Mapping allele-specific DNA methylation: a new tool for maximizing information from GWAS

Abstract

In this issue of The Journal, an article by Schalkwyk et al.(1) shows the landscape of allele-specific DNA methylation (ASM) in the human genome. ASM has long been studied as a hallmark of imprinted genes, and a chromosome-wide version of this phenomenon occurs, in a random fashion, during X chromosome inactivation in female cells. But the type of ASM motivating the study by Schalkwyk et al. is different. They used a high-resolution, methylation-sensitive SNP array (MSNP) method for genome-wide profiling of ASM in total peripheral-blood leukocytes (PBL) and buccal cells from a series of monozygotic twin pairs. Their data bring a new level of detail to our knowledge of a newly recognized phenomenon-nonimprinted, sequence-dependent ASM. They document the widespread occurrence of this phenomenon among human genes and discuss its basic implications for gene regulation and genetic-epigenetic interactions. But this paper and recent work from other laboratories(2,3) raises the possibility of a more immediate and practical application for ASM mapping, namely to help extract maximum information from genome-wide association studies.

Figure 1

Sequence-Dependent ASM as a Tool for Extracting Maximum Information from GWAS (A) In genomic imprinting, the ASM is established in gametogenesis and dictated by the parental origin of the allele, with weak or absent effects of local haplotypes. Some imprinted genes show hypermethylation on the paternal allele as shown here, whereas others show hypermethylation of the maternal allele. In successive generations, the imprint is erased and then reset appropriately in gametogenesis, according to the sex of the transmitting parent. Thus genomic imprinting is non-Mendelian. In contrast, SNP- or haplotype-dependent ASM is dictated in cis by the local DNA sequence, regardless of parent of origin. This type of ASM is transmitted in a Mendelian fashion, and its presence is an indication of nearby regulatory SNPs that function, by mechanisms still largely unknown, to confer the allelic asymmetry. Although the number of imprinted genes is reasonably well established, the number of genes with nonimprinted, sequence-dependent ASM is influenced both by tissue type and by the stringency of the cutoffs utilized for scoring the allelic asymmetry. Black circles indicate methylated CpG dinucleotides; white circles, unmethylated CpGs. IC denotes imprinting center. (B) Schema for extracting maximum information from GWAS by overlapping association signals with data from mapping ASM and ASE. Most GWAS signals, even if they are true positives, are not likely to be the most important functional SNP, but rather serve to tag a functional rSNP nearby, which can confer ASE and/or ASM. Thus, genomic regions scoring as positive by both criteria (suprathreshold or subthreshold statistical associations in GWAS and ASE or ASM by appropriate assays) are likely to be true positives harboring bona fide causal rSNPs. Avoiding false positives will require using stringent criteria for recurrent genotype-dependent ASE and ASM and validating the high-throughput data from microarrays or Nextgen sequencing by independent locus-specific assays.