Abstract
There is compelling evidence that epigenetic modifications link developmental environmental insults to adult disease susceptibility. Animal studies have associated perinatal bisphenol A (BPA) exposure to altered DNA methylation, but these studies are often limited to candidate gene and global non–loci-specific approaches. By using an epigenome-wide discovery platform, we elucidated epigenetic alterations in liver tissue from adult mice offspring (10 months) following perinatal BPA exposure at human physiologically relevant doses (50-ng, 50-μg, and 50-mg BPA/kg diet). Biological pathway analysis identified an enrichment of significant differentially methylated regions in metabolic pathways among females. Furthermore, through the use of top enriched biological pathways, 4 candidate genes were chosen to assess DNA methylation as a mediating factor linking the association of perinatal BPA exposure to metabolic phenotypes previously observed in female offspring. DNA methylation status at Janus kinase-2 (Jak-2), retinoid X receptor (Rxr), regulatory factor x-associated protein (Rfxap), and transmembrane protein 238 (Tmem238) was used within a mediational regression analysis. DNA methylation in all four of the candidate genes was identified as a mediator in the mechanistic pathway of developmental BPA exposure and female-specific energy expenditure, body weight, and body fat phenotypes. Data generated from this study are crucial for deciphering the mechanistic role of epigenetics in the pathogenesis of chronic disease and the development of epigenetic-based prevention and therapeutic strategies for complex human disease.
The developmental origins of health and disease (DOHaD) hypothesis postulates that early nutritional and environmental exposures shape health outcomes throughout the lifecourse (1). Bisphenol A (BPA), a chemical found in consumer products, including beverage and food containers and thermal papers, is an environmental exposure that has been associated with disease development in adulthood after early-life exposure (2). BPA is an endocrine active compound that interferes with estrogen (3, 4), androgen (5), and thyroid function (6), as well as transcription factors, including peroxisome x receptor and aryl hydrocarbon receptor (5, 7). Additionally, BPA has the ability to impair the development of the central nervous system, resulting in behavioral disorders (8, 9). We previously have shown altered lifecourse metabolic phenotypes, including increased activity and energy expenditure, lower body fat, and improved hormone profiles, in adult female mice following maternal dietary exposure to 3 physiologically relevant levels (measured in mouse liver and compared with human liver levels) of BPA (50-mg, 50-μg, or 50-ng BPA/kg diet) (10). Thus, BPA’s influence on normal differentiation and maturation processes during early development, with health risk manifesting later in life, is of current concern.
In studies elucidating mechanisms involved in DOHaD, epigenetic pathways, including DNA methylation, histone modifications, and chromatin modeling, are primary mechanisms of interest (11). Of these epigenetic modifications, DNA methylation is the most widely studied and best characterized, often because of its stability in stored samples. DNA methylation is generally thought to be a stable mark established early in development, but increasing evidence reveals that environmental and nutritional factors can drive sustained alterations to DNA methylation when encountered early in development. Using the viable yellow agouti (Avy) mouse model, we have shown that maternal dietary exposure to 3 physiologically relevant levels of BPA (50-mg, 50-μg, or 50-ng BPA/kg diet) results in modifications to the Avy and CabpIAP metastable epialleles and the Igf2 and H19 imprinted loci (12, 13). Perinatal exposure to 50-mg BPA/kg diet resulted in decreased DNA methylation at the Avy locus (12, 14), whereas perinatal exposure to the lower doses (50-ng and 50-μg BPA/kg diet) resulted in increased methylation at the CabpIAP locus (12).
Developmental exposures to BPA have been associated with alterations in epigenetic modifications and serve as a potential mechanism for risk for chronic disease, such as cancer, type 2 diabetes, and obesity, as well as impaired brain development and behavior (15, 16). For example, exposure to 10 µg BPA/kg body weight per day in male rats during early development resulted in hypomethylation at the phosphodiesterase type 4 variant 4 gene in prostate cancer cells during adulthood (17). In utero exposure to 5 mg/kg body weight resulted in decreased methylation in Hoxa10, a gene involved in several cancers, including endometrial carcinoma, in adult female mice (18, 19). Two additional genes implicated in prostate carcinogenesis, nucleosome binding protein-1 and hippocalcin-like 1, displayed hypo- and hypermethylation, respectively, in adult rats exposed to 10 µg BPA/kg body weight during the neonatal period (20).
Many studies that investigated the effects on the epigenome following environmental and nutritional exposures have been limited by being candidate gene–driven approaches or based on epigenetic techniques with limited genome coverage and sensitivity. Given that the main goal of DOHaD research is to improve understanding of the role that early environmental insults have on health and disease, full genome deep sequencing and tiling array technologies are valuable for elucidating key regulatory pathways involved in the etiology of disease via epigenetic mechanisms. Furthermore, following animals across the lifecourse after they were exposed to BPA during the perinatal period helps identify a role for early-life BPA exposure and later-in-life disease risk. Thus, by using an epigenome-wide platform as a discovery tool in combination with lifecourse in vivo phenotyping, we have examined regions of altered methylation (RAMs) as potential mediators on the development of metabolic phenotypes in adulthood following perinatal BPA exposure.
Materials and Methods
Mouse liver samples and DNA isolation
The liver tissues used for this experiment were obtained from BPA-exposed and control nonagouti a/a wild-type offspring that were aged out to 10 months and underwent the metabolic phenotyping, as described thoroughly elsewhere (10, 21). Briefly, dams were exposed to 1 of 4 diets (50-ng BPA, 50-ug BPA, or 50-mg BPA/kg diet or control) 2 weeks before mating and remained on their respective diet through lactation. At 3 weeks of age, all offspring were weaned onto the control diet and followed until euthanasia at 10 months of age. Offspring were measured for metabolic phenotypes (activity levels, indirect calorimetry, body weight, lean body mass, fat mass) and hormone levels (leptin, adiponectin, insulin) at 3, 6, and 9 months of age and 10 months of age, respectively.
Here, 10-month liver DNA from a subset of a/a offspring was analyzed for RAMs in murine promoter regions after (a) standard control diet (n = 6 offspring: 3 male and 3 female); (b) 50-ng BPA/kg diet (n = 6 offspring: 3 male and 3 female); (c) 50-µg BPA/kg diet (n = 5 offspring: 2 male and 3 female); (d) 50-mg BPA/kg diet (n = 6 offspring: 3 male and 3 female).
Total genomic DNA was isolated from 10-month liver tissue (n = 23) by using a standard phenol-chloroform extraction method. Briefly, about 15 mg of tissue was resuspended in 540 µL buffer Animal Tissue Lysis (ATL) and homogenized for 20 seconds at 15 Hz (TissueLyser, Qiagen, Hilden, Germany). The lysate was transferred to 60 µL of proteinase K and incubated overnight at 50°C. After overnight incubation, 12 µL of RNase A was added to lysate and incubated for 10 minutes at 37°C. Samples were extracted twice with 600 µL phenol–chloroform–isoamyl and once with 600 µL chloroform. Fifty microliters of 3M sodium acetate was added to aqueous phase and precipitated with 1 mL ice-cold 100% ethanol. For additional precipitation, 1 mL of ice-cold 75% ethanol was added to pellet twice. The pellet was air dried and resuspended with 100 µL Tris-EDTA buffer and incubated for 2 hours at 60°C with frequent mixing.
DNA fragmentation by sonication
Genomic DNA was sheared to fragment sizes between 200 and 1000 base pairs. Briefly, 17 µg of DNA was divided into equal volumes into 2 separate polymerase chain reaction (PCR) tubes (8.5 ug DNA each). DNA was sonicated with a total process time of 15 minutes with 15-seconds-on and 30-seconds-off cycles at an amplitude of 18 (Episonic 1100 series, Farmingdale, NY) in 8°C to 20°C water (monitored and cooled every 5 minutes of process time). Fragment sizes were confirmed by gel electrophoresis with ∼1 µg DNA.
Enrichment of methylated DNA
Methylated regions of fragmented DNA were enriched by methyl CpG-binding domain (MBD)–based capture by using the Methylated DNA Enrichment Kit (Epimark, New England BioLabs, Ipswich, MA). Methylated DNA binds to the MBD of the human MBD2 protein, which is coupled to paramagnetic protein A beads. First, MBD2-Fc and beads were combined, incubated by rotation for 15 minutes at room temperature, and washed twice. The fragmented DNA was then added to the MBD2-Fc/bead mixture, incubated by rotation for 20 minutes at room temperature, and washed three times to discard unbound DNA. The captured methylated CpG DNA was eluted with 100 µL of DNase-free water and 15 minutes of incubation at 65°C.
Whole genome amplification, and enrichment quality assessment via quantitative PCR
DNA enriched for CpG methylation underwent whole genome amplification according to the manufacturer’s instructions by using GenomePlex Complete Whole Genome Amplification Kit (Sigma-Aldrich, St. Louis, MO). Validation and quality assurance of the enriched fraction were performed by using a positive methylated control (Xist) and a negative unmethylated control (H19) loci via quantitative PCR. Briefly, 60 ng of the genomic sonicated DNA, noncaptured fraction from enrichment, and whole genome amplification–enriched fraction were prepared as 1:10 serial dilutions. The quantitative PCR components were prepared by using SYBR Green master mix (Qiagen Inc., Valencia, CA), forward primer (0.25 pmol), and reverse primer (0.25 pmol). This assay was performed by using the following parameters: initial denaturation at 95°C for 3 minutes, denaturation at 95°C for 10 seconds, annealing at 62°C for 30 seconds, and elongation at 72°C for 10 seconds, repeated for 39 cycles starting at the second denaturation step. All samples were run in duplicate. The cycles to threshold were averaged across duplicates and used to quantitate enrichment of the immunoprecipitation (IP) fraction compared with the input and noncaptured fraction by taking the difference of the cycles to threshold.
Hybridization and array scanning
Experimental (IP) and control (input) samples were labeled with Cy5 and Cy3, respectively, by using the NimbleGen Dual-Color DNA Labeling Kit (Roche, Indianapolis, IN) and following the NimbleGen Arrays User Guide, DNA Methylation Arrays (version 7.2. IP and input samples were pooled and cohybridized to Roche NimbleGen Mouse DNA Methylation 3x720K CpG Island Plus RefSeq Promoter Arrays, an array platform focused on 5′ promoter regions, for 16 to 20 hours. After the hybridization period was complete, arrays were washed and scanned using 2-µm, high-resolution NimbleGen MS 200 Microarray Scanner (Dr. Thomas Glover, Department of Human Genetics, University of Michigan).
Bioinformatics pipeline
All arrays were uploaded to NimbleGen DEVA software (version 2.3). An alignment grid was laid over each subarray image to extract the location of each feature as well as to generate the raw signal intensities from the IP and input samples. The appropriate normalization procedure was determined by visualizing the distribution of each subarray. The normalization process consisted of two steps: (a) The log2 ratio of the input and IP channels were calculated, and (b) the log2 ratios were corrected by subtracting the Tukey biweight estimate of the median. The Tukey biweight estimate is determined by a weight given to each data value by using a bisquare weight function. The weights assigned are greater for data values near the median of the data cluster and decrease for data away from the median, so outliers have a minimal effect on the estimate (22). The following statistical analyses were performed with R statistical software (limma R package, version 2.10.1; R Project for Statistical Computing). An empirical Bayesian t test (23) was calculated to assess the differences of the normalized log2 ratios by using dichotomous comparisons of BPA exposure groups (50-ng, 50-µg, or 50-mg BPA/kg diet) vs. the control group. Analyses were done by combining sexes as well as by stratifying by sex, and P values were adjusted by using the Benjamini-Hochberg false discovery rate approach. Statistically significant differentially methylated genes were chosen on the basis of adjusted P values <0.001.
Biological pathway analysis
RAMs with adjusted P values <0.001 were read into the Database for Annotation, Visualization and Integrated Discovery (DAVID, version 6.7; National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD) for identification of key biological pathways involved in the most differentially methylated genes. Candidate genes were chosen from enriched biological pathways that may be involved metabolic or hyperactive phenotypes. The pathway analysis was limited to females because previously observed metabolic and hormone phenotypes were exhibited strictly in female exposed offspring (10, 24).
Site-specific DNA methylation technical and biological validation
To support the RAM findings, technical and biological DNA methylation validation was performed by using the Sequenom EpiTyper, a matrix assisted laser desorption/ionization time of flight mass spectrometry–based platform. Technical and biological DNA methylation was quantified at 5, 60, 29, and 25 CpG sites in the promoter regions of Janus kinase-2 (Jak-2), retinoid X receptor (Rxr), regulatory factor x-associated protein (Rfxap), and transmembrane protein 238 (Tmem238), respectively, on the subset of 10-month liver tissue from the female a/a mice described previously (technical validation) plus additional liver tissue from the same exposure experiment (biological validation). This resulted in a total sample size as follows: (a) standard control diet (n = 17 offspring: 8 male and 9 female); (b) 50-ng BPA/kg diet (n = 16 offspring: 9 male and 7 female); (c) 50-µg BPA/kg diet (n = 16 offspring: 8 male and 8 female); and (d) 50-mg BPA/kg diet (n = 10 offspring: 5 male and 5 female). DNA was extracted from liver tissue and bisulfite converted as described previously. Bisulfite PCR amplification was performed using by FastStart Taq Polymerase (Roche Diagnostics) with a forward and a reverse primer concentration of 0.1 µM and 40 PCR cycles for the following: (a) Jak-2 (annealing temperature 60°C, primer sequences, including the forward and T7 promoter tags required for Sequenom analysis (Sequenom, San Diego, CA): 5′-AGGAAGAGAGTGAGAATAGTATATGTAGAATGGGTGTTG (forward primer) and 5′-CAGTAATACGACTCACTATAGGGAGAAGGCTAACCCAAATACCTAACTTCAATTAACC [reverse primer]); (b) Rxr (annealing temperature 60°C, primer sequences AGGAAGAGAGGTTTTTATTTTAATGTTAGGGAGGTT [forward primer] and 5′-CAGTAATACGACTCACTATAGGGAGAAGGCTAACATAAACACCAAACATTTCCTAC [reverse primer]); (c) Rfxap (annealing temperature 51°C, primer sequences 5′-AGGAAGAGAGGATTTAGATTTGGATTTGGATTTTG [forward primer] and 5′-CAGTAATACGACTCACTATAGGGAGAAGGCTCTAACAAACAATTAAAAAACAATAAATACC [reverse primer]); (d) Tmem238 (annealing temperature 57°C, primer sequences, 5′-AGGAAGAGAG GGTTTTTTGTTTGGGTTTTTTTT [forward primer] and 5′-CAGTAATACGACTCACTATAGGGAGAAGGCTAAACCACAATACCCCTCACAC [reverse primer]). Methylation analyses were performed at the University of Michigan Sequencing Core Facility following manufacturer’s recommended protocol.
Mediational regression analysis
Mediational regression analysis was conducted in a 4-part model process to determine whether DNA methylation is a mediator of BPA exposure and metabolic phenotypes in female offspring. First, a mixed-effects regression analysis was performed to identify statistically significant associations among BPA exposure and metabolic phenotypes (model 1) (spontaneous activity, energy expenditure, body composition, and hormones in 9- to 10-month-old offspring) in the female BPA-exposed offspring (50-ng, 50-µg, or 50-mg BPA/kg diet) vs. the control female offspring (10). Second, analyses from the empirical Bayesian t test, described previously, were used to determine the significance of the associations between BPA exposure and DNA methylation (model 2) in each candidate gene in the BPA-exposed groups compared with the control group. Third, a mixed-effects regression was performed to determine the associations of DNA methylation for each candidate gene and the metabolic phenotypes (model 3). Finally, if all three previous analyses resulted in a significant relationship, a final mixed-effects regression model measuring BPA exposure and its association on the metabolic phenotypes adjusting for DNA methylation was performed (model 4). Methylation was considered a mediator if (a) models 1 through 3 showed significant associations and (b) the difference estimate resulted in a ≥10% attenuation in model 4 (adjusting for methylation) compared with model 1 (without adjusting for methylation). Within-litter siblings were controlled for in the regression models. Because of the limited sample size, results were considered marginally significant at P < 0.10. Statistical analyses for models 1, 3, and 4 were completed by using SAS software, version 9.2 (SAS Institute, Cary, NC), and analyses for model 2 were completed by using R statistical software, version 2.10.1.
Results
Identification of differentially methylated regions across BPA exposed offspring
Among the 713,168 probes, 262,661 unique probes exhibited differential methylation by BPA exposure in at least 1 of our 3 comparisons (nanogram vs. control, microgram vs. control, and milligram vs. control) (Fig. 1). Among nanogram-exposed offspring (males and females; n = 6), 124,470 loci were identified as hypermethylated and 108,591 loci were hypomethylated compared with controls (Fig. 2). Among microgram-exposed offspring (males and females; n = 5), 107,303 loci were identified as hypermethylated and 89,940 loci were hypomethylated compared with controls (Fig. 2). Among microgram-exposed offspring (males and females; n = 6), 155,628 loci were identified as hypermethylated and 74,242 loci were hypomethylated compared with controls (Fig. 2). Rfxap and Tmem238 were chosen for further analysis on the basis of the lowest overall P value of the BPA-exposed females. Rfxap plays a role in the immune system, and Tmem238 expresses a transmembrane protein important in amino acid transport.
Figure 1.
Differentially methylated regions. Venn diagram illustrating the percentage of differentially methylated regions among each exposure group compared with controls following genome-wide DNA methylation analysis on 10-month liver DNA in males and females using a P value cutoff of <0.001 (control group, n = 6; ng BPA exposure, n = 6; µg BPA exposure, n = 5; mg BPA exposure, n = 6).
Figure 2.
Heat map. A supervised cluster illustrating the top 25% most differentially hypermethylated (green) and hypomethylated (red) regions among the ng, μg, and mg BPA-exposed offspring.
Biological pathways involved in differentially methylated regions
Of the significantly differentially methylated genes, key biological pathways were also drawn from these top candidate regions via DAVID for the female BPA-exposed offspring only. Two candidate genes, Jak-2 and Rxr, were chosen from enriched biological pathways because of their involvement in metabolic or hyperactive phenotypes. The pathways enriched include insulin signaling, adipocytokine signaling, Jak-STAT signaling, and tyrosine metabolism pathways (Table 1).
Table 1.
Biological Pathway Analysis in BPA-Exposed Female Offspring
| Pathway | Concept Type | No. of Genes | P |
|---|---|---|---|
| Pathways in cancer | KEGG | 305 | 1.30E-05 |
| MAPK signaling | KEGG | 250 | 1.70E-03 |
| Wnt signaling | KEGG | 144 | 8.80E-06 |
| Jak-STAT signaling | KEGG | 142 | 1.90E-03 |
| Melanoma | KEGG | 94 | 9.30E-03 |
| Adipocytokine signaling | KEGG | 62 | 9.20E-02 |
| Insulin signaling | KEGG | 124 | 8.40E-02 |
| Purine metabolism | KEGG | 142 | 4.00E-02 |
| Cell cycle | KEGG | 116 | 5.90E-02 |
| Long-term depression | KEGG | 67 | 6.00E-02 |
| Neuroactive ligand receptor | KEGG | 234 | 1.20E-02 |
| Tyrosine metabolism | KEGG | 37 | 4.20E-02 |
| Basal cell carcinoma | KEGG | 54 | 6.20E-03 |
| Gap junction | KEGG | 81 | 1.60E-02 |
Abbreviations: KEGG, Kyoto Encyclopedia of Genomes and Genes; MAPK, mitogen activated protein kinase.
Biological validation
To validate the differential methylation of our 4 candidate genes, we quantitatively measured methylation of CpG sites in female-exposed offspring. Jak and Rxr had increased methylation in the 50-mg group and decreased methylation in the microgram-exposed and nanogram-exposed groups when compared with the control group, validating differential methylation array data (Table 2). Differential Tmem methylation was validated in the nanogram-exposed group but not in the milligram- and microgram-exposed groups (Table 2). Rfxap methylation was validated in the milligram- and nanogram-exposed groups but not in the microgram-exposed group (Table 2).
Table 2.
Validation of Methylation in 4 Candidate Genes (Female Offspring Only)
| Exposure Group | NimbleGen Direction (Methylation Status Compared With Control Group) | Validation Direction (Methylation Status Compared With Control Group) |
|---|---|---|
| Jak | ||
| mg | Higher | Higher |
| μg | Lower | Lower |
| ng | Lower | Lower |
| Rxr | ||
| mg | Higher | Higher |
| μg | Lower | Lower |
| ng | Lower | Lower |
| Tmem | ||
| mg | Higher | Lower |
| μg | Higher | Lower |
| ng | Lower | Lower |
| Rfxap | ||
| mg | Lower | Lower |
| μg | Higher | Lower |
| ng | Lower | Lower |
DNA methylation mediates BPA exposure and metabolic phenotypes
Mediational regression analysis of site-specific methylation of Jak-2, Rfxap, Tmem238, and Rxr revealed that the metabolic phenotypes of oxygen consumption, carbon dioxide production, body weight, body fat, and circulating leptin levels in BPA-exposed females were mediated by differential methylation.
Specifically, after oxygen consumption was modeled as the outcome, the differences in oxygen consumption of the nanogram-exposed females compared with the control group were attenuated > 10% when adjusting for methylation at Jak-2 and Rfxap (Table 3). The differences in oxygen consumption as well as carbon dioxide production of the 50-mg–exposed females compared with the control females were also attenuated >10% after adjustment for methylation at Rxr (Tables 3 and 4). Furthermore, after modeling of carbon dioxide production as the outcome, the differences in carbon dioxide production of the nanogram-exposed females compared with the control females were attenuated >10% after adjustment for methylation at Rfxap (Table 3).
Table 3.
Mediational Regression Analysis in BPA-Exposed Female Offspring: Oxygen Consumption
| Mediator and Exposure Group | Mean Oxygen Consumption (mL/kg/h) (SEM) | Overall P | Mean Consumption Difference (SEM) vs. Control | P |
|---|---|---|---|---|
| None (model 1) | 0.02 | |||
| mg | 3901 (160) | 152 (219) | 0.48 | |
| µg | 3473 (151) | −275 (212) | 0.20 | |
| ng | 4525 (150) | 774 (212) | 0.003 | |
| Control | 3749 (151) | |||
| Jak-2 CpG site 5/6 (model 4) | 0.09 | |||
| mg | 3669 (345) | 109 (434) | 0.80 | |
| µg | 3482 (246) | −77 (363) | 0.83 | |
| ng | 4409 (297) | 849 (399)a | 0.03 | |
| control | 3559 (266) | |||
| Rfxap CpG site 1 (model 4) | 0.01 | |||
| ng | 4252 (151) | 535 (214)a | 0.01 | |
| Control | 3717 (147) | |||
| Rfxap CpG site 16 (model 4) | 0.04 | |||
| ng | 4212 (152) | 455 (217) | 0.04 | |
| Control | 3757 (150) | |||
| Rfxap CpG site 24 (model 4) | 0.03 | |||
| ng | 4236 (153) | 489 (220) | 0.03 | |
| Control | 3747 (150) | |||
| Rfxap CpG site 30 (model 4) | 0.01 | |||
| ng | 4252 (151) | 535 (214) | 0.01 | |
| Control | 3717 (147) | |||
| Rfxap CpG site 31 (model 4) | 0.04 | |||
| ng | 4212 (152) | 455 (217) | 0.04 | |
| Control | 3757 (150) | |||
| Tmem238 CpG site 9 (model 4) | 0.4 | |||
| µg | 3312 (233) | −295 (344) | 0.4 | |
| Control | 3608 (254) | |||
| Rxr CpG site 18/19/20 (model 4) | 0.74 | |||
| mg | 3750 (327) | 138 (413) | 0.74 | |
| Control | 3612 (2520) |
Abbreviation: SEM, standard error of the mean.
>10% attenuation.
Table 4.
Mediational Regression Analysis in BPA-Exposed Female Offspring: Carbon Dioxide Production
| Mediator and Exposure Group | Mean Carbon Dioxide Production (mg/kg/h) (SEM) | Overall P | Mean Production Difference (SEM) vs. Control | P |
|---|---|---|---|---|
| None (model 1) | 0.59 | |||
| mg | 3584 (182) | 320 (250) | 0.20 | |
| µg | 3692 (168) | 428 (239) | 0.08 | |
| ng | 3736 (166) | 472 (239) | 0.05 | |
| Control | 3263 (171) | |||
| Jak-2 CpG site 5/6 (model 4) | 0.17 | |||
| mg | 3833 (283) | 378 (356) | 0.29 | |
| µg | 4124 (201) | 668 (297) | 0.03 | |
| ng | 3891 (243) | 435 (326) | 0.19 | |
| control | 3455 (218) | |||
| Rfxap CpG site 1 (model 4) | 0.14 | |||
| ng | 3795 (161) | 339 (228)a | 0.14 | |
| Control | 3456 (158) | |||
| Rfxap CpG site 16 (model 4) | 0.14 | |||
| ng | 3808 (163) | 346 (232)a | 0.14 | |
| Control | 3462 (162) | |||
| Rfxap CpG site 24 (model 4) | 0.28 | |||
| ng | 3748 (151) | 234 (215)a | 0.28 | |
| Control | 3515 (149) | |||
| Rfxap CpG site 30 (model 4) | 0.14 | |||
| ng | 3795 (161) | 339 (228)a | 0.14 | |
| Control | 3456 (158) | |||
| Rfxap CpG site 31 (model 4) | 0.14 | |||
| ng | 3808 (163) | 346 (232)a | 0.14 | |
| Control | 3462 (162) | |||
| Tmem238 CpG site 9 (model 4) | 0.04 | |||
| µg | 4114 (207) | 652 (306) | 0.04 | |
| Control | 3463 (226) | |||
| Rxr CpG site 18/19/20 (model 4) | 0.14 | |||
| mg | 3871 (223) | 416 (282)a | 0.14 | |
| Control | 3454 (172) |
Abbreviation: SEM, standard error of the mean.
>10% attenuation.
After modeling of body weight and fat as the outcome, the differences in body weight and fat of the nanogram-exposed females compared with the control females were attenuated >10% after adjustment for methylation at Rfxap (Tables 5 and 6). Additionally, the differences in body weight and fat of the milligram-exposed females compared with the control females were attenuated >10% after adjustment for methylation at Tmem238 (Tables 5 and 6).
Table 5.
Mediational Regression Analysis in BPA-Exposed Female Offspring: Body Weight
| Mediator and Exposure Group | Mean Body Weight (kg) (SEM) | Overall P | Mean Body Weight Difference (SEM) vs. Control | P |
|---|---|---|---|---|
| None (model 1) | 0.06 | |||
| mg | 29 (1.5) | 3.1 (2.0) | 0.12 | |
| µg | 30.3 (1.5) | 1.8 (1.9) | 0.36 | |
| ng | 28.4 (1.5) | 3.7 (1.9) | 0.06 | |
| Control | 32.1 (1.5) | |||
| Jak-2 CpG site 5/6 (model 4) | 0.16 | |||
| mg | 29.6 (2.3) | −3.1 (2.7) | 0.27 | |
| µg | 30 (1.7) | −2.7 (2.4) | 0.27 | |
| ng | 27 (1.8) | −5.7 (2.4) | 0.03 | |
| control | 32.6 (1.6) | |||
| Rfxap CpG site 1 (model 4) | 0.15 | |||
| ng | 28.3 (1.9) | −4.1 (2.7)a | 0.15 | |
| Control | 32.5 (1.8) | |||
| Rfxap CpG site 16 (model 4) | 0.06 | |||
| ng | 27.3 (2.0) | −6 (2.9) | 0.06 | |
| Control | 33.3 (1.9) | |||
| Rfxap CpG site 24 (model 4) | 0.19 | |||
| ng | 28.3 (3.0) | −4.1 (3)a | 0.19 | |
| Control | 32.4 (1.9) | |||
| Rfxap CpG site 30 (model 4) | 0.15 | |||
| ng | 28.3 (1.9) | −4.1 (2.7)a | 0.15 | |
| Control | 32.5 (1.8) | |||
| Rfxap CpG site 31 (model 4) | 0.06 | |||
| ng | 27.3 (2.0) | −6 (2.9) | 0.06 | |
| Control | 33.3 (1.9) | |||
| Tmem238 CpG site 9 (model 4) | 0.64 | |||
| µg | 31.2 (1.1) | −0.76 (1.6)a | 0.64 | |
| Control | 32 (1.1) | |||
| Rxr CpG site 18/19/20 (model 4) | 0.17 | |||
| mg | 28.4 (2.1) | −3.9 (2.7) | 0.17 | |
| Control | 32.3 (1.6) |
Abbreviation: SEM, standard error of the mean.
>10% attenuation.
Table 6.
Mediational Regression Analysis in BPA-Exposed Female Offspring: Body Fat
| Mediator and Exposure Group | Mean Fat Mass (g) (SEM) | Overall P | Mean Fat Difference (SEM) vs. Control | P |
|---|---|---|---|---|
| None (model 1) | 0.76 | |||
| mg | 4.5 (0.73) | −1.7 (1.0) | 0.10 | |
| µg | 5.2 (0.68) | −0.93 (0.96) | 0.33 | |
| ng | 4.4 (0.67) | −1.7 (0.96) | 0.08 | |
| Control | 6.2 (0.68) | |||
| Jak-2 CpG site 5/6 (model 4) | 0.45 | |||
| mg | 4.8 (1.3) | −1.4 (1.5) | 0.37 | |
| µg | 5.1 (0.95) | −1.1 (1.3) | 0.41 | |
| ng | 4.1 (0.99) | −2.1 (1.3) | 0.13 | |
| control | 6.2 (0.86) | |||
| Rfxap CpG site 1 (model 4) | 0.29 | |||
| ng | 4.7 (0.96) | −1.5 (1.4)a | 0.29 | |
| Control | 6.2 (0.89) | |||
| Rfxap CpG site 16 (model 4) | 0.13 | |||
| ng | 4.3 (1.0) | −2.4 (1.4)a | 0.13 | |
| Control | 6.7 (0.93) | |||
| Rfxap CpG site 24 (model 4) | 0.39 | |||
| ng | 4.9 (0.99) | −1.3 (1.4)a | 0.39 | |
| Control | 6.1 (0.91) | |||
| Rfxap CpG site 30 (model 4) | 0.29 | |||
| ng | 4.7 (0.96) | −1.5 (1.4)a | 0.29 | |
| Control | 6.2 (0.89) | |||
| Rfxap CpG site 31 (model 4) | ||||
| ng | 4.3 (1.0) | 0.13 | −2.4 (1.4) | 0.13 |
| Control | 6.7 (0.93) | |||
| Tmem238 CpG site 9 (model 4) | 0.93 | |||
| µg | 5.8 (0.84) | 0.09 (1.2)a | 0.93 | |
| Control | 5.9 (0.79) | |||
| Rxr CpG site 18/19/20 (model 4) | 0.24 | |||
| mg | 4.2 (1.2) | 1.9 (1.5) | 0.24 | |
| Control | 6.1 (0.87) |
Abbreviation: SEM, standard error of the mean.
>10% attenuation.
Finally, when leptin was modeled as the outcome, the differences in leptin in the µg exposed females compared with the control females were attenuated >10% after adjustment for methylation at Tmem238 (Table 7). The remaining metabolic phenotypes did not display mediation after adjustment for methylation at these 4 candidate genes (Supplemental Tables 1–7 (29.5KB, xls) en.2016-1441.st2.xls (29.5KB, xls) en.2016-1441.st3.xls (29.5KB, xls) en.2016-1441.st4.xls (29.5KB, xls) en.2016-1441.st5.xls (29.5KB, xls) en.2016-1441.st6.xls (30KB, xls) en.2016-1441.st7.xls (29.5KB, xls) en.2016-1441.sd1.doc (30KB, doc) ).
Table 7.
Mediational Regression Analysis in BPA-Exposed Female Offspring: Leptin Levels
| Mediator and Exposure Group | Mean Leptin Level (ng/mL) (SEM) | Overall P | Mean Leptin Difference (SEM) vs. Control | P |
|---|---|---|---|---|
| None (model 1) | 0.38 | |||
| mg | 11.2 (3.1) | −6.7 (4.3) | 0.14 | |
| µg | 13.6 (2.7) | −4.4 (4.0) | 0.29 | |
| ng | 17.3 (3.2) | −0.65 (4.4) | 0.88 | |
| Control | 18 (3.0) | |||
| Jak-2 CpG site 5/6 (model 4) | 0.18 | |||
| mg | 12.6 (4.0) | −8.5 (4.8) | 0.10 | |
| µg | 13.2 (3.1) | −7.8 (4.5) | 0.10 | |
| ng | 19.9 (4.8) | −1.2 (5.8) | 0.84 | |
| control | 21.1 (3.1) | |||
| Rfxap CpG site 1 (model 4) | 0.76 | |||
| ng | 21.1 (4.2) | −1.6 (5.2) | 0.76 | |
| Control | 22.7 (3.0) | |||
| Rfxap CpG site 16 (model 4) | 0.65 | |||
| ng | 20.6 (4.1) | −2.4 (5.2) | 0.65 | |
| Control | 23 (2.9) | |||
| Rfxap CpG site 24 (model 4) | 0.89 | |||
| ng | 21.7 (4.4) | −0.8 (5.6) | 0.89 | |
| Control | 22.5 (3.1) | |||
| Rfxap CpG site 30 (model 4) | 0.76 | |||
| ng | 21.1 (4.2) | −1.6 (5.2) | 0.76 | |
| Control | 22.7 (3) | |||
| Rfxap CpG site 31 (model 4) | 0.65 | |||
| ng | 20.6 (4.1) | −2.4 (5.2) | 0.65 | |
| Control | 23 (2.9) | |||
| Tmem238 CpG site 9 (model 4) | 0.43 | |||
| µg | 15.6 (2.8) | −3.5 (4.2)a | 0.43 | |
| Control | 19.1 (3.0) | |||
| Rxr CpG site 18/19/20 (model 4) | 0.18 | |||
| mg | 11.4 (4.3) | −8.6 (5.9) | 0.18 | |
| Control | 20 (3.6) |
Abbreviation: SEM, standard error of the mean.
>10% attenuation.
Discussion
Previous work has shown that physiologically relevant levels of BPA exposure during early development result in altered levels of methylation in candidate gene and global methylation assays (12, 14, 17). In this study, we extend beyond the traditional candidate and global methylation approaches by taking advantage of an epigenome-wide platform. Several differentially methylated genomic loci in mouse liver promoter regions following perinatal BPA exposure were discovered. We further applied the discovery of regions of altered methylation for the identification of genes involved in specific disease pathways that develop after exposure to an environmental insult.
A biological pathway analysis revealed enrichment of genes involved in hormone regulation, such as insulin and leptin, as well as in neural signaling pathways (Table 1). These pathways support data from previously reported developmental exposure to BPA and resulting adult phenotypes (25–27). By using biological pathway annotation tools, studies that use epigenome-wide platforms have the opportunity to identify genomic loci that are differentially methylated in a specific disease pathway. For example, Sabunciyan et al. examined DNA methylation of the prefrontal cortex from deceased individuals known to have major depressive disorder (28). Biological pathway analysis was run on the top hit differentially methylated regions to determine which genes may be epigenetically susceptible in the progression of major depressive disorder (28). Additionally, epigenomic biomarkers were identified in patients with normal, premalignant, and oral-cavity squamous cell carcinoma tumor tissue by using an epigenome-wide platform (29). The differentially methylated genes were found to be associated with the pathways closely related to oncogenic transformation following pathway enrichment. Thus, pathway analyses provide a means to understand the biological meaning behind a large set of differentially methylated genes generated from high-density epigenome-wide platforms. Many, but not all, of the array-identified differentially methylated regions were validated via quantitative bisulfite sequencing. Lack of validation may represent a true negative result or failure to reach significance because the sample size was too small for validation. Future work should also measure messenger RNA for a gene expression to further validate direct biological effects of altered DNA methylation.
In addition to identifying differentially methylated genes by BPA exposure, we were able to combine genome-wide methylation data with lifecourse phenotyping measures. A mediational regression analysis was performed to assess whether DNA methylation mediates early-life BPA exposure and resulting adult metabolic phenotypes. The mediational regression analysis revealed that Jak-2 and Rxr (top hit candidate genes selected from metabolic pathways) and Rfxap and Tmem238 (the top two most differentially methylated genes as indicated by lowest P value) were mediators of BPA exposure and oxygen consumption, carbon dioxide production, body weight, fat, and leptin levels. Characterization of epigenetic biomarkers indicating developmental BPA exposure and predictive of disease development will result in unique opportunities to develop disease prevention strategies; this characterization may also lead to therapeutic modifications of these differentially methylated regions acting as mediators in individuals who may be predisposed to disease development or in individuals with existing disease to facilitate reversal of disease progression. Because of the plasticity of the epigenome (30), such strategies may include dietary supplementation or pharmaceutical intervention (14).
During the past decade, epigenetic technologies have evolved from once-traditional methods with restriction enzymes or candidate genes to modern technologies that allow for unbiased epigenome-wide investigation across tissues and species, yet most studies of the effects on the epigenome following environmental and nutritional exposures have been limited by being candidate gene–driven or global-driven. Because the main goal of DOHaD research is to elucidate the role that early environmental insults have on health and disease, epigenome-wide platforms, such as deep sequencing and tiling array technologies, are of value to elucidate key regulatory pathways that are involved in the etiology of disease via epigenetic mechanisms. A key initiative of future DOHaD research will be to establish epigenomic differences between target and surrogate tissues to define epigenetic profiles that are clinically relevant for disease risk or treatment because in many cases, target tissue is not feasible for population profiling. Furthermore, there is now evidence that DNA methylation may act in concert with other epigenetic mechanisms, such as histone modifications and chromatin remodeling complexes, and should also be considered in evaluation of the epigenome (31).
Acknowledgments
The authors would like to thank Dr. Thomas Glover and Sountharia Rajendran for training and equipment assistance with microarray experiments.
Acknowledgments
This work was supported by a pilot grant from the Michigan Nutrition Obesity Research Center (P30 DK089503), as well as a National Institutes of Health grant (R01 ES017524), the Michigan Lifestage Environmental Exposures and Disease National Institute of Environmental Health Sciences (NIEHS) Center of Excellence (P30 ES017885), the University of Michigan NIEHS/Environmental Protection Agency Children’s Environmental Health Formative Center (P20 ES018171/RD83480001), and University of Michigan NIEHS Institutional Training Grant (T32 ES007062).
Acknowledgments
Disclosure Summary: The authors have nothing to disclose.
Footnotes
- BPA
- bisphenol A
- DOHaD
- developmental origins of health and disease
- MBD
- methyl CpG-binding domain
- PCR
- polymerase chain reaction
- RAM
- region of altered methylation
References
- 1.Barker DJ. The developmental origins of adult disease. J Am Coll Nutr. 2004;23(6 Suppl)588S–595S. [DOI] [PubMed] [Google Scholar]
- 2.Vandenberg LN, Chahoud I, Heindel JJ, Padmanabhan V, Paumgartten FJR, Schoenfelder G. Urinary, circulating, and tissue biomonitoring studies indicate widespread exposure to bisphenol A. Environ Health Perspect. 2010;118(8):1055–1070. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Gould JC, Leonard LS, Maness SC, Wagner BL, Conner K, Zacharewski T, Safe S, McDonnell DP, Gaido KW. Bisphenol A interacts with the estrogen receptor α in a distinct manner from estradiol. Mol Cell Endocrinol. 1998;142(1-2):203–214. [DOI] [PubMed] [Google Scholar]
- 4.Kuiper GGJM, Lemmen JG, Carlsson B, Corton JC, Safe SH, van der Saag PT, van der Burg B, Gustafsson J-Å. Interaction of estrogenic chemicals and phytoestrogens with estrogen receptor β. Endocrinology. 1998;139(10):4252–4263. [DOI] [PubMed] [Google Scholar]
- 5.Krüger T, Long M, Bonefeld-Jørgensen EC. Plastic components affect the activation of the aryl hydrocarbon and the androgen receptor. Toxicology. 2008;246(2-3):112–123. [DOI] [PubMed] [Google Scholar]
- 6.Moriyama K, Tagami T, Akamizu T, Usui T, Saijo M, Kanamoto N, Hataya Y, Shimatsu A, Kuzuya H, Nakao K. Thyroid hormone action is disrupted by bisphenol A as an antagonist. J Clin Endocrinol Metab. 2002;87(11):5185–5190. [DOI] [PubMed] [Google Scholar]
- 7.Sui Y, Ai N, Park S-H, Rios-Pilier J, Perkins JT, Welsh WJ, Zhou C. Bisphenol A and its analogues activate human pregnane X receptor. Environ Health Perspect. 2012;120(3):399–405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Braun JM, Hauser R. Bisphenol A and children’s health. Curr Opin Pediatr. 2011;23(2):233–239 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Wolstenholme JT, Rissman EF, Connelly JJ. The role of Bisphenol A in shaping the brain, epigenome and behavior. Horm Behav. 2011;59(3):296–305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Anderson OS, Peterson KE, Sanchez BN, Zhang Z, Mancuso P, Dolinoy DC. Perinatal bisphenol A exposure promotes hyperactivity, lean body composition, and hormonal responses across the murine life course. FASEB J. 2013;27(4):1784–1792. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Gabory A, Attig L, Junien C. Developmental programming and epigenetics. Am J Clin Nutr. 2011;94(6 Suppl)1943S–1952S. [DOI] [PubMed] [Google Scholar]
- 12.Anderson OS, Nahar MS, Faulk C, Jones TR, Liao C, Kannan K, Weinhouse C, Rozek LS, Dolinoy DC. Epigenetic responses following maternal dietary exposure to physiologically relevant levels of bisphenol A. Environ Mol Mutagen. 2012;53(5):334–342. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Anderson O. Perinatal bisphenol A exposure: Effects on metabolic homeostasis mediated by epigenetic labile loci. University of Michigan. 2013. Available at: https://deepblue.lib.umich.edu/bitstream/handle/2027.42/100103/oliviasa_1.pdf?sequence=1. Accessed April 2, 2016.
- 14.Dolinoy DC, Huang D, Jirtle RL. Maternal nutrient supplementation counteracts bisphenol A-induced DNA hypomethylation in early development. Proc Natl Acad Sci USA. 2007;104(32):13056–13061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Jirtle RL, Skinner MK. Environmental epigenomics and disease susceptibility. Nat Rev Genet. 2007;8(4):253–262. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kundakovic M, Champagne FA. Epigenetic perspective on the developmental effects of bisphenol A. Brain Behav Immun. 2011;25(6):1084–1093. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ho SM, Tang WY, Belmonte de Frausto J, Prins GS. Developmental exposure to estradiol and bisphenol A increases susceptibility to prostate carcinogenesis and epigenetically regulates phosphodiesterase type 4 variant 4. Cancer Res. 2006;66(11):5624–5632. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Bromer JG, Zhou Y, Taylor MB, Doherty L, Taylor HS. Bisphenol-A exposure in utero leads to epigenetic alterations in the developmental programming of uterine estrogen response. FASEB J. 2010;24(7):2273–2280. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Yoshida H, Broaddus R, Cheng W, Xie S, Naora H. Deregulation of the HOXA10 homeobox gene in endometrial carcinoma: role in epithelial-mesenchymal transition. Cancer Res. 2006;66(2):889–897. [DOI] [PubMed] [Google Scholar]
- 20.Tang WY, Morey LM, Cheung YY, Birch L, Prins GS, Ho SM. Neonatal exposure to estradiol/bisphenol A alters promoter methylation and expression of Nsbp1 and Hpcal1 genes and transcriptional programs of Dnmt3a/b and Mbd2/4 in the rat prostate gland throughout life. Endocrinology. 2012;153(1):42–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Weinhouse C, Anderson OS, Bergin IL, Vandenbergh DJ, Gyekis JP, Dingman MA, Yang J, Dolinoy DC. Dose-dependent incidence of hepatic tumors in adult mice following perinatal exposure to bisphenol A. Environ Health Perspect. 2014;122(5):485–491. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Adriaens ME, Jaillard M, Eijssen LMT, Mayer C-D, Evelo CTA. An evaluation of two-channel ChIP-on-chip and DNA methylation microarray normalization strategies. BMC Genomics. 2012;13(1):42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Smyth GK;K SG . Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol. 2004;3:e3. [DOI] [PubMed] [Google Scholar]
- 24.Clayton JA, Collins FS. Policy: NIH to balance sex in cell and animal studies. Nature. 2014;509(7500):282–283. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Ben-Jonathan N, Hugo ER, Brandebourg TD. Effects of bisphenol A on adipokine release from human adipose tissue: Implications for the metabolic syndrome. Mol Cell Endocrinol. 2009;304(1-2):49–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Mizuo K, Narita M, Yoshida T, Narita M, Suzuki T. Functional changes in dopamine D3 receptors by prenatal and neonatal exposure to an endocrine disruptor bisphenol-A in mice. Addict Biol. 2004;9(1):19–25. [DOI] [PubMed] [Google Scholar]
- 27.Zhou R, Bai Y, Yang R, Zhu Y, Chi X, Li L, Chen L, Sokabe M, Chen L. Abnormal synaptic plasticity in basolateral amygdala may account for hyperactivity and attention-deficit in male rat exposed perinatally to low-dose bisphenol-A. Neuropharmacology. 2011;60(5):789–798. [DOI] [PubMed] [Google Scholar]
- 28.Sabunciyan S, Aryee MJ, Irizarry RA, Rongione M, Webster MJ, Kaufman WE, Murakami P, Lessard A, Yolken RH, Feinberg AP, Potash JB, Consortium G. Genome-wide DNA methylation scan in major depressive disorder. PLoS One 2012;7(4):e34451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Guerrero-Preston R, Soudry E, Acero J, Orera M, Moreno-López L, Macía-Colón G, Jaffe A, Berdasco M, Ili-Gangas C, Brebi-Mieville P, Fu Y, Engstrom C, Irizarry RA, Esteller M, Westra W, Koch W, Califano J, Sidransky D. NID2 and HOXA9 promoter hypermethylation as biomarkers for prevention and early detection in oral cavity squamous cell carcinoma tissues and saliva. Cancer Prev Res (Phila). 2011;4(7):1061–1072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Waterland RA, Jirtle RL. Transposable elements: targets for early nutritional effects on epigenetic gene regulation. Mol Cell Biol. 2003;23(15):5293–5300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Dolinoy DC, Weinhouse C, Jones TR, Rozek LS, Jirtle RL. Variable histone modifications at the A(vy) metastable epiallele. Epigenetics. 2010;5(7):637–644. [DOI] [PMC free article] [PubMed] [Google Scholar]