The human placenta is a temporary organ that facilitates the exchange of nutrients, gas and waste between maternal and fetal circulations. In order to carry out these functions, it is comprised of heterogeneous cell types including several trophoblast cell populations (cytotrophoblasts, extra-villous trophoblasts and syncytiotrophoblast), fibroblasts, mesenchymal cells, as well as fetal and maternal vascular tissue and blood cells. The extra-villous trophoblast cells must first invade the maternal decidua and remodel maternal arteries, to allow direct contact between maternal blood and the placental syncytiotrophoblast cell layer [1]. In addition to these traditional roles, the placenta is also important in shielding the developing fetus from the maternal immune system [2].

The placenta also undergoes several physiological changes throughout gestation, with one of the most significant being the flooding of placenta villi by maternal blood at the end of the first trimester (~12 weeks gestation), resulting in a rise in oxygen concentration as well as a decrease in trophoblast invasion. It is believed that the inability of the placenta to respond to this change in oxygen concentration can lead to placental disease, such as preeclampsia [3, 4].

The molecular mechanisms behind these morphological and functional changes are now beginning to be understood at both the gene-specific and genome-wide level. Wide-ranging genome-wide gene expression differences between placentas at different gestational ages were reported in two recent studies [5, 6]. Despite sampling from different locations within the placenta, many changes were found in common between the two studies, each of which reported changes in expression with increasing gestational age in genes involved in cell cycle and immune response. This suggests that gene expression changes are needed for physiological needs of the developing placenta, such as shielding the fetus from the maternal immune system [2]. Genes involved in Wnt signalling also showed expression changes over time [5, 6] that resulted in decreasing levels of β-catenin later in gestation, possibly linked to decreasing placental invasiveness [6].

The importance of epigenetic factors in placental development and function has long been known through the study of imprinted genes [7, 8] and it is increasingly clear that the placenta displays a unique epigenetic profile. However, the extent to which epigenetic modifications, specifically DNA methylation, contribute to placental function have only recently been widely examined (reviewed in [9].

Due to its role as the interface between the mother and fetus, the placenta is exposed to a myriad of environmental factors, some of which have been shown to alter placental gene expression, as well as epigenetic marks [10]. These include diet [11, 12], smoking [13], and assisted reproductive techniques [14, 15]. Mounting evidence implicates epigenetic marks, such as DNA methylation, in mediating environmentally-induced regulation of genome function. More studies into the effects of the environment on the placental epigenome are warranted due the importance of this organ in regulating pregnancy development.

Several genome-scale DNA methylation studies have focused on finding tissue-specific differentially methylated regions (tDMRs) between placenta and maternal blood, as a means of detecting placental pathologies and fetal chromosomal trisomies using non-invasive methods (reviewed in [16–19]. This strategy has recently resulted in the development of the first non-invasive blood test for Down syndrome [20]. However, we, and others have revealed substantial inter-individual DNA variation in placental methylation profile [21, 22], with a subset of CpG sites more likely to be differentially methylated between unrelated individuals. We proposed that these CpG sites may be especially susceptible to environmentally-induced changes associated with placental disease [21]. In a follow-up investigation, we also observed a gestational age difference in DNA methylation profile in the placenta across the third trimester [23], while others have recently reported an increase in global DNA methylation levels between pre-term (28 weeks) and full term placenta (40 weeks) [24].

The aim of the current study was to build on recent knowledge obtained through genome-scale gene expression [5, 6, 25] and DNA methylation analysis [24, 26] of the human placenta. In the current study, we used the Illumina Infinium HumanMethylation27 BeadChip platform to identify promoter regions subject to change throughout gestation. In addition we wanted to identify those that become increasingly variable between individual placentas over time, possibly in response to accumulated environmental exposures.

In this study we performed genome-wide DNA methylation analysis of gene promoter regions, using the Illumina Infinium HumanMethylation27 BeadChip, in placentas of different gestational ages (18 first trimester (8-12 weeks), 10 second trimester (17-23 weeks), and 14 third trimester samples (34-41 weeks)). We further validated the array data by targeting 12 CpG sites (covering 12 gene promoters) using the Sequenom EpiTYPER platform. The correlation of r² = 0.76 between the two platforms is comparable to that published for similar comparisons [30]. We found evidence for both a programmed change of methylation in gene promoters across gestation and an increase in variability of methylation as gestation progresses. We predict that this is directly related to the cumulative effect of intrauterine environmental exposure, however, the contribution of stochastic events to the observed variation cannot be discounted at this time. Further investigation is warranted to determine the effects of specific environmental exposures on DNA methylation patterns of the genes identified as variable in the present study.

Unsupervised clustering based on all Infinium probes (excluding those on the sex chromosomes) clearly separated all samples by gestational age, with first trimester samples clustering away from second and third trimester samples (Figure 1). This indicates that there are consistent, large scale changes in DNA methylation across gestation. There are several possible explanations for these temporal differences, one of which is the change in cell composition and differentiation from first to third trimester, especially the relative loss of cytotrophoblasts throughout gestatation, (85% of the total trophoblast population in first trimester but only 15% of the trophoblast volume at full term) [31]. In support of this playing at least a partial role in the observed change in methylation over time, we found a slightly higher correlation between the methylation profile of purified primary cytotrophoblasts with first trimester placental tissue (R² = 0.96) than with third trimester tissue (R² = 0.88; see Additional file 8). Similar data have recently been reported for second trimester placenta tissue which has been shown to be more similar to cytotrophoblasts than mesenchyme in terms of methylation profile [32]. However, the greater part of trophoblast volume at full term is syncytiotrophoblast, arising from the fusion of cytotrophoblast cells into a multinucleated layer. Thus, the observed trend of increasing methylation may equally be due to differentiation of the cytotrophoblast component, or alternatively, may even be due to other aspects of altered placental function known to occur as gestation progresses.

Despite the interspersed clustering of first trimester samples based on the overall Infinium methylation patterns, we were interested in assessing DNA methylation changes occurring at the transition from the first to the second trimester. This period of placental development is characterised by the loosening of trophoblast plugs and the associated rise in oxygen concentration from 2-3%, to 7-8% [33]. The rapid increase in oxygen levels can lead to the accumulation of reactive oxygen species (ROS), which might be linked to the reduction in trophoblast invasion and migration observed at about this time. Furthermore, it has been suggested that the inability of the placenta to adapt to this increase in oxygen concentration may lead to the development of preeclampsia [3]. In this study, we found only 12 CpG sites with Δβ > 0.2 between these time points, suggesting that DNA methylation changes are unlikely to play a major role in the physiological changes in placentation associated with transition from low oxygen to a normal oxygen environment. It must be noted however, that we do not have data on the level of vascularisation for our placental samples, so it is possible that flooding of the maternal blood may not have occurred at the time the 12 week tissue was collected.

The separation of the trophectoderm from the inner cell mass at the blastocyst stage occurs at a time when genome-wide DNA methylation levels are at their lowest [34]. The subsequent re-establishment of methylation marks occurs at a slower rate, and to a lesser extent, in the extra-embryonic lineage compared to somatic tissues [34]. This accounts for the low global DNA methylation in human placenta, which is more similar to human tumours, and is mostly due to hypomethylation of repetitive elements [35, 36]. A recent study has reported a positive correlation between global DNA methylation levels (as measured by an ELISA-like assay with 5-methylcytosine antibody) and gestational age in the placenta [24], supporting our data for an increasing level of promoter-associated DNA methylation during gestation, particularly from second to third trimester (Figure 2). This accumulation of methylation was most apparent in genomic regions that showed an intermediate level of methylation in first trimester, suggesting that these genes are the most likely to be epigenetically regulated by DNA methylation. Furthermore, these CpG sites were more likely to be in lower CpG density regions (non-CGI) (see Additional file 12) suggesting that methylation levels of CpG sites within CpG Islands are more stable across gestation in this tissue. The higher promoter methylation at term could also reflect the end point of a continual process of re-methylation in the extra embryonic lineage from the blastocyst stage, at which point the genome is almost completely hypomethylated [34].

IPA analysis of genes showing differential methylation between first and third trimesters indicated that the most affected pathways were 'communication between innate and adaptive immune cells' and other 'immune-related' pathways (Table 2). Genes in common between two or more of the top 5 immune-related pathways included immune regulators CCR7, CD28, CSF2, IFNA17, IFNA2, IFNA21, IFNB1, IL1F7, IL2, IL3, IL5, LTA, TLR6, TLR9, TNF, TNFRSF13B and TNFSF13B. This is in accordance with previous gene expression studies, which also showed substantial enrichment of immune regulators amongst the most differentially expressed genes in placenta at different gestational ages [5, 6]. Several immune-regulators showed a strong correlation between methylation and expression change from first to third trimester, including Granulysin (GNLY), which has previously been implicated in spontaneous abortions [37] (Figure 6). Given the critical role of the placenta in modulating the maternal immune response to the developing pregnancy, and mounting evidence for a role of environmental exposures in controlling both immune-system development and epigenetic profile, it is not surprising that immune regulators are amongst the most variably methylated gene groups.

In addition to identifying CpG sites that consistently change over gestation, we were also interested in CpG sites that show inter-individual variability within each gestational age. Genes that show inter-individual differences in expression in placenta have previously been described by Sood et al. [25]. We have previously reported a subset of CpG sites showing variable methylation in term placenta between unrelated individuals, and suggested that these CpG sites may be more susceptible to change under adverse conditions [21]. We have now extended these findings by demonstrating that although ~95% of all probes showed very little variation (s² < 0.01) between individuals, the number of highly variable probes (s² > 0.02) increased with gestational age. Our data support a model whereby increasing variability in methylation profile between placentas arises in response to cumulative differences in environmental exposure rather than sequence polymorphisms (Figure 5). The most variable probes in the first trimester were associated with genes involved in, 'calcium signaling' and 'nitrogen metabolism', while 'circadian rhythm signalling' and 'arachidonic acid metabolism' were enriched in both first and second trimester placenta (Table 3). The most enriched variable pathways at term included 'glutamate receptor signaling', 'valine, leucine and isoleucine degradation', and 'β-alanine', 'butanoate', and 'tyrosine' metabolism. The enrichment of pathways involved in metabolism further supports our hypothesis that the methylation variation is occurring in response to environmental influences, potentially as part of an 'adaptive' response of the developing pregnancy, thereby allowing changes in gene expression that may be more beneficial under altered environmental conditions [38]. The placental epigenome regulates placental gene expression and function, and any disruption in placental epigenetics (via either stochastic or environmental perturbation) has the potential to affect the developing fetus. Therefore it will be interesting in future to examine the potential for role of placental epigenome variation with the now widely accepted phenomenon of 'fetal programming' first described in the context of long term effects of intrauterine environment on offspring [39] and subsequently developed by Barker and colleagues in the early 1990s [40–42].

Finally, contrary to CpG sites that increase in methylation across gestation, variable CpG sites are more likely to be located within CpG islands (Additional file 12). This coupled with the finding that hypermethylated probes (β > 0.6) do not show higher variation compared to probes with low methylation levels (β < 0.2) (see Additional file 10) suggests that increased variation in late gestation is not a by-product of increasing methylation across gestation.

The direct role of DNA methylation in controlling placental global gene expression levels was examined by comparing our DNA methylation data with published gene expression data for placenta tissue of the same gestational age (Figure 6 and Additional files 13 and 14). A general trend of decreasing expression in response to increasing DNA methylation was found as expected. Further, the range of expression also decreased in the highly methylated probe group. These findings confirm that promoter DNA methylation influences global gene expression levels at all three gestational time points. However, the large range in gene expression in all four methylation quartiles, and the lack of strong correlation between methylation and expression change across gestation (Figure 6), support previous data that promoter DNA methylation is only one of the mechanisms controlling gene expression, with many other factors combining to regulate the expression of genes, including histone modifications, non-coding RNAs, and transcription factors.

In the present study, we performed genome-scale DNA methylation analysis of human placenta at three distinct gestational ages, using the Infinium HumanMethylation27 array. This platform targets over 14, 000 genes, however, is limited to the 5' promoter region. Our findings support the hypothesis that DNA methylation levels in the human placenta are dynamic and change over gestation, possibly in response to changing cellular composition and/or cumulative environmental influences. The identification of pathways that are likely to be affected by the latter (i.e. those that show higher variation at full term) will provide valuable candidates for testing in studies examining placenta-associated adverse pregnancy outcomes.