Ccr4 contributes to tolerance of replication stress through control of CRT1 mRNA poly(A) tail length

Replicating cells treated with the ribonucleotide reductase (RNR) inhibitor hydroxyurea (HU) pause or slow down replication fork progression (Branzei and Foiani, 2005). In budding yeast, replication fork stalling provokes a checkpoint response that hinges on activation of the checkpoint kinase Mec1 (ATR in mammals), and subsequent activation of the Rad53 kinase, the ortholog of the Chk2 tumour suppressor. Mec1 and Rad53 control phosphorylation and activation of the checkpoint kinase Dun1, which shares FHA and kinase domain homology with Rad53 and Chk2 and is required for the full transcriptional response to DNA damage (Zhou and Elledge, 1993; Durocher et al., 2000; Bashkirov et al., 2003). Downstream of Mec1 and parallel to Rad53 and Dun1 lies the checkpoint kinase Chk1, which phosphorylates and stabilizes the anaphase inhibitor Pds1, also known as securin, to inhibit the G2-M transition (Wang et al., 2001). Chk1 plays an unidentified role in the response to replication stress, particularly in cells lacking DUN1 (Sanchez et al., 1999; Schollaert et al., 2004).

Numerous checkpoint events converge on the RNR complex to promote survival in response to replication stress and to regulate RNR activity during a normal S phase. Two key targets of the Mec1-Rad53-Dun1 cascade have been identified: Sml1, a small physical inhibitor of RNR enzymatic activity (Zhao and Rothstein, 2002), and Crt1, a DNA-binding protein and member of the winged-helix family of transcription factors (Emery et al., 1996). Sml1 is phosphorylated by Dun1 and subsequently targeted for proteolysis, which relieves physical inhibition of RNR activity (Zhao and Rothstein, 2002). Crt1 inhibits transcription at target promoters through recruitment of the general transcriptional repressors Tup1 and Ssn6 (Huang et al., 1998; Li and Reese, 2001). Notably, Crt1 is phosphorylated in a Mec1-Rad53-Dun1-dependent manner after DNA damage or replication stress, thereby promoting its dissociation from promoter DNA and transcriptional activation (Huang et al., 1998). Crt1-repressed targets include the RNR2, RNR3 and RNR4 genes that encode the DNA-damage-inducible subunits of RNR, CRT1 itself, as well as HUG1, which encodes a small protein that negatively regulates MEC1-dependent checkpoint responses (Basrai et al., 1999). Finally, Rnr2 and Rnr4 redistribute from the nucleus to the cytoplasm after DNA damage in a Mec1-Rad53-Dun1-dependent manner (Yao et al., 2003).

The replication checkpoint response slows S-phase progression, stabilizes stalled replication forks, and increases RNR complex activity (Santocanale and Diffley, 1998; Lopes et al., 2001; Tercero and Diffley, 2001; Chabes et al., 2003). Recent data demonstrate that stabilization of replication forks is the main checkpoint function needed for cell survival in the face of replication stress. A collapsed replication fork can lead to DNA double-strand breaks, chromosome rearrangement and loss, and genome instability (Sogo et al., 2002; Tercero et al., 2003). To date, both Mec1 and Rad53 are implicated in stabilization of stalled replication forks (Lopes et al., 2001; Tercero and Diffley, 2001). Mec1-dependent signaling is also required for the fidelity of replication across regions of the genome that are prone to fork stalling and collapse, known as replication slow zones (RSZs) (Cha and Kleckner, 2002) or fragile sites (Lemoine et al., 2005; Admire et al., 2006). These fragile regions appear sensitive to deoxynucleotide levels, reduced DNA polymerase activity and compromised checkpoint function; for example, deletion of SML1 greatly reduces chromosome breakage at fragile sites in mec1 mutants (Cha and Kleckner, 2002). There is therefore a crucial link between dNTP metabolism, the replication fork and genome stability.

To identify novel regulators of the cellular response to replication stress, we performed a systematic analysis of 4812 strains in the S. cerevisiae haploid non-essential gene deletion set (Giaever et al., 2002) for sensitivity to HU. The screen identified a number of gene deletions that confer HU hypersensitivity, including CCR4 and CAF1/POP2. Ccr4 and associated factors called Not proteins are physically and functionally linked to the transcriptional machinery (Denis, 1984; Deluen et al., 2002). Ccr4 is the catalytic component of the major cytoplasmic mRNA deadenylase complex, which regulates poly(A) tail length and influences both mRNA degradation and translation (Tucker et al., 2001; Chen et al., 2002; Tucker et al., 2002). Caf1/Pop2 is a Ccr4-associated factor with suspected nuclease functions (Daugeron et al., 2001). Ccr4 and related protein complexes thus play diverse roles in processes that regulate mRNA abundance and turnover. Here, we show that CCR4 acts cooperatively with the Mec1-Rad53-Dun1 kinase cascade to help the cell cope with a variety of replication stresses. We identify mutations in CRT1 as the major suppressor of ccr4Δ HU sensitivity and demonstrate that Ccr4 controls CRT1 mRNA poly(A) tail length and influences Crt1 protein abundance. These results forge a link between mRNA poly(A) metabolism and the tolerance of replication stress.

To identify novel gene activities associated with survival of replication stress, we sought to systematically identify genes required for survival in the presence of HU. We therefore replica-pinned the entire genome-wide set of haploid deletion mutants (Giaever et al., 2002) onto plates containing 100 mM HU. Colony growth was assessed to identify candidate HU-sensitive strains, which were then confirmed in spot tests on 100 mM HU. With this approach, we identified and confirmed 49 HU-sensitive mutants (Table S1 in supplementary material).

While these studies were ongoing, a number of genome-wide phenotypic screens also examined HU sensitivity. In particular, Hartman and Tippery quantitatively profiled arrayed deletion strains for sensitivity to either 50 mM or 150 mM HU (Hartman and Tippery, 2004); Parsons et al. (Parsons et al., 2004) examined arrayed strains robotically pinned to solid media containing 100 mM HU; Bennett et al. (Bennett et al., 2001) identified diploid deletion strains cross-sensitive to both ionizing radiation (primary screen) and 100 mM HU (secondary screen). We compared our confirmed HU-sensitive deletion strains (i.e. hits from genome-wide approach corroborated by spot assay) to the results of other genome-wide studies (Fig. 1A and Table 1). Together, the compiled studies identify 118 common (i.e. found by two or more screens) and 216 unique (i.e. found by only one method) gene deletions that confer HU sensitivity (Fig. 1A). The intersection of all four screens revealed 14 HU-sensitive deletion mutants. A total of 49 high-confidence hits identified in at least three out of four screens were classified by Gene Ontology (GO) process (Table 1). As shown in Fig. 1B, many of these mutants are defective in DNA replication, recombination or repair, as expected. However, diverse cellular processes such as transcription, mRNA processing, spindle dynamics, and sister chromatid cohesion were also represented within the intersection dataset. Interestingly, tolerance to HU requires KAR3 and CIK1, which encode proteins that physically interact to form a minus-end-directed kinesin (Chu et al., 2005; Sproul et al., 2005), consistent with recent studies that implicate Mec1 and Rad53 in the regulation of spindle dynamics (Krishnan et al., 2004; Bachant et al., 2005). These genome-wide screens reveal that resistance to replication stress engages diverse cell regulatory processes.

In yeast, Ccr4 and Caf1/Pop2 physically associate to form the major cytoplasmic mRNA deadenylase complex. Ccr4 is the catalytic component of the deadenylase, and Caf1/Pop2 has suspected nuclease functions (Daugeron et al., 2001; Tucker et al., 2002). All of the genome-wide surveys described above, as well as other approaches, uncovered the HU sensitivity of the ccr4Δ strain (Fig. 1B and Table 1) (Westmoreland et al., 2004; Traven et al., 2005). The caf1Δ/pop2Δ strain scored as HU sensitive in three of the four genome-wide analyses (Fig. 1B and Table 1). As removal of either Ccr4 or Caf1/Pop2 confers HU sensitivity, we sought to further investigate their roles in the survival of replication stress. To assess the role of Ccr4 deadenylase activity in the tolerance of replication stress, Ccr4 catalytic residues E556 and D713, which are required for in vitro Ccr4 deadenylase activity, were mutated to alanine (Chen et al., 2002; Tucker et al., 2002). Strains bearing those catalytically compromised CCR4 alleles were sensitive to 100 mM HU, although less so than the full CCR4 deletion (Fig. 2A). Deadenylase-dependent and -independent activities of Ccr4 may thus contribute to survival of replication stress or DNA damage.

Since their role in the tolerance to replication stress had not been characterized, we first assessed a possible genetic role for CCR4 in known replication stress pathways. Deletion of CCR4 in combination with deletion of the Dun1 checkpoint kinase resulted in dramatic hypersensitivity to HU (Fig. 2B). In particular, the ccr4Δ dun1Δ strain was sensitive at 10 mM HU, whereas single mutants were only mildly sensitive to HU concentrations above 50-100 mM. The acute sensitivity of the ccr4Δ dun1Δ strain suggested that the mutant may be defective in stabilization of DNA replication forks, a phenotype observed in mec1 or rad53 checkpoint mutants (Lopes et al., 2001; Tercero and Diffley, 2001). We therefore examined whether ccr4Δ dun1Δ strains were irreversibly sensitive to HU (Allen et al., 1994), a phenotype associated with replication fork collapse (Branzei and Foiani, 2005; Tourriere et al., 2005). Each of the wild-type and single mutant strains tolerated an acute exposure to HU and formed colonies on medium lacking drug, whereas less than 1% of the ccr4Δ dun1Δ starting cell population was viable after exposure to HU for a 24-hour period, indicating irreversible HU sensitivity (Fig. 2C). These results suggest that Ccr4 and Dun1 collaborate to prevent catastrophic replication fork collapse following HU treatment.

In addition to irreversible HU sensitivity, replication fork collapse produces a potent replication checkpoint signal that persists after the removal of the replication stress (Tercero et al., 2003). Thus, we assessed the activation status of Rad53 in wild-type, ccr4Δ, dun1Δ and ccr4Δ dun1Δ strains both during and following acute HU treatment. Rad53 activation was observed in all strains upon addition of HU, as judged by an increase in Rad53 autophosphorylation in kinase assays performed on a membrane blot after in situ re-naturation (Fig. 3A, upper panels). The increase in Rad53 kinase activity was also evident in the appearance of lower mobility Rad53 phospho-isoforms on an anti-Rad53 immunoblot (Fig. 3A, lower panel). Compared with the wild-type control, in each of the ccr4Δ and ccr4Δ dun1Δ strains, the addition of HU elicited a further increase in the magnitude of Rad53 activation, as demonstrated by both Rad53 autophosphorylation and Rad53 mobility shift. The absence of Ccr4 thus enhanced the checkpoint signal during replication stress. Removal of HU from the wild-type cell culture resulted in rapid elimination of slower-migrating forms of phosphorylated Rad53 species, with a concomitant reduction in Rad53 kinase activity (Fig. 3A). By contrast, Rad53 dephosphorylation was delayed in the single mutant dun1Δ or ccr4Δ cells after removal of HU; this delay was exacerbated in the ccr4Δ dun1Δ strain (Fig. 3A). These observations suggested persistent checkpoint activation resulting from collapsed replication forks in the ccr4Δ dun1Δ strain. We conclude that both Ccr4 and Dun1 cooperate to promote replication fork stability in response to nucleotide depletion.

The prolonged Rad53 checkpoint kinase activation upon removal of HU in each of the ccr4Δ, dun1Δ and ccr4Δ dun1Δ strains suggested that these strains might exhibit a delay in recovering from replication stress. We therefore analyzed the DNA content of each strain after release from an HU-induced cell-cycle arrest. As expected, the wild-type, ccr4Δ, dun1Δ, and ccr4Δ dun1Δ strains exhibited an early S-phase arrest in 0.2 M HU, reflecting a lack of DNA synthesis after early origin firing (Fig. 3B). Consistent with the deactivation kinetics observed for Rad53, the wild-type strain resumed cell-cycle progression approximately 15 minutes after removal of HU, with G2-M phase DNA content readily evident at 45 minutes post release. By contrast, each of the ccr4Δ and dun1Δ strains were delayed for resumption of cell-cycle progression after release from HU, with significant G2-M DNA content appearing only at 60 minutes post release. The delay in cell-cycle progression was greatly exacerbated in the ccr4Δ dun1Δ double mutant, which has a reduced G2-M phase DNA content even at 75 minutes post release. Together these observations indicate that Ccr4 and Dun1 cooperate to enable resumption of DNA replication after a period of dNTP shortage.

To further characterize the pathway by which CCR4 acts to promote tolerance to replication stress, we tested for genetic interactions between CCR4 and genes encoding components of the replication checkpoint, replication fork machinery and replication apparatus. As the RNR enzyme complex is the molecular target of HU, we reasoned that genetic impairment of RNR activity would hamper growth of the ccr4Δ strain. Although RNR1 and RNR4 are essential genes in some genetic backgrounds, it is possible to delete either gene in other strain backgrounds, including in the BY4741 background used in this study; such rnr1Δ and rnr4Δ strains grow slowly and are hypersensitive to HU (Wang et al., 1997; Dubacq et al., 2004). We were therefore able to determine whether cells that lack Ccr4 have a genetic requirement for full RNR activity. Despite the viability of single mutant strains, we were unable to recover strains that bore both ccr4Δ and either of the rnr1Δ or rnr4Δ mutations (Fig. 4A). The inviability of the ccr4Δ rnr4Δ double mutant was rescued by expression of RNR4 from a heterologous promoter (data not shown).

We next tested whether the ccr4Δ mutation displayed synthetic HU sensitivity in combination with each of the pol32Δ, mrc1Δ, rad27Δ, tel1Δ and elg1Δ mutations. Each of these deletion strains is defective in replication and/or checkpoint function but has little or no observable growth phenotype in the absence of replication stress. Deletion of CCR4 in combination with each of the above mutations resulted in a synthetic HU sensitivity phenotype (Fig. 4B). These interactions included marked synthetic HU sensitivity of ccr4Δ in combination with the tel1Δ and elg1Δ mutations, which on their own are largely insensitive to HU (Morrow et al., 1995; Kanellis et al., 2003). Finally, since the DNA polymerase α-primase complex is linked to the DNA damage response, we tested whether deletion of CCR4 exacerbated the phenotype of a pri1-M4 temperature sensitive allele of the PRI1 gene that encodes the p48 component of the α-primase complex (Marini et al., 1997). Deletion of CCR4 reduced the permissive temperature of the pri1-M4 mutation strain from 30°C to 25°C (Fig. 4B). Taken together, these genetic interactions indicate that Ccr4 plays a role in both general DNA replication and in tolerance of replication stress, consistent with related phenotypes observed by Traven et al. (Traven et al., 2005).

Our observation that the ccr4Δ dun1Δ double mutant strain is highly sensitive to HU indicated that Ccr4 contributes Dun1-independent functions to the tolerance of replication stress. We therefore tested whether the ccr4Δ mutation interacted genetically with mutations in other replication checkpoint kinases. The ccr4Δ mutation was combined with mutations in MEC1 (mec1-21 or the catalytically inactive mec1-[D2224A] allele), RAD53 (sad1-1) (Allen et al., 1994) or CHK1 (chk1Δ). Not unexpectedly, the HU sensitivity of mec1 and rad53 (sad1) mutant strains was exacerbated by the absence of Ccr4 (Fig. 4C). Chk1 is part of a less well-characterized branch of the checkpoint response that lies parallel to Rad53 and Dun1 (Sanchez et al., 1999). Although Chk1 plays a crucial role in response to replication stress in fission yeast and in human cells, in budding yeast its role is only revealed when the Rad53-Dun1 branch is disabled; thus, chk1Δ dun1Δ strains are exceptionally sensitive to HU (Schollaert et al., 2004). As the synthetic HU sensitivity of the chk1Δ dun1Δ strain is similar to that of the ccr4Δ dun1Δ strain, we generated ccr4Δ chk1Δ, chk1Δ dun1Δ and ccr4Δ chk1Δ dun1Δ mutant strains to assess whether CHK1 lies in the same genetic pathway as CCR4. We observed that the ccr4Δ chk1Δ double mutant was no more sensitive to HU than the ccr4Δ mutant and that the ccr4Δ chk1Δ dun1Δ triple mutant strain had a similar HU sensitivity to the ccr4Δ dun1Δ strain (Fig. 4C,D). These epistatic genetic relationships suggest that CHK1 and CCR4 may function together in a novel genetic pathway that contributes to the tolerance of replication stress.

Since Ccr4 deadenylase activity is thought to negatively regulate gene expression, we reasoned that deletion of potential Ccr4 targets might suppress the HU sensitivity of ccr4Δ strains. To identify potential suppressors of the ccr4Δ HU sensitivity we undertook an insertional mutagenesis screen in a ccr4Δ strain with a genomic transposon (Tn3-LEU2) library (Ross-Macdonald et al., 1997). From a pool of 300,000 Tn3-LEU2 transformants, we obtained 30 single colony suppressors of the ccr4Δ growth defect on HU. Sequencing of the transposon insertion sites identified 24 candidate suppressor genes, of which ten independent genes accounted for the entire suppressor pool (Table 2). The two strongest suppressors were crt1::Tn3-LEU2 (eight occurrences) and pbp1::Tn3-LEU2 (six occurrences). CRT1 encodes the major transcriptional repressor of DNA-damage-regulated genes, whereas PBP1 encodes a factor that interacts with Pab1, the major poly(A) binding protein in budding yeast, and is believed to regulate polyadenylation (Mangus et al., 1998). To confirm the strongest hits from the suppressor screen, independent deletions of CRT1 and PBP1 were constructed in the context of either ccr4Δ or ccr4Δ dun1Δ strains. Deletion of CRT1 suppressed not only the HU lethality of a ccr4Δ strain but also of a ccr4Δ dun1Δ mutant (Fig. 5A and supplementary material Fig. S1A). An independent deletion of PBP1 also suppressed the HU lethality of a ccr4Δ strain, although less so than crt1Δ (Fig. 5A).

We then tested whether deletion of CRT1 also suppressed the severe and irreversible hypersensitivity to HU of the ccr4Δ dun1Δ strain. The ccr4Δ dun1Δ crt1Δ triple mutant strain retained viability even after 24-hour exposure to HU (Fig. 5B). Consistent with this bypass effect, Rad53 activity did not reach the same magnitude of induction in a ccr4Δ dun1Δ crt1Δ strain compared with a ccr4Δ dun1Δ strain, although activity did still persist at prolonged time points after HU withdrawal. The incomplete bypass and the residual checkpoint activation in the ccr4Δ dun1Δ crt1Δ strain suggested that Ccr4 and/or Dun1 have additional targets that affect checkpoint recovery. However, elimination of the characterized Dun1 target Sml1 (Zhao and Rothstein, 2002) had little effect on ccr4Δ dun1Δ strain viability on HU (supplementary material Fig. S1A). We also observed that the crt1Δ mutation suppressed the HU sensitivity of a chk1Δ dun1Δ strain (supplementary material Fig. S1B). CRT1 therefore poses a common impediment to survival of replication stress in both chk1Δ dun1Δ and ccr4Δ dun1Δ strains, consistent with the inference that CHK1 and CCR4 appear to act together in the same pathway.

Based on the above observations, we postulated that in the absence of Ccr4 the CRT1 mRNA is deregulated, which subsequently leads to HU sensitivity. To identify potential alterations in CRT1 mRNA poly(A) tail length, wild type, ccr4Δ or ccr4-1 (i.e. the catalytic mutant allele, E556A) strains were treated or not with 100 mM HU for 90 minutes and then various mRNAs analyzed for poly(A) tail length using a ligase-mediated poly(A) (LM-PAT) assay (Salles et al., 1999). Compared with the wild type control strain, CRT1 mRNA poly(A) tails were substantially longer in the ccr4Δ and ccr4-1 strains (Fig. 6A). The poly(A) tail length of other mRNAs, including RNR2, PGK1 and ACT1 were also noticeably elongated in the ccr4 mutants. However, CRT1 mRNA had the longest poly(A) tail of all mRNAs examined in the ccr4Δ strain at 45 adenosine residues, compared with 12 adenosines residues in a wild-type strain (Fig. 6A). A known Ccr4 substrate, the PGK1 mRNA, had a poly(A) tail length of 27 adenosine residues in the ccr4Δ strain; RNR2 and ACT1 poly(A) tails had lengths of 27 and 22 adenosine residues, respectively. HU treatment did not appreciably affect poly(A) status of any of the transcripts tested. Consistent with the HU sensitivity of the caf1Δ/pop2Δ strain detected in our genome-wide screen and with the known physical association of Caf1/Pop2 with Ccr4, the CRT1 mRNA poly(A) tail was also elongated to 45 adenosine residues in a caf1Δ strain (Fig. 6C). By contrast, removal of the Pan2 poly(A) nuclease had little effect on CRT1 mRNA poly(A) tail length. The CRT1 mRNA poly(A) tail was greatly elongated to 70 adenosine residues in the ccr4-1 pan2Δ double mutant strain, in accord with the dual role for the Pan2/Pan3 and Ccr4 nucleases as the major regulators of poly(A) tail length (Tucker et al., 2001).

Deadenylation of the poly(A) tail is implicated in mRNA decay (Parker and Song, 2004) and, moreover, Ccr4 is implicated in transcriptional control as well as mRNA deadenylation (Collart, 1996; Tucker et al., 2001). We therefore used quantitative real time PCR to address the possible effect of ccr4Δ on CRT1 mRNA levels, both in the absence and presence of HU. Contrary to our initial expectations, CRT1 mRNA abundance in the ccr4Δ strain was actually reduced to 25% of that in a wild-type strain during vegetative growth, with a statistically insignificant increase to 55% after HU treatment (Fig. 6B; all comparisons with untreated wild-type control). CRT1 mRNA abundance was also lowered in a dun1Δ strain to 61% of wild-type levels, consistent with the known role for Dun1 in CRT1 transcriptional induction (Huang et al., 1998). CRT1 transcript levels in the ccr4Δ dun1Δ double mutant strain did not appreciably differ from either single mutant. After HU treatment, CRT1 mRNA in the ccr4Δ dun1Δ strain also increased somewhat, to approximately 50% of WT levels. In wild-type cells, we observed little induction of the CRT1 mRNA in response to HU, a result that is at odds with previous findings that the CRT1 mRNA is induced 1.9-fold by HU relative to ACT1 mRNA levels (Huang et al., 1998). This discrepancy may in part derive from the use of different strain backgrounds and different conditions [BY4741 and 100 mM HU in this study, versus W303 and 200 mM HU in previous work (Huang et al., 1998)]. In addition, we noticed that ACT1 mRNA abundance was consistently increased in response to HU in our wild-type strain, such that possible induction of the CRT1 mRNA is masked by normalization to ACT1 (Fig. 6A; T.B. and R.W., unpublished observations). Thus, although it is evident that Ccr4 modulates both the poly(A) tail length and abundance of the CRT1 mRNA, the observed decrease in CRT1 mRNA levels is not consistent with the observed genetic hyperactivity of CRT1 in a ccr4Δ strain.

Elongated poly(A) tails are associated with increased rates of translation (Preiss et al., 1998). To address possible countervailing effects of elongated poly(A) tails and concomitantly decreased CRT1 mRNA in the ccr4Δ strain, Crt1 protein levels were assessed with a polyclonal antibody to the N-terminus of Crt1 (Zhang and Reese, 2005). In contrast to the observed drop in CRT1 mRNA, Crt1 protein was not decreased during vegetative growth in a ccr4Δ strain when compared with wild-type controls (Fig. 6D). As expected, in both the wild-type and ccr4Δ strains, Crt1 protein levels were moderately increased upon HU treatment. However, HU did not elicit as great an increase in Crt1 in a dun1Δ strain, in accord with the known role for Dun1 in CRT1 transcriptional induction (Huang et al., 1998). Importantly, the induction of Crt1 protein by HU was restored in the ccr4Δ dun1Δ strain, suggesting that the effect of the ccr4Δ mutation is independent of the checkpoint itself. As a control for these experiments, Pgk1 levels were constant in all strains and conditions tested (Fig. 6D). Taken together, the above data suggest complex effects in the ccr4Δ strain, including a possible scenario in which failure to deadenylate the CRT1 mRNA increases translational capacity per CRT1 mRNA molecule.

As pbp1 alleles were also uncovered as partial suppressors of the ccr4Δ HU lethality, and as Pbp1 is a putative regulator of both polyadenylation and translation, we examined CRT1 mRNA poly(A) tail length in a panel of gene deletion mutants, including pbp1Δ strains (Fig. 6E). However, no difference in CRT1 mRNA poly(A) tail length between wild-type and the pbp1Δ mutant was detected. In addition, no difference in CRT1 mRNA poly(A) tail length was apparent between ccr4Δ and ccr4Δ pbp1Δ strains. Thus, it does not appear that pbp1-mediated suppression of ccr4Δ HU lethality occurs at the level of the CRT1 poly(A) tail. Since Pbp1 is a ribosome-associated factor (Fleischer et al., 2006) and has been previously implicated in 3′ end-dependent translation (Tadauchi et al., 2004), we posit that the suppression of ccr4Δ HU sensitivity upon removal of Pbp1 is mediated through decreased translation of the CRT1 transcript in the ccr4Δ pbp1Δ strain.

Our genetic results suggested that deregulated CRT1 activity may cause severe HU lethality in ccr4Δ dun1Δ cells. To test this notion, we integrated a GAL1-CRT1 allele into ccr4Δ and dun1Δ strains and assessed the HU sensitivity of each strain on repressive and inducing media. As expected, repression of CRT1 on glucose medium suppressed the HU lethal phenotypes of both ccr4Δ and dun1Δ strains (Fig. 6F). Conversely, galactose induction of CRT1 hypersensitized both the ccr4Δ and dun1Δ strains, such that both were compromised even at low concentrations of HU. Overexpression of CRT1 had only a modest effect on the HU sensitivity of a wild-type strain, indicating that robust mechanisms normally keep CRT1 activity in check. These genetic data suggest that DUN1 and CCR4 represent two independent genetic pathways that repress CRT1 activity during replication stress.

Since Crt1 is a transcriptional repressor, we employed genome-wide expression profiles to identify putative Crt1 targets. We compared the transcriptional profiles of ccr4Δ and ccr4Δ dun1Δ strains to those of analogous strains in which CRT1 had been deleted, i.e. ccr4Δ crt1Δ and ccr4Δ dun1Δ crt1Δ strains in competitive hybridizations against genome-wide DNA microarrays (Fig. 7A,B). In each genetic context, deletion of CRT1 resulted in a dramatic upregulation of its known target genes, including RNR2, RNR3 and RNR4 (Fig. 7A,B; Table 3). Furthermore, HU-induced transcription of both RNR2 and RNR4 was severely impaired in a ccr4Δ dun1Δ strain, much more so than in either dun1Δ or ccr4Δ strains (Fig. 7B). Removal of CRT1 thus enables ccr4Δ and ccr4Δ dun1Δ strains to respond to and cope with replication stress through heightened induction of RNR genes. Although it has been previously reported that RNR3 is induced in a ccr4Δ strain upon addition of HU (Traven et al., 2005), our transcriptional profiles did not corroborate this result (Table S2 in supplementary material). A recent direct analysis of RNR transcripts in a variety of Ccr4-Not complex mutants concurs with our genome-wide expression profiles (Mulder et al., 2005).

The set of approximately 16 to 37 genes (depending on genetic context) induced more than twofold in the absence of CRT1 all represent candidate modulators of HU resistance in either a ccr4Δ or a ccr4Δ dun1Δ strain. To date we have been unable to identify any single multi-copy or overexpression suppressor of the ccr4Δ defect, in either direct tests with this gene set or in suppression screens with a 2-μm plasmid genomic library (data not shown). As shown in Table 3, the ORF most responsive to CRT1 deletion was YMR279c, which is also induced by MMS (Gasch et al., 2001). YMR279c putatively encodes a putative 12 transmembrane-spanning protein of unknown function but with homology to drug transporters. However, deletion of YMR279c did not affect sensitivity to HU in any background tested (data not shown).

The above genetic and gene expression data suggested that a severe defect in RNR subunit expression might be responsible for HU hypersensitivity of ccr4Δ dun1Δ cells. To test this idea, RNR2 or RNR4 were expressed from the conditional GAL1 promoter, either alone or in combination with RNR3 expressed from the constitutive GAPDH (GAP) promoter. Expression of either RNR3 alone or co-expression of RNR2 and RNR4 together from heterologous promoters only slightly improved the viability of a ccr4Δ dun1Δ strain on low concentrations of HU, whereas combined expression of GAP-RNR3 and GAL1-RNR4 further augmented survival. The combined heterologous expression of GAL1-RNR2, GAP-RNR3 and GAL1-RNR4 significantly improved viability on HU, although this bypass was still not as strong as with the crt1Δ mutation itself (Fig. 7C). In contrast to the partial bypass of the ccr4Δ dun1Δ strain HU sensitivity upon heterologous expression of RNR3, expression of RNR1 from the GAP promoter failed to improve survival of a ccr4Δ dun1Δ strain on HU (supplementary material Fig. S1C). These data suggest a defect in coordinated RNR2, RNR3 and RNR4 transcription, substantially contributing to the observed inviability of the ccr4Δ dun1Δ strain in the presence of HU. The differential expression of other Crt1 target genes may also contribute to the survival of the ccr4Δ dun1Δ crt1Δ strain on HU.

Through a systematic screen of the budding yeast gene deletion collection for HU-sensitive strains we have discovered that Ccr4, the catalytic component of the mRNA deadenylase complex, plays a crucial role in the response to replication stress. Other recent reports have also implicated Ccr4 in resistance to HU-induced replication stress (Bennett et al., 2001; Westmoreland et al., 2004; Traven et al., 2005). Because our genetic analysis shows that mec1, rad53 and dun1Δ mutants, but not chk1Δ mutants, are hypersensitized to HU upon deletion of CCR4, it appears Ccr4 operates in parallel to the Mec1-Rad53-Dun1 pathway to resolve DNA replication stress, perhaps in concert with Chk1. Consistently, we note that CCR4 and RAD9 act in the same epistasis group for resistance to ionizing radiation (Westmoreland et al., 2004) and that Rad9 physically interacts with Chk1 after DNA damage (Sanchez et al., 1999); together with our findings, these results suggest that Chk1 and Ccr4 act in the same branch of the replication checkpoint.

Our genetic and biochemical analysis implicates Ccr4 in the regulation of Crt1 function, and suggests that Crt1-mediated repression of transcription poses a major impediment to survival after exposure to HU. Crt1 activity is known to be negatively regulated by Mec1/Rad53/Dun1-dependent phosphorylation after DNA damage or replication stress, which serves to prevent Crt1 from recruiting the Ssn6-Tup1 repressor complex to target promoters (Huang et al., 1998). Prompted by our analysis of Crt1 transcriptional targets, we found that combined expression of RNR2, RNR3 and RNR4 can substantially bypass the HU sensitivity of the ccr4Δ dun1Δ double mutant. It has been recently reported that RNR gene induction after DNA damage is defective in cells disrupted for Ccr4-Not complex function (Mulder et al., 2005). The severe sensitivity of the ccr4Δ dun1Δ strain to low levels of HU suggests that Dun1 and Ccr4 converge to regulate Crt1 and promote RNR gene induction (see Fig. 8 for proposed model of Crt1 regulation).

The mechanism whereby Ccr4 limits Crt1 activity appears multifarious. As one might have predicted, CRT1 mRNA has a longer poly(A) tail in a ccr4Δ strain than in a wild-type strain. The known consequences of longer poly(A) tails are an increase in translation rate (Colgan and Manley, 1997; Preiss et al., 1998) and often an increase in mRNA stability (Tucker et al., 2001). In ccr4Δ strains, however, we found that CRT1 mRNA is actually decreased to approximately 25-50% of wild-type levels. However, we did not observe a corresponding decrease in Crt1 protein abundance in the ccr4Δ strain; indeed, under conditions of replication stress, Crt1 appeared to be subtly increased in both ccr4Δ and ccr4Δ dun1Δ strains. An increased translation rate of the CRT1 mRNA may underlie these effects; this notion is supported by isolation of pbp1Δ as a strong suppressor of ccr4Δ HU sensitivity. We note that CRT1 mRNA is normally of very low abundance in wild-type cells (Huang et al., 1998) (this study), which may be linked to its unusual sensitivity to poly(A) tail length. Regardless of the precise mechanism that hyperactivates CRT1 in the absence of CCR4, it is clear that the Crt1 level critically affects checkpoint function, RNR regulation and survival of checkpoint mutants (Huang et al., 1998).

Since Ccr4 regulates the polyadenylation state of myriad mRNAs (Muhlrad and Parker, 2005), at present it is not clear whether Ccr4 and poly(A) metabolism play a specific regulatory role - opposed to a passive role - in balancing various gene activities, including that of CRT1. It is possible that Ccr4 activity itself is regulated by replication stress either directly via post-translational modification or indirectly through the modulation of certain mRNAs species by RNA-binding proteins. The fact that CHK1 and CCR4 may act in the same pathway might argue for a specific regulatory function. Identification of crucial cis-elements on CRT1 mRNA that regulate its translation or stability will help address its role during replication stress.

In fission yeast, overexpression of the Cid13 protein, a cytoplasmic poly(A) polymerase, rescues the HU sensitivity of many checkpoint mutants and, conversely, its deletion results in synergistic HU sensitivity (Saitoh et al., 2002). These phenotypes are also consistent with the notion that cytoplasmic poly(A) tail elongation of one or more crucial target genes results in replication-checkpoint defects. Although suc22 (the Schizosaccharomyces pombe equivalent of RNR2) was identified as a candidate target for Cid13, it is unclear whether the shorter poly(A) tail of the suc22 mRNA in cid13Δ mutant is specific for this transcript or whether it is the result of a more global response to the cid13Δ mutation. All told, these observations suggest a conserved link between mRNA metabolism and replication-stress tolerance in eukaryotes.

In addition to the requirement for proper poly(A) metabolism, our results also point to a connection between transcription and replication stress. Our insertional mutagenesis screen uncovered two partial loss-of-function alleles of MOT1, which encodes an essential ATP-dependent transcription factor that regulates binding of Spt15 (TBP) to TATA boxes in target promoters (Adamkewicz et al., 2001). A recent proteome-wide approach identified Mot1 as a physical interacting partner of Crt1 (Gavin et al., 2002). Of particular interest, Ccr4-associated Not gene products have previously been functionally linked to Mot1 in the regulation of transcription (Collart, 1996). The isolation of mot1 alleles that suppress the HU sensitivity of the ccr4Δ strain provides additional evidence that transcriptional defects compromise the ability to cope with replication stress (Mulder et al., 2005). Interestingly, Ccr4 is implicated in transcriptional initiation through genetic and physical interactions with the Paf1-RNA-polymerase-II complex and a paf1Δ strain is also sensitive to HU (Chang et al., 1999; Betz et al., 2002). Defective transcriptional activation at Crt1-repressed target promoters, through either ineffective transcriptional initiation or impaired chromatin remodelling, may contribute to the HU sensitivity of the ccr4Δ strain.

While evidently complex to interpret, our results emphasize the inter-connected nature of the network that controls the response to replication stress (see supplementary material Fig. S2 for a diagram of all known network interactions). The DNA damage response relies on appropriately controlled checkpoint signal transduction, DNA repair, DNA replication, oxidative metabolism, transcription, mRNA metabolism and cell cycle control, all of which must be balanced against one another to ensure optimal genome stability. That non-obvious and often subtle changes can perturb crucial cellular responses (Davierwala et al., 2005) is epitomized by the unanticipated connections between mRNA metabolism and the response to replication stress. Determining just how Crt1 and Ccr4 establish an appropriately poised transcriptional and post-transcriptional program to enable tolerance of replication stress will require further systems level interrogation of the response.

Standard methods were used for yeast culture and genetics (Amberg, 2005). All strains were congenic with BY4741 (S288c MATa) (Brachmann et al., 1998), except where noted as congenic with W303 MATa, and are listed in supplementary material Table S3. All plasmids are listed below. Mutant CCR4 alleles were generated with the QuikChange system (Invitrogen) and cloned into EcoRI and XhoI sites of pRS416. The plasmids pGAP-RNR3 (pBAD079, TRP1 CEN GAP-RNR3); pGAP (BAD054, TRP1 CEN, empty vector); pGAP-RNR1 (pBAD070, TRP1 CEN GAP-RNR1); pGAL (pDL054, URA3 CEN); pGAL-RNR4 (pDL57, URA3 CEN GAL1-RNR4) were provided by Steve Elledge (Harvard University). pGAL-FLAG (MT3164, LEU2 CEN) and pGAL-RNR2-FLAG (MT3963, LEU2 CEN GAL1-RNR2-FLAG) were generated with the Gateway (Invitrogen) vector recombination system (Ho et al., 2002). Integrations of GAL1 promoters at genomic loci, and specific gene replacement with Schizosaccharomyces pombe his5+MX knockout cassettes were performed as described previously with the pFA6 family of plasmids (Longtine et al., 1998).

The haploid non-essential yeast deletion set (4812 strains) was arrayed by robotic pinning of colonies onto XY (YEPD + 100 mg/L adenine + 200 mg/L tryptophan) + 2% glucose medium containing 100 mM HU. Colonies were grown for one day at 30°C and subsequently re-pinned to XY-GLU+100 mM HU for another day of growth. Viability was assessed by visual comparison of colony size between exposure to drug and no drug and a relative score was assigned to arrayed colonies. All potential HU-sensitive strains were confirmed by streaking on XY-GLU or XY-GLU+100 mM HU to assess growth. HU sensitivity was confirmed using a spot test of serially-diluted cells on 100 mM HU (see below).

For assays on solid media, 5 μl of tenfold serially diluted cultures (starting OD₆₀₀ of 1.0) were spotted on XY-plates containing either 2% glucose (GLU) or 2% galactose (GAL), in the presence or absence of indicated concentrations of HU. For liquid assays, cells were grown to an OD₆₀₀ of 0.2 and HU was added to a final concentration of 200 mM. For determinations of colony-forming units (CFU), aliquots were removed from cultures at indicated intervals and spotted on XY-GLU plates with a Whitley Automatic Spiral Plater (WASP). Cells were grown at 30°C and CFU determined with WASP counting tables.

Cultures were grown in XY-2% glucose to an OD₆₀₀ of 0.3 and total RNA was prepared by glass-bead lysis in phenol (Tyers et al., 1993). 50 μg total RNA was reverse transcribed to cDNA with SuperScript II (Invitrogen) using a poly(dT) oligonucleotid and directly labelled with either dCTP conjugated to Cy3 or Cy5. Differentially labelled experimental and control cDNA preparation were pooled, heated to 65°C for 2 minutes, then cooled to 50°C. Hybridization to oligonucleotide microarray was performed overnight at 37°C in 20:1:1 DIG-Easy Hybe (Roche):yeast tRNA(10 mg/ml):ssDNA (10 mg/ml). Arrays were rinsed in 1 ×SSC 0.1% SDS at 50°C followed by two washes of 15 minutes in 1×SSC 0.1% SDS. Arrays were rinsed three times in 0.1×SSC to remove SDS and gently spun dry. Scanning was performed with a GenePix 4000B (Axon Instruments). Spots were identified and quantitated with Quantarray 1.0 software (GSI Lumonics). Expression ratios were normalized and centred on the median by Quantarray Data Handler 3.0, and analyzed with AFM 4.0 (Breitkreutz et al., 2001). For real time PCR analysis, 50 μg total RNA was reverse transcribed and 1/50 of the reverse transcriptase (RTase) reaction was serially diluted twofold for analysis of ACT1 and CRT1 levels with a Taqman 7300 Real Time PCR system (Applied Biosystems) and standard conditions, using the following primers for ACT1: FOR: 5′-TGGATTCCGGTGATGGTGTT-3′; REV: 5′-TCAAAATGGCGTGAGGTAGAGA-3′ ACT1 probe: TET fluorophore 5′-CTCACGTCGTTCCAATTTACGCTGGTTT-3′ For CRT1: FOR: 5′-CCGCCAGCATCACACACTTA-3′; REV: 5′-ACGATGATTTGCTCGCTATGG-3′ CRT1 probe: FAM fluorophore 5′-TTGCCCCCCATGTCAGTGAATATCCC-3′.

For immunoblot anaysis, 10 ml of culture was grown in XY-2% GLU to an OD₆₀₀ of 0.3, centrifuged, and the pellet was frozen. Indicated samples were treated with 100 mM HU for 90 minutes. Cell pellets were lysed with glass beads in 20% TCA (Fischer). Extracts were resolved by 8% SDS-PAGE and transferred to nitrocellulose, followed by blocking with 5% skimmed milk in TBS-0.05% Tween (TBST). Crt1 protein levels were detected with a rabbit polyclonal antibody raised against the N-terminus of Crt1, provided by Joseph Reese (Zhang and Reese, 2005), and donkey anti-rabbit IgG-HRP secondary antibody. Control Pgk1 protein levels were detected with mouse monoclonal anti-yeast Pgk1 IgG (Molecular Probes) and sheep anti-mouse IgG-HRP. Rad53 (DAB001) was detected with an affinity-purified polyclonal rabbit anti-Rad53 antibody and goat anti-rabbit-HRP secondary antibody (Kanellis et al., 2003).

Rad53 kinase activity was determined with in situ kinase assay by [γ-³²P]ATP incorporation as described (Pellicioli et al., 1999). Briefly, cultures were treated with indicated concentrations of HU and lysed with glass beads in 20% TCA. Extracts were resolved by 8% SDS-PAGE and transferred to PVDF membrane. Proteins on the membranes were denatured with guanidine hydrochloride and re-natured overnight at 4°C with gentle shaking. Membranes were incubated with [γ-³²P]ATP to detect Rad53 kinase activity in situ (Pellicioli et al., 1999).

Asynchronous cultures were grown to an OD₆₀₀ of 0.2. Synchronization in early S-phase was achieved by treatment of mid-log phase cultures with 200 mM HU for 3 hours. HU was removed by washing cell pellets twice with fresh media lacking drug. Cell were pelleted, re-suspended in 70% ethanol and processed for FACS analysis using the Sytox Green nucleic acid stain (Invitrogen) exactly as described in Kanellis et al. (Kanellis et al., 2003). Analysis was performed on a Becton-Dickinson FACScalibur flow cytometer.

Transposon insertion mutagenesis was carried out as described (Ross-Macdonald et al., 1997). Briefly, a ccr4Δ::KANMX MATa strain was transformed with 1 μg of NotI-digested transposon Tn3-LEU2 genomic library. After selection on SC minus leucine+100 mM HU for 3-5 days, putative suppressors were re-confirmed on 100 mM HU. Transposon insertions were isolated by RsaI digestion of genomic DNA followed by re-ligation. PCR was performed on ligated DNA to amplify DNA adjacent to transposon-specific sequences, followed by sequencing (Ross-Macdonald et al., 1997).

Ligase-mediated poly(A) tests (LM-PAT) were performed essentially as described (Salles et al., 1999). All strains were grown in rich media supplemented with adenine. Total RNA was isolated by the hot-phenol method and incubated with an excess of oligo(dT) primer, followed by addition of a linker oligo(dT-12) primer (5′-GCGAGCTCCGCGGCCGCGTTTTTTTTTTTT-3′) and T4 DNA ligase (Salles et al., 1999). PCR was performed as described by Salles et al. with primers designed within 100 bp upstream of the ORF stop codon. The following gene-specific primers were used for analysis: CRT1-PAT: 3′-GTCATGAAATTCGTCAATGGCC-3′; RNR2-PAT: 5′-GCCGGTGCTTTCACCTTCAACG-3′; PGK1-PAT: 5′-GATCTCCCATGTCTCTACTGGTGG-3′; ACT1-PAT: 5′-CCGCTTTGGCTCCATCTTCCATG-3′. PCR products were separated on 2% high-resolution agarose (Invitrogen) and visualized with ethidium-bromide staining followed by laser scanning (Fuji FLA-5100). Poly(A) tail lengths were calculated on the basis of a 100 bp ladder (NEB) using FUJI MultiGauge software. Sizes were determined from peak intensities.

We thank David Lee, Steve Elledge, Marco Foiani, Rodney Rothstein, Michael Snyder and Joseph Reese for generous gifts of strains, plasmids, libraries and antibodies, Howard Lipshitz for helpful discussions, and Danielle Dewar and Marc Angeli for technical assistance. This research was supported by grants from the Canadian Institutes of Health Research (M.T.) and the National Cancer Institute of Canada with funds from the Canadian Cancer Society (D.D.). T.P. acknowledges support from the Australian Research Council and the Sylvia and Charles Viertel Charitable Foundation. M.T. holds a Canada Research Chair (CRC) in Bioinformatics and Functional Genomics; D.D. holds a CRC in Proteomics, Bioinformatics and Functional Genomics; R.N.W. holds a Terry Fox Foundation/National Cancer Institute of Canada Doctoral Studentship (Biomedical); M.A.C. holds a Canada Graduate Scholarship (Doctoral) from the Natural Sciences and Engineering Research Council of Canada; T.H.B. holds an NH and MRC Howard Florey Centenary Fellowship.