2D gel electrophoresis reveals dynamics of t-loop formation during the cell cycle and t-loop in maintenance regulated by heterochromatin state

Linear chromosome ends are capped by telomeres that have been previously reported to adopt a t-loop structure. The lack of simple methods for detecting t-loops has hindered progress in understanding the dynamics of t-loop formation and its function in protecting chromosome ends. Here, we employed a classical two-dimensional agarose gel method (2D gel method) to innovatively apply to t-loop detection. Briefly, restriction fragments of genomic DNA were separated in a 2D gel, and the telomere sequence was detected by in-gel hybridization with telomeric probe. Using this method, we found that t-loops are present throughout the cell cycle, and t-loop formation tightly couples to telomere replication. We also observed that t-loop abundance positively correlates with chromatin condensation, i.e. cells with less compact telomeric chromatin (ALT cells and trichostatin A (TSA)-treated HeLa cells) exhibited fewer t-loops. Moreover, we observed that telomere dysfunction-induced foci, ALT-associated promyelocytic leukemia bodies, and telomere sister chromatid exchanges are activated upon TSA-induced loss of t-loops. These findings confirm the importance of the t-loop in protecting linear chromosomes from damage or illegitimate recombination.

Telomeres protect linear chromosomes from exonucleolytic degradation, undesirable illegitimate recombination events, and end-to-end fusions (1,2). In human cells, telomeres are composed of tandem repeats of dsDNA TTAGGG/AATCCC, a terminal ssDNA 5 3Ј-G-rich overhang, and a telomere-binding complex called "shelterin" (3,4). The ends of telomeric DNA have a propensity to form "t-loops," in which 3Ј-G-rich overhang invades and is paired with the C-rich strand of the dsDNA telomeric repeat tract in cis to form a D-loop (5). It has been proposed that t-loops may play a critical role in protecting linear chromosomes from nuclease-mediated end-resection and unscheduled DNA repair (6).
T-loops were first discovered in the nuclei of human and mouse cells (5). They were subsequently observed in other eukaryotic species, including Trypanosoma brucei (7), Pisum sativum (8), Gallus gallus (9), and Caenorhabditis elegans (10). T-loops are considered to be an evolutionarily-conserved structure for protecting linear chromosome termini. However, many questions regarding the establishment and maintenance of t-loops in cells remain to be elucidated (11). For example, very little is known about how t-loops are negotiated by DNA replication machinery during S phase. The mechanism underlying the formation and maintenance of t-loops through the cell cycle is also poorly understood. Moreover, it is not yet known what happens to telomeres/chromosome ends if massive t-loops are disrupted in vivo. The answers to these questions are the foundation for better understanding the function of t-loops in protection of chromosome ends.
One of the roadblocks to answering these critical questions is the lack of a reliable, sensitive, and commonly used method to detect t-loops in cells and/or cell-free extracts. Electron microscopy (EM) and the recently developed stochastic optical reconstruction microscopy (STORM) are valuable tools for studying and visualizing t-loops, but they are low-throughput and labor-intensive, and they require equipment that is prohibitive for most research groups (5,12). Therefore, an accurate and sensitive biochemical method to detect t-loops is needed to stimulate progress in this important research area.
Based on the structural similarity between t-loop and wellcharacterized rolling circle replication intermediates (RCIs), the RCIs migrate in a unique size-and shape-dependent sigmoid pattern in nondenatured two-dimensional (2D) agarose gel electrophoresis (13)(14)(15); therefore, we propose that t-loops might segregate from the linear telomere DNA during 2D gel electrophoresis. Here, we report the development of a nondenatured 2D agarose gel electrophoresis method (2D gel method) that readily detects t-loops from human cells. The method was carefully validated by a series of biochemical assays as well as conventional EM. Evidence presented here supports the following four major conclusions about the formation and dynamics of t-loops in human cells: 1) t-loops are present throughout the cell cycle; 2) t-loop formation is tightly coupled with telomere replication; 3) less condensed telomeric chromatin showed fewer t-loops in human ALT cells; and 4) trichostatin A (TSA)-induced hyperacetylation of telomeric nucleosomes promotes disruption of t-loops that is associated with the formation of APBs and increased frequency of T-SCE, indicating the homologue recombination occurs at unprotected telomeres.

Mobility of telomere-homologous structures during nondenatured 2D agarose gel electrophoresis
RCIs are a unique and extensively characterized type of replication intermediate. RCIs and t-loops are similar to each other, both comprising a circular dsDNA and a linear dsDNA tail. Several methods are available to detect RCIs based on their structure, including those based on characteristic gel electro-phoretic mobility. RCIs with a fixed loop size but different tail lengths are detected as an "eyebrow"-like shape (see inset in Fig.  1A) (13,16). The multiple "eyebrows" formed from RCIs with different loop sizes converged into a sigmoidal arc (Fig. 1A) (13).
Here, we exploit nondenatured two-dimensional agarose gel electrophoresis followed by hybridization to a telomere-specific probe (2D gel method) to specifically detect t-loops in human cells. Given the structural similarity between RCIs and t-loops, both consisting of loops and tails that highly variable in size (5), a telomeric sigmoidal arc is expected. As expected, a sigmoidal arc of telomere-homologous species is detected in genomic DNA from nonsynchronized HeLa cells (Fig. 1B).
To exclude the possibility that the sigmoidal arc is the unique structure of HeLa, we detected the sigmoidal arcs using a 2D gel method in a variety of human cells, including primary cells (human T cell and fibroblast BJ), telomerase-positive cells (HeLa S3 and A549), and ALT cells (VA13 and U2OS) (Fig. 1C). The sigmoidal arcs were observed in all of these cells, some of which have been observed by EM in a previous study (5). It seems that the sigmoidal arc represents a special telomeric structure, which predicted t-loops.

Characterization of DNA structures in the sigmoidal arc
The DNA in the sigmoidal arc region of the 2D gel method was characterized by biochemical methods as follows. First, the sigmoidal arc gradually decreased as the temperature rises to 65°C, the temperature at which most t-loops dissociate ( Fig.  2A), indicating the low thermal stability of t-loops. Second, the sigmoidal arc is barely detected in undigested HeLa genomic DNA (Fig. S1A), indicating that the telemetric DNA in the sigmoidal arc is not extrachromosomal DNA. Trace amounts of  (13). Inset indicates the eyebrow formed by RCIs with the same loop size but different tail lengths. B, HeLa genomic DNA was purified, digested with RsaI and HinfI, and analyzed by 2D gel method. Sigmoidal arc, t-circle-tail (35), and linear telomere were indicated. C, telomere homologous DNA was analyzed in human primary cells (T-cells and BJ fibroblast), telomerase-positive cells (HeLa S3 and A549), and ALT cells (VA13 and U2OS). Sigmoidal arc (black arrow) and t-circle (red allow) (24) were indicated.

Study t-loop with 2D gel method
the sigmoidal arc signal may be due to DNA breakage at the subtelomere or telomere during purification. Third, after a second round of purification of HeLa genomic DNA, the sigmoidal arc decreased, suggesting that this DNA fraction is not an artifact of genomic DNA caused by purification (Fig. S1B). Fourth, in the experiments described above, HeLa genomic DNA was purified without proteinase treatment at 55°C to preserve the t-loop structure. To test the susceptibility of putative t-loop structures to protease digestion, the cell lysate was digested with proteinase K at 55°C (10 min) before purified by phenol/ chloroform extraction or chromatography on Qiagen TM resin. Results showed that the sigmoidal arcs were almost the same as Fig. S1B (left panel) (Fig. S1C). This result suggests that struc-tures in the sigmoidal arc are not mediated by obligate proteinprotein or protein-DNA interactions.
In addition, HeLa genomic DNA was digested with Plasmid-Safe TM DNase, an ATP-dependent exonuclease that degrades linear dsDNA to deoxymononucleotides, but not circular dsDNA. Based solely on its enzymatic properties, Plasmid-Safe TM DNase should remove the linear tail from a t-loop, leaving a loop (circular DNA) (Fig. 2, B and C). Plasmid-Safe TM DNase digested ϳ80% of linear telomeres, removing the sigmoidal arc entirely, while generating a new signal corresponding to circular telomeric DNA (Fig. 2C).
In t-loop structure, the telomeric ssDNA 3Ј-G-rich overhang is thought to invade into the dsDNA telomeric repeat

Study t-loop with 2D gel method
tract, adapting a D-loop structure (5). Under this circumstance, the invading G-rich strand would be resistant to exonucleolytic digestion because of lacking a free 3Ј-end. To test this hypothesis, HeLa genomic DNA was digested with exonuclease I (ExoI), analyzed by 2D gel method, and hybridized under native and denatured conditions to G-strand-or C-strand-specific probes. The exact same amount of undigested DNA was used as control. Results showed that DNA in the sigmoidal arc includes G-rich ssDNA (C-probe hybridization) but not C-rich ssDNA (G-probe hybridization) (Fig. 2D). In particular, most of the signal detected by the C-rich probe in the sigmoidal arc is resistant to ExoI, whereas little of the signal detected by the C-rich probe in the linear chromosome fraction is ExoI-resistant (Fig.  2D). These results support the conclusion that the DNA structures in the sigmoidal arc are t-loops, in which the G-rich 3Ј-ss-DNA telemetric overhang invades into the dsDNA telomeric repeat tract to adapt a D-loop.
Moreover, we observed the t-loop structure in the sigmoidal arc by EM. The area of gel corresponding to sigmoidal arc (equal to linear fragments Ն7.5 kb in length) was excised (Fig.  S2, A and B), and DNA was purified and analyzed by EM. In three independent experiments, we observed about 400 DNA molecules, of which 74 were scored as t-loops ( Fig. S2C) with an average size of 11.7 kb, consistent with excised telomeric DNA ranging from 7.5 to 15 kb (Fig. S2D). The loop sizes were varied from 2 to 16 kb with a mean size of 7.8 kb, whereas the t-loop tails range from 0.3 to 11.3 kb with the average length of 3.8 kb (Fig. S2D). Tail lengths of t-loops were positively correlated with their loop sizes, indicating that the bigger loop carries a longer tail in general (Fig. S2E). Linear telomere-homologous molecules excised from the linear dsDNA region were used as a control (Fig. S2A). No t-loop-shaped structure was observed (data not shown). In conclusion, this evidence suggested that the sigmoidal arc detected by the 2D gel method represents t-loops.

Depletion of TRF2 induces the loss of t-loops
TRF2 is a component of shelterin that protects telomeres by repression of ATM signaling and non-homologous end-joining (17,18). Previous studies have revealed that TRF2 has the ability to generate t-loops in vitro (5,19). Recently, it was also demonstrated that TRF2 is required for the formation or maintenance of t-loops in vivo (12). To further verify this conclusion, we constructed TRF2 knockout HeLa cells with two stably expressed sgRNAs (targeting TRFH domain of TRF2) and inducible Cas9 (Fig. 3A), and cells with empty vector were used as control. After 7 days of doxycycline treatment, TRF2 is depleted in sgTRF2 cells (Fig. S3A), resulting in a dramatic increase of chromosome end fusion (from 0 to 32.3%) and a decrease in free chromosome end compared with the control (Fig. 3, D and E, and Fig. S3B). Results of the 2D gel method were quantified, and the percentage of t-loops was calculated as described under "Experimental procedures." Compared with control, t-loops were dramatically decreased in TRF2-deficient cells (30.7 to 12.2%) (Fig. 3, B and C). In addition, we found that inhibition of ATM phosphorylation in TRF2-deficient cells could suppress the chromosome fusion but barely rescue the t-loop disruption (Fig. 3, B-E, and Fig. S3), which means TRF2 is directly responsible for t-loop formation or maintenance, consistent with a previous study (12).

Immediate folding of t-loops after telomere replication during S phase
Our previous studies implied that telomeres must be unfolded (i.e. not in t-loops) at least twice during S phase in proliferating cells: once during S phase to permit telomere replication, and then again at the end of S phase to permit C-strand fill-in DNA synthesis (20,21). In this instance, t-loops could either refold immediately after they are replicated, or they could remain unfolded in an open linear confirmation until C-strand fill-in synthesis completed at the end of S phase. To verify the folding states of t-loops during replication, HeLa S3 cells were synchronized at G 1 /S, released into S phase, and then harvested at different time points, corresponding to early S, middle S, late S/G 2 , and G 1 phase (Fig. 4A). Genomic DNA was isolated and analyzed for the presence of t-loops by the 2D gel method. The results demonstrate that the amount of t-loops is fairly constant throughout the cell cycle (Fig. 4B), which is consistent with the hypothesis that telomeres unfold and immediately refold after being replicated during S phase; therefore, they must unfold and refold a second time to allow for C-strand fill-in DNA synthesis at late S/G 2 phase.
To further confirm that t-loop unfolding/refolding is tightly coupled to telomere replication, HeLa S3 cells were synchronized at G 1 /S phase, released into S phase for 3 h, pulse-labeled for 1 h with BrdU during mid-S phase, and harvested immediately (Fig. 4C). DNA was isolated from the pulse-labeled cells and separated by CsCl density gradient centrifugation according to density. Telomere DNA that had replicated during the 1-h pulse (labeled with BrdU) was divided from unreplicated and previously replicated DNA (unlabeled) (Fig. 4D) and analyzed by 2D gel method. It shows that both newly synthesized leading and lagging telomeres have t-loops (Fig. 4E). Because the density-labeled telomeres represent the cohort of telomeres that were replicating during the 1-h time window of exposure to BrdU, these data indicate that t-loops refold immediately following telomere replication.

Decondensation of telomeric chromatin leads to the disruption of t-loops
The structure of t-loops was compared in primary cells (T cells and BJ fibroblasts), telomerase-positive cells (HeLa S3 and A549), and ALT cells (VA13 and U2OS) (Fig. 1C). The results show that the percentage of t-loops was varied by cell type as follows: T cells from human blood, 24.5%; BJ fibroblasts, 16.9%; HeLa cells, 17.2%; A549, 17%; U2OS cells, 6.5%; VA13 cells, 7.3% (Fig. 1C). These results are consistent with a previous report that 6.9 -9.2% of telomeres have t-loops in ALT cells and 15-40% in telomerase-positive HeLa cells (5,22). t-circles were observed in VA13 and U2OS cells, but they were barely detectable in other cell lines (23,24).
In addition, it has been reported that telomeric DNA is less compacted in ALT cells than in telomerase-positive cancer cells (25). To explore whether the decondensation of telomeric chromatin is correlated with the low abundance of t-loops,

Study t-loop with 2D gel method
non-ALT HeLa cells were treated with TSA, a histone deacetylase inhibitor that regulates histone acetylation indirectly. After 48 h of treatment with TSA, the density of acetylated histones H3 and H4 on telomeres increased, suggesting the level of telo-meric heterochromatin decreased (Fig. 5, A and B). Consistently, micrococcal nuclease (MNase) assay showed that up to 60% of telomeric DNA in HeLa cells treated with TSA was digested to mononucleosomes by MNase (16-min digestion),

Study t-loop with 2D gel method
similar to U2OS cells (65% of telomeric DNA was converted to mononucleosomes by MNase digestion for 16 min), whereas only 30% of telomeric DNA in untreated HeLa cells was converted to mononucleosomes under comparable digestion conditions (Fig. 5C). When HeLa cells were exposed to TSA, the intensity of the t-loops decreased in a dose-dependent manner (Fig. 5, D and E). In addition, we also observed that the t-circle is increased after TSA treatment, a mechanism that needs further research (Fig. 5D).

Loss of t-loops is associated with activation of homologous recombination at telomeres
Linear chromosomes lacking functional telomeres induce a DNA damage response (DDR), one marker of which is formation of 53BP1 foci at DNA damage sites (26). Here, the colocalization of antibody to 53BP1 and a hybridization probe for telomeric DNA sequences were used to assess the abundance of telomere dysfunction-induced foci (TIFs) (27). The results showed that ϳ58% of TSA-treated HeLa cells have at least one TIF, compared with ϳ20% in DMSO-treated HeLa cells (Fig. 6,  A and B), indicating the DNA damage level increasing at telomere in cells exposed to TSA.
Aggregates of nuclear protein, known as PML bodies, begin to accumulate after HeLa cells are exposed to TSA (Fig. 6C). Some of these PML bodies are associated with telomeric chromatin, forming APBs. 48 h after treatment with TSA, the number of PML bodies, as well as APBs, in HeLa cells increased dramatically, to include up to 50% of telomeres (data not shown). Quantitative analysis showed Ͼ9 APBs/ cell in ϳ27% of TSA-treated cells and Ͼ4 APBs/cell in ϳ64% cells (Fig. 6D). In control cells treated with DMSO, 98% cells had Ͻ3 APBs/cell and in 42% of the cells, no APBs were detected (Fig. 6D).
APBs are often associated with telomeric recombination in human ALT cells (28). To this end, chromosome orientation-FISH (CO-FISH) was performed to determine the frequency of T-SCEs (29). By scoring the signals on 1688 and 1633 chromosomes from control cells and TSA-treated cells, respectively, we found that although T-SCEs were rare in control cells (2.4%), TSA-treated cells exhibited a more than 10-fold increase in the frequency of T-SCEs (30.7%) (Fig. 6, E and F), indicating the occurrence of HR at telomeres. In addition, it is possible that telomeric HR, which is induced by TSA-mediated chromatin decondensation, subsequently results in loss of t-loops. If so, unchanged t-loops would be expected if HR is inhibited in TSA-treated cells. To test this, HeLa cells were co-treated with TSA and B02, the latter is a specific inhibitor of Rad51 that plays an essential role in HR. We still observed a decrease of t-loops (Fig. S4, A and B) and generation of a t-circle, demonstrating that the loss of t-loops is not caused by HR. We also showed that TSA and/or B02 treatment did not significantly alter telomere length (Fig. S4C). In addition, telomere repeat-containing RNA (TERRA) level transcribed from 6p, 7p, XpYp, 7q, and 17q telomeres was not changed by TSA treatment (Fig. S4D). D, samples prepared as described in C were applied to a CsCl density gradient. Differential density separates sample into three peaks, corresponding to unlabeled telomere, leading and lagging telomere fragments, as indicated. E, T-loop formation in DNA fragments, corresponding to leading and lagging telomeres, were analyzed by 2D gel method in the same gel side-by-side.

Detection of t-loops by nondenatured 2D agarose gel electrophoresis
Neutral/neutral 2D gel method, which detects native DNA structure in a gentle condition, was widely used in studying DNA RCIs in different kinds of species (30 -33). In recent 20 years, the 2D gel method has been widely used in studying the dynamics of special telomeric structures, including G/C-rich overhang, extrachromosomal telomeric circle, and t-circle tail (34 -37). Based on the structural similarity between t-loops and RCIs, in this study, we developed a new method that employs 2D gel electrophoresis followed by Southern blotting to detect t-loops in cell-free extracts. To detect t-loops/sigmoidal signal using 2D gel electrophoresis, genomic DNA needs to be isolated under moderate conditions to avoid the acute vortex and high-temperature digestion of proteins. Moreover, electrophoresis was carried out at low-voltage/4°C to avoid the heat accumulation that may disrupt t-loop structure. The following evidence indicated that the sigmoidal arc observed in our 2D gel method represents t-loops. First, this structure is a part of the chromosome and is sensitive to heating. Second, this structure can convert into circular DNA by Plasmid-Safe TM nuclease that

Study t-loop with 2D gel method
specifically digests linear DNA but not circular DNA. Third, the structure consists of a single-stranded G-rich DNA that is resistant to ExoI digestion. Fourth, the result that ALT cells have much less t-loops compared with telomerase-positive cells observed by our method is consistent with the observations by EM (5,22). Finally, we observed t-loops by EM with DNA extracted from the sigmoidal arc, and the t-loops showed similar loop and tail sizes as a previous result from other groups (5,12). All of these results demonstrate that this method can document the t-loop structure.
In addition to the biochemical and EM validation of t-loop analysis using our 2D gel method, we also test the robustness of this method by reproduced necessity of TRF2 on stabilization of the t-loop. TRF2 has been reported to be essential for t-loop formation and maintenance (12). Here, we showed that t-loops decreased when TRF2 was depleted independent of chromosome fusion frequency. These results are consistent with previous reports (12,38,39) that further verified the authenticity of the 2D gel method.

Advantage and disadvantage of 2D gel method to detect t-loops
The study of t-loops has been challenging previously, due to lack of a sensitive and handy method. EM and STORM are reliable methods to visualize the t-loops directly (12). However, the application of these two methods is limited due to technical difficulties such as cross-linking levels and sample preparation (5,12,22). The 2D gel method we presented here is a widely applicable method for it combined 2D gel electrophoresis and in-gel hybridization with telomeric probes. This method has

Study t-loop with 2D gel method
several obvious advantages. First, it does not require special equipment and technology. Second, only a small amount of DNA (5-10 g of genomic DNA) is needed. Third, it is capable of analyzing multiple samples simultaneously in the same condition. Finally, it can be finished in a couple of days. Therefore, this method provides a convenient way to study t-loop, which is applicable to almost all researchers.
Although the 2D gel method shows many advantages as described above, a few limitations still exist. First, because the 2D gel method is performed without DNA cross-linking, a part of the t-loops would be disrupted during DNA purification, digestion, and electrophoresis, which is why we detected less t-loop (17% in HeLa) than EM results (15-40% in HeLa) (5). Second, the sigmoidal arc represents a converged fraction of physically similar t-loops (Fig. 1A), whereas parts of the t-loops migrating to the region between the circle and sigmoidal arc (gray curves in Fig. 1A) are invisible for low abundance. Third, the quantification of the sigmoidal arc (t-loop) is based on intensity analysis of Southern blotting (relative ratio), which was not equal to the molecular number. However, even in this case, it still shows a similar relative abundance of t-loops between 2D gel method and EM (5,22), suggesting that t-loop quantification using 2D gel method is reliable.

t-loop formation during cell cycle
The dynamics of t-loop formation during the cell cycle and heterochromatin regulation of t-loop structure are fundamental issues still poorly understood. Previously, it has been proposed that t-loops must unfold during S phase to allow replication processing (40), and C-rich strand fill-in also requires an "open" telomere structure (20,21). Our results showed that t-loops are consistently present at chromosome ends throughout the cell cycle, and its formation is closely coupled with telomere replication, i.e. t-loops are immediately formed after the replication is finished (Fig. 4). These results imply that t-loop refolding follows closely with telomere synthesis; in other words, t-loop unfolds during late S/G 2 phase and immediately re-folds after the C-rich strand synthesis is completed. The apparent need for the consistent presence of t-loops at chromosome ends further suggests that they play a critical role in protecting linear chromosome ends at all stages of the cell cycle.

Heterochromatin state of telomeres and the maintenance of t-loops
A previous study (41) showed that telomeres and subtelomeric regions exist as condensed heterochromatin. Mammalian telomeric repeats are characterized by low levels of acetylated H3 and H4 and high levels of trimethylated lysine 9 in histone H3 (H3K9me3) and lysine 20 in histone H4 (H4K20me3) (42). In yeast, the formation of back-folding loops depends on the activities of several histone deacetylases that play roles in chromatin silencing (43). Consistent with these findings, our data demonstrate that TSA-induced decondensation of telomeric chromatin triggers disruption of t-loops (Fig. 5). In general, heterochromatin regions are less accessible to DNA-modifying proteins and nucleases (44,45). As such, the heterochromatic state might help to protect t-loops against DNase attack, helicase unwinding, and other undesired processes.
As expected, in association with the loss of t-loops, significantly increased TIFs were observed in cells treated with TSA (Fig. 6, A and B), indicating that the deprotection of chromosome ends activates DDR. The frequency of T-SCEs also increased after TSA treatment (Fig. 6, E and F), which may be caused by the disruption of t-loops. In support of this hypothesis, human ALT cancer cells display much less t-loops and a higher frequency of telomeric HR, compared with non-ALT cells (Fig. 1C and 6E) (22,25).

Cell culture, treatment, and transfection
HeLa S3, A549, and U2OS cells were grown in DMEM (Hyclone) containing 10% calf serum (PAA), 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C under 5% CO 2 . Human foreskin fibroblasts (BJ) and VA13 were grown in DMEM (Hyclone) containing 10% FBS (PAA). For drug treatment, HeLa cells were treated with 0.5 M trichostatin A (unless indicated otherwise) or 27.4 M B02 (Millipore) for 48 h. For TRF2 knockout, we used inducible CRISPR/Cas9 system. First, HeLa cells were transfected with lentivirus carrying inducible Cas9 (pHAGE-TRE-Cas9) and treated with neomycin for 10 days to obtain a inducible Cas9-HeLa cell line. Second, two lentiviruses with individual sgRNAs targeting to TRF2 were sequentially transfected into cells, and the empty vector was used as control. Two rounds of selection with puromycin (2 g/ml) and blasticidin (10 g/ml) were used to select cells express sgRNAs stably. The sequences of sgRNAs are as follows: sgTRF2-1-F GCCTTTCGGGGTAGCCGGTA and sgTRF2-1-R TACCGGCTACCCCGAAAGGC; sgTRF2-2-F GAACCCGCAGCAATCGGGACA and sgTRF2-2-R TGTC-CCGATTGCTGCGGGTTC. Third, cells with sgTRF2 were treated with doxycycline (1 g/ml) for 7 days to induce Cas9 expression and deficient TRF2, and KU60019 (10 g/ml, inhibitor of phosphorylation of ATM) was added at the last 4 days during doxycycline treatment to inhibit ATM phosphorylation. Cells with sgCtrl were treated in same condition as controls.

Cell cycle synchronization
HeLa S3 cells were synchronized at G 1 /S phase by double thymidine block as described previously (20). Cells were then released into S phase and harvested 3, 6, 8, 10 or 15 h after release. Genomic DNA was purified from recovered cell pellets immediately after harvest. For FACS analysis, cells were fixed with 75% ethanol and stained with propidium iodide.

Genomic DNA isolation
Unless indicated otherwise, genomic DNA was extracted using AxyPrep TM blood genomic DNA miniprep kit (Axygen Biosciences, Union City, CA) following the manufacturer's instruction. All steps were performed at room temperature omitting vortexing. Genomic DNA was also isolated using the

Study t-loop with 2D gel method
DNeasy kit (Qiagen, Valencia CA) and using phenol/chloroform extraction as the primary deproteinization step (5), after which DNA samples prepared by different methods were used for side-by-side comparison.

Neutral-neutral 2D agarose gel electrophoresis and Southern blotting (2D gel method)
2D agarose gel electrophoresis was performed as described previously (46,47). Briefly, 10 g (5 g at least) of genomic DNA was digested with RsaI and HinfI (New England Biolabs) and loaded onto a 0.4% agarose gel, and electrophoresis was carried out in 1ϫ TBE at 1 V/cm for 12 h at room temperature. The lane containing DNA was excised from the gel, and the gel buffer was exchanged with 1ϫ TBE with 0.3 g/ml ethidium bromide (Sigma). The gel slice was placed and cast with 1% agarose gel in 1ϫ TBE containing 0.3 g/ml ethidium bromide. The gel was run at 4°C for 6 h at 3 V/cm. In-gel hybridization analysis of telomeric DNA was performed as described previously (35) with minor modification: the gel was dried for 1 h at 50°C and denatured for 30 min in 0.5 M NaOH and 1.5 M NaCl. The gel was rinsed several times in distilled water, neutralized with 0.5 M Tris-HCl, 1.5 M NaCl (pH 8.0), and hybridized with a telomere probe in 1ϫ hybridization buffer (2 g/ml sonicated Escherichia coli DNA, 10ϫ Denhardt's buffer, 0.5% SDS, and 5ϫ SSC) at 42°C overnight. The gel was washed four times in washing buffer (2ϫ SSC, 0.5% SDS) and exposed to a PhosphorImager screen. Telomere probes (G-rich and C-rich) were prepared as described previously (20).
The percentage of t-loops was calculated as follows: (intensity of signal in the sigmoidal arc)/(intensity of total signal) ϫ 100%.

EM
DNA from the sigmoidal arc was purified with Qiagen Min-Elute Gel-extraction kits (Qiagen, catalog no. 28604). The enriched DNA in TE buffer (10 mM Tris-HCl (pH 8.0), and 1 mM EDTA) was applied to a layer of benzalkonium chloride (BAC) (6 l of deionized formamide, 0.6 l of 40% glyoxal, 1 l of enriched DNA and 1 l of 0.02% BAC in 12 l) and transferred to pretreated thin carbon membrane-coated EM grids. After rinsing with distilled water, DNA was stained with 0.75% formic acid uranium at near neutral pH and dehydrated with 5 l of 100% ethanol. The EM grids were examined under a JEM-100CXII transmission electron microscope. Images were captured using an OSIS MEGAVIEW G2 CCD camera. The dimension of t-loops was measured by iTEM software (TEM IMAGING PLATFORM, Olympus).

ChIP
ChIP was essentially done as described previously (49). The following antibodies were used in this study: anti-H3K9ac (rabbit, Sigma); anti-H4ac (rabbit, Millipore); and human IgG (Sigma). Telomeric DNA was detected by hybridization with a telomere-specific probe.

Immunoblot
Antibody to TRF2 (mouse, Millipore) was used in Western blotting to determine the expression level of TRF2 in control and CRISPR/Cas9 knockout cells.

TRF assay
The telomere length assay was performed as described previously (35).

Statistical analysis
Two-tailed unpaired Student's t test was used for analysis between two samples (Graphpad Prism). Error bars represent the mean Ϯ S.D. of no less than three biological repeats/independent experiments: *, p Ͻ 0.05; **, p Ͻ 0.005; ***, p Ͻ 0.001. For the sample treated with more than two samples, analysis was performed by one-or two-way ANOVA (Graphpad Prism), and p values are indicated in the figure legends.