Agrochemicals with Estrogenic Endocrine Disrupting Properties: Lessons Learned?

Abstract

Many agrochemicals have endocrine disrupting properties. A subset of these chemicals is characterized as “estrogenic”. In this review, we describe several distinct ways that chemicals used in crop production can affect estrogen signaling. Using three agrochemicals as examples (DDT, endosulfan, and atrazine), we illustrate how screening tests such as the US EPA’s EDSP Tier 1 assays can be used as a first-pass approach to evaluate agrochemicals for endocrine activity. We then apply the “Key Characteristics” approach to illustrate how chemicals like DDT can be evaluated, together with the World Health Organization’s definition of an endocrine disruptor, to identify data gaps. We conclude by describing important issues that must be addressed in the evaluation and regulation of hormonally active agrochemicals incsluding mixture effects, efforts to reduce vertebrate animal use, chemical prioritization, and improvements in hazard, exposure, and risk assessments.

1. Introduction

Synthetic chemicals, including many herbicides and insecticides, play an important part in the global production of crops. Manufacturers of these so-called ‘plant protection products’ and other industry advocates argue that agrochemicals are necessary to provide sufficient food supplies, especially as the world population continues to rise (1–3). Yet, many of these chemicals are specifically designed and selected for use because they are biologically active (4), demonstrating that their use may not be innocuous. For example, numerous insecticides were specifically designed to target aspects of insect biology that are dependent on hormones such as molting, reproduction, and metamorphosis (5). Unfortunately, modern studies show structural and physiological similarities between these ‘insect’ hormones and their receptors, and the hormones and receptors used by crustaceans, birds, and mammals (6–8). Thus, in addition to their intended effects on target species, numerous pesticides affect physiological processes in non-target species, including humans (9–12).

Human epidemiology studies document associations between agrochemicals and a wide range of diseases including cancer (13–15), metabolic diseases (16–18), and neurological conditions (19,20), among others. In some cases, effects were observed in individuals with high levels of exposure due to their occupations, e.g., workers involved in pesticide spraying (15). Yet, in other cases that are perhaps more concerning, associations between pesticide exposures and diseases are observed in the general population, suggesting that even exposures that might be characterized as “low” can cause harm (21,22).

Many agrochemicals are endocrine disrupting chemicals (EDCs), e.g., compounds that interfere with one or more aspect of hormone action (23–25). Like chemicals used for other purposes (including food contact materials, plastics, personal care products, and industrial compounds), numerous pesticides mimic or block the actions of endogenous hormones, alter hormone synthesis or secretion, or alter downstream actions that would typically occur when a hormone binds its receptor. We note that the term ‘pesticide’, which encompasses many agrochemicals, is a broad term which includes insecticides, herbicides, fungicides, rodenticides, and other biocides.

Estrogenic EDCs are a subset of chemicals that, broadly defined, induce biological responses consistent with the effects of endogenous estrogens (26,27). As we discuss in greater detail below, many estrogenic chemicals are agonists of estrogen receptor (ER), although these may interact with the receptor (or the co-regulatory elements involved in estrogen signaling pathways) in ways that are distinct from endogenous hormones (28,29). Yet, others can produce estrogenic responses by upregulating the expression of ERs, increasing the cell or organism’s response to endogenous estrogens, and still others can increase the production of endogenous estrogens. Thus, the designation of an “estrogenic” chemical is not specific to one specific use (e.g., chemicals used in crop production) or to a specific mechanism of action (e.g., ER agonists).

The goal of this review is to use examples of agrochemicals with endocrine disrupting properties to provide guidance for how regulators and other decision-makers can identify and evaluate potential EDCs. We focus on three chemicals that interfere with estrogen signaling pathways, dichlorodiphenyltrichloroethane (DDT), endosulfan, and atrazine to highlight that there are several different ways that chemicals can manifest “estrogenic” behaviors. We use these three examples to demonstrate how current screening tests can serve as a good first-pass to determine if a pesticide is an EDC. We further argue that agrochemicals that act as EDCs should be evaluated in the context of the principles of endocrinology, even in the regulatory context; we show how the Key Characteristics approach can be used to differentiate chemicals that are most concerning. Finally, we briefly examine other issues that can affect the evaluation of agrochemicals including the dependence on hazard- versus risk-based approaches, mixture effects, and the global desire to reduce animal testing.

2. Model agrochemicals that interfere with estrogen signaling

We selected three agrochemicals as illustrative examples to explore how pesticides might affect estrogen signaling pathways and contribute to disease in non-target populations including humans. The first chemical, DDT (dichlorodiphenyltrichloroethane), is an organochlorine insecticide that was first synthesized in 1874 (30). Widely used in the 1930s through the early 1970s, human and wildlife exposure to DDT and its metabolite (DDE) has been documented, even in contemporary populations (31–33). The earliest evidence that DDT has estrogenic properties came in the 1960s with a rodent and avian uterotrophic test (34).

The second chemical, endosulfan (6,7,8,9,10,10-Hexachloro-1,5,5a,6,9,9a-hexahydro-6,9-methano-2,4,3-benzodioxathiepine-3-oxide) is a chlorinated hydrocarbon used as an insecticide from the 1950s until the late 2010s. Due to the widespread use of endosulfan, it is detected in remote locations that have little or no recorded use of the chemical (35); documentation of human exposure via biomonitoring studies are limited. Initial studies of endosulfan concluded that it has estrogenic activity due to increased proliferation of breast cancer cells in the E-SCREEN test following treatment with μM concentrations (36,37).

Finally, the third chemical, atrazine (6-chloro-N-ethyl-N′-isopropyl-1,3,5-triazine-2,4-diamine), is a triazine broadleaf herbicide used heavily on corn crops (38). Utilized as an agrochemical since the 1970s, atrazine continues to be applied to crops in the US, China, Australia, and elsewhere even though it has been banned in the European Union (39). The influence of atrazine on estrogen signaling pathways is more complex than the prior two examples; it does not promote cell proliferation in the E-SCREEN test (36) and has very low affinity for ER (40). Instead, it acts by increasing the expression of the cytochrome P450 enzyme, aromatase, which subsequently increases the production of estrogens (41).

In this section, we will provide a brief overview of each of these three chemicals. We note that this is not intended to be a systematic review but rather provide the reader with a general understanding of the hazards associated with exposures.

2.1. DDT, a historical example that continues to raise concerns

DDT was originally developed for military uses during WWII to control insects that act as vectors for infectious diseases including malaria, typhus and body lice (30). DDT was quite effective; malaria cases fell from 400,000 in 1946 to almost zero in 1950 (42). DDT exposures affect the nervous system by interfering with nerve impulses, causing the opening of sodium channels which blocks the ability of neurons to establish action potentials (43). Thus, its mechanism of action as an insecticide is not specific to mosquitos or other insects that carry pathogens but can affect action of sodium channels in non-target and non-insect species (44,45). Unfortunately, because of its perceived effectiveness in controlling insects and its relatively low cost to manufacture, it began to be used for agricultural crop production including cotton, peanut and soybeans and other purposes such as indoor pest control (46). In the United States, the maximum use of DDT as a pesticide occurred in 1959 (47). DDT found its way into the food chain and human dietary intake of DDT peaked in 1965 (48). As DDT use increased, insects including the target species (e.g., mosquitos) quickly developed resistance, making it less effective in controlling malaria and other infectious diseases (46,49).

Commercial DDT typically consists of mixtures of p,p’-DDT and o,p’-DDT (30). When metabolized, it creates several breakdown products including DDE (dichlorodiphenyl dichloroethylene) and DDD (dichlorodiphenyl dichloroethane). DDT and its metabolites bioaccumulate in lipids, and have a long half-life in the body (e.g., approximately 10 years in human blood and 4.2 – 5.6 years in adipose, depending on individual body composition) (50). Like other lipophilic chemicals, mobilization of fat stores (e.g., during dieting, lactation, etc.) releases DDT into the blood. It can then be excreted in the urine, feces or breast milk (51).

Because DDT bioaccumulates, it also biomagnifies in the food chain (51). Conservation biologists became concerned about DDT when its use was associated with eggshell thinning in birds due to its inhibition of prostaglandin synthesis and alterations in calcium metabolism (52). Several species, including the American bald eagle, became critically endangered as DDT use increased (53). DDT and DDE were also shown to affect reproductive health in alligators; males living in contaminated lakes in Florida, USA had decreased plasma testosterone concentrations, disorganized testes, and abnormally small penises whereas females had increased plasma estradiol concentrations and abnormal ovarian morphologies (54,55). In the US, DDT use was regulated starting in 1972 by the US EPA; yet, DDT continues to be used elsewhere, especially in sub-Saharan African nations, for the control of mosquitos that carry malaria (56).

For several decades, human exposures to DDT have been documented in the US in the National Health and Nutrition Examination Survey (NHANES), a biomonitoring and health program examining a representative population of children and adults. In the second round of the NHANES, conducted between 1976 and 1980, o,p’-DDT was detectable in 0.4% of the serum samples whereas p,p’-DDT was detected in 37.5% of the samples (32). p,p’-DDE, the primary breakdown product of p,p’-DDT, was detected in 99.5% of the samples. Since the 1970s, average serum concentrations of DDT and DDE in the U.S. population have declined by five- to ten-fold (32,33). In the fourth round of the NHANES, conducted between 1999 and 2004, o,p’-DDT was below the limit of detection but p,p’-DDT was detected in 73.8% of individuals aged 12 years and older (31). p,p’-DDE was again detected in 99.7% of persons aged 12 years and older. Thus, even though the average serum level of DDT has decreased in the US since it was banned from use as an agrochemical, the metabolites of DDT remain in the serum of many American adults.

2.1.1. DDT Mechanism of Action

In some of the earliest studies to examine the estrogenic effects of DDT, increased uterotrophic responses consistent with an ER agonist exposure were observed in rats, quail, and chickens administered o,p’-DDT, but not p,p’-DDT (34). Yet, in cell- and yeast-based assays, both o,p’-DDT and p,p’-DDT were capable of binding to the human ER (57). These results suggest some species differences in response to p,p’-DDT. Further, several DDT metabolites including o,p’-DDE and o,p’-DDD were ER agonists in cell and yeast assays using human ER. These studies revealed estrogenic effects of DDT and its metabolites, typically at high levels of exposure (to animals) or relatively high concentrations (in cells/yeast). In fact, data from several studies indicate that DDT is a “weak” estrogen compared to 17β-estradiol because of an approximately 1000-fold lower binding affinity for ER (58). Yet, even when rodents were administered lower doses of o,p’-DDT, at levels that were intended to replicate human serum concentrations, ER-mediated alterations in uterine responses were observed, consistent with a potent xenoestrogen (59). DDT can also exert estrogenic activity via mechanisms that are not dependent on the ER. For example, both o,p’-DDT and o,p’-DDD activate the transcription factor AP-1 in endometrial cells and kidney cells that are not responsive to estradiol (60). Finally, both p,p’-DDT and o,p’-DDE bind to GPR30, a transmembrane ER, and can displace 17β-estradiol (61).

2.2. Endosulfan, an insecticide being phased out of use

Endosulfan is a chlorinated hydrocarbon that was used around the world as an insecticide from its development in the 1950s until the late 2010s. With neurotoxic properties due to its agonist effects on GABA-gated chloride channels, endosulfan was widely used on food crops including potatoes, apples and tomatoes, and on cotton to control a variety of insect pests and mites (62). Over a period of five decades, cumulative production was estimated at approximately 308,000 tons (63).

Like DDT, endosulfan is a persistent organic pollutant, and because of these properties its use was phased out in agreement with the Stockholm Convention (64). Many countries, including the US, developed planned rollbacks ultimately leading to a complete ban, ending its use in the late 2010s. However, other members of the Stockholm Convention including China and India have continued its manufacture and use despite “legal” bans in those countries.

In one of the earliest studies of endosulfan in controlled laboratory studies, high doses (2.5–10 mg/kg) decreased epididymal sperm counts and altered biochemical parameters of testis health in male rats exposed during adulthood (65). When a similar study was repeated in male rats exposed during the peripubertal period, deficits in sperm count and sperm health were again observed (66). Rats exposed to endosulfan during gestation and the perinatal period also manifested decreased sperm count and other testicular parameters (67). Importantly, the effects of endosulfan on male reproductive outcomes are not limited to mammals, as they have also been observed in crocodiles (68). More recent investigations also suggest that early life exposures to endosulfan alter development, differentiation, and function of the female reproductive tract (69–73) and development of the male mammary gland (74).

2.2.1. Endosulfan mechanism of action

The earliest evaluation of the estrogenicity of endosulfan was from a test of estrogen potency, the E-SCREEN (36), which revealed that both endosulfan stereoisomers (endosulfan-α and endosulfan-β), individually or in combination, induced proliferation of MCF-7 breast cancer cells (37). Endosulfan similarly promoted proliferation of rat IEC-6 intestinal epithelial cells (75) and increased expression of estrogen-dependent genes in rat uterine cells (76). Despite these estrogen dependent endpoints, endosulfan is not considered an ER agonist; endosulfan has a binding affinity for ER that is less than 0.001% of 17β-estradiol (36,37). When HeLa cells transfected to express ER were treated with endosulfan, there was modest agonist activity of ERα and modest antagonist activity of ERβ (77).

Instead, endosulfan appears to alter estrogenic activity through modulating the expression of ER transcripts via methylation of specific promoter variants on the ER gene in a dose- and tissue-dependent manner; expression of ERα, as well as expression of ER-dependent target genes, was increased in the uterus of rats following endosulfan exposure (78). Consistent with non-genomic mechanisms, endosulfan also phosphorylates ERK in MAPK/ERK cascade signaling and increases activation of the downstream transcription factor, c-fos (79). It also activates other signaling transcription factors that represent endpoints in estrogen-sensitive MAPK, PI3K, PKA, and CaMKIV signaling pathways, and this activation was blocked by pharmacological ER antagonists (80).

2.3. Atrazine, an herbicide with continued widespread use

Atrazine is a triazine herbicide commonly used to protect plant agriculture, particularly corn, from invasion of broadleaf weeds (38). Historically, atrazine has been used heavily across the world with reports in the early 1990’s suggesting a worldwide application of 70,000 – 90,000 tons per year (81). Even after bans in the EU, it remains one of the most commonly used herbicides in the US, China and Australia (39). Its heavy use and moderate persistence have led to ubiquitous contamination of aquatic and terrestrial ecosystems.

Once it is in the body, atrazine is rapidly metabolized (half-life in humans = 10–11 hours) and it does not bioaccumulate in fat (82). However, it is semi-persistent in the environment, with no observed degradation in groundwater and a half-life of >200 days in surface water. In soil, it can remain for months or years, and it can migrate from soil into water. Atrazine is also detected in air due to its adsorption to dust and other particulate matter. In one biomonitoring study, researchers with the US CDC found detectable levels of one atrazine metabolite, atrazine mercapturate, in the urine of fewer than 5% of the population (limit of detection, 0.8 ng/ml) (83). Yet, these authors later determined that measurements of only one of the nine known metabolites of atrazine were producing underestimates of exposure (84). To our knowledge, these findings have not been applied to update estimates of human exposures in the US or elsewhere.

The effects of atrazine on amphibian health have received significant attention in the academic, industry, and regulatory communities after low doses altered male reproductive endpoints in amphibians (85–87). Perhaps the most shocking study of atrazine to date was a publication documenting the complete chemical castration of male frogs exposed to low doses of atrazine during development (88). Recent studies affirm these findings, suggesting that atrazine may have large consequences for sexual development in wildlife (89). Demasculinizing and defeminizing effects of atrazine have now been observed across taxa, with examples in mollusks (90), teleost fish (90), reptiles (68), birds (91), and mammals (92).

2.3.1. Atrazine mechanism of action

Early studies evaluating the estrogenic effects of atrazine were conducted in the 1990s when reproductive effects were reported in multiple strains of rats (93–95). Atrazine exposure induced several reproductive abnormalities including decreased length of estrous, longer periods of cornified epithelium, and longer periods of diestrus, each of which are associated with prolonged estrogen signaling. Yet, early evaluations of atrazine with the E-SCREEN assay identified no estrogenic responses in breast cancer cells (36). Further studies evaluating binding to the ER revealed it had low affinity (40). A breakthrough came in the year 2000, when an increase in expression of the cytochrome P450 enzyme aromatase was observed in H295R cells treated with atrazine (96); atrazine induces aromatase expression and increases estrogen synthesis by inhibiting phosphodiesterase (97). The effect of atrazine on aromatase expression continues to be supported with additional studies (98) and may serve as an exemplar for how a chemical can be considered “estrogenic” without directly binding to ER.

3. What makes a chemical an estrogen?

3.1. EU and US regulatory approaches to address agrochemicals that are EDCs

The presence of endocrine disrupting activity in pesticides is acknowledged in regulations in both the US and in the EU. In the European Union, EDCs are regulated by two legislative actions, the Plant Protection Products Regulation (99) which was implemented in 2009, and the Biocidal Products Regulation (100), implemented in 2012. According to these regulations, a chemical that has endocrine disrupting properties can only be used if exposures are considered negligible. In June 2016, the European Commission proposed criteria for how plant production products could be evaluated for endocrine disrupting properties, which would also require these agrochemicals to be listed as Substances of Very High Concern (SVHC) (101). Listing a chemical as an SVHC is the first step in restricting the use of a chemical in the EU. However, a set of criteria for how EDCs would be evaluated in the regulatory context was not approved by the European Parliament until late 2017.

In the US, the Food Quality Protection Act, implemented starting in 1996, required the US EPA to evaluate the estrogenic activities of pesticides, and called for the development of a screening program that would utilize validated tests (102). In the same year, the EPA was directed by amendments to the Safe Drinking Water Act to address the contamination of drinking water with estrogenic substances as well (103). In response to this legislation, to determine whether chemicals have endocrine disruptor activity, the EPA developed a two-tiered suite of assays to be implemented in the Endocrine Disruptor Screening Program (EDSP) (104,105).

3.2. Model agrochemicals and EDSP Tier 1 Assays for Endocrine Disruption

In the EDSP Tier 1 assays, in vitro and in vivo screens identify chemicals that have the potential to interact with the endocrine system; importantly, although the statutes only required testing for estrogenic activity of pesticides, the Tier 1 assays include tests to evaluate androgen receptor agonists and antagonists, ER agonists and antagonists, and thyroid hormone receptor agonists and antagonists (Table 1). Based on the intended design of the EDSP, any chemicals that are identified as showing effects consistent with endocrine disruption in one or more Tier 1 assays are supposed to be tested in Tier 2 assays. The Tier 2 assays utilize standard guideline tests (e.g., a one-generation toxicity test) to examine adverse outcomes and determine the doses at which effects are observed so that risk assessments can be performed using a “weight of evidence” approach (106,107). Importantly, there are concerns about whether these assays are appropriate or are sufficiently sensitive to assess the health effects of EDCs (24,108–115).

TABLE 1.

Summary of Tier 1 Screening Tests from the EDSP

Assay	Type of Assay	Description of Outcomes
ER binding assay	Cell fraction	Evaluates ER agonists and antagonists
Aromatase	Cell fraction	Determines whether compounds promote or interfere with enzymes that transform testosterone to estradiol
AR binding	Cell fraction	Evaluates androgen receptor agonists and antagonists
Steroidogenesis	In vitro	Determines whether compounds promote or inhibit other enzymes involved in hormone production
ER transcriptional activation	In vitro	Evaluates activation of the ER
Uterotrophic	In vivo (rodent)	Increased uterine weight, dependent on ER
Hershberger	In vivo (rodent)	Altered weight of male reproductive organs, dependent on androgens
Pubertal female	In vivo (rodent)	Examines timing of puberty and outcomes influenced by the hypothalamic-pituitary-gonadal axis. Also some thyroid outcomes.
Pubertal male	In vivo (rodent)	Examines timing of puberty and outcomes influenced by the hypothalamic-pituitary-gonadal axis. Also some thyroid outcomes.
Fish short term reproduction	In vivo (fish)	Evaluates functional outcomes of steroid hormones, or function of the hypothalamic-pituitary-gonadal axis
Amphibian metamorphosis	In vivo (Amphibian)	Evaluates thyroid hormone-dependent metamorphosis

When we examine the published literature and the available data from the EDSP Tier 1 assays, or other similar assays, a more complex story of the endocrine disrupting properties of DDT (and its metabolites), endosulfan, and atrazine emerges (Table 2). With regard to estrogenic activity: for DDT, there is strong evidence that this chemical and its metabolites are estrogenic in vitro and in vivo; for endosulfan, direct effects on the ER, as discussed above, are weak; for atrazine, , there is little if any evidence that it can bind to ER. For aromatase activity, which can produce estrogenic responses by increasing production of estradiol: endosulfan and atrazine induce aromatase expression. With regard to androgenic activity: DDT and DDE are antiandrogenic. Finally, for thyroid hormone disruption: there is some evidence that DDT can suppress thyroid function in the pubertal male assay; endosulfan increases the time to metamorphosis in amphibians, consistent with thyroid hormone disruption; atrazine also alters timing of metamorphosis in amphibians.

Table 2.

Summary of effects of three model agrochemicals in Tier 1 EDSP assays, or similar assays examining comparable outcomes

Assay	DDT	Endosulfan	Atrazine
ER binding assay	Positive for ER binding (143)	Negative for ER binding (143)	Negative for ER binding (198)
Aromatase	DDT only affects aromatase at cytotoxic concentrations (96) but DDE appears to induce hepatic aromatase in vivo (146)	Positive for weak inhibition of aromatase (199)	Positive for aromatase upregulation (200)
AR binding	Positive for AR antagonism (144)		Negative for AR binding (201)
Steroidogenesis	Positive for multiple effects on steroidogenesis in ovarian follicles (145)		Positive for indirect effect on aromatase activity (200)
ER transcriptional activation	Positive for ER transactivation (143)	Negative for ER transactivation (143)	Negative for ER transcriptional activation (202)
Uterotrophic	Positive for estrogenic uterotrophic activity (203)	Positive for some aspects of the uterotrophic test, but not uterine weight (204)	Negative for estrogenic activity Positive for anti-estrogenic activity (205)
Hershberger			Negative for androgenic activity Positive for anti-androgenic activity (183)
Pubertal female			Positive effects consistent with suppressed GnRH (206)
Pubertal male	Positive for suppression of thyroid gland function (207)		Positive effects consistent with altered hormone secretion (208)
Fish short term reproduction			Trends, but no significant effects (209)
Amphibian metamorphosis		Increased time to metamorphosis (210)	Evidence of non-monotonic effects on timing of metamorphosis (211)

Perhaps it should not be surprising that these three agrochemicals have more than one mechanism of action contributing to their endocrine disrupting properties. Evaluations examining the high throughput assays included in the EPA’s ToxCast screens have revealed that many EDCs participate in multiple endocrine pathways, as agonists or antagonists of hormone receptors, promoters or inhibitors of steroidogenic enzymes, disruptors of receptor transactivation or transcription factors involved in these processes (116–118).

3.3. Widening our view of “estrogenic” chemicals

The term “estrogenic” has overwhelmingly been used to indicate a chemical that acts as an agonist for ER. When a compound binds to the ER in a similar manner to endogenous estrogens (e.g., in the binding pocket), the bound ER translocates to the nucleus, recruits coregulators, and binds to estrogen response elements (EREs) on DNA, ultimately driving the expression of estrogen responsive genes (119–121). In this way, the activated ER complex acts as a transcription factor. This is referred to as the classical ligand-dependent estrogen receptor mediated pathway.

Signaling pathways that are initiated by estrogen binding to the nuclear ER are powerful and well characterized mechanisms of estrogen activity, but this is not the only pathway that leads to expressed genes typically driven by ER activation. Korach and colleagues (121–124) describe three additional mechanisms for estrogen/ER signaling: first, there is the ligand independent ER mediated pathway where cellular growth factors activate intracellular kinase pathways, phosphorylating and activating target proteins including the ER, causing the receptor to bind to EREs on DNA, even in the absence of a hormone ligand. Second, in ERE-independent signaling pathways, estrogens bind to the ER and then associate with alternative transcription factors (like Fos and Jun) to alter transcription of genes with alternative response elements like AP-1. This causes the ER to bind to DNA at sites distinct from the EREs, upregulating a unique set of genes. In the third pathway, the non-genomic signaling pathway, hormones bind to receptors on the cell surface, generating intracellular phosphorylation cascades that can alter methylation, and thus expression, of estrogen responsive genes (124,125).

The three model agrochemicals we have evaluated highlight different ways that environmental chemicals can ultimately produce estrogenic responses: direct binding to the ER, increased expression of the receptor itself, altered recruitment of coregulatory elements to the hormone-receptor (or EDC-receptor) complex, increased signaling via non-genomic pathways, and increased synthesis and secretion of endogenous estrogens. Thus, EDCs can act on a molecular level in distinct ways to ultimately produce estrogenic responses.

3.3.1. Key Characteristics

A recent approach to improve the characterization of EDCs suggests that EDCs can display one or more of ten “key characteristics” (126). The key characteristics include: interacts with or activates hormone receptors; antagonizes hormone receptors; alters hormone receptor expression; alters signal transduction in hormone-responsive cells; induces epigenetic modifications in hormone-producing or hormone-responsive cells; alters hormone synthesis; alters hormone transport across cell membranes; alters hormone distribution or circulating hormone levels; alters hormone metabolism or clearance; alters fate of hormone-producing or hormone-responsive cells (Table 3). Unfortunately, internationally validated assays are not yet available for most of these key characteristics (126), however, other well-developed methods are available for use (108,127–130).

Table 3.

Application of the key characteristics for EDCs to DDT

Key Char	Description	Evidence for DDT
KC1	interacts with or activates hormone receptors	Yes, ER
KC2	antagonizes hormone receptors	Yes, androgen receptor
KC3	alters hormone receptor expression	Yes, in female reproductive tissues
KC4	alters signal transduction in hormone-responsive cells	Yes, in mammary cells
KC5	induces epigenetic modifications in hormone-producing or hormone-responsive cells
KC6	alters hormone synthesis	Yes, via aromatase
KC7	alters hormone transport across cell membranes
KC8	alters hormone distribution or circulating hormone levels	Yes, androgens
KC9	alters hormone metabolism or clearance
KC10	alters fate of hormone-producing or hormone-responsive cells	Yes, in mammary gland

We have applied the Key Characteristics approach to the example of DDT (Table 3). When the evidence is evaluated collectively, and organized in a manner that allows the features of DDT to be examined, it becomes clear that this chemical is an EDC. Importantly, the Key Characteristics approach was designed to ‘free’ scientists and regulators from the task of demonstrating each molecular mechanism that is responsible for the adverse outcomes observed in exposed animals or human populations. For example, it may not be possible to know if the effect of DDT on breast cancer risk in women is due to its ability to bind ER; instead, this approach offers a “framework to evaluate biologically plausible connections between responses at different levels and from different methods” (126). Ultimately, determining whether a chemical is “estrogenic”, and evaluating the strength of the evidence in support of that conclusion, remains complex and almost certainly requires evaluations beyond those available in the EDSP.

3.4. Application of the WHO definition of an EDC

There remains some debate about how EDCs should be defined and identified; in the United States, the US EPA defined an EDC as “an exogenous agent that interferes with the production, release, transport, metabolism, binding, action, or elimination of natural hormones in the body responsible for the maintenance of homeostasis and the regulation of developmental processes” (131). In the European Union, the World Health Organization (WHO)’s definition is used, which describes EDCs as an “exogenous substance or mixture that alters function(s) of the endocrine system and consequently causes adverse effects in an intact organism, or its progeny, or (sub)populations.” (132). This definition has been interpreted as requiring that a chemical (a) produces an adverse effect, (b) has an endocrine mode of action, and (c) demonstrates a plausible relationship between the adverse effect and the endocrine mode of action (133).

Here, we apply this definition to one of our model agrochemicals, DDT to illustrate how chemicals with endocrine activities such as estrogenic activities can be characterized as EDCs. We also note that chemicals need not be developed to be endocrine active to be an EDC; DDT was never intended to interfere with hormonal action, thus its endocrine activity is an inadvertent side effect.

3.4.1. DDT: Evidence for adverse effects

Since the early 1990s, several epidemiology studies have examined the association between serum concentrations of DDT (or DDT metabolites) and breast cancer. This remains one of the best studied outcomes in human populations, with numerous prospective populations available (134). The first epidemiological study to show an association between DDT exposures and breast cancer revealed a threefold (OR = 3.68; CI: 1.01– 13.5) increase in the risk of breast cancer among those who were most heavily exposed to DDT (135). Yet, other studies reported weaker, or no, associations between DDT exposures and breast cancer. In a meta-analysis, 500 studies examining the association between DDT exposure and breast cancer were screened, and 46 case-control studies were included in the final analysis (136). Individuals with the highest levels of DDT exposure had a slightly elevated risk of developing breast cancer, but the associations were not statistically significant (DDE OR = 1.04, 95% CI: 0.94–1.15; DDT OR = 1.02, 95% CI: 0.92–1.13).

Compared to the typical case-control study, where DDT (and DDT metabolites) are measured at the same time as the breast cancer diagnosis, prospective approaches revealed different associations. For example, Cohn et al. measured DDT levels in blood samples collected from 1959–1967, and showed that women who were heavily exposed to DDT before the age 14 had a five-fold (OR = 5.4, 95% CI: 1.7–17.1) increase in the risk of developing breast cancer before menopause (137). These results indicate that age at exposure and the amount of DDT that an individual is exposed to have an impact on the risk of developing breast cancer. Follow-up studies using the same approach, examining early life exposures to DDT and examining breast cancer risk in later life, tell an equally compelling and similar story (138,139).

Although breast cancer risk is perhaps one of the best examined outcomes relevant to DDT exposures in human populations, other meta-analyses reveal associations between DDT (and its metabolites) and other endocrine diseases. For example, human and animal studies suggests that DDT and DDE promote adiposity, increased BMI, and obesity (16) although an older meta-analysis concluded DDT was not associated with an increased risk of type 2 diabetes (140). Another meta-analysis revealed non-significant increases in risk for prostate cancer compared to DDE exposures and occupational exposures to DDT (141).

3.4.2. DDT: Endocrine mechanism of action

As described above and in Table 2, DDT is widely acknowledged to be an ER agonist (142,143). Considering the key characteristics of EDCs, DDT is a androgen receptor antagonist (144), and it alters steroidogenic enzymes including aromatase (145,146). Circulating DDT exposures are associated with decreased serum testosterone levels in adult men living in a heavily contaminated area (147). Administration of DDT to rats increases expression of ER and progesterone receptor in the uterus and ovary (148); to our knowledge, expression of hormone receptors in the mammary gland of DDT-exposed animals has not been examined. DDT also activates cell-signaling cascades (e.g., the p38α MAP kinase cascade) (149), although some of these pathways may be activated independent of the ER as they cannot be blocked with cotreatment of ER antagonists. Finally, DDT exposure for one week during the pubertal period can increase proliferation and differentiation of epithelial cells in the mammary gland (150).

3.4.3. DDT: Plausible connections between endocrine activity and adverse effects

There are several aspects of DDT’s endocrine activity, including its ER agonist properties, its ability to increase ER expression in some tissues, and its promotion of proliferation in the mammary gland, that have plausible connections to increased breast cancer risk. Perhaps most important are the strong data indicating that DDT is an ER agonist. Lifetime exposure to estrogen is a risk factor for breast cancer (151); this has been demonstrated in epidemiology studies that show higher risk in women associated with early age of menarche or later age of menopause (152). Exposures to endogenous and synthetic estrogens during the perinatal period can promote breast cancer in adulthood (153–155).

When examining DDT in this way, there is a strong argument that it should be considered an EDC according to the WHO definition (132). This three-step evaluation, combined with the key characteristics approach (126), provides a strong basis to understand DDT, its effects on health, and the risks it poses to human populations. Because DDT and its metabolites are persistent, it is likely to pose risks to global human health for several more decades; further, because DDT continues to be used for malarial control in some areas of the world, some populations continue to be directly exposed.

4. Using endocrine principles to examine toxic chemicals with endocrine disrupting properties

One issue that has been raised previously is that the study of EDCs, including when decision-makers conduct hazard and risk assessments, should not rely solely on the traditional methods used to evaluate all toxicants (127,129,156–160). Instead, we and others propose that the study and regulation of EDCs should take into account the principles of endocrinology (24,161–163), including a) hormones are responsible for virtually all aspects of health and development, from conception until death; b) the effects of hormones are mediated via very specific interactions with receptors; c) hormones act at low doses; d) hormones induce non-linear, and even non-monotonic, responses; and e) the effects of hormones are dependent on life stage during exposures.

Why might it be important to consider these principles of endocrinology when evaluating agrochemicals? For DDT, the role of life stage has become very clear; when early meta-analyses did not account for the timing of exposure, a non-significant relationship between exposure and breast cancer risk was observed, yet once prospective studies were conducted and individuals with early life exposures were examined, this relationship became much clearer (137–139). This is also clearly true for endosulfan and its effects on male fertility (65–67) and the effects of atrazine on mammary gland development (164), and is likely true for most if not all EDCs (165). It is also clear from the studies of all three of the model agrochemicals that effects occur even in human populations with relatively low levels of exposure.

4.1. Improving chemical prioritization

Another challenge that must be addressed is the need to improve prioritization of agrochemicals (and other chemicals) for additional testing. Numerous strategies have been proposed to prioritize chemicals, including:

A). Prioritization of those chemicals where insufficient margins-of exposure exist.

Current regulatory approaches appear insufficient to protect human populations from EDCs (114,166–168). In the risk assessment process, chemicals are evaluated for hazards, typically with animal tests, and the doses at which harmful effects are observed are identified (169). Risk assessors then compare the doses at which harm is observed with the exposures (known or anticipated) in humans or wildlife. When a sufficient margin of exposure exists, no further action is needed. Therefore, perhaps the strongest evidence that risk assessments fail to protect public health comes from the field of environmental epidemiology. If the process of risk assessment described above is sufficient, no adverse effects should occur in humans if exposures only occur below the levels that induce harm in animals – even when uncertainties are accounted for (e.g., through the use of adjustment factors or correcting factors to calculate a tolerable daily intake from the results of a standard toxicity test) (114). Yet, hundreds of epidemiology studies show associations between agrochemicals and health outcomes, at “safe” levels of exposure, indicating that these risk assessment processes are insufficient.

B). Prioritization of those chemicals that have the most concerning physiochemical properties.

Chemicals like DDT and endosulfan were regulated, in part, because of their persistence in the environment. Other persistent chemicals, including many perfluorinated chemicals, should be prioritized for study and regulation for the same reason (170). Atrazine, which has been well studied, is no longer allowed to be used in the EU because of “ubiquitous and unpreventable” contamination of surface water, but it continues to be used in many other countries including the US (39). Because of the strong evidence that atrazine is an EDC, and additional evidence that it affects the health and development of animals (89,171), it should be prioritized for additional regulatory action.

C). Prioritization of the chemicals that have the biggest data gaps.

Many chemicals that are widely used such as glyphosate have been poorly studied, and it remains under debate whether they are EDCs (172). Yet, there is strong evidence that their use should be reduced or eliminated for other reasons (e.g., their potential to initiate or promote cancers (173,174)). Data gaps may exist for hazard identification (e.g., because insufficient testing was conducted prior to registration, or because human studies raise new concerns), for dose response characterization (e.g., because non-monotonic dose responses were not previously considered), or for exposure assessment (e.g., because biomonitoring has not been conducted, or because usage patterns have changed dramatically from when previous assessments were conducted). Importantly, the Key Characteristics approach can be a helpful way to organize mechanistic data, allowing data gaps to be more easily identified (126). The Key Characteristics approach may also allow for prioritization of chemicals for regulation by identifying those that are known “bad actors” with sufficient evidence to support the need for regulation or restrictions on use.

D). Prioritization of the chemicals that are used in the highest volumes.

In 2012, almost 3 billion kilograms of pesticides were used globally; approximately 50% of these chemicals were herbicides (175). In the USA, twelve of these pesticides were used at volumes exceeding 4.5 million kilograms per year (including glyphosate, atrazine, metolachlor-S, dichloropropene, 2,4-D, metam, acetochlor, metam potassium, chloropicrin, chlorothalonil, pendimethalin, and ethephon). Most of these chemicals had documented increases in use between 2005 and 2012; annual glyphosate use, for example, doubled over this period of time. Heavily used chemicals should be prioritized for study (176), especially if human and wildlife exposures have been documented.

4.1.1. Can we evaluate estrogenic chemicals without animal testing?

There is currently a global desire to reduce the number of animals used in chemical safety testing (177). In the EU, there are already restrictions around animal testing for a number of uses such as those used in cosmetics (178) and there are concerted efforts to reduce, replace, or refine vertebrate animal testing for other chemicals including those used as plant protection products and biocides (179). In the US, the EPA administrator has announced a commitment to reduce mammalian animal testing by 30% by 2025 and eliminate all agency requests for mammalian studies by 2035 (180).

How will chemicals be evaluated for estrogenic activity without mammalian models? There are several assays in the EDSP Tier 1 that do not require mammalian in vivo models (e.g., the cell fraction and in vitro tests, see Table 1). There are many other assays that are available through the ToxCast and Tox21 high-throughput approaches that can also be used to evaluate binding and activation of hormone receptors, as well as steroidogenesis outcomes including the activation or inhibition of aromatase (181,182).

In an evaluation of the EDSP Tier 1 data that are available for atrazine, it was argued that full completion of the EDSP Tier 1 assays is expensive and should be avoided, yet the authors also concluded that the non-animal tests are insufficient for regulatory decision-making (183). Others propose a step-wise approach to testing, whereby positive results observed in the least-expensive assays should prevent a chemical from being used for commercial purposes; therefore vertebrate animal testing would be reserved only for those chemicals that do not appear to raise concerns in cells or cell fractions (109). This type of tiered approach would significantly reduce total animal numbers.

To completely avoid animal testing for endocrine disruption, there must first be a method by which the results of in vitro and cell fraction assays can be used to characterize hazards and employed in the risk assessment process. Currently, regulatory agencies including the US EPA require that a chemical be shown to induce an adverse effect for it to be restricted from use (130,161), and it is unclear how data from non-animal testing can be used for this purpose. It is also unclear how cell cultures can account for sensitive periods of development, or the complexity of the maternal-fetal unit. Addressing these essential data gaps will be necessary before regulatory agencies can transition away from animal testing.

4.2. Brief review of other challenges for agrochemicals that are EDCs

4.2.1. Regulatory approaches are different in the EU, US and elsewhere

The different regulatory approaches used in the US and EU highlight how different jurisdictions have addressed some of the challenges related to EDCs. The hazard-based approaches used in the EU require that a chemical be removed from certain types of uses (including agrochemical pesticides) if it is demonstrated to be an EDC. The implementation of this approach will likely require that more pesticides be regulated, preventing the use of chemicals that have endocrine disrupting properties from use on crops – similar to how chemicals with carcinogenic, mutagenic, or reproductive toxicity are disallowed from use in pesticides in the EU. As stated in a 2009 European Implementation Assessment, “no approval is granted when a substance to be used in plant protection products (active substance, safener, co-formulant or synergist) is of high risk for human health (carcinogens, mutagens, toxic for reproduction, endocrine disruptors) or environment (persistent organic pollutant – POP; persistent, bioaccumulative and toxic – PBT; very persistent and very bioaccumulative – vPvB). This means that the approval process is governed, at EU level, according to a hazard-based approach (substances are eliminated from the approval process based on the hazard posed by those substances), which distinguishes the European regulatory system as drastically more strict than comparable chemical regulations outside Europe or in other EU sectors” (184).

In contrast, the risk-based approaches used in the US suggest that EDCs can be used in pesticides, as long as exposures are controlled and maintained below a “reference” dose. Unfortunately, even though the United States has a fairly robust biomonitoring program developed and implemented by the US Centers for Disease Control and Prevention (185–187), exposure data are lacking for a large number of pesticides. Further, for those pesticides that do have data available, it is not clear that they adequately account for seasonal variation in exposure, as would be anticipated for chemicals that are sprayed only during certain times of the crop growing season.

4.2.2. Other scientific challenges

There are other challenges that are relevant to protecting the public from pesticides with endocrine disrupting properties. These include the need to address agrochemicals as mixtures of chemicals, even though many laboratory studies examine only one compound at a time. There are concerns, based on studies comparing the effects of active ingredients and whole pesticide formulations (188), that studies examining only active ingredients may underestimate – or entirely mischaracterize – the effects of pesticide mixtures. Chemical mixtures, including estrogenic chemicals, can have additive and even synergistic effects in concert (189). Additionally, concerns have been raised that rodent chows are widely contaminated with pesticides (190), creating difficulties in evaluating the effects of pesticides on a non-contaminated background.

There are also issues with whether individual chemicals are characterized as the active ingredient versus an “inert” ingredient in pesticide formulations (191); “inert” compounds added to pesticides are not necessarily biologically inert. They are often adjuvants, e.g., chemicals added to make it easier to mix the formulation, to enhance the activity of the active ingredient, or compounds that cause the active ingredient to stick to an applied area; inert ingredients can also include carriers (e.g., chemicals that aid in the delivery of the active ingredient) and diluents (e.g., solvents used to dilute or dissolve the chemicals). An example of an “inert” ingredient is the surfactant polyethoxylated tallow amine, which is utilized in many pesticides. It may be an EDC because of its ability to decrease aromatase activity (192) (although the manufacturer maintains it is safe (193)). Another feature that makes it difficult to evaluate pesticide formulations is that the ingredients, and relative concentrations of these ingredients, can differ greatly between products, or even between products that appear to be the same that are purchased in different locations.

5. Lessons learned & a path forward

We selected three model agrochemicals, DDT, endosulfan, and atrazine, that are also known or putative EDCs. Each of these chemicals disrupts endocrine signaling pathways in a way that is consistent with the broader concept of “estrogenicity”, but they do this via distinct mechanisms. These chemicals are also unique because of their histories – one has been heavily regulated globally for several decades, one has more recently been regulated, and the third continues to be widely used in many countries. Lastly, these three agrochemicals are distinct because of their toxicokinetic properties; DDT and endosulfan are persistent pollutants that bioaccumulate in fat, whereas atrazine is only pseudo-persistent because of its repeated use (e.g., it is constantly found in the environment due to regular applications), and it does not bioaccumulate in fat.

We also examined the evidence that these three chemicals are EDCs. There are tools available to evaluate agrochemicals for endocrine disrupting activity including the EDSP Tier 1 Assays. Unfortunately, and contrary to the intention of the EDSP program, very few pesticides have been evaluated with the full suite of assays. Going forward, it will be necessary to evaluate more chemicals. Considering the large cost to run all eleven assays, it may make more sense to prioritize chemicals for testing based on specific criteria (e.g., human exposure levels, amounts applied yearly, or other physiochemical properties of concern), to move to high throughput screening, and to run the least expensive tests first followed only by more expensive tests if the least expensive tests suggest no harm. This will be especially important as regulatory agencies move away from mammalian and other vertebrate animal testing. The EDSP assays are not sufficient to evaluate all aspects of endocrine disruption, but they do offer a logical first-pass that allows for the evaluation of some aspects of estrogen, androgen, and thyroid signaling pathways.

We also note that the key characteristics approach can be a helpful way to organize and synthesize mechanistic data, and identify data gaps, when evaluating putative EDCs. EDCs are often characterized based on the outcomes that they disrupt or affect, such as the label of “estrogenic” chemicals, but there are multiple distinct mechanisms by which chemicals can display estrogenic properties. This is important when evaluating potential risks to exposed populations, but it also means that simplified screening programs need to account for this endocrine complexity.

We have noted throughout this review that, for many agrochemicals, human exposure data are lacking. This absence of good exposure data will ultimately affect the ability of regulatory agencies to conduct public health protective risk assessments. Further, it has been argued that EDCs may not be appropriate for risk assessments and instead should be evaluated based on hazard assessment and identification instead (194). Complications such as non-linear dose responses may make the traditional methods of extrapolation from high dose effect to low dose safety inappropriate (195). For this reason, the EU’s approach to regulate EDCs in a manner similar to carcinogenic, mutagenic, and reproductive toxicant substances should be explored across the globe, including in the US.

Finally, we acknowledge that protection of agriculture is necessary, and is often the goal of agrochemical use. However, agriculture and crop production cannot and should not be divorced from human health. While future decisions to regulate agrochemicals can take into account the costs of regulation (on farmers or on crop production), suggestions that specific agrochemicals are needed to feed a growing population should be supported by evidence. For example, economic studies in the EU following the regulation of atrazine do not support the original claims from the agrochemical industry that banning the use of atrazine would be devastating to corn production (196). It is equally important that any cost-benefit analyses conducted to justify the use of agrochemicals also carefully calculate the costs of not regulating (e.g., costs to human health, disruption to wildlife ecosystems), which can be significant (197).

We also cannot divorce the issue of agrochemicals, and synthetic chemical production more broadly, from the context of anthropogenic activities, which challenge public health. As climate change disrupts the lifecycles, foraging behaviors, and geographic range of insect pests, including insects that will destroy crops and carry diseases, societies will need to respond in ways that do not rely on pesticides that are also EDCs. Already today, from oil and gas extraction to plastics production to soil depletion to insect population decline to loss of biodiversity, human activities are having detrimental impacts on both public and environmental health. Ultimately, when societies defend and promote the use of EDCs, we not only harm the health of human and wildlife populations, we stifle progress towards healthier agrochemical options.