C57BL/6J mice originally obtained from Jackson Laboratories (Bar Harbor, ME) were bred and housed in same-sex colonies of up to four individuals with unrestricted food and water access at the Animal Care Facility at Western University (London, Ontario, Canada). Temperature and humidity were maintained at 21°C–24°C and 40%–60%, respectively, with lighting on a 14:10-h light-dark cycle. All housing, bedding, and nest material were consistent between cages. All treatment protocols complied with ethical standards established by the Canadian Council on Animal Care and were approved by the Animal Use Subcommittee at Western University (London, Ontario, Canada).

Females (10 weeks old) were housed individually and randomly assigned to ethanol or control groups. Both groups had free access to water, with the ethanol group additionally provided free access to 10% ethanol in water solution (Kleiber et al., 2011; Alberry and Singh, 2016). Ethanol was introduced with increasing concentrations of 2%, 5%, and 10% after 48 h with the previous concentration. After 10 days of continuous 10% ethanol availability, females were mated with males (15 weeks old) with only water available. Males were removed after a 24-h period and 10% ethanol was returned, which represented gestational day 0. Until postnatal day 10, 10% ethanol was freely available, with only water available for the remainder of this study. Blood alcohol levels were not taken to minimize maternal stress, voluntary consumption of 10% ethanol at 14 g ethanol per kg body weight per day produces peak blood alcohol levels of 120 mg dl^-1 (Allan et al., 2003). This study represents a modest amount of ethanol exposure, with experimental mice consuming 8 g ethanol per kg body weight per day on average, sufficient to produce behavioral deficits in their offspring, including hyperactivity in a novel environment, hypoactivity in a home cage environment, and learning deficits (Alberry and Singh, 2016).

Postnatal maternal separation and isolation were used to induce early life stress, as described previously (Savignac et al., 2011; Benner et al., 2014; Alberry and Singh, 2016). On postnatal day 2, half of each litter was randomly selected for stress, tail coloring with permanent marker was used to distinguish pups. From postnatal days 2 to 14, stress group pups were removed and isolated for 3 hours per day during the light phase in 8 oz. Dixie cups with bedding and nest material. Pups not selected for maternal separation remained with the dam and other littermates during this time. On postnatal day 21, pups were weaned and housed with same-sex littermates with 2–4 individuals per cage. Following treatments, mice belonged to one of four groups—control, ethanol, stress, or ethanol + stress.

Male mice were sacrificed via carbon dioxide asphyxiation and cervical dislocation on postnatal day 70. The hippocampus was dissected from the whole brain as described previously (Spijker, 2011). With samples stored in phosphate-buffered saline (PBS), snap-frozen in liquid nitrogen, and kept at −80°C. Samples were ground by pestle over liquid nitrogen to create a powder. Stages of buffer RLT (Qiagen, Valencia, CA) were added and mixed by pipetting. Sampled were centrifuged after a 10-min incubation period. The supernatant was loaded onto AllPrep DNA spin columns and the AllPrep DNA/RNA Mini Kit Protocol (Qiagen, Valencia, CA) was followed to isolate DNA and RNA from the same sample. DNA quantification was determined by NanoDrop 2000c Spectrophotometer (Thermo Fisher Scientific, Wilmington, DE).

Three DNA samples were randomly selected from each treatment group (12 total DNA samples from individual mice processed separately) and sent on dry ice to Arraystar Inc. (Rockville, MD), with one sample per group as an input sample for sequence control. DNA was quantified by NanoDrop 1000 (Thermo Fisher Scientific, Wilmington, DE), then fragmented to a range of 200–1,500 bp using a Diagenode Bioruptor, end-repaired, and 3' adenylated for ligation of genomic adapters. Fragments were immunoprecipitated by anti-5-methylcytosine antibody, then PCR amplified. AMPure XP beads were used to select fragments from 300–800 bp, then quantified by Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Following denaturation with 0.1 M NaOH, single-stranded DNA molecules were captured and amplified in situ on an Illumina flow cell. Libraries were sequenced on the Illumina HiSeq 4000 platform using the HiSeq 3000/4000 SBS Kit (300 cycles) protocol, then Off-Line Basecaller (OLB V1.8) was used for base calling. Reads were aligned to the mouse genome (UCSC MM10) using HISAT2 (V2.1.0) (Kim et al., 2015) after passing the Solexa CHASTITY quality filter and resulting BED files were used for differential methylation analysis. Sequences have been made available at GEO accession number GSE137984.

Aligned reads from HISAT2 were used for peak calling using Model-based analysis for ChIP-Seq (MACS) (version 1.4.2) (Zhang et al., 2008), where enriched regions (peaks) for each sample were identified by comparison to input background samples for each group using a dynamic Poisson distribution (q < 10^-5). MeDIP enriched regions were annotated to the nearest gene using UCSC RefSeq, with differentially methylated regions (DMRs) in the promoter of known genes for Ethanol vs. Control, Stress vs. Control, as well as Ethanol + stress vs. Control identified by diffReps (cut-off: log₂FC = 1.0, p = 10^-4) (Shen et al., 2013). Promoter regions are defined here as 2,000 bp upstream and downstream from the transcription start site of a gene. Generalized hypergeometric tests for enrichment of gene ontology (GO) terms and Kyoto encyclopedia of genes and genomes (KEGG) pathways were used for genes represented by promoter DMRs for each treatment group using goana and kegga functions in the limma software package (Ritchie et al., 2015) in R, filtered by significance (p < 0.05).

Genes represented by promoter DMRs for each treatment group were compared to previously reported transcript-level expression analysis from an RNA-Seq experiment using RNA isolated from the same samples (Alberry et al., 2019). Briefly, RNA-Seq was carried out by The Centre for Applied Genomics (The Hospital for Sick Children, Toronto, Ontario, Canada). Paired-end reads were pseudo-aligned to version 38 of the Ensembl mouse transcriptome via kallisto (Bray et al., 2016) and differential expression analysis was determined using sleuth (Pimentel et al., 2017; Yi et al., 2018). Gene lists used here were filtered by significance (p < 0.01), resulting in a list of 164 unique transcripts altered following ethanol exposure, 116 for stress, and 217 for the combination of ethanol + stress. Sequence data for the RNA-Seq experiment is available at the Gene Expression Omnibus (GEO), accession number GSE133369.

DMRs (DMRs) were determined for each experimental treatment group compared to controls, then associated with annotated genes via proximity to transcription start sites. Promoter regions are defined here as 2 kb upstream or downstream from the transcription start site of a gene, with genes with promoter DMRs considered for this analysis. In this way, several DMRs considered spatially distinct can be present in the promoter region of a single or multiple gene(s). A gene implicated by more than one DMR is considered separately when the DMRs are in opposition, but only once if the DMRs are in the same direction. Using promoter DMRs, 1,264 genes are implicated by prenatal ethanol treatment (Figure 1A, details in Supplementary Table 1). Following postnatal maternal separation stress, 1,472 genes are represented by promoter DMRs (Figure 1B, details in Supplementary Table 2). When mice faced both ethanol + stress, 958 genes are implicated by promoter DMRs (Figure 1C, details in Supplementary Table 3). Interestingly, while the ethanol group has a similar number of genes implicated by hypo- and hypermethylated promoter DMRs, the stress group genes are represented by 73% hypomethylated promoter DMRs. Genes implicated by promoter DMRs in the ethanol + stress group are also more evenly distributed between hypo- and hypermethylated. While a single DMR is either hypo- or hypermethylated, a single gene may be represented by multiple DMRs with different states. There are between 16 and 29 genes in each comparison that are implicated by both hypo- and hypermethylation of promoter DMRs.

Hypo- and hypermethylated DMR-associated genes for each treatment group were analyzed separately for GO and KEGG pathway enrichment. In the ethanol group, genes related to voltage-gated calcium channel activity tended to be hypomethylated at their promoter regions (p < 0.007) (Table 1). In addition, brain-related terms including NMDA receptor activity (p = 0.001), positive regulation of behavior (p = 0.002), circadian rhythm (p = 0.003), and apical dendrite (p = 0.03) were all biological processes overrepresented by genes with hypomethylated promoters following ethanol exposure. Interestingly, components responsible for transcriptional regulation are also implicated by genes with hypomethylated promoters, including PRC1 complex (p = 4.04 x 10^-4), transcription factor TFIID complex (p = 0.03), and RNA polymerase II transcription factor complex (p = 0.046). Genes related to phosphoinositide kinase (p < 0.02) and cytoskeleton regulation (p = 0.005) were hypermethylated at their promoter regions following prenatal ethanol exposure. In the stress group, genes important for histone modifications (p < 0.01) and neuronal organization (p = 0.007) were overrepresented by genes with hypomethylated promoters (Table 2). Conversely, genes with hypermethylated promoters following stress are important for cytokine interactions (p < 0.006). When mice faced both ethanol + stress, genes important for neuron migration (p < 0.005) were hypomethylated at their promoters (Table 3). As with stress alone, the combination of ethanol + stress results in promoter hypermethylation for genes responsible for cytokine interactions (p < 0.01).

While genes implicated by promoter DMRs are unique to each treatment, there are regions of overlap between treatments (Figure 2). 120 genes are implicated by promoter DMRs following ethanol, stress, as well as ethanol + stress (Figure 2A). When directionality is considered, there are 85 genes implicated by hypomethylated promoter DMRs in all three treatment groups (Figure 2B). Finally, there are 19 genes implicated by hypermethylated promoter DMRs in all three treatment groups (Figure 2C). Genes consistently implicated by promoter DMR hypomethylation tend to be important for histone methylation (p < 0.04) and cyclic adenosine monophosphate (cAMP) protein kinase activity (p < 0.04) (Table 4). Conversely, genes implicated by promoter DMR hypermethylation are important for ion channel activity (p < 0.007) as well as neuroactive ligand-receptor interactions (p = 0.04) and chemokine signaling (p < 0.003).

Since hippocampal transcriptome profiles exist for these mice, genes implicated by altered promoter methylation could be compared to genes with altered transcription. The number of genes represented by altered promoter methylation following each treatment is represented by the direction of transcriptional change following the RNA-Seq study previously reported are presented here (Table 5). While many genes are included for comparison, the overlap between genes implicated by altered DNA methylation and transcription is minimal. For ethanol mice, 1,264 genes with promoter DMRs were compared to the previously reported 164 unique transcripts altered in these samples. There are two corresponding transcripts with hypomethylated promoters and increased expression, Slc2a9, solute carrier family 2 (facilitated glucose transporter), member 9, and Blvrb, biliverdin reductase B (flavin reductase (NADPH)). Conversely, for hypermethylated promoters, there are two transcripts with decreased expression following PAE, Eif4g1, eukaryotic translation initiation factor 4, gamma 1, and Myo6, myosin VI.

In the stress group, 1,472 genes with promoter DMRs were compared to the 116 unique transcripts altered in these samples. There are three genes with hypomethylated promoters and increased transcript expression, Arpp21, cyclic AMP-regulated phosphoprotein, 21, Mark2, microtubule affinity regulating kinase 2, and Rab33b, RAB33B, member RAS oncogene family. One gene, Slc39a11, solute carrier family 39 (metal ion transporter), member 11, is hypermethylated at its promoter and has reduced transcript expression following maternal separation.

For ethanol + stress mice, 958 genes with promoter DMRs were compared to the 217 unique transcripts altered in these samples. Interestingly, Arpp21 is also hypomethylated at its promoter and overexpressed following ethanol + stress treatment alongside two additional genes, Dlg3, discs large MAGUK scaffold protein 3, and Pinx1, PIN2/TERF1 interacting, telomerase inhibitor 1. Conversely, two genes share increased promoter methylation and decreased gene expression, Slu7, pre-mRNA-splicing factor SLU7, and Unc119, unc-119 lipid binding chaperone. While the number of genes with altered promoter methylation and expression resulting from each treatment or combination of treatments is minimal, most of these genes have known brain functions in the literature, ranging from neuronal migration to synapse formation and plasticity (Table 6).

Prenatal ethanol exposure results in specific changes in promoter DNA methylation. We report promoter hypomethylation of genes important for brain function and formation as well as for transcriptional regulation following prenatal ethanol exposure. We also found promoter hypermethylation of genes important for phosphoinositide kinase activity as well as cytoskeleton regulation. These alterations may be lasting evidence of altered control of gene expression earlier during development that persists into adulthood. The ethanol exposure model employed here has been used previously, with lasting alterations to whole-brain DNA methylation profiles (Laufer et al., 2013). While we report comparable amounts of hypo- and hypermethylated gene promoters following continuous gestational ethanol exposure, previous work using targeted binge exposure trimester three equivalent models have found alcohol-induced hypermethylation in the hippocampus in both mice (Chater-Diehl et al., 2016) and rats (Otero et al., 2012). Conversely, a model of ethanol exposure more representative of trimester two equivalent exposure in mice found global hypomethylation in the neocortex, but this pattern is more complex when brain structures are independently assessed (Öztürk et al., 2017). Our work uses a continuous exposure model and is aimed at investigating promoter methylation, highlighting the importance of dose and timing of ethanol exposure as well as specific methylation patterns being observed.

We report more genes with hypo- than hypermethylated promoters following early life stress via maternal separation. Of these, hypomethylated genes tend to be important for histone methylation and chemokine activity. Meanwhile, hypermethylated promoters tend to be of genes important for cytokine interactions. Early life stress results in lasting changes to DNA methylation in the hippocampus of rats (Kosten et al., 2014). In humans, child abuse has been associated with hypermethylation of several genomic loci in the hippocampus (McGowan et al., 2009). Meanwhile, maternal separation in rats induces an inflammatory immune response in the hippocampus (Roque et al., 2016). Interestingly, early life stress has been associated with reduced serum cytokine levels, suggesting that hypermethylation of related genes may not be hippocampal specific, rather playing a more ubiquitous role in the body (Raineki et al., 2017).

The combination of prenatal ethanol exposure and maternal separation stress resulted in the hypomethylation of genes important for neuronal migration and hypermethylation of genes related to cytokine interactions. Similarly, genes that were implicated by hypermethylated promoters following any of the treatments were also important for chemokine and cytokine interactions. It is apparent from Figure 1 that the addition of early life stress to individuals that faced prenatal ethanol exposure does not result in an additive DNA methylation response. Few genes are shared between ethanol only or stress only and the combination of treatments as implicated by promoter DMRs. Prenatal ethanol exposure followed by early life stress has been associated with altered immune system development, with early life adversity differentially affecting peripheral and central immune response depending on previous ethanol exposure (Raineki et al., 2017).

We previously reported changes in gene expression in association with behavioral deficits following prenatal ethanol exposure and maternal separation in these mice (Alberry et al., 2019). Interestingly the module of genes most closely associated with treatment and anxiety-like behavior includes genes important for transcriptional regulation and neurodevelopment. With these results in mind, we examined what overlap exists between genes with altered expression and altered promoter methylation that is complementary.

Following prenatal ethanol exposure, two genes have decreased promoter methylation and corresponding increased expression—Slc2a9 and Blvrb. Slc2a9 codes for a fructose transporter that acts as a uric acid transporter (Le et al., 2008). While not much is known regarding its role in the brain, SLC2A9 SNPs have been associated with better memory performance (Houlihan et al., 2010), and social phobia (Lyngdoh et al., 2013). Blvrb codes for an oxidoreductase important for heme degradation, with no clear link to brain function. Prenatal ethanol exposure results in promoter hypermethylation and decreased transcript expression for two genes, Eif4g1 and Myo6. Eif4g1 produces a translation initiation factor, eIF4G, with mutations present in familial Parkinson Disease (Chartier-Harlin et al., 2011). In fact, this gene acts as a scaffold protein with others in the hippocampus in ways that are perturbed in neurodegeneration likely through protein misfolding (Dhungel et al., 2015). Myo6 is an actin-based motor protein that has been previously linked to stress response in the mouse hippocampus (Tamaki et al., 2008). Mice lacking this gene have fewer synapses and shorter dendritic spines (Avraham et al., 1995). Myo6 is important for postsynaptic stimulus-dependent α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid-type glutamate receptor (AMPAR) internalization (Osterweil et al., 2005), as well as for brain-derived neurotrophic factor (BDNF)-mediated glutamate release from presynaptic terminals in long-term potentiation (Yano et al., 2006). Together, these results suggest the combined accumulation of DNA methylation, followed by reduced expression of these transcripts may be important for the altered brain-related phenotypes observed in these mice following prenatal ethanol exposure.

There are three genes with promoter hypomethylation and a corresponding increase in transcript expression following maternal separation stress, Arpp21, Mark2, and Rab33b. Arpp21 is also implicated by the combination of ethanol + stress through promoter hypomethylation and increased expression. The ARPP21 protein product is composed of conserved domains for RNA-binding, with the ability to interact with eIF4G and preferentially bind to the 3' untranslated regions of mRNAs. Additionally, overexpression and knockdown experiments of ARPP21 show it positively regulates dendrite growth (Rehfeld et al., 2018). While we report increased Arpp21 expression following stress or ethanol + stress, stress-induced decreases have also been reported (Constantinof et al., 2019). Mark2 produces a serine/threonine-protein kinase involved in cell polarity and microtubule dynamics, as well as in neuron migration (Sapir et al., 2008; Mejia-Gervacio et al., 2012). Rab33b is a member of Rab family proteins that regulate intracellular vesicular trafficking, particularly at the Golgi apparatus (Zheng et al., 1998). Recently, the zebrafish ortholog of Rab33b, rab33ba, has been shown to mediate axon outgrowth in the forebrain (Huang et al., 2019). Slc39a11 is the only gene that has promoter hypermethylation and overexpression in the stress group. It codes for a Zrt-, Irt-like protein (ZIP) for zinc transport (Yu et al., 2013). Taken together, these results suggest that the loss of methylation and subsequent overexpression of these genes following postnatal maternal separation stress may lead to observed changes in behavioral phenotypes mediated by improper neuronal migration and growth.

The combination of ethanol + stress treatments results in three genes with increased expression and corresponding promoter hypomethylation, Arpp21, Dlg3, and Pinx1. Arpp21 was introduced previously as it was also implicated by stress alone. Of these, Dlg3 is one of the most significantly altered transcripts in response to Stress and Ethanol + Stress treatments that we previously reported (Alberry et al., 2019). It codes for synapse-associated protein 102, a protein important for synapse formation. Knockout mice have impaired synaptic plasticity and spatial learning (Cuthbert et al., 2007), while human DLG3 mutations result in nonsyndromic X-linked intellectual disability (Tarpey et al., 2004). Pinx1 is a telomerase inhibitor with little known brain function, although variants have been associated with late-onset Alzheimer disease (Tosto et al., 2019). Two genes share decreased promoter methylation and increased expression following the combination of treatments, Slu7 and Unc119. Slu7 is a pre-mRNA splicing factor that forms part of the spliceosome and is important for genome stability (Jiménez et al., 2019). Unc119 codes for a lipid binding chaperone required for G protein localization in sensory neurons (Zhang et al., 2011). In fact, knockdown of Unc119 results in neurotoxicity and decreased dendritic branching (May et al., 2014). These results suggest that dysregulation of genes important for neuronal migration and synapse formation as well as genome integrity via altered promoter methylation following the combination of ethanol + stress may underlie the complex and heterogeneous behavioral changes observed in these mice.

Promoter DNA methylation is an essential mechanism of epigenetic control for the expression of genes. It is affected by environmental factors, including ethanol and stress. Changes in promoter DNA methylation status can lead to the initiation and maintenance of gene expression over time. Promoter DNA methylation has been the research focus in FASD, however, DNA methylation does not act in isolation. While we focus on DMRs in gene promoters, methylation status at a single CpG site may be sufficient to influence gene expression. In this analysis, we may miss the influence of a single CpG that has this type of influence without resulting in a significantly altered region of DNA methylation. Similarly, we have focused on promoter DNA methylation for its known influence on gene expression via altered transcription factor binding. Gene body or other intergenic changes in methylation would not be captured in this analysis and may be important for gene expression. Phenotypic outcomes are often the result of interrelated epigenetic mechanisms. For example, noncoding RNAs are also important for the regulation of gene expression at the transcriptional and posttranscriptional levels. They include long noncoding RNAs (lncRNAs) (Lee, 2012), microRNAs (miRNAs) (Chuang and Jones, 2007), short interfering RNAs (siRNAs), and piwi-interacting RNAs (piRNAs) (Holoch and Moazed, 2015). In FASD research, the noncoding RNA research has been focused on the relationship between microRNAs and their mRNA gene targets. Several microRNAs are altered in reciprocal direction to their gene targets following prenatal alcohol exposure (Balaraman et al., 2013; Laufer et al., 2013), offering another potential avenue for control. Additionally, modifications to histone tails at specific genomic loci are the foundation of epigenetic mechanisms, with a wide variety of known potential modifications (Allis and Jenuwein, 2016). In FASD mouse models, altered histone modifications (Veazey et al., 2015; Chater-Diehl et al., 2016), as well as enhanced lysine dimethyltransferase activity resulting in increased histone H3 lysine 9 and 27 dimethylation and neurodegeneration have been reported (Subbanna et al., 2013). Such results suggest changes in gene-specific expression in response to prenatal ethanol exposure must be the result of a variety of epigenetic mechanisms. Although DNA methylation, noncoding RNAs, and histone modifications have all been implicated, their inter-relationship is not known.

The promoter DNA methylation results included in this report are novel in that they were obtained from the same mice that were used to assess behavioral changes (Alberry and Singh, 2016) and hippocampal gene expression (Alberry et al., 2019). This allows for assessment of the relationship between promoter DNA methylation and expression of that gene in response to prenatal ethanol exposure, postnatal maternal separation, and the combination of both treatments. Our results implicate many genes via altered expression or promoter DNA methylation, and these genes are relevant in the context of their involvement in ethanol or stress response. In fact, many genes implicated are important for brain development and function. What is apparent is that changes in expression of only a small subset of genes is compatible with changes in promoter DNA methylation. The results argue that promoter methylation is only partly involved in the environmentally induced changes in gene expression we have modeled. Epigenetic mechanisms other than promoter DNA methylation therefore are likely responsible for the lasting gene expression alterations and behavioral abnormalities seen in pups subjected to prenatal ethanol exposure, maternal separation, and the combination of the two treatments. Although logical, such conclusions must be assessed experimentally, as the hippocampus represents a heterogenous cell population. Also, the interventions used are relatively mild, applied during neurodevelopment, with methylation status assessed in adulthood. Not surprisingly, this experimental paradigm has resulted in subtle changes in both gene expression and DNA methylation. Future research must focus on experimentally establishing mechanisms involved in altered gene expression and maintenance of these changes in different cell types within the brain. This will require modified experimental protocols and the use of single cell sequencing. Greater understanding into how these changes occur and persist has the potential to lead to the development of novel strategies for the amelioration of FASD-related deficits for children born with this common and complex disorder.