Genome-wide and parental allele-specific analysis of CTCF and cohesin DNA binding in mouse brain reveals a tissue-specific binding pattern and an association with imprinted differentially methylated regions

Abstract

DNA binding factors are essential for regulating gene expression. CTCF and cohesin are DNA binding factors with central roles in chromatin organization and gene expression. We determined the sites of CTCF and cohesin binding to DNA in mouse brain, genome wide and in an allele-specific manner with high read-depth ChIP-seq. By comparing our results with existing data for mouse liver and embryonic stem (ES) cells, we investigated the tissue specificity of CTCF binding sites. ES cells have fewer unique CTCF binding sites occupied than liver and brain, consistent with a ground-state pattern of CTCF binding that is elaborated during differentiation. CTCF binding sites without the canonical consensus motif were highly tissue specific. In brain, a third of CTCF and cohesin binding sites coincide, consistent with the potential for many interactions between cohesin and CTCF but also many instances of independent action. In the context of genomic imprinting, CTCF and/or cohesin bind to a majority but not all differentially methylated regions, with preferential binding to the unmethylated parental allele. Whether the parental allele-specific methylation was established in the parental germlines or post-fertilization in the embryo is not a determinant in CTCF or cohesin binding. These findings link CTCF and cohesin with the control regions of a subset of imprinted genes, supporting the notion that imprinting control is mechanistically diverse.

Figure 1.

(A) ChIP-seq was performed for CTCF and cohesin (RAD21) on P21 brain in B × C and C × B F₁ hybrid animals. The experimental design and number of uniquely mapped reads taken forward for further analysis are shown. (B) Regions of CTCF and RAD21 binding were identified using the Useq, and regions identified with a FDR of <13 were considered significant and were tested for parent-of-origin-specific binding. (Black bar) The number of reads for each experiment that fell at, or within ±500 bp of a binding region; (white bar) indicates reads in binding regions that aligned over a SNP between C57BL/6 (Bl6) and Mus m. castaneus (cast); (gray bar) the number of reads after the paired reads are considered together and the best-quality read is used to map the read to Bl6 or cast. (Hatched bar) The final number of reads assigned. There was a consistent bias toward the reference sequence (C); however, this effect was eliminated after we combined B × C and C × B reads (D).

Figure 2.

Overlap of the 49,358 CTCF and 52,938 cohesin binding regions in mouse brain. This demonstrates that just over half of CTCF (55%) and cohesin (51%) binding sites are shared, suggesting both independent and combinatorial functions for CTCF and cohesin in the 3-wk mouse brain.

Figure 3.

CTCF binding analysis. (A) MEME motif finder was executed on the CTCF binding locations identified by ChIP-seq in brain and compared with motifs identified using previously published ES cells and liver binding locations. Each data set found the canonical motif with high degrees of certainty. (B) The level of cytosine methylation within CTCF binding sites in the brain was compared with that across the whole genome using data from Xie et al. (2012). In both CpG and non-CpG context cytosine methylation, cytosines within CTCF binding sites are methylated less than those outside CTCF binding sites (χ² contingency table tests, P < 0.001 for CpG and non-CpG context), confirming that CTCF prefers to bind to unmethylated DNA. (C) Genomic locations of CTCF binding are normalized to the proportion of the genome that constitutes each location (represented by the red line). This was considered for all CTCF peaks called with an FDR < 13 and separately for the 116 regions where CTCF binding was seen on one parental allele only (regions identified with a P < 0.001). CTCF is significantly enriched at genic regions, but depleted in distal intergenic regions. Parent-of-origin-specific CTCF binding locations are similar but show that binding is depleted in introns but not in intergenic regions.

Figure 4.

(A) Proportional Venn diagrams comparing coincidence of CTCF binding sites between ES cells, liver, and brain demonstrate significant overlap of CTCF binding in these tissues, Coincident binding was also considered after the removal of binding regions containing the consensus CTCF motif; overlap of CTCF binding in the absence of the consensus motif was much lower than when all binding sites were considered. (B) The percentages of shared peaks for each tissue type for all peaks and for nonmotif peaks.

Figure 5.

Chromosomal location of genome-wide significant parent-of-origin-specific CTCF binding regions. Where CTCF is bound on the maternally inherited allele, this is illustrated with a circle; where CTCF is bound on the paternally inherited allele, this is illustrated with a square. Only chromosomes where parent-of-origin-specific binding was seen are shown. CpG density is indicated.

Figure 6.

Multiple parent-of-origin-specific CTCF binding sites are observed on the paternal allele at the Magel2/Peg12 locus. (Triangles) Paternally bound CTCF binding sites. Genes and CpG islands are indicated. This region represents a unique example in the mouse genome of CTCF bound only on the paternal allele at eight regions in close proximity. This figure was adapted from the UCSC Genome Browser.