Human-relevant potency threshold (HRPT) for ERα agonism

HRPTs can be expressed as differences in mechanistic potency relative to well-characterized endogenous effectors or human pharmaceuticals. They are derived from mechanistically based in vitro and in vivo assays supported by empirical observations from clinical studies, or in rare cases, epidemiological evidence. To develop an HRPT, data are sought for effectors with high, intermediate, and low mechanistic potencies via the particular MoA, e.g., ERα agonism. In practice, these respective categories will include (1) well-characterized endogenous effectors and/or human pharmaceuticals that produce a predictable physiological effect specific to the MoA; (2) effectors with measurable, but weak mechanistic potencies that produce no physiological effect via the specific MoA; and (3) effectors with mechanistic potencies intermediate to categories 1 and 2 that produce physiological effects via the specific MoA only under conditions of high dose, and often only in combination with a sensitive physiological or developmental state. In addition to data for different effectors of widely varied mechanistic potency, HRPT derivation requires mechanistic potency data from a common endpoint specific to the MoA of interest. The most useful endpoints are those that measure a functional change (e.g., transcriptional activation; cellular proliferation or hypertrophy) rather than merely a molecular interaction (e.g., binding affinity), but deriving an HRPT may require correlating mechanistic potency data from the molecular and physiological levels. This is particularly important when the MoA of interest includes numerous specific molecular interactions.

To address the EC’s requirement, here we illustrate the method for deriving an HRPT by applying the concept to the estrogen MoA. For brevity, we will limit our illustration to comparisons of mechanistic potency via ERα agonists with the understanding that an HRPT derivation for the complete estrogenic system may also require comparisons of mechanistic potencies via each component of the estrogen system. As such, the HRPT proposed here is applicable only to the ERα agonist MoA. Work is ongoing within our group to develop a more comprehensive HRPT proposal for estrogen MoAs.

Several features make the estrogenic system an excellent prototype to illustrate the derivation of an HRPT. The complexity of the estrogenic system makes it one of the most challenging systems for which an HRPT might be developed, yet, the wealth of available data makes it possible to address each intricacy. To modify the estrogen system, an effector must bind to and alter the functional state of at least one of the following; (1) an estrogen receptor, (2) the system for synthesis, activation, desensitization, or breakdown of estrogen receptors, (3) an enzyme or enzymes responsible for synthesis or breakdown, or elimination of estrogens and their precursors, (4) circulating proteins that bind estrogens, (5) the synthesis, activation or destruction of these enzymes and proteins (items 2–4), (6) endocrine and other systems that modify the estrogen system (e.g., LH, FSH, GnRH, inhibin), (7) co-activators, co-repressors, co-integrators, or chaperones, (8) any system that influences the estrogen system through cross-talk, or (9) the presence and availability of estrogen receptor effector systems. These numerous complexities support the proposition that physiological effects may not be predictable across agents that can interact with the estrogenic system and that their characterization may need to proceed on a chemical by chemical basis. Screening of multiple effector systems and with different in vivo physiological models may be required (Germain et al. 2006). Since the primary determinants of mechanistic potency—affinity and activity—underlie all physiological responses produced through hormonal pathways, whether therapeutic or adverse, this implies that deriving HRPTs may require assessments of mechanistic potency for several effectors.

Although the complexity of estrogenic pathways is challenging, the basic principles of receptor, enzyme, and transport kinetics are applicable to each molecular interaction that composes the system. These principles form the conceptual basis for mechanistic potency thresholds that must be exceeded for a chemical to produce physiological effects (Borgert et al. 2013), a concept corroborated through modeling of molecular signaling networks (Zhang et al. 2014). To our knowledge, no theories have been proposed that would replace these fundamental tenets of endocrine pharmacology and toxicology, and thus, it would seem reasonable that HRPTs can be estimated if sufficient data are available. Wealth of data is available for many pharmaceuticals that target estrogenic pathways, e.g., those that are used for hormone replacement therapy, birth control, treatment of estrogen-dependent cancers, and other estrogen system modifications. Furthermore, widespread human exposure to botanical estrogens in the diet provides considerable human data on molecules that are much less potent than active pharmaceuticals, but which nonetheless may produce physiological responses under certain conditions. Finally, extensive characterization of the estrogen receptor system has produced abundant mechanistic potency data for these endogenous, pharmaceutical, and botanical estrogens, as well as for endogenous and exogenous molecules that exhibit weak interactions with components of the estrogen system but produce no discernible physiological effects. Thus, sufficient data are available for molecules spanning a wide range of estrogenic mechanistic potencies and physiological effects to allow estimation of an HRPT for most estrogenic MoAs.

The estrogen system is important for the growth, development, and regulation of the female reproductive system and in pregnancy. The estrogenic system is also important for the growth and development of the male reproductive system. It also has important effects on bone, the cardiovascular system, the brain, liver, and other tissues in both genders.

The primary source of endogenous estrogens in the non-pregnant female is the ovarian follicle and its derived tissue the corpus luteum. The three primary endogenous estrogens from this source are 17β-estradiol, estrone, and estriol. The placenta produces these estrogens during pregnancy. Estetrol is an endogenous estrogen produced from estriol and 17β-estradiol by the fetal liver during pregnancy. Chemicals that have some estrogenic activity are also produced in breasts, adrenal glands, adipose tissue, and liver. In males, follicle stimulating hormone (FSH) induces the synthesis of aromatase in the testis, and it induces this enzyme in the adrenal glands and other tissues in both genders. Aromatase converts testosterone and other androgens into estrogens. During menopause, estrone, produced in adipose tissue, is the primary estrogen. Other non-ovarian, non-pregnancy-related endogenous compounds that have some estrogenic activities include 2-hydroxyestrone, 4-hydroxyestrone, and 16-hydroxyestrone, which are metabolites of estrone, metabolites of dehydroepiandrosterone (DHEA) 7-oxo-DHEA, 7α-hydroxy-DHEA, and 16α-hydroxy-DHEA, 7β-hydroxyepiandrosterone, Δ4-androstenedione, Δ5-androstenediol, 3β-androstanediol, 3α-androstanediol, and 5α-androstane-3β,17β-diol produced from testosterone, and 27-hydroxycholesterol produced from cholesterol.

There are numerous sources of exogenous chemicals with estrogenic activities. Plants consumed as food contain a variety of phytoestrogens. Other chemicals that have estrogenic activity include pesticides such as o,p-DDT and chlordecone, surfactants such as octyl and nonyl phenol, phthalates used as UV stabilizers in plastics, polychlorinated biphenyls, once widely deployed as dielectric and coolant fluids in electrical apparatuses and other uses, drug products such as diethylstilbestrol and tamoxifen, and an extensive variety of others.

Most estrogenic chemicals produce their physiological effects via one or more of at least three estrogen receptor subtypes—nuclear receptors ERα and ERβ and the G-protein-coupled membrane receptor GPER30. The nuclear receptors can bind to estrogen response elements to either stimulate or repress gene transcription. They can also tether to other response element binding proteins to act as co-activators or co-repressors of gene transcription. The nuclear receptors can also be transported to the cell membrane where they form complexes with other membrane receptors, e.g., mGluR1, to activate the transduction systems operated by those receptors (Sinchak and Wagner 2012). Some estrogenic chemicals that act as agonists at some of the receptor subtypes under some conditions can act as antagonists at others (Ring and Dowsett 2004; Tzukerman et al. 1994; Harrington et al. 2003). When agonists bind to ERα or ERβ, helix 12 of the receptor protein relocates and seals the agonist binding site (Brzozowski et al. 1997). The bound receptor dimerizes, is phosphorylated on two serine residues, and may then complex with its target system in the cell to produce its physiological effect. The relocation of helix 12 also promotes the association of co-activator with the receptor. Phosphorylations are mediated by several second messenger protein kinase systems including those that are activated through stimulation of growth factor receptors (cross-talk). Some ER antagonists go through the same steps with the exception that helix 12 relocation to seal the binding site is prevented. This also prevents association of the complex with co-activators.

GPER30 is activated by estrogens resulting in Gβγ stimulation of a tyrosine kinase, which phosphorylates and activates SrC-1, which phosphorylates and activates ERK1 and ERK 2 protein kinases (Filardo et al. 2009). In addition to activation of ERK1,2, GPER30 activation can result in a variety of other intracellular signaling actions including formation of IP3 with resultant release of calcium from the endoplasmic reticulum, and formation of cAMP with resultant activation of protein kinase A (Filardo et al. 2009, 2002).

To clearly and concisely illustrate the process for developing an HRPT, we constrain our analysis to the ERα agonist-mediated MoA. Since ERα agonism is the putative MoA for many alleged EDCs, the HRPT for this MoA is immediately useful for hazard identification and its derivation provides a prototype for the development of HRPTs generally.

In vivo ERα agonist mechanistic potencies

When measured in vivo with the uterotrophic assay, the same relative order of mechanistic potencies is generally observed among the groups of chemicals (Table 2). Endogenous estrogenic hormones and pharmaceutical estrogens are the most potent chemicals via ERα; 17β-estradiol, 17α-ethinylestradiol and diethylstilbestrol define the supra-threshold potency range and serve as positive controls and reference chemicals for this assay (Kanno et al. 2003). Although both aromatizable and non-aromatizable androgens increase uterine weight in rats, their uterotrophic action is antagonized by anti-androgens, with little inhibition by anti-estrogens except at very high doses (Beri et al. 1998; Armstrong et al. 1976; Schmidt et al. 1976; Schmidt and Katzenellenbogen 1979). Thus, androgens appear to exert a predominantly anabolic effect on the uterus rather than an estrogenic effect, which is consistent with our contention that their in vitro mechanistic potencies at ERα (approximately 5 × 10⁻⁶ relative to 17β-estradiol; see Table 1) are insufficient to be physiologically effective.

Table 2

Uterotrophic relative potencies calculated* from values reported in the literature

Open image in new window

*Unless provided in the publication, Relative Potencies (RP) were derived by dividing an EC% value (listed or estimated) of the reference chemical ^1,2,3 by the equivalent EC% value (listed or estimated) of each test chemical. EC% values were either extracted from tables or interpolated from figures. Where necessary, concentrations were adjusted to better approximate equivalent EC% values. ND not different than control

¹Estradiol-17β (E2) used as reference chemical for calculation of relative potencies

²Ethinyl estradiol-17α (EE) used as reference chemical for calculation of relative potencies

³Diethylstilbestrol (DES) used as reference chemical for calculation of relative potencies

^aHigh content of genistin/genistein/glycitein

^bLow content of genistin/genistein/glycitein

^AJefferson et al. (2002). CD-I immature mice; SQ injection; dose range 0.01–1000 000 ng/kg/day

^BLomnitski et al. (2003). Immature CD-I mice; SQ injection; 500 and 500,000 μg/kg/day

^CSwankar et al. (2012). Ovariectomized Balb/cByJ mice; SQ injection; 5 mg/kg/day

^DHe et al. (2003). Ovariectomized B6C3F1 mice; oral dose; 1, 10, 50, 100, 250, 500, 1000 mg/kg

^ESong et al. (1999). Immature B6D2F1 mice; oral gavage; 3 mg/day

^FSelvarag et al. (2004). Ovariectomized age-matched C57BL/6 mice; SQ injection 0, 4, 8, 12, 20 mg/kg/day or in the diet 0, 500, 1000 ppm for 12 days

^GColdham et al. (1997). Prepubertal CFLP mice; SQ injection; DES doses of 0.5, 5, 50 ng; coumestrol doses of 1, 10, 100 μg

^HRachon et al. (2007). Ovariectomized SD rats; in diet; 400 mg/kg of chow

^JMcKim et al. (2001). Immature Sprague-Dawley^c and Fischer − 344^d rats; oral gavage; 10, 50, 100, 250, 500, 1000 mg/kg/day

^KYamasaki et al. (2002a). Immature Crj:CD(SD) rats; SQ injection; 0.6 ng/kg/day (EE) − 30 mg/kg/day (Genistein)

^LSchmidt et al. (2006). Ovariectomized Wistar rats; oral dose by gastric tube; 100 mg/kg

^Mde Lima Toccafondo Vieira et al. (2008). Immature Wistar rats; oral gavage; doses 125, 300, 720, 1730, 4150 mg/kg/day

^NPunt et al. (2013). Immature rats, SQ injection of 35 mg/kg/day

^OBarlas et al. (2014). Immature Wistar rats; oral gavage; 1, 10, 100 mg/kg/day

^PYamasaki et al. 2003. Immature Crj:CD(SD) rats; SQ injection; 0.6 ng/kg/day–40 mg/kg/day

^QYamasaki et al. (2002b). Immature Crj:CD(SD) rats; SQ injection; 0.2 ng/kg/day–200 mg/kg/day

^RKeiler et al. (2015). Immature Lewis rats; SQ injection; 4 μg/kg/day–15 mg/kg/day, ^e6-(l,l-dimethylallyl)naringenin

^SOverk et al. (2008). Ovariectomized Sprague Dawley rats, 200 g body weight. 50 μg/kg/day E2; 4 mg/kg/day (8-PN); 4-400 mg/kg/day (extracts). Chemicals measured in plasma (< 0.5–3.7 ng/mL), liver (< 2–4.4 ng/g) and mammary gland (< 3–0.6 ng/g). ^fMetabolite of 8-prenylnaringenin

Several botanical estrogens are more potent than androgens at ERα (Table 1), but unlike androgens, the uterotrophic effects of those tested are completely blocked by estrogen-receptor antagonists (Yamasaki et al. 2002b; Woods and Hughes 2003). Their uterotrophic potencies based on data from mouse and rat are generally consistent with or lower than their in vitro potencies, on the order of 10⁻³–10⁻⁵ relative to 17β-estradiol or 17α-ethinyl estradiol (Table 2; Woods and Hughes 2003). None of the botanical estrogens show greater in vivo mechanistic potency (Table 2) than in vitro (Table 1). 8-prenylnaringenin has the highest in vivo mechanistic potency (Table 2) of the botanical estrogens, but is more potent in vitro by an order of magnitude (Table 1). Although zearalenone appears to be the most potent botanical estrogen in vitro (Table 1), it is four orders of magnitude less potent in vivo (Table 2). Diethylstilbestrol has mechanistic potency 100–1000 times greater than those botanical estrogens in vivo. The uterotrophic potency of D4 is roughly 100 times lower than the least potent of the botanical estrogens, on the order of 10⁻⁶–10⁻⁷ relative to ethinyl estradiol in rats and 17β-estradiol in mice, but unlike the androgens, its uterotrophic activity is completely blocked by the pure estrogen antagonist ICI-182,780 and is absent in ERα-knockout mice (He et al. 2003).

HRPTs can be estimated by comparing clinical data on the physiological effects in humans of chemicals that vary widely in their relative mechanistic potencies for ERα agonism. The estrogenic effects of 17β-estradiol, and 17α-ethinyl estradiol, are well known (Simpson and Santen 2015), producing consistent effects in humans, non-human primates, and in assays specific and sensitive for ERα agonism such as ERαTA and uterotrophic assays. The minor hormonal estrogens estrone and estriol have mechanistic potencies within two orders of magnitude of 17β-estradiol (Watson et al. 2008). Hence, chemicals that reproducibly exhibit in vitro and in vivo ERα agonist potencies within two orders of magnitude of 17β-estradiol would predictably produce estrogenic effects in human males and females. Of note, α-erythroidin, an alkaloid found in certain Chinese herbal medicines, appears to have a mechanistic potency within three orders of magnitude of 17β-estradiol (Table 1) and is reported to have abortifacient activity (Djioque et al. 2014).

Soy isoflavones, which are three to five orders of magnitude less potent than 17β-estradiol (Table 1), have been extensively evaluated and have shown no clear evidence of ERα-mediated estrogenic activity at high dietary levels of exposure or at higher doses when administered as post-menopausal hormone replacement in women (Cline et al. 2001; McCarty 2006; Woods and Hughes 2003; Bedell et al. 2014; EFSA 2015). EFSA (2015) concluded that the results of four epidemiological studies (2216 isoflavone users, total) and ten interventional controlled studies (816 participants, total) did not show an association between exposure to isoflavones-containing food supplements and adverse effects in breast, and that the majority of animal studies were negative. For uterus, EFSA (2015) found a general lack of statistically significant changes in endometrial thickness compared to control. A few histopathological effects and a few cases of endometrial hyperplasia were reported, but not cases of endometrial carcinoma. EFSA (2015) found that most of the animal studies that evaluated uterine effects were also negative.

Similarly, the results of a recent meta-analysis found that ingestion of soy protein or isoflavone, even at levels that greatly exceed the typical Japanese dietary intake, has no effect on reproductive hormone concentrations in men as measured by testosterone, sex-hormone binding globulin, free plasma testosterone and free androgen index (Hamilton-Reeves et al. 2010). Similarly, soy or soy isoflavones lack significant effects on circulating estrogen levels, sperm counts or sperm parameters, and there is no compelling evidence that soy intake produces erectile dysfunction or gynecomastia in men (Messina 2010). Vandenplas et al. (2014) found no strong evidence of adverse outcomes among published cross-sectional, case–control, cohort studies or clinical trials conducted in children fed soy-based infant formulas from 1909 to 2013, and which provided information on anthropometric growth, bone health (bone mineral content), immunity, cognition, and reproductive and endocrine functions. Infants fed soy-based infant formulas had similar levels of hemoglobin, serum protein, zinc and calcium concentrations and bone mineral content as children fed formulas based on cow’s milk or vegan human-milk substitutes (Vandenplas et al. 2014).

Some lignans and essential fatty acids have ERα agonist potencies similar to the lower-potency soy isoflavones based on binding affinities and cellular responses (Jin et al. 2013; Liu et al. 2004; Pianjing et al. 2011; Rose and Connolly 1989; Menendez et al. 2004). Although their agonistic interactions with ERα are detectable in the highly sensitive ERαTA (Table 1: enterolactone, enterodiol, sesamin, sesamol, sesamolin) and the in vivo uterotrophic assay (Penttinen-Damdimopoulou et al. 2009), their mechanistic potencies appear to be too low to produce overt physiological effects in either rodents or humans (Table 3). Beneficial effects attributed to dietary lignans and dietary supplements containing lignan extracts may be due to MoAs other than ERα agonism (Adolphe et al. 2010).

Studies in non-human primates are generally consistent with the human data. Although changes in maternal and fetal 17β-estradiol and testosterone levels and proliferation of Leydig cells have been reported in some studies with non-human primates administered high doses of soy isoflavones, clear estrogenic effects have generally not been observed (Cline et al. 1996; Woods and Hughes 2003; Wood et al. 2006; EFSA 2015). Differences in dose or bioavailability seem inadequate to fully explain the contrasting estrogenic effects of soy isoflavones in rodents versus non-human primates, as a recent pharmacokinetic investigation indicates that genistein administration produces circulating blood levels in rhesus monkeys comparable to or higher than those that produce adverse reproductive effects in rodents (Doerge et al. 2016). Instead, the difference is likely due to a selectivity of soy isoflavones for the ERβ subtype in humans and non-humans primates (Jiang et al. 2013).

Important differences in pharmacokinetic parameters (Soukup et al. 2016) suggest that humans would be slightly less sensitive than rats and substantially less sensitive than mice to the effects of an equivalent internal dose of soy isoflavones. In humans, the more highly estrogenic aglycone derivatives represented a very small proportion of the total plasma isoflavones, which was a slightly lower proportion than seen in rats overall and substantially lower than in mice. The proportion of conjugated isoflavones was very high in humans, similar to rats, and greater than in mice. All male and female rats and mice converted daidzein to the more estrogenic S-equol, whereas only 36% of women and 20% of men showed this capacity (Soukup et al. 2016). These factors suggest that mechanistic potencies for the various isoflavones will not be underestimated from results of transactivation assays and rodent uterotrophic assays.

Despite the apparent lack of ERα agonist-mediated estrogenic activity in humans, botanical estrogens and isoflavones are now used as hormonal replacement therapy by millions of post-menopausal women. Initially, therapeutic benefits were questionable. Sirtori et al. (2005) concluded that the clinical epidemiological data did not support claims of therapeutic effects by soy isoflavones, even at high levels of intake, and that the physiological effects of phytoestrogens appeared to be insignificant in humans. Although intra-individual and interspecies differences in intestinal conversion and bioavailability of isoflavones and their derivatives (Woods and Hughes 2003) account for some of this variability, mechanistic potency also plays an important role. With recognition of the importance of standardizing botanical preparations to the most potent constituents (de Lima Toccafondo Vieira et al. 2008; Messina 2014) have come an increasing proportion of studies that do show efficacy. Table 4 summarizes extensive reviews published since Sirtori’s (2005) analysis that report some efficacy of soy isoflavones among 60 clinical trials and more than 25,000 subjects, but no attributable adverse effects (Messina 2014; EFSA 2015).

Standardizing preparations to the most potent estrogenic constituents has not increased the incidence of adverse side effects typically associated with pharmaceutical estrogens. This improved efficacy without adverse ERα-mediated proliferative side effects may be explained by the selectivity of isoflavones for ERβ relative to ERα (McCarty 2006). For several major isoflavones, the affinities for ERα and ERβ are similar, but the relative effects on transactivation are much different (Pfischer et al. 2008; Dornstauder et al. 2001). Thus, the potency of bioavailable isoflavones via ERβ approaches that of pharmaceutical estrogens and produces physiologically detectable effects, while their potency via ERα is lower and physiologically less significant in humans.

Relative mechanistic potencies of botanical isoflavones, as reflected in the ERα-transactivation and uterotrophic potencies of isoflavones used to standardize botanical estrogenic supplements, would thus appear to define a potency threshold below which adverse effects in humans are unlikely. Based on in vitro (Table 1) and in vivo (Table 2) relative potency values for coumestrol, equol, and genistein, and comprehensive reviews by the UK Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment (Woods and Hughes 2003) and by the European Food Safety Authority, EFSA (EFSA 2015), a conservative estimate of this potency threshold would be in the range of 1.0E−04 relative to the potency of 17β-estradiol (Tables 1, 2). It would seem improbable for chemicals with relative potencies below this level to exhibit physiological effects via ERα in humans. Indeed, we have found no literature documenting bona fide ERα agonist effects in humans produced by substances with ERα agonist potencies below this level. Figure 1 shows these ranges of potency graphically.

Open image in new window

It is important to recognize that our proposed HRPT for the ERα agonist MoA does not depend on an assumption that isoflavones pose no health risks. Our proposed HRPT for this MoA is below the potency of the more potent botanical estrogens, including genistein, equol, and coumestrol (Table 1), which are thought to be responsible for any potential estrogenic health effects, to the extent that they exist. The fact that high isoflavone intakes have not been shown to produce clinically measurable adverse effects in adults or infants, as discussed above, argues for the conservativeness of our proposed HRPT.