The solution structure of the domain from MeCP2 that binds to methylated DNA¹

Abstract

MeCP2 is an abundant mammalian protein that binds methylated CpG (mCpG) sequences within double-stranded DNA, represses transcription by recruiting histone deacetylases, and is essential for embryonic development. It is one of a family of proteins which mediate the biological consequences of DNA methylation. These proteins each possess a sequence motif of about 70 residues which, in MeCP2, form a domain necessary and sufficient for binding to mCpG. The solution structure of the mCpG-binding domain (MBD) from MeCP2 has been solved and the DNA-binding surface of the domain mapped using NMR spectroscopy. Residues 95–162 of MeCP2 adopt a novel fold forming a wedge-shaped structure. An N-terminal four-stranded antiparallel β-sheet forms one face of the wedge, while the other face is formed mainly by a C-terminal helical region. The thin end of the wedge is extended by a long loop between β-strands B and C containing many basic residues. The B-C loop together with residues in strands B, C and D, and at the N terminus of the α-helix, appears to form an interface with methylated DNA. Unstructured residues at the NH₂ terminus of the domain are also involved in formation of the complex. The presence of numerous arginine and lysine side-chains on the DNA-binding surface of MBD is consistent with the requirement for the mCpG site to be flanked by non-specific sequences of base-pairs. The absence of symmetry in the domain implies that recognition does not exploit the symmetry of the binding site. A conserved hydrophobic pocket containing the side-chains of Tyr123 and Ile125 on the positively charged β-sheet face is a candidate for the region of contact with the methyl-groups of the modified cytosine residues.

Introduction

The major covalent modification of the eukaryotic genome is the addition of a methyl group at the 5-position of cytosine in cytidine-guanosine (CpG) dinucleotide pairs. Because of its ability to repress transcription, methylation of CpG sequences has been implicated in stable modulations of gene expression in a cell-type-specific manner during development (Bartolomei & Tilghman, 1997). Mice lacking maintenance DNA methyltransferase (DNMT) are developmentally retarded and die at mid-gestation (Li et al., 1992).

While certain invertebrate genomes have a pattern of DNA methylation characterised by methylation-rich and methylation-free regions, the somatic genomes of vertebrates are consistently methylated except for regions known as “CpG islands” (Bird, 1986). CpG islands are coincident with the promoters of nearly two-thirds of human genes transcribed by RNA polymerase II (Antequera & Bird, 1994). DNA methylation can repress transcription at a distance Cedar 1988, Kass et al 1993, and repression of methylated transgenes requires prior assembly of chromatin (Buschhausen et al., 1987). Silencing of genes by methylation at CpG islands has been observed in the inactive X-chromosome (Riggs & Pfeifer, 1992), in genomic imprinting Neumann and Barlow 1996, Razin and Cedar 1994, and in transformed cell-lines and tumours (e.g. Bird, 1996).

A family of five mammalian proteins, MeCP2, MBD1, MBD2, MBD3 and MBD4, has been identified (Hendrich & Bird, 1998, and references therein) as being of likely importance in interpreting the signal that methylation of DNA represents. Each member has a stretch of 60–80 residues displaying a high level of similarity (50–70 %) between all five proteins (Figure 1). In MeCP2, this region has been identified as necessary for binding directly to methyl-CpG (mCpG, Nan et al., 1993), and it has been termed the mCpG-binding domain (MBD). MBD2 and MBD4 bind specifically to methylated DNA in vitro (Hendrich & Bird, 1998). Both proteins co-localise with foci of heavily methylated satellite DNA in mouse cells and localisation is disrupted in cells deficient in CpG methylation (Hendrich & Bird, 1998). MBD2 has been reported to possess demethylase activity (Bhattacharya et al., 1999) and MBD4 is probably a novel repair enzyme (Bellacosa et al., 1999). MBD3 apparently lacks specific methylated DNA-binding activity, although it has a high degree of sequence similarity with MBD2 (Figure 1). MBD1 and MeCP2 bind specifically to methylated DNA in any sequence context and inhibit transcription Meehan et al 1989, Lewis et al 1992, Boyes and Bird 1991, Nan et al 1997.

MeCP2 is capable of binding to a single, symmetrically methylated CpG dinucleotide and binds to chromosomes at sites known to contain methylated DNA (Nan et al., 1996). MeCP2 (Figure 1) contains both an MBD and a transcriptional repression domain. It is able to recruit histone deacetylases Nan et al 1998, Jones et al 1998, and to displace histone H1 from preassembled chromatin that contains methylated DNA. Its localisation is disrupted in embryonic stem cells (ES) lacking a functional DNMT gene (Nan et al., 1996). Like DNMT-deficient ES, MeCP2-deficient ES grow normally in culture but are incapable of supporting embryonic development (Tate et al., 1996). On the other hand, MeCP2-deficient somatic cells are viable and MeCP2 may be responsible for genome-wide transcriptional repression or “transcriptional noise reduction” (Bird, 1995), but may also effect suppression of specific genes.

The boundaries of the minimal domain of MeCP2 necessary and sufficient for binding mCpG were defined empirically using mutagenesis and band-shift assays (Nan et al., 1993). While residues 98–161 encompass the consensus MBD residues in MeCP2 (Figure 1), additional residues within MeCP2 at either the COOH or the NH₂ terminus of the consensus sequence are required for binding. The minimum length mCpG-binding sequence in MeCP2 consists of amino acid residues 78–162. A recombinant fragment corresponding to this sequence was shown to bind to a single CpG site with an affinity in the order of 1 nM, and as a monomer (Nan et al., 1993). Symmetrically methylated CpG in the context of double-stranded DNA, flanked by non-specific sequences of no less than six base-pairs is a requirement of the target (Nan et al., 1993).

Given that the MBD is pivotal to the biological manifestation of the methylation signal in eukaryotic DNA, its structure-function relationships are of interest. Here, the solution structure of a recombinant fragment of MeCP2 encompassing its MBD, and capable of recognising mCpG is described, and on the basis of chemical shift perturbations, the DNA-binding surface of the domain is identified.