Targeted Degradation of CTCF Decouples Local Insulation of Chromosome Domains from Genomic Compartmentalization

Abstract

The molecular mechanisms underlying folding of mammalian chromosomes remain poorly understood. The transcription factor CTCF is a candidate regulator of chromosomal structure. Using the auxin-inducible degron system in mouse embryonic stem cells, we show that CTCF is absolutely and dose-dependently required for looping between CTCF target sites and insulation of topologically associating domains (TADs). Restoring CTCF reinstates proper architecture on altered chromosomes, indicating a powerful instructive function for CTCF in chromatin folding. CTCF remains essential for TAD organization in non-dividing cells. Surprisingly, active and inactive genome compartments remain properly segregated upon CTCF depletion, revealing that compartmentalization of mammalian chromosomes emerges independently of proper insulation of TADs. Furthermore, our data support that CTCF mediates transcriptional insulator function through enhancer blocking but not as a direct barrier to heterochromatin spreading. Beyond defining the functions of CTCF in chromosome folding, these results provide new fundamental insights into the rules governing mammalian genome organization.

Figure 1. Acute and reversible depletion of CTCF with the auxin-inducible degron system in mESCs

(A) Deploying the AID system at Ctcf in mESCs. (B) Western-blot showing reversible loss of CTCF in CTCF-AID cells (C) Immunofluorescence staining (D) Long-term survival (12 days) is only compromised in CTCF-AID cells treated with auxin after introduction of the Tir1 transgene (E) Time-course flow-cytometry (F) Brightfield images of mESC colonies after auxin treatment indicating cells tolerate a 2-day depletion with no adverse effects on viability. See Figure S1

Figure 2. CTCF is required for accumulating loops between CTCF/Cohesin binding sites

(A-B) Snapshots of 1.3 Mb of Hi-C data at 20kb resolution CTCF-AID mESCs aligned with CTCF ChIP-seq and the Smc1a HiChIP loops identified by Mumbach et al. 2016. Normalized Hi-C counts are multiplied by 10⁵ (C) Genome-wide aggregation of normalized Hi-C signal anchored at Smc1a HiChip loops separated by 280 to 380kb (1196 loops). Similar results were obtained for smaller and larger loops. See Figure S2

Figure 3. CTCF instructs insulation of TADs

(A) Snapshots of 6Mb of Hi-C data at 20kb resolution from CTCF-AID mESCs aligned with CTCF ChIP-seq. Normalized Hi-C counts are multiplied by 10⁵ (B) Left: CTCF depletion dampens insulation at TAD boundaries (higher insulation score over 100kb surrounding boundaries). Right: residual boundaries detected after CTCF depletion (and without persistent CTCF peaks, ∼20% of total boundaries) maintain insulation independently of CTCF. Note that lower score denotes higher insulation potential (C) Snapshot of Hi-C data at the Tbx5 locus and differential contact map showing more inter-TAD (red) and fewer intra-TAD (blue) Hi-C signal after CTCF depletion (D) 3D distance measurement from DNA FISH highlighting that CTCF depletion triggers inter-TAD compaction but does not affect intra-TAD packaging at the cytological level (E-F) same as C-D at the Prdm14 locus (n=90-100 alleles, Kolmogorov-Smirnov test). See Figure S3

Figure 4. Proper insulation of TADs is not required for higher-order segregation of A and B compartments

(A) Hi-C contact maps at 100kb resolution across entire chromosome 2. Bar denotes segments called as A (green) or B (red) compartment using 20kb-cis Eigenvector 1. Normalized Hi-C counts are multiplied by 10⁵ (B) Distributions of cis Eigenvector 1 values across entire chromosome 2 are remarkably stable to depletion of CTCF (C) cis Eigenvector 1 values are not affected genome-wide by CTCF depletion (D) Overall scaling of Hi-C contact frequency as a function of genomic distance is not affect by the loss of CTCF, highlighting that CTCF does not affect general chromatin compaction. See Figure S4

Figure 5. CTCF remains essential for insulation of TADs in resting cells, and acts dose-dependently

(A) mESCs can be converted into cycling NPCs and induced to exit cell cycle by terminal differentiation into astrocytes (B-D) Extracts of restriction-fragment resolution interpolated 5C heatmaps at the Xic. LaminB1 DamID from(Peric-Hupkes et al., 2010). Color dots denote boundaries identified before CTCF depletion (E-G) log2 ratio of 100kb insulation scores from depleted versus untreated cells at boundaries identified before depletion. Plots include boundaries probed beyond the region depicted in the heatmaps (H) Titration of auxin leaves cells with intermediate CTCF levels. Percentages are relative to untreated CTCF-AID cells, where CTCF levels are 2-3 fold lower than parental untagged mESCs (I) CTCF-dependent boundaries loose insulation as a function of leftover CTCF levels (J) 5C heatmaps used to calculate insulation scores. See Figure S5.

Figure 6. CTCF and transcriptional regulation

(A) RNA-seq. fold change compared to untreated cells for genes differentially expressed at one or more time points. Wash denotes 2-day washoff after a 2-day treatment (B) RNA-seq alignment with ChIP-exo (from untreated cells) for each time-point (C) The CTCF site in the promoters of immediately down-regulated genes tends to be ∼60bp upstream of the TSS in direct orientation with transcription, and demarcates the beginning of the nucleosome-depleted region as previously measured by MNAse-seq (Teif et al., 2012) (D) Immediately up-regulated genes tend to lie at shorter genomic distance to neighboring enhancers than down-or non-regulated genes. Trend is rapidly lost over time (E) Enhancer-promoter pairs are more likely to be normally interrupted by a TAD boundary for genes that become up-regulated upon CTCF depletion. See Figure S6.

Figure 7. CTCF does not constrain H3K27me3 spreading in mESCs and summary model

(A) A subset of CTCF binding sites mark transitions in H3K27me3 patterns (B-C) CTCF depletion does not trigger H3K27me3 spreading beyond the formerly bound CTCF site itself (center) (D) Our observations are consistent with TAD formation by loop extrusion, and establish CTCF as the major factor defining domain boundaries genome-wide (E) Statistical average cartoon representation of TAD disruption and compartment preservation upon loss of CTCF See Figure S7.