• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of ajhgLink to Publisher's site
Am J Hum Genet. Mar 12, 2010; 86(3): 411–419.
PMCID: PMC2833385

Genetic Control of Individual Differences in Gene-Specific Methylation in Human Brain


We have observed extensive interindividual differences in DNA methylation of 8590 CpG sites of 6229 genes in 153 human adult cerebellum samples, enriched in CpG island “shores” and at further distances from CpG islands. To search for genetic factors that regulate this variation, we performed a genome-wide association study (GWAS) mapping of methylation quantitative trait loci (mQTLs) for the 8590 testable CpG sites. cis association refers to correlation of methylation with SNPs within 1 Mb of a CpG site. 736 CpG sites showed phenotype-wide significant cis association with 2878 SNPs (after permutation correction for all tested markers and methylation phenotypes). In trans analysis of methylation, which tests for distant regulation effects, associations of 12 CpG sites and 38 SNPs remained significant after phenotype-wide correction. To examine the functional effects of mQTLs, we analyzed 85 genes that were with genetically regulated methylation we observed and for which we had quality gene expression data. Ten genes showed SNP-methylation-expression three-way associations—the same SNP simultaneously showed significant association with both DNA methylation and gene expression, while DNA methylation was significantly correlated with gene expression. Thus, we demonstrated that DNA methylation is frequently a heritable continuous quantitatively variable trait in human brain. Unlike allele-specific methylation, genetic polymorphisms mark both cis- and trans-regulatory genetic sites at measurable distances from their CpG sites. Some of the genetically regulated DNA methylation is directly connected with genetically regulated gene expression variation.


Changes of DNA methylation at CpG dinucleotides are heritable and play an important role in gene expression, X chromosome inactivation, parental imprinting, development, and complex disease.1–3 However, the regulation of DNA methylation of specific genes is poorly understood. A pilot study of the Human Epigenome Project (HEP) showed that there is considerable interindividual variation in DNA methylation, with ~50% of CpG sites having greater than 50% variation across all samples.4 Several other studies also documented individual CpG sites that exhibit variation among individuals.5–7

A twin study showed that within the H19 differentially methylated region (DMR), the heritability of methylation of individual CpG sites ranged from 20% to 74%. For the Insulin-like growth factor 2 (IGF2, [MIM 147470]) DMR, heritability among CpG sites varied between 57% and 97%.6 Bjornsson and colleagues observed that intraindividual DNA methylation changed over time and that this variation over time may be under genetic control.2 Furthermore, there exists compelling evidence for several associations between genetic variants and DNA methylation of specific genes. For example, SNP genotypes of the IGF2/H19 locus, where degree of methylation is involved in male infertility,8 were found to be significantly associated with methylation of the IGF2 DMR.6 Allele-specific methylation (ASM) was demonstrated in 16 SNPs from a genome-wide study, meaning that one SNP allele was associated with a complete or nearly complete methylation of a nearby CpG site, and the other allele was associated with the complete unmethylated state, or the SNP itself destroyed a CpG site by changing the C or G.9

All these findings led to the hypothesis that a considerable proportion of CpG sites may be quantitative traits with regulation by specific genetic variants. Identification of genetic variants that are associated with gene-specific DNA methylation could open a new venue to the understanding of methylation regulation.

In order to test the feasibility of genetic mapping of factors regulating DNA methylation, we performed a genome-wide association study (GWAS) testing associations between SNP genotypes and DNA methylation of individual CpG sites by treating methylation as a quantitative trait. This search is not limited to SNPs close to the CpG sites, which ASM normally targets. We include here genetic variants that can be hundreds of thousands of base pairs away or on a different chromosome. DNA methylation is controlled by various additional factors and can be tissue specific. We elected to study methylation control in cerebellum of human adult brain.

Material and Methods

We obtained 164 human cerebellum samples from the Stanley Medical Research Institute (SMRI). Of these, 153 individuals of European ancestry were included in the current analysis. The brains were autopsy specimens from patients with various psychiatric disorders and normal controls. Diagnoses of patients and unaffected controls were based on structured interviews by a senior psychiatrist with family member(s), to establish or rule out axis I diagnoses.10–12 The diagnoses were made by two senior psychiatrists, who used DSM-IV criteria.13 All samples have age, gender, race, postmortem interval (PMI), brain pH, smoking and alcohol use, suicide status, and psychotic features data. The sample demographic data and covariates are summarized in Table S1 available online.

Genotyping Methods

Genomic DNA was extracted from frozen cerebellar tissues provided by the SMRI. A phenol/chloroform/isoamyl alcohol protocol was modified and followed. The DNA was resuspended in 0.1 mM EDTA TE buffer. Genomic DNA was evaluated by NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) for concentration and by 1% agarose gel to validate the DNA integrity. We used Affymetrix GeneChip Mapping 5.0K Array and Assay Kits (Affymetrix, Santa Clara, CA) for genotyping according to the Affymetrix protocol. Genotypes were called with the BRLMM-p algorithm (Affymetrix) with all arrays simultaneously.

SNPs with call rates ≥ 99%, Hardy-Weinberg equilibrium (HWE) p values ≥ 0.001, and minor allele frequencies (MAF) ≥ 10% were included in the association tests. Of the genotypes obtained, 239,834 SNPs passed quality control and were used for subsequent analyses. Principal component analysis was applied to verify sample population homogeneity by running EIGENSTRAT.14 To examine the relatedness between samples, pair-wise identity-by-state was calculated with PLINK.15 The results confirmed that the 153 selected samples were unrelated and of European ancestry (see Figures S1–S3).

Methylation Assays

Genomic DNA was quantified by PicoGreen (Invitrogen, CA) and diluted to a final concentration of 50 ng/μl. DNA methylation was assessed at the Genomics Core Facility of Northwestern University (Chicago, IL) with Illumina Infinium HumanMethylation27 BeadChips (Illumina Inc., San Diego, CA). Technical details of this array are described elsewhere.16 The BeadChip probes 27,578 CpG sites. Of those, 20,007 sites (72.5%) are located in CpG islands and 7,571 (27.5%) are not. Almost all the sites (97.7%) are located in “promoter proxy” regions (less than 1500 bp from a transcription start site). The methylation level of each interrogated CpG site was calculated as the ratio of signal from a methylated probe relative to the sum of methylated and unmethylated probes. This value, β, ranges continuously from 0 (unmethylated) to 1 (fully methylated).

Pyrosequencing was used to validate DNA methylation data obtained from Beadchips. Genomic DNA was bisulfite converted according to the protocol in EpiTect 96 Bisulfite Kit (QIAGEN) and used as PCR template. Primers were designed with Pyrosequencing Assay Design Software v1.0.6 (Biotage, Uppsala, Sweden). A full list of primer sequences can be found in Table S2. PCR amplifications were performed with a standard protocol in 25 μl reactions, containing 20 ng of bilsulfite-converted DNA, 0.02 μM tagged Primer, 0.2 μM primer, 0.18 μM universal biotin-labeled primer, 1.0 mM MgCl, 0.125 mM dNTP, 1×PCR buffer, and 0.8U Hotstart Taq polymerase (QIAGEN). PCR cycling conditions are 95°C 5 min, 50 cycles (95°C 15 s, 60°C 30 s, 72°C 30 s), and 72°C 5 min. PCR products were processed according to the manufacturer's standard protocol (Biotage). In brief, 3 μl of streptavidin-sepharose beads (Amersham Biosciences, Piscataway, NJ) and 40 μl of binding buffer (pH 7.6, 10 mM Tris-HCl, 1 mM EDTA, 2 M NaCl, 0.1% Tween 20) were mixed with 20 μl of PCR product for 10 min at room temperature. The reaction mixture was immobilized onto streptavidin-coated beads. After application of the vacuum, the beads were treated with high-purity water for 30 s, 70% ethanol for 5 s, and a denaturation solution (0.2 M NaOH) for 5 s and washed for 5 s with washing buffer (10 mM Tris-acetate at pH 7.6). The beads were then suspended with 40 μl of annealing buffer (20 mM Tris-acetate, 2 mM Mg-acetate at pH 7.6) containing 0.5 μM of sequencing primer, prefilled in a PSQ 96 Plate (Biotage). The plate with samples was heated at 80°C for 2 min and finally cooled to room temperature. Sequencing reactions were performed with a PSQ 96 SNP Reagent Kit (Biotage) according to the manufacturer's instructions. The percent methylation at each CpG site was calculated from the raw data with the Pyro-Q-CpG Software (Biotage).

Classification of CpG Sites

More than half of CpG sites assayed (54.3%) were hypomethylated (≤20% DNA methylated; see Figure S4). Studies of HEP data from three human chromosomes reported a similar pattern of methylation distribution. CpG sites near promoters were more likely to be hypomethylated.17,18

A CpG site was studied for SNP association if it had fewer than 95% of individuals with only hypomethylation or only hypermethylation (≥80% DNA methylated); 8590 CpG sites were thus included. Of those, 5097 were not within CpG islands. CpG sites within 2 kb of CpG islands were defined by Irizarry et al. as “CpG island shores.”19 We observed that as compared with CpG islands, CpG sites with greater variability were enriched both in “CpG island shores” (permutation p value < 1.0E-7) and in more distant regions (>2 kb from CpG islands, permutation p value < 1.0E-7, see Table S3).

Considering each CpG site for all samples, we generally observed unimodal (single-peak) distributions. The vast majority of the included CpG sites (92.7%) had a unimodally distribution, as observed previously,6,20 lending support to a QTL approach to genetic analysis rather than a qualitative (binary) locus approach (Figures S5 and S6).

Expression Data

Expression data from cerebellum for 45 of the same individuals is available from the SMRI Online Genomics Database. Oligonucleotide microarray chip (HGU95Av2) experiments reported in that database were carried out according to the manufacturer's protocol (Affymetrix, Santa Clara, CA).10,12 We performed RMA normalization with Partek Genomics Suite (Partek Inc., St. Louis, MO). There are technical replicates for every individual.10-12 A total of 12,625 probe sets were assayed by HGU95Av2. We selected 4648 probes that were coded as “present” (called by the Affymetrix Microarray Suite [MAS] algorithm) in ≥80% of samples.

Expression and Methylation Data Preprocessing

COMBAT21 was used to correct for batch effects within the methylation and expression array data, including 15 technical replicate pairs in the methylation data and 45 technical replicate pairs in the expression data. For later analysis of each technical replicate pair, the data were averaged for the replicated samples to obtain a single datum. In order to remove the effects of known and unknown covariates on the data, surrogate variable analysis (SVA)22 was applied and the identified surrogate variables were regressed out. We examined the effects of known variables on the methylation data pre- and post-COMBAT21 and SVA.22 In the regression analysis, quantitative and categorical covariates were used according to the data (details in Table S4). The methylation data before processing demonstrated strong batch effects (barcode of chips was a significant [p < 0.05] covariate for 91% of probes). Batch effects were present in only 2% of the probes after correction, which is close to our chance expectation (Table S4). For the expression data, brain pH and batch effects were significant (p < 0.05) in 41% and 10%, respectively, of the probes in the data prior to preprocessing but each is significant in only 1% of the probes in the corrected data.

QTL Analysis

To fit a normal distribution, quantile normalization was used for both expression and methylation residuals. Linear regression analysis was performed to test for correlation between the normalized residuals and the number of minor alleles via an additive genetic model by PLINK.15 From this analysis, an asymptotic p value from the Wald statistic was obtained as a measure of association of each SNP with methylation of any given CpG site.

Multiple Testing Correction

Three sets of permutations of phenotype were performed. Permutations for a CpG-SNP combination were calculated with the adaptive perm option of PLINK (aperm), permuting up to 1 billion replicates (EMP p value). This corrects for possible nonnormality of the phenotype distribution. Permutations correcting for multiple testing within a cis region or whole-genome scan were also performed with the max (T) permutation (mperm) option of PLINK (region-wide p for cis; genome-wide p for trans). For each phenotype, results were permuted 1000 times, with the same seed to maintain the correlation between phenotypes. To estimate phenotype-wide significance (in addition to region-wide significance), the best statistic per replicate for each phenotype was saved with the PLINK mperm-save option. Statbest, the statistic from the most significant phenotype, was defined for each replicate. Phenotype-wide corrected p values were calculated as (R+1)/(N+1) where R is the number of times the statbest exceeded the observed statistic and N is the number of permutations (1000).

cis- and trans-Regulation of Methylation

Like the classification of gene expression regulators, the regulation of methylation traits can be roughly divided into two types: cis-acting regulation by DNA elements in or adjacent to each CpG site, and trans-acting regulation by factors from the genomic regions distant from the CpG sites, including from different chromosomes. We defined the SNPs within a region bounded by 1 Mb distance from both ends of each CpG site as candidates for cis analysis. All the other SNPs were analyzed for trans-acting associations for each CpG site.

Effect of mSNPs on Gene Expression

We tested the association of corresponding gene expression and mSNPs (SNPs showing phenotype-wide significant associations with DNA methylation of CpG sites in a given gene) by linear regression with PLINK. Region-wide significance was corrected for the number of SNPs analyzed for each expression probe (for details, see above QTL analysis). Note that there were only a small number of individuals and genes with existing acceptable expression data, as described above.

Correlation Analysis of DNA Methylation and Gene Expression

We investigated the correlation between genetically determined DNA methylation, which showed phenotype-wide significant cis-mQTL association and expression of the corresponding genes. Pearson linear regression was applied to detect the correlation between DNA methylation and gene expression by R after preprocessing of expression and methylation data (see above). The multiple testing correction of p values was performed by positive false discovery rate (q value) implemented in Partek Genomics Suite.23


mQTL Analysis

In the cis analysis, 12,117 SNP-CpG pairs, consisting of 9,448 SNPs and 2,046 CpG sites (of 1,795 genes), were significantly correlated (region-wide permuted p ≤ 0.05) (Table S5). The associations of 3,323 pairs (involving 736 CpG sites of 658 genes associated with 2,878 SNPs) remained significant after correcting for the 8,590 methylation phenotypes tested (phenotype-wide p value ≤ 0.05). Among the 736 CpG sites with phenotype-wide significant cis associations, CpG sites within CGIs were more likely to have phenotype-wide significant cis-mQTLs than in non-CGI regions (permutation p value = 3.0E-4) (Table S6).

The cis associations with methylation showed effect sizes (R2) ranging from 0.17 to 0.73. The most significant association for each of the CpG sites are shown in Table 1 (top 10 probes with the smallest Wald p value) and Table S7 (highlighted in orange for the 736 CpG sites with phenotype-wide significant cis-mQTL associations). All SNPs that have region-wide significant associations with DNA methylation are in Table S5. Closer inspection on the positions of SNPs with phenotype-wide significant cis-mQTL associations (Figure 1) showed that most of the associated SNPs were near the CpG sites, 95% within a 149 kb range.

Figure 1
p Values and Distances between CpG Sites and SNPs
Table 1
Ten CpG-SNP Pairs with the Most Significant Phenotype-wide Corrected cis Associations

390 SNPs showed cis association (phenotype-wide p ≤ 0.05) with two or more CpG sites of 141 genes. Interestingly, 163 SNPs' and 85 genes' CpG sites are clustered in 37 genomic regions (Table S8). In each cluster, multiple SNPs are associated with multiple CpG sites of several different genes. For example, a 186 Kb region on chromosome 1 contains 14 SNPs that are associated with three CpG sites of three different genes (LCE1D [MIM 612606], LCE2B [MIM 612610], and LCE3A [MIM 612613]). Gene families were frequently observed in these clusters, including keratin-associated protein, claudin, killer cell lectin-like receptor, proline-rich protein BstNI subfamily, late cornified envelope protein, and other gene families. These genes in one cluster are likely to be coregulated for their DNA methylation.

In the trans analysis, 372 SNP-CpG pairs involving 368 SNPs and 246 CpG sites (of 240 genes) showed association at permutation corrected genome-wide p ≤ 0.05 (Table S9). Thirty-eight SNP-CpG trans pairs (12 CpG sites and 38 SNPs) were significant after further phenotype-wide correction (corrected p ≤ 0.05, highlighted in orange in Table S9). Table 2 showed the strongest trans associations of each CpG site. Ten of the trans associations have SNPs from different chromosomes, while two are on the same chromosomes but more than 1 Mb away from the target CpG sites.

Table 2
Phenotype-wide Significant trans Associations Showing the Best Association for Each CpG Site


In order to validate the DNA methylation measurement obtained from the microarray, we successfully designed pyrosequencing assays for five randomly selected cis-mQTL CpG sites. Each CpG site was processed in one batch for all samples, so no batch effect was involved. SVA is not applicable to analysis of a few CpG sites. Linear regression analysis was performed for the SNP that showed the best signal with a given CpG site measured by pyrosequencing without adjustment of covariates (see Table S10). Correspondingly, the association was recalculated with residuals of DNA methylation measured by Illumina Beadchips after correction for batch effect by COMBAT but no further processing of SVA. All five cis associations detected with Illumina Beadchip were confirmed by data from pyrosequencing. One example is shown in Figure 2. Minor allele T of rs10492813 significantly increased DNA methylation of cg23815491 (gray boxplot). Pyrosequencing validated the association predicted by the Beadchip data (white boxplot). We also examined the genotype clusters for the five SNPs associated with DNA methylation of the five pyrosequencing-validated CpG sites (see Figure S7). The genotype clusters were all of good quality.

Figure 2
Methylation of cg23815491 Measured by Pyrosequencing and Beadchip Associated with SNP rs10492831

Functional Effects of SNPs and DNA Methylation on Gene Expression

The known relationships between SNPs and DNA methylation and between DNA methylation and gene expression raised the question of whether SNPs associated with mQTLs might also be associated with expression of the same genes. Therefore, we further analyzed the association of mSNPs (phenotype-wide significantly associated with DNA methylation in a cis manner mentioned above) with gene expression of a given gene for which we had existing data with acceptable quality. We identified 85 genes related to phenotype-wide significant cis-mQTL and with an expression probe that met QC criteria (greater than 80% present, see Material and Methods above). These SNPs, methylation, and expression data make up 112 CpG-expression probe pairs (comprising101 CpG sites and 95 expression probes) and 550 mSNP-expression probe pairs (comprising 447 mSNPs and 95 expression probes). We looked at the relationships between these data pairs in two ways: (1) association between genotypes of mSNPs with gene expression and (2) correlation between methylation and gene expression.

Ninety-two of the 550 mSNP-expression probe pairs (comprising 59 mSNPs and 19 expression probes of 17 genes) showed a region-wide significant association between the mSNP genotypes and the expression of the corresponding gene (Table S12).

Twenty of the 112 CpG-expression probe pairs (comprising 19 CpG sites and 18 expression probes of 17 genes) showed nominally significant correlations. As expected, DNA methylation negatively correlated with gene expression for 15 pairs, a majority of the pairs. Interestingly, five pairs of four genes (ZNF266 [MIM 604751], FANCG [MIM 607139], DDT [MIM 602750], and FUT1 [MIM 211100]) showed that increased DNA methylation upregulated the genes' expression. Eleven of the 20 pairs (11 CpG sites of 10 genes and 10 expression probes of 10 genes) survive multiple test correction (FDR q value ≤ 0.05; see Table 3).

Table 3
Most Significant Correlations of Methylation with Phenotype-wide Significant cis Regulators and Expression

Putting the above two correlations with the mQTL data together, we noticed that 10 genes (involving 10 CpG sites, 11 expression probes, and 29 SNPs) showed three-way associations—the same SNP simultaneously showed a significant association with DNA methylation and with gene expression. At the same time, DNA methylation significantly correlated with gene expression of the same gene (Table S13). For example, minor allele C of SNP rs2235375 was associated with the increased methylation level of gene IRF6 (MIM 607199) (CpG site cg23283495, phenotype-wide p value < 0.001). The C allele was also associated with the reduced expression of IRF6 with region-wide significance (region-wide p ≤ 0.05). A significant linear negative correlation between methylation and expression of gene IRF6 was observed. The box plot of expression and methylation in gene IRF6 with SNP rs2235375 is presented in Figure 3. The other nine “three-way associations” are showed in Figure S8.

Figure 3
DNA Methylation and Gene Expression of IRF6 Plotted by Genotypes of rs2235375


We have observed that numerous CpG sites are regulated by genetic variants in cis and/or trans manner. Our results showed that CpG sites with extensive variability were more enriched in non-CGI (CpG islands) regions than within CGIs. Previous studies found that CpG sites in CGI were largely unmethylated,17,18 whereas CpG sites in non-CGI regions were moderately to highly methylated.16 Interestingly, CpG sites within CGIs were more likely to be phenotype-wide significantly associated with cis regulators than in non-CGI regions (permutation p value = 3.0E-4). The fact that genetic variants regulate DNA methylation of CGIs more than of CpG island shores and distant CpG sites is intriguing in light of the report by Irizarry et al.19 They reported that CpG island shores are enriched for tissue-specific methylation sites in a study comparing different tissue types.19

Our findings are concordant with the findings of a number of previous studies associating variation of candidate gene DNA methylation with SNPs in cis in human and mice.6,9,24 Kerkel et al. carried out a pioneering genome-wide survey of ASM sites.9 They found 16 sites with ASM. Eight of the 16 sites (around BCL2 [MIM 151430], CYP2A7 [MIM 608054], EFNB1 [MIM 300035], GCNT3 [MIM 606836], LTF [MIM 150210], PIM1 [MIM 164960], VNN1 [MIM 603570], and MAGEL2 [MIM 605283] genes) reported by Kerkel et al. gave at least nominally significant associations in our cis association study, even though the Kerkel et al. group studied tissues other than brain. Only the LTF cis association reached phenotype-wide significance in our study. Differences between the two studies might be attributable to tissue differences and to different statistical criteria for significance.

Kerkel et al.'s findings were probably limited to SNPs within 2 kb regions around the CpG sites, because their detection relied on short amplicons after HpaII or MspI digestion. Our association tests capture longer-distance SNP-methylation correlations: cis associations extend to SNPs within 1 Mb distance on each side of the CpG site. Most of our cis phenotype-wide significant associations (87.9%) are from regions more than 2 kb away from the CpG sites. In a few cases, effects were observable over longer distances, which were probably out of the range of linkage disequilibrium (LD) because the average r2 (r: correlation coefficient between pairs of loci, a measure of LD) for SNP pairs decreases to less than 0.1 when the distance interval is 160 kb in Europeans.25 trans association of DNA methylation largely came from different chromosomes.

DNA methylation of IGF2/H19, one of the best-characterized genetically regulated loci, was previously reported to be strongly determined by heritable factors and SNPs in cis.6 Our analysis confirmed cis association in IGF2/H19 (region-wide p < 0.05), although the SNP that showed the most significant association in a previous study was not included in this study and the association in this study did not reach our strict phenotype-wide significance level.

Several known genes (such as DNMT1 [MIM 126375], DNMT3A [MIM 602769], DNMT3B [MIM 602900], MTHFR [MIM 607093], etc.) are involved in general DNA methylation. Dnmt3a and Dnmt3b are required for de novo methylation of DNA in mammals.26–28 Dnmt1 is known to have a high preference for hemimethylated CpG sites and has an important role in maintenance of methylation.29,30 MTHFR affects global DNA methylation.31 None of these genes contain SNPs significantly associated with mQTLs in the present study. Heijmans and colleagues also failed to detect an association of MTHFR with DNA methylation of IGF2/H19.6 One possible reason is that the polymorphisms in these genes are not completely covered in our study and in Heijmans et al.'s study (for list of SNPs tested in this study, see Table S11). Another more likely explanation is that the current study identifies gene-specific DNA methylation, which is different from the global DNA methylation controlled by these known methylation pathway genes. The current results may eventually contribute to an understanding of the mechanisms of gene-specific DNA methylation.

By using available and acceptable quality gene expression data, we further found that about 13% (59/447) of genetic variants regulating DNA methylation (mSNPs) also affect gene expression. Around 18% (20/112) of CpG-expression probe pairs showed nominally significant correlations. Increased DNA methylation upregulated the gene's expression for five CpG-probe pairs of four genes. Previous studies have also reported positive correlations between DNA methylation of gene body and gene expression.32–34 Further studies are needed to explore the mechanism of hypomethylation and decreased gene expression.

Taking SNP, DNA methylation, and gene expression together, we observed 10 genes that showed three-way associations. In the case of rs2235375, the C allele was simultaneously associated with increased methylation and decreased expression of IRF6. SNP rs2235375 is an intronic polymorphism located 14 kb away from the studied promoter CpG site. This finding strongly implicates a distant intronic variation affecting DNA methylation, which may then impact expression of the gene itself. Interestingly, the C allele of SNP rs2235375 has been associated with nonsyndromic cleft lip with or without cleft palate.35 In addition, Irf6 knockout mice developed abnormal skin, limb, and craniofacial morphogenesis.36 DNA methylation regulation disturbance and consequential reduction in gene expression could be an explanation of the genetic association detected in the cleft lip study.

These associations imply that mQTLs and eQTLs of particular genes may be related but not identical. Similarly, Kerkel et al.'s study reported that two out of four genes tested showed both allelic expression and allelic methylation.9 Of course, many other factors are involved in regulating gene expression besides DNA methylation and SNPs. The contribution of genetic variants and DNA methylation to gene expression varies gene by gene. Studies of both mQTLs and eQTLs can illuminate the potential functional role of genetic variations in association studies of complex disease.

Our results reveal numerous new instances of genetic variability contributing to the variability of DNA methylation of specific genomic regions. In recent years, genome-wide association has revealed many SNPs associated with diseases or other phenotypes. In addition to effects on protein coding, RNA splicing, and microRNA targeting, the impact of genetic factors on DNA methylation is clearly another important aspect to study in the evaluation of SNP functional effects. The interactions between SNPs and target CpG sites, particularly in those related to trans signals, may lead to identification of novel gene-gene interactions. These findings could lead to the discovery of novel mechanisms that determine gene-specific DNA methylation, which has functional effects on phenotypes including disease.


The authors thank the families of the individuals involved in this study. The Stanley Medical Research Institute and its Collaborators, Drs. Elashoff, Torrey, and Webster, generously gave us access to their sample collections. This work was supported by NARSAD Distinguished Investigator Awards (to E.S.G.), the Brain Research Foundation at the University of Chicago (to C.L.), NIH MH080425 (to C.L.), and NIH 5R01 MH61613 (to E.S.G.). Support from the Geraldi Norton Foundation and the Eklund Family is also gratefully acknowledged. We declare no conflict of interest.

Supplemental Data

Document S1. Sixteen Figures and Seven Tables:
Table S5. Region-wide cis Associations of mQTL:
Table S7. cis Associations of 8590 CpG Sites:
Table S8. Coregulators of More than One CpG Site:
Table S9. Genome-wide Significant trans Associations of mQTLs:
Table S11. Coregulation of mQTLs and eQTLs in Human Cerebellum:
Table S12. Three-Way Associations of SNPs, DNA Methylation, and Gene Expression in Human Cerebellum:

Web Resources

The URLs for data presented herein are as follows:


1. Bird A. DNA methylation patterns and epigenetic memory. Genes Dev. 2002;16:6–21. [PubMed]
2. Bjornsson H.T., Sigurdsson M.I., Fallin M.D., Irizarry R.A., Aspelund T., Cui H., Yu W., Rongione M.A., Ekström T.J., Harris T.B. Intra-individual change over time in DNA methylation with familial clustering. JAMA. 2008;299:2877–2883. [PMC free article] [PubMed]
3. Robertson K.D. DNA methylation and human disease. Nat. Rev. Genet. 2005;6:597–610. [PubMed]
4. Rakyan V.K., Hildmann T., Novik K.L., Lewin J., Tost J., Cox A.V., Andrews T.D., Howe K.L., Otto T., Olek A. DNA methylation profiling of the human major histocompatibility complex: A pilot study for the human epigenome project. PLoS Biol. 2004;2:e405. [PMC free article] [PubMed]
5. Chen H., Taylor N.P., Sotamaa K.M., Mutch D.G., Powell M.A., Schmidt A.P., Feng S., Hampel H.L., de la Chapelle A., Goodfellow P.J. Evidence for heritable predisposition to epigenetic silencing of MLH1. Int. J. Cancer. 2007;120:1684–1688. [PubMed]
6. Heijmans B.T., Kremer D., Tobi E.W., Boomsma D.I., Slagboom P.E. Heritable rather than age-related environmental and stochastic factors dominate variation in DNA methylation of the human IGF2/H19 locus. Hum. Mol. Genet. 2007;16:547–554. [PubMed]
7. Oates N.A., van Vliet J., Duffy D.L., Kroes H.Y., Martin N.G., Boomsma D.I., Campbell M., Coulthard M.G., Whitelaw E., Chong S. Increased DNA methylation at the AXIN1 gene in a monozygotic twin from a pair discordant for a caudal duplication anomaly. Am. J. Hum. Genet. 2006;79:155–162. [PMC free article] [PubMed]
8. Boissonnas C.C., Abdalaoui H.E., Haelewyn V., Fauque P., Dupont J.M., Gut I., Vaiman D., Jouannet P., Tost J., Jammes H. Specific epigenetic alterations of IGF2-H19 locus in spermatozoa from infertile men. Eur. J. Hum. Genet. 2009;18:73–80. [PMC free article] [PubMed]
9. Kerkel K., Spadola A., Yuan E., Kosek J., Jiang L., Hod E., Li K., Murty V.V., Schupf N., Vilain E. Genomic surveys by methylation-sensitive SNP analysis identify sequence-dependent allele-specific DNA methylation. Nat. Genet. 2008;40:904–908. [PubMed]
10. Knable M.B., Barci B.M., Webster M.J., Meador-Woodruff J., Torrey E.F., Stanley Neuropathology Consortium Molecular abnormalities of the hippocampus in severe psychiatric illness: Postmortem findings from the Stanley Neuropathology Consortium. Mol. Psychiatry. 2004;9 609–620, 544. [PubMed]
11. Torrey E.F., Webster M., Knable M., Johnston N., Yolken R.H. The stanley foundation brain collection and neuropathology consortium. Schizophr. Res. 2000;44:151–155. [PubMed]
12. Torrey E.F., Barci B.M., Webster M.J., Bartko J.J., Meador-Woodruff J.H., Knable M.B. Neurochemical markers for schizophrenia, bipolar disorder, and major depression in postmortem brains. Biol. Psychiatry. 2005;57:252–260. [PubMed]
13. American Psychiatric Association . Fourth Edition. American Psychiatric Association; Washington, DC: 2000. DSM-IV. Diagnostic and Statistical Manual of Mental Disorders.
14. Price A.L., Patterson N.J., Plenge R.M., Weinblatt M.E., Shadick N.A., Reich D. Principal components analysis corrects for stratification in genome-wide association studies. Nat. Genet. 2006;38:904–909. [PubMed]
15. Purcell S., Neale B., Todd-Brown K., Thomas L., Ferreira M.A., Bender D., Maller J., Sklar P., de Bakker P.I., Daly M.J., Sham P.C. PLINK: A tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet. 2007;81:559–575. [PMC free article] [PubMed]
16. Bibikova M., Le J., Barnes B., Saedinia-Melnyk S., Zhou L., Shen R., Gunderson K.L. Genome-wide DNA methylation profiling using Infinium® assay. Epigenomics. 2009;1:177–200. [PubMed]
17. Eckhardt F., Lewin J., Cortese R., Rakyan V.K., Attwood J., Burger M., Burton J., Cox T.V., Davies R., Down T.A. DNA methylation profiling of human chromosomes 6, 20 and 22. Nat. Genet. 2006;38:1378–1385. [PMC free article] [PubMed]
18. Fan S., Zhang X. CpG island methylation pattern in different human tissues and its correlation with gene expression. Biochem. Biophys. Res. Commun. 2009;383:421–425. [PubMed]
19. Irizarry R.A., Ladd-Acosta C., Wen B., Wu Z., Montano C., Onyango P., Cui H., Gabo K., Rongione M., Webster M. The human colon cancer methylome shows similar hypo- and hypermethylation at conserved tissue-specific CpG island shores. Nat. Genet. 2009;41:178–186. [PMC free article] [PubMed]
20. Ladd-Acosta C., Pevsner J., Sabunciyan S., Yolken R.H., Webster M.J., Dinkins T., Callinan P.A., Fan J.B., Potash J.B., Feinberg A.P. DNA methylation signatures within the human brain. Am. J. Hum. Genet. 2007;81:1304–1315. [PMC free article] [PubMed]
21. Johnson W.E., Li C., Rabinovic A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics. 2007;8:118–127. [PubMed]
22. Leek J.T., Storey J.D. Capturing heterogeneity in gene expression studies by surrogate variable analysis. PLoS Genet. 2007;3:1724–1735. [PMC free article] [PubMed]
23. Storey J.D., Tibshirani R. Statistical methods for identifying differentially expressed genes in DNA microarrays. Methods Mol. Biol. 2003;224:149–157. [PubMed]
24. Schilling E., El Chartouni C., Rehli M. Allele-specific DNA methylation in mouse strains is mainly determined by cis-acting sequences. Genome Res. 2009;19:2028–2035. [PMC free article] [PubMed]
25. Shifman S., Kuypers J., Kokoris M., Yakir B., Darvasi A. Linkage disequilibrium patterns of the human genome across populations. Hum. Mol. Genet. 2003;12:771–776. [PubMed]
26. Gowher H., Jeltsch A. Enzymatic properties of recombinant Dnmt3a DNA methyltransferase from mouse: The enzyme modifies DNA in a non-processive manner and also methylates non-CpG [correction of non-CpA] sites. J. Mol. Biol. 2001;309:1201–1208. [PubMed]
27. Okano M., Xie S., Li E. Cloning and characterization of a family of novel mammalian DNA (cytosine-5) methyltransferases. Nat. Genet. 1998;19:219–220. [PubMed]
28. Okano M., Bell D.W., Haber D.A., Li E. DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell. 1999;99:247–257. [PubMed]
29. Fatemi M., Hermann A., Pradhan S., Jeltsch A. The activity of the murine DNA methyltransferase Dnmt1 is controlled by interaction of the catalytic domain with the N-terminal part of the enzyme leading to an allosteric activation of the enzyme after binding to methylated DNA. J. Mol. Biol. 2001;309:1189–1199. [PubMed]
30. Li E., Bestor T.H., Jaenisch R. Targeted mutation of the DNA methyltransferase gene results in embryonic lethality. Cell. 1992;69:915–926. [PubMed]
31. Friso S., Choi S.W., Girelli D., Mason J.B., Dolnikowski G.G., Bagley P.J., Olivieri O., Jacques P.F., Rosenberg I.H., Corrocher R., Selhub J. A common mutation in the 5,10-methylenetetrahydrofolate reductase gene affects genomic DNA methylation through an interaction with folate status. Proc. Natl. Acad. Sci. USA. 2002;99:5606–5611. [PMC free article] [PubMed]
32. Ball M.P., Li J.B., Gao Y., Lee J.H., LeProust E.M., Park I.H., Xie B., Daley G.Q., Church G.M. Targeted and genome-scale strategies reveal gene-body methylation signatures in human cells. Nat. Biotechnol. 2009;27:361–368. [PMC free article] [PubMed]
33. Movassagh M., Choy M.K., Goddard M., Bennett M.R., Down T.A., Foo R.S. Differential DNA methylation correlates with differential expression of angiogenic factors in human heart failure. PLoS ONE. 2010;5:e8564. [PMC free article] [PubMed]
34. Rauch T.A., Wu X., Zhong X., Riggs A.D., Pfeifer G.P. A human B cell methylome at 100-base pair resolution. Proc. Natl. Acad. Sci. USA. 2009;106:671–678. [PMC free article] [PubMed]
35. Scapoli L., Palmieri A., Martinelli M., Pezzetti F., Carinci P., Tognon M., Carinci F. Strong evidence of linkage disequilibrium between polymorphisms at the IRF6 locus and nonsyndromic cleft lip with or without cleft palate, in an Italian population. Am. J. Hum. Genet. 2005;76:180–183. [PMC free article] [PubMed]
36. Ingraham C.R., Kinoshita A., Kondo S., Yang B., Sajan S., Trout K.J., Malik M.I., Dunnwald M., Goudy S.L., Lovett M. Abnormal skin, limb and craniofacial morphogenesis in mice deficient for interferon regulatory factor 6 (Irf6) Nat. Genet. 2006;38:1335–1340. [PMC free article] [PubMed]

Articles from American Journal of Human Genetics are provided here courtesy of American Society of Human Genetics


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles
  • SNP
    PMC to SNP links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...