In vertebrates, genomic DNA is often methylated at the 5th position of cytosine in the sequence of CpG. The methyl group is transferred from S-adenosyl-L-methionine by DNA methyltransferase. This DNA methylation is the only chemical modification that genomic DNA receives under physiological conditions. The DNA methylation in vertebrates, different from prokaryotes, functions as one of the regulatory mechanisms for gene silencing. The promoter regions of genes that are actively transcribed are usually undermethylated, while those of silent genes are heavily methylated. The suppression of gene expression is mediated by inhibiting the binding of transcription factors or by recruiting the methylated DNA binding proteins to the methylated DNA. It has been reported that the methylated DNA binding proteins and the DNA methyltransferases recruit histone deacetylase (HDAC) to deacethylate core histones in nucleosome. Such the deacetylation is a trigger to induce methylation of core histones, a sign of heterochromatin. On the other hand, the methylation of histone H3 position at Lys 9 recruits DNA methyltransferase. DNA methylation is both a cause for and a result of heterochromatinization.

DNA methylation of the genome is essential for development and plays crucial roles in a variety of biological processes, such as tissue-specific gene expression, genomic imprinting, X-chromosome inactivation, timing of DNA replication, and carcinogenesis via suppression of certain genes. Our final goal is to elucidate the mechanisms how the genomic DNA methylation and the expression of methylated genes are regulated. We have obtained the following results up to the present.

1. Creation of DNA Methylation Patterns

The methylation pattern of genomic DNA is dynamic and changes during early stage of embryogenesis, cell differentiation, and formation of germ cells. In vertebrates, two types of DNA methyltransferase activities have been reported, i.e., de novo and maintenance types. In mouse, de novo type DNA methylation creates tissue-specific methylation patterns at the implantation stage, and maintenance type methylation ensures clonal transmission of lineage-specific methylation patterns during replication. Two DNA methyltransferases, Dnmt3a and Dnmt3b, are responsible for the creation of methylation patterns. Recently, Dnmt3L, a member of the Dnmt3 family, has been genetically shown to be indispensable for the global methylation in germ cells.

In mammals, genome-wide DNA methylation patterns are formed at the implantation stage and during gametogenesis. In germ cells, this global DNA methylation contributes also to the establishment of genomic imprinting. Although it has been shown that Dnmt3a and Dnmt3b are responsible for the creation of the DNA methylation patterns, it has not been elucidated that how each DNA methyltransferase methylates distinct region of the genome in vivo. Knockout of Dnmt3b gene gives a severer phenotype than that of Dnmt3a, and leads to hypomethylation in the satellite 2 and 3 regions of specific chromosomes, which is a cause of the ICF (immunodeficiency, centromeric region instability, and facial anomalies) syndrome in man. The results clearly indicate that Dnmt3a and Dnmt3b methylate distinct genomic regions in vivo. However, we have shown that the sequence specificities and kinetic parameters of the de novo DNA methylation activities of Dnmt3a and Dnmt3b are similar to each other, and thus do not explain their distinct targeting of DNA methylation in vivo.

To explain their distinct DNA methylation regions, other than the sequence specificity of the target DNA and the kinetic properties, the timing of expression of Dnmt3a and Dnmt3b must be considered. We have shown that during early embryogenesis, Dnmt3b1, one of the isoforms of Dnmt3b, is specifically expressed in totipotent embryonic cells, such as inner cell mass, epiblast and embryonic ectoderm cells, while Dnmt3a is significantly and ubiquitously expressed after 10.5-day embryos. This timing of expression of Dnmt3b1 coincides with the global methylation of genomic DNA at the implantation stage. We have also shown that Dnmt3a2, a short form of Dnmt3a, and Dnmt3L, a necessary factor for DNA methylation in germ cells, are highly expressed in a stoichiometric manner in male germ cells in 17-day embryos, at which stage genome-wide DNA methylation occurs. As the conditional knockout of Dnmt3a gene in mouse germ cells erases the methylation imprint in germ cells, this stage-specific expression of Dnmt3a2 may contribute to the global methylation of genomic DNA in germ cells. Accordingly, we propose that such distinct timing of expression of DNA methyltransferases may be one of the reasons for the distinct DNA methylation regions by these two enzymes.

It should be noted that Dnmt3a is ubiquitously expressed in somatic cells without aberrant methylation. This clearly indicates that, other than the difference in timing of expression, these enzymes should methylate the distinct DNA regions by different mechanisms in vivo. DNA in the eukaryotic nucleus is packaged into a highly compact nucleoprotein complex, i.e., chromatin, of which the elemental structure is nucleosomes. Interestingly, when the Lsh gene in mouse, the ATRX gene in man, or the DDM1 gene in Arabidopsis is mutated, the DNA methylation level decreases. The products of these genes are components of ATP-dependent chromatin remodeling factors of the SWI2/SNF2 family. In addition, Dnmt3a and/or Dnmt3b interact directly with chromatin remodeling factor SNF2H, which is a member of the ISWI family, and with factors that modify core histones such as histone deacetylase, histone methyltransferase, and HP1, which specifically binds K9-methylated histone H3. These modifications and binding proteins are the trigger for induction of the conversion of chromatin into inactive heterochromatin. Accordingly, the DNA methylation activity of Dnmt3a and Dnmt3b towards chromatin or its elemental structure, nucleosomes, should be considered to determine how specific regions of genomic DNA are methylated.

We reconstituted mononucleosomes from recombinant histones H2A, H2B, H3, and H4 with DNAs differing in length and sequence, and then used them as substrates for determination of the DNA methylation activity of recombinant Dnmt3a and Dnmt3b. The DNA methylation activity of both Dnmt3a and Dnmt3b was severely inhibited on the formation of nucleosomes. Interestingly, the DNA methylation activity towards the naked DNA region of reconstituted nucleosomes was much higher for Dnmt3a than Dnmt3b, and that towards the DNA within the nucleosome core region was significant for Dnmt3b and negligible for Dnmt3a. We propose that the preferential DNA methylation activity of Dnmt3a towards the naked part of nucleosomal DNA and the significant methylation activity of Dnmt3b towards the nucleosome core region contribute to their distinct methylation of genomic DNA in vivo. The stage-specific high-level expression of Dnmt3b contributes to the global methylation of the genome, even when it is packaged into nucleosomes; and the ubiquitous expression of Dnmt3a promptly methylates genomic DNA at the time when it becomes partially naked, for example, through the action of a chromatin-remodeling factor.

It has been genetically shown that Dnmt3L is a necessary factor for the DNA methylation in germ cells, however, it is not known what the molecular mechanism underlying the de novo DNA methylation in the presence of Dnmt3L is. We examined the in vitro effect of Dnmt3L on the DNA methylation activity of Dnmt3a and Dnmt3b using purified proteins. Dnmt3L stimulates the DNA methylation activity of both Dnmt3a and Dnmt3b, the stimulation being due to a direct interaction between Dnmt3L and Dnmt3a or Dnmt3b. Interestingly, the stimulation is not DNA sequence-specific. We determined that the region responsible for the stimulation is in the C-terminal half of Dnmt3L. Dnmt3L may contribute to the genome-wide DNA methylation in germ cells not by guiding Dnmt3a/Dnmt3b to specific target sequences but by stimulating the DNA methylation activity of Dnmt3a/Dnmt3b.

2. Protection from DNA Methylation Wave

     Differentially regulated genes in genome are separated by so called “chromatin insulator”, which is a DNA sequence that serves as a boundary element. Various insulators have been found in Drosophila and vertebrates. These insulators have two conserved properties. One is an enhancer-blocking activity: an insulator blocks enhancer and promoter interaction when placed between them. The other is protection from positional effects. A promoter/reporter cassette surrounded by insulators is expected to be isolated from the local chromosomal environment, and therefore to be protected from positional effects. The most extensively investigated vertebrate insulator, cHS4, is derived from the chicken β-globin gene locus control region. Previous reports have shown that cHS4 insulator protects a transgene from DNA methylation as well as position-effect variegation when introduced into genome.

The 573-bp fragment of sea urchin arylsulfatase (Ars) gene locus is reported to have typical features of an insulator. Since Ars insulator works in various cell types and across species, it has been suggested that this non vertebrate-derived insulator may serve as a universal insulator. We developed a Moloney murine retrovirus-based vector and found that that the Ars insulator protects the transgene from DNA methylation wave as well as from silencing.

3. Maintenance of DNA Methylation Patterns

Dnmt1 favors methylating the hemimethylated state of CpG sites, which appears just after the replication and repair steps, and this property of Dnmt1 ensures the maintenance of DNA methylation patterns. It was reported that Dnmt1 exists around replication foci, and binds to proliferating cell nuclear antigen (PCNA), a prerequisite factor for replication and repair, with the sequence motif at 160-172, which is very close to the N-terminus. It is also reported that the N-terminal 1-343 sequence binds to DNA. We have further restricted this DNA binding sequence to the N-terminal 119-197, overlapping with the PCNA-binding motif. We also found that the N-terminal sequence 1-290 forms an independent domain. Under physiological conditions, Dnmt1 methylates the hemimethylated CpG sites that are generated at the replication fork. Thus it is convenient for Dnmt1 to stay on the same DNA and methylate CpG sites in a processive manner. Further, it is reasonable to speculate that the binding of Dnmt1 to DNA and PCNA, which circles and slides on the DNA, through its N-terminal domain helps the DNA methylation activity to be processive. We addressed whether or not the DNA methylation activity of Dnmt1 is processive, and if so, whether interactions with the N-terminal DNA and PCNA binding regions are necessary for this property. We expressed and purified recombinant Dnmt1 (Dnmt1-FL) and its truncated form (Dnmt1-dN) that lacks N-terminal 290 amino acid residues. Both recombinant Dnmt1-FL and Dnmt1-dN show similar specific activities, specificity towards different types of DNA substrates including hemimethylated DNA, and processivity. The results indicate that Dnmt1 methylates hemimethylated DNA in a processive manner, and that this processivity is independent of interactions with DNA and PCNA in the N-terminal domain. In addition, Dnmt1 is shown to preferentially methylate one strand of the double-stranded DNA during its processive methylation.

4. Characterization of Methylated DNA Binding Proteins

A family of proteins containing about a 60-amino acid sequence called the methyl-CpG binding domain (MBD) specifically recognizes the methyl-CpG in DNA. At present, five genes of this family, MeCP2, MBD1, MBD2, MBD3, and MBD4, have been identified. MBD-containing proteins have been shown to suppress the transcription of methylated genes. MeCP2, MBD2, and MBD3 silence the genes by recruiting a histone deacetylase (HDAC)-containing complex, and MBD1 through other pathways. MBD4 possesses DNA glycosylase activity to excise thymine from the T:G mismatch base pair, which is thought to play an important role in the repair of the mismatch caused by deamination of the methylated cytosine. MeCP2 is an intrinsic component of the SIN3 complex, and MBD3 is of the Mi-2/NuRD complex, both of which are corepressors that contain HDAC. MBD2 is a component of a huge complex named MeCP1, and is reported to interact also with the Mi-2/NuRD complex. MBD2 is expected to recruit the Mi-2/NuRD complex to methylated DNA. Two corepressor complexes, SIN3 and Mi-2/NuRD, are involved in chromatin remodeling and transcriptional silencing. The amino acid sequences of MBD2 and MBD3 are similar to each other, however, different from MBD2, mouse MBD3 does not bind to methyl-CpG. On the contrary, Xenopus MBD3 (xMBD3) strongly binds to methylated DNA. Interestingly, at the same time, Xenopus expresses a long isoform of xMBD3, named xMBD3LF, a product of alternative splicing. xMBD3LF have a 20-amino acid sequence inserted in the middle of MBD, and cannot bind to methyl-CpG. The MeCP2 gene located on the X chromosome is the causative gene for Rett syndrome, and mouse embryos in which this gene is targeted show a neurological defect. Although MBD2 and MBD3 are similar to each other, MBD2-targeted mice are viable and fertile, while MBD3-targeted mice are embryonic lethal. These results indicate that the functions of these proteins are distinct.

We determined the expression of Xenopus MeCP2 (xMeCP2), MBD2 (xMBD2), and MBD3 (xMBD3) during development. Interestingly, not the xMBD2 but the xMBD3 protein is expressed constitutively at early stages of embryogenesis. In Xenopus eggs and embryos, equal amounts of xMBD3 and xMBD3LF proteins are expressed. xMBD3 is highly expressed in the prospective eye regions, brain, and branchial arches. Suppression of xMBD3 expression by antisense morpholino oligonucleotide injection severely affects eye formation and brain development. The overexpression of mutant xMBD3 and xMBD3LF, which cannot bind to methyl-CpG, cause an eye-defective phenotype in a dominant negative manner. We propose that xMBD3 plays an essential role in the suppression of the specific genes that are methylated by guiding the corepressor complex, Mi-2/NuRD, as an intrinsic component during an early stage of Xenopus embryogenesis.