The role of CTCF in regulating nuclear organization

At the microscopic level, the nucleus is grossly partitioned into regions of heterochromatin and euchromatin, corresponding to inactive and active portions of the genome, respectively (1). In addition, each chromosome appears to be positioned nonrandomly, adopting preferred nuclear positions relative to each other in certain cell types (2). Interestingly, gene activation or silencing is frequently accompanied by locus relocalization within the nucleus (3–5), suggesting that the spatial organization of chromatin may be important for the regulation of gene expression.

Recently developed techniques, such as 3C (chromosome conformation capture) and RNA-TRAP (tagging and recovery of associated proteins) identify interacting regions of chromatin and reveal its overall three-dimensional organization within the nucleus (6, 7). In 3C, formaldehyde–cross-linked chromatin is digested with restriction enzymes and then ligated under conditions that favor the ligation of cross-linked fragments, which can then be detected by polymerase chain reaction (PCR) and sequenced. In RNA-TRAP, in situ hybridization is used to target horseradish peroxidase activity to primary transcripts associated with a transcribed gene. Localized peroxidase activity catalyses the covalent attachment of a biotin tag to nearby chromatin, which after purification on streptavidin agarose beads can be analyzed by PCR to determine the presence of specific interactions.

The use of these techniques has identified specific interactions between distal genomic sequences, revealing another level of nuclear organization (6–12). Although many of these interactions are likely to be coincidental and are driven by the need to share common resources, as with the colocalization of genes within transcription factories (13), several of these long-range interactions have been shown to possess important biological functions (8, 11, 12). These new findings have expanded our understanding of the mechanisms regulating gene expression, and have led to the replacement of the traditional linear model of gene regulation with a three-dimensional model in which long-range intrachromosomal and interchromosomal associations orchestrate and coordinate gene expression. In this model, distal regulatory elements physically interact both in cis and in trans with the genes that they control. Perhaps most surprisingly, regulatory elements on one chromosome can also directly regulate the expression of genes on other chromosomes (8, 11, 12). These new concepts might have potential clinical importance, as they provide an explanation for the preferential chromosomal translocation partners observed in certain cancers (14). On page 785 in this issue, Majumder et al. (15) help explain how CTCF might orchestrate these long-range interactions.

A three-dimensional model of gene regulation raises the question of which factors control nuclear organization and long-range interactions. One potential candidate is CTCF, a ubiquitously expressed transcription factor that has multiple context-dependent functions. CTCF can act as an enhancer-blocking protein, can bind to boundary elements to prevent spreading of heterochromatin, and can function as both a transcriptional activator and a transcriptional silencer. How CTCF carries out these diverse functions is not completely clear, but recent work from several groups has revealed that CTCF may function through topological organization of the genome (Fig. 1). One study, for example, suggested that CTCF regulates gene expression by inducing the formation of long-range chromatin loops (16). Two years later, the first direct evidence for CTCF-mediated chromatin looping was revealed from analysis of the β-globin locus (17).

CTCF was later found to mediate interchromosomal interactions between the Igf2/H19 imprinting control region (ICR) on chromosome 7 and the Wsb1/Nf1 gene complex on chromosome 11 (12). Deletion of the Igf2/H19 ICR or abrogation of CTCF expression disrupted this interchromosomal association and altered Wsb1/Nf1 gene expression (12). More recently, CTCF has been implicated in driving X chromosome homologous pairing during the process of X chromosome inactivation (18). Collectively, these results suggest that CTCF mediates at least some of its gene regulatory functions through the three-dimensional organization of the genome.

The growing interest in CTCF is highlighted in the study by Majumder et al. (15), which describes a new model of major histocompatibility complex (MHC) class II gene regulation that is mediated by CTCF-dependent, long-range intrachromosomal interactions. Expression of two divergently transcribed MHC class II genes, HLA-DRB1 and HLA-DQA1, is driven by an intervening nuclear matrix–bound enhancer element called XL9 (19). CTCF had been shown to bind to this enhancer element, but its effect on the expression of the two class II genes had not been investigated (19). In the new study, the authors find evidence for long-range chromatin loops between the promoters of the HLA-DRB1 and HLA-DQA1 genes and the intergenic XL9 enhancer. These interactions depended on a complex consisting of CTCF, the transcription factor RFX, and the transactivator CIITA. Knocking down CTCF using RNAi reduced the long-range interactions between the XL9 enhancer element and the MHC class II genes and decreased expression of HLA-DRB1 and HLA-DQA1. These findings provide a novel model for MHC class II expression, and also provide insight into several unanswered questions about the biology of CTCF, such as how these long-range interactions are mediated and how the tissue-specific functions of CTCF are regulated.

Because CTCF can form dimers and maybe even oligomers, it is possible that CTCF molecules bound to distal elements could interact with each other, thereby driving loop formation or interchromosomal interactions (16). However, in a study by Ling et al. (12), the presence of CTCF on both interacting loci was not an absolute requirement for interchromosomal interaction, suggesting either that CTCF is part of a multi-protein bridging complex or that CTCF recruits additional factors that then drive long-range interactions. Indeed, Majumder et al. (15) showed that CTCF requires at least two additional factors to function, at least in the control of the HLA-DRB1 and HLA–DQA1 genes. CTCF, RFX, and CIITA formed a complex even in the absence of DNA, and the loss of any one of these factors abolished long-range interactions with the intergenic enhancer element, precluding MHC class II expression.

Whether RFX and CIITA form a structural part of a bridging complex or whether they simply recruit CTCF and/or additional factors to the XL9 enhancer remains to be resolved. It will be interesting to purify these and other bridging complexes and characterize their protein composition. Interestingly, recent data show that CTCF is able to recruit cohesins to specific genomic locations (20–22). Cohesins, better known for their role in mediating sister chromatid cohesion during mitosis, have proven chromatin bridging potential and are thus an obvious candidate for spatial organization of the genome.

As CTCF is a ubiquitous factor, its function cannot be regulated by differential expression. Thus, long-range intra- and interchromosomal interactions must be regulated on other levels. This control might involve direct blocking of CTCF binding by developmentally regulated or allele-specific DNA methylation. The mono-allelic nature of the interaction between the Igf2/H19 ICR and the Wsb1/Nf1 locus, for example, appears to be conferred by imprinted DNA methylation of the noninteracting loci (12).

As CTCF may require the presence of other protein factors in some contexts, another level of control might involve regulation of the expression of these factors or their ability to interact with CTCF. Indeed, Majumder et al. (15) found that interactions between XL9 and the HLA-DRB1 and HLA-DQA1 genes could be induced in A431 epithelial cells by interferon γ treatment, which is essential for CIITA expression in these cells. In untreated A431 cells, CIITA is not expressed, precluding long-range interactions between the HLA-DRB1 and HLA-DQA1 genes and the XL9 enhancer and thus MHC class II expression.

The model proposed by Majumder et al. (15) predicts that several chromatin organization states might exist for the MHC class II locus. For example, the MHC class II locus in B cells, which express CTCF, RFX, and CIITA, is in an active state with long-range interactions between the XL9 enhancer and the HLA-DRB1 and HLA-DQA1 genes driving constitutive class II expression. In contrast, most nonimmune cells do not express CIITA or MHC class II. In these cells, however, regulatory elements in the MHC class II locus are still bound by transcription factors such as RFX, CREB, and NF-Y (23, 24). And these silent MHC class II promoters display histone modifications normally associated with accessible chromatin (25), suggesting that in cells lacking CIITA, the MHC class II genes are poised and ready for class II expression. According to the data from Majumder et al. (15), the induction of CTIIA expression in these cells would lead to recruitment of CTCF to the XL9 enhancer, the formation of long-range interactions between the XL9 enhancer and the HLA-DRB1 and HLA-DQA1 genes, and initiation of MHC class II expression. However, it must be noted that in the absence of genetic evidence, it is not possible to rule out an alternative hypothesis in which the formation of long-range interactions is a consequence (rather than the cause) of MHC class II transcription.

The number of CTCF binding sites in the human genome was recently estimated to be between 14,000 and 15,000, and it is likely that a comparable number are present within the mouse genome (26, 27). Interestingly, most of these sites appear to be occupied in a cell type–independent manner (27). It is therefore tempting to speculate that CTCF is important for the maintenance of nuclear architecture and, by default, the regulation of many genes. This does not appear to be the case, however, according to Majumder et al. (15), as knocking down CTCF did not result in global changes in gene expression. Thus, it is highly likely that other factors also serve to regulate long-range interactions and nuclear structure. The protein SATB1, for example, has been implicated in the regulation of thymocyte nuclear architecture and the initiation of cytokine expression by directing long-range interactions within the T helper 2 cytokine locus on mouse chromosome 11 (28, 29). Other factors, such as cohesins, MeCP2, and MENT, which have been shown to have chromatin bridging potential, may also help organize the three-dimensional structure of the nucleus (20–22, 30, 31).

Further characterization of MHC class II gene expression may reveal additional CTCF-mediated long-range interactions within the MHC class II gene complex. Indeed, the MHC class II gene HLA-DRA has previously been shown to form DNA loops that bring other regulatory elements into close proximity with its promoter sequences (25). Additional CTCF binding sites have also been identified within the MHC class II locus, highlighting the potential for a general role for CTCF in regulating MHC class II expression (27, 32).

From a more global perspective, long-range interactions and interchromosomal associations are likely to be widespread within the nucleus, meriting genome-wide analyses to define the nuclear “interactome” (11, 33–35). The task will then be to sort through these data to identify which interactions are functional and which are coincidental. In addition, isolation and characterization of chromatin bridging complexes, such as those containing CTCF, may reveal additional players in long-range regulation. One of the most interesting questions that needs to be addressed is whether these long-range interactions are stable, static structures or ephemeral complexes that form from transient interactions.