CTCF maintains regulatory homeostasis of cancer pathways

Abstract

Background: CTCF binding to DNA helps partition the mammalian genome into discrete structural and regulatory domains. Complete removal of CTCF from mammalian cells causes catastrophic genome dysregulation, likely due to widespread collapse of 3D chromatin looping and alterations to inter- and intra-TAD interactions within the nucleus. In contrast, Ctcf hemizygous mice with lifelong reduction of CTCF expression are viable, albeit with increased cancer incidence. Here, we exploit chronic Ctcf hemizygosity to reveal its homeostatic roles in maintaining genome function and integrity.

Results: We find that Ctcf hemizygous cells show modest but robust changes in almost a thousand sites of genomic CTCF occupancy; these are enriched for lower affinity binding events with weaker evolutionary conservation across the mouse lineage. Furthermore, we observe dysregulation of the expression of several hundred genes, which are concentrated in cancer-related pathways, and are caused by changes in transcriptional regulation. Chromatin structure is preserved but some loop interactions are destabilized; these are often found around differentially expressed genes and their enhancers. Importantly, the transcriptional alterations identified in vitro are recapitulated in mouse tumors and also in human cancers.

Conclusions: This multi-dimensional genomic and epigenomic profiling of a Ctcf hemizygous mouse model system shows that chronic depletion of CTCF dysregulates steady-state gene expression by subtly altering transcriptional regulation, changes which can also be observed in primary tumors.

Keywords: CTCF; Cancer; Chromatin architecture; Chromatin state; Hemizygosity; Transcription.

Fig. 1

Ctcf hemizygosity as a model to subtly perturb nuclear homeostasis. a The engineered Ctcf locus contains loxP sites flanking the protein-coding exons of the gene (wild-type (WT), Ctcf ^+/+), which can be removed using Cre recombinase (Ctcf ^+/−). Mouse embryonic fibroblast (MEF) lines were derived from six WT and six Ctcf ^+/− littermates. Quantitative analyses of CTCF binding, transcription, proteome, chromatin state, and chromatin structure were performed in multiple biological replicates (“Methods”). b Validation of CTCF depletion in Ctcf ^+/− MEF cultures. Quantification of Ctcf deletion by qRT-PCR and quantitative western blotting experiments show that there is only partial compensation in the level of CTCF from DNA to RNA to protein

Fig. 2

Ctcf hemizygosity results in altered chromatin binding. a Differential binding analysis identified 787 CTCF binding sites differentially occupied between Ctcf hemizygous and wild-type MEFs, most of which show reduced genomic occupancy in the Ctcf ^+/− MEFs. Significant changes are shown in red (FDR < 5%). b Example genome tracks showing highly consistent loss of CTCF binding at three genomic loci overlapping the genes indicated at the top. Data are shown for the five biological replicates that passed quality control, normalized to account for sequencing depth differences. WT wild-type

Fig. 3

Differentially bound CTCF loci are found near genes, occur in shorter motifs, and have lower binding affinity and evolutionary conservation. a Differential CTCF binding sites were significantly enriched within promoters and gene bodies compared to stable CTCF binding sites (chi-square test, p = 4.9 × 10^− 10). b Stable CTCF peaks had a higher proportion of the longer (~ 33 bp) motif word compared to the differential sites. Multiple alignments of a randomly chosen subset of a hundred CTCF binding sites that are either stable or differential are shown. Each position in the alignment is colored corresponding to the nucleotide present, following the color scheme used in the CTCF motif logo shown at the top. c Binding sites susceptible to reduced CTCF concentration have significantly lower motif affinity (Mann-Whitney test, p < 2.2 × 10^− 16). d Regions bound by CTCF across the mouse lineage are less sensitive to Ctcf hemizygosity. Example tracks are shown of a stable CTCF binding site that is conserved in five species of mice, compared to a differential site that is found in only a subset of the species, M. = Mus. (M. musculus chr6:120,736,800 for the stable site and chr2:31,887,060 for the differential site). *** p value < 0.001

Fig. 4

CTCF depletion dysregulates oncogenic pathways. a Nearly 300 genes are differentially expressed between wild-type and Ctcf^+/− cells; significant changes are shown in black (FDR < 5%). b Transcriptional changes (x-axis) were highly correlated to corresponding changes at the protein level (y-axis); Spearman’s correlation coefficient is noted in the top left corner. Of all genes from a, 85% had concordant fold-change estimates in the proteomics dataset. c Gene set enrichment analysis performed on the differentially expressed (DE) genes highlights dysregulation of cancer related pathways. Representative significantly enriched terms from the Gene Ontology and KEGG pathways are shown. d Differentially expressed genes are strongly enriched for having higher numbers of CTCF binding sites (BS) than genes with stable expression, in their gene bodies or flanking 5 kb

Fig. 5

Transcriptional perturbations arise from regulatory changes in the nuclear genome. a Changes in expression of dysregulated genes are accompanied by changes in the activity of their proximal promoters, as well as enhancers linked via chromatin loops. On the right the expression differences between wild-type and Ctcf ^+/− cells are shown, ordered by increasing fold change; genes expressed at lower levels in the hemizygous cells are in blue, whereas those expressed higher are in red. To the left, a heatmap of the difference in mean abundance of H3K4me3 occupancy is shown. Each column is a 5 kb window, extending 17.5 kb up- and downstream of each gene’s transcription start site, which is in the center. On the far left, an equivalent heatmap for the difference in occupancy of H3K27ac, centered at the midpoint of the peak. Gene–enhancer pairs were inferred from significant interactions identified from Hi-C data and thus elements can be separated by large distances. For each gene, the enhancer with most regulatory potential is shown (“Methods”). The same color scale is used throughout. b Transcriptional changes are accompanied by concordant changes in the activity of their regulatory elements. In Ctcf hemizygous cells there was reduced gene expression of Dusp4, lower occupancy of promoter marks (H3K4me3 and H3K27ac), and reduced binding at the associated enhancer element (shaded boxes). Two representative replicates of equivalent sequencing depth are shown for each dataset

Fig. 6

Concordant gene alterations in diverse murine and human tumors. The comparison of the set of CTCF-dependent genes and those differentially expressed in mouse liver tumors or in human uterine and breast tumors revealed a large overlap. The majority of these changed in the same direction (concordant) in the Ctcf hemizygous MEFs and the tumor samples. Additionally, the set of genes unique to MEFs is indicated (not differential) and those that were either not expressed or did not have a one-to-one ortholog in the human genome (not common). The concordant gene changes across these diverse tumors are highly overlapping, as seen in the Venn diagram