Most human gene promoters are embedded within CpG islands that lack DNA methylation and coincide with sites of histone H3 lysine 4 trimethylation (H3K4me3), irrespective of transcriptional activity. Here, a zinc-finger protein Cfp1 is found to be associated with non-methylated CpG islands and sites of H3K4me3 genome-wide in vivo. Cfp1 is part of the Setd1 complex which trimethylates H3K4. Artificial CpG clusters are shown to recruit Cfp1, leading to novel peaks of H3K4me3. Therefore a primary function of non-methylated CpG islands might be to genetically determine the local chromatin modification state. Most human gene promoters are embedded within CpG islands that lack DNA methylation and coincide with sites at which histone H3 lysine 4 is trimethylated (H3K4me3 sites). Here, a zinc-finger protein, Cfp1, is found to be associated with non-methylated CpG islands and H3K4me3 sites throughout the genome in the mouse brain. A primary function of non-methylated CpG islands might be to genetically determine the local chromatin modification state by interaction with Cfp1 and perhaps other CpG-binding proteins. CpG islands (CGIs) are prominent in the mammalian genome owing to their GC-rich base composition and high density of CpG dinucleotides1,2. Most human gene promoters are embedded within CGIs that lack DNA methylation and coincide with sites of histone H3 lysine 4 trimethylation (H3K4me3), irrespective of transcriptional activity3,4. In spite of these intriguing correlations, the functional significance of non-methylated CGI sequences with respect to chromatin structure and transcription is unknown. By performing a search for proteins that are common to all CGIs, here we show high enrichment for Cfp1, which selectively binds to non-methylated CpGs in vitro5,6. Chromatin immunoprecipitation of a mono-allelically methylated CGI confirmed that Cfp1 specifically associates with non-methylated CpG sites in vivo. High throughput sequencing of Cfp1-bound chromatin identified a notable concordance with non-methylated CGIs and sites of H3K4me3 in the mouse brain. Levels of H3K4me3 at CGIs were markedly reduced in Cfp1-depleted cells, consistent with the finding that Cfp1 associates with the H3K4 methyltransferase Setd1 (refs 7, 8). To test whether non-methylated CpG-dense sequences are sufficient to establish domains of H3K4me3, we analysed artificial CpG clusters that were integrated into the mouse genome. Despite the absence of promoters, the insertions recruited Cfp1 and created new peaks of H3K4me3. The data indicate that a primary function of non-methylated CGIs is to genetically influence the local chromatin modification state by interaction with Cfp1 and perhaps other CpG-binding proteins.

CpG islands (CGIs) are vertebrate genomic landmarks that encompass the promoters of most genes and often lack DNA methylation. Querying their apparent importance, the number of CGIs is reported to vary widely in different species and many do not co-localise with annotated promoters. We set out to quantify the number of CGIs in mouse and human genomes using CXXC Affinity Purification plus deep sequencing (CAP-seq). We also asked whether CGIs not associated with annotated transcripts share properties with those at known promoters. We found that, contrary to previous estimates, CGI abundance in humans and mice is very similar and many are at conserved locations relative to genes. In each species CpG density correlates positively with the degree of H3K4 trimethylation, supporting the hypothesis that these two properties are mechanistically interdependent. Approximately half of mammalian CGIs (>10,000) are “orphans” that are not associated with annotated promoters. Many orphan CGIs show evidence of transcriptional initiation and dynamic expression during development. Unlike CGIs at known promoters, orphan CGIs are frequently subject to DNA methylation during development, and this is accompanied by loss of their active promoter features. In colorectal tumors, however, orphan CGIs are not preferentially methylated, suggesting that cancer does not recapitulate a developmental program. Human and mouse genomes have similar numbers of CGIs, over half of which are remote from known promoters. Orphan CGIs nevertheless have the characteristics of functional promoters, though they are much more likely than promoter CGIs to become methylated during development and hence lose these properties. The data indicate that orphan CGIs correspond to previously undetected promoters whose transcriptional activity may play a functional role during development. In the decade since the sequence of the human genome was announced, efforts have been made to annotate all genes with their regulatory sequences. CpG islands are short regions containing the sequence CG at high density that map to regions controlling the expression of most human genes (known as promoters). Using a biochemical method, we have identified and mapped all CpG islands in the human and mouse genomes and find that over half are remote from known gene promoters—so-called “orphans.” Mice, which were thought to possess far fewer CpG islands than humans, turn out to have a very similar number. Surprisingly, orphan CpG islands in both species often mark hitherto unknown promoters. The activity of these novel promoters is particularly dynamic during normal development, as they are often silenced by DNA methylation. In colorectal cancers, however, aberrant DNA methylation affects all CpG islands equally.

Trimethylation of histone H3 Lys 4 (H3K4me3) is a mark of active and poised promoters. The Set1 complex is responsible for most somatic H3K4me3 and contains the conserved subunit CxxC finger protein 1 (Cfp1), which binds to unmethylated CpGs and links H3K4me3 with CpG islands (CGIs). Here we report that Cfp1 plays unanticipated roles in organizing genome-wide H3K4me3 in embryonic stem cells. Cfp1 deficiency caused two contrasting phenotypes: drastic loss of H3K4me3 at expressed CGI-associated genes, with minimal consequences for transcription, and creation of “ectopic” H3K4me3 peaks at numerous regulatory regions. DNA binding by Cfp1 was dispensable for targeting H3K4me3 to active genes but was required to prevent ectopic H3K4me3 peaks. The presence of ectopic peaks at enhancers often coincided with increased expression of nearby genes. This suggests that CpG targeting prevents “leakage” of H3K4me3 to inappropriate chromatin compartments. Our results demonstrate that Cfp1 is a specificity factor that integrates multiple signals, including promoter CpG content and gene activity, to regulate genome-wide patterns of H3K4me3.

Human and mouse genomes contain a similar number of CpG islands (CGIs), which are discrete CpG-rich DNA sequences associated with transcription start sites. In both species, ∼50% of all CGIs are remote from annotated promoters but, nevertheless, often have promoter-like features. To determine the role of CGI methylation in cell differentiation, we analyzed DNA methylation at a comprehensive CGI set in cells of the mouse hematopoietic lineage. Using a method that potentially detects ∼33% of genomic CpGs in the methylated state, we found that large differences in gene expression were accompanied by surprisingly few DNA methylation changes. There were, however, many DNA methylation differences between hematopoietic cells and a distantly related tissue, brain. Altered DNA methylation in the immune system occurred predominantly at CGIs within gene bodies, which have the properties of cell type–restricted promoters, but infrequently at annotated gene promoters or CGI flanking sequences (CGI “shores”). Unexpectedly, elevated intragenic CGI methylation correlated with silencing of the associated gene. Differentially methylated intragenic CGIs tended to lack H3K4me3 and associate with a transcriptionally repressive environment regardless of methylation state. Our results indicate that DNA methylation changes play a relatively minor role in the late stages of differentiation and suggest that intragenic CGIs represent regulatory sites of differential gene expression during the early stages of lineage specification.

DNA is methylated at the dinucleotide CpG in genomes of a wide range of plants and animals. Among animals, variable patterns of genomic CpG methylation have been described, ranging from undetectable levels (e.g., in Caenorhabditis elegans) to high levels of global methylation in the vertebrates. The most frequent pattern in invertebrate animals, however, is mosaic methylation, comprising domains of methylated DNA interspersed with unmethylated domains. To understand the origin of mosaic DNA methylation patterns, we examined the distribution of DNA methylation in the Ciona intestinalis genome. Bisulfite sequencing and computational analysis revealed methylated domains with sharp boundaries that strongly colocalize with ∼60% of transcription units. By contrast, promoters, intergenic DNA, and transposons are not preferentially targeted by DNA methylation. Methylated transcription units include evolutionarily conserved genes, whereas the most highly expressed genes preferentially belong to the unmethylated fraction. The results lend support to the hypothesis that CpG methylation functions to suppress spurious transcriptional initiation within infrequently transcribed genes.