The effect of environmental exposures on epigenetic regulation in human populations remains poorly understood. The consequences of maternal exposure to EACs, including BPA, on the epigenetic landscape of offspring has been studied using animal models, and this includes the discovery of altered DNA methylation at candidate metastable epialleles such as CDK5 activator-binding protein (Cabp) and viable yellow Agouti (A
) alleles [17, 19]. The effects of postnatal BPA exposure on the human epigenome, however, remain largely unexplored. To date, the majority of environmental epidemiology studies that incorporate epigenetics have been conducted in the context of a cohort study, where both environmental exposures and epigenetic outcomes are quantified in healthy populations [47–49]. The majority of these studies have looked at single genes or global methylation profiles. Increasingly, epigenome-wide approaches are useful in identifying a broader constellation of epigenetic changes associated with environmental exposures . Here we measured DNA methylation levels genome-wide in 46 pre-adolescent Egyptian girls with varying concentrations of BPA exposure. The study population was homogeneous with respect to age and genetic population stratification was limited, thereby increasing our ability to identify exposure based methylation profiles.
While the overall levels of urinary BPA concentrations of Egyptian girls were low and less variable compared to age-matched American girls , we identified several CpG sites showing differential methylation upon exposure and biological pathways of interest using a genome-wide approach. This provides support for the hypothesis that even low concentrations of BPA exposure may be associated with health effects via epigenetic modification, and studies of relevant ranges of BPA exposure in human populations are warranted. These results should be interpreted cautiously as the use of spot urine for measuring BPA exposure is a limitation of our study. For example, interpersonal variation in amount of water consumed may affect urine output. Thus, we adjusted urine BPA concentrations using specific gravity. Similarly, to reduce variability arising from exposures throughout the day, the participants provided spot urine sample during a specified time period between noon to 4 pm. Additionally, previous studies on various populations have shown that individuals are exposed to mixtures of chemicals, potentially with similar mechanisms of action as BPA, such as phthalates, pesticides, and heavy metals. Future studies using more comprehensive methods of exposure assessment, including the measurement of chemical mixtures, are warranted .
Of interest, many previously identified pathways associated with BPA exposure were also identified in our analysis as top differentially methylated pathways. LRpath enrichment analysis identified BPA-induced hypomethylation among genes involved in metabolism and steroid hormones, including progesterone, testosterone, and estradiol. BPA is one of several widely known EACs or xenoestrogens [52, 53], and steroid hormones are known to influence metabolism and inflammation including various immune responses. As such, several immune-response concepts were enriched with hypomethylated CpG sites in girls in the BPA-high group. In previous studies, higher cytokine activity was observed in BPA-exposed rodents . In addition, the autoimmune thyroid disease pathway was significantly enriched with hypomethylated genes in girls categorized in the BPA-high group, and BPA is a known thyroid receptor antagonist [55, 56]. Genes involved in the N-glycan biosynthesis pathway were also found to be differentially methylated based on BPA exposure, supporting previous work linking BPA concentrations with insulin resistance and type II diabetes , For example, N-glycans are key players mediating cell-to-cell communication, interacting with glycan-binding proteins , and can be attached to human insulin receptor (IR) and interfere with the IR interaction with antibodies or binding proteins by masking the surface of IR dimers .
Prenatal BPA exposure has been associated with altered DNA methylation and expression of imprinted genes in mouse embryos [60, 61]. A methylation analysis of oocytes from CD-1 mice perinatally exposed to 20 or 40 ug BPA/kg diet identified decreased methylation among imprinted genes, including insulin like growth factor 2 receptor (IGF2R) and paternally expressed 3 (PEG3) . When pregnant JF1 mice were exposed to 200 ug BPA/kg diet, embryos showed a 10% difference in Slc22a18 expression compared with controls . When we examined DNA methylation of 34 imprinted genes (from 233 sites) in these subjects, BPA-associated changes were not observed in the imprinted genes, with the exception of a single CpG site in the NDN promoter (Additional file 2: Figure S3B). The imprinted gene, NDN, is located within the Prader-Willi/Angelman syndrome (PWS) chromosomal region . Loss of function of this gene has been associated with post-natal lethality ; thus NDN has a critical role during development. Thus, methylation profiles at imprinted genes, while labile to nutritional and environmental influences early in development [64–66], may remain stable and relatively resistant to environmental factors over the life course.
DNA methylation of CpG islands acting in concert with other epigenetic mechanisms, are known to be important in female mammalian X chromosome inactivation and maintenance . As with genomic imprinting, chromosome X inactivation may thus be potentially vulnerable to the effects of environmentally-induced epigenetic reprogramming . In our data, we observed an enrichment of hypomethylated sites on chromosome X with increasing BPA concentrations (Figure 4 and Additional file 1: Table S5). One of the significantly hypomethylated cytobands in chromosome X was Xq13 (p-value < 0.02), where the X Inactivation Center (XIC) is located. The XIC is known to play a role in Xist expression, a major regulator of X inactivation. Two of the four sites validated by quantitative bisulfite pyrosequencing were CXorf23 (Xp22.12) and BEX2 (Xq22.1) from chromosome X. Although we were unable to achieve robust statistical significance as seen in the BeadChip data, we did observe a slight decrease in DNA methylation in BPA-high samples (Figure 6). As the girls observed in our study had low urinary BPA concentrations compared to age-matched American girls , it will be important to further evaluate the chromosome X findings in other exposed populations.
A functional validation of DNA methylation targets using gene expression was not performed in this cohort due to the unavailability of field-collected RNA. To overcome this limitation, we extracted gene expression information from multiple previously conducted studies available in CTD. The increasing wealth of data in the public arena now allows for the comparison of our genome-wide signatures with those from other studies, efficiently providing additional insight into key biologically-relevant pathways. When the curated genes known to interact or harbor differential expression upon BPA exposure in the CTD were compared with the list of differentially methylated genes in our study, more than 20% of BPA-associated genes were found to overlap. The biologically-prioritized candidate gene, BEX2, is associated with estrogen, cell cycle, and apoptosis pathways  and over-expressed in Ishikawa cells, a well-differentiated human endometrial adenocarcinoma cells, treated with BPA . In addition, the biologically-prioritized candidate, HOXA10, has previously been associated with decreased DNA methylation among in utero BPA-exposed mouse samples . Altered methylation at HOXA10 was confirmed by bisulfite pyrosequencing in our Egyptian girls, despite relatively low BPA concentrations with small variation compared to US populations . Although the validation did not result in the same depth of significance as seen in the array data, it showed a clear trend toward differential methylation (Additional file 2: Figure S4). Importantly, some biologically-prioritized candidate genes are known to target the major disease pathways, such as DNA repair, cell cycle, and development. Thus, future epigenetic epidemiology studies with increased sample size and increased exposure ranges are needed to further elucidate the consequences of genome-wide altered methylation upon BPA exposure, including BRCA1, BEX2, and HOXA10.