|Home | About | Journals | Submit | Contact Us | Français|
The chromatin of naive embryonic stem cells (ESCs) has a largely open configuration, as evident by the lack of condensed heterochromatin and the hypomethylation of DNA. Several molecular mechanisms promoting this constellation were previously identified. Here we present evidence for an important epigenetic function of MAD2L2, a protein originally known for its role in DNA damage repair, and for its requirement in germ cell development. We demonstrate using super-resolution microscopy that numerous MAD2L2 microfoci are exclusively associated with euchromatin, similar to other factors of the DNA damage response. In the absence of MAD2L2 the amount of heterochromatin demarcated by H3K9me2 was significantly increased. Among the most strongly suppressed genes was Dppa3, an ESC- and germ-cell-specific gene regulating DNA methylation. In Mad2l2-deficient ESCs 5-methylcytosine levels were globally increased, while several imprinted genes became hypomethylated and transcriptionally activated. Our results emphasize the important function of MAD2L2 for the open chromatin configuration of ESCs.
Open chromatin is a global, constitutive feature of naive embryonic stem cells (ESCs), which are characterized by low levels of repressive histones and DNA methylation (Gaspar-Maia et al., 2011, Meshorer and Misteli, 2006). Diverse factors are required to maintain ESC chromatin in a decondensed state. They are important to stabilize pluripotency and to maintain the potential for self-renewal and differentiation. DPPA3 (also called PGC7 or Stella) is one of the factors known to modulate epigenetic processes in both embryonic and germ cell development (Payer et al., 2003). It is a highly specific, maternal effect gene required for early embryonic cells and primordial germ cells (PGCs), and demarcates naive versus primed pluripotency (Hayashi et al., 2008, Ohinata et al., 2008). Heterogeneous expression of DPPA3 was detected in ESC populations growing in Lif/serum as well as LIF/2i medium (Hayashi et al., 2008, Singer et al., 2014). DPPA3 heterogeneity was concluded to result from a combination of low burst frequencies and large burst sizes (Singer et al., 2014). DPPA3 is known to be involved in the maintenance of the low level of DNA methylation typical for early embryonic cells as well as ESCs (Hayashi et al., 2008). At the same time, it protects imprinted loci against demethylation (Nakamura et al., 2007). DPPA3 is required for the generation of fully reprogrammed pluripotent stem cells, in particular for the erasure of the epigenetic memory encoded in the DNA methylation pattern (Xu et al., 2015b).
Both ESCs and PGCs, the two cell types characterized by the expression of DPPA3, require the MAD2L2 protein for the proper establishment of their epigenetic status (Pirouz et al., 2013, Pirouz et al., 2015). In the absence of Mad2L2 PGCs do not reprogram toward a more flexible chromatin configuration after their induction, and ESCs are not able to propagate their typical open chromatin, which is necessary for stable pluripotency.
MAD2L2 (also called REV7 or MAD2B) was originally identified by its function in DNA repair, playing an accessory role in translesion DNA repair and an inhibitory role in the resection of 5′ ends after DNA double-strand breaks and at unprotected telomeres (Boersma et al., 2015, Xu et al., 2015a). As a downstream component of the DNA damage response pathway, MAD2L2 promotes non-homologous end joining versus homologous recombination. The role of MAD2L2 in DNA repair, ESC development, and PGC development is tightly connected to the epigenetic status of the respective cells. To investigate the relationship between MAD2L2 and the formation of open chromatin we compared wild-type and Mad2l2-deficient ESCs, employing super-resolution microscopy, transcriptome sequencing, methylome sequencing, and chromatin immunoprecipitation sequencing (ChIP-seq) analysis. We identified Dppa3 as a downstream target of MAD2L2, and derived a model explaining the epigenetic effects of MAD2L2.
ESCs are known to express variable levels of specific RNAs and/or proteins, depending among other factors on growth conditions. This indicates a highly dynamic regulation of gene expression and reflects a metastable equilibrium between distinct cellular states. We characterized the MAD2L2 protein in wild-type ESCs and compared the dynamics with classical ESC markers. Individual cells of a colony expressed significantly different levels of MAD2L2 (Figure 1A and Movie S1), either very high (4.8% out of 1,000 evaluated, cells from five different fields), high (72%), or low (23.2%), resembling the heterogeneity observed for DPPA3 expression (Figure 3A) (Hayashi et al., 2008). In contrast, both MAD2L2-positive and -negative ESCs expressed Oct4 (Figure S1A). The differentiation of ESCs into epiblast-like cells (EpiLC) represents the transition from naive to primed pluripotency (Hayashi et al., 2011). This was accompanied by a drastic downregulation of the MAD2L2 protein (Figure S1B) and Dppa3 mRNA (Figure 3C). Together, these observations suggest a dynamic fluctuation from a state of naive pluripotency demarcated by Mad2l2 and Dppa3 expression to a primed state with both markers off. EpiLCs can be further converted to unipotent PGC-like cells (PGCLCs), close in vitro equivalents of PGCs, on embryonic day 9 (Hayashi et al., 2011). PGCLCs represent exactly the developmental stage at which a crucial function of MAD2L2 was previously demonstrated (Pirouz et al., 2013). High expression levels of both MAD2L2 and DPPA3 are typical for germ cells including PGCLCs, and for naive ESCs (Figures S2A and and33C).
In ESCs the MAD2L2 protein was associated with the chromatin fraction; however, it did not decorate condensed, mitotic chromosomes (Figures 1I and 1J). Super-resolution microscopy revealed that it was distributed in microfoci, not overlapping with the DNA-enriched heterochromatin near the nuclear lamina and around the nucleolus, as identified by H3K9me2 (Figures 1B and 1C) (Willig et al., 2006). High levels of MAD2L2 protein were found in many ESCs of a colony in regions separate from transcriptionally inactive chromatin, as identified by H3K27me3 (Figures 1E and 1F). H3K27me3 accumulated in more central areas of ESC nuclei, compared with the peripheral localization of H3K9me2 (Figures 1C and 1E). ESC colonies always additionally contained cells with either predominantly MAD2L2 or predominantly H3K27me3, but not both, proteins (Figures 1D and 1F). More precisely, 70.6% of wild-type ESCs (out of 1,141 evaluated cells, from four fields) predominantly expressed MAD2L2 in the absence of H3K27me3, 23.6% had a non-overlapping expression of both MAD2L2 and H3K27me3, and 5.7% predominantly expressed H3K27me3 in the absence of MAD2L2. These numbers reflect the highly dynamic chromatin status in ESCs. Line scan histogram profiles through single nuclei demonstrated that MAD2L2 did not co-localize with compacted DNA and the heterochromatin markers H3K9me2, H3K27me3, and HP1α (Figure S1C). In contrast to its exclusion from heterochromatin, MAD2L2 co-localized completely with the euchromatin identified by H3K4me3 and AcH4, which are the dominant histone modifications in ESCs growing in LIF/2i (Figures 1G and 1H).
Like ESCs, PGCLCs also express significant levels of MAD2L2 (Figure S2A). To investigate whether MAD2L2 is instrumental for the balance between open and closed chromatin, we analyzed Mad2l2-deficient ESCs and PGCLCs. Mutant ESCs had an increased amount of heterochromatin, as indicated by the global levels of H3K9me2 and H3K27me3, and by a significantly increased number of cells with high levels of H3K9me2 and H3K27me3 (Figure 1K). The differentiation of ESCs via EpiLCs into PGCLCs was normally accompanied by the re-induction of the Oct4-GFP transgene and by a loss of the high levels of H3K9me2 typical for EpiLCs (Figure S2B). However, Mad2l2-deficient PGCLCs maintained high levels of H3K9me2 and were not able to reprogram toward H3K27me3 (Figure S2C). This resembled the initial induction of Mad2l2-deficient PGCs and their defective epigenetic reprogramming observed in vivo (Pirouz et al., 2013). Taken together, we found a correlation between the presence of Mad2l2 and an open chromatin configuration, as is typical for naive pluripotency and germ cells after their epigenetic reprogramming. In the absence of Mad2l2, ESCs accumulated heterochromatin and lost their naive pluripotency. Mad2l2-deficient PGCLCs were locked in their stably repressed chromatin configuration.
Not only MAD2L2, but also other DNA damage response factors accumulated in microfoci in the nuclei of ESCs, although there was no prior induction of DNA damage (Figures 2A–2C) (Banath et al., 2009, Turinetto et al., 2012). Phosphorylation of histone H2AX and accumulation γH2AX foci around the sites of DNA breaks are usually an immediate-early response to DNA damage. In ESCs they were mainly localized, but not necessarily restricted, to areas of low DNA density (Figure 2A). Foci of the E3 ligase RNF168 and of the large adaptor protein 53BP1 were restricted to open chromatin, and their sizes were smaller, quite similar to the MAD2L2 microfoci (Figures 2B and 2C).
ESCs were co-cultured on feeder layers consisting of mitomycin-treated (i.e., DNA damaged) murine embryonic fibroblasts (mMEFs). This allowed a side-by-side comparison of foci formed in response to DNA damage with ESC foci (Figures 2D–2F). Mitomycin C-induced MAD2L2 foci varied widely in size (1–3.5 μm) in the intensity of staining for MAD2L2 and in numbers per nucleus, clearly due to the random nature of DNA damage induction. Focus sizes in ESCs were generally smaller (approximately 0.2–0.3 μm) and the numbers per cell were drastically higher, being homogeneously distributed throughout the open chromatin regions of ESC nuclei, separate from DAPI-stained, condensed heterochromatin (Figures 2D–2F). Thus, the parallel scanning of ESCs and mMEFs revealed an inverse relationship between the level of MAD2L2 and the presence of DAPI-stained heterochromatin. It also demonstrated the principal differences underlying focus formation in response to DNA damage and in ESCs.
We had previously compared the transcriptomes of Mad2l2-deficient and wild-type ESCs, and identified major transcriptional changes (Pirouz et al., 2013). A re-evaluation of the transcriptome data (see Experimental Procedures) revealed that the fourth most significantly downregulated gene was Dppa3 (Figure 3B). Downregulation of Dppa3 in the absence of MAD2L2 was confirmed by RT-PCR for ESCs, EpiLCs, and PGCLCs (Figure 3C), and for ESCs also by immunofluorescence (Figure 3A). We used the BVSC cell line to visualize Dppa3 expression by fluorescence of the DPPA3-eCFP fusion protein. DPPA3-positive and -negative ESCs were sorted based on expression of the ECFP reporter, and qPCR confirmed a positive correlation between Dppa3 and Mad2l2 levels. When sorted ESCs with a low level of DPPA3 were recultured in LIF/2i medium, 7% of the cells had upregulated the expression of Dppa3, suggesting that Mad2l2 expression also could recover (Figures S3A–S3C).
ChIP-seq analysis of the Dppa3 locus revealed that its transcriptional repression in Mad2l2-deficient ESCs was associated with the deposition of the repressive histone H3K9me2 in two positions upstream of the transcriptional start site (Figure 3D). The expression of Dppa3 could be rescued by the introduction of a vector encoding a MAD2L2-GFP fusion protein into Mad2l2-deficient ESCs (Figure 3E). In this experiment the reactivation of the Dppa3 gene occurred in the absence of DPPA3 protein. Thus, the initial effect of MAD2L2 was exerted independently of the epigenetic function(s) of DPPA3. Once derepressed, the Dppa3 locus could be transcribed in ESCs, probably under control of the core pluripotency factor PRDM14, which was previously shown to be necessary for Dppa3 expression in PGCLCs (Magnusdottir et al., 2013). Remarkably, ESC colonies overexpressing GFP-MAD2L2 expressed DPPA3 as visualized by a Dppa3-eCFP transgene uniformly without an indication for heterogeneity (Figure 3F).
We measured a significant increase of the global levels of 5-methylcytosine in Mad2l2-deficient ESCs compared with wild-type cells, although the high level found in EpiLCs was not reached (Figure 4A). Analysis of DNA methylation by methylated DNA immunoprecipitation sequencing (MeDIP-seq) confirmed the observed global difference, and allowed the display of DNA methylation at a single-nucleotide resolution (Figure 4B). These data revealed substantial, genome-wide perturbations of methylation patterns. For example, they demonstrated the presence of three methylation peaks instead of one around transcriptional start sites in mutant ESCs (Figures 4C and S4). Noteworthy, however, is that the Dppa3 gene and its upstream region remained largely unmethylated in both wild-type and mutant ESC. In early embryonic cells several genes maintain their methylation in a widely undermethylated epigenetic landscape, and consequently are transcriptionally repressed (Habibi et al., 2013). DPPA3 was previously shown to play a role in this imprinting of genes (Bortvin et al., 2004, Kang et al., 2013, Nakamura et al., 2007). To study this process in Mad2l2-deficient, and consequently Dppa3-repressing, ESCs, we compared the RNA-seq datasets with the MeDIP-seq data. A total of 221 genes were methylated in wild-type ESCs but were unmethylated and transcriptionally upregulated in the Mad2l2-deficient ESCs (Figure 4D). Of these, the top five (Peg3, Trp73, Ampd3, Mest, and Nnat) were known imprinted genes (http://www.geneimprint.com). Taken together, the perturbation of DNA methylation in Mad2l2−/− ESCs not only resulted in a general increase of methylation, but also affected imprinting through demethylation of specific loci. Although the Mad2l2-deficient ESCs were shown to be pluripotent and capable of contributing to chimeras (Pirouz et al., 2015), they were not classified as naively pluripotent cells with regard to the increased level of DNA methylation.
The absence of MAD2L2 in mutant ESCs corresponds with the low MAD2L2 levels observed in somatic cells (Figure 4E). Cells with a low level of MAD2L2 have in common that the inhibitory binding of MAD2L2 to EHMT1 and EHMT2 is reduced (Pirouz et al., 2013). Thus, histone H3 can be efficiently methylated on residue K9, and H3K9me2 can then form a docking place for further epigenetic regulators. DNA methyltransferase DNMT1 is recruited together with UHRF1 and EHMT2 at the replication fork, where it will also bind to proliferating cell nuclear antigen, and cooperatively to hemimethylated DNA (Esteve et al., 2006, Liu et al., 2013, Rothbart et al., 2012). Together, these and other interacting factors represent an assembly of both chromatin writers and readers, which maintain and establish a cell-specific chromatin configuration (Liu et al., 2013, Shinkai and Tachibana, 2011). Collectively, the low levels of MAD2L2 and DPPA3 correspond to high levels of repressive histone modifications, substantial DNA methylation, and thus a closed chromatin configuration. In contrast, MAD2l2 levels are high in wild-type ESCs and PGCLCs, downstream of a cascade resembling the DNA damage response pathway (Figures 1A–1I and S2A). In such cells MAD2L2 can directly bind to both EHMT1 and EHMT2, suppress EHMT2 expression, and thus inhibit histone H3K9 and DNA methylation (Pirouz et al., 2013). By antagonizing the formation of H3K9me2, MAD2L2 will derepress the Dppa3 gene. DPPA3 can then displace DNMT1 from UHRF1 at the replication fork, so that the growing strand cannot be methylated (Figure 4F; Funaki et al., 2014). In any case, the combined presence of MAD2L2 and DPPA3 in ESCs correlates with their generally low levels of both DNA and histone H3K9 methylation, i.e., a generally open chromatin configuration. In conclusion, the global promotion of open chromatin, and the inhibition of heterochromatin in ESCs and germ cells (PGCLCs and PGCs) depend critically on MAD2L2. The local opening of chromatin observed around DNA breaks may involve MAD2L2 and a similar mechanism (Noon et al., 2010). It is conceivable that the major importance of MAD2L2 lies in its epigenetic functions, be it in a stem or germ cell context, or in response to DNA damage.
Mad2l2-deficient ESCs carrying an Oct4-GFP allele (Pirouz et al., 2015), and BVSC ESCs, with an insertion of the CFP reading frame into the Dppa3 locus (Ohinata et al., 2008) were cultured in LIF/2i medium (Ying et al., 2008). For PGCLC formation, ESCs were first differentiated for 2 days into EpiLCs in the presence of activin and fibroblast growth factor, and subsequently for 6 days in PGCLC differentiation medium as described by Hayashi et al. (2011). See Supplemental Experimental Procedures for further details.
Three independent biological replicates of 10,000 cells each were read by an Accuri C6 cytometer (BD Biosciences). See Supplemental Experimental Procedures for further details.
Three independent biological replicates were prepared from 108 cells each, and were fractionated into cytoplasmic, nucleoplasmic, and chromatin fractions. See Supplemental Experimental Procedures for further details.
At least three independent replicates, with two technical replicates each, were performed using the following primer pairs: Gapdh (5′-GTC GTG GAG TCT ACT GGT GTC-3′ and 5′-GAG CCC TTC CAC AAT GCC AAA-3′), Dppa3 (5′-CGT CCT ACA ACC AGA AAC AC-3′ and 5′-CTG CTC AAT CCG AAC AAG TC-3′), and Mad2l2 (5′-GTT GCC TTG AGT CCC TAC AG-3′ and 5′-TCC ACA TCG TTC TTC TCC AG-3′).
Immunocytochemistry was performed as specified in Supplemental Experimental Procedures. Sections were analyzed by confocal or stimulated emission depletion microscopy.
Dot-blot analysis was performed as described by Hayashi et al. (2011). Three independent replicates of 50 ng DNA were applied to a membrane filter, and the methylation level was detected with an antibody against 5mC (Abcam).
M.K. initiated, led, and managed the project, and wrote the paper. A.R. co-designed the study, performed the majority of experiments and the bioinformatics analyses, and contributed critically to text and figures. M.P. generated the Mad2l2-deficient ES line, and was involved in the early concept of the project. M.d.V. helped with interpretations. D.K. performed stimulated emission depletion microscopy.
The work was supported by the Max Planck Society. M.D.V. is a Helmholtz Association young investigator. We thank M. Saitou (Kyoto University) and M. Boiani (MPI Muenster) for ESC lines, G. Salinas (Goettingen University) for sequencing, and H. Jäckle for his hospitality.
Published: March 16, 2017
Supplemental Information includes Supplemental Experimental Procedures, four figures, and one movie and can be found with this article online at http://dx.doi.org/10.1016/j.stemcr.2017.02.011.
Sequencing data are available from the European Nucleotide Archive under accession number ENA: PRJEB19344 at http://www.ebi.ac.uk/ena/data/view/PRJEB19344.