CTCF: master weaver of the genome

Abstract

CTCF is a highly conserved zinc finger protein implicated in diverse regulatory functions, including transcriptional activation/repression, insulation, imprinting, and X chromosome inactivation. Here we re-evaluate data supporting these roles in the context of mechanistic insights provided by recent genome-wide studies and highlight evidence for CTCF-mediated intra- and interchromosomal contacts at several developmentally regulated genomic loci. These analyses support a primary role for CTCF in the global organization of chromatin architecture and suggest that CTCF may be a heritable component of an epigenetic system regulating the interplay between DNA methylation, higher-order chromatin structure, and lineage-specific gene expression.

Figure 1. Allele-specific chromatin contacts at an imprinted locus

(a,b) Linear depiction of the mouse H19/Igf2 locus. The maternally expressed non-coding H19 gene is located approximately 90kb downstream from the gene encoding Insulin-like growth-factor 2 (Igf2) that is expressed exclusively from the paternal allele. The imprinting control region (ICR) ~2kb upstream of H19 contains four CTCF binding sites and is essential for regulation of the entire locus. Differentially methylated regions (DMRs), such as DMR1 upstream of Igf2 promoters (P1, 2, 3) and DMR2 within Igf2 exon 6, act in concert to regulate reciprocal, allele-specific expression patterns from a shared set of downstream enhancers at 8kb (Ee: endodermal tissue enhancer) and 25kb (Em: mesodermal tissue enhancer) downstream of the H19 gene. –CH₃, DNA methylation. Green ovals, enhancers. (c,d) Schematic 3-D models illustrating allele-specific patterns of CTCF binding, DNA methylation, and chromatin looping. Although loops are illustrated here via CTCF multimerization, it is not yet clear if these long-range interactions can be attributed to CTCF binding to all sites and subsequent dimerization or if CTCF detection via chromatin immunoprecipitation is due to indirect interactions via looping.

Figure 2. Cell type-specific interchromosomal interactions at a developmentally regulated locus

(a) Schematic representation of the mouse β-globin locus. Four globin genes (solid green arrows) are embedded within a larger olfactory receptor genes cluster (open arrows). Developmentally regulated globin expression (εy and βh1 in primitive erythroid cells; β-major and β-minor in definitive erythroid cells) is regulated in part by a series of cis-acting regulatory elements surrounding the locus. An upstream locus control region (LCR) containing 6 DNase I-hypersensitive sites (HSs) is required for high-level transcription. Three CTCF binding sites have been identified upstream (5′HS85, 5′62/60, and 5′HS5) and one 20kb downstream (3′ HS1) of the gene. Black arrows, HSs. (b–d) Diagrams illustrating lineage-specific CTCF binding patterns, 3C-based intra-chromosomal interactions, and globin gene expression profiles in erythroid progenitors (b), definitive erythroid cells (c), and non-erythroid brain cells (d). Although loops are illustrated here via CTCF multimerization, it is not yet clear if CTCF binds directly to DNA at each site or if detection by chromatin immunoprecipitation is due to indirect interactions via looping.

Figure 3. An inducible chromatin loop

Model for cytokine-induced loop formation at the human major histocompatibility complex class II (MHC-II) locus. CTCF binds to the XL9 enhancer element between two co-regulated genes, HLA-DRB1 and HLA-DQA1, driven by divergent promoters. Transcription factors RFX, CREB, and NF-Y bind to regulatory sequences within the proximal promoter of MHCII genes when they are transcriptionally inactive. Interferon γ (IFNγ) treatment induces transcription in non-expressing cell types by upregulating the non-DNA binding co-activator CIITA, which subsequently forms a heteromultimer with RFX-CREB-NF-Y-bound promoters and the CTCF-bound enhancer in parallel with HLA-DRB1 and HLA-DQA1 gene activation.

Figure 4. Potential classes of CTCF-mediated contacts

Experimental evidence for certain sub-classes of CTCF loops exists (a–f), whereas others can be hypothesized based on genome-wide distribution patterns (g–l). (a) Anchoring via direct attachment to subnuclear structures such as the nucleolus and/or nuclear matrix; (b) Transcriptional regulation via contact between intergenic locus control region and promoter-proximal regulatory element; (c) Active chromatin hub around multiple co-regulated genes via contact between multiple distal CTCF binding sites; (d) Mono-allelic gene expression via allele-specific contacts between multiple imprinted regulatory elements; (e) X-chromosome inactivation or mono-allelic gene expression via interchromosomal contacts between regulatory elements in trans; (f) Global nuclear organization via demarcation of lamina-associated domains (LADs; (g) RNA polymerase II pausing and/or termination via intragenic contacts between introns and exons; (h) RNA processing or transcriptional re-initiation via a single gene 5′-3′ loop; (i) Alternative promoter selection via contact between two insulator elements demarcating transitions in chromatin structure; (j) Boundary/barrier loops to demarcate independently regulated chromatin domains containing a co-regulated gene-dense cluster via contact between two insulator elements; (k) Enhancer blocking loops that topologically separate inappropriate enhancer-promoter interactions via contact between two insulator elements; (l) Interchromosomal translocations via contacts between two regulatory elements in trans.

Figure 5. Possible mechanisms for developmentally-regulated loop formation

Previously identified mechanisms for mammalian insulator regulation are broadly categorized according to four general organizing principles, including those that alter CTCF occupancy and those that may facilitate the de novo formation, maintenance, or stabilization of CTCF loops without altering CTCF binding to its cognate consensus site. Models that generally illustrate specific regulatory mechanisms within each category are provided. -CH₃, DNA methylation. Purple squares, CTCF consensus sites. Blue squares, thyroid hormone response element. Green ovals, enhancers. Blue ovals, CTCF. Red rings, cohesin. Pink dots, PARylation mark. Red ovals, Factor X representing any CTCF binding partner. Orange ovals, thyroid hormone receptor.

Figure 6. Epigenetic inheritance of CTCF-mediated chromatin loops

Data are consistent with a model centered on the interplay between DNA methylation, poly(ADP-ribosyl) ation, and higher-order chromatin loops. Three classes of CTCF consensus sites (non-CpG, unmethylated CpG-containing, and methylated CpG-containing) display differential patterns of CTCF binding and ‘heritability’ during cell division. In the case of general structural loops mediated by CTCF-bound non-CpG consensus sites (top panel), the majority would lose CTCF during mitosis. CTCF re-binding would generally be the default state and structural contacts across the cell would be re-established after every cell division. Conversely, a smaller subset of binding sites with unmethylated CpG-consensus variants or CpG dinucleotides in the surrounding 50–60 bp insulator element (middle, bottom panels) may retain CTCF binding through the cell cycle to protect against de novo methylation. Higher-order chromatin structures mediated by these particular elements would retain the potential for heritability through mitosis via cell type- and locus-specific post-translational modifications such as PAR or recruitment of protein binding partners such as cohesin. Among the putative classes of unmethylated CpG-consensus-mediated contacts, constitutive structures may be observed around key allele-specific imprinted genes (bottom panel), whereas acquisition of DNA methylation in response to developmental and/or environmental cues would result in abrogated CTCF binding and, consequently, potentially permanent disruption of contacts at a smaller subclass of loops encompassing developmentally-regulated loci (middle panel). Although only CpG-containing consensus sites retain the potential for heritable loop structures in this model, we cannot rule out the possibility of PAR-stabilized contacts between CpG- and non-CpG consensus sites or methylation-independent structures between non-CpG sites stabilized by PAR, cohesin, and/or additional unknown mechanism(s). PAR is shown only stabilizing loops between CpG consensus sites, but this has not yet been proven. Purple squares, non-CpG consensus sites. Red-hashed purple -squares, CpG-containing consensus sites. –CH₃, DNA methylation. Red ovals, PARylation mark.