The binding sites for the chromatin insulator protein CTCF map to DNA methylation-free domains genome-wide

Abstract

All known vertebrate chromatin insulators interact with the highly conserved, multivalent 11-zinc finger nuclear factor CTCF to demarcate expression domains by blocking enhancer or silencer signals in a position-dependent manner. Recent observations document that the properties of CTCF include reading and propagating the epigenetic state of the differentially methylated H19 imprinting control region. To assess whether these findings may reflect a universal role for CTCF targets, we identified more than 200 new CTCF target sites by generating DNA microarrays of clones derived from chromatin-immunopurified (ChIP) DNA followed by ChIP-on-chip hybridization analysis. Target sites include not only known loci involved in multiple cellular functions, such as metabolism, neurogenesis, growth, apoptosis, and signalling, but potentially also heterochromatic sequences. Using a novel insulator trapping assay, we also show that the majority of these targets manifest insulator functions with a continuous distribution of stringency. As these targets are generally DNA methylation-free as determined by antibodies against 5-methylcytidine and a methyl-binding protein (MBD2), a CTCF-based network correlates with genome-wide epigenetic states.

Figure 1

Characterization of the CTCF target-site library. (A) The bulk of the CTCF target-site library of mouse fetal liver interacts with CTCF in vitro as determined by bandshift analysis. Lane 1 depicts inserts from the library cut with NotI as probe and no protein; lane 2 shows band-shift with recombinant CTCF. The specificity of the band shift was ascertained by including a 100-fold molar excess of cold H19 ICR as competitor (lane 3). (B) Band shift assays with nine randomly picked individual clones. The first eight clones are positive, whereas clone 1006 is negative for band-shift. The – and + symbols indicate absence and presence of recombinant CTCF, respectively. (C) Microarray analysis of the CTCF target-site library. From left to right, the images show total DNA estimation using an oligo hybridization probe, pattern of in vivo (using ChIP DNA from mouse fetal liver) and in vitro (using recombinant CTCF) interactions. Each clone is represented by four adjacent spots that were printed in duplicates on each microarray. (D) The random amplification of ChIP DNA is unbiased for five randomly selected sequences. The first five lanes show the genomic DNA concentration to ascertain that the PCR amplifications were performed under semiquantitative conditions. (E) Comparison of two independent ChIP-on-chip hybridization experiments. (F) Quantitative scatterplots of in vivo/in vitro binding patterns distributed over different classes of sequences as indicated in the images. The red lines represent an estimate of in vivo/in vitro CTCF binding efficiency based on the 10 highest values.

Figure 2

Immunofluorescent analysis of CTCF and HP1β in murine lung fibroblast cells. Colocalization of CTCF (Cy-3) and HP1β (FITC) clusters within the nuclei (DAPI) was seen after double immunostaining using rabbit anti-CTCF and goat anti-HP1β antibodies. Magnification ×1000. From left to right, images show merged CTCF/HP1/DAPI, CTCF (red), HP1 (green), DAPI (blue).

Figure 3

Depiction of positions of CTCF target sites showing significant in vivo binding. A 300-kb window surrounding each target site (red arrow) is displayed for a selection of loci, including those with known functions in oncogenesis. Additional loci are described in Table 1 and in the Supplemental material.

Figure 4

The insulator trap assay. (A) Schematic maps of the various constructs used in the classical insulator study. Symbols explained at the bottom of the panel. Each construct is linked to its performance in the enhancer-blocking assays, which were normalized to RNA input and episome copy number. The SV40 enhancer-driven expression of the pREPH19A construct was assigned a value of 100 whereas all other samples were normalized relative to this value. The mean deviation of three different experiments is indicated for each vector construct. (B) Schematic maps of the different pREPtox vectors. Cerise circle: the position of the SV40 enhancer. Green square: the H19 promoter. Pink and red blocks: the different inserts from clones (indicated by its original number) and H19 ICR, respectively. The numbers of the surviving clones were estimated from a colony count assay. (C) Outline of the strategy of the toxin-A assay and its application in microarray analysis of the CTCF target-site library. (D) An example of hybridization with input library sequences, affinity-purified (with recombinant CTCF) CTCF target sites, and the selection of clones with enhancer-blocking properties. (E) Presents scatter plot analyses of insulator strength, determined from the microarray analysis, and in vitro binding patterns, broken down into different sequence categories of the CTCF library. (F) Shows a scatter plot analysis between the insulator/in vitro binding ratios and DNA content of the corresponding spots of the microarrays, as determined by oligo hybridization.

Figure 5

Cross-referencing methylation status with CTCF occupancy. (A) shows that an antibody against 5-methylcytidine immunopurifies only the methylated paternal H19 ICR allele if the maternally inherited allele is of the wild type. Conversely, when the mutated H19 ICR allele is inherited maternally (labeled 142* and unable to interact with CTCF in vivo while displaying massive de novo methylation; Pant et al. 2003), both alleles are brought down as determined by using PCR primers spanning CTCF target site #3 and a diagnostic EcoRV site (Pant et al. 2003). (B,C) Scatterplot analyses comparing CTCF in vivo occupancy/insulator strength vs. single CpG methylation (B, using an antibody against methylated cytidine) and clustered (C, using an antibody against MBD2) CpG methylation states in mouse fetal liver.