Methyl CpG binding proteins: coupling chromatin architecture to gene regulation

Transcription

1 (2001) 20, 3166 ± 3173 ã 2001 Nature Publishing Group All rights reserved 0950 ± 9232/01 $ : coupling chromatin architecture to gene regulation Paul A Wade*,1 1 Emory University School of Medicine, Department of Pathology and Laboratory Medicine, Woodru Memorial Research Building ± Room 7105B, 1639 Pierce Drive, Atlanta, GA 30322, USA A correlation between DNA methylation and transcriptional silencing has existed for many years. Recently, substantial progress has been reported in the search for proteins that interpret the regulatory information inherent in DNA methylation and translate this information into functional states, resulting in the identi cation of a family of highly conserved proteins, the MBD family. Direct connections between these proteins and histone modi cation enzymes have emerged as a common theme, implying that DNA methylation exerts its e ects primarily through repressive chromatin architecture. Recent structural determinations of the DNA binding domain of two MBD family members, MeCP2 and MBD1, provide a framework to model the interactions of this family with DNA. Comparative sequence analysis and experimental DNA binding data can be interpreted using this structural framework allowing one to contrast the members of the MBD family with each other and to predict the properties of new family members. The identi cation of mutations in MeCP2, the founding member of this family, as causal for the neurological developmental disorder Rett Syndrome provides additional information regarding amino acid residues crucial to the functions of this interesting protein family. (2001) 20, 3166 ± Keywords: DNA methylation; MeCP2; Rett Syndrome; histone deacetylase; chromatin Methylation is a common form of DNA modi cation in animals, occurring at the position ve of cytosine residues almost exclusively within the context of CpG dinucleotides. CpG methylation is non-random and the majority of potential sites in mammalian genomes are modi ed (Cooper and Krawczak, 1989). Methylation participates in the partitioning of genomes into active and inactive functional compartments (Cross and Bird, 1995; Razin, 1998). The constitutents of the inactive compartment associated with DNA methylation include imprinted genes, the inactive female X chromosome, and parasitic DNA elements (Bestor, 2000). The essential nature of DNA methylation in mammals has *Correspondence: been substantiated via mutation of the methyltransferase enzymes themselves which leads to developmental defects and death prior to or shortly after birth (Okano et al., 1999; Li et al., 1992). Several excellent recent reviews describe the DNA methyltransferases and their biology (Bestor, 2000; Robertson and Wol e, 2000; Hendrich and Bird, 2000), this review will focus instead on proteins that function downstream of the DNA methylation signal. While higher plants clearly possess sophisticated gene regulatory circuits featuring DNA methylation, this review will focus on methyl CpG binding proteins in animals. Formally, DNA methylation might lead to transcriptional repression through multiple mechanisms. Methylation is known to interfere with the ability of some transcription factors to bind their cognate recognition sequences. It might also result in structural e ects on nucleosomes themselves or e ects on nucleosome positioning, nucleosome stability, or assembly of higher order chromatin structures. However, work over the last decade has led to the accumulation of a body of data suggesting that the functional properties of methylated DNA result primarily from the action of a conserved family of proteins that selectively bind methylated CpG dinucleotides (Bird and Wol e, 1999). MeCP2 ± the prototype methyl CpG binding protein The prototype methyl CpG binding protein is MeCP2, a polypeptide capable of binding selectively to a single symmetrically methylated CpG (Lewis et al., 1992). This protein is associated with chromosomes throughout the cell cycle, colocalizes with methyl CpG rich DNA (Lewis et al., 1992), and consists of two functional domains (Figure 1). The methyl CpG binding domain or MBD is su cient to direct speci c binding to methylated DNA (Nan et al., 1993). Regions outside the MBD contribute to overall binding energy through non-speci c, presumably electrostatic interactions (Meehan et al., 1992). A second functional domain, the transcriptional repression domain or TRD, is required for transcriptional repression in vitro and in vivo (Nan et al., 1997; Jones et al., 1998; Kaludov and Wol e, 2000). As expected for a chromosomal protein, MeCP2 is released from nuclei by treatment with nucleases or by extraction with salt,

2 3167 Figure 1 Mammalian MBD family members. The gure depicts the mammalian MBD family in cartoon fashion. Notable sequence motifs are indicated in the gure and discussed in the text although it presents a biphasic extraction pro le from rat brain nuclei (Meehan et al., 1992). In addition, puri ed recombinant Xenopus MeCP2 binds to methyl CpG dinucleotides in a nucleosomal context. The isolated MBD domain retains the capacity to bind methylated nucleosomes albeit with reduced a nity compared to the intact protein (Chandler et al., 1999). Several lines of evidence contributed to the notion that MeCP2 might function to repress transcription within a chromatin infrastructure. Methylated DNA injected into either mammalian cells or into Xenopus oocyte nuclei is transcriptionally repressed relative to unmethylated controls. This repression, however, requires su cient time to permit chromatin assembly (Buschausen et al., 1987; Kass et al., 1997). Further, reactivation of genes silenced by aberrant promoter methylation in cancer cell lines requires inhibition of both DNA methyltransferases and histone deacetylases (Cameron et al., 1999). The subsequent nding that MeCP2 is physically associated with the transcriptional corepressor Sin3 and histone deacetylases in both mammalian cells and Xenopus oocytes (Nan et al., 1998; Jones et al., 1998) identi ed candidate regulatory enzymes involved in the assembly of a specialized chromatin state at methylated loci. The region of interaction of Sin3 with mammalian MeCP2 was found to be largely coincident with the previously de ned TRD (Nan et al., 1998). Importantly, in both mammalian cells and Xenopus oocytes, arti cial recruitment of the MeCP2 TRD to a promoter leads to transcriptional repression that is partially relieved by inhibitors of histone deacetylase (Nan et al., 1998; Jones et al., 1998). The solution structure of the MBD domain from rat MeCP2 has recently been solved (Wake eld et al., 1999). The MBD is a wedge shaped structure (Figure 2) with one face of the wedge composed of a beta sheet and the other face consisting of an alpha helix and hairpin loop (Wake eld et al., 1999). The vertex of the wedge is extended by a long loop between two of the beta strands that contains several basic residues (Wake eld et al., 1999). Addition of methyl CpG containing DNA results in changes in the resonance of several residues in this loop, in b strands 2 ± 4 and in the alpha helix. These residues de ne a surface of the wedge that likely interacts with DNA (Wake eld et al., 1999). This surface presents a set of basic residues (in strand b2 and the exible loop) immediately anking a hydrophobic patch in strand b3. Additional basic residues line the opposite side of the hydrophobic patch (Figure 2). The importance of structural exibility in the loop has been con rmed by mutation of a conserved glycine residue to proline, which results in a minimally 25-fold reduction in binding a nity (Free et al., 2000). Three solvent exposed hydrophobic residues (Tyrosine 123, Isoleucine 125, and Alanine 131) are postulated to interact with the methyl groups in the major groove, mutation of these residues individually results in reductions in binding a nity (Free et al., 2000). The MBD family of proteins A survey of EST databases with the MBD domain of MeCP2 led to the identi cation of an additional family member initially termed PCM1 (Cross et al., 1997). Subsequent searches of the expanding EST databases identi ed four proteins in mammals that all possessed the MBD sequence motif, MBD1 (identical to PCM1), MBD2, MBD3, and MBD4 (Hendrich and Bird, 1998). Outside of the MBD domain itself, these proteins bear little obvious resemblance to each other (Figure 1; Hendrich and Bird, 1998). The single exception to this rule is the high degree of similarity between MBD2 and MBD3 which are approximately 70% identical in the region de ned by MBD3 (Hendrich and Bird, 1998). The MBD proteins are ubiquitously expressed in somatic tissues, while ES cells fail to express MBD1 and express very low levels of MBD2 (Hendrich and Bird, 1998). Additionally, all four of these MBD proteins are alternatively spliced with some splice variants being tissue speci c and others clearly

3 3168 Figure 2 Three dimensional structure of the MBD domain from MeCP2. The structural coordinates for the rat MeCP2 MBD domain (Wake eld et al., 1999) and the human MBD1 MBD domain were utilized in the Ribbons program to generate the cartoon. Open squares superimposed on the structure are basic residues making up the putative DNA interaction surface. The `hydrophobic patch' residues are indicated by open circles. The conserved tyrosine residue in strand b3 is indicated a ecting the MBD domain (Hendrich and Bird, 1998). The genomic structures of the human and murine MBD1-MBD4 genes have been determined, all have an intron within the MBD domain itself and the human and mouse genes are highly similar in their exon/intron organization (Hendrich et al., 1999a). MBD1 MBD1 is the largest member of the family and contains a sequence motif, the CXXC motif, shared with DNA methyltransferase I (Cross et al., 1997). The mrna for human MBD1 is found in at least ve alternatively spliced forms, a ecting the carboxyl terminus as well as the CXXC motifs (Fujita et al., 1999; Ng et al., 2000). Murine MBD1 apparently does not express the full complement of splice variants found in the human mrna (Hendrich et al., 1999a). Speci c binding to methylated DNA substrates requires only the MBD domain and the CXXC motifs are unimportant for DNA binding in vitro (Cross et al., 1997). In mammalian cells, MBD1 is localized throughout euchromatin with additional concentrations at a subset of pericentromeric regions on mitotic chromosomes (Fujita et al., 1999; Ng et al., 2000). Initially, MBD1 was de ned as a component of the MeCP1 complex (Cross et al., 1997). A more recent reexamination of MBD1 using new reagents revealed that it is not a component of MeCP1 and that its native molecular mass from HeLa nuclear extracts is 200 ± 400 kda, inconsistent with the properties of MeCP1 (Ng et al., 2000). While the identities of the proteins associated with MBD1 in this complex remain elusive, MBD1 is not depleted by antisera speci c for MTA2, HDAC1 and SAP30, implying that it is also not a component of the previously de ned Mi-2/NuRD and Sin3 complexes (Ng et al., 2000). Recombinant MBD1 can repress transcription of methylated, but not unmethylated, templates in vitro much like MeCP2 (Cross et al., 1997; Fujita et al., 1999). The various isoforms of MBD1 repress transcription in both mammalian and insect cells (Fujita et al., 1999; Ng et al., 2000). The carboxyl terminus of the protein contains a transcriptional repression domain with functions analogous to that of MeCP2 despite a lack of obvious sequence similarity (Ng et al., 2000). Like MeCP2, transcriptional repression mediated by MBD1 is sensitive to inhibitors of histone deacetylase although the precise HDAC associated with MBD1 remains to be elucidated (Ng et al., 2000). The multiple splice variants of MBD1 in human cells also appear to di er substantially in their a ects on transcription in transient assays in both mammalian and insect cells (Fujita et al., 1999). The solution structure of the MBD domain of MBD1 has also been solved (Ohki et al., 1999). The fold of this domain is essentially identical to that of MeCP2, indicating that all the MBD proteins are likely to adopt a similar conformation (Figure 2). The placement and orientation of charged and hydrophobic residues in strands b2 and b3 and the exible loop are highly similar to the MeCP2 structure (Ohki et al., 1999; Wake eld et al., 1999). In MeCP2, three solvent exposed hydrophobic residues were predicted to directly contact the methyl groups in the major groove (Wake eld et al., 1999). In MBD1 however, only one of these positions contains a hydrophobic amino acid (tyrosine 34). In fact, this residue is absolutely conserved in all members of the MBD family that bind methylated DNA. Mutation of this tyrosine to alanine in MeCP2 results in a 10-fold reduction in binding a nity (Free et al.,

4 2000), the same mutation in the context of MBD1 virtually abolishes binding (Ohki et al., 1999), highlighting its functional importance to interaction of MBD proteins with methylated DNA. It seems likely that the hydrophobic residues on strand beta three contact the methyl groups in the major groove of DNA and the charged residues in strand b2, b4, and the exible loop interact with the DNA backbone. MBD2 MBD2 is highly similar to MBD3 in a large region corresponding roughly to amino acids 140 ± 400 (Hendrich and Bird, 1998). However, the MBD2 mrna codes for *140 amino acids preceding this conserved region, a similar sequence is not found in MBD3. This region of MBD2 contains a repeat consisting of glycine and arginine residues (Figure 1, Hendrich and Bird, 1998). Compared to MBD3, MBD2 has a more restricted pattern of expression and an alternatively spliced mrna is evident in testis (Hendrich and Bird, 1998). MBD2 binds methylated DNA in a manner very similar to the isolated MBD domain of MeCP2 (Hendrich and Bird, 1998; Wade et al., 1999). Surprisingly, MBD2b (a version lacking the amino terminal 140 amino acids) has been reported to possess DNA demethylase activity (Bhattacharya et al., 1999), although this nding has been questioned (Wade et al., 1999; Ng et al., 1999). Tethering of MBD2 near a promoter via a heterologous DNA binding domain results in moderate transcriptional repression that is sensitive to Trichostatin A (Ng et al., 1999). In cotransfection experiments MBD2b increased transcriptional repression observed on methylated reporter constructs (Boeke et al., 2000). Surprisingly, mapping of the portion of MBD2b required for transcriptional repression in this assay identi ed a small region that partially overlaps the MBD domain (Boeke et al., 2000). Immunoprecipitation studies demonstrate that MBD2 is physically associated with HDAC1 in mammalian cells and implicate MBD2 as the longsought methyl CpG binding component of the MeCP1 complex (Ng et al., 1999). Further, while MBD2 is associated with HDAC1 and with RbA p48/p46, by coimmunoprecipitation analysis it is not a component of the previously de ned Sin3 and Mi-2/NURD complexes (Ng et al., 2000). However, a direct interaction of Sin3A with MBD2b in the region su cient to direct transcriptional repression has also been described (Boeke et al., 2000). It seems likely that the biochemical properties of MBD2 and its associations with other proteins, like the case of MBD1, are not yet fully appreciated. MBD3 MBD3 is the smallest member of the MBD family, coding for a protein of about 30 kda. The coding sequence of MBD3 is highly similar to that of MBD2 throughout its length (Hendrich and Bird, 1998). The mrna is relatively abundant in most somatic tissues and also in ES cells (Hendrich and Bird, 1998). MBD3 presents a relatively rich variety of splice variants (Figure 3). The MBD is encoded by portions of exons 1 and 2 (Figure 2; Hendrich et al., 1999a). A subset of mammalian MBD3 RNAs utilize an upstream splice donor in the rst exon resulting in deletion of major portions of the MBD domain (Figure 3), this form of the protein has been termed MBD3D (Hendrich and Bird, 1998; Zhang et al., 1999). Xenopus laevis presents an additional alternatively spliced isoform (Figure 3), MBD3 Long Form, that inserts 20 amino acids between exons 1 and 2 (Wade et al., 1999). Both the MBD3D and MBD3 Long Form splice variants fail to bind methylated DNA substrates (Wade et al., 1999; Zhang et al., 1999). While MBD2 maintains the same exon/intron organization as MBD3, there are no known examples of MBD2 splice variants that a ect the integrity of the MBD domain (Hendrich et al., 1999a). This suggests that MBD3 has assumed at least some functions that are unrelated to DNA methylation since the gene duplication event resulting in these two very similar proteins. Indeed, the binding properties of MBD3 appear to vary with species. The mammalian protein will only bind methylated DNA in vitro under certain conditions, even then its selectivity is poor when compared to MBD2 or MeCP2 (Hendrich and Bird, 1998; Wade et al., 1999). In contrast, Xenopus MBD3 binds methylated DNA with an a nity quite similar to the isolated MBD domain from MeCP2 (Wade et al., 1999). Careful comparison of the sequences of the MBD domain in proteins experimentally proven to bind selectively to methylated DNA reveals two highly conserved residues which are altered in mammalian MBD3 (Figure 4). First, an absolutely conserved tyrosine residue in beta strand 3 is changed to phenylalanine. This residue is largely solvent exposed in both the MeCP2 and MBD1 structures and mutations result in a dramatic decrease in binding (Wake eld et al., 1999; Ohki et al., 1999). Involvement of the tyrosine hydroxyl in a hydrogen bond with a DNA base could account for the absolute conservation of this residue. The second highly conserved residue substituted in mammalian MBD3 is a conserved lysine or arginine residue near the amino terminus of strand b3 (Figure 4). If this residue makes an ionic contact with the DNA backbone, substitution with histidine might su ciently decrease binding energy to result in a qualitatively poor protein-dna interaction. Indeed, mutation of the homologous residue in MBD1 to alanine virtually abolishes DNA binding (Ohki et al., 1999). The amino acid sequence of Zebra sh MBD3 very closely resembles that of Xenopus MBD3. Speci cally, both the conserved tyrosine and lysine/arginine residues are present in the sh protein. It seems likely that this protein, like Xenopus MBD3, will bind methylated DNA selectively. Should this prediction 3169

5 3170 Figure 3 Exon/Intron organization of the MBD domain of MBD3. The exons of various species' MBD3 are indicated in the gure. Exon 1 is solid black, Exon 2 is an open white rectangle. The alternative exon of MBD3 Long Form is indicated by a shaded rectangle with black outlining. Sequence present only in the Drosophila MBD homolog is indicated by shaded rectangles without outlining. Secondary structure features as extrapolated from the published structures of MeCP2 and MBD1 are depicted above each exon Figure 4 Sequence alignment of MBD family members. The predicted protein sequences of several MBD family members are aligned in the gure. The upper set of proteins all bind methylated DNA speci cally. Residues conserved in all these proteins that di er in mammalian MBD3 are in bold in the mammalian MBD3 sequence. MeCP2 residues mutated in Rett Syndrome are in bold in the human MeCP2 sequence. Secondary structures as predicted from the structures of MeCP2 and MBD1 are indicated at the top of the alignment hold true, it suggests that at least some vertebrates retain a requirement for a subset of MBD3 proteins to interact with methylated DNA speci cally. As EST sequences for MBD2 homologs exist in both these organisms, this requirement is not likely to result from a lack of MBD2. However, both Zebra sh (Macleod et al., 1999) and Xenopus (Stancheva and Meehan, 2000 and references therein) lack the global demethylation event characteristic of early mammalian development. MBD3 is a component of a very abundant multiprotein complex containing a histone deacetylase and a chromatin remodeling enzyme, the Mi-2 complex, in Xenopus eggs (Wade et al., 1998). The developmental transcription pattern of Xenopus and by extension of zebra sh undoubtedly utilizes DNA methylation in a manner distinct from mammals. The ability of Xenopus MBD3 to bind methylated DNA and the prediction regarding the zebra sh protein strongly suggest that recruitment of the Mi-2 complex to methylated loci de nes a crucial regulatory pathway during development of these organisms. MBD4 MBD4 is the only known member of the MBD family not associated with histone deacetylase activity. While the MBD domain of MBD4 has the highest similarity to MeCP2 in the MBD family, the carboxyl terminus has homology to bacterial DNA repair enzymes (Hendrich and Bird, 1998). MBD4 is expressed in most human tissues and splice variants are evident

6 (Hendrich and Bird, 1998). In keeping with its similarity to MeCP2, an MBD4-GFP fusion is localized at densely methylated sequences in mouse cells (Hendrich and Bird, 1998). Several lines of evidence implicate a role for MBD4 in DNA repair. It interacts with the DNA repair protein MLH1in a two-hybrid screen (Bellacosa et al., 1999). Recombinant MBD4 induces nicks and linearization of supercoiled plasmids (Bellacosa et al., 1999) although the protein lacks endonuclease activity on oligonucleotide substrates (Hendrich et al., 1999b). MBD4 clearly has speci c DNA N-glycosylase activity with a strong preference for G : T mismatches (Hendrich et al., 1999b; Petronzelli et al., 2000). Surprisingly, the MBD domain does not appear to in uence the speci city of the glycosylase activity for G : T mismatches (Hendrich et al., 1999b; Petronzelli et al., 2000). However, while the MBD domain binds to symmetrically methylated CpG dinucleotides, it prefers 5-methyl CpG paired with TpG (Hendrich et al., 1999b). As this particular G : T mismatch is the expected product of deamination of a single 5-methyl C in a symmetrically methylated CpG dinucleotide, MBD4 has been designated a repair enzyme speci c for methylated DNA (Hendrich et al., 1999b; Petronzelli et al., 2000). Interestingly, mutations in the MBD4 gene have been isolated in carcinomas with microsatellite instability, where DNA mismatch repair activity is defective (Bader et al., 1999; Riccio et al., 1999). Rett syndrome and MeCP2 In October 1999, the surprising nding that several patients with the neurological disorder Rett Syndrome had mutations in the MeCP2 gene was announced (Amir et al., 1999). Rett Syndrome is a neurological disorder occurring predominantly in females characterized by normal early development followed by a period of regression. Patients lose speech and purposeful hand movements while acquiring a variety of neurological symptoms. Stabilization generally occurs and most patients survive into adulthood (Rett, 1966; Hagberg et al., 1983; Hagberg, 1985). In the past year, several groups have reported similar ndings, expanding the number of MeCP2 mutations reported to in excess of 90 mutations in more than 400 individuals (Amir et al., 2000; Buyse et al., 2000; Bienvenu et al., 2000; Cheadle et al., 2000; Hampson et al., 2000; Huppke et al., 2000; Kim and Cook, 2000; Obata et al., 2000; Wan et al., 1999; Xiang et al., 2000). Most groups report that more than 80% of classic Rett cases examined have mutations in the X-linked MeCP2 gene. Additionally, a large number of non-disease associated polymorphisms in MeCP2 have been reported (see for example, Wan et al., 1999; Buyse et al., 2000). In some cases, skewed X- inactivation pro les result in asymptomatic carrier females (Wan et al., 1999; Amir et al., 2000; Dragich et al., 2000). There appears to be little correlation between phenotype of Rett patients and speci c mutations in MeCP2, in fact, patients with identical mutations exhibit markedly di erent symptoms (Huppke et al., 2000). Several recent reports have con rmed MeCP2 mutations in males with mental defects (Orrico et al., 2000; Meloni et al., 2000). Rett syndrome mutations in MeCP2 tend to cluster in three distinct locations in the coding region: the MBD, the TRD, and the C terminus. The properties of a subset of these mutations have been examined. In the context of Xenopus MeCP2, missense mutations in the MBD domain (R106W, R133C, F155S, and T158M) reduced the ability of the protein to bind methylated DNA (Ballestar et al., 2000). Selective binding was essentially eliminated in all the mutants examined with the exception of T158M, which reduced binding a nity by a factor of two (Ballestar et al., 2000). An analysis of Rett mutants in the context of the MBD domain of rat MeCP2 revealed that the missense mutants R106W, R133C and F155S have a signi cant e ect on DNA binding a nity (Free et al., 2000). Of these mutations, both R106W and F155S appear to reduce binding a nity through structural perturbations in the MBD fold (Free et al., 2000). Mutation of arginine 133 to cysteine reduces binding a nity and also perturbs tyrosine 123 ± a residue postulated to interact with the methyl groups on the DNA substrate (Free et al., 2000; Wake eld et al., 1999). Finally, the T158M mutation had near wild-type a nity for methylated DNA (Free et al., 2000). In the context of full length human MeCP2, none of the missense mutations in the MBD domain, including T158M, retained the capacity to bind methylated DNA, emphasizing the importance of context for functional analysis of Rett mutants (Yusufzai and Wol e, 2000). As expected, truncation mutations in the TRD have no discernible e ect on selectivity for methylated DNA, although overall binding a nity is a ected (Yusufzai and Wol e, 2000). A set of TRD nonsense mutations were unable to repress transcription in a tethering assay in Xenopus oocytes, although no defect was observed for the R306C missense mutation (Yusufzai and Wol e, 2000). Interestingly, truncation mutations lacking the intact carboxyl terminus were seen to be signi cantly less stable in oocytes, suggesting involvement of this region in global folding and stability of MeCP2 (Yusufzai and Wol e, 2000). The defects observed in MeCP2 function as a result of these mutations lead to the inevitable conclusion that the neurological defects in Rett Syndrome result from loss of MeCP2 function. Further experiments are clearly required to de ne precise molecular mechanisms resulting in pathology. However, it also seems that most Rett mutations do not result in subtle changes in the properties of MeCP2, but rather catastrophically impact DNA binding, protein stability, or both. Perspectives The past decade has seen considerable progress in the understanding of how DNA methylation is translated into functional states in the genome. A family of 3171

7 3172 proteins has been identi ed that bind to methylated DNA, these proteins are found in association with enzymes that alter the fundamental properties of chromatin. At a super cial level, one can attribute the functional properties of methylated DNA to these MBD associated proteins. However, a number of interesting questions remain unanswered. Only one member of the family, MBD3, has been described in biochemical detail. Proteins associated with MBD1 and MBD2, in particular, remain to be identi ed. A second major unanswered question is the distribution of MBD proteins within the genome. While these ve proteins share a common DNA binding surface and recognize the same binding site, they are clearly di erentially partitioned. Finally, what are the consequences of disruption of this family, as occurs in Rett Syndrome? The answers to these and other unresolved issues will be crucial in understanding the functions of these key regulatory molecules. Acknowledgments I thank Timur Yusufzai and Alan Wol e for communication of results prior to publication. I thank Dr Paula Vertino and Dr Xiaodong Cheng for critical reading of this manuscript. Research in my laboratory is supported by grants from the National Institute of Child Health and Human Development (1 K22 HD ) and from the Rett Syndrome Research Foundation. I gratefully acknowledge The Massachusetts Rett Syndrome Association for nancial support. References Amir RE, Van den Veyver IB, Schultz R, Malicki DM, Tran CQ, Dahle EJ, Philippi A, Timar L, Percy AK, Motil KJ, Lichtarge O, Smith EO, Glaze DG and Zoghbi HY. (2000). Ann. Neurol., 47, 670 ± 679. AmirRE,VandenVeyverIB,WanM,TranCQ,FranckeU and Zoghbi HY. (1999). Nat. Genet., 23, 185 ± 188. Bader S, Walker M, Hendrich B, Bird A, Bird C, Hooper M and Wyllie A. (1999)., 18, 8044 ± Ballestar E, Yusufzai TM and Wol e AP. (2000). Biochemistry, 39, 7100 ± Bellacosa A, Cicchillitti L, Schepis F, Riccio A, Yeung AT, MatsumotoY,GolemisEA,GenuardiMandNeriG. (1999). Proc. Nat. Acad. Sci. USA, 96, 3969 ± Bestor TH. (2000). Hum. Mol. Genet., 9, 2395 ± Bhattacharya SK, Ramchandani S, Cervoni N and Szyf M. (1999). Nature, 397, 579 ± 583. BienvenuT,CarrieA,deRouxN,VinetMC,JonveauxP, Couvert P, Villard L, Arzimanoglou A, Beldiord C, Fontes M, Tardieu M and Chelly J. (2000). Hum. Mol. Genet., 9, 1377 ± Bird AP and Wol e AP. (1999). Cell, 99, 451 ± 454. Boeke J, Ammerpohl O, Kegel S, Moehren U and Renkawitz R. (2000). J. Biol. Chem., 275, ± Buschhausen G, Wittig B, Graessmann M and Graessmann A. (1987). Proc.Natl.Acad.Sci.USA,84, 1177 ± Buyse IM, Fang P, Hoon KT, Amir RE, Zoghbi HY and Roa BB. (2000). Am.J.Hum.Genet.,67, 1428 ± CameronEE,BachmanKE,MyohanenS,HermanJGand Baylin SB. (1999). Nat. Genet., 21, 103 ± 107. Chandler SP, Guschin D, Landsberger N and Wol e AP. (1999). Biochemistry, 38, 7008 ± Cheadle JP, Gill H, Fleming N, Maynard J, Kerr A, Leonard H, Krawczak M, Cooper DN, Lynch S, Thomas N, Hughes H, Hulten M, Ravine D, Sampson JR and Clarke A. (2000). Hum. Mol. Genet., 9, 1119 ± Cooper DN and Krawczak M. (1989). Hum. Genet., 83, 181 ± 188. Cross SH and Bird AP. (1995). Curr. Opin. Genet. Dev., 5, 309 ± 314. Cross SH, Meehan RR, Nan X and Bird A. (1997). Nat. Genet., 16, 256 ± 259. Dragich J, Houwink-Manville I and Schanen C. (2000). Hum. Mol. Genet., 9, 2365 ± Free A, Wake eld RI, Smith BO, Dryden DT, Barlow PN and Bird AP. (2000). J. Biol. Chem., 276, 3353 ± Fujita N, Takebayashi S, Okumura K, Kudo S, Chiba T, Saya H and Nakao M. (1999). Mol. Cell. Biol., 19, 6415 ± Hagberg B. (1985). Acta Paediatr. Scand., 74, 405 ± 408. Hagberg B, Aicardi J, Dias K and Ramos O. (1983). Ann. Neurol., 14, 471 ± 479. Hampson K, Woods CG, Latif F and Webb T. (2000). J. Med. Genet., 37, 610 ± 612. Hendrich B and Bird A. (1998). Mol. Cell. Biol., 18, 6538 ± Hendrich B and Bird A. (2000). Curr. Top. Microbiol. Immunol., 249, 55 ± 74. Hendrich B, Abbott C, McQueen H, Chambers D, Cross S and Bird A. (1999a). Mamm. Genome, 10, 906 ± 912. HendrichB,HardelandU,NgHH,JiricnyJandBirdA. (1999b). Nature, 401, 301 ± 304. Huppke P, Laccone F, Kramer N, Engel W and Hanefeld F. (2000). Hum. Mol. Genet., 9, 1369 ± Jones PL, Veenstra GJ, Wade PA, Vermaak D, Kass SU, Landsberger N, Strouboulis J and Wol e AP. (1998). Nat. Genet., 19, 187 ± 191. Kaludov NK and Wol e AP. (2000). Nucl. Acids Res., 28, 1921 ± Kass SU, Landsberger N and Wol e AP. (1997). Curr. Biol., 7, 157 ± 165. Kim SJ and Cook Jr EH. (2000). Hum. Mutat. (Online), 15, 382 ± 383. Lewis JD, Meehan RR, Henzel WJ, Maurer-Fogy I, Jeppesen P, Klein F and Bird A. (1992). Cell, 69, 905 ± 914. Li E, Bestor TH and Jaenisch R. (1992). Cell, 69, 915 ± 926. Macleod D, Clark VH and Bird A. (1999). Nat. Genet., 23, 139 ± 140. Meehan RR, Lewis JD and Bird AP. (1992). Nucl. Acids Res., 20, 5085 ± Meloni I, Bruttini M, Longo I, Mari F, Rizzolio F, D'Adamo P, Denvriendt K, Fryns JP, Toniolo D and Renieri A. (2000). Am.J.Hum.Genet.,67, 982 ± 985. Nan X, Campoy FJ and Bird A. (1997). Cell, 88, 471 ± 481. Nan X, Meehan RR and Bird A. (1993). Nucl. Acids Res., 21, 4886 ± Nan X, Ng HH, Johnson CA, Laherty CD, Turner BM, Eisenman RN and Bird A. (1998). Nature, 393, 386 ± 389. Ng HH, Jeppesen P and Bird A. (2000). Mol. Cell. Biol., 20, 1394 ± 1406.

8 NgHH,ZhangY,HendrichB,JohnsonCA,TurnerBM, Erdjument-Bromage H, Tempst P, Reinberg D and Bird A. (1999). Nat. Genet., 23, 58 ± 61. Obata K, Matsuishi T, Yamashita Y, Fukuda T, Kuwajima K, Horiuchi I, Nagamitsu S, Iwanaga R, Kimura A, Omori I, Endo S, Mori K and Kondo I. (2000). J. Med. Genet., 37, 608 ± 610. Ohki I, Shimotake N, Fujita N, Nakao M and Shirakawa M. (1999). EMBO J., 18, 6653 ± Okano M, Bell DW, Haber DA and Li E. (1999). Cell, 99, 247 ± 257. Orrico A, Lam C, Galli L, Dotti MT, Hayek G, Tong SF, Poon PM, Zappella M, Federico A and Sorrentino V. (2000). FEBS Lett., 481, 285 ± 288. Petronzelli F, Riccio A, Markham GD, Seeholzer SH, Stoerker J, Genuardi M, Yeung AT, Matsumoto Y and Bellacosa A. (2000). J. Biol. Chem., 275, ± Razin A. (1998). EMBO J., 17, 4905 ± Rett A. (1966). Wien. Med. Wochenschr., 116, 723 ± 726. Riccio A, Aaltonen LA, Godwin AK, Loukola A, Percesepe A, Salovaara R, Masciullo V, Genuardi M, Paravatou- Petsotas M, Bassi DE, Ruggeri BA, Klein-Szanto AJ, Testa JR, Neri G and Bellacosa A. (1999). Nat. Genet., 23, 266 ± 268. Robertson KD and Wol e AP. (2000). Nat. Rev. Genet., 1, 11 ± 19. Stancheva I and Meehan RR. (2000). Genes Dev., 14, 313 ± 327. WadePA,GegonneA,JonesPL,BallestarE,AubryFand Wol e AP. (1999). Nat. Genet., 23, 62 ± 66. Wade PA, Jones PL, Vermaak D and Wol e AP. (1998). Curr. Biol., 8, 843 ± 846. Wake eld RI, Smith BO, Nan X, Free A, Soteriou A, Uhrin D, Bird AP and Barlow PN. (1999). J. Mol. Biol., 291, 1055 ± Wan M, Lee SS, Zhang X, Houwink-Manville I, Song HR, Amir RE, Budden S, Naidu S, Pereira JL, Lo IF, Zoghbi HY, Schanen NC and Francke U. (1999). Am. J. Hum. Genet., 65, 1520 ± Xiang F, Buervenich S, Nicolao P, Bailey ME, Zhang Z and Anvret M. (2000). J. Med. Genet., 37, 250 ± 255. Yusufzai TM and Wol e AP. (2000). Nucl. Acids Res., 28, 4172 ± Zhang Y, Ng HH, Erdjument-Bromage H, Tempst P, Bird A and Reinberg D. (1999). Genes Dev., 13, 1924 ±