DNA methylation is essential for development in the mouse and plays an important role in inactivation of the X-chromosome and genomic imprinting (10Jaenisch R Trends Genet. 1997; 13: 323-329Abstract Full Text PDF PubMed Scopus (319) Google Scholar). It may also contribute to immobilization of mammalian transposons, suppression of transcriptional noise, and the control of tissue-specific gene expression, but decisive evidence on these points is lacking. The theme that is common to all these phenomena is transcriptional repression. Work on animals, plants, and fungi now leaves little doubt that gene silencing is a major biological consequence of DNA methylation (5Colot V Rossignol J.L Bioessays. 1999; 21: 402-411Crossref PubMed Scopus (271) Google Scholar). Methylation-mediated silencing also plays a part in both the etiology of human disease and attempts at therapeutic intervention. The methylation of tumor suppressor gene promoters contributes directly to progression of some cancers (12Jones P.A Laird P.W Nat. Genet. 1999; 21: 163-167Crossref PubMed Scopus (2055) Google Scholar), while the methylation of exogenous DNA introduced into cells compromises efforts at gene therapy (8Garrick D Fiering S Martin D Whitelaw E Nat. Genet. 1998; 18: 56-59Crossref PubMed Scopus (532) Google Scholar). To understand the DNA methylation system, we need to know how the methylation of CpG dinucleotides is targeted and maintained, how the methyl-CpG signal is read, and how this leads to repression of transcription. These questions are finally yielding to experimental scrutiny because several new protein players have been identified. Two years ago there was one mammalian DNA cytosine methyltransferase (DNMT1) and one generalized methyl-CpG-binding protein (MeCP2). Now there are four mammalian DNMTs, five methyl-CpG-binding proteins and a candidate demethylase (see Figure 1). Two human genetic diseases have recently been attributed to mutations in a methyl-CpG-binding protein (MeCP2; 1Amir R.E Van den Veyver I.B Wan M Tran C.Q Francke U Zoghbi H.Y Nat. Genet. 1999; 23: 185-188Crossref PubMed Scopus (3800) Google Scholar) and the DNA methyltransferase DNMT3b (16Okano M Bell D.W Haber D.A Li E Cell. 1999; 99: 247-257Abstract Full Text Full Text PDF PubMed Scopus (4486) Google Scholar). This review will assess recent progress in understanding how some of these proteins bring about transcriptional repression. The most direct mechanism by which DNA methylation can interfere with transcription is to prevent the binding of basal transcriptional machinery or ubiquitous transcription factors that require contact with cytosine in the major groove of the double helix. Most mammalian transcription factors have GC-rich binding sites and many have CpGs in their DNA recognition elements. Binding by several of these factors is indeed impeded or abolished by methylation of CpG, though some factors, notably Sp1, are indifferent to methylation status. Although direct interference is likely to play a part in suppression of methylated genes, its role may be minor overall, as several densely methylated genes can be transcribed when chromatin and/or methyl-CpG binding proteins are absent. An alternative route by which CpG methylation can inactivate genes is by direct exclusion of the transcriptional machinery from methylated promoter DNA. For example the presence of methyl-CpGs could influence nucleosome stability or positioning to deny the access of transcription factors to a promoter. Methylation of multiple CpGs can exclude core histones from particular sequences while attracting them to others (for example, 6Davey C Pennings S Allan J J. Mol. Biol. 1997; 267: 276-288Crossref PubMed Scopus (82) Google Scholar). Earlier studies have shown, however, that CpG methylation does not have a general influence on the stability of nucleosomes. Each nucleosome includes a single molecule of a linker histone such as H1. The possibility that histone H1 preferentially associates with methylated DNA to facilitate repression is the subject of disagreement in the literature. Differences in experimental results may depend on choice of naked or nucleosomal templates, particular promoters, and specific variants of H1. Xenopus oocytes microinjected with methylated templates selectively repress their transcription dependent on nucleosomal assembly in the absence of the types of H1 normally associated with transcriptional repression (13Kass S.U Landsberger N Wolffe A.P Curr. Biol. 1997; 7: 157-165Abstract Full Text Full Text PDF PubMed Scopus (305) Google Scholar). In fungal systems, a distinctive repression mechanism has been found, as methylation leads to a change in chromatin structure that is independent of transcriptional state. Transcription initiation is not influenced directly by CpG methylation, but instead the elongating RNA polymerase cannot progress though a methylated gene (5Colot V Rossignol J.L Bioessays. 1999; 21: 402-411Crossref PubMed Scopus (271) Google Scholar, 18Selker E.U Cell. 1999; 97: 157-160Abstract Full Text Full Text PDF PubMed Scopus (136) Google Scholar). This mechanism may not apply in mammals, as most gene exons are methylated at CpG, and even the presence of a densely methylated CpG island within the body of the mouse Igf2r gene does not prevent Igf2r expression. The highest density of nonmethylated CpGs in the vertebrate genome are found in CpG islands, which usually contain promoter or other regulatory DNA that is required for active transcription of a gene. CpG island chromatin is enriched in hyperacetylated histones and deficient in linker histones. These are essential features of transcriptionally competent chromatin templates. In contrast, chromatin assembled on artificially methylated DNA becomes associated with hypoacetylated histones, refractory to nuclease or restriction endonuclease digestion and transcriptionally silent. Microinjection of methylated or unmethylated templates into mammalian cells or Xenopus oocytes has shown that DNA methylation and nucleosomal chromatin can interact to bring about transcriptional silencing (13Kass S.U Landsberger N Wolffe A.P Curr. Biol. 1997; 7: 157-165Abstract Full Text Full Text PDF PubMed Scopus (305) Google Scholar). Methylated templates are initially transcribed, but become progressively repressed as nucleosomes are assembled. Eventually the transcriptional machinery, including RNA polymerase, is erased from the promoter. Erasure does not occur on unmethylated templates, even when they are also assembled into nucleosomes. How might chromatin and methylation cooperate to bring about repression? The properties of the methyl-CpG-binding proteins are proving to be the key to interpreting the connection between DNA methylation and transcriptional silencing. The founder member of the family is MeCP2 (Figure 1; 9Hendrich B Bird A Mol. Cell. Biol. 1998; 18: 6538-6547Crossref PubMed Scopus (1068) Google Scholar and references therein), which consists of a single polypeptide that contains both a methyl-CpG-binding domain (MBD) and transcriptional repression domain (TRD). MeCP2 is capable of binding to a single symmetrically methylated CpG both in naked DNA and within chromatin. The TRD confers repression when tethered to a Gal4 DNA–binding domain and can operate over a distance of several hundred base pairs to silence transcription. Immunoprecipitation experiments have established that the TRD interacts with Sin3A in mammalian and Xenopus cells. Sin3A interacts with histone deacetylase, which is known from work on other repression systems to aid establishment of a repressive chromatin environment (see 14Knoepfler P.S Eisenman R.N Cell. 1999; 99 (this issue): 447-450Abstract Full Text Full Text PDF PubMed Scopus (310) Google Scholar [this issue of Cell]). Consistent with this model, the inhibition of histone deacetylase using trichostatin A partially relieves transcriptional repression by the tethered TRD of MeCP2 (reviewed in 17Razin A EMBO J. 1998; 17: 4905-4908Crossref PubMed Scopus (661) Google Scholar). MeCP2 is clearly not the sole connection between DNA methylation and transcriptional repression. Cells in which MeCP2 is deficient (e.g., HeLa cells) nevertheless repress methylated reporter constructs (15Ng H.-H Zhang Y Hendrich B Johnson C.A Burner B.M Erdjument-Bromage H Tempst P Reinberg D Bird A Nat. Genet. 1999; 23: 58-61PubMed Scopus (0) Google Scholar). Recent evidence implicates other methyl-CpG binding activities in mediating the effects of methylation on transcription. One of these is MeCP1, which exists in many cell types and was implicated several years ago in repression of methylated genes. A search of the EST database with the methyl-CpG-binding domain of MeCP2 has identified four new proteins with closely related domains: MBDs 1–4 (Figure 1; 9Hendrich B Bird A Mol. Cell. Biol. 1998; 18: 6538-6547Crossref PubMed Scopus (1068) Google Scholar). One of these, MBD2, turns out to be a component of the MeCP1 complex, together with histone deacetylases HDAC1 and HDAC2, and RbAp46 and RbAp48 (15Ng H.-H Zhang Y Hendrich B Johnson C.A Burner B.M Erdjument-Bromage H Tempst P Reinberg D Bird A Nat. Genet. 1999; 23: 58-61PubMed Scopus (0) Google Scholar). As in the case of MeCP2, a Gal4-MBD2 fusion protein represses promoters with nearby Gal4-binding sites and in some cases TSA can reduce repression. It was initially thought that MBD1 (see below) was a component of the MeCP1 complex, but this possibility has now been ruled out by several experimental criteria (15Ng H.-H Zhang Y Hendrich B Johnson C.A Burner B.M Erdjument-Bromage H Tempst P Reinberg D Bird A Nat. Genet. 1999; 23: 58-61PubMed Scopus (0) Google Scholar). MBD2 and MBD3 are the only two members of the current MBD protein family that show sequence similarity outside the methyl-CpG-binding domain. They are in fact closely related proteins, although MBD3 has a characteristic C-terminal tail. Perhaps surprisingly, MBD2 and MBD3 appear to reside in distinct complexes in the cell. MBD3 is a component of the Mi2/NuRD deacetylase complex (20Wade P.A Gegonne A Jones P.L Ballestar E Aubry F Wolffe A.P Nat. Genet. 1999; 23: 62-66PubMed Google Scholar, 21Zhang Y Ng H.-H Erdjument-Bromage H Tempst P Bird A Reinberg D Genes Dev. 1999; 13: 1924-1935Crossref PubMed Scopus (931) Google Scholar). This complex is the most abundant macromolecular form of histone deacetylase found in Xenopus eggs and embryos and in mammalian cells. The Mi-2 protein is a member of the SWI2/SNF2 superfamily of ATPases that disrupt histone–DNA interactions (see 19Tyler J.K Kadonaga J.T Cell. 1999; 99 (this issue): 443-446Abstract Full Text Full Text PDF PubMed Scopus (204) Google Scholar [this issue of Cell]). In this context, Mi-2 has to disrupt the nucleosome to allow access of RbAp48 to the histone-fold domain of histone H4, which normally lies sequestered inside the coils of nucleosomal DNA. RbAp48 interacts with histone deacetylase directly and enhances enzymatic activity presumably by tethering the deacetylase next to the target site for deacetylation at the N-terminal tail of histone H4 (Figure 2). Are the proposed chromatin-modifying attributes of Mi2/NuRD at the service of DNA methylation? In Xenopus the answer appears to be yes, as xMBD3 binds to methylated DNA preferentially in Southwestern and bandshift assays and the Mta-like component also shows a measurable preference for binding to methylated DNA in vitro (20Wade P.A Gegonne A Jones P.L Ballestar E Aubry F Wolffe A.P Nat. Genet. 1999; 23: 62-66PubMed Google Scholar). In mammals, opinion is divided on the methyl-CpG binding credentials of mammalian MBD3 (mMBD3). According to one view, neither mMBD3 protein itself nor native Mi2/NuRD complex (9Hendrich B Bird A Mol. Cell. Biol. 1998; 18: 6538-6547Crossref PubMed Scopus (1068) Google Scholar, 21Zhang Y Ng H.-H Erdjument-Bromage H Tempst P Bird A Reinberg D Genes Dev. 1999; 13: 1924-1935Crossref PubMed Scopus (931) Google Scholar) have a significant affinity for methylated DNA sequences in vitro, and an mMBD3-GFP fusion does not localize to densely methylated regions in vivo (9Hendrich B Bird A Mol. Cell. Biol. 1998; 18: 6538-6547Crossref PubMed Scopus (1068) Google Scholar). On the other hand, mMBD3 has been shown to have a modest preference for particular methylated DNA sequences by bandshift assays (20Wade P.A Gegonne A Jones P.L Ballestar E Aubry F Wolffe A.P Nat. Genet. 1999; 23: 62-66PubMed Google Scholar). To complicate matters further, Mi-2/NuRD can interact with MBD2 in vitro (21Zhang Y Ng H.-H Erdjument-Bromage H Tempst P Bird A Reinberg D Genes Dev. 1999; 13: 1924-1935Crossref PubMed Scopus (931) Google Scholar). For mammals, therefore, there are three suggested roles of mMBD3 and the Mi2-NuRD complex vis-a-vis repression by DNA methylation: (1) that Mi2-NuRD plays no part; (2) that Mi2-NuRD can become involved if recruited by the methyl-CpG-binding protein MBD2; and (3) that Mi2-NuRD is a methyl-CpG binding repressor complex in its own right. In Xenopus, xMBD3 shows a strong preference for methylated DNA in vitro, encouraging the view that role (3) applies in this vertebrate. MBD1 binds selectively to methylated DNA and represses transcription from a naked methylated promoter in vitro. Recently, MBD1 has been shown to repress transcription in vivo in a methyl-CpG-directed manner using Drosophila and mammalian cells (7Fujita N Takebayashi S Okumura K Kudo S Chiba T Saya H Nakao M Mol. Cell. Biol. 1999; 19: 6415-6426Crossref PubMed Scopus (171) Google Scholar). MBD1 resembles MeCP2 in being a chromosome-bound protein. Methylation may not be the sole determinant for its localization in the nucleus, however, as overexpressed protein can bind to heterochromatic sites even in methylation-deficient mouse cells that fail to localize MeCP2 or MBD2 (9Hendrich B Bird A Mol. Cell. Biol. 1998; 18: 6538-6547Crossref PubMed Scopus (1068) Google Scholar). There are almost certainly far more methyl-CpGs in the genome (∼2 × 107) than molecules of either MeCP2 or MBD1. Whether MBD1 and MeCP2 randomly associate with sites, or segregate due to other constraints is not yet known. Of the five proteins that have the methyl-CpG-binding domain (Figure 1), four (MeCP2, MBD1, MBD2, and MBD3) are implicated in transcriptional repression. MBD2b—a translation product resulting from initiation at an internal AUG (see Figure 1)—is also reported to be a DNA demethylase (2Bhattacharya S.K Ramchandani S Cervoni N Szyf M Nature. 1999; 397: 579-583Crossref PubMed Scopus (545) Google Scholar), and MBD4 is a thymine DNA glycosylase. Three out of the four repressors are in complexes that contain histone deacetylases (whether MBD1 exists in a complex is not yet clear), indicating that chromatin modification is an important feature of the silencing mechanism. Accordingly, methylated transfected genes can be reactivated by treatment with TSA (reviewed in 17Razin A EMBO J. 1998; 17: 4905-4908Crossref PubMed Scopus (661) Google Scholar). The effects of TSA on naturally methylated CpG island promoters, however, suggest that histone deacetylation is only a part of the story. Methylated CpG islands from the promoters of tumor suppressor genes (3Cameron E.E Bachman K.E Myohanen S Herman J.G Baylin S.B Nat. Genet. 1999; 21: 103-107Crossref PubMed Scopus (1678) Google Scholar) and the methylated fragile X mental retardation gene, FMR1 (4Coffee B Zhang F Warren S.T Reines D Nat. Genet. 1999; 22: 98-101Crossref PubMed Scopus (267) Google Scholar), cannot be reactivated by simple application of TSA. These specific findings parallel the general observation that treatment of animal cells with TSA leads to surprisingly minor changes in the expression of individual genes. Thus, although there is compelling evidence for histone acetyltransferases and deacetylases having a pervasive role in eukaryotic gene control, for unknown reasons this is often not reflected in the response to TSA. Remarkably, TSA addition, when coupled to minimal demethylation by treatment with 5-aza-2′ deoxycytidine, does lead to robust reexpression of a variety of methylated tumor suppressor promoters, although neither drug treatment alone reactivates (3Cameron E.E Bachman K.E Myohanen S Herman J.G Baylin S.B Nat. Genet. 1999; 21: 103-107Crossref PubMed Scopus (1678) Google Scholar). These data suggest that while recruiting deacetylase is important, other mechanisms augment DNA methylation–dependent gene silencing. The other mechanisms could include direct interference of methylation with transcription factor binding or chromatin structure (see above). Alternatively, repression by the methyl-CpG-binding proteins could occur via mechanisms that do not involve the histones. It would not be surprising if these mediators of repression utilize several independent mechanisms for gene silencing, of which histone deacetylation is but one. The attractive feature of Mi2-NuRD is the presence of a chromatin remodeling engine that may facilitate the energy-dependent deacetylation of nucleosomal histones. It is presently less obvious how the MeCP1 and MeCP2 complexes could modify histones within chromatin. The experiments examining the role of DNA methylation, MBD2, or the TRD of MeCP2 in chromatin modification and transcription use transfected or microinjected DNA. Under these experimental conditions, chromatin is only slowly assembled through a succession of intermediate structural states that may well retain accessibility to the deacetylase enzyme. Under more natural chromosomal circumstances, chromatin is assembled at the replication fork during S phase. The DNMT1 methyltransferase is sequestered at the replication fork and rapidly reestablishes symmetric CpG methylation on hemimethylated substrates. If either the MeCP1 or MeCP2 complexes could associate with the replication fork they might be able to use the replication process as a means to scan the entire genome for methylated DNA, rapidly deacetylate nascent chromatin, and release repressed chromatin into the nucleoplasm. Alternatively, these complexes may come associated with their own independent chromatin remodeling machinery. The biochemical characterization of three chromatin remodeling and histone deacetylase complexes that contain methyl-CpG-binding proteins provides potential mechanisms for DNA methylation to contribute to both the global repression of transcriptional noise and the targeted repression of genes. DNA methylation may encourage promoters that are destined for repression to become even more stably silenced than they would be by the association of chromatin with unmethylated DNA (Figure 3). In addition, DNA methylation may more efficiently exclude regulatory transcriptional activators and the basal transcriptional machinery from the vast bulk of nonproductive sites in chromatin. As a consequence, the attention of these factors may be focused on the relatively small fraction of the genome associated with potentially active promoters within the nonmethylated CpG islands. Methyl-CpG-binding repressors presumably go where methylation takes them, but what determines where in the genome methyl-CpGs lie? Our profound ignorance of the mechanisms involved may be short-lived. Expectations have been raised by the discovery of two new mammalian cytosine DNA methyltransferases, DNMT3a and 3b, that are important in establishing embryonic methylation patterns (16Okano M Bell D.W Haber D.A Li E Cell. 1999; 99: 247-257Abstract Full Text Full Text PDF PubMed Scopus (4486) Google Scholar). The gene for each protein is essential for normal mouse development and analysis of DNAs from double mutant (Dnmt3a−/−, Dnmt3b−/−) embryos and embryonal stem cells show that de novo methylation of certain sequences is defective. DNMT1 is often considered as a maintenance methyltransferase that can complete hemimethylated sites following DNA replication, but can not transfer methyl groups to nonmethylated DNA. In vitro, however, DNMT1 can methylate nonmethylated DNA. Whether there are two or three mammalian de novo methyltransferases remains to be determined. The big question, however, is this: how are de novo methyltransferases recruited to, or excluded from, particular regions of the genome in normal cells? What dictates whether CpG islands (e.g., those on the X chromosome, at imprinted genes and other loci), repetitive DNA, viral DNA sequences, or mobile elements should either evade methylation, or succumb to it? Although the signals that cause susceptibility or resistance to methylation are still unknown, genetic approaches in plants are leading the way forward. Short inverted repeats of DNA sequence and double-stranded RNA can promote methylation of homologous sequences (18Selker E.U Cell. 1999; 97: 157-160Abstract Full Text Full Text PDF PubMed Scopus (136) Google Scholar). Intriguingly, a SWI/SNF family member has genetic connections to DNA methylation patterns in Arabidopsis (11Jeddeloh J.A Stokes T.L Richards E.J Nat. Genet. 1999; 22: 94-97Crossref PubMed Scopus (562) Google Scholar); could this mean that chromatin disruption is a prerequisite for de novo methylation in this plant? An alternative possibility is that DNA methylation is the default state of parts of the genome and that demethylation, either active or passive, is the key determinant of methylation patterns. We can look forward to exciting revelations as the knowledge vacuum in this area is filled over the next few years.
更多
原文链接
分享
引用
