Differential DNA methylation in placenta associated with maternal blood pressure during pregnancy

Abstract

Abnormal blood pressure during pregnancy is associated with impaired fetal growth, predisposing the offspring to cardiometabolic abnormalities over the life-course. Placental DNA methylation may be the regulatory pathway through which maternal blood pressure influences fetal and adult health outcomes. Epigenome-wide association study of 301 participants with placenta sample examined associations between DNA methylation and mmHg increases in systolic and diastolic blood pressure in each trimester. Findings were further examined using gene expression, gene pathway and functional annotation analyses. CpGs known to be associated with cardiometabolic traits were evaluated. Increased maternal systolic and diastolic blood pressure were associated with methylation of 3 CpGs in the first, 6 CpGs in the second, and 15 CpGs in the third trimester at 5% FDR (P-values ranging from 6.6×10⁻¹⁵ to 2.3×10⁻⁷). Several CpGs were enriched in pathways including cardiovascular-metabolic development (P=1.0×10⁻⁴⁵). Increased systolic and diastolic blood pressure were associated with increased CpG methylation and gene expression at COL12A1, a collagen family gene known for regulatory functions in the heart. Out of 304 previously reported CpGs known to be associated with cardiometabolic traits, 36 placental CpGs were associated with systolic and diastolic blood pressure in our data. The present study provides the first evidence for associations between placental DNA methylation and increased maternal blood pressure during pregnancy at genes implicated in cardiometabolic diseases. Identification of blood pressure-associated methylated sites in the placenta may provide clues to early origins of cardiometabolic dysfunction and inform guidelines for early prevention.

Clinical Trial Registration:

(ClinicalTrials.gov, NCT00912132.

Introduction

Abnormal blood pressure during gestation is associated with impaired fetal growth,^{1, 2} and increases the lifetime risk of hypertension and cardiovascular diseases in the offspring.^{3,4, 5} Pregnancy-related hypertensive disorders including preeclampsia and gestational hypertension are associated with higher blood pressure in childhood⁶ and may result in persistent dysfunction in systemic and pulmonary circulation in the child.⁷ Placental perfusion is controlled by maternal blood flow and fetal circulation.⁸ Insufficient perfusion leads to secretion of proinflammatory molecules that damage maternal endothelial cells and disrupt implantation.^{8, 9} Disruptions in placental pathophysiology may alter arterial structure in the child and is associated with hypertension in adulthood.¹⁰ However, little is understood about the mechanisms that may explain the developmental origins of the cardiometabolic manifestations in the offspring.

DNA methylation at cytosine-(phosphate)-guanine (CpG) dinucleotide markers is associated with alterations in vascular function in response to elevated blood pressure.^{11, 12} Maternal cardiometabolic factors are associated with DNA methylation in the placenta.^13–15 Differential DNA methylation in first trimester placental trophoblasts of women with preeclampsia alters expression of genes, (e.g. cadherin 11 [CDH11]), demonstrating that epigenetic modifications early in pregnancy can have effects on placental function.^{13, 16} Discovering placental epigenetic changes associated with maternal blood pressure may facilitate efforts to identify diagnostic markers and clinical therapeutics, informing guidelines for early prevention of cardiometabolic dysfunctions.^{12, 17} However, limited research exists in assessing the relationship between DNA methylation and blood pressure levels,¹¹ and there are no placental epigenome-wide studies (EWASs) of blood pressure during pregnancy.

The objectives of this study are: 1) to perform EWASs examining the associations of maternal systolic blood pressure (SBP) and diastolic blood pressure (DBP) during 1^st, 2^nd, and 3^rd trimesters of pregnancy with CpG methylation in the placenta, 2) to test gene expression changes per millimeters of mercury (mmHg) increase in SBP and DBP at each trimester 3) and to evaluate associations of previously known cardiometabolic trait-associated CpGs with blood pressure in an attempt to shed light on the mechanisms of developmental origins of cardiometabolic diseases.

Methods

The data that support the findings of this study are available from the corresponding author upon reasonable request. The DNA methylation data are available through dbGaP with accession number phs001717.v1.p1.

Study Design and Population

The current study included 312 women from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Fetal Growth Studies, a cohort study of women without major pre-existing medical conditions who provided placenta samples at delivery.^{18, 19} The study population is described in detail in the Online Supplement Methods section. The study was approved by institutional review boards at the NICHD and each of the participating clinical sites. Written informed consent was obtained from each woman who participated in the study.

Measurement of Maternal Variables

SBP and DBP measurements in mmHg were abstracted from routine blood pressure measurements taken during prenatal visits and measurements scheduled at 8–13, 16–22, 24–29, and 34–37 weeks of gestation. The 1^st, 2^nd and 3^rd trimester SBP was calculated to be the average of record-abstracted SBP measurements taken during 0–13 weeks, 14–27 weeks and 28–40 weeks of gestation, respectively.^20–22 Trimester-specific DBP was also calculated similarly. Gestational age was determined using the date of the last menstrual period and confirmed by ultrasound.^{18, 19}

Placental DNA and RNA Extraction and Measurement

Placental parenchymal biopsies were obtained at the fetal side from 312 mothers. After quality control (detailed in the Online Supplement Methods), 301 mothers with placenta sample were included in the EWAS. DNA was extracted as previously described.²³ DNA methylation was assayed using Illumina’s Infinium Human Methylation450 Beadchip (Illumina Inc., San Diego, CA) array. Standard Illumina protocols were followed for background correction, and normalization to internal control probes.²⁴ RNA was extracted from 80 placentas using TRIZOL reagent (Invitrogen, MA), and sequenced using the Illumina HiSeq2000 system. The expression of the transcripts were quantified using Salmon.²⁵ Details on quality control and quantification of RNA and methylation data are provided in the Online Supplement Methods section.

Placental SNP genotyping was done using HumanOmni2.5 Beadchips (Illumina Inc., San Diego, CA). Genotype-based principal components (PCs) were calculated from genome-wide autosomal SNP data. The R package “prcomp” was used to calculate PCs on the samples’ percent methylation profiles.²⁶ A total of 75 samples with both methylation and RNA-seq data were included in the association tests for correlation between DNA methylation and gene expression, as well as covariate-adjusted associations of SBP and DBP with gene expression.

Epigenome-wide Analyses

We performed six epigenome-wide analyses by fitting robust linear regression models for each CpG site as the response variable on the M-value scale and each of the 1^st, 2^nd and 3^rd trimester SBP and DBP variables as the primary predictors. All analyses included maternal age, race/ethnicity, fetal sex, five methylation sample plate (to adjust for potential batch effects), ten genotype-based PCs, three methylation-based PCs and putative cell-mixture estimated using surrogate variable analysis (SVA)²⁷ components (n=20) as adjustment factors in the models. Adjustment for genotype-based PCs has been demonstrated to provide additional benefit in accounting for population stratification.²⁸ We also included race/ethnicity in our models to potentially account for socio-economic differences. The quantile-quantile plots of P-values are presented in Figure S1. Additionally, we implemented Differentially Methylated Regions analysis described in our Online Supplement Methods section.²⁹

Gene Expression Analyses

First, we tested the Pearson correlation between the DNA methylation levels (M-values) of the top-significant CpGs and mRNA levels of their corresponding mapped genes. Second, we tested the associations between the mRNA levels of the genes with SBP and DBP measurements using DESeq2,³⁰ adjusting for maternal age, race/ethnicity, fetal sex, and 10 genotype-based PCs. The Benjamini-Hochberg adjusted P-value of the Wald test was used to correct for multiple testing and evaluate significant associations (P-value<0.05) between the blood pressure measurements and mRNA levels (hereafter referred as gene expression).

Canonical Pathway and Functional Annotation Analyses

To evaluate whether the top-significant CpGs are abundant in island and transcript start site (TSS) promoter regions, the chi-squared test was used. Functions of genes near the top 200 CpGs selected based on nominal EWAS P-value were explored to identify the most enriched canonical pathways and gene networks using Ingenuity Pathway Analysis (IPA, QIAGEN Redwood City, CA, USA, www.qiagen.com/ingenuity). Further details on pathway and functional annotation analyses are provided in the Online Supplement Methods section.

Evaluation of previously reported cardiometabolic trait-associated CpGs

Previously reported CpGs that were associated with cardiometabolic traits by EWASs were identified through PubMed using keywords and MeSH terms on DNA methylation or epigenetics and blood pressure, hypertension or preeclampsia studies published between January 01, 2013 and January 01, 2019.^{11, 13–15, 17, 31–34} We identified 312 cardiometabolic trait-associated CpGs, out of which 304 unique CpG sites were found in our data. To evaluate the associations of these 304 CpGs with SBP and DBP in our study, we considered CpGs with P-values <0.05 and effect directions (hyper- or hypo-methylation status) identical to the reported effect to be statistically significant (details are provided in the Online Supplement Methods section).

Results

Participants’ Characteristics

A total of 301 pregnant women who self-identified their race/ethnicity as non-Hispanic White (25.6%), non-Hispanic Black (23.9%), Hispanic (33.9%) and Asian/Pacific Islander (16.6%) participated in the study. Women were on average (mean ± s.d.) 27.7±5.3 years old and delivered at 39.5±1.1 gestational weeks (Table 1, S1 and S2). There were 8 (2.7%) gestational hypertensive and 5 (1.7%) preeclamptic women. The mean SBP at 1^st, 2^nd, and 3^rd trimester was 109.3±10.7, 108.7±9.0 and 111.1±9.0 mmHG, respectively, and the mean DBP was 66.6±7.5, 64.7±6.0, 67.1±6.3 mmHG, respectively.

Table 1.

Characteristics of study participants

Characteristics	Mean ± SD or n (%) (n=301)
Maternal age, y	27.7 ± 5.3
Gestational age at delivery, wk	39.5 ± 1.1
Self-reported race/ethnicity, n (%)
White	77 (25.6)
Black	72 (23.9)
Hispanic	102 (33.9)
Asian/Pacific Islander	50 (16.6)
Gestational hypertension, n (%)	8 (2.7)
Preeclampsia, n (%)	5 (1.7)
Systolic blood pressure, mmHG
1st Trimester (0–13 wk)	109.3 ± 10.7
2nd Trimester (14–27 wk)	108.7 ± 9.0
3rd Trimester (28–40 wk)	111.1 ± 9.0
Diastolic blood pressure, mmHG
1st Trimester (0–13 wk)	66.6 ± 7.5
2nd Trimester (14–27 wk)	64.7 ± 6.0
3rd Trimester (28–40 wk)	67.1 ± 6.3

Associations of maternal SBP and DBP during pregnancy with DNA Methylation in Placenta

Maternal SBP and DBP during pregnancy were associated with DNA methylation at a total of 24 CpG sites in placenta (3 in the 1^st, 6 in the 2^nd and 15 in the 3^rd trimester) (BACON-corrected FDR P-value<0.05; Empirical P-value ranged from 6.6×10⁻¹⁵ to 2.3×10⁻⁷) (Table 2, S3, Figure S2). Notably, higher 1^st trimester SBP and higher 1^st trimester DBP were associated with increased methylation at cg06880028 (COL12A1/FILIP1). Moreover, higher SBP or DBP at both 2^nd and 3^rd trimesters were significantly associated with increased methylation at cg14203118 (EEF1D), cg04867934 (RABGAP1L), cg09559971 (TRAF7 and SNORD60), and cg19565306 (RABL2B), and decreased methylation at cg02918224 (TP53INP2). Out of the 24 CpGs that were significantly associated with SBP and DBP in our data, 18 were enriched in CpG island (Chi-squared test P=0.002) and 11 were in promoter transcript start site (TSS) (Chi-squared P=1.4×10⁻⁶) regions. Among the top-significant CpGs in Table 2, cg10752545 (NAV2), cg05903251 (TBL3), cg06570818 (AGPAT1/RNF5), cg03712991 (SNX18), cg24195524 (RNF5/RNF5P1), cg17813879 (SH2D4A), cg14203118 (EEF1D), and cg13392293 (CDK5RAP2) showed directionally consistent associations with change in blood pressure in repeated measurement analyses after Bonferroni-correction (Table S4). Significantly differentially methylated regions associated with blood pressure for regions that span our top-significant CpGs in Table 2 included NAV2, HOXC4, TP53INP2, AGPAT1, RNF5, and RNF5P1 (Adjusted P-value<0.05) (Table S5).

Table 2.

Top-significant CpGs (FDR P <0.05) associated with maternal SBP and DBP during pregnancy

CpG	Gene	Chr: Position*	Relation to Gene/Feature	Relation to Island	Δβ†	P‡	FDR§	Gestation Week||	Phenotype
cg11702145	HCN3	1:155247214	TSS200	Island	0.021	1.6E-08	2.1E-02	28–40 wk	SBP
					0.039	1.5E-08	1.2E-03	28–40 wk	DBP
cg04867934	RABGAP1L	1:174129130	5’UTR	Island	0.015	9.0E-08	1.6E-04	14–27 wk	SBP
					0.013	6.1E-08	3.0E-02	28–40 wk	SBP
cg23618323	EDEM3	1:184723951	1stExon; 5’UTR	Island	0.016	2.4E-08	7.6E-03	28–40 wk	SBP
cg10752545	NAV2; LOC100126784	11:19734701	TSS200; Body	Island	0.027	4.4E-11	2.4E-06	28–40 wk	SBP
					0.041	1.4E-11	7.6E-04	28–40 wk	DBP
cg26201952	HOXC4	12:54446253	Nearest gene	N_Shore	−0.003	1.8E-08	2.4E-03	28–40 wk	SBP
					−0.005	9.0E-10	2.5E-04	28–40 wk	DBP
cg05130406	CAPS2	12:75723912	TSS200	Island	0.009	2.9E-08	6.0E-03	28–40 wk	SBP
					0.015	8.4E-09	3.7E-03	28–40 wk	DBP
cg23314283	SUGT1	13:53226751	5’UTR	Island	0.017	3.7E-11	5.5E-03	0–13 wk	SBP
cg05903251	TBL3	16:2022307	Body	Island	0.025	1.7E-15	3.3E-02	28–40 wk	SBP
					0.043	3.6E-14	1.1E-02	28–40 wk	DBP
cg09559971	TRAF7; SNORD60	16:2205259	TSS1500	Island	0.022	2.6E-09	8.5E-04	14–27 wk	SBP
					0.029	1.0E-11	1.0E-07	28–40 wk	SBP
					0.043	8.1E-13	1.2E-07	14–27 wk	DBP
					0.049	1.2E-12	2.2E-08	28–40 wk	DBP
cg00912407	ZDHHC7	16:84998665	Unclassified	OpenSea	−0.002	3.4E-09	2.1E-02	28–40 wk	SBP
					−0.004	1.5E-09	6.3E-03	28–40 wk	DBP
cg23685529	C17orf49	17:6918055	TSS200	Island	0.03	6.0E-10	3.7E-02	28–40 wk	DBP
cg13679837	XRN2	20:21283918	TSS200	Island	0.01	3.7E-09	1.1E-02	28–40 wk	DBP
cg02918224	TP53INP2	20:33298052	Nearest gene	Island	−0.009	2.3E-12	2.4E-04	28–40 wk	SBP
					−0.012	4.3E-08	3.0E-02	14–27 wk	DBP
					−0.013	1.3E-09	3.7E-03	28–40 wk	DBP
cg20351327	EIF3D	22:36925411	TSS200	Island	0.017	3.2E-13	3.4E-04	28–40 wk	SBP
					0.024	7.7E-09	5.8E-03	28–40 wk	DBP
cg19565306	RPL23AP82; RABL2B	22:51222011	Body; 1stExon; TSS200; 5’UTR	Island	0.028	4.0E-08	1.0E-04	14–27 wk	SBP
					0.027	3.4E-08	4.2E-04	28–40 wk	SBP
cg25627055	TRIM41	5:180650368	5’UTR; 1stExon	Island	0.025	2.4E-10	4.6E-04	28–40 wk	SBP
					0.035	1.5E-07	4.3E-02	14–27 wk	DBP
cg03712991	SNX18	5:53813394	TSS200	Island	0.006	2.3E-07	4.0E-02	28–40 wk	SBP
cg06570818	AGPAT1; RNF5; RNF5P1	6:32146466	Nearest gene	OpenSea	−0.006	1.3E-08	1.2E-03	28–40 wk	DBP
cg24195524	RNF5; RNF5P1	6:32147809	Nearest gene	OpenSea	0.002	1.2E-07	2.5E-02	28–40 wk	SBP
cg06880028	COL12A1; FILIP1	6:75918165	Nearest gene	S_Shore	0.004	3.2E-09	2.2E-03	0–13 wk	SBP
					0.006	1.1E-07	3.7E-03	0–13 wk	DBP
cg13534424	CAMK2B	7:44365292	TSS200	Island	0.01	2.2E-08	6.8E-04	0–13 wk	SBP
					0.009	2.9E-08	1.1E-02	28–40 wk	SBP
cg14203118	EEF1D	8:144680893	TSS1500	Island	0.026	2.2E-10	6.0E-08	14–27 wk	SBP
					0.031	1.3E-13	3.2E-14	28–40 wk	SBP
cg17813879	SH2D4A	8:19252188	Nearest gene	OpenSea	0.011	6.6E-15	9.1E-04	14–27 wk	DBP
cg13392293	CDK5RAP2	9:123342137	Body	Island	0.027	5.4E-14	2.0E-02	14–27 wk	SBP

^*Build 37 chromosome position

^†Methylation fold change/delta beta per 1 mmHg increase. Estimates that are statistically significant after FDR correction are shown in bold

^‡Empirical P-value

^§Bacon-corrected FDR P-value

^||0–13 wk, 1^st trimester; 14–27 wk, 2^nd trimester; 28–40 wk, 3^rd trimester

Canonical Pathway Analysis

Genes near top 200 EWAS CpGs were enriched in disease pathways including cardiovascular and metabolic diseases (Table S6). These included 15 molecules from 1^st trimester SBP EWAS that were involved in cardiovascular disease functions. Additionally, 13 molecules from 1^st trimester DBP EWAS were involved in cardiovascular system development and function. From 2^nd trimester SBP EWAS, 17 molecules were involved in cardiovascular disease and 15 molecules were involved in cardiac arrhythmia, cardiovascular and metabolic disease functions. Lastly, metabolic disease function was among the top canonical pathways that involved 31 molecules identified from 3^rd trimester SBP EWAS.

Functional Annotations

In the mQTL database, 201 SNPs were reported to be cis-meQTL with cg06880028 (COL12A1; FILIP1), which was associated with 1^st trimester SBP and DBP in our study. Another locus (rs4417778) was cis-meQTL with cg02918224 (TP53INP2), which was associated with 2^nd and 3^rd trimester DBP and 3^rd trimester SBP in our study (Table S7). The identified cis-meQTL SNPs overlap with regulatory motifs and showed the highest enrichment of cell type-specific enhancers (after Bonferroni correction of the binomial test p-values, P<3.9×10⁻⁴) for fetal heart and psoas muscle cells (Table S8). The expression of genes (n=28) near top-significant CpGs (n=25) in 53 different tissues from GTEx v7 Ensembl version 92 database are shown in Figure S3. FILIP1 gene is highly expressed in the artery, left ventricle of the heart and uterine tissues (Figure S4). COL12A1’s RNA expression is highest in the endometrium and its protein is moderately expressed in the placenta (Figure S5).

Correlations between DNA methylation and Gene Expression in Placenta

Higher 3^rd trimester SBP and DBP were significantly associated with higher expression of COL12A1 in placenta (log2-fold change=0.03 [P=0.03] and log2-fold change=0.05 [P=0.03] per 1 mmHG increase in SBP and DBP, respectively) (Table S9). No significant correlations were found between DNA methylation and gene expression levels of the annotated genes for each of the 24 CpG sites (Figure S6). However, there was a positive relationship between DNA methylation at cg06880028 (which was hypermethylated with higher BP) and COL12A1 gene expression levels (r=0.16, P=0.16) (Figure S6).

Evaluation of Previously Reported Cardiometabolic-trait-associated CpG Sites

Out of 304 CpG sites that were implicated in cardiometabolic traits in previous EWASs and evaluated in our study, 36 CpGs (11.8%) showed directionally consistent associations with maternal SBP and DBP in our data. These included 12 CpGs in placenta associated with preeclampsia, two CpGs in maternal leukocyte associated with preeclampsia, one CpG in peripheral blood associated with hypertensive pregnancy, one CpG in adult blood associated with SBP and DBP, and 20 CpGs in adult blood associated with myocardial infarction (Table S10-S11). Figure 1 shows a Venn Diagram of the CpGs that exhibited overlapping associations with cardiometabolic traits in published EWAS and with maternal SBP and DBP during pregnancy in our study.

Figure 1.

Illustration of overlaps between significantly differentially methylated CpGs in the present study and previously known cardiometabolic trait-related CpGs.

Discussion

To our knowledge, this is the first EWAS of trimester-specific maternal blood pressure in placenta. We identified 24 novel CpG sites associated with trimester-specific maternal systolic and diastolic blood pressures. These blood pressure-associated placental CpGs were abundant in CpG islands and promoter transcription start sites, which suggest their potential roles in functional regulation of gene transcription.³⁵ The genes annotating the top 200 EWAS CpGs showed enrichment in pathways including cardiovascular-metabolic development. Nearly 12% of CpGs previously implicated in cardiometabolic traits were found to be associated with maternal SBP and DBP in our data. These findings provide clues to specific epigenetic modifications in the placenta that may mediate the effect of maternal blood pressure on fetal development and disease risk in later life.³²

We observed corroborating evidence from DNA methylation, gene expression, and meQTL analyses for the role of the COL12A1 locus in blood pressure regulation. Specifically, higher SBP and DBP were associated with higher methylation at cg06880028 (near COL12A1 or FILIP1) in the 1^st trimester, and higher expression of COL12A1 gene in the 3^rd trimester. Several common variants were cis-meQTL with cg06880028 in blood across pregnancy, childhood, adolescence and middle age, suggesting the role of genetic regulation on DNA methylation across the life-course.³⁶ COL12A1 encodes the alpha chain of type XII collagen, which is highly expressed in the endometrium and is involved in embryonic development.³⁷ Weaker expression of collagen XII has been found in the intima of atherosclerotic plaques.³⁸ In mice, reduced mRNA expression of COL12A1 involved in endothelial-to-mesenchymal transition (EMT) was observed in the heart endothelial cells.³⁹ EMT is a fundamental process in establishing an early placental blood supply, and in atherosclerotic lesions associated with plaque stability.⁴⁰ EMT-derived fibroblastic-like cells can enhance clinical disease progression towards a collagen-poor, rapture prone plaque phenotype.⁴⁰ Together, these data suggest type XII collagen, involved in EMT, may play a role in preventing formation of atherosclerosis.

FILIP1 encodes filamin A binding protein and is highly expressed in fetal heart cells. FILIP1 transcripts related to fatty acid metabolism can be observed as early as 10 weeks during fetal development and were shown to increase as fetal heart matures.⁴¹ Fetal cardiomyocyte growth regulation is critical since the number of cardiac cells at birth predetermines the number of cardiomyocytes in later life,⁴² a known predictor of cardiac disease.⁴³ Our study’s finding of associations of higher SBP and DBP in the 1^st trimester with higher methylation of CpGs near FILIP1 was also corroborated by positive but non-significant correlation between 3^rd trimester SBP and DBP and expression of the FILIP1 gene.

There are a few placenta-specific EWASs in blood pressure-related traits such as preeclampsia and gestational hypertension.^13–15 Yeung et al¹⁵ identified 63 differentially methylated CpGs in the placenta associated with preeclampsia, 11 of which showed directionally consistent associations with higher SBP and DBP across the trimesters in our study. Anton et al¹³ identified 229 CpGs differentially methylated in preterm preeclamptic placentas (. Notable findings from our study when evaluating the reported CpGs include, associations of increased methylation of cg26624576 (CDH11) with increased SBP and DBP across the trimesters. CDH11 (Cadherin-11) regulates anchoring trophoblasts to the decidua.⁴⁴ Hypermethylation of cg26624576 in preeclamptic placenta has been validated by pyrosequencing by Anton et al¹³, and contributes to the decreased trophoblast invasiveness associated with preeclampsia.¹³

In maternal leukocytes, increased methylations of cg24030449 (FGF20) and cg01485645 (MLLT6) were observed among women with preeclampsia.³³ In our study, increased methylation of the same CpGs in placenta was associated with increased SBP in the 2^nd and 3^rd trimesters. FGF20 is involved in embryonic development, cell growth, and tissue repair.

In offspring blood, lower methylation of cg09735274 (TRIM31) was observed in hypertensive pregnancies.³² In our study, we increased 3^rd trimester SBP and DBP was associated with lower methylation of cg09735274 in placenta. TRIM31 gene expression in placenta has been associated with preeclampsia.⁴⁵ It is noteworthy that some of the differentially methylated CpG cites that were associated with levels of blood pressure in our study overlapped in their associations with hypertensive disorders of pregnancy.

In adult blood of 17,010 individuals, higher SBP and DBP has been associated with lower methylation of cg14476101 (PHGDH).¹¹ In our study, higher 3^rd trimester DBP was associated with lower methylation of cg14476101 in placenta. Among Swedish adults with history of myocardial infarction (MI), 211 CpG sites were differentially methylated in blood,¹⁷ of which 20 CpG sites showed directionally consistent associations with increased SBP and DBP in our study. In particular, methylation of cg03135061 (DYSF) was associated with MI,¹⁷ and increased maternal 1^st trimester DBP was associated with increased methylation of the same CpG in our study. Additionally, MI-associated increased methylations of two CpGs (cg19866478 and cg08301572), near zinc finger protein genes (ZNF480 and ZNF730), were associated with increased trimester-specific SBP and DBP in our study. Zinc finger proteins are involved in regulation of several cellular processes and development of diabetes.⁴⁶

Epigenetic marks have been shown to mediate the effect of pregnancy hypertension on risk of vascular disease later in life.³² These marks can mediate gene expression patterns that control embryonic development and organogenesis,⁴⁷ playing a central role in the developmental programming of adult disease.³² Our finding for blood pressure-CpGs methylation associations in the placenta as well as evidence from published cardiometabolic traits associated with epigenetic modifications in several tissues and stages of life support the developmental origins and health and diseases (DoHaD) hypothesis.^3,4

Our study has several limitations. Placental DNA methylation and gene expression profile varies during pregnancy.⁴⁸ Therefore, there may be a potential complex relationship between maternal blood pressure and placental DNA methylation levels across gestation that could not be distinguished by our study for which all placentas were collected at or near term. We are not able to distinguish whether differential methylation is a cause or consequence of maternal blood pressure change.³² We observed outliers and large ranges of values in gene expression (e.g. RBX1) that may be due to cell-type heterogeneity by sample. Potential heterogeneity in cell composition of our placenta samples may have been exacerbated by placental histopathological differences between sub-groups of individuals, however, in our previous study that used the same study participants,⁴⁹ we did not find evidence for differences of gross placental histopathological characteristics by race/ethnicity or sex.

To account for confounding by cell-type, we adjusted for cell-type heterogeneity based on data-driven estimation using the SVA method. Ability to adjust for genotype-based ancestry was another important strength in our study as it allowed effective removal of bias due to population stratification.²⁸ With further application of a Bayesian method, the inflation factors of P-values in the association results were minimized.⁵⁰ Moreover, our EWAS findings were corroborated with downstream functional annotations and directionally consistent replications of adult BP-related CpGs. Finally, our study included diverse race/ethnic study participants with longitudinal measurement of maternal blood pressure throughout pregnancy.

Conclusions

In this study we report evidence for novel placental epigenetic changes associated with increased maternal blood pressure during pregnancy at genes including COL12A1 and FILIP1 that are implicated in cardiometabolic diseases. Epigenetic changes may explain the mechanisms of the associations between increased maternal blood pressure, fetal development and adult cardiometabolic diseases. Identification of blood pressure-induced methylated sites in the placenta may provide clues to early origins of cardiometabolic diseases and inform guidelines for early prevention.

Perspectives

Placental DNA methylation is a suggested regulatory mechanism through which maternal blood pressure influences fetal and adult health outcomes. We conducted the first epigenome-wide association study to examine the associations of maternal systolic and diastolic blood pressure during 1^st, 2^nd, and 3^rd trimesters of pregnancy with DNA methylation in the placenta. Findings reported herein provide strong evidence for placental DNA methylation of genes involved in cardiovascular-metabolic development pathways. Placental CpGs that were known to be associated with cardiometabolic factors in other studies, conferred associations with systolic and diastolic blood pressure in our data. Findings on blood pressure-associated methylated sites in the placenta may provide clues to early origins of cardiometabolic disease risk and inform guidelines for early prevention.

Novelty and Significance.

What is new?

• This is the first placental EWAS of maternal blood pressure during pregnancy.

What is relevant?

• Higher maternal blood pressure during pregnancy is associated with DNA methylation changes in the placenta.

• Placental DNA methylation loci have relevance for cardiometabolic diseases.

Summary

• Placental methylation signatures of higher blood pressure may underlie molecular pathways in early origins of cardiometabolic disease risk.