To initially assess the extent to which epimutations and dysregulated gene expression can be induced in cells maintained in culture, we exposed mouse iPSCs to three doses of BPS (1, 50, and 100 μM) and measured the impact on the epigenome and transcriptome using the Illumina Infinium Mouse Methylation BeadChip Array and RNA-seq, respectively (Figure 1). All three doses of BPS induced individual differentially methylated CpGs (DMCs) as well as differentially methylated regions (DMRs) (Figure 1a), and differentially expressed genes (DEGs) (Figure 1b) when compared to control mouse iPSCs treated with vehicle (EtOH) only. The extent of this exposure-specific epimutagenesis was correlated with the dose of BPS used. Information regarding the overlap of DMCs/DMRs/DEGs identified for each dose of BPS is shown in Figure 1—figure supplement 1. We observed BPS-induced DMCs within both promoter and gene body regions of a portion of DEGs (17-38%) (Table 1). Importantly, in many cases, we observed a correlation between differential expression of a gene and the presence of DMCs in the promoter region of that gene (Figure 1—figure supplement 2a). These results provided an initial proof of concept that exposure of cells maintained in vitro to an EDC such as BPS can disrupt the epigenome and transcriptome in a dose-dependent manner. Interestingly, exposure to 1 µM BPS induced very few DMRs, but did induce widespread DMCs and DEGs. Because 1 μM BPS is below the FDA’s suggested safe environmental level established for exposure of humans or intact animals to BPA (Ribeiro et al., 2019), and was sufficient to induce DMCs and dysregulated gene expression on all chromosomes in our cultured iPSCs, we utilized this dose and focused solely on DMCs when assessing epimutations in all subsequent experiments (Table 1).

We next sought to determine if different key cell types – somatic, pluripotent or germ – are differentially susceptible to induction of epimutations in response to direct exposure of each to a similar dose of BPS. Thus, we exposed pluripotent (iPSCs), somatic (Sertoli and granulosa cells), and germ (PGCLCs) cell types to 1 μM BPS and measured changes in DNA methylation patterns. We identified exposure-specific DMCs in each exposed cell type relative to its corresponding control (exposed to carrier only; Figure 2a). We observed overall differences among the different cell types in total numbers of DMCs, with iPSCs showing the highest number of DMCs, followed by Sertoli cells and granulosa cells, and then PGCLCs, respectively (Table 2).

Interestingly, among the exposure-specific DMCs identified in each cell type, we observed predominantly hypomethylated DMCs in the somatic and pluripotent cell types, but predominantly hypermethylated DMCs in PGCLCs (Table 2). These findings confirm that there are both quantitative differences among cell types in overall susceptibility to epimutagenesis, and qualitative cell-type specific differences in the prevalence of hypo- versus hypermethylated DMCs following exposure to BPS. The latter likely reflects the fact that the epigenome in PGCs is naturally more hypomethylated than that in pluripotent or somatic cell types (Hajkova, 2011), enhancing the likelihood that most changes in DNA methylation induced in PGCLCs will necessarily involve hypermethylation.

As we only found an average of 11.05% direct overlap in DMCs between two or more cell types and no overlapping DMCs shared between all cell types (Figure 2—figure supplement 1), we next analyzed annotations associated with genomic sites of BPS exposure-specific DMCs in each cell type. We mined annotations associated with genomic regions included on the Infinium array and found that a substantial group of exposure-specific DMCs was associated with enhancer regions in the two somatic (Sertoli and granulosa) and one pluripotent (iPS) cell types exposed to BPS, while in PGCLCs BPS-induced DMCs were more prevalent at promoter regions containing transcription start sites (TSSs) and transcription factor binding sites (Figure 2b), indicative of yet another qualitative cell-type specific difference in induction of epimutations by the same dose of BPS.

As EDCs are thought to induce epimutations via disruption of classical hormonal signaling (Henley and Korach, 2010; Kelce et al., 1995; Swedenborg et al., 2009; vom Saal and Hughes, 2005; You et al., 1998), we next sought to determine if the differential extent of BPS induction of epimutations we observed in different cell types was associated with differential expression of relevant hormone receptors in each. We performed immunocytochemistry (ICC) to detect the presence of the relevant estrogen receptors – ERα and ERβ, while co-staining for cell-type specific markers (WT1 [Sertoli cell marker], FSHR [granulosa cell marker], FUT4 [iPSC marker], and NANOG [PGCLC or endogenous PGC marker]) to confirm the identity of the four cultured cell types examined in this study (Sertoli cells, granulosa cells, iPSCs and PGCLCs), as well as endogenous mouse PGCs (Figure 3a). Both somatic cell types (Sertoli and granulosa) showed positive immunolabeling for both ERα and ERβ, whereas the pluripotent cells showed positive immunolabeling for ERβ, but negative immunolabeling for ERα, and the PGCLCs and endogenous PGCs showed negative immunolabeling for both ERα and ERβ. Importantly, the latter result is consistent with previous reports of lack of expression of either ERα or ERβ at the protein level in endogenous PGCs (Meccariello et al., 2014).

These results demonstrate cell-type specific differences in expression of hormone receptors (ERα and ERβ) which are potentially relevant to the disruptive action of the estrogen mimetic, BPS. We noted that expression of at least one potentially relevant hormone receptor (ERβ) in either somatic or pluripotent cells correlated with a higher incidence of DMCs induced by exposure of somatic or pluripotent cell types to 1 µM BPS relative to the incidence of DMCs induced by exposure of PGCLCs, which do not express either estrogen receptor, to the same dose of BPS (Table 2). Nevertheless, we did still observe induction of DMCs when PGCLCs were exposed to BPS, demonstrating that expression of relevant canonical hormone receptors is not an absolute requirement for induction of epimutations in response to an EDC. To exclude the possibility that the observed susceptibility in germ cells could be correlated with the presence of other endocrine receptors that could interact with BPS, we performed additional ICC for the presence of AR, PPARγ, and RXRα and found they were all absent as well in endogenous PGCs (Figure 3—figure supplement 1). This result and our observation noted above that somatic and pluripotent cell types showed a higher incidence of epimutations at apparent enhancer regions while germ cells showed a higher epimutation incidence at apparent promoter regions reveal cell-type specific differences in susceptibility to induction of epimutations by exposure to the EDC, BPS.

The relatively higher incidence of BPS-induced epimutations in cell types expressing one or both estrogen receptor(s) suggests BPS-induced epimutagenesis may be manifest, at least in part, through canonical endocrine signaling pathways. If this is the case, we might expect to see elevated induction of epimutations in genomic regions enriched for relevant HREs which, for interactions with the estrogen mimetic, BPS, would be EREs. Previous studies have defined a full ERE consensus sequence (Bourdeau et al., 2004), but other reports have indicated that estrogen receptors can often bind to ERE half-sites (Mason et al., 2010; Figure 3b). Indeed, when we mined publicly available ERα ChIP-seq peaks from the UCSC genome browser database (Dunham and Kundaje, 2012; Myers et al., 2011; Sloan et al., 2016; Wang et al., 2013; Wang et al., 2012) and performed motif enrichment, we identified two distinct ERE half-sites within regions enriched for DMCs rather than one full-sized ERE consensus sequence, with the half-sites appearing to be more biologically relevant to interaction with estrogen or its mimetics (Figure 3b). We then plotted the frequency of ERE half-sites in genomic regions within 500 bp of BPS-induced epimutations (Figure 3c and d). We found an increase in the frequency of ERE half-sites identified within 500 bp of BPS-induced DMCs genome-wide in all four cell types investigated, but the frequency was notably lower in germ cells (Figure 3c). Thus, this higher frequency of ERE half-sites within 500 bp of BPS-induced DMCs was conserved in three of the four cell types – somatic (Sertoli and granulosa) and pluripotent (iPSCs) when we focused solely on apparent enhancer regions, but not in germ cells (PGCLCs) where the majority of DMCs occurred at apparent promoter regions (Figure 3d). We also mined the UCSC mouse genome sequence for genome-wide prevalence of ERE half-sites and found that while these sites occur in both promoter and enhancer regions, they are nearly fourfold more frequent in enhancers (Figure 3—figure supplement 4 and Table 3). Thus, it is perhaps not surprising that BPS-induced epimutations were more prevalent in enhancer regions in somatic and pluripotent cell types that express one or both ERs, than in germ cells that do not express either ER.

Taken together, these results suggest a relationship between (i) expression of either ERβ alone or ERα and ERβ together, (ii) elevated susceptibility to BPS-induced epimutagenesis, and (iii) the occurrence of BPS-induced DMCs in genomic regions – particularly enhancers – containing EREs. These observations are consistent with the notion that one mechanism contributing to EDC-based induction of epimutations involves disruption of canonical endocrine signaling pathways. However, the fact that exposure to BPS also induced epimutations in germ cells, even in the absence of expression of estrogen-related receptors, and generated many DMCs in regions not inclusive of EREs, suggests that disruption of canonical endocrine signaling pathways is not the only mechanism by which exposure to an EDC can induce epimutations.

To further interrogate cell-type specific differences in the genesis of epimutations following exposure to BPS, we compared DNA methylation patterns detected in the control (vehicle only) samples to identify naturally occurring, cell-type specific DMCs inherently associated with each different cell fate. Interestingly, of the 297,415 distinct CpGs interrogated by the Illumina Infinium Mouse Methylation BeadChip Array used for this analysis, >240,000 showed some degree of inherent differential methylation among the four cell types tested, suggesting they are tied to cell-fate specific differential DNA methylation (Figure 4a). We next determined the extent to which DMCs induced specifically by exposure of each cell type to BPS occurred at CpG dinucleotides that were among these naturally occurring cell-type specific DMCs. We found that a large majority of BPS exposure-specific epimutations detected in each cell type occurred at CpGs that also show inherent, cell-type specific variation in DNA methylation (>95% of BPS-induced epimutations in somatic and pluripotent cell types and ~89% in germ cells; Figure 4a). As with the overall pattern of BPS-induced epimutations shown in Figure 2, BPS-induced epimutations at CpGs showing inherent cell-type specific variation were enriched in apparent enhancer regions that occurred near EREs in somatic and pluripotent cell types, whereas those in germ cells were found predominantly in promoter regions containing significantly fewer EREs (Figure 4b and d). However, the low percentage of BPS-induced epimutations that occurred at CpGs that did not show inherent cell-type specific variation in DNA methylation patterns were found to be enriched in promoter regions lacking nearby EREs in all four cell types (Figure 4c and d). Thus, this latter group of BPS-induced epimutations appears to represent a small core group that arises following exposure to BPS via a mechanism that does not rely upon disruption of canonical endocrine signaling, and so is common to all cell types, regardless of expression of relevant endocrine receptors or nearby residence of relevant HREs within the genome.

We next sought to determine if the cell-type specific differential susceptibility to BPS-induced epimutagenesis translated to a similar extent of dysregulation of gene expression in each cell type. We found that there was a similar relationship between the presence of DMCs in the promoter region of a gene and differential expression of that gene in all four cell types examined (Figure 1—figure supplement 2b). Somewhat surprisingly, RNA-seq analysis of gene expression patterns in each exposed cell type relative to its corresponding control (same cell type exposed to carrier only) revealed the greatest number of dysregulated genes in PGCLCs, despite PGCLCs being the cell type that showed the lowest number of epimutations following exposure to BPS (Tables 2 and 4). iPSCs showed the second highest level of dysregulated genes, while somatic Sertoli and granulosa cells showed relatively low levels of dysregulated gene expression (Table 4). Indeed, numbers of dysregulated genes were three orders of magnitude lower in differentiated somatic cells than in either pluripotent cells or germ cells (Table 4).

Because BPS-exposed PGCLCs showed the highest level of dysregulated genes, as well as the highest enrichment of DMCs occurring primarily at promoters (Figure 2b), we next assessed the general proximity between DMCs and promoter regions in each cell type to determine if DMCs in promoter regions may be more likely to predispose dysregulated gene expression than DMCs elsewhere in the genome (Figure 5a). Indeed, we found that in all four cell types, the smaller the median distance between BPS-induced DMCs and neighboring promoter regions, the larger the number of BPS-induced DEGs (Figure 5b). Thus, it appears that while PGCLCs showed the fewest overall BPS-induced DMCs among the four cell types exposed to BPS, this exposure induced a higher proportion of epimutations in regions within or adjacent to promoters in this cell type. This, and the relatively high extent of decondensed chromatin genome-wide in fetal germ cells, appear to have contributed to the higher incidence of dysregulated gene expression in BPS-exposed PGCLCs than that observed in the other three BPS-exposed cell types, even though the overall numbers of BPS-induced epimutations were greater in the other cell types.

To identify potential mechanisms by which BPS exposure may induce epimutations via disruption of non-canonical endocrine signaling pathways, we mined our RNA-seq data to identify genes dysregulated independently of expression of relevant hormone receptors or presence of nearby EREs. We identified a set of genes in all four cell types tested that displayed promoters enriched for apparent endocrine-signaling independent DMCs (Figure 6—figure supplement 1). Gene ontology (GO) analysis indicated that several of the 1957 genes we identified were associated with ubiquitin-like protease pathways, including 124 of 800 genes that regulate protein degradation and 5 of 9 genes involving 1-phosphatidylinositol-3 kinase activities, linked to the PI3K/AKT signaling pathway (Figure 6a).

Previous reports have established a link between ubiquitin-like protease pathways and classical ER signaling (Beamish and Frick, 2022; Kabir et al., 2015), suggesting that genes related to ubiquitin-like protease pathways that have promoters lacking EREs may be regulated by factors that are themselves encoded by genes with promoters containing EREs, and so could be indirectly activated or repressed by disruption of classical ER signaling. However, involvement of PI3K/AKT pathway signaling has previously been linked to estrogen signaling through non-canonical G-protein coupled receptors such as GPER1, which was sufficient to induce estrogen signaling in ER KO mouse cell lines which lack the capacity for classical ER signaling (Filardo et al., 2002; Molina et al., 2017). To determine if our data support the suggestion that the non-canonical BPS-induced changes were linked to involvement of PI3K/AKT pathway genes, we performed GO analysis on the 1890 DEGs identified in PGCLCs in which epimutations appeared to be induced independent of classical ER signaling. Interestingly, we detected no apparent involvement of pathways involving ubiquitin-like proteases, but we did detect differential expression of four of seven genes associated with a pathway involving phosphatidylinositol-4 phosphatase signaling which intersects with the PI3K/AKT signaling pathway (Figure 6b). Finally, while we did not detect expression of Gper1 transcripts, we did observe differential expression of genes encoding other less well studied G-protein-coupled receptors, including Gprc5a, Gprc5b, Gpr89, Gpr107, and Gpr161 in all four cell types exposed to BPS. These receptors have all been previously shown to bind either BPA or 17β-estradiol in rat models and could be potential targets of non-canonical BPS-induced estrogen signaling (Figure 6c) as described in the following links: (Rat Genome Database, 2024a; Rat Genome Database, 2024b; Rat Genome Database, 2024c; Rat Genome Database, 2024d; Rat Genome Database, 2024e).

Transitions between pluripotent and germ cell fates, or vice versa, are accompanied by large-scale epigenetic reprogramming in vivo (Cantone and Fisher, 2013; Lee et al., 2014; Santos et al., 2002; Sanz et al., 2010), and these are recapitulated during similar transitions induced in vitro (Ishikura et al., 2016). To determine the extent to which BPS-induced epimutations persist during a pluripotent to germline transition in vitro, we first exposed iPSCs to 1 μM BPS and then differentiated the exposed iPSCs first into epiblast-like cells (EpiLCs) and then into PGCLCs, recapitulating the early germline epigenetic reprogramming event that normally occurs in vivo (Kurimoto and Saitou, 2018; Ohta et al., 2017; Figure 7a). We then used genome-wide analyses by EM-seq and RNA-seq to compare numbers of exposure-specific DMCs and DEGs, respectively, in PGCLCs derived from BPS-exposed iPSCs with those in the directly exposed iPSCs and found lower, but still substantial numbers of both (28,168 vs 38,105 DMCs, and 1437 vs 1637 DEGs in the derived PGCLCs compared to the directly exposed iPSCs, respectively; Figure 7b and c).

When we compared the specific DMCs and DEGs that were detected in the BPS-exposed iPSCs with those that were detected in the subsequently derived PGCLCs, we found that >90% of each were not conserved during the pluripotent to germline transition in cell fate. Specifically, only 3.7% of the DMCs and 8.4% of the DEGs detected in the BPS-exposed iPSCs were also detected in the PGCLCs derived from the exposed iPSCs. Among the small portion of specific 138 DEGs (Supplementary file 1) that were conserved from exposed iPSCs to derived PGCLCs, several (12.32%) (Cdkn1a, Ccnd2, Plk2, Tgfbr1, Gadd45g, Lck, Ltbr, Mad2l1, Ap3m2, Ctsz, Tcirg1, Gusb, Id2, Lefty2, Gstm7, Acsl1, Slc39a14) were involved in cell cycle and apoptosis pathways which could potentially be linked to cancer development (Figure 7—figure supplement 1). Thus, of the 38,105 DMCs and 1637 DEGs induced by exposure of iPSCs to 1 µM BPS, 36,688 and 1499, respectively, did not persist during differentiation of iPSCs to form PGCLCs, and so were apparently corrected by germline reprogramming. Simultaneously however, 26,752 novel DMCs and 1299 novel DEGs appeared in the derived PGCLCs that were not present in the BPS-exposed iPSCs, so were apparently generated de novo during the germline reprogramming process. These results are consistent with the notion that germline epigenetic reprogramming corrected many of the epimutations that were present in the BPS-exposed pluripotent cells from which they were derived, but that exposure of cells to EDCs may disrupt the underlying chromatin landscape in a way that then interferes with subsequent reprogramming such that in addition to correcting many previously induced epimutations, the germline reprogramming process, acting on a disrupted chromatin landscape, also generates many novel epimutations de novo during the pluripotent to germline transition.

Mice were used as the primary source of cells for the establishment of cultures for the analysis of differential cell-type specific susceptibility to exposure to the EDC BPS at the cellular level. All mice were euthanized prior to dissection and isolation of the tissue or cell type to be used for experiments. All animals were bred on-site and maintained in one of the UTSA on-campus vivaria under controlled temperature and humidity conditions in a fixed 12 hr light, 12 hr dark cycle with free access to 5V5R extruded food and non-autoclaved conventional RO water. 5V5R has been verified to contain a targeted level of 50 PPM total phytoestrogenic isoflavones (genistein, daidzein, and glycitein) and has been certified for use with estrogen-sensitive protocols limiting the effect of these plant-derived compounds. Euthanasia was performed by trained personnel using continuous CO₂ exposure at a rate of 3 L/min until one minute after breathing had stopped. Euthanasia was confirmed by cervical dislocation. Tissue from euthanized mice was used to obtain somatic (Sertoli, granulosa, and mouse embryonic fibroblast [MEF]) cells, and primordial germ cells [PGCs]. Both Sertoli and granulosa cells were isolated from respective male and female mice euthanized at postnatal day 20 (P20). MEFs and PGCs were isolated from fetuses at embryonic day 13.5 postcoitum (E13.5).

Cells were immunolabeled as previously described (Rodig, 2022) to validate the purity of cell primary cultures using known cell specific markers (Figure 2—figure supplement 2) to detect the presence or absence of relevant endocrine receptors ERα and ERβ at the protein level. Pluripotent (iPS) and somatic (Sertoli and granulosa) cells were grown on 13 mm Thermanox plastic coverslips (174950) prior to fixation, permeabilization, and immunolabeling (Nalge Nunc International, NY USA). Non-adherent germ (PGC and PGCLC) cells were spun down onto poly-L-lysine coated slides (63410–01) at 400 RPM for one minute using a Thermo Shandon CytoSpin III Cytocentrifuge from Rakin (Electron Microscopy Sciences, PA USA & Rakin Biomedical Corporation, MI USA). Cells were fixed with 4% formaldehyde for 10 min at room temperature (RT) and then washed three times for five minutes each with ICC buffer (PBS containing 0.01% Triton X-100 detergent) to permeabilize the cell and nuclear membranes prior to blocking with 5% goat serum which was added to the ICC buffer and incubated for 1 hr at RT. Following blocking, primary antibodies, in ICC buffer, were added to the slides and left to incubate overnight at 4 °C. The following day, slides were washed three times with ICC buffer, and fluorescent secondary antibodies in ICC buffer were then added to the slides and left to incubate in the dark for 1 hr. After secondary antibody incubation, cellular nuclei were stained with DAPI at a 1:1000 dilution in ICC buffer for 7 min, followed by three final washes with ICC buffer for 5 min. Coverslips then were transferred to microscope slides and mounted with 5–10 μL of VECTASHEILD Antifade Mounting Medium (H-1000) and sealed with clear nail polish before being imaged on a Zeiss AXIO Imager.M1 Fluorescence Microscope (Vector Laboratories Inc, CA, USA and Zeiss Group, Oberkochen DE). Images were processed for contrast and brightness enhancement and for the addition of scale bars using Fiji (RRID:SCR_002285; Schindelin et al., 2012). Information about the primary and secondary antibodies along with catalogue numbers and dilutions used can be found in the Key resources table.

Total RNA was extracted from 3 replicate preps of cells as described previously (Rio et al., 2010) and treated with 1.5 U/μg total RNA RQ1 DNase1 (M6101) to remove contaminating genomic DNA (Promega Corporation, WI USA). Fifty ng of cleaned RNA was retrotranscribed with SuperScript III as recommended by the manufacturer (Invitrogen, MA, USA). Primers were designed using Primer-BLAST (RRID:SCR_003095; Ye et al., 2012) from NCBI. A complete list of all primer sequences used in this study can be found in Supplementary file 6. Relative expression levels of selected genes were assessed by real-time PCR using the PowerTrack SYBR Green Master Mix according to the manufacturer’s instructions (Applied Biosystems, MA USA) then run on a QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific, MA, USA) and analyzed with QuantStudio Design and Analysis Software. Each sample was normalized based on constitutive expression of the Gusb reference gene to obtain the ΔCt [(2-Ctgene-CtGusb)].

Despite numerous studies illustrating the dangers of estrogenic mimetic EDCs, the EPA has currently not published any limits concerning the maximum acceptable dose of BPS below which exposure on a daily basis is considered to be safe. Thus for this study, we established our initial testing range of 1 μM, 50 μM, and 100 μM of BPS based on the limit established for BPA at ≤ 4.44 μM (Ribeiro et al., 2019) selecting 1 dose below the established safe limit and two doses that exceed that limit. For all cell types except PGCLCs, BPS was dissolved in absolute ethanol and added to media gassed with 5% CO₂, 5% O₂, and balanced N₂ and then injected into three replicate sealed T-25 cell culture flasks and left to incubate for 24 hr (Figure 1—figure supplement 3). The media was gassed to ensure enrichment to 5% CO₂, which is normally regulated by the cell incubator, limiting the potential for the pH to change in the media during this exposure period due to the absorption of CO₂ by incubating cells. The concentration of either BPS dissolved in ethanol for treatment groups or ethanol vehicle alone for control groups made up 0.02% of the total media volume. Following 24 hr of exposure, media containing BPS was removed and cells were washed and allowed to recover in fresh untreated media (without BPS or EtOH) for 24 hr prior to harvesting. PGCLCs were not suitable for this exposure paradigm as EpiLCs undergo differentiation to PGCLCs in cell aggregates formed in low-adherence round-bottom 96-well plates and cannot be maintained in T-25 sealed flasks. Therefore, diluted BPS was added to PGCLC media and added to cells in three replicate batches of round-bottom 96-well plates to incubate for 24 hr prior to a shortened wash and ‘chase’ period of 8 hr prior to cell sorting.

A total of 1 μg of extracted genomic DNA from each of three replicate exposure experiments for each cell type was bisulfite-converted with the EZ DNA Methylation Kit (D5001) and modified according to the manufacturer’s recommendations (Zymo Research, CA USA and Illumina, CA USA). These samples were run on the Infinium Mouse Methylation BeadChip Array following the Illumina Infinium HD Methylation protocol. This beadchip array includes 297,415 cytosine positions within the mouse genome (CpG sites, non-CpG sites, and random SNPs). The methylation score for each CpG is represented as a β-value which is a ratio of the fluorescence intensity ranging between 0 (unmethylated) and 1 (methylated). Arrays were scanned by HiScan (Illumina, CA, USA). Quality control (Figure 2—figure supplement 4) and downstream data processing of the.idat files was using the Sensible Step-wise Analysis of DNA Methylation BeadChips (SeSAMe) Bioconductor package (Ding et al., 2023; Triche et al., 2013; Zhou et al., 2022; Zhou et al., 2018). DNA methylation levels of differentially methylated cytosines (DMCs) are determined using mixed linear models. This general supervised learning framework identifies CpG loci whose differential methylation is associated with known control vs. treated co-variates. CpG probes on the array were defined as having differential changes that met both p-value and FDR ≤ 0.05 significant thresholds between treatment and control samples for each cell type analyzed. Additionally, we followed up our DNA methylation analysis of individual dCpGs by identifying differentially methylated regions (DMRs). DMRs were created by grouping all CpGs measured on the array into clusters using Euclidean distance (Ding et al., 2023; Triche et al., 2013; Zhou et al., 2022; Zhou et al., 2018). The p-values from the differential methylation of individual CpGs within the resulting CpG clusters were aggregated, and clusters were then filtered selecting for regions that contained a p-value ≤ 0.05 (Ding et al., 2023; Triche et al., 2013; Zhou et al., 2022; Zhou et al., 2018).

Total RNA was extracted from 3 replicate preps of cells using Trizol as previously described (Rio et al., 2010). Contaminating genomic DNA was removed by RQ1 DNase (M6101) treatment according to the manufacturer’s instructions (Promega Corporation, WI USA). RNA concentration was determined using Qubit (Q32855) and RNA integrity (RIN) scores were determined using tape station (5067–5576) according to the manufacturer’s instructions (Thermo Fisher Scientific Inc, MA, USA and Agilent Technologies, Inc CA, USA). Strand-specific RNA-seq libraries were prepared with the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina sequencing (E7760S) according to the manufacturer’s protocol (New England Biolabs, MA, USA). Briefly, this process consisted of poly(A) RNA selection, RNA fragmentation, and double stranded cDNA generation using random oligo(dT) priming followed by end repair to generate blunt ends, adaptor ligation, strand selection, and polymerase chain reaction amplification to generate the final library. Distinct index adaptors were used for multiplexing samples across multiple sequencing lanes. Sequencing was performed on an Illumina NovaSeq 6000 instrument yielding sequences of paired end 2x50 base pair runs. Demultiplexing was performed with the Illumina Bcl2fastq2 program (Illumina, CA, USA).

The quality of the fastq reads was checked using FastQC (RRID:SCR_014583; Andrews et al., 2023; de Sena Brandine and Smith, 2019; Figure 7—figure supplement 4). Reads were aligned to the mm10 mouse reference genome using the Rsubreads package to produce read counts (RRID:SCR_016945; Liao et al., 2019). These were then used for differential gene expression analysis using the edgeR package (RRID:SCR_012802; Chen et al., 2016; McCarthy et al., 2012; Robinson et al., 2010). Briefly, gene counts were normalized using the trimmed mean of M-value normalization (TMM) method before determining counts per million (CPM) values (Robinson and Oshlack, 2010). For a gene to be classified as showing differential gene expression between BPS-treated and EtOH vehicle-only control samples, a threshold of both a Benjamini-Hochberg adjusted p-value ≤0.05 and a false discovery rate (FDR)≤0.05 had to be met.

Genomic DNA was extracted from three replicates of cells as previously described (Sambrook and Russell, 2006). Smaller fragmented DNA (≤10 kb) and contaminating RNA were removed by cleaning the genomic DNA on a genomic DNA clean and concentrator-10 column (D4011), according to the manufacturer’s instructions (Zymo Research, CA, USA). DNA concentration was determined using a Qubit (Q32850) and genomic DNA (100 ng) was sheered using a Bioruptor and the size of the sheered DNA was determined using a TapeStation 4200 according to the manufacturer’s instructions in the UTSA Genomics Core (Diagenode Inc, NJ, USA and Agilent Technologies, Inc CA, USA). Large size (470–520 bp) EM-seq libraries were prepared with the NEBNext Enzymatic Methyl-seq kit (E7120S) according to the manufacturer’s protocol (New England Biolabs, MA, USA). Briefly, this process consisted of A-tailing, adaptor ligation, DNA oxidation by TET2 initiated by the addition of Fe (II), strand denaturization with formamide, deamination by APOBEC3A, polymerase chain reaction amplification, and bead selection to generate the final libraries. Distinct index adaptors were used for multiplexing samples across multiple sequencing lanes. Sequencing was performed on an Illumina NovaSeq 6000 instrument yielding sequences of paired end 2x150 base pair runs. Demultiplexing was performed with the Illumina Bcl2fastq2 program (Illumina, CA, USA).

EM-seq data was processed using the comprehensive wg-blimp v10.0.0 software pipeline (Lehle and McCarrey, 2023; Wöste et al., 2020). In brief, reads were trimmed prior to initiating the wg-blimp pipeline using Trim Galore (RRID:SCR_011847; https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/). Sequenced reads were aligned to the mm10 genome using gemBS (King et al., 2020; Merkel et al., 2019; Schilbert et al., 2020). The BAM files from alignment underwent a series of QC tests including read quality scoring by FastQC (RRID:SCR_014583; Andrews et al., 2023), overall and per-chromosome read coverage calculation, GC content, duplication rate, clipping profiles by Qualimap (RRID:SCR_001209; Okonechnikov et al., 2016), and deduplication by the Picard toolkit (RRID:SCR_006525; Figure 7—source data 1). Methylation calling was performed by MethylDackel (Ryan, 2023a; https://github.com/dpryan79/MethylDackel, copy archived at Ryan, 2023b) and statistically significant DMC/DMR calling was performed by the metilene (Jühling et al., 2016) and BSmooth (RRID:SCR_005693; Hansen et al., 2012) algorithms. Metilene uses a binary segmentation algorithm combined with a two-dimensional statistical test that allows the detection of DMCs/DMRs in large methylation experiments with multiple groups of samples. BSmooth uses a local-likelihood approach to estimate a sample-specific methylation profile, then computes estimates of the mean differences and standard errors for each CpG to form a statistic similar to that used in a t-test. Finally, potential regulatory regions were identified through the use of MethylSeekR (RRID:SCR006513; Burger et al., 2013). Results from the pipeline were displayed in the wg-blimp interactive results web browser that was built using the R Shiny local browser hosting framework (RRID:SCR_001626; Chang, 2023).