Eukaryotic checkpoint genes regulate multiple cellular responses to DNA damage. In this report, we examine the roles of budding yeast genes involved in G2/M arrest and tolerance to UV exposure. A current model posits three gene classes: those encoding proteins acting on damaged DNA (e.g. RAD9 and RAD24), those transducing a signal (MEC1, RAD53 and DUN1) or those participating more directly in arrest (PDS1). Here, we define important features of the pathways subserved by those genes. MEC1, which we find is required for both establishment and maintenance of G2/M arrest, mediates this arrest through two parallel pathways. One pathway requires RAD53 and DUN1 (the ‘RAD53 pathway’); the other pathway requires PDS1. Each pathway independently contributes ∼50% to G2/M arrest, effects demonstrable after cdc13‐induced damage or a double‐stranded break inflicted by the HO endonuclease. Similarly, both pathways contribute independently to tolerance of UV irradiation. How the parallel pathways might interact ultimately to achieve arrest is not yet understood, but we do provide evidence that neither the RAD53 nor the PDS1 pathway appears to maintain arrest by inhibiting adaptation. Instead, we think it likely that both pathways contribute to establishing and maintaining arrest.
DNA damage or blocks to DNA replication cause all eukaryotic cells to delay cell cycle progression; these delays are mediated by controls called checkpoints (reviewed in Elledge, 1996; Page and Orr‐Weaver, 1997; Paulovich et al., 1997a; Longhese et al., 1998; Weinert, 1998). Following damage, normal cells arrest before mitosis while checkpoint‐deficient cells proceed through mitosis, consequently suffering genomic instability or cell death. That DNA damage‐induced cell cycle checkpoints maintain a stable genome has broad implications for human disease, particularly cancer. The mammalian p53 and ATM genes regulate checkpoint responses and, when mutated, cause genomic instability and predisposition to cancer (Hartwell and Kastan, 1994). Mutations in ATM also lead to other pathologies, although the relationships of these to DNA damage checkpoints are more elusive (Xu et al., 1996; Rotman and Shiloh, 1997).
Checkpoint controls responding to DNA damage have now been studied in a number of organisms, including budding and fission yeast, flies, filamentous fungi and mammalian cells (Hari et al., 1995; Elledge, 1996; Ye et al., 1997). Because both the types of responses and the genes regulating them are highly conserved, there is reason to believe that knowledge of checkpoint pathways in yeast may lead ultimately to an understanding of and perhaps treatments for cancer and other human diseases.
In mitotic cells, the conserved responses to DNA damage include at least four distinct cell cycle delays (Weinert, 1998): one in G1 (the G1/S checkpoint); two responses during S phase, inhibition of DNA replication (the S phase progression checkpoint) or of mitosis (the S/M checkpoint); and one in G2 (the G2/M checkpoint). Checkpoint controls also mediate a delay during meiotic recombination (Lydall et al., 1996). The genes required for cell cycle delays have additional cellular roles. These include regulation of transcriptional induction of repair genes (Huang et al., 1998; for reviews, see Bachant and Elledge, 1998; Weinert, 1998), an essential function which appears to involve control of dNTP synthesis (Zhao et al., 1998), and mediation of chromosome recombination during meiosis (Grushcow et al., 1999).
The genes regulating checkpoint responses are likewise conserved. As an example, a putative protein kinase encoded by the MEC1 gene in budding yeast regulates all checkpoint responses; MEC1 homologs subserve similar functions in other eukaryotic cells (e.g. rad3+ in fission yeast, ATM and ATR in mammalian cells, and mei‐41 in flies, as reviewed in Carr, 1997; Hoekstra, 1997). Conservation, however, does not extend to all details of molecular mechanisms. For example, the G1/S and G2/M arrests involve cyclin‐dependent kinase (CDK)–cyclins in both budding yeast and human cells, although the details of CDK–cyclin regulation appear to be quite different (reviewed in Nurse, 1997; Weinert, 1998).
This study focuses on how checkpoint proteins in budding yeast govern the G2/M arrest response after DNA damage. Previous studies forged the working model tested here. In the model, distinct sets of proteins recognize damage, transduce a signal or mediate arrest more directly. The model is based largely on genetic and physiological studies; the molecular details are largely unknown. Proteins that may act directly on damage include Rad9, Rad24, Rad17, Mec3 and Ddc1 (reviewed in Longhese et al., 1998; Weinert, 1998). Their direct involvement with DNA damage is inferred from several observations: mutations in those genes alter degradation of double‐stranded to single‐stranded DNA (Garvik et al., 1995; Lydall and Weinert, 1995), the Rad17 protein has sequence similarity to Rec1, a bona fide 3′–5′ exonuclease from Ustilago maydis (Thelen et al., 1994; Lydall and Weinert, 1995), and Rad24 has sequence similarity to replication factor C proteins (Griffiths et al., 1995; Lydall and Weinert, 1997). The signal transducers include the protein kinases Mec1, Rad53 and Dun1. A member of the so‐called phosphatidylinositol 3‐kinase family (for reviews, see Carr, 1997; Hoekstra, 1997), Mec1 regulates all known checkpoint responses (reviewed in Paulovich et al., 1997a; Weinert, 1998), whereas other checkpoint genes have more circumscribed functions. The conventional protein kinases, Rad53 and Dun1 (Zhou and Elledge, 1993; Allen et al., 1994), act in several but not all MEC1‐dependent pathways, an issue clarified herein. Finally, a possible target of the G2/M pathway is the Pds1 protein, which is required for both the spindle assembly and DNA damage checkpoints (Yamamoto et al., 1996a,b).
The order of function amongst the budding yeast checkpoint proteins (e.g. Rad9, Mec1, Rad53 and Pds1) remains inferential, derived from both genetic data and biochemical studies of protein phosphorylation. Rad53, for example, is inferred to act downstream of Mec1 because phosphorylation of Rad53 after damage requires an intact MEC1 gene (Sanchez, et al., 1996; Sun et al., 1996). With regard to transcriptional induction, Dun1 is hypothesized to act downstream of Rad53 because phosphorylation of Dun1 after damage requires an intact RAD53 gene (Zhou and Elledge, 1993; Allen et al., 1994). In a similar fashion, Pds1 phosphorylation requires an intact MEC1 gene. Therefore, Pds1 may also act downstream of Mec1 (Cohen‐Fix and Koshland, 1997). Based on studies of DNA degradation (Lydall and Weinert, 1995), Rad9 is inferred to act upstream of Mec1, yet results from studies of protein phosphorylation (Emili, 1998; Vialard et al., 1998) imply a downstream position. However, because phosphorylation may be the consequence of feedback, interpretations of gene function based on that criterion alone are necessarily ambiguous (Zhou et al., 1993). An example of possible feedback phosphorylation in the G2/M checkpoint pathway in yeast has already been reported (Paciotti et al., 1998).
The objective of this report is to examine the order of function and specific roles of genes acting in the G2/M (or mitotic) arrest pathway, using genetic epistasis coupled with physiological tests of arrest after several kinds of DNA damage. Our results reconfigure central features of earlier checkpoint pathway models; therefore, for clarity of presentation, we now introduce the model we presently favor (Figure 1). Based on the observations reported herein, we suggest that MEC1 mediates G2/M arrest after damage by activating two parallel pathways. One of the two pathways requires both the Rad53 and Dun1 protein kinases; the other pathway requires PDS1. Previously, Cohen‐Fix and Koshland (1997) suggested a similar model, based upon a different experimental approach, namely examining protein phosphorylation; that kind of approach can only imply the existence of functional independent pathways. Here, we directly test the physiological consequences of disrupting the two pathways. Our report and the earlier one of Cohen‐Fix and Koshland are thus complementary; their data provide a biochemical explanation for the physiological effects disclosed by genetically disrupting the two pathways. We show here that disruptions of either pathway lead to an arrest defect of ∼50%. Each pathway likewise contributes independently to survival after UV exposure. Finally, the hierarchy of RAD53 and DUN1 has been redefined: both equally participate in the G2/M arrest component subserved by that branch of the checkpoint pathway.
Assaying the roles of checkpoint genes in G2/M arrest
To determine the roles of genes in cell cycle arrest, we primarily employed a quantitative arrest assay based on the cdc13 mutation (arrest assays utilizing other kinds of DNA damage are described later). CDC13 encodes a gene product that binds to the ends of chromosomes. When defective, its failure to bind (Lin and Zakian, 1996; Nugent et al., 1996) leads to generation of single‐stranded DNA (ssDNA) near chromosome ends (Garvik et al., 1995), causing a robust, prolonged cell cycle arrest (Weinert and Hartwell, 1993; Lydall and Weinert, 1995). We analyzed (see Materials and methods) the kinetics of G2/M arrest in cdc13‐damaged cells proceeding synchronously through one cell cycle (Figure 2A), as described previously (Lydall and Weinert, 1995). Logarithmically growing cells synchronized in G1 are released at a temperature (usually 36°C) restrictive for cdc13; during the ensuing S phase, DNA damage is generated. Checkpoint‐proficient cells arrest at the G2/M stage as large‐budded cells with an undivided nucleus (Figure 2A). Checkpoint‐deficient cells, however, proceed past G2/M. To prevent entry into the next cell cycle, we included a second mutation, cdc15, which causes arrest in late mitosis as large‐budded cells with a divided nucleus (Figure 2A).
MEC1 and RAD9 are required for complete arrest
Using the cdc13‐based assay, we evaluated the roles in cell cycle arrest of certain genes proposed to detect DNA damage directly (e.g. RAD9) or to transduce a signal (e.g. MEC1). Mec+ cells displayed a robust damage‐induced response in which >80% of cells arrested 2 h after release from G1 and remained so for at least 4 h, the duration of a typical assay (Figure 2B). In contrast, rad9 cells did not arrest (Figure 2B), but instead progressed through mitosis (Figure 2C) at a rate comparable with that of CDC13+ (i.e. undamaged) cells, as we have reported previously (Lydall and Weinert, 1995); mec1‐1 mutants likewise were completely defective for arrest (Figure 2B). That the observed failure of rad9 and mec1 mutants to arrest reflected their progression through G2/M to the next stage of cell division—and not cell death or some other artifact of the assay—is evident from the contemporaneous scoring of post‐mitotic cells (Figure 2C). Again note that mec1 and rad9 mutant strains progressed just as rapidly through mitosis as cells without damage (Figure 2C).
The mec1‐1 strains used in this study (Table I) carry a second mutation, sml1, which suppresses the essential defect of mec1‐1 mutants. Using a temperature‐sensitive allele of mec1, we demonstrated that sml1 did not affect the checkpoint or transcriptional functions of MEC1 (unpublished observations), a conclusion also supported by genetic analyses of the sml1 mutation in mec1‐1 mutant strains (Paulovich et al., 1997b; Zhao et al., 1998).
RAD53 provides only a partial arrest
Next, we examined the role of RAD53. Surprisingly, after cdc13‐induced DNA damage, rad53‐11 mutant cells did in fact arrest for ∼2 h (Figure 3A), after which approximately half of the cells proceeded through mitosis (Figure 3B). The rad53‐11 allele is complex: the checkpoint response appears defective at all temperatures but its essential function is temperature sensitive (Weinert et al., 1994). To ascertain whether the hypomorphic nature of the rad53‐11 mutation might account for the partial delay, we generated a rad53Δ strain kept alive by a high copy plasmid encoding the RNR1 gene (L.Vallen and F.Cross, personal communication; Sanchez et al., 1996; Desany et al., 1998). The rad53Δ pRNR1 strain likewise exhibited a partial arrest defect (Figure 3C and D).
Is the partial delay in rad53Δ cells due to residual checkpoint signaling or, alternatively, might it reflect another aspect of complex mutant cell physiology (for example, poor growth, a delay in S phase or even a delay caused by the pRNR1 suppressor plasmid)? We reasoned that a DNA damage‐specific delay would require intact genes acting upstream of RAD53, for example RAD9 or RAD24 (both rad9Δ and rad24Δ strains are completely arrest defective; Lydall and Weinert, 1995). In a rad24Δ rad53Δ pRNR1 strain, the partial delay after damage previously observed was completely eliminated (Figure 3C and D). We therefore conclude that the partial delay seen in rad53Δ mutants is in fact the consequence of residual signaling by a checkpoint pathway.
We also considered whether RAD53's partial role in arrest might reflect a peculiarity of cdc13‐induced DNA damage (Garvik et al., 1995). To do so, we examined the arrest response after a second kind of DNA damage. Using the HO endonuclease (see Materials and methods), a single, irreparable double‐stranded break was generated. Both the rad9 and mec1 mutants were again completely arrest defective and, again, rad53Δ mutants showed a partial arrest defect (Figure 4). We conclude that the mechanism of arrest after DNA damage, be it from cdc13‐induced damage or a double‐stranded break, proceeds completely through RAD9, RAD24 and MEC1, but partially through RAD53—and an additional pathway.
DUN1 and RAD53 act in the same pathway for G2/M arrest
DUN1 encodes a protein kinase initially reported to act specifically in transcriptional induction of repair genes (Zhou and Elledge, 1993). However, when we re‐examined the role of DUN1 in arrest, we found that dun1 mutants also exhibited a partial defect quite similar to that of rad53 mutants, after either cdc13‐induced damage (Figure 5A) or an irreparable double‐stranded break (Figure 4). Recently, dun1 mutants independently were reported to be partially arrest defective after a third kind of DNA damage, UV irradiation (Pati et al., 1997).
Since DUN1 and RAD53 appear to have similar roles in providing part of the G2/M arrest and since both genes encode protein kinases with similar motifs beyond the kinase domains (see Discussion), do the two have overlapping roles? If so, a dun1 rad53 double mutant might be completely arrest defective, indicating that DUN1 and RAD53 lie in parallel pathways. We found that rad53‐11 dun1Δ double mutants were non‐viable; however, viability could be restored by transforming with a high copy plasmid containing RNR1, thus permitting the cell cycle experiments to be performed. Surprisingly, the rad53Δ dun1Δ pRNR1 strain exhibited a partial arrest defect (Figure 5B) quite similar to that of either single mutant (compare Figure 5B with Figures 5A and 3C). The partial delay in rad53 dun1 double mutants required an intact RAD24 gene (Figure 5B) and is therefore the consequence of checkpoint signaling. We conclude, with regard to G2/M arrest, that RAD53 and DUN1 act in a single pathway (hereafter termed the ‘RAD53 pathway’), that the RAD53 pathway elicits only a partial arrest and, consequently, that a second pathway independent of the two protein kinases most likely exists.
The role of PDS1 in a parallel pathway of arrest
PDS1 was identified initially as a gene required for arrest at both the spindle assembly and the DNA damage checkpoints. Yamamoto et al. (1996b) reported that pds1 mutants partially arrest after X irradiation (this conclusion was inferred from an analysis of arrest kinetics, as deduced from bud morphology, of pds1 mutants compared with rad9 mutants). The nature of the residual delay in pds1 mutants after damage was unknown. Unfortunately, quantifying arrest in pds1 mutants after cdc13‐induced DNA is complicated by an additional defect, encountered at 36°C, which completely blocks progression through mitosis independently of DNA damage (Yamamoto et al., 1996a). To circumvent this problem, we performed the assays at 30°C, a temperature permissive for this essential function of PDS1 yet restrictive for cdc13 (realizing, of course, that cell cycle arrest by the cdc15 mutation would not be as effective at 30°C as at 36°C, which did complicate the scoring at later time points). At 30°C, a pds1 CDC13+ strain (therefore without DNA damage) exhibits a modest mitotic delay (Figure 6A). The delay is comparable with that seen in a rad9 pds1 CDC13+ strain (Figure 6A) and is therefore neither DNA damage‐ nor checkpoint gene‐dependent.
After cdc13‐induced DNA damage, we observed that pds1 mutants experienced a substantially longer G2/M delay (Figure 6B) than that seen in pds1 CDC13+ cells (Figure 6A). The cdc13‐dependent, differential delay in pds1 mutants is due to checkpoint signaling because that portion of the delay does require an intact RAD9 gene. There is no additional delay in pds1 cdc13 rad9 cells (Figure 6B) compared with pds1 CDC13+ rad9 cells (Figure 6A). We conclude that PDS1 contributes to G2/M arrest and another pathway accounts for the residual checkpoint‐dependent delay in pds1 mutants.
G2/M arrest is eliminated in pds1 rad53 and in pds1 dun1 cells
Because the RAD53 and PDS1 pathways each account for only part of the arrest response, we determined if the two sets of genes act in the same or in parallel pathways. We found that the checkpoint‐dependent cell cycle arrest was completely eliminated in rad53‐11 pds1Δ (Figure 7A), rad53Δ pds1Δ pRNR1 (Figure 7B) and dun1Δ pds1Δ (Figure 7C) cells; data from the corresponding pds1Δ CDC13+ strains, which control for the non‐checkpoint‐dependent mitotic delay of pds1, are shown for comparison in Figure 7D–F.
An examination of all possible diploid combinations of pds1 and dun1 hetero‐ and homozygotes (Figure 8) lends further support to the parallel pathway model. The double homozygote exhibited less of an arrest than either single homozygote (Figure 8, bars 2–4). Interestingly, dun1 did not appear to be completely recessive (compare bars 5–7) although the imperfect synchrony afforded by starvation release (see Materials and methods) made it impossible to conclude this with statistical certainty.
Because the non‐checkpoint‐dependent mitotic delay of pds1 complicated the arrest assays in strains bearing that mutation, we sought additional confirmation of the order of gene function using constitutive overexpression of the non‐degradable allele PDS1mdb (Cohen‐Fix et al., 1996) or of RAD53 in mutants defective for other components of the checkpoint pathway. The G2/M arrest produced by overexpression of PDS1mdb was independent of RAD9, MEC1, RAD53 and DUN1; congruently, overexpression of RAD53 caused arrest independently of MEC1, DUN1 and PDS1 (data not shown).
Based on the three sets of genetic evidence described above, we conclude that the DNA damage‐induced G2/M checkpoint arrest, which is mediated entirely through MEC1, is subserved independently by the RAD53 and PDS1 pathways downstream of MEC1.
A possible role for two pathways: testing establishment versus maintenance of arrest
There is evidence in yeast for separate controls that establish or maintain arrest. One manifestation of that separation is the phenomenon of ‘adaptation’, discussed below, in which arrested cells eventually resume cell cycle progression even though the damage persists (Sandell and Zakian, 1993; Toczyski et al., 1997).
In our model (Figure 1), MEC1 is a central regulator for the arrest response after DNA damage. As such, MEC1 might have a role in maintaining arrest as well as initiating it; in this model, after damage, MEC1 would signal G2/M arrest continually via the RAD53 and/or PDS1 pathways. Alternatively, MEC1 might be required simply to initiate arrest; a pathway downstream of MEC1 would serve to maintain arrest but MEC1 itself would not be required. If the first alternative were true, then loss of MEC1 activity would abrogate arrest. The latter possibility predicts that loss of MEC1 activity would not terminate arrest.
A conditional mec1ts allele was employed to distinguish between the two models. A double‐stranded DNA break was effected with the HO endonuclease (the strain also carried a rad54 mutation to prevent repair of the double‐stranded break). Arrest after an HO‐induced double‐stranded break was profound and persistent in Mec+ cells (Sandell and Zakian, 1993) and in mec1ts cells at the permissive temperature of 23°C (Figure 9). To test if MEC1 were required to maintain arrest, the mec1ts strain was HO‐induced with galactose at the permissive temperature until at least 80% of the cells were arrested (usually ∼6 h), then shifted to the restrictive temperature of 36°C (Figure 9). After 2–4 h at 36°C, the mec1ts strain resumed mitosis, in contrast to Mec+ cells which remained arrested (Figure 9). The simplest explanation for this observation is that Mec1p is required for both the establishment and the maintenance of arrest. This conclusion is in contrast to the proposed role of the fission yeast MEC1 homolog, Rad3 (Martinho et al., 1998). Martinho et al. concluded that Rad3 is required to establish but not to maintain arrest after DNA damage. We are unable to resolve the apparent contradiction between these two observations; the dichotomy may be ascribable to differences in experimental conditions (e.g. we employed an irreparable break, whereas the damage used in their study i.e. γ or UV irradiation, was apparently repaired in <1 h) or to species differences.
One plausible explanation for the existence of the two parallel pathways downstream of MEC1, then, is that one pathway might serve to establish arrest while the other acts to maintain it. Toczyski et al. (1997) had identified a specific mutation in the yeast CDC5 gene which abrogates adaptation; when cdc5‐ad mutant cells suffer damage, they arrest persistently without adaptation. Using the cdc5‐ad mutation, we examined the possibility that one of the two pathways for G2/M arrest might establish arrest and the other might inhibit adaptation (at least during the first 10 h when CDC5+cells with damage remain arrested). If either the RAD53 or the PDS1 pathway maintains arrest by inhibiting adaptation, then that specific pathway would not be required for arrest in cdc5‐ad mutants. We observed, however, that pds1 cdc5‐ad, dun1 cdc5‐ad and rad53 cdc5‐ad mutants all had partial arrest phenotypes (Figure 10). The simplest explanation for these findings is that RAD53, DUN1 and PDS1 do not maintain arrest by inhibiting adaptation, at least via CDC5; rather, both pathways contribute to establishing arrest.
UV resistance mediated by checkpoint pathways
If the RAD53 and PDS1 pathways contribute independently to cell cycle arrest after damage, each should also contribute independently to survival after DNA damage. To test this prediction, we determined survival after UV irradiation in a series of mutants (Figure 11A and B, summarized in Figure 11C). A pds1 mutant was only very slightly UV sensitive (Figure 11A), consistent with its specific role in partial G2/M arrest; dun1 (Figure 11A) and rad53 (Figure 11A and B) single mutants exhibited greater sensitivity than did pds1 mutants, concordant with their dual roles in G2/M arrest and transcriptional induction. Disrupting both G2/M arrest pathways, i.e. pds1 dun1 or pds1 rad53 double mutants (Figure 11A–C), increased UV sensitivity compared with cells with a defect in only one pathway, consonant with each contributing independently to UV resistance.
We observed that rad53 mutants were more UV sensitive than dun1 mutants (Figure 11A), perhaps because of RAD53's additional roles in DNA replication. RAD53 is required for the S/M checkpoint and for another aspect of DNA replication, but DUN1 is apparently not. We also found that double mutant dun1 rad53 strains (Figure 11A and B) were more UV sensitive than either single mutant, again suggesting diverse roles for each of the two genes in addition to their common roles in the G2/M arrest and transcriptional induction pathways.
Finally, the UV sensitivities of mec1 mutants (in strains containing sml1, Figure 11A, or the high copy suppressor pRNR1, Figure 11B) were substantially greater than that of pds1 rad53 double mutants or even the pds1 dun1 rad53 triple mutant (Figure 11C). That observation implies that MEC1 has RAD53‐, DUN1‐ and PDS1‐independent roles in tolerance to DNA damage.
MEC1‐dependent G2/M checkpoint pathways
Based on experiments employing two kinds of DNA damage, cdc13‐induced telomere‐related damage and a single, double‐stranded break caused by the HO endonuclease, we find that MEC1 regulates G2/M arrest by acting through parallel pathways, one regulated by RAD53 and the other by PDS1 (Figures 3,4,5,6,7,8). The two‐pathway hypothesis is supported further by experiments involving overexpression of RAD53 and PDS1 (data not shown) and by UV survival curves which demonstrate that defects in both pathways render cells more UV sensitive than do mutations in just one pathway (Figure 11). Previously, Cohen‐Fix and Koshland (1997) independently proposed that RAD53 and PDS1 exist in independent pathways downstream of MEC1, based on phosphorylation data. In this report, we describe two physiological consequences—impaired G2/M arrest and diminished survival after UV exposure—of disrupting either or both pathways.
The conclusion that RAD53 and PDS1 act downstream of MEC1 does, however, carry an important caveat. We have not shown that the demonstrable cell cycle arrests are indeed due to activation of bona fide checkpoint pathways and cannot do so conclusively because the genes upon which RAD53 and PDS1 act have yet to be identified. Nonetheless, the simplest interpretation of the various genetic tests described herein is that RAD53 and PDS1 act downstream from MEC1, as well as the other genes tested (Figure 1).
How might two pathways regulate G2/M arrest in budding yeast?
We considered and provide evidence (Figure 10) arguing against one possible model for two pathways: one pathway might establish arrest, the second maintain it by inhibiting adaptation (the resumption of cell cycle progression in the face of persistent damage). We also considered a similar model in which MEC1 is required only for the initiation of arrest while both the RAD53 and PDS1 pathways are required only for maintenance of arrest in a manner independent of MEC1. However, our finding (Figure 9) that MEC1 must act continuously for G2/M arrest suggests that it signals continuously through at least one of the two parallel pathways. Two additional models are considered in more detail below.
Model 1: convergent pathways. An entirely plausible model is that the two pathways converge on a common regulator(s) coordinately to achieve optimal G2/M arrest. In fission yeast, G2/M arrest does appear to occur through convergent regulators that hold p34CDC2 in its inactive state by maintaining phosphorylation of its Tyr15 (reviewed in Nurse, 1997). Maintaining Tyr15 phosphorylation involves an active protein kinase and an inactive phosphatase. Do checkpoint pathways in budding yeast likewise converge on CDC28 (the CDC2 homolog) and, if so, do they regulate its phosphorylation? Although CDC28 seems to be involved in G2/M arrest (Li and Cai, 1997), biochemical and genetic evidence strongly argue that arrest does not occur through inhibitory tyrosine phosphorylation (Amon et al., 1992; Sorger and Murray, 1992). Another possible target, as yet untested, is cyclin localization which, in mammalian cells, is linked to G2/M arrest (Kao et al., 1997; Jin et al., 1998). Pds1 probably acts on Esp1, among other proteins, to regulate mitosis (Ciosk et al., 1998). Regardless, the targets of the RAD53 and PDS1 pathways, be they CDC28, ESP1 or other as yet unspecified genes, must first be established in order to test a convergent pathway model definitively.
Model 2: two‐stage model. Alternatively, we suggest that the RAD53 and PDS1 pathways might achieve arrest by individually inhibiting two sequential steps in mitosis. For example, RAD53 might inhibit a G2 to metaphase transition while PDS1 might block a metaphase to anaphase transition. This model is supported by several pieces of circumstantial evidence, which suggest that damaged yeast cells can, in fact, arrest in either an interphase‐like (G2) or a metaphase‐like (M) state. First, when yeast cells are DNA damaged during interphase, they may arrest in the subsequent G2 stage. That determination is based on two kinds of observations indicative of cells in interphase, namely that at least some Cdc28 is phosphorylated (Amon et al., 1992; Sorger and Murray, 1992) and some chromosomal condensation appears interphase‐like (Guacci et al., 1994). Secondly, DNA‐damaged yeast cells can also arrest in metaphase. Benomyl‐treated (metaphase) and then X‐irradiated yeast cells show a robust RAD9dependent arrest (Weinert and Hartwell, 1988). The benomyl‐arrested state appears to be metaphase because, relative to interphase cells, Cdc28 is dephosphorylated (Sorger and Murray, 1992) and chromosomes are condensed (Guacci et al., 1994). Additionally, Pds1 appears to be an inhibitor of the metaphase–anaphase transition, as argued from other observations (Yamamoto et al., 1996a,b). Finally, that DNA‐damaged yeast cells may arrest in metaphase is supported by the observation that cells may experience a RAD9‐dependent delay in mid‐anaphase (Yang et al., 1997).
We therefore speculate, as a rationale for two pathways, that normal, DNA‐damaged budding yeast cells may arrest in either a G2‐like or a metaphase‐like state and that each arrest state is achieved by one of the two pathways. How then might this two‐stage arrest model be reconciled with the observation that irradiated metaphase cells achieve a complete arrest (Weinert and Hartwell, 1988), which we have concluded requires both the RAD53 and the PDS1 pathways? A complete metaphase arrest becomes explicable if cells are presumed to be in equilibrium, i.e. if they are able to oscillate between the G2‐ and metaphase‐like states, and hence are periodically subject to regulation by either pathway. Mammalian cells with DNA damage do in fact exhibit retrograde movement from early prophase to interphase (Carlson, 1969a,b; discussed in Weinert and Lydall, 1993). Testing the two‐stage arrest hypothesis in budding yeast must await experiments designed to determine if damaged metaphase cells may return to a G2‐like state and whether that retrogression requires, for example, RAD53 but not PDS1.
The roles of RAD53 and DUN1
A second feature of the current model is that RAD53 and DUN1 are both essential for function of the ‘RAD53’ G2/M arrest pathway. Although earlier studies of these two protein kinases had indicated that both have roles in transcriptional responses to DNA damage (Zhou and Elledge, 1993; Allen et al., 1994), the role of DUN1 in arrest was appreciated only recently (Pati et al., 1997). Here, we extend previous observations by showing that, first, dun1 mutants fail to arrest after two additional types of damage and, secondly, that RAD53 and DUN1 act in a single G2/M arrest pathway. How the two protein kinases might function in a single pathway is unknown but could be in a manner analogous to MAP kinases cooperating to achieve optimal responses (Ferrell, 1996; Ferrell and Machleder, 1998). Alternatively, the two kinases may phosphorylate unique substrate targets, each being required for function of that arrest pathway. That RAD53 and DUN1 are both required for arrest certainly adds another level of complexity to studying this pathway; events determined to be RAD53‐independent need also to be tested for a possible role for DUN1. By way of example, Wolter et al. (1996) reported that the damage inducibility of SNM1 was independent of SAD1 (RAD53) but when they tested SNM1 inducibility in a dun1 strain, they in fact found dependency. This was surprising in light of earlier results that had placed DUN1 downstream of RAD53 in the transcription induction pathway (Zhou and Elledge, 1993).
The roles of specific genes in tolerance to DNA damage
The UV sensitivities of a series of single and multiple mutants provides further corroboration of the roles of two pathways in tolerance of DNA damage (Figure 11). For example, mutating both arrest pathways (e.g. a pds1 dun1 strain) resulted in greater UV sensitivity than disrupting either arrest pathway alone.
However, the UV sensitivity profiles also indicate that certain checkpoint proteins have multiple roles in tolerance of DNA damage. For example, RAD53 and DUN1 must do more in this regard than simply contributing to G2/M arrest since the rad53 dun1 double mutant is more UV sensitive than either single mutant. Indeed, the UV sensitivity of the rad53Δ dun1Δ double mutant was so profound (Figure 11B) in comparison with the respective single mutants that any additional effect of pds1Δ which was expected in the triple mutant strain was not observable (Figure 11B).
How the two protein kinases Rad53 and Dun1 mediate G2/M arrest and other responses is largely unknown. Both have FHA domains (fork head‐associated; Hofmann and Bucher, 1995), which are also found in the fission yeast RAD53 homolog, cds1+ (Murakami and Okayama, 1995; Lindsay et al., 1998). Mutations in the Rad53 FHA2 domain (Rad53 has two such domains) abolish G2/M arrest and disrupt its binding to Rad9 in vivo (Sun et al., 1998), yet render cells only mildly UV sensitive (Fay et al., 1997). Comparable studies of the FHA domain of Dun1 have not been performed. Much of Rad53′s contribution to UV survival must therefore derive from its role in responses other than arrest. Consequently, we think it likely that the overall contribution of G2/M arrest to UV resistance is less substantial than that of other effects mediated via the two parallel pathways.
The UV sensitivity profiles illustrate yet another feature of checkpoint gene function. Most MEC1‐dependent pathways described thus far also involve RAD53 (reviewed in Paulovich et al., 1997a; Weinert, 1998). However, MEC1 must promote tolerance to DNA damage via pathways independent of RAD53 (as well as DUN1 and PDS1), because mec1 pRNR1 single mutants are substantially more UV sensitive than even the rad53 dun1 pds1 triple mutant (Figure 7). One such pathway may relate to the transcriptional induction of the MEC1 and RAD53 genes, which requires MEC1 but not RAD53 (Kiser and Weinert, 1996). MEC1 may also regulate other, as yet unidentified, pathways.
Materials and methods
Strains, plasmids and media
The strains used (Table I) are isogenic with either W303 or A364a, as indicated. The 2 μ plasmids containing RNR1 were provided by L.Vallen and F.Cross (personal communication) or S.Elledge (Desany et al., 1998). A plasmid with the RAD53 gene under control of the GAL1 promoter was provided by D.Stern (Stern et al., 1991). The GAL‐PDS1mdb allele was provided by O.Cohen‐Fix and D.Koshland (Cohen‐Fix et al., 1996). The dun1Δ (Zhou and Elledge, 1993), rad9Δ, rad24Δ and rad53Δ alleles (Weinert and Hartwell, 1990; Lydall and Weinert, 1995, 1997) were integrated into yeast strains by transformation (Schiestl and Gietz, 1989) and/or genetic crosses. After transformation, mutants were selected by prototrophy, appropriate phenotypes confirmed, and genomic structures verified by Southern blot analysis. The pds1Δ allele was introduced by genetic crosses of strains provided by O.Cohen‐Fix and D.Koshland. The mec1‐1 sml1 and rad53‐11 mutants were generated in A364a (Weinert et al., 1994) and introduced into W303 by backcrosses (Lydall and Weinert, 1995). Identification and characterization of the sml1 mutation present in mec1‐1 strains is described elsewhere (Paulovich et al., 1997b; Gardner, 1998; Zhao et al., 1998). W303 strains containing cdc13‐1 and cdc15‐1 mutations were based on mutants used in a previous study (Lydall and Weinert, 1995).
Rich (YEPD), complete and minimal media (Sherman et al., 1986) were supplemented with adenine 20 μg/ml to promote optimal growth of W303 ade2 strains. Appropriate minimal media were used for strains containing plasmids. Alpha factor (Sigma), made up in a stock solution of 5×10−4 M, typically was used at a concentration of 1:25 000 in bar1 or 1:500 in BAR1+ strains. Carbon sources (dextrose, raffinose or galactose) were in final concentrations of 2% each.
G2/M arrest assays using cdc13‐induced damage
Quantitative assays of cell cycle arrest were performed as previously described (Lydall and Weinert, 1995). Cells were grown at 23°C (permissive temperature for cdc13 and cdc15) in rich media to mid‐log concentration. Cells were synchronized in G1 by treatment with alpha factor for 2.5–3 h at 23°C. The G1 (unbudded) cells were released from arrest by washing three times with medium, then shifted to 36°C (30°C for pds1Δ strains, see Results) in fresh rich medium. Aliquots of cells were taken at appropriate time points, fixed with 70% ethanol and stained with 4,6‐diamino‐2‐phenylindole (DAPI; Pringle et al., 1989). Cell and nuclear morphologies were scored in at least 100 cells from each sample, under light and fluorescence microscopy.
Arrest assays of diploid cells were performed in a similar fashion, with two modifications. First, synchrony (∼80% G1 cells) was accomplished by starvation (growth in YEPD for 48–60 h); the cells were released by resuspending them in fresh YEPD at 30°C. Secondly, the assays were performed in a microtiter‐type device. The nine diploid strains were assayed simultaneously in a single device.
The arrest assay was modified as follows for cells overexpressing RAD53 and PDS1mdb under control of the galactose promoter. Cells were pre‐grown in selective media containing raffinose to permit subsequent induction of the GAL promoter, then simultaneously treated with alpha factor and resuspended in YEP plus raffinose and galactose. Washing and subsequent resuspension of cells were also in YEP raffinose plus galactose. These cell cycle experiments were done at 30°C, except for RAD3 overexpression in the pds1Δ strain, which was tested at 23°C to minimize the mitotic delay of pds1Δ mutants.
Cell survival after UV exposure
Mid‐log cells grown in rich media were plated on solid media and irradiated with appropriate doses using a Stratalinker 1800. Despite efforts to prepare the Stratalinker uniformly (e.g. ‘pre‐warming’ for 3 min), we encountered day‐to‐day variability in the absolute, but not relative, levels of cell death for all strains. Since the relative sensitivities of our strains were highly reproducible, the data shown on each graph in Figure 11 are from experiments performed on the same day and irradiated at the same time for each dose. The results from a single, typical trial are shown; results from other trials are available upon request.
Cell cycle arrest after a double‐stranded break
The strains, isogenic with W303, were MATa GAL‐HO::URA3 rad54::LEU2 and contained a mutation in a checkpoint gene. Cells were grown in YEP raffinose at 30°C, synchronized in G1 with alpha factor for 2 h, then were incubated for an additional 2 h following addition of galactose (2%) to induce transcription of the HO endonuclease. The induced G1 cells were released from arrest by washing and resuspended in YEP raffinose plus galactose media; cell cycle progression was analyzed in ethanol‐fixed and DAPI‐stained cells. Each strain experienced ∼70% lethality from the irreparable HO cleavage event, indicating a similar rate of HO‐induced DNA breakage in all strains.
This work was supported by a grant (GM 45276) to T.W. from the National Institutes of Health, a Cancer Biology Training Fellowship (CA 09213) to R.G. and a National Institutes of Health, National Research Service Award, GM 19370, from the National Institute of General Medical Sciences to C.W.P.
- Copyright © 1999 European Molecular Biology Organization