The causal effect of red blood cell folate on genome-wide methylation in cord blood: a Mendelian randomization approach
© Binder and Michels; licensee BioMed Central Ltd. 2013
Received: 16 October 2012
Accepted: 19 November 2013
Published: 4 December 2013
Investigation of the biological mechanism by which folate acts to affect fetal development can inform appraisal of expected benefits and risk management. This research is ethically imperative given the ubiquity of folic acid fortified products in the US. Considering that folate is an essential component in the one-carbon metabolism pathway that provides methyl groups for DNA methylation, epigenetic modifications provide a putative molecular mechanism mediating the effect of folic acid supplementation on neonatal and pediatric outcomes.
In this study we use a Mendelian Randomization Unnecessary approach to assess the effect of red blood cell (RBC) folate on genome-wide DNA methylation in cord blood. Site-specific CpG methylation within the proximal promoter regions of approximately 14,500 genes was analyzed using the Illumina Infinium Human Methylation27 Bead Chip for 50 infants from the Epigenetic Birth Cohort at Brigham and Women’s Hospital in Boston. Using methylenetetrahydrofolate reductase genotype as the instrument, the Mendelian Randomization approach identified 7 CpG loci with a significant (mostly positive) association between RBC folate and methylation level. Among the genes in closest proximity to this significant subset of CpG loci, several enriched biologic processes were involved in nucleic acid transport and metabolic processing. Compared to the standard ordinary least squares regression method, our estimates were demonstrated to be more robust to unmeasured confounding.
To the authors’ knowledge, this is the largest genome-wide analysis of the effects of folate on methylation pattern, and the first to employ Mendelian Randomization to assess the effects of an exposure on epigenetic modifications. These results can help guide future analyses of the causal effects of periconceptional folate levels on candidate pathways.
Identification of the protective effect of periconceptional folic acid supplementation against neural tube defects in neonates led to the mandated fortification of flours and other grain products in several countries [1-4]. In addition to the prevention of neural tube defects, folic acid supplementation has been associated with decreased risk of other congenital malformations, such as heart defects and oral clefts [5-8]. Despite these benefits, concern has been raised to possible adverse effects. In mouse models, maternal methyl donor supplementation was associated with increased risk of allergic airway disease [9, 10]. However, subsequent human studies of possible detrimental effects have been relatively inconclusive [11-14]. An understanding of the biological mechanism by which folate acts to affect fetal development can inform appraisal of expected benefits and risk management, and is ethically imperative due to the ubiquity of fortified foods. Folate, an essential vitamin that can be obtained from diet and synthetic supplements, is an important component in the one-carbon metabolism that frees up methyl goups for DNA methylation. Thus, epigenetic modifications provide a putative molecular mechanism mediating the effect of folic acid supplementation on neonatal and pediatric outcomes. Prior observational studies have reported inconsistent associations between maternal folic acid supplementation and maternal folate levels with infant DNA methylation, specifically among imprinted genes [15-17]. One difficulty when studying the association between folic acid supplementation and DNA methylation is the possibility of effects being obfuscated by the influence of diet on total maternal folate levels contributing to dose misclassification. Analyses of the effects of folate levels may also be biased due to the unmeasured confounding by multifaceted environmental exposures associated with socio-economic status that may also influence epigenomic profile. Therefore new approaches are necessary to consistently estimate the causal effect of folate on DNA methylation.
Where Z is our instrument, A is our exposure of interest, and Y is our outcome. In economic literature, Z is referred to the exogenous variable, i.e. explained by variables outside the model, and A is the endogenous variable, explained by other variable in the model. In the first stage (1), the exposure is regressed on the instruments. The second stage (2) regresses the outcome on the fitted values () from the first stage. Given the instrument Z meets the conditions outlined above, the parameter estimate from fitting model (2) will provide a consistent estimate of the causal effect of our exposure on our outcome. Similar to the other techniques to control for confounding in observation studies, these assumptions are unverifiable, but reasonable when the instrument is a genetic polymorphism. Using a genetic variant as the instrument, also known as Mendelian Randomization, is an appealing approach to establish temporality and due to the lack of common causes of the instrument and the outcome aside from population stratification [21-23]. Although the instrumental variable estimate will be asymptotically unbiased, in finite samples the instrumental variable estimates will be biased towards the observed confounded association. This bias arises because the true relationship between the instrument and the exposure in the first stage of the analysis is unknown and it must be estimated, resulting in model over-fitting. The magnitude of this bias depends on the strength of the association between the instrument and the exposure [19, 24]. Weak instrument bias, which is often a concern for Mendelian Randomization studies, can be minimized and precision increased by including measured confounders in the two-stage analysis .
Using common methylenetetrahydrofolate reductase (MTHFR) polymorphisms as our instrument, Mendelian Randomization provides one method to assess the causal effect of maternal folate on epigenetic profile. MTHFR catalyzes the synthesis of 5-methylTHF, which is the coenzyme required for homocysteine remethylation to methionine, the precursor for the DNA methylating agent S-adenosylmethionine. Two common polymorphism, C677T and A1298C, are associated with reduced MTHFR enzymatic activity, resulting in higher homocysteine levels [26-28]. In this study we used a Mendelian Randomization approach to assess the effect of red blood cell (RBC) folate on genome-wide DNA methylation in cord blood. The mother’s MTHFR genotype was utilized as our instrument, given the efficiency of maternal folate metabolism would be expected to modify developmental exposure. A long-term measure of folate intake, RBC folate has been demonstrated to be responsive and sensitive to inter-individual differences in controlled folate intake [29-31]. In a prior study of the association between of maternal and newborn folate status, early maternal plasma folate (~18 weeks) and self-reported folic acid supplementation were found to be significantly associated with cord blood folate levels . In our study population of 50 newborns, site-specific CpG methylation within the proximal promoter regions of approximately 14,500 genes was analyzed using the Illumina Infinium Human Methylation27 Bead Chip. With the Mendelian Randomization approach we were able to identify the causal effects of folate on epigenetic modifications that would have been substantially biased given a standard regression analysis. The possible utility of Mendelian Randomization to investigate the causal structure of disease etiology mediated by epigenetic modifications in observational studies has been discussed by others [33-36], but this is the first study to apply this approach.
The epigenetic birth cohort
Data and biospecimens were collected between June 2007 and June 2009 on the labor and delivery floor of the Department of Obstetrics, Gynecology and Reproductive Biology at Brigham and Women’s Hospital in Boston as previously described . Briefly, pregnant women were invited to participate in our study, and 1941 completed a questionnaire, and agreed to donate placenta and cord blood samples. Maternal and cord blood collected from the base of the umbilical cord was stored in EDTA tubes. Blood was processed immediately and buffy coats were stored at -80°C. From this cohort, a subset of 50 cord blood samples were analyzed for genome-wide DNA methylation associated with RBC folate. The study protocol was approved by the Institutional Review Board of the Brigham and Women’s Hospital.
Genome-wide methylation analysis
For the current study, 1 μg of cord buffy coat DNA from each of the 50 individuals was processed at the USC Epigenome Center (Los Angeles, CA, USA) as previously described . For comprehensive analysis of genome-wide methylation, the Illumina Infinium Human Methylation27 Bead Chip was used to simultaneously interrogate methylation at 27,578 CpG sites, spanning 14,495 RefSeq genes. On average, the Infinium array targets 2 CpG sites per gene, with higher coverage (3-20 CpGs) for cancer-related and imprinted genes. Data was assembled at the Epigenome Center by converting fluorescence intensities from methylated (M) and unmethylated (U) alleles to methylation level given by β = (M)/(M + U + 100). If signal intensity was not significantly different from background measurements, the β-value was recorded “NA.” The statistical analysis was restricted to autosomal CpGs with unique probe target sequences. The 50-mer oligonucleotide probes were removed from further analysis if the probe had (a) cross-reactive target(s) with at least 40 matching bases, at least 90% identity, end-nucleotide match, and gapless sequence alignment against the target sequence . Further restriction to CpG sites with no missing data reduced the data analysis from 23,682 to 23,264 autosomal CpGs. Methylation level for these loci was square-root arcsine transformed to stabilize the variance. Transformed loci that were not normally distributed at the 0.05 α-level were then removed from the dataset (Shapiro-Wilk test). In total, 16,989 loci were analyzed in subsequent statistical analyses.
Genotyping of the maternal MTHFR C677T (rs1801133) and A1298C (rs1801131) SNPs was performed on the ABI PRISMs 7300 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). Primers and probes were ordered from TaqMan® SNP Genotyping Assays (Applied Biosystems) MTHFR C677T (assay ID: C___1202883) and A1298C (assay ID: C____850486). The PCR was performed in 25 ul, with each reaction containing 25 ng gDNA, 1.25 ul of 20× Assay Mix, 12.5 ul of TaqMan genotyping master mix and q.s. with PCR grade water. Cycling conditions were as follows: 50°C for 2 min, 95°C for 10 min, and 40 cycles of 92°C for 15 s and 60°C for 1 min. After the amplification, plates were scanned by the ABI PRISMs 7500 PCR system to determine genotypes by allelic discrimination. Hardy-Weinberg assumptions were assessed using a Chi-square test.
Red blood cell folate
Cord blood RBC folate was measured on the Roche E Modular system (Roche Diagnostics, Indianapolis, IN) in the laboratory of Dr. Nader Rifai at Children”s Hopital in Boston, MA, USA. Serum and plasma folate are sensitive to day-to-day variation in intake, reflecting short-term diet [40, 41]. We chose to assay RBC folate given is more stable marker of long-term patterns . Red blood cells were first lysed with ascorbic acid, and folate was then measured on the hemolysate. Hemoglobin was also measured on this hemolysate to standardize RBC folate per gram of hemoglobin. The sample was treated with monothioglycerol and sodium hydroxide to release the folate from endogenous binding proteins, and incubated with a ruthenium labeled folate binding protein, forming a folate complex. Biotinylated folate and streptavidin-coated magnetic microparticles were then added to the reaction mixture. Ruthenium labeled biotin complexes bound to the magnetic microparticles, and unbound reagents and sample were washed away. A chemiluminescent reaction was electrically stimulated to generate light, the intensity being indirectly proportional to the amount of folate present in the sample. This assay is approved by the Food and Drug Administration for clinical use. The lowest detection limit of this assay is 0.6 ng/mL, and the day-to-day imprecision values at concentrations of 7.6, 14.3 and 19.2 ng/mL are 3.9, 3.1 and 2.0%, respectively. The normal range is 3.1 to 17.5 ng/mL. The positively skewed RBC folate measurements were log transformed for subsequent statistical analyses. In this cohort, log RBC folate ranged from 6.13 to 7.54 log (ng/mL), with a median level of 6.664 log (ng/mL).
To estimate the effect of RBC folate on DNA methylation, we exploited Mendelian Randomization methods using the two-stage least squares (TSLS) approach. In the first stage, log transformed RBC folate was regressed against MTHFR genotype modeled additively. Predicted values from this first stage were then used to model square-root arcsine transformed methylation levels. An indicator for whether conception was planned, a putative confounder, was included in both stages to increase precision and decrease weak instrument bias. The effect of RBC folate on methylation level using the TSLS approach was estimated for each CpG locus independently using the AER package in R . Among the sites that had significant changes in methylation at the 0.05 α-level, the effect of log transformed RBC folate on untransformed β-values was estimated using TSLS. While this yielded valid, more interpretable parameter estimates, inference was biased due to the violation of the ordinary least squares normality assumptions. Therefore robust 95% confidence intervals for these estimates were generated by bootstrapping 1000 replicates. The UCSC Genome Browser database was used to characterize the location of the significant CpG sites in relation to annotated features and CpG islands . Functional enrichment among the genes in closest proximity to the significant sites compared to all Gene Ontology (GO) annotated genes represented on the Illumina microarray was evaluated based on biological process using the GOstats package provided by Bioconductor . Overrepresentation in the data set was assessed using a conditional Hypergeometric test, which considers the relationships between the GO terms and conditions on the significant child terms.
In both the exposure and outcome models, the confounder u i and error term were independently normally distributed. Parameters and estimated from each significant TSLS model were assumed to be the true population values, and were used to create the simulated outcome and exposure values. Genotype, planned vs. unplanned conception, and RBC folate corresponded to the observed values of each individual. The distribution of effect estimates were assessed among the 10,000 simulated datasets using Mendelian Randomization and ordinary least squares approaches adjusting for conception, for α c = β c = 0.1 and 0.2. All statistical analyses were conducted using R version 2.15.2.
Four of CpG loci indicated to be significantly associated with RBC folate by the MR approach were selected for verification by pyrosequencing, chosen to reflect varying effect sizes. These regions were in proximity to the transcription start site of four genes: AIRE, GPR12, OBFC2B, and SMG6. Pyrosequencing of the regions surrounding these significant loci was performed by a commercial service (EpigenDx, Worcester, MA) on a subset of 38 samples in duplicate. EpigenDx performed the bisulfite conversion, designed the assays, and pyrosequenced the regions, including high and low methylation controls. The association between methylation and folate level at each locus interrogated by pyrosequencing was estimated by MR as previously described. Regional changes were visualized by a loess curve of the site-specific estimates plotted against genomic location.
Results and discussion
The global minor allele frequencies estimated in the 1000 Genome phase 1 population are approximately 33% for C677T (rs1801133) and 23% for the A1298C (rs1801131). Both SNPs were in Hardy-Weinberg equilibrium among our study population (p = 0.129; p = 0.433). Except for a very rare cis 677 T/1298C haplotype found in some parts of the United Kingdom and Canada, most 677 T and 1298C alleles are associated with 1298A and 677C alleles, respectively . Thus modeling MTHFR genotype additively corresponded to modeling the risk associated with having zero, one, or two common trans MTHFR variants. Modeling the number of variants co-dominantly did not significantly explain more variation in log transformed RBC folate than the additive model (F = 1.335, p = 0.254).
Genes in proximity to loci with a significant change in methylation levels per one unit change in log (RBC folate)
UCSC annotated features
Bootstrap 95% CI
5′UTR; 1st Exon; Enhancer
Within 100 bp of TSS
Within 500 bp of TSS; 5′UTR; 5′UTR
Body; Within 1000 bp of TSS
Within 500 bp of TSS; 5′UTR
Biological process Gene Ontology (GO) enrichment among the 7 GO annotated genes in closest proximity to the loci significantly associated with RBC folate-level
Gene ontology term
Number in significant subset
Number in all GO annotated genes analyzed
Regulation of dephosphorylation
Nuclear-transcribed mRNA catabolic process, nonsense-mediated decay
mRNA export from nucleus
mRNA catabolic process
Humoral immune response
Nucleobase, nucleoside, nucleotide and nucleic acid transport
G-protein coupled receptor protein signaling pathway
Genes in proximity to the 10 loci with the most significant change in square-root arcsine methylation levels per one unit increase in log(RBC folate) using the ordinary least squares approach
UCSC annotated features
Within 1000 bp of TSS; 3′UTR; Enhancer
Within 100 bp of TSS
Within 1000 bp of TSS
Within 1500 bp of TSS
Within 100 bp of TSS
Within 500 bp of TSS
DNA methylation has become increasing integrated into public health studies as a modifiable indicator of the underlying biologic changes mediating the effects of endogenous and exogenous exposures on subsequent disease risk. However, the field of epigenetic epidemiology is still in its infancy and the effects of many exposures on methylation profile have yet to be explored and verified. In observational studies, the validity of estimated associations is always susceptible to criticism, given the possibility of residual or unmeasured confounding. Due to an array of factors that may influence both methylation and folate levels, inconsistencies among previously reported associations between maternal folate intake and neonate methylation patterns may be the result of bias. Using an instrumental variable approach, we were able to estimate the causal effect of cord RBC folate on DNA methylation level across the epigenome. While instrumental variable analysis also requires strong unverifiable assumptions, these assumptions are more cogent when genetic polymorphisms are employed as the instrument. Several papers have outlined the potential utility of a Mendelian Randomization approach in the context of epigenetic epidemiology [34, 36]. These results demonstrate in practice, Mendelian Randomization can be a robust method to assess epigenetic modifications compared to the standard ordinary least squares approach. The cost of this improved internal validity is decreased precision. The reduction in power associated with the two-stage estimation precluded detection of significant changes in methylation level after correcting for multiple testing. By accounting for both measured and unmeasured common causes of exposure and outcome, we assert the loci with the most significant associations identified using the MR approach represent the most appropriate candidates for validation. Due to the use of a weak instrument, there is the possibility of residual bias in the direction of the confounded relationship . However, simulation studies have demonstrated that even weak instruments substantially decrease the bias relative to the confounded association [25, 51]. Our sensitivity analyses similarly illustrated this robustness in the context of epigenetic modifications. An additional drawback to this MR analysis is the possibility of over-fitting the data by estimating the association between the genetic variant and intermediate phenotype in the same cohort. To reduce this potential bias, future studies applying MR may consider a two-sample approach, using an external dataset to estimate the instrument-exposure association. However, the application of the two-sample approach is dependent on an additional, generally unverifiable assumption that the independent cohorts are drawn from the same underlying population. Building off the ‘Causality Equivalence Theorem’ presented by Chen , another recently suggested method to infer causal indirect effects of genotype on outcome relies on a series of models to statistically test necessary conditional independencies between covariates . However, the application of this approach was deemed inappropriate given the associations between RBC folate and methylation level conditional on MTHFR genotype were likely influenced by common unmeasured causes of folate and methylation levels.
Although the Mendelian Randomization method provides valid estimates of the causal effect of RBC folate on methylation level across the genome, there is some uncertainty as to the functional implications. A majority of the significant loci were within the 5′ UTR of RefSeq annotated genes, while a few were located within the gene body or associated with a regulatory element, for which methylation may have disparate effects on transcriptional regulation. However, future studies can utilize these results to guide investigation of potential pathways mediating the influence of folate levels on developmental outcomes. These results should also be considered in the context of the samples analyzed, considering even closely related cell types, such as haematopoietic lineages, exhibit discrepancies in the methylome . While this MR approach is robust to underlying disparities in cell type distribution that may influence both folate and methylation levels, it does not adjust for any downstream effects of RBC folate on cell type distribution that may impact observed changes in methylation. As a mediator, the analysis of the association between RBC folate and methylation is still valid without adjustment for cell type distribution. In this context, both fluctuations in cell type distribution and cell type-specific methylation may be meaningful, but these forms of variation are not partitioned by this analysis. Given this change in cell type distribution may be a component of the total effect RBC folate on methylation level and the inability to disambiguate the temporality of RBC folate and cell type distribution, we did not adjust for cell type distribution through recently proposed regression calibration techniques . However, future studies of DNA methylation should be cognizant the potential impact of cell distribution when defining biologically meaningful variation given the specific research question.
Using a novel application of Mendelian Randomization methods to DNA methylation data, this study was able to provide insight into the biological mechanism mediating the effects of maternal folate on fetal development. An array of socioeconomic and culture factors influence individual diet and other environmental exposures that may also alter the epigenome. Given the many putative determinants of DNA methylation levels, there is a high likelihood of unmeasured confounding using standard regression techniques to assess the association between RBC folate and methylation levels. This study demonstrated that the amalgamation of these unmeasured predictors of folate and methylation level generally biases effect estimates towards the null. Using Mendelian Randomization methods it was possible to identify a significant, appreciable effect of RBC folate on methylation level of several CpG loci in cord blood that would have otherwise been obfuscated using standard regression techniques. Future studies will be able to use these results to guide additional analysis of the effects of periconceptional folate on candidate pathways. More generally, this study demonstrated the utility of Mendelian Randomization for the assessment of epigenetic modifications.
This work was supported by Public Health Research Grant R21CA128382 (P.I.: KBM) from the National Cancer Institute, National Institutes of Health. AMB was supported by Training Grant T32HD060454 in Reproductive, Perinatal and Pediatric Epidemiology from the National Institute of Child Health and Human Development, National Institutes of Health.
- Rieder MJ: Prevention of neural tube defects with periconceptional folic acid. Clin Perinatol. 1994, 21 (3): 483-503.PubMedGoogle Scholar
- Pitkin RM: Folate and neural tube defects. Am J Clin Nutr. 2007, 85 (1): 285S-288S.PubMedGoogle Scholar
- De Wals P, Tairou F, Van Allen MI, Uh SH, Lowry RB, Sibbald B, Evans JA, Van den Hof MC, Zimmer P, Crowley M, Fernandez B, Lee NS, Niyonsenga T: Reduction in neural-tube defects after folic acid fortification in Canada. N Engl J Med. 2007, 357 (2): 135-142. 10.1056/NEJMoa067103.View ArticlePubMedGoogle Scholar
- Anonymous: Prevention of neural tube defects: results of the Medical Research Council Vitamin Study. MRC Vitamin Study Research Group. Lancet. 1991, 338 (8760): 131-137.View ArticleGoogle Scholar
- Wilcox AJ, Lie RT, Solvoll K, Taylor J, McConnaughey DR, Abyholm F, Vindenes H, Vollset SE, Drevon CA: Folic acid supplements and risk of facial clefts: National population based case-control study. BMJ. 2007 Mar 3, 334 (7591): 464-10.1136/bmj.39079.618287.0B.PubMed CentralView ArticlePubMedGoogle Scholar
- Bailey LB, Berry RJ: Folic acid supplementation and the occurrence of congenital heart defects, orofacial clefts, multiple births, and miscarriage. Am J Clin Nutr. 2005, 81 (5): 1213S-1217S.PubMedGoogle Scholar
- Czeizel AE: Periconceptional folic acid and multivitamin supplementation for the prevention of neural tube defects and other congenital abnormalities. Birth Defects Res A Clin Mol Teratol. 2009, 85 (4): 260-268. 10.1002/bdra.20563.View ArticlePubMedGoogle Scholar
- Goh YI, Bollano E, Einarson TR, Koren G: Prenatal multivitamin supplementation and rates of congenital anomalies: a meta-analysis. J Obstet Gynaecol Can. 2006, 28 (8): 680-689.PubMedGoogle Scholar
- Hollingsworth JW, Maruoka S, Boon K, Garantziotis S, Li Z, Tomfohr J, Bailey N, Potts EN, Whitehead G, Brass DM, Schwartz DA: In utero supplementation with methyl donors enhances allergic airway disease in mice. J Clin Invest. 2008, 118 (10): 3462-3469.PubMed CentralPubMedGoogle Scholar
- Miller RL: Prenatal maternal diet affects asthma risk in offspring. J Clin Invest. 2008, 118 (10): 3265-3268.PubMed CentralPubMedGoogle Scholar
- Magdelijns FJ, Mommers M, Penders J, Smits L, Thijs C: Folic acid Use in pregnancy and the development of atopy, asthma, and lung function in childhood. Pediatrics. 2011, 128 (1): e135-e144. 10.1542/peds.2010-1690.View ArticlePubMedGoogle Scholar
- Whitrow MJ, Moore VM, Rumbold AR, Davies MJ: Effect of supplemental folic acid in pregnancy on childhood asthma: a prospective birth cohort study. Am J Epidemiol. 2009, 170 (12): 1486-1493. 10.1093/aje/kwp315.View ArticlePubMedGoogle Scholar
- Haberg SE, London SJ, Nafstad P, Nilsen RM, Ueland PM, Vollset SE, Nystad W: Maternal folate levels in pregnancy and asthma in children at age 3 years. J Allergy Clin Immunol. 2011, 127 (1): 262-264.e1-PubMed CentralView ArticleGoogle Scholar
- Martinussen MP, Risnes KR, Jacobsen GW, Bracken MB: Folic acid supplementation in early pregnancy and asthma in children aged 6 years. Am J Obstet Gynecol. 2011, 206 (1): 72.e1-72.e7-PubMed CentralGoogle Scholar
- Hoyo C, Murtha AP, Schildkraut JM, Jirtle RL, Demark-Wahnefried W, Forman MR, Iversen ES, Kurtzberg J, Overcash F, Huang Z, Murphy SK: Methylation variation at IGF2 differentially methylated regions and maternal folic acid use before and during pregnancy. Epigenetics. 2011, 6 (7): 928-936. 10.4161/epi.6.7.16263.PubMed CentralView ArticlePubMedGoogle Scholar
- Steegers-Theunissen RP, Obermann-Borst SA, Kremer D, Lindemans J, Siebel C, Steegers EA, Slagboom PE, Heijmans BT: Periconceptional maternal folic acid use of 400 microg per day is related to increased methylation of the IGF2 gene in the very young child. PLoS One. 2009, 4 (11): e7845-10.1371/journal.pone.0007845.PubMed CentralView ArticlePubMedGoogle Scholar
- Ba Y, Yu H, Liu F, Geng X, Zhu C, Zhu Q, Zheng T, Ma S, Wang G, Li Z, Zhang Y: Relationship of folate, vitamin B12 and methylation of insulin-like growth factor-II in maternal and cord blood. Eur J Clin Nutr. 2011, 65 (4): 480-485. 10.1038/ejcn.2010.294.PubMed CentralView ArticlePubMedGoogle Scholar
- Greenland S: An introduction to instrumental variables for epidemiologists. Int J Epidemiol. 2000, 29 (6): 1102-10.1093/oxfordjournals.ije.a019909.View ArticlePubMedGoogle Scholar
- Martens EP, Pestman WR, de Boer A, Belitser SV, Klungel OH: Instrumental variables: application and limitations. Epidemiology. 2006, 17 (3): 260-267. 10.1097/01.ede.0000215160.88317.cb.View ArticlePubMedGoogle Scholar
- Hernan MA, Robins JM: Instruments for causal inference: an epidemiologist’s dream?. Epidemiology. 2006, 17 (4): 360-372. 10.1097/01.ede.0000222409.00878.37.View ArticlePubMedGoogle Scholar
- Smith GD, Ebrahim S: Mendelian randomization: prospects, potentials, and limitations. Int J Epidemiol. 2004, 33 (1): 30-42. 10.1093/ije/dyh132.View ArticlePubMedGoogle Scholar
- Davey Smith G, Ebrahim S: What can mendelian randomisation tell us about modifiable behavioural and environmental exposures?. BMJ. 2005, 330 (7499): 1076-1079. 10.1136/bmj.330.7499.1076.PubMed CentralView ArticlePubMedGoogle Scholar
- Didelez V, Sheehan N: Mendelian randomization as an instrumental variable approach to causal inference. Stat Methods Med Res. 2007, 16 (4): 309-330. 10.1177/0962280206077743.View ArticlePubMedGoogle Scholar
- Bound J, Jaeger DA, Baker RM: Problems with instrumental variables estimation when the correlation between the instruments and the endogeneous explanatory variable is weak. J Am Statistic Assoc. 1995, 90 (430): 443-Google Scholar
- Burgess S, Thompson SG: CRP CHD genetics collaboration: avoiding bias from weak instruments in Mendelian randomization studies. Int J Epidemiol. 2011, 40 (3): 755-764. 10.1093/ije/dyr036.View ArticlePubMedGoogle Scholar
- Frosst P, Blom HJ, Milos R, Goyette P, Sheppard CA, Matthews RG, Boers GJ, den Heijer M, Kluijtmans LA, van den Heuvel LP: A candidate genetic risk factor for vascular disease: a common mutation in methylenetetrahydrofolate reductase. Nat Genet. 1995, 10 (1): 111-113. 10.1038/ng0595-111.View ArticlePubMedGoogle Scholar
- van der Put NM, Gabreels F, Stevens EM, Smeitink JA, Trijbels FJ, Eskes TK, van den Heuvel LP, Blom HJ: A second common mutation in the methylenetetrahydrofolate reductase gene: an additional risk factor for neural-tube defects?. Am J Hum Genet. 1998, 62 (5): 1044-1051. 10.1086/301825.PubMed CentralView ArticlePubMedGoogle Scholar
- Davis SR, Quinlivan EP, Shelnutt KP, Maneval DR, Ghandour H, Capdevila A, Coats BS, Wagner C, Selhub J, Bailey LB, Shuster JJ, Stacpoole PW, Gregory JF, Davis SR, Quinlivan EP, Shelnutt KP, Maneval DR, Ghandour H, Capdevila A, Coats BS, Wagner C, Selhub J, Bailey LB, Shuster JJ, Stacpoole PW, Gregory JF: The methylenetetrahydrofolate reductase 677C- > T polymorphism and dietary folate restriction affect plasma one-carbon metabolites and red blood cell folate concentrations and distribution in women. J Nutr. 2005, 135 (5): 1040-1044.PubMedGoogle Scholar
- Guinotte CL, Burns MG, Axume JA, Hata H, Urrutia TF, Alamilla A, McCabe D, Singgih A, Cogger EA, Caudill MA: Methylenetetrahydrofolate reductase 677C- > T variant modulates folate status response to controlled folate intakes in young women. J Nutr. 2003, 133 (5): 1272-1280.PubMedGoogle Scholar
- Perry CA, Renna SA, Khitun E, Ortiz M, Moriarty DJ, Caudill MA: Ethnicity and race influence the folate status response to controlled folate intakes in young women. J Nutr. 2004, 134 (7): 1786-1792.PubMedGoogle Scholar
- Hung J, Yang TL, Urrutia TF, Li R, Perry CA, Hata H, Cogger EA, Moriarty DJ, Caudill MA: Additional food folate derived exclusively from natural sources improves folate status in young women with the MTHFR 677 CC or TT genotype. J Nutr Biochem. 2006, 17 (11): 728-734. 10.1016/j.jnutbio.2005.11.009.View ArticlePubMedGoogle Scholar
- Hay G, Clausen T, Whitelaw A, Trygg K, Johnston C, Henriksen T, Refsum H: Maternal folate and cobalamin status predicts vitamin status in newborns and 6-month-old infants. J Nutr. 2010, 140 (3): 557-564. 10.3945/jn.109.117424.View ArticlePubMedGoogle Scholar
- Ogbuanu IU, Zhang H, Karmaus W: Can we apply the Mendelian randomization methodology without considering epigenetic effects?. Emerg Themes Epidemiol. 2009, 6: 3-10.1186/1742-7622-6-3.PubMed CentralView ArticlePubMedGoogle Scholar
- Relton CL, Davey Smith G: Two-step epigenetic Mendelian randomization: a strategy for establishing the causal role of epigenetic processes in pathways to disease. Int J Epidemiol. 2012, 41 (1): 161-176. 10.1093/ije/dyr233.PubMed CentralView ArticlePubMedGoogle Scholar
- Kirkbride JB, Susser E, Kundakovic M, Kresovich JK, Davey Smith G, Relton CL: Prenatal nutrition, epigenetics and schizophrenia risk: can we test causal effects?. Epigenomics. 2012, 4 (3): 303-315. 10.2217/epi.12.20.PubMed CentralView ArticlePubMedGoogle Scholar
- Relton CL, Davey Smith G: Epigenetic epidemiology of common complex disease: prospects for prediction, prevention, and treatment. PLoS Med. 2010, 7 (10): e1000356-10.1371/journal.pmed.1000356.PubMed CentralView ArticlePubMedGoogle Scholar
- Michels KB, Harris HR, Barault L: Birthweight, maternal weight trajectories and global DNA methylation of LINE-1 repetitive elements. PLoS One. 2011, 6 (9): e25254-10.1371/journal.pone.0025254.PubMed CentralView ArticlePubMedGoogle Scholar
- Bibikova M, Le J, Barnes B, Saedinia-Melnyk S, Zhou L, Shen R, Gunderson KL: Genome-wide DNA methylation profiling using Infinium(R) assay. Epigenomics. 2009, 1 (1): 177-200. 10.2217/epi.09.14.View ArticlePubMedGoogle Scholar
- Chen YA, Choufani S, Ferreira JC, Grafodatskaya D, Butcher DT, Weksberg R: Sequence overlap between autosomal and sex-linked probes on the Illumina HumanMethylation27 microarray. Genom. 2011, 97 (4): 214-222. 10.1016/j.ygeno.2010.12.004.View ArticleGoogle Scholar
- Jacques PF, Sulsky SI, Sadowski JA, Phillips JC, Rush D, Willett WC: Comparison of micronutrient intake measured by a dietary questionnaire and biochemical indicators of micronutrient status. Am J Clin Nutr. 1993, 57 (2): 182-189.PubMedGoogle Scholar
- Selhub J, Jacques PF, Wilson PW, Rush D, Rosenberg IH: Vitamin status and intake as primary determinants of homocysteinemia in an elderly population. JAMA. 1993, 270 (22): 2693-2698. 10.1001/jama.1993.03510220049033.View ArticlePubMedGoogle Scholar
- Herbert V: Recommended dietary intakes (RDI) of folate in humans. Am J Clin Nutr. 1987, 45 (4): 661-670.PubMedGoogle Scholar
- Kleiber C, Zeileis A: Applied Econometrics with R. 2008, New York: Springer-VerlagView ArticleGoogle Scholar
- Fujita PA, Rhead B, Zweig AS, Hinrichs AS, Karolchik D, Cline MS, Goldman M, Barber GP, Clawson H, Coelho A, Diekhans M, Dreszer TR, Giardine BM, Harte RA, Hillman-Jackson J, Hsu F, Kirkup V, Kuhn RM, Learned K, Li CH, Meyer LR, Pohl A, Raney BJ, Rosenbloom KR, Smith KE, Haussler D, Kent WJ: The UCSC genome browser database: update 2011. Nucleic Acids Res. 2011, 39: D876-D882. 10.1093/nar/gkq963. Database issuePubMed CentralView ArticlePubMedGoogle Scholar
- Falcon S, Gentleman R: Using GOstats to test gene lists for GO term association. Bioinformatics. 2007, 23 (2): 257-258. 10.1093/bioinformatics/btl567.View ArticlePubMedGoogle Scholar
- Ogino S, Wilson RB: Genotype and haplotype distributions of MTHFR677C > T and 1298A > C single nucleotide polymorphisms: a meta-analysis. J Hum Genet. 2003, 48 (1): 1-7. 10.1007/s100380300000.View ArticlePubMedGoogle Scholar
- Relton CL, Wilding CS, Laffling AJ, Jonas PA, Burgess T, Binks K, Tawn EJ, Burn J: Low erythrocyte folate status and polymorphic variation in folate-related genes are associated with risk of neural tube defect pregnancy. Mol Genet Metab. 2004, 81 (4): 273-281. 10.1016/j.ymgme.2003.12.010.View ArticlePubMedGoogle Scholar
- Castro R, Rivera I, Ravasco P, Jakobs C, Blom HJ, Camilo ME, de Almeida IT: 5,10-Methylenetetrahydrofolate reductase 677C- > T and 1298A- > C mutations are genetic determinants of elevated homocysteine. QJM. 2003, 96 (4): 297-303. 10.1093/qjmed/hcg039.View ArticlePubMedGoogle Scholar
- Bailey LB, Gregory JF: Polymorphisms of methylenetetrahydrofolate reductase and other enzymes: metabolic significance, risks and impact on folate requirement. J Nutr. 1999, 129 (5): 919-922.PubMedGoogle Scholar
- Relton CL, Pearce MS, Burn J, Parker L: An investigation of folate-related genetic factors in the determination of birthweight. Epidemiol. 2005, 19 (5): 360-367.Google Scholar
- Burgess S, Thompson SG: Bias in causal estimates from Mendelian randomization studies with weak instruments. Stat Med. 2011, 30 (11): 1312-1323. 10.1002/sim.4197.View ArticlePubMedGoogle Scholar
- Chen LS, Emmert-Streib F, Storey JD: Harnessing naturally randomized transcription to infer regulatory relationships among genes. Genome Biol. 2007, 8 (10): R219-10.1186/gb-2007-8-10-r219.PubMed CentralView ArticlePubMedGoogle Scholar
- Millstein J, Zhang B, Zhu J, Schadt EE: Disentangling molecular relationships with a causal inference test. BMC Genet. 2009, 10 (1): 23-10.1186/1471-2156-10-23.PubMed CentralView ArticlePubMedGoogle Scholar
- Houseman EA, Accomando W, Koestler D, Christensen B, Marsit C, Nelson H, Wiencke J, Kelsey K: DNA methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinforma. 2012, 13: 86-10.1186/1471-2105-13-86.View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.