SIRT6 facilitates directional telomere movement upon oxidative damage

Oxidative damage to telomeres leads to telomere attrition and genomic instability, resulting in poor cell viability. Telomere dynamics contribute to the maintenance of telomere integrity; however, whether oxidative damage induces telomere movement and how telomere mobility is regulated remain poorly understood. Here, we show that oxidative damage at telomeres triggers directional telomere movement. The presence of the human Sir2 homolog, Sirtuin 6 (SIRT6) is required for oxidative damage-induced telomeric movement. SIRT6 knock out (KO) cells show neither damage-induced telomere movement nor chromatin decondensation at damaged telomeres; both are observed in wild type (WT) cells. A deacetylation mutant of SIRT6 increases damage-induced telomeric movement in SIRT6 KO cells as well as WT SIRT6. SIRT6 recruits the chromatin-remodeling protein SNF2H to damaged telomeres, which appears to promote chromatin decondensation independent of its deacetylase activity. Together, our results suggest that SIRT6 plays a role in the regulation of telomere movement upon oxidative damage, shedding new light onto the function of SIRT6 in telomere maintenance.

The ends of chromosomal DNA are organized into specialized DNA-protein structures, telomeres, which consist of a shelterin protein complex and capped TTAGGG DNA repeats. The highly conserved shelterin complex protects chromosome ends from degradation and an inappropriate DNA damage response (DDR), thus preserving chromosome stability and integrity 1,2 . Telomeres typically display random dynamic movement within the nucleus. In cells using the alternative telomere lengthening 3 pathway, long-range directional movement of telomeres results in telomere clusters and their association with PML bodies 4 . When imaged and traced with PNA probes in ALT-positive U2OS cells, the majority of telomeres show slow diffuse movement confined to a radius of <5 μm. However, up to 15% of telomeres display non-confined movement with a high mobility, with some showing significant directional movement within 1 hr after damage 5 ; this movement is likely damage-induced. Telomere movement is also affected by the chromatinized telomeric structure, the association of telomeres with the nuclear matrix, the interaction of telomeres with certain proteins, and by meiosis and mitosis 6 . Thus, telomere movement must be an intricately regulated process. Although increased telomere movement facilitates efficient telomere damage repair or telomere elongation in ALT cells, the mechanisms by which telomere mobility is regulated remain unclear. Oxidative damage to telomeres leads to telomere attrition and genomic instability and is most likely a major cause of incomplete ends of replicated chromosomes 7 . Indeed, oxidative damage has been shown to induce telomere shortening 8 . However, whether oxidative damage affects telomere mobility still needs to be explored, and again, it is not known which regulatory mechanisms are involved in telomere movement.
To address these questions, we utilized a unique approach, DNA damage targeted at telomeres (DART at telomeres) [9][10][11][12] . In this approach, KillerRed (KR)-tagged shelterin proteins act as "controllable telomere bombs" to generate localized oxidative damage at telomeres upon activation by light. The telomere specificity and light controllability of the KillerRed system make it an ideal tool to investigate the mechanisms by which telomere integrity is maintained in response to oxidative damage immediately after such damage 9-12 . SIRT6 is known to be a multifunctional protein implicated in DNA repair and the maintenance of telomere integrity 13 . SIRT6 functions as an NAD + -dependent histone (H3K9 and H3K56) deacetylase of telomeric chromatin [14][15][16] . Although SIRT6 deficiency leads to genomic and telomeric instability, metabolic defects, and aging-related degenerative pathologies in mice 15,[17][18][19][20] , it is not known whether and how SIRT6 is involved in telomere dynamics. Here, we found that oxidative DNA damage at telomeres enhances directional telomere movement within 1 min and that SIRT6 is essential for this process. Chromatin decondensation at damaged telomeres was observed in WT but not in SIRT6 KO cells. We found that SIRT6 recruits SNF2H, an ATP-dependent chromatin-remodeling factor, to damaged telomeres. Our results have shed new light on how telomere stability, movement, and chromosomal condensation are regulated by SIRT6 in the presence of oxidative damage at telomeres.

Results
Oxidative damage at telomeres leads to increased directional telomere movements in the short term. We previously found that oxidative damage, the most frequent form of DNA damage, induces telomere shortening and cell death 8 . To determine whether telomere oxidative damage affects telomere mobility, KR tagged TRF1 (KR-TRF1) was transfected into cells for telomere labeling and telomere-specific oxidative damage induction. KillerRed is a 550-580 nm light-activated fluorescent protein which releases localized superoxide upon activation. In our system, half of a cell nucleus was irradiated with a full power 559 nm laser in a confocal microscope to activate KR; consequently, only telomeres within the region of the light illumination are damaged. The activation process takes a few seconds while the other half of the cell nucleus remains an undamaged control with the same basal cell movement. The real time telomeric movements were manually tracked for 120 frames with a time interval of 0.429 s, in two dimensions (2D), immediately after damage induction (Fig. 1a). The mobility was quantified as the mean square displacement (MSD; described in Materials and Methods), which measures the square of the distance that one particle travels between two images within a time interval. The definition and calculation of MSD in this study follow a previous study that measured telomere dynamics in cells 21 . To evaluate the system drift, we used fixed U2OS cells (transiently expressing KR-TRF1) or fluorescence nanobeads and imaged as described. The measured MSD from the system drift was less than approximately 8.8 × 10 3 nm 2 , which is significantly smaller than the measured telomere movement in live cells (Figs 1b and S1). Therefore, such a small drift can be ignored compared to damage induced telomere dynamics (see below).
Representative tracking maps of individual telomeres immediately after damage induction, with a time interval of 0.429 s for 120 frames, are shown in Fig. 1c. Similar to the described behavior of genome foci with variable trajectories 22 , telomeres in live cells also display intrinsic trajectories. Some moved large distances with potential directionality, while others moved randomly or freely within a confined region inside the nucleus. This is expected since chromatin and telomeres are moving around during cell division and other cell proliferation processes.
According to the movement model described for genome foci 22 and the method used in single particle tracking analysis in 2D 23 , when the MSD/time is fitted into a power law function model where the value of α is used to define the type of telomeric movement, α > 1 indicates directional movement displaying a curve with an increasing slope, α = 1 indicates a simple random walk, and α < 1 indicates a random walk in a confined region (Fig. 1c). Movement of each individual telomere before and after damage induction is shown in Fig. 1d, indicating a trend of increase of MSD after damage (Fig. 1e).
Oxidative damage-induced telomere mobility is not dependent on specific telomere probes.
To confirm that the increased directional movement of telomeres is caused by damage, RFP-TRF1 was used to fluorescently label telomeres to exclude the effects of TRF1 expression and light illumination. RFP-TRF1 was imaged within 1 min as shown in Fig. 1a, and the mobility was quantified. The MSD due to telomere movement observed in RFP-TRF1 tagged telomeres in live U2OS cells was substantially greater than 8.8 × 10 3 nm 2 (Fig. 2a), thus confirming that telomere movement occurred in the nucleus in undamaged cells. The MSD/time curve and maximum MSD value of RFP-TRF1 labeled telomeres showed no significant difference with and without the light activation, indicating that TRF1 expression and the 559 nm laser light scanning does not affect the pattern of telomere movement (Fig. 2a), suggesting that RFP-TRF1 does not affect telomere movement in a light-dependent manner. Again, when damaged by light-activated KR-TRF1, telomeres in U2OS cells showed increased mobility within 1 min of damage induction, as indicated by higher average MSDs of damaged telomeres relative to undamaged telomeres (Fig. 2b).
To prove that the effects of KR-TRF1 on telomere movement are not specific to this particular fused protein, we used KR-TRF2 to generate telomere oxidative damage in U2OS cells (Fig. 2c) and KR-dCas9 and a telomere-targeting sgRNA to track telomere movement (Fig. 2d). Our previous studies indicate that both KR-TRF2 and KR-dCas9 + sgRNA at telomeres induce localized telomere damage specifically [9][10][11][12] . Increased short-term telomere mobility and increased telomeric movement were observed after damage induction in both systems (Fig. 2c,d). These data together indicate that oxidative telomeric damage immediately induces increased telomere mobility regardless of telomere probes. Accordingly, the MSD/time of damaged telomeres showed a shift towards directional rather than random movement compared with undamaged telomeres (Fig. 2e). As shown in Fig. 1c,d, undamaged telomeres do not significantly reach MSD 5 × 10 4 nm 2 ; we therefore set a threshold of MSD = 5, and values above this threshold indicate a telomere that displays particularly high dynamics. We compared the rate of MSD >5 × 10 4 nm 2 with/without damage induction by light illumination in KR-TRF1 cells, and  SIRT6 is critical for directional movement in DNA damage repair. SIRT6 deficiency leads to telomeric instability and aging-related degenerative pathologies in mice 15,[17][18][19][20] . To understand whether the function of SIRT6 is involved in regulating telomere mobility in response to oxidative damage, WT and SIRT6 KO mouse fibroblast (MEF) cells were damaged with light-activated KR-TRF1. Damaged telomeres in WT MEF cells showed increased movement following damage induction (Fig. 3a). SIRT6 KO MEF cells exhibited an increased basal level of mobility compared to that in WT cells before damage; however, the cells did not show the damage-induced increase in total movement after damage (Fig. 3b). Overexpression of GFP-SIRT6 in SIRT6 KO cells leads to increased movement after damage, showing that SIRT6 increases the short-term telomere mobility in response to telomere-specific oxidative damage (Fig. 3c). Given that SIRT6 functions as an NAD + -dependent histone (H3K9 and H3K56) deacetylase of telomeric chromatin [14][15][16] , we tested the effects of the SIRT6 deacetylase mutant H133Y 24,25 . The SIRT6 H133Y mutant increases the telomere movement after damage as well as the WT SIRT6 (Fig. 3d,e), indicating that the deacetylase activity of SIRT6 is not essential for SIRT6-dependent telomere mobility.
SIRT6 cooperates with SNF2H in the regulation of telomere chromatin structure. SIRT6 is also known to interact with the chromatin remodeling factors γH2AX and SNF2H 26,27 . Our previous study indicates that SNF2H is necessary for the efficient repair of double strand breaks (DSBs) via its ATPase chromatin remodeling activity which relaxes chromatin after damage 28 . SIRT6 recruits SNF2H to DSBs at genomic sites, where it functions in chromatin decondensation and subsequent recruitment of repair factors 27 . Since the deacetylase activity of SIRT6 is not essential for damage-induced telomere movement (Fig. 3e), we next examined if SNF2H is involved in chromatin decondensation at damaged telomeres as well as telomere movement. As indicated in a previous study and proven by our experiments, SNF2H is down-regulated in SIRT6 KO cells 27 . We overexpressed GFP-SNF2H in SIRT6 KO cells and found that SNF2H expression in SIRT6 KO cells partially restored increased telomere movement after damage (Fig. 4a), indicating that the function of SIRT6 in telomere movement might be partially mediated by SNF2H. Next, we examined the damage response of SNF2H at telomeres. Both GFP-SNF2H and SIRT6 are recruited to sites of oxidative telomeric damage but not to sites tagged with DsRed-TRF1 (DR-TRF1) or to KR-TRF1 without light activation (Fig. 4b,c); importantly, efficient recruitment of SNF2H to damaged telomeres is dependent on SIRT6 (Fig. 4d). Consistent with our finding that deacetylase activity is not required for damage-induced movement, SIRT6 H133Y complements both the recruitment of and interaction with SNF2H (Fig. 4e,f). These results suggest that the cooperation of SIRT6 and SNF2H at damaged telomeres does not need the deacetylase activity of SIRT6. Since SNF2H functionally rescues the SIRT6 deficiency and responds to telomere damage, these results together indicate that SNF2H might be a mediator of SIRT6 in the regulation of telomere movement through chromatin remodeling.
Chromatin decondensation contributes to telomere movement after damage. SNF2H has been shown to mediate chromatin remodeling and decondensation 29 . To understand whether chromatin decondensation is linked to telomere movement, we treated cells with the histone deacetylation inhibitor Trichostatin A (TSA) to artificially decondense chromatin [30][31][32] , and observed the mobility of damaged telomeres (Fig. S2). H3K9 acetylation (H3K9Ac) generally marks an opened chromatin structure. TSA treatment led to an increased H3K9Ac, confirming that the overall chromatin structure is largely decondensed. We accordingly observed a significantly increased movement of telomeres upon TSA treatment, indicating that chromatin decondensation might contribute to telomere movement (Fig. S2).
Next, we investigated whether the increased telomere movement was coupled with decondensation of chromatin at damaged telomeres, as indicated in Fig. 5a. To examine chromatin changes in response to oxidative telomeric damage, fluorescence lifetime imaging (FLIM) was used to assess the chromatin decondensation by fluorescence lifetime (FL), in which increased FL indicates decondensed chromatin structure 33,34 . In WT cells, telomere damage clearly led to chromatin decondensation compared to cells without damage (Fig. 5a). Furthermore, in contrast, SIRT6 KO cells did not show any changes of FL after damage (Fig. 5b), supporting the notion that SIRT6 is necessary for chromatin decondensation after damage. Interestingly, the quantification of local FL at damaged telomeres in MEF WT/SIRT6 KO cells shows that telomeres in SIRT6 KO cells are less decondensed (Fig. 5c), further confirming the lack of chromatin decondensation at damaged telomeres in SIRT6 KO cells. Thus, oxidative damage to telomeres leads to decondensed chromatin in WT cells, and chromatin decondensation could not occur in the absence of SIRT6. Statistical analysis of the FL of global chromatin shows that the levels of decondensation also align with the levels of telomere movement in SIRT6 WT and KO cells (Fig. 3). All of these data together indicate that SIRT6 might regulate telomere movement after oxidative telomeric damage, and this process might be mediated by promoting chromatin decondensation locally and globally.
To further evaluate the chromatin states at telomeres, we stained H3K9Ac before and after damage induction. In U2OS cells, we observed an increased H3K9Ac level at KR-TRF1 damaged telomeres compared to RFP-TRF1 labeled telomeres (Fig. 5d), which is consistent with the observed chromatin decondensation at telomeres after damage. H3K9Ac staining at telomeres significantly increased in WT MEF cells after damage; in contrast, the Movement of telomeres with KR-TRF1 with/without damage was analyzed as described in Fig. 1c. Red curve represents directional movement; blue curve represents random walk. MSD of undamaged telomeres is lower than 5 × 10 4 nm 2 , which is indicated as a gray dotted line in the graph. (e) The MSD-time fit; MSD of damaged telomeres (red) versus the undamaged ones (black) was plotted.

Discussion
Maintenance of telomere integrity is associated with telomere dynamics. Here, we showed that oxidative damage at telomeres leads to increased directional telomere movements in the short-term immediately after damage induction. Damage-induced telomere movement is known to be required for damage repair. For example, DSBs, the most deadly type of DNA damage, have been reported to increase telomere mobility in ALT positive cells over 1 hr in dynamic tracking experiments 35 . ALT-positive cells use ALT preferentially when telomerase is nonfunctional or not present in the cell, and thus lengthen telomeres using homologous recombination (HR) machinery coupled with directional telomeric movement for a homology search over a relatively long term. We observed increased telomere mobility immediately after damage induction, which was reflected in increased speed and increased potential directionality within seconds. Regarding the question of how fast telomeres respond to DNA damage, it is worth noting that for the first time, the KillerRed targeted damage induction system made it possible to observe telomere dynamics directly following damage in a real-time fashion. Previous telomere labeling using the PNA probe cas9 or YFP-TRF1 was unable to induce specific damage at telomeres 21 . While the flag-TRF1-folk1 system can induce DSBs by folk1 at telomeres, folk1 is an endonuclease, and thus its exact cutting time is variable, even with drug induction, making it inefficient for studying long-term effects of DSB induction 35 . Control of KR-TRF1 activation and reactive oxygen species (ROS) release through light allowed for carefully controlled damage induction and observation of dynamics directly following oxidative telomeric damage. Importantly,  Fig. 1a. Damage was induced using a 559 nm laser light full-power scan, and average MSD was calculated (left). p < 0.001. (c) U2OS cells were transfected with KR-TRF2 and movement was measured as described in Fig. 1. KR-TRF2 labeled telomeres show a similar trend of increased mobility after KR induced oxidative telomere damage. p < 0.001. (d) Cas9 + GFP labeled telomeres show increased mobility after KR-induced oxidative telomere damage. U2OS cells were co-transfected with Cas9 + GFP, sgRNA for telomeres, and KR-TRF1. Telomeres were tracked with GFP-Cas9 after KR activation. (e) The percentage of directional movement in damaged telomeres vs. undamaged ones from three independent experiments (n is from 97-147 in each group of experiment). p = 0.0068. ** indicates p < 0.05 (f). Percentage of the MSD distribution of telomeres with/ without damage in KR-TRF1 cells from the same set of data in Fig. 1e. p = 0.0052; ** indicates p < 0.05.  Fig. 1a. Damage was induced using a 559 nm laser light full-power scan 36 hr after transfection, and the average MSD was calculated. p < 0.001. (b) MEF SIRT6 KO cells expressing the KR-TRF1 telomere label were imaged using the method shown in Fig. 1a. Damage was induced using a 559 nm laser light full-power scan, and average MSD was calculated. (c,d) KR-TRF1 labeled/ damaged telomeres in MEF SIRT6 KO cells transfected with GFP-SIRT6 or a GFP-vector were imaged directly after damage induction using a 559 nm laser light full-power scan, and average MSD was calculated. p < 0.001. (e) WB (left) shows the expression of GFP-SIRT6 and GFP-H133Y in MEF SIRT6 KO cells. KR-TRF1 labeled/ damaged telomeres in MEF SIRT6 KO cells transfected with GFP-SIRT6 or GFP-H133Y were imaged directly after damage induction using a 559 nm laser light full-power scan, and average MSD was calculated. One hundred telomeres were used for analysis and three independent experiments were done in this analysis.  increased telomere dynamics were observed regardless of probe or cell line, when we labeled the damaged telomeres with either TRF1 or TRF2, or sgRNA + dCas9 labeled telomeres, in several cell lines via dynamic tracking of telomeres within 1 min (Fig. 2). Oxidative telomeric damage alters telomere movement in both mode (from random walk to directional movement) and speed (from slow to fast), which appears to present an efficient defense of telomeres upon damage in a real-time fashion.
In exploring how oxidative DNA damage-induced telomere dynamics over the short term are regulated, we found that SIRT6, a multitasking protein related to DNA damage repair and aging, is an important factor for this process. At the cellular level, SIRT6 KO causes increased telomere instability, promoting telomere damage coupled with increased 53BP1 foci, telomere fusions, and telomere shortening. This leads to severe aging pathology in mice 17 . Our results extend our knowledge of the roles of SIRT6 at telomeres and indicate that SIRT6 regulates telomere mobility in response to oxidative telomere damage. SIRT6 KO cells exhibit decreased chromatin decondensation at telomeres after damage and less recruitment of SNF2H to telomere damage. SNF2H is one of the ISWI chromatin remodeling complexes. It is known that human ISWI complexes are targeted by SMARCA5 ATPase and SLIDE domains to help resolve lesion-stalled transcription 36 . The function of SIRT6/SNF2H is also required to form efficient γH2AX foci 37 . The observation that SNF2H is recruited to damaged telomeres in a SIRT6-dependent manner indicates that SIRT6 might coordinate with SNF2H to regulate chromatin states after damage. Future studies are needed to directly assess the mechanisms by which SIRT6 and SNF2H function at telomeres in chromatin decondensation. How damage-dependent telomere movement affects DNA repair efficiency also needs further exploration.
Finally, we propose a model of how SIRT6 regulates telomere dynamics (Fig. 6). In WT cells, SIRT6 recruits and is coordinated with SNF2H. SNF2H contributes to chromatin decondensation. In SIRT6 KO cells, SNF2H is not efficiently recruited to telomere damage, chromatin is not decondensed, and telomere movements are impaired (right arrow). Thus, we speculate that SIRT6-dependent recruitment of SNF2H results in local chromatin decondensation at telomeres, facilitating telomere dynamics. While up-regulation of SIRT6 may be a viable mechanism for increased cellular proliferation, it also makes it a viable chemotherapeutic target. Since the deacetylase activity of SIRT6 is not required for maintaining telomere dynamics and tumor cells generally encounter increased oxidative stress, our study also suggests the possibility of inhibiting telomere lengthening and cellular proliferation in tumors by targeting SIRT6 at telomeres but not its deacetylase activity.
Live cell imaging. Cells were transfected with KR-TRF1 17 hr after passaging and 24 hr prior to imaging with the Olympus FV/1000 confocal microscopy system (Olympus) and FV/1000 software. Half-cells were selected and irradiated with a 100% power 559 laser (20 scan, 5 mW/scan) to activate the KR-TRF1 marked telomeres in the half-cell, or in the whole nucleus to induce damage. Then, telomere movement was tracked in the cell within the 1:8 zoom in 256 × 256 vision with a 2 μs/pixel scan speed and ZDC on. The time interval of the image acquisition with the above confocal setting is 0.429 s, and the total numbers of frames are 120. All of the experiments follow the same confocal setting as described. Cells were incubated at 37 °C on a thermos plate in normal media during imaging. The movement trend is exactly the same in the whole cell with/without activation of KR-TRF1 as in the half cell. System drift was analyzed using fixed KR-TRF1 or a four-color fluorescent bead (0.1 μm diameter, blue/green/orange/dark red fluorescence, Cat: T7279, Thermo Fisher Scientific).

Mobility analysis.
For the mobility analysis, only cells that were stably attached to the dish surface and showed little deformation were evaluated. Some cells were analyzed after cell movement correction using ImageJ. The analysis of fluorophores (telomere and centromere) tracking was based on the particle localization algorithm, MrSE (maximum radial symmetry estimator). The positions of the fluorophores were determined by calculating the maximum radial symmetry center of the fluorophore. Trajectories were created by linking the nearest fluorophores of the two sequential frames with a maximum tolerance distance of 3 pixels (~300 nm). MSD (mean squared distance) was calculated by: where Δ is the frame cycle (1 sec), n is the number of frames in a time delay, N is the total number of frames, and P (iΔ) is the position at the i th frame. Note that only the fluorophores detected during the entire frames were used to calculate the MSD. For comparison of treatment conditions, nuclei were segmented in each field of view to isolate only the nuclear pixel signal for data analysis using MATLAB. The fluorescence lifetime was fit using a χ 2 test, with DAPI best modeled by a single exponential decay. The average fluorescence lifetime was calculated from all nuclear pixels. Magnitudes of the average fluorescence lifetimes for segmented nuclei in each sample condition were compared using Student's t-test. The fluorescence lifetime at distinct DNA damage sites was determined by obtaining the fluorescence lifetime values at corresponding DNA damage sites labeled for TRF1-KR manually in the Becker & Hickl SPCImage Software. The average fluorescence lifetime of DNA damage sites in SIRT6 WT/KO cells was calculated and compared using Student's t-test. For global KR activation, U2OS cells were transfected with KR-TRF1. After 24 hr, cells were exposed to a 15 W Sylvania cool white fluorescent bulb for 10 min on a stage UVP (Uvland, CA, USA), in which the distance to light was 15 cm, to activate KillerRed.