Endometriosis, endocrine disrupters, and epigenetics: an investigation into the complex interplay in women with polybrominated biphenyl exposure and endometriosis

Abstract

Purpose

Endocrine disrupting compounds (EDCs) have been shown to affect multiple biologic processes especially steroid-hormone processes. We sought to determine differences in DNA methylation exists between women with and without endometriosis following exposure to polybrominated biphenyl (PBB).

Methods

Cross-sectional study of 305 females in the Michigan PBB Registry. DNA was extracted, and DNA methylation was interrogated using the MethylationEPIC BeadChip (Illumina, San Diego, California). Demographic data was analyzed using Chi-squared and T tests. Linear regressions were performed for each cytosine-guanine dinucleotide (CpG) site, modeling the logit transformation of the β value as a linear function of the presence of endometriosis. Sensitivity analyses were conducted controlling for estradiol levels and menopausal status. Replication study performed evaluating for any association between CpGs reported in the literature and our findings.

Results

In total, 39,877 CpGs nominally associated with endometriosis (p < 0.05) after adjusting for age and cellular heterogeneity, although none remained significant after correction for multiple comparisons (FDR < 0.05). Pathway analysis of these CpGs showed enrichment in 68 biologic pathways involved in various endocrine, immunologic, oncologic, and cell regulation processes as well as embryologic reproductive tract development and function (FoxO, Wnt, and Hedgehog signaling). We identified 42,261 CpG sites in the literature reported to be associated with endometriosis; 2012 of these CpG sites were also significant in our cohort.

Conclusion

We found 39,877 CpG sites that nominally associated with endometriosis (p < 0.05) after adjusting for age and cellular heterogeneity; however, none remained significant after correction for multiple comparisons (FDR < 0.05).

Electronic supplementary material

The online version of this article (10.1007/s10815-020-01695-9) contains supplementary material, which is available to authorized users.

Introduction

Endometriosis is a chronic disease that is estimated to affect approximately 6–10% of reproductive age women, between 50 and 80% of women with pelvic pain, and up to 50% of women with infertility [1–5]. Effects of the disease on physical, mental, and social wellbeing are well established, and impaired quality of life and work productivity across countries and ethnicities has been noted [6–9]. Furthermore, women with endometriosis have a two- to threefold increase in absolute risk of developing epithelial ovarian cancer, especially clear cell and endometrioid subtypes. Increased risk for other malignancies has been suggested, including endometrial and breast [10, 11]. Studies also suggest that women with endometriosis may be at higher risk of developing cardiovascular disease and other promoting factors, including hypertension and hypercholesterolemia [12, 13]. These elevated risks also persist beyond menopause, when symptoms from endometriosis typically abate.

The defining characteristic of endometriosis is ectopic growth of endometrial stromal and glandular cells outside of the uterine cavity, typically in the peritoneal cavity [4, 14]. Common explanations include Sampson’s theory of retrograde menstruation, immune-mediated, hematologic spread of menstrual tissue, and contribution of mesenchymal stem cells or ceolomic metaplasia [14–18]. An environmental basis has also been suggested in the pathogenesis of endometriosis [4, 19–21]. The etiology is complex and multi-faceted; however knowledge of the health effects of endocrine disrupting compounds (EDCs), particularly estrogen-like EDCs, and their impact on gene expression continues to be researched. Thus, a shift towards understanding the genetic and/or environmental components of the disease, particularly through the study of epigenetic mechanisms such as DNA methylation, has occurred with promising momentum.

EDCs can affect multiple biologic processes especially steroid-hormone dependent human reproductive tract development and adult reproductive function [22–26]. One particular endocrine disrupting compound, a brominated flame retardant (BFR), was introduced into the food supply in rural Michigan farms in 1973 after the accidental contamination of livestock feed with FireMaster FF-1, a polybrominated biphenyls (PBB). Shortly thereafter, significant health effects were noted in the livestock, and the Michigan Department of Community Health (MDCH) began enrolling Michigan residents who lived on affected farms and/or had been exposed to PBB through the contamination of meat and dairy products as a result of this contamination into the Michigan PBB Registry, a cohort with currently over 7000 people [27, 28].

While PBBs have been banned and are no longer in production, significant concerns over their health effects remain because they are persistent in biological systems, accumulate in adipose tissue, and have long half-lives of 11–29 years in females [27, 29–32]. Specifically, in the Michigan PBB cohort, while the exposure occurred over 40 years ago, reproductive system health effects are evident in women exposed to PBB through this accident and their children. The main congener of FireMaster FF-1, PBB153, is a diortho-substituted congener which may be neurotoxic and estrogenic, and earlier studies in the Michigan PBB cohort have revealed earlier menarche, and earlier pubic hair stage was noted in pubertal girls with PBB exposure in utero and through breast milk [33–36]. Some other findings to date include increased risk for spontaneous abortion and increased risk of genitourinary conditions in children [28, 37–41]. However, it is still not clear what links PBB exposure to its associated health effects. One study recently found 1890 CpG sites to be associated with total PBB levels in participants in the Michigan PBB cohort. While these CpGs were not enriched in any particular biological pathway, they were more likely to be in estrogen receptor binding sites [42].

Regarding endometriosis, evidence linking classes of environmental chemicals to endometriosis in humans is also rapidly emerging. Phthalates, dioxin, and particular congeners of polychlorinated biphenyl (PCB) have been found to be associated with endometriosis in animal studies [9, 43–45]. Earlier epidemiologic studies in the Michigan PBB cohort did not find any association between endometriosis and PBB but did find an increased incidence of endometriosis correlating to PCB exposure [46]. Furthermore, studies in DNA methylation have found particular CpGs to be associated with endometriosis [47–51]. Yotova et al. [47] compared CpG methylation in isolated endometrial stromal cells obtained from eutopic endometria with and without endometriosis; Dyson et al. [49] and Yamagata et al. [48] also included ovarian cysts. The objective of our study was to determine whether differences in DNA methylation were present in women exposed to PBB and endometriosis.

Materials and methods

Study population and participants

Participants for this study were selected from the Michigan Polybrominated Biphenyl (PBB) Registry. Original enrollment in this registry was conducted by the Michigan Department of Community Health (MDCH; currently MDHHS, Michigan Department of Health and Human Services) and primarily included people who lived on farms affected by PBB, bought food from affected farms, or worked at the chemical plant that produced PBB and their family members [28, 52, 53]. These participants have been followed since the 1970s and have answered questionnaires about their health status. Recently, this registry has been transferred to Emory University, and original registry members’ children and other members of the community who were also exposed to PBB have been enrolled (http://pbbregistry.emory.edu/).

For the purposes of this study, participants were selected who had (1) available epigenetic data, (2) self-reported as being female, and (3) answered questionnaire information concerning whether they had ever received a doctor diagnosis of endometriosis. For cases, individuals had to indicate that they had received a doctor diagnosis of endometriosis at any time; and for controls, individuals had to indicate that they did not receive a doctor diagnosis of endometriosis on a questionnaire. All included individuals completed a questionnaire within a year of their blood sample being taken for epigenetic analysis. There were 65 cases and 240 controls who met these criteria. Whether these individuals were menopausal was also of interest. Individuals who reported being menopausal, or who were over the age of 55 years at sample collection were classified as menopausal (N = 140). Women self-reporting as non-menopausal and < 55 years of age were classified as pre-menopausal (N = 163). Self-reported menopausal status for two women under the age of 55 years was missing from their health questionnaire. Other obstetric data including whether the subject experienced a miscarriage or pre-term birth; and reproductive health information including whether the subject experienced a fertility problem or early menopause (before 40 years of age), was also collected from the health questionnaires.

Measurement of PBB exposure

Given the molecular structure of PBB, a total of 209 possible congeners exist and are distinguished by the total number and position of the bromine atoms around the biphenyl ring [54]. PBB-153 (2,2′,4,4′,5,5′-hexa-bromobiphenyl) is the predominant PBB congener in biological matrices and made up an average of 61% of the fire retardant mixture involved in the Michigan exposure [36, 40]. In addition to PBB-153, exposure to three other common congeners of PBB (PBB-77, PBB-101, PBB-180) was previously assessed in members of this cohort using gas chromatography-tandem mass spectrometry (GC-MS/MS) [55]. The limit of detection (LOD) for PBB-153 in the serum was 0.002 ng/mL; for PBB-77 it was 0.0045 ng/mL; for PBB-101 it was 0.0039 ng/mL; for PBB-180 it was 0.0056 ng/mL [55]. The reliability of this method was assessed by running repeat measurements against known quantities of PBB as previously described [55]. The four imputed congeners were then summed to give a total PBB value per person. The total PBB was then transformed using a natural log in order to have a less skewed distribution of values. The extraction recovery ranged from 83.2–99.2%, the accuracy ranged from 89 to 119%, and the precision ranged from 2.8–8.5%, as previously described [55].

DNA extraction

Whole blood was centrifuged at 3000 rotations per minute (rpm) to separate the plasma from the buffy coat. Buffy coats were aliquoted and stored at − 80 °C. DNA was extracted from buffy coat samples using the QIAamp DNA Blood Mini Kit (QIAGEN, Hilden, Germany) according to manufacturer’s instructions. DNA was extracted from the blood sample from which the PBB exposure levels were calculated.

DNA methylation

DNA methylation was interrogated using the Infinium MethylationEPIC BeadChip (Illumina, San Diego, CA) as previously described [42]. For each of the 850,000 CpG dinucleotides queried on the array, a β value was generated that approximates, based on relative intensity values from bisulfite-treated and control DNA, the proportion of DNA methylated at each CpG site. After excluding samples with low signal and any samples with missing data or missing probes, our dataset contained a total of 816,999 CpG sites. Finally, we used corrected methylated (M) and unmethylated (U) signals to compute β values (estimates of the methylation proportion β = M/(U + M). β values were then logit transformed. Estimated cell type proportions were calculated for each sample using Houseman’s method that uses methylation signals that are distinct in each blood cell type [56, 57]. The DNA methylation data is publicly available through NCBI’s Gene Expression Omnibus (GSE116339).

Statistical analyses

All analyses were performed in R version 3.5.0. Demographic data was analyzed using Chi-squared and T tests. We used linear regressions to test for the association between each CpG site and endometriosis. Each linear model was adjusted for covariates including age at blood draw and estimated cell type proportions (CD8T, CD4T, natural killer cells, B cells, and monocytes). Sensitivity analysis was performed to determine if these associations were independent of PBB and estradiol levels using linear regressions controlling for the previously described covariates and imputed estradiol levels. Estradiol levels were imputed directly from DNA methylation data using the method described via personal communication.

To evaluate whether menopausal status affected epigenetic patterns in women with endometriosis, we evaluated pre- and post-menopausal women separately and then determined which CpG sites were common to both groups and significantly associated with endometriosis. Next, we used the package missMethyl to determine if the CpG sites associated with endometriosis were enriched in any particular biological pathways [58]. Lastly, we compared the CpG sites documented in the literature to be associated with endometriosis to our array by performing a linear regression in the same fashion as our primary analysis [47–49]. We were unable to obtain copies of the supplemental files to include the data from Saare et al. [50] and Rahmioglu et al. [51]. For all epigenetic analyses, a Storey’s q value of 5% was used as the false discovery rate (FDR) to correct for multiple comparisons (FDR < 0.05) [59].

Ethical approval

The Institutional Review Boards at Emory University approved the protocols, and participants gave written informed consent.

Results

Cohort demographics

Sixty-five women (21.3% of our cohort) reported a physician-made diagnosis of endometriosis. Age at survey completion and blood draw, average age at time of PBB exposure, and imputed estradiol levels were all comparable between women with and without endometriosis (Table 1). Women with endometriosis had a higher incidence of reported early menopause (p = 0.001), fertility problems (p = 2.6 × 10⁻⁶), and reported history of miscarriage (p = 0.002) than women without endometriosis (Tables 1 and 2). Interestingly, women with endometriosis also had slightly higher total PBB levels (0.43 ng/mL vs. 0.30 ng/mL), although not statistically significant (p = 0.08). No differences were noted in cell type composition between women with and without endometriosis in CD8T (p = 0.15), CD4T (p = 0.70), B cell (p = 0.26), natural killer cells (p = 0.99), monocytes (p = 0.51), and granulocyte concentrations (p = 0.28).

Table 1.

Cohort demographics for PBB exposed women with and without endometriosis

Demographic	Endo (N = 65) Mean (SD)	No endo (N = 240) Mean (SD)	95% CI	T statistic	p value
Age, years	48.9 (9.2)	50.0 (11.5)	− 1.61, 3.79	0.80	0.43
Age at PBB exposure	10.3 (7.5)	11.4 (9.7)	− 1.21, 3.24	0.90	0.37
Total PBB a,b,c (ng/mL)	0.43 (4.09)	0.30 (3.99)	− 0.74, 0.04	− 1.78	0.08
Imputed estradiola,d (ng/dL)	60.27 (1.53)	63.48 (1.50)	− 0.06, 0.17	0.88	0.38

Endo endometriosis, SD standard deviation

^aGeometric mean and geometric standard deviation

^bNanogram per milliliter (ng/mL)

^cAdjusting for lipid levels, BMI, and age attenuated the difference in PBB between subjects with endometriosis and controls (p = 0.41)

^dNanogram per deciliter (ng/dL)

Table 2.

Reproductive demographics of PBB exposed women with and without endometriosis

Reproductive demographics	Endo N = 65 (%)	No endo N = 240 (%)	χ2	p value
Early menopause	17 (26)	23 (9.5)	10.91	0.001
Fertility problema	27 (42)	36 (15)	22.09	2.6 × 10−6
Number of women reporting a preterm birth	14 (22)	40 (17)	1.29	0.26
Number of women reporting a miscarriage	25 (38)	48 (20)	9.74	0.002

Endo endometriosis, χ² Chi-squared

^aNumber of women diagnosed with infertility by a physician

p value calculated by Chi-squared test

DNA methylation and endometriosis

In total, 39,877 CpG sites nominally associated with endometriosis after adjusting for age and cellular heterogeneity (4.9%; p < 0.05), although none remained significant after correction for multiple comparisons (FDR < 0.05) (Table 3; Supplemental Table 1). We evaluated whether CpGs associated with endometriosis were enriched in any particular biological pathways. We identified 68 pathways involved in various other endocrine, immunologic, oncologic, and cell regulation processes (Supplemental Table 2). As estrogen levels vary across the menstrual cycle and endometriosis is an estrogen-dependent disease, we wanted to identify if estradiol levels had any influence on the DNA methylation findings in our cohort. In the sensitivity analysis controlling for estradiol levels, we found no significant difference in the CpGs associated with endometriosis, confirming that findings were independent of estradiol levels (Supplemental Table 3). The nominally significant CpG results for endometriosis with and without controlling for imputed estradiol levels were highly correlated (r = 0.99, p < 2.2e-16, Fig. 1).

Table 3.

Top CpG sites associated with endometriosis

CpG site	Position	Gene	T statistic	p value	FDR
cg04883592	chr18: 55988293	NEDD4L	− 5.45	1.08 × 10−7	0.09
cg18268085	chr5: 168081123		− 4.95	1.26 × 10−6	0.51
cg25542772	chr1: 10273965	KIF1B	− 4.79	2.65 × 10−6	0.59
cg00454719	chr13: 37453500	SMAD9	− 4.77	2.90 × 10−6	0.59
cg00610519	chr9: 139288284	SNAPC4	− 4.67	4.49 × 10−6	0.60
cg01359081	chr6: 146942205	ADGB	4.65	5.04 × 10−6	0.60
cg00718959	chr5: 168080990		− 4.64	5.35 × 10−6	0.60
cg10501758	chr2: 128223957		4.61	5.91 × 10−6	0.60
cg16257868	chr1: 220446207	RAB3GAP2	4.52	8.85 × 10−6	0.72
cg04702354	chr6: 151410066	MTHFD1L	− 4.45	1.22 × 10−5	0.72

Fig. 1.

Effect size for endometriosis CpGs with and without estradiol levels. A sensitivity analysis controlling for estradiol levels showed no significant difference in the CpGs associated with endometriosis, confirming that our findings were independent of estradiol levels. The nominally significant CpG results for endometriosis with and without controlling for imputed estradiol levels were highly correlated (r = 0.99, p < 2.2e-16)

DNA methylation and menopausal status

We assessed the degree to which DNA methylation patterns in pre-menopausal women compared to those of post-menopausal women in lieu of the markedly different estrogen environments. A total of 140 women (46% of our cohort) were menopausal at the time of assessment and blood draw. When evaluating whether DNA methylation associates with menopausal status, a total of 37,172 CpGs were nominally associated (p < 0.05). When the association between endometriosis and DNA methylation was analyzed separately in pre- (n = 163) and post-menopausal women, a total of 30,315 CpG sites were nominally associated with endometriosis (p < 0.05) in pre-menopausal women (Supplemental Table 4), whereas 57,442 CpG sites were nominally associated (p < 0.05) in post-menopausal women (Supplemental Table 5). However, only a total of 1838 identical CpG sites were nominally associated with endometriosis (p < 0.05) in both pre- and post-menopausal women, which is less than expected by chance (OR = 0.84, 95% CI = 0.80–0.89, p = 2.00e-11; Fig. 2).

Fig. 2.

When the association between endometriosis and DNA methylation was analyzed separately in pre- (n = 163) and post-menopausal women, a total of 30,315 CpG sites were nominally associated with endometriosis in pre-menopausal women (p < 0.05), whereas 57,442 CpG sites were nominally associated in postmenopausal women (p < 0.05). Only a total of 1,838 identical CpG sites were nominally associated with endometriosis in both pre- and post-menopausal women (p < 0.05), which is less than expected by chance (OR = 0.84, 95% CI = 0.80–0.89, p = 2.00e-11)

DNA methylation and PBB

As this was a PBB-exposed population, PBB was examined as a covariate to assess the degree to which PBB-associated CpG sites, reported by Curtis et al. [42], shared DNA methylation patterns with endometriosis in this same cohort of women. We found that a total of 4893 (12.3%) of the above 39,877 CpG sites were associated with both endometriosis and PBB (OR = 0.97, 95% CI = 0.93–0.99, p = 0.03) and that the effect sizes were negatively correlated (r = − 0.19, p < 2.2e-16, Fig. 3). When controlling for PBB level, the DNA methylation results did not change substantially in our subset of women with endometriosis (r = 0.98, p < 2.2e-16).

Fig. 3.

Effect size of shared nominally significant CpGs with PBB and endometriosis. In analyzing the degree to which PBB-associated CpGs [42] shared DNA methylation patterns with endometriosis in this same cohort of women, we found that a total of 4,893 (12.3%) of the 39,877 CpGs were associated with both endometriosis and PBB (OR = 0.97, 95% CI = 0.93–0.99, p = 0.03) and that the effect sizes were negatively correlated (r = − 0.19, p < 2.2e-16, Fig. 3). When controlling for PBB level, the DNA methylation results did not change substantially in our subset of women with endometriosis (r = 0.98, p < 2.2e-16)

Replication of endometriosis-associated CpGs in the literature

We identified 42,261 CpGs reported in the literature that were previously found to be associated with endometriosis [47–49]. Of those, 38,747 (92%) CpGs were present on our EPIC array and passed quality control (QC) and could be assessed in our cohort. We found 2012 of the 38,747 CpG sites to be nominally associated (p < 0.05) in our cohort as well, though none remained significant after correction for multiple comparisons (FDR < 0.05; Supplemental Table 6).

Discussion

To our knowledge, this is the first study exploring epigenetic differences associated with endometriosis in a population highly exposed to PBB. Our findings suggest that while some CpG sites were nominally associated with endometriosis, no significant CpG sites were identified by investigation of DNA methylation data obtained via peripheral blood. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment of the nominally significant endometriosis CpGs was significant for 68 biologic pathways important for endocrine, immune, metabolic, and oncologic functions (FDR < 0.05). We found that PBB levels were slightly higher in women with endometriosis; however, they did not reach the threshold of statistical significance when compared to women without endometriosis. There was also no statistically significant difference when controlling for estradiol levels, nor were differences detected when controlling for menopausal status at time of sample collection. Differences in DNA methylation patterns did persist in post-menopausal women with endometriosis. Women with endometriosis were also noted to have higher incidences of self-reported early menopause, miscarriage, and fertility problems in this cohort.

Epigenetic mechanisms, which lead to modifications of gene activity not accompanied by changes in gene sequence, have been recognized as important factors in the development of a broad range of human diseases including endometriosis [20, 60, 61]. Several studies have identified differences in DNA methylation profiles in the endometrium of women with endometriosis [62–64]. These DNA methylation patterns subsequently lead to altered gene expression in cell proliferation, inflammation/immune response, angiogenesis, and steroid hormone response [60, 65, 66]. Altered expression/production of transcription factors, growth factors, cell cycle regulators, and cytokines/chemokines can subsequently influence specific target genes and proteins that play central roles in the development of endometriosis (i.e., reduced expression of decidualization genes) [64, 67]. For example, several genes whose products are critical for normal endometrial response have been found to have aberrant methylation in women with endometriosis, including HOXA10, FOXO1, COX-2, PGRB, estrogen receptor 2 (ESR2), steroidogenic-factor-1 (SF-1), GREB1, VEZT, and WNT4 [60, 65, 68–78].

In our study, we found 39,877 differentially methylated CpGs that were nominally associated with endometriosis, of which, 4893 associated with both endometriosis and PBB (12.3%; p < 0.05) [42]. PBB was included as a covariate primarily because this was an exposed population. Our findings are consistent with a prior study demonstrating no association between PBB and endometriosis [46]. Therefore, this variable is unlikely to be affecting the results of our epigenome-wide association study (EWAS). It also verifies that our findings are independent of PBB and truly reflective of endometriosis. While no CpG sites were statistically significant in our population of women with endometriosis, we did find that out of the 39,877 CpGs nominally associated with endometriosis, several CpGs were associated with the HOXA10, FOXO1, COX-2, ESR2, ESR1, GREB1, VEZT, and the WNT4 genes (all CpGs p < 0.05), consistent with findings from prior studies. In addition, one of the top 10 nominally associated CpGs in our population of women with endometriosis (cg00454719) is associated with the SMAD9 gene, which is affected by progesterone and may affect endometrial receptivity [79–81].

We also found enrichment for 68 KEGG pathways among our endometriosis CpGs (FDR < 0.05). These pathways were involved with several different cancers (including endometrial and breast), as well as with alterations of cell-cycle regulation, signaling, transcriptional regulation, and cellular metabolism. Pathways involving genes critical to proper reproductive system, embryologic development, and function (FoxO, Wnt, and Hedgehog signaling) were also among the significant pathways. Given that women with endometriosis have a higher risk of developing different cancers, these findings are consistent and meaningful as they may offer deeper understanding to steps that may be involved in this pathophysiologic process. Other significant pathways involved key steps or components in endometrial cyclic remodeling including pathways involved with endocytosis, adherens junctions, focal adhesions, and EGFR tyrosine kinase inhibitors. In addition, pathways involved with type II diabetes mellitus, calcium, insulin, fatty acid, and other metabolic pathways were also significant in our study population. These findings are meaningful given the increased risk for women with endometriosis to develop cardiovascular disease [12]; [13]. This may serve in part to explain why these health risks persist throughout the woman’s lifetime, as opposed to exclusively in pre-menopausal years. Furthermore, these relevant and significant pathways provide two additional layers of support that the CpG results found in our cohort are not by chance alone and are independently associated with endometriosis.

It is now understood that both estrogen and progesterone signaling pathways, receptor expression, and ratio are altered in women with endometriosis [64, 82, 83]. In particular, substantial epigenetic dysregulation of decidualization susceptibility genes leading to insufficient decidualization enhanced estrogenic activity, and reduced progesterone responsiveness have been found [63, 64, 83]. Alterations in DNA methylation in the endometrium have been identified throughout the menstrual cycle in small studies, with greatest variability in the late secretory and early follicular phase [50, 51, 65]. In attempt to account for this, we performed a sub-analysis controlling for estradiol level at time of blood draw as well as menopausal status. No significant differences were noted between the two groups when controlling for imputed estradiol levels. Interestingly, a total of 30,315 CpG sites were nominally significant (p < 0.05) in pre-menopausal women, whereas 57,442 CpG sites were nominally significant in post-menopausal women. This is reasonable to expect given the wide changes that have been reported across a menstrual cycle at times of differing hormonal dominance [84]. To our knowledge, our study is the first to include investigation of DNA-methylation in post-menopausal women with history of endometriosis.

More than 40,000 CpGs have been identified and reported in the literature as differentially methylated in endometriosis via the direct study of endometrial or ovarian tissue. We sought to further explore whether any correlation was noted in DNA methylation findings given the difference in measurement medias (tissue vs. peripheral blood samples), by comparing our findings to reported CpG sites associated with endometriosis in the literature [47–49]. Interestingly, when using a significance level of p < 0.05, there were 2012 CpG sites that were similarly significant in our cohort. Our findings are important as, in contrast to one’s genotype, epigenetic DNA methylation is more variable and is affected by many factors. DNA methylation, in particular, is cell-type specific [85]. Thus, investigating peripheral venous blood and correlating the findings in the context of known endometrial or ovarian DNA methylation changes with endometriosis has to be done with caution. While we did control for cell types and found no difference in women with or without endometriosis, our findings may have only reached nominal significance due to the predominant involvement of a very specific cell type in endometriosis (endometrial cells), as opposed to greater systemic DNA methylation changes.

Strengths

This study lends valuable insight as identification of more specific causes of endometriosis and its pathophysiologic sequelae are actively being sought in order to develop more targeted treatment modalities. The genetic and epigenetic theory of endometriosis has been increasingly investigated, and the ability to analyze such extensive genome sequences via a less invasive manner (peripheral blood sampling vs. endometriotic tissue sampling) is very valuable. Our analysis of peripheral blood also was performed on the largest and most comprehensive array, unique from other published studies in the literature. Data from blood samples may be effective for diagnosis or for performing larger population-based studies in the future. Blood may also better reflect the immune and systemic effects of endometriosis which may not be fully accounted for in direct tissue studies. Our findings do lend themselves to support, at least in part, an epigenetic role in the development of endometriosis; whether PBB exposure increased these individuals’ susceptibility to endometriosis, increased the severity of their pathology, or neither is impossible to determine with the current understanding of the disease. Lastly, in regard to PBB, this study population is the largest direct exposure group to our knowledge.

Weaknesses

One limitation of our study includes having only imputed systemic levels of estradiol as controls. Endometriosis has a well-documented increased local production of estradiol which may not be reflected systemically [5, 19]. Prior studies investigating DNA methylation in women with endometriosis, however, standardized their data to the subject’s menstrual cycle phase. This data was not available in our cohort as intake questions focused on lifetime history of endometriosis instead of capturing the subject in a time when they were actively experiencing symptoms or complications as a result of their endometriosis. For this reason, we sought to control for estradiol level at time of intake; however, we acknowledge that this may not be the most accurate measure given the absence of cycle data in our cohort. We also acknowledge that our replication analysis involved the comparison of DNA methylation obtained in two different modalities (blood vs. endometrial tissue). Smoking status and body mass index (BMI) could not reliably be included in our models due to missing self-reported data. Given the limited cultural diversity and distinct PBB exposure episode in our sample population, results may also not be applicable to the general population. Small sample size, recall bias or inaccurate self-report, and inability to confirm endometriosis and other diagnoses with medical documentation are further limitations to our study.

Conclusion

The significance of this study population cannot be understated as it presents a rare opportunity to study susceptibility and to uncover the role of these factors in PBB-associated health effects. Because of their similarities in structure, function, and toxicity, PBBs are also an excellent model for persistent organic chemicals that are currently in use. To our knowledge, our study was the first to explore epigenetic differences associated with endometriosis in the context of PBB exposure, in particular via peripheral blood sampling. No significant CpG sites were identified, and no significant overlap was noted in comparing our findings in women with endometriosis to women with PBB exposure suggesting that PBB and endometriosis have distinct pathways. Overlap was also noted when comparing our findings to CpGs reported in the literature, suggesting that while tissue data is the most specific means of measuring DNA methylation, there may be a role for peripheral blood analysis. KEGG pathway enrichment of the nominally significant endometriosis CpG sites was significant for 68 biological processes including pathways involved with various cancers and oncologic processes, aberrant cell cycle regulation, and metabolic processes consistent with the literature of women with endometriosis having an increased susceptibility to cardiovascular disease and certain cancers. While we found a higher incidence of endometriosis in women with PBB than previously reported, taken together, our findings are not suggestive of causal relationship between PBB exposure and endometriosis.

Funding information

This project was financed by the following grants: National Institutes of Environmental Health Sciences (5RO1ES024790, 5RO1ES025775, R24ES028528, 5P30ES019776) and the National Institute of General Medicine Sciences (T32GM008490).

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.