DDT Exposure during Pregnancy and DNA Methylation Alterations in Female Offspring in the Child Health and Development Study

Abstract

Studies measuring dichlorodiphenyltrichloroethane (DDT) exposure during key windows susceptibility including the intrauterine period suggest that DDT exposure is associated with breast cancer risk. We hypothesized that prenatal DDT exposure is associated with DNA methylation. Using prospective data from 316 daughters in the Child Health and Development Study, we examined the association between prenatal exposure to DDTs and DNA methylation in blood collected in midlife (mean age: 49 years). To identify differentially methylated regions (DMRs) associated with markers of DDTs (p,p’-DDT and the primary metabolite of p,p’-DDT, p,p’-DDE, and o,p’-DDT, the primary constituents of technical DDT), we measured methylation in 30 genes important to breast cancer. We observed DDT DMRs in three genes, CCDC85A, CYP1A1 and ZFPM2, each of which has been previously implicated in pubertal development and breast cancer susceptibility. These findings suggest prenatal DDT exposure may have life-long consequence through alteration in genes relevant to breast cancer.

Introduction

The pesticide dichlorodiphenyl-trichloroethane (p,p’-DDT, referred to generically as the pesticide, DDT), its primary metabolite (dichlorodiphenyldichloroethylene, p,p’-DDE) and an isomer that was a contaminant of commercial DDT (orthoparaDDT or o,p’-DDT) are endocrine disrupting chemicals (EDCs).(1) DDT was used widely in the U. S. beginning in 1945, peaking in 1959 and was banned in 1972.(2) Evidence from both in vivo and in vitro assays demonstrates that DDT can impact estrogen signaling through the ligand binding domain of ERα and ERβ.(3–8) DDT and o,p´-DDT, the most estrogenic components, also support the growth of breast tumors in animal models.(9, 10) Thus, DDT exposure may affect breast cancer risk both as a direct carcinogen and through alterations in hormonal signaling.(10, 11)

The mammary gland undergoes several developmental changes during the intrauterine period, puberty, childbearing and menopause which may be more susceptible windows for exogenous or endogenous carcinogenic influences.(12–15) In a prospective study of pregnant women in the Child Health and Development Studies (CHDS), Cohn and colleagues reported that high levels of serum p,p’-DDT were associated with a 5-fold increase in risk of breast cancer in women exposed to DDT prior to age 14 years, while there was no association in women who were not exposed before age 14.(16) Furthermore, CHDS daughters who had high in utero exposures to o,p’-DDT had a nearly four-fold increased risk of breast cancer after 54-years of follow-up.(17) These findings support the hypothesis that environmental exposures during windows of susceptibility when the breast is developing may increase the risk of breast cancer.(18) Women highly exposed to DDT in utero include women born in the 1960s, before DDT was banned. These women are just now approaching the age of increased risk for breast cancer. Thus the potential association of prenatal DDT exposure with breast cancer is relevant to the search for midlife biomarkers of risk.

Epigenetic biomarkers, such as DNA methylation, can change the activity of a DNA segment without changing the underlying DNA sequence and are essential to developmental processes and genomic imprinting.(19–21) White blood cells (WBCs) are common and readily accessible sources of DNA to determine methylation differences related to exposures.(22–25) Measuring methylation in repetitive elements in WBC DNA, we previously found overall methylation levels were lower in women with breast cancer compared with their unaffected sister controls.(26, 27) Increasing evidence also suggests that many breast cancer susceptibility genes may be altered through epigenetic alterations. For example, silencing of BRCA1 by methylation has been observed in breast cancer tissues,(28–31) and in WBCs.(32–34)

Fetal development is a critical time period when most of the epigenetic landscape is established,(21) and environmental exposures during this lifestage in addition to other vulnerable windows of susceptibility may alter epigenetic alternations.(35–37) For example, many epigenome-wide association studies reported maternal cigarette smoking during pregnancy was associated with altered DNA methylation in specific CpG sites in infants, children and adolescents.(38–42) Using information from the New York Women’s Birth Cohort, we previously reported that there are persistent DNA methylation changes in these specific CpG sites in midlife associated with prenatal smoking exposure.(43) These observations suggest that maternal smoking during pregnancy is associated with offspring DNA methylation in WBC across the lifecourse. Animal studies have demonstrated that EDCs may affect long-term health outcomes through epigenetic mechanisms (reviewed in (44)). Some limited evidence exists for DDT exposure and DNA methylation alterations in humans.(45–48)

In this report we test the hypothesis that DDT-associated changes in DNA methylation at midlife could account for DDT associations with breast cancer risk we observed in the Child Health and Development Studies cohort (CHDS). We investigated this hypothesis by examining the relation of in utero DDT exposure to methylation of breast cancer-associated genes at midlife in CHDS daughters.

Materials and methods

Study participants

Female offspring who were born into the Oakland, California based CHDS pregnancy cohort from 1959–1967 and who participated in the “Three Generations of Breast Cancer (3Gs)” or the Health Disparities Study (DISPAR), in adulthood at ages 44–54, between 2010 and 2013, formed the basis for this study. Details of the 3Gs and DISPAR study recruitment and response are described elsewhere.(49, 50) In order to be eligible for the current study, Prenatal Environmental Determinants of Intergenerational Risk (PEDIGREE), participants were further required to complete a home visit and/or provide a bio-specimen (saliva or blood sample) and to have an existing or pending mammogram within a year of recruitment. Using these criteria, two groups were invited to participate: daughters whose mothers had been diagnosed with breast cancer (n=231) and daughters whose mothers had not been known to have a breast cancer diagnosis as of the time of recruitment (n=281). Authorization to collect mammography was received from 491 (96%) and mammograms were successfully collected for 397 (81%). For this study, WBC DNA with consent was available for 335 (84%) participants. We further required available data on all study variables (organochlorine measures, age, race/ethnicity and body mass index) to achieve our final analysis sample (n=316). The study was approved by the Institutional Review Boards of Columbia University and the Public Health Institute (Oakland, California).

Exposure constructs.

We measured DDT using maternal serum samples collected from each mother 1–3 days after she gave birth.(16) Specifically, we measured p,p´-DDT, the active ingredient of DDT; o,p´-DDT, a low concentration contaminant; and p,p´-DDE, the most abundant p,p´-DDT metabolite (for details see(16, 51)).

Blood-based Biomarkers in Midlife.

We profiled DNA methylation levels in 30 candidate genes selected based on published genetic or epigenetic association with at least one or more of the following categories: (i) genes that are associated with breast cancer identified in genome-wide association studies (GWAS), or mutations related to breast cancer risk,(52, 53) (ii) genes related to age at menarche from GWAS,(54) (iii) genes related to growth and development from GWAS(55) or EWAS(56) (iv) genes involved in DNA recombination and repair (http://sciencepark.mdanderson.org/labs/wood/dna_repair_genes.html), and (v) selected candidate CpG sites from a previous study of adolescents that showed evidence of methylation changes between girls with and without a breast cancer family history in our exploratory genome-wide DNA methylation profiling from pilot work in 48 girls,(57) but outside the above selection criteria (Supplementary Table 1). For each locus, we selected the bis-seq primer locations based on the chromatin states defined by Ernst et al.(58) and available in the UCSC human genome browser.(59) We focused on active promoters and enhancers, insulators, and poised chromatin, since these regions are implicated in gene regulation in cancer (60–62) and are often enriched in disease‐associated DMRs.(63–65) In addition, we covered a few loci with repressed chromatin or transcription associated chromatin states. When indicated, annotation of GWAS peaks was performed using the NHGRI-EBI catalog (53) and literature searches for smaller-scale genetic association studies.

DNA extraction and bisulfite treatment

We extracted genomic DNA from whole blood samples by a salting out procedure; lysing cells with SDS in a nuclei lysis buffer and treating with RNase A (final 133 μg/mL) and RNase T1 (final 20 units/mL) to remove RNA. We coprecipitated proteins with NaCl (330 μL of saturated NaCl added per 1mL solution) by centrifugation and recovered genomic DNA from the supernatant by precipitation with 100% ethanol, washed it in 70% ethanol, and dissolved the DNA in the Tris-EDTA buffer. We bisulfite-converted genomic DNA (500 ng) using the EpiTect Bisulfite Kit (Qiagen, Germantown, MD), as per the manufacturer’s instructions and resuspended the DNA in 20uL of distilled water with storage at −20°C. We performed all DNA methylation assays blinded to DDT exposure data.

Targeted bisulfite DNA sequencing

We examined DNA methylation at the 30 selected loci using targeted massive parallel bisseq on genomic DNA from the WBC samples, as described.(66) Oligonucleotide primers were designed in MethPrimer(67) (Supplemental Table 1). Bisulfite-converted DNA was amplified and barcoded by PCR on a Fluidigm AccessArray high throughput PCR machine, followed by Nextgen (Illumina MiSeq) sequencing, as described.(66) After trimming for adaptors, low-quality bases (Phred score<30) and sequenced DNA fragments (reads) with a length <40 bp with TrimGalore (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/), the reads were aligned to the human genome (GRCh37 build) using Bismark aligner.(68) We used the Bismark extractor for methylation calling and since low coverage and low PCR complexity can lead to less representative estimates of DNA methylation levels, we filtered out values where the coverage was less than 100 reads and the complexity for each amplicon (estimated by the number of DNA methylation patterns represented at least 10 times) was less than the median complexity score of the amplicon across samples +/−1 standard deviation (SD). To rule out technical outliers due to low PCR complexity, which can occur due to amplification of limiting amounts of staring DNA, we applied a filtering algorithm based on the complexity of the sequenced DNA fragments. The complexity score was estimated by the number of DNA methylation patterns represented at least 10 times. Values where the complexity score was less than the mean score of the given amplicon across samples minus 1 SD were filtered out. Post complexity QC filtering showed lower but still high variance in methylation level. Therefore, all analyses presented in this paper have been performed on the filtered data. We used Bismark software to determine the percentage of methylation at each CpGs and then calculated the mean percentage methylation in each amplicon/gene by averaging the percentages of methylation across all CpGs in the amplicon/gene.

Statistical methods

We calculated the percentage of methylation at each CpG position by dividing the total number of methylated reads by the total number of reads (sum of methylated and unmethylated reads). While methylation differences at single CpG sites might have biological relevance, alteration of regulatory DNA elements often induces methylation changes affecting multiple contiguous CpGs spanning up to several kb of DNA. Each amplicon/gene contained a different number of CpG sites (Supplemental Table 1). We averaged percentages of methylation at each CpG across all CpGs in the amplicon/gene and use that value as an indication of percentage of methylation for each amplicon/gene for further data analysis. To assess the association between DNA methylation and in utero DDT exposure, we compared percentage of methylation for each amplicon/gene across different levels of markers of DDT exposure in maternal serum using the Kruskal-Wallis test. The rejection of the null hypothesis concludes that there is no difference in percentage of DNA methylation of the gene among the levels of DDT exposure. We divided each daughter participant into different DDT exposure groups using the cutpoint values of each markers of DDTs from our previous study:(17) for p,p´-DDT (low exposure, < 8.09 μg/L; median exposure, 8.09–13.90 μg/L; high exposure, > 13.90 μg/L), o,p´-DDT (low exposure, ≤ 0.42 μg/L; median exposure, 0.43–0.72 μg/L; high exposure, > 0.72 μg/L) and p,p´-DDE (low exposure, ≤ 35.23 μg/L; median exposure, > 35.23–58.49 μg/L; high exposure, > 58.49 μg/L). We used these cuptoint values as our previous study showed an association of prenatal DDT exposure with breast cancer and thus might have biological relevance. Both age and race/ethnicity were associated with DDT exposure,(16, 49) and DNA methylation.(69, 70) For each DMR, we carried out a multivariable analysis adjusting for age and race/ethnicity using linear regression models with percentage of DNA methylation as the outcome, and categorical variables for markers of DDT exposure. We also tested interaction with family history of breast cancer. All analyses were performed with SAS software 9.4 (SAS Institute, Cary, NC) and R.3.10.

Results

Table 1 presents the distributions of selected characteristics and levels of prenatal DDT exposure of participants. The mean age of daughter participants when blood was drawn was 49.3 years (SD=2.0). The mean levels of prenatal DDT exposure were 12.4 (SD=7.7, μg/L) for p,p’-DDT, 47.0 (SD=20.4, μg/L) for p,p’-DDE and 0.51 (SD=0.44, μg/L) for o,p’-DDT. Overall, the percentages of daughter participants with high prenatal DDT exposure were 30.4% for p,p’-DDT, 21.8% for p,p’-DDE and 21.8% for o,p’-DDT. Supplemental Table 2 present the spearman correlation of prenatal DDT exposure with age and race/ethnicity. Consist with previous studies, (16, 49), prenatal DDT exposure were associated with age and race/ethnicity.

Table 1.

Demographic characteristics and in utero DDT exposure of daughters in the Child Health and Development Studies Pregnancy Cohort (N=316)

Variable	Mean, SD
Age	49.3 (2.0)
BMI	28.0 (7.3)
p-p’-DDT, μg/L	12.4 (7.7)
p-p’-DDE, μg/L	47.0 (20.4)
o-p’-DDT, μg/L	0.51 (0.44)
Race/Ethnicity of mother	No, %
White	226 (71.5)
Black	49 (15.5)
Hispanic	12 (3.8)
Asian	14 (4.4)
Other	15 (4.8)
p-p’-DDT
Low (<8.09, μg/L)	112 (35.4)
Median (8.09–13.90, μg/L)	108 (34.2)
High (>13.90, μg/L)	96 (30.4)
p-p’-DDE
Low (≤35.23, μg/L)	108 (34.2)
Median (>35.23–58.49, μg/L)	139 (44.0)
High (>58.49, μg/L)	69 (21.8)
o-p’-DDT
Low (≤0.42, μg/L)	166 (52.5)
Median (0.43–0.72, μg/L)	81 (25.6)
High (>0.72, μg/L)	69 (21.8)
Mother has breast cancer
No	173 (54.8)
Yes	143 (45.2)

SD: standard deviation

The distribution of methylation in the 30 genes by prenatal DDT exposure are summarized in Tables 2–4. Using the Kruskal-Wallis test to compared values of percentage of methylation across different DDTs exposure groups without adjusting any covariate, we observed higher prenatal p,p’-DDT exposure was associated with higher methylation in IGF1 and ZFPM2. Methylation in IGF1 was 3.2% higher in daughter participants with higher p,p’-DDT than participants with lower exposure (81.1±8.4%, 83.8±4.9%, and 84.2±5.8% for <8.09, 8.09–13.90 and >13.90, μg/L of p,p’-DDT, p=0.03). The median of percentage of methylation levels in ZFPM2 were 44.1± 11.2%, 42.5 ±12.5% and 46.6 ±11.0% (p=0.009) for <8.09, 8.09–13.90 and >13.90, μg/L of p,p’-DDT (Table 2 and Figure 1). Higher methylation in CCDC85A, and CYP1A1 were associated with higher prenatal p,p’-DDE exposure; methylation in CCDC85A, and CYP1A1 were 1.7%, and 5.0% higher in daughter participants with the highest p,p’-DDE than participants with lowest exposure (Table 3 and Figure 2). Methylation in CCDC85A, RECQL4 and ZFPM2 were associated with intrauterine o,p’-DDT exposure (Table 4 and Figure 3). The difference in methylation between highest and lowest o,p’-DDT groups ranged from 0.9% for RECQL4 to 5.6% for ZFPM2.

Table 2.

Distribution of methylation levels in 30 genes by p-p’-DDT exposure of daughters in the Child Health and Development Studies Pregnancy Cohort

	p-p’-DDT <8.09, μg/L		p-p’-DDT: 8.09–13.90, μg/L		p-p’-DDT >13.90, μg/L
Gene	Median, %	IQR	Median, %	IQR	Median, %	IQR	P.value
ADRB1	24.4	9.0	24.5	9.4	24.3	8.7	0.76
ARHGEF7	44.7	10.3	46.5	8.3	44.4	9.4	0.61
BRCA1	81.2	11.2	82.4	12.5	83.1	11.0	0.47
CCDC85A	34.7	7.7	34.5	9.9	35.7	8.2	0.57
CCNL1	0.2	0.6	0.5	1.1	0.2	0.5	0.05
CDH1	7.4	2.1	7.1	2.4	7.7	2.9	0.07
CELF4	56.6	10.7	55.6	11.9	54.7	12.3	0.60
chr11	7.7	4.1	8.2	4.3	8.4	5.0	0.26
chr12	78.5	4.9	78.3	5.4	78.4	4.8	0.60
chr16	81.7	5.4	82.3	6.2	81.4	5.8	0.73
chr4	81.9	8.6	81.2	8.9	81.8	9.0	0.44
CYP1A1	51.2	20.4	50.8	17.5	53.1	16.2	0.09
DLGAP2	97.1	1.1	97.1	1.2	96.9	1.1	0.24
ESR1	54.2	20.3	55.1	19.7	55.2	21.3	0.69
GAB2	96.8	2.6	97.2	2.8	97.0	3.5	0.30
GNA12	38.4	17.9	35.0	25.9	38.8	23.5	0.47
IGF1	81.1	8.4	83.8	4.9	84.2	5.8	0.03
MCHR2	13.3	3.8	13.3	3.5	12.9	4.4	0.85
OBSCN	92.4	1.8	92.8	1.9	92.4	2.0	0.09
PCDHGB1	39.5	7.8	41.3	7.7	39.7	8.3	0.15
PEX14	66.7	7.4	66.2	6.2	65.8	8.2	0.67
RAD51L1	98.2	2.0	98.2	1.2	98.0	1.5	0.44
RECQL4	86.7	4.6	87.2	4.3	87.5	3.8	0.25
SLC39A14	87.3	10.1	87.2	9.8	87.3	12.1	0.63
SLC6A3	73.8	4.7	73.8	6.5	74.6	5.2	0.41
TCF7L2	52.4	14.2	49.9	16.2	51.4	15.3	0.44
TERT	96.6	3.0	96.9	2.4	96.7	2.9	0.56
XRCC3	79.7	3.3	79.5	2.8	79.9	4.1	0.86
ZFPM2	44.1	11.2	42.5	12.5	46.6	11.0	0.0090
ZNF483	35.7	4.3	36.4	3.1	35.4	3.2	0.56

p value of Kruskal-Wallis test

IQR: The interquartile range

Table 4.

Distribution of methylation levels in 30 genes by o,p’-DDT exposure of daughters in the Child Health and Development Studies Pregnancy Cohort

	o,p’-DDT ≤0.42, μg/L		o,p’-DDT: 0.43–0.72, μg/L		o,p’-DDT >0.72, μg/L
Gene	Median	IQR	Median	IQR	Median	IQR	P.value
ADRB1	24.3	8.9	24.8	8.6	25.1	11.6	0.43
ARHGEF7	45.5	8.6	44.4	9.3	44.6	9.2	0.90
BRCA1	81.9	12.1	81.2	12.8	82.3	10.5	0.97
CCDC85A	34.3	8.0	34.8	8.0	36.4	8.5	0.01
CCNL1	0.3	0.8	0.3	0.7	0.3	0.6	0.90
CDH1	7.4	2.2	7.1	3.0	7.9	2.9	0.05
CELF4	56.5	10.0	56.3	12.6	55.0	12.4	0.40
chr11	8.0	4.4	8.2	4.7	7.5	5.8	0.90
chr12	78.6	5.0	78.1	5.2	78.1	4.9	0.95
chr16	81.6	5.5	82.3	6.2	81.9	5.8	0.17
chr4	81.4	8.6	81.5	9.6	82.1	8.3	0.88
CYP1A1	50.3	19.7	51.9	16.4	54.0	14.5	0.05
DLGAP2	97.1	1.0	97.1	1.2	97.0	1.3	0.75
ESR1	55.0	19.2	54.9	22.1	55.2	21.1	0.98
GAB2	97.0	2.7	96.9	2.7	97.1	3.5	0.96
GNA12	36.0	20.8	37.8	23.8	42.4	26.8	0.35
IGF1	83.2	5.5	83.4	7.2	83.1	9.5	0.92
MCHR2	13.2	3.5	12.7	4.5	13.4	3.9	0.37
OBSCN	92.7	1.8	92.5	1.7	91.9	2.1	0.05
PCDHGB1	40.1	7.7	40.2	7.9	40.5	9.4	0.66
PEX14	65.6	7.2	67.8	7.5	66.5	8.1	0.37
RAD51L1	98.1	1.8	98.3	1.3	98.1	1.8	0.33
RECQL4	86.6	4.6	87.8	4.0	87.5	3.9	0.03
SLC39A14	87.2	8.6	87.6	10.1	87.2	15.9	0.45
SLC6A3	73.5	5.2	74.5	6.2	74.5	5.1	0.13
TCF7L2	52.5	13.0	49.6	15.8	51.4	16.2	0.28
TERT	96.7	2.8	96.6	2.8	96.8	2.8	0.96
XRCC3	79.7	3.2	79.7	3.7	79.7	4.5	0.81
ZFPM2	43.3	11.2	44.6	11.9	48.9	13.5	0.01
ZNF483	36.1	3.6	35.5	3.1	35.9	4.0	0.23

p value of Kruskal-Wallis test

IQR: The interquartile range

Figure 1. Methylation differences by p,p’-DDT in IGF1, and ZFPM2.

The boxplot displays the distribution of CpG methylation values (percent) by different p,p’-DDT exposure group (low exposure, < 8.09 μg/L; median exposure, 8.09–13.90 μg/L; high exposure, > 13.90 μg/L) in 2 gene regions (IGF1 and ZFPM2) examined. The middle bold line represents the median methylation. Each dot represents the methylation value of each samples. The distribution of percentage of IGF1 methylation (left) by each p,p’-DDT exposure group: low exposure (median: 81.1%, Q1-Q3=(75.8%, 84.3%), Min-Max (68.2%, 93.8%)); median exposure (median:83.8%, Q1-Q3=(81.0%, 85.9%), Min-Max=(70.6%, 92.6%)); high exposure (median: 84.2%, Q1-Q3=(80.9%, 86.7%), Min-Max=(56.0%, 91.1%)). The distribution of percentage of ZFPM2 methylation (right) by each p,p’-DDT exposure group: low exposure (median: 44.1%, Q1-Q3=(38.4%, 49.6%), Min-Max= (17.4%, 82.9%); median exposure (median: 42.5%, Q1-Q3=(36.1%, 48.6%), Min-Max=(16.9%, 69.6%)); high exposure (median:46.6%, Q1-Q3=(41.1%, 52.1%), Min-Max=(26.9%, 75.3%)).

Table 3.

Distribution of methylation levels in 30 genes by p,p’-DDE exposure of daughters in the Child Health and Development Studies Pregnancy Cohort

	p,p’-DDE ≤35.23, μg/L		p,p’-DDE :35.23–58.49, μg/L		p,p’-DDE >58.49, μg/L
Gene	Median, %	IQR	Median, %	IQR	Median, %	IQR	P.value
ADRB1	23.5	9.3	24.3	8.6	26.5	9.8	0.14
ARHGEF7	44.0	9.5	45.9	8.9	44.5	9.1	0.29
BRCA1	80.7	14.0	82.4	10.2	82.9	11.3	0.30
CCDC85A	34.7	7.7	34.2	8.2	36.4	8.4	0.02
CCNL1	0.3	0.8	0.3	0.8	0.2	0.6	0.49
CDH1	7.3	2.0	7.3	2.6	7.7	3.0	0.19
CELF4	55.7	11.1	57.3	11.3	55.7	13.1	0.54
chr11	7.7	3.9	8.5	5.2	8.2	5.2	0.32
chr12	78.7	4.6	78.1	5.1	78.2	5.2	0.32
chr16	81.6	5.4	82.0	6.4	82.1	6.3	0.21
chr4	82.3	8.0	81.4	8.4	80.9	9.8	0.36
CYP1A1	50.0	20.1	51.8	17.3	55.0	19.2	0.03
DLGAP2	97.2	1.1	96.9	1.2	97.1	1.0	0.17
ESR1	55.1	19.6	55.0	20.6	56.0	23.7	0.88
GAB2	96.9	2.4	97.2	2.7	97.0	3.4	0.70
GNA12	37.9	21.1	35.7	21.4	39.7	32.4	0.63
IGF1	82.2	4.3	83.2	7.4	84.5	6.6	0.23
MCHR2	13.4	3.5	12.2	3.9	13.8	3.9	0.07
OBSCN	92.5	1.7	92.5	2.1	92.2	1.7	0.25
PCDHGB1	40.6	7.8	40.3	8.0	39.9	9.1	0.73
PEX14	67.0	7.1	65.8	7.5	67.5	8.9	0.25
RAD51L1	98.1	1.5	98.1	1.7	98.0	1.9	0.97
RECQL4	86.7	4.6	86.8	4.5	87.5	3.6	0.16
SLC39A14	87.3	9.7	87.3	10.0	87.2	10.9	0.25
SLC6A3	73.6	4.9	74.5	6.0	74.1	5.5	0.72
TCF7L2	52.3	13.5	50.3	13.9	54.5	16.6	0.23
TERT	96.8	2.7	96.6	2.9	97.1	2.5	0.50
XRCC3	79.6	3.6	79.9	3.2	79.7	3.3	0.94
ZFPM2	44.2	11.4	44.5	11.9	44.6	10.9	0.59
ZNF483	35.9	3.4	36.0	3.4	35.4	4.0	0.86

p value of Kruskal-Wallis test

IQR: The interquartile range

Figure 2. Methylation differences by p,p’-DDE in CCDC85A, and CYP1A1.

The boxplot displays the distribution of CpG methylation values (percent) by different p,p’-DDE exposure group (low exposure, ≤ 35.23 μg/L; median exposure, > 35.23–58.49 μg/L; high exposure, > 58.49 μg/L).) in 2 gene regions (CCDC85A and CYP1A1) examined. The middle bold line represents the median methylation. Each dot represents the methylation value of each samples. The distribution of percentage of CCDC85A methylation (left) by each p,p’-DDE exposure group: low exposure (median: 34.7%, Q1-Q3=(30.7%, 38.5%), Min-Max(22.5%, 48.3%)); median exposure (median:34.2%, Q1-Q3=(30.7.0%, 38.9%), Min-Max=(21.0%, 49.9%)); high exposure (median: 36.4%, Q1-Q3=(32.3%, 40.7%), Min-Max=(19.8%, 50.9%)). The distribution of percentage of CYP1A1 methylation (right) by each p,p’-DDE exposure group: low exposure (median: 50.0%, Q1-Q3=(38.1%, 58.2%), Min-Max= (16.8%, 76.1%); median exposure (median: 51.8%, Q1-Q3=(41.0%, 58.3%), Min-Max=(9.5%, 82.5%)); high exposure (median:55.0%, Q1-Q3=(45.4%, 64.6%), Min-Max=(20.2%, 79.6%)).

Figure 3. Methylation differences by o,p’-DDT in CCDC85A, RECQL4 and ZFPM2.

The boxplot displays the distribution of CpG methylation values (percent) by different o,p’-DDT exposure group (low exposure, ≤ 0.42 μg/L; median exposure, 0.43–0.72 μg/L; high exposure, > 0.72 μg/L) in 3 gene regions (CCDC85A, RECQL4 and ZFPM2) examined. The middle bold line represents the median methylation. Each dot represents the methylation value of each samples. The distribution of percentage of CCDC85A methylation (top left) by each o,p’-DDT exposure group: low exposure (median: 34.3%, Q1-Q3=(30.5%, 38.5%), Min-Max(21.0%, 48.3%)); median exposure (median:34.8%, Q1-Q3=(31.0%, 39.1%), Min-Max=(19.8%, 50.9%)); high exposure (median: 36.4%, Q1-Q3=(33.0%, 41.5%), Min-Max=(21.0%, 49.3%)). The distribution of percentage of RECQL4 methylation (top right) by each o,p’-DDT exposure group: low exposure (median: 86.7%, Q1-Q3=(84.4%, 89.0%), Min-Max= (79.4%, 95.3%); median exposure (median: 87.8%, Q1-Q3=(85.8%, 89.8%), Min-Max=(80.1%, 95.4%)); high exposure (median:87.5%, Q1-Q3=(85.5%, 89.4%), Min-Max=(80.9%, 94.3%)). The distribution of percentage of ZFPM2 methylation (down left) by each o,p’-DDT exposure group: low exposure (median: 43.3%, Q1-Q3=(37.8%, 49.0%), Min-Max(16.9%, 82.9%)); median exposure (median:44.6%, Q1-Q3=(37.8%, 49.7%), Min-Max=(17.4%, 67.8%)); high exposure (median: 48.9%, Q1-Q3=(41.0%, 54.5%), Min-Max=(25.7%, 75.3%)).

We present analyses adjusted for age and race/ethnicity in Table 5; the associations with CCDC85A, CYP1A1 and ZFPM2 and DDT remained. Higher p,p’-DDT, and p,p’-DDE were associated with 2.05% and 1.89 % higher CCDC85A methylation. Methylation in CYP1A1 was 3.68% (95%CI=−0.19, 7.56) higher in daughter participants with highest p,p’-DDT, 4.30% (95%CI=0.27,8.32) higher in daughter participants with the highest p,p’-DDE and 4.14% (95%CI=0.33, 7.94) higher daughter participants with the highest o,p’-DDT. Daughter participants with the highest p,p’-DDT had 3.79% (95%CI=0.76, 6.81) higher methylation in ZFPM2, and daughter participants with the highest o,p’-DDT had 4.31% (95%CI=1.33, 7.29) higher methylation. We did not see any significant interaction with a family history of breast cancer.

Table 5.

Multivariable Association between DNA Methylation and DDT exposure of daughters in the Child Health and Development Studies Pregnancy Cohort

		Median Exposure*		High Exposure*
Gene	DDT Markers	β Estimate	95% CL	β Estimate	95% CL
CCDC85A	p-p’-DDT	0.50	(−1.21, 2.21)	0.35	(−1.50, 2.20)
	p-p’-DDE	−0.46	(−2.00, 1.08)	2.05	(0.17, 3,94)
	o-p’-DDT	0.82	(−0.83, 2.46)	1.89	(0.08, 3.70)
CYP1A1	p-p’-DDT	0.12	(−3.48, 3.71)	3.68	(−0.19, 7.56)
	p-p’-DDE	0.78	(−2.48, 4.04)	4.30	(0.27, 8.32)
	o-p’-DDT	0.40	(−3.09, 3.89)	4.14	(0.33, 7.94)
IGF1	p-p’-DDT	2.80	(0.33, 5.26)	2.37	(−0.31, 5.05)
	p-p’-DDE	1.14	(−1.14, 3.42)	1.50	(−1.21, 4.22)
	o-p’-DDT	−0.65	(−3.14, 1.85)	0.04	(−2.55, 2.63)
RECQL4	p-p’-DDT	−0.22	(−1.09, 0.66)	0.53	(−0.41, 1.48)
	p-p’-DDE	−0.14	(−0.94, 0.66)	0.40	(−0.59, 1.38)
	o-p’-DDT	0.86	(0.01, 1.71)	0.72	(−0.21, 1.65)
ZFPM2	p-p’-DDT	−1.52	(−4.33, 1.29)	3.79	(0.76, 6.81)
	p-p’-DDE	1.21	(−1.40, 3.82)	0.62	(−2.64, 3.88)
	o-p’-DDT	−0.55	(−3.33, 2.23)	4.31	(1.33, 7.29)

Adjust for age and race/ethnicity

Cl: confidence interval

^*The cutpoint value of median exposure for p,p´-DDT (low exposure, < 8.09 μg/L; median exposure, 8.09–13.90 μg/L; high exposure, > 13.90 μg/L), o,p´-DDT (low exposure, ≤ 0.42 μg/L; median exposure, 0.43–0.72 μg/L; high exposure, > 0.72 μg/L) and p,p´-DDE (low exposure, ≤ 35.23 μg/L; median exposure, > 35.23–58.49 μg/L; high exposure, > 58.49 μg/L).

Discussion

Due to the estrogenic properties of DDT compounds,(3–8) a number of epidemiological studies have investigated DDT exposure in relation to breast cancer (Reviewed in(14)). Most evidence is based on DDT measured in blood specimens collected at midlife or later during time periods after DDT was banned. Evidence from these studies is weak and inconsistent.(16, 17, 71–80) The strongest evidence for DDT associations with breast cancer in human populations is based on the CHDS where DDT could be measured in blood specimens obtained from young women during active DDT use.(16) DDT levels measured at the time of cancer diagnosis were associated with stage/ aggressiveness of breast cancer.(81, 82) We observed the largest DDT associations with breast cancer for women exposed to DDT in utero or before puberty.(16, 17, 80) Here we investigated the possible contribution of DDT-related changes in DNA methylation of breast-cancer associated genes to breast cancer risk in the CHDS.

Increasing evidence suggests epigenetic effects of EDCs on human health.(45–48, 83, 84) We compared the methylation status of breast-cancer associated genes for daughter participants with different levels of prenatal DDT exposure to identify specific DMRs associated with breast cancer susceptibility. We observed three DDT DMRs located in genes that are involved in growth and development and breast cancer susceptibility, CCDC85A, CYP1A1 and ZFPM2, respectively.(85–89) All three DDT DMRs consistently showed higher methylation in daughter participants with higher prenatal DDT exposure measured by the primary constituents of commercial DDT; o,p´-DDT and p,p´-DDT, and p,p’-DDE, the primary metabolite of p,p´-DDT. Higher prenatal exposure to p,p´-DDT or o,p´-DDT was associated with higher methylation in the promoter regions of CCDC85A and ZFPM2, genes related to puberty development, and CYP1A1 which is a breast cancer susceptibility gene. Although the biological function of the DMRs in these regions is unclear, DNA methylation in promoter regions usually correlates inversely with gene expression.(90, 91) Methylation of the DDT associated-DMRs in ZFPM2, CCDC85A and CYP1A1 showed high inter-individual variability (Figures 1–3), suggesting the presence of methylation quantitative trait loci (mQTLs; a.k.a. meQTLs).

Animal and epidemiological studies implicate EDCs as a significant concern to public health.(92) EDCs can interfere with the endocrine system, resulting in adverse health outcomes.(93) Many animal studies have linked EDCs such as diethylstilbestrol (DES), polychlorinated biphenyl (PCBs) and DDT exposure to epigenetic modifications including DNA methylation changes, resulting in alteration in gene expression and chromosomal stability.(94–99) In particular, the mouse model had clearly established the ability of environmental factors to influence epigenetics thus promoting phenotypic changes later in development.(100, 101) Compared with controls, young rats exposed to DDT had lower methylation overall.(98) Rats treated in utero and postnatally with organochlorine pesticides and PCBs also showed decreased methylation in the tumor suppressor gene p16 (INK4a) compared to controls.(99) An in vitro study of rat ovarian cells observed that o,p’-DDT can suppress the expression of selected genes (such as cytochrome P450 side chain cleavage enzyme (P450scc), progesterone receptor (PR), and epidermal growth factor epiregulin (EREG)) in very low doses through an estrogen receptor-independent pathway.(102)

DNA methylation may play an important role in causing disease by silencing genes through hypermethylation or activating genes through hypomethylation.(103–106) Moreover, global decrease in 5-methylcytosine content might contribute to the reactivation of transposable elements and the generation of chromosomal instability.(45, 46, 107, 108) Many epidemiological studies of DNA methylation have focused on global DNA methylation measured as methylation value in repetitive elements such as LINE-1 and Alu or in CCGG sequences which is quantified by the LUMA assay as an indication of overall DNA methylation value in the sample. Studies focusing on identification of DMRs can measure epigenetic wide DNA methylation profile using sequence- and array-based technologies.(109) Advantages of the bis-seq method are the ability to examine methylation across multiple CpGs. Evidence related to EDCs and DNA methylation in humans is more limited. Using information from the Sister Study of 200 women, half of whose mothers retrospectively reported taking DES during pregnancy, Harlid et al.(110) examined the association of genome-wide DNA methylation with intrauterine DES exposure. Although they found 22 CpGs had nominal p values < 10⁻⁴, none achieved genome-wide significance after considering multiple comparisons (q<0.05).(110) The association of prenatal exposure to DDT and other organic pollutants measured in blood from mothers during pregnancy or at delivery with global DNA methylation markers including LINE1 and Alu in cord blood DNA was examined in a birth cohort of Mexican-American children. The cohort was established in 1999–2000 when DDT use was continuing in Mexico for Malaria control.(107) Higher prenatal DDT exposure was associated with lower Alu methylation at birth; with an β value (95%CI) of −0.37 (−0.69,−0.05) for o,p′ -DDT, and an β value (95%CI) of −0.33 (−0.64,−0.01) for p,p′-DDE.(107) They also examined prenatal DDT exposure and DNA methylation in 9-year old children, and found the same but nonsignificant trend of lower Alu methylation with prenatal polybrominated diphenyl ethers (PBDE) exposure.(107) In contrast, other studies examined global methylation in adults and found lower methylation in Alu was associated with high level of serum DDT.(45, 46) Our study found prenatal DDT exposure is associated with persistent change in DNA methylation of breast-cancer associated genes, in particular genes that are important to puberty development.

CYP1A1, the main cytochrome P450 enzymes, play an important role in the detoxification of environmental carcinogens.(111) Increasing evidence suggests that exposures including smoking can modify methylation in CYP1A1. Multiple CpG sites in the promoter regions of CYP1A1 have been associated with maternal smoking during prenatal exposure and some of these associations have extended to midlife.(43, 112, 113) Microarray analyses in human endometrial endothelial cells revealed that DDT affected biological processes such as the cell cycle, cell division, and lipid metabolism.(114) Wójtowicz et al. (115) studied the effects of DDT on gene expression in placental cells and found DDT/DDE inhibited the expression of CYP1A1 and AhR within 48 h after treatment. This study suggests that DDT exposure might affect the AhR/CYP1A1 signaling pathway. Studying the association between genetic polymorphisms and gene expression of CYP1A1 in breast tissue, Goth-Goldstein et al. (116) found there is a large variation in CYP1A1 expression in breast tissue and the variations could not explained by the variant genotype. Alteration in DNA methylation might be the underlying mechanism by which DDT regulates gene expression. Higher methylation in the enhancer of CYP1A1 was associated with lower mRNA levels.(117)

The breast undergoes many changes in early life and environmental factors may have a stronger effect on breast cancer risk during development and maturation of breast tissue.(78) The effect of DDT on breast cancer observed in the developmental stage of exposure such as prenatal and early life;(16, 17) contrasts with the lack of a clear association when DDT is measured in midlife may be due to this being outside of specific windows of susceptibility.(75, 76) Exposures during a crucial time of development can alter genome activity associated with the differentiation programming of cells or organ systems through epigenetic mechanism. Modification of the epigenome may continue throughout development, subsequently affecting the adult transcriptome and making tissues, such as the breast, susceptible to developing disease.(118)

A key strength of our study is that we prospectively examined the association between in utero exposure to DDT measured in maternal early postpartum serum with adult DNA methylation changes. However, DNA methylation was only estimated one time using blood collected at midlife in daughters. To better understand the long-term effect of prenatal DDT exposure on DNA methylation changes, repeated measurements of DNA methylation profiles across the lifecourse including infancy, childhood, adolescence, and adulthood are needed. Although we selected the list of candidate genes based on a priori considerations, if we were to conservatively divide the p value by the number of tests we conducted, our results may be due to chance. In this study, the DNA methylation profile was measured in DNA that was derived from whole blood which contains a mixture of different cells. As DNA methylation profile is cell-type specific,(119) we recognize that the abundance of specific cell subtypes which may have different levels of methylation in the genes of interest may impact our results. However, comparing methylation in these DDT associated-DMRs in ZFPM2, CCDC85A and CYP1A1 across different cell types using data from Illumina HumanMethylation450 BeadChips (GEO accession: GSE35069), we did not find any differences in methylation by cell type including WBC, peripheral blood mononuclear cells, monocytes and neutrophils,(120) suggesting the methylation in these three DMRs are not different by blood cell types.

The ability of prenatal DDT exposure to alter methylation in breast cancer genes has been suggested. A recent study using Illumina Infinium HumanMethylation 450 K BeadChips measured cord blood from 24 subjects and examined the association of prenatal DDT exposure with fetal genome-wide DNA methylation.(121) Comparing DNA methylation profiles between subjects with no detectable DDT and subjects with detectable DDT, the authors identified 1,131 CpG sites with differences≥5% associated with intrauterine DDT exposure. Most of these CpG sites were located in the open sea regions and only 22% DMRs were in CpG islands. These CpG sites included 690 hypermethylation sites and 441 hypomethylation sites. The authors further validated the association between methylation of BRCA1 and intrauterine DDT exposure in another group of 126 subjects and found both DNA methylation and gene expression were statistically significantly different between the different exposure groups of p,p′-DDE, o,p′-DDD, o,p′-DDT, and p,p′-DDT (P < 0.05).(121) Unfortunately, our BRCA1 amplicon is about 700 bp away from their amplicon so we were unable to replicate this result in our study.

Our prospective study examining the effect of prenatal DDT exposure on adult DNA methylation changes suggests the persistent effect of environmental exposure may be through epigenetic alteration. The observation of alteration in DNA methylation profiles in genes important to breast cancer such as genes associated with menarche suggests a potential molecular mechanism involved in prenatal DDT exposure in breast cancer development. If verified, DNA methylation patterns associated with intrauterine DDT exposure may become clinically useful biomarkers of risk in current populations of women in midlife who were heavily exposed to DDT in utero in the 1960’s. Verification should include investigation of biomarker relevance to future risk in animal, in vitro, and clinical studies and eventually investigation of the potential for mitigating risk via restoring methylation marks to lower risk patterns.

Highlights.

• Prenatal DDTs exposure is associated with DNA methylation in key genes that are potentially important to breast cancer.

• Differentially methylated regions (DMRs) in CCDC85A, CYP1A1 and ZFPM2 are associated with markers of DDTs

• Higher methylation in genes that are involved in growth and development and breast cancer susceptibility are associated with higher in utero DDT exposure

• Prenatal DDT exposure may have life-long consequence through alteration in genes relevant to breast cancer.

Abbreviations:

CHDS

Child Health and Development Studies

bis-seq

bisulfite sequencing

BMI

body mass index

DDT

dichlorodiphenyltrichloroethane

DMR

differentially methylated regions

EDCs

endocrine disrupting chemicals

o,p´-DDT

[1,1,1-trichloro-2-(p-chlorophenyl)-2-(o-chlorophenyl) ethane]

p,p´-DDE

[1,1´-dichloro-2,2´-bis(p-chlorophenyl) ethylene]

p,p´-DDT

[1,1,1-trichloro-2,2-bis(pchlorophenyl) ethane]

WBC

white blood cells