Epigenetic modifications are changes to the genome that can affect gene expression without altering DNA sequence. Like DNA itself, certain epigenetic modifications can be copied faithfully when cells divide, allowing daughter cells to retain this information from their parents. This ensures that gene expression is maintained in a stable manner down cell lineages. One such modification is methylation, whereby methyl groups are added to DNA. Two papers in this issue1,2 show that there is a massive loss of DNA methylation from most of the genome immediately after fertilization in human embryos. Thus, methylation memory is erased on a global scale — an epigenetic reprogramming step that seems to be fundamental in mammals.
DNA methylation usually represses transcription, and primarily occurs on cytosine bases in the dinucleotide sequence cytosine–guanine (CpG, where p denotes the phosphate backbone of DNA, indicating that the nucleotides are on the same DNA strand). Because Watson–Crick base-pairing dictates that C pairs with G on complementary DNA strands, CpG sequences align and both strands are methylated in the same place. Therefore, methylation patterns can be passed on when cells divide, through the CpG 'memory module'. This inheritance of epigenetic information is vital in specialized cell lineages, which must maintain their identity as they divide — for example, dividing blood cells maintain their epigenetic identity to give rise to daughters that are also blood cells.
Guo et al.1 (page 606) and Smith et al.2 (page 611) analysed genome-wide DNA methylation in early-stage human embryos by high-throughput sequencing. They studied eggs, sperm, fertilized eggs (zygotes) and embryos at various stages of development, including the blastocyst stage, which occurs just before the embryo becomes implanted in the uterus, and a post-implantation stage. Both groups found that the DNA of sperm was highly methylated and that of eggs moderately so (much like mouse sperm and eggs3,4,5). However, zygotes and two-cell embryos had lost a large proportion of this methylation. In particular, Guo et al. observed marked demethylation of the paternal, sperm-derived genome, compared with more-modest demethylation of the maternal genome.
At the blastocyst stage, methylation levels remained low. This was true in all blastocyst cell types, including the cells of a structure called the inner cell mass, which are pluripotent — they can give rise to every cell of the body. Previous research indicates6 that epigenetic memory must be erased for embryonic cells to achieve pluripotency, providing a possible explanation for global demethylation. By contrast, both groups observed that, after implantation, when cells had begun to adopt tissue-specific identities, DNA methylation rapidly rose to a level characteristic of differentiated cells. After a near-total wipe-out, the epigenetic memory system was back in place (Fig. 1).
These results, combined with those from mice7,8 and other mammals9,10, suggest that global methylation reprogramming after fertilization is evolutionarily conserved. Perhaps this is because early-stage mammalian embryos undergo rapid transcriptional activation, together with early diversification of cell types — factors that necessitate a transient pluripotent state. Nonetheless, it is remarkable that the demethylation kinetics of mouse and human embryos are so similar, given that other aspects of their early development are less conserved. For example, the major transcriptional activation of the embryonic genome occurs at the two-cell stage in mouse embryos, whereas in humans it takes place at the transition between four and eight cells.
It is exciting that Guo and colleagues detected an alternative form of epigenetic modification called hydroxymethylation preferentially in the paternal genome, because hydroxymethylation is implicated in demethylation in mice11,12,13. This reinforces the idea that major mechanisms of epigenetic reprogramming are conserved in mammals. The studies did not address the mechanisms of demethylation further. Such analyses are challenging in human embryos, but Smith and co-workers have taken a first step, growing pluripotent embryonic stem cells derived from blastocyst-stage embryos in vitro, and finding that the cells become rapidly remethylated. This might be a viable system for manipulating and so studying genome-wide methylation and demethylation in human embryos, as is possible in mice.
Genome-wide analyses permit a detailed survey of distinct regions of DNA sequence whose function is known to be modified by methylation, allowing investigation of how they behave in the face of global demethylation. Such regions include 'imprinted' genes, CpG-rich genetic regions called CpG islands, and transposons (DNA sequences that can move about the genome).
Imprinted genes are those that are expressed preferentially from one parental chromosome (maternal or paternal), unlike most genes, which can be expressed from both chromosomes. Unusually, epigenetic memory in imprinted sequences is retained throughout development. The two groups confirmed this in human embryos, which they found carried methylation memories from the embryos' parents in conserved imprinted regions.
The authors found that, in contrast to sperm, human eggs had hundreds of methylated CpG islands that differed from those in mouse eggs3,4,5 and, as a general rule, these maternal epigenetic marks were not well maintained after fertilization in the embryos of mice14 or humans. Perhaps this reflects a difference in the development of the egg in the two species that is no longer relevant after fertilization. Alternatively, some of these maternal epigenetic signals may be required only in the early embryo, and thus could contribute to species differences in imprinting, particularly in the placenta15.
Transposons need to be treated with caution during reprogramming, because demethylation might cause their transcriptional activation. If they are evolutionarily 'young' and relatively unmutated, this might lead to their being able to move around in the genome, which could result in unwanted mutations. Guo and colleagues investigated one class of transposon, LINE elements, and found that evolutionarily young elements were more resistant to demethylation than their older counterparts.
The new studies provide an atlas of methylation reprogramming in early human embryos and hence a foundation for studying epigenetic regulation of human development. This is vital if we are to understand the epigenetic mechanisms that control pluripotency and differentiation. Such understanding will also help in assessing the long-term consequences of fertility interventions, including in vitro fertilization, for human health.