DNA methylation is an epigenetic mechanism involved in transcriptional silencing of imprinted genes, genes located on the inactive X chromosome, and a number of tumour suppressor genes in cancer. MBD (methyl-CpG-binding domain) proteins selectively bind to methylated DNA and recruit chromatin remodelling and transcriptional repressor complexes, thereby establishing a repressive chromatin state. MBD2, a member of the MBD protein family, binds to methylated promoter CpG islands (clusters of high-density CpG dinucleotides) and acts as a methylation-dependent transcriptional repressor. Previous work has demonstrated that decreased CpG island methylation in mice lacking the DNA methyltransferase DNMT1 is associated with impaired tumorigenesis when crossed on the tumour-susceptible ApcMin/+ background. Mbd2 deficiency also dramatically reduces adenoma burden and extends life span in a gene dosage-dependent manner in this mouse model. Mbd2 is therefore essential for tumorigenesis in the murine intestine, although it is dispensable for the viability of the host animals. These findings validate MBD2 as a potential target for therapeutic intervention in colorectal cancer.
In addition to the information encoded by the DNA sequence, a distinct mechanism has evolved for storing and passing on information. This so-called epigenetic information can be stably inherited through many cell generations. One type of epigenetic information comprises the addition of a methyl group to the 5′-position of the cytosine ring. In plants, cytosines can be methylated in almost any sequence context, whereas in vertebrates methylation is restricted mainly to the palindromic sequence CpG . The methyl-CpG-dinucleotide represents a mutational hotspot due to its capacity to undergo spontaneous deamination, giving rise to thymine. Evolutionary pressure has therefore led to an under-representation of CpGs in the vertebrate genome. Approximately 70% of the remaining CpG dinucleotides are methylated in vertebrates . However, CpG islands (clusters of high-density CpG dinucleotides) are mainly unmethylated and are found in approx. 60% of human promoters . Methylation of promoter CpG islands and subsequent inactivation of the associated genes have been found to occur in genes located on the inactive X chromosome, genes silenced by imprinting and in some tumour suppressor genes that are aberrantly silenced in malignant cells .
MBD (methyl-CpG-binding domain) proteins
The fact that methylated CpG islands are associated with transcriptional repression led to the search for nuclear factors that are able to read and interpret this epigenetic information. The first entity shown to bind to methylated DNA and mediate transcriptional repression was named methyl-CpG-binding protein, MeCP1 . Later, this factor was shown to be a protein complex consisting of a number of proteins, one of which binds directly to methylated DNA. The second methyl-CpG-binding factor identified was named MeCP2, and represents the founding member of the family of MBD proteins . This family consists of five members: MeCP2, MBD1, MBD2, MBD3 and MBD4. They each contain a common MBD . Sequence similarity between MBD proteins is limited to the MBD itself, with the exception of MBD2 and MBD3, which display approx. 70% overall amino acid identity. This diversity predicts that MBD proteins exert different functions, but biochemical analysis revealed that they also differ in their ability to bind to methylated DNA. Three members of this family, MeCP2, MBD1 and MBD2, can all bind to methylated DNA and suppress transcription from a methylated target gene due to the presence of a TRD (transcriptional repressor domain), which promotes interactions with other proteins. MeCP2 is able to bind to a single symmetrically methylated CpG . It interacts with a transcriptional repressor complex containing HDACs (histone deacetylases) and the transcriptional co-repressor Sin3a [8,9]. This important finding provided the first link between two global mechanisms of gene regulation, DNA methylation and histone modifications. MeCP2 has also been shown to bind to a histone methyltransferase activity targeting Lys9 on the histone H3 tail , a mark commonly associated with repressed chromatin .
MBD1 has been demonstrated to form a complex with the histone H3 K9 methylase, SETDB1, and the CAF-1 (chromatin assembly factor 1) . This complex is thought to be involved in the maintenance of histone methylation patterns during DNA replication. MBD1 contains, in addition to the MBD and the TRD, up to three zinc-co-ordinating CXXC domains, depending on the splice variant. The third CXXC domain has been shown to bind specifically to non-methylated CpGs. Binding of MBD1 to non-methylated reporter genes via the CXXC domain can induce transcriptional silencing, but MBD1 can also silence methylated reporter genes via its MBD . The biological significance of this dual DNA-binding capability of MBD1 is currently unknown.
MBD3 is the smallest member of the MBD family, coding for a protein of approx. 30 kDa. Due to changes in two highly conserved amino acids, mammalian MBD3 has lost its ability to bind to methylated DNA, whereas Xenopus MBD3 has an affinity for methylated CpGs similar to that of MeCP2 . MBD3 is a component of the Mi-2–NuRD complex, a transcriptional repressor complex containing HDACs, a chromatin remodelling ATPase and other proteins . Clearly, MBD3 is crucial to normal mammalian development as MBD3 knockout mice fail to develop to term . Unlike mammalian MBD3, MBD4 can bind to methylated DNA, but it is not involved in transcriptional repression. It contains a glycosylase domain, and has a role in the repair of methyl-CpG/TpG mismatches that can arise from spontaneous deamination .
MBD1, MBD2 and MBD3 are expressed ubiquitously in mouse tissues, but show lower expression levels in mouse embryonic stem cells. Splice variants exist for all MBDs. MBD2 occurs as a full-length protein (MBD2a) and as an N-terminal truncation (MBD2b) that arises through the usage of an alternative translational start codon. In testis, MBD2 is expressed as a shorter variant due to the usage of an alternate third exon . As mentioned above, MBD2 shares high sequence identity with MBD3, but it possesses a 140 amino acid-long N-terminus, which is absent from MBD3. This region of MBD2 contains a repeat consisting of glycine and arginine residues, the function of which is currently unknown. The truncated MBD2b protein lacking the N-terminus was reported to demethylate DNA , but this finding could not be confirmed by others [14,19,20]. Biochemical analysis revealed that MBD2 represents the methyl-CpG-binding activity of the MeCP1 complex, which represses transcription from methylated reporter genes in an HDAC-dependent manner . In a later study, the repressor complex originally termed MeCP1 was shown to consist of MBD2 and the chromatin remodelling complex Mi-2–NuRD, which contains HDACs, MBD3, the chromatin remodeller Mi-2 and other proteins . This demonstrates that MBD proteins in general provide a link between two important epigenetic mechanisms: CpG methylation and histone modification.
MBD2 and transcriptional regulation
To further establish the function of MBD2, knockout mice were generated . Whereas mice lacking Mbd3 died during early embryogenesis, Mbd2-null mice displayed a surprisingly weak phenotype. Overall DNA methylation levels and imprinting were not affected by the absence of MBD2 protein, though it was observed that Mbd2-deficient mice tend to be smaller. Cross-fostering experiments revealed that this low weight gain phenotype depended on inadequate feeding by Mbd2-null mothers, and was independent of the genotype of the pups. The exact function of MBD2 protein in maternal behaviour is still unknown. To examine whether MBD2 protein is required for appropriate repression of methylated promoters, fibroblast cell lines from wild-type and Mbd2-null mice were established and assayed for their ability to repress transcription of methylated reporter genes. Significantly reduced repression of these reporter genes was seen in two independent Mbd2-deficient cell lines, but expression reached only 25% of the level of unmethylated reporters. This finding indicates a possible redundancy between MBD proteins in the repression of methylated promoters.
An interesting study investigated the role of MBD2 protein in T helper cell differentiation . Naive, undifferentiated T helper cells express very low levels of the cytokines IFN-γ (interferon-γ) and IL-4 (interleukin-4). Differentiation to Th1 cells correlates with up-regulation of IFN-γ levels, whereas Th2 cells are characterized by high IL-4 expression levels. When naive T cells obtained from Mbd2-null mice were stimulated to undergo Th2 differentiation, higher protein levels of IFN-γ per cell and more IFN-γ-expressing cells were seen. Mbd2-null Th1 cells also showed higher IL-4 expression. Interestingly, Mbd2-null cells also reproducibly expressed low levels of ectopic IL-4 under Th1 conditions. This effect was strictly dependent on MBD2 levels, since Mbd2-heterozygous cells showed intermediate effects. ChIP (chromatin immunoprecipitation) experiments revealed that MBD2 binds to the regulatory region of the Il4 locus where it appears to act as a silencer under non-stimulating conditions. When Th2 differentiation is induced, MBD2 is displaced from the Il4 locus by the transcription factor GATA-3, and the Il4 gene can be transcribed. Therefore MBD2 acts as repressor for Il4 during T cell differentiation, but the effect of MBD2 absence is a subtle one, since only a slight derepression is observed. Furthermore, Mbd2−/− mice show normal development of lymphoid organs and major lymphocyte subsets . This study identified Il4 as a direct target gene of MBD2. It also strengthened the notion that there may be a redundancy between members of the MBD family, since Mbd2 deficiency resulted only in weak Il4 derepression.
So far, no sequence specificity for MBD2 binding has been observed. This raises the question of whether MBD2 randomly binds to methyl-CpGs and acts as a general repressor for methylated regions, or if it is targeted in any way to selected sequences. Using the yeast two-hybrid system, a novel protein that interacts with MBD2 was identified . This protein, named MIZF (MBD2-interacting zinc finger), can repress transcription in an HDAC-dependent manner and enhances MBD2-mediated repression in co-transfection experiments. In a later study, the same group showed that MIZF binds to a specific 5 bp recognition sequence and can repress transcription of genes containing this consensus sequence in their promoter, e.g. the retinoblastoma gene . According to these results, it is possible that MIZF may recruit MBD2, and potentially also the Mi-2–NuRD repressor complex, to specific target sequences and thereby allow for sequence-dependent repression of methylated regions.
MBD2 and tumorigenesis
Promoter methylation and subsequent silencing of tumour suppressor genes can be found in a variety of tumours, thereby indicating that a perturbation of epigenetic gene regulation may be involved in tumorigenesis. After the discovery of the family of MBD proteins, their function as a read-out mechanism of epigenetic marks instigated a number of studies aiming to examine their role in tumorigenesis. However, no clear correlation between MBD expression levels and tumour risk or aggressiveness could be established. One study demonstrated that the detection of high levels of MBD proteins, which were sometimes reported in malignant cells compared with normal cells, depended on the choice of reference gene. When transcript levels were normalized against β-actin, an apparent increase in all five MBDs and all three DNMTs (DNA methyltransferases) in human lung cancer cell lines compared with normal lung tissue was observed. However, when mRNA levels were normalized against the PCNA (proliferating cell nuclear antigen), a marker for cell proliferation, these differences disappeared. The authors concluded that up-regulation of DNMTs and MBDs probably only reflects the increased cell proliferation typical for tumour cells . A number of publications demonstrate conclusively that MBD2 is involved in the suppression of aberrantly methylated tumour suppressor genes by binding to methylated regulatory regions. For example, MBD2 has been shown to bind to the aberrantly methylated promoter regions of the tumour suppressor genes p14/ARF and p16/Ink4A in a colon cancer cell line. Treatment with 5-aza-2′-dC, a demethylating agent, results in re-expression of these genes as well as dissociation of MBD2 from the promoter regions . Another study investigated the role of MBD2 in silencing of the hypermethylated tumour suppressor gene GSTP1 (glutathione S-transferase pi) in a human breast cancer cell line. ChIP revealed that both MBD2 and DNMT1 bind to its promoter region. Transfection of siRNAs (small interfering RNAs) complementary to MBD2 mRNA reduced MBD2 transcript levels and stimulated GSTP1 expression .
These results prompted study of the role of MBD2 in tumorigenesis in more detail. For this purpose, Mbd2-deficient mice were crossed on to an ApcMin/+ (or Min) background . ApcMin/+ mice carry a heterozygous mutation in the tumour suppressor gene Apc, which renders them very susceptible to the development of intestinal tumours . Mice that were homozygous, heterozygous or wild-type with respect to Mbd2, and carried the Min mutation were aged. Mbd2−/− mice survived significantly longer than control Min mice, whereas Mbd2+/− mice showed intermediate survival. This indicates that the dosage of Mbd2 is crucial for this tumour-resistant effect. Furthermore, the number of tumours per mouse as well as the size of the individual tumours was dramatically reduced in Mbd2−/− mice, indicating that Mbd2 is necessary not only for tumour development but also for tumour growth. These results, together with the fact that Mbd2−/− mice are viable and fertile, make MBD2 an attractive potential target for therapeutic intervention in cancer.
Mice deficient for the maintenance of DNMT1 display a similar tumour resistance phenotype on an ApcMin/+ background , presumably because lower DNMT1 levels result in a decrease in the aberrant methylation of tumour suppressor genes . This led to the hypothesis that MBD2, one of the read-out mechanisms of DNA methylation, works in the same pathway. If this were the case, one would predict that tumours arising in Mbd2-deficient mice would show reduced levels of tumour suppressor gene hypermethylation, because this mark could not be interpreted correctly in the absence of MBD2, and there would be a selection for tumours arising through different mechanisms. However, the few tumours that developed in ApcMin/+Mbd2−/− mice did not show reduced promoter methylation of tumour suppressor genes . Another study has demonstrated that depletion of MBD2 using antisense inhibitors suppresses growth of human lung and colorectal cell lines in vitro and human cancer xenografts in vivo . The exact mechanism of suppression of tumorigenesis by MBD2 deficiency still remains unclear.
Another finding argues against an overlap in MBD2 and DNMT1 function in tumorigenesis. Reduced levels of DNMT1 seem to inhibit the formation of epithelial tumours (carcinogenesis), but additionally they lead to genomic hypomethylation, which has also been associated with the development of some types of tumours . Two studies demonstrated that under certain circumstances, reduced DNMT1 levels result in increased or accelerated tumorigenesis. In a mouse model for sarcomas, a malignant tumour arising from connective tissue, Dnmt1-deficient mice developed sarcomas at an earlier age compared with littermates wild-type for Dnmt1 . This phenotype was accompanied by increased chromosomal instability. Another study analysed the tumour susceptibility of mice carrying a hypomorphic and a null allele for Dnmt1, which resulted in a dramatic reduction of overall DNA methylation levels . These mice developed aggressive T cell lymphomas at approx. 6 months of age. Again, this seemed to correlate with genomic instability, which has been linked to global hypomethyation. In order to investigate whether MBD2 protein has a similar role in the development of soft tumours, Mbd2-null mice were crossed with p53-deficient mice, a model for lymphomagenesis . In contrast with Dnmt1-deficiency, loss of Mbd2 does not accelerate lymphomagenesis. Thus it seems that reduced DNMT1 levels alleviate the silencing of tumour suppressor genes caused by aberrant CpG methylation, which can protect against the development of epithelial cancer. At the same time, reduced DNMT1 levels lead to global hypomethylation, which induces chromosomal instability. This in turn can facilitate the formation of lymphomas and other soft tissue tumours. Lack of MBD2, on the other hand, does not lead to hypomethylation and genomic instability, but it efficiently inhibits the formation of intestinal tumours in the Min mouse. This makes MBD2 a potential therapeutic target for colorectal carcinoma.
Stem Cells and Development: A Focus Topic at BioScience2005, held at SECC Glasgow, U.K., 17–21 July 2005. Edited by T. Kouzarides (Cambridge, U.K.), S. Newbury (Newcastle upon Tyne, U.K.), B. Richardson (University College London, U.K.), R. Sablowski (John Innes Centre, Norwich, U.K.), D. Tosh (Bath, U.K.), M. Welham (Bath, U.K.) and A. Willis (Nottingham, U.K.).