Schizosaccharomyces pombe Cds1^Chk2 regulates homologous recombination at stalled replication forks through the phosphorylation of recombination protein Rad60

Eukaryotic cells need to precisely duplicate their genomes during S phase in every cell cycle. However, exogenous and endogenous sources of DNA damage can block the progression of DNA replication forks, potentially resulting in a range of replication-associated DNA structures including single-strand lesions and double-strand breaks (DSBs) (Lambert and Carr, 2005; Lambert et al., 2007). To cope with replication fork blocks, eukaryotic cells employ a DNA-structure-dependent checkpoint pathway (Carr, 2002; Elledge, 1996). In the fission yeast Schizosaccharomyces pombe, Rad3 (a homolog of human ATR) plays a major role in the response to replication fork stalling. Hydroxyurea (HU), which causes deoxyribonucleoside triphosphate starvation by inactivating ribonucleotide reductase, is most widely used to investigate the cellular response to replication fork stalling. In the presence of HU, Rad3^ATR is activated and phosphorylates multiple target proteins. This results in activation of checkpoint kinase Cds1, the S. pombe homolog of Chk2. Once activated, Cds1^Chk2 phosphorylates further downstream targets to regulate cell-cycle progression and DNA repair mechanisms (Furuya and Carr, 2003; Kai and Wang, 2003). S. pombe rad60 was originally identified by screening for genes that are required for homologous recombination (Morishita et al., 2002). The rad60⁺ gene is essential for growth and encodes a protein that belongs to the RENi (Rad60, Esc2p, and Nip45) family (Novatchkova et al., 2005). The Rad60 protein is involved in DNA repair through the homologous recombination pathway. Genetic and biochemical analysis demonstrates that it functions in concert with the Smc5/6 complex (Miyabe et al., 2006; Morishita et al., 2002).

Smc5/6 is one of the three structural maintenance of chromosome (SMC) complexes. Cohesin, composed of Smc1 and Smc3, maintains the link between two sister chromatids until cells undergo mitosis, whereas condensin, composed of Smc2 and Smc4, is required for condensation of chromosomal DNA during mitosis (Hirano, 2005). S. pombe smc6 was first identified as a gene that complemented a DNA-damage-sensitive mutant (Fousteri and Lehmann, 2000; Lehmann et al., 1995). Subsequent analysis revealed that the Smc5/6 complex is required for DNA repair by homologous recombination and has an additional essential function (Murray and Carr, 2008). Recent studies suggest that Smc5/6 has multiple functions in homologous recombination (Ampatzidou et al., 2006; Irmisch et al., 2009; Miyabe et al., 2006; Murray and Carr, 2008). rad60-1, a temperature-sensitive hypomorph of rad60, shows mutual genetic interactions with smc6, and the Rad60 protein was shown to physically interact with Smc5/6 complex (Boddy et al., 2003; Morishita et al., 2002). These observations suggested that rad60 not only shares functions with smc6 in homologous recombination, but is also required for the essential function of Smc5/6, which is less well characterized.

Rad60 interacts with the forkhead-associated (FHA) domain of Cds1^Chk2 and is phosphorylated in a Cds1-dependent manner (Boddy et al., 2003; Raffa et al., 2006). Rad60 protein disperses throughout the cell when cells are challenged with HU, whereas it normally localizes within the nucleus of unperturbed cells throughout the cell cycle. In cds1-fha1 mutant cells, Rad60 remains in the nucleus even when cells are challenged with HU. Immunoprecipitated Cds1^Chk2 phosphorylates the N-terminus of Rad60 in vitro, indicating that Cds1^Chk2 directly phosphorylates Rad60 to regulate its localization (Boddy et al., 2003; Raffa et al., 2006).

Recent studies have identified a substrate preference for Cds1^Chk2 (O'Neill et al., 2002; Seo et al., 2003), showing that an arginine residue at the –3 position is the most important residue for the substrate preference. Peptide phosphorylation is substantially decreased when alanine is substituted for arginine at the –3 position. Thus, Cds1^Chk2 preferentially phosphorylates serine or threonine residues in an RxxS/T motif. To gain further insight into the regulation and significance of Rad60 phosphorylation by Cds1^Chk2, we substituted alanine for serine or threonine in the Cds1^Chk2 consensus target motifs in Rad60. We identified two residues in Rad60 that are targeted for phosphorylation in a Cds1^Chk2-dependent manner and found that these are responsible for the re-localization of Rad60 protein in response to HU treatment. In the absence of this re-localization we observed no quantifiable HU sensitivity or sensitivity to a range of DNA-damaging agents. However, we find that the rad60 phosphorylation-site mutations suppress specific smc6 mutations, indicating that Rad60 is required to regulate homologous recombination at stalled replication forks by controlling Smc5/6 activity.

In vitro, Cds1^Chk2 kinase exhibits a preference for the substrate motif RxxS/T (O'Neill et al., 2002; Seo et al., 2003). We found three potential phosphorylation sites matching this motif in the Rad60 protein (Fig. 1A). Threonine 72 (T72) and serine 126 (S126) are located in the N-terminal domain and T365 is in the SUMO-like domain 2 at the C-terminus. To examine whether these residues are phosphorylated by Cds1^Cds1, we individually or in combination replaced them with alanine and introduced the mutations into the genomic rad60⁺ locus. Because the phosphorylated form of Rad60 protein is known to show a significant hypermobility shift, we resolved Rad60 by SDS-PAGE and detected the protein using an anti-Rad60 antibody. As shown in Fig. 1B, four distinct forms of Rad60 were detected, in agreement with previously published results (Raffa et al., 2006), and the level of Rad60 protein was unaffected by the mutations introduced. After treatment with HU, both Rad60-T72A and Rad60-S126A showed an intermediate hypershift (form 1 to form 3 and form 2 to form 3, respectively), whereas almost all of the wild-type Rad60 was converted to form 4. The hypershift essentially disappeared in the rad60-T72A S126A double-mutant cells (we will refer to the rad60-T72A S126A double mutant as rad60-2A and the rad60-T72A S126A T365A triple mutant as rad60-3A).

Because the Rad60 hypershift is known to be dependent on Cds1^Chk2, we next performed the kinase assay in vitro using recombinant Rad60 as substrate for immunoprecipitated Cds1^Chk2. Wild-type Rad60 was efficiently phosphorylated in vitro and this phosphorylation was dependent on Cds1^Chk2 (Fig. 2A). The efficiency of phosphorylation was significantly decreased for Rad60-2A and Rad60-3A proteins, suggesting that T72 and/or S126 are direct targets of Cds1^Chk2. Next, we examined the effect of each single mutant on phosphorylation in vitro (Fig. 2A, right panels). T72A decreased the phosphorylation signal to a similar extent as the Rad60-2A mutant protein. S126A also apparently decreased the signal but the effect was less significant. To further clarify the phosphorylation of S126, we expressed truncated proteins to separate S126 from T72. The N-terminal fragment (N1) and middle fragment (M1) encompass T72 or S126, respectively. Both fragments were phosphorylated by Cds1^Chk2 in vitro, and the corresponding alanine mutants (N1A and M1A) significantly decreased the signal (Fig. 2B). Together with the phosphorylation-dependent hypershift data (Fig. 1) these data led us to conclude that both T72 and S126 are direct targets of Cds1^Chk2.

Rad60 is diffused throughout the whole cell in response to HU treatment, whereas it is localized in the nucleus during the normal cell cycle. This HU-dependent re-localization of Rad60 is dependent on Cds1^Chk2 (Boddy et al., 2003). We thus examined the effects of T72 and S126 phosphorylation-site mutants on the localization of Rad60. Myc-tagged wild-type and mutant Rad60 were expressed from the rad60 genomic locus and stained with anti-myc antibody. Both the wild-type and mutant versions of Rad60 were localized in the nucleus in the absence of HU (Fig. 3A). Wild-type Rad60 was dispersed throughout the whole cell following treatment with HU and showed only a weak nuclear signal, as previously described. The rad60-S126A mutation showed the most striking effect on localization. Rad60-S126A was found only in the nucleus, even after treatment with HU. The Rad60-T365A protein behaved in an identical manner to wild-type Rad60 whereas the Rad60-T72A protein displayed an intermediate pattern of localization. These in vivo results suggest that phosphorylation of S126 is the primary requirement for the re-localization of Rad60 in response to Cds1^Chk2 activation following HU treatment. Interestingly, we observed that T72 was a better substrate for Cds1^Chk2 than S126 in vitro. It has been reported that phosphorylation of T72 is required for the interaction of Rad60 with the FHA domain of Cds1 (Raffa et al., 2006). Thus, phosphorylation of T72 might be necessary for efficient phosphorylation of S126 and thus for efficient re-localization of the protein.

rad60-T72A and rad60-S126A mutations both affected the re-localization of Rad60. However, cells expressing either single-mutant proteins or double- and triple-mutant proteins from the rad60 locus, and at equivalent levels to wild-type protein, are not sensitive to HU, methyl methanesulfonate (MMS) or mitomycin C (MMC) even at concentrations sufficient to reduce the viability of wild-type cells (Fig. 3B). One explanation for this could be that there are alternative redundant mechanisms that result in the same phenotypic effect as the absence of Rad60 phosphorylation. We therefore examined the effects of deleting various genes (e.g. rhp51, rhp18, mus81, rqh1, srs2, brc1, slx1) on the sensitivity of rad60-2A to HU. However, loss of Rad60 phosphorylation caused no significant effect in any of these backgrounds (data not shown). There are several independent mechanisms for maintaining or repairing the stalled replication forks, and this could complicate our efforts to detect the effect of phosphorylation of a single DNA repair protein.

Previous studies have shown that Rad60 interacts with the Smc5/6 protein complex both physically and genetically (Boddy et al., 2003; Morikawa et al., 2004; Morishita et al., 2002). We therefore examined whether rad60-2A affected the sensitivity of smc6 mutants to genotoxic stress. As shown in Fig. 4A, rad60-2A suppressed the HU sensitivity of the smc6-X mutant. Similar, but less pronounced suppression was observed for the smc6-74 mutant. rad60-2A failed to suppress the UV sensitivity of these smc6 mutants, consistent with the fact that UV irradiation does not induce Rad60 phosphorylation (data not shown). Smc5/6 has been proposed to function during DNA repair by homologous recombination. However, rad60-2A could not suppress the sensitivity of rhp51Δ cells to DNA damage (data not shown). Thus, rad60-2A does not bypass the requirement for homologous recombination when the Smc6 protein is dysfunctional, but appears to enhance functions of the hypomorphic mutant Smc6.

These results reminded us of the fact that rad60 has been shown to act as a multicopy suppressor of smc6-X (Morishita et al., 2002). However, we observe that multicopy rad60 suppresses both the HU and UV sensitivity of smc6 mutants (Fig. 4B). This suppression is less pronounced for smc6-74 than for smc6-X, which is consistent with the suppression by rad60-2A. These results suggest that the suppression of smc6 mutants by rad60-2A is due to an excess of Rad60 protein in the nucleus in the presence of HU.

To verify the suppression of smc6 mutants by rad60-2A we performed an assay to measure loss of the ura4⁺ gene integrated within one copy of the ribosomal DNA. smc6-X cells show a significantly elevated frequency of ura4⁺ loss from the ribosomal DNA (Irmisch et al., 2009). In this assay, loss of the ura4⁺ gene is probably due to ectopic sister chromatid recombination or intrachromosomal recombination. As shown in Fig. 4C, the frequency of ura4 loss is elevated more than tenfold in smc6-X cells and this was partially suppressed by rad60-2A. Specifically, the frequency of HU-induced ura4⁺ loss was reduced by a factor of ∼50% in the double mutant. Suppression can also be seen in untreated cells, though it is less dramatic (Fig. 4C). These results led us to conclude that the suppression of phenotypes of smc6 mutants by rad60-2A is significant. Interestingly, even in the smc6⁺ background, the frequency of HU-induced ura4 loss in rad60-2A was significantly lower than that in the wild type (2.4 ±0.3% and 3.7 ±0.9%, respectively). This suggests that the phosphorylation of Rad60 is required for cells to correctly regulate recombination at ribosomal DNA after replication stalls, although phosphorylation of Rad60 does not detectably affect cell viability.

In this study, we have employed site-directed mutagenesis to identify Cds1^Chk2 target residues in Rad60. Analysis of phosphorylation in vivo and in vitro has identified two target residues, T72 and S126. Phosphorylation of T72 has previously been reported to be required for correct binding between Rad60 and Cds1^Chk2 (Boddy et al., 2003; Raffa et al., 2006). The FHA domain of Cds1^Chk2 preferentially binds to TxxD motifs in which threonine is phosphorylated (Durocher et al., 2000). Rad60-T72 is not only located in a putative Cds1^Chk2 target motif, RxxS/T, but also encompasses the Cds1^Chk2 FHA binding motif, TxxD. Because we used recombinant protein purified from E. coli extract for the kinase assay, it is unlikely that Cds1^Chk2 binds phosphorylated T72 and phosphorylates another residue in vitro. Therefore, we conclude that T72 is a direct target of Cds1^Chk2.

Phosphorylation of S126 is less efficient in vitro than that of T72 although the S126A mutation causes a more striking effect on the re-localization of Rad60 in response to HU treatment (Fig. 2A,B; Fig. 3A). These observations suggest that Cds1^Chk2 first phosphorylates T72 and stabilize its association with Rad60 and that this stabilization is required for efficient phosphorylation of S126, which is responsible for the re-localization of Rad60. Consistent with this model, S126 is inefficiently phosphorylated in vitro when Rad60-T72A protein is used as a substrate (Fig. 2A). The model is also consistent with the observation that defects in the hypershift of Rad60 are more severe in the rad60-T72A mutant than in the rad60-S126A mutant (Fig. 1B).

The target consensus recognition of Cds1^Chk2 is influenced not only by the residue at the –3 position but also, to a lesser extent, by that at position –5. Leucine at –5 increases the phosphorylation of peptide, although the effect is much less dramatic than that of arginine at –3 (O'Neill et al., 2002). There is a leucine at –5 of T72 but not S126. Stable Cds1^Chk2 binding might be required for S126 to be efficiently phosphorylated. The dispersal of Rad60 is inhibited in the presence of leptomycin B (supplementary material Fig. S1), indicating that Crm1-dependent nuclear export is involved in this dispersal. Although Crm1 requires a nuclear export signal (NES) to export target proteins, S126 is not located in an apparent NES. An NES predictor (NetNES) predicts a putative NES in Rad60 at the C-terminus. However, it has not been determined whether this putative NES is functional for nuclear export. It is also possible that another Rad60-interacting protein containing an NES is required for the nuclear export of Rad60. Further study is thus needed to elucidate the mechanism relating Cds1^Chk2-dependent Rad60 phosphorylation to Rad60 re-localization.

Here, we have shown that the hypershift of Rad60 is completely abolished in rad60-2A mutant cells and we have failed to detect Cds1^Chk2-dependent phosphorylation of the N-terminal portion of Rad60 in vitro when the T72 and S126 residues were changed to alanine. On the other hand, Raffa and collegues identified phosphorylation of S32 and S34 and showed that these residues affect the hypershift of Rad60 in response to HU (Raffa et al., 2006). These observations suggest that Rad60 is tightly regulated by post-translational modifications.

The rad60-2A mutation suppressed the HU sensitivity of smc6 mutants whereas the rad60-2A mutant cells were not hypersensitive to DNA-damaging agents (Fig. 3B; Fig. 4A). It has been proposed that Rad60 functions in concert with the Smc5/6 complex because Rad60 physically interacts with Smc5/6 and because hypomorphic mutants of rad60 show mutual genetic interactions with smc6. Our results support a model in which Rad60 assists Smc5/6 in its functions. When the function of Smc5/6 is compromised by hypomorphic mutation, an additional quantity of nuclear Rad60 appears to facilitate the response to replication stress. In addition to the suppression of the frequency of ura4⁺ loss from ribosomal DNA in smc6-X mutants by the rad60-2A mutation, we also observed that the frequency of ura4⁺ loss was decreased in rad60-2A single mutant cells (Fig. 4C).

It has been reported that cohesin, another SMC complex that is required for sister chromatid cohesion, regulates the length of ribosomal DNA repeats in S. cerevisiae (Kobayashi and Ganley, 2005). On the one hand, Smc5/6 also localizes on ribosomal DNA in S. cerevisiae and is required for proper separation of this region during mitosis (Torres-Rosell et al., 2005). On the other hand, Smc5/6 is shown to be required for efficient sister chromatid recombination: in smc6 mutants, ectopic recombination is elevated whereas sister chromatid recombination is decreased (De Piccoli et al., 2006). Elevated ectopic recombination in the ribosomal DNA at the expense of sister chromatid recombination is consistent with the increased ura4⁺ loss we observed in S. pombe. Thus, to ensure ectopic recombination in specific situations, such as the maintenance of ribosomal DNA repeats, cells might need to regulate Smc5/6 by reducing the concentration of Rad60 in the nucleus.

In S. pombe, a recombination protein Mus81 is also phosphorylated in Cds1^Chk2-dependent manner (Boddy et al., 2000). The mus81-T239A mutation abolishes the interaction of Mus81 with Cds1 in a similar manner to the abolition of the Rad60-Cds1^Chk2 interaction in the rad60-T72A mutant. Interestingly, the regulation of Mus81 in response to replication stress closely resembles that of Rad60. Mus81 dissociates from chromatin in the presence of HU whereas Mus81-T239A protein remains chromatin-associated. However, mus81-T239A mutation enhances recombination frequency of cells after HU treatment (Kai et al., 2005). This is exactly the opposite of what we have observed in rad60-2A cells. However, mus81 is essential for growth of rad60 mutants (Boddy et al., 2003; Morishita et al., 2002), suggesting that these genes have overlapping functions. Slx1/4, another structure-specific endonuclease, has also been shown to be involved in recombination at ribosomal DNA repeats in S. pombe (Coulon et al., 2004; Coulon et al., 2006). In S. cerevisiae, the non-catalytic subunit Slx4 is phosphorylated in an Mec1^ATR-dependent manner and is required for phosphorylation of ESC4 protein (Flott and Rouse, 2005; Roberts et al., 2006), which is required for restart of stalled replication forks (Rouse, 2004). ESC4 is a homolog of S. pombe brc1, a multicopy suppressor of smc6-74 (Lee et al., 2007; Sheedy et al., 2005; Verkade et al., 1999). Multicopy brc1 suppresses the sensitivity of smc6-74 but not smc6-X to DNA damage, but multicopy rad60 suppresses smc6-X more dramatically than it does smc6-74 (Fig. 4B). Taken together, these data clearly indicate that checkpoint responses regulate multiple pathways to overcome the difficulties induced by replication stress. Understanding the intricacies of this regulation is a complex but important job.

In this report, we have shown that checkpoint kinase Cds1^Chk2 regulates homologous recombination at the ribosomal DNA repeats through the phosphorylation of Rad60 at T72 and S126. The phosphorylated form of Rad60 disperses from the nucleus and this dispersal appears to promote ectopic recombination, possibly by influencing the function of Smc5/6. In S. cerevisiae and S. pombe, Smc5/6 has been shown to localize at centromeric and repeated sequences, including the ribosomal DNA repeats (Pebernard et al., 2008; Torres-Rosell et al., 2005). It has also been reported that Smc5/6 accumulates at the sites of DNA double-strand breaks or collapsed replication forks (Lindroos et al., 2006). S. cerevisiae ESC2 is a putative homolog of rad60, and has been shown to be involved in chromatin silencing and sister chromatid cohesion (Dhillon and Kamakaka, 2000; Ohya et al., 2008). However, there is no evidence that S. pombe Smc5/6 has similar activity, and a correlative interaction between ESC2p and Smc5/6 has not been reported. One possibility is that Rad60 plays a role in regulating the localization of Smc5/6 on chromatin. Further studies are required to elucidate the function of Rad60, and how this relates to Smc5/6 functions.

The S. pombe strains used in this study are listed in Table 1. S. pombe cells were grown in medium supplemented with yeast extract (YE) or in Edinburgh minimal medium (EMM). Standard genetic and molecular procedures were employed as described previously (Moreno et al., 1991). To examine the sensitivity to drugs, serial dilutions of cells were spotted on YE plates (YEA) containing each drug, and incubated at 30°C for 3-4 days. NLS sequences were identified using NetNES predictor (la Cour et al., 2004).

Total protein was extracted in buffer G (50 mM Tris-HCl, 100 mM NaH₂PO₄, 6 M guanidine hydrochloride, pH 8.0). For SDS-PAGE, proteins were precipitated with trichloroacetic acid and resuspended in 1× SDS-PAGE sample buffer. Subsequently, proteins were transferred to PVDF membranes and probed with affinity-purified anti-Rad60 (BioAcademia, Osaka, Japan). Detection was performed with HRP-conjugated secondary antibody and ECL Advance Western Blot Detection Kit (GE Healthcare).

GST-Rad60 was expressed in E. coli, and purified on glutathione Sepharose (GE healthcare). Purified protein was incubated with immunoprecipitated Cds1^Chk2 as described previously (Lindsay et al., 1998). Samples were subjected to 10% SDS-PAGE and gels were dried and exposed to phospho-image screens after staining with Coomassie brilliant blue.

Strains expressing Myc-tagged Rad60 from the native locus were used for indirect immunofluorescence. Cells were fixed with 3.7% formaldehyde and processed as described previously (Caspari et al., 2000). Processed cells were stained with anti-Myc monoclonal antibody (9E10) and Alexa-Fluor-488-conjugated goat anti-mouse IgG (Molecular Probes, Eugene, OR). Stained cells were observed under an epifluorescence microscope and photographed.

In three separate experiments, 11 independent single colonies for each strain were inoculated into 10 ml YE (2% YE, 6% sucrose, pH 5.5) medium and grown to stationary phase. 1×10³ cells were plated on YEA, grown for ∼5 days at 30°C and then the colonies were replica plated to medium without uracil. To assay for ura4 loss after treatment with HU, colonies were inoculated as above, grown to mid-log phase and treated with 10 mM HU for 4 hours. Cells were then washed, resuspended in YE and grown to stationary phase. The ura4 loss assay measures marker loss from a single ura4+ gene integrated into one ribosomal DNA repeat.