Human pluripotent stem cells (PSCs) are subject to the appearance of recurrent genetic variants on prolonged culture. We have now found that, compared with isogenic differentiated cells, PSCs exhibit evidence of considerably more DNA damage during the S phase of the cell cycle, apparently as a consequence of DNA replication stress marked by slower progression of DNA replication, activation of latent origins of replication, and collapse of replication forks. As in many cancers, which, like PSCs, exhibit a shortened G1 phase and DNA replication stress, the resulting DNA damage may underlie the higher incidence of abnormal and abortive mitoses in PSCs, resulting in chromosomal non-dysjunction or cell death. However, we have found that the extent of DNA replication stress, DNA damage, and consequent aberrant mitoses can be substantially reduced by culturing PSCs in the presence of exogenous nucleosides, resulting in improved survival, clonogenicity, and population growth.

We postulate that exit from pluripotency involves intermediates that retain pluripotency while simultaneously exhibiting lineage-bias. Using a MIXL1 reporter, we explore mesoderm lineage-bias within the human pluripotent stem cell compartment. We identify a substate, which at the single cell level coexpresses pluripotent and mesodermal gene expression programmes. Functionally these cells initiate stem cell cultures and exhibit mesodermal bias in differentiation assays. By promoting mesodermal identity through manipulation of WNT signalling while preventing exit from pluripotency using lysophosphatidic acid, we ‘trap’ and maintain cells in a lineage-biased stem cell state through multiple passages. These cells correspond to a normal state on the differentiation trajectory, the plasticity of which is evidenced by their reacquisition of an unbiased state upon removal of differentiation cues. The use of ‘cross-antagonistic’ signalling to trap pluripotent stem cell intermediates with different lineage-bias may have general applicability in the efficient production of cells for regenerative medicine.

The timing of DNA replication in mammals is crucial for minimizing errors and influenced by genome usage and chromatin states. Replication timing in the newly formed mammalian embryo remains poorly understood. Here, we have investigated replication timing in mouse zygotes and 2-cell embryos, revealing that zygotes lack a conventional replication timing program, which then emerges in 2-cell embryos. This program differs from embryonic stem cells and generally correlates with transcription and genome compartmentalization of both parental genomes. However, consistent and systematic differences existed between the replication timing of the two parental genomes, including considerably later replication of maternal pericentromeric regions compared to paternal counterparts. Moreover, maternal chromatin modified by Polycomb Repressive Complexes in the oocyte, undergoes early replication, despite belonging to the typically late-replicating B-compartment of the genome. This atypical and asynchronous replication of the two parental genomes may advance our understanding of replication stress in early human embryos and trigger strategies to reduce errors and aneuploidies.

Copy number variants (CNVs) are genomic rearrangements implicated in numerous congenital and acquired diseases, including cancer. The appearance of culture-acquired CNVs in human pluripotent stem cells (PSCs) has prompted concerns for their use in regenerative medicine. A particular problem in PSC is the frequent occurrence of CNVs in the q11.21 region of chromosome 20. However, the exact mechanism of origin of this amplicon remains elusive due to the difficulty in delineating its sequence and breakpoints. Here, we have addressed this problem using long-read Nanopore sequencing of two examples of this CNV, present as duplication and as triplication. In both cases, the CNVs were arranged in a head-to-tail orientation, with microhomology sequences flanking or overlapping the proximal and distal breakpoints. These breakpoint signatures point to a mechanism of microhomology-mediated break-induced replication in CNV formation, with surrounding Alu sequences likely contributing to the instability of this genomic region.

Epigenomes of mammalian oocytes and embryos undergo major transitions essential for successful development. Here, we provide genome-wide maps of histone variant H2A.Z during twelve stages of mouse oogenesis and preimplantation embryo development and relate it to histone marks and genomic features. This revealed that major waves of H2A.Z incorporation occur early in growing oocytes, forming distinct patterns of maternal, embryonic, and persistent H2A.Z enrichment. Late maternal enrichment is inherited by the zygote and precedes reduced formation of lamina associated domains and early replication in the maternal genome of 2-cell embryos. Persistent H2A.Z enrichment is strongly associated with CpG islands and H3K4me3 near transcription start sites of active genes, but thousands of maternal and embryonic H2A.Z incorporation sites exist elsewhere, frequently at transposable elements. The persisting H2A.Z enrichments across related developmental stages enable preservation of epigenetic information despite major concurrent changes in H3K4me3, H3K27me3, and DNA methylation. Altogether, this advances our understanding of how histone variants contribute to epigenetic reprogramming during mammalian oogenesis and early development.