Background: DNA methylation patterns are initiated by de novo DNA methyltransferases DNMT3a/3b adding methyl groups to CG dinucleotides in the hypomethylated genome of early embryos. These patterns are faithfully maintained by DNMT1 during DNA replication to ensure epigenetic inheritance across generations. However, this two-step model is based on limited data. Results: We generated base-resolution DNA methylomes for a series of DNMT knockout embryonic stem cells, with deep coverage at highly repetitive elements. We show that DNMT1 and DNMT3a/3b activities work complementarily and simultaneously to establish symmetric CG methylation and CHH (H = A, T or C) methylation. DNMT3a/3b can add methyl groups to daughter strands after each cycle of DNA replication. We also observe an unexpected division of labor between DNMT1 and DNMT3a/3b in suppressing retrotransposon long terminal repeats and long interspersed elements, respectively. Our data suggest that mammalian cells use a specific CG density threshold to predetermine methylation levels in wild-type cells and the magnitude of methylation reduction in DNMT knockout cells. Only genes with low CG density can be induced or, surprisingly, suppressed in the hypomethylated genome. Lastly, we do not find any association between gene body methylation and transcriptional activity. Conclusions: We show the concerted actions of DNMT enzymes in the establishment and maintenance of methylation patterns. The finding of distinct roles of DNMT1-dependent and -independent methylation patterns in genome stability and regulation of transcription provides new insights for understanding germ cell development, neuronal diversity, and transgenerational epigenetic inheritance and will help to develop next-generation DNMT inhibitors.

Trimethylated histone H3 lysine 27 (H3K27me3) is linked to gene silencing, whereas H3K4me3 is associated with gene activation. These two marks frequently co-occupy gene promoters, forming bivalent domains. Bivalency signifies repressed but activatable states of gene expression and can be resolved to active, H3K4me3-prevalent states during multiple cellular processes, including differentiation, development and epithelial mesenchymal transition. However, the molecular mechanism underlying bivalency resolution remains largely unknown. Here, we show that the H3K27 demethylase UTX (also called KDM6A) is required for the resolution and activation of numerous retinoic acid (RA)-inducible bivalent genes during the RA-driven differentiation of mouse embryonic stem cells (ESCs). Notably, UTX loss in mouse ESCs inhibited the RA-driven bivalency resolution and activation of most developmentally critical homeobox (Hox) a–d genes. The UTX-mediated resolution and activation of many bivalent Hox genes during mouse ESC differentiation were recapitulated during RA-driven differentiation of human NT2/D1 embryonal carcinoma cells. In support of the importance of UTX in bivalency resolution, Utx-null mouse ESCs and UTX-depleted NT2/D1 cells displayed defects in RA-driven cellular differentiation. Our results define UTX as a bivalency-resolving histone modifier necessary for stem cell differentiation.

Over the past decades, genome-wide association studies (GWAS) have identified thousands of phenotype-associated DNA sequence variants for potential explanations of inter-individual phenotypic differences and disease susceptibility. However, it remains a challenge for translating the associations into causative mechanisms for complex diseases, partially due to the involved variants in the noncoding regions and the inconvenience of functional studies in human population samples. So far, accumulating evidence has suggested a complex crosstalk among genetic variants, allele-specific binding of transcription factors (ABTF), and allele-specific DNA methylation patterns (ASM), as well as environmental factors for disease risk. This review aims to summarize the current studies regarding the interactions of the aforementioned factors with a focus on epigenetic insights. We present two scenarios of single nucleotide polymorphisms (SNPs) in coding regions and non-coding regions for disease risk, via potentially impacting epigenetic patterns. While a SNP in a coding region may confer disease risk via altering protein functions, a SNP in non-coding region may cause diseases, via SNP-altering ABTF, ASM, and allele-specific gene expression (ASE). The allelic increases or decreases of gene expression are key for disease risk during development. Such ASE can be achieved via either a "SNP-introduced ABTF to ASM" or a "SNP-introduced ASM to ABTF." Together with our additional in-depth review on insulator CTCF, we are convinced to propose a working model that the small effect of a SNP acts through altered ABTF and/or ASM, for ASE and eventual disease outcome (named as a "SNP intensifier" model). In summary, the significance of complex crosstalk among genetic factors, epigenetic patterns, and environmental factors requires further investigations for disease susceptibility.