Introduction

The original proposal of the gene-gating hypothesis more than 30 years ago1 outlined the principle that the juxtaposition of inducible genes to the nuclear pore would facilitate their reactivation and the rapid nuclear export of derived mRNAs upon repeated stimuli2,3. While such events have subsequently been well documented in both yeast4 and flies5, they have been much more elusive in mammalian cells. This situation likely reflects that the nuclear volumes of mammalian cells are vastly larger6 to compound direct comparisons with yeast, for example. We have earlier provided indirect evidence that the colorectal super-enhancer recruited the transcriptionally active MYC alleles to nuclear pores from intra-nucleoplasmic positions to increase the rate of nuclear export of MYC mRNAs7. This process increased the overall MYC expression level by enabling the exported mRNA to escape from the more rapid decay kinetics in the nucleus compared to the cytoplasm7. Importantly, the facilitated nuclear export rate of MYC mRNAs was the sole parameter responsible for the increased MYC expression in colon cancer cells (HCT-116) compared to primary human colon epithelial cells. Although the suggested link of this process to the distal (>500 kb) OSE7 is congruent with a version of the gene-gating phenomenon also in human cancer cells, the underlying mechanisms are not known. To deal with these shortcomings, we focus here on the potential regulatory cis elements within the colorectal OSE. Among the candidates, the CTCFBS, positioned within an eRNA gene, CCAT1, stood out for a number of reasons. First, CTCF has been described as a master regulator of the genome, attributed to a range of pivotal processes8. These include the organization of chromatin insulators, boundaries9, enhancer–gene interactions via the cohesin complex10, as well as the mediation of the rhythmic recruitment of active circadian genes to the lamina for subsequent repression11. Second, it binds to a region within the OSE that shows physical proximity with MYC in colon cancer cells7, and third, CTCF is linked with long-range regulation of MYC expression12,13. Specifically, it has been argued that OSE-MYC interactions are facilitated by the CCAT1 eRNA when complexed with CTCF14,15.

To assess if this CCAT1-specific CTCFBS plays a functional role in the gene-gating process, we have edited 8 bps within its site following a previously used strategy16,17,18,19, by using the CRISPR technology. The comparison of two cell clones (D3 and E4) carrying the mutated OSE allele with the wild type (WT) parental HCT-116 cells provides genetic evidence that the CTCFBS is necessary for the WNT-regulated increase in the nuclear export rate of MYC mRNAs. This CTCF function reflects, moreover, its ability to direct the trafficking of the active MYC allele to the nuclear periphery in a stepwise manner. Thus, it is responsible for the WNT-dependent activation of CCAT1 eRNA that we could link to the repositioning of MYC from the nuclear interior to a location juxtaposed to the periphery. The next step, promoting MYC to migrate from this position to the nuclear periphery/pore (encompassing <1 μm), is triggered by the WNT-regulated interaction between CTCF and AHCTF1—a key nucleoporin essential for the gating process7—potentially facilitated by ß-catenin. Thus, the mutant OSE allele is, in contrast to the WT allele, unable to efficiently recruit AHCTF1, resulting in a significantly reduced representation of both the OSE and MYC regions at nuclear periphery/pore-proximal positions in mutant cells. However, the lack of a functional CTCFBS does not affect the ability of the OSE to physically interact with MYC, documenting a function of the OSE-specific CTCFBS, which is independent of the well-known ability of CTCF to connect distant chromatin fibers9,17,18. Importantly, the WT HCT-116 cells display a proliferative advantage over the cells carrying the edited OSE alleles due to reduced MYC expression levels in the mutant cells. Since the gating process is specific for colon cancer cells7, these findings open up potential strategies to both diagnose cancer cells at risk of developing pathological gene-gating and antagonize pathological MYC expression without affecting the normal MYC function. Finally, the designation of the gating function to the versatile CTCF20 raises the question whether this feature is not restricted to the OSE, but applies genome wide.

Results

CTCF binding to an oncogenic super-enhancer distal to MYC confers an excessive growth advantage to colon cancer cells

While CTCF binding to a site within the OSE-specific eRNA gene (CCAT1) is prominent in HCT-116 cells, it was absent from the corresponding region in normal colon epithelial cells (Fig. 1a, b). Since the normal cell counterparts (human colon epithelial cells, HCECs) lack functional MYC gating7, this correlation encouraged us to target this particular CTCFBS using the CRISPR technique to generate HCT-116 cells with edited OSE alleles that no longer bind CTCF. We chose to target the C/G sites within the CTCFBS (the edited sequences are marked in gray in Fig. 1a), similar to a strategy that successfully identified in vivo a chromatin insulator in the mouse17,21. Two clones (D3 and E4) were generated, with identical mutations within the CCAT1-specific CTCFBS in all alleles (Fig. 1c). To assess the genome-wide off-target effects of CRISPR–Cas9 editing, we sequenced the genomes of the WT, D3 and E4 cells and adapted a genome wide off-target detection pipeline modified from GOTI22 (Supplementary Fig. 1a). Allowing for a filter of 100% sequence similarity for a blast word size ≥12 nucleotides, we found no sgRNA-associated target for the D3 and E4 cell clones in the blast results. We further analyzed the potential off-targets using Cas-OFFinder (see Methods) allowing for 5 mismatches in the settings to directly align the sgRNA with the genome. No overlap was found between the variant location and potential off-target sites to indicate that the editing process did not as such generate genome wide variants. However, we did find a total of 91 indels and 10 SNVs that were shared between the D3 and E4 cell clones. Two of the indels that were common to both cell clones mapped to the vicinity of CTCFBSs on chromosome 8 and 22. ChIP-seq analyses showed that these changes did not antagonize CTCF binding (Supplementary Fig. 1b). Since neither of these regions interacted with MYC or the OSE, they are unlikely to be involved in the MYC gating process.

Fig. 1: The CTCFBS within the OSE-specific eRNA gene (CCAT1) confers a proliferative advantage to the HCT-116 cells.

a The position of the CCAT1-specific CTCFBS within the OSE is indicated (black arrow). The core binding sequence was modified at 8 bases, as marked by gray boxes in the panel, by CRISPR editing. Orange boxes depict enhancer regions. b ChIP analyses of the occupancy of the CTCFBS within the OSE. The MYC promoter and the H19 ICR were used as positive controls. Neg. CTCF negative site (see Methods). c DNA sequences in the edited CTCFBS in comparison to WT HCT-116 cells. d ChIP-seq profiles of CTCF binding patterns to a region encompassing the OSE (upper) and MYC (lower) regions. The ChIP-seq data, which is visualized in relation to a genome browser snapshot, was normalized from three independent experiments for WT HCT-116, D3, and E4 cells. The boxed motif, representing the CCAT1 gene region, is enlarged to identify the edited CTCFBS within its intron. The arrows identify the edited CCAT1-specific CTCFBS. e Co-cultures of wild type and mutant HCT-116 cells harvested at the indicated time points, followed by qPCR analyses of the proportion of WT and mutant CTCFBSs, respectively. f The effect of BC21 on the relative growth rate of the WT, D3, and E4 cells. All the genomic coordinates use hg19 as a reference genome. The data represent in all instances the average of three independent experiments with indicated standard deviation. The p values were calculated by the two-tailed Student’s t test.

We next examined if the edited CTCFBS retained an ability to interact with CTCF. ChIP assays showed that the mutated OSE allele lost >90% of CTCF binding compared to the WT allele in the parental HCT-116 cell population (Fig. 1b). In contrast, the level of CTCF binding to internal controls, such as the MYC promoter and the H19 imprinting control region, was statistically indistinguishable between the WT and mutant HCT-116 cells (Fig. 1b). To generate an overview of the CTCF binding patterns in the OSE-MYC region, the WT cells and D3/E4 cell clones were subjected to ChIP-seq analyses. The patterns, resulting from the normalization of three independent ChIP-seq experiments, showed that, apart from the CCAT1-specific CTCFBS, all other CTCFBS within the OSE-MYC domain bound CTCF equally well in WT and D3/E4 cells (Fig. 1d).

To assess if the mutation of the CTCFBS translated into a phenotypic trait, we developed primers and PCR conditions to determine the allele-specific representation of the WT and mutant (D3/E4) OSE alleles (Supplementary Fig. 1c) in co-cultures that were maintained for up to 2 weeks. We chose this strategy over the analyses of growth curves for the individual cell populations, as it provided an internal control directly enabling a comparison of the proliferation rate of WT and mutant cells. By scoring for the ratios between WT and D3 or E4 OSE alleles we could determine that the WT CCAT1-specific CTCFBS confers excessive growth advantage to the WT cells that out-grew the D3/E4 clones already within a week (Fig. 1e). Given the link between pathological WNT signaling and MYC gating7, we next addressed if the inhibition of the canonical ß-catenin function would specifically target this growth advantage. Since ß-catenin is mutated in HCT-116 cells23, we aimed at antagonizing its function in an as direct manner as possible. To this end, we used the BC21 drug, which we have earlier shown to specifically antagonize the proximity between ß-catenin and TCF4 as well as to evict ß-catenin from the OSE chromatin proximal to the CTCFBS7. The BC21 drug was optimized to antagonize the physical interaction between ß-catenin and TCF4 (Supplementary Fig. 1d) without interfering with MYC transcription (see below). Figure 1f shows that BC21 treatment indeed reduced the growth advantage of the WT HCT-116 cells in both the WT/D3 and WT/E4 co-cultures ca 6-fold during a 2-week co-culturing. We therefore conclude that the binding of CTCF to the intron of the CCAT1 gene endows colon cancer cells with a WNT signaling-dependent growth advantage that efficiently outcompetes a cell population lacking this functional CTCFBS.

The CCAT1-specific CTCFBS within the oncogenic super-enhancer controls the nuclear export rate of the mRNAs produced from the interacting, distal MYC, and FAM49B genes

We have earlier shown that the OSE controls MYC expression post-transcriptionally in a WNT signaling-dependent manner in HCT-116 cells7. To assess the potential involvement of the CCAT1-specific CTCFBS in this process, we first addressed if the facilitated rate of nuclear export of MYC mRNAs, a hallmark of the mammalian gating principle7, was impaired in the mutant D3 and E4 cells in comparison to the WT HCT-116 cells. This assay was performed by determining the cytoplasmic/nuclear ratio of newly synthesized mRNAs over time, as described earlier7. We explored the nuclear export rate of both the MYC and FAM49B mRNAs, since the OSE physically interacts also with an enhancer proximal to FAM49B, as determined by Nodewalk analyses24, and because the MYC and FAM49B gene products are functionally intertwined25,26. Figure 2a shows that both mutant cell clones displayed more than threefold reduction in the rate of nuclear export of MYC and FAM49B mRNAs, without displaying any significant change in the overall transcriptional rate of either MYC or FAM49B (Fig. 2b). Moreover, the nuclear export rate in the mutant cells was, in contrast to the parental WT HCT-116 cells, unaffected when treated with BC21 (Fig. 2a). Of note, the level of MYC transcription per cell was unaffected by the editing of the CCAT1-specific CTCFBS (Fig. 2b). The presence of two MYC/FAM49B alleles per mutant cell as opposed to three copies in the WT cells (see Source data) indicated that MYC/FAM49B transcription per allele might be higher in the mutant cells (see also Supplementary Fig. 2a, b). Although this observation is not statistical significant, it suggests that the OSE-specific CTCFBS does not reduce MYC and FAM49B transcription per se. As could be expected, the reduced rate of nuclear export of MYC and FAM49B mRNAs in the mutant cells correlated with significantly reduced total MYC and FAM49B mRNA expression levels, which were similar for both D3 and E4 clones, in comparison with WT HCT-116 cells (Fig. 2c). We next modeled the observed reduction of MYC expression in the mutant cells using the parameters of transcription, mRNA decay and export rates, as previously described7. Figure 2d demonstrates that, based on these parameters, the simulated MYC expression difference between the WT HCT-116 cells and mutant D3 or E4 clones agreed with the experimental data. We thus conclude that the CCAT1-specific CTCFBS increases MYC expression solely by facilitating the nuclear export rate of its derived mRNA - a feature that provides the WT HCT-116 cells with a strong proliferative advantage.

Fig. 2: The CCAT1-specific CTCFBS increases MYC and FAM49B expression by facilitating the nuclear export of MYC and FAM49B mRNAs in a WNT-dependent manner.

a The rate of nuclear export of newly synthesized MYC and FAM49B mRNAs in WT and mutant HCT-116 cells in the absence or presence of BC21 (see Methods). b The transcriptional rate of the MYC and FAM49B genes. The data were generated by qRT-PCR analyses of MYC/FAM49B transcription using newly synthesized (30 min ethynyl-uridine pulse) RNA as template and normalized to ACTB transcription. c The steady state levels of cytoplasmic MYC and FAM49B mRNAs in WT and mutant HCT-116 cells normalized to TBP expression and external “spike in” RNA controls. The levels of TBP and ß-actin mRNA expression, markers used to normalize the mRNA export and transcription rates, were not significantly different between the WT and D3/E4 cells and correctly estimated the number of input cells (Supplementary Fig. 2a, b). d Comparison between observed and simulated cytoplasmic MYC RNA levels in WT HCT-116 and E4 cells. The simulation used the parameters of nuclear export of MYC mRNA (a), transcriptional rate (b), and the kinetics of MYC mRNA decay in the nuclear and cytoplasmic compartments, as described before7. The data represent in all instances the average of three independent experiments with indicated standard deviation. The p values were calculated by the two-tailed Student’s t test.

ß-catenin and CTCF collaborate in recruiting the nuclear pore-anchor AHCTF1 to the oncogenic super-enhancer

Given that AHCTF1 is both a key factor in the anchoring of the OSE to nuclear pores and essential for the trafficking of MYC in colon cancer cells7, we next explored potential relationships between CTCF and AHCTF1 using co-immunoprecipitation analyses. Remarkably, the results showed that the recovery of AHCTF1 in these samples generally exceeded that of CTCF (Fig. 3a, b). This observation strongly indicated that a subpopulation of CTCF physically interacts with oligomers of AHCTF1. Of note, the expression levels (Supplementary Fig. 3a, b) and complex formation between CTCF and AHCTF1 by co-immunoprecipitation analyses (Supplementary Fig. 3c) were very similar between WT and D3/E4 cells to rule out clonal variation as an underlying cause of the effects of the edited CTCFBS. Since AHCTF1 is part of the nuclear pore as 16-mers27, a fraction of CTCF might interact directly or indirectly with the nuclear pore and/or a pre-nucleopore complex. In line with this reasoning, the relatively high frequency of interactions between CTCF and NUP133 might thus reflect that a pre-nucleopore complex28 interacts with CTCF via AHCTF1. Our earlier observation that BC21 evicted not only ß-catenin but also AHCTF1 and NUP133 from the OSE chromatin7 indicated that ß-catenin is also involved in the CTCF-AHCTF1 complex formation. We therefore examined if the physical link between CTCF and AHCTF1 could be sensitive to BC21 treatment, again using co-immunoprecipitation analyses. Figure 3c shows that BC21 indeed counteracted the formation of the CTCF-AHCTF1 complex. Moreover, as BC21 evicted AHCTF1 from the OSE chromatin (Fig. 3d) without negatively affecting CTCF binding to the CTCFBS (Supplementary Fig. 3d), ß-catenin might facilitate the recruitment of AHCF1 to the CCAT1-specific CTCFBS. To scrutinize the role of CTCFBS-bound CTCF in more detail, we examined if a reduction in CTCF expression levels by siRNA treatment would affect the presence of AHCTF1 at this CTCFBS. Figure 3e shows that reduced expression of CTCF (Supplementary Fig. 3e) impaired the binding of both CTCF and AHCTF1 in WT HCT-116 cells to the CCAT1-specific CTCFBS. Since the mutated CTCFBS also displayed a reduced ability to bind AHCTF1 in both the D3 and E4 clones (Fig. 3f), we conclude that the CTCF-CTCFBS complex likely collaborates with ß-catenin to promote the recruitment of AHCTF1 to the OSE.

Fig. 3: CTCF and ß-catenin recruit AHCTF1 to the oncogenic super-enhancer to promote its ability to reach the nuclear pore.

a Co-immunoprecipitation analyses of physical interactions between CTCF, ß-catenin, NUP133 and AHCTF1. IgG negative control. b Quantification of the CTCF-bound complexes shown in (a). c Co-immunoprecipitation analyses of the physical interactions between AHCTF1 and CTCF in WT HCT-116 cells in the absence or presence of BC21. d ChIP analyses of the binding of AHCTF1 to the oncogenic super-enhancer in DMSO control or BC21-treated WT HCT-116 cells. e ChIP analyses of CTCF and AHCTF1 binding to the CCAT1-specific CTCFBS in cells transfected with siGFP or siCTCF. The signals were normalized to the siGFP controls. The average siCTCF-mediated reduction in CTCF expression was 85% (Supplementary Fig 3d). f ChIP analyses of AHCTF1 binding to the CTCFBS and the CCAT1 promoter within the OSE in WT HCT-116 and mutant (D3/E4) cells. g The knock-down of AHCTF1 expression by siRNA using a siGFP as control. h 3D DNA FISH analyses of the proximity between the OSE and the nuclear periphery in HCT-116 cells in the presence or absence of AHCTF1 (reduced to 72% in comparison to controls7). The bars represent the sum of two independent experiments (219 and 201 alleles, respectively) for siGFP and siAHCTF1-treated cells. i Box-and-whisker plots show median values, interquartile ranges and Tukey whiskers of the distribution of the OSE within 0.7 μm from the nuclear periphery. j In situ proximity ligation assay (ISPLA) of the proximity between CTCF and AHCTF1 in the absence or presence of BC21 in WT HCT-116 cells. Overviews of the DMSO, BC21, and no primary antibody control motifs (upper row), with the the lower row shows magnifications of focal planes marked in the upper row. Bar = 5 μm. k The quantification of the ISPLA signals. The data is based on three independent experiments counting a total number of 710 alleles. C CTCF antibody, A AHCTF1 antibody, No ab no primary antibody. All the data (except for h) represent the average of three independent experiments with indicated standard deviations. The p values for (b–g, k) were calculated by the two-tailed Student’s t test whereas the p value for (i) was calculated using the two-sided KS test.

To explore if AHCTF1 has any role in the distribution of the OSE within the nucleus, we performed DNA FISH analyses of formaldehyde-fixed WT HCT-116 cells transfected with siGFP or siAHCTF1. The DNA FISH signals scoring for the distances of the OSE alleles from the nuclear periphery are presented in a bar diagram with distance windows ranging from <0.3 to >1.5 μm, as described before7. The results show that a reduction in cellular AHCTF1 expression (Fig. 3g) impedes the ability of the OSE to make the final stretch to the periphery/pore (Fig. 3h). The window of 0,7 micrometers from the periphery represented the highest statistical significance (Fig. 3i) in altered OSE distribution upon AHCTF1 downregulation. This data therefore suggests that the interaction between CTCF and AHCTF1 most likely occurs prior to the final anchoring of the OSE to the nuclear pore. Of note, although AHCTF1 is primarily localized at nuclear pores, a considerable fraction can also be found in the nucleoplasm28. Indeed, in situ proximity analyses (ISPLA)11 showed that the highest potential for interactions between CTCF and AHCTF1 spanned a region 1-2 micrometers distal to the nuclear periphery (Fig. 3j, k and Supplementary Fig. 3f), similar to the CTCF-NUP133 ISPLA signals (Supplementary Figs. 3f, 4a, b). Strikingly, this potential for interaction was significantly reduced in WT cells treated with BC21 (Fig. 3j, k), reinforcing the notion that CTCF and the ß-catenin/TCF4 complex collaborate in the complex formation with AHCTF1.

All of these observations converge on a model where CTCF and ß-catenin act as major players in recruiting AHCTF1 to the OSE and hence the completion of the gating of MYC to nuclear pores. Interestingly, both TCF4, which has binding motifs on either side of the CCAT1-specific CTCFBS (Supplementary Fig. 5a), and ß-catenin showed prominent binding to the OSE not only in WT HCT-116, but also in D3 and E4 cells (Supplementary Fig. 5b) that were shown above to display reduced AHCTF1 binding to the CTCFBS. Taken together with that: (i) BC21 evicts AHCTF1 (Fig. 3d) and ß-catenin7, but not CTCF, from the OSE (Supplementary Fig. 3d), (ii) BC21 impairs the physical interaction between CTCF-AHCTF1 as well as the proximities between CTCF and AHCTF1 when juxtaposed to the periphery (Fig. 3c, j, k), and (iii) knock-down of CTCF expression reduces the presence of AHCTF1 at the OSE (Fig. 3e), we submit that CTCF and ß-catenin join forces in promoting the presence of AHCTF1 at the OSE. We argue, moreover, that despite the absence of a strong interaction between ß-catenin and CTCF in HCT-116 cells (Fig. 3a, b), these factors collaborate in recruiting and/or stabilizing the presence of AHCTF1 at the OSE with AHCTF1 likely functioning as a bridge between these factors (Fig. 4). According to this scenario, the hypothesized accumulation of AHCTF1 at the CTCFBS as oligomers might enable its lateral distribution to the CCAT1 promoter within the OSE (Fig. 3d, f).

Fig. 4: Schematic model of the recruitment of AHCTF1 to the OSE-specific CTCFBS.

Both WT HCT-116 and D3/E4 cell clones displayed prominent TCF4/ß-catenin binding to the CCAT1-specific CTCFBS region, independently of the mutation in the CTCFBS (Supplementary Fig. 5b). Since both the mutation of the CTCFBS (Fig. 3d, f) and the disruption of ß-catenin-TCF4 complex7 lead to a reduction in AHCTF1 presence at the OSE, we propose that the juxtaposed CTCF and TCF4-binding sites (Supplementary Fig. 5a) collaborate to stabilize the presence of AHCTF1 at the CCAT1-specific CTCFBS via a CTCF-AHCTF1-ß-catenin complex. The timing of these events is largely unknown and are thus hypothetically visualized in the image.

CTCF coordinates the proximity between the OSE and MYC regions at the nuclear periphery, but is not directly responsible for their interactions

To explore how the gating process is coordinated with interactions between the OSE-MYC regions, we first addressed their distributions within the nuclear architecture using 3D DNA FISH probes covering 8 and 9 kb of the OSE and MYC regions, respectively (Fig. 5a)7. To visualize the relationships between the generated FISH signals, we calculated the value “c” from the equation c = b − a, where “b” and “a” are the distances in micrometers between the nuclear periphery and the MYC or the OSE signals, respectively7. The data was further stratified by including information about single-single (primarily G1), single-double (primarily early S phase) or double-double (primarily late S + G2) DNA FISH signals to explore any relationship to the cell cycle. Figure 5a, b shows that the OSE and MYC alleles approach the nuclear periphery in a coordinated manner (c ≈ 0 at the periphery), which is in keeping with the previously reported observation that they are generally closest to each other when approaching the nuclear periphery in WT HCT-116 cells7. The proximity of both the OSE and MYC regions to the periphery as well as the coordination of their recruitment to the nuclear periphery (as indicated by the “c” value) were, however, strongly reduced in both the E4 (Fig. 5a, b) and D3 (Supplementary Fig. 6) mutant cells. The statistical significance of the differential sub-nuclear localization of the OSE and MYC between WT and mutant HCT-116 cells was determined by examining the cumulative distribution of the OSE and MYC alleles for the WT, D3 and E4 cell populations spanning one micrometer from the nuclear periphery. Of note, only the un-replicated alleles showed a significant effect on the proximity between the OSE and the nuclear periphery (Fig. 5c, d and Supplementary Fig. 6), suggesting that the gating principle is specific for the G1 phase of the cell cycle. We also observed that the presence of the MYC alleles at the periphery was significantly reduced in both D3 and E4 cells (Fig. 5c, d) to reinforce our earlier observation that it is the OSE region that brings MYC to the nuclear periphery/pore and not vice versa7.

Fig. 5: The CCAT1-specific CTCFBS influences the proximity between the OSE and MYC at the nuclear periphery, but not their overall interaction frequency.

a Schematic map (to scale) of the OSE and MYC regions with the position of the DNA FISH probes indicated. b Analysis of the “c” value (scoring for the difference in the distances of MYC and the OSE from the nuclear periphery) in relation to the proximity of the OSE to the nuclear periphery in control and mutant HCT-116 cells for un-replicated alleles (MYCsingle/OSEsingle). The replication state of the MYC and OSE regions are indicated by the number of replicated alleles (see Supplementary Fig. 6 for additional information). A total of 1085 (Ctrl), and 740 (E4) alleles were counted from two independent experiments. c The overall proximity between the OSE and MYC regions from the nuclear periphery were stratified into three distances. S/S = un-replicated alleles; D(MYC)/S(OSE) = The MYC allele replicated before the OSE allele; S(MYC)/D(OSE) = The OSE allele replicated before the MYC allele; D(MYC)/D(OSE) = Both MYC and OSE alleles replicated. d The cumulative distribution of un-replicated MYC and OSE alleles within one micrometer from the nuclear periphery in WT HCT-116, D3, and E4 cells. The numbers have been derived from the source data of (b, c). e Chromatin fiber interaction analyses (Nodewalk)7,24 showing the percentage of sequences within the OSE region (hg19:chr8:128192176-128309374) that interacted with the MYC anchor (hg19:chr8:128,746,000-128,756,177). f Graphic representation of the interaction patterns in the 5’-flank of the MYC anchor. The data is the average of unique ligation events normalized from three independent experiments. The p values indicated in panels b and d were determined using the two-sided KS test, while the p values in panel e were determined by the two-tailed Student’s t test.

Although such data would seem to indicate that CTCF directly facilitates communications between the OSE and MYC13,14, this is likely not the case. The reason for this conclusion is based on data generated by the Nodewalk technique7,24, a 3C-like method, which analyses physical interactions between distal regions within the living cell in an ultra-sensitive manner. Using MYC as an anchor, the Nodewalk analyses generated a total of 22972 deduplicated reads or ligation events from a total input representing ca 60000 WT, D3 and E4 cells in total, respectively, from three independent experiments. When plotted in bar diagrams, showing the percentage of reads mapping to the OSE (Fig. 5e), or mapped on the physical map of the OSE-MYC region (Fig. 5f), it appears clear that the CCAT1-specific CTCFBS has no role in the overall interaction patterns between MYC and the OSE. This observation is not entirely surprising given the scarcity of CTCF binding sites within the OSE and the multiple contacts between the entire OSE and the MYC anchor. Indeed, the Mediator complex that covers the OSE29 is known to mediate enhancer–gene interactions30. Nonetheless, the CCAT1-specific CTCFBS might indirectly promote physical interactions between the OSE and MYC regions by directing their recruitment to the crowded environment represented by the nuclear periphery5.

WNT-dependent activation of CCAT1 transcription is mediated by CTCF to drive the peripheral positioning of the oncogenic super-enhancer

Recent observations have implicated the CCAT1 eRNA as a mediator of OSE-MYC interactions14, potentially via RNA–RNA interactions15. The involvement of specific CCAT1 eRNA isoforms in these processes, with different 3′-end (CCAT1-L) or 5′-ends (CCAT1-5L), complicates, however, interpretation (Fig. 6a). In Supplementary Fig. 7a we show that while the WT HCT-116 cells do not express the CCAT1-5L version, they prominently express the CCAT1-L version. To assess the potential ability of this transcript a