Telomere length is maintained by the de novo addition of telomere repeats by telomerase, yet recombination can elongate telomeres in the absence of telomerase. When the yeast telomerase RNA component, TLC1, is deleted, telomeres shorten and most cells die. However, gene conversion mediated by the RAD52 pathway allows telomere lengthening in rare survivor cells. To further investigate the role of recombination in telomere maintenance, we assayed telomere length and the ability to generate survivors in several isogenic DNA recombination mutants, including rad50, rad51, rad52, rad54, rad57, xrs2, and mre11. The rad51, rad52, rad54, and rad57 mutations increased the rate of cell death in the absence of TLC1. In contrast, although the rad50, xrs2, and mre11 strains initially had short telomeres, double mutants with tlc1 did not affect the rate of cell death, and survivors were generated at later times than tlc1 alone. While none of the double mutants of recombination genes and tlc1 (except rad52 tlc1) blocked the ability to generate survivors, a rad50 rad51 tlc1 triple mutant did not allow the generation of survivors. Thus RAD50 and RAD51 define two separate pathways that collaborate to allow cells to survive in the absence of telomerase.
TELOMERES ensure chromosome stability by protecting chromosome ends from fusions, recombination, and degradation (reviewed in Greider 1996). In wild-type yeast Saccharomyces cerevisiae, the repeated telomere sequence contains ∼300 bp of d(C1-3A)/d(G1-3T) sequence. This telomere repeat sequence is bound by Rap1p and other telomere-binding proteins that participate in telomere length homeostasis (reviewed in Zakian 1996). Telomeres are elongated by telomerase, a ribonucleoprotein polymerase that contains an essential RNA and protein component(s) (reviewed in Greider 1996; Nugent and Lundblad 1998). When the gene for either the RNA component (TLC1) or a protein component (EST2) of yeast telomerase is deleted, cells exhibit progressive telomere shortening and ultimately cell death (Singer and Gottschling 1994; Landvay et al. 1996; Lingneret al. 1997). However, “survivors” that have rearranged and amplified telomere regions appear in late passage cultures. These survivors are dependent on the RAD52-mediated yeast recombination system (Lundblad and Blackburn 1993; Lendvayet al. 1996).
Recombination has been thought to play a role at telomeres for many years (reviewed in Blackburn and Szostak 1984; Zakian 1989). The repetitive nature of telomeric and subtelomeric sequences makes them a good target for recombination (Horowitz and Haber 1984; Louis and Haber 1990a). Indeed, a recombination model for telomere maintenance based on a gene conversion-type mechanism was proposed before telomerase activity was identified (Bernardset al. 1983; Walmsleyet al. 1984). The ability of linear plasmids to exchange foreign telomere repeats also suggested that recombination might play a role in telomere maintenance (Pluta and Zakian 1989). The requirement for RAD52 in the generation of survivors in late passage tlc1 and in est1, est2, est3, and est4 yeast cultures also suggested a role for recombination in telomere length maintenance (Lundblad and Blackburn 1993; Lendvayet al. 1996). Recently, mutations in the genes encoding the yeast Ku proteins HDF1 and HDF2 (also known as YKU70 and YKU80) as well as RAD50, XRS2, and MRE11 were shown to affect telomere length maintenance (Boulton and Jackson 1996, 1998; Porteret al. 1996; Kironmai and Muniyappa 1997; Gravelet al. 1998; Nugentet al. 1998; Polotniankaet al. 1998), which suggests that more than one recombination pathway may play a role in telomere maintenance.
Genes in the yeast RAD52 epistasis group (RAD50, RAD51, RAD52, RAD54, RAD55, RAD57, XRS2, MRE11) are involved in both mitotic and meiotic recombination in S. cerevisiae (reviewed in Peteset al. 1991; Haber 1998). They can be divided into three subgroups on the basis of their mutant phenotypes. RAD52 appears to be essential for all forms of mitotic homologous recombination, which include both spontaneous and double-strand break (DSB)-induced gene conversion, single-strand annealing between directly repeated sequences, and break-induced replication (Rudinet al. 1989; McDonald and Rothstein 1994; Rattray and Symington 1994, 1995; Sugawaraet al. 1995; Malkovaet al. 1996a). In contrast, although rad51, rad54, rad55, and rad57 mutants are defective in repair of chromosomal DSBs by gene conversion, these mutations do not prevent such repair in plasmid substrates with an apparently less constrained chromatin structure (Sugawaraet al. 1995). Moreover, these four genes are also not required for single-strand annealing or for break-induced replication (McDonald and Rothstein 1994; Ivanovet al. 1996; Malkovaet al. 1996a; L. Signon and J. E. Haber, unpublished results). A third group, RAD50, MRE11, and XRS2 has a much less profound role in homologous recombination (Ivanovet al. 1994; Malkovaet al. 1996b; Tsubouchi and Ogawa 1998), but mutations in these three genes severely impair nonhomologous DNA end-joining (Ogawaet al. 1995; Moore and Haber 1996). The division of these proteins into subgroups reflects their underlying biochemical associations. Rad51p, Rad52p, Rad54p, Rad55p, and Rad57p proteins exhibit a variety of interactions (Donovanet al. 1994; Hayset al. 1995; Jianget al. 1996; Cleveret al. 1997; Sung 1997). Rad50p, Xrs2p, and Mre11p proteins interact to form a separate complex (Johzuka and Ogawa 1995), which has been associated with exonucleolytic activity both in vivo and in vitro (Ivanovet al. 1994; Furuseet al. 1998; Tsubouchi and Ogawa 1998; Usuiet al. 1998; Moreauet al. 1999).
Telomerase is activated in 80–90% of human tumor cells and immortal cell lines and is thought to be required for the long-term growth of these cells (reviewed in Autexier and Greider 1996). However, cell lines and tumors that are telomerase negative have extremely long telomeres, suggestive of a recombination-mediated telomere lengthening (reviewed in Reddelet al. 1997). In addition, the ability of telomerase null mouse cells from late generation animals with short telomeres to form tumors may be through a recombination-mediated pathway (Blascoet al. 1997). Thus, understanding the role of recombination in telomere maintenance in the absence of telomerase may have implications for tumor growth in cells that lack telomerase or where telomerase has been inhibited.
To understand the role of DNA recombination in telomere maintenance more completely, we examined telomere length in a set of isogenic yeast mutants of the RAD52 epistasis group in both the absence and presence of the telomerase RNA component TLC1. Our results suggest that RAD50 and RAD51 define two separate RAD52-dependent homologous recombination pathways that collaborate to allow telomere maintenance in the absence of telomerase.
MATERIALS AND METHODS
Yeast strains, plasmids, and methods: The yeast strains used in this study are listed in Table 1. The set of isogenic strains with mutations in different RAD genes has been previously described (Moore and Haber 1996). These strains were maintained in standard growth media and grown at 30°. The tlc1::URA3 disruption plasmid (pLS132A) was created by replacing an Asp718-NsiI fragment of the wild-type TLC1 gene with a HindIII fragment of the URA3 gene in pLS131, a pUC119-based plasmid that contains a full-length TLC1 EcoRI-XbaI fragment. Yeast strains were constructed by standard genetic manipulations (Roseet al. 1988). To construct double or triple mutants, various heterozygous rad diploid cells were transformed with an EcoRI-XbaI fragment of pLS132A to knock out the TLC1 gene, and Ura+ cells were checked by Southern blot to verify the correct tlc1::URA3 integration. The resulting diploids were sporulated, tetrads were dissected, and spore clones were tested for the appropriate markers (some by Southern analysis). Recombination-deficient cells were also confirmed by their sensitivity to methyl methanesulfonate (MMS).
Cell culture: Cells were passaged in YPD liquid media by serial dilution to examine cell viability and telomere length. When a single spore from a dissection plate was grown to a colony we assumed ∼15–20 doublings, depending on the colony size. Cells were then streaked further for single colonies and were estimated to have undergone 20–25 divisions per streakout, depending on colony size. In Figure 4b, a cell spotting assay was used to examine cell viability. Equivalent numbers of cells (measured by hemocytometer) of different genotypes were diluted serially by 10-fold in YPD medium, spotted on a YPD plate, and incubated at 30° for 2 days. To assay cell viability in culture (Figures 2 and 4a), cells were grown to saturation (108 cells/ml), and each day were diluted to a concentration of 5 × 105 cells/ml in fresh YPD media and the cell density measured 24 hr later by hemocytometer (Singer and Gottschling 1994). This cycle was repeated for 4–9 days. Using this assay, each day of growth represents up to 10 generations, depending on the cell density of the culture when diluted. In Figure 3, approximate cell doubling numbers were estimated on the basis of actual cell density. At each dilution point, cells were also plated to examine colony size and possible contamination, and some cells were pelleted and frozen for later telomere length analysis.
Telomere length analysis: Yeast genomic DNA was isolated, and ∼2–4 μg DNA was digested with XhoI or PvuII and separated on a 1% agarose gel. DNA were then transferred to Hybond N+ (Amersham, Piscataway, NJ) membrane, UV cross-linked, and hybridized with a random primed telomeric poly(d[GT/CA]) (Pharmacia, Piscataway, NJ) probe in Church and Gilbert hybridization solution (1 mm EDTA, 0.5 m Na2HPO4 pH 7.2, 7% SDS, and 1% BSA) at 60° overnight. The blots were washed once at room temperature and three times at 60° with 2 × SSC and 0.1% SDS and exposed to a Fuji BAS 2000 PhosphorImager screen.
Rad50, xrs2, and mre11 mutants have shortened telomeres while rad51, rad52, rad54, and rad57 mutant cells do not: Deletion of the telomerase RNA gene TLC1 leads to loss of viability only after a phenotypic lag of 40–60 cell divisions in yeast (Singer and Gottschling 1994). Interestingly, est1 rad52 or tlc1 rad52 double mutant cells die much more quickly than the tlc1 or est1 single mutants (Lundblad and Blackburn 1993; Lendvayet al. 1996). One explanation for the rapid death of rad52 tlc1 double mutants might be that the telomere length was shorter in rad52 cells compared to wild type. To investigate this we examined telomere length in a set of isogenic strains that were mutated in genes that are known to be involved in recombination (RAD50, RAD51, RAD52, RAD54, RAD57, XRS2, and MRE11). For each mutant, telomere length was measured on Southern blots after prolonged cell growth by streaking cells on plates six successive times (Figure 1). Genomic DNA was prepared after one and six streakouts, digested with XhoI, and probed with a telomere repeat sequence. The broad band near 1.3 kb is derived from telomeres located adjacent to subtelomeric Y′ elements. The larger bands represent non-Y′-containing telomeres (Louis and Haber 1990b). rad52 cells did not have shorter telomeres; however, rad50, xrs2, and mre11 mutants had significantly shorter telomeres than wild-type telomeres as previously described (Boulton and Jackson 1998; Nugentet al. 1998). The other mutants examined, rad51, rad54, or rad57, and rad1, had telomere lengths similar to wild-type cells (Figure 1). In some samples telomere length appeared to increase or decrease between the first and sixth streakout (compare bands between 2 kb and 4 kb in Figure 1). These size variations are likely due to telomere length heterogeneity in a population of cells and the founder effect of single-colony analysis. Because a single cell gives rise to a colony, the initial cell might have had a very long telomere within the initial broad distribution. This leads to the initial long telomere length of the resulting colony (Shampay and Blackburn 1988). This effect is not seen when cells are passaged continuously in liquid media (data not shown), because founder effects do not operate under those conditions. To further resolve differences in telomere length, yeast genomic DNA was digested with PvuII that does not cut within the Y′ subtelomeric region and thus allows individual telomeres to be resolved. Again rad50, xrs2, and mre11 cells showed distinctly shorter telomeres than wild type (data not shown); similar results were also recently reported by other groups (Kironmai and Muniyappa 1997; Boulton and Jackson 1998; Nugentet al. 1998).
Viability declines sooner in double mutants of tlc1 and rad51, rad52, rad54, and rad57 than in tlc1 single mutants: To investigate the role of recombination in telomere maintenance in the absence of telomerase, we constructed a set of isogenic strains deleted for both TLC1 and RAD50, RAD51, RAD52, RAD54, RAD57, XRS2, and MRE11, respectively. To generate these strains, the TLC1 gene was deleted from diploids heterozygous for each mutant recombination gene and then double mutants were obtained from meiotic segregants (see materials and methods). Cell viability was examined by growing cells in liquid culture, measuring the density of the culture daily before diluting, and passing to a new culture tube (Singer and Gottschling 1994). We found that rad51 tlc1, rad52 tlc1, rad54 tlc1, and rad57 tlc1 cells showed more rapid cell death than tlc1 (Figure 2). The rad50 tlc1, xrs2 tlc1, and mre11 tlc1 cells died with similar kinetics to tlc1 mutant cells (data not shown). The fact that rad51 tlc1, rad54 tlc1, and rad57 tlc1 cells died earlier than tlc1 cells, like the rad52 tlc1 mutants, implies a role for these recombination genes in telomere maintenance.
Mutations in RAD50 group genes retard generation of tlc1 survivors compared to tlc1 mutants alone, while mutations in RAD51 group genes accelerate generation of survivors: It was previously shown that survivors are generated in tlc1 (and est1, est2, est3, and est4) late passage cultures and this process is dependent on the RAD52 gene product (Lundblad and Blackburn 1993; Singer and Gottschling 1994; Lendvayet al. 1996). These survivor cells show significant amplification and rearrangement of both telomeric repeats and subtelomeric regions, and it is presumed that recombination-mediated telomere lengthening allows their sustained growth. We examined the requirement of other RAD52 epistasis group genes for the generation of tlc1 late passage survivors. Double mutant rad50 tlc1, rad51 tlc1, rad52 tlc1, rad54 tlc1, rad57 tlc1, xrs2 tlc1, and mre11 tlc1 cells from freshly dissected spore clones were passaged in liquid culture continuously for >130 generations. Cell aliquots were taken every 15 generations to examine colony size by plating on YPD plates and telomere length was measured on Southern blots (see materials and methods). All double mutants grew well directly after germination and then showed a period of slow growth. Interestingly, after the slow growth period, all cultures except rad52 tlc1 cells regained a fast growth rate that is indicative of the generation of survivors (data not shown).
To determine if the different rates of cell death in the two subgroups of recombination mutants might be due to a direct effect on telomeres, telomere lengths were examined at several representative passage points (Figure 3). For all double mutants, telomere length initially decreased as the cells were passaged (Figure 3 and data not shown). In the late passage cultures, amplification and rearrangement of the telomeric and subtelomeric region was seen in rad50 tlc1, rad51 tlc1, rad54 tlc1, rad57 tlc1, xrs2 tlc1, and mre11 tlc1 cells. New bands with greater intensity of hybridization to the telomeric probe appeared in the late, but not the early passages, similar to the behavior of est1 and tlc1 late generation survivors (Lundblad and Blackburn 1993; Lendvayet al. 1996). The different recombination mutants generated survivors with different kinetics. rad51 tlc1, rad54 tlc1, and rad57 tlc1 cells showed aberrant-sized telomere bands indicative of survivors 72–75 cell divisions after germination (Figure 3). This is earlier than tlc1 single mutant cells, which did not show such aberrant telomere bands until generation 85. In contrast, rad50 tlc1, xrs2 tlc1, and mre11 tlc1 generated survivors later than tlc1 cells (around generation 100), although the major XhoI band that represents many telomeres was initially shorter than that in tlc1 cells (Figure 3, lanes 21, 24, 27, and see Figure 4). Thus the kinetics of generating telomeric amplifications correlated with the initial kinetics of cell death but did not correlate with telomere length in rad51 tlc1, rad54 tlc1, and rad57 tlc1 mutants.
As reported previously, rad52 tlc1 showed continuous telomere shortening (Figure 3, lanes 10–12) until generation 62. No survivors were generated and no evidence of telomeric amplifications or rearrangements was evident. Interestingly, the average telomere length in the last viable passage was longer than in some of the double mutants, such as xrs2 tlc1 and mre11 tlc1 (compare shortest telomere in Figure 3, lanes 4–6 with lanes 24 and 27), which again suggests that senescence is not strictly correlated with average telomere length. This is consistent with the observation that certain tlc1 mutants with shorter telomeres show good viability (Prescott and Blackburn 1997).
Mutations in both RAD50 and RAD51 block the generation of tlc1 late passage survivors: rad51 and rad50 mutations have different effects on the kinetics of cell death and the appearance of tlc1-independent survivors. To test whether these two mutations are part of the same pathway or different pathways in generating survivors, we made a rad50 rad51 tlc1 triple mutant (see materials and methods). We assayed the growth of the cells in liquid culture assay (Singer and Gottschling 1994) as well as in a cell culture spotting assay (Nugentet al. 1998). In the liquid growth assay the triple mutant rad50 rad51 tlc1 failed to generate survivors (Figure 4A). Each of the double mutants rad50 tlc1 and rad51 tlc1 generated survivors with the characteristic kinetics as described above. The rad50 tlc1 cells generated survivors later than tlc1 alone and the rad51 tlc1 cells generated survivors earlier than tlc1 seen in the previous experiment.
As an additional method to evaluate the relative viability of these cultures, we used the cell spotting assay described by Nugent et al. (1998). This assay was carried out on cells that had been grown for ∼60 generations (similar to days 5–6 in Figure 4A). Tenfold dilutions of each culture were spotted on YEPD plates to determine cell viability. As seen previously, tlc1 and rad50 tlc1 were slightly less viable than wild type and rad51 tlc1, rad54 tlc1, and rad57 tlc1 showed significantly reduced viability. Both the triple mutant rad50 rad51 tlc1 and rad52 tlc1 were inviable after this extent of cell growth (Figure 4B). Telomere length analysis on Southern blots showed no evidence of telomeric and subtelomeric amplification in the rad50 rad51 tlc1 triple mutant cells (data not shown). The failure to recover survivors with telomeric amplification in the rad50 rad51 tlc1 triple mutant is similar to what occurs in a rad52 tlc1 strain. This result suggests that rad50 and rad51 mutations affect two different pathways of recombinational maintenance of telomeres.
We have analyzed telomere length and cell viability in recombination-deficient strains, which include rad50, rad51, rad52, rad54, rad57, xrs2, and mre11, in both the presence and absence of telomerase. These genes can be divided into three groups on the basis of their telomere phenotypes (Table 2). First, rad50, xrs2, and mre11 single mutants showed initial telomere shortening followed by stabilization. When combined with tlc1 mutations, this group generated survivors later than tlc1 alone. Second, rad51, rad54, and rad57 alone had wild-type telomere length, but when combined with tlc1, the double mutants died sooner than tlc1. Interestingly, survivors appeared at earlier generations than in tlc1 single mutants. This earlier generation of survivors may be a direct result of the strong selection imposed by the early cell death in this group of mutants. The final group consists of rad52 and the rad50 rad51 double mutant; these cells died much earlier than tlc1 single mutants and failed to generate survivors.
The grouping of these RAD genes into these categories on the basis of telomere phenotypes parallels the grouping on the basis of their recombinational phenotypes and their biochemical properties (reviewed in Kanaar and Hoeijmakers 1997). First, mutations in RAD50, XRS2, and MRE11 have similar genetic effects on recombination (Maloneet al. 1990; Ogawaet al. 1995; Ivanovet al. 1996). These proteins interact to form a complex (Johzuka and Ogawa 1995; Usuiet al. 1998) that appears to provide or regulate an exonuclease function (Sugawara and Haber 1992; Sharples and Leach 1995; Ivanovet al. 1996; Tsubouchi and Ogawa 1998), but they also act together in other processes that do not appear to rely on exonuclease activity (Alaniet al. 1990; reviewed in Haber 1998; Tsubouchi and Ogawa 1998; Moreauet al. 1999). Second, genetic evidence indicates that the RAD51, RAD54, RAD55, and RAD57 genes function together in one pathway (Rattray and Symington 1995), and the proteins encoded by these genes interact physically (Hayset al. 1995; Jianget al. 1996; Cleveret al. 1997; Sung 1997). Although Rad52p and Rad51p proteins have been shown to interact both physically and genetically (Shinoharaet al. 1992; Milneet al. 1996), it is clear that their roles in recombination are significantly different (Rattray and Symington 1994; Malkovaet al. 1996a). There is increasing evidence of a RAD51-independent homologous recombination pathway that requires both RAD52 and RAD59 (Bai and Symington 1996).
Multiple recombination pathways play a role in telomere length: Genetic evidence indicates that there are at least three separate pathways involved with homologous recombination in yeast and that the RAD52 gene is involved in all three pathways (reviewed in Haber 1995; Rattray and Symington 1995; Malkovaet al. 1996a). Our data suggest that at least two independent recombination pathways are involved in telomere length maintenance. The group of mutants represented by rad50 had short telomeres and generated survivors late. The fact that survivors were generated at all suggests that another pathway, or gene product, in addition to the RAD50 pathway plays a role in generating survivors. Similarly, survivors were still generated when mutations of the RAD51 group genes were combined with tlc1. In contrast, neither rad52 tlc1 nor rad50 rad51 tlc1 triple mutants generated survivors. One interpretation of these results is that two different pathways defined by the RAD50 and RAD51 genes are needed to generate survivors and that RAD52 plays an essential role in both pathways.
There are two additional interpretations of the result that rad50 rad51 tlc1 triple mutants fail to generate survivors. One is that the non-homologous-end-joining (NHEJ) and homologous recombination are both needed for the generation of survivors. However, because RAD52 affects multiple homologous recombination pathways but has little or no effect on NHEJ pathways (reviewed in Peteset al. 1991; Krameret al. 1994; Moore and Haber 1996; Tsukamotoet al. 1996), it is hard to argue that rad52 tlc1 cells are deficient in both NHEJ and homologous recombination.
Another possible interpretation of the triple mutant phenotype is that RAD50 and RAD51 are in the same homologous recombination pathway but neither deletion mutant alone inactivates the pathway completely; only the double mutant completely knocks out the pathway. However, results from earlier experiments support the notion that RAD50 and RAD51 define two alternative homologous recombination pathways. First, a rad51 mutation reduces spontaneous recombination between heteroalleles as much as 10-fold, while a rad50 mutation increases recombination by about the same extent (Ivanovet al. 1994; Rattray and Symington 1995). Second, for interchromosomal gene conversion induced by a DSB, RAD51 is essential, while the absence of RAD50 has only a small effect on the efficiency of repair, although the kinetics of repair are delayed (Ivanovet al. 1994; Malkovaet al. 1996b). Third, the discovery that RAD59 defines a RAD51-independent, but RAD52-dependent pathway of recombination (Bai and Symington 1996) argues that there are indeed multiple homologous repair pathways. Finally, although a rad51 mutation prevents gene conversion, it still can carry out breakinduced replication, where one end of a broken chromosome invades a homolog and promotes replication all the way to the end of the chromosomal template (Malkovaet al. 1996a; Morrowet al. 1997; Bosco and Haber 1998). Thus, we interpret the lack of survivors in the rad50 rad51 tlc1 triple mutant as the elimination of two separate RAD52-dependent pathways, one requiring RAD51 and the other requiring RAD50. Although the data cited above support a model where RAD52 plays a central role, at the present time we cannot exclude a model in which the lack of survivors in the rad50 rad51 tlc1 triple mutant is independent of the lack of survivors in rad52 tlc1.
Mechanism of recombination-mediated telomere elongation: Recombination-mediated telomere elongation likely occurs by a mechanism similar to break-induced replication. If a chromosome end is lost or severely shortened, the broken chromosome will only have sequences on one side of the break that are homologous to any template from which it can be repaired; thus conventional gene conversion mechanisms will not allow repair (reviewed in Szostaket al. 1983). A recombination-dependent replication mechanism, however, can accomplish such repair (Malkovaet al. 1996a; Morrowet al. 1997; Bosco and Haber 1998). In wild-type cells, an HO-induced DSB is nearly always repaired, most often by a gene conversion process in which the intact homologous chromosome is used as a template to copy sequences around the break. In a rad52 diploid, >99% of the cells lose the broken chromosome. In contrast, in a rad51 strain, half of the cells are able to repair the DSB. However, the repair events are not conventional gene conversions; instead, the broken chromosome end proximal to the DSB apparently invades the intact homologous chromosome and replicates all the sequences (100 kb or more) to the end of the chromosome. This RAD52-dependent break-induced replication mechanism (Malkovaet al. 1996a) is analogous to the break-copy repair mechanism proposed by Morrow et al. (1997). Recent studies have found that this type of replication is efficient in wild-type cells in circumstances where gap repair cannot occur, such as when the end of a chromosome has been lost (Bosco and Haber 1998). This mechanism is similar to earlier proposals for the repair and maintenance of telomere ends (Bernardset al. 1983; Walmsleyet al. 1984; Dunnet al. 1985). The fact that break-copy repair events can apparently occur after several cell cycles (Malkovaet al. 1996a) suggests that chromosomes lacking telomeres in yeast might be stable through several rounds of replication, as has been previously proposed (Sandell and Zakian 1993). Thus the evidence suggests that break-induced replication is the mechanism that elongates telomeres in the absence of telomerase.
rad51, rad52, rad54, and rad57 mutations accelerate the death of tlc1 mutants: We began this analysis with the observation that the rad52 tlc1 double mutant dies at earlier generations than mutants in tlc1 alone. We hypothesized that if death is simply due to telomere loss, the rad52 tlc1 double mutant might lose telomeres at a faster rate. We saw no evidence of this in Southern blot analysis. The double mutants for rad51 tlc1, rad54 tlc1, and rad57 tlc1 also showed a faster rate of death than tlc1, and yet they also showed no increase in the rate of average telomere shortening. Because our method measures the average length of telomeres, we cannot rule out a subtle effect on a subset of chromosome ends that is not evident in the Southern blot analysis.
If there is no increased rate of telomere shortening, why do the double mutants rad51 tlc1, rad54 tlc1, and rad57 tlc1 die sooner? Additive effects of independent pathways that both affect chromosome stability may play a role. There is an accelerated rate of chromosome loss in rad51, rad52, and rad54 mutant cells (reviewed in Friedberg 1988) and this rate is further increased by X-irradiation, which leads to an increased rate of death (Mortimeret al. 1981). Assuming that tlc1 cells die, at least in part, from chromosome loss, increasing the chromosome loss rate by combining two mutations, both of which cause loss, may lead to more rapid cell death. Alternatively, DNA repair mechanisms that can normally maintain viability in the face of DNA damage may be overwhelmed when faced with too much damage.
RAD50, XRS2, and MRE11 play a role in telomere maintenance even in the presence of telomerase: The fact that telomeres are shorter in rad50, xrs2, mre11, hdf1, and hdf2 mutants suggests that these genes are involved in telomere maintenance (Porteret al. 1996; Kironmai and Muniyappa 1997; Boulton and Jackson 1998; Nugentet al. 1998). The RAD50, XRS2, and MRE11 genes are involved in both homologous recombination and DNA end-joining (reviewed in Moore and Haber 1996; Kanaar and Hoeijmakers 1997; Tsukamotoet al. 1997; Tsukamoto and Ikeda 1998). Mutations in HDF1 and HDF2 confer an altered chromosome end structure, which suggests that the Ku proteins function in telomere end maintenance (Gravelet al. 1998; Polotniankaet al. 1998). Double mutants of rad50 tlc1, xrs2 tlc1, and mre11 tlc1 generated survivors after more cell divisions than double mutants of tlc1 with rad51, rad52, rad54, and rad57. This may be analogous to the mechanism by which rad50 and xrs2 mutants delay but do not prevent mating-type switching in yeast (Ivanovet al. 1994). Perhaps unprocessed ends are not able to take part in a normal recombination process at telomeres, or, as our evidence now suggests, the absence of a RAD50-dependent homologous recombination pathway blocks telomere gene conversion (break-induced replication) at normal telomeres even in the presence of telomerase.
The analysis of telomere length in recombinationdeficient mutants presented in this article suggests a role for DNA recombination in normal telomere maintenance in yeast. Recombination has been implicated in telomere length maintenance under special circumstances in human cells (Bryan et al. 1995, 1997). Some of the RAD52 group genes have homologues in higher eukaryotes, including human (reviewed in Ivanov and Haber 1997; Kanaar and Hoeijmakers 1997). Thus it will soon be possible to test whether DNA recombination plays a similar role in mammalian telomere length maintenance.
We thank Dr. Jef Boeke and the Greider and Haber laboratory members for helpful discussions and critical reading of the manuscript. We thank Qijun Chen for assistance with tetrad dissection and Sang Eun Lee for providing the yeast strain ySL43. This work was supported by National Institutes of Health (NIH) grants GM43080 and CA16519 to C.W.G. and by Department of Energy grant 1ER61235 to J.E.H. S. Le was supported by an NIH postdoctoral fellowship (CA 68736).
Communicating editor: L. S. Symington
- Received September 11, 1998.
- Accepted February 11, 1999.
- Copyright © 1999 by the Genetics Society of America