Whole-genome bisulfite sequencing (WGBS) is becoming an increasingly accessible technique, used widely for both fundamental and disease-oriented research. Library preparation methods benefit from a variety of available kits, polymerases and bisulfite conversion protocols. Although some steps in the procedure, such as PCR amplification, are known to introduce biases, a systematic evaluation of biases in WGBS strategies is missing. We perform a comparative analysis of several commonly used pre- and post-bisulfite WGBS library preparation protocols for their performance and quality of sequencing outputs. Our results show that bisulfite conversion per se is the main trigger of pronounced sequencing biases, and PCR amplification builds on these underlying artefacts. The majority of standard library preparation methods yield a significantly biased sequence output and overestimate global methylation. Importantly, both absolute and relative methylation levels at specific genomic regions vary substantially between methods, with clear implications for DNA methylation studies. We show that amplification-free library preparation is the least biased approach for WGBS. In protocols with amplification, the choice of bisulfite conversion protocol or polymerase can significantly minimize artefacts. To aid with the quality assessment of existing WGBS datasets, we have integrated a bias diagnostic tool in the Bismark package and offer several approaches for consideration during the preparation and analysis of WGBS datasets.

Mouse embryonic stem cells are dynamic and heterogeneous. For example, rare cells cycle through a state characterized by decondensed chromatin and expression of transcripts, including the Zscan4 cluster and MERVL endogenous retrovirus, which are usually restricted to preimplantation embryos. Here, we further characterize the dynamics and consequences of this transient cell state. Single-cell transcriptomics identified the earliest upregulated transcripts as cells enter the MERVL/Zscan4 state. The MERVL/Zscan4 transcriptional network was also upregulated during induced pluripotent stem cell reprogramming. Genome-wide DNA methylation and chromatin analyses revealed global DNA hypomethylation accompanying increased chromatin accessibility. This transient DNA demethylation was driven by a loss of DNA methyltransferase proteins in the cells and occurred genome-wide. While methylation levels were restored once cells exit this state, genomic imprints remained hypomethylated, demonstrating a potential global and enduring influence of endogenous retroviral activation on the epigenome.

Primordial germ cell (PGC) development is characterized by global epigenetic remodeling, which resets genomic potential and establishes an epigenetic ground state. Here we recapitulate PGC specification in vitro from naive embryonic stem cells and characterize the early events of epigenetic reprogramming during the formation of the human and mouse germline. Following rapid de novo DNA methylation during priming to epiblast-like cells, methylation is globally erased in PGC-like cells. Repressive chromatin marks (H3K9me2/3) and transposable elements are enriched at demethylation-resistant regions, while active chromatin marks (H3K4me3 or H3K27ac) are more prominent at regions that demethylate faster. The dynamics of specification and epigenetic reprogramming show species-specific differences, in particular markedly slower reprogramming kinetics in the human germline. Differences in developmental kinetics may be explained by differential regulation of epigenetic modifiers. Our work establishes a robust and faithful experimental system of the early events of epigenetic reprogramming and regulation in the germline.

Transposable elements (TEs) regulate diverse biological processes, from early development to cancer. Expression of young TEs is difficult to measure with next-generation, single-cell sequencing technologies because their highly repetitive nature means that short complementary DNA reads cannot be unambiguously mapped to a specific locus. Single CELl LOng-read RNA-sequencing (CELLO-seq) combines long-read single cell RNA-sequencing with computational analyses to measure TE expression at unique loci. We used CELLO-seq to assess the widespread expression of TEs in two-cell mouse blastomeres as well as in human induced pluripotent stem cells. Across both species, old and young TEs showed evidence of locus-specific expression with simulations demonstrating that only a small number of very young elements in the mouse could not be mapped back to the reference with high confidence. Exploring the relationship between the expression of individual elements and putative regulators revealed large heterogeneity, with TEs within a class showing different patterns of correlation and suggesting distinct regulatory mechanisms.

Erasure of DNA methylation and repressive chromatin marks in the mammalian germline leads to risk of transcriptional activation of transposable elements (TEs). Here, we used mouse embryonic stem cells (ESCs) to identify an endosiRNA-based mechanism involved in suppression of TE transcription. In ESCs with DNA demethylation induced by acute deletion of Dnmt1, we saw an increase in sense transcription at TEs, resulting in an abundance of sense/antisense transcripts leading to high levels of ARGONAUTE2 (AGO2)-bound small RNAs. Inhibition of Dicer or Ago2 expression revealed that small RNAs are involved in an immediate response to demethylation-induced transposon activation, while the deposition of repressive histone marks follows as a chronic response. In vivo, we also found TE-specific endosiRNAs present during primordial germ cell development. Our results suggest that antisense TE transcription is a "trap" that elicits an endosiRNA response to restrain acute transposon activity during epigenetic reprogramming in the mammalian germline.

Maternal overnutrition has been associated with increased susceptibility to develop obesity and neurological disorders later in life. Most epidemiological as well as experimental studies have focused on the metabolic consequences across generations following an early developmental nutritional insult. Recently, it has been shown that maternal high-fat diet (HFD) affects third-generation female body mass via the paternal lineage. We showed here that the offspring born to HFD ancestors displayed addictive-like behaviors as well as obesity and insulin resistance up to the third generation in the absence of any further exposure to HFD. These findings, implicate that the male germ line is a major player in transferring phenotypic traits. These behavioral and physiological alterations were paralleled by reduced striatal dopamine levels and increased dopamine 2 receptor density. Interestingly, by the third generation a clear gender segregation emerged, where females showed addictive-like behaviors while male HFD offspring showed an obesogenic phenotype. However, methylome profiling of F1 and F2 sperm revealed no significant difference between the offspring groups, suggesting that the sperm methylome might not be the major carrier for the transmission of the phenotypes observed in our mouse model. Together, our study for the first time demonstrates that maternal HFD insult causes sustained alterations of the mesolimbic dopaminergic system suggestive of a predisposition to develop obesity and addictive-like behaviors across multiple generations.