Report

CTCF Mediates Interchromosomal Colocalization Between Igf2/H19 and Wsb1/Nf1

See allHide authors and affiliations

Science  14 Apr 2006:
Vol. 312, Issue 5771, pp. 269-272
DOI: 10.1126/science.1123191

Abstract

Gene transcription may be regulated by remote enhancer or insulator regions through chromosome looping. Using a modification of chromosome conformation capture (3C) and fluorescence in situ hybridization, we found that one allele of the insulin-like growth factor 2 (Igf2)/H19 imprinting control region (ICR) on chromosome 7 colocalized with one allele of Wsb1/Nf1 on chromosome 11. Omission of CCCTC-binding factor (CTCF) or deletion of the maternal ICR abrogated this association and altered Wsb1/Nf1 gene expression. These findings demonstrate that CTCF mediates an interchromosomal association, perhaps by directing distant DNA segments to a common transcription factory, and the data provide a model for long-range allele-specific associations between gene regions on different chromosomes that suggest a framework for DNA recombination and RNA trans-splicing.

Igf2 and H19 are coordinately regulated adjacent imprinted genes located ∼80 kb apart on mouse chromosome 7 (1). An ICR located between the genes contains four binding regions for CTCF (25), a zinc finger–binding protein that binds to a variety of DNA sequences and serves as an insulator and regulator of gene transcription (68). Using 3C (9), Reik and colleagues demonstrated that on the paternal chromosome, differentially methylated region DMR2 loops out to interact with the methylated ICR, pushing the Igf2 promoter into contact with the H19 enhancer, which results in Igf 2 transcription. On the maternal chromosome, DMR1 interacts with the unmethylated ICR, partitioning the Igf2 promoter into a silent loop, inhibiting Igf 2, and promoting H19 transcription (1). To determine whether the ICR has other long-range associations, we developed an associated chromosome trap (ACT) assay to identify previously unknown, remote interacting sequences (fig. S1).

We identified three bands corresponding to unique DNA sequences that appear to interact with the ICR (Fig. 1A). One sequence was identified as Igf 2 DMR1 on chromosome 7 (1012). A DNA fragment, termed IAS1 (ICR-associated site), corresponded to an intergenic sequence on chromosome 11, located between the Wsb1 and Nf1 genes (Fig. 1B; fig. S2), and DNA fragment IAS2 was localized to a gene-poor region on chromosome 6 (Fig. 1B).

Fig. 1.

Associated chromosome trap (ACT) assay searching for DNA segments interacting with Igf2/H19 ICR in mouse bone marrow fibroblast cell line, BMM3-4. (A). Nested polymerase chain reaction (PCR) products were run on 5% polyacrylamide electrophoresis gels and analyzed with a PhosphoImager. Three bands appear on the gel after the nested PCR reaction; one is an amplified fragment between ICR and DMR1; the other two bands are IAS1 and IAS2. (B) IAS1 and IAS2 sequences (linker sequence not included). Upper case letters, DNA fragment between Bgl II (agatct) and Msp I (CCGG) sites ligated with ICR; lower case, ICR partial sequence.

A chromatin immunoprecipitation (ChIP) assay using a CTCF-specific antibody identified two regions in the ICR, corresponding to CTCF binding sites numbers 1 and 3 (Fig. 2A; figs. S2 and S3), as well as a region around IAS1, but no interaction with IAS2 was found. CTCF binds to the maternal ICR allele only, which leads to looping and interactions with DMR1. The interaction of CTCF with IAS1 was restricted to the paternal chromosome (Fig. 2B; figs. S2B, S4, and S5).

Fig. 2.

ChIP assay with mouse CTCF-specific antibody in BMM3-4, FSKN2, and SF1-G cells. (A) ChIP PCR results. Boxes in broken lines represent IAS1 regions identified by CTCF-specific antibody in ChIP. ICR of Igf2/H19; IAS1 on chromosome 11; IAS2 on chromosome 6; marker, 100–base pair DNA marker; one primer pair, 2930/2941 (see table S1 for primer sequences) encompassing Igf2 promoter 2 region (Pr2) was used in ChIP PCR as a negative control. (B) Allele-specific binding of ICR (left) and IAS1 (right) with CTCF. PCR products were subjected to digestion by allele-specific restriction enzymes BsmA I for ICR and Mse I for IAS1. Only maternal ICR and paternal IAS1 showed specific binding. Genomic DNA served as a control.

These allele-specific interactions suggest colocalization or a close juxtapositioning of ICR on chromosome 7 with IAS1 on chromosome 11. In order to demonstrate this physical colocalization, we used fluorescence in situ hybridization (FISH) with bacterial artificial chromosome (BAC) probes for each locus. According to a strict definition of colocalization, one and only one pair of alleles colocalized in each of three cell lines in 30 to 42% of all cells examined (Fig. 3; fig. S6), thereby demonstrating the close association between these two chromosomes.

Fig. 3.

FISH results in BMM3-4, FSKN2, SF1-G, BMshRNA4-4, and ICR-deleted cells. H19-BAC1 probe is labeled by Spectrum Red deoxyuridine triphosphate (dUTP); RP23-256H2 probe is labeled by Spectrum Green dUTP. Fifty random nuclei from each cell line were counted on each of three slides for each cell line to determine the colocalization percentage of the two alleles. Colocalization is defined as overlapping or touching of the two signals; any separation of the signals (i.e., any intervening pixels) is defined as nonoverlapping. The degree of colocalization (average of three slides) of the two loci is BMM3-4 cells (15 out of 50; 30%), FSKN2 cells (16 out of 50; 32%), SF1-G cells (21 out of 50; 42%), BMshRNA4-4 cells (0 out of 50; 0%), paternal ICR-deleted (+/Δ) cells (8 out of 50; 16%), and maternal ICR-deleted (Δ/+) cells (0 out of 50; 0%).

When the 3.8-kb maternal ICR was deleted (13), no interchromosomal interaction was detected, but when the paternal ICR was deleted, the colocalization of these interchromosomal regions was preserved (Fig. 3), which confirmed the allele-specific requirement for the CTCF-binding maternal allele. Igf2 was biallelically expressed in the cells in which the maternal, but not the paternal, ICR was deleted, and H19 was biallelically expressed in both ICR-deleted cell lines. Nf1 and Wsb1 mRNA abundance was slightly higher in the maternally deleted ICR cells compared with the paternally deleted ICR cells (figs. S7 and S8).

Finally, the role of CTCF in promulgating this interchromosomal interaction was examined by knocking-down endogenous CTCF levels. In cells expressing very low levels of CTCF (Fig. 4), FISH analysis revealed absence of colocalization between IAS1 and the ICR (Fig. 3), and the expression of Nf1 and Wsb1 from the paternal allele was reduced by ∼50%, while the expression of these genes from the maternal allele was unchanged (Fig. 4; fig. S9). Loss of Igf 2, but not H19, imprinting was detected in the BMshRNA4-4 cell line (fig. S10).

Fig. 4.

Decreased expression of Nf1 and Wsb1 paternal alleles in CTCF knocked-down BMshRNA4-4 cells. (A) CTCF expression by PCR analysis in BMM3-4, BMpSM2c, and BMshRNA4-4 cell lines. BMM3-4 has no vector; BMpSM2c cells: BMM3-4 cells in which an empty pSM2c vector is transferred to serve as a control; BMshRNA4-4 cells: BMM3-4 cells in which a pSMshRNA4-4 vector that expressed CTCF shRNA is transferred to knock-down CTCF expression. (B) Quantitative PCR results show mouse CTCF mRNA abundance was inhibited by 89% in BMshRNA4-4 cells. (C) Western blot analysis of BMM3-4 and BMshRNA4-4 using mouse CTCF monoclonal antibody. No CTCF protein could be detected in BMshRNA4-4 cells. (D and E) Allele-specific PCR of Nf1 and Wsb1 in BMM3-4 and BMshRNA4-4 cells. (F) Gapdh expression by PCR (as a control). (G) Allele-specific quantitative PCR. Abundance of Nf1 and Wsb1 in BMM3-4 and BMshRNA4-4 cells (n = 6).

We developed the ACT assay to identify distant DNA regions that are in propinquity or interact with a defined target DNA region, and ACT predicted the colocalization of Igf 2/H19 ICR with Wsb1/Nf1. Using FISH, we showed that one Igf 2/H19 ICR allele on chromosome 7 was frequently in close association with one allele of Wsb1/Nf1 IAS1 on chromosome 11. We hypothesize that the maternal allele on chromosome 7 interacts with the paternal allele of chromosome 11, because only these parental alleles bind CTCF, and the interaction disappears when CTCF levels are diminished or when the maternal ICR is deleted. Moreover, when CTCF abundance is reduced by short hairpin–mediated RNA interference (shRNA)-induced knockdown, there is both loss of imprinting of Igf 2 and decreased expression of Wsb1 and Nf1.

The existence of long-range interactions between regions of a chromosome separated by ∼80 kb led to the discovery of intrachromosomal loops that juxtapose downstream enhancers to promoter regions to increase gene transcription (1, 1012). Intrachromosomal looping has been described for the Hgb locus and for the Dlx5/Dlx6 imprinted dyad, as well as Igf 2/H19 (1418). Associations between chromosomes were recently described by Spilianakis et al. (19), who demonstrated a regulatory interaction between the interferon-γ promoter region on chromosome 10 and regions of the T helper cell TH2 cytokine locus on chromosome 11. They suggested that interchromosomal associations might also be seen in other systems where gene expression is monoallelic.

In intermitotic cells, much of the chromatin is decondensed into a three-dimensional chromosomal territory within a highly organized nuclear architecture (20). There is evidence that each chromosome occupies a distinct territory, and three-dimensional maps have been proposed that delineate precise neighborhoods for each chromosome. Repressed genes may be localized to loops that are positioned within silent compartments inside or outside of the chromosome territory. Actively transcribed genes may reside on loops that extend from the chromosomal territory into interchromatin compartments, where they may share transcriptional machinery with other nearby genes or, potentially, with active genes that are located on nearby chromosomes.

Although it is likely that these two genes on chromosomes 7 and 11 physically interact with and regulate each other's expression with CTCF as a necessary intermediary, it is also possible that these DNA regions are present on loops that occupy the same transcription factory and that they are in close juxtaposition, but do not directly interfere with or enhance each others' transcriptional activity (2123). Transcription factories in which numerous preassembled transcription complexes are concentrated are presumed to exist in the nucleus. As there are fewer factories than transcribed genes in each cell, several genes probably share common factories (21, 24). Transcriptionally active genes are recruited into these factories and then may leave when transcription ceases. Osborne et al. (21) have recently demonstrated that genes over 40 megabases (Mb) of mouse chromosome 7 can interact in cis in these factories, and they postulate that chromatin loops from nearby chromosomes could also occupy the same transcription factory. In our experiments, the absence of CTCF abrogated the association of the interchromosomal regions and decreased Nf1 and Wsb1 mRNA abundance, suggesting that the chromosome 11 loop may only be recruited to the factory when bound to CTCF. The absence of the maternal ICR on chromosome 7 also abolished the conformational propinquity of these two chromosomal loops, suggesting that the mutated maternal allele of Igf 2/H19 may no longer be present in the factory. The two regions may also interact in a more direct way, perhaps by binding to dimerized CTCF (25) or to another common factor and stabilizing each others' presence in the factory; in the absence of this factor or the ability to bind this factor, the presence of one or both chromosome loops within the factory may be compromised.

The decrease of paternal Nf1 and Wsb1 expression in the cells in which CTCF was knocked down indicates that the colocalization and/or the lack of CTCF regulates paternal expression of these genes. In cells in which the maternal ICR has been deleted, expression of Nf1 and Wsb1 are increased. It is possible that the amount of CTCF at the transcription factory is limiting, and maternal chromosome 7 ICR and paternal IAS1 on chromosome 11 actively compete for CTCF. In the absence of the maternal ICR, there might be more CTCF available to IAS1, which would lead to increased gene transcription. Because there is strict parental allele specificity for CTCF binding (paternal chromosome 11 and maternal chromosome 7), it is tempting to implicate this interchromosome association in the imprinting process. There was loss of Igf 2 imprinting in cell lines in which CTCF was knocked down and in which the maternal ICR was deleted (figs. S8 and S10).

Our data suggest that CTCF acts as an organizer of higher-order chromatin structure, directing DNA segments into transcription factories and/or facilitating interactions with other DNA segments. In this way, CTCF may enable epistasis in addition to its other important roles of regulating transcription and serving an insulator function. Absence of CTCF resulted in the disassociation of the Wsb1/Nf1 locus from the Igf 2/H19 locus, which altered gene expression.

The association between the Igf 2/H19 locus on chromosome 7 and the Wsb1/Nf1 locus on chromosome 11 provides an ideal model to study the relation between spatial genome organization and gene expression. The ACT assay can discover new interchromosomal associations or confirm the existence of previously suspected ones. The close association of different chromosomes could provide a model to elucidate RNA trans-splicing leading to interchromosomal mRNA fusion (26, 27). Interchromosomal associations, such as those discovered in TH cells (19), demonstrate the complexity and sophistication of remote regulation of gene expression and indicate the need to study gene expression in the context of whole-nucleus architecture. By so doing, we can enhance our understanding of gene regulation and interchromosomal DNA recombination in normal physiology and diseases caused by translocation and transposition (28, 29).

Supporting Online Material

www.sciencemag.org/cgi/content/full/312/5771/269/DC1

Materials and Methods

Figs. S1 to S10

Tables S1 and S2

References

References and Notes

View Abstract

Navigate This Article