DNA damage checkpoint maintenance through sustained Chk1 activity

The cellular response to DNA damage involves both the repair of lesions and the activation of signalling pathways, known as checkpoints, that delay cell cycle progression until the completion of repair (Hartwell and Weinert, 1989). DNA damage checkpoints share the common architecture of detectors, signal transducers and effectors (O'Connell et al., 2000). Conceptually, the checkpoint delay can be divided into three phases: initiation; maintenance during repair; and termination once repair is completed to allow cell cycle progression.

The G2 DNA damage checkpoint acts to prevent mitotic entry in the presence of DNA damage and is highly conserved from fission yeast to humans. Initiation of this checkpoint in mammalian cells involves activation of the ATM and ATR protein kinases. These serine/threonine kinases share sequence homology with the phosphoinositide-3 lipid kinases (PI-3Ks) (O'Connell et al., 2000) and are therefore termed PI-3K-related kinases (PIKKs). In fission yeast, the earliest known biochemical marker of checkpoint initiation is the activation of the ATR homologue Rad3 (Edwards et al., 1999). In addition, PIKK family members phosphorylate several checkpoint proteins, including the ATR-interacting protein ATRIP (Cortez et al., 2001), Rad26 in Schizosaccharomyces pombe and members of RFC- and PCNA-related complexes, which are required for transduction of the checkpoint signal (O'Connell et al., 2000).

PIKK family members ultimately inhibit entry into mitosis by activation of the effector kinase Chk1, an event that is dependent on all known upstream checkpoint proteins (Martinho et al., 1998; Walworth et al., 1993; Walworth and Bernards, 1996). This activation is achieved via the Rad3/ATR-dependent phosphorylation of key serine residues located in the C-terminal regulatory domain, specifically S345 in S. pombe and S317 and S345 in human cells (Capasso et al., 2002; Guo et al., 2000; Liu et al., 2000; Lopez-Girona et al., 2001; Zhao and Piwnica-Worms, 2001). In S. pombe, Chk1 activation also requires the BRCT-domain protein Crb2 (Esashi and Yanagida, 1999; Saka et al., 1997), which transiently binds to phosphorylated Chk1 during a checkpoint arrest (Mochida et al., 2004). In both humans and S. pombe, Chk1 phosphorylation is also required for its interaction with 14-3-3 proteins, which are essential for checkpoint function (Capasso et al., 2002; Chen et al., 1999; Jiang et al., 2003).

Chk1 elicits a G2 arrest through regulation of the Wee1 tyrosine kinases and Cdc25 tyrosine phosphatases. Together, these proteins control the activity of the cyclin-dependent kinase Cdc2 (Furnari et al., 1997; O'Connell et al., 1997; Raleigh and O'Connell, 2000), which is universally required for mitotic entry (Nurse, 1990). Deletion of chk1 results in checkpoint failure, whereas Chk1 overexpression elicits a sustained G2 arrest that is independent of its phosphorylation or the presence of upstream checkpoint proteins (Capasso etal., 2002; Ford et al., 1994; Lopez-Girona et al., 2001; O'Connell et al., 1997; Walworth et al., 1993).

A genetic analysis of S. pombe Rad3 has shown that this protein is required only for the initiation and not the maintenance of the G2 DNA damage checkpoint. If a temperature-sensitive (ts) allele of Rad3 is inactivated before irradiation, checkpoint initiation fails and cells prematurely enter mitosis, leading to the accumulation of mitotic abnormalities and a reduction in cell viability. However, if Rad3 is inactivated following irradiation, the checkpoint response is unaffected (Martinho et al., 1998). These experiments have established that the initiation and maintenance of the checkpoint are distinct events. Although the mechanism(s) by which the checkpoint is maintained is currently unclear, there are several plausible means by which this could occur. One possibility is that sustained Chk1 activity ensures maintenance of Cdc2 inhibition. Alternatively, Chk1 might be dispensable following the phosphorylation of Cdc25 and Wee1, and checkpoint maintenance might be imposed at this level of the cascade, via dephosphorylation, subcellular localization or interaction with regulatory proteins such as 14-3-3. To date, there is no evidence to support or refute any mechanism of checkpoint maintenance

We have developed an in vitro kinase activity assay for S. pombe Chk1 and shown that Chk1 is activated following irradiation in a manner dependent on Rad3-mediated phosphorylation of serine-345 (Capasso et al., 2002). Using this assay, we now show that Chk1 is rapidly activated following irradiation, and that this activity is maintained for the duration of a checkpoint arrest. Coincident with mitotic entry, there is a moderate but highly reproducible reduction of Chk1 activity and phosphorylation. These findings indicate that Chk1 function extends beyond initiation of the checkpoint. Furthermore, using ts alleles of chk1, we show that inactivation of Chk1 at any point after checkpoint initiation leads to lethal mitotic entry. This establishes the requirement for Chk1 for checkpoint initiation and maintenance in S. pombe. Given the highly conserved nature of the G2 DNA damage checkpoint pathway, we predict that human cells will be regulated in a similar fashion.

All strains are derivatives of wild-type S. pombe 972 h^– and 975 h⁺. Standard procedures and media were used for culture growth, transformation, microscopy and genetic analysis (Moreno et al., 1991).

Synchronous cultures were prepared by centrifugal elutriation and irradiated as described previously (O'Connell et al., 1997; Verkade et al., 1999) or by a cdc10-M17 block-release-block protocol (Verkade and O'Connell, 1998). Cell cycle progression was followed by septation, binucleate or mitotic indices, averaged from three counts of at least 100 4′,6-diamino-2-phenylindole (DAPI) stained cells. Samples for estimation of cell number were fixed in 3.7% formaldehyde and counted as described previously (O'Connell et al., 1997). For 150 Gy ionizing radiation, cells at a density of 4×10⁶ cells ml^–1 were irradiated using a Varian Linear Accelerator with a 6 MeV electron beam at a dose rate of 38 Gy minute^–1. The source was set at 80 cm above the flasks that were covered with a sheet of Perspex. Alternatively, cells were irradiated using a ¹³⁷Cs source at a dose rate of 0.7 Gy minute^–1.

For ultraviolet-C (UV-C) assays, cells were plated at various densities on YES agar [yeast extract plus supplements, described in (Moreno et al., 1991)] and irradiated in a Stratalinker (Stratagene). For ionizing radiation assays, cells were irradiated using the Linear Accelerator as described above. For methyl methanesulfonate (MMS) sensitivity assays, cultures were diluted to 4×10⁶ cells ml^–1 and 5 μl of tenfold dilutions were spotted onto YES agar containing the indicated concentration of MMS. For all survival assays, temperature was maintained during damage treatment, sample collection, plating and incubation. Colonies were counted and percentage survival was expressed as a proportion of unirradiated controls.

Cells were disrupted in immunoprecipitation (IP) buffer (10 mM NaPO₄ pH 7.0, 0.15 M NaCl, 1% NP-40, 10 mM EDTA, 50 mM NaF, 2 mM DTT, 1 mM PMSF, 3 mg ml^–1 N-p-Tosyl-L-phenylalanine chloromethyl ketone (TPCK), 10 mg ml^–1 E64, 100 mM benzamidine, 2 μg ml^–1 aprotinin, 2 μg ml^–1 leupeptin, 2 μg ml^–1 pepstatin) using acid-washed glass beads (Sigma) and a cell homogenizer. Five μg anti-haemagglutinin (anti-HA) 12CA5 monoclonal antibody (Roche) was added per milligram of protein in a total volume of 1 ml (with IP buffer). Samples were incubated at 4°C on a rotator for 1-2 hours, then 20 μl of protein-A/Sepharose (50% v/v slurry in IP buffer) was added and samples incubated for a further 30-60 minutes at 4°C on a rotator. Protein-A/Sepharose beads were washed five times with IP buffer and three times with 1× kinase buffer (50 mM Tris-HCl pH 7.0, 1 mM DTT, 5 mM MgCl₂, 0.4 mM MnCl₂, 25% glycerol (G5516; Sigma), 0.1% Triton X-100, 100 μM ATP). The washed Sepharose was resuspended in 24 μl kinase reaction buffer [1× kinase buffer, 12 μg peptide (RIARAASMAALARK), 10 μCi [γ-³²P]ATP (3000 Ci mmole^–1)] and incubated at 30°C for 15 minutes in an Eppendorf thermomixer. The Sepharose was pelleted and 20 μl of the supernatant was spotted onto P81 paper (Whatmann), which was then washed three to five times in ∼200 ml 0.5% orthophosphoric acid, rinsed with ethanol and dried. The dried p81 papers were then mixed with 5 ml Ready-safe scintillant (Beckman) and counted in a liquid scintillation counter. For synchronous cell populations, Chk1 activity was expressed as activity detected above background, where background was measured as activity in kinase assays using an isotype control IgG. For assays performed on asynchronous cell populations, Chk1 activity was expressed as activity detected above background, where background was measured as the activity detected for chk1 kinase dead cells. Each assessment of non-specific background produced essentially the same value.

For detection of Chk1-HA, extracts were prepared using acid-washed glass beads and cell homogenization. For asynchronous cells, extracts were prepared in 8 M urea, 100 mM Na₂HPO₄, 10 mM Tris-HCl, pH 8.0. For synchronous cells, extracts were prepared for immunoprecipitation and assay of Chk1 activity (as above) and aliquots retained and used for immunoblotting. A total of 20-50 μg protein was separated by sodium-dodecyl-sulfate polyacrylamide-gel electrophoresis (SDS-PAGE) on a 7% gel run at 200 V for 45 minutes and transferred to nitrocellulose in 25 mM Tris, 192 mM glycine, 20% methanol at 60 V for 45 minutes. The HA tag was detected with 12CA5 monoclonal antibody and a horseradish-peroxidase-coupled secondary antibody (Amersham) and proteins were detected using enhanced chemiluminescence (ECL) reagent (Roche).

In response to irradiation, Rad3-mediated phosphorylation of Chk1 at serine-345 leads to induction of Chk1 kinase activity (Capasso et al., 2002). Using a HA-tagged allele of chk1 (chk1:ep), the hyperphosphorylated form of the protein can be detected as a slower migrating species by SDS-PAGE (Walworth and Bernards, 1996). Rad3 is required only to initiate the G2 DNA damage checkpoint, demonstrating that checkpoint initiation and maintenance are separable processes (Martinho et al., 1998). Checkpoint maintenance could operate through Chk1 and/or further downstream, through its substrates Cdc25 and Wee1. Therefore, we asked whether Chk1 was required simply as a checkpoint initiator or for checkpoint initiation and maintenance.

If Chk1 were required for maintenance of the G2 DNA damage checkpoint, one would expect its kinase activity to remain elevated for the duration of a checkpoint-induced arrest. Previously, Chk1 phosphorylation has been used as a marker of checkpoint activation (Walworth and Bernards, 1996). However, it has not been shown conclusively that Chk1 phosphorylation strictly correlates with Chk1 activity. Here, we have used a recently developed Chk1 kinase assay (Capasso et al., 2002) to determine whether Chk1 is active for the duration of a checkpoint arrest and to assess the relationship between Chk1 phosphorylation and kinase activity throughout a checkpoint delay.

We prepared synchronous cultures of G2 cells by centrifugal elutriation, which were then irradiated with 150 J m^–2 UV-C or left untreated. In these experiments, in which a large population of cells was required for biochemical analysis, synchrony was somewhat compromised. Thus, the duration of the checkpoint arrest was approximately 60-75 minutes, which is slightly longer than previously described for smaller, more synchronous cultures (O'Connell et al., 1997; Raleigh and O'Connell, 2000) (Fig. 1A). Most of both unirradiated and irradiated populations completed mitosis, as evidenced by a doubling in cell number. No significant change in Chk1 activity was observed in unirradiated cultures for the duration of the experiment (data not shown). In the irradiated population, both the phosphorylation and induction of Chk1 kinase activity occurred rapidly and were maximal by ∼30 minutes after irradiation (Fig. 1A,B). As cells entered mitosis, Chk1 activity was reduced by ∼40% but it remained significantly higher than the basal level as the population passed into the next cell cycle. This was coupled with a reduction in Chk1 phosphorylation, although hyperphosphorylated Chk1 remained visible in all time points after irradiation and levels increased as cells passed into the next cell cycle. Owing to diminished synchrony in the larger cell populations used, it is likely that the magnitude of the changes in Chk1 activity and phosphorylation as cells pass mitosis have been underestimated. The induction of Chk1 activity and phosphorylation was not significantly increased with doses of UV-C radiation above 50 J m^–2 (Fig. 1C) (Walworth and Bernards, 1996), despite the substantially longer cell cycle delay observed (Al-Khodairy and Carr, 1992).

We conclude that Chk1 phosphorylation and activity are induced by irradiation and that this induction is saturated at relatively low experimental doses of UV-C. Furthermore, both Chk1 activity and Chk1 phosphorylation remain high for the duration of the checkpoint, suggesting a requirement for Chk1 function beyond checkpoint initiation.

To address directly the requirement for Chk1 function for G2 DNA damage checkpoint maintenance, we sought to inactivate Chk1 conditionally following the establishment of a checkpoint arrest. To this end, we performed two screens to isolate ts alleles of chk1. The first screen involved the site-specific mutagenesis of the non-catalytic domain of HA-tagged Chk1 (J. Ross and M. O'Connell, unpublished), followed by integration of the mutants into the endogenous chk1 locus. These integrants were then screened for hypersensitivity to the DNA-damaging drug MMS at 36°C, but not at 25°C. This screen resulted in the isolation of chk1-ts1 (E472D). The second screen involved the mutagenesis of pREP1::chk1 in XL-1Red Escherichia coli (Stratagene), which caused the accumulation of point mutations in this plasmid. The resultant plasmids were then transformed into chk1Δ cells and transformants were incubated at 25°C in the presence of thiamine to repress the nmt1 promoter and prevent Chk1 expression. Colonies were then replica plated to 25°C and 36°C in the absence of thiamine, to allow promoter derepression. Those colonies that failed to grow at 25°C in the absence of thiamine (owing to the induction of a cell cycle arrest by chk1 overexpression) but grew at 36°C (owing to the inactivation of the mutant chk1 at this temperature) were retained. The resultant ts alleles were sequenced and the mutations identified were reintroduced into the endogenous chk1 locus. From this screen, the alleles chk1-ts2 (E111R), chk1-ts3 (L151S), chk1-ts4 (I389S) and chk1-ts5 (ΔI484) were isolated.

To determine whether these mutants were suitable for experiments that examine the role of chk1 in checkpoint maintenance, we analysed the function of these chk1-ts alleles in response to DNA damage induced by ionizing radiation, UV-C radiation and MMS at both 25°C and 36°C. In all cases, chk1-ts1 cells were found to be hypersensitive to DNA damage at 36°C, and to be slightly more sensitive than wild-type cells at 25°C (Fig. 2A,B). Therefore, the chk1-ts1 allele essentially phenocopies chk1Δ at 36°C and wild-type chk1 at 25°C. The remaining four alleles (chk1-ts2 to chk1-ts5) behaved similarly, but only when expressed from the endogenous chk1 promoter carried on a multicopy plasmid (data not shown).

To determine whether the increased sensitivity of chk1-ts1 to DNA-damaging agents at 36°C was caused by checkpoint failure, the DNA-damage response of the chk1-ts1 cells was examined within a single cell cycle. Despite many attempts, we were unable to synchronize this strain by centrifugal elutriation, despite obtaining populations of cells of equal size. We do not know the reason for this effect, although backcrossing did not alleviate the problem. We were unable to use cdc25-22 block and release to generate synchrony, given the temperature-sensitive nature of chk1-ts1 and the possibility that this protocol might inappropriately induce synthetic checkpoint defects (Harvey et al., 2004). Therefore, asynchronously growing cultures were treated with 150 J m^–2 UV-C radiation to determine the checkpoint proficiency at 25°C and 36°C. Mitotic progression in these experiments was measured by septation (that is, the presence of a medial septum, which appears transiently ∼20 minutes after mitosis). The mitotic delay induced by a checkpoint arrest is accompanied by a transient loss of septated cells.

At 25°C, there was a clear cell cycle delay and decrease in septation in both wild-type (chk1:ep) and chk1-ts1 cells ∼45 minutes after irradiation (Fig. 3A). The arrest was shorter in chk1-ts1 than in the wild type, suggesting a weak checkpoint defect. In both strains, septation returned to basal levels from ∼90 minutes onwards, indicating the resumption of cell cycle progression. There was no decrease in septation observed in chk1Δ cells, which fail to arrest in response to DNA damage. At 36°C, wild-type cells arrested normally and remained arrested for ∼60 minutes before re-entering the cell cycle (Fig. 3A), whereas chk1Δ cells and cells carrying the chk1-ts1 allele failed to arrest in response to irradiation. When wild-type S. pombe cells arrest in response to DNA damage, cells continue to grow but fail to divide, resulting in cell elongation. Therefore, we examined the length of chk1-ts1 cells in response to UV-C radiation at 25°C and 36°C. In contrast to cells treated at 25°C, UV-C irradiated chk1-ts1 cells at 36°C were not elongated and binucleate cells were evident (Fig. 3B). This is consistent with a lack of checkpoint function in chk1-ts1 cells at 36°C. In addition, there were mitotic abnormalities observed in UV-C-irradiated chk1-ts1 cells at 36°C, indicative of checkpoint failure. We conclude that chk1-ts1 is conditionally checkpoint defective at 36°C.

Considering the position of the chk1-ts1 mutation in the non-catalytic domain (E472D), we wished to determine whether Chk1 activity was altered in this strain. Therefore, we performed Chk1 kinase assays on wild-type (chk1:ep) and chk1-ts1 cells irradiated with 150 J m^–2 UV-C radiation or 150 Gy ionizing radiation. Chk1-ts1 was found to be induced following irradiation at 25°C, although much less than wild-type Chk1 (Fig. 4A), consistent with its modest checkpoint defect at this temperature. At 36°C, although there was a reduced induction of in vitro kinase activity detected in irradiated wild-type cells compared with that at 25°C, activity above basal levels was readily detectable. By contrast, there was no detectable Chk1 activity in irradiated chk1-ts1 cells grown and irradiated at 36°C. It is notable that the higher temperature reduced both viability and Chk1 activation of both strains following UV-C irradiation, although to different extents (Fig. 2A, Fig. 4A). Because 36°C is the highest tolerable temperature for S. pombe, it is probable that there are multiple causes for the observed reduction in the viability and it cannot be attributed solely to decreased Chk1 kinase activity, especially given that these viability effects extend to chk1Δ cells.

To determine whether Chk1-ts1 was phosphorylated similarly to wild-type Chk1 in response to DNA damage, we also analysed Chk1 phosphorylation by western analysis in these experiments. For wild-type cells, the hyperphosphorylated form of Chk1 was detected in irradiated samples at both 25°C and 36°C (Fig. 4B). However, we were unable to detect any phosphorylated species of the Chk1-ts1 mutant protein following irradiation at 36°C, although there was a very slight mobility shift of Chk1-ts1 at 25°C after ionizing radiation.

If Chk1 were required for checkpoint maintenance then one would expect that inactivation of Chk1 after the initiation of the G2 DNA damage checkpoint would result in premature mitotic entry. Therefore, we determined the effect of inactivation of Chk1 during a checkpoint that had been initiated at 25°C using the chk1-ts mutants (Fig. 5A). First, we used MMS or ionizing radiation to inflict DNA damage in these experiments. We pre-established a robust checkpoint arrest at 25°C with either a 16 hour exposure to 0.0075% MMS or treatment with a cumulative 250 Gy dose of ionizing radiation delivered over 6 hours at 0.7 Gy minute^–1. During these treatments at 25°C, in which there is a 4 hour cell cycle, checkpoint induction in wild-type (chk1:ep) cells led to significant cellular elongation owing to continued growth during the imposed cell cycle delay (Fig. 5B). A similar elongation was observed for chk-ts1, but not for chk1Δ, because these cells continued into mitosis. The cultures of checkpoint-arrested cells were then shifted to 36°C and the proportion of cells with aberrant mitoses was determined every 15 minutes for 1 hour. In treated samples, t₀ indicates the effect of the DNA damage induced at 25°C. No aberrant mitoses were observed in wild-type cells over the time course for ionizing-radiation-treated cells, although a background of ∼6% aberrant mitoses were seen in MMS-treated cells (Fig. 5A). The chk1Δ cells treated with MMS or ionizing radiation showed many aberrant mitotic figures at t₀ owing to the absence of a checkpoint, and this increased only slightly over the hour at 36°C. However, in chk1-ts1 cells, there was a rapid accumulation of aberrant mitoses for both MMS- and ionizing-radiation-treated cells following Chk1-ts1 inactivation at 36°C. The alleles chk1-ts2 to chk1-ts5 behaved in a similar manner when expressed from a multicopy plasmid (data not shown). We conclude that inactivation of Chk1 leads to a loss of checkpoint maintenance.

To corroborate these findings, we undertook a second series of experiments. In this case, wild-type (chk1:ep), chk1-ts1 and chk1Δ cells were mock irradiated or exposed to 150 Gy ionizing radiation, delivered rapidly at 38 Gy minute^–1 at 25°C (total 3.95 minutes). The cultures were left at 25°C or shifted to 36°C at 15 minutes (t₁₅) after irradiation. As a measure of cells progression into mitosis, the number of binucleate cells and the number of aberrant mitoses were monitored for each strain and untreated controls.

In untreated cells incubated at 25°C, there was no evidence of a decrease in the number of binucleate cells for any strain (Fig. 6A). However, the shift to 36°C led to a decrease in the proportion of binucleate cells in all strains regardless of radiation dose. This decrease can be attributed to a heat shock, which causes a cell cycle delay in S. pombe (Polanshek, 1977). Therefore, the response to the temperature shift in irradiated populations included a component of heat-shock delay. It should be realized that such a heat-shock effect is not evident in cells that have been previously arrested in the cell cycle (Verkade and O'Connell, 1998) (Fig. 5).

In cells irradiated at 25°C, a cell cycle delay resulting from initiation of the G2 DNA-damage checkpoint was observed for both wild-type (chk1:ep) and chk1-ts1 cells, albeit that chk1-ts1 cells had a shortened delay compared to the wild type (Fig. 6B). Consistent with the presence of a functional checkpoint in wild-type and chk1-ts1 cells at 25°C, there were few aberrant mitoses (<2%) in these strains (Fig. 6C). In chk1Δ cells, no delay was observed and there was a considerable number of aberrant mitoses in irradiated cultures at 25°C. Therefore, as expected from the previous experiments, the chk1-ts1 allele is also checkpoint proficient at 25°C by this protocol.

In chk1-ts1 cultures shifted to 36°C following irradiation, the proportion of binucleate cells decreased in a manner similar to that of wild-type cells (Fig. 6B). However, chk1-ts1 cells consistently re-entered the cell cycle earlier than wild-type cells at this temperature, indicating that this lack of Chk1 function caused cells to enter mitosis prematurely, following a chk1-independent heat-shock delay. This was accompanied by an increased number of aberrant mitoses in chk-ts1 cells, to a level similar to that observed for chk1Δ cells, accounting for approximately 50% of mitoses observed (Fig. 6C). In wild-type cells, the number of aberrant mitotic figures remained low at 36°C (Fig. 6C). In chk1Δ cells, the number of aberrant mitoses following ionizing radiation was reduced in cultures shifted to 36°C compared with those incubated at 25°C. This was due to the cell cycle delay imposed by the heat shock in these cultures, which, like a G2 delay enforced by inactivation of Cdc25, rescues the mitotic defects of irradiated chk1Δ cells (Walworth et al., 1993). Therefore, chk1-ts1 initiates a checkpoint response at 25°C, allowing time for DNA repair and resulting in a low number of aberrant mitoses (Fig. 6C). When shifted to 36°C, chk1-ts1 is rapidly inactivated, causing premature resumption of the cell cycle.

In a third series of experiments, a cdc10-M17 block-release-block protocol was used to synchronize cells in G2 (Verkade and O'Connell, 1998). Cdc10 is a transcription factor required for passage of the G1-S transition (Simanis and Nurse, 1989) and thus cdc10-M17 cells arrest in G1 at the non-permissive temperature of 36°C. However, these cells need only to be shifted down to 25°C for approximately 30 minutes for Cdc10-M17 to allow passage of G1 and S phase into G2 (Verkade and O'Connell, 1998; Verkade et al., 2001). Owing to time and temperature restraints for synchrony, it was only feasible to use UV-C irradiation in these experiments. Under this protocol, chk1-ts1 cells had a clear UV-C-induced checkpoint delay at 25°C, as observed previously, which was shorter than that of wild-type cells (Fig. 7). In asynchronously growing cells, the survival of this strain under UV-C irradiation mirrors that of wild-type cells (Fig. 4), suggesting that this shorter delay is sufficient to maintain viability or that the synchronization protocol is affecting the duration of the delay. When cells were shifted to 36°C 20 minutes after irradiation, there was no evidence of a checkpoint arrest for chk1-ts1 and there was a concomitant increase in aberrant mitoses (data not shown). An arrest was retained in wild-type controls under the same conditions. Under both conditions, either no delay or a delay of one time point was observed in chk1Δ cells (Fig. 7), which, in synchronous cultures, exhibit a short but reproducible delay only to UV-C irradiation (Al-Khodairy et al., 1994; Capasso et al., 2002).

Together, these experiments demonstrate a role for chk1 not only in checkpoint initiation but also in checkpoint maintenance. Furthermore, they demonstrate that failure to maintain the G2 DNA-damage checkpoint following initiation of a checkpoint response and temporary cell cycle delay leads to a loss of mitotic fidelity.

A physiological response to DNA damage detected in G2 phase involves the initiation of a checkpoint arrest and the maintenance of this cell cycle delay until the completion of DNA repair. Failure to do so results in mitotic entry with unrepaired or partially repaired DNA, leading to defects in chromosome segregation and a consequent reduction in cell viability. In recent years, a sophisticated model has emerged of how the many checkpoint proteins co-operate to ensure the activation of Chk1 (O'Connell et al., 2000), the effector kinase for the G2 DNA-damage checkpoint and thus initiation of the checkpoint signalling cascade. However, what role, if any, Chk1 might play in maintenance of the checkpoint has not been investigated.

We have shown here that Chk1 activity is dynamically regulated following DNA damage and that this activity is required for the duration of a G2 checkpoint. In synchronous cultures, Chk1 kinase activity and phosphorylation were rapidly induced following DNA damage, and remained high for the duration of a checkpoint arrest. Chk1 activity was reduced as cells recover from the checkpoint arrest, although to levels significantly higher than basal, and there was a concomitant reduction in Chk1 phosphorylation as cells re-enter the cell cycle. We further demonstrated that the inactivation of Chk1 activity during an already-established DNA damage checkpoint leads to checkpoint failure, resulting in premature mitotic entry and mitotic abnormalities. Chk1 is therefore required for not only the initiation of the G2 DNA-damage checkpoint but also checkpoint maintenance, during which we propose that Chk1 is responsive to the detection of ongoing DNA repair.

It is unclear how Chk1 activity is maintained to ensure that cells remain checkpoint arrested until the completion of DNA repair. Chk1 activation requires the phosphorylation of its C-terminal domain by Rad3 (Capasso et al., 2002; Lopez-Girona et al., 2001; Martinho et al., 1998). However, Rad3 is only required for checkpoint initiation and its inactivation following DNA damage does not affect the ability of a cell to maintain a checkpoint arrest (Martinho et al., 1998). Thus, it must be possible to maintain Chk1 activity independent of Rad3 function. Perhaps the phosphorylated epitope of Chk1 is protected or Chk1 dephosphorylation is prohibited for the duration of a checkpoint arrest. Alternatively, another kinase might phosphorylate Chk1. Three proteins have been reported to interact with Chk1. The 14-3-3 proteins Rad24 and Rad25 interact specifically with the S345 phosphorylated form of Chk1 (Capasso et al., 2002; Chen et al., 1999). The BRCT-protein Crb2 has been found to interact with Chk1 in the yeast two-hybrid system, is required for Chk1 phosphorylation and associates with phosphorylated Chk1 (Esashi et al., 2000; Mochida et al., 2004; Saka et al., 1997). Therefore, maintaining a phosphorylation-dependent interaction between Chk1 and either 14-3-3 or Crb2 might contribute to checkpoint maintenance. In vivo, this might include the interaction of the C-terminal regulatory domain of Chk1 with these or other proteins. In this regard, it is notable that three of the five chk1-ts alleles we identified were mutations in this domain.

Having demonstrated a role for Chk1 in checkpoint maintenance, it follows that Chk1 is an obvious target for signalling pathways involved in checkpoint release and resumption of the cell cycle. We show here that a decrease in Chk1 phosphorylation and activity occurs coincident with cell cycle re-entry, yet both remain above basal. This might suggest that only a proportion of Chk1 within the cell, for example that which is in active complexes, need be inactivated for cell cycle resumption. Alternatively, Chk1 regulation might be more complicated than simple phosphorylation events.

The chk1-ts1 mutant we have isolated has substantially reduced kinase activity compared with the wild type at 25°C, yet is checkpoint proficient at this temperature. This implies that more Chk1 is activated in the cell than is required for a checkpoint arrest. Moreover, the observed SDS-PAGE mobility shift of Chk1-ts1 in response to DNA damage is minimal compared with that of the wild type. This might indicate that activation of the Chk1-ts1 allele requires a reduced threshold of Rad3-dependent phosphorylation. However, a number of mutations in Chk1 can affect the magnitude of this mobility shift (Capasso et al., 2002) and so these data must be interpreted carefully. We also find that maximal Chk1-ts1 activity is significantly lower than that at which wild-type cells re-enter the cell cycle. This suggests that Chk1 regulation in vivo is more complicated than currently thought. However, by necessity, we have used in vitro assays of Chk1 activity and this might only partially reflect the activity in vivo. Further to this, it has been shown that S345 phosphorylation is not required for the activity of recombinant Chk1 (O'Connell et al., 1997), nor for the G2 arrest elicited by Chk1 overexpression (Lopez-Girona et al., 2001) or the basal activity observed in unirradiated cells (Capasso et al., 2002). Chk1 regulation might involve mechanisms other than S345 phosphorylation, such as localization and interaction with substrates or regulatory proteins.

We have shown here that regulation of Chk1 activity is important for checkpoint maintenance, but is it the duration of Chk1 activity or the level of Chk1 activity that is crucial in the checkpoint signalling cascade? We show that the level of Chk1 activation is saturated at relatively low doses of UV-C and did not thereafter increase with increasing doses of UV-C radiation. Therefore, rather than the absolute level of Chk1 activity determining the duration of a checkpoint delay, it appears that Chk1 activity must reach a critical threshold for checkpoint maintenance. This, together with our finding that Chk1 activity is transiently reduced upon mitotic entry, indicates that the duration of Chk1 activity is the crucial determinant of checkpoint duration.

In addition to finding a role for Chk1 activity in checkpoint maintenance, we were surprised that Chk1 phosphorylation and activity increases following a brief decrease observed upon checkpoint release. This might indicate that a low level of unrepaired lesions persists in the cell. The signal generated from these lesions must be insufficient to block mitosis but might be more potent in the subsequent G1 and S phases, when cells are particularly sensitive to DNA damage (Al-Khodairy et al., 1994; Verkade et al., 2001). Consistent with this notion, we have only observed the transient loss of hyperphosphorylated Chk1 in synchronous cultures as they pass through mitosis, not in asynchronous cultures. In irradiated asynchronous cultures at any time after irradiation, there are a significant number of interphase cells with activated Chk1. This would mask the modest reduction in Chk1 phosphorylation and activity that we observe at mitotic entry in synchronous cultures and might explain why this change in Chk1 activity and phosphorylation has not been reported previously.

Our finding that Chk1 activity is required throughout the DNA-damage checkpoint has important consequences for anticancer therapeutics. Most tumour cells have lost G1 checkpoints and resort only to a G2 arrest in response to genotoxic stress. Abolishing their G2 checkpoint might selectively sensitize such tumour cells to treatment, because normal cycling cells will continue to use their G1 checkpoint (Koniaras et al., 2001). There is emerging evidence that Chk1 inhibition is a selective way to fulfil this objective (Jackson et al., 2000; Zachos et al., 2003; Zhao et al., 2002). Our finding that Chk1 is required for both checkpoint initiation and maintenance provides a broad window during which the specific inactivation of Chk1 by inhibitory drugs might be of benefit as an anticancer regime.

We thank J. Hagekyriakou for his assistance in experiments involving ionizing radiation, and U. John and R. Pearson for advice on Chk1 kinase assays. We are also grateful to S. Harvey for critical reading of this manuscript and M. Krien for his assistance in microscopy, advice and critical reading of the manuscript. C.L. is a recipient of an Australian Postgraduate Award, N.d.E. is supported by an NHMRC Peter Doherty Fellowship, and M.O'C. is a Scholar of the Leukemia and Lymphoma Society. This work was supported by the NHMRC (114229) and the NIH/NCI (CA100076-01).