Adaptations of placental and cord blood ABCA1 DNA methylation profile to maternal metabolic status

Abstract

In utero environmental perturbations have been associated with epigenetic changes in the offspring and a lifelong susceptibility to cardiovascular diseases (CVD). DNA methylation at the ATP-binding cassette transporter A1 (ABCA1) gene was previously associated with CVD, but whether these epigenetic marks respond to changes in the maternal environment is unknown. This study was undertaken to assess the associations between the maternal metabolic profile and ABCA1 DNA methylation levels in placenta and cord blood. Placenta and cord blood samples were obtained at delivery from 100 women including 26 with impaired glucose tolerance (IGT) diagnosed following a 75 g-oral glucose tolerance test (OGTT) between week 24 and 28 of gestation. ABCA1 DNA methylation and mRNA levels were measured using bisulfite pyrosequencing and quantitative real-time PCR, respectively. We report that ABCA1 DNA methylation levels on the maternal side of the placenta are correlated with maternal high density lipoprotein cholesterol (HDL-C) levels (r < –0.21; P < 0.04) and glucose levels 2 h post-OGTT (r = 0.25; P = 0.02). On the fetal side of the placenta, ABCA1 DNA methylation levels are associated with cord blood triglyceride levels (r = –0.28; P = 0.01). ABCA1 DNA methylation variability on both sides of the placenta are also associated with ABCA1 mRNA levels (r < –0.35; P = 0.05). As opposed to placenta, cord blood DNA methylation levels are negatively correlated with maternal glucose 2 h post-OGTT (r = –0.26; P = 0.02). In conclusion, the epivariations observed in placenta and cord blood likely contribute to an optimal materno–fetal cholesterol transfer. These in utero epigenetics adaptations may also potentially trigger the long-term susceptibility of the newborn to dyslipidemia and CVD.

Keywords: :

• Gestational diabetes
• high-density lipoprotein cholesterol
• fetal programming
• lipid metabolism
• epigenetics

Introduction

Pregnancy is associated with a number of metabolic adaptations that support a normal fetal growth. These adaptations include maternal weight gain, changes in blood lipid profile, and insulin resistance.¹ If the pancreatic β-cells cannot face the increased demand of superimposed gestational insulin resistance, pregnant women develop gestational diabetes mellitus (GDM). GDM, defined as impaired glucose tolerance with first recognition during pregnancy, is the most frequent metabolic perturbation observed in pregnant women with a prevalence ranging from 2 to 20% according to the population studied and diagnostic criteria used.² Fetal exposure to GDM is associated with a long-term increased risk for obesity, type 2 diabetes (T2D) and cardiovascular disease (CVD).^3-5 This concept of fetal programming of adult metabolic diseases (originating from Barker’s hypothesis⁶) is well supported by several studies and thus now widely accepted. However, the molecular mechanisms involved have not been clearly defined so far. Recent studies have suggested that epigenetic modifications might explain, to some extent, the link between suboptimal fetal conditions and long-term risks for cardio-metabolic disorders in the newborn.⁷

Epigenetics refers to DNA modifications independent of the primary sequence that are involved in gene transcription regulation.⁸ DNA methylation is the most stable and studied epigenetic modification and is generally associated with a repression of gene expression.⁹ The addition of a methyl group is catalyzed by the DNA methyltransferases and occurs at position 5' of the pyrimidine ring of the cytosine upstream of a guanine (CpG dinucleotide).¹⁰ Cytosine methylation in a CpG context has been shown to be associated to environmental stimuli including the hyperglycemic intrauterine environment.^11-13 This is now supported by an in vitro study showing that a chronic exposure to hyperglycemia induces persistent DNA methylation and gene expression profile alterations of key metabolic genes.¹⁴

Epigenetic adaptations of the placenta and newborn appear to be important for the regulation of placental functions, fetal growth, and development. They have been associated with risk factors (birth weight, child body mass index [BMI] and adiposity) known to trigger an increased long-term susceptibility to metabolic disorders.^15-19 The newborn epigenetic signature acquired in the womb also appears to be persistent throughout adult life in animal models and in human epidemiological studies.^20-22

Previously, a study carried on by Tobi et al.²² has reported that DNA methylation levels at the ATP-binding cassette transporter A1 (ABCA1) gene locus are higher in leukocytes of adults exposed prenatally to the ending World War II Dutch famine, suggesting that the ABCA1 DNA methylation profile might be sensitive to in utero adverse environmental exposure. ABCA1 transfers cholesterol from cells to apolipoproteins A-1 (apo-A1), thus contributing to the formation of nascent high-density lipoproteins (HDL).²³ In human placenta, ABCA1 is highly expressed in endothelial cells of fetal capillaries and also in the syncytiotrophoblast cells on the maternal side of the placenta where it regulates materno-fetal cholesterol transport.^24,25 ABCA1 is thus a crucial regulator of the lipoprotein metabolism. Recently, our group reported that ABCA1 epipolymorphisms are associated with HDL particle profile (especially HDL cholesterol [HDL-C] concentrations) as well as history of coronary artery disease in adults with familial hypercholesterolemia.²⁶ Hence, ABCA1 DNA methylation profile appears to be sensitive to in utero environment and might have long-term functional impact on lipid profile and the onset of CVD. We have thus hypothesized that the newborn ABCA1 DNA methylation profile could be disturbed by the maternal metabolic status. Accordingly, the main objective of the current study was to determine the associations between the maternal glucose status and lipid profile and ABCA1 DNA methylation profile in placenta and cord blood.

Results

Table 1 shows the characteristics of mothers and their newborns according to the maternal glucose tolerance status. Women were on average 29 y old, were slightly overweight and had normal fasting glucose concentrations and lipid profile. Of the 100 women recruited, 26 had impaired glucose tolerance (IGT), whereas 74 were normoglycemic (NGT). Among the women with IGT, 13 were recommended to follow a diet and 12 were treated with insulin in addition to the diet (treatment was unknown for one woman). Parity was higher in women with IGT compared with NGT women (P = 0.02). Women with IGT gained less weight on average than NGT women during pregnancy. ABCA1 DNA methylation levels were found to be significantly lower in the blood of IGT mothers than with NGT mothers (P = 0.04). Cord blood ABCA1 DNA methylation levels also tended to be lower in newborn exposed to IGT (P = 0.06) (Table 1). Moreover, newborns from the IGT and NGT groups had similar gestational age, birth weight, and cord blood lipid profile. Interestingly, children born to mothers with IGT had significantly higher cord blood c-peptide levels than those born to NGT.

Table 1. Maternal and newborn characteristics according to glucose tolerance status

	NGT (n = 74)		IGT (n = 26)
	Mean ± SD	Range	Mean ± SD	Range
Maternal Characteristics
Personal history of gestational diabetes	2 (2.7%)	-	8 (30.8%)**	-
Gravidity
0	35 (47.3%)	-	9 (34.6%)	-
1	20 (27.0%)	-	7 (26.9%)	-
2	16 (21.6%)	-	8 (30.8%)	-
3	3 (4.1%)	-	2 (7.7%)	-
Parity
0	43 (58.1%)	-	9 (34.6%)*	-
1	18 (24.3%)	-	14 (53.8%)*	-
2	13 (17.6%)	-	3 (11.5%)*	-
Mode of delivery
C-section	14 (18.9%)	-	3 (11.5%)	-
Vaginal	60 (81.1%)	-	23 (88.5%)	-
1st trimester
Age (years)	28.8 ± 3.9	22.0–39.0	29.1 ± 4.0	20.0–38.0
BMI (kg/m^2) Waist circumference (cm)	25.4 ± 5.4 90.1 ± 14.6	16.6–43.0 62.0–137.0	25.6 ± 3.7 87.5 ± 8.4	19.6–32.5 68.0–102.0
Fasting glucose (mmol/L)	4.46 ± 0.32	3.80–5.20	4.43 ± 0.50	3.40–5.40
Fasting insulin (mU/L)^a HOMA-IR^a	5.88 ± 4.39 1.16 ± 0.92	1.70–23.90 0.34–4.78	5.89 ± 3.95 1.15 ± 0.75	2.30–16.70 0.46–2.97
Total cholesterol (mmol/L)	4.73 ± 0.77	3.30–7.20	4.75 ± 0.90	3.10–8.00
HDL-C (mmol/L) LDL-C (mmol/L)	1.56 ± 0.28 2.53 ± 0.74	1.00–2.35 1.20–5.00	1.62 ± 0.34 2.62 ± 0.72	1.25–2.31 1.30–5.10
TG (mmol/L)^a	1.07 ± . 0.48	0.40–3.00	1.24 ± 0.50	0.60–2.70
2nd trimester
2 h-post-OGTT glucose concentration (mmol/L)	6.16 ± 0.90	3.80–7.70	8.43 ± 0.61**	7.80–10.80
Weight gain between 1st and 3rd trimester (% of initial weight)	18.8 ± 7.0 (n = 73)	1.6–34.1	13.5 ± 5.9** (n = 22)	1.5–25.3
Newborn characteristics
Sex
Male	35 (47.3%)	-	13 (50.0%)	-
Female	39 (52.7%)	-	13 (50.0%)	-
Gestational age (weeks)	39.3 ± 1.2	36.0–41.6	39.1 ± 1.5	36.1–41.6
Birth weight (kg)	3.45 ± 0.46	2.00–4.27	3.33 ± 0.53	2.13–4.58
Placenta weight (g)	590 ± 123 (n = 66)	360–992	584 ± 115 (n = 25)	370–883
Fetal-placental weight ratio	5.97 ± 0.84 (n = 66)	3.74–7.82	5.78 ± 0.83 (n = 25)	3.46–7.50
Cord blood total cholesterol (mmol/L)	1.63 ± 0.36 (n = 64)	0.93–2.89	1.64 ± 0.37 (n = 23)	1.2–2.54
Cord blood LDL-C (mmol/L)	0.85 ± 0.24 (n = 59)	0.50 ± 1.90	0.80 ± 0.21 (n = 23)	0.51–1.23
Cord blood HDL-C (mmol/L)	0.62 ± 0.19 (n = 64)	0.28–1.11	0.63 ± 0.21 (n = 23)	0.38–1.14
Cord blood TG (mmol/L)^a	0.40 ± 0.24 (n = 64)	0.13–1.82	0.43 ± 0.22 (n = 23)	0.22–0.94
Cord blood c-peptides (pmol/L)	112.7 ± 71.9 (n = 40)	1.0–354.0	160.8 ± 76.3* (n = 23)	15.0–308.0
ABCA1 DNA methylation (%)
Maternal Blood – Mean	33.1 ± 6.9	20.0–45.6	29.9 ± 5.6*	18.6–39.9
Maternal side of the placenta – CpG5	15.6 ± 3.8	8.5–25.7	16.7 ± 4.4	8.4–23.7
Maternal side of the placenta- CpG6	14.6 ± 3.6	7.7–25.8	15.1 ± 3.9	7.9–21.8
Fetal side of the placenta – CpG5	15.7 ± 4.8	4.0–26.1	16.2 ± 3.4	9.9–24.7
Fetal side of the placenta- CpG6	16.6 ± 4.2	7.5–27.0	15.8 ± 2.3	10.6–19.4
Cord blood - Mean	19.1 ± 5.6	7.7–32.2	16.6 ± 4.9^†	8.0–24.6

^a Geometric mean and p-value obtained after log₁₀-transformation of the variable. Means are different between NGT and IGT (**P ≤ 0.01;*P ≤ 0.05; ^†P ≤ 0.10).

ABCA1 genomic context and DNA methylation analyses

Overall, 22 CpG dinucleotides within four regions of the ABCA1 gene promoter CpG island were first epigenotyped in ten placenta samples (maternal and fetal sides), maternal blood at first trimester and cord blood samples (Fig. 1). The CpG sites that were found to be unmethylated (<10.0%) or fully methylated (>90.0%) were excluded from the current study as they showed low variability and were unlikely to explain the phenotypic variability. Hence, ABCA1 loci B, C, and D, mostly unmethylated (<10%) and almost invariable in all four tissues, were not further analyzed. ABCA1-A locus (8 CpGs), located just upstream the first exon, was the only region showing interindividual variability in DNA methylation levels across all tissues and was consequently analyzed in the remaining samples.

Figure 1.ABCA1 CpG island proximal promoter region and localization of the 4 loci epigenotyped in maternal and fetal placenta, as well as in maternal and cord blood samples. The ABCA1-A locus shows DNA methylation variability in both fetal and maternal sides of the placenta (especially at CpG5 and CpG6), as well as in maternal and cord blood samples.

DNA methylation levels of maternal blood samples at the ABCA1-A locus were highly correlated between CpG sites 3 to 8 (r = 0.80 to 0.97; P < 0.001), whereas CpG sites 1 and 2 were less strongly correlated with the 6 others (r = 0.58 to 0.87; P < 0.001) (Fig. S1). In cord blood samples, DNA methylation levels of the 8 CpGs analyzed were found to be highly correlated with each other (r = 0.83 to 0.97; P < 0.001) (Fig. S1). Mean DNA methylation level values of the 8 CpGs epigenotyped were thus computed in maternal and cord blood samples and used for further analyses as previously described by Guay et al.²⁶ and Tobi et al.²²

Mean ABCA1 DNA methylation levels were higher in maternal blood as compared with cord blood (31.9 ± 7.3% vs. 18.0 ± 5.7%, P < 0.001). Interestingly, mean ABCA1 DNA methylation levels were found to be well correlated between maternal blood and cord blood samples (r = 0.50; P < 0.001). Maternal blood ABCA1 DNA methylation levels explained up to 20.0% (R²) of the DNA methylation variability at ABCA1 locus in cord blood.

On average, ABCA1-A locus DNA methylation was found to be lower than 10% in maternal (8.8 ± 2.6%) and fetal (9.7 ± 2.7%) placenta samples. Nevertheless, mean ABCA1-CpG5 and ABCA1-CpG6 DNA methylation levels were between 14.7% and 15.9% on average with good variability (s.d. > 3%) (Fig. 1). CpG sites 1 to 4, 7, and 8 were excluded from the analysis as they were considered unmethylated (DNA methylation < 10%) and showed low variability (s.d. < 3%) (Fig. 1). As they were only modestly correlated with each other (maternal side: r = 0.62 and fetal side: r = 0.54), ABCA1-CpG5 and ABCA1-CpG6 were investigated separately in both maternal and fetal side of the placenta. DNA methylation levels at ABCA1-CpG5 and CpG6 in maternal and cord blood were not associated to DNA methylation levels in either the maternal (r < |0.12|; P > 0.29) or fetal side of the placenta (r < |0.10|; P > 0.31).

Placental ABCA1 DNA methylation levels according to the maternal and newborn metabolic profiles

We first assessed whether the maternal metabolic profile might be associated with placental ABCA1 DNA methylation variability. On the maternal side of the placenta, we observed that higher maternal HDL-C concentrations at first trimester of pregnancy were associated with decreased DNA methylation levels at both ABCA1-CpG5 (r = –0.21; P = 0.04) and ABCA1-CpG6 (r = –0.24; P = 0.02) (Fig. 2A and B). Consistently, the maternal placental ABCA1-CpG6 locus was found to be differentially methylated according to maternal HDL-C level tertiles at first trimester (right section of Fig. 3, P = 0.05). Indeed, mothers with the highest HDL-C levels (>1.67 mmol/L, n = 35) had 2.3% lower DNA methylation levels than women with the lowest concentrations (<1.44 mmol/L, n = 33).

Figure 2. Placental ABCA1-CpG5 and -CpG6 DNA methylation levels according to maternal and newborn metabolic profile. Higher maternal HDL-C levels at first trimester of pregnancy (A and B) and lower maternal glucose concentrations 2 h post-OGTT (C) were both associated with lower ABCA1 DNA methylation levels in the maternal side of the placenta. Panel D shows the negative correlation between ABCA1-CpG5 DNA methylation levels in the fetal side of placenta and cord blood triglyceride levels. ^aIncluded age, BMI, TG at first trimester and history of gestational diabetes as covariates. ^bAdjusted for gestational age and newborn sex. ^cCorrelation coefficient and P value obtained after log10-transformation of the cord blood triglyceride levels.

Figure 3. DNA methylation levels at ABCA1-CpG6 on the maternal side of the placenta according to glucose tolerance status and HDL-C levels. Women with impaired glucose tolerance (IGT) and low HDL-C levels (<1.44 mmol/L) had the highest ABCA1-CpG6 DNA methylation levels. Maternal glucose status and HDL-C levels interact to increase ABCA1 DNA methylation on maternal side of placenta (P = 0.04). ^aP values obtained in the Bonferroni post-hoc analysis by comparing low (<1.44 mmol/L) and medium (1.44–1.67 mmol/L) levels of HDL-C to levels >1.67 mmol/L in IGT group. ^bP value obtained in the Bonferroni post-hoc analysis by comparing HDL-C levels <1.44 mmol/L to levels >1.67 mmol/L in the entire cohort (n = 100).^cANCOVA adjusted for age and BMI at first trimester and history of GDM.

Moreover, we observed a positive correlation between maternal placenta ABCA1-CpG5 DNA methylation levels and maternal glucose 2 h post-OGTT levels (r = 0.25; P = 0.02) (Fig. 2C). However, ABCA1 DNA methylation levels were found to be similar in the IGT and NGT groups on both sides of the placenta (Table 1). In addition, DNA methylation levels at ABCA1-CpG6 were significantly different among IGT women according to HDL-C level tertiles at first trimester (left section of Fig. 3; P = 0.04). In women with IGT and high HDL-C levels (>1.67 mmol/L, n = 9), ABCA1-CpG6 was found to be hypomethylated when compared with the two lower tertiles of HDL-C (Fig. 3). For women with NGT, no significant difference in DNA methylation was observed among the HDL-C levels tertiles (middle section of the Fig. 3). Of interest, an interaction effect was observed between maternal glucose status and HDL-C levels on DNA methylation levels at ABCA1-CpG6 (P = 0.04) (Fig. 3). A trend for interaction on DNA methylation levels at ABCA1-CpG5 was also observed between these two factors (P = 0.07), even though no significant DNA methylation difference was observed between groups (HDL-C levels and glucose tolerance status) for this CpG on the maternal side of placenta (data not shown). Results remained unchanged after consideration for potential confounding factors (age, BMI and triglycerides [TG] at first trimester of pregnancy, history of GDM, and weight gain between first and third trimester).

On the fetal side of the placenta, the maternal metabolic profile did not contribute significantly to ABCA1 DNA methylation level variability. Nevertheless, DNA methylation at fetal placental ABCA1-CpG5 was significantly correlated with cord blood TG concentrations (r = –0.28; P = 0.01) (Fig. 2D), whereas a trend for association was also observed with cord blood HDL-C levels (r = 0.19; P = 0.09) when newborn sex and gestational age were taken into account. All the associations remained similar after statistical adjustments for maternal age, BMI at first trimester, history of GDM and weight gain between first and third trimester. DNA methylation levels on the maternal and fetal side of the placenta were not found to be associated with other maternal or fetal metabolic variables (Tables S2 and S3).

According to the results of the stepwise multivariable linear regression analyses of placental ABCA1 DNA methylation level predictors, maternal and fetal characteristics explained respectively up to 11.4% and 5.9% of the ABCA1 DNA methylation variability observed in the maternal or fetal side of placenta (Table 2). These results remained unchanged when the forward and backward multivariable linear regression models were applied (data not shown).

Table 2. Multivariable linear regression analyses of ABCA1 DNA methylation levels predictors in placenta and cord blood

	Maternal side of placenta		Fetal side of placenta	Cord blood
Predictors	ABCA1-CpG5 (n = 96)	ABCA1-CpG6 (n = 96)	ABCA1-CpG5 (n = 79)	Mean ABCA1 DNA methylation (n = 71)
1st trimester HDL-C level (mmol/L)	β = -0.26 P = 0.008	β = -0.29 P = 0.005	NS	NS
1st trimester TG level (mmol/L)^†	NS	NS	NS	β = -0.20 P = 0.06
2nd trimester 2 h-post-OGTT glucose level (mmol/L)	β = 0.26 P = 0.008	β = 0.17 P = 0.08	NS	β = -0.18 P = 0.08
Cord blood TG level (mmol/L)	NS	NS	β = -0.27 P = 0.02	NS
Placental weight (g)	NS	NS	NS	β = 0.25 P = 0.01
Maternal blood ABCA1 DNA methylation level	NS	NS	NS	β = 0.41 P < 0.001
R^2 (%) R^2adj (%) SEest	13.3 11.4 3.7 P = 0.001	10.8 8.8 3.5 P = 0.005	7.1 5.9 4.6 P = 0.02	37.8 34.0 4.5 P < 0.001

The input-independent variables tested were maternal characteristics (age, BMI, fasting triglyceridemia, HDL-C, 2 h post-OGTT glucose, and blood ABCA1 DNA methylation levels) and fetal characteristics (gestational age, sex, placental weight, cord blood HDL-C and TG levels). All β are standardized coefficient values. ^†Results obtained after log₁₀-transformation of the TG level. NS = P > 0.1.

Correlations between cord blood ABCA1 DNA methylation levels, maternal characteristics, and newborn metabolic profiles

Next, we investigated whether maternal and newborn characteristics might be associated with cord blood ABCA1 DNA methylation level variability. First, the maternal glucose metabolism at second trimester of pregnancy was found to be associated with the DNA methylation profile of the cord blood ABCA1-A locus. Indeed, we observed a negative correlation between cord blood ABCA1-A DNA methylation levels and maternal glucose concentrations (2 h post-OGTT) (Fig. 4), as well as an association with fasting TG at first trimester (r = –0.24; P = 0.03) when all maternal potential confounders variables were included in the statistical model. Moreover, higher cord blood ABCA1-A DNA methylation levels were significantly associated with higher placenta weight (r = 0.29; P = 0.01) (Table S2) and lower fetal-placental weight ratio (r = -0.25, P = 0.03; Fig. 4) at term when both gestational age and newborn’s sex were considered. These correlations remained unchanged after the inclusion of the following variables: maternal age, BMI, blood ABCA1 DNA methylation, TG at first trimester, glucose levels 2 h post-OGTT at second trimester and weight gain between first and third trimester. No additional correlation was observed between cord blood ABCA1 DNA methylation levels, maternal and newborn characteristics (Tables S2 and S3). Overall, the combination of maternal and fetal characteristics might explain up to 34.0% of the cord blood ABCA1 DNA methylation variability (Table 2).

Figure 4. Cord blood ABCA1-A locus DNA methylation levels according to maternal and newborn characteristics. Pearson’s correlations between cord blood mean ABCA1 DNA methylation levels and: (A) 2 h post-OGTT maternal glucose concentrations, and (B) placental efficiency (estimated by fetal-placental weight ratio). ^aAdjusted for newborn’s sex and gestational age, as well as maternal age, BMI, TG at first trimester and history of gestational diabetes.

Correlation between ABCA1 DNA methylation and mRNA levels on the maternal and fetal side of the placenta

To assess the functional impact of the ABCA1 DNA methylation level variability, ABCA1 mRNA levels were quantified in maternal and fetal placenta. Relative ABCA1 gene expression was found to be negatively correlated with ABCA1-CpG6 DNA methylation levels on the maternal (r = –0.35; P = 0.05) and fetal side of the placenta (r = –0.36; P = 0.05) (Fig. 5). Moreover, a trend for association was observed between DNA methylation at ABCA1-CpG5 and ABCA1 gene expression in fetal placenta (r = –0.35; P = 0.06). Of note, potential binding sites for transcription factors such as AP-2, AP-4, ETS1, SRY, and STAT were found closed to ABCA1-CpG5 and CpG6, within or near ABCA1-A locus.²⁶

Figure 5. Functional impact of the ABCA1 DNA methylation level variability. Pearson’s correlations between ABCA1-CpG6 locus DNA methylation and mRNA levels on (A) the maternal side and (B) the fetal side of the placenta.

Discussion

Previous studies suggested that in utero environment perturbations, such as maternal dyslipidemia and hyperglycemia, contribute to the development of adult chronic diseases (obesity, diabetes, CVD),^5,27 possibly through epigenetic adaptations in the placenta and offspring.^11,12,28 Leukocyte DNA methylation variability at ABCA1 gene locus was previously associated to prenatal famine exposure, blood lipid profile and CVD.^22,26 These results suggest that ABCA1 epigenetic profile is sensitive to the in utero environment and that methylation at this locus could explain, to some extent, the link between suboptimal maternal conditions and the metabolic health programming of the newborn. In the current study, we showed that higher ABCA1 DNA methylation levels on the maternal side of the placenta were associated with lower maternal HDL-C levels and higher maternal glucose concentrations 2 h post-OGTT at first trimester and second trimester of pregnancy respectively (Fig. 6).

Figure 6. Maternal and fetal metabolic variables associated with placental and cord blood ABCA1 DNA methylation variations. Maternal (glucose 2 h post-OGTT and HDL-C levels) and fetal metabolic profile (cord blood triglyceride levels) are respectively associated with ABCA1 DNA methylation levels in the maternal and fetal side of the placenta. Higher maternal glucose 2 h post-OGTT and triglyceride levels are associated with lower cord blood ABCA1 DNA methylation levels, whereas maternal epigenetic profile is positively associated with cord blood ABCA1 DNA methylation.

In our cohort, IGT women gave birth to children with higher cord blood c-peptide levels even if they gained less weight throughout their pregnancy possibly because of their required clinical follow-up.^29,30 These results suggest that child born to IGT mother have higher insulin production and confirm that they were exposed to higher maternal glucose concentrations in utero. A hyperglycemic in utero environment has already been associated with impaired fetal and placental metabolism and epigenomic changes. Indeed, several epidemiological studies have associated maternal glucose with fetal overgrowth, excess adiposity in childhood and long-term susceptibility of offspring to obesity and metabolic disorders.^5,31,32 Moreover, a recent in vitro study has reported that chronic exposure to hyperglycemia leads to persistent DNA methylation modifications of key metabolic genes in Schwann cells.¹⁴ Likewise, associations between maternal hyperglycemia exposure and placental DNA methylation variations at LEP, ADIPOQ and MEST gene loci have been previously reported by our group and El Hajj et al.^11-13 The maternal lipid profile (HDL-C, total cholesterol, and TG), in pregnancies complicated or not by IGT, was also shown to be a determinant of birth weight and lifelong susceptibility to cardio-metabolic disorders.^33-35 Our results support these observations and suggest that the effects of maternal lipid and glucose changes could partly be mediated through epigenetic modifications at specific loci, such as ABCA1 and other genes.

The exact role of ABCA1 is not clearly defined on the maternal side of the placenta. Indeed, ABCA1 appears to regulate both the placental cholesterol uptake and cholesterol efflux toward maternal circulation.^36,37 We showed that mothers with low HDL-C and high glucose levels at first and second trimester of pregnancy respectively had higher ABCA1 DNA methylation levels and a concomitant reduction in ABCA1 expression in the placenta. This interaction might be attributed to an overall less favorable metabolic profile, often characterized by the combination of low HDL-C and high glucose levels. Moreover, the reduction of placental ABCA1 mRNA has already been reported in obese and normal weight subjects with IGT and suggested to impair placental cholesterol-uptake and transport.³⁸ Although ABCA1 DNA methylation levels on the maternal side of the placenta were not associated with birth weight or the cord blood lipid profile, this suggests that other key players of the placental cholesterol metabolism may compensate for the loss of function of ABCA1 (e.g., ATP-binding cassette G1 [ABCG1] or scavenger receptor class B member 1 [SRB1]). Furthermore, we cannot exclude that the ABCA1 DNA methylation epivariations we observed on the maternal side of the placenta may also contribute to the maternal HDL-C level variability.

The role of ABCA1 in the endothelial cells of the fetal side of the placenta is more clearly defined.³⁹ Indeed, Stefulj et al.³⁹ have reported that ABCA1 regulates the cholesterol efflux from the placenta to apo-A1 and apo-E for HDL biogenesis and subsequent release into fetal circulation. In the current study, ABCA1 DNA methylation levels on the fetal side of the placenta were not associated with maternal metabolic profile, but they were associated with cord blood lipid profile (Fig. 6). Increased ABCA1 DNA methylation levels and the expected decreased expression were observed when cord blood HDL-C concentrations were higher and TG levels were lower. Interestingly, ABCA1 DNA methylation on the maternal side of the placenta was not associated with newborn weight or cord blood lipid profile. As the fetus can synthesize de novo cholesterol, the existence of fetal stimuli to prevent excess cholesterol delivery from the placenta has been proposed.⁴⁰ Our results support this hypothesis and suggest that cord blood lipid concentrations could prevent excess cholesterol transfer from the fetal side of the placenta to fetal circulation through epigenetic modifications at the ABCA1 gene locus. The ABCA1 epigenetic profile on the fetal side of the placenta may be more susceptible to fetal lipid level adaptations than the maternal side since it appears to be one of the most important regulators of placental-fetal cholesterol transfer and likely of newborn cholesterol homeostasis. However, the cord blood lipid profile only explains a modest fraction of ABCA1 DNA methylation variability in the fetal side of the placenta suggesting that other factors might affect this epipolymorphism and the cholesterol transfer.

In a previous study carried on leukocytes, we failed to find a significant association between DNA methylation at the ABCA1-A gene promoter locus and ABCA1 mRNA levels.²⁶ The results of the current study are the first to suggest that the ABCA1 DNA methylation locus tested in our previous study (Guay et al.²⁶) and that of Tobi et al.²² is associated with its gene expression regulation, at least in a tissue-specific manner in placenta. Further analyses in others tissue will be needed to confirm the possible tissue-specific impact of DNA methylation at ABCA1 gene promoter locus.

Cord blood samples were also epigenotyped to determine the relationship between the blood lipid profile and maternal IGT. We first observed that ABCA1 DNA methylation levels were associated with maternal blood leukocyte ABCA1 DNA methylation. Based on correlation analyses, we estimated that the familial resemblance or the multifactorial heritability of leukocyte ABCA1 DNA methylation might be as high as 20%. This is concordant with previous studies, which have lately estimated the whole-blood DNA methylation heritability to 18% at the genome-wide level in twin pairs, and 20% and 15–48% at LEP and FABP3 gene loci respectively in family member comparison studies.^41-43 Furthermore, it suggests that the newborn epigenetic profile at the ABCA1 gene locus is partially inherited and thus likely affected by the environment to which the mother has been exposed throughout her own life (Fig. 6).

Although newborns’ methylome is only moderately inherited from the mother, our results also suggest that it may vary according to the intrauterine environment to which it has been exposed. Indeed, cord blood ABCA1 DNA methylation was found to be negatively correlated with maternal fasting triglyceridemia and glucose levels 2 h post-OGTT at first and second trimester of pregnancy respectively (Fig. 6). Interestingly, the correlation between maternal glucose and ABCA1 DNA methylation levels showed opposite associations in cord blood and on the maternal side of the placenta. Accordingly, newborns’ methylome may differently respond to maternal glucose in order to overcome placental ABCA1 DNA methylation modifications and impaired materno-fetal cholesterol transfer. On the maternal side of the placenta, we observed increased DNA methylation levels at ABCA1-A locus when maternal glucose concentrations were higher at second trimester. As increased placental ABCA1 DNA methylation levels are associated with decreased mRNA levels, this suggests that high maternal glucose concentrations could reduce materno-fetal cholesterol transfer. Hence, the newborn ABCA1 gene promoter locus might be less methylated in order to increase endogenous HDL-C biogenesis and to overcome the decrease in placental cholesterol transfer. Moreover, the association found between cord blood ABCA1 DNA methylation levels and the fetal-placental weight ratio (a placental efficiency marker representing the feto-placental development occurring throughout pregnancy to meet fetal growth demands⁴⁴) also suggests that the ABCA1 epigenetic profile of newborns might adapt to or influence placental efficiency to ensure an adequate lipid supply. Therefore, as observed on the fetal side of the placenta, newborns’ ABCA1 DNA methylation profile may also serve as a regulatory mechanism to ensure that a sufficient quantity of endogenous and/or exogenous cholesterol is available to the fetus for its optimal growth and development. Nevertheless, the ABCA1 epigenetic “compensatory” signature acquired in utero may have lifelong consequences on newborns’ lipid metabolism possibly triggering their susceptibility to obesity, dyslipidemia, and CVD.²⁶

Interestingly, Tobi et al.²² have observed that prenatal exposure to ending World War II-related famine in the Netherlands was associated with increased DNA methylation levels at the ABCA1 gene promoter locus in leukocytes of adults. This reinforces the potential role of ABCA1 epigenetic adaptations to in utero environment as DNA methylation changes seem to be opposing in conditions where nutrients are restricted (famine) or in excess (hyperglycemia); two conditions associated with a long-term increased susceptibility to CVD.^45-49 Moreover, it suggests that dysregulation (hypo or hypermethylation) of the ABCA1 epigenetic profile could impair ABCA1 function and possibly increase the risk of CVD. Indeed, we have previously reported increased ABCA1 DNA methylation levels in patients with a previous history of coronary artery disease (Guay et al.).²⁶ Further studies will be needed to assess whether epigenetic perturbations of ABCA1 gene observed in newborns might contribute or not to the development of chronic diseases in children and adults, but the current study and the literature provide good evidence in support of the existence of this relationship.

One strength of this study is the longitudinal follow-up of women from first trimester of pregnancy to delivery. It allowed us to accurately identify mothers developing IGT during pregnancy and avoid their misclassification with women who were diabetic pre-pregnancy. Moreover, the study design, along with our fairly large sample size, has allowed us to consider potential confounding factors in our statistical models. ABCA1 was epigenotyped in four different tissues thus providing insights on tissue-specific regulation. It also gives us the overall picture of the feedback regulatory mechanisms that might exist at epigenetic levels between these inter-connected tissues. Nevertheless, as placenta and cord blood samples were analyzed cross-sectionally at birth, the causal variable (placenta or cord blood) in the compensatory mechanisms discussed remains to be identified. Moreover, ABCA1 DNA methylation levels will need to be assessed in these children in a few years to determine methylation stability at this locus and its association with developmental indices and lipid profile throughout infancy and childhood. ABCA1 DNA methylation levels were tested for association with numerous closely related and highly correlated metabolic variables (e.g., HOMA-IR, insulin and glucose; or total cholesterol, HDL-C, LDL-C, and TG). Bonferroni correction for multiple testing would be likely considered too stringent in this context and would thus potentially generates false-negative results.⁵⁰ It remains that some of the reported associations are quite significant (P < 0.01) and that this study is the third one putting forward the association between ABCA1 DNA methylation levels and suboptimal in utero exposure or adverse cardiometabolic outcomes. In addition, as the results observed are associative, further studies (such as the one carried by Kim et al.¹⁴) will need to assess the functional impact of the maternal metabolic profile on placental and newborn methylome.

The current study reports for the first time that maternal metabolic profile is associated with ABCA1 DNA methylation levels in the maternal side of placenta and cord blood. Our results also suggest that ABCA1 DNA methylation profile changes in response to in utero environment and fetal stimuli to ensure optimal cholesterol delivery to the fetus and then cellular intake. These epigenetic adaptations could potentially trigger a long-term susceptibility of the newborn to obesity, dyslipidemia and CVD. This study also suggests that epigenetic modifications are involved in fetal programming and raises the possibility that other gene loci will respond to maternal conditions and could explained the link between suboptimal in utero environment and metabolic health programming of the newborn. Our results thus emphasize the need for the instigation of better maternal care programs early during pregnancy.

Materials and Methods

Sample and clinical data

One hundred women with a singleton pregnancy were selected from a cohort recruited in a founder population of French-Canadian origin from the Saguenay-Lac-Saint-Jean region of Québec (Canada). The women included in this study were aged between 18–40 y, with no personal history of pre-gestational type 1 or 2 diabetes, polycystic ovary syndrome (PCOS), or familial hypercholesterolemia. Women with a positive history of alcohol and/or drug abuse during the current pregnancy were also excluded. All women provided a written informed consent before their inclusion, and all clinical data were de-identified. This project was approved by the Chicoutimi Hospital Ethics Committee, in accordance with the Declaration of Helsinki.

Women were met at the ECOGENE-21 Clinical Research Center at the first (11–14 wk), second (24–28 wk), and third (36–39 wk) trimester of pregnancy. A registered nurse performed a complete anthropometric and metabolic profiling. Anthropometric variables and blood pressure were measured using standardized procedures.⁵¹ Blood samples were obtained after a 12 h overnight fast from the antecubical vein into vacutainer tubes containing EDTA at each gestational trimester and at the time of delivery (cord blood). Plasma cholesterol, TG, and glucose levels were enzymatically measured using a Beckman analyzer (models DXC and LX20). Total cholesterol was determined in plasma and HDL particles after precipitation of very low-density lipoprotein (VLDL) and low-density lipoprotein (LDL) particles (d > 1.006 g/ml) in the infranatant with heparin and MnCl₂. Plasma LDL-cholesterol (LDL-C) levels were estimated using the Friedewald formula.⁵² Insulin measurements were performed using a radioimmunoassay method (Advia Centaur, Simmens). IGT was defined as a 2 h glucose ≥ 7.8 mmol/L following a 75-g OGTT performed at 24–28 weeks’ gestation according to the WHO criteria.⁵³ Insulin resistance levels were assessed using HOMA-IR (fasting glucose [mmol/L] × fasting insulin [mU/L] / 22.5) at the second trimester.⁵⁴ Clinical information about the newborn was collected at birth from medical files and included gestational age, sex of the newborn and anthropometric measurements. The gestational age was calculated at the first visit from the date of the last menstrual period and was corrected afterward, as required, based on the date of the ultrasound scans. Cord blood lipid and glucose levels were assessed as described for the mother. In addition, cord blood C-peptide levels, measured by ELISA, were used as a marker of the newborn’ insulin-secretory activity.⁵⁵

ABCA1 DNA methylation and mRNA measurements

Placenta and cord blood samples were taken within minutes of delivery and placental expulsion (on average < 15 min) by well-trained clinicians (MD). Cord blood samples collected were mixed arterial and venous blood. On the fetal side of the placenta, biopsies were taken near the insertion point of the umbilical cord and consisted of the chorionic plate with the chorionic villi. Maternal placenta biopsies were sampled on the opposite side of the placental cord blood insertion point and were composed of the basal plate and the chorionic villi. The biopsies were washed in PBS 1× to remove cord/maternal blood, dissected to remove conjunctive tissues and kept in RNAlater (Qiagen, #76106) at –80 °C until nucleic acid extraction. Placental DNA and RNA were purified using the All Prep DNA/RNA/Protein Mini Kit (Qiagen, #80004). DNA was purified from whole blood and cord blood samples with the Gentra Puregen Blood Kit (Qiagen, #158389). Maternal blood samples at first trimester were available for 96 of the 100 mothers, whereas 89 cord blood samples were available for DNA methylation analysis. Placental RNA quality was assessed with Agilent 2100 Bioanalyser RNA Nano Chips (Agilent Technologies, #5067–1511). RNA samples on both the maternal and fetal side of the placenta showed good quality with a mean RNA integrity number (RIN) ranging from 6.9 to 9.3 (mean: 8.1 ± 0.7) and 6.5 to 9.5 (mean: 8.2 ± 0.9) respectively.

We used the gold-standard pyrosequencing technology, an accurate and quantitative sequencing assay, to determine base-specific cytosine methylation levels at four different loci within the CpG island of the ABCA1 gene proximal promoter (Fig. 1). The epigenotyped regions were the same as those previously described.²⁶ Pyrosequencing assays combine sodium bisulfite DNA conversion chemistry (EpiTech Bisulfite Kits; Qiagen; #59104), polymerase chain reaction (PCR) amplification (Pyromark PCR Kit; Qiagen; #978703), and sequencing by synthesis assay (Pyromark Gold Q24 Reagents; Qiagen; #978802) of the target sequence. Sodium bisulfite preferentially deaminates unmethylated cytosine residues to thymines (after PCR amplification), whereas methyl cytosines remain unmodified. Each of the pyrosequencing runs performed included a negative PCR control and a sequencing primer control as well as sodium bisulfite conversion controls. Additionally, pyrosequencing quality control (peak height, deviation from reference sequence pattern and unexpected peak height) was assessed for each sample using Pyromark Q24 Analysis Software (v1.0.10.134). Most of the samples were not run in replicates but 100% of the samples passed the pyrosequencing quality control. PCR primers were selected using the PyroMark Assay Design software v2.0.1.15. The PCR and pyrosequencing primers for the amplification of the ABCA1 gene CpG island loci are described in Table S1.

ABCA1 mRNA levels were measured in a randomly selected subsample of 31 placental samples from the maternal and fetal sides. One sample from the fetal side was not available for mRNA quantification. Complementary DNA (cDNA) was generated from placental total RNA using a random primer hexamer provided with the High Capacity cDNA Archive kit (Life Technologies, #4368814). Equal amounts of cDNA were run in duplicate and amplified in a 20 µl reaction containing 10 µL of 2X Universal PCR Master Mix (Life Technologies, #4366072). Primers and Taqman probes were designed to cover exon boundaries (Life Technologies, ABCA1: Hs01059118_m1). Each sample was calibrated to the YWHAZ housekeeping gene (endogenous control; YWHAZ: Hs00237047_m1).^56-58 Relative quantification estimations were performed using an Applied Biosystems 7500 Real Time PCR System as recommended by the manufacturer. ABCA1/GAPDH C_t ratio (1/x) values were used in the analyses.

Statistical analyses

The normal distribution of all variables was assessed using the Kolmogorov-Smirnov test. The following variables were found normally distributed after they were log₁₀-transformed: fasting TG levels, insulin levels and HOMA-IR. Partial Pearson correlation was used to investigate the association between DNA methylation levels and the main variables of interest: glucose concentrations 2 h post-OGTT, fasting glucose and insulin levels, maternal and cord blood lipid profile, as well as ABCA1 mRNA levels. Categorical variables were compared using a Pearson χ²-statistic, whereas mean differences between groups for continuous variables were compared with a Student t-test (NGT vs. IGT) or an ANOVA, followed by the Bonferroni’s post-hoc comparison test (between the HDL-C levels tertiles). The statistical models were controlled for the following potential confounding factors when appropriate: maternal age, BMI, and TG at first trimester and previous history of GDM as well as newborn’s sex and gestational age. Results remained unchanged after consideration for potential confounding factors (age, BMI, TG at first trimester of pregnancy, smoking, gravidity, parity, mode of delivery, newborn sex, history of GDM and weight gain between first and third trimester). Results were considered statistically significant when P values were <0.05 (two-sided).

Stepwise multivariable linear regression was used to identify predictors of the placental and cord blood ABCA1 DNA methylation level variability. The input-independent variables were: maternal characteristics (age, BMI, weight gain between first and third trimester, fasting triglyceridemia and HDL-C at first trimester, 2 h post OGTT glucose at second trimester and blood ABCA1 DNA methylation levels) and fetal characteristics (gestational age, sex, placental weight, cord blood HDL-C and TG levels). All the variables with a P ≤ 0.10 were included in the regression model. All statistical analyses were performed with the IBM SPSS Statistics 20 software (release 20.0.0, SPSS).

Acknowledgments

The authors acknowledge the contribution of Cécilia Légaré, Sébastien Claveau, MSc; Nadia Mior; Jeanine Landry, RN, and Chantal Aubut, RN from the ECOGENE-21 Clinical Research Center for their dedicated work in this study. The authors also express their gratitude to Céline Bélanger, Chicoutimi Hospital, for her thoughtful language revision of the manuscript. Bouchard L and Hivert MF are junior research scholars from the Fonds de la recherche du Québec en santé (FRQS). Baillargeon JP is senior research scholars from the FRQS. Hivert MF is also supported by a Canadian Diabetes Association clinical scientist award. Bouchard L, Hivert MF, Perron P, and Baillargeon JP are members of the FRQS-funded Centre de recherche clinique Étienne-Le Bel (affiliated to the Centre hospitalier universitaire de Sherbrooke). During this research, Guay SP was the recipient of a doctoral research award from the Canadian Institutes for Health and Research (CIHR) and Houde AA was a recipient of a FRQS doctoral training award. Houde AA and Desgagné V were supported by Diabète Québec. This project was supported by ECOGENE-21, the Canadian Institutes of Health Research (CIHR team in community genetics [grant #CTP-82941]), FRSQ, and Diabète Québec.

10.4161/epi.26554

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Authors’ Contributions

Houde AA and Guay SP contributed to the study design, performed the data collection, the data analysis/interpretation and wrote the manuscript. Desgagné V contributed to data collection and revised the manuscript. St-Pierre J, Hivert MF, Baillargeon JP, and Perron P participated in study design conception and revised the manuscript. Gaudet D contributed to the study design, supplied research infrastructure and revised the manuscript. Brisson D and Bouchard L conceived the study design, participated to data analysis/interpretation process and revised the manuscript.

Supplemental Materials

Supplemental materials may be found here: www.landesbioscience.com/journals/epigenetics/article/26554