In the original version of the above article, a recent publication and its findings had not been acknowledged. The online and print versions have now been corrected, and the corrected sentence reads ''. but the importance of loss of H3K9me2 is not clear (Walter et al., 2016).'' The full citation to the reference has been added to the reference list: Walter, M., Teissandier, A., Pé rez-Palacios, R., Bourc'his, D. (2016). An epigenetic switch ensures transposon repression upon dynamic loss of DNA methylation in embryonic stem cells. Elife 5.

SummaryGlobal demethylation is part of a conserved program of epigenetic reprogramming to naive pluripotency. The transition from primed hypermethylated embryonic stem cells (ESCs) to naive hypomethylated ones (serum-to-2i) is a valuable model system for epigenetic reprogramming. We present a mathematical model, which accurately predicts global DNA demethylation kinetics. Experimentally, we show that the main drivers of global demethylation are neither active mechanisms (Aicda, Tdg, and Tet1-3) nor the reduction of de novo methylation. UHRF1 protein, the essential targeting factor for DNMT1, is reduced upon transition to 2i, and so is recruitment of the maintenance methylation machinery to replication foci. Concurrently, there is global loss of H3K9me2, which is needed for chromatin binding of UHRF1. These mechanisms synergistically enforce global DNA hypomethylation in a replication-coupled fashion. Our observations establish the molecular mechanism for global demethylation in naive ESCs, which has key parallels with those operating in primordial germ cells and early embryos.

Genomic DNA methylation contributes substantively to transcriptional regulations that underlie mammalian development and cellular differentiation. Much effort has been made to decipher the molecular mechanisms governing the establishment and maintenance of DNA methylation patterns. However, little is known about genome-wide variation of DNA methylation patterns. In this study, we introduced the concept of methylation entropy, a measure of the randomness of DNA methylation patterns in a cell population, and exploited it to assess the variability in DNA methylation patterns of Alu repeats and promoters. A few interesting observations were made: (i) within a cell population, methylation entropy varies among genomic loci; (ii) among cell populations, the methylation entropies of most genomic loci remain constant; (iii) compared to normal tissue controls, some tumors exhibit greater methylation entropies; (iv) Alu elements with high methylation entropy are associated with high GC content but depletion of CpG dinucleotides and (v) Alu elements in the intronic regions or far from CpG islands are associated with low methylation entropy. We further identified 12 putative allelic-specific methylated genomic loci, including four Alu elements and eight promoters. Lastly, using subcloned normal fibroblast cells, we demonstrated the highly variable methylation patterns are resulted from low fidelity of DNA methylation inheritance.

Repression of endogenous retroviruses (ERVs) in mammals involves several epigenetic mechanisms. Acute loss of the maintenance methyltransferase Dnmt1 induces widespread DNA demethylation and transcriptional activation of ERVs, including CpG-rich IAP (intracisternal A particle) proviruses. Here, we show that this effect is not due simply to a loss of DNA methylation. Conditional deletions reveal that both Dnmt1 and Np95 are essential for maintenance DNA methylation. However, while IAPs are derepressed in Dnmt1-ablated embryos and embryonic stem cells (ESCs), these ERVs remain silenced when Np95 is deleted alone or in combination with Dnmt1. This paradoxical phenotype results from an ectopic interaction between NP95 and the H3K9 methyltransferase SETDB1. Normally, SETDB1 maintains silencing of IAPs, but in the absence of DNMT1, prolonged binding of NP95 to hemimethylated DNA transiently disrupts SETDB1-dependent H3K9me3 deposition. Thus, our observations reveal an unexpected antagonistic interplay between two repressive pathways involved in retroviral silencing in mammalian cells.

DNA methylation, the only known covalent modification of mammalian DNA, occurs primarily in CpG dinucleotides. 51% of CpGs in the human genome reside within repeats, and 25% within Alu elements. Despite that, no method has been reported for large-scale ascertainment of CpG methylation in repeats. Here we describe a sequencing-based strategy for parallel determination of the CpG-methylation status of thousands of Alu repeats, and a computation algorithm to design primers that enable their specific amplification from bisulfite converted genomic DNA. Using a single primer pair, we generated amplicons of high sequence complexity, and derived CpG-methylation data from 31 178 Alu elements and their 5′ flanking sequences, altogether representing over 4 Mb of a human cerebellum epigenome. The analysis of the Alu methylome revealed that the methylation level of Alu elements is high in the intronic and intergenic regions, but low in the regions close to transcription start sites. Several hypomethylated Alu elements were identified and their hypomethylated status verified by pyrosequencing. Interestingly, some Alu elements exhibited a strikingly tissue-specific pattern of methylation. We anticipate the amplicons herein described to prove invaluable as epigenome representations, to monitor epigenomic alterations during normal development, in aging and in diseases such as cancer.