In budding yeast, RAD9 and RAD24/RAD17/MEC3 are believed to function upstream of MEC1 and RAD53 in signalling the presence of DNA damage. Deletion of any one of these genes reduces the normal G1/S and G2/M checkpoint delays after UV irradiation, whereas in rad9Δ–rad24Δ cells the G1/S checkpoint is undetectable, although there is a residual G2/M checkpoint. We have shown previously that RAD9 also controls the transcriptional induction of a DNA damage regulon (DDR). We now report that efficient DDR induction requires all the above‐mentioned checkpoint genes. Residual induction of the DDR after UV irradiation observed in all single mutants is not detectable in rad9Δ–rad24Δ. We have examined the G2/M checkpoint and UV sensitivity of single mutants after overexpression of the checkpoint proteins. This analysis indicates that RAD9 and the RAD24 epistasis group can be placed onto two separate, additive branches that converge on MEC1 and RAD53. Furthermore, MEC3 appears to function downstream of RAD24/RAD17. The transcriptional response to DNA damage revealed unexpected and specific antagonism between RAD9 and RAD24. Further support for genetic interaction between RAD9 and RAD24 comes from study of the modification and activation of Rad53 after damage. Evidence for bypass of RAD53 function under some conditions is also presented.
Cells have evolved multiple strategies for tolerating genomic damage. The most important of these are numerous repair systems that remove or bypass potentially mutagenic DNA lesions (Friedberg et al., 1995). However, other cellular responses to DNA lesions are also important for the maintenance of genome stability. Delays at multiple cell‐cycle transitions have long been recognized after exposure of both mammalian and yeast cells to DNA‐damaging agents (Burns, 1956; Brunborg and Williamson, 1978; Murnane, 1995). The genetic control of these delays, termed ‘checkpoints’, was first established in budding yeast by Weinert and Hartwell (1988) who showed that the RAD9 gene functions in G2/M arrest after irradiation with X‐rays. Subsequently, it has become clear that RAD9 also functions at the G1/S (Siede et al., 1993, 1994), intra‐S (Paulovich et al., 1997) and mid‐anaphase (Yang et al., 1997) checkpoints. Defects in checkpoint regulation can lead to genome instability and, in higher eukaryotes, neoplastic transformation (Hartwell and Kastan, 1994; Morgan and Kastan, 1997).
Genetic analysis in the budding yeast has identified a further six genes, RAD17, RAD24, RAD53, MEC1, MEC3 and DDC1, in addition to RAD9, that appear to be required for efficient DNA damage‐dependent checkpoints at G1/S, intra‐S and G2/M (Elledge, 1996; Longhese et al., 1997). Of these genes, RAD53 and MEC1 are also required for an S/M checkpoint in which the presence of unreplicated DNA prevents the onset of mitosis (Elledge, 1996). Preliminary evidence for the order of function for this DNA structure checkpoint pathway(s) has come from analysis of the radiation‐sensitive phenotype of certain mutants. Thus, RAD9 is in an epistasis group distinct from RAD24, and RAD17, MEC3 and DDC1 are members of the RAD24 epistasis group (Eckardt Schupp et al., 1987; Lydall and Weinert, 1995; Longhese et al., 1997). The position of MEC1 relative to RAD9 and the RAD24 epistasis group is, however, not clear from these analyses. RAD53 is believed to function further downstream, as the Rad53 protein is phosphorylated after DNA damage in a MEC1‐dependent manner (Sanchez et al., 1996) and overexpression of RAD53 can bypass deficiencies in RAD9 and MEC1 (Allen et al., 1994; Sanchez et al., 1996).
DNA damage in Saccharomyces cerevisiae results in a rapid transcriptional response (Friedberg et al., 1995). Recently, we described a role for RAD9 in this transcriptional response leading to the induction of a large regulon of >15 genes with roles in DNA repair (Aboussekhra et al., 1996). The observation that this DNA damage regulon (DDR) is genetically and co‐ordinately controlled in a RAD9‐dependent manner is reminiscent of the SOS response of Escherichia coli where a regulon of similar genes with roles in DNA repair is also induced rapidly after DNA damage (Walker, 1984; Shinagawa, 1996). Interestingly, the MEC1 and RAD53 genes are themselves in the DDR (Aboussekhra et al., 1996; Kiser and Weinert, 1996), suggesting a degree of positive feedback regulation in this pathway.
Here we demonstrate that single mutants in either RAD9 or the RAD24 epistasis group are not completely defective for checkpoint activity in either G1/S or G2/M, whereas in the rad9Δ–rad24Δ double mutant the G1/S checkpoint is not detectable and there is only a residual G2/M checkpoint. This indicates that RAD9 and RAD24 control the majority of the DNA damage checkpoint capacity in S.cerevisiae. We have used overexpression analysis to demonstrate that RAD9 and the RAD24 epistasis group are partially redundant in function, operate upstream of RAD53, as expected, and also of MEC1. This analysis also suggests an order of function within the RAD24 epistasis group. Furthermore, we extend our previous observations on the RAD9‐controlled DNA damage‐dependent transcriptional response by showing that RAD9, RAD24, RAD17, MEC3, MEC1 and RAD53 are equally important for induction of the DDR. The residual transcriptional induction observed in single mutants is not observed in a rad9Δ–rad24Δ double mutant. Thus RAD9 and the RAD24 epistasis group contribute additively and equivalently to both checkpoint delays and the transcriptional response after DNA damage. Overexpression studies also demonstrate a specific and reciprocal antagonism between RAD9 and RAD24. Support for this data suggesting genetic interaction between RAD9 and RAD24 comes from study of the phosphorylation of Rad53. The phosphorylation and activation of Rad53 kinase is dependent on both RAD9 and the RAD24 epistasis group but, under some circumstances, the checkpoint pathway can function in the absence of active Rad53 kinase. Together, our data indicate that the efficient functioning of two upstream branches, defined by RAD9 and RAD24, of the DNA damage checkpoint pathway is required for a fully integrated response to DNA damage.
The rad9Δ–rad24Δ double mutant abolishes the G1/S checkpoint and virtually all of the G2/M checkpoint after UV damage
We have shown previously that rad9Δ cells are not completely defective in the G2/M checkpoint (Aboussekhra et al., 1996), suggesting a RAD9‐independent contribution. We considered the possibility that both RAD9 and the RAD24 epistasis group would be required for fully functional checkpoints, both in G1/S and in G2/M. We examined the G1/S checkpoint after UV irradiation in wild‐type, rad9Δ, rad24Δ and rad9Δ–rad24Δ cells by monitoring the appearance of budded cells after release from a G1 block (Figure 1A). The appearance of budded cells indicates the release of cells from the G1 block into the cell cycle. Similarly to previous reports (Siede and Friedberg, 1990; Siede et al., 1993, 1994; Longhese et al., 1996), resumption of the cell cycle after release from the α‐factor block in irradiated wild‐type cells was significantly delayed by 30–40 min compared with the non‐irradiated control. This delay in appearance of budded cells as a response to UV treatment was reduced to ∼20 min in both rad9Δ and rad24Δ single mutants. In contrast, the rad24Δ–rad9Δ double mutant showed no detectable G1/S delay after UV irradiation; rather the kinetics of appearance of budded cells was not significantly different from that of unirradiated wild‐type cells, indicating that the G1/S checkpoint is not detectable in these cells.
To support these data, we performed a second set of experiments that examined the G2/M checkpoint (Figure 1B). Exponentially growing wild‐type, rad9Δ, rad24Δ and rad9Δ–rad24Δ cells were UV irradiated and the proportion of G2 cells monitored. The UV‐irradiated culture of wild‐type cells displayed a marked G2/M checkpoint, producing a maximum of 50% G2/M‐blocked cells after 3 h. Interestingly, by monitoring the proportion of unbudded or small budded cells under these conditions, we have never observed evidence for a significant G1/S checkpoint (data not shown), indicating the dominance of the G2/M checkpoint. In agreement with our previous observations (Aboussekhra et al., 1996), a residual G2/M cell cycle delay was observed for rad9Δ cells, corresponding in this experiment to ∼20% G2/M cells after 3 h. A similar G2/M delay was observed in rad24Δ (Figure 1B). However, the residual G2/M checkpoint delay apparent in single null mutants of RAD9 or RAD24 is reduced further in rad9Δ–rad24Δ cells. Similarly, we have demonstrated a residual G2/M checkpoint in single mutants, which reduced still further in the double mutant, by monitoring the delayed release from a nocodazole block after UV irradiation (data not shown).
We have demonstrated that rad9Δ–rad24Δ cells are more defective in their G1/S and G2/M checkpoints than the respective single mutants. These data suggest a model that places RAD9 and RAD24 on separate but additive branches of the DNA damage‐dependent checkpoint pathway. To confirm this model further, we also examined mec3Δ and rad9Δ–mec3Δ cells for their radiation sensitivity and G1/S and G2/M checkpoints. We observed that these phenotypes were more defective in rad9Δ–mec3Δ cells than in either rad9Δ or mec3Δ single mutants (data not shown). rad9Δ–mec3Δ cells, however, showed less marked phenotypes than the rad24Δ–rad9Δ cells, indicating that a strictly linear pathway within the RAD24 epistasis group is unlikely.
RAD24, RAD17, MEC3, MEC1 and RAD53 have equivalent roles to RAD9 in regulation of the transcription response to DNA damage
RAD9 is required for the transcriptional induction of the DDR after DNA damage (Aboussekhra et al., 1996). We therefore investigated the role of the RAD24 epistasis group in the regulation of the DDR; MEC1 and RAD53 were also tested for this response. Exponentially growing rad9Δ, rad24Δ, rad17Δ, mec3Δ and mec1‐1 cells were arrested in G1 with α‐factor, and either irradiated or mock irradiated while maintaining the G1 arrest. This protocol allows damage‐specific transcriptional induction to be examined in the absence of cell‐cycle regulation and outside of S phase. We examined transcription of several genes representing different functional groups within the DDR: CDC9 (DNA ligase); DUN1 and RAD53 (protein kinases involved in cellular responses to DNA damage); RAD51 (a recA homologue); and RNR1 (a large subunit of ribonucleotide reductase). RNR3 was also probed with qualitatively similar results (data not shown). A clear defect in the transcriptional response of all the genes tested was observed in rad9Δ, rad24Δ, rad17Δ, mec3Δ and mec1‐1 mutants compared with the wild‐type strain (Figure 2A). We also observed similar transcriptional defects using the sad1‐1 mutant allele of RAD53 constructed in the W303 background (data not shown, but for the RNR1 transcript see also Figure 4D). The transcriptional induction defect was similar among all of the mutants. The highest absolute levels of induction detected in wild‐type cells were with the RNR1, RNR3 (data not shown) and RAD51 transcripts (Figure 2B). Although the absolute levels of induction observed in the wild‐type for the other inducible genes tested, CDC9, DUN1 and RAD53, were less dramatic, the dependency on an intact transcriptional pathway was always observed (Figure 2B). We also tested the transcriptional inducibility of the RAD9, RAD24, RAD17 and MEC3 checkpoint genes themselves. Although RAD9 was clearly non‐inducible, we found that RAD24, RAD17 and MEC3 were slightly (∼2‐fold) UV inducible in a manner that was also dependent on an intact checkpoint pathway (data not shown). This, together with the inducibility of RAD53 and DUN1 (Figure 2A), suggests a degree of positive feedback regulation within this pathway.
The rad9Δ–rad24Δ double mutant abolishes any residual transcriptional response after UV damage
In all the single mutants tested, the genes of the DDR, in particular RNR1, RAD52 and CDC9, showed residual induction after UV radiation (Figure 2B). This residual induction was not detected in rad9Δ–rad24Δ cells (Figure 2B). Statistical analysis of the levels of induction of these three transcripts in rad9Δ and rad24Δ single mutants compared with rad9Δ–rad24Δ cells demonstrated significant differences. Thus, the RAD9 and RAD24 genes also function separately and additively in the transcriptional response to DNA damage.
In wild‐type cells, elevated levels of Rad9, Rad24, Rad17 or Mec3 have minor effects on the G2/M checkpoint after DNA damage and no effect on cell‐cycle progression
We used overexpression analysis both to address the possibility that RAD9 and the RAD24 epistasis groups are interconnected and also in an attempt to order their function. First, we examined the G2/M checkpoint after UV irradiation of exponentially growing wild‐type cells transformed with vectors in which the RAD9, RAD24, RAD17 and MEC3 checkpoint genes were placed under the transcriptional control of the inducible GAL1 promoter. In all cases, the presence of high levels of the encoded proteins was confirmed by Western blotting (data not shown). We observed that the normal wild‐type G2/M checkpoint and cell survival responses to UV irradiation were only modestly perturbed by elevated levels of Rad9, Mec3, Rad24 and Rad17 (Figure 3). Furthermore, increased levels of these proteins in wild‐type cells did not affect growth or cell‐cycle kinetics measurably (data not shown).
Elevated levels of Rad9, or any one of Rad24, Rad17 or Mec3, rescue the G2/M checkpoint and survival phenotypes observed in null mutants of both the RAD9 and the RAD24 epistasis categories
We next examined the effects of overexpression of RAD9 or three of the known members of the RAD24 epistasis group in cells harbouring null mutations in these genes. (i) RAD9 overexpression not only rescued the G2/M checkpoint and survival defects observed in rad9Δ cells, but also suppressed these defects in mec3Δ, rad24Δ and rad17Δ cells (Figure 3B, and compare with the curves for the respective single mutants in Figure 3A, D, G and J). Interestingly, in rad24Δ cells containing overproduced Rad9, the G2/M checkpoint was ∼50% more marked than the wild‐type (Figure 3B). (ii) We obtained similar results with MEC3 overexpression. Both the G2/M checkpoint (Figure 3E) and survival defects (Figure 3F) observed in rad9Δ, rad17Δ, rad24Δ and mec3Δ cells were significantly complemented. [Although MEC3 overexpression only fully rescued the G2/M checkpoint in mec3Δ cells, in rad9Δ, rad24Δ and rad17Δ cells 66–75% rescue relative to the wild‐type was observed (Figure 3E).] In addition, MEC3 overexpression somewhat delayed exit from the G2/M checkpoint in rad24Δ and rad17Δ cells, but not in rad9Δ cells (Figure 3E). (iii) RAD24 overexpression fully rescued the G2/M checkpoint deficiency associated with its own null mutant and significantly rescued (81%) this defect in rad9Δ cells (Figure 3H), thereby indicating that another member of the RAD24 epistasis group can bypass checkpoint deficiencies associated with null mutation of RAD9. RAD24 overexpression suppressed the survival defects of all mutants to significant extents (Figure 3I). The G2/M checkpoint rescue in rad17Δ (56% rescue) and mec3Δ (49% rescue) mutants was less than that obtained with MEC3 and RAD9 overexpression (Figure 3B and E). (iv) Overexpression of RAD17 significantly complemented the G2/M checkpoint and survival defects associated with the rad17Δ and rad9Δ backgrounds (∼90% rescue in both cases). Interestingly, with the rad24Δ and mec3Δ mutants (Figure 3K and L), no significant rescue of the G2/M checkpoint defect was observed, suggesting that RAD17 requires both RAD24 and MEC3 for transduction of the damage signal. (v) The RAD53 gene is believed to function downstream of the RAD9 and RAD24 epistasis groups in the DNA damage response pathway (Navas et al., 1996; Sanchez et al., 1996), although the relative position of MEC1 has not been determined precisely. Overexpression of RAD9 or RAD24 did not rescue the UV sensitivities (Figure 3O) or G2/M checkpoint defects (compare Figure 3M and N) observed in sad1‐1 (an allele of RAD53) and mec1‐1 cells, suggesting that MEC1 functions downstream of RAD9 and the RAD24 epistasis group. Neither of these mutations completely abrogate the G2/M checkpoint (Figure 3M), most likely because they may be hypomorphic alleles retaining partial function. However, a MEC1‐ and RAD53‐independent contribution to this checkpoint is formally possible.
In summary, all known members of the RAD24 epistasis group can bypass defects caused by the rad9Δ mutation and, reciprocally, RAD9 overexpression can bypass null mutations in the RAD24 epistasis group efficiently. However, the checkpoint defects and UV sensitivity of sad1‐1 and mec1‐1 cells cannot be overridden by overexpression of either RAD9 or RAD24, demonstrating that RAD53 and MEC1 are essential components downstream of RAD9 and the RAD24 epistasis group of genes for these phenotypes.
Overproduced Rad9 or Rad24 rescues the defect in the transcriptional response to DNA damage observed in null mutants of both RAD9 and the RAD24 epistasis category
We analysed the induction of the DDR in G1‐arrested checkpoint mutants in which RAD9 and RAD24 were overexpressed continuously from the GAL1 promoter. Transcripts were examined prior to, and 30 min after, UV irradiation. Figure 4 shows the results obtained with the RNR1 transcript, but qualitatively similar results were also obtained for CDC9 and RAD53 (data not shown). RAD9 and RAD24 overexpression not only rescued the defect in the transcriptional response observed in their own null mutant backgrounds, but slightly enhanced (2‐ to 3‐fold) the response observed in wild‐type cells (Figure 4A and B), suggesting that they are somewhat rate limiting for this response. RAD9 overexpression resulted in rescue of the transcriptional induction defect found in null mutations of all tested members of the RAD24 epistasis group to wild‐type levels (Figure 4A). Although RAD24 overexpression complemented the transcriptional induction defect observed in the rad24Δ and rad17Δ strains to the same level as found in the wild‐type (Figure 4B), it did not rescue the defect seen in mec3Δ cells. RAD17 overexpression complemented the transcriptional defect in its own null mutant to near wild‐type levels and the defect in rad9Δ cells to ∼70% of wild‐type levels (Figure 4C). However, it only weakly rescued the transcriptional defect of rad24Δ cells (50% rescue) and did not significantly rescue this defect in mec3Δ cells (Figure 4C). Unlike RAD9 and RAD24 overexpression, RAD17 overexpression did not cause an increase in transcription in the wild‐type background, suggesting that it is not rate limiting for the transcriptional response. A very striking result was obtained when RAD9 was overexpressed in rad24Δ cells, or when RAD24 was overexpressed in rad9Δ cells (Figure 4A and B). Elevated Rad9 levels in rad24Δ cells caused a 35‐fold transcriptional induction relative to unirradiated cells of RNR1 after UV irradiation (Figure 4A). Similarly, elevated Rad24 levels in rad9Δ cells caused this induction to be increased to 46‐fold (Figure 4B). Wild‐type cells typically show a 6‐fold induction of RNR1 after UV irradiation of G1‐arrested cells (Figure 4A–D). These data can be explained by negative regulatory interactions between RAD9 and RAD24. As the enhanced transcriptional response observed when RAD9 is overexpressed in rad24Δ cells is not apparent in rad17Δ and mec3Δ cells (Figure 4A), nor is it observed after RAD17 overexpression in any of the mutant backgrounds (Figure 4C), these interactions appear to be specific to RAD24 and RAD9. Finally, overexpression of RAD9 or RAD24 did not bypass the transcriptional induction defect seen in mec1‐1 and sad1‐1 mutant cells (Figure 4D), once more placing MEC1 and RAD53 function downstream of the RAD9 and RAD24 epistasis categories.
Modification and activation of the Rad53 protein after damage is dependent on RAD9, RAD24, RAD17 and MEC3 and requires the simultaneous presence of both Rad9 and Rad24
Rad53 is phosphorylated in a RAD9‐dependent manner under conditions in which G1 synchronized cells are UV irradiated and immediately released from the G1 block (Navas et al., 1996). Similar modification of Rad53 has also been shown to be MEC3 dependent after methylmethane sulfonate (MMS) treatment of cells blocked in G2 with a microtubule inhibitor (Sun et al., 1996). This DNA damage‐dependent modification of Rad53 is MEC1 dependent (Sanchez et al., 1996), but its significance is not well understood. Treatment of cells with hydroxyurea, an inhibitor of DNA synthesis, results in Rad53 phosphorylation accompanied by a modest increase in its kinase activity (Sun et al., 1996), and Rad53 kinase activity has been shown to be required for full checkpoint pathway function (Fay et al., 1997; Kim and Weinert, 1997).
We examined Rad53 modification after UV irradiation in wild‐type and checkpoint mutant cells continuously blocked in G1 or in G2. As phosphorylation of Rad53 is known to be cell‐cycle regulated close to the G1/S boundary (Sun et al., 1996), these experiments focus on damage‐dependent modification of Rad53 independent of cell‐cycle position. Essentially identical results were obtained for the G1 and G2 block experiments (Figure 5A). In both experiments, the DNA damage‐dependent modification of Rad53 is clearly evident in wild‐type cells 30 min after UV irradiation as more slowly migrating bands (lanes 2 and 8). However, in rad17Δ (lanes 4 and 10) and rad9Δ–rad24Δ double mutant cells (lanes 6 and 12), this modification was not detectable. In addition, Rad53 modification was not detected in rad9Δ, rad24Δ (Figure 5B) or mec3Δ cells (data not shown). To determine whether modification of Rad53 correlates with activation of Rad53 as a kinase, we performed histone H1 kinase assays with Rad53 immunoprecipitates (Figure 5D). Rad53 kinase activity was detected after UV irradiation in wild‐type cells (lanes 3), but not in rad24Δ (lane 5) or in rad9Δ (data not shown) cells, and correlated with phosphorylation of Rad53 (compare Figure 5A and B with D). This kinase activity is Rad53‐ and UV‐specific because it is not detectable in irradiated pre‐immune precipitates (lane 1) nor in unirradiated immunoprecipitates (lane 2). Furthermore, it is not detectable in an irradiated immunoprecipitate from a strain carrying an apparently ‘kinase‐dead’ allele of RAD53, rad53K277A (lane 4). Similar results were obtained by examining Rad53 autophosphorylation (data not shown). Thus, detectable damage‐dependent modification of Rad53 in G1‐ or G2‐blocked cells is dependent on intact RAD9 and RAD24 branches of the DNA damage checkpoint pathway, and this modification results in activation of Rad53 kinase.
In Figure 4, we present results that suggest specific antagonistic interactions between RAD9 and RAD24. The phosphorylation of Rad53 is another cellular response to DNA damage downstream of RAD9 and the RAD24 epistasis group. We examined this modification in UV‐irradiated cells that were held in G1 in the presence of overproduced Rad9 or Rad24. Overproduction had no effect on Rad53 modification in the absence of DNA damage (data not shown). In wild‐type cells, the modification of Rad53 first appears immediately after irradiation, seems to peak at 60 min, but is still detectable after 6 h (Figure 5B). When either Rad9 or Rad24 were overproduced in wild‐type cells, the phosphorylated form of Rad53 also appeared immediately after irradiation but then continued to accumulate to high levels for the duration of the experiment (Figure 5B). In rad9Δ or rad24Δ cells, no Rad53 modification is detectable (Figure 5B), and this phenotype can be rescued by overproduction of Rad9 or Rad24 respectively (data not shown). Rad9 overproduction can bypass the defect in Rad53 modification after damage found in mec3Δ and rad17Δ (Figure 5C, lanes 1–4). Similarly, Rad24 overproduction rescues this defect in rad17Δ (Figure 5C, lanes 7 and 8). However, Rad9 overproduction in rad24Δ cells or Rad24 overproduction in rad9Δ cells did not rescue the defect in Rad53 modification after UV irradiation (Figure 5B), although the G1/S checkpoint (data not shown), the transcriptional response (Figure 4), the G2/M checkpoint (Figure 3) and cellular viability (Figure 3) were restored under these conditions. These data suggest that under these conditions Rad9 and Rad24 are required simultaneously for Rad53 phosphorylation. The lack of Rad53 phosphorylation in rad24Δ cells overexpressing RAD9 correlates with a lack of kinase activity in these cells (Figure 5D, lane 6), indicating that Rad53 kinase activity is not required for functioning of the checkpoint pathway under these conditions. Interestingly, Rad24 overproduction did not rescue Rad53 modification in mec3Δ cells (Figure 5C). This final observation is consistent with our previous data demonstrating that Rad24 overproduction cannot restore the G2 checkpoint (Figure 3) or the transcriptional response (Figure 4) in mec3Δ cells, and supports the hypothesis that MEC3 functions downstream of RAD24 within this epistasis group.
RAD9 and RAD24 define two converging branches of the DNA damage checkpoint that function additively to produce most of the checkpoint signal
RAD9, RAD24, RAD17, MEC3 and DDC1 are thought to be involved in a signalling cascade that links damaged DNA to presumptive signal transducing kinases encoded by MEC1 and RAD53. These kinases then regulate, via poorly understood mechanisms, the downstream biological effects of the DNA damage‐dependent checkpoint pathway, namely cell‐cycle delay, transcription of the DDR and possibly DNA repair (for a recent review, see Elledge, 1996). RAD24, RAD17, MEC3 and DDC1 have been placed into an epistasis group distinct from RAD9 with respect to their UV or X‐ray sensitivity phenotypes (Eckardt Schupp et al., 1987; Lydall and Weinert, 1995; Longhese et al., 1997). Epistasis categories are usually indicative of distinct biochemical pathways; however, both RAD9 and the RAD24 epistasis group are required for efficient cell‐cycle arrest after DNA damage in G1/S (Siede et al., 1993) and G2/M (Weinert and Hartwell, 1988; Weinert et al., 1994; Yang et al., 1997; see also Figures 1 and 3). Both epistasis groups are also required after damage for efficient induction of the DDR (Figure 2), which includes RAD53 and phosphorylation of the Rad53 protein (Figures 2 and 5A). These data are best interpreted by the assumption that RAD9 and the RAD24 epistasis group define biochemically distinct branches of the DNA damage‐dependent checkpoint pathway. These branches probably function at, or close to, DNA damage and subsequently converge. The point of convergence is thought to be at or above MEC1, with RAD53 functioning downstream, as Rad53 overproduction suppresses the mec1‐1 allele and Rad53 phosphorylation after damage depends on MEC1 function (Sanchez et al., 1996). Our data place RAD9 and the RAD24 epistasis group upstream of MEC1, as overexpression of either RAD9 or RAD24 cannot rescue the G2/M checkpoint (Figure 3M and N), the UV sensitivity (Figure 3O) or the defect in the transcriptional response to DNA damage (Figure 4C) observed in mec1‐1 strains. Additionally, the transcriptional induction after UV treatment of RAD53 itself is severely affected in a mec1‐1 mutant (Figure 2).
That each upstream branch of the DNA damage‐dependent checkpoint pathway is partially redundant and quantitatively equivalent in terms of the signal transduced from damaged DNA is demonstrated by two observations. First, RAD9 overexpression can bypass mutations in the RAD24 epistasis group and members of the RAD24 epistasis group are capable of bypassing deficiencies in RAD9 for checkpoint, survival and transcription phenotypes (Figures 3 and 4). Secondly, single mutants in either branch have indistinguishable phenotypes with respect to the transcriptional induction of the DDR (Figure 2), checkpoints (both G1/S and G2/M, Figures 1 and 3) and survival after damage (Figure 3). It is particularly interesting that after UV damage a rad9Δ–rad24Δ double mutant has no detectable transcriptional induction of the DDR (Figure 2B), has no apparent G1/S checkpoint (Figure 1), only a residual G2/M checkpoint (Figure 2), and is more sensitive to DNA damage than either single mutant (Eckardt Schupp et al., 1987; Lydall and Weinert, 1995; our unpublished observations). Similarly, the rad9Δ–mec3Δ double mutant is also more defective in the G1/S and G2/M checkpoints and has an additively worse rad phenotype than either single mutant (data not shown). These data suggest that the two branches are responsible for the majority of the signal emanating from damaged DNA. The remaining G2/M checkpoint observed in these double mutants is probably due to S‐phase‐specific pathways (Navas et al., 1996).
Using overexpression analysis, we observed a hierarchy of suppression phenotypes within the RAD24 epistasis group (Figure 3). Overexpression of MEC3 is the most effective at rescuing the G2/M checkpoint, followed by RAD24 and then RAD17. Thus it is likely that Rad17 requires both Rad24 and Mec3 for function, suggesting that the RAD24 epistasis group of genes can be functionally ordered with RAD17 and RAD24 upstream of MEC3. This ordering within the RAD24 epistasis group is supported by the following observations. Overexpression of RAD17 is not very effective at rescuing the survival defect of rad24Δ and, particularly, mec3Δ cells. Similarly, it cannot rescue the deficiency in the transcriptional response to DNA damage of mec3Δ cells and appears to rescue this defect weakly in rad24Δ cells (Figure 4C). RAD24 overexpression cannot rescue the DNA damage‐dependent phosphorylation of Rad53 in mec3Δ cells and only weakly rescues this modification in rad17Δ cells (Figure 5C). Such functional ordering does not necessarily preclude a previously proposed model which puts the gene products of the RAD24 epistasis group into a putative protein complex (Lydall and Weinert, 1995), nor does it imply that there is a strictly linear relationship between the members of the RAD24 epistasis group. Indeed, we have observed slightly less penetrant checkpoint and survival phenotypes with rad9Δ–mec3Δ cells compared with rad9Δ–rad24Δ cells, consistent with a degree of non‐linearity of function within this group of genes. On the other hand, the more penetrant phenotypes observed with the rad9Δ–rad24Δ mutant, compared with the rad9Δ–mec3Δ mutant, may be due, at least in part, to loss of specific RAD9–RAD24 interactions (see below). The data discussed above support the model outlined in Figure 6. RAD9 and the RAD24 epistasis group are placed onto two separate branches of the signalling pathway, upstream of MEC1 and RAD53, with both normally contributing equally to the signal sent from damaged DNA. Resolution of the details of this pathway will require future biochemical investigation.
We have shown that cells harbouring a single null mutation in any of RAD9, RAD17, RAD24 and MEC3, or the previously isolated alleles of RAD53 and MEC1, sad1‐1 and mec1‐1 respectively, have a severely diminished transcriptional response to DNA damage. Previously, using cells partially deficient in nucleotide excision repair, Kiser and Weinert (1996) described RAD53‐ and MEC1‐dependent transcriptional induction of RNR3 after UV irradiation, but they observed only partial dependency upon RAD9 and the RAD24 epistasis group. However, an independent study indicated that RNR3 induction after DNA damage was highly dependent on RAD9, RAD24 and MEC3 (Navas et al., 1996). Furthermore, essentially identical residual G1/S and G2/M checkpoints are observed in all the single mutants we have tested (Figures 1 and 3, and data not shown). Our data indicate that any of the disruptions to the DNA damage‐dependent checkpoint pathway we have used result in a quantitatively similar defect in the transcriptional as well as the checkpoint responses to DNA damage. Interestingly, all the checkpoint genes tested in this study appear to control the damage‐dependent transcriptional induction of RAD53 itself, and also of DUN1 (Figure 2). Both encode related protein kinases with roles in the cellular response to DNA damage (Zhou and Elledge, 1993; Allen et al., 1994; Sun et al., 1996); RAD53 has roles in all known responses whereas DUN1 is thought to function primarily in regulating the transcriptional response to damage of the RNR1, 2 and 3 genes. [However, recent evidence suggests that DUN1 also has a significant role in checkpoint regulation (Pati et al., 1997).] Thus a degree of positive feedback regulation of the checkpoint pathway is possible. This possibility is further supported by the reproducible, albeit only 2‐fold, transcriptional induction after UV irradiation of RAD24, RAD17 and MEC3 (data not shown).
Evidence for specific genetic interactions between RAD9 and RAD24 required for phosphorylation and activation of Rad53
We made an unexpected observation while studying the effects of overexpression of RAD9 or RAD24 on the transcriptional response to DNA damage. A dramatically increased transcriptional response was observed when RAD9 was overexpressed in rad24Δ cells (Figure 4A). This result indicates that RAD9 can act negatively on RAD24. Lydall and Weinert (1995) have also obtained evidence that RAD9 acts negatively on RAD24. Any effects of RAD9 on RAD17 or MEC3 were not examined in their study. Their observation has led to the model that Rad9 functions as a negative regulator of a RAD24‐dependent exonuclease activity, although its role is unlikely to be limited to negative regulation as null mutants of RAD9 and RAD24 have similar phenotypes and RAD9 overexpression rescues rad24Δ phenotypes (Figure 3). Using the transcriptional response to DNA damage, it was also possible for us to obtain the reciprocal result, i.e. RAD24 acts negatively on RAD9 function. Removal of RAD9 from cells in which RAD24 was overexpressed caused a similarly dramatic increase in the transcriptional response in comparison with the response in wild‐type cells (Figure 4B). Interestingly, the antagonistic effect observed between RAD9 and RAD24 was specific for these two genes. Overexpression of RAD9 in rad17Δ and mec3Δ cells, or overexpression of RAD17 in rad9Δ cells, did not cause an enhanced transcriptional response (Figure 4A and C). The antagonism between RAD9 and RAD24 that we have observed by examination of the transcriptional response to DNA damage seems surprising as we do not see dramatic rescue of other phenotypes under these conditions (Figure 3). A possible explanation for these observations may be that the dramatic transcriptional rescue phenotypes are related to conditions in which the cell cycle is arrested. This hypothesis is supported by our observation that antagonism between RAD9 and RAD24 in the transcriptional response is difficult to detect in asynchronous, exponentially growing cells (M.‐A.de la Torre‐Ruiz, unpublished results). Significantly, the cdc13 single‐stranded DNA accumulation assay used by Lydall and Weinert (1995) also involves cell‐cycle arrest, this time in G2. Thus, in the absence of cell cycling, the antagonistic interactions specific to RAD9 and RAD24 may be detected more easily. Nevertheless, evidence that such antagonism can occur in cycling cells comes from the observation of an enhanced (50% greater) G2 checkpoint when RAD9 was overexpressed in rad24Δ cells (Figure 3B).
Support for the specific genetic interaction between RAD9 and RAD24 also comes from our studies of Rad53 modification after DNA damage. UV irradiation of wild‐type cells causes modification of Rad53 (Figure 5A), activation of its kinase activity (Figure 5D), checkpoint arrest in both G1/S and G/M (Figures 1 and 3) and transcriptional induction of the DDR (Figure 2). The UV‐dependent modification of Rad53 is not detectable in any single mutants arrested in G1. Furthermore, rad24Δ cells do not have detectable Rad53 kinase activity under these conditions (Figure 5D); nevertheless, single checkpoint mutants have residual G1 checkpoint activity and there is residual induction of the DDR. These residual activities are due either to undetectable but sufficient phosphorylation and activation of Rad53 kinase or to unknown RAD53‐independent mechanisms. In the rad9Δ–rad24Δ double mutant, modification of Rad53 and its kinase activity are also undetectable, but residual checkpoint and transcriptional induction are now abolished. This may be explained by complete loss of Rad53 activation or by the absence of RAD53‐independent mechanisms in the double deletion. Support for the latter of these possibilities comes from the observation that Rad53 kinase activity in single and double mutants is as low as in an apparently ‘kinase‐dead’ allele of RAD53, rad53K227A (Figure 5D).
The lack of detectable Rad53 phosphorylation after UV irradiation of the single mutants can be rescued in most cases by overexpression of RAD9 or RAD24. The principal exception to this occurred when Rad9 was overproduced in rad24Δ cells or Rad24 in rad9Δ cells. Under these conditions, neither Rad53 modification nor its kinase activity are detectable (Figure 5). This is despite full, or even greater than full, activity of the checkpoint pathway under these conditions, as measured by survival (Figure 3), the G1/S checkpoint (data not shown), the G2/M checkpoint (Figure 3) and the transcriptional response (Figure 4). Thus the downstream consequences of checkpoint pathway activation are not always dependent on wild‐type levels of activation of RAD53 function, suggesting that under certain conditions this function can be bypassed. This bypass may allow partial rescue of checkpoint function when one of the two upstream branches of the checkpoint pathway is perturbed (the single mutants). A more pronounced bypass is observed, again without Rad53 phosphorylation and activation, in the absence of antagonistic interactions between RAD9 and RAD24 together with overproduction of either Rad9 or Rad24. Interestingly, the regulation of a downstream effector of the G2 checkpoint, PDS1, has recently been shown to be RAD53 independent (Cohen‐Fix and Koshland, 1997). Furthermore, DUN1 is a partial functional homologue of RAD53 for checkpoint function and there is some evidence for an alternative checkpoint pathway in budding yeast that can be activated by expression of a human cDNA, CHES1 (Pati et al., 1997). Thus DUN1‐dependent mechanisms or even alternative pathways could conceivably be used for RAD53 bypass under some circumstances. However, the biochemical mechanism and biological relevance of the proposed bypass pathway(s) will require future investigation.
In summary, our data indicate that RAD9 and the RAD24 epistasis group can be placed onto two branches that converge on the same target. Both branches act in an additive manner, being responsible for most of the checkpoint signal in exponentially growing cells; moreover, they display partial functional redundancy. Additionally, there is evidence for specific antagonistic interaction between RAD9 and RAD24 after DNA damage. At least under certain conditions, interaction between RAD9 and RAD24 is also required for phosphorylation of Rad53 and activation of Rad53 kinase. Intriguingly, the downstream biological consequences of this pathway are not always correlated with this kinase activity.
Materials and methods
RAD9, RAD24, RAD17 and MEC3 deletions were generated in the W303‐1a background (Mata; ho; ade2‐1; can1‐100; his3‐11; leu2‐3,115; trp1‐12; ura3), using the direct gene replacement technique previously described (Baudin et al., 1993). The URA3 marker was used to delete RAD24, RAD17 and MEC3, whereas RAD9 was deleted with HIS3. We used the following specific oligos corresponding to the 5′‐ and 3′‐terminal coding sequences of the genes to be deleted (upper case) followed by 17–21mers designed to anneal the selected marker (either URA3 or HIS3) to be deleted (lower case). RAD9: sense oligo, 5′TCTTCAACATCAGGGCTATGTCAGGCCAGTTAGTTCAATGGAAAAGCTCTCatgcg gcatcagagcag‐3′; antisense oligo, 5′‐TAATTTCATCTAACCTCAGAAATAGTGTTGTATATATCATTGTCCGTAATcttacgcatctgtgcgg‐3′. RAD24: sense oligo, 5′‐ATGTACAAGAAGCTTTAGATGCCATGTTTTTACCTAACGCCAAGCATAGGatgcggcatcagagcag‐3′; antisense oligo, AGTTAGAGTATTTCCAGATCTGAATCT‐ GAAAGG GACTCACTGATAACTGcttacgcatctgtgcgg. RAD17: sense oligo, 5′‐GGTGTGGAAACAAAGTAGTTGAAGGATTTCAACTATGCGAATCAACAATGgattcggtaatctccgaac‐3′; antisense oligo, TCTGCGTTTTCTGCGATGCTGGATATTGACTTAAAAAAATATAGGAATATattgaagctctaatttgtgag. MEC3: sense oligo: 5′‐ATGAAATTAAAATTGATAGTAAATGGTTGTGAAGCACCTGATGATTATAAgattcggtaatctccgaac‐3′; antisense oligo, 3′‐TACAAGCCCTTCGCTCTTGCTATAATATATGATTTGTCCTCTTTCCCattgaagctctaatttgtgag.
PCR fragments were transformed into the strain W303‐1a. Transformants selected for URA+ or HIS+ prototrophy were checked by UV sensitivity, Southern blotting and diagnostic PCR. Cells were grown in YEPD or minimal medium and plasmids introduced as previously described (Aboussekhra et al., 1996). Synthetic α‐factor was used at a final concentration of 20 μg/ml in the W303 background. The sad1‐1 and mec1‐1 strains were kind gifts from Ted Weinert (Tucson, AZ) and Errol Friedberg (Dallas, TX). mec1‐1 was backcrossed a further three times into our W303 background, scoring for MMS sensitivity each time. The strain obtained reverts at high frequency and was scored for MMS sensitivity prior to each use. The sad1‐1 strain is in a W303‐related background but in all experiments it was also compared with the isogenic parental background, in addition to our W303‐1a. Additionally, a rad53‐1 strain behaved identically to sad1‐1 in the transcriptional response.
For cloning the RAD9, RAD24, MEC3 and RAD17 coding sequences under the GAL1 promoter, we PCR amplified the four open reading frames corresponding to each gene using Pfu thermostable DNA polymerase (Stratagene). The amplified fragments were then subcloned into either the BamHI (RAD9, RAD17 and MEC3) or the SpeI (RAD24) sites of YCpIF vectors (Foreman and Davis, 1994) as follows: pGAL‐RAD9 (in YCpIF16), sense oligo 5′‐CGCGGATCCATGTCAGGCCAGTTAGTTC‐3′ and antisense 5′‐CGCGGATCCTCTAACCTCAGAAATAGTGTTG‐3′; pGAL‐RAD24 (in YCpIF15), sense oligo 5′‐CGTCAAC‐ TAGTTGATAGTACGAATTTGAACAAACGGCCC‐3′ and antisense 5′‐CCAGTACTAGTAGTATTTCCAGATCTGAATCTGAAAGGGAC‐3′; pGAL‐MEC3 (in YCpIF10), sense oligo 5′‐CGTCAGGATCCGAAATTAAAATTGATAGTAAATGG‐3′ and antisense 5′‐CCAGTGGATCCTTACAAGCCCTTCGATCTTGC; pGAL‐RAD17 (in YCpIF15), sense oligo 5′‐CGTCAGGATCCGCGAATAAACAGTGAGCTAGCG‐3′ and antisense 5′‐CCAGTGGATCCGTAAAAAATATAGGAATATCCTTTGTTGG‐3′.
Overexpression of any of these checkpoint genes fully complemented the cell cycle and survival deficiencies after UV irradiation of their own mutant backgrounds, indicative of correct function. Sequence analyses did not reveal any introduced mutations. Overproduction of each encoded protein was confirmed by Western blotting with the 12CA5 monoclonal antibody directed against the haemagglutinin (HA) epitope included in the YCpIF plasmids.
Irradiation procedures, α‐factor and nocodazole blocks and survival curves
G1 blocks with α‐factor and UV irradiation were performed as previously described (Aboussekhra et al., 1996). For survival curves, cells were grown to log phase (1×107 cells/ml), diluted appropriately, plated either in YPD or minimal medium, allowed to dry briefly and then irradiated at the doses indicated. Cells arrested in G2 after UV were monitored by nuclear and cell morphology. Nuclei were visualized by staining with 4′,6‐diamino‐2‐phenylindole (DAPI, Sigma). Large budded cells corresponding to mother and daughter cells of essentially equivalent volume with the nucleus at the neck of the bud were scored as cells blocked in G2/M, prior to anaphase. This phenotype is indicative of the G2/M checkpoint. Fluorescence‐activated cell sorting (FACS) analysis was also used to confirm this G2/M block. An advantage of this method is that the only perturbation to the cultures required is the UV treatment, thereby allowing a more physiological examination of any resulting cell cycle arrest. Nocodazole (Sigma) was prepared in dimethylsulfoxide (DMSO) and used at a final concentration of 5 μg/ml.
Northern blot analysis
Total RNA preparations were performed either by the hot phenol extraction or using the RNeasy Mini Kit (Qiagen). Northern blots were prepared, hybridized and quantitated by phosphorimaging (Molecular Dynamics) all as previously described (Aboussekhra et al., 1996).
Rad53 antibody, Western blotting and kinase assay
A C‐terminal fragment of Rad53 (amino acids 468–835) was expressed in E.coli using the pET21b expression vector. This was affinity purified using a nickel‐NTA–agarose column (Qiagen #30410) and used to immunize rabbits. For Western analysis, total yeast protein extracts were run on SDS–PAGE gels with 6.5% acrylamide and 80:1 acrylamide:bis‐acrylamide. The rabbit polyclonal serum was used at a final dilution of 1:5000 in 0.5% fat‐free milk in phosphate‐buffered saline (PBS) containing 0.02% Tween‐20 for a primary incubation of 12 h. Horseradish peroxidase‐linked secondary antibody (Amersham #NA934) was used at 1:10 000 in PBS/0.02% Tween for a 1 h incubation. Chemiluminescent detection was performed using the ECL kit (Amersham #RPN2106). Immunoprecipitations were performed from 200 μg of pre‐cleared total yeast extract with a final dilution of polyclonal antibody of 1:100. The protein A beads were washed extensively including high salt and RIPA buffer washes. For the kinase assay, histone H1 (Sigma #H5505) and [γ‐32P]ATP were added in reaction buffer (25 mM MOPS pH 7.2, 5 mM EGTA, 15 mM MgCl2, 60 mM β‐glycerophosphate, 0.1 mM Na orthovanadate, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 2× protease inhibitor cocktail) and the reaction was incubated at 30°C for 30 min. Kinase assays were run on 15 or 10% SDS–PAGE gels, dried and exposed to X‐OMAT film for 1–4 h.
We are grateful to J.J.Blow, A.M.Carr, J.F.X.Diffley, J.Q.Svejstrup, R.D.Wood and the members of the CDC laboratory for critical reading of this manuscript, and to L.Serrano‐Endolz for help with statistical analyses. We thank T.A.Weinert, E.C.Friedberg, M.P.Longhese and M.Foiani for yeast strains. M.A.de la T.‐R. is a Marie Curie Fellow supported by EU HC and M Network, contract number CHRX CT94 0685.
- Copyright © 1998 European Molecular Biology Organization