Integrated epigenomic analyses of neuronal MeCP2 reveal a role for long-range interaction with active genes

Abstract

Mutations in MECP2 cause the autism-spectrum disorder Rett syndrome. MeCP2 is predicted to bind to methylated promoters and silence transcription. However, the first large-scale mapping of neuronal MeCP2-binding sites on 26.3 Mb of imprinted and nonimprinted loci revealed that 59% of MeCP2-binding sites are outside of genes and that only 6% are in CpG islands. Integrated genome-wide promoter analysis of MeCP2 binding, CpG methylation, and gene expression revealed that 63% of MeCP2-bound promoters are actively expressed and that only 6% are highly methylated. These results indicate that the primary function of MeCP2 is not the silencing of methylated promoters.

Fig. 1.

The majority of MeCP2-binding sites are intergenic or intronic. MeCP2 sites from 26.3 Mb of human genomic sequence at the moderate L3 level of statistical significance are represented. Similar analyses at all four significance levels (L1–L4) are shown in SI Table 1. (Center) A total of 170 MeCP2 sites were divided into intragenic (blue) and intergenic (red) or both (purple). (Left) Intragenic sites were further subdivided into exclusive intronic or exonic or overlapping both introns and exons (shades of blue). (Right) Intergenic sites were categorized according to distances from gene transcription start sites (43) or transcription end sites (TES) (shades of red).

Fig. 2.

Validation of ChIP–chip analysis on established MeCP2-binding sites. A representative histogram of MeCP2 log2 signal ratio values is shown below the University of California, Santa Cruz, Genome Browser window containing known genes, CpG islands (green boxes), and L3 sites (black and pink boxes) for each MeCP2 target gene locus. For DLX5 and DLX6 controls, one site was previously identified in mouse (right pink box) (19) and validated (SI Fig. 7), and another site in DLX6 (left pink box) was also identified in mouse and validated by conventional ChIP analysis (19). For BDNF, an MeCP2 site in the first intron (pink box) was observed at the same position as previously shown in mouse and rat Bdnf (21, 22) and appears to control activation-induced transcription of the gene in neurons (44). Additional new MeCP2 sites were observed and validated (black boxes). A site in SNRPN that serves as the paternal imprinting control region for 15q11–13 was included (pink box) as an additional control. Analyses of this site in both mouse and human by conventional ChIP revealed binding by MeCP2 (16, 17). However, binding of MeCP2 to this site does not affect transcription of SNRPN (16, 19). The SNRPN site was validated by ChIP–chip at levels L1–L4.

Fig. 3.

JUNB is a partially methylated, actively expressed gene regulated by MeCP2 binding. Three previously undescribed MeCP2-binding sites (black boxes) were found upstream of JUNB, with another in the first intron of RNASEH2A, overlapping with a CpG island (green box). (Upper) Shown in the histograms, high levels of MeCP2 binding (red histograms) were correlated with high levels of Pol2 binding (green histograms), a correlation also observed in ChIP–chip analysis of MeCP2 and Pol2 binding to genome-wide promoters. (Lower) Shown is bisulfite analysis of JUNB sites 1 and 2 compared with higher resolution of MeCP2 ChIP–chip histograms (open circles, unmethylated; filled circles, methylated CpG sites). Partial methylation of CpGs was observed in undifferentiated (UD) and differentiated (D) SH-SY5Y cells, especially methylation of CpGs next to AT runs (40), which are indicated by an asterisk. In contrast, the proximal CpG island promoter of this actively expressed gene was unmethylated and de-enriched for MeCP2 binding relative to the methylated upstream regions.

Fig. 4.

The majority of MeCP2-bound promoters genome-wide are expressed genes. Comparison of the strongest MeCP2-bound promoters with genes expressed in differentiated SH-SY5Y neurons is shown. (Center) Shown is a Venn diagram of 1,524 MeCP2 promoter hits with 11,247 expressed genes (blue circle), which reveals that 954 or 62.6% of MeCP2-bound promoters are transcriptionally active genes. (Left) The strongest MeCP2-bound promoters were identified by comparing the top 2,600 bound promoters, removing duplicate promoters to obtain 2,563 and 2,567 hits, respectively, for two replicate arrays, ChIP-1 and ChIP-2. (Right) Expression microarray analysis of 48-h-differentiated SH-SY5Y cells in three biologic replicates (arrays 1–3) (20) was reanalyzed using a determination of expressed genes by Affymetrix mismatch analysis to identify a conservative common set of 11,247 expressed genes in SH-SY5Y neurons.

Fig. 5.

Highly methylated promoters are not bound by MeCP2. (Center) The 1,524 promoters with the highest levels of MeCP2 binding identified in Fig. 4 were compared with the 4,062 most highly methylated promoters, revealing an overlap of only 91 genes. Therefore, only 2.2% of highly methylated promoters were bound by MeCP2, and only 6.0% of MeCP2-bound promoters were highly methylated. (Left and Right) Two replicate MeDIP assays were performed on genomic DNA from SH-SY5Y cells (Right) differentiated identically to those used for MeCP2 promoter ChIP–chip (Left) and expression profiling (see Materials and Methods). The median top 5,000 log2 promoter signals were selected from each experiment, and, after the removal of duplicate promoters, the overlap produced 4,062 genes in common.