Loss of Yeast Peroxiredoxin Tsa1p Induces Genome Instability through Activation of the DNA Damage Checkpoint and Elevation of dNTP Levels

Peroxiredoxins are a family of antioxidant enzymes critically involved in cellular defense and signaling. Particularly, yeast peroxiredoxin Tsa1p is thought to play a role in the maintenance of genome integrity, but the underlying mechanism is not understood. In this study, we took a genetic approach to investigate the cause of genome instability in tsa1Δ cells. Strong genetic interactions of TSA1 with DNA damage checkpoint components DUN1, SML1, and CRT1 were found when mutant cells were analyzed for either sensitivity to DNA damage or rate of spontaneous base substitutions. An elevation in intracellular dNTP production was observed in tsa1Δ cells. This was associated with constitutive activation of the DNA damage checkpoint as indicated by phosphorylation of Rad9/Rad53p, reduced steady-state amount of Sml1p, and induction of RNR and HUG1 genes. In addition, defects in the DNA damage checkpoint did not modulate intracellular level of reactive oxygen species, but suppressed the mutator phenotype of tsa1Δ cells. On the contrary, overexpression of RNR1 exacerbated this phenotype by increasing dNTP levels. Taken together, our findings uncover a new role of TSA1 in preventing the overproduction of dNTPs, which is a root cause of genome instability.

Loss-of-function studies in mice implicated an essential role of peroxiredoxins in antioxidant defense and tumor suppression [14]. Particularly, peroxiredoxin 1-knockout mice not only suffered from severe anemia due to oxidative stress, but were also susceptible to several types of malignant tumors [15]. Consistent with this, genome-wide screening revealed that yeast peroxiredoxin TSA1 was a strong suppressor of gross chromosomal rearrangements and spontaneous mutations [16,17]. In addition, a mutator phenotype was observed in yeast cells lacking one or more peroxiredoxins. The phenotype could be rescued by yeast peroxiredoxin Tsa1p or mammalian Prx1, but not by their active-site mutants defective for peroxidase activity [18,19]. In further support of a role of TSA1 in the maintenance of genome stability, many genetic interaction partners of TSA1 identified through synthetic genetic array analysis were components of DNA repair machinery or DNA checkpoints [20,21]. For example, TSA1 was found to interact genetically with REV1/ REV3 and OGG1, which are critically involved in translesion synthesis (TLS) and the repair of oxidative DNA damage, respectively [22,23]. However, the exact mechanism by which Tsa1p suppresses genome instability remains to be fully understood.
Intracellular dNTP levels are one important determinant of cellular response to DNA damage [24]. For yeast cells to survive DNA damage, increased dNTP production would be allowed to facilitate replication, but with a trade-off of high spontaneous mutation rate [25]. In other words, abnormally high dNTP levels are causally associated with genome instability [24,26].
We previously demonstrated that yeast Tsa1p is a housekeeping peroxiredoxin which sufficiently suppressed the mutator phenotype [18]. Although both an aberrantly high level of ROS and an imbalance in free radical contents, which is caused by compensational activation of other antioxidants such as Sod1p [27], could underlie the mutator phenotype of tsa1D cells, additional events subsequent to the disruption of TSA1 might also be influential in the induction of genome instability. In this study we asked whether perturbation of dNTP pools might contribute to the mutator phenotype observed in tsa1D cells. We then investigated the cause of dNTP pool expansion. Our findings suggested that constitutive activation of the DNA damage checkpoint and consequent overproduction of dNTPs are the root cause of genome instability in tsa1D cells.

Results
Deletion of DUN1, SML1, or CRT1 Modulates Sensitivity of tsa1D Cells to DNA Damage Yeast peroxiredoxin TSA1 was found to be a strong suppressor of mutations and gross chromosomal rearrangements [16][17][18]. In addition, further deletion of another gene involved in DNA repair or DNA checkpoints caused synthetic growth defect or lethality in tsa1D cells [21,22]. Bearing these findings in mind, here we sought to dissect the interaction of TSA1 with the DNA damage checkpoint and particularly the machinery of dNTP synthesis, in order to understand the role of TSA1 in the maintenance of genome stability.
We first examined the sensitivity of tsa1D cells to various DNA damaging agents. tsa1D cells were sensitive to hydroxyurea (HU), 4-nitroquinoline 1-oxide (4NQO) and ultraviolet (UV) irradiation ( Figure 1A, lanes 1 and 5; Figure 1B; Figure 1C and Figure 1D, lanes 1 and 2). Re-expression of TSA1 in tsa1D cells suppressed the sensitivity phenotype ( Figure 1C and Figure 1D, lane 3). This suppression required the catalytic cysteine (Cys47) of Tsa1p, but not the C-terminal cysteine (Cys170), pointing to the importance of the antioxidant property of Tsa1p in the protection against DNA damage ( Figure 1C, lanes [3][4][5]. The sensitivity of tsa1D cells to DNA damage prompted us to investigate further the genetic interactions between TSA1 and components of the DNA damage checkpoint. In light of the finding that TSA1 genetically interacts with DNA damage checkpoint genes DUN1 and SML1 [22], we chose these two genes and their effector CRT1 for further analysis. Dun1p is a checkpoint kinase that phosphorylates and regulates ribonucleotide reductase (RNR) inhibitor Sml1p [28]. Dun1p also inhibits Crt1p, a transcriptional corepressor of RNR, through phosphorylation [29,30]. Deletion of DUN1, SML1 or CRT1 in tsa1D cells exerted a significant impact on their sensitivity to HU, 4NQO and UV irradiation. Loss of DUN1 further sensitized tsa1D cells to H 2 O 2 , HU, 4NQO and UV ( Figure 1A, lanes 1, 2, 5 and 6). In support of the specificity of effect, this sensitization was reversed upon expression of TSA1 or DUN1 in tsa1D dun1D cells ( Figure 1D, lanes [6][7][8]. Conversely, loss of SML1 or CRT1 rescued the sensitivity phenotype of tsa1D cells to 4NQO and UV ( Figure 1A, lanes 5, 7 and 8). It is noteworthy that such reversion of sensitivity was not observed in cells treated with H 2 O 2 ( Figure 1A, lanes 5, 7 and 8), suggesting that the effect might be specific to DNA damaging agents and was not caused directly by ROS. These observations supported the notion that TSA1 interacts specifically with the DNA damage checkpoint in a manner that is not mediated directly through ROS.
Although the sensitivity pattern of the different mutant strains in the spot assay was highly reproducible, a more quantitative comparison of these strains is desired. Hence, survival curves of strains in the presence of 4NQO and UV were also obtained ( Figure 1B). Dose-dependent killing of the strains by 4NQO and UV was observed. At all doses tested, the degrees of sensitivity of different strains to either 4NQO or UV were in the same order as shown in the spot assay. In particular, the survival curves indicated a further enhancement of the sensitivity phenotype in tsa1D dun1D versus tsa1D cells and a suppression of sensitivity in tsa1D sml1D cells ( Figure 1B). Collectively, our results demonstrated that the survival of tsa1D cells under DNA damage was decreased upon deletion of DUN1, but enhanced when either SML1 or CRT1 was genetically disrupted.
We also compared the phenotypes of tsa1D cells and cells lacking Sod1p, another key antioxidant enzyme [31]. In contrast to the genetic interactions observed in tsa1D cells, deletion of DUN1, SML1 or CRT1 in sod1D cells enhanced its sensitivity to HU and 4NQO ( Figure 1E, lanes 6, 7 and 8). Thus, TSA1 and SOD1 interact with DUN1, SML1 and CRT1 through different mechanisms.
Deletion of DUN1, SML1, or CRT1 Modulates Mutator Phenotype of tsa1D Cells We next investigated whether compromising DNA damage checkpoint genes in tsa1D cells might also alter their mutator phenotype. In agreement with previous reports [16][17][18], tsa1D cells exhibited high rates of spontaneous mutations in both canavanineresistant (CAN R ) and 5FC-resistant (5FC R ) assays (Figure 2A, Figure 2B and Figure 2D, groups 1 and 5). On the other hand, deletion of DUN1 did not significantly affect spontaneous mutation rates (Figure 2A, Figure 2B and Figure 2D, groups 1 and 2), whereas loss of SML1 or CRT1 caused a mild increase in CAN R mutation rates in WT cells ( Figure 2A and Figure 2D, groups 1, 3 and 4). However, the disruption of DUN1, SML1 and CRT1 in tsa1D cells modulated the mutator phenotype in opposite directions ( Figure 2A and Figure 2B, columns 5-8). Whereas reduction of spontaneous mutation rates was observed in tsa1D dun1D cells (Figure 2A and Figure 2B, columns 5 and 6), deletion of SML1 or CRT1 in tsa1D cells significantly enhanced the mutator phenotype ( Figure 2A and Figure 2B, columns 1, 7 and 8). Complementation of the reduction of mutation rate in tsa1D dun1D cells by re-introduction of DUN1 or TSA1 further verified the specificity of effect ( Figure 2C, columns 6 and 8). Thus, the mutation rates of tsa1D cells correlated with the activity of the DNA damage checkpoint.
In all individual deletion mutants, tsa1D cells displayed the highest mutation rate ( Figure 2). We postulated that this might be attributed either directly or indirectly to the elevation of intracellular ROS levels in these cells [4]. If that is the case, challenging the other DNA damage checkpoint mutants with ROS might have an impact on the mutator phenotype. To test this idea, we treated the cells with low-dose H 2 O 2 and assessed the impact on CAN R mutation rates. Interestingly, mutation rates increased in WT and dun1D cells to comparable levels ( Figure 2D, groups 1 and 2). In contrast, a further increase in mutability was observed when SML1 or CRT1 was comprised ( Figure 2D, groups 1, 3 and 4). Although the mutation rates of sml1D and crt1D cells in the presence of H 2 O 2 were still not as high as that of tsa1D cells in the absence of H 2 O 2 ( Figure 2D, groups 3-5), our results did suggest

Author Summary
Peroxiredoxins are a family of antioxidant enzymes highly conserved from yeast to human. Loss of peroxiredoxin in mice can lead to severe anemia and malignant tumors, but the underlying cause is not understood. One way to derive new knowledge of peroxiredoxins is through genetic analysis in yeast. We have shown that loss of peroxiredoxins in yeast is associated with an increase in mutation rates. Here, we demonstrate that this elevation of mutation rates in yeast cells lacking a peroxiredoxin is due to increased production of deoxyribonucleoside triphosphates (dNTPs), the building blocks of DNA. Our findings suggest a new model in which compromised antioxidant defense causes accumulation of damaged DNA and activation of the DNA damage checkpoint. For yeast cells to survive DNA damage, dNTP production is increased to facilitate DNA replication, but at the price of high mutation rates. This new model could lead to a better understanding of human diseases including cancer. Figure 1. Influence of DUN1, SML1, or CRT1 deletion on the sensitivity of tsa1D cells to DNA damage and replicative stress. (A) Spot tests. Ten-fold serial dilutions of strains BY4741 (WT), dun1D, sml1D, crt1D, tsa1D, tsa1D dun1D, tsa1D sml1D, and tsa1D crt1D were spotted on YPD medium containing indicated doses of H 2 O 2 , HU or 4-NQO. Cells were exposed to the indicated dose of UV after plating. Plates were incubated for 4 days at 30uC. Experiments were repeated for six times and similar results were obtained. (B) Survival curves. Logarithmically growing yeast cells, BY4741 (WT), dun1D, sml1D, tsa1D, tsa1D dun1D, and tsa1D sml1D, in YPD were treated with indicated doses of 4-NQO for 90 min before plating on YPD agar. For UV treatment, cells were first plated onto YPD agar followed by UV irradiation at the indicated doses. Plates were incubated for 3 days at 30uC, and then counted for survival. The number of colonies from untreated plates was taken as 100%. Experiments were repeated for three times and similar results were obtained. (C) TSA1 catalytic cysteine mutant cannot complement the HU sensitivity in tsa1D cells. Tenfold serial dilutions of the indicated strains transformed with pRS415, pTSA1, pTSA1C47S, or pTSA1C170S were spotted on SC-Leu plates containing 2% glucose and the indicated doses of diamide or HU. Note that cells grew more slowly on SC plates than on YPD plates as shown in (A). (D) Influence of DUN1, SML1, or CRT1 deletion on the sensitivity of sod1D cells to DNA damage. (E) Complementation of drug sensitivity in tsa1D dun1D cells by TSA1 or DUN1. doi:10.1371/journal.pgen.1000697.g001 that ROS could differentially modulate the mutator phenotype of different mutants.

Loss of TSA1 Elevates Cellular dNTP Production
We next investigated the mechanism that underlies the correlation of DNA damage checkpoint activity in tsa1D cells with drug sensitivity and mutator phenotype ( Figure 1 and Figure 2). DUN1, SML1 and CRT1 are regulators of RNR, the rate-limiting enzyme in dNTP synthesis [32][33][34]. Considered together with the model that elevated dNTP levels are required for surviving DNA damage in yeast at the price of increasing mutation rates [24], we asked whether the mutator phenotype of tsa1D cells would be due to alteration in cellular dNTP production. Thus, we measured dNTP levels in our mutants. Surprisingly, tsa1D cells produced significantly more dNTPs than wild type (WT) cells ( Figure 3A, groups 1 and 5). The magnitude of dNTP overproduction in tsa1D cells was comparable to that in sml1D cells ( Figure 3A, groups 3 and 5), in which the removal of Sml1p activates RNR leading to the rise in dNTP levels [28].
To shed further light on the roles of dNTP production in the generation of mutator phenotype, we compared the dNTP levels in other mutant cells. As expected, dun1D cells produced less dNTPs than WT cells ( Figure 3A, groups 1 and 2), since Dun1p is required for phosphorylation and subsequent removal of the RNR inhibitor Sml1p [28]. Loss of CRT1 was also found to increase cellular dNTP production ( Figure 3A, groups 1 and 4), as Crt1p is a transcriptional corepressor of RNRs [29]. However, loss of DUN1 reduced cellular dNTP production in tsa1D cells ( Figure 3A, groups 5 and 6), whereas deletion of SML1 or CRT1 in tsa1D cells further increased dNTP levels ( Figure 3A, groups 5, 7 and 8). Noteworthily, increased production of dNTPs in tsa1D cells could be fully complemented by Tsa1p, but not by its catalytic cysteine mutant C47S ( Figure 3B). Thus, the antioxidant property of Tsa1p was likely required for preventing overproduction of dNTPs.
We then asked whether the reduction of dNTP pools in tsa1D dun1D cells would be associated with a further drop in intracellular ROS levels in the absence of DUN1. Interestingly, tsa1D dun1D cells exhibited a higher level of intracellular ROS over WT, tsa1D or dun1D cells ( Figure 3C, columns 1, 2, 5 and 6), suggesting that the mutator phenotype in tsa1D and tsa1D dun1D cells correlates directly with dNTP production, but not generation of ROS. On the other hand, loss of SML1 or CRT1 did not alter the ROS levels in either WT or tsa1D cells ( Figure 3C, columns 1, 3, 4, 5, 7 and 8). Thus, in addition to the accumulation of ROS, elevation of dNTP production might also contribute to genome instability in tsa1D cells.
While deletion of DUN1, SML1 or CRT1 has an impact on dNTP production, they are multifunctional proteins that might also affect other biological processes [28][29][30]. To address this concern, we modulated the production of dNTP more directly by overexpressing RNR1 gene in tsa1D and sod1D cells. This overexpression has previously been shown to elevate intracellular dNTP levels substantially [24,26]. Indeed, when we induced the expression of Rnr1-3MYCp in WT cells ( Figure 4A), the spontaneous mutation rate was increased ( Figure 4B, columns 1 and 2). Furthermore, overexpression of RNR1 also exacerbated the mutator phenotype in tsa1D and sod1D cells ( Figure 4B, columns [3][4][5][6]. This lent additional support to the importance of dNTP overproduction in the induction of genome instability.  If the mutability of tsa1D cells is indeed caused by dNTP overproduction, the mutations generated would primarily be base substitutions rather than large deletion and gross chromosomal rearrangements [24]. With this in mind, we examined the types of mutations arisen in tsa1D and other mutants (Table 1). We noted that the majority of mutations found in tsa1D cells were base substitutions (83.3%) and frameshifts (13.3%). Large deletions were very rare in tsa1D cells (3.3%) when compared with WT (13.3%) cells. In addition, all of the mutations found in tsa1D sml1D cells with high dNTP levels ( Figure 3) were base substitutions (90%) and frameshifts (10%), whereas more deletions (13.3%) were detected in tsa1D dun1D cells (Table 1) with low dNTP concentrations (Figure 3). In keeping with previous findings [16], relatively more deletions (10%) were also observed in sod1D cells (Table 1). Generally, base substitutions were more prevalent in the strain when dNTP levels were high (Figure 3), whereas the incidences of deletions correlated negatively with dNTP concentrations. Therefore, the mutation spectra of tsa1D and other strains are consistent with the notion that elevation of dNTP levels is the underlying cause of genome instability in the absence of TSA1.

Loss of TSA1 Activates DNA Damage Checkpoint Leading to Elevation in dNTP Production
Above we demonstrated the elevation of dNTP levels in tsa1D cells ( Figure 3). In addition, our results also indicated the genetic interaction of TSA1 with DNA checkpoint genes (Figure 1). This led us to further investigate whether elevated production of dNTPs in the absence of TSA1 might be explained by the activation of the DNA damage checkpoint. As a first step, we assessed checkpoint activation by examining the steady-state levels of Rad53p, the yeast ortholog of human CHK2 kinase whose phosphorylation and activation are pivotally involved in the control of checkpoint response to DNA damage [35,36]. Particularly, Rad53p is a master regulator of Dun1p, Sml1p and Crt1p [35].
In this analysis we included the sod1D control strain, in which the effectors of the Mec1p-dependent DNA damage checkpoint were previously shown to be downregulated [31]. Phosphorylated Rad53p species were more evident in tsa1D cells compared to WT and sod1D cells ( Figure 5A, lanes 1, 3 and 5; Figure 5B, lanes 1 and 2; Figure 6A, lanes 1 and 2; and Figure 7A, lanes 1 and 3). This difference became more pronounced in the presence of H 2 O 2 ( Figure 5A, lanes 2, 4 and 6; Figure 5B, lanes 5 and 6; Figure 6A, lanes 3 and 4; and Figure 7A, lanes 5 and 7).
The steady-state levels of phosphorylated Rad53p in DNA damage checkpoint mutants were also compared. Whereas deletion of DUN1 triggered phosphorylation of Rad53p ( Figure 5C, lanes 1 and 2), an observable increase in phosphorylated Rad53p species was not found in sml1D or crt1D cells ( Figure 5C, lanes 1, 3 and 4). Notably, loss of DUN1 in tsa1D cells further enhanced the activation of Rad53p ( Figure 5C, lanes 5 and 6), whereas tsa1D sml1D and tsa1D crt1D cells had similar levels of phosphorylated Rad53p compared to tsa1D cells ( Figure 5C, lanes 5, 7 and 8).
In addition to Rad53p, we also checked for the status of Rad9p, a more upstream transducer in the DNA damage checkpoint pathway [37]. Rad9-13MYCp was found to be activated in tsa1D cells ( Figure 6B, lanes 1, 2, 4 and 5) and this activation could not be reversed by the C47S mutant of Tsa1p ( Figure 6B, lanes 2, 3, 5  and 6). As a marker for DNA double-strand breaks (DSBs) [38], the level of cH2A was also found to be elevated in tsa1D cells as compared to WT ( Figure 5B, lanes 1 and 2; Figure 6A, lanes 1 and 2; and Figure 6B, lanes 1 and 2). This agrees with a recent report that tsa1D cells displayed an increased number of Rad52-YFP foci indicative of DNA damage [23]. The levels of cH2A in other DNA damage checkpoint mutant cells were also examined. Among dun1D, sml1D and crt1D cells, an elevation in cH2A level was only found in dun1D cells ( Figure 5C, lane 2). In addition, disruption of DUN1, SML1 or CRT1 in tsa1D cells did not affect cH2A levels significantly ( Figure 5C, lanes 5-8). Noteworthily, although phosphorylated Rad53p and cH2A were abundant in dun1D and tsa1D dun1D cells ( Figure 5C, lanes 2 and 6), their mutation rates remained low (Figure 2A and Figure 2B, columns 2 and 6) plausibly due to the low levels of dNTPs ( Figure 3A, groups 2 and 6). In other words, elevation of dNTP levels might be the direct cause of genome instability.
Consistent with the activation of Rad53p and Rad9p, the levels of Rad53p target Sml1-3HAp were diminished in tsa1D cells in the presence ( Figure 5B, lanes 5 and 6) and absence of H 2 O 2 ( Figure 5B, lanes 1 and 2). Importantly, all of the above changes in tsa1D cells could be fully complemented by TSA1 ( Figure 5B, lanes 2 and 3), but not by its C47S mutant ( Figure 5B, lanes 2 and 4). On the other hand, in agreement with previous findings [31], we did not observe a significant change of Sml1p level in sod1D cells (data not shown).
RNR is an important downstream effector of the DNA damage checkpoint which mediates the production of dNTPs [33,34]. Since the expression of RNR genes is transcriptionally activated in response to DNA damage [29], we used semi-quantitative RT-PCR to determine the relative levels of RNR1/2/3/4 transcripts in the presence and absence of HU. In this analysis we included an additional control termed HUG1, a target of Mec1p induced highly by DNA damage [39]. As shown in Figure 6C, RNR transcripts were induced to higher levels in tsa1D cells than in WT and sod1D cells. The induction of RNR1 and RNR3 was greatest in both untreated and HU-treated tsa1D cells. The level of HUG1 transcript was also elevated in tsa1D cells and this was more pronounced in the presence of HU ( Figure 6C). In sharp contrast, sod1D cells treated with HU showed a lower magnitude of induction of RNR1, RNR3 and HUG1 mRNAs ( Figure 6C). These results obtained from sod1D cells were generally consistent with previous findings [31]. Thus, the pattern of RNR induction in tsa1D cells was not ascribed to a general effect caused by the lack of any antioxidant enzyme, but was highly specific. Collectively, our results suggested that loss of TSA1 induces the activation of the DNA damage checkpoint leading to the induction of RNR and consequent overproduction of dNTPs.

Mutation of Rad53p Suppresses Genome Instability in tsa1D Cells
If activation of the DNA damage checkpoint in tsa1D cells is really important to the generation of genome instability caused by dNTP overproduction, genetic disruption of the checkpoint would be able to reverse the mutator phenotype of tsa1D cells.  To test this hypothesis, we employed a RAD53 mutant termed rad53AA, in which both T354 and T358 in the activation loop of Rad53p had been replaced by alanine, thereby abrogating the autophosphorylation activity in response to DNA damage [40]. This defective rad53AA allele similar to rad53-11 is thought to be associated with reduced dNTP production due to high abundance of Sml1p [40,41]. Thus, we set out to characterize the phenotypes of tsa1D cells carrying the rad53AA allele.
We next characterized the sensitivity of these mutants towards H 2 O 2 , UV and HU. TSA1 rad53AA cells were sensitive to HU ( Figure 7B, lanes 1 and 2) as previously described [40]; while tsa1D RAD53 cells were sensitive to H 2 O 2 , UV and HU ( Figure 7B, lanes 1 and 3) similar to tsa1D cells in BY4741 background (Figure 1,  lanes 1 and 5). Resembling tsa1D dun1D cells in BY4741, tsa1D rad53AA cells in W303 background displayed further sensitivity to H 2 O 2 and UV when compared to tsa1D RAD53 cells ( Figure 7B,  lanes 3 and 4).
We then looked at the effect of a defective DNA damage checkpoint on genome instability in tsa1D cells. Intriguingly, tsa1D rad53AA cells displayed a significantly reduced (,50%) rate of spontaneous 5FC R mutations over tsa1D RAD53 cells ( Figure 7C). On the other hand, both tsa1D rad53AA and tsa1D RAD53 cells had high levels of intracellular ROS over WT cells as measured by DCF fluorescence ( Figure 7D). These observations suggested that rad53AA mutation can suppress the mutator phenotype in tsa1D cells without affecting cellular redox environment. This generally agrees with the phenotypes of tsa1D dun1D cells (Figure 1 and Figure 2), lending further support to the concept that intracellular dNTP levels are an important determinant in the induction of genome instability in tsa1D cells.

Discussion
Here, we provided the first evidence that loss of yeast peroxiredoxin TSA1 causes genome instability through constitutive activation of the DNA damage checkpoint leading to overproduction of intracellular dNTPs. There are two salient points in our work. First, we demonstrated the elevation of dNTP levels in tsa1D cells and its direct correlation with the mutator phenotype ( Figure 2, Figure 3, Figure 4). Second, we demonstrated the activation of the DNA damage checkpoint in tsa1D cells in relation to elevated production of dNTPs ( Figure 1, Figure 5, Figure 6, Figure 7). Our findings suggested a new model for the role of peroxiredoxins in the maintenance of genome integrity, which has implications in the understanding of human diseases including cancer.
In agreement with our findings on the accumulation of cH2A and activation of the DNA damage checkpoint in tsa1D cells, several lines of evidence in the literature supported the role of Tsa1p and other peroxiredoxins in the protection of cells against DNA damage. First, tsa1D cells produce significantly more ROS [4], which cause DNA and protein damage [27,42,43]. Second, loss of TSA1 results in increased formation of Rad52-YFP foci, an indicator of DNA DSBs [23]. Third, tsa1D cells are highly sensitive to the functional state of DNA repair and checkpoints [22]. In particular, tsa1D is synthetically lethal with rad51D mutation, indicating that the viability of rad51D cells deficient in recombination repair requires TSA1 function [44]. Finally, human  Figure 5B. (B) Spot assay. Ten-fold serial dilutions of the indicated strains were spotted on YPD medium containing the indicated doses of H 2 O 2 or HU. Some cells were exposed to UV after plating. (C) RAD53 mutation suppresses mutator phenotype in tsa1D cells. Mutation rates were calculated as in Figure 2B. (D) ROS detection. DCF fluorescence was measured as in Figure 3C. doi:10.1371/journal.pgen.1000697.g007 peroxiredoxins have been implicated in cellular defense against oxidative DNA lesions [45]. In this context, the activation of the DNA damage checkpoint in tsa1D cells demonstrated in our study highlights the pivotal roles of the checkpoint in cell survival and provides an explanation for the synthetic lethality seen in various double deletion mutants involving TSA1 and another DNA repair or checkpoint gene [21].
Deletion of TSA1 in yeast cells has previously been shown to result in both a mutator phenotype and an increase in gross chromosomal rearrangements [16,22,23]. Although the causes and origin of gross chromosomal rearrangements remain poorly understood, oxygen metabolism and ROS production are implicated in the prevalence of these rearrangements in tsa1D cells [23]. Noteworthily, base substitution, but not chromosomal rearrangement, was the predominant type of mutation found in our analysis of mutation rates (Table 1). Thus, the major type of genome instability analyzed in our study is an increased rate of point mutations, but not gross chromosomal rearrangements involving more complex alterations such as translocations, large deletions and amplifications.
Our findings point to a role of dNTP levels in determining the mutation rate of tsa1D cells. Strong genetic interactions between TSA1 and four RNR regulators DUN1, SML1, CRT1 and RAD53 were observed in the context of sensitivity to DNA damage ( Figure 1 and Figure 7), spontaneous mutability ( Figure 2 and Figure 7) and dNTP production ( Figure 3). Although the catalytic cysteine of Tsa1p is required for the suppression of mutator phenotype, the mutability of tsa1D cells correlated directly with dNTP concentrations (Figure 2, Figure 3, Figure 4), but not with high ROS levels ( Figure 3 and Figure 7). One plausible explanation is that the loss of TSA1 might cause accumulation of both ROS [4] and DNA damage ( Figure 5). This activates the DNA damage checkpoint through Rad53p, Rad9p and Sml1p ( Figure 5, Figure 6, Figure 7) leading to transcriptional activation of RNR genes ( Figure 6) and elevated production of dNTPs ( Figure 3). Once at high dNTP levels, replicative and TLS polymerases by-pass DNA lesions more efficiently to promote survival, but only at the price of increasing mutation rates [24,25]. This model implicates dNTP pool expansion as the major culprit in the induction of genome instability in tsa1D cells. Indeed, reducing dNTP levels without affecting ROS production was sufficient to reverse the mutator phenotype of tsa1D cells (Figure 3). In particular, tsa1D dun1D cells have high levels of ROS ( Figure 3C), phosphorylated Rad53p ( Figure 5C) and cH2A ( Figure 5C). However, these cells showed a low mutation rate ( Figure 2C) because the dNTP levels were also low ( Figure 3A). On the contrary, increasing dNTP levels by overexpressing RNR1 aggravated the mutator phenotype ( Figure 4). Furthermore, point mutations but not deletions were predominantly found in tsa1D cells (Table 1), implicating a role for dNTP overproduction in compromising genome stability. In further support of this model, compromise of TLS polymerases also suppressed CAN R mutations in tsa1D cells [23].
We found that the levels of dNTPs in tsa1D cells were as high as those in sml1D cells ( Figure 3A). This finding revealed an unexpected role of TSA1 in the maintenance of dNTP pools in eukaryotic cells. We further observed transcriptional activation of RNR genes in tsa1D cells (Figure 6), which could be mediated through the activation of Rad53p checkpoint. Although this might provide an explanation for the overproduction of dNTPs, exactly how Tsa1p is mechanistically involved in regulating RNR expression remains to be further investigated.
Consistent with previous findings [26], elevation of intracellular dNTPs over a particular threshold level by overexpressing Rnr1p can sufficiently induce a mutator phenotype (Figure 4). Plausibly, the dNTP levels in tsa1D sml1D and tsa1D crt1D cells might have reached the threshold level causing a dramatically increased mutation rate (Figure 2 and Figure 3). When the elevation of dNTP levels have not reached the threshold as in the case of sml1D, crt1D and tsa1D cells, accumulation of intracellular ROS might serve to trigger or aggravate the mutator phenotype. In tsa1D cells, ROS levels were constantly high ( Figure 3C) causing severe DNA damage ( Figure 5). In contrast, ROS levels were low ( Figure 3C) and DNA damage was not detected ( Figure 5C) in sml1D or crt1D cells. This might explain the higher mutation rate in tsa1D cells versus sml1D or crt1D cells (Figure 2). Further exacerbation of the mutator phenotype of sml1D and crt1D cells by ROS such as H 2 O 2 ( Figure 2D) lent some credence to this model.
We demonstrated the requirement of the catalytic cysteine for the ability of Tsa1p to modulate dNTP production ( Figure 3). Through irreversible hyperoxidation, this residue can act as a redox sensor, which triggers the switch of peroxiredoxin from peroxidase to chaperone activity under stress [12,13]. In this connection, it would be of great interest to understand whether and how the chaperone activity of Tsa1p might be involved in the regulation of dNTP production.
Activation of Rad53p by upstream kinase Mec1p requires adaptor proteins Rad9p and Mrc1p [46,47]. We noted that Rad53p phosphorylation was dramatically increased in tsa1D versus WT cells ( Figure 6A and Figure 7A). In contrast, the increase in Rad9p phosphorylation in the absence of TSA1 was less pronounced ( Figure 6B). Although additional experiments are required to investigate the cause of this difference between Rad53p and Rad9p, one possibility is that the deletion of TSA1 might exert a stronger effect on Mrc1p activity.
Hypermutability or genome instability is a hallmark of cancer [48]. Mammalian Prx1 is a candidate tumor suppressor gene [15,49]. Because peroxiredoxins are highly evolutionarily conserved proteins, an understanding of the mechanism by which yeast Tsa1p protects cell from genome instability might derive novel insight into the tumor suppressive role of Prx1 in mammalian cells. Our work demonstrates the importance of high dNTP levels in the mutability of tsa1D cells. Further analysis of dNTP concentrations of Prx1-null mouse cells will reveal whether increased production of dNTPs might be a general mechanism for the generation of genome instability in higher eukaryotes.

Strains and Plasmids
S. cerevisiae strains BY4741 [50] and W303-1a, and their isogenic strains ( Table 2) were used. All knockout mutants were constructed by one-step gene deletion method [51]. Primers were listed in Table 3. Expression vector for DUN1 was derived from pRS415. Expression plasmids for TSA1 and its mutants have been described [18].

Measurement of Mutation Rates, dNTP, and ROS Levels
Rates of spontaneous forward mutations to confer CAN R or 5FC R were measured as described [18,53]. Spectra of CAN R mutations were determined by DNA sequencing. Ten independent cultures were analyzed in each experiment. Cell extracts were prepared and dNTP levels were measured with Klenow enzyme and [ 3 H] labeled dATP or dTTP (PerkinElmer) as described [54]. Standard curves were used to estimate the cellular dNTP levels. Three independent cultures were analyzed in each experiment. Intracellular ROS levels were measured by fluorimetry using DCF (Molecular Probes) as described [4,18].

RNA Analysis
Total RNA was extracted by phenol/freeze RNA preparation method as described [55]. For RT-PCR, 3 mg of total RNA was used for cDNA synthesis. Semi-quantitative PCR was performed and optimized to ensure that the amplification was in the linear range. PCR primers were listed in Table 3.

Western Blotting
Western blot analysis was performed essentially as described [31]. Yeast cells were harvested by centrifugation, followed by trichloroacetic acid extraction with the help of glass beads.