|Home | About | Journals | Submit | Contact Us | Français|
A cause of aging in Saccharomyces cerevisiae is the accumulation of extrachromosomal ribosomal DNA circles (ERCs). Introduction of an ERC into young mother cells shortens life span and accelerates the onset of age-associated sterility. It is important to understand the process by which ERCs are generated. Here, we demonstrate that homologous recombination is necessary for ERC formation. rad52 mutant cells, defective in DNA repair through homologous recombination, do not accumulate ERCs with age, and mutations in other genes of the RAD52 class have varying effects on ERC formation. rad52 mutation leads to a progressive delocalization of Sir3p from telomeres to other nuclear sites with age and, surprisingly, shortens life span. We speculate that spontaneous DNA damage, perhaps double-strand breaks, causes lethality in mutants of the RAD52 class and may be an initial step of aging in wild-type cells.
One of the hallmarks of aging in most organisms is that mortality rate increases exponentially with age (24). Because yeast cells divide asymmetrically, mother and daughter cells can be separated microscopically at each cell division, and such experiments reveal that mothers have a fixed division capacity, called their life span. A number of morphological changes occur as mother cells grow older: slowing of the cell cycle, enlargement of cell size, loss of mating ability, and accumulation of intracellular granules (64, 66, 87). The daughter cells from old mothers have a reduced life span potential, hinting that a dominant cytoplasmic senescence factor asymmetrically accumulates in old mother cells and that this factor can leak to daughter cells from old mothers (21, 44).
A genetic study has revealed that the allele of SIR4 affects life span (45). Null alleles cause a shortened life span, and a gain-of-function allele gives rise to an extended life span. The SIR2/3/4 gene products are normally positioned at telomeres and HM loci, where they mediate transcriptional silencing (35, 54). SIR2 also plays a role at the nucleolus, the site of repeated copies of ribosomal DNA (rDNA), to suppress recombination and mediate silencing (14, 32, 88). In aging cells, the sir complex at telomeres and HM loci relocates to the nucleolus (46). This relocalization is mimicked constitutively by the gain-of-function allele of SIR4 that extends life span (45). Thus, the relocalization of the Sir complex to the nucleolus extends life span in wild-type yeast cells.
Studies of the human WRN gene, recessive mutations in which cause the disease Werner syndrome (99), further support the close link between the nucleolus and aging. Individuals with Werner syndrome show symptoms of accelerated aging, including hair graying and loss, atherosclerosis, bilateral ocular cataracts, diabetes, and osteoporosis (23, 76). WRN has greatest homology with genes encoding DNA helicases of the RecQ family, including Saccharomyces cerevisiae SGS1 (30), Escherichia coli recQ (50, 68), Schizosaccharomyces pombe rqh1 (89), Xenopus laevis FFA-1 (98), and human BLM and RecQL (22, 73). The WRN protein has been demonstrated previously to have ATP-dependent DNA helicase activity and 3′→5′ exonuclease activity (33, 37, 81). Importantly, WRN protein is localized in the nucleolus in human cells (34, 58), suggesting that its role in promoting longevity may be linked to a nucleolar function.
The sgs1 mutation suppresses the slow growth and hyperrecombination at the rDNA caused by a top3 mutation, and Sgs1p interacts with both Top2p and Top3p (30, 96). The sgs1 mutation also causes genomic instabilities, including hyperrecombination at rDNA and other loci and a reduction in fidelity of both mitotic and meiotic chromosome segregation (30, 95, 96). Interestingly, like the WRN mutation, the sgs1 mutation accelerates aging: it decreases the average life span of yeast cells by 60% and accelerates the onset of age-associated phenotypes, including sterility and the redistribution of the Sir proteins from telomeres to the nucleolus (86). Sgs1p, like WRN protein, is concentrated in the nucleolus (86). In addition, expression of the WRN protein in the sgs1 mutant suppresses the hyperrecombination phenotype (97).
Microscopic analysis has revealed that the nucleolus in old mother cells is enlarged and fragmented (86). These changes are caused by the genomic instability in the tandem repeats of rDNA. Midway in the life span of mother cells, an extrachromosomal rDNA circle (ERC) is excised from the genome (85). Each ERC contains an ARS sequence, and plasmids containing such sequences autonomously replicate and segregate asymmetrically in mother cells (67). Thus, mother cells build up ERCs to very high levels, and daughters are ERC free (85). The release of a single ERC in young cells is sufficient to shorten life span by 40%, proving that ERC accumulation causes senescence. ERCs can leak into daughters of very old mothers, consistent with the view that they are the previously described senescence factor (21, 44).
Since the sgs1 mutant displays hyperrecombination at the rDNA, it is possible that cellular recombination mechanisms lead to the formation of ERCs. We thus sought to understand how ERCs were formed. Here, we analyze the effects of mutations that cause a defect in recombination on ERC formation and aging. Our findings show a link between ERC formation and the RAD52 pathway of homologous recombination. Further, our results suggest that DNA breaks might be an early event in the aging process, which then triggers the formation of ERCs and the release of the Sir protein complex from telomeres in aging cells.
Yeast strains used in this study are listed in Table Table1.1. All strains are isogenic. Strains were cultured at 30°C with standard media (82). For isolation of old cells, yeast extract-peptone-dextrose with 2.5% glucose was used as previous described (85). PPY74 (RDN1::ADE2) was constructed by transforming PSY316α with pDS40 (85). PPY16 and PPY103 were constructed by transforming PSY316a and PPY74 with pBS/SK-E1/E2-I3/I4-3ARU (a gift of A. Lau and S. Bell) cut with KpnI/NotI. This transformation replaced the region from the HMR-I to the HMR-E silencer with the URA3 gene (hmrΔ1::URA3), which is not silenced. PPY27 was constructed by transforming PPY16 with a gel-purified, BamHI/PstI ADE2 fragment from pURADE2 (85). PPY35 and PPY143 were made by transforming PPY27 and PPY103 with pPP46 cut with PshAI/AatII and uncut pRS315 (84). The transformed cells were first grown on a Leu− plate to select for cells that had acquired pRS315 and replica plated onto a Leu− 5-fluoroorotic acid plate to select for Ura− cells among the Leu+ cells. Ura− Leu+ cells were screened by PCR to check for the correct transformants (hmrΔ2::ADH1-GFP). PPY56 (sir3Δ::URA3) was constructed by transforming PPY35 with pDM42 (55). To construct PPY70 (sgs1Δ::hisG), after transformation of PPY35 with pPP69 cut with NotI, a correct transformant was passed over 5-fluoroorotic acid to eliminate the URA3 gene (10). PPY98 (rad52::LEU2) was constructed by transforming PPY35 with pSM20 (D. Schild).
Disruption of rad genes was carried out by the one-step transplacement method (7, 75). The following regions within coding sequences of rad genes were replaced with HIS3: from +40 to +3285 for rad1Δ, from +21 to +1648 for rad7Δ, from +49 to +3253 for rad26Δ, from +165 to +3843 for rad50Δ, from +1 to +1196 for rad51Δ, from +63 to +1488 for rad52Δ, from +43 to +1379 for rad57Δ, and from +39 to +668 for rad59Δ.
pPP46 was constructed by the following procedures. (i) pJR1426 (a gift of M. Foss and J. Rine) contains a 5.1-kb fragment of HMRα in a pRS316 backbone (84). The HMRα fragment contains the a1 and a2 genes replaced with the α1 and α2 genes but contains intact HMR-E and HMR-I silencers. pJR1426 was first cut with SacI/SmaI, and the ends were blunted with the Klenow fragment and ligated, resulting in pPP21. These procedures removed the linker sequences, including the XbaI site, between the two restriction sites that are outside of the HMRα fragment. (ii) pPP21 was digested with BclI/XbaI, which removed α1 and α2 genes, and ligated with polylinker insert cut with BclI/XbaI, resulting in pPP33. This polylinker insert, containing many restriction sites, was created by annealing and extending the following two oligonucleotides with Taq DNA polymerase: 5′-GCGCGTCGGCCGCTGATCAGTCGACTCGCGATCGATCCTAGGCTAGCGAATTCAGATCTTCCGGA-3′ and 5′-GCTGGC TCTAGAGCATGCGGCCGGTTAACCCGCGGTCCGGAAGATCTGAATT CGCTAGCCTAGGA-3′. (iii) pPP16 was constructed by inserting the HindIII fragment containing the soluble green fluorescent protein (GFP) (S65T and V163A mutant) gene amplified by PCR from pJK19-1 (43) into pSP400. pSP400 was constructed by moving the entire promoter and terminator of ADH1 in pDB20 (9) into pRS306 (84). (iv) pPP46 was made by inserting the entire promoter and terminator of ADH1 including soluble GFP of pPP16 cut with SmaI/XbaI into pPP33 cut with HpaI/XbaI.
pSGS12μ was constructed by inserting the NotI fragment containing the SGS1 coding sequence amplified by PCR from a cosmid into the NotI site of pDB20. pMM2 was created by replacing the region between bp +481 and +4026 of the coding region with a hisG::URA3::hisG fragment (3). The NotI fragment of pMM2 was cloned into the NotI site of pTKS(+) (38), producing pPP69.
Life span analysis was performed by counting the number of daughter cells that bud off from a virgin mother cell before cessation of cell division, as previously described (44). The sample size for each life span analysis was 43 to 51 cells. Each life span analysis was carried out at least two independent times.
Old cells were obtained as previously described (85), except that for some experiments Sulfosuccinimidyl-6-(biotinamido)-6-hexanimide hexanoate (Pierce, Rockford, Ill.) was used for biotinylation, instead of sulfosuccinimidyl-6-(biotinamido)hexanoate.They both gave a similar yield of old cells.
Immunofluorescence experiments were performed as described elsewhere (31, 46), except that anti-Sir3p used in this study was generated by immunizing a rabbit with the full-length Sir3p (60a). Optical sections of images were obtained with the CELLscan system (Scanalytics, Billerica, Mass.) as previously described (46). Strains that do not contain the GFP gene at HMR were used for these studies.
DNA used for gel analysis was prepared as previously described (85), except that no phenol extraction was performed. Total DNA (5 μg) for each sample was electrophoresed without ethidium bromide at 1 V/cm for 24 to 30 h. Young cells used in this experiment were free cells that were removed when old cells were magnetically sorted.
Young or old cells (about 2 × 106 cells) were resuspended in 1 ml of phosphate-buffered saline containing 5 μg of propidium iodide (PI) (Sigma, St. Louis, Mo.) per ml and sonicated briefly to separate the clumped cells. To ensure appropriate comparison with old cells, young cells were biotinylated and incubated with streptavidin-coated magnetic beads (PerSeptive Biosystems, Cambridge, Mass.). The only exception was sir3 cells that were from a log-phase culture. The magnetic beads present along with the cells did not interfere with the fluorescence-activated cell sorting (FACS) analysis; young cells with and without beads gave similar results. The level of green fluorescence of each cell was determined by using FACScan (Becton Dickinson, San Jose, Calif.). Dead cells were first excluded from the analysis by being stained with PI, which was measured with a 650 long-pass filter (15, 18). PI preferentially stains dead cells that have porous membranes and will not diffuse appreciably into intact cells. Then, the level of green fluorescence of 105 live cells was measured with a 530/30 band pass filter, and their fluorescence was displayed in a histogram with the CellQuest Analysis program (Becton Dickinson).
The significance of differences in mean life span between two strains was determined as previously described (45).
To investigate the possibility that ERCs are excised from the rDNA locus through intrachromosomal homologous recombination, we tested whether ERC formation requires RAD52, a gene needed for most homologous recombinational events (71, 83). Age-matched wild-type and mutant cells that had divided on average seven to eight generations were magnetically sorted, and their DNA was analyzed by gel electrophoresis. While the old wild-type cells clearly accumulated ERC species, ERCs were undetectable in the old rad52 cells (Fig. (Fig.1A).1A). A small amount of ERCs could have been present in the old rad52 cells and gone undetected. Thus, we also searched for ERCs in sgs1 rad52 double mutant cells. Although the old sgs1 cells accumulated slightly more ERCs than did the age-matched wild-type cells, the old sgs1 rad52 cells showed no detectable ERCs (Fig. (Fig.1A).1A). Thus, the RAD52 gene and, presumably, homologous recombination are required for the formation of ERCs.
Since ERCs are a cause of aging and the old rad52 cells do not accumulate ERCs, one might predict that the rad52 mutant would have a long life span. To test this possibility, we performed a life span analysis on rad52, sgs1, and sgs1 rad52 cells. Contrary to the prediction, the average life span of rad52 cells (average = 7.1 generations) was about 70% shorter than that of the wild-type strain (average = 23.5 generations) (Fig. (Fig.1B).1B). It was even shorter than the average life span of sgs1 cells (average = 9.8 generations). Interestingly, the sgs1 rad52 cells (average = 5.5 generations) had a slightly shorter average life span than the rad52 cells, indicating synthetic shortening of life span by each mutation.
rad52 cells lose chromosomes at an elevated rate (63). To determine whether chromosome loss was responsible for the premature death of rad52 cells, we compared the life spans of rad52 haploid cells and homozygous rad52 diploid cells. Chromosome loss in diploid cells should not lead to lethality. The wild-type diploids showed a life span similar to that of the wild-type haploids (Fig. (Fig.1C),1C), as previously reported (45, 65). The homozygous rad52 diploids (average = 7.5 generations) also displayed a life span similar to that of the rad52 haploids (Fig. (Fig.1C).1C). Thus, rad52 cells do not appear to die due to loss of essential genes caused by a chromosome loss.
Interestingly, 70 to 80% of both rad52 haploid and diploid mother cells ceased dividing as large-budded cells, while only 15 to 25% of wild-type cells arrested as large-budded cells. Therefore, most of the rad52 cells arrested at the G2/M phase of the cell cycle, perhaps due to a failure to adapt after DNA damage-induced checkpoint arrest (51, 77). These findings suggest that premature death in rad52 mutant cells could be caused by double-strand breaks (DSBs) which go unrepaired. In wild-type cells, these breaks would be repaired by using sister chromatids through homologous recombination. Moreover, the repair of breaks in the rDNA might also generate ERCs if repaired with another rDNA repeat on the same chromosome.
We then investigated whether other age-associated phenotypes are still present in cells that do not accumulate ERCs with age. One of the hallmarks of yeast aging is the gradual increase in the number of cells that lose silencing at the HM loci (87). We investigated the state of silencing in old rad52 cells. Because the previously used assay to determine age-specific phenotype of sterility is laborious (86, 87), we developed an assay to easily detect the state of silencing at the HM loci. We replaced the a1 and a2 genes at the HMR locus with the GFP gene driven by the constitutive ADH1 promoter (Fig. (Fig.2A).2A). We postulated that in young cells GFP expression would be silenced, while in old cells GFP would be expressed, giving rise to green fluorescence.
Indeed, GFP was efficiently silenced in young cells as measured by FACS. Young cells with GFP at HMR had a profile similar to that of cells without GFP, except for a small subpopulation of cells with slightly higher fluorescence (Fig. (Fig.2B).2B). GFP expression at HMR was, as expected, silenced in a Sir-dependent manner: young sir3 cells showed about 10-fold-higher fluorescence than did young wild-type cells. Moreover, in aging cells with GFP (average, seven to eight generations old), a subpopulation of the cells showed higher fluorescence, indicated by the rightward shift of the fluorescence histogram (Fig. (Fig.2B).2B). It is known that old cells become enlarged, which might cause an increase in autofluorescence. However, this does not account for the increase in fluorescence in old cells with GFP because the difference between young and old cells with GFP (4.95) is more than threefold higher than that between young and old cells without GFP (1.31). This assay is thus effective in determining the age-specific phenotype of loss silencing at HMR.
FACS analysis of the old rad52 cells that were on average seven to eight generations old also showed that a high proportion of cells (average fluorescence = 26.08) have lost silencing compared to the age-matched, wild-type cells (average fluorescence = 9.60) (Fig. (Fig.2B).2B). Again, the increase in the average fluorescence seen for old rad52 cells is not due to the enlargement of cells (Fig. (Fig.2).2). Thus, although devoid of rDNA circles, rad52 cells prematurely lose HMR silencing as they age.
Loss of HM silencing in old wild-type cells is likely due to the redistribution of Sir3 proteins from the telomeres and HM loci to the nucleolus (46). Thus, we examined Sir3p localization in old rad52 cells by indirect immunofluorescence with anti-Sir3p antibody. The nucleus is stained with DAPI (blue) (4′,6-diamidino-2-phenylindole), and the nucleolus is stained with anti-Nop1p (red) in this experiment. In young rad52 cells, Sir3p was found at three to seven bright perinuclear foci (green), characteristic of telomeres (Fig. (Fig.3).3). This pattern was indistinguishable from that observed for young wild-type cells. Old wild-type cells (about 18 generations old) showed a nucleolar relocalization of Sir3p (yellow in the merged image), as previously described (46). Distinct from the old wild-type cells, 20 to 30% of sorted, old rad52 cells (average, seven to eight generations old) showed a diffuse, nuclear pattern of Sir3p staining that included the nucleolus. About a third of the cells that displayed a diffuse, nuclear pattern showed many bright foci, some of which could be telomeric foci. The remaining cells showed a telomeric staining like young cells. We believe that those cells that showed a diffuse, nuclear pattern are cells that have reached the end of their life span. They are not likely to be dead cells because very old wild-type cells (18 generations old) did not give a similar staining. The pattern of nuclear staining is consistent with the movement of Sir3p away from telomeres and HM loci and could explain the premature loss of silencing seen at HMR (Fig. (Fig.2B).2B). We speculate that the Sir proteins, which play a role in DNA repair through nonhomologous end joining (11, 93), leave the telomeres and HM loci to repair DNA damage, perhaps DSBs, that occur elsewhere in old rad52 cells.
We then set out to determine if other genes important for homologous recombination played a role in ERC formation. RAD51 encodes a RecA homolog, and RAD57 shows RecA homology (42, 83, 90). Both display a partial defect in homologous recombination (2, 74). In contrast to the old rad52 cells, old rad51 and rad57 cells had a detectable level of ERCs, albeit lower than that of the age-matched wild-type cells (Fig. (Fig.4A).4A). Also, both rad51 (average = 13.0 generations) and rad57 (average = 12.5 generations) mutations shortened life span by about 40% (Fig. (Fig.4B).4B). The less severe shortening of life span seen for rad51 and rad57 mutants than for the rad52 mutant (70% shorter) (Fig. (Fig.1A)1A) correlates with their lesser degree of deficiency in homologous recombination. FACS analysis of age-matched rad51 (average fluorescence = 20.30) and rad57 cells (average fluorescence = 19.56) showed a premature loss of silencing compared to the age-matched wild-type cells (Fig. (Fig.2B).2B).
We also examined another member of the RAD52 epistasis group, RAD50, which plays a role in resection of broken ends by a 5′-to-3′ exonuclease activity during DSB-induced homologous recombination (41, 83). Strikingly, the rad50 mutant had a life span (average = 7.3 generations) similar to that of the rad52 mutant (average = 7.1 generations) (Fig. (Fig.4B4B and and1B)1B) and showed a low but visible amount of ERCs (Fig. (Fig.4A).4A). While indirect immunofluorescence assay performed on young rad50 cells showed a pattern of staining similar to that of young wild-type cells, old rad50 cells showed a diffuse, nuclear localization of Sir3p (Fig. (Fig.3).3). As in the rad52 mutant, a defect in homologous recombination may play a role in the shortening of life span in the rad50 mutant. Since RAD50 also has roles in illegitimate recombination, telomeric maintenance, and checkpoint function (4, 11, 62, 69), it is also possible that a disruption in these functions contributes to the shortened life span in the mutant.
We then investigated if the effect of mutations in DNA repair genes on longevity is restricted to mutants defective in homologous recombination. Another form of repair that applies to repeated DNA sequences is single-strand annealing (SSA) (36, 70). SSA occurs between homologous regions flanking a DSB, by annealing of complementary DNA after extensive 5′-to-3′ degradation extending away from the break (8, 25). RAD1 encodes an endonuclease that can remove nonhomologous single-strand ends of a DSB, and it is required for SSA (39, 40). The rad1 mutation did not have a significant effect on life span (Fig. (Fig.5),5), indicating that SSA is not necessary for normal life span.
Since RAD1 is also required for nucleotide excision repair (1, 72), we infer that this form of repair is also not germane to aging. Consistent with this conclusion, mutation in another gene involved in nucleotide excision repair, RAD7 (60, 72), did not affect life span (Fig. (Fig.5).5). Finally, the rad26 mutation, which causes a defect in transcription-coupled repair (94), also had no effect on life span (Fig. (Fig.5).5). In summary, the shortening of life span by mutations in the DNA repair genes examined is specific to those affecting homologous recombination.
In this paper, we have determined the effects of mutations in various DNA repair genes on the formation of ERCs and life span. Interestingly, mutations in RAD52, RAD50, and RAD51 (or RAD57), all of which affect homologous recombination, gave a total, severe, or partial reduction, respectively, in the formation of ERCs. Thus, the formation of ERCs requires the activity of the RAD52-dependent pathway of homologous recombination. Surprisingly, these mutations did not extend the life span of mother cells but, rather, shortened their life span.
Since ERCs are a cause of aging in wild-type mother cells, how can we explain the shortened life spans of these mutants? For the rad52, rad51, and rad57 mutants, the degree of shortening correlates well with the severity in the reduction of homologous recombination in these mutants. In fact, it is this reduction in homologous recombination that governs the lower rate of generation of ERCs in these mutants.
Surprisingly, the rad50 mutation, which has a modest effect or none on the intrachromosomal recombination rate (29, 36, 74), had a severe reduction in ERC accumulation. Heteroallelic interchromosomal recombination is increased by 10-fold in the rad50 mutant (56, 57). It is possible that RAD50, along with MRE11 and XRS2, regulates the balance of intrachromosomal and interchromosomal recombination events. In the absence of RAD50 function, the frequency of the interchromosomal events could increase at the expense of reduction in intrachromosomal events. If so, the severe reduction in ERC formation observed in the rad50 mutant may be due to such dysregulation. Since RAD50 also has roles in illegitimate recombination, telomeric maintenance, and checkpoint function (4, 11, 62, 69), it is possible that a disruption in these functions contributes to the reduction in ERC formation and shortening of life span, in addition to the disruption in homologous recombination.
Since DNA lesions such as DSBs are not repaired efficiently in the mutants defective in homologous recombination, these mutants are likely to be dying prematurely due to unrepaired DSBs. Consistent with this idea, unlike wild-type cells, most of the rad52 cells ceased dividing as large budded cells (G2/M). The premature death of rad52 cells does not appear to be due to chromosome loss, since the rad52 haploid cells lived as long as did the rad52 diploid cells. Introduction of two unrepairable DSBs that cannot be repaired through homologous recombination in wild-type cells causes an adaptation failure and a permanent G2/M arrest (51). We speculate that old rad52 cells cease dividing and permanently arrest at G2/M because of multiple unrepaired DSBs.
In wild-type cells, these DSBs and other lesions are repaired efficiently so that cells escape early death. However, as a by-product of those repair events, ERCs can be generated by homologous recombination in the rDNA, and these ERCs then carry out the proposed gradual aging program (Fig. (Fig.6).6). By this view, the generation of an ERC in a mother cell at once corrects the acute problem of a DNA break in the rDNA at the price of establishing the mortality of that mother cell lineage.
Mutations in other RAD genes were also examined but did not affect life span. These include mutants defective in SSA (rad1), nucleotide excision repair (rad1 and rad7), and transcription-coupled repair (rad26). Thus, the only DNA repair genes examined that shortened life span were a part of the RAD52 pathway of homologous recombination. Finally, mutation in the RAD52 homolog RAD59 (6), had little effect on life span (data not shown).
In wild-type cells, the Sir complex bound at telomeres is redistributed to the nucleolus approximately midway in the life span of mothers (46). This relocalization leads to the appearance of the sterile phenotype because of a loss of silencing at HML and HMR, from which the Sir complex has been removed (45). We have speculated that the generation or accumulation of ERCs is slowed down by the redirected Sir complex, explaining the life span extension (19, 85). In rad mutants defective in homologous recombination, immunostaining with anti-Sir3p antibodies showed that the Sir complex, in contrast with old wild-type cells, is present diffusely throughout the nucleus including the nucleolus. This relocalization is probably responsible for the loss of silencing with age, as determined by the expression of a GFP marker inserted at HMR (Fig. (Fig.22).
Why is there a redistribution of the Sir complex away from telomeres and HM loci with age in rad mutants? We infer that the relocalization of the Sir complex in rad mutant cells is caused by DSBs at the rDNA and elsewhere in the genome, most likely during DNA replication (Fig. (Fig.6)6) (59, 80, 100). The Sir proteins, along with Ku70/80, have been implicated in the repair of DSBs by an end-joining reaction in yeast (93). Recent findings show that the induction of DSBs by EcoRI endonuclease also elicits the movement of the Sir complex away from telomeres (61). We speculate that diffuse nuclear staining is not observed in the wild type because DSBs are repaired efficiently by homologous recombination between sister chromatids in the S or G2 phase of the cell cycle.
The redistribution of the Sir complex to the nucleolus in wild-type cells may be triggered by DSBs that occur specifically in the rDNA. The recruitment of the Sir complex to the nucleolus in aging wild-type cells may be an attempt to employ end joining to supplement the repair of rDNA breaks by homologous recombination (Fig. (Fig.6).6). The repair of such damage by other repair pathways, including SSA and end-joining pathways, would avoid the possibility of generating ERCs by homologous recombination in the tandemly repeated rDNA. A DSB occurring at rDNA has been shown to be efficiently repaired in rad52 cells through the SSA pathway (70). The deletion of the RAD1 gene does not have a significant effect on life span (Fig. (Fig.5),5), suggesting either that the contribution of SSA in DSB repair in the rDNA is small or that RAD1 is not required for SSA in the rDNA.
However, it is also possible that the redistributed Sir complex extends the life span of wild-type cells through other mechanisms (19): by bolstering Sir2p-mediated suppression of rDNA recombination and thus reducing ERC formation (26, 32), by slowing the replication of ERCs that have already formed, or by reducing the bias in the segregation of these plasmids for mother cells (5).
Our results argue that DNA damage, probably DSBs, occurs throughout the genome of a wild-type yeast cell during its life span and is normally repaired efficiently through homologous recombination. Two lines of evidence suggest that yeast rDNA is particularly prone to DSBs. First, the continuous activity of topoisomerases, perhaps along with Sgs1p, is required for the transcription of rDNA (79), and for the maintenance of stability of the rDNA repeats in the genome (16, 17, 30, 47). These findings suggest that a high rate of rDNA transcription may pose unusual topological problems. Second, yeast cells accumulate arrested replication forks at rDNA (12, 13), which in E. coli are known to generate DSBs (59, 80). Concordantly, mutation in the FOB1 gene (48, 49, 53), which is required for replication fork blocking and HOT1 recombination activities, extends the life span of mother cells (20).
The identification of a DNA lesion that triggers the formation of ERCs may be important in the larger context of aging in higher organisms. In mammals, the repair of DNA breaks by homologous recombination is weaker than in yeast (52, 78). Rather, the nonhomologous end-joining pathway involving DNA-PK and Ku appears to be as important as the homologous recombination pathway (92). Any break in mammalian rDNA, therefore, may be repaired to yield not an ERC but a deletion within the genomic rDNA array.
A human homolog of SGS1, WRN, is defective in people with the premature aging disease Werner syndrome. It is of interest that WRN protein, like Sgs1p, is concentrated in the nucleolus in human cells (34, 58, 86). Moreover, deletions in genomic DNA occur at an elevated frequency in Werner syndrome cells (27, 28), although a specific effect on rDNA has yet to be demonstrated. It will be of interest to determine whether deletions resulting from DSBs accumulate with age in mammalian rDNA and whether a progressive loss in functional rDNA copies is a plausible explanation of aging-related changes.
P. U. Park and P.-A. Defossez contributed equally to this work.
We thank David Sinclair, Kevin Mills, Brad Johnson, David McNabb, and members of the Guarente laboratory for advice and stimulating discussions. Anna Lau, Steve Bell, David Sinclair, Jasper Rine, James Broach, Pam Silver, Sally Pak, and Mitch McVey generously provided plasmids. Many thanks go to Ed Hurt and Kevin Mills for antibodies. We also thank Glenn Paradis and Michael Jennings for technical assistance in FACS analysis and the Kaiser lab for the use of their Axioscope. P.U.P. thanks Y. S. Park for support. P.-A.D. thanks Guillaume Adelmant and Ezra Aksoy for advice and support.
P.U.P. is supported by the National Science Foundation Predoctoral Fellowship, and P.-A.D. is supported by INSERM. The Guarente lab is supported by National Institutes of Health grant AG11119.