The SIR2/3/4 complex and SIR2 alone promote longevity in Saccharomyces cerevisiae by two different mechanisms

The SIR2/3/4 complex regulates life span

To examine roles of the Sir proteins in yeast aging, null alleles of either sir2, sir3, or sir4 were introduced into a W303 MATa RAD5 strain (W303R). Disruption of either sir3 or sir4 resulted in a 20% decrease in mean life span (Fig. 1A). In contrast, the sir2 mutant strain demonstrated a more severe reduction in mean life span of ∼50% (Fig. 1A). To ascertain whether the short life span of a sir mutant could be the result of a defect in NHEJ, hdf1 and hdf2 were disrupted in W303R, and the life spans of the mutants were determined. Both hdf mutants had life spans that were significantly shorter than wild type (Fig. 1B). Thus, these observations initially raised the possibility that the short life span of sir mutants was due to a defect in NHEJ.

Because hdf1 and hdf2 mutants are also defective for telomere maintenance and are temperature sensitive (Feldmann et al. 1996; Boulton and Jackson 1996a,b; Barnes and Rio 1997; Gravel et al. 1998), we wished to measure the life span of a mutant specifically defective for NHEJ, without the pleiotropic effects associated with hdf mutations. The yeast ligase IV gene, DNL4, is required for NHEJ (Schar et al. 1997; Teo and Jackson 1997; Wilson et al. 1997), but is not required for growth at 37°C in the W303R strain background (not shown). The life span of the dnl4 mutant was indistinguishable from wild type (Fig. 1C), demonstrating that NHEJ does not affect aging, and suggesting that the short life span of a sir mutant is not caused by a defect in NHEJ.

The short life span of sir3 and sir4 mutant haploid strains is due to the simultaneous expression of a and α mating-type genes

In addition to their role in NHEJ, the SIR genes are also required for transcriptional repression. Mutation in any of the sir genes results in a loss of silencing at the HM loci, coexpression of a and α mating-type genes, and a corresponding sterility phenotype in haploid S. cerevisiae. This raised the possibility that the short life span of sir mutants was due to the inappropriate expression of loci that are normally silenced in a SIR+ strain (e.g., see Astrom et al. 1999). Therefore, we wished to test whether the simultaneous expression of a and α mating-type genes affected life span.

HMLα was deleted from each of the sir mutant strains above. As expected, deletion of HMLα resulted in a restoration of the a mating phenotype, as demonstrated by the ability of the sir hml strains to mate with a MATα strain (not shown). Strikingly, haploid sir4 hml and sir3 hml strains showed a significant extension of mean life span relative to the sir4 and sir3 strains (Fig. 2A,B), and were indistinguishable from W303R. In contrast, the sir2 hml strain had a life span that was indistinguishable from the sir2 strain (Fig. 2C). Deleting HML from the wild-type parent did not alter the life span (not shown). Similarly, the sir4 gene was disrupted in an isogenic strain of mating type α. The HMRa locus was then deleted from this strain, restoring the ability of these cells to mate. Once again, a 20% shortening of life span was observed in the sir4 mutant, and this life span defect was suppressed by deletion of hmr (not shown). Thus, it appears that the life span defect caused by a sir4 or sir3 mutation, but not a sir2 mutation, is due to derepression of the HM loci.

To determine whether mating-type heterozygosity in an otherwise wild-type background (SIR+) is sufficient to reduce haploid longevity, we directly compared the life spans of the sir4, sir3, and W303R strains with an isogenic strain with MATα integrated at the URA3 locus (W303Ra/α). The mean life span of the W303Ra/α strain was significantly shorter than the wild-type strain and indistinguishable from the sir3 and sir4 strains (Fig. 2D). Thus, simultaneous expression of a and α mating-type information causes a short life span even in the presence of a functional Sir2/3/4 complex.

To further test the hypothesis that mating-type heterozygosity shortens life span, we constructed homozygous sir4/sir4 and sir2/sir2 mutant diploids. The life span of the homozygous sir4/sir4 diploid was indistinguishable from the wild-type diploid, and ∼20% shorter than isogenic a or α haploid strains (Fig. 3A). We have found previously in other strain backgrounds that the life spans of a/α diploids and isogenic haploids were not significantly different (Kennedy et al. 1995; Park et al. 1999). The effect of a/α heterozygosity in diploid cells therefore appears to be strain specific, for reasons that are presently unclear.

In contrast, the homozygous sir2/sir2 diploid displayed a life span significantly shorter than the wild-type or sir4/sir4 homozygous strains (Fig. 3B), consistent with the conclusion that the short life span of the sir2 mutant is not related to cell mating type. Interestingly, the sir2/SIR2 heterozygous diploid had a life span that was much closer to the sir2/sir2 homozygote than to wild type, suggesting that the levels of Sir2p are limiting for life span, at least in diploid strains.

The short life span of sir2, sir3, or sir4 haploids is likely due to an increase in the rate of ERC formation

As deleting HMLα extends the life span of a sir3 or a sir4 mutant, but fails to suppress the life span defect of a sir2 mutant, Sir2p must function to forestall senescence by a mechanism(s) that does not involve silencing at HM loci. Sir2p is known to localize to the rDNA independently of Sir3p and Sir4p, in which it functions to repress homologous recombination (Gottlieb and Esposito 1989). Because homologous recombination is absolutely required for ERC formation (Park et al. 1999), the short life span of a sir2 mutant is likely to be caused, at least in part, by an increased rate of ERC formation. As a direct test of this hypothesis, we quantitated ERC levels in age-matched W303R and the isogenic sir2 hml mutant cells. In both young (unsorted) and old (sorted) cells, the sir2 hml strain accumulated a greater number of high molecular weight ERCs than W303R (Fig. 4). The fact that we can detect more ERCs in an unsorted population of sir2 hml cells, relative to wild-type, indicates that a significant fraction of the sir2 hml cells form an ERC in an early mother cell division. This suggests that increased homologous recombination at the rDNA in a sir2 mutant directly results in the formation of ERCs, which shortens the life span.

We were not able to observe a clear increase in the levels of ERCs in sir3 or sir4 mutants relative to wild type (not shown). This does not rule out the possibility that sir3 or sir4 mutations increase ERC formation, because an undetectable difference in levels of ERCs might be able to cause the small difference in the life spans of these sir mutants relative to wild type. As a more accurate measure of recombination rates in the rDNA, we measured the loss rate of the ADE2 marker integrated into a single rDNA repeat. This marker was lost at a two- to threefold higher frequency in a sir3 or sir4 mutant strain compared with wild type (Table 1). Furthermore, this elevated rate of rDNA recombination was suppressed by deleting HMLα to restore the a mating type. These data suggest that the short life span in the sir3 or sir4 nonmating haploids is caused by a small increase in recombination at the rDNA due to the expression of both a and α mating-type genes.

Also consistent with the notion that sir3 and sir4 mutants have an elevated rate of ERC formation, the release of an ERC into wild-type or sir4 mother cells caused an identical shortening of life span. For this experiment, we used the method of Sinclair and Guarente (1997) to release an ERC into the nucleus of wild-type and sir4 virgin cells. On release of an ERC, W303R+ERC cells had a mean life span that was shortened by ∼25% and did not differ significantly from the sir4+ERC strain (Fig. 5). This finding indicates that the sir4 (and presumably sir3) mutation does not affect the rate of accumulation of ERCs and strengthens the idea that it increases the rate of formation of ERCs.

Deletion of FOB1 suppresses the rDNA hyperrecombination phenotype and the life span defect of a sir2 mutant

A key gene that regulates rDNA recombination and life span is FOB1. Deletion of fob1 gives rise to bidirectional rDNA replication (Kobayashi and Horiuchi 1996), decreases the rate of rDNA recombination (Kobayashi et al. 1998), dramatically reduces ERC formation, and extends life span by 30%–60% (Defossez et al. 1999). We therefore wished to examine the relationship between sir2 and fob1 with respect to life span, rDNA recombination, and ERC formation.

To eliminate any effects of the sir2 mutation on HM expression, the sir2 hml strain was used. When fob1 was deleted in this strain, the short life span of the sir2 mutant was largely suppressed, giving rise to a life span extended by ∼50% (Fig. 6). The mean and maximum life span of the sir2 fob1 hml strain was comparable with W303R, but slightly shorter than the fob1 single mutant. This suppression, along with the experiment described below (Fig. 7B), suggests that SIR2 and FOB1 act in the same pathway to determine yeast life span (see Discussion).

Consistent with the life span data, deletion of fob1 also suppressed the rDNA hyperrecombination phenotype of the sir2 hml strain (Table 1). An ADE2 marker inserted into the rDNA array was lost at an ∼10-fold higher rate in the sir2 hml strain than in W303R. Deletion of fob1 in this background reduced this high rate of marker loss to approximately that of the SIR2 FOB1 strain.

To test whether the fob1 deletion reduced the rate of ERC accumulation in the sir2 mutant, we determined ERC levels in young and old sir2 hml fob1 cells. Deletion of fob1 reduced ERC levels in both young and old cells, even below those seen in W303R (Fig. 4). A similarly dramatic reduction in ERCs is observed when fob1 is mutated in a wild-type background (Defossez et al. 1999). The fact that the sir2 hml fobl strain does not live significantly longer than W303R, even though the levels of ERCs are lower than wild type, suggests that Sir2p may play an additional role as a regulator of longevity.

Increasing the dosage of SIR2 extends life span

The above analysis of SIR2 suggested that this gene played a central role in yeast aging. The short life span of the SIR2/sir2 heterozygous diploid further suggested that the levels of this protein may be limiting. If this were so, then increasing the levels of Sir2p might extend the life span of wild-type cells.

Thus, a second copy of SIR2 was integrated into the genome of W303R and W303R sir2 (see Materials and Methods), and the life spans of the resulting transformants were determined. The integrated SIR2 restored the wild-type life span (Fig. 7A) and mating ability (not shown) to the sir2 mutant, indicating that the cloned SIR2 was functional. Strikingly, integration of a second copy of SIR2 into the wild-type strain gave rise to an ∼30% extension of life span (Fig. 7A,B). This finding shows that the levels of Sir2p are limiting for the life span of wild-type mother cells.

The above finding afforded another possibility to test whether FOB1 and SIR2 function in the same pathway. If the role of SIR2 were to prevent recombination at blocked replication forks in the rDNA, then a second copy of SIR2 should not extend further the elongated life span in a fob1Δ strain. Thus, a second SIR2 copy was integrated into the fob1Δ strain and life spans were determined (Fig. 7B). The increased levels of Sir2p did not extend further the long life span in the fob1Δ strain, providing additional evidence that the role of SIR2 in extending life span involves preventing recombination at the Fob1p-arrested replication fork.

The role of cell type in aging

Several previous experiments led us to propose that the role of the Sir complex in yeast longevity might be to repair damage in the rDNA. First, the wild-type Sir complex redistributes from telomeres to the nucleolus in aging mother cells (Kennedy et al. 1997). Second, the sir4-42 mutation, identified previously as an allele-specific suppressor for the short life span of a strain carrying a carboxy-terminal deletion in uth4 (Kennedy et al. 1995), results in the constitutive localization of the Sir complex to the nucleolus. One specific possibility was that the Sir complex might function by repairing DSBs in the rDNA via NHEJ. Although one study indicated that the role of the Sir complex in NHEJ was an indirect consequence of the a/α cell type (Astrom et al. 1999), more recent studies showed that the Sir complex redistributes directly from telomeres to induced DSBs and likely functions directly in promoting their repair (Martin et al. 1999; Mills et al. 1999).

Mutations in hdf1 and hdf2 shorten life span significantly. However, the short life spans of these mutants do not result from a defect in NHEJ. This follows from the fact that the dnl4 mutant is also defective in NHEJ, but has a wild-type life span. Hdf mutants show an aberrant organization of telomeres (Gravel et al. 1998) and are temperature sensitive for growth (Feldmann et al. 1996; Boulton and Jackson 1996a,b; Barnes and Rio 1997). Thus, the short life spans of hdf1 and hdf2 strains are caused by defects in processes other than NHEJ, perhaps related to telomere maintenance or function.

Rather, it is the expression of both a and α mating-type genes in sir3 or sir4 mutants that appears to cause the short life span in these strains. Deletions of the HM loci that preserve haploid mating reverse the short life span of sir3 or sir4 mutants. Further, homozygous deletion of sir4 in an a/α diploid does not shorten life span.

How might a/α expression in haploids shorten life span? It has been suggested that RAD52-mediated recombination is increased in cells expressing both a and α mating-type genes (Heude and Faber 1993; Schild 1995; Fasullo et al. 1999), thus accounting for the increased resistance of diploid cells to certain DNA damaging agents. Consistent with these findings, we demonstrate an increase in the rate of marker loss from the rDNA due to a mutation in sir4, and this increase is suppressed by deleting HMLα (thereby preventing a/α coexpression). In addition, RAD52 is absolutely required for the formation of ERCs (Park et al. 1999). Thus, we propose that ERCs are generated at a higher rate in sir mutants, and that this increase in ERC formation results in a shortened life span (Fig. 8).

How then can we understand the sir4-42 mutation and the redistribution of the wild-type Sir complex in old cells? We speculate that the constitutive localization of the sir4-42 mutant complex to the rDNA might alter the recombinational properties in that locus to forestall the generation of ERCs. In this way, the mutation is really a neomorph and does not correspond to any normal activity of the Sir complex. In addition, accumulation of ERCs in old cells results in an amplification of rDNA sequence, which may titrate the Sir2/3/4p complex to the nucleolus.

A direct role of Sir2p in forestalling aging

The extremely short life span of a sir2 mutant is not a consequence of the a/α cell type. Rather, elevated levels of rDNA recombination (Gottlieb and Esposito 1989) result in an increase in the rate of formation of ERCs and thus accelerated aging. Strikingly, deletion of the replication fork block protein gene, FOB1, largely suppresses the short life span of a sir2 mutant. This epistasis of FOB1 to SIR2 implies that these genes function in the same pathway. The inability of a second SIR2 copy to extend life span in a fob1 strain strengthens this view, although we cannot exclude the possibility that fob1 mutant cells senesce for reasons unrelated to SIR2 function.

We propose that Sir2p stabilizes blocked replication forks in the rDNA to prevent DNA breaks, recombination, and the generation of ERCs (Fig. 8). In the absence of Fob1p, the replication fork is not blocked in the rDNA, and Sir2p is not required to prevent the formation of ERCs. This suppression of the short life span of sir2 mutants by mutation in fob1 contrasts with the short life span in mutants lacking the Sgs1p DNA helicase, which is not suppressed by fob1 (Defossez et al. 1999). Therefore, unlike the case in a sir2 mutant, the short life span of an sgs1 mutant is not caused by a Fob1p-blocked replication fork.

Further support for this model comes from observations that rad52 or fob1 mutations suppress the rDNA hyperrecombination phenotype of a sir2 mutant (Gottleib and Esposito 1989; Table 1). This demonstrates that in the absence of Sir2p, the predominant form of repair at the rDNA is through the RAD52 pathway of homologous recombination, most likely at the replication fork block. Correspondingly, ERCs are dramatically reduced in fob1 and sir2 fob1 hml mutant strains (Defossez et al. 1999; Fig. 4) and eliminated in a rad52 strain (Park et al. 1999).

However, because the sir2 fob1 hml strain has a shorter life span than the fob1 mutant, Sir2p must function to extend life span by an, as yet undefined, additional mechanism. Thus, FOB1 is only partially epistatic to SIR2 for life span, but is completely epistatic to SIR2 for ERC formation. Therefore, we propose that SIR2 promotes longevity by at least three distinct mechanisms; first, Sir2p functions, along with Sir3p and Sir4p, to silence HM loci; second, Sir2p directly inhibits ERC formation by antagonizing homologous recombination at the blocked replication fork in the rDNA; finally, SIR2 apparently helps extend life span by a third mechanism that is independent of ERCs or cell type.

Sir2p as the limiting component in yeast aging

It is of interest that introduction of a second copy of SIR2 into the yeast genome extends the life span 30% beyond wild type. Sir2p is, therefore, limiting in wild-type cells, and rDNA recombination at the replication fork block is not fully suppressed. In the fob1 mutant, in which the replication fork block is absent and life span is extended, a second copy of SIR2 does not extend life span further. The short life span of a sir2/SIR2 heterozygous diploid reveals a haplo insufficiency, also indicating that levels of Sir2p are limiting for life span. This is consistent with the observation that rDNA silencing is extremely sensitive to Sir2p dosage (Smith et al. 1998), and suggests that rDNA silencing, rDNA recombination, and life span are all tightly coupled to nucleolar levels of Sir2p.

Of all of the SIR genes, SIR2 is evidently the only one conserved in higher eukaryotes. Whereas the function of SIR2-related genes in mammals is not presently known, it is interesting to speculate that an alteration in Sir2p levels or activity may alter the pace of aging in these organisms, as well. If so, a key molecular link in the aging process of many organisms would be highlighted in the form of the function of this gene.

Strains, plasmids, and media

The yeast strains used in this work are listed in Table 2. Strain W303R is described in Mills et al. (1999), in which it is referred to as W303AR5. The sir3Δ::URA3 disruption was generated as described in Sinclair and Guarente (1997). The sir4Δ::HIS3, sir2Δ::TRP1, and hmlΔ::LEU2 disruptions were generated as described in Mills et al. (1999).

All other gene disruptions were generated by targeted PCR disruption. Marker genes were amplified from the appropriate pRS integrating vector (Sikorski and Heiter 1989) with primers designed to incorporate 40 nucleotides of homology to genomic DNA at each end of the disruption cassette. In each case, the entire coding region of the disrupted gene and up to 100 nucleotides of flanking sequence were removed. All disruptants were verified by PCR prior to subsequent analysis. Additionally, hdf1 and hdf2 mutations were verified by the inability of these strains to grow at 37°C. Lack of NHEJ in the dnl4 and hdf1 mutants was verified by a plasmid end-joining assay as described (Mills et al. 1999).

Integration of SIR2 at URA3 was accomplished by transforming cells with plasmid p306SIR2 digested with XcmI. In addition to the entire coding region of SIR2, ∼1000 nucleotides of upstream sequence and 200 nucleotides of downstream sequence are present. Plasmid p306SIR2 was constructed by ligating the PCR-amplified SIR2 gene into the XbaI and XhoI sites of pRS306.

Unless otherwise stated, all strains were cultured at 30°C and grown on YEPD containing 2% glucose.

Life span analysis

Micromanipulation and life span analysis were performed as described previously (Kennedy et al. 1994), with the following modifications. Prior to analysis, strains were patched onto solid YPD for at least two consecutive days. Cells were then transferred at low density to fresh YPD and allowed to grow for ∼3 hr. Virgin daughter cells were then isolated as buds from mother cells and subjected to life span analysis. Each experiment consisted of 37–50 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. Each figure represents data derived from a single experiment.

Release of ERCs into virgin daughter cells

The method of Sinclair and Guarente (1997) was used to release an ERC into virgin cells, with the following modifications. Life spans were performed on plates lacking adenine and containing 1% galactose and 2% raffinose. Mother cells whose daughters formed large colonies on this medium failed to induce the Cre recombinase and were therefore not included in the data set.

Determination of rDNA recombination rate

Strains carrying an ADE2 marker integrated into the rDNA array were grown overnight and then plated onto solid YPD. Colonies were allowed to grow for 48 hr at 30°C then placed at 4°C for 48 hr. The number of half-red/half-white colonies was determined; each was assumed to represent a marker loss event during the first cell division after plating. The number of half-sectored colonies divided by the total number of colonies (not including entirely red colonies) was reported as the rate of marker loss.

Purification of old cell DNA and ERC analysis

Old cells were obtained by the biotin-labeling procedure (Smeal et al. 1996) and age was determined by counting bud scars (Sinclair et al. 1997). ERC levels were detemined as described (Defossez et al. 1999), with the following modifications. ERC species were separated on a 0.6% agarose gel (without ethidium bromide) run at 1 Volt/cm for 44 hr.