Long-Range Chromatin Interactions Drive Mutant TERT Promoter Activation

Telomerase is a reverse transcriptase that elongates telomeres and thus maintains genomic integrity (1–3). It is minimally composed of the catalytic protein component TERT and a template forming RNA component, TERC. Although TERC is ubiquitously expressed (4), TERT is epigenetically silenced in most adult somatic cells, limiting their replicative lifespan. Up to 90% of human cancers reactivate TERT expression transcriptionally to reconstitute telomerase enzyme activity, which subsequently enables replicative immortality (5).

Recently, two cancer-specific somatic mutations in the TERT promoter were identified (6, 7). These mutations are particularly common in a subset of cancers including melanoma (74%), glioblastoma (83%), hepatocellular carcinoma (44%), and urothelial bladder carcinomas (53.5%; refs. 6–10). These mutations cause a Cytosine (C) to Thymidine (T) transition at -124bp (-124C>T) and -146bp (-146C>T) upstream of the ATG start site, resulting in the creation of novel E-Twenty-Six (ETS) transcription factor binding motifs (6). The ETS family includes 27 members, among which GABPA has been shown to specifically associate with the de novo motif created by mutations in several cancer cell types (11).

Because the mechanisms of activating transcription from wild-type (WT) TERT promoter in stem cells could be vastly different from those employed by mutant TERT promoters, understanding how these promoters function is of critical importance. Chromatin remodeling and epigenetic mechanisms, particularly histone modifications and DNA methylations, have been shown to regulate TERT transcription (12). -146C>T and -124C>T mutations are located close to the transcription start site (TSS) of the TERT gene. It is intriguing to speculate that these promoter mutations may affect recruitment of epigenetic regulators, which modulate the chromatin structure in order to drive TERT expression. These mutations are mostly heterozygous wherein only mutant allele specifically recruits GABPA/B1 and RNA polymerase 2 (POL2) to drive transcription and also displays active histone marks (13). But what initiates these epigenetic changes and the nature of these epigenetic changes that eventually convert the inactive TERT promoter to an active one are not understood at all. In this study, we addressed these questions and observed that binding of the transcription factor GABPA to proximal mutant TERT promoter, specifically, initiates and mediates long-range chromatin interactions and enrichment of active histone marks on these promoters and hence drives its transcription. The specificity of these events at mutant TERT promoter provides insights into telomerase reactivation and suggests therapeutic opportunities for cancer therapy for a wide range of cancers.

Cancer-specific mutations in the TERT promoter (-146C>T and -124C>T) create consensus-binding sites for ETS transcription factors (14). ETS transcription factors are autoinhibited, and binding of these factors alone is not sufficient to drive transcriptional activation, and it is known that they need to heterodimerize with other factors for activating transcription (15–17). GABPA and ETS1 have been reported to specifically interact with the proximal mutant TERT promoters (11, 18), but the mechanism(s) by which they activate the mutant promoters is not completely understood. It is known that levels of these factors do not dramatically differ between normal and cancer cells. Furthermore, there are several ETS/GABPA binding sites on the WT promoter which remain dormant in most cell types. Clearly, just creating one more binding of ETS/GABPA on the mutant TERT promoters does not fully explain the mechanism of TERT reactivation by these single-nucleotide changes. Are there uncharacterized events specifically initiated by the mutant sites and their chromatin context which drive TERT transcription from mutant TERT promoters? To address these issues, we first asked if cancer-specific point mutations alone are associated with activating epigenetic changes on the mutant TERT promoters. Figure 1A shows the cell lines used in the study along with TERT promoter mutation status. Figure 1B shows the schematic outline of the TERT promoter. In this study, we divide the promoter into three distinct regions: proximal promoter (up to -1 kb), distal promoter (up to -5 kb), and long-distance elements (beyond -5 kb), upstream of the TSS. The most common TERT promoter mutations are shown at the -146C>T and -124C>T positions with respect to the ATG translation start site. We examined the active histone marks H3K4Me3 and H3K9Ac in BLM, A375, T98G (bearing -146C>T mutation in TERT promoter), and U251 (bearing -124C>T mutation in TERT promoter) cell lines up to -5 kb upstream of the TERT gene. Interestingly, we observed enrichment of both H3K4Me3 and H3K9Ac in the proximal TERT promoter region of these mutant cell lines (Fig. 1C; Supplementary Fig. S1A). We next performed similar experiments in primary melanocytes with the WT TERT promoter (Fig. 1D), and additionally we investigated multiple cancer cells, including PC3 (prostate cancer), Fadu (head and neck cancer), and the HCT116 colon cancer cell line (Fig. 1D; Supplementary Fig. S1B), wherein TERT is driven by a WT promoter. We observed a trend of lower enrichment of the active histone marks in the active WT TERT promoter from primary melanocytes and other cancer cells (Fig. 1D). It must be noted that average TERT expression levels from a panel of cells with mutant and active WT TERT promoter are not correlative with the mutational status of the promoter as reported previously (Supplementary Fig. S1C; refs. 13, 19). These results suggest that comparing TERT levels and epigenetic marks across cancer cell types is not ideal, and the mechanistic insights regarding the functioning of the mutant and WT TERT promoters could only be inferred using a set of isogenic lines wherein the cause and effect of making these alterations can be evaluated.

A recent study by Barutcu and colleagues suggested that GABP was specifically enriched in Topologically Associated Domains (TAD) in breast cancer cells, suggesting that it may also mediate interactions between domains at long distances (20). To investigate if GABPA bound to mutant TERT promoter could engage in long-range chromatin interactions, we performed the Circular Chromosome Conformation Capture (4C) assay (21). Figure 1E depicts the intrachromosomal interactions of the TERT promoter (chromosome 5) in BLM and A375 cell lines as brought forth by the 4C assay. The y axis indicates the read counts and the x axis indicates distance upstream and downstream of the TERT promoter which are aligned with respect to the dashed red line. The top panel (Fig. 1E) indicates the RefSeq genes on chromosome 5. Significant interactions are represented in different shades of red and blue dots for A375 and BLM cells, respectively. Darker dots indicate more significant interactions, indicated as q values in the color-coded legends. We observed multiple significant interactions occurring along chromosome 5 upstream and downstream of -146C>T TERT promoter. Taken together, these findings suggest that mutant TERT promoter displays active histone marks and long-range chromatin interaction.

To address if the histone marks and chromatin interactions observed on the mutant promoters are specifically driven by the point mutation in proximal mutant promoters, we reversed the mutated nucleotide to the WT residue by genome editing of promoters using the CRISPR/Cas9 system (Fig. 2A). The mutant TERT promoter (-146C>T labeled as red color) was targeted by single-guide RNA adjacent to the mutant region. The WT nucleotide was introduced using a repair template containing homology regions as shown in Fig. 2A. We generated isogenic lines with the mutant TERT promoter (-146C>T), hereafter referred to as BLM6, and with the WT TERT promoter (-146C), hereafter referred to as BLM14. Genetic status was confirmed by Sanger sequencing as shown in Fig. 2B (top left). As compared with the BLM6, TERT expression was dramatically reduced in the BLM14 cells (Supplementary Fig. S2A). Reversal of mutant to WT TERT promoter also leads to reduced telomerase activity (Supplementary Fig. S2B). Early-passage cells did not show significant differences in telomere length; however upon long-term culture, significant telomere attrition was observed in engineered cells (Supplementary Fig. S2C and S2D), suggesting the functionality of this mutation reversal. We also analyzed BLM14 cells for the Alternative Lengthening Telomeres (ALT) phenotype. There was no colocalization between TRF2 and PML in BLM6 or BLM14 cells (Supplementary Fig. S2E). U2OS was used as the positive control where we observed very strong colocalization as reported previously for ALT cell lines (22). We performed RNA sequencing from BLM6 and BLM14 cells in order to eliminate possible off-target effects of the CRISPR/Cas9 method. No significant differences were detected in genes that were predicted to have off-target effects for guide RNA used in CRISPR/Cas9 editing by RNA sequencing (Supplementary Table S1). Next, we investigated the epigenetic status of the proximal TERT promoter in these cell lines. Interestingly, we observed that the proximal active histone marks were significantly reduced upon single-nucleotide reversal (Fig. 2B).

Next, we mapped the long-range intrachromosomal interactions in these cell lines using the 4C assay. Fig. 2C–E show differential intrachromosomal interactions of the TERT promoter between BLM6 and BLM14 cells. Red dots indicate significant and black dots indicate nonsignificant interactions between mutant and WT TERT promoters. The intrachromosomal interactions were dramatically reduced in BLM14 (Fig. 2D) as compared with BLM6 cells (Fig. 2C). Fold interaction differences are indicated as a log₂ scale that is shown in a lower panel together with RefSeq genes that are located on chromosome 5 (Fig. 2E). As 4C is a semi-quantitative method, to accurately quantify the differences of these chromatin interactions, we selected 12 of the interactions that were obtained from 4C analysis and performed chromosome conformation capture (3C)-qPCR assays. Figure 2F shows the strongest five regions interacting with the mutant TERT promoter. The x axis indicates the distance of the regions from -146C>T site of the TERT promoter, which is represented as “0” along with its coordinates below, and the y axis indicates the binding strength in log₁₀ scale. The strongest interaction was observed between the TERT promoter and a region 300 kb upstream of promoter (chr5:1,556,087–1,558,758), hereafter referred to as T-INT1 (Fig. 2F). The 3C-qPCR assay verified that the mutant -146C>T mutant TERT promoter in BLM6 cells showed significantly higher interaction with the T-INT1 intergenic region when compared with the -146C WT TERT promoter in BLM14 cells (Fig. 2G). Supplementary Table S2 shows the list of regions with genomic coordinates that were found to differentially interact with the mutant TERT promoter in BLM6 cells. Because a GABPA binding site is the only difference between WT and mutant TERT promoters, we speculated that histone marks and long-range interactions uniquely apparent on the mutant promoters might be initiated/mediated by GABPA binding to this site in the proximal promoter.

As the mutant TERT promoter creates a binding site for the ETS family member GABPA transcription factor, we analyzed the GABPA recruitment in BLM6 and BLM14 cells. As shown in Fig. 3A, GABPA was specifically enriched at the proximal mutant TERT promoter in BLM6 cells. A similar enrichment pattern was obtained for RNA POL2 (Fig. 3B). We next explored the possibility that GABPA recruitment specifically drives the epigenetic marks in the proximal TERT promoter region. We depleted GABPA by siRNA in BLM6 cells (Fig. 3C). Chromatin immunoprecipitation (ChIP) was performed in GABPA-depleted cells, which showed reduced recruitment of GABPA to the mutant TERT promoter along with reduced POL2 recruitment (Fig. 3D). This was associated with reduced TERT expression in GABPA-depleted cells (Fig. 3E). Most importantly, GABPA depletion was associated with reduced enrichment of active histone marks in the proximal TERT promoter (Fig. 3F).

Mutant TERT promoter showed long-range interaction as observed using 4C assays (Fig. 2C–E), and GABP has been shown to have the potential for mediating long-range interactions (20). To investigate if GABPA is a key mediator, we performed the 3C-qPCR assay with GABPA-depleted BLM6 cells. We observed significant reduction of intrachromosomal interaction between the TERT promoter and the T-INT1 region upon GABPA depletion (Fig. 3G). To understand the molecular basis of how GABPA could mediate this long-range interaction, we analyzed the T-INT1 region for the presence of potential GABPA motifs using the transcription factor motif finder database Jaspar (23). We found multiple GABPA motifs in this region (Supplementary Fig. S3A). It is well known that GABPA factors work as tetramers (15–17). As this region is about 2.5 kb long and GABPA sites are spread along the whole sequence, we designed ChIP qPCR primers specific to the 5′ region, central region, and 3′ region. GABPA was specifically enriched at the 5′ region (data not shown). We examined the GABPA binding to this region in BLM6 and BLM14 cells and found that it was significantly increased in BLM6 cells (Supplementary Fig. S3B). Overall, these results suggest that GAPBA dimers/tetramers bound at long distance, the T-INT1 region (particularly GABPA sites located in the 130 to 290bp region according to chr5:1,556,087) and -146C>T mutant site in the TERT promoter provide the stable binding required for GABPA-mediated transcription, and this might be the molecular basis of the specific interaction of the mutant TERT promoter with the 300 kb region upstream of its TSS. Indeed, the presence of several GABPA sites in 5′ of the T-INT1 region may explain why this region might specifically interact with the mutant TERT promoter. Because the WT TERT promoter does not possess the GABPA motif at the -146 position, such long-range interaction does not take place and the chromatin interaction around the TERT promoter does not get stabilized for further modification by epigenetic factors. To our knowledge, this is the first report showing GABPA can indeed function as a mediator of long-distance interactions.

Active histone marks are known to recruit chromatin remodelers such as BRD4 (24), so we analyzed whether BRD4 is associated with active TERT promoter in BLM6 cells. We observed strong enrichment of BRD4 at the TERT promoter in BLM6 cells as compared with BLM14 cells (Fig. 3H). Similar patterns of BRD4 recruitment were observed upon reducing GABPA expression, suggesting the essential role for GABPA in initiating the establishment of the chromatin marks on this promoter (Fig. 3I).

The BET domain family inhibitor JQ1 has been shown to regulate GABPA expression (25). Among BET family members, BRD4 has been shown to be the most potent target of JQ1 inhibitor. Thus, to investigate whether BRD4 affects GABPA expression, we knocked down BRD4 in BLM6 and BLM14 cells. Figure 3J shows the knockdown efficiency in BLM6 and BLM14 cells. We observed significant reduction in GABPA expression, suggesting that BRD4 regulates GABPA expression independent of TERT promoter status (Fig. 3K). In addition, we assessed direct BRD4 recruitment to the GABPA promoter and observed that it was highly enriched, suggesting that BRD4 indeed directly regulates GABPA expression (Fig. 3L). However, reducing BRD4 reduced TERT expression in BLM6 cells (Fig. 3M). This was corroborated with reduction in telomerase activity upon knocking down GABPA or BRD4 in BLM6 cells (Supplementary Fig. S4A). Because there is no GABPA recruitment on TERT proximal promoter of BLM14, GABPA reduction due to BRD4 knockdown had no effect on BLM14 cells. Moreover, the epigenetic status of the TERT promoter affects the proliferation rate in BLM6 but not BLM14 cells with WT TERT promoter (Supplementary Fig. S4B and S4C). Furthermore, knocking down BRD4 led to significant reduction in active histone marks in BLM6 cells (Fig. 3N). We also investigated the long-range interaction changes in the TERT promoter upon BRD4 knockdown. We observed significant reduction in association of the proximal TERT promoter with the T-INT1 region as measured by the 3C-qPCR assay, when BRD4 levels were reduced (Fig. 3O). As a control, we also analyzed the effect of BRD4 knockdown on TERT expression in Fadu cancer cells bearing active WT TERT promoter. TERT expression remained unaffected upon BRD4 knockdown in Fadu cells (Supplementary Fig. S4D and S4E). There was no effect of knocking down BRD4 or GABPA on alternate splicing of TERT mRNA (Supplementary Fig. S4F). This suggests that depletion of GABPA or BRD4 affects TERT transcription from the mutant TERT promoter. These results could be used to guide us toward therapeutic modalities.

To further validate the novel molecular mechanism mediated by long-range interactions observed in our assays, and to find the general significance of these findings, we performed CRISPR-mediated reversal of -146C>T mutation in A375 melanoma cells to generate mutant WT TERT promoter–containing cells (A375 -146C). Reversal of TERT promoter mutation in A375 cells reduced the TERT expression and telomerase activity and significantly diminished the active histone marks (Fig. 4A–C). ChIP-qPCR analysis showed that POL2 binding was significantly reduced in A375 -146C cells (Fig. 4D). Furthermore, GABPA and BRD4 were significantly enriched in A375 -146C>T cells as compared with A375 -146C cells (Fig. 4E and F). We performed the 3C-qPCR assay in these cells, and the A375 -146C cells showed significant reduction in interaction with the T-INT1 region (Fig. 4G). We also generated WT TERT promoter–containing glioblastoma cell line T98G (T98G -146C) using the CRISPR/Cas9 method (18). T98G cells naturally contain a -146C>T mutation in the TERT promoter. We found that T98G -146C cells showed reduced active histone marks, POL2, GABPA, and BRD4 in the proximal TERT promoter region (Fig. 4H–J). The 3C-qPCR assay revealed that the WT TERT promoter in T98G -146C cells showed significant reduction in interaction with the T-INT1 region as compared with mutant TERT promoter (Fig. 4K). We conclude that the mechanism by which -146C>T mutation operates is most likely conserved across cancer types.

To investigate if the cancer-associated single-point mutations are sufficient for causing the changes described above, we next introduced the -146C>T mutation in HCT116 colon cancer cells with the WT TERT promoter to create isogenic lines (HCT116 -146C and HCT116 -146C>T; Fig. 5A). Cell lines were verified by sequencing, and the HCT116 -146C>T cell line was homozygous (data not shown). Introduction of the TERT promoter mutation in HCT116 cells increased the TERT expression and Telomeric Repeat Amplification Protocol (TRAP) activity, and significantly enriched the active histone marks (Fig. 5B–D). ChIP analysis showed that POL2 binding was significantly increased at the mutant TERT promoter (Fig. 5E). Furthermore, GABPA and BRD4 were significantly enriched in the mutant TERT promoter as compared with the WT TERT promoter (Fig. 5F and G). GABPA was also strongly associated with the T-INT1 region in HCT116 -146C>T cells (Supplementary Fig. S5). We further investigated whether HCT116 -146C and HCT116 -146C>T cells show differential long-range interaction. The 3C-qPCR assay revealed that the mutant TERT promoter in HCT116 -146C>T cells showed significant interaction with the T-INT1 region (Fig. 5H), which was also the strongest interaction observed in 3C analysis of BLM cells (Fig. 2G).

As summarized in Fig. 5I, reversing -146C>T mutation to -146C WT in BLM, A375, and T98G cells causes reduction in GABPA recruitment and loss of long-range interaction. This is concomitant with reduction in active histone marks, POL2 recruitment, TERT expression, and telomerase activity. In contrast, introduction of -146C>T mutation in HCT116 cells causes an increase in GABPA recruitment and formation of long-range interaction. This is concomitant with an increase in active histone marks, POL2 recruitment, TERT expression, and telomerase activity as compared with the WT isogenic counterpart. We conclude that the -146C>T TERT promoter mutation could be sufficient to initiate long-range interaction of this promoter, and it could also be sufficient for activating this promoter via epigenetic modifications initiated and promoted by stabilizing long-range interaction with the T-INT1 region of the TERT promoter (Fig. 1B).

To verify whether mutant TERT promoters are indeed activated via long-range interaction with the T-INT1 region, we created cells wherein we deleted the T-INT1 region while keeping the proximal mutant TERT promoter intact. We also investigated the generality of our observations by also including a cell line with -124C>T mutation in these assays. Figure 6A shows the location of the mutation site (-146C>T or -124C>T) on the TERT promoter and the location of the T-INT1 region. Putative GABPA motifs in the T-INT1 region are shown as red dots. CRISPR-mediated removal of the interacting region from three cell lines (BLM and A375 cell lines which have the -146C>T mutation in the TERT promoter, and U251 cell line which has the -124C>T mutation in the TERT promoter) led to the generation of respective isogenic lines that do have the T-INT1 region (T-INT1 WT) or cells where the T-INT1 region is deleted (T-INT1 KO; Fig. 6A). Genetic status was confirmed by genotyping PCR as shown in Supplementary Fig. S6A. In BLM cells, upon removal of the T-INT1 region, TERT expression and telomerase activity reduced dramatically (Fig. 6B and C) without any change in GABPA expression (Supplementary Fig. S6B). We observed a significant reduction in the active histone mark H3K4Me3 in the proximal TERT promoter (Fig. 6D). Also POL2, GABPA, and BRD4 recruitment to TERT promoter was also reduced significantly (Fig. 6E and F; Supplementary Fig. S6C). Removal of the T-INT1 interaction region in A375 was also associated with reduced TERT expression and telomerase activity (Fig. 6G and H). The active histone mark H3K4Me3 was reduced significantly in the proximal TERT promoter of A375 (Fig. 6I). POL2 and GABPA recruitment to the TERT promoter was also reduced significantly (Fig. 6J and K).

Interestingly, we observed a dramatic reduction in TERT expression and telomerase activity in the U251 T-INT1 KO cell line which harbors the -124C>T mutation (Fig. 6L and M). Similar to melanoma cells, deletion of the T-INT1 region decreased the enrichment of H3K4Me3, POL2, and GABPA in the proximal TERT promoter of U251 cells (Fig. 6N–P).

These results demonstrate that GABPA recruitment to the mutant TERT promoter enables long-range interaction of the proximal TERT promoter with the T-INT1 region and facilitates generation of associated epigenetic changes that drive its transcription. These results also suggest that this is a general mechanism used by both the -124C>T and the -146C>T mutant promoters. Our results are presented in the form of a model in Figure 7. 

Cancer-specific TERT promoter mutations offer an invaluable starting point to understand the mechanistic basis of activation of the dormant TERT promoter in cancers. Our study reveals a novel mechanism utilized by GABPA to activate TERT transcription. Using the 3C assay, we have shown that GABPA mediates long-range chromatin interaction to form a stable complex that drives TERT transcription. Reversal of the mutation in the TERT promoter or removal of the mutant TERT promoter–interacting region (T-INT1) using genome editing results in loss of GABPA binding, depletion of epigenetic marks, and reduction of TERT transcription. Interestingly, the introduction of the -146C>T TERT promoter mutation in HCT116 cells could specifically enrich histone marks and enable long-range interaction. Our findings suggest that GABPA binding to the mutant TERT promoter is an initiating and sufficient event to drive the epigenetic status of the TERT promoter.

The introduction of the -146C>T and -124C>T TERT promoter mutations in human embryonic stem cells and their differentiation of them into fibroblasts and neural progenitor cells revealed that these mutations were capable of overcoming epigenetic silencing (26). Moreover, the TERT expression in these cells was comparable to that found in cancer cells and could significantly delay telomere shortening–induced senescence. Furthermore, the high frequency of mutations in the TERT promoter mostly at two nucleotide positions (-146C>T, -124C>T) strongly implicates them as driver events, appearing upon tumor initiation or possibly later in tumor development (27, 28).

Human TERT promoter harbors a G-rich region of 12 sequential G-tracts, encompassing three SP1 binding sites, and has the ability to form multiple G-quadruplexes. Masking of the SP1 binding sites by G-quadruplex structure is predicted to produce significant inhibition of TERT promoter activity (29). The G-quadruplex structure may also have an important role in the regulation of the mutant TERT promoter, and it is suggested that TERT promoter mutations may have an impact on this predicted secondary structure and that the complex relationship between secondary structure and GABP recruitment could also play a role in regulating TERT expression (11). The complex regulatory cross-talk between long-range interaction and G-quadruplex structure in regulating mutant versus WT TERT promoters needs further investigation.

Furthermore, it would also be interesting to understand what cofactor(s) cooperate with GABPA or other ETS family members on the -146C>T and the -124C>T mutant sites. Clearly, having the know-how and the ability to target distinct mechanisms which operate in cancer cells which are driven by the mutant TERT promoter would be a step forward in designing therapeutics against telomerase in cancer. Targeting mutant TERT promoter–specific pathways that do not impinge on the WT TERT promoter will help in designing drugs that are less toxic and can be taken over a longer term at a higher dose. With the realization that telomerase inhibition could yield a magic bullet to treat most human cancers of any origin and carrying any genetic alteration, many pharmaceutical companies have developed and tested various versions of telomerase inhibitors over the last decade. Many refinements were made, and the most successful inhibitor, by Genron, was believed to reach the clinic. But this inhibitor, which inhibits enzymatic function of telomerase, failed clinical trials early last year. Based on the understanding of how mutant TERT promoters operate, a new set of inhibitors that limit TERT transcription only in cancer cells with -124C>T and -146C>T mutations could be designed rapidly. Because these new transcriptional inhibitors will work to limit levels of TERT and hence telomerase (indirectly limiting the activity) only in cells with mutation, namely the cancer cells, they may not cause toxicity to stem cells or other normal somatic cells where continued telomerase activity is necessary for physiology.

The A375 and HCT116 cell lines were originally purchased from the ATCC and were a gift from Dr. Shang Li (DUKE-NUS Medical School, Singapore). Fadu cells were obtained from the ATCC. BLM cells and primary melanocytes were a gift from Dr. Birgitte Lane (Institute of Medical Biology, Singapore), and T98G-WT (CRISPR-engineered WT TERT promoter) was produced previously (18). PC3 was a gift from Dr. Ernesto Guccione (Institute of Molecular and Cell Biology, Singapore). All cells were maintained in DMEM except PC3, which was maintained in RPMI and supplemented with 10% FBS (Hyclone), penicillin and streptomycin (Gibco) using standard tissue culture techniques. Cell lines were not authenticated by us.

Genomic DNAs of cell lines were isolated with tail lysis buffer supplemented with Proteinase K. DNA was precipitated with isopropanol, and the TERT promoter was sequenced as described previously (18).

The -146C>T TERT promoter mutation in the BLM and A375 cell lines was converted back to the WT promoter sequence (-146C), and -146C was mutated to -146C>T in HCT116 cells as described previously (18). Briefly, guide RNA specific to the TERT promoter was cloned into pX458 (GFP) plasmid and was cotransfected with repair template (130bp) containing WT (-146C) or mutant (-146C>T) TERT promoter sequence. GFP-positive cells were seeded into 96-well plates (1 cell/well) by FACS after 48 hours, and each clone was screened by PCR and Sanger sequencing.

pX458-GFP plasmid was modified by removing Cas9 and GFP with EcorI-AgeI restriction enzymes and inserting Ds-Red monomer in order to prevent off-target effect of excess Cas9 enzyme in the cells (30) and also to be able to observe transfection efficiency of both plasmids. Guide-RNA-1 CACCGCCACAAGGAATGCCGTACAT and Guide-RNA-2 CACCGCCGGCGTGGAGCAATTCCAC targeting TERT interaction region (Chr5:1,556,087–1,558,758) were cloned into pX458 (GFP) and pX458 (Ds-Red) plasmids, respectively. Cells were transfected in 6-well plate by X-tremeGENE 9 transfection reagent (Roche). GFP and Ds-Red–positive cells were sorted into 96-well plates (1 cell/well) after 48 hours by FACS, and each clone was genotyped by PCR with primers surrounding the T-INT1 region: Forward: GTCTGCATGCAATGCTGTC and reverse: GGGGACATTTATGTCTTCTGC. T-INT1 WT PCR product was approximately 4530bp, and T-INT1 KO PCR product was 550bp.

The RNA-sequencing library was prepared from two replicates of the BLM6 and BLM14 cell lines by the Illumina TruSeq Total RNA Sequencing Kit according to the manufacturer's instructions.

Raw reads were aligned to the human genome hg19 by Tophat2. Differential gene expression was performed by Cufflinks and Cuffdiff with default parameters.

For siRNA treatment, cells were transfected when they reached approximately 60% confluency with siControl (Dharmacon; D001810-10), siGABPA (Dharmacon; L011662), or siBRD4 (Dharmacon; L-004937) using X-tremeGENE siRNA transfection reagent (Roche) according to the manufacturer's recommendations. Medium was replaced with fresh medium after 8 hours of incubation, and cells were harvested 24 to 48 hours after transfection. For gene expression analysis, RNAs and cDNAs were prepared as described previously (31).

ChIP was performed as described previously (32). Anti-H3K4Me3 (Millipore; 04-745), anti-H3K9Ac (Millipore; 07-352), anti-H3K14Ac (Millipore; 07-353), anti-POL2 (Santa Cruz Biotechnology; sc-899), anti-GABPA (Santa Cruz Biotechnology; sc-22810), anti-BRD4 (Bethyl; A301-985A), and IgG (Santa Cruz Biotechnology) antibodies were used for the immunoprecipitation. All antibodies were used at a concentration of 1 μg per 1 million cells. After elution, chip eluate was used for ChIP-qPCR with primers (33) targeting TSS, 0.8 kb, 1 kb, 2.5 kb, and 5 kb upstream regions of the TERT gene and/or TERT interaction region. Primers are indicated in Supplementary Table S3.

4C was performed as described previously (34). Briefly, 10⁷ cells were fixed with 1% formaldehyde for 10 minutes, and nuclei pellets were isolated after cell lysis with cold lysis buffer (50 mmol/L Tris-Hcl pH 7.5, 150 mmol/L NaCl, 5 mmol/L EDTA, 0.5% NP-40, 1% Triton X-100) supplemented with protease inhibitors. First step digestion was performed overnight at 37°C with 200 unit HindIII enzyme. Digestion efficiency was measured by RT-qPCR with HindIII site–specific primers. After phenol-chloroform extraction, DNA was ligated overnight at 16°C by T4 DNA ligase (Fermentas). Following de-crosslinking, DNA was processed for second digestion with 50 units of DpnII enzyme overnight at 37°C. After final ligation, 4C template DNAs were quantified by the Qubit dsDNA High Sensitivity Kit (Thermo Fischer) and preceded for library preparation for sequencing using TERT promoter–specific primers (Supplementary Table S3) with Illumina Nextera adapters.

The quality of the FASTQ files was evaluated using FastQC v0.11.3 (35). Illumina Nextera adaptor sequences were removed using scythe v0.991 (36). Tagdust2.3 was used to extract reads that were sandwiched between the bait region (TERT promoter + HindIII) and a DpnII cut site (37). The extracted reads were aligned to the hg19 genome using Bowtie 2 (38) with very sensitive parameters in unpaired mode. The aligned reads were filtered for mapping qualities of > 30. r3CSeq (39) was used in batch mode to call interactions for A375 and BLM cells. Plots were generated by FourCSeq of the differential intrachromosomal interactions of the TERT promoter between replicates of BLM6 CRISPR mutant and BLM14 CRISPR WT (40). The green line shows the distance-dependent fit, and the blue dashed line indicates the fit of z-score > 2. Interactions detected by z-score > 2 (P-adjusted value < 0.05) for at least one replicate are shown as red dots. Fragments not called as interactions and that do not show significant change between conditions are shown as black dots. The calculated log₂ fold change is shown above the Hg19 RefSeq genes in the regions 1 megabase from the TERT promoter.

3C was performed as described previously (41). Cells were processed in the same way as described in 4C. Following HindIII digestion and ligation, 3C template DNAs were quantified by the Qubit dsDNA High Sensitivity Kit (Thermo Fisher). In order to optimize real-time PCRs and to estimate the minimal amount of 3C template for quantification in a linear range, we prepared control template including equal amounts of target regions that we obtained in 4C analysis. These regions were amplified by region-specific primers and purified from gel and mixed in equimolar amounts. After digestion of these fragments with HindIII, fragments were ligated by T4 DNA ligase. 3C DNA templates were used for normalization of 3C-qPCRs. 3C-qPCR reactions were performed by the Sybr-Greener Kit (Invitrogen), and samples were normalized by the levels of TERT promoter region. 3C-qPCR primers are indicated in Supplementary Table S3.

Real-time TRAP assay was performed as described previously (42).

Colony formation assay was performed by crystal violet BLM6, and BLM14 cells were transfected with siControl, siGABPA, and siBRD4. After 24 hours, 500 cells were seeded into a 6-well plate. Culture media were refreshed every 2 days, and after 7 days, cells were fixed with 75% ethanol for 30 minutes and stained with 0.2% crystal violet dye.

ALT-associated PML bodies were visualized via TRF2 and PML immunofluorescence staining. Cells were grown on chamber slides (Millipore EZ slides) and fixed with 4% paraformaldehyde for 10 minutes at room temperature. After washing with PBS, cells were permeabilized with 0.2% Triton X-100 solution in PBS for 5 minutes at room temperature and were blocked with PBS supplemented with 0.2% Triton X-100 and 1% BSA (blocking buffer) for 30 minutes at room temperature. Following incubation, cells were incubated with the primary antibodies TRF2 (Millipore; 05-521) and PML (Santa Cruz Biotechnology; sc-5621), 1/200 and 1/150 dilutions, respectively, in blocking buffer overnight at 4°C. Cells were washed 4 times with blocking buffer for 8 minutes and incubated with secondary antibodies AF488 and AF555 (1/2,000 dilution; Invitrogen; A11001 and A21428) for 1 hour at room temperature. After cells were washed 4 times, image acquisition was performed with Zeiss LSM800, Plan-Apochromat 63x/1.40 aperture of the objective lenses.

The Student t test (two-tailed) was performed to determine the significance of difference for ChIP-qPCR, qPCR, and RT-TRAP experiments. Results of each ChIP, gene expression, 4C, and 3C assays were obtained from at least two or three independent experiments, as indicated in the figure legends.

All data have been uploaded to GEO (GSE77265).

No potential conflicts of interest were disclosed.

Conception and design: S.C. Akıncılar, E. Khattar, V. Tergaonkar

Development of methodology: S.C. Akıncılar

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.C. Akıncılar, B. Unal

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.C. Akıncılar, E. Khattar, P.L.S. Boon, M.J. Fullwood, V. Tergaonkar

Writing, review, and/or revision of the manuscript: S.C. Akıncılar, E. Khattar, P.L.S. Boon, B. Unal, M.J. Fullwood, V. Tergaonkar

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.C. Akıncılar, V. Tergaonkar

Study supervision: V. Tergaonkar

S.C. Akıncılar and B. Unal are supported by the SINGA scholarship. We thank the Agency for Science Technology and Research, Singapore (A*STAR) for funding and support to the V. Tergaonkar laboratory. We thank Phua Qian Hua for her help during the genotyping process and Dr. Shang Li (DUKE-NUS Medical School, Singapore) for the southern blot telomere length experiment.

This work was supported by the core budget from the Institute of Molecular and Cell Biology (IMCB), Singapore. P.L.S. Boon and M.J. Fullwood are supported by the National Research Foundation (NRF) Singapore through an NRF Fellowship awarded to M.J. Fullwood (NRF-NRFF2012-054) and Nanyang Technological University School of Biological Sciences start-up funds awarded to M.J. Fullwood, as well as by funding given to the Cancer Science Institute, National University of Singapore (NUS), by the NRF and the Ministry of Education, Singapore, under the Research Center of Excellence funding, and the RNA Biology Center at the Cancer Science Institute of Singapore, NUS, as part of funding under the Singapore Ministry of Education's Tier 3 grants.