Heat induction of a novel Rad9 variant from a cryptic translation initiation site reduces mitotic commitment

Temperature has the enigmatic ability to inactivate DNA break repair in human cells (Pandita et al., 2009). The Mre11-Rad50-Nbs1 (MRN) complex recruits the DNA damage sensor ATM^Tel1 kinase to broken chromosomes (Falck et al., 2005), where the kinase phosphorylates the histone variant γ-H2AX within megabase regions surrounding the lesion (Rogakou et al., 1998). This chromatin modification attracts additional repair factors like the scaffold protein 53BP1^Crb2 (Ward et al., 2003) which form a large checkpoint complex to inactivate the cell cycle regulator Cdc2^CDK1 (Gould and Nurse, 1989; O'Connell et al., 1997; Smith et al., 2010). A small rise in temperature from ∼37°C to ∼40°C blocks this signalling process by inducing the rapid relocalisation of the MRN complex from the nucleus to the cytoplasm (Seno and Dynlacht, 2004) and by delaying the recruitment of 53BP1 to chromatin (Laszlo and Fleischer, 2009).

Unexpectedly, ATM kinase remains active under heat stress conditions still modifying histone γ-H2AX despite the absence of detectable DNA breaks (Hunt et al., 2007). Although the biological details of ATM activation are unknown, the underlying mechanism may be equivalent to its stimulation by oxidative stress. An increase in reactive oxygen activates ATM directly in the absence of the MRN complex by inducing the formation of a disulfide-crosslinked dimer (Guo et al., 2010). The cellular targets of ATM at high temperatures remain to be identified, but they may be linked with apoptosis (Furusawa et al., 2011) or a prolonged G2-M arrest (Zölzer and Streffer, 2001).

While ATM signals unprocessed DNA double-strand breaks, the related kinase ATR^Rad3 binds via its partner protein ATRIP^Rad26 directly to the ssDNA Binding Protein (RPA) after the ends of the break have been converted to ssDNA tails (Zou and Elledge, 2003; Sartori et al., 2007; Zierhut and Diffley, 2008). This end processing is crucial for dsDNA break repair by homologous recombination in G2 (Caspari et al., 2002; Ferreira and Cooper, 2004; Huertas et al., 2008). The resulting junctions between ssDNA and dsDNA are independently recognised as a damage signal by the Rad17 complex that loads the Rad9-Rad1-Hus1 ring (9-1-1) next to ATR-ATRIP (Bermudez et al., 2003). The 9-1-1 ring resembles the replicative clamp PCNA (Caspari et al., 2000a; Xu et al., 2009; Doré et al., 2009), but in contrast to PCNA, contains an extended and highly flexible domain which is provided by the tail of the Rad9 subunit. This tail domain, which is phosphorylated during the unperturbed cell cycle and in response to DNA damage (Caspari et al., 2000b; St Onge et al., 2001; Roos-Mattjus et al., 2003), co-activates ATR^Rad3 jointly with the scaffold protein Rad4^TopBP1 that is preloaded onto chromatin at the start of S phase (Furuya et al., 2004; Delacroix et al., 2007; Navadgi-Patil and Burgers, 2009; Zegerman and Diffley, 2007). Assembly of the ATR^Rad3 complex in S phase and G2 signals to Cdc2 via the checkpoint kinase 1 (Chk1), whereas ATM^Tel1 utilises checkpoint kinase 2 (Chk2^Cds1) independently of the cell cycle stage (Smith et al., 2010).

We report here a novel requirement of the DNA damage checkpoint kinases Rad3^ATR, Tel1^ATM, Cds1^Chk2 and Chk1 for the maintenance of a heat-induced G2 arrest. Our experiments reveal striking differences between genotoxic and temperature stress. While detection of DNA lesions requires the recruitment of Rad9 to the chromatin, the 9-1-1 loader Rad17 is obsolete for heat signalling. Consistent with the conclusion that the normal 9-1-1 complex is not involved, cells induce a N-terminally truncated Rad9 variant (Rad9-M50) at elevated temperatures by utilising a cryptic start site in the rad9 mRNA (AUG-50). Thus, the ability of Rad3^ATR to delay mitosis at high temperatures requires a different Rad9 protein than its checkpoint function in response to DNA damage. Induction of Rad9-M50 ensures that Chk1 remains in a hypo-phosphorylated state by facilitating the removal of its DNA-damage induced modification under heat stress conditions.

Rad9-M50 was discovered whilst conducting experiments at elevated temperatures with the well characterised rad9-HA strain. This strain expresses a Rad9 protein with a C-terminal hemagglutinin (HA) tag from its endogenous locus on chromosome 1 (Caspari et al., 2000a; Harris et al., 2003; Furuya et al., 2004). Exposure of rad9-HA cells to 40°C for 20 min resulted in the appearance of a smaller band (Fig. 1A). Induction of this variant was independent of the growth medium and occurred to a similar extent between 37°C and 40°C. Its expression was effectively blocked by the ribosome inhibitor cycloheximide (Fig. 1B) strongly indicating that active translation is important. Expression of the isoform was independent of the HA tag and occurred also when Rad9 was C-terminally fused to GFP (supplementary material Fig. S1A).

Given that the Wis1-Sty1 MAPK pathway detects environmental stress (Shiozaki and Russell, 1995; Shiozaki et al., 1998; Millar et al., 1995), we investigated Rad9-M50 expression in the presence of high osmotic and oxidative conditions. As shown in Fig. 1C, induction was specific to temperature stress at 40°C and not observed when rad9-HA cells were exposed to 0.6 M KCl, 1 M sorbitol or 0.3 mM H₂O₂ for 30 min at 30°C. Neither mutation of the MAPK kinase (MAPKK) Wis1 nor loss of the MAPK Sty1 reduced Rad9-M50 expression (supplementary material Fig. S1B,C). We also excluded an involvement of the nutrient sensing Tor1 pathway (Petersen and Nurse, 2007) and of the Srk1 kinase acting down-stream of the Wis1-Sty1 pathway (Lopez-Girona et al., 1999) (supplementary material Fig. S1D).

Because inhibition of translation blocked synthesis of this variant (Fig. 1B), we decided to mutate all internal methionine codons to alanine to test whether they act as a start site at high temperatures (Fig. 1D). Mutant alleles were generated by fusion PCR and integrated at the rad9 chromosomal locus using the Cre-lox cassette exchange technique (Watson et al., 2008). This technique allows for the targeted integration of rad9 genes at its chromosomal locus which was replaced by a marker gene (ura4+) flanked by the recognition sites (loxP, loxM) of the Cre recombinase. Transformation of this strain with a plasmid that encodes Cre and contains a rad9-HA allele flanked by the same lox sequences results in the exchange of rad9 with the chromosomal marker. The integrated rad9-HA alleles were amplified from genomic DNA and sequenced to confirm the mutation.

Intriguingly, only replacement of methionine 50 (M50A) abolished expression of the variant, thereafter referred to as Rad9-M50 (Fig. 1E). Since AUG-50 is down-stream of the first intron (Fig. 1D) and because this intron is retained in some rad9 cDNA clones (Murray et al., 1991), we deleted this sequence in-frame. Usage of AUG-50 was however not influenced by this intron (supplementary material Fig. S1D).

Ribosomes could either reach AUG-50 by moving past the first AUG codon in a process known as leaky ribosome scanning (Kochetov, 2008) or they could enter the transcript downstream of the first initiation site. Leaky ribosome scanning is used, for example, to synthesise a mitochondrial and cytoplasmic variant of the S. cerevisiae glutaredoxin-2 enzyme. Translation from the first AUG produces the mitochondrial variant, while initiation at AUG-35 deletes the leader peptide allowing the protein to remain in the cytoplasm (Porras et al., 2006). Internal ribosome entry sites (IRES) are normally located up-stream of the first AUG (King et al., 2010) making it less likely that such a sequence exists between AUG-1 and AUG-50. However, RNA hairpin structures can recruit ribosomes independently of IRES (Zu et al., 2011) and such secondary structures may form in a temperature-dependent manner.

To test whether Rad9-M50 is post-translationally modified, we exposed rad9-HA cells to 30°C or 40°C for 1 hour and subjected soluble protein extracts to isoelectric focusing on an immobilised pH gradient (non-linear, pH 3–10) prior to electrophoresis on a 10% SDS PAGE. While the phospho-isoforms of full-length Rad9 do not separate under these conditions, two distinct isoforms of Rad9-M50 were present at 40°C (Fig. 1F). Although this indicates that the variant is phosphorylated, we failed to detect a change in cells devoid of Rad3 kinase, the 9-1-1 subunit Hus1, the 9-1-1 loader Rad17 or the MAPK Sty1 (supplementary material Fig. S1E,F).

To investigate whether expression of Rad9-M50 is dependent on the full-length protein, we mutated the first methionine to alanine (rad9-M1A-HA) (Fig. 2A). As expected, this rad9-M1A-HA strain lacked Rad9 rendering cells highly DNA damage sensitive and checkpoint deficient (Fig. 2C; Fig. 3C-E). Loss of Rad9 had no influence on expression of Rad9-M50 which was fully induced at 40°C (Fig. 2B). Removal of Rad9, whilst allowing cells to express Rad9-M50, had an unexpected effect on hydroxyurea (HU) sensitivity. As shown in Fig. 2C, rad9-M1A cells lose viability faster than rad9 deletion cells (Δrad9). This implies that the truncated protein, which is expressed at low levels at 30°C in rad9-M1A cells (Fig. 2B), may substitute for Rad9 thereby acting in a dominant negative manner. This genotoxic function of Rad9-M50 was not abolished upon deletion of rad1 (Fig. 2B) suggesting that it does not depend on the formation of a 9-1-1 like complex.

Since the RNA sequence between methionine-1 and serine-49 may be involved in the regulation of AUG-50, we deleted this sequence from the genomic rad9 gene using the Cre-lox technique (Fig. 2A). Interestingly, the basal levels of Rad9-M50 increased sharply in this rad9-Δ1-49-HA strain at 30°C thereby reducing the degree of induction at 40°C (Fig. 2B). This suggests a role of the mRNA segment between M1 and S49 in the suppression of AUG-50 usage at low temperatures.

We also mutated AUG-1 and AUG-50 simultaneously to alanine to test whether ribosomes could utilise any of the remaining AUG codons (Fig. 2A). This rad9-M1A+M50A-HA strain neither expressed Rad9 nor Rad9-M50, but weakly induced a smaller variant at 40°C which was also detectable in the absence of AUG-50 (Fig. 2B,D). These observations imply that heat relaxes the usage of internal initiation sites in the rad9 transcript allowing ribosomes to initiate down-stream of AUG-1. This conclusion was confirmed by the absence of any inducible band in cells expressing a rad9 gene in which M50 and M74 were both replaced by an alanine residue (rad9-M50A+M74A-HA) or in which all remaining methionine codons were mutated to alanine (rad9-M50A+M74A+M311A+M312A+M357A) (Fig. 2D).

Alignment of the N-terminal sequences of Rad9^Sp, Rad9A^Hs and Rad9B^Hs shows that M50 is replaced by a cysteine in both human proteins (Fig. 3A). Interestingly, the Ensembl database (release 64 – Sep. 2011) curates a yet uncharacterised splice variant of Rad9B (ENSP00000387329; Rad9B-001; 345 aa) that lacks the first 72 aa starting at methionine 73. Further work is however required to establish whether this splice variant is a functional paralog of Rad9-M50. We also analysed Rad9 proteins from other yeast species across the Ascomycota group and found that M50 is the only internal start site that is conserved across diverse clades (supplementary material Fig. S1G). For example, Rad9 from Kluyveromyces thermotolerans shares only 19.9% identity with S. pombe Rad9 but has a methionine residue at position 51 (supplementary material Fig. S1G).

To test whether Rad9-M50 can replace the full-length protein in the 9-1-1 ring, we took advantage of the rad9Δ1-49-HA strain that lacks Rad9 but expresses high levels of the variant at 30°C (Fig. 2B). As shown in Fig. 3B, loss of the N-terminal domain would delete an internal segment of Rad9 without affecting the contact interfaces with Hus1 and Rad1. Hence, the truncated protein may still be able to associate with both proteins. If such a complex were to exist, it does not respond to DNA damage since rad9Δ1-49-HA cells are highly DNA damage sensitive (Fig. 3C) and lack a cell cycle arrest when DNA replication was challenged with the Ribonucleotide Reductase (RNR) inhibitor hydroxyurea (HU) (Fig. 3E). To measure the checkpoint arrest, wild-type (rad9-HA) and rad9Δ1-49-HA cells were both enriched in early G2 by isolating small G2 cells from a lactose gradient (Forsburg and Rhind, 2006), and released into rich medium with and without 12mM HU at 30°C. While wild-type cells delayed cell cycle progression (Fig. 3D), rad9Δ1-49-HA cells were as deficient as cells devoid of the full-length protein (rad9-M1A-HA) (Fig. 3E).

We employed size fractionation chromatography to test whether rad9Δ1-49-HA cells, which only express Rad9-M50, form the 9-1-1 complex. As reported previously (Caspari et al., 2000), the complex was present in fractions 12 to 14 (Superdex-200 column) when protein extracts from rad9-HA wild-type cells (30°C) were used. Rad9 was however absent from these fractions when extracts from rad9-Δ1-49-HA cells (30°C) were analysed. The variant eluted instead in fractions 8 to 11 that contain protein complexes larger than 400 kDa (Fig. 3F). To find out whether the same changes apply to the endogenous variant, we fractionated an extract from rad9-HA cells after having heat shocked the cells at 40°C for 30 min. The induced variant possessed a very similar elution profile as Rad9-M50 in rad9-Δ1-49-HA cells, although fractions 12 and 13 contained some of the variant (Fig. 3G). Deletion of the 9-1-1 subunit Rad1 caused the expected loss of Rad9 from fractions 12 to 14 and induced a shift of the full-length protein to the high molecular weight fractions 8 to 11 (Fig. 3H). This confirms that fractions 12 to 14 contain the 9-1-1 complex and it suggests that Rad9 and Rad9-M50 present in fractions 8 to 11 form alternative protein complexes. While detection of Rad9-M50 in the high molecular weight fractions was not affected by loss of Rad1, the small amount detected in fractions 12 and 13 was lost (Fig. 3H). Taken together, some Rad9-M50 may assemble with Rad1 in a 9-1-1 like complex (frac. 12 and 13), but the majority of the protein appears to be in alternative protein complexes (frac. 8-11).

Consistent with the inability of Rad9-M50 to substitute for Rad9, Chk1 phosphorylation, which is dependent on the 9-1-1 ring (Capasso et al., 2002), was abolished when rad9Δ1-49-HA cells were treated with the topoisomerase I inhibitor campothecin (CPT) (Fig. 3I).

The first insight into the biological roles of Rad9-M50 came from the observation that its induction level dropped with an increase in cell number. To systematically analyse this observation, we grew a rad9-HA wild-type culture in rich medium at 30°C from logarithmic into stationary phase and exposed cells at 4 different stages to 40°C for 30 min (Fig. 4A,B). Intriguingly, induction of Rad9-M50 declined whilst cells exited the logarithmic growth phase (Fig. 4B; time points 1, 2 and 3). Once cells entered stationary phase, expression of the full-length protein started to decline as well (Fig. 4B; time point 4). Since this suggests that only cycling cells express Rad9-M50, we resorted to defined minimal medium to compare cycling with non-cycling cells. While phenylalanine as a nitrogen source still permits slow progression through the cell cycle, the absence of nitrogen arrests cells in G1 (Fantes and Nurse, 1977). In agreement with the earlier observation, only slowly cycling cells, but not arrested cells, induced Rad9-M50 (Fig. 4C).

Intrigued by the disappearance of Rad9 in stationary cells, we extended this analysis to Rad1, Hus1, Rad3, Rad17 and Chk1. Interestingly, only the 9-1-1 subunits showed a significant drop in expression once cells entered stationary phase (supplementary material Fig. S2). The amount of Chk1 and Rad17 declined to a much smaller extent, and the level of Rad3 kinase remained constant (supplementary material Fig. S2). The situation for Hus1 was even more intriguing. As previously reported (Caspari et al., 2000), S. pombe cells constitutively express three Hus1 isoforms with yet unknown functions in addition to the full-length protein. While variant B remained largely unchanged, the other three isoforms (A, C and D) disappeared in stationary phase (supplementary material Fig. S2C). The loss of the 9-1-1 complex in non-cycling cells may explain why so far no mutation in this complex has been linked with a disorder in post-mitotic cells (O'Driscoll and Jeggo, 2003).

Since it was previously reported that asynchronous S.pombe cultures arrest cell cycle progression at elevated temperatures (Nurse, 1975; Petersen and Hagan, 2005), we wanted to know whether induction of Rad9-M50 blocks the cell cycle at 40°C. To this end, we synchronised wild-type cells (rad9-HA), cells unable to express Rad9-M50 (rad9-M50A-HA), and cells devoid of full-length Rad9, but able to induce Rad9-M50 (rad9-M1A-HA), in G2 using lactose gradients. Small G2 cells were released into rich medium at either 30°C or 40°C, and samples were withdrawn over a period of 300 min (Fig. 4D-F). At 30°C, wild-type cells progressed through two cell cycle rounds as indicated by the two peaks of septation that coincide with G1/S phase (Mitchison and Nurse, 1985). At 40°C, wild-type cells remained in G2 for up to 200 min before re-entering the cell cycle (Fig. 4D). In contrast, cells devoid of Rad9-M50 (rad9-M50A-HA), but still expressing Rad9, terminated this heat-induced arrest ∼40 min prematurely (Fig. 4E). On the contrary, cells lacking Rad9, but expressing Rad9-M50 (rad9-M1A-HA), showed a wild-type like arrest (Fig. 4F). We repeated this experiment with backcrossed strains and obtained the same results (Fig. 4H).

Since wild-type cells suppress expression of Rad9-M50 at 30°C (Fig. 1A), we tested whether its untimely induction at 30°C would enforce a G2 arrest. To do this, we took advantage of the novel urg1 (uracil regulated gene 1) expression system (Watson et al., 2011). Expression of urg1 is rapidly induced upon addition of uracil to cells grown in minimal medium. Using the Cre-lox cassette exchange technique, we integrated a rad9-M50-EGFP fusion gene down-stream of the urg1 promotor on chromosome 1 in rad9⁺ and rad9 deleted cells (supplementary material Fig. S3A). Addition of uracil induced expression of Rad9-M50-EGFP within 15 min at 30°C (supplementary material Fig. S3B) closely resembling its rapid induction by heat stress (Fig. 1A). Interestingly, its upregulation in G2 synchronised cells had no effect on cell cycle progression at 30°C, but extended the heat-induced G2 arrest by ∼40 min at 40°C (Fig. 4I,J). This cell cycle effect was independent of the presence of the endogenous rad9 gene (supplementary material Fig. S3C,D). Overexpression of Rad1 had no effect on the G2 arrest at 40°C showing that this is a specific function of Rad9-M50 (supplementary material Fig. S3E,F). These experiments confirm a role of Rad9-M50 in cell cycle regulation and they also show that Rad9-M50 requires heat shock conditions to be active.

Very little is currently known about the role of DNA damage checkpoint proteins in the response to heat stress (Pandita et al., 2009). To find out whether Rad9-M50 acts in the context of the checkpoint or independently of it, we measured the G2 arrest at 40°C of strains either deleted for a checkpoint gene or carrying a kinase-dead (KD) allele of checkpoint kinases. Each experiment was repeated at least three times with two independently isolated mutants to cater for any inter-experimental variations (Fig. 5). This analysis revealed three different phenotypes: (i) cells deficient in Tel1^ATM or Rad3^ATR kinase entered mitosis on average 60 min earlier than wild-type cells (Fig. 5B,C), (ii) cells deleted for rad9, hus1, rad1, cds1^Chk2, chk1 or crb2^53BP1 entered mitosis ∼40 min earlier (Fig. 5A,G,H,J,L,N), and (iii) cells without the 9-1-1 loader Rad17 displayed a normal G2 arrest (Fig. 5I). On balance, these results reveal a novel requirement of the DNA damage checkpoint for the maintenance of a heat-induced G2 arrest. But in contrast to genotoxic stress, the checkpoint genes are not required for its induction. This is not the only striking difference between genotoxic and thermal stress. While Rad17 is essential to load the 9-1-1 ring onto damaged chromatin, the loader is dispensable under heat stress conditions (Fig. 5I). To confirm this important observation, we repeated this experiment with cells devoid of rad17 and cells lacking both Rad17 and Rad9-M50 (Δrad17 rad9-M50A). While Δrad17 cells showed a wild-type-like arrest, cells devoid of Rad17 and Rad9-M50 displayed a shorter arrest (Fig. 5P). Although this shows that Rad9-M50 acts independently of Rad17, we were surprised to find that deletion of rad1 or hus1 shortened the G2 arrest to a similar extent as loss of Rad9-M50. This could be explained by an alternative complex containing these proteins which is not dependent on Rad17. The requirement of Cds1 kinase for this G2 arrest was also unexpected (Fig. 5L), because this kinase acts normally in S phase in response to DNA replication stress (Lindsay et al., 1998). Premature entry into mitosis did not correlate with a temperature sensitivity of the checkpoint mutants (supplementary material Fig. S2D). The only exception were cells deleted for crb2 which are temperature sensitive. The latter observation points towards an additional function of this protein outside of the normal checkpoint response.

While human ATM^Tel1 performs a dominant DNA damage checkpoint role, S. pombe Tel1 is much less important as long as Rad3 is active (Furuya et al., 2004). To probe their relationship in response to heat stress, we combined the tel1 deletion with a kinase-dead allele of rad3. In contrast to the single mutants, the Δtel1 rad3-KD double mutant never arrested completely slowly entering mitosis at 40°C (Fig. 5F). This shows that both kinases contribute more equally to a heat-induced arrest than to a G2 delay triggered by genotoxic stress.

Given that the 9-1-1 ring activates Rad3-to-Chk1 signalling (Furuya et al., 2004; Navadgi-Patil and Burgers, 2009), we combined the rad9-M50A mutation with gene deletions in chk1, cds1 and crb2, and with a kinase-dead allele of rad3. While loss of the variant had no additive effect in the absence of Rad3, Crb2 or Chk1, its ablation shortened the arrest in a cds1 mutant to a similar extent as observed in rad3 and tel1 mutants (Fig. 5D,K,M,O). As summarised in Fig. 5Q, these findings imply that Rad9-M50 acts in the same heat response pathway as Rad3, Crb2 and Chk1, but in parallel to Cds1.

Given that Chk1 is phosphorylated at S345 by Rad3 kinase whilst bound to damaged chromatin (Capasso et al., 2002; Kosoy and O'Connell, 2008), we analysed the modification status of Chk1 in response to genotoxic and heat stress utilising normal SDS PAGE and isoelectric focusing. We used two different types of genotoxic stress, the chronic modification of Chk1 in cells with hyper-active Cdc2 kinase (Capasso et al., 2002) and the induced modification upon inhibition of topoisomerase 1 [camptothecin (CPT)] (Walworth et al., 1993).

To this end, we grew Chk1-HA and Chk1-HA cdc2.1w cells at 30°C, and shifted samples to 40°C for 1 hour. The Cdc2.1w kinase harbours a point mutation (G146D) that renders the cell cycle regulator hyper-active (Booher and Beach, 1986). As reported previously, this aberrant increase in Cdc2 activity triggered the constitutive phosphorylation of Chk1 at 30°C resulting in a slower migrating band (Fig. 6A). Unexpectedly, exposure to 40°C suppressed this band shift suggesting that heat reverses Chk1 modifications triggered by genotoxic stress. Heat itself failed to produce a band shift in Chk1 wild-type cells (Fig. 6A). The same extracts were then subjected to isoelectric focusing using a non-linear pH gradient from 3 to 10. This assay revealed 4 isoforms of Chk1 at 30°C in the absence of DNA damage. This implies that Chk1 can be multiply phosphorylated, and that some of these phosphorylation events occur in undamaged cells (Fig. 6B, panel 1). While isoforms 2, 3 and 4 remained unchanged at 40°C, the abundance of isoform 1 declined (Fig. 6B, panel 2). Consistent with the band shift in a cdc2.1w background at 30°C, we observed significant changes to Chk1. The abundance of isoform 4 strongly increased at the expense of isoform 3, and a novel, more alkaline isoform was present (Fig. 6B, panel 3). These changes were reversed at 40°C as indicated by the presence of only two isoforms (No 2 and 3) (Fig. 6B, panel 4). These findings imply that isoform 4 and the more alkaline isoform correspond to the conformational changes induced by phosphorylation of Chk1 at S345 (Kosoy and O'Connell, 2008).

To test whether heat also suppresses Chk1 phosphorylation induced by acute DNA damage, we grew chk1-HA cells at 30°C in presence of 40 µM CPT for 3 hours before splitting the culture. One sample was shifted to 40°C for 1 hour while the other remained at the lower temperature (Fig. 6C). Although CPT was present throughout the experiment, heat efficiently suppressed the modification of Chk1 resulting in the disappearance of the slower migrating band (Fig. 6D).

Given that heat also induces Rad9-M50, we compared the kinetics of its upregulation with the suppression of Chk1 phosphorylation upon a temperature shift to 40°C. To this end, chk1-HA and rad9-HA cells were pre-incubated for 3 hours at 30°C in the presence of 40 µM CPT, and samples were withdrawn 0, 10, 20 and 30 min after a shift to 40°C. As shown in Fig. 6E, the slower migrating Chk1 band started to decline after 20 min, at the same time as Rad9-M50 appeared. Interestingly, this decline was specific to Chk1 and not observed for the slower migrating Rad9 bands (Fig. 6E, lower panel). Isoelectric focusing of Chk1 extracts taken at 0 min and 30 min after the shift revealed that CPT triggered a strong increase in isoform 4 (Fig. 6F; 0 min) that was reversed by heat stress (Fig. 6F; 30 min). Interestingly, the more alkaline isoform, present in cdc2.1w cells at 30°C, was absent after CPT treatment suggesting that chronic and acute genotoxic stress affect Chk1 in different ways.

To test whether Rad9-M50 is linked with Chk1 de-phosphorylation, we combined the chk1-HA gene with the rad9-M50A-HA allele in the same strain (chk1-HA rad9-M50A-HA). These cells are able to phosphorylate Chk1 in the presence of CPT, because they express Rad9, but they should be unable to reverse this modification since Rad9-M50 is absent. Isoelectric focusing of protein extracts obtained from chk1-HA rad9-M50A-HA cells, which were pre-incubated for 3 hours at 30°C in the presence of 40 µM CPT before a shift to 40°C for 30 min, showed the expected increase in isoform 4 at 0 min (Fig. 6G, panel 1), but in contrast to wild-type cells, the intensity of isoform 4 did not decline at 40°C. On the contrary, the damage-induced modifications became more abundant as indicated by a significant increase in the alkaline isoform (Fig. 6G, panel 2).

We concluded from these data that Rad9-M50 signalling modulates the DNA damage response at elevated temperatures by promoting the removal of modifications from Chk1 kinase that were induced by genotoxic stress (Fig. 6H).

In summary, our data entertain a model such that heat-induction of Rad9-M50 from a cryptic translation initiation site results in the assembly of an alternative Rad9-Rad1-Hus1 complex that activates Rad3 in a chromosomal context which is inaccessible to Rad17. Activation of Rad3 ensures that Chk1 kinase remains in a hypo-phosphorylated state thereby preventing premature mitosis under heat stress conditions.

These findings are consistent with a recent report showing that ATR^Rad3-to-Chk1 signalling is activated when human cells are exposed to elevated temperatures (Furusawa et al., 2011). Which heat alterations stimulate ATR is unknown and its signalling output is the phosphorylation of Chk1 at S345 and not its dephosphorylation. Why S. pombe cells remove this modification at elevated temperatures is currently unclear, but it may allow them to modulate Chk1 activity to maintain a heat-induced G2 arrest (Fig. 5). Chk1 kinase has a basal activity level which increases ∼5–10 fold upon phosphorylation of S345 in its C-terminal domain (Capasso et al., 2002; Kosoy and O'Connell, 2008). The details of its activation remain to be resolved, but it is generally believed that modification of S345 by ATR^Rad3 releases the C-terminal domain from the N-terminal catalytic domain thereby stimulating kinase activity (Tapia-Alveal et al., 2009). DNA damage-induced phosphorylation at S345 is removed by Dis2 phosphatase in S. pombe (den Elzen and O'Connell, 2004) and by protein phosphatase 2A (PP2A) in human cells (Leung-Pineda et al., 2006). Our data show that Chk1 dephosphorylation is limited to the DNA damage-induced modifications (Fig. 6). This selectivity could be achieved by different phosphatases, by different adapter proteins for the same phosphatase or by conformational changes within the kinase shielding some phosphate groups. The precise mechanism remains to be uncovered, but our data show a strong correlation between Chk1 dephosphorylation and the induction of Rad9-M50. The requirement of Rad3, Rad1, Hus1 and Crb2 for the heat-induced G2 arrest suggests that Rad9-M50 targets Chk1 indirectly.

Our conclusion that Rad9-M50 reprograms Rad3 to maintain Chk1 in a hypo-modified state is in line with the observation that ATR^Rad3 kinase stimulates PPA2 to dephosphorylate Chk1 in human cells (Leung-Pineda et al., 2006). Taken together, Rad3 may target a protein phosphatase like PPA2 or Dis2 to remove the damage-induced modifications from Chk1 at elevated temperatures.

How heat activates ATM^Tel1 and ATR^Rad3 remains a mystery. Given that S phase is the most temperature-sensitive cell cycle stage (VanderWaal et al., 2001), heat could cause chromosomal alterations during DNA replication. Human cells arrest S phase at elevated temperatures upon the release of nucleolin from the nucleolus (Wang et al., 2001). A role of Rad9-M50 in S phase would be consistent with its induction in cycling cells (Fig. 4). Perhaps heat causes a DNA alteration which is not accessible to the normal 9-1-1 complex. This would explain the need for an alternative variant. Both, human Rad17 and the 9-1-1 complex are stimulated by ssDNA Binding Protein (RPA) (Zou et al., 2003), but RPA is the target of nucleolin upon its release from the nucleolus at high temperatures (Wang et al., 2001). Hence, binding of nucleolin to RPA may interfere with the loading of the 9-1-1 ring at elevated temperatures. Alternatively, hyperthermia could directly affect the activity of enzymes like topoisomerases which are involved in DNA replication (Bromberg and Osheroff, 2001).

On the other hand, there is evidence that heat damages DNA directly. For example, 8-oxoguanine accumulates in DNA at elevated temperatures in the presence of reactive oxygen (Bruskov et al., 2002) and heat-labile repair intermediates caused by DNA methylation are converted into DNA breaks (Lundin et al., 2005). Whether any of these changes lead to the heat-activation of ATR^Rad3 and ATM^Tel1 kinase remains to be discovered.

Intriguingly, human cells also respond to cellular stress by synthesising N-terminally truncated proteins and many of them regulate mitotic commitment. Genotoxic stress activates a cryptic cdc25B promotor in a Chk1-dependent manner to produce a shorter variant of this phosphatase allowing cells to exit the G2 arrest (Jullien et al., 2011). Stress caused by the accumulation of unfolded protein in the endoplasmatic reticulum triggers binding of MDM2 to the p53 mRNA thereby inducing a p53 variant (p53/47) that lacks the first 39 aa and arrests cells in G2 (Bourougaa et al., 2010). Finally, heat stress induces a shorter variant of the transcription factor Oct4, Oct4B1, which regulates cell cycle progression in stem cells (Farashahi Yazd et al., 2011). Although the induction mechanisms are different, the processes are limited to dividing cells and the variants are involved in cell cycle regulation.

Further work is however necessary to explore whether human cells use a similar Rad9 variant to activate ATR kinase under heat stress conditions.

Wild type (h- ade6-M216 leu1-32 ura4-D18), rad9-HA (h- ade6-M216 leu1-32 ura4-D18 rad9-3HA-kanMX4), Δchk1 (h- ade6-M216 chk1::ura4+ leu1-32 ura4-D18), chk1-KD (h- ade6-M216 chk1::loxP-chk1-D155E-HA-loxM leu1-32 ura4-D18), chk1-S345A (h- ade6-M216 chk1::loxP-chk1-S345A-HA-loxM leu1-32 ura4-D18), Δrad9 (h- ade6-M216 rad9::ura4+ leu1-32 ura4-D18), Δcrb2 (h- ade6-M216 crb2::ura4+ leu1-32 ura4-D18), Δtel1 (h- ade6-M216 tel1::ura4+ leu1-32 ura4-D18), Δrad1 (h- ade6-M216 rad1::ura4+ leu1-32 ura4-D18), Δrad17 (h- ade6-M216 rad17::ura4+ leu1-32 ura4-D18), Δcds1 (h- ade6-M216 cds1::ura4+ leu1-32 ura4-D18), Δhus1 (h- ade6-M216 hus1::leu2+ leu1-32 ura4-D18), Chk1-HA (Walworth et al., 1993), Hus1-Myc (Caspari et al., 2000), Rad3-Myc and rad3-D2249E (Bentley et al., 1996), Rad17-GFP-HA (h90 ade6-216 leu1-32 lys1-131 ura4-D18 rad17::rad17-GFP-HA-kanMX4), Rad1-GFP-HA (h90 ade6-216 leu1-32 lys1-131 ura4-D18 rad1::rad1-GFP-HA-kanMX4), Rad9-GFP-HA (h90 ade6-216 leu1-32 lys1-131 ura4-D18 rad9::rad9-GFP-HA-kanMX4) and Rad9-M1A-CFP (h- ade6-M216 rad9::loxP-rad9-M1A-CFP-loxM leu1-32 ura4-D18).

As described by Watson and colleagues (Watson et al., 2008), the loxP sequence was integrated 181 nt upstream of the rad9 start condon at position 1,714,271 and the loxM sequence was placed 136 nt down-stream of the stop codon at position 1,715,672 on chromosome 1. All rad9 mutant strains described in this report are variants of this ‘base’ strain (h- ade6-M216 rad9::loxP-ura4+-loxM leu1-32 ura4-D18). The mutated rad9-HA alleles were constructed using fusion PCR. The internal mutant primers (35–45 bp) had the complementary sequences accommodating the desired mutation, the forward primer (CGATAGTGGCATGCTAGAAAACACCACATTATAGATTTACC) contained two tandem SphI sites and the reverse primer (GCTATCACACTAGTCAGATCTATATTACCCTGTTATCCC) contained two tandem SpeI sites. Equal amounts of the 2 overlapping DNA fragments were fused in two steps. 10 cycles: 5 µl 2 mM dNTPs, 10 µl 5×GC Buffer, 0.5 µl Phusion DNA Polymerase in 50 µl. 30 cycles in the same PCR tube: plus 2.5 µl 10 µM forward primer, 10 µM primer reverse primer, 5 µl 2 mM dNTPs, 10 µl 5×GC Buffer, 0.5 µl Phusion DNA Polymerase and water to a final volume of 100 µl [98°C 30 s, 56°C 30 s, 72°C 90 s]. The resulting SphI-rad9-HA-SpeI fragments were cloned between the loxM and loxP repeats in the plasmid pAW8, which also contains the cre recombinase gene. The promotor sequence of 181 nt between loxP and ATG-1 (or ATG-50 in rad9-Δ1-49-HA) was restored. Integrated rad9 alleles were amplified and sequenced.

The urg1::rad9-M50-EGFP and the urg1::rad1-HA strains were both constructed as described by Watson and colleagues (Watson et al., 2011). ATG-50 is the first start codon in the urg1::rad9-M50-EGFP construct down-stream of the urg1 promotor and ATG-1 is the first start codon in the urg1::rad1-HA construct. Both ORFs end with the last codon before the stop codon to allow for translation into the linker and the tag sequences.

Protein extracts for the 2D protein electrophoresis were prepared from 5×10⁸ cells as described by Schmidt and colleagues (Schmidt et al., 2007). Between 10 µg and 15 µg protein was loaded onto Immobiline^TM DryStrip gels pH 3–10NL 7 cm (GE Healthcare) in Destreak^TM rehydration solution with 0.5% of the corresponding IPG buffer. Strips were rehydrated for 12 hours at 50 V on a Biorad PROTEAN IEF cell and focused using the rapid ΔV (for Rad9-HA) or linear ΔV (for Chk1-HA) preset method (10,000 Vh). Before applying strips onto a 10% SDS PAGE, they were sequentially incubated in a IPG tray on a orbital rocking platform for 10 min in 2.0 ml of equilibration buffer I [6 M urea 0.375 M Tris-HCl (pH 8.8), 2% SDS, 20% glycerol, 2% (w/v) DTT] and 2.0 ml equilibration buffer II [6 M urea, 0.375 M Tris-HCl (pH 8.8), 2% SDS, 20% Glycerol, 2.5% (w/v) Iodoacetamide].

Lactose gradients were performed as described previously (Forsburg and Rhind, 2006) with the following changes. Cells were grown at 30°C in rich medium to a low cell number 10⁶–10⁷ cells/ml, and 5×10⁸ cells were harvested from these cultures. Lactose gradients were centrifuged at 750 rpm for 7 min in a Sorvall RT Legend bench top centrifuge and small G2 cells were taken from the top of the cell cloud. G2 cells were washed in rich medium and split into two equal 1 ml samples. One sample was incubated at 30°C, whereas the second sample was re-suspended in pre-warmed rich medium and incubated at 40°C. 40 µl aliquots were withdrawn in 20 min intervals and added to 200 µl methanol. Cells were pelleted and stained with 30 µl of a Hoechst (1∶1000)-calcofluor (1∶100) solution (stocks: calcoflour 1 mg/ml in 50 mM sodium citrate, 100 mM sodium phosphate pH 6.0; Hoechst 10 mg/ml in water) prior to scoring under a fluorescence microscope.

The urg1 expression strains were grown in minimal medium minus uracil at 30°C, harvested and loaded onto a lactose gradient prepared with the same minimal medium. Small G2 cells were split into four samples in minimal medium. Two samples were incubated at 30°C and two samples at 40°C. Uracil was added to a final concentration of 0.25 mg/ml to only one sample at either temperature.

Preparation of both, total and soluble protein extracts, and performance of size fractionation are described by Caspari and colleagues (Caspari et al., 2000). The anti-HA antibody (HA.11, clone 16B12, Covance Ltd) was used to detect Rad9-HA, the anti-HA antibody (ab9110, ABCAM) was used to detect Chk1-HA.

We are grateful to Mrs Muneera Hamdi Alghannami for her assistance constructing the Rad9 quintuple mutant and to Mr H. M. Syfuddin for his assistance constructing the Rad9M1A-CFP strain. We would like to thank the Yeast Genetic Resource Center (YGRC) based at Osaka City University and Osaka University for strains. We would like to acknowledge Adam Watson for his advice on the Cre-Lox and urg1 systems. We also thank Dr Nancy Walworth, Dr Claudia Barros and Dr Nia Whitely for helpful comments on the manuscript. This manuscript is dedicated to the late Per Christensen. The authors declare that they have no conflict of interest.