Cdk1 restrains NHEJ through phosphorylation of XRCC4-like factor Xlf1.

Eukaryotic cells use two principal mechanisms for repairing DNA double-strand breaks (DSBs): homologous recombination (HR) and nonhomologous end-joining (NHEJ). DSB repair pathway choice is strongly regulated during the cell cycle. Cyclin-dependent kinase 1 (Cdk1) activates HR by phosphorylation of key recombination factors. However, a mechanism for regulating the NHEJ pathway has not been established. Here, we report that Xlf1, a fission yeast XLF ortholog, is a key regulator of NHEJ activity in the cell cycle. We show that Cdk1 phosphorylates residues in the C terminus of Xlf1 over the course of the cell cycle. Mutation of these residues leads to the loss of Cdk1 phosphorylation, resulting in elevated levels of NHEJ repair in vivo. Together, these data establish that Xlf1 phosphorylation by Cdc2(Cdk1) provides a molecular mechanism for downregulation of NHEJ in fission yeast and indicates that XLF is a key regulator of end-joining processes in eukaryotic organisms.

Introduction

The ability to repair DNA damage is critically important for the preservation of genomic integrity. DNA double-strand breaks (DSBs) can be repaired by two different cellular pathways: homologous recombination (HR) and nonhomologous end-joining (NHEJ) (Symington and Gautier, 2011). HR processes use undamaged homologous DNA sequences—typically from the sister chromatid—as a repair template, thus enabling error-free repair. NHEJ can also restore chromosome integrity by religation of DSB ends (Chiruvella et al., 2013) in the absence of homologous sequences but is potentially more error prone. While core factors such as Ku, XRCC4, XRCC4-like Factor (XLF), and DNA ligase 4 are required for all NHEJ repair reactions, accessory factors, including polymerases and nucleases, are also needed to process termini of imprecise DSBs into ligatable substrates.

The relative preference for break repair pathways differs between eukaryotes. Mammalian cells use NHEJ as the predominant DSB repair mechanism, where the pathway is available throughout the cell cycle. Yeast prefer to repair DSBs by HR (Manolis et al., 2001). Nevertheless, most eukaryotes utilize both NHEJ and HR; therefore, the choice of repair pathway is crucial for cell survival. DSB repair pathway selection is regulated in the cell cycle, with NHEJ predominating in G1 phase and HR restricted to the G2 and S phases of the cell cycle (Ferretti et al., 2013). Cyclin-dependent kinase 1 (Cdk1) plays a key role in regulating end resection during HR. Resection is strongly inhibited by low Cdk1 activity in G1 and can be reduced in G2 by Cdk1 inhibition (Aylon et al., 2004; Ira et al., 2004). In mammalian cells and budding yeast, the main target of CDK phosphorylation is CtIP/Sae2, which facilitates DSB end resection (Huertas et al., 2008). Cdk1 phosphorylation also influences later steps in HR, as well as expression levels of HR proteins.

In budding yeast, the initiation of DSB resection is normally suppressed in G1 due to low Cdk1 activity and depends on the MRX complex (Clerici et al., 2008). However, this dependence on Cdk1 activity can be overcome by deletion of Ku, suggesting that it induces indirect control over NHEJ by affecting HR instead. Several potential Cdk1 phosphorylation sites have been found in budding yeast Ku70/Ku80; however, their mutation did not affect NHEJ activity (Zhang et al., 2009). Thus, direct Cdk1 targets for NHEJ regulation have not yet been identified. In fission yeast, there is a reciprocal relationship between the deployment of the two major DSB pathways with NHEJ functioning during G1 and HR predominant in G2 cells (Ferreira and Cooper, 2004). It has been proposed that Cdk1 may influence this pathway selection, but a mechanism has not been identified.

Xlf1 is the fission yeast homolog of XLF/Cernunnos (Hentges et al., 2006; Cavero et al., 2007), a core NHEJ factor that binds to DNA and stimulates end-joining. In the present study, we identify Xlf1 as a key regulator of NHEJ activity in the cell cycle. We report that Cdk1 phosphorylates specific residues in the C terminus of Xlf1 over the course of the cell cycle. Using phospho-null and phosphomimic mutant strains, we demonstrate that Xlf1 phosphorylation inhibits the NHEJ pathway. We also identify effects on the checkpoint response and cellular events related to DSB resection. Together, these data establish that Xlf1 phosphorylation by Cdc2Cdk1 provides a molecular mechanism for the downregulation of NHEJ in fission yeast and offers insights into how this pathway may be regulated in other eukaryotic organisms.

Results and Discussion

Cdk1 Phosphorylates Xlf1 In Vitro

NHEJ is tightly regulated in fission yeast, but the mechanism is unknown (Ferreira and Cooper, 2004). To identify if posttranslational modifications regulate NHEJ in S. pombe, we analyzed the sequences of the core factors (Ku, Lig4, and Xlf1). The C-terminal region of Xlf1 has two sites, T180 and S192, that conform to [ST]-P-x-[KR], a consensus motif for phosphorylation by the Cdc2Cdk1 kinase. These potential phosphorylation sites are conserved in other fission yeasts (Figure 1A). To test if these sites serve as substrates for Cdc2Cdk1, we mutated them to alanine and performed kinase assays. Wild-type (WT) and mutated Xlf1 proteins were incubated with mammalian Cdk1 kinase (Figure 1B) or S. pombe Cdc2Cdk1 complex (Figure 1C). Similarly to histone H1, a known Cdc2Cdk1 substrate (Moreno et al., 1989), Xlf1 was phosphorylated, establishing it as an in vitro substrate for the mammalian and the fission yeast kinase. Phosphorylation of single point mutants (Xlf1.T180A or Xlf1.S192A) was markedly reduced compared to WT Xlf1, suggesting that the two sites can be phosphorylated. When both residues were mutated to alanine (Xlf1.T180A.S192A or Xlf1.AA), Xlf1 phosphorylation was abolished. Thus, Cdc2Cdk1 can phosphorylate Xlf1 on two conserved C-terminal residues in vitro.

Figure 1

Xlf1 Is Phosphorylated by Cdk1

(A) Alignment of the C-terminal portion of four fission yeast xlf1 homologs (S. pombe, S. octosporus, S. cryophilus, and S. japonicus), displaying two conserved cdc2cdk1 phosphorylation motifs (yellow boxes): [ST]Px[KR].

(B) In vitro Cdk1 kinase assay. Recombinant Xlf1 protein was incubated with mammalian Cdk1 in the presence of γ-32P-ATP and analyzed by autoradiography. Proteins used were the Cdk1 substrate histone H1 (positive control, 32 kDa), wild-type Xlf1 (27 kDa), single mutations of T180A and S192A, and double mutations T180A.S192A (AA). Recombinant Xlf1 is susceptible to cleavage between T180 and S192, and the cleavage product (indicated by an asterisk) is visible in the Coomassie-stained loading control.

(C) In vitro kinase assay using S. pombe Cdc2. Kinase assay with recombinant protein was conducted as in (B), except using Cdc2 complex purified from S. pombe cells.

(D) In vivo phosphorylation of Xlf1. GFP-tagged Xlf1 was immunoprecipitated from cell extracts of wild-type and xlf1.T180A.S192A (xlf1.AA), treated or mock treated with lambda phosphatase, and separated by SDS-PAGE in the presence of the phosphate-binding retardant Phos-tag. Phosphorylated Xlf1 is indicated by an arrow.

(E) GFP-tagged wild-type Xlf1 was immunoprecipated from cells containing the Shokat active site mutation cdc2.F84G (Dischinger et al., 2008) that had either been treated or mock treated with the inhibitor 1NM-PP1. Immunoprecipitates were treated or mock treated with lambda phosphatase and analyzed by SDS-PAGE and immunoblotting in either the presence or the absence of Phos-tag.

(F) A temperature-sensitive mutation was used to block cdc10-M17 nmt41-GFP.xlf1 cells in G1 phase and then released into the cell cycle. Phosphorylation of GFP-Xlf1 was analyzed by SDS-PAGE and immunoblotting of cell extracts in the presence or absence of Phos-tag. A nonspecific band detected by the GFP antibody is indicated by an asterisk.

Xlf1 Is Phosphorylated by Cdk1 in a Cell-Cycle-Dependent Manner

Next, we sought to investigate if phosphorylation of Xlf1 occurs in vivo. While phosphorylation had no apparent effect on the migration of Xlf1 in standard SDS-PAGE (Figures 1B and 1E), addition of a phosphate-binding agent, Phos-tag (Kinoshita et al., 2006), resolved phosphorylated Xlf1 as a distinct band. GFP-tagged Xlf1 was immunoprecipitated from asynchronous cultures and analyzed by Phos-tag western blotting. The slower migrating species was abolished by treatment with lambda phosphatase (Figure 1D), confirming that this represented phosphorylated cellular Xlf1. This species was absent in xlf1.AA, a strain that expresses Xlf1.AA, regardless of treatment with phosphatase. Together, these results demonstrate that Xlf1 is phosphorylated at T180 and S192 in vivo and indicate that further Cdc2Cdk1 kinase sites are unlikely.

To determine if Cdc2Cdk1 is responsible for Xlf1 phosphorylation in vivo, we used a cdc2-as mutant strain in which we could inhibit Cdc2Cdk1 activity (Dischinger et al., 2008). Treatment of cdc2as GFP-xlf1 cultures with the inhibitor (1NM-PP1) caused a significant reduction of the phosphoband of Xlf1 in Phos-tag western blots, which could be further reduced by phosphatase treatment (Figure 1E). This establishes that Xlf1 is phosphorylated by Cdc2Cdk1 in unperturbed asynchronous cultures.

Cdc2Cdk1 activity increases from a minimum in G1 to levels peaking in G2, triggering entry into mitosis. To study changes in Xlf1 phosphorylation status as cells progressed through the cell cycle, we used a temperature-sensitive cdc10 mutant (cdc10-M17) to synchronize GFP-xlf1 cultures in G1. A phospho-Xlf1 species was not detectable in G1-arrested cells (Figure 1F); however, upon release from the arrest, phosphorylated Xlf1 appeared and increased after 120 min, as cells entered G2 phase. These data indicate that Cdc2Cdk1 phosphorylates Xlf1 in a cell-cycle-dependent manner.

Xlf1 Phosphorylation by Cdc2Cdk1 Alters the Repair of DSBs by NHEJ

As NHEJ is most active in G1 but inhibited in S/G2 (Ferreira and Cooper, 2004), we predicted that Xlf1 phosphorylation inhibits NHEJ. To test this hypothesis, we examined the ability of xlf1 phosphorylation mutants to religate linearized plasmid DNA (Manolis et al., 2001). Leucine auxotrophic cells were transformed with linearized plasmid DNA containing the LEU2 marker. Their ability to religate ends and form colonies on selective plates lacking leucine was assessed by comparison to cells transformed with uncut plasmid. Log-phase cultures in which Cdc2Cdk1 phosphorylation of Xlf1 had been abolished (xlf1.AA′) displayed a ∼2.5-fold increase in end-joining compared to WT cells (Figure 2A). In contrast, the Xlf1 phosphomimetic mutant xlf1.EE showed a moderate decrease in plasmid end-joining. The religation levels were unaffected by the nature of the ends (blunt or overhangs) in xlf1 mutants. These results indicated that an inability to phosphorylate Xlf1 leads to increased NHEJ activity and supports the hypothesis that Cdc2Cdk1 phosphorylation of Xlf1 inhibits NHEJ.

Figure 2

Phospho-Null xlf1 Increases DNA End-Ligation Activity and Attenuates Damage Responses

(A) Religation of linearized plasmids containing blunt ends, 3′ and 5′ overhangs. Leu+ selection used to monitor religated plasmids. Data are reported as the mean ± 95% confidence interval.

(B) Religation of linearized plasmids with hygromycin selection within 2 hr of electroporation. Data are reported as the mean ± 95% confidence interval.

(C) Chk1-HA phosphorylation in response to ionizing radiation was monitored in wild-type and nmt41-xlf1.AA cells. Cells irradiated with 50 Gy, 100 Gy, and 200 Gy were analyzed 30 min after irradiation. To characterize the kinetics of the checkpoint response, cells irradiated with 50 Gy were also analyzed at 3.5 min, 7 min, and 10 min after irradiation. Phosphorylated Chk1 is indicated.

(D) The formation and persistence of Rad52-GFP foci following 50 Gy irradiation were monitored in wild-type and nmt41-xlf1.AA cells.

The resection of DSBs is thought to make them unsuitable for end-joining. Therefore, we asked if we could increase levels of end-joining of linearized plasmids by impairing the resection of DNA ends. We repeated the plasmid religation assays in a strain lacking the resection gene ctp1. Deleting ctp1 had no effect on plasmid religation levels (Figure 2B), suggesting that preventing Ctp1-dependent resection did not channel plasmid DSBs into NHEJ. However, when ctp1 was deleted in xlf1.AA cells, end-joining levels increased >4-fold compared to WT levels, significantly above an ≈2-fold increase caused by xlf1.AA in the presence of functional Ctp1. This finding suggested that, in log-phase cells, Ctp1 and phosphorylated Xlf1 synergistically counteract end-joining of linearized plasmid DNA.

Xlf1.AA Slows Down Cellular Responses to DSBs

Next, we characterized cellular response to ionizing radiation (IR)-induced DSBs in Xlf1 phospho-null strains. A key event in the cellular response to DNA damage is the activation of DNA damage checkpoints that arrest progress in the cell cycle to ensure that DSBs can be repaired. We monitored the phosphorylation status of Chk1 kinase, a marker of G2 checkpoint activation (Walworth and Bernards, 1996), in response to IR. IR-induced Chk1 phosphorylation was observed in nmt41-xlf1.AA chk1-HA cultures at all IR doses, with no discernible difference to WT chk1-HA cultures (Figure 2C). We also characterized the kinetics of the checkpoint response. In chk1-HA cells, Chk1 phosphorylation reached a maximum value within 10 min. In contrast, the phospho-Chk1 band did not appear until 10 min after irradiation, rising further at later time points. This observation suggests that nonphosphorylatable Xlf1 causes a deceleration of the checkpoint response, though full induction of the G2 checkpoint, as measured by Chk1 phosphorylation, is still achieved within 30 min.

Next, we studied the processing of IR-induced DSBs. Rad52 foci form after strand resection has begun and channeled into recombination processes (Symington and Gautier, 2011). In order to study the possible interference of phospho-null Xlf1 on such processes, we analyzed Rad52-GFP foci formed in live cells following IR. We irradiated rad52-GFP and nmt41-xlf1.AA rad52-GFP cells with 50 Gy and monitored the formation and persistence of Rad52-GFP foci using live-cell imaging. While Rad52 foci were observed in most cells with both WT xlf1 and nmt41-xlf1.AA, foci formation was slower in nmt41-xlf1.AA (Figure 2D). Rad52 foci formation peaked ∼30 min after IR in WT cells but 40–70 min after IR in nmt41-xlf1.AA, in which Rad52 foci also persisted for longer. This observation is consistent with HR processes, including DSB resection, being slowed in nmt41-xlf1.AA cells. We did not observe a general decrease of Rad52 foci in nmt41-xlf1.AA. As Rad52 foci are associated with resection, and as resected DSBs are not a suitable substrate for NHEJ, the observed deceleration of HR is unlikely to be the result of the overall balance of DSB repair pathways being tipped in favor of NHEJ.

Overexpression of Xlf1.AA Sensitizes Cells to DNA Damage

As the balance between HR and NHEJ is tightly regulated in the cell cycle, we next asked if deregulation of repair pathways resulted in altered sensitivity to DNA damage. To address this question, we characterized the IR sensitivity of Xlf1 phosphomutants in a stationary state, not requiring cell cycle regulation. Spores from homozygous crosses of xlf1.AA displayed an IR sensitivity similar to that of WT spores, in contrast to a small but reproducible increase in IR sensitivity with xlf1.EE spores (Figure S1). This suggests that NHEJ is active but can be attenuated, as observed in the phosphomimetic xlf1 mutant. Similarly, while deletion of ctp1 increased IR sensitivity of spores, there was a small but reproducible increase in radioresistance in xlf1.AA ctp1d, whereas the opposite effect was evident with xlf1.EE ctp1d spores (Figure S1). The IR sensitivity of ctp1d spores may be due to damage other than DSBs induced in the spore state, such as single-strand breaks and other lesions impeding the first round of replication following germination. Overexpression of WT xlf1 and xlf1.AA from nmt41 increased the IR resistance in spores above WT levels. This is again compatible with the notion that Xlf1 is a limiting factor regulating NHEJ levels.

In contrast to spores, mutation of the Xlf1 phosphorylation sites did not have a discernible effect on the sensitivity of vegetative cells to a range of DNA damage treatments (data not shown). However, overexpression using the medium-level nmt41 promoter of Xlf1 in vegetative cells rendered nmt41-xlf1.AA cells mildly sensitive to a variety of DNA-damaging agents, such as camptothecin, tert-butyl hydroperoxide, phleomycin, and methyl methanesulfonate in comparison to nmt1-xlf1 (WT) (Figure S1). The sensitivity was increased in nm41-xlf1.AA. No increase in xlf1.AA DNA damage sensitivity was observed in strains in which rad50 had been deleted (Figure S1), revealing an epistatic relationship with rad50. In addition, we noted that high-level overexpression from the nmt1 promoter of xlf1.AA, but not with WT xlf1, caused cells to become inviable (Figures S1 and S2).

NHEJ Is Hyperactivated in an Xlf1 Cdk1 Phosphorylation Null Mutant

As plasmid religation assays have relaxed requirements for NHEJ (Almeida and Godinho Ferreira, 2013), we sought another in vivo assay to study the role of Xlf1 phosphorylation by Cdc2Cdk1 on chromosomal DSBs. Chromosome end fusions can be generated by NHEJ in cells with unprotected telomeres, such as taz1d (Ferreira and Cooper, 2004). However, chromosome end fusions are only generated if taz1d cells are arrested in G1 but not during S/G2, because of the downregulation of NHEJ activity in S/G2 phases. If cell cycle inhibition of NHEJ activity restricts taz1d chromosome fusions to G1 phase, disinhibition of NHEJ should lead to chromosome fusions in S/G2-phase taz1d cells. Therefore, taz1d strains provide a suitable assay to study the potential inhibitory effect of Xlf1 phosphorylation on NHEJ in vivo. We first compared G1-arrested (nitrogen-starved) and S/G2 (log-phase) cultures in which telomeres were unprotected due to taz1 deletion. Chromosome fusions were detected in taz1d cells in G1 but not in S/G2 cells (Figure 3A). In contrast, no fusions appeared in G1-arrested taz1d xlf1d cells, because of downregulation of NHEJ, as expected. Chromosome fusions were also observed in taz1d xlf1.AA cells when arrested in G1 phase, showing that mutation of the phosphorylation sites preserves the ability to carry out NHEJ. It is important to note, however, that fusions were not detected in taz1d xlf1.AA during S/G2 phases. Thus, contrary to our prediction, disinhibition of NHEJ by preventing Xlf1 phosphorylation is not sufficient to fuse unprotected chromosome ends in log-phase cultures.

Figure 3

xlf1.AA Promotes NHEJ Fusion of Unprotected Telomeres

(A) Ligation of chromosome ends in xlf1 mutants in taz1d strains. Scheme of telomeric NotI restriction fragments. Chromosomes I and II each release two telomeric restriction fragments (C, I, L, and M). Chromosome III lacks NotI restriction sites; NotI digests of genomic DNA of the indicated strains were separated by PFGE, and chromosomal end-to-end fusions were detected by Southern blotting with a telomere probe (arrows indicate positions of resolved telomere fusions). Nitrogen-starved taz1d used as positive control for fusions.

(B) PFGE analysis reveals that Xlf1.AA promotes NHEJ-mediated telomeric fusions in cycling cells in taz1d ctp1d background.

(C) Xlf1 overexpression is sufficient to promote telomeric fusions in taz1d cycling cells, and xlf1.AA mutation increases the amount of these fusions. Ligation of chromosome ends in taz1d strains overexpressing xlf1 mutants from the nmt41 promoter was analyzed by PFGE. taz1+ control strains contain the same amount of DNA, though the signal from the telomeric probe is weaker as a result of telomere elongation in taz1d.

(D) Assaying for NHEJ and HR activities in parallel, using hygromycin resistance and leu1 integration of linearized plasmid fragments. Data are reported as the mean ± 95% confidence interval.

See also Figure S3C.

NHEJ and HR are regulated independently in the fission yeast cell cycle, as inactivation of HR does not lead to increased use of NHEJ in G2 cells (Ferreira and Cooper, 2004). Conversely, we expected that abnormal activation of NHEJ in G2 cells will take place in the presence of activated HR, since the disinhibition of NHEJ alone in the xlf1.AA mutant would not affect HR activity. Therefore, we reasoned that competition for DSBs between NHEJ and HR is likely taking place in taz1d xlf1.AA cells, potentially masking the disinhibition of NHEJ, which may only become detectable once HR is inactivated. To test this hypothesis, we repeated the experiment in a background in which ctp1 was deleted. Ctp1 is essential for the initiation of resection, targeting DSBs toward HR (Limbo et al., 2007). When the chromosome fusion assays were repeated, no chromosome fusions were detected in ctp1d taz1d xlf1 during S/G2 phases (Figure 3B), establishing that inactivation of Ctp1-dependent resection alone is not sufficient to cause fusion of unprotected chromosome ends. Strikingly, inactivation of HR in ctp1d taz1d xlf1.AA cells led to substantial telomere fusions. We analyzed seven independent clones (Figure S3A) and found that all contained intra- and/or interchromosomal end fusions, including clones with circular chromosomes (both termini of a chromosome are fused) during S/G2 phases, a striking effect of NHEJ activity as confirmed by its absence in the corresponding lig4d strain (Figure S3B). This indicates that xlf1.AA prevents the inhibition of NHEJ in G2 cells when HR is inactivated. In contrast, no chromosome fusions were detected in the corresponding phosphomimicking xlf1 mutant, ctp1d taz1d xlf1.EE. These observations support the hypothesis that Cdc2Cdk1 phosphorylation of Xlf1 switches off NHEJ during the cell cycle.

To further explore the nature of competition between HR and NHEJ when both DSB repair pathways are active in the xlf1.AA phospho-null mutant, we speculated that overexpression of xlf1 may overcome the requirement to inactivate HR. We constructed taz1d strains in which xlf1 is controlled by the medium-strength nmt41 promoter integrated at the xlf1 locus. Chromosome fusions could be detected in log-phase nmt41-xlf1.AA taz1d cells, even though HR was functional (Figure 3C). Overexpression of WT xlf1 similarly caused telomere fusions in the presence of HR function, though to a lesser extent than in the phospho-null mutant. This suggests that xlf1 overexpression contributes to overcoming recombinogenic mechanisms that prevent chromosome fusions, tipping the HR/NHEJ balance in favor of end-joining, either by promoting NHEJ before DSB becomes subject to HR or by separately inhibiting HR. These observations imply that xlf1 is an important regulator in the balance of HR versus NHEJ during the cell cycle.

Xlf1.AA Affects Both HR and NHEJ of Linearized Plasmid DNA

The observation that chromosome end ligations (taz1d cells) in the xlf1.AA phosphomutant require inactivation of HR indicates crosstalk between NHEJ and HR. To verify if a similar effect could be identified in the processing of plasmid DNA ends by the two DSB repair pathways, we designed an assay in which the levels of HR and NHEJ can be assessed in parallel. Cells were electroporated with two different linearized plasmids. One plasmid, containing WT leu1 but no ARS, could be integrated at the leu1-32 locus via an HR-dependent mechanism, giving rise to Leu+ colonies. The second plasmid, containing the antibiotic resistance gene hph and an S. pombe ARS, could be recircularized by NHEJ and stably maintained, giving rise to hygromycin-resistant colonies. Hygromycin selection was applied after 2 hr. Cells were transformed in parallel; the frequency of leu1+ colonies and hygromycin-resistant colonies were determined in relation to an uncut plasmid control as a measure of HR and NHEJ activity, respectively. Deletion of xlf1 decreased plasmid religation ∼50 fold (Figure 3D), whereas deletion of ctp1 decreased plasmid integration ∼50 fold, indicative of inactivation of NHEJ and HR, respectively. Deletion of xlf1 or lig4 also caused a small but reproducible decrease in HR-dependent plasmid integration. Plasmid religation was increased in xlf1.AA mutant cells, though less than in the assay used in Figure 3D, presumably because of the much-reduced time available for religation with antibiotic rather than auxotrophic selection. Unexpectedly, plasmid integration was reduced by two-thirds in xlf1.AA (endogenous promoter), suggesting that disabling the phosphorylation of Xlf1 leads to inhibition of HR. In addition, while deletion of ctp1 by itself had no significant impact on NHEJ, deletion of ctp1 in an xlf1.AA mutant led to a much larger increase in plasmid religation than in HR-competent xlf1.AA. Together, these observations provide further evidence that phospho-null xlf1.AA affects the levels of HR.

Cell cycle regulation of DSB repair pathway selection by Cdk1 was first established with the discovery that DSB resection requires Cdk1 activity (Aylon et al., 2004; Ira et al., 2004). However, while Cdk1 has been shown to control the function of several HR factors, a reciprocal regulation of NHEJ by Cdk1 has not been reported. While budding yeast Ku70 and Ku80 contain several potential Cdk1 phosphorylation sites, their mutation does not affect NHEJ activity (Zhang et al., 2009). Moreover, the dependence of resection for Cdk1 activity can be overcome by the deletion of Ku, suggesting an indirect control over NHEJ by modulation of HR (Clerici et al., 2008). Nej1, the budding yeast XLF homolog, was discovered as a factor downregulating NHEJ in diploid cells (Frank-Vaillant and Marcand, 2001; Kegel et al., 2001; Valencia et al., 2001). Lif1, the S. cerevisiae XRCC4 homolog, has been found to be phosphorylated by Cdk1, but this phosphorylation has little effect on the levels of classical NHEJ and, instead, affects a Sae2ctp1-dependent resection-mediated imprecise joining pathway (Matsuzaki et al., 2012). Our study shows that Cdk1 phosphorylation of a core NHEJ factor directly regulates classical NHEJ during the cell cycle. We show that Xlf1 becomes phosphorylated by Cdc2Cdk1 on T180 and S192 and that the levels of phosphorylation increase through the cell cycle as Cdc2Cdk1 activity increases. These results establish that phosphorylation of Xlf1 has an inhibitory effect, as there is a reduction in the levels of religation of linearized DNA with the phosphomimic xlf1.EE but an increase with the phospho-null xlf1.AA mutant. Together, these observations allow us to propose a model in which Xlf1 functions as a cellular switch for NHEJ (Figure 4), with nonphosphorylated Xlf1 representing the on state (NHEJ active) and phosphorylated Xlf1 representing the off state (NHEJ inactive). The NHEJ repair pathway is fully active in G1, but as cells advance in the cell cycle, Cdc2Cdk1 levels rise and Xlf1 becomes increasingly phosphorylated, leading to inactivation of end-joining. Although this report establishes that Xlf1 is an important NHEJ regulator, further studies are required to establish how Cdk1 phosphorylation of this factor restrains NHEJ in cycling cells and whether this key regulatory function is also conserved in other eukaryotic organisms.

Figure 4

Model of Cdk1 Regulation of the Balance between NHEJ and HR

The activity of CDK1 is a major determinant of DSB repair pathway choice. When CDK1 is low, nonphosphorylated Xlf1 exists in the cell and NHEJ is fully active. As CDK1 activity rises, phosphorylation of Xlf1 increases, leading to the inactivation of NHEJ. The phosphorylation of CDK1 targets in the HR pathway leads to the activation of HR.

Experimental Procedures

Standard methods used for strain construction, western blotting, and microscopy are detailed in the Supplemental Experimental Procedures. Strains used are listed in Table S1.

Phosphatase Treatment of nmt41-GFP-Tagged Xlf1

A total of 109 log-phase cells grown without thiamine was resuspended in lysis buffer (50 mM Na phosphate [pH 7], 150 mM NaCl, 50 mM NaF, 10 mM EDTA, 10% glycerol, 0.5% NP40, Roche protease inhibitor) and broken with glass beads. Cleared lysate containing 15 μg total protein was incubated for 2 hr at 4°C with 15 μl GFP-Trap A magnetic beads (Chromotek) per immunoprecipitation. Beads were washed with lysis buffer, washed three times in PMP buffer (New England Biolabs; NEB) plus MnCl2, resuspended in 100 μl, divided into two, and either mock treated or incubated with 800 units of lambda phosphatase (NEB) at 30°C for 30 min. Beads were resuspended in Laemmli buffer and boiled. Samples were separated by SDS-PAGE on 12% gels containing 25 μM Phos-tag (AAL-107 Wako) and 50 μM MnCl2. Prior to transfer onto polyvinylidene fluoride, gels were incubated for 10 min in transfer buffer with 1 mM EDTA and then without EDTA. GFP-Xlf1 was detected using anti-GFP antibody (Invitrogen, 1:2,500 dilution).

In Vitro Cdc2Cdk1 Kinase Assay

Active Cdc2 enzyme was isolated from a lysate of 5 × 108 WT S. pombe cells resuspended in 400 μl of HB buffer (25 mM Tris [pH 7.5], 15 mM EGTA, 15 mM MgCl2, 0.1% NP40, protease inhibitor cocktail) broken using glass beads. Per kinase reaction, cell lysate containing 1 mg total cellular protein was mixed with 40 μl of p13suc1 agarose conjugate (Millipore), incubated with rotation at 4°C for 3 hr, washed three times with HB buffer, and then washed once with kinase buffer (10 mM HEPES [pH 7.5], 75 mM KCl, 5 mM MgCl2, 1 mM dithiothreitol). For each kinase assay reaction, beads were then mixed in a total volume of 15 μl containing 1.5 μg purified protein (either recombinant 6His-Xlf1 or histone H1), 20 μM ATP, and 5μCi γ-32P-ATP, all diluted in kinase buffer. The same method was used with Cdk1 (NEB). The kinase reaction was allowed to proceed for 10 min at room temperature and then was stopped by adding SDS-PAGE loading buffer and heated to 90°C for 5 min. Protein was separated on 15% gels and subjected to Coomassie staining and autoradiography.

Author Contributions

P.H. and A.J.D. designed the key experiments and wrote the manuscript with advice from M.G.F. and C.C.R. H.W. designed and carried out hygromycin plasmid assays and spore survivals and made chk1 and rad52 strains. Experimental work using Taz1d was designed by M.G.F. and C.C.R. C.C.R carried out and made strains needed for these taz1 experiments. P.H. carried out all other experimental work and strain production.