Twenty years of transcriptomics, 17alpha-ethinylestradiol, and fish

Abstract

In aquatic toxicology, perhaps no pharmaceutical has been investigated more intensely than 17alpha-ethinylestradiol (EE2), the active ingredient of the birth control pill. At the turn of the century, the fields of comparative endocrinology and endocrine disruption research witnessed the emergence of omics technologies, which were rapidly adapted to characterize potential hazards associated with exposures to environmental estrogens, such as EE2. Since then, significant advances have been made by the scientific community, and as a result, much has been learned about estrogen receptor signaling in fish from environmental xenoestrogens. Vitellogenin, the egg yolk precursor protein, was identified as a major estrogen-responsive gene, establishing itself as the premier biomarker for estrogenic exposures. Omics studies have identified a plethora of estrogen responsive genes, contributing to a wealth of knowledge on estrogen-mediated regulatory networks in teleosts. There have been ~40 studies that report on transcriptome responses to EE2 in a variety of fish species (e.g., zebrafish, fathead minnows, rainbow trout, pipefish, mummichog, stickleback, cod, and others). Data on the liver and testis transcriptome dominate in the literature and have been the subject of many EE2 studies, yet there remain knowledge gaps for other tissues, such as the spleen, kidney, and pituitary. Inter-laboratory genomics studies have revealed transcriptional networks altered by EE2 treatment in the liver; networks related to amino acid activation and protein folding are increased by EE2 while those related to xenobiotic metabolism, immune system, circulation, and triglyceride storage are suppressed. EE2-responsive networks in other tissues are not as comprehensively defined which is a knowledge gap as regulated networks are expected to be tissue-specific. On the horizon, omics studies for estrogen-mediated effects in fish include: (1) Establishing conceptual frameworks for incorporating estrogen-responsive networks into environmental monitoring programs; (2) Leveraging in vitro and computational toxicology approaches to identify chemicals associated with estrogen receptor-mediated effects in fish (e.g., male vitellogenin production); (3) Discovering new tissue-specific estrogen receptor signaling pathways in fish; and (4) Developing quantitative adverse outcome pathway predictive models for estrogen signaling. As we look ahead, research into EE2 over the past several decades can serve as a template for the array of hormones and endocrine active substances yet to be fully characterized or discovered.

1. Introduction

The past twenty years of endocrine disruption research has revealed that several veterinary and human pharmaceuticals are present in aquatic environments at concentrations sufficient enough to elicit adverse effects in a range of species, from the smallest microorganisms up to the largest of aquatic mammals. These waterborne pharmaceuticals can be detrimental to relationships within aquatic food webs [15] and can impact populations for generations [4, 113]. Arguably, no pharmaceutical has received as much scientific nor public attention as 17alpha-ethinylestradiol (EE2), which is used in the pharmaceutical industry as a surrogate for 17beta-estradiol (E2) in birth control pills. In the early 1990s, researchers became increasingly aware of how ubiquitous this pharmaceutical was in water systems, and over the past twenty years, significant efforts have been made in characterizing aquatic EE2 exposure, as well as improving its removal from wastewater treatment facilities in order to protect aquatic wildlife in receiving waters.

There is compelling evidence that EE2 exposure can lead to tissue damage [121], reproductive dysfunction [55], disrupted tissue steroidogenesis [69, 104], altered spawning [23], behavioral changes [94], and population level consequences, as noted by Kidd et al. (2007) in their seminal work in the Experimental Lakes Area [63]. Many of these aforementioned effects were noted at environmentally relevant concentrations of EE2 (less than 5 ng/L). While apical responses in fish to EE2 are well documented, omics studies continue to reveal new molecular and cellular insights into underlying mechanisms. Both research avenues highlight the legacy of EE2 and its negative effects in aquatic organisms, prompting a movement to address other pharmaceuticals present in the aquatic environment. Acknowledging this is important, we continue to produce new pharmaceuticals that not only impact reproduction, but also those that control blood pressure, impede cancer cell growth (i.e. antineoplastics), regulate lipids, and manage depression, to name but a few. In this mini review, we highlight research conducted on the omics of EE2-mediated toxicity in fish to identify knowledge gaps and future directions, and to define directions that research from which other pharmaceuticals can benefit.

2. A problem discovered: A brief history of EE2 in ecotoxicology

In the late 1980s, researchers recognized that estrogenic chemicals were present in various watersheds. Studies associated adverse effects of waterborne estrogenic chemicals with changes in gonad size and reproductive development in fish near pulp mills in Sweden [97] and Canada [82]. Initial studies investigating effluents demonstrated impacts at several points along the pituitary-gonad axis in fish such as the white sucker (Catostomus commersoni) [112], and estrogen receptor activation assays in cell lines revealed estrogenic responses to pulp and paper mill effluent [122]. Additional studies determined that some of these biological responses could be replicated with exposure to β-sitosterol, a plant sterol with estrogen-like activity [70], although the responses were not always identical to those from estrogens as sterols with estrogen-like activities also interfere with other hormonal systems in fishes, depending on the concentration and reproductive maturation of the individual [83]. Studies first started to look at linkages to vitellogenin (described below) and pulp and paper mill effluent in Finland in the late 1990s [103].

Parallel to the studies in Canada on pulp mills, studies in the United Kingdom were finding intersex in fish associated with sewage effluent discharges. A low incidence of intersex had been noticed back to the mid-1980s [111]. While early studies suspected estrogens (reviewed in [90]), initial attention focused largely on the role of detergents and alkylphenols [59], although it quickly expanded to examine other chemicals. By 1996–98, attention was shifting to estrogens in the effluent [110] and studies found widespread evidence of intersex in fish near sewage outfalls [40, 58]. Concerns are now global, agreed upon by many that xenoestrogens, specifically the pharmaceutical EE2, can exert widespread reproductive effects in aquatic wildlife [77].

The question morphed from incidence to biomarker development for estrogenic exposure and potential endocrine activity. Importantly, teleost fish produce vitellogenin in the liver, which travels through the blood to the ovary. Vitellogenin is largely under the control of estrogens and is eventually incorporated into growing oocytes via receptor-mediated actions [31, 100]. Vitellogenin and other lipids in the yolk sac provide a rich source of nutrients for embryos, and the early development of fish depends upon an adequate source of lipids for energy. Sumpter and Jobling [105] pointed out that this endogenous reproductive process in female fish could be used as a biomarker for estrogenic exposures in environments. This was a breakthrough in ecotoxicology - males also carry the genes to naturally produce vitellogenin, albeit at relatively low levels compared to females. However, when males are exposed to weak estrogens or estrogens at low concentrations, vitellogenin can be rapidly and actively transcribed at elevated levels. Thus, vitellogenin induction, both at the messenger and protein level, became a reliable and robust biomarker for estrogenic exposures; even today it remains the gold standard as a molecular indicator of exposure in aquatic organisms [3, 10, 25, 34, 109, 123]. Building upon this, researchers realized that additional estrogen-responsive genes needed to be identified to understand the molecular basis of estrogenic actions [54]. Over the past decade, a repertoire of estrogen-responsive genes has been revealed in teleost fish, and these responses have advanced our understanding of endogenous processes regulated by estrogens, as well as the potential health impacts of exogenous endocrine active substances.

3. Transcriptomics and adverse outcome pathways

Toxicology research over the years has shifted from measuring lethality using high doses of contaminants to investigating sub-lethal doses that potentially impact development, reproduction, health, and susceptibility to disease. This focus on mechanistic models of toxicology expanded at the same time that “omics” technologies were being developed in human medicine to provide a more comprehensive analysis of molecular impairments that lead to disease. The holistic approach that these methods provide spurred the idea of being able to link a molecular initiating event to downstream changes at the cellular, tissue, organismal, and population level for aquatic organisms. This new linkage paradigm, referred to as “Adverse Outcome Pathways (AOP)” [5], has reshaped how ecotoxicologists think about pharmaceuticals in the environment. In fact, it can be thought of as a new way of approaching effects-based toxicology.

Omics technologies today include broad methods that evaluate whole genomes (DNA), transcriptomes (mRNAs and non-coding RNAs), proteomes (proteins) and metabolomes (metabolites) that are altered upon interaction with a chemical substance. Omics technologies deliver specific and relevant information at the molecular level about how a compound interacts with its target, and many of the earlier case studies in environmental science involved endocrine active substances that were known to interact with nuclear receptors and induce gene transcription. The estrogen receptor itself is induced in the liver, along with a large number of other genes that are regulated by estrogens or estrogen mimics. The initial binding of an estrogen to the estrogen receptor sets a particular pathway in motion. Thus, it has been a broadly applied tool to identify estrogen responsive genes in multiple species and tissues.

Transcriptomics methods rely on genome-wide measurements of changes in the levels of mRNAs in tissues of exposed animals. For example, microarrays are made commercially, with the probes printed directly to the glass slides with knowledge of the coordinates for each. RNA that is extracted from controls and chemical treated organisms is obtained, copied into cDNA and labeled with a fluorescent probe and then hybridized to the arrays. The amount of fluorescent probe per spot on the array is used to quantify the amount of the message present in the tissue of origin. The ability to print thousands of cDNA or oligonucleotides onto a glass slide paved the way to generate a wealth of molecular data, increasing understanding of E2-responsive genes and networks in a variety of fish tissues. Only recently and within the past few years, researchers have moved toward the use of RNA sequencing (RNAseq) to study endocrine-mediated responses. RNAseq is a non-targeted method for determining changes in the transcriptome. In this method, total RNA is prepared, and mRNAs with poly A tails are sequestered from the total RNA by binding to poly dT oligonucleotides attached to magnetic beads. The mRNAs are then converted to cDNAs, fragmented and prepared for sequencing by Next Generation Sequencers. For this process, the Illumina sequencer is most often selected today due to sample cost and experimental throughput, although other platforms (e.g. PacBio) are becoming more popular. The analysis of the data includes matching the reads to a reference genome and then counting how many copies of each transcript is present in the output. Put together, both microarray and RNAseq technologies have advanced our understanding of EE2 action in fish, and we outline some of these efforts below.

4. Omics, 17alpha-ethinylestradiol, and fish

At the turn of the century, researchers began to apply gene array technology to study xenoestrogens. These earlier efforts involved the printing of cDNA molecules onto nylon membranes; radiolabeled pieces of cDNA would find their complementary targets on the nylon membrane, yielding a signal that was proportional to the expression levels of the cDNAs present in the sample. Typically, these membranes, or macroarrays, contained less than 50 genes and included transcripts expected to be responsive to xenoestrogens. These early macroarrays were applied to screen estrogens and EE2 in sheepshead minnow (Cyprinodon variegatus variegatus) [68] and plaice (Pleuronectes platessa) [16] and opened the door for more elaborate printing technologies onto glass slides (i.e. microarrays). In teleost fish, microarrays were also quickly leveraged to study the effects of EE2 in the mid-2000s. Microarrays containing 15,000 to 64,000 probes were used to assess environmental estrogens, quantifying molecular responses in context of phenotypic anchors. Taking a step back, we conducted a search using Pubmed in June of 2019 using the terms “fish + ethinyl estradiol + transcriptomics or microarray” and identified 39 published transcriptomic studies in fish; some of these first studies were published in 2006 and as of 2019, there were new reports published on the actions of EE2 at the transcriptome level (Table 1). These new studies leverage RNA-seq approaches, a more robust technology with higher sensitivity and accuracy compared to microarrays. Due to increased depth, RNAseq is expected to reveal new pathways and E2-responsive interactomes involving EE2 and other xenoestrogens.

Table 1:

Studies investigating 17alpha-ethinylestradiol in fish tissues using transcriptomics approaches (i.e. specifically microarray and next generation sequencing).

Species	Scientific Name	Year	Tissue	Sex	Marine or Freshwater	Life stage	Type of study	Method	Publication
Chub mackerel	Scomber japonicus	2019	liver	male	saltwater	prob adult based on size	lab	RNA-seq	[91]
Japanese medaka	Oryzias latipes	2019	whole body	male	freshwater	embryos-larvae	lab	RNA-seq	[1]
Pacific sardine	Sardinops sagax	2019	liver	male	saltwater	prob adult based on size	lab	RNA-seq	[91]
Rainbow trout	Oncorhynchus mykiss	2019	posterior kidney	unknown	freshwater	adults (160 days at least)	lab	RNA-seq	[9]
Atlantic cod	Gadus morhua	2018	liver slice exposure	male, female	saltwater	juveniles	lab	RNA-seq	[120]
Zebrafish	Danio rerio	2018	liver	male	freshwater	adult	lab	microarray and RNA-seq	[52]
Zebrafish	Danio rerio	2018	testis	male	freshwater	adult	lab	RNA-seq	[87]
Fathead minnow	Pimephales promelas	2017	liver	male	freshwater	adult	lab	microarray	[38]
Guppy	Poecilia reticulata	2017	whole brain	male, female	freshwater	adult	lab	RNA-seq	[96]
Rare minnow	Gobiocypris rarus	2017	gonads	male, female	freshwater	adult	lab	RNA-seq	[41]
Three-spined stickleback	Gasterosteus aculeatus	2017	liver	male	freshwater	likely mixed	wild-caught, lab exp.	microarray	[119]
Zebrafish	Danio rerio	2017	whole brain	male, female	freshwater	adult	lab	RNA-seq	[88]
Fathead minnow	Pimephales promelas	2016	testis	male	freshwater	adult	lab	microarray	[37]
Three-spined stickleback	Gasterosteus aculeatus	2016	testis	male	freshwater	adult	lab	microarray	[89]
Gulf pipefish	Syngnathus scovelli	2015	liver	male, female	saltwater	adult	wild-caught, lab exp.	RNA-seq	[92]
Korean rose bitterling	Rhodeus uyekii	2015	liver, skin	male	freshwater	adult?	wild-caught, lab exp.	RNA-seq	[65]
Rainbow trout	Oncorhynchus mykiss	2015	hepatocytes	unknown	freshwater	juvenile	lab	microarray	[53]
Rainbow trout	Oncorhynchus mykiss	2015	testis	male	freshwater	juvenile	lab	microarray	[27]
Rainbow trout	Oncorhynchus mykiss	2015	testis	male	freshwater	juvenile	lab	microarray	[26]
Largemouth bass	Micropterus salmoides	2014	liver, ovary	female	freshwater	adult	lab	microarray	[20]
Yellow catfish	Pelteobagrus fulvidraco	2014	gonads	male, female, super male		1 year old	lab	RNA-seq	[57]
Coho salmon	Oncorhynchus kisutch	2013	pituitary	female	saltwater	sub-adult	lab	RNA-seq	[46]
Mummichog	Fundulus heteroclitus	2013	liver	female	estuarine	adult	lab	microarray	[32]
Zebrafish	Danio rerio	2013	embryo	unknown	freshwater	embryos	lab	microarray	[99]
Japanese Medaka	Oryzias latipes	2012	testis	male	freshwater	adult	lab	microarray	[79]
Japanese Medaka	Oryzias latipes	2012	testis	male	freshwater	adult	lab	microarray	[48]
Stickleback	Gasterosteus aculeatus	2010	liver	male	freshwater	adult	wild-caught, lab exp.	microarray	[60]
Zebrafish	Danio rerio	2010	liver	male, female	freshwater	adults	lab	microarray	[24]
Fathead minnow	Pimephales promelas	2009	testis	male	freshwater	adult	lab	microarray	[42]
Rainbow trout	Oncorhynchus mykiss	2008	liver	unknown	freshwater	immature	lab	microarray	[49]
Zebrafish (lab wild-type strain)	Danio rerio	2008	brain, gonads	male, female	freshwater	mature	lab	microarray	[118]
Zebrafish	Danio rerio	2008	brain, gonads	male, female	freshwater	adult	lab	microarray	[117]
Common carp	Cyprinus carpio	2007	liver	unknown	freshwater	immature	lab	microarray	[81]
Rainbow trout	Oncorhynchus mykiss	2007	hepatocytes	male	freshwater	adult?	lab	microarray	[39]
Rainbow trout	Oncorhynchus mykiss	2007	liver	male, female	freshwater	juvenile	lab	microarray	[45]
Zebrafish	Danio rerio	2007	liver, telencephalon	male	freshwater	adult	lab	microarray	[74]
Zebrafish	Danio rerio	2007	gonads	male, female	freshwater	adult	lab	microarray	[98]
Common carp	Cyprinus carpio	2006	liver	unknown	freshwater	juvenile	lab	microarray	[80]
Goldfish	Carassius auratus	2006	brain	male	freshwater	adult?	lab	microarray	[75]
Rainbow trout	Oncorhynchus mykiss	2006	hepatic	male	freshwater	adult, juvenile	lab	microarray	[101]

The species that have been studied for their response to EE2 exposure include both freshwater and marine, for example goldfish (Carassius auratus), zebrafish (Danio rerio), flounder (Platichthys flesus), Atlantic cod (Gadus morhua), stickleback (Gasterosteus aculeatus), mummichog (Fundulus heteroclitus), rainbow trout (Oncorhynchus mykiss), and largemouth bass (Micropterus salmoides), among others (Figure 1A; Table 1). However, more than 85% of the transcriptomics studies with EE2 have been conducted in freshwater species (e.g., zebrafish (Danio rerio), guppy (Poecilia reticulata), rare minnow (Gobiocypris rarus), and largemouth bass (Micropterus salmoides) compared to saltwater species such as Pacific sardine (Sardinops sagax), Atlantic cod (Gadus morhua), gulf pipefish (Syngnathus scovelli), sheepshead minnow (Cyprinodon variegatus), and mummichog (Fundulus heteroclitus). Moving forward, additional data to address how saltwater species are affected by pharmaceutical estrogens like EE2 are required to fill gaps in our understanding of estrogen-mediated gene networks. Physiological differences in ion regulation and hormone signaling are reported between freshwater and saltwater fish and this may translate into sensitivity differences in gene expression to pharmaceutical estrogens [2, 50]. What is impressive is that an array of species has been investigated at the transcriptomics level, and this presents a rich comparative perspective of EE2-induced gene expression. Moreover, there have been a range of tissues investigated at the molecular level following exposure to EE2, the most predominant tissues being the liver and the testes (Figure 1B). Moving forward, additional data on spleen, kidney [9], pituitary [46] and early embryogenesis would be needed to strengthen understanding of the molecular pathways altered by EE2 in addition to revealing novel mechanisms for estrogens. A last point to make is that, while the majority of studies are in male fish, there are some data available on female fish, which offer a unique perspective on sex-specific responses. EE2 has been studied in female zebrafish, Atlantic cod, largemouth bass and mummichog (listed in Table 1), which provides useful information on sex differentiation, as well as sex-specific biomarkers for estrogenic exposures.

Figure 1:

Percent of studies reporting on transcriptomics responses to 17alpha-ethinyestradiol (EE2) in (A) fish and (B) tissues. Rainbow trout and zebrafish have been the dominant species studied, while the liver and the testis are often the most studied tissues when investigating molecular responses to EE2.

5. Transcriptomics reveals mechanisms of 17alpha-ethinylestradiol action in fish

EE2 induces both intersex and sex change in fish [28, 121], but the mechanisms by which EE2 induces these conditions, both in the laboratory and in the field, have been a significant question by scientists globally [8]. In one study, Feswick et al. [37] used a short-term (96 hour) exposure of male fathead minnows to environmentally-relevant levels (15 ng/L) of EE2, to identify early transcriptional changes potentially related to the initiation of intersex. Exposed males did not exhibit any significant change in testes morphology nor gonadosomatic index during the short exposure, but they did show a reduction in both testosterone and 11keto-testosterone. Transcriptomic profiling in the testes revealed that gene networks associated with male reproduction (e.g. sperm motility, insemination, male sex determination) were rapidly suppressed with EE2 exposure while gene networks related to female reproduction (e.g. ovary function, ovary follicle development, and granulosa cell development) were rapidly increased following exposure to EE2. Moreover, networks involved in steroid biosynthesis and steroid metabolism were suppressed. Gene networks centered around key transcription factors such as foxl2, which signals ovarian follicle development and granulosa cell development, as well as dmrt1 and sox9, two key proteins in male sex determination, were rapidly regulated by EE2. Studies such as these are important because they begin to identify molecular initiating events prior to the appearance of intersex.

Feswick and colleagues [36, 38] also conducted a broad scale study using EE2 to compare interlaboratory reliability and reproducibility for gene expression data. Laboratories were tasked to identify estrogen-responsive genes and networks in fathead minnow liver following exposure to EE2. The objective was to identify E2-responsive genes in the liver that could yield new insights into the regulation of hepatic physiology and to identify reliable biomarkers for EE2 exposure (i.e., those identified in all laboratories). Microarrays revealed EE2 exposure increased processes such as protein folding and amino acid activation in the liver, whereas gene networks associated with blood clotting and coagulation, as well as the alternative and classical complement activation pathways, triglycerides storage, and xenobiotic clearance and metabolism were decreased by EE2. Laboratories were consistent in identifying a number of estrogen-responsive genes in the liver, including apolipoprotein E, apolipoprotein A1, insulin growth factor 1, x-box binding protein 1, and estrogen receptor alpha. Notably, there was high variability, both biological and technical, for vitellogenin, which can impede interpretation of the data [38]. High variability in vitellogenin has been observed by others [11]. Jastrow and colleagues [56] recently presented some guidelines for addressing variability in the vitellogenin biomarker with estrogenic exposures.

From an environmental perspective, the realization that specific genes show consistent and reproducible responses to EE2 leads to the question of whether E2-responsive gene networks are more appropriate for biomonitoring programs, compared to individual biomarkers such as vitellogenin. Based on these data from the interlaboratory study, and those collected from the Comparative Toxicogenomics Database, an estrogen-responsive network was developed (including estrogen receptor alpha, transferrin, myeloid cell leukemia 1, insulin like growth factor 1, and methionine adenosyltransferase 2A, among other genes) [38]. Thus, considering multiple lines of evidence, we continue to hone our knowledge of estrogen-responsive gene networks, and these networks have been proposed in environmental monitoring programs to improve decision-making and to increase our ability to detect estrogens in the environment [73]. However, in order to better characterize risk from endocrine active substances, including xenoestrogens, an increasing reliance on integrative and computational approaches may be needed within the context of the adverse outcome pathway framework.

Knowledge as to the role of xenoestrogens in the process of both intersex and sex reversal is not limited to the gonads, and researchers now appreciate the role that estrogens play in shaping the central nervous system of fishes. In the guppy, RNASeq was used to document the effect of 8 ng/L and 38 ng/L EE2 on the brain transcriptome of both males and females [95]. Not surprisingly, the male brains exposed to EE2 exhibited gene expression changes which were more similar to female brains, suggesting a feminizing effect. While the researchers did not conduct pathway analysis nor gene set enrichment, the study nevertheless revealed that a significant number of genes affected in the brain were related to glutamate and nuclear receptor signaling, chromatin organization, and LINE transposable elements, which showed treatment and sex specific responses.

Exposure of males to EE2 during development has revealed that even transient, low-level (environmentally relevant) exposure during critical developmental periods can have irreversible reproductive consequences into adulthood. For example, using RNAseq, a host of genes associated with spermatogenesis, steroid synthesis, and testis development and function were differentially expressed in zebrafish exposed to 1.2 and 1.5 ng/L EE2 from fertilization to 80 days of age followed by depuration for 82 days [87]. Other biological processes affected by EE2 in the study included lipid and carbohydrate metabolic process, protein and nucleic acid metabolic processes, gene regulation, response to hormone, response to stress, and circadian rhythm. These processes are therefore hypothesized to be related to lower fertility in male adult zebrafish that persist over time, in the absence of EE2. Many of the these biological processes are consistent with those reported in other studies investigating effects of EE2 [37]. The study by Porseryd et al. [87] suggests that non-coding sequence perturbations by EE2 is a potential mechanism of disruption in fish, and this would not have been revealed without transcriptome data. The role of non-long coding RNA in estrogen-mediated responses is anticipated to be a new avenue of research in fish endocrinology.

6. Non-genomic signaling by estrogens

There is a body of evidence showing that vertebrates, including fish, have membrane receptors for estradiol and other sex steroids, in addition to soluble nuclear receptors [106]. These receptors bind their ligands with very high affinity (Kd’s in the 1–5 nM range) and have very specific functions in reproduction. Data on the characterization of specific membrane receptors for estradiol, progestins, and testosterone have been presented previously [19, 61, 62, 86].

Initial studies by the Thomas laboratory characterized the progestin membrane receptor alpha (mPRa) and showed that it regulated oocyte maturation and in fact was the maturation inducing steroid (MIS) 20β–S receptor [124]. In addition to effects in the ovary, MIS also is responsible for inducing sperm hypermotility in males, and thus the mPRa receptor has also been found in the membranes of sperm. Specific membrane receptors for testosterone were also identified [13].

Membrane estradiol receptors have also been characterized in fish. The receptor identified by the Thomas group was GPER, also known as GPR30 [85]. This receptor is found on the surface of fish oocytes and when it is bound by estradiol shows inhibitory action on oocyte maturation. It is possible that a truncated version of ERa may also be expressed tethered to membranes.

A general characteristic of the membrane receptors is that they are composed of 7 transmembrane segments and are part of a G-protein superfamily. They signal through activation of intracellular second messenger pathways and the signaling occurs very quickly. Through in situ hybridization, membrane steroid receptors have been found in several tissues, including the brain [71].

In a recent experiment, fathead minnow were exposed to 5 ng/L EE2 or 100 ng/L levonorgestrel (the progestin portion of the birth control pill) for 30 min to evaluate signaling cascades using a phosphoproteomics approach [102]. Changes in phosphorylation patterns of brain proteins were distinct for the two chemicals, with some overlap. Both estradiol and the progestin altered phosphorylation patterns in proteins generally involved in neurogenesis and synaptic activity, but the specific group of proteins affected by each was different. In some cases, the directionality of phosphorylation on the same protein was in opposite directions by the two chemicals. Each chemical also showed some unique phosphorylation pattern cascades. EE2 was involved mostly in neuronal processes and neuroprotection, while levonorgestrel altered phosphorylation patterns for proteins involved in axon cargo transport and calcium signaling.

It is intriguing to understand how soluble nuclear receptors and their membrane counterparts work together to signal tissues in fish brains and reproductive tissues. Clearly, more research is required to get the complete picture of how they support each other to maintain homeostasis and control reproduction.

7. Computational endocrinology: New view of estrogens

Various data sources are available to examine estrogen-mediated effects, and it is important to leverage data incorporating reliable endpoints with direct mechanistic and/or (sub)population relevance. In fish, these endpoints are primarily derived from reproductive parameters in chronic studies. For example, there are parameters with high mechanistic specificity for estrogen activity, such as male vitellogenin production. On the contrary, there are parameters affected by several modes of action. For example, fecundity is an apical endpoint regulated by estrogen, but systemic toxicity can also affect reproductive output (ECHA/EFSA, 2018). Thus, it is important to consider the mechanistic specificity of these parameters, as they have different levels of utility to identify potential estrogens.

Fortunately, many guideline studies include informative parameters, and there are various resources to access these data, as well as non-guideline studies. For example, the fish short term reproduction assay (OECD TG 229), 21-day fish assay (OECD TG 230), fish sexual development test (OECD TG 234), and medaka extended one-generation reproduction test (OECD TG 240) are harmonized guideline studies with parameters indicative of estrogen activity. In addition, the fish life cycle toxicity test (OPPTS 850.1500) is a common assay that can be modified to include mechanistic parameters, such as vitellogenin. There are several major data sources to access this information, including the eChemPortal database from REACH, the ECOTOX database from USEPA, the METI database from the Japanese Ministry of the Environment, the Pesticide Ecotoxicity Database from USEPA, and peer-reviewed literature, to name a few. A recent effort has consolidated these data sources, as well as others, into a curated database called EnviroTox (https://envirotoxdatabase.org/) that contains >91,000 aquatic toxicity records for >1,500 species and >4,000 CAS numbers [21]. Ultimately, these resources are valuable to identify ecotoxicology studies for risk assessment and other applications. Not only are these data useful to identify toxicity thresholds for individual substances, they may also be used to compare effects between species and chemicals. For example, probabilistic approaches such as species sensitivity distributions have been used to identify sensitive taxa and predicted-no-effect concentrations (i.e., HC5 values) for estrogens [18], and chemical toxicity distributions have been used to compare the sensitivities of common in vitro and in vivo estrogen agonist assays [30]. Thus, data mining offers a useful approach to address certain information needs related to estrogenic effects in fish.

The molecular targets of estrogens are similar between fish and other vertebrates, and the growth of new assessment methodologies (NAMs), including omics, has offered several data-driven approaches to better understand how molecular diversity affects responses to estrogens. Contrary to most vertebrates, which have 2 nuclear estrogen receptors, there are 3 nuclear estrogen receptors in teleost fish: esr1, esr2a, and esr2b, the latter of which arose from esr2 in a whole genome duplication event that occurred approximately 350 million years ago [44, 47]. Indeed, these nuclear receptors have unique effects, primarily via genomic mechanisms, and these activities complement those by membrane estrogen receptors, which act via intracellular signaling cascades [84]. While these targets have been susceptible to molecular evolution, there remains a high level of molecular and functional conservation for estrogen receptors among vertebrates, including fish. One way to examine this conservation, and by extension taxonomic susceptibility to estrogens, is to leverage omics data to compare molecular target sequence similarity.

This concept has been materialized in the SeqAPASS (Sequence Alignment to Predict Across Species Susceptibility) tool by USEPA, which quantitatively compares protein sequence/structural similarity across species to identify taxonomic sensitivity for a given target. The tool uses 3 levels of analysis to determine susceptibility: 1) primary amino acid sequence similarity, 2) functional domain sequence similarity, 3) amino acid residue similarity. Put together, this information is evaluated to set a susceptibility “cut-off” that is compared across taxa [67]. In the case of estrogens, this approach has been used to identify susceptible taxa based on the human estrogen receptor. Indeed, it was found that fish (class Actinopterygii) met the “cut-off”, and this susceptibility was confirmed with reproductive toxicity data associated with estrogen receptor activity [66]. While this current approach examines susceptibility at a broad taxonomic level, in the long term, omics data will become further useful to refine the relative sensitivity of species to estrogens through more advanced comparative bioinformatics approaches.

For ecotoxicologists, evidence of estrogen receptor conservation supports the use of read across approaches that complement or expand our understanding of estrogen-mediated effects in fish. For example, many high-throughput screening assays in the ToxCast/Tox21 programs [29, 107] have been used to screen chemicals for endocrine activity, including estrogen receptor agonism/antagonism [51, 93]. While these in vitro assays are based on mammalian models, they have screened thousands of compounds and offer useful data to characterize estrogen activity. Furthermore, this information has been used to develop computational models that integrate multiple assay responses to predict in vivo estrogen-mediated responses, such as those from the mammalian uterotrophic assay [17]. Likewise, since these assays screen large chemical libraries with high structural diversity, this information is also useful to construct robust quantitative structure-activity relationship (QSAR) models that predict estrogen receptor binding and activity [72]. In the near future, these assays are expected to include transcriptomics [78], which will offer high-dimensional response profiling for many chemicals, including potential estrogens. These datasets will be useful to identify chemicals affecting estrogen signaling pathways and genes, among others.

Certainly, these NAMs offer valuable information, and it remains important to validate these approaches to assess estrogen-mediated effects in fish. It is evident that there is strong structural conservation of ERα among species, which allows similar compounds to bind to fish and human ERα [7], although at different affinities [76]. Still, mammalian in vitro models have been useful to predict non-mammalian responses. For example, there is a significant relationship between the relative potency of compounds in ToxCast/Tox21 estrogen agonist assays, the ToxCast ER bioactivity model, and in vivo vitellogenin induction in male fish [33]. Likewise, in Tier 1 Screening of the Endocrine Disruptor Screening Program, there are similar outcomes between in vivo mammalian and fish assays [6]. Thus, there is growing evidence that comparisons across approaches, especially those utilizing NAMs (e.g., omics/HTS), will become increasing useful to predict estrogenic responses in non-target species, such as fish (Figure 2).

Figure 2:

A framework for assessing environmental estrogens in the context of adverse outcome pathways. Omics can be leveraged to support read across from other taxa, including mammalian data (e.g., high-throughput in vitro assays).

Considering the growing complexity of ecotoxicological data – especially omics data – adverse outcome pathways (AOPs) offer a useful framework to organize and integrate numerous lines of evidence. An AOP consists of a series of key events (KE) linking a molecular initiating event (MIE) to an adverse outcome (AO) through a series of causal key event relationships (KERs) [114]. In this framework, omics data primarily describe cellular KEs, but they are also useful to inform on MIEs. For example, proteomics data are useful to identify molecular target sequences that are conserved across species (e.g., SeqAPASS), and this information can define the taxonomic domain of applicability for an AOP [14]. Accordingly, QSAR models may be useful to identify structural features, and thus related chemicals, that trigger a MIE in a particular AOP [35, 108]. Downstream, omics data offer a useful perspective to identify cellular departures from homeostasis. Given sufficient magnitude, these responses may trigger subsequent KEs and potentially lead to an AO. In addition, gene expression profiles may also be associated with a particular mode of action. For example, gene set classifiers have been developed in zebrafish for endocrine disrupting chemicals, including estrogen agonists (e.g., EE2) [116]. Likewise, estrogen-responsive interactomes have been developed to complement vitellogenin as a diagnostic biomarker for estrogenicity in fish [38]. Thus, omics data are useful for several applications in the AOP framework, and this information will be important to improve AOPs for estrogen-mediated effects in fish.

A major goal of the AOP framework is to identify causal pathways between mechanistic and apical responses. In many cases, these pathways are not linear and may involve larger networks that include several MIEs and/or AOs [64]. Thus, it will be important to define critical paths in these networks to identify those associated with estrogen versus alternative or confounding paths [115]. In addition, a major goal is to better define our quantitative understanding of these biological relationships. To this end, quantitative AOPs (qAOPs) have been developed for reproductive outcomes in fish [22], and additional efforts will be required to consider compensatory and recovery processes. While recovery processes have been a research topic for estrogens and fish [12, 43], it remains a challenge to integrate this information into AOPs. Regardless, the AOP framework has been useful to organize effects data, and these efforts will ultimately lead to a better understanding of estrogen-mediated effects in fish to support ecological risk assessment.

8. Conclusions

The past several years has yielded a rich source of comparative data for EE2 in various teleost fishes. Studies investigating the pharmaceutical EE2 have yielded important clues into estrogen action. Major steps needed moving forward include: (1) Establishing conceptual frameworks for incorporating estrogenic-responsive networks into environmental monitoring programs; (2) Data mining (ECOTOX, EnviroTox) to identify effect thresholds for estrogens in fish; (3) Characterizing novel estrogen receptor signaling pathways in fish, including both nuclear and membrane receptor activity; (4) Leveraging comparative bioinformatics to identify susceptible taxa; (5) Integrating comparative lines of evidence (e.g., mammalian data, in vitro assays) to identify chemicals likely to affect estrogen-mediated endpoints in fish; (6) Incorporating various lines of evidence (described above) to construct qAOPs for estrogen-related effects in fish that include compensatory/recovery processes. As we look ahead, research into EE2 and other environmental estrogens can serve as a template for other potential endocrine active substances.

Highlights.

17alpha-ethinylestradiol (EE2) is one of the most widely studied pharmaceuticals in fish.

Transcriptome studies have revealed mechanisms of action in numerous fish species and tissues.

While data are prevalent for liver, brain, and gonad, less is known about EE2 action in kidney and pituitary.

Transcriptomics will contribute to quantitative adverse outcome pathways for estrogen signaling.