We report here a DNA methylation profiling of promoters and 5' end of genes in sperm and embryos developing through the MBT, a period of intense chromatin remodeling [7
]. Both sperm and embryonic genomes are marked by areas of methylation and hypomethylation; yet a striking observation is the significant proportion of genes methylated in sperm and unmethylated in pre-MBT, MBT and post-MBT embryos. In the context of the global demethylation and remethylation of the zebrafish genome that follows fertilization [7
], it is possible that these genes escape post-fertilization remethylation. Their hypomethylated state argues for transcriptional permissiveness, in line with their homeostatic and developmental functions.
Nevertheless, nearly all genes methylated in embryos are also methylated in sperm. This overlap is consistent with a transgenerational inheritance of DNA methylation states through fertilization. This view is supported by the incomplete erasure of parental promoter methylation in the mouse [2
] and the lack of global demethylation in Xenopus
]. Our data are also compatible with a resetting of DNA methylation after fertilization, as promoter methylation changes reported here parallel global methylation changes [7
], implying demethylation and remethylation phases. This interpretation is in line with post-fertilization methylation events observed in mouse [2
] and Xenopus
] embryos. A demethylation/remethylation model would nevertheless argue that a site-specific methylation determinant be perpetuated through fertilization. A high CG content may be a key genetic determinant for inheritance of the hypomethylated state, while determinants of the methylation state may speculatively include non-coding RNAs, which have been identified in sperm [30
]. This exciting question remains to be examined closely.
A large fraction of genes hypomethylated in embryos are also hypomethylated in sperm ([27
] and this paper). Because these genes are enriched in transcription regulation and developmental functions, hypomethylation may confer, already in sperm, an instructive role for the developmental gene expression program. Similarly, hypomethylation and homeostatic functions of maternally expressed genes [10
] support a model of an instructive role of the hypomethylated state in female gametes. One such key instructive function may be the targeting of H3K4me3 after fertilization, which takes place on DNA hypomethylated regions, preferentially occurs on maternally expressed genes, and has been proposed to confer a propensity for transcriptional activation after ZGA onset [13
Several loci of developmentally important genes stand out as hypomethylated in sperm and embryos, but are methylated in ZF4 fibroblasts. These genes are largely CG-rich. Interestingly, genes associated with clusters of three or more CGIs have been shown to be genetically marked in a manner that distinguishes them from genes associated with a single CGI [28
]. Genes in CGI clusters mostly encode transcription factors with sequence-specific DNA binding properties (such as hox
genes), while single CGI genes mainly harbor housekeeping functions [28
]. We find that genes in CGI clusters overlap with regions hypomethylated only in embryos and are methylated in fibroblasts; this is strongly suggestive of a net methylation gain on developmentally regulated genes that no longer need to be expressed in fibroblasts, in order to elicit long-term repression of the developmental functions these genes are involved in. In contrast, genes with single CGIs hypomethylated in embryos remain hypomethylated in fibroblasts, consistent with the cellular homeostatic functions of these genes (these observations also incidentally indicate that methylation detected in ZF4 fibroblasts is not a mere consequence of their cultured state, but reflects a differentiation process). Differential methylation of genes associated with single versus multiple CGIs elicited by differentiation may be due to developmental stage-dependent or tissue-specific alternative Dnmt usage [16
]. An intriguing question is whether genetic determinants also exist for other types of epigenetic marks and if such a genomic component might have a stronger impact on the epigenome during early development than in differentiated cells.
We previously showed that a greater proportion of maternal genes than zygotic genes are marked by H3K4me3 after ZGA [13
]. Our DNA methylation profiling provides an epigenetic basis for this observation by showing that maternal promoters are enriched in hypomethylated HCPs, and that most hypomethylated promoters are marked by H3K4me3. Indeed, in mammalian cells CpG density within CGIs correlates with H3K4me3 levels [32
], and unmethylated CGIs are sufficient to target SETD1-mediated H3K4me3 in concordance with recruitment of the adaptor protein CFP1 [26
]. Our RNA-seq data indicate that, in zebrafish, both homologues of the human H3K4 histone methyltransferase SETD1A
genes are maternally expressed and transcripts remain detected at high levels until the MBT, after which they are degraded (data not shown) [10
]. Interestingly, a similar transcript profile occurs for cfp1
, suggesting that the Setd1/Cfp1 histone methyltransferase complex may operate post-fertilization in zebrafish, for H3K4 trimethylation on hypomethylated promoters in the pre-MBT period.
Cohorts of genes differentially expressed through the MBT display distinct methylation patterns related to CG content rather than transcription status. Maternal genes are enriched in HCPs, suggesting that a sequence determinant may define their maternal status. Their hypomethylated state may enable transcription during oogenesis; however, we show that their transcriptional fate is not determined by methylation state. In addition, zygotic genes contain a higher proportion of methylated promoters than maternal genes, nearly as high as non-expressed genes, consistent with a lack of correlation between promoter methylation and embryonic gene expression (see also [6
]). In the mouse, however, promoter methylation seems to restrict gene activation, until demethylation associated with tissue-specific differentiation [2
]. Anamniote vertebrate development may thus mainly rely on changes in histone modifications rather than DNA methylation to regulate gene expression changes at the onset of ZGA [14
]. Additional experiments will be required to dissect the genetic and epigenetic determinants of chromatin states and gene expression patterns during zebrafish development.