DNA Methylation and Expression of Estrogen Receptor Alpha in Fathead Minnows exposed to 17α-ethynylestradiol

Abstract

The gene expression response thought to underlie the negative apical effects resulting from estrogen exposure have been thoroughly described in fish. Although epigenetics are believed to play a critical role translating environmental exposures into the development of adverse apical effects, they remain poorly characterized in fish species. This study investigated alterations of DNA methylation of estrogen receptor alpha (esr1) in brain and liver tissues from 8–10 month old male fathead minnows (Pimephales promelas) after a 2d exposure to either 2.5ng/L or 10ng/L 17α-ethynylestradiol (EE2). Changes in the patterns of methylation were evaluated using targeted deep sequencing of bisulfite treated DNA in the 5’ region of esr1. Methylation and gene expression were assessed at 2d of exposure and after a 7 and 14d depuration period. After 2d EE2 exposure, males exhibited significant demethylation in the 5’ upstream region of esr1 in liver tissue, which was inversely correlated to gene expression. This methylation pattern reflected what was seen in females. No gene body methylation (GBM) was observed for liver of exposed males. Differential methylation was observed for a single upstream CpG site in the liver after the 14d depuration. A less pronounced methylation response was observed in the upstream region in brain tissue, however, several CpGs were necessarily excluded from the analysis. In contrast to the liver, a significant GBM response was observed across the entire gene body, which was sustained until at least 7d post-exposure. No differential expression was observed in the brain, limiting functional interpretation of methylation changes. The identification of EE2-dependent changes in methylation levels strongly suggests the importance of epigenetic mechanisms as a mediator of the organismal response to environmental exposures and the need for further characterization of the epigenome. Further, differential methylation following depuration indicates estrogenic effects persist well after the active exposure, which has implications for the risk posed by repeated exposures.

1. Introduction

The prevalence of contaminants in the environment and the potential risk they pose to exposed organisms has been an area of concern to researchers for decades. These include pharmaceutical, household, and industrial chemicals that enter the environment after passing through wastewater treatment processes (Jarosova et al., 2014; Kolpin et al., 2002). Commonly found in wastewater effluent, estrogens are well known to function as endocrine disrupting compounds (EDCs) for exposed fish and other vertebrates because of their interference with normal reproductive and hormonal physiological processes (Armstrong et al., 2016; Jackson and Klerks, 2019; Kidd et al., 2007; Ye et al., 2018). More specifically, exposure of fish populations to exogenous estrogens is associated with reproductive problems such as decreased fecundity, fitness, and sperm production, as well as feminization of males (Kristensen et al., 2005; Parrott and Blunt, 2005). Both natural and synthetic forms of estrogen persist in aquatic environments. Naturally produced estrone and 17β-estradiol, as well as the synthetically produced active ingredient in many types of birth control,17α-ethynylestradiol (EE2) are some of the most potent estrogens (Cargouet et al., 2004; Jarosova et al., 2014).

Estrogen receptors, nuclear receptors which are bound by estrogens and estrogen mimics, are transported to the nucleus where they bind estrogen response elements (EREs) and initiate transcription (Nelson and Habibi, 2013). Genes coding for the estrogen receptor have important roles in reproduction. For example, estrogens stimulate the synthesis of vitellogenin, a precursor egg yolk protein, produced in oviparous vertebrates including fish (Nicolas, 1999). For genes that are specifically involved in reproduction, expression can be induced upon exposure to exogenous estrogens such as EE2. Three nuclear estrogen receptors are recognized in teleost fishes, estrogen receptor alpha (esr1), and two estrogen receptor beta subtypes (esr2a, esr2b) which are thought to have occurred through gene duplication (Filby and Tyler, 2005; Hawkins et al., 2000; Menuet et al., 2004). Additionally, GPR30 is a transmembrane G protein-coupled estrogen receptor involved in non-genomic estrogen-initiated cell signaling cascades and in fish is thought to have a role in maintenance of oocyte meiotic arrest (Filardo and Thomas, 2005; Thomas, 2017).

Altering transcriptional regulation via epigenetic mechanisms is one method by which an organism can adapt to changing conditions, including environmental stressors such as toxins (Mirbahai and Chipman, 2014; Brander et al., 2017). Epigenetic changes, including mechanisms such as non-coding RNAs, histone modification, and DNA methylation, stably alter gene function without changing the underlying DNA sequence and are largely found to occur across taxa (Brander et al., 2017). Among these, DNA methylation is the most well-studied and easily accessible experimentally (Gutierrez-Arcelus et al., 2013; Jones, 2012; Mirbahai and Chipman, 2014). In vertebrates, DNA methylation refers to the addition of a methyl group to carbon 5 of cytosine residues, most often where the cytosine precedes a guanine (known as CpG sites) (Mirbahai and Chipman, 2014).

The location of DNA methylation in the genome affects its function. Methylation of CpG islands near transcriptional start sites (TSS) of promoter regions are strongly correlated with closed chromatin structure and gene silencing (Jones, 2012; Lou et al., 2014). Thus, DNA methylation is important for regulating transcription (Jones, 2012; Stromqvist et al., 2010). Methylation profiles can change in response to external environmental stimuli including toxins and toxicants and are associated with the development of diseases. DNA methylation patterns may provide a means to predict the sensitivity of organisms to environmental stressors such as exogenous estrogens.

DNA methylation studies in fish are limited, and most do not evaluate methylation at the single base resolution in a tissue-specific manner. It is not known whether EE2 affects DNA methylation level and/or pattern in the upstream region of esr1 in fish. In a transgenerational study, early life exposure of Menidia beryllina to environmentally relevant levels of EE2 was associated with differential methylation in the gene body of GPR30 (a non-nuclear estrogen receptor) in the F0 and F2 generations as well as other genes potentially responsive to endocrine disrupting chemicals (Major et al., 2020). Limited work has been done with other estrogen mimics, such as bisphenol A (BPA) in zebrafish which found significantly decreased mean methylation levels approximately 2kb upstream of exon 1 of esr1. This same study found no difference in mean methylation level of upstream regions of esr2 (Zhao et al., 2017). In another study, Strömqvist et al. (Stromqvist et al., 2010) found that induced vitellogenin expression in male and female zebrafish resulting from EE2 exposure was associated with significantly different methylation levels compared to control. Additionally, methylation levels of CpG sites decreased significantly following depuration.

The endocrine disrupting nature of exogenous estrogens and the negative impacts on exposed animals necessitates a better understanding of the mechanistic basis for gene expression changes. Estrogen induced changes in the epigenome may underlie observed changes in gene expression levels and apical effects. The current work aims to characterize changes in the patterns and levels of DNA methylation in the upstream and 5’ gene body of esr1 in the liver and brain of mature Pimephales promelas (FHM) exposed to EE2 for 2 d and to determine if these correlate to changes in esr1 gene expression. Further, we seek to characterize the kinetics of observed changes following depuration for seven and 14 days. Lastly, esr1 methylation patterns of EE2 exposed males are compared to adult females to determine if males adopted a feminized methylation pattern.

2. Materials and Methods

2.1. Water Chemistry.

Nominal concentrations of EE2 used for this study were 2.5 and 10 ng/L. Measured EE2 levels were taken using pooled water samples across tanks for each dose over 3 days. Water chemistry analysis found mean EE2 levels of 1.4 ± 0.31 and 3.5 ± 1.34 ng/L, respectively. Hereafter treatment groups will be referred to as “low” and “high”. Control water tanks contained no measurable levels of EE2.

2.2. EE2 Exposure & necropsy.

Reproductively mature fathead minnows used for this study (8–10 months of age) were supplied from an in-house culture (Andrew W. Breidenbach Environmental Research Center, Cincinnati, OH). This is not considered an inbred population so genetic variants are expected to be present and detectable. Past estimates based on microsatellite data indicate high levels of heterozygosity at 0.67 (unpublished data, personal correspondence). All fish were treated humanely and according to established IACUC protocols and approved by the AWBERC Animal Use and Care Committee. Fish exposures and necropsies were carried out according to previously established protocols (Flick et al., 2014). Male fathead minnows were separated from females for two weeks prior to the start date of the exposure. Male fish were exposed for two days to two doses of EE2 (low, high) or to water-only negative control. Exposures were performed in a flow-through diluter system in 30, 10-L glass aquariums, with five fish randomly assigned to each; treatments were assigned to tanks in an alternating block design. Approximately 143 fish were used in this study and comprised three groups; control (n=44), low dose EE2 (n=46) and high dose EE2 (n=45). Eight unexposed females were included for comparison. Fish were monitored and fed brine shrimp (Artemia nauplii) daily. Temperature, water flow rate, and water samples were taken daily.

Following the 2d exposure, three fish were randomly selected from each tank and anesthetized in 400 mg/l tricaine methanesulfonate (MS-222). Brains and livers were dissected from each fish and immediately frozen in liquid nitrogen and stored at −80 °C. The remaining ~60 fish were transferred to tanks receiving dechlorinated tap water supplemented with CaCO₃ to a hardness of 180 mg/L for depuration. Half of these fish were depurated for 7 days and half for 14 days, brain and liver tissue samples were taken as described above.

2.3. Putative esr1 gene structure and annotation.

Esr1 gene structure was determined through alignment of assembled full length RNA transcripts to a recently updated FHM genome assembly (Martinson et al. 2020, in review). Eight putative splice variants were identified and the variant displaying the greatest similarity in protein sequence to Danio rerio (ZF) esr1 (GenBank AB037185) was selected as the model for analysis.

Methylation analysis targeted the 5’ region of the esr1 gene and incorporated approximately 750 bp of upstream of the putative TSS through exon 2. Methprimer was used to identify CpG islands in the targeted region (Li and Dahiya, 2002). Identification of cis-regulatory elements within the targeted region was based on functional analysis of esr1 cis elements in zebrafish (Menuet et al., 2004). FHM cis-elements were identified by aligning zebrafish cis-element sequences identified in (Menuet et al., 2004) to the corresponding region in the FHM esr1. Additionally, regulatory elements were predicted using Promo (Farre et al., 2003; Messeguer et al., 2002).

2.4. Nucleic Acid Extractions.

Total RNA and DNA were extracted using the Allprep DNA/RNA Mini Kit (Qiagen, Hilden, Germany). Tissues were homogenized in 1 ml Buffer RLT Plus and processed following manufacture’s protocol. RNA and DNA were quantified spectrophotometrically using a Synergy HTX multi-mode reader (BioTek, Winooski, VT) and stored at −80 °C and −20 °C, respectively, until further use. Both RNA and DNA were quality checked using the TapeStation 4200 (Agilent Technologies, Santa Clara, CA).

2.5. Esr1 Expression.

100ng of total RNA was used as input in 20 μl reverse transcription (RT) reactions. Reverse transcription reactions were carried out in duplicate and in accordance with previously established protocol (Flick et al., 2014). No-template controls and no-amplification controls (reverse transcriptase omitted from reactions) were included with each set of samples being processed. The resulting cDNA was diluted 1:5 with nuclease-free water and used for qPCR.

Primers designed for 60S ribosomal protein (rpl8; forward: TCAAGGGGATTGTGAAGGAC and reverse: TCACGGAAAACCACCTTAGC) gene (GenBank accession no. AY919670) were used to normalize esr1 expression levels (Flick et al., 2014). Primer sequences for esr1 were: esr1 forward: GAAGCATTCAGGGTCACAATGAC and esr1 reverse: CTTCATAACACTTGCGTAGTCTGCAT. cDNA from brain and liver were PCR amplified with Power SYBR green (ThermoFisher, Waltham, MA) using an Applied Biosystems 7900HT Fast Real-Time PCR System (Life Technologies, Carlsbad, CA) in a 384-well format. All primers amplified specific amplicons indicated by a single band of appropriate length when analyzed on an Agilent 2100 Bioanalyzer using a DNA 1000 kit (Agilent Technologies). PCR amplification efficiency was >92%. The ΔΔCT method was used to quantify relative gene expression (Livak and Schmittgen, 2001). Significant differences in gene expression among groups were determined by ANOVA followed by Tukey multiple comparisons of means post hoc test.

2.6. Bisulfite Sequencing.

DNA from both brain and liver tissue was bisulfite-converted using EZ-96 DNA Methylation-Lightning MagPrep Kit (Zymo Research, Irvine, CA) according to manufacturer’s protocol. Following bisulfite treatment, CpG-enriched regions in the targeted region were PCR amplified using bisulfite-specific primers (Table 1). Primers were designed using Bisulfite Primer Seeker 12S (Zymo Research) and included 5’ tags specific for Illumina-based DNA sequencing (5’-ACACTGACGACATGGTTCTACA-[TS-For]-3’, 5’-TACGGTAGCAGAGACTTGGTCT-[TS-Rev]-3’). All primer products were validated using agarose gel electrophoresis to check for discrete bands of the correct size.

Table 1.

Primer names, size, and sequence used for targeting CpG sites for bisulfite sequencing.

Primer Name	Length	CpG sites	Amplicon Size (bp)	Sequence
Forward 1	37	14	350	GTTAGAYGAGGTGTTTAGGTAGTAGGATATTTGGTGG
Reverse 1	25	14	350	TCCTCACCRCCCACTACCCACAAAC
Forward 2	29	9	328	GGGTGAYGTGGTGTTATAGGTTTGTGTGG
Reverse 2	33	9	328	AAAAAAAAACCRAACTTAAAAAAACAAACAATC
Forward 3	31	2	251	TTATTYGTTAGAGTAGGTTATGGTAATTAGG
Reverse 3	38	2	251	TAACCACRACCTCCCATACAAAATTAAATAATAAAAAC
Forward 4	29	12	309	TYGAGGTATGTAGTAAAGTATTTGTGAAG
Reverse 4	34	12	309	CTCTCCTAATCRAATTCATCAAAATAAACAAATC
Forward 5	37	6	320	GTAATTTTGAGYGTGGTATGGTTGTTGGTGTTAGATG
Reverse 5	36	6	320	TATAATACATCCRCCTCAAAATTATACATATTATAC

Primary PCR products from all five amplified regions were pooled per individual fish. Pooled PCR products were validated with 1% agarose gel electrophoresis and quantified using the Quant-iT Picogreen assay (Thermofisher, Waltham, MA). Pools were normalized and 250bp paired-end sequences of amplicon pools were generated using an Illumina MiSeq (Illumina, San Diego, CA). Theoretical sequencing depth was approximately 500X.

2.7. Bioinformatic Pipeline.

All bioinformatic analyses were conducted using Red Hat Enterprise Linux (release 7.6). FastQC (v0.11.8) was used to quality check MiSeq output files. Adapter sequences and primers were removed using Cutadapt (v1.18). Reads were aligned with Bismark Bisulfite Mapper (Krueger and Andrews, 2011) (v0.19.0) and Bowtie2 v2.3.1 (Langmead and Salzberg, 2012). Reads were mapped to esr1 in the FHM genome assembly (NCBI accession WIOS00000000). Resulting coverage files from Bismark were analyzed using MethylKit (Akalin et al., 2012) in R 3.6.1 (2019) to apply logistic regression to detect significantly differentiated CpG sites/regions. Samples were filtered to remove bases with coverage of less than 10 reads as well as bases that exceeded the 95th percentile of coverage in each sample. After filtering, samples were normalized using default settings to normalize the coverage distributions among samples. The Chi-square option was selected to analyze the deviances of the null and alternative models. Overdispersion correction was applied to the samples using “MN” and p-values were adjusted for multiple hypothesis testing using the Benjamini Hochberg (BH)(Benjamini and Hochberg, 1995) correction. A BH value of < 0.1 was selected as the significance threshold. Analytical parameters were selected based on previous comparative analyses which found high performance for this method; indicated by high sensitivity while still retaining specificity when combined with high effect size (Wreczycka et al., 2017). MethylKit provides the functionality to analyze methylation data at single base pair resolution as well as over predetermined sections of sequence using tiling windows. To gain a clear understanding of DNA methylation level and pattern, the data was analyzed both at the single base pair resolution as well as in tiling windows of 200bp lengths chosen specifically to capture regions of biological interest.

3. Results

3.1. Esr1 putative cis-elements and promoter.

Sequence alignments were conducted between the targeted gene region and cis-elements upstream region of the zebrafish esr1 promoter that were functionally evaluated in (Menuet et al., 2004) (Figure 1). The spacing of cis-elements in intron 1 was highly conserved between zebrafish and FHM with a full length estrogen response element (ERE), flanked by 5’ and 3’ ½ ERE/AP-1 binding sites which were directly upstream from an initiator element (Inr) sequence. Sequence alignments conducted using sequences identified in (Menuet et al., 2004) over the entire targeted region identified several other ½ ERE sites in Exon 1 and in the upstream promoter region, which were not identified in (Menuet et al., 2004)(Table 2). The presence of these sites were supported using an in silico cis-element prediction algorithm PROMO. Sequence alignment also identified a second Inr (C/T,C/T,A,N,T/A,C/T,C/T) at position −702 of the targeted FHM upstream region. A single CpG island of 100 bp was identified within the gene body, starting at position 1299.

Figure 1.

Putative esr1 gene model diagram (top) showing a targeted region including a 5′ upstream region (orange), exons (blue), and intronic regions (black line), and putative transcription factor binding sites. CpG sites are shown with corresponding pie charts representing 2d liver tissues (A) control (water only), (B) low dose males (1.4 ± 0.31 ng/L EE2), (C) high dose males (3.5 ± 1.34 ng/L), (D) females (no exposure), and the same treatments in brain tissues (E) through (H), respectively (n=27–29 per group for males, n=8 for females). Mean methylation level for each CpG site is depicted by the colored region of each pie chart, and significance is denoted by asterisks (FDR=0.1, differential methylation calculated by logistic regression). CpG sites lacking sequence data are shown by empty pie charts.

Table 2.

Transcription factor binding sites and locations as predicted by Promo (Messeguer et al., 2002) or by global alignment of TF sites identified in (Menuet et al., 2004).

Name	Type	Location
Estrogen Response Element (ERE)	cis-regulatory element	−402 – −3963, −158 – −1523, 35 – 412, 293 – 2993, 318 – 3333 *, 342 – 3483
Initiator site (Inr)	Initiator site	−728 – −7221, 367 – 3741
C-ets-1	Transcription factor	427 – 4332, 653 – 6592, 669 – 6753, 745 – 7512, 786 – 7922, 1324 – 13302
AP-1	Transcription factor	292 – 3031, 340 – 3521

¹Identified by sequence alignments based on Menuet

²Identified by PROMO

³Identified by both Menuet and Promo

^*Full ERE site

3.2. Sequencing and Mapping.

Raw coverage was approximately 500x across all samples (Supplementary material Figure 1). Overall Q30 scores were ≥ 79% for liver samples and ≥ 65% for brain samples. Raw clusters generated ranged from 5–9 million reads per library. Mapping efficiency using Bismark ranged from 52–73% for liver and 34–70% for brain samples, which is within the expected range for bisulfite converted sequence reads (Chatterjee et al., 2013; Tran et al., 2014). In total, approximately 34 CpG sites were targeted in this study. After filtering and normalizing across treatment groups, 25 CpG sites had adequate sequencing depth to be included in the analysis; 10 CpG sites in the 5’ upstream region of esr1, and 15 within the gene body (Figure 1). Methylation variance was low in the upstream region. In contrast, variance of gene body CpGs was relatively high and demonstrated some location-dependence (Figure 2). Similar trends in variability in methylation levels at non-island CpG sites have been previously found in humans (Bock et al., 2008).

Figure 2.

Standard deviation from the mean at each CpG site for 2d fish (n=83) following exposure to EE2 for (A) liver and (B) brain tissues. CpG sites −739, −463, −415, and −298 in brain and site −646 in liver were excluded from analysis due to limited sequencing depth. All CpG sites oriented left of the TSS at 0bp are upstream, and all CpG sites located to the right of the TSS are in the gene body of esr1.

3.3. Esr1 Expression.

In liver, relative esr1 expression was significantly higher than control in both treatment groups following 2d exposure to EE2 (p<0.001, Figure 3A), which was no longer evident at either depuration time point. No significant differences in relative esr1 expression at any time point were observed in brain tissue (Figure 3B).

Figure 3.

Relative expression of esr1 in liver (A) and brain (B) tissues from male fathead minnows. X-axes indicate timepoint (D2, D7, and D14) and treatment group (control, low (1.4 ng/L ± 0.31) EE2 dose, high (3.5 ± 1.34 ng/L) EE2 dose). At Day 2 (n=27–29 per group), both treatment groups differed significantly from control (p<0.001, ANOVA with Tukey test for multiple comparisons of means). There were no significant differences in treatment groups relative to control after depuration (D7 (n=8–9 per group), D14 (n=8–10 per group)). Significance is denoted by asterisks and outliers in the data are shown as black dots.

3.4. DNA methylation of individual CpG sites in 5′ upstream region and gene body of esr1.

3.4.1. 2d - upstream gene region.

To determine whether exposure of males to EE2 resulted in a feminized methylation pattern in the esr1 gene, methylation patterns were compared among male treatment groups and untreated females relative to control males. In liver, females displayed lower levels of methylation relative to control males in the 5’ upstream region of esr1 at the three CpG sites (−463, −415, and −298) located closest to the TSS (Figure 4). Similarly, reduced methylation of CpG at position −415 was observed in both the EE2 treatments (Table 3; Figure 4). An additional CpG site (−463) was found to be hypomethylated in liver of males treated with high levels of EE2. While the CpG located at position −298 was just above the BH cutoff for significance (Table 2), its proximity to the only other differentially methylated CpG sites in the upstream region coupled with the consistency in the direction of differential methylation among these sites both with treatment and across sexes, strongly suggests that it is part of a differentially methylated cluster. These results suggest a dose relationship of EE2 concentrations on differential methylation, both in terms of the magnitude of the response and the number of effected CpGs. EE2-dependent methylation patterns in males were highly similar to the female pattern, suggesting EE2 induces a feminized pattern of methylation in the proximal promoter region.

Figure 4.

Percent methylation differences in liver tissue for 2d males (n=27–29 per group) and females (n=8) as modeled by logistic regression of three CpG sites found to be closest to the predicted TSS. * indicates statistically significant differences (FDR < 0.10). Low EE2 dose is measured at 1.4 ± 0.31 ng/L and high EE2 dose is 3.5 ± 1.34 ng/L.

Table 3.

Heatmap showing p-values after Benjamini Hochberg adjustment for each CpG site for both liver and brain tissue, timepoint, and dose. Color coding is relative to adjusted p value (FDR=0.1).

The region encompassing the three CpG sites (−463, −415, and −298) shown to be differentially methylated in the liver did not meet the sequencing depth cutoff and was thus excluded from analysis in the brain. However, site −472, which is immediately 5’ to (within 9 bp) the cluster identified in liver samples, demonstrated reduced methylation levels. Though this CpG was not differentially methylated in either liver or female tissue, its proximity to other demethylated CpGs may suggest a regional consistency between liver and brain responses.

In order to determine the potential functional significance of demethylation, differentially methylated CpGs were mapped relative to putatively identified cis-elements. The three differentially methylated CpGs identified in the liver and the one in the brain flank a partial ERE (−402 – −396)site in the upstream region (Table 2). The second partial ERE, which is site located more proximally to the TSS is approximately 140 bp 3’ of the closest differentially methylated CpG.

3.4.2. 2d — gene body.

In the liver, females displayed significantly lower gene body methylation (GBM) levels across all 15 CpG sites in the gene body compared to control males (Figure 1; Figure 5). In contrast, no differences in methylation were observed in liver from males from either EE2 exposure (Figure 5). In brain samples, no differences in female methylation patterns were observed relative to control males. However, methylation patterns in the gene body of males varied considerably with dose. While no differential methylation was observed in the high dose, methylation increased in 12 of 15 CpG sites in low dose relative to control and several others in the region approached significance (Figure 5).

Figure 5.

Bar plots showing percent differences in methylation generated from logistic regression across CpG sites in the gene body of esr1. (A) indicates percent methylation differences in liver at Day 2 across treatment groups and females compared to control. (B) indicates percent methylation differences across treatment groups and females compared to control at Day 2 in brain. All significant CpG sites are denoted by asterisks (FDR < 0.10). Low EE2 dose is measured at 1.4 ± 0.31 ng/L and high EE2 dose is 3.5 ± 1.34 ng/L.

3.5. Depuration.

In order to characterize the kinetics of DNA methylation, methylation was evaluated in fish depurated for 7 or 14 days. For practical reasons, smaller sample sizes were used for these comparisons (n=8–10 fish/group), which, coupled with the considerable amount of inter-individual variation in methylation, may have limited statistical power.

No significant differences in methylation were in the upstream region for either tissue at day 7 (Table 3). In liver, a single CpG (−642) was hypermethylated relative to controls at day 14 (Table 3, Supplementary Figure 2B). This CpG was not differentially methylated either during the active exposure or at 7d post-exposure. In brain, the CpG at site −472 was hypermethylated at 14 d post exposure, opposite of what was observed at the 2d active exposure.

Consistent with what was observed at 2d, no GBM was observed at either depuration time point in the liver. In the brain, however, a more robust methylation response was observed in the gene body (Table 3, Supplementary Figure 2C). Eight of the CpGs that were differentially methylated during the active low dose exposure, were also found to be differentially methylated in the high dose treatment at 7 d (Table 3). This was lost by 14d post-exposure. The direction of the 7d response paralleled what was observed during the active exposure (Supplementary Table 2B).

3.6. DNA methylation using tiled window analysis.

3.6.1. 2d.

In addition to conducting single base-pair resolution analysis, tiled windows of 200bp with 50bp steps were used to capture differential methylation over defined regions. In liver, 2d low dose males were significantly less methylated across a tiled window from −425 to −226. Similarly, males exposed to the high dose of EE2 were significantly less methylated across −475 –276 (p < 0.05) and −425 – 226 (p < 0.001). This region contained CpG site −298, which was just above the BH signficance threshold in the per CpG analysis. Results for females were consistent with the single base-pair resolution analysis; all CpG sites were significantly less methylated compared to untreated males (−475 – 226, p < 0.05 for upstream region; 875 – 1474, p < 0.001 for gene body).

In brain, low dose males displayed significanly less methylation compared to control in upstream region (tiled window of −475 – −276 (p< 0.001). In contrast, methylation levels were increased in the gene body of low dose males for specific tiled windows (875 – 1074 and 925 –1124, p < 0.05. Though individual CpGs located between 1125 – 1474 were differentially methylated, when considered as a region, no differential methylation was observed, though it approached signficance at p = 0.0535.

3.6.2. Depuration.

For both brain and liver, tiled window analysis revealed no significant differences in methylation level for either timepoint (7d, 14d).

3.7. Methylation of esr1 is negatively correlated with gene expression in liver tissues.

A correlation analysis was conducted to determine the relationship between esr1 methylation gene expression in the 2d liver samples. A moderate inverse relationship was observed for individual CpG sites in both the promoter and gene body regions and gene expression (Pearson’s correlation coefficient r(81) = -.31, p < 0.01).

We further examined this relationship using methylation levels from the tiled window analysis. No differential methylation was observed in the gene body of the liver, however, three adjacent tiled windows spanning 1024 – 1624 of the targeted region were found to negative correlation with gene expression (Figure 6), suggesting the statistical threshold used in differential methylation analysis may have been overly stringent. No correlation were observed in the farthest upstream tiled windows from −676 – 476 or from gene body region 2100 – 2300.

Figure 6.

Pearson correlations between tiled windows and esr1 gene expression in male liver tissue. Pearson correlation analysis was conducted between CpG levels of individual tiled windows and gene expression in livers of males exposed to either low (1.4 ± 0.31 ng/L or high (3.5 ± 1.34 ng/L EE2. Specific region locations are indicated on the y axis of each panel

4. Discussion

The objectives of this study were to assess liver and brain tissues from reproductively mature male fathead minnows in order to 1) characterize changes in the pattern and level of DNA methylation for esr1 in fish exposed to EE2 and determine whether changes correlate with esr1 gene expression levels from EE2 treated fish, 2) determine whether methylation changes persist after the chemical challenge is withdrawn, and 3) determine if EE2-dependent methylation differences in treated males display a feminized pattern.

4.1. Upstream region

Increased methylation levels were observed in the upstream region of esr1 in both liver and brain tissue. In the liver, both the low and high EE2 treatments resulted in reduced methylation levels at CpG sites located most proximally to the TSS. While only one CpG was demethylated in the low exposure, two were found to be demethylated in the high treatment and a third was just above the significance threshold. This and the increased magnitude of demethylation observed in the high dose exposure suggested some dose dependence in the methylation response (Figure 4, Table 1), which is also reflected in the increased magnitude of esr1 gene expression (Figure 2). The pattern observed in the high dose males reflected the pattern seen in females relative to control males, suggesting a feminized response (Figure 4). The concentration used in the high dose (3.5 ng/L EE2) has previously been shown to induce feminization in sexually mature male FHM, as evidenced by reduced gonadal somatic indices and tubercle grade, albeit at longer exposure durations (Filby et al., 2007).

Increased promoter methylation is thought to reduce transcriptional activity by inducing steric changes in DNA which inhibits the binding of TF required for transcriptional initiation (Heberle and Bardet, 2019; Kribelbauer et al., 2019). Additionally, methylated CpGs serve as binding sites for methyl binding proteins which can directly interact with regulatory proteins and can alter chromatin structure (Fournier et al., 2012)(Boxer et al., 2020). In the present study, promoter methylation of esr1 was negatively correlated with its gene expression in liver of males treated with either concentration of EE2. These results are consistent with previous studies that have also demonstrated a negative correlation between esr1 promoter methylation and transcriptional repression (Lung et al., 2020). Differentially methylated CpGs were found to occur in the proximity of two predicted partial estrogen response elements (ERE) (Figure 1,Table 3), which is consistent with autoregulation of esr1 gene expression driven by activated estrogen bound ESR1 (Tian et al., 2015). Menuet et al (Menuet et al., 2004) functionally validated a regulatory cassette, including a full-length ERE flanked by two partial EREs, located just 5’ of the ZF TSS. Though that same regulatory module is found in the FHM esr1 gene (Figure 1, Table 3), it resides in intron 1 of the FHM esr1 gene model. This suggests that the FHM esr1 contains an additional exon and that transcription is initiated via an upstream alternative promoter. Notably, the use of alternative promoters is very common (Kimura et al., 2006) and for the esr1 it has been observed in a number of taxa including fish species (Cotter et al., 2013; Ishii and Sakuma, 2011; Kos et al., 2001; Pakdel et al., 2000). The consequence of alternative promoter usage of esr1 is unclear, however, it is associated with tissue and regional specificity, as well as disease (Grandien, 1996; Kos et al., 2001; Osterlund et al., 2000). Disruption of methylation in the upstream region of esr1 has the potential to negatively impact apical endpoints. ESR1 is known to play a regulatory role in liver metabolism (Qiu et al., 2017) and its disruption in the liver is linked to metabolic disease in humans (Efstathiadou et al., 2015). ESR1 has also been identified as a master regulator of CYP3A4 expression, the primary drug metabolizing enzyme (Wang et al., 2019). This has the potential to add uncertainty to risk estimates of wastewater treatment plant (WWTPs) discharge, given that WWTP effluent is known to be a major sources of EE2, other estrogens, and pharmaceutical and personal care products into surface waters (Tiedeken et al., 2017).

During depuration, a single CpG dinucleotide at position −642 was identified as differentially methylated in the low exposure after 14 d of depuration, however, it was not shown to be differentially methylated during either the active exposure or at 7 d post-exposure (Table 2). The relative methylation change between the control and low exposure was relatively modest (<1.5%), however, highly consistent. Methylation among all replicates within the control treatment were lower than all treated replicates. Though the magnitude of methylation was modest, several studies have shown that small, robust changes in methylation can have an outsized effect on relevant endpoints (reviewed in (Breton et al., 2017). The position of the methylated CpG is proximal to both the upstream putative Inr and the ERE sequences (Figure 2). The proximity to the Inr site may suggest a role in alternative promoter usage rather than regulation of transcriptional initiation. The biological relevance of punctuated CpG methylation and the mechanism is unknown, however, it may suggest that acute exposure to potent estrogens is sufficient to initiate a lasting biological cascade that continues to modify CpG methylation well after the cessation of exposure. This is consistent with previous studies that have characterized lasting differential methylation and phenotypes resulting from transient exposure to environmental stimuli (Ruiz-Hernandez et al., 2015).

Direct comparison of the effects of EE2 on either methylation or gene expression between the brain and liver is hampered by the exclusion of CpGs in the brain that overlapped with differentially methylated CpGs in the liver and a lack of observed differential expression. The lack of differential expression in brain tissue was somewhat surprising since the esr1 is primarily regulated on the transcriptional level (Lung et al., 2020) and the brain is a well-established estrogen responsive tissue (McEwen, 2001). Unlike the liver, however, the brain is highly heterogeneous with distinct regional transcriptional profiles (Sun et al., 2012). This suggests that the current study may have lacked the spatial resolution to detect regional expression differences and the lack of observed differential expression may have resulted from the averaging of expression levels across different brain regions. Only one CpG (−472bp) was hypomethylated in the low treatment exposure group in the brain and was just above the significance threshold for the high exposure group (Table 2). This CpG was also hypomethylated at 14 d, but not 7 d, post-exposure. It is located directly upstream of differentially methylated CpGs in the liver, suggesting it may be part of the same regulatory region. Similar to what was observed with gene expression, identification of differentially methylated CpGs may have been affected by averaging across regional differences in methylation. Estrogens play many functional roles in brain tissue (McEwen, 2002), suggesting that disruption of the signaling pathway could potentially negatively impact an equally diverse set of endpoints. Though preliminary, our results suggest the importance of further evaluation of EE2 effects in the brain on methylation, transcription, and downstream signaling with finer spatial resolution

4.2. Gene Body Methylation (GBM).

GBM differed considerably between the brain and liver tissue. In liver, no significant GBM differences were observed for any treatment or time point, however, liver GBM was found to be negatively correlated to gene esr1 gene expression. The gene bodies of livers from females, however, were significantly less methylated relative to control males across the entire targeted gene region. The lack of any EE2 related GBM in males may suggest the EE2 induced feminized methylation pattern is limited to the upstream gene region in the liver. In contrast to the liver, an increase in GBM was evident across most of the CpG sites assayed in brain tissue of males exposed to the low dose of EE2. Interestingly, this pattern was not observed in either the high treatment or in the female (Table 2; Figure 5). Differential methylation in the low dose exposed males was not sustained post-exposure, however, in the high exposure group, the region was largely differential methylation at 7 d post-exposure was consistent with what was observed in the low dose during active exposure (Table 2). The consistency of differential methylation in the gene body across time points, suggests that GBM post-exposure may have some functional significance and is further evidence of a sustained effect of EE2. This has potentially negative implications given pleiotropic effects of ESR1 and the probability that exposed organisms may encounter a second exposure before the effects of the initial exposure have fully dissipated. GBM methylation is conserved across species and has been associated with multiple aspects of transcriptional and translational regulation, including alternative splicing, use of alternative promoters, elongation efficiency, initiation and rate (Maunakea et al., 2010). Though the relationship between GBM and expression levels is still controversial (Ball et al., 2009; Lou et al., 2014; Major et al., 2020), it is often associated with increased expression (Ball et al., 2009; Jones, 2012; Maunakea et al., 2010). Gene location dependent methylation has been previously investigated for the esr1 gene in humans (Shenker et al., 2015). In cell culture systems and epithelial cells derived from breast milk, relative promoter methylation levels of esr1 were related to both the estrogen receptor status and esr1 expression. ER-negative groups displayed higher promoter methylation and lower esr1 expression relative to ER-positive groups, however, the opposite trend was observed in gene bodies (Shenker et al., 2015). Further, esr1 expression increased in cell culture systems concomitant with demethylation induced by decitabine in promoters of ER-negative samples yet decreased with reduced GBM in ER-positive samples, suggesting a mechanistic link between GBM and increased esr1 expression (Shenker et al., 2015). As stated above, no differential expression of the esr1 gene was observed with EE2 treatment in the brain, limiting interpretation of the biological significance of GBM methylation. However, the consistency of the methylation pattern during the active EE2 exposure and depuration suggests further study may be warranted.

Conclusions:

We have demonstrated that EE2 exposure results in the differential methylation of CpG dinucleotides in the esr1 gene in both brain and liver tissue. Promoter methylation patterns were largely similar between tissues in exposed males and, in liver, appeared to exhibit a dose-dependent shift toward a feminized pattern. Promoter methylation in the liver was also anti-correlated to esr1 gene expression. GBM patterns of treated males differed between tissues. Lastly, some methylation effects of EE2 remained after chemical withdraw. Though the current study has identified EE2-induced methylation changes in the proximal promoter and 5’ intragenic regions of esr1, interpretation must be considered in the larger context of the complexity of esr1 transcriptional regulation. The likelihood of estrogenic exposure to aquatic organisms in surface waters coupled with the methylation changes identified in the present work, strongly suggests the need for a more comprehensive evaluation of estrogenic effects on the methylome at finer spatial scales, as well as the experimental validation of the functional significance of identified methylation changes.