Putative Adverse Outcome Pathway Development Based on Physiological Responses of Female Fathead Minnows to Model Estrogen versus Androgen Receptor Agonists

Abstract

Several adverse outcome pathways (AOPs) have linked molecular initiating events like aromatase inhibition, androgen receptor (AR) agonism, and estrogen receptor (ER) antagonism to reproductive impairment in adult fish. Estrogen receptor agonists can also cause adverse reproductive effects, however, the early key events (KEs) in an AOP leading to this are mostly unknown. The primary aim of this study was to develop hypotheses regarding the potential mechanisms through which exposure to ER agonists might lead to reproductive impairment in female fish. Mature fathead minnows were exposed to 1 or 10 ng 17α-ethynylestradiol (EE2)/L or 10 or 100 μg bisphenol A (BPA)/L for 14 d. The response to EE2 and BPA was contrasted with the effects of 500 ng/L of 17β-trenbolone (TRB), an AR agonist, as well as TRB combined with the low and high concentrations of EE2 or BPA tested individually. Exposure to 10 ng EE2/L, 100 μg BPA/L, TRB, or the various mixtures with TRB caused significant decreases in plasma concentrations of 17β-estradiol. Exposure to TRB alone caused a significant reduction in plasma vitellogenin (VTG), but VTG was unaffected or even increased in females exposed to EE2 or BPA alone or, in most cases, in mixtures with TRB. Over the course of the 14-d exposure, the only treatments that clearly did not affect egg production were 1 ng EE2/L and 10 μg BPA/L. Based on these results and knowledge of hypothalamic–pituitary–gonadal axis function, we hypothesize an AOP whereby decreased production of maturation inducing steroid leading to impaired oocyte maturation and ovulation, possibly due to negative feedback or direct inhibitory effects of membrane ER activation, could be responsible for causing adverse reproductive impacts in female fish exposed to ER agonists.

1. Introduction

An endocrine disrupting chemical (EDC) has been defined as ‘an agent that interferes with the synthesis, secretion, transport, binding, or elimination of natural hormones in the body that are responsible for the maintenance of homeostasis, reproduction, development and/or behavior’ (Kavlock et al., 1996). EDCs either can be natural or synthetic and can derive from a wide range of uses, including pesticides in agriculture and pharmaceuticals in healthcare. It has been well established that EDCs are pervasive in many environments with the potential to adversely affect both humans and ecologically relevant species (Hotchkiss et al., 2008; Kabir et al., 2015). Both the European Union and United States have developed programs designed to assess risks and support the regulation of EDCs. A critical aspect of most extant or anticipated regulatory programs for EDCs require some degree of evidence demonstrating that an observed adverse effect can be mechanistically linked to evidence of endocrine disruption (Kassotis et al., 2020; Matthiessen et al., 2017). Charged originally with screening pesticides (and later many more chemicals) for possible endocrine-disrupting activities, in 1996 the US Environmental Protection Agency (EPA) chartered the Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC) to support their Endocrine Disruptor Screening Program (https://www.epa.gov/endocrine-disruption/endocrine-disruptor-screening-and-testing-advisory-committee-edstac-final). The EDSTAC proposed a tiered testing of EDCs for potential effects in humans and wildlife which, while technically robust, is time-consuming and resource and animal intensive (Borgert, et al., 2011; Coady et al., 2017; Willett, 2011). Consequently, there has been a strong motivation to expand the predictive utility of in vitro high throughput and in silico methods, sometimes termed new approach methodologies (NAMs), to support regulatory decision-making with respect to EDCs (Browne et al., 2015; Coady et al., 2017; Judson et al., 2020; Kleinstreuer et al., 2018; Mansouri et al., 2020). A challenge to the use of NAMs, however, involves making the link between molecular and biochemical measures (or predictions) of possible changes in endocrine function and actual adverse effects. The adverse outcome pathway (AOP) framework (Ankley and Edwards, 2018; Ankley et al., 2010a) provides the basis for achieving this through definition of causal associations between key events (KEs) spanning biological levels of organization, ranging from initial molecular changes to adverse outcomes in individuals and populations (Browne et al., 2015; Coady et al., 2017; Knapen et al., 2015; 2020).

Several AOPs have linked molecular/biochemical changes caused by endocrine-active chemicals to reproductive impairment in sexually mature fish. For example, there are AOPs for aromatase (steroid synthesis) inhibitors (AOP 25; https://aopwiki.org/aops/25), androgen receptor (AR) agonists (AOP 23; https://aopwiki.org/aops/23), and estrogen receptor (ER) antagonists (AOP 30; https://aopwiki.org/aops/30). There is strong evidence that ER agonists can also cause reproductive disruption in adult fish (e.g., Kidd et al., 2007; Kramer et al., 1998; Matthiessen and Sumpter, 1998). However, intermediate KEs causally linking ER activation to impaired reproduction in female fish have not been documented, which represents an important knowledge gap relative to AOP development. Given the prevalence of estrogens/estrogenic activity in the environment (e.g., Blackwell et al., 2019; Desbrow et al., 1998; Leusch et al., 2018; Snyder et al., 2001; Ternes et al., 1999), AOPs for this class of EDCs could enhance the use of mechanistic data for decision-making. The goal of the present study was to develop hypotheses for the mechanism(s) through which exogenous ER agonists might lead to reproductive impairment in female fish. This serves as an initial step toward identifying and expanding the potential intermediate KEs for associated AOPs and developing evidence for biological plausibility linking these KEs to one another. Additional measurable biological responses along the pathway causally linking ER agonism to reproductive impairment in female fish would have utility as part of tiered testing strategies for the risk assessment of estrogenic chemicals. Finally, we were able to use data from the current study to assess the utility of a hypothalamic-pituitary-gonadal (HPG) axis model (Li et al., 2011) for predicting the impacts of chemicals on plasma vitellogenin (VTG, egg yolk protein) and 17β-estradiol (E2) concentrations. Plasma E2 and VTG are important components of several AOPs (e.g., AOPs 23, 25, 30; https://aopwiki.org), and further assessment of the Li et al. (2011) model supports the development of quantitative AOPs for single chemicals and mixtures (Conolly et al., 2017).

2. Methods

2.2. Overview

Fay et al. (2017) described different approaches for the derivation of AOPs, including those that feature “bottom-up” (starting with a molecular initiating event) and “top-down” (starting with a defined adverse outcome) strategies. In the current analysis, both the initiating event (ER activation) and outcome of concern (reduced egg production) are known, but the intermediate KEs connecting the two are not. To start exploring this connection, critical endocrine KEs (e.g., plasma E2 and VTG concentrations, reproductive output) were measured in female fathead minnows (Pimephales promelas) exposed to various EDCs. Over two experiments, fish were exposed to either a potent, highly specific ER agonist (17α-ethinylestradiol [EE2], Hannah et al., 2009; Lӓnge et al., 2001; Pawlowski et al., 2004) or a somewhat less potent (and specific) ER agonist, bisphenol A (BPA; Ma et al., 2019). Results observed for the estrogenic compounds were compared to those in fish exposed to 17β-trenbolone (TRB), a strong AR agonist which acts through a well-documented existing AOP (AOP 23; https://aopwiki.org/aops/23). Additionally, effects of the individual endocrine-active chemicals were compared to their combined effects in a mixture. In combination with existing knowledge of the HPG axis, the comparison was intended to aid development of a putative AOP linking ER agonism to reproductive toxicity in fish, which could be evaluated through subsequent targeted testing.

2.3. Chemical Preparation and Test Organisms

Fathead minnows were obtained from an on-site culture unit at the US EPA Great Lakes Toxicology and Ecology Division (Duluth, MN). Fish were reproductively mature, and 5–6 months old. All laboratory procedures involving fathead minnows were reviewed and approved by the Animal Care and Use Committee in accordance with the Guide for the Care and Use of Laboratory Animals (National Research Council, 2011). Stock solutions of TRB, EE2, and BPA (purity >95, 98, and 99%, respectively; Sigma Aldrich, St. Louis, MO, USA) were prepared in filtered Lake Superior water, without the use of carrier solvents, by continuous stirring in 20 L glass carboys.

2.4. Experimental Design

Experiment 1 (EE2-TRB) –

Fish were exposed for 14 d to control (Lake Superior water), 500 ng TRB/L, low (1 ng/L) or high (10 ng/L) concentrations of EE2, or a mixture of EE2 low/TRB and EE2 high/TRB. There were six tanks per treatment group, each divided in two by a mesh screen. One male and one female fathead minnow were randomly paired and added to each side of the mesh screen of the tank (n=12 males and females per treatment) and provided a breeding substrate. Because each breeding pair was separated, the number of eggs produced by individual females could be tracked. The test concentration of TRB was selected based on a 21-d reproduction study with fathead minnows where significant effects on egg production and plasma E2 and VTG levels in females were noted at 500 ng/L (Ankley et al., 2003). The 10 ng/L concentration of EE2 was based on treatments used in prior studies that had consistently induced VTG in male fathead minnows (Ekman et al., 2008), while the low concentration was expected to yield a more modest induction of VTG in the fish (Armstrong et al., 2016; Flick et al., 2014).

Experiment 2 (BPA-TRB) –

Fish were exposed for 14 d to control (Lake Superior water), 500 ng TRB/L, low (10 μg/L) or high (100 μg/L) concentrations of BPA, or a mixture of BPA low/TRB and BPA high/TRB. The experiment employed a group spawning design in which there were four replicate tanks per treatment, with four males and four females present in each tank, along with four breeding substrates. In this design, breeding pairs were not separated, so egg production could only be tracked at the tank level for each group (n=4 per treatment). However, 16 males and 16 females per treatment could be sampled for other endpoints such as plasma E2 and VTG. Bisphenol A concentrations were selected based on a concentration-response of VTG induction in carp (Mandich et al., 2007).

Other than the concentration and experimental design differences noted above, the exposure conditions and sample collection for Experiments 1 and 2 were the same. Fish were fed adult brine shrimp (Artemia) twice daily and held at 25±0.5 °C with a 16:8 light/dark photoperiod. Fish were acclimated to test conditions for 12 d before exposures began. During the acclimation period water quality parameters (e.g., dissolved oxygen, pH, temperature), mortality and egg production were recorded. Exposures began by delivering the chemicals dissolved in Lake Superior water to the test tanks at a flow rate of approximately 45 mL/min. Fish were inspected for survival daily, and any eggs deposited on the breeding substrates were gently rolled off the tiles, counted, and the tiles replaced.

After the 14 d of exposure, fish were anesthetized using buffered tricaine methane sulfonate (MS-222 100 mg/L; sodium bicarbonate, 200 mg/L). Following anesthetization, blood was collected from the caudal vasculature using microhematocrit tubes. Plasma was isolated by centrifugation and stored at −80 °C until further analysis. Fish were weighed and nuptial tubercle scores were recorded using methods described elsewhere (Jensen et al., 2001; USEPA, 2002). Ovaries were removed and weighed before being divided into several pieces. About 15 mg of ovary tissue was saved for an ex vivo steroid production assay.

2.5. Steroid and Vitellogenin Measurements

Ex vivo ovarian steroid production was determined using methods previously described by Martinovic et al. (2008), which were originally adapted from McMaster et al. (1995). Briefly, approximately 15 mg subsample of ovary from each fish was transferred to glass culture tubes (Fisher Brand 12×77 mm; 14-961-25) on ice with 500 ml of Medium 199 (M2520; Sigma) with 0.1 mM isobutylmethylxanthine (IBMX; Sigma I7018) and 1 mg 25-hydroxycholesterol (Sigma H1015)/mL. After incubation at 25 °C for 12 h, media was removed and stored at −20 °C until steroid hormones were extracted and analyzed. Controls containing supplemented medium without gonad tissue were also incubated and analyzed as blanks. Liquid-liquid extraction with diethyl ether was used to extract target steroids (E2, testosterone [T]) from the medium and quantified using radioimmunoassay as describe in Jensen et al. (2001) and USEPA (2002).

Plasma E2 and T in the females were extracted and quantified using the same techniques as for the ex vivo assay (Jensen et al., 2001). When plasma volume was limited, only E2 was quantified. Quantification of plasma VTG was achieved using an enzyme-linked immunosorbent assay with a polyclonal antibody to fathead minnow VTG, and purified fathead minnow VTG as a standard (Korte et al., 2000; USEPA, 2002).

2.6. HPG Axis Model Simulations

The HPG axis model developed by Li et al. (2011) was used to predict the expected impacts of EE2 and TRB, as well as a binary mixture of the two steroids, on plasma E2 and VTG. Model simulations used the mean measured concentrations of EE2 and TRB in water and the distribution of whole body and gonad weights of females in Experiment 1. One thousand simulations were performed using the best-fit parameter sets from Li et al. (2011) for each fish in each treatment group and the prediction of E2 and VTG plotted against the empirical results. Experiment 2 was not simulated.

2.7. Statistical Methods

All statistical analyses were performed using GraphPad Prism 9.0.2. The Brown-Forsythe test was used to evaluate homogeneity of variance. Significant differences between treatments for physiological data (steroids, VTG) were determined using a one-way analysis of variance. When the Brown-Forsythe tests indicated that variances were not homogeneous, the Kruskal-Wallis Test was used. Multiple comparison testing (Šidák’s or Dunn’s Tests) was then employed to compare all treatments. Spawning data were square root transformed before statistical tests were performed. Differences among treatments were considered significant at p ≤ 0.05. Given the differences in exposure design between the experiments (pair-spawning versus group spawning) statistical comparisons were limited to comparison within each experiment.

2.8. Exposure Verification

Water concentrations of the test chemicals were measured periodically over the course of both Experiments 1 and 2. Concentrations of EE2 were determined using solid-phase extraction (SPE) and liquid chromatography-mass spectrometry (LC/MS). Water samples (100 mL) were taken from odd numbered replicate tanks from each EE2 treatment on test days 1 and 9. Even numbered replicate tanks were sampled on days 6 and 13. Each treatment not containing EE2 was sampled once during the experiment. Samples were loaded onto a Strata X, 3 ml, 200 mg SPE column (Phenomenex, Torrance, CA, USA) at a flow rate of 5 mL/min. Columns were rinsed twice with analytical grade water, then dried under vacuum. Columns were eluted with methanol (4 × 1 ml). The methanol fraction was brought to dryness using nitrogen and resuspended in 10% methanol/water for LC/MS analysis. Sample extract (100 μl) was injected onto a Zorbax SB C-18 2.1 × 75 mm (Agilent, Wilmington, DE, USA) column at 30 °C. The mobile phase composition began at 50% methanol/water and a linear gradient method finished at 95% methanol/water in 15 min at a flow rate of 200 μl/min. Positive ion mass spectral data were acquired using a photoionization source with a toluene dopant at the flow rate of 50 μl/min. Quantifications were made using the 279 ion (M-H2O+H). The detection limit was 1.0 ng/L. Average (±SD) spike recoveries were 65±11% (n=8). Average (± SD) duplicate agreement for EE2 water measurements was 92±9.1% (n=8).

Water concentrations of BPA were measured using high-performance liquid chromatography (HPLC) with fluorescence detection (Ankley et al., 2010b). Tank water was directly injected onto a Synergi Hydro RP 2.0 × 500 mm column (Phenomenex) and eluted isocratically at 40 °C using 40% 0.015 M phosphate buffer (pH 3.3)/acetonitrile at a flow rate of 200 μl/min. Fluorescence at 275 nm excitation and 300 nm emission was used for quantification, with a detection limit of 0.1 μg/L. Average (±SD) spike recovery was 98±1.6% (n=10). Average (± SD) duplicate agreement was 99±0.4% (n=20).

Direct injection HPLC with fluorescence detection was used to determine water concentrations of TRB (Ankley et al., 2003; 2010b). Approximately 100–500 μL of exposure water was sampled from each tank containing TRB five times throughout each experiment (only half of the tanks for each TRB treatment were sampled on day 2 in experiment 1) and tanks not containing TRB were sampled once during the experiments. Samples were injected into a Synergo-Hydro RP-C18 column (3 × 75 mm; Phenomenex), and isocratically eluted using 70% methanol/water with a flow rate of 0.25 mL/min at 30 °C. TRB was detected using excitation and emission wavelengths of 364 and 460 nm, respectively. An external standard method of concentration quantification was used with a detection limit of 150 or 300 ng/L for Experiments 1 and 2, respectively. Average (±SD) duplicate agreement for TRB water measurements across all treatments containing TRB was 94±7.0% (n=15) in Experiment 1 and 92±8.4% (n=15) in Experiment 2. Spike recoveries were 94±8.9% (n=5) in Experiment 1 and 105±9.6% (n=5) Experiment 2.

3. Results

3.1. Experiment 1 (EE2-TRB)

Temperature and dissolved oxygen were recorded daily, and pH was measured every 3–4 d. Overall mean ±SD for water quality parameters from Experiment 1 were: temperature 24.9±0.19 °C, dissolved oxygen 6.4±0.24 mg/L, and pH 7.9±0.05. There were no substantive differences in these parameters among treatment groups or over the course of the experiment.

The concentration of TRB in Experiment 1 was initially below the target concentration (500 ng/L) with an average across six treatment tanks of 184±13 ng/L (mean±SD). Following adjustment of stock solution delivery rates on day 1, the concentration for the remainder of the experiment increased to 432 ng/L on day 2 and averaged 358±99 ng/L across all treatment tanks and sampling days over the course of the test (Table 1). EE2 low concentrations averaged 1.7±0.4 ng/L, while EE2 high concentrations averaged slightly below the target of 10 ng/L at 7.8±1.2 ng/L (Table 1). The EE2 and TRB concentrations in the mixture treatment groups were, on average, 13–23.5% greater than in the corresponding single chemical treatments (Table 1). No TRB or EE2 was detected in either control tanks or tanks not intentionally dosed with the chemicals.

Table 1:

Measured water concentrations (ng/L) of 17α-ethynylestradiol (EE2) and 17β-trenbolone (TRB) during 14-day exposure (Experiment 1).

	Water concentration 17α-ethynylestradiola
	Day of exposure
Treatment	1	2	6	9	13	Meanb
Control	NDc	NSd	ND	ND	ND
EE2 Low	2.3 (0.2)	NS	1.5 (0.1)	1.4 (0.3)	1.6 (0.2)	1.7 (0.4)
EE2 High	6.8 (0.4)	NS	7.1 (0.2)	9.4 (0.6)	8.0 (1.0)	7.8 (1.2)
EE2 L+TRB	2.2 (0.2)	NS	1.7 (0.1)	2.2 (0.3)	2.4 (0.1)	2.1 (0.3)
EE2 H+TRB	6.8 (0.6)	NS	7.9 (0.5)	10.5 (0.2)	10.2 (0.7)	8.8 (1.7)
	Water concentration 17β-trenbolonee
TRB 500	184 (13)	432 (19)	390 (51)	419 (9)	403 (18)	358 (99)
EE2 L+TRB	305 (22)	460 (13)	444 (20)	460 (17)	469 (9.5)	424 (67)
EE2 H+TRB	259 (52)	468 (21)	461 (18)	467 (22)	463 (19)	419 (91)

^aConcentrations are the mean of samples taken from three of six replicate tanks for each treatment.

^bMean (standard deviation) of all measurements over the 14 day exposure

^cND: Not detected (detection limit 1.0 ng/L)

^dNS Not sampled

^eConcentrations are the mean from all six exposure tanks except day 2 (sampled from three exposure tanks

There were two mortalities over the course of the 14-d exposure: one control female and one high EE2 female. Cumulative eggs per female calculations included data from these females until they died, at which point the total number of females was reduced to account for mortality. Fourteen d of exposure to EE2 (low or high) did not significantly impact fecundity, although there appeared to be a modest reduction in the number of eggs produced in fish in the high (10 ng/L) EE2 treatment group (Fig. 1A). Egg production was significantly reduced in all the TRB treatments, both alone and in combination with low or high EE2 (Fig. 1A).

Figure 1.

Fecundity in fathead minnows exposed to (A) 17α-ethynylestradiol (EE2), 17β-trenbolone (TRB) or their mixtures, or (B) bisphenol A (BPA), TRB, or their mixtures for 14 days. Bars represent mean ± standard error of the mean (n = 11–12 pairs [A] or 4 replicate tanks containing 3–4 females [B] per treatment; * p < 0.1; ** p < 0.05; *** p < 0.01).

Ex vivo T production by ovarian tissue was not affected by EE2 but was significantly reduced by all treatments that included TRB (Fig. 2A). Exposure to the mixture of TRB and the high concentration of EE2 caused a significant reduction in ex vivo production of E2 by ovary tissue (Fig. 2B). Ex vivo production of E2 in the high EE2 treatment was reduced considerably compared to controls but was just above the threshold of statistical significance (p=0.0505; Fig. 2B). The potential veracity of this reduction was supported by a statistically significant decrease in plasma E2 concentration in females from the high EE2 treatment (Fig. 2C). All treatments containing TRB also caused significant decreases in plasma E2 (Fig. 2C). The greatest contrast between the responses of the fish to EE2 and TRB involved plasma VTG. Vitellogenin concentrations in EE2-treated females appeared to increase in a dose-dependent manner relative to the controls; while this increase was not statistically significant it contrasted sharply from the decrease in plasma VTG caused by TRB alone (Fig. 2D). Notably, a significant increase in plasma VTG was observed for fish exposed to a mixture of TRB and the high concentration of EE2.

Figure 2.

(A) Ex vivo production of testosterone (T), (B) 17β-estradiol (E2), and (C) plasma concentrations of E2 and (D) vitellogenin (VTG) measured for female fathead minnows exposed to 17α-ethynylestradiol (EE2), 17β-trenbolone (TRB), or a mixture of the two for 14 d. Bars represent mean ± standard error of the mean (n = 10–12 per treatment). Columns with the same letter are not significantly different from one another (p < 0.05).

3.2. Experiment 2 (BPA-TRB)

In Experiment 2, water temperature was 24.9±0.14 °C, dissolved oxygen 6.02±0.38 mg/L, and pH 7.0±0.09, none of which differed notably among treatments.

Average water concentrations of TRB over the 14-d exposure were within 10% of the 500 ng/L target concentration in all treatments containing TRB (Table 2). There were very low, but quantifiable levels of BPA both in the control tanks and the TRB-only treatments, however, the observed mean concentrations (0.3 and 0.4 μg/L, respectively) were well over a magnitude lower than BPA low treatment. Average water concentrations of BPA were slightly (10–20%) higher than the 10 μg/L target concentration in the BPA low treatment groups (i.e., both with and without TRB), while measured BPA concentrations in the BPA high treatments were within 10% of the 100 μg/L target concentration (Table 2).

Table 2:

Measured water concentrations of bisphenol A (BPA; μg/L) and 17β-trenbolone (TRB; ng/L) during 14-day exposure (Experiment 2).

	Water concentration bisphenol Aa (μg/L)
	Day of exposure
Treatment	2	5	7	9	12	Meanb
Controlc	0.1 (0)	0.4 (0.04)	0.3 (0.2)	0.04 (0.04)	1.25 (0.2)	0.3 (0.4)
BPA Low	7.7 (0.4)	11.0 (0.3)	11.5 (0.5)	12.1 (0.6)	13.5 (0.3)	11.2 (2.0)
BPA High	69.9 (2.2)	111 (1.1)	104 (1.0)	112 (1.0)	111 (1.0)	102 (17)
TRBc	0.1 (0)	0.4 (0.3)	0.4 (0.08)	0.1 (0.1)	0.95 (0.2)	0.4 (0.3)
BPA L+TRB	11.2 (0.4)	12.3 (0.5)	12.4 (0.4)	12.8 (0.4)	14.4 (1.1)	12.6 (1.2)
BPA H+TRB	66.2 (2.2)	99.1 (1.1)	98.4 (0.6)	104 (1.3)	101 (1.2)	93.7 (14)
	Water concentration 17β-trenbolonea (ng/L)
TRB	512 (53)	550 (33)	522 (31)	552 (17)	550 (32)	537 (35)
BPA L+TRB	522 (17)	558 (26)	532 (28)	558 (24)	532 (37)	541 (28)
BPA H+TRB	460 (32)	495 (42)	438 (33)	488 (22)	493 (59)	475 (42)

^aConcentrations are the mean of samples taken from four replicate tanks for each treatment.

^bMean (standard deviation) of all measurements over the 14 day exposure

^cLimit of quantification = 0.1 μg/L; concentrations below limit of quantification were set to 0.025 for the generation of summary statistics.

There was no fish mortality in any treatment group. The greatest decreases in egg production occurred in the two TRB and BPA co-exposure groups (Fig. 1B). Fish in the TRB alone and high BPA treatments produced notably fewer eggs than the controls, but the data were sufficiently variable that the decreases were not significant (p=0.08 and 0.06, respectively). Cumulative egg production in the low BPA treatment was similar to controls (Fig. 1B).

As observed in Experiment 1, all treatments of the fish with TRB resulted in significantly reduced ex vivo ovarian production of T (Fig. 3A). Exposure to BPA alone reduced ex vivo ovarian production of both T and E2 in a concentration-dependent manner (Fig. 3A, B). Exposure to TRB alone did not significantly affect ex vivo E2 production, but when BPA was combined with TRB, ovarian production of E2 was again decreased in a concentration-dependent fashion (Fig. 3B). Exposure to TRB alone and in combination with BPA significantly reduced plasma E2 concentrations in the females (Fig. 3C). The high BPA treatment also reduced plasma E2 in the fish. As in Experiment 1, TRB alone significantly reduced plasma VTG concentrations in the females; in Experiment 2, TRB combined with low BPA also decreased plasma VTG (Fig. 3D). The high BPA treatment alone significantly increased VTG in the females and in the TRB-high BPA mixture group, BPA seemed to obviate the effects of TRB in that plasma VTG concentrations were similar to the controls (Fig. 3D).

Figure 3.

(A) Ex vivo production of testosterone (T), (B) 17β-estradiol (E2), and (C) plasma concentrations of E2 and (D) vitellogenin (VTG) measured for female fathead minnows exposed to bisphenol A (BPA), 17β-trenbolone (TRB), or a mixture of the two for 14 d. Bars represent mean ± standard error of the mean (n = 14–16 per treatment). Columns with the same letter are not significantly different from one another (p < 0.05).

3.3. HPG Axis Model Simulations

While the primary goal of the work described herein was to support AOP development through assessing experiments with known estrogens and an androgen, the resultant dataset also provided an opportunity to evaluate a multicompartment computational model previously developed by Li et al. (2011). This physiologically-based model was designed to predict the effects of estrogens, androgens, or their mixtures on plasma E2 and VTG in spawning fathead minnow females. Li et al. (2011) calibrated and then evaluated their model using control and treatment data (EE2 or TRB exposures) for fathead minnows from 18 different studies. Additionally, Li et al. (2011) simulated the expected response to a mixture of EE2 and TRB, but, at the time, there were no matched empirical data with which to evaluate the accuracy of the predictions for a mixture. In the present study, the model was used to simulate the conditions employed in Experiment 1, namely the same concentrations and exposure duration. In general, the model predicted plasma E2 concentrations in the control and TRB-exposed fish reasonably well (Supplementary Fig. S.1). However, the model generally over-predicted the effect of EE2 exposure on plasma E2 concentrations. Measured concentrations were generally greater than the model predicted (Supplementary Fig. S.1). The median model predicted VTG concentrations were generally higher than the observed plasma VTG concentrations. There was some overlap of the lower model predictions with the observed data but predicted VTG concentrations were generally higher than measured, except in the case of the TRB-EE2-high mixture (Supplementary Fig. S.1). Overall, while the model did perform reasonably from a qualitative perspective (i.e., showing similar trends), it did not accurately predict the empirical results, despite the extensive calibration. This indicates that more optimization and refinement (e.g., modifying the model structure/equations to include additional feedback mechanisms) is needed before it can be confidently used in predictive assessments such as the development of quantitative AOPs.

4. Discussion

4.1. Effects of Model Estrogens on Female Fathead Minnows

Strong ER agonists such as the synthetic steroid EE2 have been shown to alter several facets of reproductive endocrinology in a variety of fish species (e.g., Lӓnge et al., 2001; Parrott and Blunt, 2005; Parrott et al., 2003; Pawloski et al., 2004; Peters et al., 2007; Van den Belt et al., 2001; Xu et al., 2008; Zha et al., 2008). Many of these studies have been conducted using fathead minnows. For example, in a fathead minnow life-cycle study (Lӓnge et al., 2001), males exposed to 4.0 ng EE2/L did not develop normal secondary sex characteristics and only 50% of females at this same concentration reproduced successfully when paired with control males. In a 21-d study, female fathead minnows had a significant decrease in the mean number of spawned eggs when exposed to 10 ng EE2/L and a decrease in their gonadosomatic (GSI) index at 100 ng/L (Pawlowski et al., 2004). Pawlowski et al. (2004) also noted an increase in plasma VTG in females exposed to >1 ng EE2/L. A 21-d exposure of adult fathead minnows to 0.47 ng EE2/L resulted in a significant reduction in egg production when compared to a 14-d pre-test acclimation period with the same fish with no treatment (Armstrong et al., 2016). While we did not observe effects on egg production in fathead minnows exposed to 1 ng EE2/L, the higher treatment (10 ng/L) may have reduced egg production and did significantly affect other aspects of reproductive endocrinology in the females (e.g., plasma E2 levels). Collectively, these studies establish that EE2 affects reproductive endocrinology and egg production in adult female fathead minnows.

A variety of other model estrogens also can lead to reproductive impairment in female fish, including fathead minnows. For example, exposure to the potent natural estrogen E2 has been linked to inhibition of egg production, histological changes in ovaries, and reductions in GSI in fathead minnows (Damman et al., 2011, Kramer et al., 1998, Miles-Richardson et al., 1999). Bisphenol A is a less potent ER agonist than EE2 or E2, but nonetheless has been shown to affect aspects of reproductive endocrinology in both male and female fish (for review see Faheem and Bhandari, 2021), including fathead minnows. For example, Sohoni et al. (2001) reported that exposure of fathead minnow adults to BPA for up to 164 d deceased egg hatching success and egg production at water concentrations of 640 and 1280 μg BPA/L, respectively. Sohoni et al. (2001) also reported increased plasma VTG concentrations in females exposed to 640 μg BPA/L. Mihaich et al. (2012) conducted a 164-d exposure of fathead minnows to BPA; at a water concentration of 640 μg/L they found a decrease in early vitellogenic cells in the gonads of the females, although there were no significant effects on egg production. Mihaich et al. (2012) also reported that a water concentration of 64 μg BPA/L increased plasma VTG in females, which compares favorably to our observation of elevated VTG in females exposed to 100 μg/L. Overall, prior studies in conjunction with the present work, demonstrate that BPA can affect reproductive endocrinology and egg production in female fathead minnows.

Together, the experiments provided evidence that exposure to the potent and specific xenoestrogen EE2 as well as the less potent and less specific BPA could impact reproductive endocrinology and egg production in females. Nonetheless, males were also exposed to both xenoestrogens as well as TRB. As such, even though analyses and discussion of these results focused on females and female-relevant AOP development, it is acknowledged that some of the effects observed in females in the present study could be influenced by effects of the test chemicals on males. For example, some studies have suggested that exposure to EE2 at concentrations similar to those employed here can impact territorial behaviors and aggression in male fathead minnows (e.g., Majewski et al., 2002; Salierno and Kane 2009). Likewise, there is at least potential for alterations in pheromonal signaling. We did not explicitly examine male behaviors and/or other reproductive signaling in the present study. Consequently, it cannot be ruled out that some aspects of the effects observed in females may be mediated or modulated by impacts on the co-exposed males. Nonetheless it was clear that female physiology was impacted.

4.2. Consideration of the Experimental Data for AOP Development

Although ER agonists such as EE2 and BPA clearly can affect reproduction of female fish, there is not a sufficient understanding of how the chemicals impact HPG function to develop full AOPs linking ER activation to decreased egg production by females. An aim of the present study was to address this knowledge gap. Specifically, we undertook an analysis to examine data from a series of experiments conducted in our lab with fathead minnows that measured a range of responses across biological levels of organization, including sex steroid (E2, T) metabolism/status, VTG production, and egg output. These parameters are, or reflect, well-established KEs in several other AOPs capturing the effects of endocrine-active chemicals on reproduction in female fish (e.g., AOPs 23, 25, 30; https://aopwiki.org). Studies were conducted both with a strong ER agonist (EE2) and a chemical with lesser ER binding affinity (BPA). We also conducted exposures with the androgen TRB, which operates via a well-documented AOP linking AR activation to reproductive effects in female fish (AOP 23; https://aopwiki.org), as a contrast to the effects of the estrogens. To help further compare/contrast effects of ER versus AR agonists we conducted the experiments both with single chemicals and as binary (estrogen and androgen) mixtures.

Responses of the fish to TRB were largely consistent with expectations based on KEs and KE relationships in AOP 23 (https://aopwiki.org). In both experiments, the androgen reduced ovarian T production and plasma E2 concentrations, an observation consistent with the fact that T is the metabolic precursor to E2. The decrease in plasma E2 in the females corresponded with a substantial reduction in plasma VTG, which is produced in the liver of fish via activation of the ER(s) by E2. The reduction in VTG, in turn, resulted in decreased egg production by the female fathead minnows. One observation from the TRB studies that could be considered inconsistent with expectations was a lack of effects of the androgen on ex vivo E2 production at 14 d. However, extensive time-course studies with fathead minnows exposed to a variety of endocrine-active chemicals, including TRB, have shown that effects on ex vivo steroid production and changes in expression of genes coding for key steroidogenic enzymes generally occur early in an exposure and can be transient, or even cyclical, as the fish HPG axis undergoes compensation/adaptation (Ankley and Villeneuve, 2015).

Direct comparison of responses of the fish to EE2 versus BPA is complicated somewhat by differences in the potency of the compounds, as well as the fact that BPA can affect multiple endocrine pathways in fathead minnows. For example, Ekman et al. (2012) found that BPA appears to act both as an ER agonist and an AR antagonist in fathead minnow males. Nonetheless, there were similarities in responses of the fish to the two ER agonists. Both appeared to decrease ex vivo E2 production and, subsequently, plasma E2 concentrations. In addition, there were trends toward increased plasma concentrations of VTG in both EE2- and BPA-exposed females. Finally, both the ER agonists caused modest (albeit non-significant) reductions in egg production at their highest concentrations tested.

The most substantive difference in response of the fish to EE2 and BPA versus TRB involved plasma VTG. When the chemicals were tested individually the two ER agonists tended to increase VTG, while TRB significantly reduced plasma VTG concentrations. This occurred even though EE2 and BPA both reduced plasma E2 concentrations, and likely is a consequence of activation of ER(s) by the exogenous estrogens (Jones, 2000; Norris and Carr, 2020). The differential response of the fish in terms of VTG status carried through to the mixture studies. For most of the endpoints—ex vivo T production, plasma E2 concentration, and egg production—the effects of TRB, whether combined with EE2 or BPA, seemed to dominate observed responses in fish exposed to the various mixtures. That is, the profile of effects in the mixture experiments largely mirrored that seen in fish exposed to TRB alone. A notable exception involved VTG concentrations, where the high EE2-TRB treatment caused a marked increase in plasma VTG and the high BPA-TRB mixture resulted in VTG levels comparable to controls, indicating that BPA essentially negated TRB-induced decreases in VTG production.

4.3. Consideration of model predictions

The mechanistic model of Li et al. (2011) had mixed success simulating the results of experiment 1. Although the model was reasonably effective at simulating the plasma E2 and VTG concentrations in the control population and that exposed to TRB alone, the model did not accurately simulate the effect of the EE2 treatments in a quantitative sense. In particular, the model under-estimated plasma E2 concentrations in the EE2-exposed females, but at the same time generally over-estimated the plasma VTG concentrations. This result is somewhat counter-intuitive given that E2 is expected to drive VTG production. One may have expected both plasma E2 and VTG would have been under-estimated. For EE2-exposed fish, Li et al (2011) was calibrated with plasma VTG data only since plasma E2 was not measured in model calibration studies. The model accounts for both the direct effects of EE2 as well as impacts on endogenous hormones, e.g., the inhibition of testosterone production by bound ER. The inhibition constant was set at 0.016 based on the value used in a male FHM model (Watanabe et al., 2009). The model predictions suggest that the inhibition of testosterone production by bound ER is too large causing depression of plasma E2 that is too great at the tested EE2 concentrations. With respect to plasma VTG predictions, the results could reflect some over-estimation of EE2’s potency to induce VTG in the females. Regardless, the results highlight the challenges associated with developing effective mechanistic models for use in quantitative AOP applications (e.g., Conolly et al. 2017), particularly when the underlying biological events are incompletely understood.

4.4. Proposed Putative AOPs

Based on the experimental data, it appears that the role of feedback in the HPG axis in fish exposed to ER agonists has many similarities to that hypothesized for AR agonists (AOP 23), where excess levels of an exogenous androgen lead to a negative feedback response that reduces endogenous steroid (T, E2) production. The reduction in circulating E2 in AOP 23 leads to decreased VTG production by females and subsequent reductions in oocyte growth and egg production. In the present investigation, while EE2 and BPA both decreased E2 in a manner similar to TRB, plasma VTG concentrations were not reduced in the females. Overall, in the current study production of VTG occurred in females exposed to exogenous ER agonists for 14 d, but egg production still appeared to decrease. Consequently, egg production/release in female fish seemed to be affected by KEs not directly related to the concentration of VTG in the fish.

For fish reproduction to occur, ovulation and final oocyte maturation must take place (Lessman, 2009; Nagahama and Yamashita, 2008). Cytochrome P450 aromatase is rate limiting for production of estrogens by the gonad (Simpson et al., 1994), while 20β-hydroxysteroid dehydrogenase plays a critical role in catalyzing the production of progesterone derivatives that stimulate final oocyte maturation (Nagahama, 1997; Tanaka et al., 2002). Final oocyte maturation and ovulation rely on a shift from aromatase expression in the granulosa cells, supporting endogenous E2 production, to the expression of cytochrome P450 c17α-hydroxylase, 17, 20-lyase and/or 20β-hydroxysteroid dehydrogenase, which support the synthesis of maturation-inducing steroids from progesterone (Clelland and Peng, 2009; Nagahama and Yamashita, 2008). We hypothesize that the predominantly negative feedback signal elicited by the on-going presence of an exogenous estrogen or androgen, which leads to an overall reduction in endogenous steroid synthesis, also impairs the production of the maturation-inducing steroids required to trigger final oocyte maturation and ovulation. This then ultimately leads to reduced egg reproduction. So, although levels of VTG needed for egg production are sustained in female fish exposed to exogenous estrogens, reproductive output is nonetheless impaired, potentially due to reduction in final maturation of the oocyte.

Based on these results, we propose a putative AOP in which elevated activation of the ER by exogenous estrogen reduces gonadotropin-releasing hormone from the hypothalamus and/or gonadotropin release from the pituitary (Fig. 4). Altered secretion of gonadotropins is hypothesized to down-regulate expression of transcripts coding for steroidogenic enzymes, ultimately leading to reduced endogenous steroid production and potentially impairing a luteinizing hormone-mediated surge in 20β-hydroxysteroid dehydrogenase, thereby impacting the local synthesis and release of maturation inducing steroids (Fig. 4). This is expected to impair oocyte maturation and ovulation, ultimately leading to reduced egg production (Fig. 4). While our experimental data are consistent with this interpretation, the present study was not specifically designed to test the hypothesized AOP. For example, we do not have direct measurements of maturation-inducing steroids or oocyte maturation from the present study, nor were the transcripts for key steroidogenic enzymes such as 20β-hydroxysteroid dehydrogenase measured. Likewise, the specific mechanisms by which ER agonism elicits negative feedback on the HPG axis remain somewhat uncertain. These are endpoints that will be targeted in future investigations and may also shed additional insights into refinements to the HPG-axis model that could improve its performance.

Figure 4.

Putative adverse outcome pathways linking (A) estrogen receptor (ER) or (B) G protein estrogen receptor (GPER) activation to decreased female spawning. 20β-HSD = 20β-hydroxysteroid dehydrogenase.

There also is limited evidence suggesting the possibility of another AOP via which estrogen exposure may lead to impaired spawning in female fish. Specifically, work in several fish species including common carp (Cyprinus carpio), yellowfin seabream (Acanthopagrus latus), zebrafish (Danio rerio), and Japanese medaka (Oryzias latipes) suggest that actions of estrogens on G protein-coupled estrogen receptors (GPERs) on the membranes of fish oocytes may impair oocyte maturation (Jeng et al., 2020; Majumder et al., 2015; Miyaoku et al., 2021; Pang and Thomas, 2010). Thus, the potential actions of estrogens on the GPERs could be a contributor to lack of spawning. Unfortunately, the results of the current study again do not provide a way to discriminate or confirm these possibilities. However, the current experimental results, along with emerging knowledge concerning fish reproductive endocrinology, yield two hypothesized AOPs that can be evaluated in further experimentation. Additional understanding of intermediate KEs in the pathway(s) linking ER activation to reproductive impairment in female fish would help inform the development of additional assays and endpoints for EDC screening and risk assessment.