ABSTRACT
A hallmark of imprinted genes in mammals is the occurrence of parent-of-origin-dependent asymmetry of DNA cytosine methylation (5mC) of alleles at CpG islands (CGIs) in their promoter regions. This 5mCpG asymmetry between the parental alleles creates allele-specific imprinted differentially methylated regions (iDMRs). iDMRs are often coupled to the transcriptional repression of the methylated allele and the activation of the unmethylated allele in a tissue-specific, developmental-stage-specific and/or isoform-specific fashion. iDMRs function as regulatory platforms, built through the recruitment of chemical modifications to histones to achieve differential, parent-of-origin-dependent chromatin segmentation states. Here, we used a comparative computational data mining approach to identify 125 novel constitutive candidate iDMRs that integrate the maximal number of allele-specific methylation region records overlapping CGIs in human methylomes. Twenty-nine candidate iDMRs display gametic 5mCpG asymmetry, and another 96 are candidate secondary iDMRs. We established the maternal origin of the 5mCpG imprints of one gametic (PARD6G-AS1) and one secondary (GCSAML) iDMRs. We also found a constitutively hemimethylated, nonimprinted domain at the PWWP2AP1 promoter CGI with oocyte-derived methylation asymmetry. Given that the 5mCpG level at the iDMRs is not a sufficient criterion to predict active or silent locus states and that iDMRs can regulate genes from a distance of more than 1 Mb, we used RNA-Seq experiments from the Genotype-Tissue Expression project and public archives to assess the transcriptional expression profiles of SNPs across 4.6 Mb spans around the novel maternal iDMRs. We showed that PARD6G-AS1 and GCSAML are expressed biallelically in multiple tissues. We found evidence of tissue-specific monoallelic expression of ZNF124 and OR2L13, located 363 kb upstream and 419 kb downstream, respectively, of the GCSAML iDMR. We hypothesize that the GCSAML iDMR regulates the tissue-specific, monoallelic expression of ZNF124 but not of OR2L13. We annotated the non-coding epigenomic marks in the two maternal iDMRs using data from the Roadmap Epigenomics project and showed that the PARD6G-AS1 and GCSAML iDMRs achieve contrasting activation and repression chromatin segmentations. Lastly, we found that the maternal 5mCpG imprints are perturbed in several hematopoietic cancers. We conclude that the maternal 5mCpG imprints at PARD6G-AS1 and GCSAML iDMRs are decoupled from parent-of-origin transcriptional expression effects in multiple tissues.
