Dynamic CTCF binding directly mediates interactions among cis-regulatory elements essential for hematopoiesis

Abstract

While constitutive CCCTC-binding factor (CTCF)-binding sites are needed to maintain relatively invariant chromatin structures, such as topologically associating domains, the precise roles of CTCF to control cell-type-specific transcriptional regulation remain poorly explored. We examined CTCF occupancy in different types of primary blood cells derived from the same donor to elucidate a new role for CTCF in gene regulation during blood cell development. We identified dynamic, cell-type-specific binding sites for CTCF that colocalize with lineage-specific transcription factors. These dynamic sites are enriched for single-nucleotide polymorphisms that are associated with blood cell traits in different linages, and they coincide with the key regulatory elements governing hematopoiesis. CRISPR-Cas9-based perturbation experiments demonstrated that these dynamic CTCF-binding sites play a critical role in red blood cell development. Furthermore, precise deletion of CTCF-binding motifs in dynamic sites abolished interactions of erythroid genes, such as RBM38, with their associated enhancers and led to abnormal erythropoiesis. These results suggest a novel, cell-type-specific function for CTCF in which it may serve to facilitate interaction of distal regulatory emblements with target promoters. Our study of the dynamic, cell-type-specific binding and function of CTCF provides new insights into transcriptional regulation during hematopoiesis.

Graphical abstract

Figure 1.

Dynamic CTCF occupancy during hematopoiesis. (A) Schematic of the primary human blood cells analyzed. HSPCs, T cells, B cells, and monocytes were purified by their corresponding purification beads. Erythroblasts were differentiated in vitro from HSPCs. (B) Heatmap in each panel represents the CTCF ChIP-seq signals within the 2-kb window centered on ChIP-seq peak summits specific to different blood cell types. (C) Genome browser tracks showing an erythroid-dynamic CTCF-binding site (highlighted in pink). (D) Top overrepresented motifs among the dynamic CTCF-binding sites from different types of blood cells. (E) The color chart displays the DNA sequences around the summits of CTCF-bound peaks, aligned with the DNA sequence of core CTCF motif and upstream motifs. Red represents A, blue represents C, green represents T, and yellow represents G. (F) The heatmap represents the ChIP-seq signals within a 2-kb window centered on dynamic CTCF-binding sites and the location of TF binding site motifs in different blood cell types. The DNA segments in each panel are in the same order for each cell type, decreasing from the highest aggregated CTCF ChIP-seq signals within a 2-kb window. The graphs above each heatmap show the aggregated ChIP-seq signal or frequency of matches to motifs. Ery, erythroblast; mono, monocyte.

Figure 2.

Functional importance of dynamic CTCF-binding sites. (A) Heatmap showing the unsupervised hierarchical clustering based on Pearson correlation of ATAC-seq signals on dynamic CTCF peaks across different blood lineages. The hierarchical cluster was generated based on the Pearson correlation of normalized ATAC-seq signals within dynamic CTCF sites. The chromatin accessibility within dynamic CTCF sites in each blood cell types were calculated based on ATAC-seq data generated in the previous publication. (B) Functional annotation results of dynamic CTCF sites for the top Gene Ontology cellular component and phenotype from GREAT. The radius of the circle represents the significance, −log10(FDR), of the enrichment. (C) Enrichment significance of trait-associated SNPs within different types of dynamic CTCF-binding sites. (D) Schematic of the pooled sgRNA library design and screening procedure used to experimentally validate the function of dynamic CTCF-binding sites. (E) Pie chart showing the numbers of dynamic CTCF-binding sites with different functions based on sgRNA perturbation results. CLP, common lymphoid progenitor; CMP, common myeloid progenitor; GMP, granulocyte-monocyte progenitor; HSC, hematopoietic stem cell; LMPP, lymphomyeloid primed progenitor; MEP, megakaryocytic-erythroid progenitor; MPP, myeloid primed progenitor; NK, natural killer.

Figure 3.

Epigenetic and 3D chromatin features of dynamic CTCF occupancy. (A) Heatmap showing the signals of ATAC-seq, H3K27ac, and H3K27me3 around the 20-kb windows centered on the peak summits of GOSs (top) and LOSs (bottom) in HSPCs and erythroblasts. (B) Heatmap showing the methylation levels of CpG sites within the CTCF motifs in HSPC and erythroblast dynamic sites. Each row represents 1 CpG site. Left panel represents the CpG sites in GOSs. Right panel represents the CpG sites in LOSs. (C) Aggregated distribution of different groups of CTCF-binding sites within a 2-Mb window of the TAD boundary in erythroblasts identified by Hi-C. (D) Pie chart showing the proportion of CTCF peaks overlapping with the anchors of chromatin loops identified by H3K27ac HiChIP using HSPC differentiated erythroblasts from the same donor. (E) Box plot showing the distribution of the number of loops per anchor region for GOSs and CONs in erythroblasts.

Figure 4.

Disrupting dynamic CTCF binding leads to abnormal hematopoiesis. (A) Signal tracks for CTCF and GATA1 ChIP-seq and ATAC-seq at the RBM38 gene locus in HSPCs and erythroblasts. The dynamic site that was genome edited to remove the CTCF motif is highlighted. (B) CTCF ChIP-seq signals in WT HUDEP-2 cells (WT) and HUDEP-2 cells with a 19-bp deletion of the CTCF motif near RBM38 promoter (mut) after 3 days of inducted erythroid maturation. The dynamic CTCF site that harbors the deleted motif is highlighted. (C) Representative flow cytometric plot of Band3 and CD49d expression in gated CD235a⁺ WT HUDEP-2 cells and 2 mut clones after 3 days of induced maturation. (D) Cell pellets of WT HUDEP-2 cells and 2 mut clones after 4 days of induced maturation. (E) Relative expression level of HBB in WT HUDEP-2 cells and 2 mut clones, as compared with GAPDH before differentiation (D0) and 3 days after induced maturation (D3). D0, n = 2; D3, n > 3. *P < .05, paired 1-tailed t test. (F) Representative flow cytometric plot of annexin V and propidium iodide expression in WT HUDEP-2 cells and 2 mut clones after 3 days of induced maturation. (G) Genome-edited CD34⁺ cells were grown in erythroid differentiation medium for 12 days. Representative flow cytometric plots of Band3 and CD49d levels in gated CD235a⁺ cells. (H) Genome-edited CD34⁺ cells were seeded onto methylcellulose. Box plot shows the numbers of burst-forming unit–erythroid colonies after 14 days in culture. n = 6, *P < .01, unpaired 2-tailed t test. NT, nontargeting control gRNA; PI, propidium iodide.

Figure 5.

Regulatory mechanism of dynamic CTCF binding. (A) Relative expression of RBM38 in WT HUDEP-2 cells and 2 mut clones, as compared with HPRT1 before differentiation (D0) and 3 days after induced maturation (D3). D0, n = 2; D3, n > 3. **P < .01, paired 2-tailed t test. (B) ChIP-seq signal tracks of GATA1 and H3K27ac in WT HUDEP-2 cells and 2 mut clones after 3 days of induced maturation. The dynamic CTCF site that harbors the deleted motif is highlighted. (C) Capture Hi-C results with baits targeting the promoter of the RBM38 gene in WT HUDEP-2 cells and a mut clone before differentiation (D0) and 3 days after induced maturation (D3). The targeted CTCF-binding site is highlighted in pink. (D) Chromatin loops identified by H3K27ac HiChIP that are anchored within the targeted dynamic CTCF-binding site in WT HUDEP-2 cells and a mutant clone after 3 days of induced maturation. The top panel shows a 2.8-Mb window, and the bottom panel zooms in on a 265-kb window. The targeted CTCF-binding site is highlighted. (E) Schematic model of the regulatory mechanism of dynamic CTCF-binding site. chr, chromosome.