CTCF binding: features and consequences

CTCF binds with a high-affinity to a nonpalindromic canonical motif in its binding site with a sequence consensus referred to as M1[269, 296-299] (Figure 1.7a). Studies have shown that the central zinc fingers, 4-7/8, are needed for this interaction[300].

This 20 bp core motif is common to nearly all known CTCF binding sites as identified by various immunoprecipitation methods, and the involvement of nonspecific zinc fingers, other than the ones mentioned previously, with the surrounding DNA sequence helps stabilize the binding[297]. A second 10-bp motif, termed M2, is found upstream of the canonical M1 separated by a DNA spacer[269, 297, 301], where it interacts with the 9-11 zinc fingers[302]. Findings suggest that the M2 motif is in conjunction with the M1 in 15%-25% of all CTCF binding sites, whereby CTCF binds with high affinity depending on the spacer between the motifs[303].

The presence of a CpG in the canonical motif’s consensus sequence lends support to the idea that methylation of cytosine residues at carbon 5 of the base to form 5-methylcytosine (5mC) in CGI-harbouring CTCF binding sites may underlie CTCF selectivity in different cellular contexts[304]. Studies support a model where DNA methylation is a common regulatory measure to control CTCF occupancy at many loci, such as CDKN2A, B-cell CLL/lymphoma 6 (BCL6) and brain-derived neurotrophic factor (BDNF)[305-307]. Comparison of DNA methylation patterns in 19 human cell lines with mapped CTCF occupancy showed that 41% of tissue-specific CTCF binding sites are linked to differential DNA methylation[293]. On the other hand, 67% of those sites that were linked with variable DNA methylation, the presence of 5mC correlated with a corresponding downregulation of cell-type-specific CTCF binding. CTCF also forms a complex with poly(ADP-ribose) polymerase 1 (PARP1) and DNA (cytosine-5)-methyltransferase 1 (DNMT1), activating PARP1, which in turn inactivates DNMT1 by poly(ADP-ribosyl)ation, maintaining methyl-free CGIs in the genome[308, 309]. Furthermore, other studies in mammals have observed that CTCF can also cooperate with RNAs to stabilize its interactions with other protein complexes, such as the DEAD-box RNA helicase and p68 and their associated ncRNA[310]. These, along with more recent findings that demonstrate that CTCF binds to the Jpx RNA, indicate that ncRNA are involved in the stabilising of interactions mediated by CTCF and its protein partners[311]. Saldana-Meyer et al.

reported at least 17,000 genomic RNAs that interact with CTCF[287].

One of the most interesting findings of recent years is that a pair of CTCF binding sites will only engage to fold chromatin, forming long-range loop interactions if they are in a convergent, linear orientations, producing asymmetrical insulator

orientation of the looping and disrupt the packaging of the underlying chromosome segmentation pattern into an insulated TAD, proving that the proper arrangement of binding sites is crucial for the correct functioning of CTCF[244, 279, 282, 285]. In addition, the deletion of a TAD boundary in the vicinity of the Xist locus on chromosome X results in ectopic loop interactions general mis-regulation of gene expression[78]. Analysis of the Six homeodomain locus in zebrafish unveiled CTCF binding sites in oriented convergently with TADs at TAD boundaries, and attempted deletion of any of these boundaries results in erratic interdomain enhancer-promoter interactions[313].

Other regulatory factors may also contribute to augmenting or modulating CTCF function[314]. For example, Smad proteins interact with CTCF at the Igf2/H19 imprinted control region[315]. Similarly, at the Igf2/H19 locus, p68 helps, along with the long noncoding RNA SRA, to stabilize cohesin binding and create an effective insulator. DNA-bound CTCF/cohesin complexes recruit the core promoter factor TFIII to helps stabilize CTCF binding at specific promoter-proximal regions at many loci in ESC[316, 317]. CTCF also associates with PARP1 to establish inter-chromosomal contacts during the circadian cycle[318].

Homozygous knockout of CTCF is embryonic-lethal[319-321], and partial deletion of CTCF leads to an altered gene expression pattern, yet with more limited phenotypic impact, increasing radiation sensitivity, defective DNA-repair mechanism, and cell cycle arrest[80, 322]. Full removal of CTCF results in total loss of nearly all loop interactions in a highly dose-dependent manner[247, 323, 324]. Conditional Ctcf knockouts in a tissue-specific context, such as in oocytes, lymphocytes, neurons, and cardiomyocytes, lead to organ failures[276, 325-327]. Acute depletion of CTCF in vitro by both RNAi and transient auxin-mediated in mouse ESC yields full removal of CTCF from the nucleus, disruption of loop structures and TADs, yet high-order chromosome compartmentalization is maintained[269, 323, 328]. Although Ctcf hemizygous mice undergo normal development, they exhibit an increased predisposition to tumours[329]. Even though halving of CTCF protein concentration is physiologically tolerated, the process reduces the overall fitness of the organism.

CTCF has also been shown to be a haploinsufficient tumour suppressor gene in human cancers[329-331]. A recent study observed that Ctcf hemizygous cells show modest but consistent changes in almost 1000 CTCF binding sites that are of lower affinity and weaker evolutionary conservation across the murine lineage. This coincided with dysregulation of several hundred genes' expression, which are ontologically enriched in cancer-related pathways. Chromatin configuration is, however, unaffected apart from disruption to some loop domains[332]. Mutations of CTCF motifs lead to oncogene dysregulation in some cancers[294], and defective limb development in humans and

mice[244]. Unlike germline variants, somatic missense and nonsense mutations of CTCF are abundant in human tumours [333, 334]. Hyper-methylation of the GC-rich CTCF binding motif was observed to decrease CTCF occupancy in glioma, and constitutive CTCF–CTCF binding site interactions are reportedly deleted in T-cell acute lymphoblastic leukaemia, resulting in oncogenic upregulation[245, 323].

Figure 1.6: CTCF regulates 3D chromatin architecture

(a) An interaction heat map of a ~2.5-Mb chromosome segment. TADs, their

at TAD borders contribute to its formation. CTCF may act an as an enhancer blocker (left). On the other hand, CTCF bound inside TADs may act as an enhancer facilitator through looping the DNA with the help of cohesin. Blue boxes denote gene promoters, and black boxes denote genes. (c) The TAD borders in mammals are enriched for housekeeping and tRNA genes, SINEs and CTCF-binding sites. Figure adapted from Ong and Corces [80].

Intriguingly, Satou et al. recently found that CTCF binds to a motif in the Human T lymphotropic virus type 1 (HTLV-1; the human T-cell leukemia virus), when HTLV-1 is inserted into the host cell genome[335]. It is hypothesised that CTCF binding to the provirus can promote abnormal chromatin looping by dimerizing with CTCF in the surrounding host genome. The presence of a single CTCF-dependent chromatin loop in vitro T cell line has since been demonstrated[335, 336].