Chromatin Insulators: A Role in Nuclear Organization and Gene Expression

Abstract

Chromatin insulators are DNA-protein complexes with broad functions in nuclear biology. Based on the ability of insulator proteins to interact with each other, it was originally thought that insulators form loops that could constitute functional domains of co-regulated gene expression. Nevertheless, data from genome-wide localization studies indicate that insulator proteins can be present in intergenic regions as well as at the 5′, introns or 3′ of genes, suggesting a broader role in chromosome biology. Cells have developed mechanisms to control insulator activity by recruiting specialized proteins or by covalent modification of core components. Recent results suggest that insulators mediate intra- and inter-chromosomal interactions to affect transcription, imprinting and recombination. It is possible that these interactions set up cell-specific blueprints of nuclear organization that may contribute to the establishment of different patterns of gene expression during cell differentiation. As a consequence, disruption of insulator activity could result in the development of cancer or other disease states.

INTRODUCTION

Patterns of transcription required for cell differentiation are initially established by specific transcription factors. The maintenance of these patterns of gene expression is then carried out by alterations in chromatin structure that are epigenetically inherited between cell generations. These changes in chromatin organization take place at the level of the 10 nm fibre and include covalent histone modifications, DNA methylation and alterations induced by ATP-dependent remodelling complexes. In addition, recent evidence suggests that the three-dimensional organization of the genome within the nucleus of eukaryotic cells may also be critical for achieving proper spatial and temporal patterns of gene expression during development. The factors and processes involved in the establishment, maintenance and regulation of specific states of nuclear organization are largely unknown but insulators are emerging as likely candidates to play this crucial role (Bushey, Dorman et al. 2008; Phillips and Corces 2009).

Insulators are DNA-protein complexes experimentally defined by their ability to block enhancer-promoter interactions and/or serve as barriers against the spreading of the silencing effects of heterochromatin. Not all sequences described as insulators display these two properties; Felsenfeld and colleagues have proposed to classify these sequences as “enhancer blocking” of “barrier” insulators depending on whether they interfere with enhancer promoter interactions or antagonize spreading of heterochromatin under experimental conditions (Gaszner and Felsenfeld 2006). Barrier insulators appear to function by recruiting enzymes involved in histone modifications that stop the spreading of silencing marks (Huang, Li et al. 2007; Dhillon, Raab et al. 2009) and will not be considered in this review. Instead, we concentrate our discussion on the function of enhancer blocking insulators with special emphasis on recent data supporting a role in nuclear organization. These insulators were originally discovered in Drosophila, which has been a particularly good model system in which to analyze insulator function. Several different insulators have been identified in this organism whereas vertebrates appear to mostly rely on the CTCF insulator (Phillips and Corces 2009). We conclude by proposing that the primary role of insulators may not be to regulate enhancer-promoter interactions or heterochromatin spreading. Rather, insulators may mediate intra- and inter-chromosomal interactions with the primary goal of organizing the eukaryotic genome into epigenetically heritable states. This insulator-mediated organization may be important to regulate DNA function at multiple levels, including transcription initiation, elongation and DNA recombination. Miss-regulation of insulator function may lead to alterations in gene expression and the development of disease states and cancer.

INSULATOR STRUCTURE: CORE COMPONENTS

There are several types of insulators in Drosophila that have been studied in detail. They include the scs and scs’ sequences originally discovered flanking the heat shock hsp70 locus (Kellum and Schedl 1992; Zhao, Hart et al. 1995), the gypsy insulator first found in the gypsy retrotransposon (Geyer and Corces 1992; Hoover, Chien et al. 1993), the Fab-7/Fab-8/Mcp insulators located in the bithorax complex (Gyurkovics, Gausz et al. 1990; Karch, Galloni et al. 1994) and the SF1 insulator described in the Antennapedia complex (Belozerov, Majumder et al. 2003). Each of these insulators consists of a DNA sequence and a specific DNA binding protein that interacts with this sequence; additional factors that interact with the DNA binding protein have been characterized for some of these insulators (Figure 1).

Figure 1.

Diagram showing the structure of different Drosophila and vertebrate insulators. A. Each Drosophila insulator subclass contains a different binding protein that may define the specific function of the corresponding subclass. All insulators share the common protein CP190, although the role of this protein in the function of the GAGA insulator has not been demonstrated experimentally. In addition, all subclasses may also have one Mod(mdg4) isoform. The gypsy/Su(Hw) insulator contains Mod(mdg4)2.2. The dCTCF insulator lacks this isoform but contains a different variant of Mod(mdg4) (T. Gerasimova, E. Lei and V. Corces, unpublished observations). It is not known whether the BEAF insulator contains a Mod(mdg4) variant but GAGA has been shown to interact with Mod(mdg4)2.2. B. Structure of the vertebrate CTCF insulator.

In the case of the scs insulator, the DNA binding protein component is Zeste-White 5 (ZW5), which is a Zn finger protein required for cell viability. Null mutations in the zw5 gene are recessive lethal, but hypomorphic alleles display a variety of pleiotropic effects on wing, bristle and eye development (Gaszner, Vazquez et al. 1999). The scs’ sequences interact with a protein called Boundary Element Associated Factor 32 (BEAF 32). The BEAF 32 gene encodes two different proteins named BEAF 32A and BEAF 32B, which are present at hundreds of sites on Drosophila polytene chromosomes (Zhao, Hart et al. 1995). The two isoforms differ at the N terminal DNA binding domain (BED Finger domain). The common C terminal region is involved in protein-protein interactions between the two isoforms. Analysis of mutations in the BEAF 32 gene shows that BEAF 32B is required for viability whereas BEAF 32A mutations do not show significant phenotypic defects. Expression of a dominant negative form of BEAF 32 results in changes in chromosome structure and cell viability(Gilbert, Tan et al. 2006; Roy, Gilbert et al. 2007; Roy, Tan et al. 2007).

The gypsy insulator contains binding sites for the Suppressor of Hairy-wing [Su(Hw)] protein, which is an 12 zinc finger DNA binding protein. Mutations in the su(Hw) gene cause female sterility but do not result in lethality (Harrison, Gdula et al. 1993). Su(Hw) interacts with two other components of the gypsy insulator, Mod(mdg4)2.2 and CP190 (Gause, Morcillo et al. 2001; Ghosh, Gerasimova et al. 2001). Mod(mdg4)2.2 does not bind to DNA directly but interacts with Su (Hw) through its carboxy-terminal domain. In addition, Mod(mdg4)2.2 also contains a BTB domain in the N-terminal domain that mediates homo and hetero-multimerization with other insulator components. The mod(mdg4) gene encodes approximately 29 different isoforms that arise by alternative cis- and trans-splicing (Buchner, Roth et al. 2000; Labrador, Mongelard et al. 2001; Labrador and Corces 2003); null mutations in the gene result in lethality, but mutations affecting the Mod(mdg4)2.2 isoform are viable and show defects in gypsy insulator function (Gerasimova, Gdula et al. 1995). CP190 also contains a BTB domain as well as 3 zinc fingers and it interacts with both Su(Hw) and Mod(mdg4)2.2. The BTB domains of Mod(mdg4)2.2 and CP190 also mediate homo di- and multi-merization of these two proteins. CP190 binds DNA with low affinity and specificity but it does not interact directly with insulator sequences present in the gypsy retrotransposon, where it is recruited through interactions with Su(Hw) and Mod(mdg4)2.2 instead. Mutations in the CP190 gene are lethal (Pai, Lei et al. 2004).

The bithorax complex of Drosophila contains an intricate collection of transcriptional regulatory sequences that orchestrate the complex spatio-temporal expression of the three genes present in the complex. The proper interplay between these regulatory sequences requires the function of several insulators of which Fab-8 has been studied in most detail. Fab-8 sequences interact with the Drosophila homolog of the vertebrate CTCF insulator protein. Drosophila CTCF (dCTCF) has 12 zing fingers. Mutations in dCTCF are lethal and show abdominal hometic phenotypes (Gerasimova, Lei et al. 2007; Mohan, Bartkuhn et al. 2007). dCTCF is also found in the Mcp and Fab-6 insulators present in the bithorax complex but not Fab-7 (Holohan, Kwong et al. 2007). Fab-7 may represent a fourth class of insulators that use the GAGA factor (GAF) as a DNA binding protein that also contains a BTB domain. Mutations in the trl gene, which encodes GAF, affect Fab-7 insulator activity (Schweinsberg, Hagstrom et al. 2004). In addition, GAF is present and required for the function of the SF1 insulator found in the Antennapedia complex (Belozerov, Majumder et al. 2003).

If the functions of all Drosophila insulators have converged into that of CTCF in vertebrates, one may expect to find some shared protein components among Drosophila insulators. This is indeed the case. The CP190 protein, first found in the gypsy/Su(Hw) insulator, also interacts with dCTCF (Gerasimova, Lei et al. 2007; Mohan, Bartkuhn et al. 2007). Genome-wide mapping of dCTCF and CP190 sites also support this conclusion (Bartkuhn, Straub et al. 2009; Bushey, Ramos et al. 2009). These studies have also shown that BEAF and CP190 co-localize at hundreds of sites throughout the genome (Bushey, Ramos et al. 2009). These results suggest that the insulators defined by these three different DNA binding proteins, Su(Hw), dCTCF and BEAF share the BTB domain-containing protein CP190 and may therefore use similar mechanisms to effect their insulator function. On the other hand, GAF does not appear to interact directly with CP190 but has been shown to interact with Su(Hw) and Mod(mdg4)2.2 (Melnikova, Juge et al. 2004); since these two proteins can in turn interact with CP190, GAF insulators may act mechanistically like the other three types (Figure 1).

Vertebrates appear to rely mostly on the widespread CTCF insulator (Wallace and Felsenfeld 2007) and we will concentrate our discussion on this sequence. CTCF is a highly conserved protein containing an eleven zinc finger central DNA binding domain, displaying close to 100% homology between mouse, chicken, and human, embedded within slightly more divergent N- and C-termini (Ohlsson, Renkawitz et al. 2001). On the basis of its ability to bind to a wide range of variant sequences as well as specific co-regulatory proteins through combinatorial use of different zinc fingers, CTCF was originally described as a multivalent factor (Filippova, Fagerlie et al. 1996). This structural feature may explain the ability of vertebrate cells to fulfil all insulator functions with just one insulator whereas Drosophila cells require multiple insulators. CTCF homozygous knockout mice exhibit early embryonic lethality prior to implantation, highlighting the critical importance of CTCF in diverse cellular processes (Splinter, Heath et al. 2006; Heath, Ribeiro de Almeida et al. 2008).,

Interestingly, the Mod(mdg4) and CP190 proteins have not been conserved in vertebrates. Instead, the vertebrate CTCF protein interacts with cohesin, which has recently emerged as a critical partner of CTCF in mediating inter- and intra-chromosomal interactions necessary for transcription and insulator function. Cohesin is a ring-shaped complex that holds chromatids together between the S and M phases of the cell cycle. In vertebrates, cohesins are present at CTCF target sequences in the genome and this localization depends on CTCF (Parelho, Hadjur et al. 2008; Rubio, Reiss et al. 2008; Stedman, Kang et al. 2008; Wendt, Yoshida et al. 2008). Cohesin is necessary for nearly all the intra-chromosomal interactions attributed to CTCF, such as those observed at the H19/IGF2, β-globin, IFNG and APO loci; knockdown of cohesin impairs chromatin interactions at these loci (Hadjur, Williams et al. 2009; Mishiro, Ishihara et al. 2009; Nativio, Wendt et al. 2009; Hou, Dale et al. 2010). These observations have led to the proposal that the ring-like structure of the cohesin complex can mediate interactions between distant genomic sites by a mechanism similar to that employed to maintain sister chromatids together.

INSULATOR STRUCUTRE: ACCESORY FACTORS

CTCF has been found to interact with other nuclear factors in addition to cohesins, although the broad relevance of some of these interactions to general aspects of CTCF function is unclear. For example, CTCF interacts in vivo with CHD8, which is a member of the chromodomain helicase (CHD) family and has an SNF2-like helicase/ATPase domain. CHD8 has been shown to be present at some CTCF target sites, including the H19/IGF2 Imprinting Control Region (ICR), the promoter regions of the BRCA1 and C-MYC genes, and the 5′HS5 insulator of the β-globin locus in human cells (Ishihara, Oshimura et al. 2006). Like CTCF, CHD8 is also required for the insulator activity of the H19/IGF2 ICR. Loss of CHD8 leads to expression of IGF2 from the maternal and paternal alleles (Ishihara, Oshimura et al. 2006). CHD8 together with CTCF can affect aspects of chromatin structure such as CpG DNA methylation and histone acetylation. Knockdown of CHD8 leads to hypermethylation of CpGs at CTCF binding sites in the promoter regions of the BRCA1 and C-MYC genes (Ishihara, Oshimura et al. 2006). In addition to CHD8, CTCF can also interact in vitro with SIN3A which recruits a histone deacetylase activity necessary for the silencing function of CTCF (Lutz, Burke et al. 2000); this result has not been confirmed by in vivo experiments. These observations suggest that CTCF may have context-dependent functions that are mediated by different protein partners.

CTCF has been also found to interact with transcription factors such as YY1, which is a zinc finger protein capable of activating or repressing transcription depending on the promoter context. Like CTCF, YY1 also binds to DNA in a methylation-sensitive manner. YY1 can bind the PEG3 ICR and regulate its parent of origin-dependent expression. Insulator assays with deletion constructs of YY1-binding sites indicate that the region functions as a methylation-sensitive insulator like the H19/IGF2 ICR (Kim, Kollhoff et al. 2003). YY1 is a required cofactor for CTCF in processes such as X chromosome inactivation. Both YY1 and CTCF bind at Tsix DNA, and a deficiency of YY1 leads to aberrant Tsix and Xist expression, aberrant XCI patterns, and results in a deficit of male and female embryos. YY1 and CTCF together function as Tsix transcriptional activators in ES cells (Donohoe, Zhang et al. 2007). It is not yet clear whether YY1 alone can act as an insulator or only contributes to the insulator activity of CTCF as a cofactor. In addition to YY1, CTCF can interact with other transcription factors such as YB1, Oct4, Kaiso and thyroid hormone receptor (TR). CTCF interacts with YB1 in vitro and in vivo and functions as a co-repressor at the C-MYC promoter (Chernukhin, Shamsuddin et al. 2000). CTCF interacts with Kaiso in vivo and both bind to the chicken HS4 insulator. Kasio inhibits enhancer the blocking activity of CTCF in transgenic assays (Defossez, Kelly et al. 2005). CTCF interacts with Oct4 and control X-chromosome pairing during X-chromosome inactivation (Donohoe, Silva et al. 2009). A subset of CTCF binding sites is found next to TR response elements in the mouse C-MYC and the human APP genes. Functional TR binding is required for enhancer blocking activity (Weth, Weth et al. 2010). Intriguingly, CTCF has been also found to interact with RNA polymerase II. The largest subunit of Pol II associates with CTCF target sites in vivo and this interaction is dependent on the presence of intact CTCF binding sequences. A single CTCF binding site in the C-MYC gene (N-MYC-CTCF), but not amutant site deficient for CTCF binding, is sufficient to activate transcription from a promoterless reporter gene in stably transfected cells (Chernukhin, Shamsuddin et al. 2007). This suggests that CTCF may recruit Pol II and activate transcription. The extent of such a role for CTCF is unclear, since genome-wide studies of CTCF sites suggest that only a subset of CTCF sites are located at promoters in the human genome (Kim, Abdullaev et al. 2007).

The function of Drosophila and vertebrate insulators appears to rely in part on their ability to interact with the nuclear lamina. Drosophila insulator proteins fractionate with components of the nuclear lamina, suggesting a direct or indirect interaction between the two (Byrd and Corces 2003). Furthermore, the dTopors protein, which is located in the nuclear periphery, interacts with both Mod(mdg4)2.2 and Lamin, and mutations in the Lamin Dm0 gene results in disruption of insulator activity (Capelson and Corces 2005). It appears that interaction with a nuclear substrate is also required for the proper function of the CTCF insulator. For example, it has been shown that CTCF interacts with nucleophosmin, a nuclear matrix protein that is concentrated at the surface of the nucleolus and associates with CTCF at the chicken HS4 insulator in the β-globin locus. Consistent with this distribution, transgenic copies of reporter genes containing the chicken HS4 insulator show preferential localization to the outer part of the nucleolus, and this localization depends on the presence of an intact CTCF binding site (Yusufzai, Tagami et al. 2004). In addition to nucleophosmin, other nuclear matrix proteins such as Lamin may also be important for CTCF function. In human cells, many of the Lamin B1-associated domains (LADs) in the genome are flanked by CTCF (Guelen, Pagie et al. 2008). Lamin A also works together with CTCF to establish nucleolar localization. The D4Z4 human subtelomeric repeats localize telomeres to the nuclear periphery and display insulator activity in transgenic cell lines. CTCF and Lamin A both bind to D4Z4 and are required for proper perinuclear localization. Reduction of either protein results in a decrease in the localization of telomeres at the nuclear periphery and impairs insulator function (Ottaviani, Rival-Gervier et al. 2009; Ottaviani, Schluth-Bolard et al. 2009). It is thus possible that the interaction between CTCF and other nuclear matrix proteins contributes to chromatin localization and organization in the nucleus in a variety of cell processes.

These results suggest that Drosophila and vertebrate insulators may play a variety of functions in nuclear biology. Many of these functions can be explained by the ability of CTCF and Drosophila insulator proteins to mediate interactions between two or more chromosomal locations. Different Drosophila insulators may have specialized roles in effecting distinct nuclear functions based on the presence of particular protein components. In vertebrates, it is not clear whether CTCF acts by different mechanisms that are context dependent and vary based on its association with the different partners described above. It is nevertheless possible that these different functions have a common requirement for the ability of CTCF to mediate interactions between distant sites in the genome; this may be also the case in some of the more classical roles characteristic of standard transcription activators and repressors. Although the DNA recognition sequence of CTCF appears to be conserved at different genomic locations (Kim, Abdullaev et al. 2007), the distinct context-dependent roles of this protein must be mediated by the different protein partners with which it associates. The nature of the determinants responsible for these context-dependent interactions is unknown and should be an important issue for future investigation.

GENOMIC DISTRIBUTION OF INSULATORS

The existence of several Drosophila insulator subclasses with different DNA binding proteins but sharing some functional components raises the question of whether they all have the same role in the regulation of gene expression or whether there is a functional specialization in their tasks. The possibility of such specialization is highlighted by results showing differential localization of insulator subclasses with respect to genomic landmarks. For example, Su(Hw) and dCTCF are preferentially excluded from exonic regions (mostly 5′ and 3′ UTRs), with only 8%, 16%, and 17% of sites found within exons, respectively, whereas BEAF sites are enriched in UTRs (Bushey, Ramos et al. 2009). When the location of these proteins is compared with the location of genes, few Su(Hw)-binding sites are found in the 1 kb regions flanking genes. However, dCTCF and BEAF show a distribution that is highly skewed toward the 5′ end of genes and is enriched in the first 200 bp just upstream of the transcription start site. Insulator proteins also show a compartmentalized distribution in relation to the level of gene expression. For example, 83% of dCTCF sites and 89% of BEAF sites at the 5′ end of genes localize to genes that are highly expressed. However, Su(Hw)-binding sites are most often found near genes with low expression levels. Finally, different insulator proteins appear to associate with genes involved in different cellular processes. Genes containing dCTCF in the 200 bp region upstream of their transcription start site are mostly involved in developmental processes, whereas genes containing BEAF in this region are mostly involved in metabolic processes. Both dCTCF and BEAF are enriched near or at genes involved in cell cycle processes, whereas Su(Hw)-containing genes show little significant clustering based on biological process (Emberly, Blattes et al. 2008; Bushey, Ramos et al. 2009; Jiang, Emberly et al. 2009). These observations suggest a division of labour among Drosophila insulators, both with respect to gene function as well as specific aspects of cell function. Given the fact that all insulators share CP190 and perhaps Mod(mdg4), it is likely that all of them use the same mechanism to perform their function, namely bring together different regions of the genome. Nevertheless, Drosophila cells appear to use a variety of DNA binding insulator proteins to recruit other insulator components to mediate these interactions. Therefore, the specific outcome of these interactions may be determined by where in the genome the binding sites for each of these proteins are localized. In the case of Su(Hw) and a subset of dCTCF sites, their localization in intergenic regions suggests that their role may be to form loops that may represent independent functional domains. The rest of dCTCF sites and BEAF sites are located around promoter regions and their function may rely on the same type of interactions to bring these regions of genes to specific nuclear compartments such as transcription factories.

Since Drosophila has several insulators with distinct localization patterns with respect to genomic features, it is interesting to contrast this information with the genome wide localization of CTCF in vertebrates. The distribution of CTCF in mouse embryonic stem cells and in various human cell lines has been studied using ChIP-chip and ChIP-seq approaches (Barski, Cuddapah et al. 2007; Kim, Abdullaev et al. 2007; Chen, Xu et al. 2008; Jothi, Cuddapah et al. 2008; Cuddapah, Jothi et al. 2009). The number of sites uncovered varies depending on the approach, from ca. 13,800 in IMR90 human fibroblasts using ChIP-chip (Kim, Abdullaev et al. 2007) to ca. 39,600 in mouse ES cells using ChIP-seq (Chen, Xu et al. 2008). In general, the distribution of CTCF sites on each chromosome correlates with gene density. However, different from transcription factors, CTCF binding sites are generally located an average of 48 kb away from promoters. Approximately 46% of the sites are located in intergenic regions, consistent with the classical role of insulators in establishing chromatin domains, whereas ca. 20% of the sites display promoter proximal localization. As in Drosophila, a significant number of the sites fall within genes, with 22% in introns and 12% in exons in human fibroblast cells (Kim, Abdullaev et al. 2007). Analyses in resting human CD4+ T cells have found a similar genomic distribution of CTCF sites, with 45% intergenic, 7% 5′UTR, 3% exonic, 29% intronic, 2% 3′UTR, and 13% within 5 kb of the transcription start site. Although CTCF sites generally correlate with gene density, it is possible to find patterns in their distribution, other than location with respect to genes, suggestive of various roles in gene expression. For example, CTCF depleted domains, which exhibit lower-than-average CTCF density, tend to include clusters of related gene families and genes that are transcriptionally coregulated. In addition, some CTCF sites are located in intergenic regions at the transition of chromatin domains with different epigenetic status. For example, there are 793 CTCF sites flanking H3K27me3 domains in CD4⁺ T cells and this occupancy may be cell type specific (Cuddapah, Jothi et al. 2009). CTCF can also bind 5–10 kb outside Lamina-associated domains (LADs), which contain low gene density and low expression genes, representing a repressive chromatin environment; out of 1344 LADs found, 333 have CTCF binding at least at one side of the border (Guelen, Pagie et al. 2008). In all three cases, the distribution of CTCF agrees with a classical insulator role for this protein in the establishment or maintenance of functional transcriptional domains. On the other hand, CTCF enriched domains, which have higher than average CTCF binding, often have multiple alternative promoters (Kim, Abdullaev et al. 2007). These and other sites located close or within genes may play roles in chromosome biology different from those classically assigned to insulators.

The genome occupancy of CTCF sites is similar in different cell types but a significant fraction appears to be cell-type specific. For example, 40–60% of CTCF sites are common among CD4⁺ T, HeLa and Jurkat cells (Cuddapah, Jothi et al. 2009) and the rest of the sites occupied by CTCF vary between the different cell types. Similarly, around 30% of CTCF sites are different in human IMR90 fibroblasts and U937 erythroid progenitor cells (Kim, Abdullaev et al. 2007). Nevertheless, as is the case in Drosophila, a number of CTCF sites in mammals are cell type-specific and may play a role in establishing patterns of gene expression required for cell differentiation.

INSULATORS MEDIATE INTRA- AND INTER-CHROMOSOMAL INTERACTIONS

All Drosophila insulators, with perhaps the exception of scs/ZW5 for which data is unavailable, share the BTB domain proteins CP190 and Mod(mdg4), although the specific isoform of Mod(mdg4) appears to be different in different insulators. The BTB domains of these two proteins as well as GAF can interact with each other in various in vitro or in vivo assays, suggesting that insulator proteins may mediate intra- and inter-chromosomal interactions among insulator sites throughout the genome. Various types of observations support this conclusion. For example, Su(Hw), Mod(mdg4)2.2, dCTCF and CP190 show a punctuate distribution in the nuclei of diploid cells. These sites, called insulator bodies, appear to contain multiple individual insulator sequences and their morphology is disrupted by mutations in insulator components (Gerasimova, Byrd et al. 2000; Pai, Lei et al. 2004; Gerasimova, Lei et al. 2007). Furthermore, FISH experiments have shown that DNA sequences contained between two insulators form a loop, which becomes two smaller loops when a new insulator is inserted in the middle of the DNA (Byrd and Corces 2003). Although Drosophila insulator bodies are present throughout the nucleus, they seem to localize preferentially in the nuclear periphery. This localization may be mediated by the protein dTopors, which has been shown to interact with Su(Hw) and Mod(mdg4)2.2 as well as with Lamin Dm0 (Capelson and Corces 2005; Capelson and Corces 2006). Therefore, dTopors may serve as an anchor to attach insulator sites to the nuclear lamina or nuclear matrix (Capelson and Corces 2005). In addition to this type of information based on FISH and immunolocalization experiments, results from 3C experiments have shown that scs and scs’ sequences are present in close proximity in the nucleus forming a loop of the intervening sequence. These contacts may be mediated by CP190 or may be direct, since ZW5 and BEAF 32 have been shown to interact both in vitro and in vivo in Drosophila embryos (Blanton, Gaszner et al. 2003).

The role of insulators in mediating intra- and inter-chromosomal interactions has been better established in vertebrates using 3C, 4C and FISH approaches in a number of different systems. The β-globin locus contains several CTCF sites that are conserved in mice and humans. These sites are located within two DNase I hypersensitive regions that flank the β-globin locus: the 5′HS5 located in the locus control region (LCR) and the downstream 3′HS1; additional CTCF sites are located further away at either end of the locus. The 5′HS5 and 3′HS1 sites interact with each other as well as with the more distal CTCF sites flanking the locus and these interactions are severely weakened in the absence of the CTCF protein. The interactions with 5′HS5 or 3′HS1 are cell type specific but they are not required for proper transcription of the β-globin gene. These interactions exist in both erythroid cells, in which the β-globin gene is transcribed and in erythroid progenitor and fibroblast cells in which the gene is not expressed, but not in non-expressing brain cells (Tolhuis, Palstra et al. 2002; Splinter, Heath et al. 2006; Hou, Dale et al. 2010). Mutation of 3′HS1 destabilizes the long-range interactions among these CTCF sites but mutation or deletion of 3′HS1 or 5′HS5 can neither affect the expression kinetics nor levels of the β-globin genes in erythroid cells (Farrell, Grinberg et al. 2000; Splinter, Heath et al. 2006). Nevertheless, a number of CTCF sites present outside of the β-globin locus interact with each other but not with 5′HS5 or 3′HS1, and some of these interactions are both cell type specific and transcription relevant.

These interactions can only be detected in either erythroid cells or non-erythroid cells. Knockdown of CTCF, which leads to a global reduction of both groups of interactions, can negatively affect β-globin gene transcription (Hou, Dale et al. 2010). It is possible that these interactions are required for β-globin expression, although it is also possible that the observed changes in transcription are due to indirect effects of CTCF knockdown. Therefore, it appears that CTCF-mediated interactions involving 5′HS5, 3′HS1 and other CTCF sites create a three-dimensional organization of the β-globin locus before the commitment to the erythroid lineage that is not necessary or sufficient for the activation of the globin genes. The fact that CTCF-mediated interactions among these sites are not observed in brain cells suggests that this organization may play an earlier role during differentiation after the commitment to ectodermal or mesodermal fates. New interactions established later during erythroid differentiation and involving a different set or more distally located CTCF sites may be involved in the establishment of a more global architecture that is actually more directly responsible of the expression of the β-globin locus.

The requirement of CTCF-mediated interactions for gene expression has been studied in several other loci. At the cytokine interferon-γ (IFNG) locus there are three conserved CTCF binding sites located upstream, downstream and within the gene 1.5 kb from the transcription start site (TSS) in human and mouse cells. The CTCF sites display cell type-specific interactions in specialized T helper 1 (Th1) cells and these interactions parallel differences in CTCF occupancy (Hadjur, Williams et al. 2009; Sekimata, Pérez-Melgosa et al. 2009). These CTCF sites also interact with the enhancers present in the locus, although the enhancers do not have CTCF binding sites. The three-dimensional conformation of the locus and robust Ifng expression in Th1 cells are dependent on the presence of CTCF (Sekimata, Pérez-Melgosa et al. 2009). It is possible that interactions among the CTCF sites help recruit the enhancers and activate gene expression at the locus. The apolipoprotein (APO) gene (APOA1/C3/A4/A5) cluster in humans is another interesting example of CTCF-mediated loops that are required for proper gene expression. Results from 3C experiments indicate that CTCF sites in this region interact with each other in Hep3B cells. These interactions lead to the formation of two transcribed loops with APOC3/A4/A5 in one loop and APOA1 in the other. Consistent with the model based on 3C data, CTCF knockdown leads to a decrease of APOC3/A4/A5 but an increase of APOA1 expression (Mishiro, Ishihara et al. 2009). Major histocompatibility class II (MHC-II) genes are also regulated by intra-chromosome interactions mediated by CTCF. The XL9 element contains CTCF sites and is located in the intergenic region between the MHC-II genes HLA-DRB1 and HLA-DQA1. Knockdown of CTCF diminishes expression of these genes. 3C experiments detect an interaction between XL9 and proximal promoter elements of these two MHC-II genes. These interactions also depend on the class II transactivator (CIITA), which can associate with CTCF (Majumder, Gomez et al. 2008). The chromatin conformation mediated by the associated between CIITA and CTCF is required for transcription of the MHC-II genes. In addition to its role in processes that result in increased transcription of genes, CTCF may also mediate chromatin high order structure that results in gene silencing. For example, 3C experiment suggest that silenced Hox genes in human cells are spatially clustered via a specific three-dimensional architecture of the locus; CTCF is present at the sites of contact and appears to be responsible for the establishment and/or maintenance of this architecture (Ferraiuolo, Rousseau et al. 2010).

The three-dimensional arrangement of the chromatin fibre created by CTCF-mediated interactions also plays an important role in imprinted gene expression at the H19/IGF2 locus. The imprinting control region (ICR) immediately upstream of H19 contains CTCF sites that are crucial for the imprinting of these two genes. Results from 3C experiments demonstrate that these CTCF sites can mediate allele-specific chromosome interactions that seem to control the accessibility of the IGF2 promoter to the shared enhancer. On the maternal allele of both mouse and human cells the ICR is not methylated and CTCF binds to DNA. In the mouse, the ICR interacts with CTCF sites present in the upstream DMR1 and downstream MAR3 sites that flank the IGF2 gene. The three-dimensional structure formed as a consequence of these interactions keeps IGF2 in an enclosed domain (Murrell, Heeson et al. 2004; Kurukuti, Tiwari et al. 2006). In human cells, the ICR interacts with a CTCF Downstream Site (CTCF DS) located downstream of the shared enhancer; this interaction creates a loop that encloses the enhancer (Nativio, Wendt et al. 2009). In both mouse and human maternal chromosomes the enhancer is unable to interact with the IGF2 promoter. In mouse, the 142^* allele has a mutated CTCF binding site in the ICR, and in mice that inherit this allele from their mother, the ICR interacts with DMR2 instead of DMR1 and MAR3; this is accompanied by biallelic expression of IGF2. On the other hand, there is no effect if 142^* is paternally inherited (Pant, Mariano et al. 2003; Kurukuti, Tiwari et al. 2006). On the paternal chromosome, the ICR is methylated and CTCF does not bind. In mouse cells the ICR now interacts with the DMR2 site located downstream of IGF2 and allows the enhancer to interact with the IGF2 promoter (Murrell, Heeson et al. 2004; Kurukuti, Tiwari et al. 2006) whereas in human cells the ICR cannot interact with the CTCF DS (Nativio, Wendt et al. 2009). These results suggest that imprinting of the H19/IGF2 locus is dependent on a specific three-dimensional organization of the region mediated by interactions between CTCF sites. Other studies suggest that CTCF-mediated interactions are not limited to the locus but also extend to CTCF sites located in other chromosomes. Using 3C or 4C, the H19 ICR on chromosome 7 has been shown to interact with genes on different chromosomes, such as WSB1/NF1 on chromosome11, ABCG2 on chromosome 6 and OSBPL1a on chromosome 18 (Ling, Li et al. 2006; Zhao, Tavoosidana et al. 2006). The majority of these inter-chromosomal interactions take place primarily with the maternally-inherited H19 ICR. Knockdown of CTCF or mutation of CTCF sites on the H19 ICR in the maternal chromosome abrogates these associations, suggesting that these interactions are CTCF dependent (Ling, Li et al. 2006; Zhao, Tavoosidana et al. 2006). Interestingly, imprinted loci are strongly overrepresented among the regions involved in inter-chromosome interactions with the H19 ICR. The clustering of these imprinting genes by inter-chromosome interactions, which is termed ‘imprinting interactome’, may facilitate the regulation of these genes in trans. When the CTCF site is mutated in the maternal H19 ICR, the interactions are abrogated and the transcription of these imprinted genes is also altered; the paternal expression of the WSB1/NF1 gene is reduced by 50-fold, expression of Impact is reduced by 50% and transcription of OSBPL1a is increased by 2.5-fold (Ling, Li et al. 2006; Zhao, Tavoosidana et al. 2006). In addition, the asynchronous replication timing of these genes in spermatogonia is switched from late to early (Sandhu, Shi et al. 2009). Therefore, CTCF mediated inter-chromosomal interactions are important for the nonallelic regulation of the epigenetic status of multiple genes in trans. One important advantage for such regulation is that the interactions can be reprogrammed during germline development when the epigenetic states of imprinted domains are reprogrammed. In the testis, when the maternal allele is turned to the paternal mode, the interactome can be observed in spermatogonia, but not in spermatocytes and round spermatids where the reprogramming is complete (Sandhu, Shi et al. 2009).

Another interesting example of CTCF-mediated inter-chromosome interactions takes place during X chromosome pairing, which is important for X inactivation. To achieve the mutually exclusive designation of active X (Xa) and inactive X (Xi), it is necessary for the two X chromosomes to communicate in trans through homologous pairing. Pairing depends on a 15-kb region within the Tsix and Xite loci. ChIP experiments show that CTCF binds both Tsix and Xite elements in female embryonic stem cells. CTCF knockdown reduces the frequency of X-X pairing to background levels in wild type embryonic stem cells and embryoid bodies (Xu, Donohoe et al. 2007). It is not clear whether CTCF mediates the pairing directly or by recruiting other factors; Oct4, which can associate with CTCF, is also needed for this pairing (Donohoe, Silva et al. 2009).

It is possible that CTCF-mediated interactions have a broader role than just regulation of transcription. One interesting example of the diverse roles that CTCF may play in nuclear biology is that of V(D)J recombination, a process that is highly regulated during B cell development. IgH rearrangement in pro-B cells begins with D_H to J_H rearrangement followed by rearrangement of a V_H gene segment to D_HJ_H. The complexity of the mechanisms controlling V_H to D_HJ_H rearrangement is in part due to the organization of the genes involved in the process. In mice, there are more than 100 V_H genes spanning a 2.5Mb region; the 96 V genes cover 3.1 Mb whereas the J genes occupy a 2 kb region. How do all the V genes access the small J cluster in the Igh and Ig loci? It has been recently shown that CTCF is present across the Igh, Ig, and Ig loci and plays a critical role in bringing these different loci together in the nucleus and that the cohesin subunit Rad21, which interacts with CTCF in a cell lineage-specific manner, plays a regulatory role in the process (Degner, Wong et al. 2009). Although CTCF does not display obvious cell lineage-specific binding as does Rad21, it is indispensable for the cell-lineage specific contraction of the Igh locus. Preliminary results for 3D-FISH indicate that Igh locus contraction is decreased in pro-B cells treated with CTCF shRNA (Degner-Leisso and Feeney 2010).

Taken together, these results suggest that CTCF can mediate long range intra- and inter-chromosome interactions at different loci throughout the genome either by interactions with other CTCF sites or association with other factors. The result of these interactions is a specific three-dimensional arrangement of the chromatin that can have different effects on chromosome biology, depending on the nature of the sequences brought together by CTCF.

MECHANISMS OF REGULATION OF INSULATOR FUNCTION

If insulators mediate inter- or intra-chromosomal interactions that result in the formation of chromatin loops, which in turn may be attached to the nuclear matrix, it is possible that the resulting structures determine a particular pattern of nuclear organization that may be important for gene expression. For example, it is possible that as cells differentiate, insulator-mediated changes in nuclear organization precede or accompany cell differentiation and may be crucial in the establishment and/or maintenance of specific patterns of gene expression. If this is the case, cells must possess mechanisms to regulate insulator activity in order to establish distinct nuclear architectures that are cell fate-specific.

Evidence for the existence of mechanisms to control insulator function comes in part from genome-wide studies of insulator protein localization in Drosophila cell lines of different tissue origin. Studies in Kc cells, which have a neural origin, indicate that there are 3747 Su(Hw), 2266 dCTCF, 2995 BEAF and 5272 CP190 sites where these proteins are present in the genome. Of these, 47% of Su(Hw), 62% of dCTCF and 71% of BEAF sites colocalize with CP190 sites (Bushey, Ramos et al. 2009). Since CP190 is required for insulator function, this observation suggests that cells may control the activity of these various insulators by regulating the recruitment of CP190. In addition, comparison of the genomic location of different insulator proteins in Kc and Mbn2 cells (a hematopoietic cell line), reveals that while many sites are constant, a fraction of the localization sites for each of the four insulator proteins is different between the two lines. For example, 18% of Su(Hw) sites in Kc cells and 5% of Su(Hw) sites in Mbn2 cells are cell type-specific. This is also the case for dCTCF, for which 18% of sites in Kc cells and 37% in Mbn2 cells are cell-unique, whereas the number of cell type-specific BEAF sites is 11% in Kc cells and 11% in Mbn2 cells. In the case of CP190, which is found at all three insulator subclasses, 17% of sites present in Kc cells and 14% in Mbn2 cells were found to be cell type-specific (Bushey, Ramos et al. 2009). These results suggest that cells may regulate insulator activity by controlling the recruitment of the DNA binding proteins to their target sites in the genome in addition to controlling the recruitment of CP190.

Several proteins have been characterized in Drosophila that may play a role in regulating insulator function. dTopors, in addition to serving as an attachment point for insulators to the nuclear lamina, it has E3 ubiquitin ligase activity. This activity is required for proper insulator function. Its substrate has not been clearly identified but Su(Hw) is a likely candidate, since over-expression of dTopors enzymatic activity reverses the effect of mod(mdg4) mutations on the ability of Su(Hw) to interact with chromatin (Capelson and Corces 2005). In addition, modification of Mod(mdg4)2.2 and CP190 by sumoylation inhibits insulator function. Disruption of the SUMO conjugation pathway improves the enhancer-blocking function of a partially active insulator, indicating that SUMO modification acts to negatively regulate the activity of the gypsy insulator. Sumoylation does not affect the ability of CP190 or Mod(mdg4)2.2 to bind chromatin (Capelson and Corces 2006). Interestingly, dTopors inhibits sumoylation of Mod(mdg4)2.2 and CP190. Therefore, this protein may have a double effect on insulator function by ubiquitinating some insulator components and inhibiting the sumoylation of others. A second candidate protein with a possible role in regulating insulator function is the Rm62 RNA helicase. Insulator activity decreases in the presence of mutations in components of the RNAi machinery; insulator function is restored by mutations in Rm62. These observations have led to a model suggesting that insulator bodies contain RNA whose synthesis requires RNAi proteins. Rm62 may interact with this RNA to decrease insulator function (Lei and Corces 2006).

dTopors and Rm62 have been only shown to affect the function of the gypsy/Su(Hw) insulator but their potential role in regulating the activity of other insulator subclasses has not been tested. There are currently no characterized mechanisms that control the activity of the dCTCF insulator. O-glycosylation of BEAF can be detected in Drosophila embryonic cells in a domain of the protein that is required for association with the nuclear matrix; however, it is not clear whether glycosylation is required for scs’ insulator function (Pathak, Rangaraj et al. 2007). A second possible candidate to regulate the BEAF insulator is the DREF protein. DREF has been characterized as a transcription factor that shares binding sites with BEAF 32. It is possible that DREF regulates BEAF binding through competition for the same DNA sequences (Hart, Cuvier et al. 1999).

In vertebrates, the CTCF insulator can also be regulated at different levels. As in Drosophila, one strategy is to control the interaction between CTCF and its DNA target sequence. The best understood mechanism to control CTCF occupancy is through DNA methylation on CpG dinucleotides within and around the CTCF binding site. This has been well studied in the H19/IGF2 locus, where CTCF only binds to the unmethylated maternal allele but does not bind to the methylated paternal allele (Bell and Felsenfeld 2000; Hark, Schoenherr et al. 2000). CTCF binding to methylation-free sites is also documented for other genes, including both imprinted and non-imprinted loci, like RASGRF1, KvDMR1, GRB10, INK/ARF and DM1 (Filippova, Thienes et al. 2001; Hikichi, Kohda et al. 2003; Yoon, Herman et al. 2005; Fitzpatrick, Pugacheva et al. 2007; Rodriguez, Borgel et al. 2010). Therefore, regulation of CTCF binding via DNA methylation may be a general strategy to control the function of this insulator. In addition, CTCF function can be regulated by post translation modification. CTCF can be phosphorylated at the C-terminus by the protein kinase CK2. Mutation of the phosphorylated residue in CTCF does not alter its nuclear localization or DNA binding in vitro, but enhances repression of the C-MYC promoters. A phospho-mimetic mutant behaves in the opposite fashion (Klenova, Chernukhin et al. 2001; El-Kady and Klenova 2005). Thus, phosphorylation may affect the interaction of CTCF with other proteins and modulate its activity.

CTCF can also be covalently modified by poly(ADP-ribosyl)ation. CTCF interacts with Poly(ADP-ribose) polymerase (PARP)-1 both in vivo and in vitro (Guastafierro, Cecchinelli et al. 2008). Both proteins colocalize at the IGF2/H19 ICR and overlap genome wide at more than 140 CTCF target sites in the mouse genome, including both imprinted and non-imprinted loci (Yu, Ginjala et al. 2004). Poly(ADP-ribosyl)ation does not affect the ability of CTCF to bind DNA but it is required for its insulator function. For example, mutation of the poly(ADP-ribosyl)ation site in CTCF or treatment with the PARP inhibitor ABA, can compromise imprinting of the H19/IGF2 locus, resulting in biallelic expression of IGF2, but CTCF binding to the H19/IGF2 ICR is not affected (Yu, Ginjala et al. 2004; Farrar, Rai et al. 2010). In addition to Poly(ADP-ribosyl)ation of CTCF, PARP-1 can also modify itself. The modified PARP-1 can inhibit DNMT1 activity and control DNA methylation patterns (Guastafierro, Cecchinelli et al. 2008). This may in turn control the binding of CTCF. Thus poly(ADP-ribosyl)ation may affect the function of CTCF by modulating its interaction with DNA and with other proteins.

Another mechanism to regulate insulator function that is shared by Drosophila and vertebrates is the use of homologous proteins that can compete for DNA binding. Like DREF and BEAF in Drosophila, there is also a homologue of CTCF in vertebrates named CTCFL or BORIS. These two proteins share homology of the central 12 zinc finger DNA binding domain and they can recognize the same DNA binding sequences (Loukinov, Pugacheva et al. 2002). BORIS is normally present only in the testis during germ-line development where it can stimulate the histone-methyltransferase activity of PRMT7 and contribute to the efficient DNA methylation of BORIS/CTCF sites in male germ line (Jelinic, Stehle et al. 2006). Therefore, BORIS can play a role in writing the epigenetic marks that will be read by CTCF. Usually, BORIS is expressed in a mutually exclusive manner with CTCF and only influences CTCF function by its effects in the germ line (Loukinov, Pugacheva et al. 2002). However, BORIS can be aberrantly expressed under particular conditions; for example, reduction of CTCF in normally BORIS-negative human fibroblasts results in derepression of BORIS expression (Renaud, Pugacheva et al. 2007). Abnormally expressed BORIS may affect the activity of CTCF by various mechanisms. First, BORIS can compete with CTCF for binding sites and lead to alterations in gene expression; although BORIS shares the zinc finger domain with CTCF, the two proteins are different in the N and C terminal regions that constitute approximately two thirds of the full length amino acid sequences of these proteins and could recruit different functional partners. For example, the transcriptional factor Sp1 can interact with BORIS but not CTCF (Hong, Kang et al. 2005; Kang, Hong et al. 2007). Second, the two proteins differ in their ability to bind methylated DNA. In contrast to CTCF, the binding of BORIS to its target site is methylation-independent in vitro and methylation-preferential in vivo at the H19/IGF2 ICR (Nguyen, Cui et al. 2008). As a consequence, BORIS could bind to methylated CTCF/BORIS sites, where CTCF is not normally present, and affect the expression of adjacent genes. Third, aberrantly expressed BORIS together with ubiquitously present PRMT7 might change the methylation status of previously non-methylated sites and evict CTCF. The binding of BORIS and the different transcription factors recruited to genes harbouring the CTCF/BORIS sites could lead to totally different expression patterns from those observed when only CTCF is present in the cell (Hong, Kang et al. 2005). Not surprisingly then, aberrant expression of BORIS can lead to disease and cancer as we discuss below.

Protein partners of CTCF may be also indispensable for the establishment and/or maintenance of chromosome interaction mediated by this protein and it is therefore plausible that regulation of these factors could change CTCF activity. Knockdown of CHD8 has been shown to impair imprinting of IGF2 (Ishihara, Oshimura et al. 2006). Depletion of cohesin components without affecting CTCF expression can also lead to disruption of chromatin interactions and changes in the expression of genes under CTCF control (Hadjur, Williams et al. 2009; Mishiro, Ishihara et al. 2009; Nativio, Wendt et al. 2009; Hou, Dale et al. 2010). In the Igh locus, although the binding of CTCF is not altered during B cell differentiation, the recruitment of cohesin is cell type specific and appears to be responsible for the hypothesised role of CTCF in V(D)J recombination (Degner, Wong et al. 2009). It is therefore possible that control of the expression or post translation modification of CTCF partners will regulate the different cellular activities of CTCF. The complexity of mechanisms by which CTCF activity can be regulated is showcased by a recent report showing that RNA helicase p68 (DDX5) and its associated noncoding RNA, steroid receptor RNA activator (SRA), bind to CTCF, and are both essential in vivo for insulator function at the H19/IGF2 ICR. Contrary to the inhibitory role of the RNA helicase Rm62 on insulator function in Drosophila, p68 is needed to positively regulate insulator function in vertebrate cells. Knockdown of p68 decreases the insulator activity of the H19/IGF2 ICR and increases the expression of IGF2 as previously observed in CTCF knockdown cells. In Drosophila cells, Rm62 interacts with CP190, while in vertebrates p68 is required for the recruitment of cohesin (Yao, Brick et al. 2010). In spite of the opposite effect of the two helicases on insulator function, the conservation of this mechanism between insects and vertebrates is striking.

INSULATORS, CANCER AND DISEASE

CTCF affects the expression of genes such as hTERT or C-MYC that regulate cell cycle processes that are important for cell growth, differentiation and apoptosis. An appropriate balance between these processes is important for normal development, whereas an imbalance can lead to tumor development. hTERT is the human telomerase reverse transcriptase, a catalytic subunit and limiting factor for telomerase activity, which is important during the cell cycle. CTCF binds to sequences of the hTERT gene located in the first two exons and represses its expression. CTCF only binds to hTERT in cells where the gene is not transcribed but not in telomerase-expressing cells. Knockdown of CTCF derepresses hTERT gene expression in normal telomerase-negative cells (Renaud, Loukinov et al. 2005). hTERT is a general cancer susceptibility locus, and its abnormal expression may disturb appropriate control of the cell cycle and induce aberrant cell growth and (Johnatty, Beesley et al. 2010). CTCF also binds constitutively to regulatory elements close to the C-MYC P2 promoter (Gombert and Krumm 2009). Binding of CTCF can affect the expression of C-MYC, although the molecular mechanism underlying this effect is not well understood. Deletion of the two CTCF sites results in reduced expression of the C-MYC gene, while CTCF knockdown leads to an increase in C-MYC protein levels (Torrano, Chernukhin et al. 2005; Gombert and Krumm 2009). The C-MYC protein is important for cells during the transition from proliferation to differentiation. Ectopic expression of CTCF in K562 cells leads to growth retardation and promotion of differentiation associated with increased C-MYC expression, whereas CTCF knockdown significantly inhibits differentiation with reduced C-MYC transcription (Torrano, Chernukhin et al. 2005). Thus, the loss of CTCF function may disturb the balance between cell growth, differentiation and apoptosis due to its effects on the expression of cell growth regulators. This observation has been extended to other genes whose expression depends on CTCF such as the tumor repressor retinoblastoma (Rb) gene. Mutations and deletions of the Rb gene have been associated with a number of inherited malignancies. CTCF binds in vitro and in vivo to the human Rb promoter and this binding is required for Rb expression. When the CTCF binding site is removed or mutated, reporter gene expression decreases (De La Rosa-Velázquez, Rincón-Arano et al. 2007). Some other genes encoding regulators of the cell cycle, such as p19(ARF), p16(INK4a), PLK, BRCA1, p53, p27, and IGF2, are also growth suppressors frequently silenced in cancer whose expression is also controlled by CTCF (Filippova 2007). For example, BRCA1 is expressed in normal cells but it is silenced in some cancer cells. CTCF binds at the promoter region of BRCA1 only in expressing cells but not bind tumor cells in which BRCA1 is silenced (Darci, Debora et al. 2004; Xu, Huo et al. 2010). The INK4B–ARF–INK4A (INK/ARF) locus contains three tumor suppressor genes that are kept silenced by DNA methylation in different types of cancer. The p16(INK4a) tumor suppressor gene is a frequent target of epigenetic inactivation in cancers such as breast, lung, colorectal and multiple myeloma. CTCF bind upstream of the p16(INK4a) promoter and the absence of binding is associated with silencing of p16(INK4a) expression in breast cancer and multiple myeloma cells. Moreover, ablation of CTCF protein function from p16(INK4a)-expressing cells by shRNA results in epigenetic changes in the p16(INK4a) promoter and loss of transcription (Witcher and Emerson 2009; Rodriguez, Borgel et al. 2010). Conditional over-expression of CTCF in B cells enhances expression of p27, p21, p53, and p19(ARF), followed by inhibition of cell growth and induction of apoptosis, while knockdown of CTCF results in inhibition of these genes (Qi, Martensson et al. 2003). These results suggest that loss of CTCF function can lead to the silencing of growth suppressor genes and contribute to carcinogenesis. Whether these effects are due to changes in the three-dimensional organization of the DNA or more local effects on transcription is unclear at this time.

CTCF also plays a role in repeat-associated diseases. The affected genes in myotonic dystrophy (DM1), spinocerebellar ataxia 7 (SCA7), SCA2, dentatorubral-pallidoluysian atrophy (DRPLA) and Huntington’s (HD) contain CTCF sites on one or both sides of the CTG/CAG repeat-containing region (Filippova, Thienes et al. 2001). It has been reported that CTCF can control noncoding transcription of the repeats. For example, at DM1, the two CTCF sites flanking the CTG repeat can function as an insulator to restrict the antisense transcription and constrain the heterochromatin state to the repeats without affecting the nearby DMPK and SIX5 genes, whose silencing contributes to myotonic myopathy, cataracts and cardiac-conduction defects in the disease (Filippova, Thienes et al. 2001; Cho, Thienes et al. 2005). In contrast, cells from affected individuals show a loss of CTCF binding, which is associated with expansion of the repeats, spread of heterochromatin, and regional CpG methylation (Cho, Thienes et al. 2005). Therefore, CTCF can cis-regulate the stability of the trinucleotide repeat. In transgeneic mice carrying a SCA7 genomic fragment with a CTCF binding site, mutation or methylation of the DNA sequence at the binding site can promote repeat expansion (Libby, Hagerman et al. 2008). However, the mechanism by which CTCF affects the stability of the repeats is not clear. A recent report suggests that CTCF can contribute to repeat stability through regulation of DNA replication. CTCF sites located between a replication origin and the repeats can slow or pause the progression of the replication machinery and enable safe passage of the replication fork. Without CTCF, the replication fork proceed in a manner that allows for slippage, hairpin formation, fork reversal and other replication errors (Cleary, Tome et al. 2010). Thus, loss of CTCF flanking the repeats may affect repeat instability. Aberrant methylation of CTCF binding sites may lead to eviction of the protein, repeat expansion and abnormal transcription of the locus.

CTCF can regulate the expression and epigenetic features of imprinted genes both in cis and in trans through intra- or inter-chromosomal interactions as discussed above. Almost all imprinted genes identified to date can be classified as regulators of embryonic growth, placental growth or adult metabolism (Jelinic and Shaw 2007). It is then not surprising that loss of imprinting can lead to various human cancers and disease. In fact, the earliest and most common alteration observed in human cancers is loss of imprinting (LOI), which has been documented in 100% of chronic myeloid leukaemia, 80% of ovarian tumors, 70% of Wilm’s tumors, 66% of colorectal cancer, 56% of Barrett’s oesophagus, 50% of renal-cell carcinomas, 50% of oesophageal cancer, 47–85% of lung adenocarcinoma, and 30% of meningiomas. Many human diseases are also associated with altered expression of imprinted genes, including Beckwith-Wiedemann (BWS), Prader-Willi, Angelman, Silver-Russell, immunodeficiency syndrome (ICF), Rett, Albright hereditary osteodystrophy, and hydatidiform mole (Jelinic and Shaw 2007). Although the role of CTCF on imprinting has only been studied in detail in the H19/IGF2 locus, similar CTCF-associated mechanisms appear to also play a role in imprinting at the RASGRF1, DLK1/GTL2, WSB1/NF1 and KvDMR loci (Yoon, Herman et al. 2005; Ling, Li et al. 2006; Fitzpatrick, Pugacheva et al. 2007). This suggests a widespread function of CTCF in the regulation of genomic imprinting and, as a consequence, the development of imprinting-associated human cancer and other diseases.

Alterations of CTCF function that lead to disease development can have both genetic and epigenetic origins. The CTCF gene maps to the cancer-associated human chromosome locus 16q22.1, which is the smallest overlap region of a variety of deletions found in breast, prostate, ovarian and Wilm’s tumors. Since CTCF null mutations are lethal, most CTCF mutations involved in human disease appear to cluster in the zinc finger domain of the protein. Instead of causing a complete loss of function, mutations in this domain alter the binding ability of CTCF. For example, one of the mutations identified abrogates CTCF binding to a subset of target sites in certain genes involved in the regulation of cell proliferation (C-MYC, ARF, PIM1, PLK, and IGF2) but does not alter binding to other regions of the genome, including the β-globin insulator, the lysozyme silencer, or the APP promoter (Filippova, Qi et al. 2002). The selective loss of CTCF function caused by mutations in the zinc finger region may be tolerated because it does not affect cell viability, but results instead in transformation to a malignant phenotype. Since only selective changes of CTCF function can be tolerated in cells, it is not surprising that CTCF mutations are infrequently discovered in cancer or disease states. Instead, it is possible that mutations leading to changes of CTCF function could affect regulation of CTCF function rather than its structure. This alterations could be epigenetic, gene specific and relatively tolerable. Most CTCF-mediated cancers may occur through aberrant methylation of CTCF binding sites, which results in loss of CTCF binding to the DNA. For example, abnormal methylation patterns of CTCF targets involved in H19/IGF2 imprinting control have been shown to be associated with colorectal, Wilm’s, and bladder tumors (Klenova, Morse et al. 2002). The methylation pattern at the KvDMR locus was found to be altered in a child with Beckwith-Wiedemann syndrome (Fitzpatrick, Pugacheva et al. 2007). This alteration of methylation at CTCF sites has also been observed at non-imprinted genes. For example, in mice lung tumors, CpG sites harbouring CTCF binding sequences are hypermethylated at the INK/ARF locus, resulting in the absence of CTCF binding and reduced expression of the tumor suppressor genes at this locus (Tam, Devereux et al. 2003; Rodriguez, Borgel et al. 2010). In some human tumors, the promoter of the BRCA1 gene is methylated, CTCF is evicted and BRCA1 is not expressed (Darci, Debora et al. 2004; Xu, Huo et al. 2010).

In addition to DNA methylation, other regulatory mechanisms of CTCF function can be altered and lead to cancer and disease. For example, testicular and ovarian tumor cells exhibit the same methylation profile as normal cells, but the BORIS protein, which recognizes the same DNA binding sites as CTCF, is expressed abnormally. Both CTCF and BORIS bind to DNA in cancer cells but BORIS inhibits the function of CTCF and leads to expression of hTERT (Renaud, Loukinov et al. 2010). Telomerase activity is not detectable in most somatic cells of adult humans but is found in highly proliferative cells, such as germ cells and stem cells, and 85–95% of cancers. The expression of BORIS in normal cells is sufficient to allow hTERT transcription and to extend their lifespan in vitro (Renaud, Loukinov et al. 2010). Reciprocal binding of CTCF and BORIS has also been observed at the NY-ESO-1 promoter and leads to derepression of this cancer-testis gene in lung tumors (Hong, Kang et al. 2005). BORIS is also aberrantly expressed in 71 % (41 of 58 cases) of breast tumors. High levels of BORIS correlate with high levels of progesterone receptor (PR) and oestrogen receptor (ER). The link between BORIS and PR/ER was further confirmed by the ability of BORIS to activate the promoters of the PR and ER genes in reporter gene assays (D’Arcy, Pore et al. 2008). Defects in other regulatory mechanisms that alter CTCF function can also lead to cancer. For example, alterations in the poly(ADP-ribosylation) pathway results in the absence of CTCF PARlation and reduction of p16(INK4a) and RASSF1A expression (Witcher and Emerson 2009).

Understanding of the different mechanism that alter CTCF function has opened new possibilities in the design of treatments for cancer or other diseases. For example, Trichostatin A (TSA), an inhibitor of histone deacetylase activity, is a well-known antitumor agent that effectively and selectively induces arrest of tumor growth and apoptosis. hTERT appears to be one of the primary targets for TSA-induced apoptosis in cancer cells. TSA induces demethylation of CpGs present in the binding site of CTCF on the hTERT promoter, leading to repression of hTERT (Choi, Min et al. 2010). CTCF can contribute to the regulation of a variety of genes whose proper expression is required for normal cell differentiation. Both genetic and epigenetic changes of CTCF function can lead to the miss-expression of these genes, resulting in the development of a malignant phenotype. Understanding the mechanisms by which CTCF function can be regulated may result in the development of novel methods to treat cancer and other diseases.

FUTURE QUESTIONS

Chromatin insulators are important regulatory sequences present in the genome of most eukaryotes. Although they are defined experimentally based on their ability to affect enhancer-promoter interactions and interfere with the spreading of repressive signals from heterochromatin, their role appears to be more general. Intra- and inter-chromosomal interactions mediated by insulator proteins may establish a web of contacts between individual insulator sites that give rise to specific patterns of nuclear organization. Insulator-mediated nuclear structures may be regulatable by controlling the interaction between insulator DNA binding proteins and their cognate sequences. In addition, recruitment of insulator components involved in mediating inter-insulator interactions may represent a second level of controlling the function of these elements. These two levels of regulation may be the result of specific covalent modifications of insulator proteins. The specific outcome of inter-insulator interactions is a consequence of the location of the particular insulator sequences with respect to specific genome features, and interference with enhancer-promoter interactions may be just one of these outcomes. It is unclear at this time whether the three-dimensional pattern of nuclear organization created by the insulator interactome is different in various cell types and whether this organization carries epigenetic information. Most studies to date have concentrated on the analysis of alterations of insulator function at the local level. These studies suggest that aberrant insulator function can lead to alterations in gene expression and, when the affected genes are involved in cell growth processes, to cancer. Nevertheless, it is possible that changes in insulator function cause more global effects on transcription. Understanding the role of nuclear organization in gene expression and cell differentiation remains the main issue for future investigation.

Figure 2.

Mechanisms by which CTCF affects gene expression by mediating intra- and inter-chromosomal interactions. A. Anchoring loops via direct attachment to subnuclear structures such as the nucleolus and/or nuclear matrix. B. Monoallelic gene expression via allele-specific contacts between multiple imprinted regulatory elements. C. Transcriptional regulation via contacts between intergenic enhancers and promoter-proximal regulatory elements. D. Global nuclear organization via demarcation of lamina-associated domains (LADs). E. X chromosome inactivation or monoallelic gene expression via inter-chromosomal contacts between regulatory elements in trans.

Figure 3.

Similarities in the regulatory mechanisms of insulator function in Drosophila and vertebrates. A. Regulation of insulator activity in Drosophila. The left panel shows two active insulators coming together to make a functional loop and the right panel displays two inactive insulators unable to form a loop. At an active insulator site, dTopors is present, Rm62/Lip is not present, Su(Hw) is ubiquitinated, Mod(mdg4)2.2 and CP190 are not sumoylated and dTopors serves as a bridge to the nuclear lamina/matrix. At inactive insulator sites, dTopors is absent and Su(Hw) is not ubiquitinated, whereas Mod(mdg4)2.2 and CP190 are sumoylated. Rm62/Lip is present and bound to RNA. Under these conditions, the two insulator sites cannot interact to form a loop. Absence of dTopors also precludes interactions with the nuclear lamina. B. Regulation of insulator activity in vertebrates. The left panel shows two active CTCF insulators coming together to make a functional loop and the right panel displays two inactive CTCF insulators unable to form a loop. At an active insulator site, CTCF is present and poly(ADP-ribosyl)ated and, directly or indirectly, interacts with the nuclear matrix/lamina; in addition, the SRA RNA and p68 are present while the cohesin complex brings together the two DNA molecules forming the base of the loop. At inactive insulator sites, CTCF is either not bound because of the DNA is methylated or it is bound but not modified by poly(ADP-ribosyl)ation; cohesin, SRA RNA and/or p68 are absent. Under these conditions the two insulator sites cannot interact to form a loop.