Methylation of CpG dinucleotides in the mammalian genome is a heritable epigenetic mark and also serves as an important mediator between the environment and genome function. Epigenetic modifications in regulatory and growth-related genes influence fetal development, with effects on important traits later in life (Bird, 2007). DNA methylation of cytosine residues, particularly at CpG dinucleotides, has been identified as an important regulatory mechanism of genome function, mediating genomic imprinting, embryo development (Razin and Shemer, 1995) and X-chromosome inactivation (Heard et al., 1997). CpG methylation may change through the course of development and aging (Madrigano et al., 2012), and in response to environmental conditions (e.g., nutrition; Murdoch et al., 2016) or disease (Bind et al., 2012; Leonard et al., 2012). Errors in methylation programming may cause developmental abnormalities and problems during pregnancy (Banister et al., 2011; Lambertini et al., 2011; Rexhaj et al., 2011; Jakovcevski and Akbarian, 2012). DNA methylation is mostly erased shortly after conception, followed by an increase in global DNA methylation around embryo implantation (Smith et al., 2012; Guo et al., 2014). During mammalian development the DNA methylome is then extensively remodeled in different cell lineages during differentiation. These dynamic changes lead to unique DNA methylation signatures in adult tissues, consistent with cell lineage and function (Reik et al., 2001; Bird, 2002; Smith and Meissner, 2013; Messerschmidt et al., 2014). Recent whole genome studies of humans (Yuen et al., 2011; Eckmann-Scholz et al., 2012; Slieker et al., 2015) and mice (Kremenskoy et al., 2003; Song et al., 2009) have identified tissue specific differentially methylated regions (tDMRs) during early development. The level of methylation of these regions then declines in adult tissues following an active demethylation wave during growth which continues gradually with aging.

In this work, variations in methylation among tissues during early mammalian development was assessed in all three germ layers of sheep fetuses by whole-genome bisulfite sequencing (WGBS).

The objective was to identify genomic regions with specific epigenetic signatures that differ between tissues. Sheep fetuses were used because of their similarity to human fetuses and potential for high external stressor exposure on farms. In particular we selected sheep since the developmental biology of the ovine fetus resembles closely the human situation displaying a comparable birth weight, organogenesis, and growth rate (Vonnahme et al., 2003; Wolfs et al., 2012). Moreover due to its long gestational period and tolerance for intrauterine surgery, among livestock species, the pregnant sheep is routinely used to investigate fetal programming, fetal growth, metabolism, and nutrition (Gopalakrishnan et al., 2004; McMillen and Robinson, 2005; Sinclair et al., 2007; Hyatt et al., 2008). For obvious ethical reason previous works describing tissue specific DMR on humans have been done in adult tissues (Lokk et al., 2014) or in fetal tissues collected from abortion material (Slieker et al., 2015), therefore important information about dynamics of DMR changes during normal development are still needed. In addition, these data provide a reference to investigate epigenetic changes induced by, e.g., environmental or nutritional stress during gestation (Waterland and Michels, 2007; Gluckman et al., 2009) which has not been investigated in fetal tissues of livestock species. The influence of external factors such as nutritional status during development that lead to epigenetic modifications and subsequently phenotypic variation, may in part be responsible for discrepancies between predicted breeding values and observed performance and is hence of interest for improving animal breeding and management (Murdoch et al., 2016).

In this work 4X coverage WGBS profiling was used to characterized patterns of CpG methylation in three different tissues from sheep fetuses. This coverage is more than the 2.5–3X which has been shown to be sufficient for the identification of DMRs with large methylation differences (Ziller et al., 2015).

Distribution of methylated sites observed across different gene features was similar to the DNA methylation landscape reported for human fetal tissue obtained with a 450 k methylation array (Huse et al., 2015; Slieker et al., 2015). CGIs across the genome have been found to have low DNA methylation in fetal tissues, while the level of CpG methylation is higher in shores and shelves moving away from CGIs, which is consistent with the stably methylated sequences surrounding CpG islands reported by Edgar et al. (2014).

Hypo-methylated CGIs become established during the early phase of cell lineage differentiation, and appears to be highly stable across diverse developmental states and cell types (Straussman et al., 2009). This was also observed in the present study with more than 84% hypo-methylated CGIs shared among fetal tissues which tended to be near TSSs. TSSs hypomethylation is likely established during early development and TSSs have been proposed as potential regulatory domains that mediate transcription (Deaton and Bird, 2011).

In contrast to the relatively subtle changes in hypomethylated CGIs, only a small proportion of methylome (32%) with higher level of methylation in CGIs was conserved among the three tissues studied. About 20% of the CGIs were differentially methylated among the three tissues. In a study of nearly 800 genes, Yuen et al. (2011), found that 23% of these were differentially methylated between different somatic tissues.

Each of the tissues examinated in the present study was composed of several cell types and the observed methylation profile was therefore an average. Assessment of ELISA-based global DNA methylation quantification in fetal tissues at the first trimester showed variations in the methylation level in the different tissues (Fa et al., 2016). In the present work, the endoderm had a lower overall level methylation whereas mesoderm showed fever DMRs.

Although tissue-specific methylation of CGIs was observed at many loci that are essential for development including HOX and PAX family members (Illingworth et al., 2008), it has also been shown that gene-body DNA methylation plays a role in tissue-specific gene regulation (Yagi et al., 2008; Wan et al., 2015).

Tissue-specific methylation was observed for both CpG in CGIs and CpG in gene bodies outwith CGIs. GO analysis for DMRs located in gene bodies and CGIs showed an overall enrichment of genes involved in tissue morphogenesis, regulation and development. DMRs in CGIs overlapped many genes belonging to cholinergic signaling, including ache, chrnA4, chrnB2, chrnB4. This pathway seems to be involved in nervous system development and behavior: analysis of methylation variation in dorsolateral prefrontal cortex from individuals ranging from fetal to 84 years old, showed that methyl-CpG levels of chrnB4 change over the human life span (Torabi Moghadam et al., 2016). Mice null for chrnA4, display an increased anxiety (Ross et al., 2000), and variation in CpG methylation of chrnA4 in ventral-hippocampal granule cells and neurons is also linked to anxiety (Oh et al., 2013). The nicotine receptor subunits are crucial for the correct nervous system development in the brain of rat fetuses. Previous studies suggested that prenatal exposure to nicotine significantly increased mRNA expression of brain nicotine receptor subunits α2, α4, α7, and β2 units and is associated with abnormal development in fetuses, including fetal brain damage (Lv et al., 2008).

Comparison between fetal tissue in the present study revealed variation of CpG methylation in the gene bodies of several cell adhesion proteins, including; ephA5; NRXN2, 3, wnt7A, cgh12, 13, and FZD3. Maternal environment induces offspring methyl-CpG variation in related genes including ephA4, A8, B1,B2; NRXN1, 2; wnt2, 2b, 3, 7a, 7b, 10b, 11; cdh 9, 11, 13, and FZD7 (Oh et al., 2013). Adhesion proteins play an essential role during fetal development. The variations in the methylation status of cell adhesion protein genes in placenta have been associated to pre-eclampsia, a condition during pregnancy causing a sudden rise in blood pressure and signs of damage to other organ systems, especially the kidney (Anton et al., 2014). CpGs methylation and mRNA expression in genes associated to Wnt signaling (wnt family) and neural development such as nrxn2, has been shown to be sensitive to environmental condition such as prenatal exposure to infectious or inflammatory insults (Richetto et al., 2017).

In the present study differences in CpG methylation were also found in genes related to cardiac muscle tissue development; MYO18B, MEF2A, MYH11. Deficiency of MYO18B in mice results in disruption of myofibrillar structures in embryonic cardiac myocytes (Ajima et al., 2008) while MEF2A deficiency is associated with myofibrillar disarray in embryonic heart development (Naya et al., 2002). However, no data are available linking methylation of MYP18B and MEF2A with expression in fetal tissues. Hypomethylation of MYH11 has been observed in muscle tissue from pigs that have undergone constant heat stress and were affected in muscle development (Hao et al., 2016), although MYH11 expression seems to be regulated by chromatin remodeling of histone acetylation rather than methylation status (Gomez et al., 2015).

The WGBS profiling of ovine fetal tissues presented here provides a catalog of DNA methylation patterns across the sheep genome related to three developmental layers of the embryo. Many DMRs were layer specific and enriched in genes often closely related to specific developmental functions. Knowledge of the epigenetic status of genes potentially involved in the maintenance and regulation of normal developmental and differentiation provides a starting reference point for studies on the effects of environmental and nutritional stressors on development.

PL, JW, PA-M, and AS conceived the study and revised the manuscript. PT and PS isolated fetal tissue. EC performed DNA extraction, libraries preparation and sequencing. MDC and BL carried out the bioinformatic analysis. EC and PT wrote the manuscript. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.