In mammals, cytosine methylation is predominantly restricted to CpG dinucleotides and stably distributed across the genome, with local, cell type-specific regulation directed by DNA binding factors1-3. This comparatively static landscape dramatically contrasts the events of fertilization, where the paternal genome is globally reprogrammed. Paternal genome demethylation includes the majority of CpGs, though methylation is maintained at several notable features4-7. While these dynamics have been extensively characterized in the mouse, only limited observations are available in other mammals, and direct measurements are required to understand the extent to which early embryonic landscapes are conserved8-10. We present genome-scale DNA methylation maps of human preimplantation development and embryonic stem cell (ESC) derivation, confirming a transient state of global hypomethylation that includes most CpGs, while sites of persistent maintenance are primarily restricted to gene bodies. While most features share similar dynamics to mouse, maternally contributed methylation is divergently targeted to species-specific sets of CpG island (CGI) promoters that extend beyond known Imprint Control Regions (ICRs). Retrotransposon regulation is also highly diverse and transitions from maternally to embryonically expressed, species-specific elements. Together, our data confirm that paternal genome demethylation is a general attribute of early mammalian development that is characterized by distinct modes of epigenetic regulation.

Human pluripotent stem cells (hPSCs) can self-renew indefinitely, making them an attractive source for regenerative therapies. This expansion potential has been linked with acquisition of large copy number variants (CNVs) that provide mutant cells with a growth advantage in culture1–3. However, the nature, extent, and functional impact of other acquired genome sequence mutations in cultured hPSCs is not known. Here, we sequenced the protein-coding genes (exomes) of 140 independent human embryonic stem cell (hESC) lines, including 26 lines prepared for potential clinical use4. We then applied computational strategies for identifying mutations present in a subset of cells5. Though such mosaic mutations were generally rare, we identified five unrelated hESC lines that carried six mutations in the TP53 gene that encodes the tumor suppressor P53. Notably, the TP53 mutations we observed are dominant negative and are the mutations most commonly seen in human cancers. We used droplet digital PCR to demonstrate that the TP53 mutant allelic fraction increased with passage number under standard culture conditions, suggesting that P53 mutation confers selective advantage. When we then mined published RNA sequencing data from 117 hPSC lines, we observed another nine TP53 mutations, all resulting in coding changes in the DNA binding domain of P53. Strikingly, in three lines, the allelic fraction exceeded 50%, suggesting additional selective advantage resulting from loss of heterozygosity at the TP53 locus. As the acquisition and favored expansion of cancer-associated mutations in hPSCs may go unnoticed during most applications, we suggest that careful genetic characterization of hPSCs and their differentiated derivatives should be carried out prior to clinical use.

Summary Although distinct human induced pluripotent stem cell (hiPSC) lines can display considerable epigenetic variation, it has been unclear if such variability impacts their utility for disease modeling. Here, we show that although low passage female hiPSCs retain the inactive X-chromosome of the somatic cell they are derived from, over time in culture they undergo an “erosion” of X-chromosome inactivation (XCI). This erosion of XCI is characterized by loss of XIST expression and foci of H3-K27-trimethylation, as well as transcriptional de-repression of genes on the inactive X that cannot be reversed by either differentiation or further reprogramming. We specifically demonstrate that erosion of XCI has a significant impact on the use of female hiPSCs for modeling Lesch-Nyhan syndrome. However, our finding that most genes subject to XCI are de-repressed by this erosion of XCI suggests that it should be a significant consideration when selecting hiPSC lines for modeling any disease.