The FETAX developmental assay has been the most commonly utilized of all of the discussed assays with reptiles and amphibians; however, it would need to be modified to include reproductive endpoints to be used as an effective screen for (anti-) estrogens/androgens. The most commonly used in vivo reptile models have been embryonic turtles or alligators exposed before sex determination, followed by incubation at male- or female-producing temperatures, prior to evaluating sex ratios and gonadal histopathology. This type of testing clearly is sensitive to EDCs, but the assays are lengthy and the availability of large numbers of uncontaminated eggs for commonly tested species is limited. The red-eared slider turtle may be the model of choice for this type of testing due to its availability, long breeding season, and potential for breeding in captivity. Several of the other assays listed above (e.g., sex reversal) also show good promise as screens for reptiles and amphibians, but further characterization would be required before routine implementation would be possible.

INVERTEBRATE TESTS

Background

Invertebrates, which constitute more than 95% of all animals, cannot be omitted from discussion of endocrine disruption in wildlife because of their importance to both ecosystem and human health. Invertebrates serve as pollinators and pest controllers, facilitate the trophic transfer of nutrients, and serve as an integral food source. Not only is there great biodiversity within invertebrate classes, but these organisms are ubiquitous, making them likely candidates for sentinels of environmental stress. Another reason to consider invertebrates as targets of endocrine disruption is that both field and laboratory data indicate EDC effects in these animals. For example, tributyltin and testosterone have been shown to induce an intersex condition in some invertebrates, called imposex, in which females develop male sex organs [214]. This condition has been observed in gastropods exposed to antifouling agents in the environment and may be responsible for decreases in populations [14]. Pseudohermaphroditic conditions also have been observed in copepods in contaminated environments [215]. The mechanism of imposex induction has been hypothesized to be metabolic androgenization, that is, elevated levels of endogenous androgenic hormones [214,216,217].

From the standpoint of the Kansas City workshop, however, conscious effort was made to limit discussions concerning invertebrate testing for two reasons. First, uncertainty exists concerning the actual role of androgens and estrogens in invertebrate development and reproduction [218]. Second, given the huge diversity of invertebrate species, we did not feel that this important area could be adequately assessed in the limited time available for the workshop.

Aspects of endocrinology and biology

Although their role is somewhat uncertain, androgenic and estrogenic hormones have been found in every invertebrate class examined thus far [219], and invertebrates are capable of generating hydoxylated, glucosylated, sulfated, and reduced/dehyrogenated metabolites of testosterone [220,221]. In some instances these compounds seem to have effects analogous to those in vertebrates; for example, in certain species androgens control male secondary sex characteristics, and in echinoderms and gastropods, estrogens increase oocyte growth (for review, see [219]).

Invertebrates also have systems uniquely different from vertebrates that could be greatly affected by environmental EDCs. For example, ecdysones found in arthropods, nematodes, and molluses mediate differentiation, growth, reproduction, vitellogenesis, and molting [222,223]. Nonsteroidal farnesyl hormones including juvenile hormone, farnesyl acetone, farnesonic acid, and methyl farnesoate are used by invertebrates to regulate embryogenesis, development, and reproduction [224,225]. The fact that several of these critical hormones are intentionally mimicked by pesticides is important from an environmental perspective [225].

Existing tests with invertebrates

Short-term toxicity tests have been conducted with a variety of aquatic invertebrates, but with a few exceptions [226], the most common chronic reproduction tests used for regulatory purposes are with parthenogenic stages of various daphnid species [92–94,227]. This, of course, limits the ability to detect compounds that might affect sexual reproduction [228].

Potential assays and endpoints for detecting EDCs

Several endpoints are potentially useful for detecting (anti-) estrogenic/androgenic effects in invertebrates, including induction of imposex and alteration of steroid metabolism. However, in general, little is known about the specificity of these endpoints relative to estrogens and androgens in invertebrates. For example, whether other environmental factors (e.g., temperature, nutrition) induce imposex is unknown. Of the limited number of chemicals examined, an excellent correlation exists between chemicals that adversely affect reproduction in daphnid chronic tests and those observed to interfere with steroid metabolism in daphnids [219]. Thus, alteration of steroid metabolism may be a general response useful for screening EDCs.

Conclusions

Direct and indirect evidence suggests that invertebrates are sensitive to chemicals in the environment that disrupt endocrine systems; however, from the perspective of screening for (anti-) estrogenic/androgenic effects, enough uncertainty exists concerning the role of these systems in invertebrates that it is premature to recommend specific tests/endpoints. However, from the standpoint of existing tests, we suggest that sexually reproductive stages be considered in the chronic reproductive tests to increase sensitivity to EDCs that cause reproductive and developmental toxicity. Induction of imposex and steroid metabolism in different species might have potential as markers of (anti-) estrogenic/androgenic effects, but uncertainties exist regarding the causes and biological significance of these endpoints. From the broader perspective of environmental EDCs, disruption of systems controlled by ecdysteroids and farnesyl hormones in invertebrates should be examined. Given the uniqueness of the invertebrate endocrine system, extrapolation to and from other classes most likely is not possible. Because of these points, the consensus of participants at the meeting was that a separate workshop is needed focused specifically on research needs and methods for assessing the ecological risk of EDCs to invertebrates.

UNCERTAINTIES AND RESEARCH NEEDS

Clearly, identification of specific screening assays and associated test conditions for detecting EDCs in wildlife species is limited by a lack of information in several areas. For example, knowledge concerning similarities/differences among vertebrate species/classes is sufficiently uncertain that it is difficult to gauge the degree to which extrapolation should serve as an underpinning for screening for (anti-) estrogenic/ androgenic effects. Ideally, a sensitive assay/endpoint in either a mammalian or nonmammalian system would be reflective of potential mechanism-specific activity across all vertebrates. For example, if ER agonists could reliably and consistently be detected by vitellogenin induction in an oviparous species, this might obviate the need, for example, of using mammalian screens such as uterine weight gain and ER binding assays to detect agonists [21]. The only way to address the issue of among-species extrapolation uncertainty is through further work in the area of comparative physiology and toxicology from a mechanistic perspective. This is a particularly critical issue for vertebrate classes (e.g., amphibians, reptiles) that have received little or no attention in terms of toxicology and/or endocrinology [16].

Uncertainties also exist as to how best to conduct screening tests. For example, although from the standpoint of cost-effectiveness limiting the number of doses tested in a screening paradigm would be desirable, selection of appropriate doses is at present problematic. Similarly, selection of the best route(s) of exposure (e.g., water vs injection in fish tests) and associated dosing regimes, although conceptually straightforward, still is unclear in a practical sense.

Finally, consideration should be given to expanding the role of in vitro tests and SAR models, relative to in vivo assays, in the screening process. For example, many of the in vivo tests recommended herein and by Gray et al. [21] as screens, although likely effective in terms of identifying (anti-) estrogenic/androgenic chemicals, could prove too resource-intensive to evaluate the thousands of chemicals of potential concern. However, until further work is done to address potential barriers (e.g., metabolism, toxicokinetic differences) to extrapolation across biological levels of organization (i.e., from in vitro to in vivo), the latter will remain the most conservative, and hence preferred, system for identifying EDCs.

Acknowledgements

Acknowledgement—We thank the many people involved in helping conduct the workshop and prepare this report, including Julie Gay, Sally Solomon, and Lee Salamone. We also thank Pat Schmieder, Steve Bradbury, and two anonymous reviewers for helpful comments on an earlier draft of the manuscript. Gerald Ankley, U.S. Environmental Protection Agency, NHEERL, Duluth, Minnesota 55804–1136; Ellen Mihaich, Rhǒne-Poulenc, Research Triangle Park, North Carolina 27709, USA; Ralph Stahl, DuPont Corporate Remediation, Wilmington, Delaware 19805, USA; Donald Tillitt, NBS, Midwest Science Center, Columbia, Missouri 65201, USA; Theo Colborn, World Wildlife Fund, Washington, DC 30037, USA; Suzzane McMaster, U.S. Environmental Protection Agency, NHEERL, Research Triangle Park, North Carolina 27709; Ron Miller, Dow Chemical, Midland, Michigan 48674, USA; John Bantle, Oklahoma State University, College of Arts&Sciences, Stillwater, Oklahoma 74078, USA; Pamela Campbell, Proctor&Gamble Technical Center, Strombeek-Bever, Belgium; Nancy Denslow, University of Florida, Department of Biochemistry, Gainsville, Florida 32160, USA; Richard Dickerson, Clemson University, ENTOX/TIWET, Pendelton, South Carolina 29670, USA; Leroy Folmar, U.S. Environmental Protection Agency, NHEERL, Gulf Breeze, Florida 32561; Michael Fry, University of California-Davis, Center for Avian Biology, Davis, California 95616, USA; John Giesy, Michigan State University, Department of Zoology, East Lansing, Michigan 48824–1222, USA; L. Earl Gray, U.S. Environmental Protection Agency, NHEERL, Research Triangle Park, North Carolina 27709; Patrick Guiney, SC Johnson Wax, Racine, Wisconsin 53402–5011, USA; Thomas Hutchinson, Zeneca, Brixham Environmental Lab, Freshwater Quarry, Brixham, Devon TQ5 8BA, England; Sean Kennedy, Canadian Wildlife Service, Hull, Quebec K1A 0H3, Canada; Vincent Kramer, Rohm&Haas, Department of Toxicology, Spring House, Pennsylvania 19477–0904, USA; Gerald LeBlanc, North Carolina State University, Department of Toxicology, Raleigh, North Carolina 27695, USA; Monte Mayes, Dow Chemical, Health&Environmental Sciences, Indianapolis, Indiana 46268–0511, USA; Alison Nimrod, University of Mississippi, Natural Product Center, University, Mississippi 38677, USA; Reynaldo Patino, Texas Tech University, Cooperative Fish&Wildlife Unit, Lubbock, Texas 79409–2120, USA; Richard Peterson, University of Wisconsin-Madison, School of Pharmacy, Madison, Wisconsin 53706, USA; Richard Purdy, 3M Company Environmental Laboratory, St. Paul, Minnesota 55144, USA; Robert Ringer, 625 High Point Drive, Mount Dora, Florida 32757, USA; Peter Thomas, University of Texas-Austin, Marine Science Institute, Port Aransas, Texas 78373, USA; Les Touart, U.S. Environmental Protection Agency, Washington, DC 20460; Glen Van Der Kraak, University of Guelph, Zoology, Guelph, Ontario, N1G 2W1, Canada; Tim Zacharewski, University of Western Ontario, Medical Sciences Building, London, Ontario N6A 5C1, Canada.

Ancillary