CDC6 interaction with ATR regulates activation of a replication checkpoint in higher eukaryotic cells

In eukaryotic cells, chromosomal DNA replication is tightly regulated during the cell cycle and closely coordinated with other cell-cycle events, such as mitosis and DNA-damage response. The periodic assembly and disassembly of pre-replication complexes (pre-RCs) at replication origins is a central mechanism that ensures that there is only one DNA replication per single cell cycle (Bell and Dutta, 2002; Diffley, 2004; Fujita, 2006). During the period of low cyclin-dependent kinase (Cdk) from late mitosis through G1 phase, the origin recognition complex (ORC), CDC6 and Cdt1 load a presumptive replicative helicase, the MCM2-7 complex, onto chromatin to assemble the pre-RC. During the S, G2 and M phases, inappropriate reassembly of the pre-RC and subsequent re-replication are strictly prohibited by redundant suppression mechanisms against multiple MCM loaders. In human cells, Cdt1 activity is very tightly restricted by multiple pathways (Fujita, 2006). Human ORC1 is degraded through ubiquitylation by SCF^Skp2 (Méndez et al., 2002). CDC6 is a target of Cdk and, although this leads to degradation in yeast (Drury et al., 1997; Jallepalli et al., 1997; Jallepalli et al., 1998; Kominami and Tada, 1997), the outcome in human cells is nuclear export (Fujita et al., 1999; Jiang et al., 1999; Petersen et al., 1999; Saha et al., 1998). Whereas phosphorylation of human CDC6 by Cdk negatively regulates its function in pre-RC formation (Sugimoto et al., 2009), it is noteworthy that a significant proportion of CDC6 proteins remains in the nuclei during S phase (Fujita et al., 1999), suggesting a potential CDC6 function(s) other than pre-RC assembly.

When chromosomal DNA is damaged, for example by DNA double-strand breaks (DSBs) and replication stress, cells activate DNA-damage response cascades to halt the cell cycle, repair the damage and, if necessary, induce apoptosis (Bartek and Lukas, 2003; Burrows and Elledge, 2008; Kastan and Bartek, 2004; Nyberg et al., 2002). Ataxia telangiectasia mutated (ATM) and ATM- and Rad3-related (ATR) kinases are phosphatidylinositol-3-kinase-related kinases and are major upstream kinases that signal several downstream pathways by phosphorylating many effector molecules, including the Chk1 and Chk2 kinases. The activated Chk1 and Chk2 eventually delay cell-cycle progression to facilitate DNA repair, mainly through inhibiting CDC25 phosphatases required for Cdk activation (Bartek and Lukas, 2003). ATM primarily responds to DSBs, and mainly phosphorylates and activates Chk2. Several other checkpoint factors co-function with ATM. For example, The Mre11-Rad50-Nbs1 complex is a sensor for DSBs, substrate for ATM and regulator of ATM activity (Lee and Paull, 2005).

ATR primarily responds to replication stress and mainly phosphorylates and activates Chk1 (Burrows and Elledge, 2008). A current model for ATR recruitment and activation at damaged sites is as follows: ATR forms a stable complex with the ATRIP (ATR-interacting protein; Cortez et al., 2001). When DNA replication is impeded, extensive single-stranded DNA (ssDNA) is generated through discordance between DNA polymerases and the MCM helicase, and then coated by replication protein A (RPA), which in turn recruits ATR through ATRIP binding (Namiki and Zou, 2006; Zou and Elledge, 2003). However, the need for the final step has been challenged by the recent identification of ATRIP mutants that are deficient in RPA binding and, thus, do not support stable ATR retention at damaged sites but still promote Chk1 phosphorylation (Ball et al., 2005; Kim et al., 2005). Also in this pathway, several crucial cofactors are required for optimal activation of ATR-Chk1. The Rad9-Rad1-Hus1 complex is a PCNA-like clamp loaded by Rad17-RFC onto damaged sites (Burrows and Elledge, 2008). TopBP1 promotes ATR kinase activity (Hashimoto et al., 2006; Kumagai et al., 2006), through interactions with Rad9-Rad1-Hus1 and ATR-ATRIP (Mordes et al., 2008). Claspin is a mediator that facilitates Chk1 phosphorylation by ATR (Lee et al., 2003). However, the mechanisms for ATR recruitment and activation at damaged sites have yet to be completely elucidated.

Several studies in higher eukaryotic cells have implicated the CDC6 proteins in replication-checkpoint activation. In human cells, overexpression of CDC6 in G2 cells blocks mitotic entry by activating Chk1 (Clay-Farrace et al., 2003). In addition, Lau et al. showed that, despite CDC6 depletion in S-phase cells slowing replication, it does not activate Chk1 and leads to cells that prematurely enter mitosis with DNA synthesis (Lau et al., 2006). In an in vitro DNA-replication system with Xenopus egg extracts, Xenopus CDC6 (XCDC6) was shown to be required for XChk1 activation induced by replication inhibition (Oehlmann et al., 2004). Nevertheless, the molecular mechanism by which CDC6 contributes to replication-checkpoint activation remains unknown. Here, we show that human CDC6 physically interacts with ATR in a Cdk-phosphorylation-stimulated manner and that this interaction is required for proper activation of the replication checkpoint induced by modest replication stress. Further experiments suggested that a similar mechanism might also operate in Xenopus cells. In fission yeast, Cdc18 (CDC6) physically interacts with Rad3-Rad26 (ATR-ATRIP) via Rad26 to sustain the replication-checkpoint activation (Hermand and Nurse, 2007). Thus, involvement of CDC6/Cdc18 interaction with ATR-ATRIP in a replication checkpoint might be a conserved feature throughout eukaryotes.

We investigated whether human CDC6 physical interacts with certain checkpoint protein(s) by GST-CDC6 pulldown assays with HeLa cell nuclear extracts followed by immunoblotting. ATM and ATR, but not Chk1, Chk2, TopBP1, claspin, Nbs1 or Mre11, were pulled down specifically with GST-CDC6 (Fig. 1AII). Rad9-Hus1-Rad1 proteins also did not bind to GST-CDC6 (data not shown). In these assays, cyclin A and Cdh1, proteins that bind to CDC6 (Mailand and Diffley, 2005; Petersen et al., 1999), served as controls. To confirm whether endogenous proteins are associated with CDC6 in vivo, HeLa cell nuclear extracts were immunoprecipitated with an anti-CDC6 antibody. As shown in Fig. 1B, ATR co-precipitated with CDC6. Thus far, we have been unable to detect co-immunoprecipitation between endogenous CDC6 and ATM. Therefore, the CDC6-ATR interaction is considered to be physiologically important.

We then mapped the interaction domain(s) of CDC6 with ATR (Fig. 1C,D). GST-CDC6 and the truncated derivatives studied were mixed with HeLa cell nuclear extracts and bound proteins were analyzed. ATR was found to bind to CDC6 fragments (amino acids) 89-268 and 180-268, but the binding was significantly weaker with fragment 1-180 (Fig. 1D). Binding studies to further narrow the ATR-binding domain found that CDC6 180-220, which contains a Walker A motif of the AAA ATPase family (Fig. 1C) (Williams et al., 1997; Herbig et al., 1999), is sufficient for ATR interaction (Fig. 1D). As shown below, this fragment was also able to bind to Xenopus ATR. In these assays, cyclin A, which binds to CDC6 via the Cy motif, was retained with CDC6 1-180 and 89-268, but not with CDC6 180-268 and 180-220, as expected. ATR forms a complex with ATRIP (Cortez et al., 2001; Zou and Elledge, 2003). In the studies reported here, it was difficult to determine whether ATRIP is retained with full-length GST-CDC6 because our antibody against ATRIP (86 kDa) cross-reacted with GST-CDC6 (Fig. 1E). However, using truncated GST-CDC6 89-268 and 180-220, which could bind to ATR, we detected specific retention of ATRIP (Fig. 1E). Similar to ATR, there was only weak binding between ATRIP and GST-CDC6 1-180.

When CDC6 is overexpressed in G2 cells, it blocks mitotic entry through Chk1 activation (Clay-Farrace et al., 2003). CDC6 has three Cdk-phosphorylation sites (S54, S74 and S106) (Jiang et al., 1999; Petersen et al., 1999) and Cdk phosphorylation is required for CDC6 activity to induce G2 arrest (Clay-Farrace et al., 2003). Therefore, we investigated whether CDC6-ATR interaction might be regulated by Cdk phosphorylation. Wild-type CDC6, a phosphorylation-deficient CDC6 mutant (S54A, S74A, S106A) and a phosphorylation-mimic mutant (S54E, S74E, S106E) (Jiang et al., 1999) were tagged with tandem hemagglutinin (HA) at the N-terminus (2HA-CDC6 WT, AAA and EEE, respectively), and the vectors expressing CDC6 were transiently transfected into 293T cells. The nuclear extracts were then immunoprecipitated with anti-HA antibodies. ATR co-precipitated with 2HA-CDC6 WT and EEE, but little co-precipitated with the AAA mutant (Fig. 2A). However, the steady-state levels of 2HA-CDC6 AAA proteins and, therefore, the levels of immunoprecipitated AAA proteins were lower than 2HA-CDC6 WT and EEE upon transient expression (Fig. 2A), as expected from the fact that Cdk phosphorylation attenuates APC/C^Cdh1-mediated CDC6 degradation (Mailand and Diffley, 2005). Therefore, the amount of co-precipitated ATR was normalized to that of immunoprecipitated CDC6. These results demonstrated that the efficacy of ATR co-precipitation with 2HA-CDC6 AAA is reduced to ~40% compared with 2HA-CDC6 WT and EEE (Fig. 2A, lower panel). To demonstrate reciprocal co-immunoprecipitation, 293T cells were transfected with a vector expressing FLAG-ATR along with 2HA-CDC6 WT, AAA or EEE and immunoprecipitated with anti-FLAG antibodies (Fig. 2B). 2HA-CDC6 WT was co-precipitated specifically with FLAG-ATR. Immunoblotting with anti-phospho-S54-CDC6 antibodies revealed that CDC6 proteins phosphorylated by Cdk were co-precipitated. Although FLAG-ATR was precipitated with comparable efficacy in all experiments, the efficacy of 2HA-CDC6 AAA co-precipitation relative to the input was decreased to one-third that of the WT and, conversely, that of 2HA-CDC6 EEE increased threefold.

We studied this result further with a different system. 293T cells transiently expressing 2HA-CDC6 were treated with 10 mM purvalanol A, a specific inhibitor of Cdk1 and Cdk2 (Gray et al., 1998; Villerbu et al., 2002), or left untreated, and then immunoprecipitated with anti-HA antibodies. Also, in these experiments, the steady-state levels of 2HA-CDC6 were reduced by Cdk inhibition (Fig. 2C). Therefore, the amount of co-precipitated ATR was normalized to that of immunoprecipitated 2HA-CDC6. The efficiency of ATR co-precipitation was found to be reduced to ~25% by purvalanol-A treatment (Fig. 2C, lower panel). Cells were also treated simultaneously with purvalanol A and MG132, an inhibitor of 26S proteasomes, and immunoprecipitated with anti-HA antibodies. Simultaneous MG132 treatment decreased the reduction in 2HA-CDC6 by Cdk inhibition (Fig. 2C). Even under such conditions, the efficacy of ATR co-precipitation decreased to ~30% by purvalanol-A treatment (Fig. 2C). Immunoblotting with anti-phospho-S54-CDC6 antibodies revealed that purvalanol-A treatment indeed inhibited phosphorylation of 2HA-CDC6 by Cdk (Fig. 2C). In addition, we found that the interaction between ATR and 2HA-CDC6 EEE was not affected by purvalanol-A treatment (Fig. 2D), further supporting the notion that this drug affects ATR-CDC6 through inhibition of CDC6 phosphorylation by Cdk.

Taken together, these data led us to conclude that interaction of ATR with CDC6 is enhanced by CDC6 phosphorylation by Cdk. When cells are synchronized in early-mid G1 phase (6 hours after release from nocodazole arrest), CDC6 protein levels are remarkably decreased by APC/C^Cdh1-mediated ubiquitylation (data not shown) (Méndez and Stillman, 2000; Mailand and Diffley, 2005). As a result, in the present study, CDC6 immunoprecipitation with anti-CDC6 antibodies was remarkably reduced and ATR co-precipitation was undetectable in such cells (data not shown). Therefore, the observed co-precipitation between CDC6 and ATR in asynchronous cells might mainly represent interactions in S-phase cells. When cells were treated with 0.1 mM hydroxyurea (HU) for 4 hours, a condition we used to investigate CDC6 involvement in replication-checkpoint activation (see below), co-precipitation of ATR with CDC6 was not enhanced (Fig. 2E). Even with 2.5 mM HU treatment for 16 hours, the efficacy of the interaction was not changed (Fig. 2F).

To determine whether CDC6 depletion by siRNA affects activation of a replication checkpoint in human cells, we designed siRNAs targeting 5′ untranslated regions of CDC6 mRNA. Treatment with the CDC6 siRNAs (6 nM) for 48 hours reduced the level of total CDC6 proteins to ~30% in HeLa cells compared with control-siRNA-treated cells (Fig. 3A). The effect was specific, because the level of MCM7 was not affected. Subcellular fractionation experiments showed that the level of chromatin-bound MCM7 proteins was decreased to ~70% by CDC6 depletion (Fig. 3A). Fluorescence-activated cell-sorting analysis showed that the cell-cycle distribution pattern is only slightly affected by CDC6 depletion (Fig. 3B). Cell growth was also largely unaffected (data not shown). Under the present experimental conditions in HeLa cells, CDC6 depletion did not induce spontaneous Chk1 phosphorylation (Fig. 3C), as reported previously (Lau et al., 2006; Kan et al., 2008). By contrast, it was reported that, when CDC6 is further depleted with more siRNAs (~100 nM), MCM loading is further reduced, leading to cell-cycle arrest around G1-S and/or cell death (Feng et al., 2003; Lau et al., 2009). Under such conditions, spontaneous Chk1 phosphorylation is observed.

We then investigated whether CDC6 depletion impairs ATR-mediated Chk1 phosphorylation induced by replication stress, such as polymerase inhibition. For these studies, CDC6-depleted or control-siRNA-treated HeLa cells were further treated with 2.5 mM HU for 4 hours and Chk1 phosphorylation was monitored with anti-phospho-S345-Chk1 antibodies (Zhao and Piwnica-Worms, 2001). As shown in Fig. 3C, robust phosphorylation of Chk1 was induced by 2.5 mM HU treatment and was not affected by CDC6 depletion, in agreement with previous findings that CDC6 depletion does not abolish Chk1 phosphorylation induced by treatment with 5 μg aphidicolin/ml in HeLa cells (Lau et al., 2006). Given the report that Chk1 phosphorylation induced by 1.5 mM HU treatment in HeLa cells is suppressed by 5 mM caffeine (Rodríguez-Bravo et al., 2006), an inhibitor for ATM and ATR, the above data and that of Lau et al. (Lau et al., 2006) suggest that CDC6 does not play a pivotal role in ATR activation by the stalled forks generated by 2.5 mM HU or 5 μg aphidicolin/ml.

However, Lau at el. also reported that CDC6-depleted HeLa cells (not treated with any drugs) prematurely enter mitosis despite the existence of actively replicating DNA (Lau et al., 2006). This suggested to us that CDC6 might be involved in replication-checkpoint activation elicited by modest replication stress. Extensive RPA-coated ssDNA is currently considered to be a crucial intermediate for ATR activation (Burrows and Elledge, 2008; Namiki and Zou, 2006; Zou and Elledge, 2003). We hypothesized that, although CDC6 is not necessarily required for activation of a replication checkpoint when there is sufficient RPA-coated ssDNAs, CDC6 might be required when replication stress is relatively modest. Therefore, because one crucial response elicited by replication-checkpoint activation involves a mitotic block (Kelly and Brown, 2000), we treated control-siRNA-transfected HeLa cells with various concentrations of HU for a relatively short period (i.e. 4 hours) and examined the effect of this on mitosis block by immunostaining the cells for phosphorylation of histone H3 at S10, a mitotic marker (data not shown). As expected, 2.5 mM HU treatment reduced the mitotic index to ~40% of that of untreated cells. Interestingly, we found that, even with 0.1 mM HU treatment, which reduces DNA synthesis by about 50% (Ge et al., 2007), HeLa cells were significantly inhibited in mitotic entry to a mitotic index of ~60%, suggesting that the replication checkpoint might be activated by low-dose HU treatment. We then tested whether CDC6 depletion affects the mitosis block. Although CDC6 depletion seemed to abrogate the mitotic block, at least partially, at all the HU concentrations tested, the effect was most prominent when cells were treated with 0.1 mM HU (data not shown). Therefore, we examined in more depth the influence of CDC6 depletion on mitotic block elicited by a 4-hour 0.1 mM HU treatment. As shown in Fig. 3D, CDC6 depletion increased the percentage of phospho-histone H3 (pHH3)-positive, mitotic cells by ~30% in cells treated with 0.1 mM HU, whereas it increased mitotic cells by only ~15% on treatment with 2.5 mM HU, suggesting that CDC6 plays a crucial role in replication-checkpoint activation by modest replication stress. Furthermore, we examined the effect of CDC6 depletion on mitotic block induced by 0.1 mM HU treatments at various time points (Fig. 3E). Longer incubation (8 and 12 hours) in the presence of 0.1 mM HU further decreased the amount of mitotic cells. Also under such conditions, CDC6 depletion resulted in a statistically significant increase in mitotic cells (Fig. 3E). However, the extent seemed to be limited compared with the 4-hour incubation.

Although the ATR-Chk1 pathway is a major component in replication-checkpoint activation (Guo et al., 2000; Hekmat-Nejad et al., 2000; Kumagai et al., 1998; Nghiem et al., 2001), several previous studies have indicated the existence of another pathway(s) to prevent premature mitotic entry upon replication stress (Brown and Baltimore, 2003; Rodríguez-Bravo et al., 2006; Zachos et al., 2003; Zachos et al., 2005). For example, it was previously shown that inhibition of mitotic entry in HeLa cells caused by 1.5 mM HU treatment for 25 hours is dependent on Chk1 and claspin but not on ATR and Rad17 (Rodríguez-Bravo et al., 2006). Therefore, we examined whether the replication-checkpoint response elicited by 0.1 mM (or 2.5 mM) HU for 4 hours depends on ATR by co-treating cells with HU and 5 mM caffeine. In order to compare the effect of caffeine with CDC6 depletion that occurs before the HU treatment (data not shown), we treated cells with 5 mM caffeine from 6 hours before the addition of HU. Caffeine was able to overcome the inhibition of mitotic entry induced by 0.1 mM HU (Fig. 3F), suggesting that activation of the replication checkpoint elicited by these experimental conditions depends on ATR (or possibly ATM). We found no detectable increase in Chk1 phosphorylation at S345 upon treatment with 0.1 mM HU for 4 hours (Fig. 3C). Therefore, even by modest replication stress in which Chk1 phosphorylation is not increased to an experimentally detectable level, an ATR-dependent replication checkpoint is activated and elicits a G2 arrest. This seems consistent with the fact that, in Xenopus egg extracts, ATR is required to prevent premature mitotic entry not only when replication is inhibited by exogenous stress but also during normal DNA replication, which might generate some internal replication stress (Hekmat-Nejad et al., 2000). Caffeine increased mitotic cells by ~40% also with 2.5 mM HU treatment (Fig. 3F) and the extent of recovery was higher than with CDC6 depletion (Fig. 3D), suggesting that CDC6 has rather a minor role in replication-checkpoint activation by strong replication stress.

To determine whether the abrogation of the checkpoint response is specifically due to the CDC6 depletion, we performed ‘add-back’ experiments. Because CDC6 interaction with ATR is enhanced by Cdk phosphorylation, we compared CDC6 EEE and CDC6 AAA. We established HeLa cell lines stably expressing 2HA-CDC6 AAA or EEE at similar levels (Fig. 4A). Analysis for subcellular localization of 2HA-CDC6 showed that almost all CDC6 AAA is bound to the chromatin and nuclear matrix, whereas most CDC6 EEE is in the soluble fraction with the rest bound to the chromatin and nuclear matrix (Sugimoto et al., 2009). These cells were then transfected with siRNAs targeting 5′ untranslated regions of CDC6 mRNA, or control siRNA. Immunoblot analysis showed that exogenous 2HA-CDC6 protein levels were not affected by CDC6 siRNA, whereas endogenous CDC6 decreased to ~30% (Fig. 4A).

Cells were further treated with 0.1 mM HU for 4 hours and mitotic indexes were determined. It has been reported that ionizing-radiation-induced DSBs activate the DNA-damage checkpoint, lead to Cdk downregulation and consequently induce APC/C^Cdh1-mediated CDC6 degradation (Duursma and Agami, 2005; Mailand and Diffley, 2005). CDC6 is also destabilized upon ultraviolet radiation (UV) and methyl-methane-sulfonate treatment through Huwe1-mediated ubiquitylation (Hall et al., 2007). We therefore examined the effects of HU treatment on CDC6 stability. The levels of endogenous CDC6 proteins in HeLa cells were not affected by either 0.1 mM or 2.5 mM HU treatment for 4 hours (Fig. 3C, Fig. 4A). Furthermore, the stability of transfected CDC6 AAA and CDC6 EEE mutant proteins was not differentially affected in response to 0.1 mM HU treatment for 4 hours (Fig. 4A). We also found that CDC6 is not destabilized by an 18-hour 2.5 mM HU treatment in normal human fibroblasts (data not shown). As reported, 10 or 25 J/m² UV irradiation resulted in CDC6 destabilization (data not shown). Therefore, replication-checkpoint activation by HU treatment might inhibit Cdk activity only to the extent that is insufficient for mitotic entry but enough for S-phase maintenance. It is generally considered that mitotic entry requires high Cdk activity.

In control HeLa cells with the backbone vector, CDC6 depletion abrogated the checkpoint response induced by 0.1 mM HU (Fig. 4B), as with parental HeLa cells. By contrast, despite CDC6 siRNA treatment, mitotic entry was inhibited in HeLa cells with 2HA-CDC6 EEE (Fig. 4B), clearly demonstrating the CDC6 requirement for the ATR-dependent mitotic-block response. Interestingly, CDC6 AAA could not rescue the checkpoint defect (Fig. 4B). It was previously shown that Xenopus CDC6 mutants deficient in Cdk phosphorylation support normal replication with an efficiency comparable to that of wild type (Pelizon et al., 2000). We also found that 2HA-CDC6 AAA induces stronger re-replication than CDC6 EEE when transiently overexpressed with Cdt1 (Sugimoto et al., 2009). Therefore, CDC6 AAA might represent a novel mutant that specifically loses the ability to regulate ATR-dependent replication-checkpoint activation but retains the pre-RC-formation function. Taken together, these data strongly indicate that, in human cells, Cdk-phosphorylation-stimulated CDC6 interaction with ATR has a crucial role in replication-checkpoint activation elicited by modest replication stress.

The ATR-dependent checkpoint is also activated by UV irradiation. Thus, we tried to evaluate the role of CDC6 in UV-induced checkpoint activation. HeLa cells were irradiated with various doses of UV and, after 4 hours, examined for mitotic block and Chk1 phosphorylation. UV irradiation blocked mitotic entry in a dose-dependent manner, and Chk1 phosphorylation was also observed in a dose-dependent manner (data not shown). We then investigated the effects of CDC6 depletion on such UV-activated checkpoint responses. In contrast to HU treatment, CDC6-depleted HeLa cells were inhibited from mitotic entry in response to UV irradiation with comparable efficiency to control-siRNA-treated cells. Likewise, Chk1 phosphorylation was not affected by CDC6 depletion (data not shown). These data suggest that CDC6 might not be involved in UV-induced checkpoint responses. It is possible that intermediates for checkpoint activation and subsequent signaling pathways are different in some aspects between nucleotide-depletion-induced and UV-induced damage. For example, although 25 J/m² UV induced Chk1 phosphorylation at a comparable level to that by 2.5 mM HU treatment for 4 hours, the former inhibited mitotic entry more strongly (~100% versus ~60% inhibition) and destabilized CDC6 (data not shown). Whereas 0.1 mM HU treatment for 4 hours induced an ~40% mitotic block without detectable Chk1 phosphorylation (Fig. 3C-F), 5 J/m² UV inhibited mitosis by only ~20% despite obvious Chk1 phosphorylation (data not shown). Further in-depth studies will be required to dissect the molecular mechanisms.

Although our data indicate that CDC6-ATR interaction plays a role in the appropriate regulation of replication-checkpoint activation in human cells, it remains unclear how such interaction is involved in ATR activation. To address this issue, we first examined the possibility that CDC6 enhances ATR kinase activity as does TopBP1. Human ATR was immunopurified and the kinase activity was measured using PHAS1 as substrate. As previously reported (Hashimoto et al., 2006; Kumagai et al., 2006), addition of Xenopus Cut5 (XCut5; TopBP1) significantly increased the ATR kinase activity. Under these experimental conditions, GST-CDC6 EEE did not enhance the kinase activity. Addition of GST-CDC6 EEE to the reactions containing XCut5 also did not further increase the activity. Similar results were obtained with wild-type CDC6 (data not shown).

In fission yeast, Cdc18 stabilizes Rad3-Rad26 (ATR-ATRIP)-chromatin binding through physical interaction with Rad26 (Hermand and Nurse, 2007). Therefore, we examined whether stable ATR-chromatin binding is affected by CDC6 depletion in HeLa cells by immunodetecting the formation of bright nuclear ATR foci that might represent high-level accumulation of ATR at sites of DNA damage (Ball et al., 2005). The number of cells with bright ATR foci increased slightly with 0.1 mM HU treatment for 4 hours and increased significantly with 2.5 mM HU treatment. However, the increase in number of ATR foci was not changed much by CDC6 depletion, suggesting that CDC6 is dispensable for stable high-level accumulation of ATR at sites at which DNA is extensively damaged. Also, clear colocalization of ATR foci with CDC6 was not observed (data not shown). Recently, it has been suggested that stable high-level retention of ATR at damaged sites is not necessarily required for some checkpoint responses (Ball et al., 2005). Therefore, it is possible that, although CDC6-ATR interaction does not mediate stable ATR retention at damaged sites, it does regulate dynamic association between ATR and small lesions. In this regard, it might be of interest that the human RPA complex binds to GST-CDC6 in pulldown assays, through the CDC6 region 1-180, which does not have a major role in CDC6-ATR interaction (data not shown).

In an in vitro DNA-replication system with Xenopus egg extracts, XCDC6 is required for XChk1 phosphorylation induced by polymerase inhibition (Oehlmann et al., 2004). Therefore, because XCDC6-XATR interaction might be conserved and play a crucial role in replication-checkpoint activation in Xenopus cells, we examined whether XCDC6 binds to XATR by pulldown assays. Xenopus egg extracts were mixed with GST or GST-XCDC6 and bound proteins were analyzed. XATR was found to bind to GST-XCDC6, but not to GST (Fig. 5A). In addition, XATR was found to co-immunoprecipitate with XCDC6 from egg extracts (Fig. 5B). Unfortunately, so far we could not determine whether the XCDC6-XATR interaction is affected by Cdk, mainly because of difficulty in specifically immunoprecipitating S-phase chromatin-bound XCDC6 with our anti-XCDC6 antibodies. Because CDC6-ATR interaction might be important for checkpoint activation, short CDC6 fragments that bind to XATR could inhibit XATR-dependent checkpoint responses by preventing physiological XCDC6-XATR interaction. Therefore, we repeated the pull-down experiments using Xenopus egg extracts and human CDC6 fragments fused to GST. As shown in Fig. 5C, XATR specifically bound to GST—human-CDC6. More importantly, XATR, similar to human ATR, bound to human CDC6 180-220 but not to CDC6 1-180 (Fig. 5C). The amino acid sequence of fragment 180-220 is conserved in XCDC6 (61% in identity). Taken together, these results indicate that CDC6-ATR interaction is a conserved feature in higher eukaryotic cells.

We first examined whether XCDC6 immunodepletion during S phase affects XATR recruitment onto chromatin in response to polymerase inhibition (Hekmat-Nejad et al., 2000). To address this, we adapted the nuclear transfer experiments (Fig. 6A). Licensed chromatin was isolated under high-salt conditions to remove chromatin-bound XCDC6 and then transferred to mock- or XCDC6-depleted extracts supplemented with aphidicolin to block DNA polymerases. After incubation for a further 45 minutes, chromatin was isolated and chromatin binding of XATR was analyzed by immunoblotting. Aphidicolin-induced recruitment of ATR to chromatin was significantly compromised by XCDC6 depletion (Fig. 6A). As previously reported (Oehlmann et al., 2004), DNA synthesis was not impaired by XCDC6 depletion from the second extracts (data not shown).

We found that the human CDC6 fragment 180-220 can partially prevent XATR binding to GST-XCDC6 when added to the egg extracts (Fig. 6B). Therefore, we tested whether this human CDC6 fragment can inhibit replication-checkpoint activation (Fig. 6C). For these studies, sperm chromatin was added to the egg extracts, which were then incubated for 15 minutes to allow pre-RC formation. Purified CDC6 fragment 180-220 with the SV40 T-antigen nuclear localization signal was then added to the reaction mixture at 10 ng/μl. The molar amount of CDC6 fragment was adjusted to 20-30 times higher than that of endogenous XCDC6 (Coleman et al., 1996). Aphidicolin (4 μg/ml) was also added to each reaction mixture and, after 45 minutes, isolated nuclei were immunoblotted with anti-phospho-S344-XChk1 antibody. Compared with control reactions with buffer alone or CDC6 fragment 1-180, which does not bind to XATR, phosphorylation of XChk1 induced by aphidicolin treatment was reduced by ~40% (Fig. 6D). We also investigated whether addition of CDC6 180-220 affects aphidicolin-induced recruitment of XATR to chromatin by immunoblotting analyses of chromatin fractions. Consistent with the reduction in XChk1 phosphorylation, aphidicolin-induced XATR chromatin recruitment was reduced by ~40% in the presence of CDC6 fragment 180-220 (Fig. 6E). In separate experiments, [³²P]-dCTP was added and the reactions were allowed to proceed for a further 45 minutes (without aphidicolin) after addition of the CDC6 fragments to monitor DNA replication. In all these reactions, DNA replication occurred at comparable levels (Fig. 6F), showing that the addition of CDC6 180-220 did not impair either pre-RC assembly or other replication reactions. Thus, human CDC6 fragment 180-220 specifically inhibits XATR-dependent XChk1 phosphorylation, probably through sequestrating XATR-XCDC6 interaction and XATR chromatin retention. Taken together, these data provide further support for our conclusion that CDC6 interacts with ATR and the CDC6-ATR interaction plays a crucial role in the regulation of replication-checkpoint activation in higher eukaryotic cells.

We have shown here that human CDC6 physically interacts with ATR in a Cdk-phosphorylation-stimulated manner (Figs 1, 2) and that CDC6 is required for the ATR-dependent replication-checkpoint response (i.e. inhibition of premature mitotic entry) activated by modest replication stress (Fig. 3). Interestingly, the defect in the checkpoint response caused by CDC6 depletion was rescued by the CDC6 EEE mutant mimicking Cdk phosphorylation, but not by the AAA mutant deficient in Cdk phosphorylation and ATR interaction (Fig. 4). Therefore, the CDC6 AAA mutant represents a mutant in which two functions of CDC6, replication licensing and replication-checkpoint regulation, are separated at the molecular level. In fission yeast, such mutant Cdc18 (CDC6) has not been identified yet. The region that we identified as an ATR-binding region on CDC6 does not include Cdk-phosphorylation sites (Fig. 1C). CDC6 is exported to the cytoplasm through Cdk phosphorylation but potential nuclear export signals on CDC6 are located at amino acids 462-488 (Delmolino et al., 2001), relatively distant from the Cdk-phosphorylation sites. However, it is conceivable that Cdk phosphorylation alters the high-order structure of CDC6 and thereby exposes the hidden nuclear export signals and ATR-binding region. In addition, the data reported here indicate that regulation of replication-checkpoint activation by CDC6-ATR interaction also occurs in Xenopus egg extracts: the XCDC6-XATR interaction was conserved (Fig. 5), and sequestration of the interaction by a human CDC6-derived peptide that binds to XATR reduced XATR chromatin recruitment and XChk1 phosphorylation induced by polymerase inhibition (Fig. 6).

These data, together with previously reported data in fission yeast (Hermand and Nurse, 2007), indicate that regulation of ATR-dependent replication-checkpoint activation through interaction between CDC6/CDC18 and ATR-ATRIP might be a conserved feature throughout eukaryotes. However, there are also differences among organisms. In the human cells studied here, this mechanism seems to play a crucial role for the replication checkpoint to be activated by modest replication stress, such as that elicited by 0.1 mM HU treatment (Fig. 3D,E, Fig. 4B), which reduces DNA synthesis by about half (Ge et al., 2007). Because CDC6 depletion alone (without additional replication perturbations) results in cells that prematurely enter mitosis with active DNA synthesis (Lau et al., 2006), the CDC6-ATR system might also operate against internal replication stress generated during the normal replication process. However, CDC6 in human cells seems dispensable for robust, ATR-dependent Chk1 phosphorylation induced by stronger replication stress, such as that induced by 2.5 mM HU treatment (Fig. 3C) (Lau et al., 2006). In addition, CDC6 is not required for ATR to accumulate at high levels in bright nuclear foci. CDC6 depletion somewhat affects ATR-dependent mitotic block activated by strong replication stress (i.e. 2.5 mM HU for 4 hours) but the relative contribution seems small (Fig. 3D,F).

A model that could reconcile these findings is depicted in Fig. 7. When replication stress is strong and, therefore, extensive RPA-coated ssDNA is generated, ATR-ATRIP molecules accumulate to high levels at the damaged sites through ATRIP-RPA interaction without the help of CDC6, forming bright ATR foci detectable by immunostaining, and robustly phosphorylate Chk1 (Fig. 7, left pathway). However, such high-level stable accumulation at damaged sites might be dispensable for ATR to be activated (Ball et al., 2005; Kim et al., 2005), although ATR foci formation could support some specific aspects of checkpoint responses. In other words, it is possible that only dynamic interaction between ATR and lesions is sufficient for ATR activation. In addition, in the presence of such extensive damage, another replication-checkpoint pathway(s) involving Chk1, but not ATR, could also be activated, leading to an ATR-independent (and thus caffeine-insensitive) block to mitosis (Rodríguez-Bravo et al., 2006). The p38 MAP kinase might be involved in such a pathway (Bulavin et al., 2001). Overall, CDC6 might have a minor role in checkpoint activation under such conditions.

However, even when cells are exposed to modest replication stress, ATR can be activated and delay mitotic entry despite low-level ATR foci formation and Chk1 phosphorylation (Fig. 3C,D). It is surprising that such low-level ATR activation can delay the cell cycle but, as in the Xenopus egg extract system, ATR is required to prevent premature mitotic entry not only when replication is inhibited by exogenous stress but also during normal DNA replication (Hekmat-Nejad et al., 2000). Presumably, fine regulatory control of the ATR-Chk1 pathway might function when cells encounter modest exogenous replication stress or endogenous replication stress. Under such circumstances, CDC6 and CDC6-ATR interaction might be required for ATR to elicit a mitotic block efficiently (Fig. 3D,E, Fig. 4B), presumably through regulating dynamic interaction between ATR and the damaged sites (Fig. 7, right pathway). Potential CDC6-RPA interactions could also play a role in this context. Extensive future studies will be required to clarify these points. Our present analyses of UV-damage-induced checkpoint responses found no indication of CDC6 involvement in these pathways. Although the details remain unclear, intermediates for checkpoint activation and subsequent signaling pathways might differ in some aspects between nucleotide-depletion-induced and UV-induced damage, as discussed above.

In contrast to human cells, in the Xenopus egg extract system, robust ATR-dependent Chk1 phosphorylation induced by 40 μM aphidicolin treatment seems dependent on XCDC6 (Oehlmann et al., 2004) and, as shown here, might be regulated by XCDC6-XATR interaction (Figs 5, 6). ATR chromatin recruitment induced by polymerase inhibition also depends on CDC6-ATR interaction in the Xenopus egg system (Fig. 6). The reason for the difference between the two systems, human cultured cells and Xenopus egg extracts, is not known at present. However, it should be emphasized that, even in the egg extract system, ORC-dependent, stable chromatin binding of XCDC6 is dispensable for XATR activation (Oehlmann et al., 2004), suggesting that the interactions between XCDC6 and the replication-checkpoint machinery are dynamic.

There is also a difference between human and fission yeast cells in terms of the CDC6-associated replication checkpoint. In yeast cells, Cdc18 (CDC6) is degraded during S phase depending on Cdk phosphorylation (Jallepalli et al., 1997; Jallepalli et al., 1998). Therefore, Cdc18 could not be involved in initial activation of the Rad3 (ATR)-dependent checkpoint. Attenuation of Cdk activity by the initial checkpoint activation then re-stabilizes Cdc18, which in turn interacts with Rad3-Rad26 (ATR-ATRIP) to sustain checkpoint activation (Hermand and Nurse, 2007). Therefore, these molecules seem to constitute a positive-feedback loop. However, in human cells, a significant part of CDC6 phosphorylated by Cdk remains in nuclei during S phase (Fujita et al., 1999) and CDC6 interaction with ATR is augmented by Cdk phosphorylation, as shown here (Fig. 2). Therefore, CDC6 might be involved in initial activation of an ATR-dependent checkpoint. When cells were treated with 0.1 mM HU for 4 hours, a condition under which we could definitely detect CDC6 involvement in replication-checkpoint activation, Cdk-phosphorylation-dependent CDC6 stabilization was not perturbed (Fig. 3C, Fig. 4A) and interaction of CDC6 with ATR was not affected (Fig. 2E). Therefore, ATR- and CDC6-dependent replication-checkpoint activation by modest replication stress might inhibit Cdk activity only to the extent that is insufficient for mitotic entry but enough for maintaining S-phase activity. If strong damage activates the replication checkpoint robustly and consequently inhibits Cdk activity markedly, then ATR-CDC6 interaction could be attenuated. However, under such conditions, CDC6 might be dispensable for checkpoint activation.

HeLa and 293T cells were grown in Dulbecco's modified Eagle's medium with 8% fetal calf serum. The chemicals used were: MG132 (Calbiochem), purvalanol A (Calbiochem), hydroxyurea (Sigma), caffeine (Nacalai) and aphidicolin (Sigma).

GST-fusion proteins were bacterially produced and purified. GST pulldown assays were performed as described previously (Sugimoto et al., 2008), with 200 μg GST-fusion proteins per 2 ml extract from 1×10⁸ cells. For pulldown assays with Xenopus egg extracts, 20 μl extracts were 40-fold diluted and then 6 μg GST-fusion proteins were added to the samples. Binding reactions for GST-fusion proteins were performed for 5 hours at 4°C. Human CDC6 fragments added to the Xenopus egg extracts were purified from the GST-fusion proteins by digestion with PreScission Protease (GE Healthcare).

293T cells were transfected as described previously (Sugimoto et al., 2008). HeLa cells were infected with retroviruses expressing the 2HA-CDC6 AAA mutant or EEE mutant, or with control retroviruses and selected with puromycin (0.5 μg/ml) as described (Sugimoto et al., 2008).

Cells were extracted with modified CSK buffer containing 500 mM NaCl, 0.1% Triton X-100, 1 mM dithiothreitol (DTT) and multiple protease inhibitors (Sugimoto et al., 2008). Aliquots of the extracts were then immunoprecipitated with the indicated antibodies and protein-G—Sepharose beads (Amersham Bioscience). The beads were washed three times with NET gel buffer (Sugimoto et al., 2008). Immunoblotting and quantitation of the band signals were performed as described previously (Sugimoto et al., 2008).

siRNA oligonucleotides (IDT) were synthesized with the following sequences (sense strand): CDC6-1 (5′-AGACUAUAACUCUACAGAUUGUGdAdA-3′), CDC6-3 (5′-GGAGGACACUGGUUAAAGAAUUUdAdT-3′) and control scrambled (5′-CUUCCUCUCUUUCUCUCCCUUGUdGdA-3′). HeLa cells (4×10⁴ cells per well in 12-well plates) were transfected with a 1:1 mixture of CDC6-1 and CDC6-3 or control siRNA duplexes (total 12 pmol; 6 nM) using HiPerFect transfection Reagent (Qiagen).

HeLa cells were washed, fixed with 3.7% formaldehyde, extracted with modified CSK buffer containing 0.1% Triton X-100, and then incubated overnight at 4°C with mouse anti-Ser10-phosphorylated histone H3 antibody. The samples were further incubated for 2 hours with Alexa-Fluor-594-conjugated goat anti-mouse IgG antibody (Molecular Probes), counterstained with DAPI and analyzed with a Leica FW4000 microscope.

Preparation of interphase egg extracts and de-membranated sperm nuclei, and the in vitro DNA-replication experiments were performed as described previously (Yoshida et al., 2005). Chromatin transfer experiments were performed as described previously (Hashimoto et al., 2006). Briefly, licensed sperm chromatin obtained after 15 minutes pre-incubation were diluted with ten volumes of EB buffer containing 0.1 mM NaCl and 2 mM DTT (Yoshida et al., 2005) and then centrifuged through a 20% sucrose layer at 10,000 g at 4°C for 5 minutes. The pellets were washed with EB containing 2 mM DTT and resuspended in second egg extracts. Immunodepletion of XCDC6 was carried out using 75 μl of anti-XCDC6 antisera for 100 μl egg extract (Yoshida et al., 2005). Nuclear and chromatin fractions were prepared as described previously (Kumagai et al., 1998; Yoshida et al., 2005). For immunoprecipitation, extracts were diluted with EB and then immunoprecipitated with the indicated antibodies and protein-G—Dynabeads (Invitrogen). The beads were washed three times with EB buffer containing 130 mM KCl.

For bacterial expression of GST-CDC6, pGEX6P-1-CDC6 was constructed by inserting human CDC6 cDNA into pGEX6P-1 (GE Healthcare). CDC6 truncated fragments were amplified by PCR using combinations of the following primers: 5′-1 primer, 5′-CGGGATCCGCTGTCGTCATGCCTCAAACCCG-3′; 5′-89 primer, 5′-CGGGATCCCATACACTTAAGGGACGAAGATTGG-3′; 5′-180 primer, 5′-CGGGATCCGATGTCATCAGGAATTTCTTGAGGG-3′; 3′-180 primer, 5′-TTAAGCGGCCGCCTAATCCATCTCCCTTTCCCTGGC-3′; 3′-220 primer, 5′-TTAAGCGGCCGCCTACTTGAGGTCTTGCAGAATCCG-3′; and 3′-268 primer, 5′-TTAAGCGGCCGCCTATTCCAATTTCCTCATCATGTCCTTC-3′. These PCR products were digested by BamHI and NotI and subcloned into pGEX6P-1. For construction of fused SV40 T-antigen nuclear localization signal (NLS)-CDC6 fragments, an oligonucleotide containing the SV40 T-antigen NLS/BamHI site/HindIII site sequence (5′-GATCGCCTAAAAAGAAGCGTAAAGTCGGATCCAAGCTTG-3′) and its complementary oligonucleotide were synthesized, annealed, inserted into the pGEX6P-1 BamHI-EcoRI site, and CDC6 fragments were then introduced into the BamHI-NotI site. pGEX6P-2-XCDC6 was provided by Haruhiko Takisawa (Osaka University, Osaka, Japan). Retrovirus vector pQCXIP-2HA-CDC6 was constructed by inserting CDC6 cDNA tagged with tandem HA at its N-terminus into pQCXIP with a puromycin-resistance gene (Clontech). The cDNAs encoding Cdk-phosphorylation-deficient CDC6 AAA (Ser54, Ser74 and Ser106 were replaced with alanines) and phosphorylation-mimic EEE (Ser54, Ser74 and Ser106 were replaced with glutamates) mutants were gifts from Wei Jiang (Burnham Institute for Medical Research, La Jolla, CA). pBJ5.1-FLAG-ATR (Cliby et al., 1998) was kindly provided by Karlene Cimprich (Stanford University, Palo Alto, CA).

Preparation of antibodies against human CDC6, human MCM7, Xenopus ATR and Xenopus Chk1 were described previously (Hashimoto et al., 2006; Sugimoto et al., 2008). Antibodies against human Rad9, Hus1 and Rad1 were provided by Toshiki Tsurimoto (Kyushu University, Fukuoka, Japan). Other antibodies were purchased as follows: ATM (2C1, GeneTex), ATR (sc1887, Santa Cruz Biotechnology), ATRIP (number 2737, Cell Signaling), Cdh1 (c7855, Sigma), Chk1 (06-965, Upstate), Chk2 (05-649, Upstate), Claspin (ab3720, Abcam), cyclin A (sc596, Santa Cruz Biotechnology), Mre11 (611366, BD Bioscience), FLAG-M2 (F3165, Sigma Aldrich), Nbs1 (611870, BD Bioscience), Ser10-phosphorylated histone H3 (number 9706, Cell Signaling), Ser345-phosphorylated-Chk1 (number 2341 or 2348, Cell Signaling), Ser54-phosphorylated-CDC6 (sc12920R, Santa Cruz Biotechnology) and TopBP1 (ab2402, Abcam).

Statistical analysis was performed by a two-tailed Student's t-test to validate the data. P-values of <0.05 were considered statistically significant.

We thank Haruhiko Takisawa, Toshiki Tsurimoto, Wei Jiang, Koji Ode and Karlene Cimprich for Materials. We are also grateful to Satoko Yoshida, Chie Ogawa and Akiko Noguchi for assistance. This work was supported in part by a Grant to M.F. from the Ministry of Education, Culture, Sports, Science and Technology of Japan.