Mitosis Inhibits DNA Double-Strand Break Repair to Guard Against Telomere Fusions

See allHide authors and affiliations

Science  11 Apr 2014:
Vol. 344, Issue 6180, pp. 189-193
DOI: 10.1126/science.1248024

Shutting Down Repair to Protect

Cells repair DNA double-strand breaks (DSBs) by halting the cell cycle and activating the machinery involved in mending the breaks. However, during mitosis neither the DNA damage checkpoint nor DSB repair occur, apparently leaving the cell extremely vulnerable to DSBs. Orthwein et al. (p. 189, published online 20 March) found that the DSB response was blocked by the phosphorylation of two crucial repair factors, RNF8 and PB531, preventing their recruitment to the site of damage. Restoring DSB repair during mitosis caused end-to-end chromosome fusions, which are catastrophic for chromosome segregation and normal cell division, explaining why the repair machinery is shut down during cell division.


Mitotic cells inactivate DNA double-strand break (DSB) repair, but the rationale behind this suppression remains unknown. Here, we unravel how mitosis blocks DSB repair and determine the consequences of repair reactivation. Mitotic kinases phosphorylate the E3 ubiquitin ligase RNF8 and the nonhomologous end joining factor 53BP1 to inhibit their recruitment to DSB-flanking chromatin. Restoration of RNF8 and 53BP1 accumulation at mitotic DSB sites activates DNA repair but is, paradoxically, deleterious. Aberrantly controlled mitotic DSB repair leads to Aurora B kinase–dependent sister telomere fusions that produce dicentric chromosomes and aneuploidy, especially in the presence of exogenous genotoxic stress. We conclude that the capacity of mitotic DSB repair to destabilize the genome explains the necessity for its suppression during mitosis, principally due to the fusogenic potential of mitotic telomeres.

We hypothesized that the inactivation of mitotic double-strand break (DSB) repair (13) might be caused by the failure to recruit the DSB repair factors BRCA1 and 53BP1 to DNA damage sites during mitosis (47). 53BP1 and BRCA1 promote DSB repair by nonhomologous end joining (NHEJ) and homologous recombination, respectively (8). Both proteins accumulate at DSB sites downstream of a common pathway consisting of the ataxia telangiectasia mutated–dependent phosphorylation of H2AX (forming γ-H2AX) followed by MDC1, RNF8, and RNF168 recruitment (8). RNF168 ubiquitylates H2A (9, 10), which triggers the recruitment of 53BP1 (11) and also BRCA1 (12). Mitosis severs this pathway upstream of RNF8 recruitment to DSB sites as the formation of γ-H2AX and MDC1 ionizing radiation (IR)–induced foci, which denote accumulation at break sites, are unaffected by mitotic entry (4).

To elucidate how mitosis blocks RNF8 recruitment to DSB sites, we tested whether the RNF8-MDC1 interaction (1315) is disabled in mitosis (see supplementary materials and methods). We observed that whereas RNF8 and MDC1 interact in asynchronously dividing cells after irradiation, their interaction is suppressed during M (mitotic) phase (Fig. 1A and fig. S1A; details of all synchronization and treatments are depicted in fig. S2). Because RNF8 recognizes redundant, phosphorylated epitopes on MDC1, we assessed whether mitosis inhibits the ability of RNF8 to recognize phospho-MDC1. MDC1-derived phosphopeptides that encompass its Thr752 (T752) phosphorylation site (pT752) are unable to retrieve RNF8 from mitotic extracts, whereas they readily retrieve RNF8 from extracts of asynchronously dividing cells (Fig. 1B and fig. S1B). This inhibition is due to cyclin-dependent kinase 1 (CDK1)–dependent phosphorylation, because pretreating mitotic extracts with PP1, a Ser/Thr phosphatase, or treating cells with the CDK1 inhibitor RO-3306 before harvesting restored the RNF8-pT752 interaction (Fig. 1B).

Fig. 1 Mitotic cells inactivate the RNF8-MDC1 interaction through phosphorylation.

(A) Immunoprecipitation (IP) of MDC1 or RNF8 from asynchronous (ASN) and mitotic U2OS cell extracts. IgG, immunoglobulin G. (B) Pull-down (PD) assay of RNF8 from ASN or mitotic extracts with unphosphorylated (T) or phosphorylated (pT) peptides encompassing the MDC1 T752 residue. INH, phosphatase inhibitor cocktail; CDKi, CDK inhibitor. (C) (Top) RNF8 domain organization. FHA, forkhead-associated domain; N, N terminus; C, C terminus. (Bottom) Whole-cell extracts of U2OS cells released from a double-thymidine block were probed using an antibody against RNF8 pT198. siRNF8, small interfering RNA against RNF8. (D) RNF8-depleted U2OS cells stably expressing green fluorescent protein (GFP)–RNF8 or its mutants were synchronized by double-thymidine block before release and irradiation (10 Gy). Interphase (Interph.) and mitotic (M) cells were processed for GFP, γ-H2AX, and BRCA1 or 53BP1 immunofluorescence [mean ± SD (error bars); N = 3 biological replicates]. (E) Representative micrographs of (D). Scale bars, 5 μm. DAPI, 4′,6-diamidino-2-phenylindole.

To test whether CDK1 could directly inhibit the RNF8-MDC1 interaction, recombinant human and mouse RNF8 proteins fused to glutathione S-transferase (GST) were subjected to phosphorylation by CDK1-cyclin B or CDK2-cyclin A before pull-down assays with pT752. CDK1, but not CDK2, could phosphorylate RNF8, which suppressed the ability of RNF8 to bind to pT752 (fig. S1, C to E). The reconstitution of the CDK1-dependent inhibition of the RNF8-pT752 interaction identified T198 as the main CDK1 site on RNF8 (fig. S1F). An antibody against the phosphorylated T198 residue (RNF8 pT198) confirmed a mitosis-specific phosphorylation of T198 (Fig. 1C and fig. S1G). Mutation of T198 to alanine (yielding RNF8 T198A) rendered the RNF8-pT752 interaction insensitive to CDK1 in pull-down assays (fig. S1E). In contrast, the T198E (E, Glu) mutation, which mimics T198 phosphorylation, constitutively inhibited RNF8 binding to phosphopeptides (fig. S1E).

We tested whether the RNF8-T198A mutation restored RNF8 recruitment to DSB sites in mitotic U2OS cells. Reintroduction of wild-type (WT) RNF8 in RNF8-depleted cells (fig. S3) enabled the formation of IR-induced foci in interphase but not in mitosis (Fig. 1, D and E, and fig. S4). In contrast, RNF8-T198A accumulated at DSB sites in both cell cycle phases (Fig. 1, D and E, and fig. S4). Conversely, the RNF8-T198E mutation suppressed IR-induced focus formation in interphase (Fig. 1D and fig. S4). RNF8-T198 phosphorylation is therefore the main mechanism by which mitosis suppresses RNF8 recruitment to DSB sites.

In the same experiments, we tested whether the T198A mutation restored BRCA1 and 53BP1 accumulation at mitotic DSB sites. Reintroducing RNF8-T198A restored BRCA1 recruitment to DSB sites but failed to restore 53BP1 IR-induced focus formation during M phase (Fig. 1, D and E). These results indicate that in addition to RNF8 phosphorylation, mitotic cells inactivate the recruitment of 53BP1 via a second inhibitory mechanism.

To identify this additional regulatory step, we mapped mitotic phosphorylation sites on 53BP1 by mass spectrometry. This approach identified a number of sites and, among them, T1609 and S1618 (S, Ser) that map to the 53BP1 ubiquitination-dependent recruitment (UDR) motif (Fig. 2A and fig. S5). The UDR allows 53BP1 to bind to ubiquitylated H2A and is necessary for its accumulation at DSB sites (11). Immunoblotting of extracts derived from cells released from a double-thymidine block confirmed that both sites are simultaneously phosphorylated in mitosis (Fig. 2A and fig. S6A). T1609 is likely a target of a proline-directed kinase such as CDK1 or p38, whereas S1618 is a PLK1 target (16). Using a recombinant 53BP1 protein fragment comprising the Tudor and UDR motifs (11), we confirmed that CDK1 could phosphorylate T1609 and that PLK1 targets S1618 (fig. S6, B and C). No substantial phosphorylation of the 53BP1 Tudor-UDR fragment was observed in the T1608/S1619A (TASA) double mutant (fig. S6, B and C). These results indicate that the 53BP1 UDR is phosphorylated in mitosis by proline-directed kinases, such as CDK1, and PLK1.

Fig. 2 Mitotic cells inactivate 53BP1 by phosphorylating the UDR.

(A) (Top) 53BP1 domain organization. BRCTs, BRCA1 C-terminal domains. (Bottom) Extracts from U2OS cells released from a G1/S block were probed by immunoblotting for S1618 and simultaneous T1609/S1618 phosphorylation. (B) Pull-down assay of RNF168-ubiquitylated NCPs with the indicated GST-Tudor-UDR constructs. Where indicated, the GST-Tudor-UDR was preincubated with CDK1 or PLK1 before the binding assays. (C) Quantitation of GFP-53BP1 focus-positive cells expressing the indicated Flag-RNF8 and GFP-53BP1 constructs in early or late mitosis [mean ± SD (error bars); N = 3]. (D) Representative micrographs of (C). Scale bar, 5 μm.

To test the importance of the mitotic phosphorylation of the UDR, we assessed the ability of the 53BP1 Tudor-UDR to interact with nucleosome core particles (NCPs) containing a dimethylated H4 Lys20 (K20) mimic (H4Kc20me2) and H2A K15-ubiquitin, as previously reported (11). The 53BP1 Tudor-UDR protein binds specifically to such dually modified NCPs (Fig. 2B and fig. S7). However, upon CDK1 and PLK1 phosphorylation, NCP binding was abrogated (Fig. 2B and fig. S7). The inhibition required both kinases (fig. S7) and the T1609/S1618 residues (Fig. 2B). Conversely, phosphomimetic mutations that convert T1609 and S1608 to glutamic acid led to the inhibition of 53BP1 binding to RNF168-ubiquitylated NCPs that was strongest with the T1609E/S168E (TESE) double mutant (fig. S7).

The T1609A, T1609E, S1618A, and TASA mutations have little to no impact on 53BP1 focus formation in interphase cells (fig. S8, A and B). However, the TESE mutation profoundly affects 53BP1 recruitment to DSB sites (fig. S8, A and B) and the ability of 53BP1 to restore IR resistance in 53bp1–/– DT40 avian cells (fig. S8, C and D). Together, these results indicate that T1609/S1618 phosphorylation inhibits 53BP1 accumulation and function at DSB sites. When mitotic cells expressing the combination of RNF8-T198A and 53BP1-TASA mutants were irradiated, 53BP1 recruitment to DSB sites was restored nearly to the level observed for the WT protein in interphase cells (Fig. 2, C and D). Under these conditions, RIF1, a 53BP1 effector during NHEJ (1721), was also recruited to DSB sites (Fig. 2D). We therefore conclude that the block to 53BP1 recruitment in mitosis is the result of RNF8 phosphorylation on T198 and 53BP1 phosphorylation on T1609/S1618.

We next assessed whether the combination of RNF8-T198A and 53BP1-TASA mutants restored DSB repair in mitosis, using the neutral comet assay. Asynchronously dividing cells expressing RNF8-T198A and 53BP1-TASA displayed DSB repair kinetics that were comparable to WT cells (Fig. 3A and fig. S9A). As expected (1, 2), M-phase cells expressing WT proteins did not repair their chromosomal breaks over the duration of the experiment (Fig. 3A). In contrast, the RNF8-T198A/53BP1-TASA–expressing mitotic cells clearly repaired DSBs over time (Fig. 3A). In two orthogonal assays, γ-H2AX focus formation/dissolution and TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling), we confirmed that expression of RNF8-T198A and 53BP1-TASA reactivated mitotic DSB repair (fig. S9, B to D). These results demonstrate that RNF8 pathway restoration reactivates DSB repair in mitotic cells.

Fig. 3 Reactivation of mitotic DSB repair is deleterious.

(A) Neutral comet analysis of asynchronously dividing or mitotic U2OS cells stably expressing the indicated Flag-RNF8 and GFP-53BP1 proteins before and after irradiation with a 10-Gy dose (N = 30). Statistical significance was determined by one-way analysis of variance (ANOVA). ns, not significant; t, time. (B) Clonogenic survival assays of the cells described in (A) in the presence or absence of NU7441 [mean ± SD (error bars); N ≥ 3]. (C) Formation of MNi in U2OS cells stably expressing the indicated proteins after irradiation (0.5 Gy) and released from the mitotic block. Scale bars, 5 μm. (D) Quantitation of MNi formation in mitotic U2OS cells stably expressing RNF8-T198A/53BP1-TASA and treated as indicated. (Left) Mitotic cells were obtained after nocodazole treatment. (Right) Mitotic cells were obtained after release from a double-thymidine block [mean ± SD (error bars); N = 3].

Cells in M phase are hypersensitive to IR, presumably because they inactivate DSB repair (4). To test whether mitotic DSB repair affects IR sensitivity, we carried out clonogenic survival assays with low doses of radiation [0.25 to 2 grays (Gy)]. As expected, mitotic cells were more IR-sensitive than their asynchronously dividing counterparts (Fig. 3B). Asynchronously dividing cells expressing RNF8-T198A/53BP1-TASA had a clonogenic survival profile identical to that of WT RNF8/53BP1-expressing cells. In stark contrast, RNF8-T198A/53BP1-TASA–expressing mitotic cells displayed marked IR hypersensitivity compared with mitotic cells expressing WT proteins (Fig. 3B). To determine whether mitotic DSB repair was responsible for the increased IR-sensitivity of the RNF8-T198A/53BP1-TASA cells, we treated cells with the DNA-PKcs inhibitor NU7441 to inhibit NHEJ, the DSB repair pathway mediated by 53BP1. NU7441 rescued the IR hypersensitivity of RNF8-T198A/53BP1-TASA–expressing mitotic cells to the levels of those expressing their WT counterparts (Fig. 3B). These results indicate that mitotic DSB repair leads to IR hypersensitivity and is therefore deleterious.

Repair of DSBs during cell division might be harmful due to defective chromosome segregation. To test this hypothesis, we monitored the formation of micronuclei (MNi), as they are caused by the mis-segregation of chromosomes or their fragments (22). We subjected mitotic cells expressing various combinations of RNF8 (wild type or T198A) and 53BP1 (wild type or TASA) proteins to a 0.5-Gy x-ray dose. Cells were released from the mitotic block and analyzed 6 hours later. In cells expressing RNF8-T198A/53BP1-TASA, we observed a selective and marked increase in CREST (i.e., kinetochore)–positive MNi, indicative of whole-chromosome mis-segregation, without a concomitant increase in kinetochore-negative MNi (Fig. 3, C and D). This increase was independent of the synchronization procedure but stimulated by irradiation and RNF168, which recruits 53BP1 to DSB sites (Fig. 3D and fig. S10). Furthermore, the combined depletion of RIF1 and PTIP, the 53BP1 effectors in DSB repair (1721, 23), along with the inhibition of DNA-PKcs, suppressed the formation of kinetochore-positive MNi (Fig. 3D). These results indicate that the reactivation of mitotic DSB repair impairs chromosome segregation, causing aneuploidy.

Together, PTIP and RIF1 promote telomere fusions (23), a chromosome rearrangement that generates dicentric chromosomes. To assess whether mitotic DSB repair produces telomere fusions, we examined metaphase telomeres by fluorescence in situ hybridization (FISH) in telomerase-negative (IMR90 E6/E7) and -positive (RPE1-hTERT) cells. We observed an increase that was greatly stimulated after IR treatment in telomere fusions in untreated RNF8-T198A/53BP1-TASA mitotic cells, but not in interphase cells (Fig. 4, A and B, and fig. S11, A to C). The presence of telomere fusions was confirmed by Southern blotting and BAL-31 nucleolytic digestion (fig. S11, D and E). RNF8-T198A/53BP1-TASA also produced high levels of NHEJ-dependent dicentrics, identified as anaphase bridges containing two kinetochores (fig. S12, A and B). Together, these results suggest that the aneuploidy caused by mitotic DSB repair is the consequence of sister telomere fusions.

Fig. 4 Mitotic DSB repair causes telomere fusions.

(A) RNF8- and 53BP1-depleted IMR90 (E6/E7) or RPE1-hTERT metaphase cells expressing the indicated constructs were exposed to IR (0.5 Gy) and collected 3 hours later. Mitotic spreads were subjected to telomere-FISH and counterstained with DAPI. Where indicated, the irradiation was carried out with cells blocked in G1, and metaphases were collected after release. Hesperadin is an Aurora B kinase inhibitor. Each circle represents a counted cell; blue bars indicate the mean. Statistical significance was determined by one-way ANOVA. (B) Representative micrographs of (A). Arrowheads indicate sister chromatid fusions. Scale bar, 5 μm. (C) Mitotic or interphase IMR90 (E6/E7) cells expressing the indicated constructs were mock-treated or irradiated (0.5 Gy) in the presence of Hesperadin, as indicated. Three hours after irradiation, cells were processed for γ-H2AX immunofluorescence and telomere-FISH. Each circle represents a counted cell; blue bars indicate the mean.

Because telomeres normally inhibit DSB repair (24), our results inferred that mitotic telomeres might be prone to deprotection and that genotoxic stress stimulates telomere uncapping. The irradiation of mitotic, but not interphase, cells induced telomere uncapping, as measured by γ-H2AX colocalization with telomeres, independently of the RNF8 and 53BP1 status (Fig. 4C and fig. S13). This phenomenon was reminiscent of the recently described Aurora B kinase–dependent telomere uncapping seen during prolonged mitoses (25). In support of the possibility that mitotic and genotoxic stress causes mitotic telomere uncapping via a common mechanism, we found that telomere deprotection (Fig. 4C and fig. S13) and fusions (Fig. 4A) were also dependent on Aurora B activity, as they were suppressed by hesperadin treatment.

We conclude that cells must suppress DSB repair in mitosis because M-phase telomeres are prone to fusions. We propose that mitotic telomeres exist in an underprotected state due to Aurora B–dependent repression of Shelterin activity. This condition can transition to a fully deprotected state, either as a low-frequency spontaneous event or at high frequency after mitotic or genotoxic stress. Because mitotic inhibition of DSB repair has been observed in a number of metazoan species, it appears that cells have not been able to evolve a better solution to the problem caused by underprotected mitotic telomeres other than transiently inactivating this major genome-maintenance pathway.

Supplementary Materials

Materials and Methods

Figs. S1 to S13

References (2629)

References and Notes

  1. Acknowledgments: All primary data are archived at the Lunenfeld-Tanenbaum Research Institute. We thank R. Szilard and Durocher lab members for critically reading the manuscript; A. Nussenzweig, J. Karlseder, D. Xu, J. Lukas, A.-C. Gingras, W. Dunham, M. Cook, A. Rosebrock, L. Harrington, F. Rossiello, F. d’Adda di Fagagna, S. Lawo, and L. Pelletier for reagents or technical advice; and D. Chowdhury for the pT1609/pS1618 antibody and for sharing unpublished data. A.O. and A.F.-T. received postdoctoral fellowships from the Canadian Institutes of Health Research (CIHR); S.M.N. is a scholar of the EIRR21st training program; C.M.B. received a fellowship from the Swiss National Science Foundation; J.S. received a doctoral award from the CIHR; and D.D. is the Thomas Kierans Chair in Mechanisms of Cancer Development and a Canada Research Chair (Tier 1). The work was supported by CIHR grant MOP89754 and by the Krembil Foundation.
View Abstract

Navigate This Article