INTRODUCTIONCongenital heart defects (CHDs) are the most prevalent and serious of birth defects, occurring in 8 to 10 of every 1,000 live births in the United States (Moller et al., 1993). The molecular mechanisms leading to CHDs are complex, and the majority of cardiac malformations that occur in humans are of unknown etiology. Multiple factors including genetic susceptibilities, epigenetic mechanisms, and environmental influences may contribute to the development of non-syndromic CHDs.
Maternal folate deficiency and alterations in one-carbon metabolism have been shown to increase the risk for several birth defects including CHDs (Hobbs et al., 2005; Hobbs et al., 2010; Rosenquist and Finnell, 2001; Shaw et al., 1995; van Driel et al., 2008). During pregnancy there is an increased requirement for folate-dependent nucleotide synthesis and DNA methylation (Oommen et al., 2005), and previous studies suggest that altered DNA methylation may be an underlying mechanism in the development of birth defects (Blom et al., 2006; Li et al., 2005; Okano et al., 1999). Various maternal factors such as diet (Waterland and Jirtle, 2003), genotype (Li et al., 2005), and environmental exposures (Ozanne and Constancia, 2007) have been implicated in abnormal fetal development. These same factors have been shown to affect DNA methylation patterns (Cooney et al., 2002; Candiloro and Dobrovic, 2009; Baccarelli et al., 2009). A combination of maternal genes and environmental exposures may alter the intrauterine environment and thereby modify the fetal phenotype (Furness et al., 2008; MacLennan et al., 2004). In addition to genetic variants, alterations in epigenetic phenomena, such as DNA methylation, could play a crucial role in determining the fetal phenotype. The impact of maternal genetics and epigenetics in the intrauterine environment and subsequent fetal development is not well understood. Little is known about the association between CHDs and maternal DNA methylation despite evidence linking CHDs with maternal folate supplementation and alterations in folate related pathways (Shaw et al., 1995).
DNA methylation is a form of epigenetic gene regulation. The added methyl group can interfere with transcription factor binding, thereby regulating transcription (Comb and Goodman, 1990). Additionally, DNA methylation changes can affect histone modifications and chromatin structure, which can alter gene expression. Hypomethylation of DNA has been reported to contribute to chromosome instability and to alter gene expression, cell differentiation, and apoptosis during embryogenesis (Ehrlich, 2003). Additionally, mutations in the murine DNA methyltransferase (DNMT) gene have been shown to result in embryonic lethality (Li et al., 1992), implicating the importance of DNA methylation in embryogenesis.
The long interspersed nucleotide element (LINE-1) constitutes 17% to 25% of the human genome, with an estimate of over 500,000 LINE-1 elements dispersed throughout the genome (Kazazian, Jr., 2004). Methylation of LINE-1 has been shown to correlate with global DNA methylation (Weisenberger et al., 2005; Yang et al., 2004). Demethylation of these elements, and subsequent activation of retrotransposon activity, can lead to alterations in expression in various genes (Sassaman et al., 1997). LINE-1 methylation of DNA isolated from peripheral blood cells has been shown to be a potential marker of exposure and disease (Bollati et al., 2007; Hsiung et al., 2007).
In the present study we sought to determine if maternal LINE-1 DNA hypomethylation was associated with the occurrence of CHD-affected pregnancies. We also attempted to determine the relationship between LINE-1 methylation and selected metabolites in the homocysteine-methionine pathway.
SUBJECTS AND METHODSStudy populationThe National Birth Defects Research and Prevention Study (NBDPS) is an ongoing multi-site population-based case-control study investigating the etiology of 30 non-syndromic birth defects. The NBDPS is the largest case-control study of birth defects ever conducted in the United States and has been previously described (Yoon et al., 2001). For the current study, case women were Arkansas residents who participated in the NBDPS and who delivered a singleton live birth with a non-syndromic CHD. Cases for whom the pregnancy was also affected by a known single-gene disorder, chromosomal abnormality, or syndrome were excluded from the NBPDS. All diagnostic tests on cardiac NBDPS case infants were reviewed by a pediatric cardiologist to ensure uniform criteria were used for diagnoses. To be eligible for the study described in this manuscript, case infants were required to have at least one cardiac lesion, which included conotruncal, septal, and/or obstructive lesions. Using a classification system developed for the NBDPS, which incorporated three dimensions of cardiac phenotype, cardiac complexity, and extracardiac anomalies (Botto et al., 2007), case women for the current study included those who carried fetuses with simple and association CHDs. Control women were Arkansas residents who had a singleton live birth without birth defects during the same period as the cases, who participated in the NBDPS, and were randomly selected from birth certificate data or hospital discharge logs. Cases and controls spoke either English or Spanish. Blood samples were collected by nurses during home visits in Arkansas that occurred between 2001 and 2008; the research nurses obtained blood samples by venipuncture after acquiring written consent. Blood samples were collected after pregnancy. All participants signed informed consent approved by the University of Arkansas for Medical Sciences Institutional Review Board.
From 548 cases and 243 controls meeting the inclusion criteria, 180 cases and 187 controls were randomly selected using the random number generator in Stata statistical package version 11 (StataCorp, College Station, TX).
CovariatesInformation regarding selected lifestyle factors was obtained from NBDPS-structured computer-assisted telephone interviews and by in-home interviews conducted at the time of blood draw by a research nurse using a Block Abbreviated Food-Frequency Questionnaire (Block et al., 1990). The information on covariates reflects up to one month prior to the visit. Covariates were selected based on their potential effects on DNA methylation status and the occurrence of heart defects; covariates included were age, race, body mass index (BMI), and the use of multivitamins, cigarettes, and alcohol.
Sample preparation and processingFasting blood samples were collected in EDTA-Vacutainer tubes and immediately chilled on ice before they were centrifuged at 4,000 × g for 10 minutes at 4°C to obtain blood plasma. The blood cell pellets and plasma samples were stored at −80°C until DNA isolation and high performance liquid chromatography (HPLC) analyses.
DNA isolation and bisulfite treatmentDNA was isolated from frozen blood samples according to the manufacturer’s protocol for the PureGene DNA isolation kit (Gentra Systems, Minneapolis, MN). Genomic DNA was quantified using the ABI TaqMan RNase P Quantification Kit (Applied Biosystems, Forest City, CA) according to the manufacturer’s protocol. Using the EZ DNA Methylation-Direct Kit (Zymo Research, Orange Country, CA), 500 ng of genomic DNA underwent sodium bisulfite modification. The converted DNA was resuspended in 20 µl of TE buffer and stored at −80°C until the samples were ready for analysis.
DNA methylation assessment with MethylightMethylight methodology provides a quantitative methylation-specific PCR assay that allows rapid analysis of methylated and unmethylated alleles via fluorescent detection. The merits of this technology have been well-characterized (Weisenberger et al., 2005; Ogino et al., 2006; Eads et al., 2000). This process uses bisulfite treatment of DNA, which converts unmethylated cytosines to uracils, leaving the methylated cytosines protected from this conversion (Frommer et al., 1992). Since the primers lack CpG sites, the methylated and unmethylated primer binding sites will be identical after bisulfite conversion. The specificity of the assay occurs during probe hybridization by the use of separately labeled probes. The improved sequence specificity allows relative quantification of methylated and unmethylated alleles in a genomic DNA sample (Zeschnigk et al., 2004). The VIC-labeled probe hybridizes to the sequence of the methylated LINE-1 allele, and the FAM-labeled probe hybridizes to the sequence of the unmethylated LINE-1 allele. Once the probes bind to their selected sites, they are cleaved by Taq polymerase during the PCR reaction, separating the fluorescent tag from the quencher. The amount of resulting fluorescence detected, either VIC or FAM, is directly proportional to the amount of PCR product generated from methylated or unmethylated DNA, respectively (Eads et al., 2000). The probe sequences cover multiple CpG sites within the LINE-1 repetitive sequences, which reflect genome-wide methylation density.
Generation of methylated and unmethylated DNA standardsMethylated and unmethylated standards were generated in order to test the specificity of the PCR reactions. Fully methylated DNA was generated from DNA provided by healthy volunteers. The DNA was incubated with M.SssI enzyme (New England Biolabs, Ipswich, MA) at a concentration of 1U/µg of DNA and with 0.16 mM of SAM overnight at 37°C. Extra SAM and M.SssI enzyme was added at the same concentrations for a second overnight incubation at 37°C. The samples were stored at −20°C. To generate unmethylated DNA, healthy volunteer DNA was amplified using the Sigma Whole Genome Amplification (WGA2) kit (Sigma-Aldrich, St. Louis, MO). The manufacturer’s protocol was used. The product was stored at −20°C. The mixing of methylated and unmethylated DNA as standards for DNA methylation assays has been used successfully by other research groups (Richards, 2006; Yideng et al., 2007).
Real Time Methylation-Specific PCR (Methylight)PCR reactions were performed with a final reaction volume of 25 µl in 96-well sealed plates on the ABI 7900 HT Real Time PCR System (Applied Biosystems, Forest City, CA). Samples contained 12.5 µl of TaqMan Universal Master Mix, 2 µl of bisulfite-treated DNA, 2.5 µM of each of the primers, and 150 nM of TaqMan MGB probes corresponding to the methylated and unmethylated LINE -1 sequence after bisulfite treatment. Real-Time PCR reactions were performed simultaneously on the same 96-well plate. The cycle conditions for the LINE-1 assay were as follows: Initial denaturation at 95°C for 10 minutes to activate Taq polymerase was followed by 50 cycles of denaturation at 94°C for 15 seconds followed by 60°C for 1.5 minutes. Equal PCR efficiency between plates was ensured by including a duplicated sample that contained equal amounts of methylated and unmethylated DNA. The primer and probe sequences are as follows: Line-1fw, 5′–TTATTAGGGAGTGTTAGATAGTGGG–3′;
Line-1rev, 5′–CCTCTAAACCAAATATAAAAT ATAATCTC–3′;
Line-1met, VIC–TACTTCGACTCGCGCAC GATA-BHQ–3′;
Line-1unmet, 6FAM–CCTACTTCAACTCACACA-BHQ–3′.
Generation of percent DNA methylation valuesThe amount of methylated and unmethylated DNA was calculated for each sample by interpolation of a standard curve that was established for both methylated and unmethylated DNA. Serial dilutions of methylated and unmethylated DNA were prepared in order to ensure that all sample values could be interpolated from the standard curve. The percent methylation was calculated by dividing the amount of methylated DNA by the amount of total DNA (methylated plus unmethylated DNA). Each sample was assayed in duplicate and the replicates achieved strong correlation (r=0.84).
Biomarker measurementsTotal plasma homocysteine, methionine, SAM, and SAH were determined by HPLC as described elsewhere by our research group (Hobbs et al., 2005). The methodologic details for metabolite elution and electrochemical detection were described previously (Melnyk et al., 1999; Melnyk et al., 2000). Plasma folic acid concentrations were measured by using Quantaphase II radioimmunoassay kit according to the manufacturer’s protocol (Biorad Laboratories, Hercules, CA).
Statistical analysisSelected characteristics of case and control subjects were compared with a Fisher’s exact test for categorical variables. LINE-1 methylation values for cases and controls were compared using a non-parametric two-sample Wilcoxon rank-sum test. The Wilcoxon rank-sum test was also used to compare LINE-1 methylation between smokers and non-smokers, and between those with normal BMIs (18.5–24.9) and obese individuals (≥30). Multivariable logistic regression was used to compute odds ratios and 95% confidence intervals for the association between LINE-1 methylation and case-control status while adjusting for age, race, vitamin supplementation, smoking, alcohol consumption, and BMI. The frequency distributions of LINE-1 methylation values for cases and controls were graphically examined and compared using a quantile-quantile (Q-Q) plot and a frequency histogram. If the distribution of methylation values was similar for cases and controls, the Q-Q plot would closely follow the diagonal line x=y, whereas deviations from the diagonal line indicate different distributions of methylation values between cases and controls. The relationships between LINE-1 DNA methylation and selected homocysteine-methionine pathway biomarkers were determined by computing Spearman’s partial correlation coefficient after covariate adjustments. The comparison between cases and controls for selected biomarkers was conducted using a multiple linear regression model adjusting for the covariates previously listed.
Analyses were performed using the Stata statistical package version 11 (StataCorp, College Station, TX).
RESULTSStudy populationLINE-1 DNA methylation was measured in 367 participants (180 cases and 187 controls). Table 1 presents selected characteristics of study participants. The majority of cases (61.7%) and controls (60.4%) were less than 30 years old. The study population consisted mostly of Caucasian women. The frequencies of most characteristics were similar between cases and controls. However, smoking was significantly more prevalent in cases (30.0%) than in controls (18.7%; p=0.015), and the median time interval between the end of pregnancy and blood draw was significantly longer for controls (16.5 months; range = 0.13–65.3 months) than for cases (15.4 months; range = 4.3–63.3 months; p=0.035). Additional analyses to determine if smoking and the interval between end of pregnancy and blood draw were confounders was conducted. Specifically, neither smoking nor time of blood draw was associated with LINE-1 methylation status in either cases, controls, or both (data not shown).
Selected metabolite analysisPlasma concentrations of selected metabolites involved in one-carbon metabolism are summarized in Table 2. As we have previously reported, we observed significant differences between cases and controls for the biomarkers analyzed after adjusting for multiple covariates. When compared to controls, cases had higher mean concentrations of homocysteine and SAH and lower mean concentrations of methionine, SAM, the SAM/SAH ratio, and folate. The metabolite concentrations in the cases were all reflective of a decreased cellular methylation capacity.
LINE-1 DNA methylation analysisTable 3 presents summary statistics of LINE-1 DNA methylation in cases and controls. DNA methylation was significantly lower (hypomethylation) in cases when compared to controls (p=0.049). To control for potential confounders, we constructed a model to adjust for age, race, body mass index (BMI), and the use of multivitamins, cigarettes, and alcohol. The significant association between LINE-1 methylation and case/control status remained after covariate adjustment (p=0.010).
To further explore the relationship between LINE-1 methylation and CHDs in our population, we conducted analyses stratified for smoking and obesity. Smoking was more prevalent among cases than controls, and thus we stratified our analyses by smoking status. We did not observe a significant relationship among LINE-1 methylation and smoking status (p=0.974). We also compared LINE-1 methylation among individuals with a normal BMI and obese individuals, but the result was not statistically significant (p=0.557).
To determine if the associations between DNA methylation and cardiac defects were related to specific cardiac phenotypes, an analysis stratified by cardiac phenotypes was conducted. Cardiac phenotypes included conotruncal, septal, and obstructive defects. Cases that had more than one cardiac defect type were included in all relevant strata. Due to a small number of cases, the 14 cases that did not fall within one of the main defect groups were excluded from the stratified analysis. Among cases, 47.2% had atrial or ventricular septal defects, 35.4% had right- and left-sided obstructive defects, and 16.7% had conotruncal defects.
After covariate adjustment, LINE-1 methylation was significantly lower among women with pregnancies affected by septal defects (mean=81.26, SD=11.44) compared to controls (mean=83.64, SD=10.16; p=0.043). LINE-1 methylation was lower among conotruncal and obstructive cases compared to controls (p=0.085 and 0.086, respectively), but these results did not achieve statistical significance. LINE-1 methylation was lowest among conotruncal cases (mean=80.0%, SD=11.2). Because LINE-1 methylation did not deviate significantly between cardiac phenotypes, cases were combined for subsequent analyses to maximize power to test study hypotheses. A Q-Q plot and histograms were created to examine and compare the frequency distributions of LINE-1 methylation in cases and controls (Figure 1). Both the Q-Q plot and the histogram revealed marked deviations in methylation between cases and controls at the lower values of the distribution.
Based on these plots, LINE-1 DNA methylation was further compared between cases and controls by computing adjusted odds ratios and 95% confidence intervals at various cutoffs on the basis of methylation levels within controls (Table 4).
Of the 180 cases, 55 (30.6%) cases had with LINE-1 methylation values below the 20th percentile of controls and 33 (18.3%) cases had LINE-1 methylation values below the 10th percentile of controls, indicating a shift to lower values in LINE-1 methylation in cases relative to controls. Women in this bottom decile were almost twice as likely to have a CHD-affected pregnancy compared to women above this decile (OR= 1.91; 95% CI: 1.03, 3.58).
These findings suggested that a lower DNA methylation status may be associated with an increased risk of CHDs.
LINE-1 methylation and selected metabolites correlationThe correlations between LINE-1 DNA methylation and selected plasma metabolites were computed to determine if metabolites were predictive of DNA methylation status among cases and controls (Table 5). Homocysteine had the strongest correlation with LINE-1 methylation. As LINE-1 DNA methylation decreased, total homocysteine concentration increased, resulting in a negative correlation (Spearman‘s r=−0.135, p=0.013). A separate model for cases and controls correlating LINE-1 and homocysteine was computed for cases and controls. The slopes from each model were not significantly different (data not shown). Concentrations of methionine, s-adenosylmethionine (SAM), s-adenosylhomocysteine (SAH), and folate were not significantly correlated with LINE-1 methylation status.
DISCUSSIONTo our knowledge, our findings represent the first published study in which DNA methylation of LINE-1 was compared between mothers of CHD-affected pregnancies and controls. DNA hypomethylation of the LINE-1 repetitive element was observed in case mothers when compared to control mothers. Placement in the lowest deciles of methylation resulted in an increased risk of CHD-affected pregnancies. LINE-1 DNA methylation status was not strongly correlated with selected methionine pathway plasma metabolites.
The magnitude of difference between case and control LINE-1 methylation was relatively small compared to differences reported in other disease models, for example, between normal and tumor tissue samples (Iacopetta et al., 2007). However, a small magnitude of difference in peripheral blood LINE-1 methylation was observed in a previous case/control study (Hsiung et al., 2007). During pregnancy, increases in cellular proliferation and one-carbon metabolism are required for uterine enlargement, expansion of blood volume, placental development, and fetal growth (Bruinse and van den, 1995; Tamura and Picciano, 2006). For the developing embryo, the increased demand for methyl groups for nucleotide synthesis and DNA methylation may result in a small threshold between sufficient and insufficient methylation capacity. Insufficient methylation capacity may cause deleterious changes in developmental programming resulting in CHDs and other birth defects.
LINE-1 methylation has shown to be significantly correlated with global methylation levels measured by HPLC in cancer models. Additionally, demethylation of repetitive elements can ultimately lead to aberrant expression in other genes (Asada et al., 2006). It has been suggested that hypomethylation of repetitive elements may influence genomic instability and may consequently cause genome instability in the progeny (Filkowski et al., 2010). Although speculation regarding transgenerational epigenetic inheritance exists, the majority of evidence exists in animal models (Morgan and Whitelaw, 2008). Correlations between parent-child pairs in LINE-1 methylation suggest the possibility of transgenerational epigenetic inheritance being associated with disease risk (Mirabello et al., 2010).
A decrease in the incidence of CHDs and neural tube defects (NTDs) can be achieved with preconception folic acid supplementation (Hobbs et al., 2010). The shared biological sensitivity for these separate anatomical sites is thought to involve their common neural epithelial cell origin (Rosenquist and Finnell, 2001). Thus, CHDs and NTDs may share multiple similar pathophysiological pathways. It has been postulated that inhibition of the methylation cycle may contribute to the development of NTDs through the disruption of crucial reactions involving the methylation of DNA and other biomolecules (Dunlevy et al., 2006). Recent studies have shown an increased risk of NTDs with decreased LINE-1 and global methylation status (Chen et al., 2010; Wang et al., 2010). These studies provide evidence that genomic methylation may play a role in the development of NTDs. Additional studies to confirm the role of DNA methylation in CHD risk should be pursued as well.
The LINE-1 repetitive element has been used as a measure for genomic methylation status for methylation-specific PCR assays (Hsiung et al., 2007; Iacopetta et al., 2007; Yideng et al., 2007) and widely used in pyrosequencing assays (Wang et al., 2010; Figueiredo et al., 2009; Bollati et al., 2007). LINE-1 methylation patterns appear to be tissue-specific and vary depending on cell type (Chalitchagorn et al., 2004), thus the values obtained from methylation studies may vary greatly depending on the processed biological sample and the assay performed. The range of LINE-1 methylation values presented in this study are consistent with two previously published studies assessing LINE-1 methylation (Iacopetta et al., 2007; Wilhelm et al., 2010). We observed a large proportion of LINE-1 methylation values greater than 80%, and this observation is in agreement with other published studies (Iacopetta et al., 2007; Jin et al., 2009; Richards et al., 2009).
An inverse correlation between cord plasma homocysteine and LINE-1 methylation in humans has been recently shown (Fryer et al., 2009). We also observed a modest inverse correlation between LINE-1 methylation and plasma homocysteine. The lack of a strong correlation between LINE-1 methylation and selected biomarkers may be explained by the hypothesis proposed by Kim (Kim, 2005) stating that the effects of folate deficiency and the supply of DNA methyl groups may be gene- and site-specific and dependent on the cell type, organ, stage of transformation, and degree and duration of folate depletion. Previous studies have also shown a lack of correlation between LINE-1 methylation and methionine pathway metabolites (Figueiredo et al., 2009; Wang et al., 2010). Since folate pathway metabolites lacked strong correlation with LINE-1 methylation, we postulate that the genomic methylation differences observed may be due to other unmeasured factors. Single nucleotide polymorphisms, additional metabolites, and environmental exposures are factors that may explain the alterations in DNA methylation observed in the study.
Certain methodological limitations of this study should be considered. Subjects were enrolled in a case-control study after completion of pregnancy. Given the rare prevalence of CHDs, the resources and sample size required to conduct a cohort study among women enrolled before conception and followed to the completion of pregnancy are enormous; thus, case-control design is the preferred epidemiological design for rare outcomes. Thus, we did not measure DNA methylation changes during pregnancy and fetal organogenesis, when changes in methylation status would be expected to have the greatest effect. We have previously stated that the plasma biomarker concentrations in our study design may not represent biomarker concentrations at the time of organogenesis, but may identify women who have persistent metabolic imbalances (Hobbs et al., 2005). It has been shown that homocysteine concentrations and dietary patterns remain relatively stable from preconception throughout pregnancy and postpartum (Cikot et al., 2001; Cuco et al., 2006). We found no correlation between maternal age and LINE-1 methylation (data not shown). Thus, LINE-1 methylation observed after pregnancy may reflect stable adult profiles that were present at the time of conception. Another limitation of our experimental design is that only allowed the analysis of only fully methylated and unmethylated alleles in the LINE-1 repetitive element. If partial methylation occurs at the level of probe discrimination, the sequence derived after bisulfite conversion is recognized by neither of the two probes. Thus, partial methylation may affect the ability of the assay to detect global methylation density via the amplification of the LINE-1 consensus sequence. This design is in contrast to other methods that measure the methylation level at several single CpG sites. An advantage of Methylight is its high sensitivity and low input DNA requirement, but it is prone to higher variability that other methods (Aparicio et al., 2009).
To our knowledge, our study is the first to measure maternal DNA methylation in a case-control study of CHDs. The study has notable strengths including the analysis of DNA methylation in a relatively large population of mothers in which information regarding various lifestyle factors and plasma metabolite concentrations were available. Our study design allowed the assessment of the relationship between LINE-1 DNA methylation and the risk of CHDs. Our findings indicate maternal LINE-1 hypomethylation is associated with an increased risk of CHD-affected pregnancies. Alterations in maternal DNA methylation converge with evidence from previous studies that indicate folate-dependent genetic and metabolic susceptibilities increase the risk of CHDs. These results suggest that future studies involving repetitive, global, and gene-specific maternal DNA methylation and CHDs are warranted. The ability to possibly restore aberrant DNA methylation to normal levels via nutritional intervention makes DNA methylation an attractive potential therapeutic target. Identification of genetic and epigenetic profiles associated with an increased risk of CHDs may add new dimensions to assessing preconception risk and may prove to be instrumental in elucidating the mechanisms that underlie the development of CHDs.