|Home | About | Journals | Submit | Contact Us | Français|
Telomeres protect the normal ends of chromosomes from being recognized as deleterious DNA double-strand breaks. Recent studies have uncovered an apparent paradox: although DNA repair is prevented, several proteins involved in DNA damage processing and checkpoint responses are recruited to telomeres in every cell cycle and are required for end protection1. It is currently not understood how telomeres prevent DNA damage responses from causing permanent cell cycle arrest. Here we show that fission yeast (Schizosaccharomyces pombe) cells lacking Taz1, an orthologue of human TRF1 and TRF2 (ref. 2), recruit DNA repair proteins (Rad22RAD52 and Rhp51RAD51, where the superscript indicates the human orthologue) and checkpoint sensors (RPA, Rad9, Rad26ATRIP and Cut5/Rad4TOPBP1) to telomeres. Despite this, telomeres fail to accumulate the checkpoint mediator Crb253BP1 and, consequently, do not activate Chk1-dependent cell cycle arrest. Artificially recruiting Crb253BP1 to taz1Δ telomeres results in a full checkpoint response and cell cycle arrest. Stable association of Crb253BP1 to DNA double-strand breaks requires two independent histone modifications: H4 dimethylation at lysine 20 (H4K20me2) and H2A carboxy-terminal phosphorylation (γH2A)3–5. Whereas γH2A can be readily detected, telomeres lack H4K20me2, in contrast to internal chromosome locations. Blocking checkpoint signal transduction at telomeres requires Pot1 and Ccq1, and loss of either Pot1 or Ccq1 from telomeres leads to Crb253BP1 foci formation, Chk1 activation and cell cycle arrest. Thus, telomeres constitute a chromatin-privileged region of the chromosomes that lack essential epigenetic markers for DNA damage response amplification and cell cycle arrest. Because the protein kinases ATM and ATR must associate with telomeres in each S phase to recruit telomerase6, exclusion of Crb253BP1 has a critical role in preventing telomeres from triggering cell cycle arrest.
Taz1 is required for proper telomere DNA replication and prevents end-joining reactions at chromosome ends that occur in the G1 phase of the cell cycle7–9. As fission yeast spend most of the cell cycle in the S and G2 phases, chromosome-end fusions and loss of viability are undetectable in taz1Δ mutants. Despite this, taz1Δ chromosome ends undergo frequent rearrangements10. To determine the nature of these rearrangements, we used a combination of live-cell imaging and chromatin immunoprecipitation (ChIP) to monitor the recruitment of various DNA repair and checkpoint proteins to telomeres in taz1Δ cells.
We endogenously tagged the homologous recombination protein Rad22RAD52 with green fluorescent protein (GFP) and, using Pot1 tagged with monomeric red fluorescent protein (mRFP) as a marker for telomeres (Supplementary Fig. 1), we studied its co-localization to chromosome ends (Fig. 1a). Few wild-type cells have Rad22–GFP foci at non-telomeric sites but these become prevalent when exposed to the radiomimetic drug bleomycin. In contrast, taz1Δ cells had Rad22–GFP foci at telomeres throughout the cell cycle (Fig. 1a). Using ChIP, we also found that taz1Δ cells strongly accumulate another homologous recombination repair protein, Rhp51RAD51, at telomeres (Fig. 1b). Thus, DNA repair is ongoing at taz1Δ chromosome ends.
DNA damage initiates signal transduction pathways, known as checkpoints, which culminate in cell cycle arrest. Telomeres protect chromosome ends from DNA repair and do not induce cell cycle arrest via checkpoint pathways. Unexpectedly, taz1Δ telomeres undergo the DNA damage response and yet these cells did not show any cell cycle delay. We were also able to rule out adaptation phenomena by analysing newly generated taz1Δ mutants; they were also unable to elicit a checkpoint response (Fig. 1c, d).
Next we examined the phosphorylation status of the Chk1 protein kinase, a well-characterized checkpoint target responsible for inhibiting Cdk1 and preventing entry into mitosis in response to DNA damage. Consistent with a lack of checkpoint activation in taz1Δ cells, phosphorylated Chk1–Myc was absent in taz1Δ cell extracts (Fig. 1e). As expected, exposure to bleomycin caused strong Chk1 phosphorylation in both wild-type and taz1Δ cells (Fig. 1e). We performed similar experiments to investigate the phosphorylation status of Cds1, which is involved in DNA-replication checkpoints, and failed to detect any activation of this branch of the checkpoint pathway unless hydroxyurea was added (Supplementary Fig. 2). Thus, even though taz1Δ cells are checkpoint proficient, they are unable to arrest the cell cycle in response to telomeres undergoing the DNA damage response.
The absence of cell cycle delay prompted us to examine the checkpoint pathway to identify at which step it was disrupted. Telomeres in taz1Δ cells are long and accumulate single-stranded DNA throughout the cell cycle11. We started probing the checkpoint pathway (Fig. 2a) at its upstream point by looking for the accumulation of RPA at telomeres by co-localization of Rad11–GFP, the largest subunit of RPA, with Pot1–mRFP. Pot1 protects telomeres from triggering cell cycle arrest via the ATR checkpoint pathway12–14. It has been proposed that Pot1 accomplishes this function directly by outcompeting RPA from telomere ends, as both molecules bind to telomeric single-stranded DNA13. Surprisingly, Pot1–mRFP co-localized with Rad11–GFP in taz1Δ cells (Fig. 2b), indicating that Pot1 does not abrogate checkpoint signalling by exclusion of RPA from dysfunctional telomeres. We quantitatively confirmed increased recruitment of RPA to telomeres in taz1Δ compared with wild-type cells using ChIP (Fig. 2c). Because telomere-associated RPA greatly increases during S phase15, we conclude that we are able to measure protein interaction with telomeres in S-phase cells using an asynchronously growing population.
Next we analysed Rad26ATRIP, Rad9 (a component of the 9-1-1 complex) and Cut5/Rad4TOPBP1. In contrast to wild type, the majority of taz1Δ cells had telomeric foci for all three components (Fig. 2b), indicating that taz1Δ dysfunctional telomeres show increased recruitment for components of the DNA-damage checkpoint pathway. ChIP analysis revealed that Cut5TOPBP1 is also strongly recruited to wild-type telomeres (Fig. 2c), consistent with previous observations that telomeres normally engage DNA damage response components15,16. Rad3ATR kinase was active in growing taz1Δ cells, because hyperphosphorylation of its known substrates, Rad26ATRIP and Rad9, was detected in untreated taz1Δ extracts as in control cells exposed to bleomycin (Fig. 2d). Further analysis revealed that phosphorylation levels of haemagglutinin (HA) tagged Rad26ATRIP observed in taz1Δ cells were otherwise sufficient to trigger cell cycle arrest. Exposure to increasing doses of bleomycin revealed that, at levels of DNA damage where phosphorylation of Rad26ATRIP–HA becomes detectable (0.5–1.0 mU bleomycin; Fig. 2e), Chk1–Myc is clearly hyperphosphorylated. Thus, Rad3ATR activity equivalent to that observed in taz1Δ cells is sufficient to activate a full checkpoint response if DNA damage occurs at non-telomeric regions.
Upon Rad3ATR activation and Cut5TOPBP1 recruitment, Crb253BP1 is brought to sites of DNA damage and, on phosphorylation by Rad3ATR, induces activation of the Chk1 kinase and cell cycle arrest17. Unlike other upstream checkpoint proteins, yellow fluorescent protein (YFP)-tagged Crb2 protein did not co-localize with Pot1–mRFP in most taz1Δ cells (Fig. 3a). In accordance with this, Crb2 was not phosphorylated in taz1Δ cells (Fig. 3b), even though treatment with bleomycin strongly induces Crb2 phosphorylation. We were also unable to detect telomere-bound tandem affinity purification (TAP)-tagged Crb2 in either wild-type or taz1Δ cells using ChIP, whereas controls showed clear recruitment of TAP–Crb2 to an HO-endonuclease-induced DNA double-strand break (Fig. 3c). Therefore, unlike other DNA damage response components, Crb2 is unable to stably bind to telomeres in either wild-type or taz1Δ cells, indicating that the checkpoint signalling pathway is blocked at the step between Cut5TOPBP1 and Crb253BP1.
We next investigated what prevents the stable association of Crb253BP1 with telomeres. Apart from Cut5, Crb2 requires two histone modifications for stable association to double-strand breaks: C-terminal phosphorylation of γH2A (γH2AX in higher eukaryotes) and H4K20me2 (refs 3–5). These modifications show different responses to DNA damage: whereas γH2A/X is locally phosphorylated by ATM or ATR at DNA double-strand breaks, H4K20me2 levels do not change on DNA damage18. Instead, H4K20me2 seems to be ubiquitously distributed even though it participates in the DNA damage response and is essential for sustained checkpoint activation19. Consistent with the observed recruitment of Rad3ATR–Rad26ATRIP to telomeres, we readily detected γH2A by ChIP at both wild-type and taz1Δ telomeres (γH2A; Fig. 3c). In contrast, H4K20me2 was undetectable at telomeres in either wild-type or taz1Δ cells (H4K20me2 telomere; Fig. 3c and Supplementary Fig. 3). The absence of H4K20me2 could not be explained by exclusion of the methyltransferase Set9/Kmt5 from chromosome ends. Telomeres show Set9-dependent mono- and trimethylated forms of H4K20 even though taz1Δ telomeres lacked the latter (H4K20me1 and H4K20me3; Fig. 3c). Similarly, Clr4 methyltransferase, responsible for H3K9 methylation, is not required to maintain the pattern of H4K20 methylation, despite a marked reduction in H4K20me3 (Supplementary Fig. 3). Thus, although H3K9me3 inversely correlates with H4K20me2 at fission yeast telomeres, these are likely to be independently regulated.
Because checkpoint signalling seems to be achieved through the exclusion of Crb2, we reasoned that artificial recruitment of Crb2 to taz1Δ telomeres might be sufficient to induce cell cycle arrest. Taz1 binds to telomeres via its C-terminal Myb domain (MybTaz1); MybTaz1 has been previously used to tether Rap1 directly to taz1Δ telomeres20,21. Likewise, we ectopically expressed a MybTaz1–YFP–Crb2 chimaeric protein in fission yeast. On induction of MybTaz1–YFP–Crb2, taz1Δ cells became extremely elongated (Fig. 3d, e), denoting a full checkpoint response confirmed by Chk1 phosphorylation (Fig. 3f). None of these phenomena was visible in wild-type cells expressing the chimaeric protein (Fig. 3e, f). Also consistent with a checkpoint response, recruitment of MybTaz1–YFP–Crb2 to telomeres in taz1Δ rad3Δ cells resulted in neither cell elongation nor Chk1–Myc phosphorylation (Fig. 3e, f). These data demonstrate that Crb2 recruitment to telomeres in taz1Δ cells is sufficient to restore the full checkpoint pathway.
Next we investigated what marks telomeres different from the rest of the genome. Pot1 and its interacting protein Ccq1 are likely candidates as they bind telomeres independently of Taz1 and prevent checkpoints at telomeres12–14,22. Indeed, in contrast to wild type, both germinating pot1Δ and taz1Δ lig4Δ pot1Δ cells derived from heterozygous diploids became elongated and presented Chk1–Myc phosphorylation (Fig. 4a, b). Cell elongation was dependent on Rad3ATR, indicating an activation of this checkpoint pathway (Supplementary Fig. 4). Therefore, Pot1 is required to prevent telomeres from eliciting Rad3ATR-dependent checkpoints not only in wild-type but also in taz1Δ cells. Ccq1 is required for telomerase recruitment and inhibition of Rad3ATR checkpoints at fission yeast telomeres22,23. Loss of Ccq1 results in progressive telomere shortening and checkpoint activation before complete telomere erosion. Pre-senescent ccq1Δ and trt1Δ single mutants have comparable short telomeres that initiate checkpoints on telomere erosion (Fig. 4c–e). Likewise, ccq1Δ taz1Δ and trt1Δ taz1Δ double mutants undergo ALT-like telomere elongation, preserving long and heterogenous telomeres (Fig. 4c). However, in contrast to trt1Δ taz1Δ double mutants, ccq1Δ taz1Δ mutants had a strong immediate checkpoint response, becoming extremely elongated with strong Chk1 phosphorylation (Fig. 4d, e). As expected, quantification of plasmid-borne YFP–Crb2 revealed that both ccq1Δ and ccq1Δ taz1Δ cells had YFP–Crb2 foci in the majority of observed cells (Fig. 4f). Moreover, ccq1Δ telomere deprotection partially requires H4K20 methylation. Preventing H4K20 methylation at telomeres by removing Set9 from either ccq1Δ or ccq1Δ taz1Δ cells reduces checkpoint activation and YFP–Crb2 foci formation (Supplementary Fig. 5). Thus, absence of the Pot1–Ccq1 complex results in prompt Crb2 foci formation, Chk1–Myc phosphorylation and cell cycle arrest, indicating that Crb2 telomere exclusion constitutes a barrier imposed by normal telomeres to a full checkpoint response. However, loss of Ccq1 in wild-type cells results in a weaker and delayed checkpoint response than in taz1Δ mutants, establishing that, in addition to Pot1–Ccq1, the Taz1 complex participates in checkpoint inhibition at normal telomeres (Fig. 4d, e).
Our work provides new insight into how chromosome ends are protected from DNA damage checkpoints (Fig. 4g). Instead of preventing the detection of DNA damage, fission yeast telomeres constitute a chromatin-privileged region on the chromosome that blocks transduction of an active checkpoint signal. This is achieved by preventing stable association of Crb2 with telomeres, a checkpoint adaptor protein required for the delivery and activation of Chk1. Studies in budding yeast using a short telomere seed adjacent to a HO break postulated an anticheckpoint activity at telomeres24. Following studies showed that checkpoint repression in budding yeast is primarily regulated by the inhibition of Mec1ATR recruitment to telomeres25–27. Fission yeast telomeres initiate a checkpoint response that fails to phosphorylate Crb2 and Chk1. However, rather than preventing recruitment and activation of Rad3ATR, exclusion of Crb253BP1 from telomeres represents the critical step in the inhibition of the checkpoint pathway.
The inability of telomeres to stably recruit Crb253BP1 is probably due to the lack of H4K20me2 epigenetic marks at chromosome ends. Studies in both mammalian cells and in fission yeast demonstrate that H4K20me2 is the relevant histone modification for Crb2 binding and is required for the stable recruitment of Crb2 to sites of DNA damage3,4,19,28. Although H4K20me2 is the most abundant form of H4K20me18, how it is regulated or localized is not understood. Our finding that telomeres have low levels of H4K20me2 provides a rationale for its ubiquitous nature throughout the genome. Conceivably, H4K20me2 marks regions of the genome required for chromosome contiguity, and cell division should not be attempted unless the DNA damage response is disengaged. In contrast, the absence of H4K20me2 marks regions where DNA perceived as damaged, such as chromosome ends, would not interfere with genome stability thus precluding a full checkpoint response.
Normal telomeres undergoing DNA replication temporarily lose DNA protection. Studies in both yeast model systems and mammalian cells show that during S phase, concomitant with the replication fork passage, chromosome ends engage ATR and ATM15,16,29,30, which are redundantly required for telomerase recruitment6. However, the periodic activation of ATM and ATR at telomeres does not result in either Chk1 or Chk2 phosphorylation or in cell cycle delay. As dividing cells use cell-cycle-staged telomere exposure to process chromosome ends and engage telomerase, it is likely that these mechanisms have evolved to prevent cell cycle arrest in response to normal replicating telomeres.
All experiments were performed using the Schizosaccharomyces pombe strains listed in Supplementary Information. Checkpoint studies were performed using a combination of microscopy, ChIP and western blotting. Telomere length was analysed by Southern blotting using telomere probes. For details of the methods used, see Methods.
Full Methods and any associated references are available in the online version of the paper at www.nature.com/nature.
We thank J. Cooper, K. Tomita and the rest of the Telomere Biology Laboratory (Cancer Research UK, London) for support at the start of this project. We thank J. Cooper, T. Wolkow, P. Russell and A. M. Carr for strains and plasmids. We thank W. Kaufman for her initial effort in the establishment of HO strains. We are grateful to S. Grewal for sharing unpublished results and to D. Lydall for insights on the quantitative analysis of checkpoint activation. We thank K. Labib, L. Jansen, K. Xavier, R. Martinho and S. Lopes for critically reading the manuscript. T.C. and C.C.R. are supported by Fundação para a Ciência e a Tecnologia (FCT) postdoctoral fellowships. T.M.N. was supported by the Sidney Kimmel Scholar Program and his laboratory is supported by NIH grant GM078253. This work was supported by the FCT (PTDC/BIA-BCM/67261/2006) and the Association for International Cancer Research (06-396).
Author Contributions T.C. helped with the design and executed most experiments. L.K. performed ChIP experiments. C.C.R. performed western and Southern blotting experiments. V.B. performed live cell analysis. B.A.M. established the ChIP and HO assays. T.M.N. contributed to the design of the ChIP and HO assays. All authors contributed with strains and data analysis. M.G.F. conceived the study, performed live cell analysis and wrote the paper.
Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of this article at www.nature.com/nature.