DNA-binding factor CTCF and long-range gene interactions in V(D)J recombination and oncogene activation

Development of B and T lymphocytes requires proper spatial and temporal control of gene expression. This is achieved through the coordinate binding of lineage-specific transcription factors to cis-regulatory elements. In many developmentally regulated gene clusters (eg, the antigen receptor and cytokine loci), gene regulation is particularly complex and involves cell type-specific interactions over large genomic distances between gene promoters and remote enhancers, silencers, or locus control regions (LCRs).^1,2  Regulation of higher-order chromatin structures is mediated by protein complexes composed of lineage-specific and ubiquitously expressed nuclear factors. It has recently become apparent that the DNA-binding protein CCCTC-binding factor (CTCF) is a leading candidate for the regulation of gene expression at complex loci during lymphoid differentiation.

CTCF is a ubiquitously expressed and highly conserved 11-zinc finger protein that was originally identified as a negative transcriptional regulator of the c-Myc oncogene.³  CTCF has since been implicated in a variety of regulatory functions, including transcriptional activation and repression, enhancer blocking, gene insulation, gene silencing, genomic imprinting, X chromosome inactivation, and long-range chromatin interactions.⁴ 

CTCF occupancy studies in various human and murine cell types, including embryonic stem cells and T lymphocytes, revealed approximately 14 000 to approximately 40 000 CTCF-binding sites, sharing a 15- to 20-bp core consensus sequence.^5–8  Although the genome-wide distribution of these sites strongly correlates with gene density, it suggests a role for CTCF distinct from that of canonical transcription factors: binding patterns do not correlate with transcriptional start sites and do not appear to predict cell-specific gene expression levels. Domains with few or no CTCF-binding sites tend to include clusters of transcriptionally coregulated genes that are often flanked by CTCF-binding sites. Conversely, CTCF-rich regions contain genes displaying extensive alternative promoter usage, including the T-cell receptor (TCR) and immunoglobulin (Ig) gene loci.⁴ 

Many studies have suggested that CTCF is the archetypal vertebrate protein that binds insulator elements. Insulators affect gene expression by blocking inappropriate communication between neighboring regulatory elements (enhancer-blocking activity, when positioned between enhancer and promoter) and by preventing the spread of repressive heterochromatin (barrier function).⁹  Based on studies showing that a 5′ boundary element of the chicken β-globin locus serves as an insulator in erythroid cells and protects against position effect in Drosophila, Chung et al were the first to hypothesize that insulator function may depend on interactions between distant elements that result in looping out of intervening sequences.¹⁰  The first association of CTCF with enhancer blocking activity was proposed on the basis of CTCF-binding to the DNAse hypersensitive site (5′HS4) upstream of the chicken β-globin locus and another insulator sequence at the 3′ end.¹¹  Based on functional analysis in transgenic systems, CTCF has been demonstrated to mediate insulation at several model loci discussed below in detail. In addition, computational analyses showing CTCF binding to highly conserved noncoding elements across 12 mammalian species provided evidence supporting a global role for CTCF as an insulator protein.¹² 

CTCF drives enhancer-blocking insulation through long-range chromatin interactions, in particular through 3-dimensional loop formation between distal elements.^4,13 

CTCF has been implicated in genomic imprinting, the epigenetically regulated process that causes genes to be expressed in a parental-origin–specific manner rather than from both chromosomes.¹⁴  In the H19/Igf2 locus, an imprinting control region, containing 4 CTCF-binding sites, is located immediately upstream of the H19 gene and is essential for regulation of imprinted expression of the H19 and Igf2 genes (Figure 1A). On the maternally inherited allele, the imprinting control region is unmethylated, allowing for CTCF binding and the formation of a tightly coiled chromatin loop insulating the Igf2 gene and preventing its promoter to access the enhancers downstream of H19. On the paternal allele, however, the imprinting control region is methylated, which abrogates CTCF binding, resulting in functional interactions between the Igf2 promoters and the enhancer elements (Figure 1A). Moreover, loss of CTCF binding is also associated with H19 promoter methylation leading to H19 gene silencing on the paternal allele.

CTCF-mediated contacts are associated with transcriptional activation of genes in the extensively characterized β-globin locus in erythroid cells.¹⁵  Developmentally regulated expression of individual globin genes is controlled by many cis-acting elements, the most important of which are located in the upstream LCR (Figure 1B). CTCF binds to 3 DNAse hypersensitive sites upstream in the LCR and 1 downstream (3′HS1) of the mouse β-globin locus in a cell type–specific manner.¹⁶  During erythroid differentiation, CTCF-bound regulatory sequences throughout the locus come into close spatial proximity to form an “active chromatin hub,” as was shown by chromosome conformation capture (3C) technology detecting long-range DNA interactions.^15–17  Formation of this “active chromatin hub” allows stage-specific interaction of active globin genes with the LCR and concomitant looping-out of transcriptionally silent isoforms (Figure 1B). Conditional CTCF deletion or mutation of the 3′HS1 CTCF-binding site resulted in disruption of CTCF contacts in erythroid progenitors but surprisingly had no effects on kinetics or levels of globin expression during erythroid differentiation.¹⁶  Because of the high density of CTCF-binding sites in the β-globin locus, this finding is explained by redundancy of chromatin contacts, which may not be abrogated by mutations in a single regulatory element. Genetic disruption of CTCF-binding to 3′HS1 also did not affect transcription of the neighboring olfactory receptor genes.¹⁶  Collectively, these findings indicate that correct globin gene expression can be achieved by different CTCF-dependent loops and thus by different chromatin conformations.

CTCF organizes the MHC class II locus into a conformation that rearranges in the presence of the crucial non-DNA–binding transcriptional coactivator CIITA. In the human locus, CTCF binds to the intergenic enhancer XL9 that coregulates expression of the HLA-DRB1 and HLA-DQA1 genes^18,19  (Figure 1C). Induction of CIITA in non-MHC class II–expressing cells results in the initiation of interactions between XL9 and the promoters upstream of HLA-DRB1 and HLA-DQA. Hereby chromatin loop formation depends on XL9-bound CTCF, CIITA, as well as the transcription factor RFX, which is constitutively bound to regulatory sequences within the HLA-DRB1 and HLA-DQA1 proximal promoters. Therefore, in the human MHC class II locus, CTCF is involved in the control of gene expression through loop formation via hetero-multimerization.

The key transcription factor of T helper-1 (Th1) cells, T-bet, promotes the expression of IFN-γ in part by facilitating CTCF binding in the Ifng locus.²⁰  CTCF acts to establish a Th1 cell–specific Ifng locus architecture, which helps to drive the juxtaposition of T-bet–binding enhancers, the flanking CTCF-binding elements, and CTCF sites in the first exon of the Ifng gene (Figure 1D). Cooperation between CTCF and T-bet is required for robust Th1 cell–specific IFN-γ expression,²⁰  which is also supported by our finding of an approximately 50% reduction in IFN-γ–producing cells when naive CTCF-deficient CD4⁺ T cells were cultured under Th1-polarized conditions.²¹  CTCF may act in a similar fashion in the Th2 cytokine locus containing the genes for IL-4, IL-5, and IL-13. Indeed, Th2 cytokines were strongly reduced in CTCF-deficient T lymphocytes cultured under Th2-polarized conditions, despite proper induction of the key Th2 transcriptional regulators Gata3 and Satb1.²¹ 

Its capacity to form dimers and to participate in multiprotein complexes allows CTCF to organize long-range 3-dimensional looping of genomic regions. Interestingly, genome-wide identification of binding sites for CTCF and for cohesin, the protein complex responsible for sister chromatid cohesion, showed that a large proportion of these sites are shared.^22,23  CTCF enables cohesin recruitment to specific sites, and cohesin is required for CTCF-mediated long-range intrachromosomal interaction (eg, at the H19/Igf2, Ifng, or MHC class II loci).^18,22,24  Because cohesin forms a large ring structure that holds sister chromatids together during mitosis, it is conceivable that cohesin may function to stabilize loops that have been initiated by CTCF.

Several lines of evidence support a role for CTCF as a crucial genome-wide organizer of chromatin architecture: (1) the high number of CTCF-binding sites in the genome, both near transcription start sites, in exons, in introns, and in intergenic regions⁴ ; (2) the essential function of CTCF in the orchestration of long-range chromatin interaction at model loci; (3) CTCF has many binding partners, including transcription factors, chromatin-modifying proteins, the nucleolar protein nucleophosmin, RNA polymerases, and Smad signaling proteins^4,25 ; and (4) the recent genome-wide identification of CTCF-mediated chromatin interactions implicates CTCF in the global organization of chromatin.^26,27  Based on combining CTCF binding and mapping of histone modifications, CTCF often demarcates the boundaries of repressed chromatin associated with nuclear lamina domains.²⁶  Recent 3C-based analyses showed that indeed CTCF-mediated loops separate H3K4 active domains from H3K27me3 repressive domains, which coincide with lamina domains.²⁷  These findings demonstrate that CTCF is involved in the division of the mammalian genome into large, discrete domains that are units of chromosome organization within the nucleus.

Lymphocyte development is a stepwise process involving ordered rearrangement of variable (V), diversity (D), and joining (J) antigen receptor gene segments, also known as V(D)J recombination.^1,28  Because these gene segments are spread over several megabases of the genome, it is obvious that locus compaction and looping are essential for the site-specific DNA recombination reactions that create a diverse repertoire of Ig and TCRs. The question how V(D)J recombination is regulated over great distances is, however, largely unanswered. Various studies over the last year have identified CTCF as a crucial regulator of long-range chromatin interactions in antigen receptor loci.

The lymphoid-specific recombination activation gene endonucleases Rag1 and Rag2 cooperate to induce double-strand DNA breaks at recombination signal sequences flanking the V, D, and J gene segments, which are subsequently joined by the nonhomologous end-joining DNA-repair machinery.^1,28  Cell cycle–dependent regulation of Rag expression during lymphocyte development is critical for preventing Rag-mediated DNA breaks leading to translocation in lymphocytes undergoing active V(D)J recombination.²⁹  B-cell development is initiated in pro-B cells in the bone marrow by sequential D-to-J_H and V-to-DJ_H joining. Successful rearrangement of the Ig heavy chain (Igh) locus results in IgH μ protein deposition on the surface of pre-B cells, together with the preexisting surrogate light chain proteins λ5 and VpreB, as the pre-B cell receptor (pre-BCR) complex. Pre-BCR signaling controls the selective expansion of cells that have a functionally rearranged IgH chain and redirects V(D) recombination activity from the Igh to the Igκ or Igλ light chain loci.^30,31  Successful Igκ/Igλ chain rearrangement results in surface expression of the BCR and additional checkpoints follow to limit autoreactivity.³²  Likewise, in the thymus, V(D)J recombination events in the TCR loci accompany the development of mature T cells, expressing an αβ TCR or, in a minor subset, a γδ TCR on the cell surface.³² 

Given that all Ig and TCR loci are rearranged by a common recombination machinery, an essential level of regulation of the V(D)J recombination process relies on lineage and developmental stage-specific changes in the accessibility of antigen receptor loci for recombination. The “accessibility hypothesis” was proposed based on the initial observation that rearranging gene segments are concomitantly undergoing germline transcription.^28,33  It has been demonstrated that transcription itself can direct recombinase targeting: blockade of transcriptional elongation through the mouse TCR-α locus suppressed Vα to Jα recombination and chromatin remodeling of Jα segments.³⁴  Germline transcription increases recombination of Jα segments located close to an upstream crucial promoter, called TEA, and prevents activation of downstream promoters through transcriptional interference.³⁴  Accessibility is controlled by local cis-regulatory elements (promoters and enhancers) and involves subnuclear relocation, histone acetylation and/or methylation, DNA demethylation, and locus compaction.^1,28  These processes are dependent on various lineage-specific transcription factors that bind to enhancer elements present in the Ig and TCR loci.^35–37  Among these transcription factors, Pax5, EBF, and E2A play a central role in B-cell differentiation, as they also establish a hierarchical regulatory network for specification and commitment to the B-cell fate.³⁸  Next to CTCF, only a few general nuclear factors that modulate antigen receptor loci topology have been identified to date, including YY1 and Ezh2.^35,37,39,40  However, the precise mechanism by which these factors act in concert with lineage-specific transcription factors to regulate chromatin structure at the Ig loci remains unclear.

The Igh locus in the mouse spans an approximately 2.7-Mbp region and is composed of more than 100 V_H gene segments and D_H, J_H, and C_H clusters located downstream. The locus contains 2 important regulatory elements: the intronic enhancer (iEμ), located 3′ of the J_H genes, and the 3′ regulator region (3′RR; Figure 2A).²⁸  The stepwise activation of the locus, reflected by D_H-to-J_H preceding V-to-D_HJ_H rearrangement, is paralleled by germline transcription. First, Iμ transcripts are generated from iEμ, μ0 transcripts from the promoter of the most 3′ D_H segments, and antisense intergenic transcripts throughout the D_H and J_H domains.^28,41,42  After D_H-to-J_H recombination, noncoding sense and antisense transcripts are generated from the V_H genes.^28,33,42,43  Three-dimensional FISH experiments have elucidated why pro-B cells deficient for Pax5, Yy1, or Ikaros only recombine D-proximal V_H genes, suggesting that locus compaction mediated by these proteins is necessary to bring distal V_H genes in close proximity to D_H-J_H.^35,37,39  Spatial distance measurements of genomic markers spanning the entire Igh locus have demonstrated that this locus is organized into compartments that contain clusters of loops or rosette-type structures separated by linkers.⁴⁴  This topology compacts the V_H region and should ensure that V_H gene segments scattered over approximately 2.5 Mbp encounter D_H-J_H segments or enhancers with appropriate frequencies.

Emerging data have revealed CTCF as an attractive candidate for controlling Igh locus topology. CTCF and cohesin bind to multiple sites throughout the V_H region,⁴⁵  implicating CTCF in Igh chromatin loop formation and effective distal V_H-D_HJ_H recombination.^46,47  However, CTCF occupancy in the V_H region is largely unchanged between pro-B and pre-B cells,⁴⁵  making it unlikely that CTCF alone accounts for Igh locus topology changes. Consistently, 3C-based experiments demonstrated that shRNA-mediated CTCF depletion only modestly affected Igh locus contraction.⁴⁸  In agreement with these findings, on mb1-Cre–mediated conditional deletion of the Ctcf gene in the B-cell lineage, Igμ⁺ pre-B cells were still generated and both distal and proximal V_H gene segments were used.⁴⁹ 

Using chromatin immunoprecipitation coupled to microarray analysis (ChIP-chip) technology, Ebert et al have mapped Igh active histone modifications and transcription factor occupancy.⁵⁰  They hereby identified 14 copies of a potential regulatory Pax5-activated intergenic repeat (PAIR) element in the distal Igh region, upstream of V_H3609 genes and interspersed among V_HJ558 genes. These PAIR elements contain Pax5-dependent active chromatin and binding sites for E2A, CTCF, and the cohesin component Rad21. Detailed analysis of 3 PAIR elements showed that immediately upstream of the CTCF-binding sites noncoding antisense transcription is initiated exclusively in pro-B cells. The authors hypothesize that PAIR elements mediate looping of distal V_H genes in pro-B cells, whereby Pax5 orchestrates the formation of a bridging complex that mediates long-range interactions between distal and proximal V_H gene regions. In contrast, on the basis of anti-CTCF Chip-loop assays, Guo et al⁵¹  presented evidence for a CTCF-mediated but iEμ-independent multiloop domain of approximately 500 kb in the 5′ V_HJ558/V_H3609 region containing the PAIR elements, whereby locus contraction may result from interaction between Eμ-bound³⁹  Yy1 and Yy1 sites in the distal V_H region. These interactions may be increased by heterotypic interactions with CTCF sites or by inter-PAIR interactions in the distal V_H region.⁵¹  Future experiments should focus on functional characterization of the PAIR elements, which would require deletion of these proposed regulatory elements to test the effects on V(D)J recombination and V_H repertoire in vivo.

In addition, CTCF-binding sites are present in the intergenic control region I (IGCR1), which is located 3 to 5 kb upstream of the most 5′ D_H element (Figure 2A).^45,52,53  These CTCF-binding sites have enhancer-blocking activity and mark a sharp decrease in Eμ-dependent antisense D_H-J_H transcription.⁵²  Targeted IGCR1 deletion in the mouse leads to D_H-J_H antisense transcripts extending into proximal V_H segments and concomitant aberrant rearrangement of the proximal V_H gene segments in developing thymocytes.^53,54  Importantly, IGCR1 balances proximal versus distal V_H usage by inhibiting germline transcription and rearrangement of D-proximal V_H7183 gene segments (Figure 2A). Moreover, IGCR1 maintains ordered V(D)J recombination by suppressing V_H joining to D segments that are not joined to J_H. Both FISH and 3C-based experiments showed that CTCF-binding sites in IGCR1 are in close spatial proximity with a region immediately downstream of 3′RR with a high density of CTCF-binding sites (Figure 2A).^48,51,55  These CTCF-mediated interactions would create a domain that contains all of the D_H, J_H, and C region elements, as well as Eμ and 3′RR. It therefore seems probable that CTCF binding to IGCR1 prevents chromatin accessibility extending into the independently activated V_H chromatin domains. This notion is supported by the findings that (1) insertion of a V_H gene segment just 5′ of the D_H cluster resulted in loss of ordered and lineage-specific Igh locus rearrangement,⁵⁶  and (2) mice with an approximately 8-kb deletion encompassing the CTCF-binding sites downstream 3′RR showed a 2-fold increase in the usage of V_H7183.⁵⁷ 

In summary, by binding to IGCR1 and downstream of 3′RR, CTCF acts as an insulator that ensures ordered and lineage-specific V(D)J recombination at the Igh locus.

The Igκ locus in the mouse spans approximately 3.2 Mbp and contains approximately 100 V_κ gene segments, a J_κ cluster and a single C_κ region.²⁸ Igκ locus rearrangement and expression are positively regulated by an intronic enhancer (iEκ) located between J_κand C_κ, and 2 enhancers downstream of C_κ (3′Eκ and Ed; Figure 2B). As D elements are lacking, κ light chains can be expressed on a single productive V_κ-to-J_κ recombination event and autoreactive BCR specificities can be replaced by “receptor editing,” the process of ongoing Ig light chain gene recombination.^1,28  Germline κ⁰ transcripts from promoters located upstream of J_κ (κ⁰ 1.1 and κ⁰ 0.8) or V_κ gene segments indicate the accessibility over J_κ and V_κ regions, respectively.⁵⁸  A recombination silencer element in the V_κ-J_κ region, called “silencer in intervening sequence” (SIS), negatively regulates Igκ rearrangement by targeting of the nonrearranging allele to centromeric heterochromatin (Figure 2B).⁵⁹  Deletion of the SIS element resulted in decreased monoallelic Igκ heterochromatin localization, reduced κ⁰ germline transcription, and enhanced proximal V_κ usage.^59,60 

Next to strong CTCF binding at the 5′ and 3′ boundaries and the SIS element,⁴⁹  we identified approximately 60 CTCF-binding sites in the V_κ region (in contrast to the previously reported low density of CTCF occupancy in the Igκ locus⁴⁵ ). CTCF-binding sites were not evenly distributed over the Igκ locus. Detailed analysis of long-range interactions in pre-B cells revealed a correlation between V_κ usage, local density of CTCF-binding sites, and the frequencies of contacts with SIS, iEκ, or 3′Eκ.⁴⁹  Using 3C-seq, we observed that long-range interactions between the SIS-, iEκ-, or 3′Eκ- and V_κ region were not uniformly affected by conditional deletion of the CTCF gene in the B-cell lineage. Loss of CTCF resulted in significantly increased interactions with the most proximal V_κ region containing the V_κ3 family. In line with these findings, CTCF-deficient pre-B cells showed significantly increased V_κ3 germline transcription and recombination.⁴⁹  On the other hand, loss of CTCF resulted in reduced contacts with the 2 most distal V_κ genes, V_κ2-137 and V_κ1-135, which are often rearranged in control but not in CTCF-deficient (pre-)B cells.

Based on these studies,^49,59,60  we proposed that CTCF binding in the SIS element provides enhancer-blocking activity that limits interactions between the κenhancers and the proximal V_κ promoters. CTCF, however, plays a dual role, as, next to the enhancer-blocking activity, it tethers remote V_κ gene promoters to the κenhancers.

The enhancer blocking activity of CTCF at the SIS element in the V_κ-J_κ intergenic region in the Igκ locus^49,60  remarkably parallels the role of CTCF at IGCR1 in the V_H-D intergenic region in the Igh locus.⁵³  Likewise, in thymocytes, CTCF colocalizes with cohesin at the TCR-α regulatory elements TEA just upstream of J_α. Loss of cohesion in double-positive thymocytes led to reduced long-range TEA-Eα interactions and significantly decreased germline transcription and rearrangement over the 3′J_α gene segments.⁶¹  Cohesin and CTCF also colocalize at 5′ boundary sites just downstream of the Eα enhancer, demarcating the TCR-α locus from the neighboring gene Dad1,⁶¹  a region with CTCF-dependent insulator activity.⁶² 

On virtually complete deletion of CTCF protein in the B-cell lineage in our conditionally targeted mice, we found that productive Igh and Igκ rearrangements were still generated, demonstrating that CTCF is not required for V(D)J recombination per se. This is also consistent with the presence of TCR-α– or TCR-β–chain rearrangement in CTCF-deficient thymocytes in vivo.⁶³  The Igh and Igκ loci both contain approximately 60 CTCF-binding sites that influence the genomic architecture of these loci. The identification in the Igh locus of PAIR elements just upstream of distal V_H3609 genes,⁵⁰  together with the observation that many proximal V_H gene segments are located within 100 bp of CTCF-binding sites,⁶⁴  suggests that proximity of a CTCF-binding site may affect the probabilities of individual V_H genes to encounter a D_H-J_H element for recombination. In this aspect, the Igh and Igκ loci are remarkably different: whereas approximately 27% of V_H genes have a CTCF-binding site within 1-kb distance, this is the case for only 2 V_κ genes (Figure 3). Moreover, only approximately 11% of V_H genes but approximately 36% V_κ genes do not have a CTCF-binding site in the intervening sequences up to their 5′ or 3′ neighboring V genes. Thus, although CTCF binds at many sites throughout the Igκ locus, most CTCF occupancy is located between V_κ segments and not adjacent to the V_κ segments.^1,49  It has been suggested that V regions are organized as rosettes by CTCF, whereby CTCF-binding adjacent to a V gene increases its recombination probability.⁶⁴  Because the Igh and Igκ loci show large differences in the proximity of CTCF-binding sites to V genes, in such a model the 2 loci would be very differently organized in 3-dimensional space to provide appropriate access of individual V regions to the proposed recombination center.⁶⁵  Interestingly, it was recently proposed that E2A proteins may modulate Igκ locus topology by acting as anchors in a mechanism similar to that put forward for CTCF in the Igh locus.¹  This hypothesis still needs to be tested but is supported by the remarkable, nonrandom distribution of E2A binding sites across the Igκ locus, in particular within 200 bp of the 5′ or the 3′ end of Vκ regions.^1,46,47 

Many leukemias an lymphomas contain chromosomal translocation involving Ig or TCR loci whereby the strong enhancers in these loci are placed in a new genomic context. Because CTCF and cohesin have been implicated in interchromosomal interactions, they probably contribute to physical proximity of particular translocation targets in lymphocytes. For example, in many Burkitt B-cell lymphomas, translocations place the MYC oncogene up to approximately 200 kb upstream of the IgH 3′RR (Figure 4A). In normal B cells, IgH and MYC are preferentially positioned in close spatial proximity relative to each other and are present in the same transcription factories.^66–68  Recently developed new genome-wide methods that measure primary chromosomal rearrangements in the absence of growth selection have shown that translocations between IGH and MYC and all other genes are indeed directly related to their contact frequencies.^69–72  However, 3C-based measurements showed that approximately 30% of all genes interact with IGH at equal or higher frequency than MYC.⁷¹  Therefore, growth selection and AID-mediated DNA damage rather than a high contact frequency between IGH and the translocations partners account for their high rate of translocation.

An oncogenic role of the IGH locus 3′RR has been defined based on long-range activation of translocated MYC genes.⁷³  In this context, an important function of CTCF is to restrict enhancer interaction to Ig loci⁴⁹  (Figure 4). Obviously, CTCF-mediated restriction might be lost when strong Ig enhancers appear in an aberrant genomic context in chromosomal translocations, which may then lead to activation of oncogenes. In addition, CTCF sites clustered around Ig enhancers may affect local chromatin conformation to promote novel promoter-enhancer interactions. The genes encoding c-Myb and c-Myc, 2 essential hematopoietic regulators and oncogenic transcription factors, harbor multiple CTCF-binding sites in their vicinity (Figure 4B). Indeed, CTCF was originally discovered as a transcriptional repressor of c-Myc.⁷⁴  Both genes are regulated at the level of transcriptional elongation by attenuator elements located in their first introns,^75,76  close to CTCF-binding sites^77,78  (Figure 4B). The exact role played by CTCF at these sites is still unclear, but the recent demonstration that CTCF regulates RNA polymerase II processing⁷⁹  suggests that CTCF could directly affect elongation. Loss of intronic CTCF binding resulting from translocations could disrupt normal locus structure and transcriptional control, allowing the juxtaposed IGH enhancers to drive MYC expression. Because MYB seems to be regulated by a similar mechanism, this may also occur within the MYB locus, when translocated to the TCR-β locus in acute T-cell leukemia.⁸⁰ 

Interestingly, based on published CTCF-binding sites,⁸¹  translocation breakpoints at the CRLF2 and BCL3 loci found in several types of B-cell malignancies are both confined to an approximately 25-kb region between the transcriptional start site and an upstream CTCF element.^82,83  Translocations therefore result in loss of the upstream CTCF-binding site, which may allow strong Ig enhancers to interact and activate the CRLF2 and BCL3 promoters. In contrast, many other loci frequently involved in leukemic translocations (eg, BCL2 and MYC),^84,85  show more complex breakpoint patterns and harbor numerous CTCF-binding sites (Figure 4B).

Expression of BCL6 in diffuse large B cell lymphomas is maintained through hypermethylation of intragenic CpG islands in its first intron.⁸⁶  Hypermethylation, which has been observed in a wide range of cancer cells, prevents CTCF-mediated silencing of BCL6 because CTCF binds these CpG islands in a methylation-sensitive fashion. Furthermore, long-range regulatory regions approximately 150 to 260 kb upstream have been implicated in BCL6 regulation,⁸⁷  possibly representing another layer of control in which CTCF might play an important role (Figure 4B).

In conclusion, based on studies in the Igh and Igκ, it has become clear that CTCF-binding provides enhancer-blocking activity that restricts Ig enhancer activity to the Ig loci. Nevertheless, CTCF also serves to bring remote V gene promoters into close proximity of Ig enhancers. It will be challenging to identify the exact mechanism by which CTCF regulates V gene choice because individual developing B cells are expected to manifest distinct and highly dynamic interactions between V regions and Ig enhancers. Moreover, the high density of CTCF-binding sites and their anticipated partial redundancy may complicate the interpretation of in vivo targeted deletion experiments.

Chromosomal translocations present in lymphoid malignancies will (1) remove CTCF sites close to the oncogenes, (2) bring in enhancer-associated CTCF sites, or (3) dramatically alter the spacing between CTCF sites and their cognate genes. These events are likely to influence local communication between cis-regulatory elements by imposing new structural constraints and thereby increasing oncogene expression. Additional epigenetic changes, including DNA methylation and histone modifications, might also interfere with CTCF function. For example, the BCL6 oncogene is involved in chromosomal translocations through breakpoints frequently found near the intronic CpG islands where CTCF binds in a methylation-sensitive manner.^86,88 

In a normal context, oncogene expression is tightly regulated in response to extracellular signaling events. It is an intriguing question why these signals are no longer able to efficiently regulate expression of the translocation products. Trying to address this issue might reveal additional roles for CTCF in the integration of signaling to chromatin. Future research should clarify the complex roles played by CTCF in the wiring of cis-DNA regulatory elements and in organizing networks of genomic interactions. The field of spatial chromatin organization research is currently exploding with the development of new exciting technologies, allowing high-throughput analysis of chromatin organization of the whole genome.^27,89,90  As the resolution of these technologies increases, they might rapidly become valuable tools to reveal the complex 3-dimensional changes in cancer genomes and unravel their relationships with critical regulatory factors, including CTCF.

This work was supported in part by Fundação para a Ciência e a Tecnologia (C.R.d.A.), Royal Netherlands Academy of Arts and Sciences (R.S.), the Center of Biomedical Genetics, the Cancer Genomics Center, and the EuTRACC Consortium (E.S. and R.S.).

Contribution: C.R.d.A. wrote the manuscript and designed figures; R.S. analyzed and interpreted data, wrote the manuscript, and designed figures; S.T. analyzed data; E.S. wrote the manuscript; and R.W.H. designed figures and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

The current affiliation for C.R.d.A. is Genome Function Group, Medical Research Council Clinical Sciences Centre, Imperial College School of Medicine, London, United Kingdom.

Correspondence: Rudi W. Hendriks, Department of Pulmonary Medicine, Room Ee2251a, Erasmus MC Rotterdam, PO Box 2040, NL 3000 CA Rotterdam, The Netherlands; e-mail: r.hendriks@erasmusmc.nl.