Evidence from many organisms indicates that the conserved RecQ helicases function in the maintenance of genomic stability. Mutation of SGS1 and WRN, which encode RecQ homologues in budding yeast and humans, respectively, results in phenotypes characteristic of premature aging. Mutation of SRS2, another DNA helicase, causes synthetic slow growth in an sgs1 background. In this work, we demonstrate that srs2 mutants have a shortened life span similar to sgs1 mutants. Further dissection of the sgs1 and srs2 survival curves reveals two distinct phenomena. A majority of sgs1 and srs2 cells stops dividing stochastically as large-budded cells. This mitotic cell cycle arrest is age independent and requires the RAD9-dependent DNA damage checkpoint. Late-generation sgs1 and srs2 cells senesce due to apparent premature aging, most likely involving the accumulation of extrachromosomal rDNA circles. Double sgs1 srs2 mutants are viable but have a high stochastic rate of terminal G2/M arrest. This arrest can be suppressed by mutations in RAD51, RAD52, and RAD57, suggesting that the cell cycle defect in sgs1 srs2 mutants results from inappropriate homologous recombination. Finally, mutation of RAD1 or RAD50 exacerbates the growth defect of sgs1 srs2 cells, indicating that sgs1 srs2 mutants may utilize single-strand annealing as an alternative repair pathway.
THE use of model organisms has greatly contributed to the identification of some of the underlying causes of aging. In particular, the budding yeast Saccharomyces cerevisiae has proven to be an effective system for studying the mechanisms of cellular senescence. Many genes that affect yeast life span have been identified (reviewed in Jazwinski 1999). Of particular interest is SGS1, a member of the RecQ family of DNA helicases. Mutation of SGS1 results in a 60% reduction in mean life span (Sinclairet al. 1997). Furthermore, many phenotypes of yeast aging occur early in sgs1 mutants, including sterility, fragmentation of the nucleolus, and movement of the Sir proteins to the nucleolus. In humans, mutation of the RecQ homologues WRN, BLM, or RECQL4 results in genetic disorders characterized by increased genomic instability, a predisposition to certain types of cancers, and in the case of WRN, hallmarks of premature aging (reviewed in Karowet al. 2000). Therefore, a clear understanding of the cellular roles performed by Sgs1p in yeast could provide clues to the molecular basis of these diseases.
Sgs1p has demonstrated 3′-5′ DNA helicase activity (Bennettet al. 1998). It binds preferentially to branched DNA substrates, including duplex structures with 3′ single-stranded overhangs and DNA junctions with multiple branches (Bennettet al. 1998). In vitro, it is more efficient at unwinding G-G paired DNA compared to duplex DNA (Sunet al. 1999). Yeast cells lacking Sgs1p have a mitotic hyperrecombination phenotype, with increases in both intra- and interchromosomal homologous recombination (Wattet al. 1996). This increased recombination occurs both at the highly repetitive ribosomal DNA (rDNA) locus and at other sites in the genome (Gangloffet al. 1994; Wattet al. 1996). Because the rDNA is G-rich, one proposed function for Sgs1p is the prevention of recombination within the rDNA, possibly by the resolution of G-G paired DNA duplexes (Sunet al. 1999).
Intrachromosomal recombination within the rDNA can result in the formation of extrachromosomal ribosomal DNA circles (ERCs), which have been shown to be a cause of aging in yeast (Sinclair and Guarente 1997). Therefore, it was proposed that Sgs1p promotes longevity in yeast by decreasing the rate of recombination and the formation of ERCs. However, recently published data demonstrate that sgs1 mutants do not accumulate ERCs more rapidly than wild-type cells (Heoet al. 1999), suggesting additional mechanisms may promote premature aging in sgs1 cells.
Further insight into the role(s) of Sgs1p comes from its interactions with the topoisomerases. Topoisomerases are able to relieve torsional stress in DNA double helices and separate intertwined DNA molecules (Wang 1996). Sgs1p interacts with both Top2p and Top3p (Gangloffet al. 1994; Wattet al. 1995; Bennettet al. 2000). In addition, sgs1 mutations suppress the slow growth and genomic instability of top3 mutants (Gangloffet al. 1994) but cause slow growth in a top1 mutant background (Luet al. 1996). Conditional top2 mutants and sgs1Δ null mutants have elevated levels of mitotic and meiotic chromosome missegregation (Holmet al. 1989; Wattet al. 1995). Both Top2p and Top3p have been postulated to function in the decatenation of newly replicated chromosomes prior to their segregation (Gangloffet al. 1994; Spell and Holm 1994; Wattet al. 1995). Together, these data suggest that Sgs1p may be involved with one or more topoisomerases in the separation of chromosomes at the late stages of DNA replication (Watt and Hickson 1996).
Srs2p, like Sgs1p, is a DNA helicase with 3′-5′ polarity (Rong and Klein 1993). A mutation in SRS2 was initially identified as a suppressor of rad6 and rad18 UV sensitivity (Aboussekhraet al. 1989). Srs2p is proposed to control the sorting of spontaneous DNA lesions into the postreplication repair pathway (Schiestlet al. 1990; Ronget al. 1991). In the absence of Srs2p, these lesions are purportedly channeled into a RAD52-dependent homologous recombination pathway. In addition, SRS2 has been implicated in nonhomologous end-joining (Hegde and Klein 2000), single-strand annealing (Sugawaraet al. 2000), and the intra-S checkpoint (Liberiet al. 2000).
Yeast cells lacking both the Sgs1 and Srs2 helicases have been reported to be inviable in one study (Leeet al. 1999) and slow growing in another (Gangloffet al. 2000). Impressively, mutation of genes involved in the initial stages of homologous recombination, including RAD51, -55, and -57, suppresses these defects (Gangloffet al. 2000). Therefore, it has been suggested that Sgs1p and Srs2p are required either for the prevention of inappropriate recombination and/or for the processing of recombination intermediates.
We wished to further characterize the causes of premature aging in sgs1 mutants and determine whether Srs2p also plays a role in the promotion of wild-type longevity. To accomplish this, we have characterized the life spans of sgs1 and srs2 mutants in detail. Here, we report that both the sgs1 and srs2 mutants have a short life span that can be defined by two separate components. A majority of mutant cells stops dividing stochastically due to mitotic cell cycle arrest, probably due to unrepaired DNA damage. Cells that escape the arrest senesce prematurely due to normal aging processes, most likely caused by ERC accumulation. Genetic analysis using the sgs1 srs2 double mutant suggests that the mitotic arrest is due to the consequences of inappropriate homologous recombination and can be rescued by diverting DNA repair substrates into other pathways.
MATERIALS AND METHODS
Strains, plasmids, and media: The yeast strains used in this work are detailed in Table 1. The mutant rad5-535 allele in W303 has been characterized previously (Fanet al. 1996). The sgs1:HIS3 disruption removes nucleotides 317–3752 of the SGS1 open reading frame (ORF; Gangloffet al. 1994), while the sgs1Δ::hisG-URA3 disruption removes nucleotides 481–4026. SRS2 was disrupted by digesting plasmid pt2Δ2R (Aboussekhraet al. 1992) with PstI and then transforming yeast, thereby replacing nucleotides 528–1935 of the SRS2 ORF with the LEU2 gene. Disruption of sgs1 and srs2 was verified by PCR. RAD9 was disrupted by integration of a PCR product containing the kanamycin resistance gene (Wachet al. 1994) and selection on SC medium containing 300 μg/ml geneticin (GIBCO, Gaithersburg, MD). Disruption of FOB1 was accomplished by transformation with a PCR product containing the URA3 or TRP1 marker gene flanked by 40 nucleotides of FOB1 sequence. Integration of SIR2 at URA3 or LEU2 was accomplished by transforming cells with p305SIR2 or p306SIR2 digested with XcmI (Kaeberleinet al. 1999). The rad1Δ::HIS3, rad50Δ::HIS3, rad51Δ::HIS3, rad52Δ::HIS3, rad57Δ::HIS3, and rad59Δ::HIS3 null alleles are described in Park et al. (1999). The hdf1Δ::HIS3, dnl4Δ::HIS3, mre11Δ::HIS3, msh2Δ::HIS3, and rad18Δ::HIS3 alleles were constructed by PCR disruption, removing the entire ORF in all cases.
pSGS1f2 was constructed by PCR amplification of full-length SGS1 from the yeast genome, including 300 nucleotides (nt) upstream and 350 nt downstream of the coding region. The resulting 5.0-kb fragment was ligated into the polylinker of the ARS-CEN plasmid pRS316. The plasmid complemented an sgs1 deletion when tested for sensitivity to 0.015% methyl methanesulfonate in agar plates.
All yeast strains were cultured at 30° and grown on YEP medium containing 2% glucose. Diploids were sporulated in 1% potassium acetate medium for 2 days at 30°, and tetrad analysis was performed using standard methods.
Life span and terminal phenotype analysis: Micromanipulation and life span analysis were performed as described previously (Kaeberleinet al. 1999). Each experiment consisted of 45–50 mother cells and was carried out at least twice independently. Statistical significance was determined by a Wilcoxon rank sum test. Average life span is stated to be different for P < 0.05.
For terminal phenotype analysis, mother cells were observed at least 24 hr after their last division and scored as either unbudded, small budded (diameter of daughter cell less than one-fourth diameter of mother cell), or large budded (diameter of daughter cells equal to or greater than one-fourth diameter of mother cell). Although each life span and terminal morphology distribution figure represents data from a single experiment, each experiment was repeated at least two times with similar results.
Immunofluorescence: Cells were grown to midlog phase and fixed for 20 min in 3.7% formaldehyde. Fixed cells were incubated for 10 min in 0.1 m EDTA with 10 mm dithiotreitol and then washed twice with cold YPD containing 1 m sorbitol. Spheroplasts of fixed cells were obtained by treatment with 0.3 mg/ml zymolyase-100T (ICN Biomedicals) for 30 min at 30°. The cells were then analyzed by immunofluorescence as previously described (Millset al. 1999). DNA was visualized by staining with 4′,6-diamidino-2-phenylindole (DAPI), and microtubules were visualized using the anti-TUB2 antibody BIB2 (a gift from F. Solomon) diluted 1:200 in PBS/1% BSA/0.1% Triton X-100 and a Cy3-conjugated anti-mouse secondary antibody (Amersham, Piscataway, NJ). Digital images were obtained using a CCD camera controlled by OpenLab image acquisition software.
Cell cycle analysis: Strains were grown to early-log phase (OD600 = 0.3) and arrested with α-factor at a final concentration of 13 μg/ml for 3 hr. The α-factor was removed by two washes in YPD and 1-ml aliquots were placed immediately on ice at indicated time points. Cells were processed for fluorescence-activated cell sorting (FACS) as described in Mills et al. (1999), except all centrifugations were performed at 3000 × g to prevent cell clumping.
Short life span caused by sgs1 or srs2 mutations: Previously, it was demonstrated that sgs1 mutants have a mean life span that is ∼40% of wild-type strains (Sinclairet al. 1997). Therefore, if Srs2p is performing a function similar to Sgs1p, a similar reduction in life span would be expected for srs2 mutants. To test this, we performed life span analysis on isogenic wild-type, sgs1, and srs2 mutants. Mutation of SRS2 indeed causes a shortening of life span relative to wild type. The mean life span of an srs2 mutant is similar to that of an sgs1 mutant, 10.2 vs. 8.6 generations, respectively (Figure 1A).
Different groups have reported that sgs1 srs2 mutants are either slow growing (Gangloffet al. 2000) or inviable (Leeet al. 1999). We introduced sgs1 and srs2 null mutations into strains of opposite mating type, mated them, sporulated them, and analyzed the meiotic products. Approximately 55% of the time, the double mutant spores were inviable, forming microcolonies of between 2 and 100 cells. The other 45% of spores were able to form slow-growing colonies. Microscopic examination of the colonies revealed a high percentage of large-budded cells, suggesting that the slow growth was related to the unusual cellular morphology. When cells from these colonies were subjected to life span analysis, their average life span was approximately three generations (Figure 1A).
Cells lacking Sgs1p or Srs2p stop dividing with a terminal morphology distribution distinct from wild-type cells: We wished to characterize the reasons for the premature senescence of sgs1 and srs2 cells. Therefore, in addition to counting the number of times that each sgs1 mother cell divided, we observed the point in the cell cycle at which the mother cell ceased division (hereafter referred to as its terminal morphology). During the course of the life span analysis, we observed that a majority of sgs1, srs2, and sgs1 srs2 mother cells stopped dividing with large buds that could not be separated by micromanipulation (Figure 1B). This process appeared to be stochastic, occurring at a rate of ∼0.08 per division for the single mutants and 0.25 for the double mutant. In contrast, the majority of wild-type mother cells possessed an unbudded terminal morphology at the end of their life span.
To further investigate whether any cells in the sgs1 and srs2 populations senesce in a manner analogous to wild-type cells, we examined the terminal morphologies of both wild-type and mutant cells that had completed various percentages of their normal life spans. The average maximum life span of wild type, as defined by the decile of cells with the greatest number of divisions, is ∼32 generations. The average maximum life span of sgs1 and srs2 mutants is ∼18 generations. Therefore, we examined the terminal morphology distribution, in both mutant strains, of cells that had stopped dividing either early (≤75% of average maximum life span) or late (>75% of average maximum life span; Figure 1C). The wild-type terminal morphology distributions were nearly identical for both populations of cells, with most cells halting division as unbudded or small budded (Figure 1C). However, in sgs1 and srs2 mutant cells that stopped dividing early, a majority of cells ceased division with large buds. In striking contrast, late-generation mutant cells were usually unbudded or small budded when they senesced, similar to wild type. This analysis suggests that the distribution of terminal phenotypes observed with sgs1 and srs2 cells is composed of two different components. The first component is a G2-arrest that is stochastic and age independent, while the second is a G1-arrest that is age dependent and may be related to wild-type yeast aging.
A portion of sgs1 and srs2 mutant cells senesce due to normal causes of aging: Because the terminal morphology distribution of late-generation sgs1 and srs2 cells is remarkably similar to the distribution of wild-type cells, we wished to determine whether this subpopulation of helicase mutants was dying due to normal causes of aging. The average and maximum life spans of a wild-type haploid yeast strain can be extended by either deleting the FOB1 gene or by introducing a second copy of the SIR2 gene (Defossezet al. 1999; Kaeberleinet al. 1999). In fob1 mutants, the life span extension is correlated with a decrease in the accumulation of ERCs, and a similar mechanism of extended longevity has been proposed in the case of SIR2 overexpression.
Therefore, we tested the effects of the fob1 mutation and SIR2 overexpression in our sgs1 and srs2 strains. Previously, we reported that deleting FOB1 gave rise to only a modest increase in the average life span of the sgs1 mutant (Defossezet al. 1999). However, those experiments were confounded by the high rate of stochastic arrest we characterize here. In fact, a reexamination of the earlier data does suggest a possible selective extension in the life span of late generation sgs1 cells. Life span analysis performed upon several independent isolates of sgs1 fob1 mutants shows that the life span of sgs1 mutants was extended by the fob1 mutation (Figure 2A). However, the extension was primarily observed in the latter half of the life span curve, suggesting that deletion of FOB1 cannot extend the life span of cells that stop dividing due to the stochastic mechanism.
The overexpression of the SIR2 gene alone or in combination with the fob1 deletion extended the maximum life span of the sgs1 mutant in an identical manner. These life span extensions were reproducible and statistically significant (Figure 2B). Analogously, the latter half of the srs2 survival curve was extended by deleting FOB1 or by overexpressing SIR2 (data not shown). These experiments offer strong evidence that the fraction of sgs1 and srs2 cells that avoid the stochastic arrest do age in a manner analogous to wild-type cells and that their life span can be extended by mutations that also extend wild-type life span.
Cell cycle arrest of sgs1 and srs2 mutants occurs in mitosis: To further characterize the large-budded phenotype, wild-type, sgs1, srs2, and sgs1 srs2 strains were grown to midlog phase, fixed in formaldehyde, and stained with DAPI to visualize DNA. Approximately 15% of the sgs1 srs2 cells were large budded, with a single nucleus at the neck of the mother cell or in a “bow-tie” conformation stuck between the mother and daughter cells (Figure 3). In the latter case, the nucleus had clearly not yet separated into two distinct nuclei, suggesting that the cells had arrested at a point in mitosis prior to anaphase. This phenomenon was also observed to a smaller extent in sgs1 and srs2 single mutants but was noticeably absent from the wild-type population.
To determine the stage of mitosis at which the cells arrested, we costained wild-type and mutant cells with DAPI and antitubulin antibodies to visualize the mitotic spindle. For those mutant cells that had bow-tie nuclei, the spindle was short and extended through the bud neck, characteristic of cells in metaphase (Figure 3). Therefore, it is likely that the early senescence observed in sgs1 and srs2 mutants is due to a mitotic checkpoint that is triggered either prior to or at the beginning of chromosome segregation.
The mitotic arrest of sgs1 and srs2 mutants is RAD9 dependent: The cause of the cell cycle arrest in sgs1 and srs2 mutants could conceivably be due to DNA damage and/or failure to complete DNA synthesis. In budding yeast, the G2/M DNA damage checkpoint requires the RAD9 gene (Weinert and Hartwell 1988). Therefore, if the cell cycle arrest observed in sgs1 and srs2 mutants is due to the presence of DNA damage, deletion of RAD9 should alleviate the arrest phenotype. Indeed, when a rad9 null mutation was introduced into the sgs1, srs2,or sgs1 srs2 strains and life span analysis was carried out, we observed that the rad9 mutation partially suppressed the aberrant terminal morphology distribution of sgs1, srs2, and sgs1 srs2 mutants (Figure 4A).
However, the rad9 mutation did not suppress the life span defects of these strains (Figure 4, B and C). In fact, deletion of RAD9 further shortened the life span of sgs1 and srs2 single mutants. This may be a result of DNA damage that is not repaired prior to completion of mitosis in the sgs1 rad9 and srs2 rad9 cells, leading to cell death. Interestingly, the rad9 mutation slightly increased the life span of an sgs1 srs2 mutant (Figure 4C). Perhaps in sgs1 srs2 mutants, irreversible mitotic checkpoint arrest due to DNA damage occurs very frequently. Mutation of RAD9 bypasses this checkpoint, allowing the cells to divide one or more times, thereby extending their life span.
To further confirm that the primary cause of mitotic arrest in sgs1 srs2 cells is due to DNA damage and not to a gross defect in DNA replication, we monitored cell cycle progression by FACS in populations of cells that were synchronized in G1 by exposure to α-factor. Wild-type and sgs1 and srs2 single mutants progressed through S phase by ∼40 min after release from α-factor arrest (Figure 5). After ∼75 min, many cells had progressed through mitosis, as demonstrated by an increase in the 1n peak. The double sgs1 srs2 mutant was also able to replicate the bulk of its DNA, although we cannot exclude the possibilities that replication was aberrant or incomplete in a manner that we cannot detect by FACS.
Closer inspection of the FACS plots suggests that many sgs1 srs2 cells are unable to progress through mitosis. An increase in cells with 1n DNA content is observed in the wild-type population at ∼75 and 150 min after release, indicating that they have undergone nuclear division. This increase is qualitatively less noticeable in the sgs1 srs2 mutant cells, in agreement with the observation that ∼25% of these cells arrest during mitosis at each cell division. In addition, a portion of both the single and double sgs1 and srs2 mutants have DNA content >2n at time points past 90 min. This could be due to cells that have adapted to the mitotic arrest and are attempting to replicate their DNA in spite of having not completed cell division. Interestingly, rad9 sgs1 srs2 cells showed a persistent 1n DNA peak. The most probable explanation for this is the presence of dead cells in the culture that underwent an aberrant mitosis prior to α-factor arrest.
Mutation of genes involved in homologous recombination suppresses sgs1 srs2 slow growth: The bow-tie appearance of the nuclei in terminally arrested sgs1 srs2 cells suggests that the chromosomes are physically unable to separate. Conceivably, this could occur if the double mutant cells attempt to repair DNA damage using homologous recombination, but are unable to complete the process. By this logic, the cell cycle defect in sgs1 srs2 cells should be rescued by mutation of genes involved in homologous recombination, including RAD51, RAD52, RAD55, and RAD57. To test this idea, we constructed diploid strains heterozygous for sgs1, srs2, and either rad51, -52, or -57 mutations. Following sporulation and tetrad dissection, we determined the genotypes of the resulting haploid spores. Strikingly, mutation of RAD51, -52,or-57 fully suppressed the slow growth of an sgs1 srs2 mutant to the level of the corresponding rad single mutant (Figure 6A). In addition, nearly 100% of the triple mutants were viable, as compared to ∼45% of the sgs1 srs2 double mutants (Table 2). These results are in good agreement with recently published data (Gangloffet al. 2000). Finally, FACS analysis using sgs1 srs2 rad51 and sgs1 srs2 rad52 strains demonstrated that these triple mutants were fully able to progress through mitosis without arresting (data not shown). Thus, we postulate that the process of homologous recombination is responsible for the formation of aberrant DNA structures that cause sgs1 srs2 mutant cells to arrest during mitosis.
Identification of genes with synthetic effects in sgs1 srs2 mutants: The increased viability and growth rate of sgs1 srs2 rad51 strains implies that there are pathways, in addition to homologous recombination, that sgs1 srs2 cells can use to survive. We hypothesized that these additional pathways might involve alternative methods to repair double-strand breaks (DSBs), such as nonhomologous end-joining (NHEJ) or single-strand annealing (SSA). To identify such pathways, we crossed various mutations in genes important for other DNA repair pathways into sgs1 srs2 and sgs1 srs2 rad51 (or rad52) mutant backgrounds. The triple and quadruple heterozygous mutants were then sporulated and the meiotic products were analyzed. Because sgs1 srs2 mutant spores obtained in this manner are viable ∼45% of the time, the effects of additional mutations could be readily deduced by scoring their effects on the percentage of viable spores.
First, we tested the hypothesis that sgs1 srs2 cells could be using a NHEJ pathway to repair DSBs and promote viability. Genes specific to this pathway include HDF1 and DNL4 (Milneet al. 1996; Wilsonet al. 1997). Mutating either of these genes in an sgs1 srs2 background had no effect on the percentage of viable spores obtained (Table 2). In addition, mutation of DNL4 in an sgs1 srs2 rad52 background did not affect the suppression of the sgs1 srs2 slow growth by rad52 (Figure 6B). Therefore, sgs1 srs2 cells cannot be using NHEJ as a significant alternative to homologous recombination for repair of DSBs.
Next, we investigated the effects of mutating other genes involved in recombination. RAD50, MRE11, and XRS2 are known to be important for a variety of repair pathways, including homologous recombination resulting in gene conversion, SSA, and NHEJ (reviewed in Haber 1998). Upon sporulation of diploids triply heterozygous for the sgs1, srs2, and rad50 or mre11 mutations, we were unable to obtain viable triple mutants (Table 2). Spores with this inferred genotype divided to form microcolonies of 1–100 cells that stopped dividing with large buds. Given that srs2 rad50 double mutants grow extremely slowly (Hegde and Klein 2000), it is perhaps not surprising that the triple sgs1 srs2 rad50 and sgs1 srs2 mre11 mutants are inviable. As rad50 and mre11 are reported to be important for adaptation from G2/M arrest after DSB formation by HO endonuclease (Leeet al. 1998), perhaps the sgs1 srs2 rad50 (or mre11) mutants quickly lose viability due to permanent cell cycle arrest.
We also probed the effects of mutating RAD1, RAD59, and MSH2, three genes that are important for the removal of 3′ nonhomologous tails during both gene conversion and SSA (reviewed in Paques and Haber 1999). Strikingly, we found that sgs1 srs2 rad1 mutant spores were never viable, while mutation of RAD59 resulted in a statistically significant reduction in the percentage of viable spores (P < 0.05; Table 2). Furthermore, sgs1 srs2 msh2 spores that were viable formed smaller colonies than their sgs1 srs2 counterparts (Figure 6C).
The synthetic lethality of sgs1 srs2 with the rad1 and rad50 mutations was confirmed by transforming the appropriate heterozygous diploids with a plasmid containing SGS1 and URA3 (pSGS1f2). After sporulation, several independent sgs1 srs2 rad1 (or rad50) segregants bearing the plasmid were unable to form colonies on 5-fluoroorotic acid (5-FOA) medium (data not shown). Because 5-FOA is lethal to cells expressing the URA3 gene product, we conclude that sgs1 srs2 rad1 and sgs1 srs2 rad1 mutants die because they are unable to lose the SGS1 plasmid. Thus, RAD1 and RAD50 are required for viability in an sgs1 srs2 mutant.
RAD1 is required for DSB repair by SSA (Ivanov and Haber 1995), while RAD59 and MSH2 are important for the efficiency of SSA (Sugawara et al. 1997, 2000). In addition, mutation of RAD50 significantly impairs the kinetics of SSA (Ivanovet al. 1996). Thus, our results can be explained by a model in which sgs1 srs2 mutants use SSA as an alternative to homologous recombination for DSB repair (Figure 7). By this model, sgs1 srs2 cells that attempt to repair DSBs by homologous recombination occasionally arrest permanently due to an inability to either reverse or complete the process. However, sgs1 srs2 mutants that manage to repair the lesion by SSA are fully able to process the SSA intermediate and do not arrest. Alternatively, RAD1, RAD50, and RAD59 may be required for the processing of homologous recombination intermediates after the action of RAD51. In their absence, sgs1 srs2 cells process DSBs to an intermediate that cannot be resolved, resulting in permanent cell cycle arrest.
Surprisingly, introduction of rad1 or rad50 mutations into sgs1 srs2 rad51 or sgs1 srs2 rad52 backgrounds had no effect on viability or growth rate (Figure 6D). The quadruple mutants are presumably lacking in both homologous recombination and in SSA, but form large colonies. This suggests that other pathways for DSB repair, in addition to SSA, can be mobilized in the absence, but not in the presence, of RAD51/52-dependent homologous recombination in sgs1 srs2 mutants (Figure 7).
Two components of sgs1 and srs2 aging: The data presented here argue that aging in cells lacking the Sgs1 DNA helicase may be more complex than initially proposed (Sinclairet al. 1997). Careful analysis of the terminal morphology distribution of sgs1 mutants (Figure 1) reveals that the life span curve of an sgs1 mutant is a composite of two different phenomena. The early part of the curve can be closely approximated by a stochastic decline in viability, which we believe is due to cell cycle arrest in G2/M. The latter part of the sgs1 survival curve is composed mostly of cells that stop dividing in G1, suggesting that they senesce due to factors other than the mitotic arrest. Deletion of another DNA helicase, SRS2, results in a similar shortening of life span. The extension of life span observed in late generation sgs1 and srs2 cells when FOB1 is deleted or SIR2 is overexpressed argues that they are aging in a manner analogous to wild-type cells.
Previously, we suggested that sgs1 mutants age prematurely due to a faster rate of accumulation of ERCs (Sinclair and Guarente 1997). In fact, we and others have now found that old wild-type and sgs1 cells that have divided an equal number of times have approximately equivalent levels of circles as quantitated by Southern blotting (Heoet al. 1999; our unpublished data). In sgs1 mutants, if DSBs in the rDNA are largely processed by SSA, the result could well be an increase in recombination, as measured by marker loss, without a concomitant increase in ERCs. In addition, sgs1 mutants, rather than accumulating more ERCs, may be hypersensitive to ERCs and senesce at levels that are tolerable in wild-type cells. A synthetic effect between the sgs1 mutation and ERCs is reasonable, since both could place a burden on the DNA replication machinery of the cell. It is still likely that ERCs play a role in aging in SGS1 wild-type cells, since, first, the ectopic generation of ERCs shortens life span (Sinclairet al. 1997), and, second, deleting FOB1 both reduces ERCs and extends life span (Defossezet al. 1999).
The effect of a rad9 mutation on sgs1 and srs2 life spans: RAD9 comprises one of two branches of a checkpoint pathway that operates to halt cell cycle progression in the presence of DNA damage (Weinert and Hartwell 1988). We observed that deletion of RAD9 was able to partially suppress the cell cycle arrest observed in sgs1 and srs2 cells. However, the viability of sgs1 rad9 and srs2 rad9 cells was reduced relative to their RAD9 counterparts. This is consistent with a model in which lack of Sgs1p or Srs2p creates DNA damage that is recognized by a RAD9-dependent checkpoint. The damage could be due to incomplete DNA replication (Frei and Gasser 2000; Liberiet al. 2000) or replication fork pausing (Chakraverty and Hickson 1999), which has been shown to cause DSBs in Escherichia coli (Michelet al. 1997).
Upon detection of a DSB, the ensuing cell cycle arrest in RAD9 cells provides sgs1 and srs2 mutants with a period of time in which to repair the DNA damage and resume cell division, presumably with a fully intact genome. Indeed, during life span analysis of both sgs1 and srs2 strains, some cells were observed to temporarily arrest and resume cell division up to several hours later. In the absence of Rad9p, sgs1 and srs2 mutants do not arrest and divide with unrepaired DNA damage. In many cases, this probably results in catastrophic loss of genetic information and cell death.
Interestingly, the very short life span of the sgs1 srs2 strain was actually extended by a few generations by deletion of RAD9. We suggest that this extension occurs because sgs1 srs2 cells that would otherwise terminally arrest in G2/M are able to divide one or more additional times in the absence of Rad9p. However, these cells quickly die due to catastrophic DNA damage that is left unrepaired in the absence of the RAD9 checkpoint.
Possible roles for Sgs1p and Srs2p in homologous recombination: Both DSBs and stalled replication forks, which can be processed into structures that topologically resemble Holliday junctions (Seigneuret al. 1998), can be repaired through a recombinational repair pathway. Sgs1p binds to cruciform structures, including synthetic Holliday junctions, with high affinity and unwinds them (Bennettet al. 1999). Therefore, Sgs1p may function, in parallel with Srs2p, to disrupt these Holliday junctions and thereby prevent promiscuous recombination at stalled replication forks or DSBs (Figure 7). In addition, Sgs1p may be required to process the Holliday junctions and resolve the recombination intermediates. Since srs2 mutants have a greater tendency to repair DSBs by homologous recombination (Schiestlet al. 1990), the inability of sgs1 srs2 cells to disrupt or process these intermediates will be exacerbated.
We propose that the mitotic arrest observed frequently in the sgs1 srs2 mutant is due to the processing of DSBs by the homologous recombination machinery into chromosomal intermediates that cannot be separated prior to mitosis. These irresolvable structures could be the cause of the bow-tie nuclei observed in G2/M-arrested sgs1 srs2 cells. In strong support of this model, mutation of genes involved in homologous recombination suppresses the slow growth and mitotic arrest of sgs1 srs2 cells (Gangloffet al. 2000; our data).
The observation that mutation of genes important for SSA, including RAD50, MRE11, RAD1, and RAD59, causes synthetic lethality (or reduced spore viability in the case of RAD59) in an sgs1 srs2 background suggests that in the absence of the two helicases, cells may use SSA to repair DSBs. SSA differs from homologous recombination in that Holliday junctions are not formed in SSA (Linet al. 1984). Thus, Sgs1p would not be required for the resolution of SSA intermediates. This model predicts that Sgs1p could play a particularly important role in the rDNA and at telomeres, where the potential for homologous recombination is high. Cells with sgs1 mutations have an ∼7-fold higher rate of marker loss within the rDNA (Gangloffet al. 1994). Intriguingly, sgs1 srs2 mutants have a 40-fold increase in rDNA marker loss (our unpublished data). This increase is partially suppressed by mutation of RAD51 and is further reduced by mutation of both RAD1 and RAD51. Thus, in cells lacking Sgs1p, perhaps DSBs at the rDNA are processed by SSA instead of by homologous recombination, resulting in deletions and an increased rate of marker loss.
The robust growth of an sgs1 srs2 rad1 rad52 strain, in which homologous recombination and SSA are absent, demonstrates that additional pathways can be recruited for the repair of DSBs. These pathways are not available to sgs1 srs2 cells except in the absence of homologous recombination, since sgs1 srs2 rad1 cells (deficient in SSA) are inviable. Possibly, in an sgs1 srs2 mutant background, DSBs are initially channeled into the homologous recombination and SSA pathways. If these options are unavailable, cells may employ other repair pathways that operate with slower kinetics.
In summary, the data presented here argue that the short life span of sgs1 and srs2 mutants is largely due to an age-independent permanent mitotic arrest that is the result of promiscuous homologous recombination. These findings raise doubts about the classification of sgs1 as a true premature aging mutant. Nonetheless, further study using the sgs1 mutant as a model may provide important clues about the underlying causes of human diseases, such as Werner and Bloom syndromes, which are caused by mutations in the RecQ homologues.
We thank Peter Park for yeast strains and Brad Johnson, Dave McNabb, and other members of the Guarente lab for discussion and suggestions that contributed to this research. We also thank Angelica Amon and Jim Haber for assistance in data interpretation, Frank Solomon for antitubulin antibodies, and Francis Fabre for the srs2 disruption construct. This work was supported by grants from the National Institutes of Health, the Seaver Foundation, the Ellison Medical Foundation, and the Linda and Howard Stern Fund to L.G. H.A.T. is supported by the Helen Hay Whitney Foundation.
Communicating editor: A. G. Hinnebusch
- Received September 25, 2000.
- Accepted January 10, 2001.
- Copyright © 2001 by the Genetics Society of America