CTCF, a Candidate Trans-Acting Factor for X-Inactivation Choice

See allHide authors and affiliations

Science  11 Jan 2002:
Vol. 295, Issue 5553, pp. 345-347
DOI: 10.1126/science.1065982


In mammals, X-inactivation silences one of two female X chromosomes. Silencing depends on the noncoding gene, Xist(inactive X-specifictranscript), and is blocked by the antisense gene,Tsix. Deleting the choice/imprinting center inTsix affects X-chromosome selection. Here, we identify the insulator and transcription factor, CTCF, as a candidatetrans-acting factor for X-chromosome selection. The choice/imprinting center contains tandem CTCF binding sites that function in an enhancer-blocking assay. In vitro binding is reduced by CpG methylation and abolished by including non-CpG methylation. We postulate that Tsix and CTCF together establish a regulatable epigenetic switch for X-inactivation.

Dosage compensation ensures equal expression of X-linked genes in XX females and XY males. In mammals, this process results in inactivation of one female X chromosome (XCI) (1) in a random or imprinted manner. In the random form (eutherian), a zygotic counting mechanism initiates dosage compensation and enables a choice mechanism to randomly designate one active (Xa) and one inactive (Xi) X [reviewed in (2)]. In the imprinted form, zygotic counting and choice are superseded by parental imprints that direct exclusive paternal X-silencing (3,4). Imprinted XCI is found in ancestral marsupials (3) but vestiges remain in the extraembryonic tissues of eutherians such as mice (4).

An epigenetic mark for random and imprinted XCI has long been postulated (2). The marks are placed at the X-inactivation center (Xic) (5), which includes thecis-acting noncoding gene, Xist (6,7), and its antisense counterpart, Tsix(8). Xist RNA accumulation along the Xi initiates the silencing step (9, 10), whereas Tsixrepresses silencing by blocking Xist RNA accumulation (11, 12). A cis-acting center for choice and imprinting lies at the 5′ end of Tsix, as its deletion abolishes random choice in epiblast-derived cells to favor inactivation of the mutated X (11, 13) and disrupts maternalXist imprinting in extraembryonic tissues (14,15). Thus, while imprinted XCI is parentally directed and random XCI is zygotically controlled, both work through Tsix to regulate Xist.

To date, only X-linked cis-elements have been identified as XCI regulators. Yet, virtually all models invoketrans-acting factors which interact with the X-linked sites. In one model for imprinted XCI, a maternal-specifictrans-factor confers resistance to XCI (16). In models for random XCI, an autosomally expressed “blocking factor” protects a single X from silencing (2). We have proposed that Tsix is thecis-target of both trans-factors (11,14).

To isolate candidate trans-factors, we now used computational analysis (Fig. 1) to identify mouse-to-human conserved elements within the 2- to 4-kilobase (kb) sequence implicated in choice and imprinting (11, 1315), a region including DXPas34(17). We found that the region is composed almost entirely of 60– to 70–base pair (bp) repeats with striking resemblance to known binding sites for CTCF, a transcription factor with a 60-bp footprint and 11 zinc fingers that work in various combinations to generate a wide range of DNA-binding activities (18). CTCF functions as a boundary element at the globin locus (19), regulates enhancer access to theH19-Igf2 imprinted genes (20–23), and associates with CTG/CAG repeats atDM1 (24). Murine Tsix contains >40 CTCF motifs and the human sequence has >10 (Fig. 1A). Dotplot analysis indicated a contiguous head-to-tail arrangement of highly homologous DXPas34 repeats (25). This clustering is rare, with only three other loci of comparable density (40 sites per 1629 bp) occurring in 40.4 Mb of available sequence (ScanACE, The clustering of nine human elements is not above genome average (test of 933 random 100-kb fragments; random sequence selection program, J. Aach). CTCF function, however, does not require a clustering of sites (20–23).

Figure 1

Tandem CTCF-like binding sites in the Tsiximprinting/choice center. (A) Histogram of conserved human and mouse sites with 0 to 3 mismatches to the CTCF consensus (20,21). Open and shaded bars represent two orientations. (B) Alignment of mouse Tsix, H19,DM1, and chicken β-globin sites. Shading indicates identity with the consensus. (C) Clustering of CTCF motifs. ΔCpG (11, 14), ΔDXPas34 (13), and AA2Δ1.7 (15). Filled triangles, sites with 0 to 3 mismatches. Open triangles, sites in the center with >3 mismatches. Forward sites, gray; reverse sites, black. Tested CTCF sites are indicated by red letters.

To determine if the sites could bind CTCF in vitro, we performed gel retardation analysis of representative sites A, B, C, and D (Fig. 1, B and C). Using in vitro–translated murine CTCF, we observed a protein-DNA complex at all sites that was eliminated by unlabeled self-competitor DNA (Fig. 2A). The complex migrated more rapidly than that formed by H19, possibly due to differential binding of CTCF isoforms (Fig. 2A; SDS-PAGE) or differential DNA bending induced by CTCF (22). Unprogrammed lysates did not shift the probe, indicating that the activity was specific to CTCF. HeLa extracts yielded two bands (Fig. 2B), one similar to that seen with in vitro–synthesized CTCF and one of lower intensity with a mobility similar to that forH19 (this band was not always seen, e.g., Fig. 2D). Preincubation with polyclonal anti-CTCF antibodies blocked complex formation (Fig. 2, A and B). Mutating the 14-bp consensus (20,21) within the 70-bp sites reduced binding (Fig. 2C) and unlabeledH19 DNA effectively competed against Tsix for CTCF binding (Fig. 2D). Thus, CTCF specifically binds Tsixin vitro.

Figure 2

Tsix elements bind CTCF in vitro and in vivo. (A) Gel-shift assay of P32-labeled Tsixoligos and CTCF protein. Reactions were carried out for 30 min at room temperature with 0.5 to 5.0 μl in vitro–synthesized CTCF protein (see SDS-PAGE) and 10 fmol double-stranded DNA probes in 20 mM HEPES (pH 7.5), 50 mM KCl, 5 mM MgCl2, 1 mM dithiothreitol, 0.3 mg/ml BSA, 5% glycerol, 0.5% Triton X-100, and 1 μg poly-dI:dC before resolution in 5% acrylamide, 0.5× TBE gels at 4°C. Cold competitors here and below (comp) were added at 200× molar excess. Supershifts were carried out using normal IgG or COOH-terminal CTCF antibodies (19). Site A, 5′-TGGAGCCTAAACCTGTCTGTCTCTTTACCAGACGCAGGGCAGCCAGAAGGCAGCCATTCACAATCCAGGAAGACAGGAAGGG-3′; site B, GGGGTTGGTTATAAGGCAGGGATTTTAGCGATCTCCCCAGGTCCCTGGCGGCGGCAGGCATTTTAGTGATAGCCCAGGTCCCCG; site C, ATTTTGGCTCCAGGACCCAGCAGACATTTTAGTTATTCCTCCGTTATGCGGCAGGCATTTTAACTATCGGTTCGGGACTACGCAGG; site D, CAGATCCCCAGTGGCAGACATTTTAGTGATAGCCCAGGTCCCCGGTGGCAGGCATTTTAGTGATAGCCCAGGTCCCCGGTGGCA. H19, MS1 (20). Arrowhead, Tsix DNA-protein complex. (B) An activity in HeLa nuclear extract (1 to 2 μg/reaction) also binds Tsix sites. (C) Mutated CTCF sites show reduced binding. Mut, mutated; WT, wild type. MutA, 5′-TGGAGCCTAAACCTGTCTGTCTCTTTACCAGTAATAGAATTCATGTAATATATCCATTCACAATCCAGGAAGACAGGAAGGG-3′; MutB, GGGGTTGGTTATAAGGCAGGGATTTTAGCGATCTCCCCAGGTCTAATAGAATTCATGGCATTTTAGTGATAGCCCAGGTCCCCG; MutC, ATTTTGGCTCCAGGACCCAGCAGACATTTTAGTTATTCCTTAATAGAATTCATGGCATTTTAACTATCGGTTCGGGACTACGCAGG; MutD, CAGATCCCCAGTGGCAGACATTTTAGTGATAGCCCAGTAATAGAATTCATGGCATTTTAGTGATAGCCCA- GGTCCCCGGTGGCA. (D) Unlabeled H19 sites compete against Tsix sites for CTCF. (E) CTCF bindsTsix in vivo (female fibroblasts) using ChIP analysis as described (28). Immunoprecipitations were performed overnight at 4°C with anti-CTCF antibodies (Upstate) or normal IgG. Primers pairs GTGTGTCATAGCTCAAGAGG, GGAGCCTAAACCTGTCTGTC (site A); AATGCTTGCCAGCTATGCGG, TAACCACCTGTAAGGGACAG (site C).

To test if CTCF binds Tsix in vivo, we carried out chromatin immunoprecipitation (ChIP) using anti-CTCF antibodies followed byTsix-specific polymerase chain reaction in female mouse fibroblasts. Because the CTCF sites are tandemly repetitive, only sites A and C could be tested. Like the H19 site [MS2 (20)], both sites were specifically coimmunoprecipitated with CTCF (Fig. 2E). In contrast, random loci on mouse chromosome 12 (MT498; and in Xist (cDNA bp 13,177 to 13,428) did not coimmunoprecipitate (MT498 shown). Thus, CTCF complexes withTsix DNA in vivo.

At some loci, CTCF sites act as chromatin insulators (19–21). In the established assay, insertion of these sites between the globin LCR and a neomycin (neo)–resistance reporter results in fewer neo-resistant K562 colonies (26). When a 4.3-kb Bam HI–Bam HI fragment containing all the Tsix sites was tested, we observed a dramatic reduction in colony number which was stronger in the R-orientation (Fig. 3, A and B). A 1.1-kb Pml–Age I fragment containing only sites B, D, and DXPas34also reduced colony number more strongly in the R-orientation (Student's t test, P < 0.0001; ANOVA,P < 0.0001). This modest orientation-dependent effect is consistent with published reports (19–23). The greater activity in the Bam HI–Bam HI fragment might be attributable to additional CTCF sites outside of DXPas34 or to possible unmapped Tsix promoter activity in the Bam HI–Bam HI fragment that would be antisense to Neo. Individual sites A, B, and C each exhibited fewer colonies relative to mutated site B (Fig. 3C; t test, P < 0.05; ANOVA, P < 0.05). Thus, Tsix can block enhancer-promoter interaction and insulating activity correlates with CTCF binding in vitro.

Figure 3

The 5′ end of Tsix contains enhancer-blocking activity. (A) The enhancer-blocking assay (26) for Tsix sites in K562 cells. Sites A, B, and C are indicated by black boxes. Fragments in both forward (F) and reverse (R) orientations (“F,” Tsix and Neotranscription in same direction) were inserted between the β-globin LCR and a neomycin-resistance reporter (Neo). Flanking globin insulators (Ins) protects against position effects (26). + control, globin insulators (pJC13-1) (26). (B) Results of enhancer-blocking assay. We transfected 1.5 pmol each of test plasmid and pTK-Hygromycin (transfection efficiency control). Neo-resistant colonies were counted 2 to 3 weeks after transfection and normalized to hygromycin-resistant colonies. Three to four experiments were averaged. P-values, unpaired one-tailed Student's t test in pairwise comparisons against the no-insulator control. (C) Enhancer-blocking activities for sites A, B, C, and mutated B. Constructs contained 1.5 kb of spacer to maintain equal distance.P-values, unpaired one-tailed Student's t test in pairwise comparisons against mutated B.

Since CTCF responds to CpG methylation at some loci (20–22), we tested methylation-sensitivity atTsix using gel retardation analysis. Unexpectedly, CTCF binding was only partially blocked by CpG methylation but was abolished when non-CpG methylation was included (Fig. 4). This contrasted with total inhibition at H19 by CpG methylation alone. Relevant to this, H19 sites contain three to four CpG's (20, 21), whereas many Tsix sites contain zero or one CpG in the consensus despite being strongly C-rich (Fig. 1B). These findings raised the possibility that non-CpG- together with CpG-methylation might regulate CTCF binding to Tsix. Notably, recent bisulfite sequencing has not uncovered differential CpG methylation in DXPas34 (27). In light of our findings, the methylation status of non-CpG sites in the CTCF array will be critical in future work.

Figure 4

CTCF binding is sensitive to DNA methylation in vitro. Gel-retardation analysis using Tsixprobes which were unmethylated (–), methylated at CpGs only (+), or methylated at all C-nucleotides (++). Cold competitor (CH3-comp) at 200× was methylated at all Cs. CpG methylation, achieved by SssI methylase and confirmed by insensitivity to HpaII or AciI digestion. Non-CpG methylation, achieved by direct synthesis. Arrow, Tsix DNA-protein complex.

In summary, we have identified CTCF as a binding protein for the cis-acting choice/imprinting center inTsix. We propose that CTCF and Tsix coordinately establish the epigenetic switch for Xist (Fig. 5). Because knocking out the CTCF array (choice/imprinting center) results in inactivation of the mutated X (11, 1315), we favor a model in which binding of CTCF designates the future Xa. In this model, the zygotic blocking factor and the maternal protective factor work through CTCF to promote Tsix expression on the Xa. CTCF could directly stimulate Tsix transcription or do so by default through blocking Xist's access to unidentified shared enhancers (20–23).Tsix transcription would in turn block Xist RNA accumulation (12). On the Xi, CTCF binding is excluded fromTsix, possibly by methylation (CH3) of the CTCF array, thereby allowing the up-regulation of Xist. In the future, finer mutational analysis and the identification of differentially methylated regions will be required to test details of the model. Because CTCF is ubiquitous, developmental specificity must be achieved combinatorially with stage- and locus-specific factors. Identification of these protein-protein interactions will be instrumental in defining the long-postulated zygotic and maternal factors.

Figure 5

Model of a regulatable epigenetic switch created by CTCF and Tsix.

  • * To whom correspondence should be addressed. E-mail: lee{at}


View Abstract

Navigate This Article