CRISPR Inversion of CTCF Sites Alters Genome Topology and Enhancer/Promoter Function

Abstract

CTCF and the associated cohesin complex play a central role in insulator function and higher-order chromatin organization of mammalian genomes. Recent studies identified a correlation between the orientation of CTCF-binding sites (CBSs) and chromatin loops. To test the functional significance of this observation, we combined CRISPR/Cas9-based genomic-DNA-fragment editing with chromosome-conformation-capture experiments to show that the location and relative orientations of CBSs determine the specificity of long-range chromatin looping in mammalian genomes, using protocadherin (Pcdh) and β-globin as model genes. Inversion of CBS elements within the Pcdh enhancer reconfigures the topology of chromatin loops between the distal enhancer and target promoters and alters gene-expression patterns. Thus, although enhancers can function in an orientation-independent manner in reporter assays, in the native chromosome context, the orientation of at least some enhancers carrying CBSs can determine both the architecture of topological chromatin domains and enhancer/promoter specificity. These findings reveal how 3D chromosome architecture can be encoded by linear genome sequences.

Figure 1. Two Distinct CTCF/cohesin-mediated Chromatin Domains in the Pcdh Locus

(A) Diagram showing the Pcdh α and βγ CCDs in the three mouse Pcdh gene clusters. The variable (Var) and constant (Con) exons are also indicated. The CBSs and their orientations are indicated as arrowheads. Different types of Pcdh CBSs are represented by differently colored arrowheads. The dark and light blue CBSs represents the CSE and eCBS, respectively, for each of the twelve “alternate promoters” (α1-α12) of the α cluster. The 21 tandem green arrowheads represent the CSE for each member of the β cluster (except β1). The yellow and red arrowheads represent CSEs for γa and γb, respectively. The two gray arrowheads represent the C-type CSEs (αc1 and γc3). The two CBS sites (a, b) downstream of the α cluster and the eight CBS sites (a–h) downstream of the γ cluster are indicated in black arrowheads. The DNaseI hypersensitive sites (HS) in the α and βγ regulatory regions are also shown. (B–G) Relative distributions of the 4C reads per million (RPM) obtained in human SK-N-SH cells (B), mouse N2A cells (C) and brain tissues (D) using the HS5-1 enhancer as an anchor. 4C interaction profiles in human SK-N-SH cells (E), mouse N2A cells (F) and brain tissues (G) with the regulatory region downstream of the γ cluster as an anchor are also shown. The significance of interactions (P value) is shown under the reads density for each panel. (H) Showing the forward orientation of CBS sites in Pcdh promoters and reverse orientation of CBS sites in Pcdh enhancers. See also Figure S1.

Figure 2. Inversion of the Pcdh HS5-1 Enhancer with CBSs Switches DNA Looping Direction and Alters Gene Expression

(A) Long-range chromatin-looping interaction profiles of the HS5-1 anchor in wild-type control (Ctr) or in a HS5-1 inversion (Inv) cell line generated from subcloned HEC-1B cells by CRISPR engineering. The log2 ratio between inversion and control is also shown. (B) The relative crosslinking frequency measured by quantitative 3C assays in the control or inversion cell lines with HS5-1 as an anchor (HS5-1 is within the same 3C restriction fragment in the genomes of both Ctr and Inv cell lines). Data are means ± SEM (n=4). *P < 0.05 and **P < 0.01. (C) Control experiments showing functional CTCF/cohesin binding after inversion. **P < 0.01. (D) RNA-Seq experiments showing expression reduction of the α, β, and γ clusters (except γc3) after inversion. *P < 0.05, **P < 0.01, and ***P < 0.001. See also Figure S2.

Figure 3. CTCF Recognition of the HS5-1b Site in Only One Direction

(A) Showing the HS5-1b CBS sequence (double-stranded) of the reverse orientation (indicated above by a red arrow) with the palindromic core highlighted. The double-stranded reverse complement HS5-1b CBS sequences (along with three probes with core sequences mutated) are also shown below the CBS consensus. The nucleotides that match to the CBS consensus are indicated by vertical lines. Note that mut2 and mut3 are exactly the same for the palindrome core sequence. (B) The wild-type (WT) or mutant (Mut) sequences of HS5-1b probes (shown in the reverse complement). (C) Gel-shift assays of the wild-type HS5-1b probe using a set of recombinant CTCF proteins with sequentially-deleted zinc-finger domains. (D–F) Gel-shift assays using recombinant CTCF proteins with probes of Mut1-3 (D), Mut4 (E), Mut5 and Mut6 (F). See also Figure S3.

Figure 4. The Role of CBS Location and Orientation in CTCF-mediated Genome-wide DNA Looping

(A) Diagram of CTCF-mediated long-range chromatin-looping interactions between CBS pairs in the forward-reverse orientations. The color charts represent 19,532 interactions of CBS pairs in K562 cells. The number and percentage of CBS pairs in the forward-reverse (FR), forward-forward (FF), reverse-reverse (RR), and reverse-forward (RF) orientations are shown. (B) The percentage of CBS pairs in the forward-reverse orientations increases from 67.5% to 90.7% as the chromatin-looping strength is enhanced. (C) Schematic of the two topological domains in the HoxD locus. The orientations of CBSs are indicated by arrowheads. CTCF/cohesin-mediated looping interactions and the two resulting topological domains are also shown. (D) Cumulative patterns of CBS orientations of topological domains in the human genome. (E) Distribution of genome-wide orientation configurations of CBS pairs located in the boundaries between two neighboring domains in the human K562 genome. Note that the vast majority (90.0%) of boundary CBS pairs between two neighboring domains are in the reverse-forward orientation. See also Figure S4 and Tables S1–S6.

Figure 5. CRISPR Inversion of CBS13-15 in the Human β-globin Cluster Confirms the CTCF/cohesin-mediated Directional Looping Mechanism

(A) Diagram of the human β-globin region. Predicted looping interactions and topological domains are shown, based on CTCF occupancy in HEK293 cells. (B) The predicted interactions (left) and the altered looping directions (right) in the three subcloned CRISPR cell lines with inversion of CBS13-15 (E28, E79 and F6) are confirmed by 4C with CBS13-15 as an anchor. The looping interactions of three mock controls are also shown. The average log2 ratios of interactions between inversions and controls are also indicated. **P < 0.01. See also Figures S2 and S5.

Figure 6. A Model of CTCF/cohesin-mediated topological 3D genome folding and gene regulation

In mammalian genomes, CTCF directionally recognizes CBSs by distinct combinations of its 11 ZF domains, and asymmetrically recruits the cohesin complex to CBS sites through its C-terminal domain (Xiao et al., 2011). CTCF together with the cohesin complex establishes specific long-range chromatin-looping interactions between CBS pairs in the forward-reverse orientations to form distinct topological domains (domains 1 and 2, see the upper right inset). The weak interactions between the two CBSs in the same forward orientation in topological domain1 may be the consequence of their looping interactions with a common CBS in the reverse orientation (Guo et al., 2012). The two CBSs in the reverse-forward orientations form a boundary insulator element between the two neighboring domains 1 and 2, blocking remote enhancers located within one domain from aberrantly activating promoters located in the neighboring domain, and thus “indirectly” ensuring proper activation of cognate promoters by distal enhancers within the same topological domain (see Inset).