Phosphorylation of histone H2A or H2AX is an early and sensitive marker of DNA damage in eukaryotic cells, although mutation of the conserved damage‐dependent phosphorylation site is well tolerated. Here, we show that H2A phosphorylation is required for cell‐cycle arrest in response to DNA damage at the G1/S transition in budding yeast. Furthermore, we show that the tandem BRCT domain of Rad9 interacts directly with phosphorylated H2A in vitro and that a rad9 point mutation that abolishes this interaction results in in vivo phenotypes that are similar to those caused by an H2A phosphorylation site mutation. Remarkably, similar checkpoint defects are also caused by a Rad9 Tudor domain mutation that impairs Rad9 chromatin association already in undamaged cells. These findings indicate that constitutive Tudor domain‐mediated and damage‐specific BRCT domain–phospho‐H2A‐dependent interactions of Rad9 with chromatin cooperate to establish G1 checkpoint arrest.
The eukaryotic DNA‐damage response involves the ataxia‐telangiectasia‐mutated (ATM) and ATM‐ and Rad3‐related (ATR) protein kinase‐dependent signal‐transduction pathways that preserve genomic integrity (Shechter et al, 2004; Lavin et al, 2005). ATM/ATR phosphorylation of H2AX on serine 139 is required for retention of the check‐point mediator proteins p53 binding protein 1 (53BP1) and mediator of DNA‐damage checkpoint 1 (MDC1) on chromatin flanking DNA double‐strand break (DSB) sites, and it acts to promote both checkpoint signalling and DSB repair (Stucki & Jackson, 2006). Similarly, the Saccharomyces cerevisiae ATM/ATR orthologues Tel1 and Mec1 also phosphorylate histone H2A (Downs et al, 2000) on serine 129 (S129) to recruit chromatin remodelling factors to sites of DSBs (Downs et al, 2004; Morrison et al, 2004; van Attikum et al, 2004) and promote efficient DSB repair (Downs et al, 2000; Redon et al, 2003).
Other histone modifications also participate in the DNA‐damage response. For example, DSB‐induced acetylation of lysine residues in histones H3 and H4 affects repair efficiency (Bird et al, 2002; Jazayeri et al, 2004; Masumoto et al, 2005), and histone lysine methylation is linked to the activation of checkpoint signalling (Huyen et al, 2004; Sanders et al, 2004; Wysocki et al, 2005; Botuyan et al, 2006). Human 53BP1, S. cerevisiae Rad9 and Schizosaccharomyces pombe Crb2 contain tandem Tudor domains (Tudor2 domains) that directly recognize methylated lysine residues (Huyen et al, 2004; Botuyan et al, 2006; Du et al, 2006). In budding yeast, mutation of either histone H3 lysine 79 (K79) or the Rad9 Tudor2 domain causes G1‐ and S‐phase checkpoint defects but does not abolish the RAD9‐dependent G2/M arrest (Wysocki et al, 2005). Therefore, we tested whether other known DNA damage‐associated histone modifications might elicit cell‐cycle phase‐specific responses. We show that defective H2A S129 phosphorylation impairs checkpoint signalling in G1 but does not compromise the G2 checkpoint, and that phospho‐H2A‐dependent G1 checkpoint signalling involves the direct interaction between phosphorylated H2A and the tandem BRCT domain (BRCT2) of Rad9.
Results And Discussion
H2A phosphorylation is required for the G1 checkpoint
Although histone H2A phosphorylation promotes DNA repair in yeast, its role in checkpoint control has not been thoroughly explored (Downs et al, 2000; Redon et al, 2003). As both DSB‐induced checkpoint signalling and DSB‐induced H2A S129 phosphorylation in G1 cells depend on the MRX‐Tel1 pathway (Grenon et al, 2001; Shroff et al, 2004), we used, for checkpoint analyses, an allele of H2A (hta‐S129*) in which S129 is mutated to a STOP codon that removes the last four residues (Downs et al, 2000).
Notably, G1‐arrested wild‐type cells showed the characteristic phosphorylation‐dependent Rad53 mobility shift in response to the radiomimetic drug bleomycin, whereas this Rad53 shift was markedly diminished in G1‐arrested hta‐S129* cells (Fig 1A, top panel). By contrast, the extent of Rad53 hyperphosphorylation following bleomycin treatment in G2/M‐arrested hta‐S129* cells was very similar to that in wild‐type cells, with only a modest defect at very low doses (Fig 1A, bottom panel). Similarly, hta‐S129* cells were severely defective in DNA‐damage‐induced hyperphosphorylation of Rad9, the key mediator of Rad53 activation (Sweeney et al, 2005); again this effect was more pronounced during G1 than in G2/M‐arrested cells (Fig 1B). Consistent with this, wild‐type cells delayed bud emergence—a marker of progression past START—following exposure to ionizing radiation during G1, whereas this response was completely absent in hta‐S129* cells (Fig 1C). By contrast, hta‐S129* cells were as proficient as wild‐type cells in delaying the onset of anaphase following bleomycin treatment of G2/M‐arrested cells (Fig 1D). On the basis of these findings, we conclude that the H2A C terminus, including the S129 phosphorylation site, is required for efficient damage‐induced hyperphosphorylation of Rad9 and Rad53, and that this function is specifically crucial for the G1 checkpoint.
Rad9:phospho‐H2A binding controls the G1 checkpoint
The tandem BRCT (BRCT2) domain‐dependent interaction of the human Rad9 homologue MDC1 with phosphorylated H2AX is required for the efficient localization of MDC1 to DSBs (Stucki et al, 2005). Therefore, we investigated whether the Rad9 BRCT2 domain could interact in a similar manner with phosphorylated H2A to mediate G1 checkpoint signalling. Phosphopeptide pull‐down experiments showed that the purified Rad9 BRCT2 domain formed a relatively salt‐stable interaction with a phosphorylated peptide comprising the C‐terminal 13 residues of histone H2A surrounding S129, but not with an unphosphorylated control peptide (Fig 2A,B). By contrast, no phosphopeptide binding was observed for a BRCT2 mutant in which Lys 1088 was changed to methionine (Rad9‐K1088M)—a mutation analogous to K1937M of MDC1 that abolishes H2AX phosphopeptide binding (supplementary Fig 1 online; Stucki et al, 2005). The BRCT2 domains of BRCA1 and MDC1 did not interact with this phosphopeptide (Fig 2C) and, unlike the phospho‐H2A peptide itself, unrelated phospho‐serine‐containing peptides did not competitively inhibit binding of the Rad9 BRCT2 domain to beads bearing the phospho‐H2A peptide (Fig 2D), underscoring the specificity of this interaction. These data therefore indicate that the Rad9 BRCT2 domain interacts specifically with phosphorylated H2A in a structurally analogous manner to the interaction between phosphorylated H2AX and MDC1.
To investigate the functional importance of the phospho‐dependent interaction between the Rad9 BRCT2 domain and H2A, we assessed the phenotypes of cells bearing the rad9‐K1088M mutation. Similar to hta‐S129* cells, G1‐arrested rad9‐K1088M cells were completely unable to delay bud emergence following ionizing radiation treatment (Fig 3A). We also analysed the role of the Rad9 BRCT2 domain in other Rad9‐dependent checkpoints. Notably, wild‐type cells slowed DNA replication in medium containing 0.02% methyl methane sulphonate (MMS), whereas rad9‐K1088M cells rapidly completed S phase in the presence of MMS, similar to rad9Δ cells (Fig 3B), indicating that the BRCT2 domain is also involved in at least some aspects of the intra‐S‐phase checkpoint. As hta‐S129A cells are reported to have an intact intra‐S‐phase checkpoint (Redon et al, 2003), this suggests that the Rad9 BRCT2 domain might have other targets in addition to phospho‐S129 H2A. However—and again in a manner similar to the hta‐S129* phenotype but in contrast to rad9Δ—the rad9‐K1088M cells efficiently delayed anaphase following treatment with bleomycin during G2/M (Figs 1D,3C). This, in conjunction with the fact that Rad9‐K1088M protein levels are similar to those of the wild‐type protein (data not shown), indicates that rad9‐K1088M is not a null allele. Consistent with the similar G1 checkpoint defect, rad9‐K1088M cells also had G1‐specific Rad9 and Rad53 phosphorylation defects similar to hta‐S129* cells. Compared with wild‐type cells, rad9‐K1088M cells were almost completely defective for bleomycin‐ and ultraviolet‐induced Rad9 hyperphosphorylation in G1 (Fig 4A, left panels; Fig 4B), but showed essentially normal levels of bleomycin‐induced Rad9 hyperphosphorylation during G2/M (Fig 4A, right panels). Rad53 phosphorylation was also severely impaired in rad9‐K1088M cells following ionizing radiation or ultraviolet irradiation of G1 cells (Fig 4C). Together with the phosphopeptide binding analyses, these results strongly indicate that the Rad9 BRCT2 domain mediates the phospho‐S129 H2A signal to effect G1 checkpoint arrest.
Rad9 Tudor2 and BRCT2 domains functionally cooperate
G1/S checkpoint defects similar to those identified here for BRCT2/H2A mutations were recently reported for substitution of the conserved tyrosine 798 (Y798) within the Tudor2 domain of Rad9 or loss of the Tudor2 domain ligand—H3 methyl‐K79 (Wysocki et al, 2005). For direct comparison, we created an alanine substitution of RAD9 Y798 (rad9‐Y798A) and a rad9‐Y798A, K1088M double mutant. Strikingly, the phenotypes of rad9‐K1088M cells and rad9‐Y798A cells were virtually identical, with defective G1/S and intra‐S‐phase checkpoints but normal G2/M checkpoints (Fig 3), as well as impaired Rad9 and Rad53 hyperphosphorylation in response to DNA damage during G1 but not G2 (Fig 4A). Importantly, the phenotype of the rad9‐Y798A, K1088M double mutant, with respect to Rad9 and Rad53 phosphorylation following DNA damage in G1, was similar to the two single‐domain mutants (Fig 4A,C), indicating that the Tudor2 and BRCT2 domains function together to mediate efficient G1 checkpoint activation.
The similar checkpoint defects of Rad9 Tudor2 and BRCT2 domain mutations raised the possibility that both domains might be needed to target Rad9 to sites of damaged chromatin. To test this possibility, we treated G1‐arrested spheroplasts with ultraviolet irradiation, and then separated extracts into soluble and chromatin‐enriched fractions to analyse Rad9 association with chromatin (Fig 5). Surprisingly, we found that significant amounts of Rad9 were already associated with chromatin in extracts from undamaged G1 cells (Fig 5A). By contrast, Rad53 was found almost exclusively in the soluble fraction. After ultraviolet irradiation, wild‐type Rad9 was depleted from the soluble pool and redistributed almost exclusively to the chromatin‐enriched fraction (Fig 5A). Strikingly, mutation of the Tudor2 domain almost completely abolished Rad9 chromatin association in undamaged cells and, unlike the wild‐type protein, Rad9‐Y798A was not further recruited into chromatin (Fig 5A). This suggests that a constitutive chromatin association of Rad9 in undamaged cells is a prerequisite for its hyperphosphorylation and concomitant enrichment in chromatin following DNA damage. In contrast to the Tudor2 domain mutant, the basal chromatin association of the Rad9‐K1088M BRCT2 mutant in undamaged cells was undiminished compared with the wild type, but it also failed to be further enriched in damaged chromatin or to be hyperphosphorylated after DNA damage (Fig 5B). The simplest explanation for these data is that the Rad9 Tudor2 domain mediates a constitutive association of Rad9 with chromatin during G1 phase in undamaged cells, as a prerequisite for its DNA damage‐specific BRCT2‐dependent enrichment in the vicinity of DNA damage. Although we cannot exclude the possibility that the observed distribution of Rad9 in otherwise undamaged G1 cells reflects a possible disruption of chromatin fibres that is recognized as DNA damage during extract preparation, we feel that this is unlikely and note that others did not observe a similar association of Rad9 in undamaged G2‐arrested cells using an almost identical lysis procedure (Toh et al, 2006).
Together, our results indicate that histone H2A phosphorylation is required for establishment of the G1 DNA‐damage checkpoint and support a model in which this function involves a direct interaction between phosphorylated H2A and the Rad9 BRCT2 domain. We have also found that both the Tudor2 and BRCT2 domains of Rad9 are required for the G1 checkpoint response, and that Tudor2 domain‐dependent chromatin localization of Rad9 in undamaged cells is crucial for the establishment of G1 checkpoint signalling. The nuclear redistribution of Rad9 in G1 cells after DNA damage is in marked contrast to that observed previously in G2 cells in which hypophosphorylated forms of Rad9 were retained only at later time points after ionizing radiation (Toh et al, 2006). This difference indicates that, in contrast to G1 cells, sustained chromatin binding is dispensable for correct checkpoint activation in G2, perhaps reflecting a requirement for a lower activation threshold.
Our results thereby confirm and extend the previously published finding that the Rad9 Tudor2 domain is required for the G1 checkpoint (Wysocki et al, 2005). While this manuscript was in preparation, our initial findings were independently confirmed by a report that H2A phosphorylation and H3 K79 methylation both control G1/S checkpoint arrest and Rad9 occupancy at DSBs in G1 (Javaheri et al, 2006). However, in contrast to their finding of a partial G1 arrest defect in hta‐S129A strains—and more consistent with the findings reported in a previous study (Redon et al, 2003)—we have found that hta‐S129* cells are completely G1 checkpoint‐defective. The simplest explanation for this apparent discrepancy could be the higher dose of ionizing radiation used by Javaheri et al (2006).
On the basis of the available data, in particular the constitutive chromatin localization of Rad9 (Fig 5), we speculate that the Rad9 Tudor2 domain—and, by analogy, the equivalent domains of Rad9 counterparts in other organisms—might recognize multiple histone methylation marks to enable association with bulk chromatin. Indeed, in budding yeast, two distinct histone lysine methylases are required for the G1 checkpoint response (Giannattasio et al, 2005; Wysocki et al, 2005), and the Tudor2 domain of 53BP1 recognizes at least three histone methylation marks in vitro (Kim et al, 2006) and might recognize multiple ligands in vivo (Botuyan et al, 2006). Constitutive chromatin ‘scanning’ could enhance the speed and efficiency of the DNA‐damage response, particularly after low doses of DNA damage. H2A phosphorylation might then act to concentrate Rad9 near sites of damage through the BRCT2 domain and thus allow efficient Rad9 hyperphosphorylation and signalling. At present, it is unclear why G1 checkpoint activation requires two histone‐binding modules that are largely dispensable in G2. As Rad9 is phosphorylated in normal post‐G1 cells, and cyclin‐dependent kinase (CDK) phosphorylation of the Rad9 orthologue Crb2 is required for the correct repair of ionizing radiation induced lesions (Caspari et al, 2002) and also for ionizing radiation induced focus formation in the absence of ligands for the Tudor2 and BRCT2 domains (Du et al, 2006), it is tempting to speculate that CDK activity controls both the choice of DNA‐repair pathway and the mode of checkpoint activation. As Rad9 is also required for efficient DNA repair (Toh et al, 2006), such controls could act to regulate and coordinate checkpoint arrest and DNA repair mechanisms.
Yeast strain construction and checkpoint experiments. Strains were constructed using PCR‐based allele replacement techniques (Erdeniz et al, 1997) and are listed in supplementary Table 1 online. Analysis of yeast checkpoint protein phosphorylation and cell‐cycle progression was carried out as detailed in the supplementary information online.
GST fusion expression and purification. Sequences encoding residues 981–1,309 of Rad9 and residues 1,634–1,863 of BRCA1 encompassing the BRCT2 domains were cloned into the vector pGEX4T1 for bacterial expression and purified using glutathione‐Sepharose (GE Healthcare, Rydalmere, Australia). Thrombin‐cleaved BRCT2 domains were purified by cation‐exchange chromatography, as detailed in the supplementary information online.
Peptide‐binding assays. Synthetic biotinylated peptides (W. Mawby, University of Bristol, UK) were coupled to streptavidin dynabeads M‐280 (Invitrogen, Mt Waverley, Australia) in TBS‐T (Tris‐buffered saline Tween‐20) in excess, such that all available biotin‐binding sites were blocked. The beads were then washed and incubated with purified BRCT2 domains (∼1–4 μg) in TBS‐T for 90 min at 4°C, followed by extensive washing before analysis by SDS–polyacrylamide gel electrophoresis and Coomassie blue staining. For competition assays, BRCT2 domains were preincubated with 100 μg competitor peptides at 4°C for 2 h before addition to the immobilized phosphopeptide as above.
Chromatin fractionation. Chromatin fractionation was carried out, essentially as described previously (Donovan et al, 1997), on G1‐arrested cells that were held in G1 for the entire procedure by inclusion of alpha‐factor in all buffers. For ultraviolet irradiation, spheroplasts were washed and resuspended in PBS‐S (PBS containing 1 M sorbitol), and treated with 30–100 J/m2 ultraviolet radiation (depending on the ultraviolet source) and then allowed to recover for 30 min at 30°C in YPD‐S (YPD containing 0.7 M sorbitol, 25 mM Tris, pH 7.4) before washing and extract preparation. Effective fractionation was assessed using antibodies to either Orc6 (Weinreich et al, 1999) or Orc2 (Duncker et al, 2002).
Supplementary information is available at EMBO reports online (http://www.emboreports.org).
Supplementary Fig 1
We thank N. Lowndes (Rad9, Rad53), B. Stillman (Orc6) and B. Duncker (Orc2) for antibodies, A. Traven for critical review of the manuscript and N. Tenis for technical support. A.H. and J.H. are supported by funding from the National Health and Medical Research Council, and the Clive and Vera Ramaciotti Foundation. Work in the S.P.J. laboratory was made possible by a Programme Grant from Cancer Research UK, and core infrastructure funding from Cancer Research UK and the Wellcome Trust.
- Copyright © 2007 European Molecular Biology Organization