A role for the elongator complex in zygotic paternal genome demethylation

The life cycle of mammals begins when a sperm enters an egg. Immediately after fertilization, both the maternal and paternal genomes undergo dramatic reprogramming to prepare for the transition from germ cell to somatic cell transcription programs. One of the molecular events that takes place during this transition is the demethylation of the paternal genome. Despite extensive efforts, the factors responsible for paternal DNA demethylation have not been identified. To search for such factors, we developed a live cell imaging system that allows us to monitor the paternal DNA methylation state in zygotes. Through short-interferingRNA-mediated knockdown in mouse zygotes, we identified Elp3 (also called KAT9), a component of the elongator complex, to be important for paternal DNA demethylation. We demonstrate that knockdown of Elp3 impairs paternal DNA demethylation as indicated by reporter binding, immunostaining and bisulphite sequencing. Similar results were also obtained when other elongator components, Elp1 and Elp4, were knocked down. Importantly, injection of messenger RNA encoding the Elp3 radical SAM domain mutant, but not the HAT domain mutant, into MII oocytes before fertilization also impaired paternal DNA demethylation, indicating that the SAM radical domain is involved in the demethylation process. Our study not only establishes a critical role for the elongator complex in zygotic paternal genome demethylation, but also indicates that the demethylation process may be mediated through a reaction that requires an intact radical SAM domain. Global removal of the methyl group from 5-methyl-CpG (5mC) of DNA has been observed in at least two stages of embryogenesis. One occurs in zygotes when the paternal genome is preferentially demethylated; however, imprinted genes are resistant to this wave of DNA demethylation. Instead, this group of genes is actively demethylated in primordial germ cells from embryonic day (E)10.5 to E12.5, which results in the establishment of gender-specific methylation patterns. Given the importance of active DNA demethylation in embryogenesis, reprogramming, cloning and stem cell biology, the identification of the putative demethylase has been one of the major focuses in the field. The first molecule for which a claim was made for DNA demethylase activity was the methyl-CpG binding protein Mbd2 (ref. 8). However, Mbd2 is not required for paternal genome demethylation, as normal demethylation is still observed in Mbd2-deficient zygotes. Several recent studies in plants, zebrafish and mammalian cells have suggested that active DNA demethylation can occur through various DNA repair mechanisms. However, it is not known whether any of these proteins affect paternal genome demethylation. Both Gadd45a and Gadd45b have been implicated in DNA demethylation in somatic cells, but the role of Gadd45a in DNA demethylation has been challenged by some recent studies. To determine whether Gadd45 proteins are involved in paternal DNA demethylation in zygotes, we performed quantitative polymerase chain reaction with reverse transcription (RT–qPCR) and found that Gadd45b is the most highly expressed gene among the Gadd45 family members in zygotes (Supplementary Fig. 1a). Because Gadd45b has been shown to affect DNA demethylation in mature nonproliferating neurons, we examined whether loss of Gadd45b function affects zygotic paternal DNA demethylation. Immunostaining with the 5mC antibody indicates that paternal DNA demethylation is not affected by Gadd45b knockout, indicating that Gadd45b is not required for paternal DNA demethylation (Supplementary Fig. 1b). To facilitate the identification of factors involved in paternal DNA demethylation, we attempted to develop two molecular probes (Supplementary Fig. 2a, b). The methyl-CpG-binding domain (MBD) of Mbd1 and the CxxC domain of Mll1 have high affinity towards methyl-CpG and non-methyl-CpG, respectively. Expectedly, EGFP–MBD exhibited a nuclear dotted pattern, whereas CxxC– EGFP exhibited diffuse nuclear staining in wild-type mouse embryonic fibroblasts (MEFs; Supplementary Fig. 2c, d). In contrast, almost 100% of Dnmt1-null MEFs that lack CpG methylation exhibited punctate nuclear localization of CxxC–EGFP. Unexpectedly, the nuclear dotted pattern of EGFP–MBD was still maintained in ,60% of the double knockout cells (Supplementary Fig. 2c, d). This result indicates that when compared to EGFP–MBD, CxxC– EGFP is the better probe whose subcellular localization pattern can reflect the DNA methylation state. We further confirmed the utility of the CxxC–EGFP reporter by demonstrating that 5-Aza-dCmediated DNA demethylation resulted in a clear increase in the number and intensity of GFP bright dots in NIH3T3 cells (Supplementary Fig. 2e). We next tested whether the CxxC–EGFP probe can accurately ‘report’ paternal genome demethylation. Because injected plasmid DNA is transcriptionally inactive in 1-cell zygotes, we adapted an mRNA injection technique that allows for visualization of molecular events in the mammalian zygote as early as 3 h after introduction. We generated poly(A) mRNAs for the CxxC–EGFP as well as H2B– mRFP1 (monomeric red fluorescent protein 1) by in vitro transcription (Supplementary Fig. 2b). Using the procedure outlined in Supplementary Fig. 3a, the mRNAs were co-injected into the zygotes immediately after in vitro fertilization. Time-lapse imaging of the injected zygotes indicates that CxxC–EGFP is visible at the pronuclear stage 2 (PN2) and accumulates throughout the PN3–4 and PN5 stages in the paternal pronucleus (Supplementary Fig. 3b). The dynamics of the paternal pronuclear CxxC–EGFP accumulation mimic paternal DNA demethylation dynamics reported previously. On the basis of this result, we conclude that paternal genome demethylation can be monitored by injection of CxxC–EGFP mRNA in zygotes.