Telomerase in Saccharomyces cerevisiae binds and preferentially elongates short telomeres, and this process requires the checkpoint kinase Tel1. Here we show that the Mre11 complex bound preferentially to short telomeres, which could explain the preferential binding of Tel1 to these ends. Compared to wild-type length telomeres, short telomeres generated by incomplete replication had low levels of the telomerase inhibitory protein Rif2. Moreover, in the absence of Rif2, Tel1 bound equally well to short and wild-type length telomeres, suggesting that low Rif2 content marks short telomeres for preferential elongation. In congenic strains, a double-strand break bound at least 140 times as much Mec1 in the first cell cycle after breakage as did a short telomere in the same time frame. Binding of replication protein A was also much lower at short telomeres. The absence of Mec1 at short telomeres could explain why they do not trigger a checkpoint-mediated cell-cycle arrest.
In the 1930s, when Muller was unable to generate terminally deleted chromosomes in irradiated flies, he reasoned that telomeres, a name he coined for the natural ends of chromosomes, must be essential to maintain stable chromosomes1. Soon after, it was found that broken chromosomes fuse with other breaks whereas telomeres do not fuse either with each other or with double-strand breaks (DSBs). These data suggested that telomeres are important to distinguish natural chromosome ends from DSBs. The cell's ability to distinguish between telomeres and DSBs is particularly remarkable because many proteins that are involved in sensing and repairing DNA damage also affect telomeres2. Molecular studies show that telomeres in most organisms consist of repeated DNA in which the strand that comprises the 3′ end of the chromosome is G-rich and extended to form a 3′ single-strand tail. In S. cerevisiae, the ~300-base-pair (bp) C1–3A/TG1–3 duplex region and the TG1–3 single-strand tail are bound, respectively, by two sequence-specific DNA binding proteins, Rap1 and Cdc13 (ref. 3).
Another key function of telomeres is to compensate for the incomplete replication of chromosome ends1. Owing to the biochemical properties of DNA polymerases, a short gap is left at the 5′ ends of newly replicated strands when the most distal RNA primer is removed. In most eukaryotes, this end replication problem is solved by a telomere-dedicated reverse transcriptase called telomerase. However, unlike conventional semi-conservative DNA replication, telomerase does not act on each telomere in every cell cycle. Rather, in yeasts and mammals, telomerase preferentially elongates the shortest telomeres in the cell4,5,6. In yeast, the frequency and extent of elongation, as well as telomerase processivity, are all greater at telomeres shorter than 125 bp (refs. 5,7).
The chromatin structure of short telomeres has been investigated to determine whether it differs from that of wild-type-length telomeres in a manner that might explain why telomerase preferentially lengthens the former. Two telomerase subunits, Est2 and Est1, bind preferentially to short telomeres, as does the Tel1 kinase8,9,10. Moreover, at a truncated left telomere of chromosome VII, preferential binding of Est2 and Est1 depends on Tel1, the yeast equivalent of the ATM checkpoint kinase, and binding of Tel1 requires the C terminus of Xrs2 (ref. 8), a subunit of the heterotrimeric Mre11 complex. Tel1 is also required for processive telomerase action at short (≤125 bp) telomeres7.
Here we show that the MRX complex, which comprises Mre11, Rad50 and Xrs2, bound preferentially to short telomeres. This result can explain how Tel1 and hence telomerase are targeted to short telomeres. However, it raises the question of how MRX recognizes short telomeres. Rap1 and its associated proteins Rif1 and Rif2—two negative regulators of telomerase—are brought to telomeres in a mutually exclusive manner by protein-protein interactions with the C terminus of Rap1 (ref. 3). As Rap1 binding sites are distributed at ~18-bp intervals throughout the yeast telomere11, by definition telomeres lose Rap1 binding sites as they shorten. Thus, differences in Rif1 or Rif2 occupancy, or both, are an appealing explanation for how cells distinguish short from wild-type length telomeres.
Here we report that Rif2 (but not Rif1) was less abundant at two natural telomeres that had been shortened from their ends by incomplete replication. In addition, Tel1 no longer bound preferentially to short telomeres in cells lacking Rif2 (rif2Δ) but not in rif1Δ cells. Thus, Rif1 and Rif2 act by different mechanisms to inhibit telomerase, and seem to be distributed differently along the length of the telomere. Likewise, when a DSB is introduced adjacent to a 162-bp tract of telomeric DNA, Rif2 inhibits the addition of telomere to the break whereas Rif1 has a much more modest effect, which supports the idea that the two proteins act by different mechanisms12. Moreover, although the checkpoint kinase Mec1 bound robustly to an induced DSB, Mec1 binding was ≥140 times lower at a short telomere in a tel1Δ strain in which Mec1 is required for telomerase activity13,14. Replication protein A (RPA), a non-sequence-specific, single-strand DNA binding protein that has essential roles in DNA replication and repair, showed a pattern similar to that of Mec1 (high binding to DSBs and low binding to telomeres). Together, our data not only provide a molecular explanation for how short telomeres are targeted for preferential elongation, but also suggest a mechanism for how cells distinguish telomeres from DSBs.
Mre11 complex binds preferentially to short VII-L telomeres
The preferential binding of telomerase to a short chromosome VII-L telomere lacking subtelomeric repeats requires the Tel1 kinase, which itself binds preferentially to this telomere8. As binding of Tel1 to DSBs15,16 and short telomeres8 requires the carboxyl end of Xrs2, we investigated whether the MRX complex also binds preferentially to a short telomere. Each of the three MRX subunits, Mre11, Rad50 and Xrs2, was tagged at its C terminus with 13 Myc epitopes and expressed from its own promoter at its endogenous chromosomal locus.
To compare the binding of MRX to short and wild-type length telomeres, we used a strain with an inducible short telomere4 (Fig. 1a). In the experimental version of this strain (Fig. 1a), the left telomere on chromosome VII contains two recognition sites for the site-specific FLP recombinase (FRT sites). FLP is expressed under the control of a galactose-inducible promoter. FLP-mediated recombination between the two FRT sites excises a subtelomeric fragment that reduces the size of the VII-L telomere to only ~100 bp, compared to ~300 bp of telomeric repeats on all other telomeres (Fig. 1a and Supplementary Fig. 1). As a control, we used an otherwise isogenic strain that also has two FRT sites in the subtelomeric region of the VII-L chromosome (Fig. 1a), but in which FLP does not affect the length of the VII-L telomere4 (Supplementary Fig. 1).
The experimental and control strains, which expressed the same epitope-tagged protein, were arrested in parallel in late G1 phase by incubation with the yeast pheromone alpha factor. Galactose was added to the G1-arrested cells to induce FLP, and the extent of recombination was assessed by Southern blotting (Supplementary Fig. 1). After recombination, cells were released from G1 arrest and followed through the subsequent synchronous cell cycle8. Samples were taken at regular intervals and processed for chromatin immunoprecipitation (ChIP) to determine the association of the epitope-tagged protein with the VII-L telomeres and for fluorescent activated cell sorting (FACS) to assess cell cycle position. As an additional control, we examined the association of epitope-tagged proteins with the chromosome VI-R telomere, which is of wild-type length in both the experimental and control strains.
The profiles of telomere binding were similar for each of the MRX subunits (Fig. 1b,c). At the wild-type length VII-L telomere in the control strain, binding was low throughout the cell cycle, although there was significant telomere binding for each subunit at the 60-min time point, coincident with the time of telomere replication (Fig. 1b). We found similarly low but significant binding of each MRX subunit at the wild-type VI-R telomere in both the experimental (open triangles) and control (closed triangles) strains (Fig. 1c). However, each MRX subunit showed robust binding to the short VII-L telomere (Fig. 1b). This binding was significant even in S phase but increased markedly as cells neared the end of the cell cycle. The binding of each MRX subunit was four to six times higher to the short VII-L telomere than to the wild-type length VII-L telomere in the control strain or to the VI-R telomeres in either the control strain or the experimental strain. The binding of Xrs2-Myc remained high for at least two cell cycles after telomere shortening (Fig. 1d). As the carboxyl end of Xrs2 is required for Tel1 to bind to the telomere8, the preferential binding of Xrs2 to the short VII-L telomere could explain the preferential binding of Tel1 to these ends.
Mec1 does not bind short telomeres even in tel1Δ cells
Telomeres can be maintained by telomerase in tel1Δ cells as long as MEC1 is present17. To determine whether Mec1, like Tel1, binds preferentially to short telomeres, we introduced three hemagglutinin (HA) epitopes into an internal region of the protein (Mec1-HA)18. As with the MRX experiment (Fig. 1), cells were arrested using alpha factor, and we determined the association of Mec1 with telomeres throughout the cell cycle in both the control and experimental strains (Fig. 2).
Mec1-HA bound equally to the short (open squares) and wild-type (closed squares) length VII-L telomeres (Fig. 2a). Binding to both VII-L telomeres was at background levels early in the cell cycle and through much of S phase. As cells completed S phase, Mec1-HA binding increased modestly until at the end of the cell cycle, it was four-fold higher than at the nontelomeric ARO1 control sequence. Thus, although Mec1-HA binds to both the control and experimental VII-L telomeres, it does not bind preferentially to short telomeres.
We also examined the binding of Mec1 to two wild-type length telomeres, VI-R (open triangles) and XV-L (open circles; Fig. 2b). As the pattern and extent of binding of Mec1-HA to the two telomeres was indistinguishable in the two strains, only the data for the experimental strain are shown. This binding was indistinguishable from the binding of Mec1-HA to the nontelomeric ARO1 locus to which the values are normalized. Although the binding of Mec1-HA to the short VII-L telomere was modestly higher than to the VI-R or XV-L telomeres, this difference was significant only at 75 min (P = 0.01) and 90 min (P = 0.03). Thus, significant binding of Mec1-HA was detected only at telomeres that were acted upon by the FLP recombinase, and even this binding was significant only at two time points.
We also examined the binding of Mec1-HA to the short and wild-type length VII-L telomeres in tel1Δ cells, in which Mec1 is essential for telomere maintenance (Fig. 2c). We could not detect binding of Mec1-HA at either the control or experimental VII-L telomeres in tel1Δ cells (Fig. 2c). These data suggest that the low binding of Mec1-HA to the FLP-generated short and wild-type length VII-L telomeres was mediated by Tel1 and was probably due to FLP-mediated DSB breakage at individual FRT sites in the absence of synapsis19. Finally, in the experimental strain, the binding of Tel1 to the short VII-L or wild-type VI-R telomere was not significantly increased in the absence of Mec1 (Fig. 2d).
Mec1 binding is much higher at DSBs than at short telomeres
To serve as a positive control for the detection of Mec1 at telomeres (Fig. 2a), we determined the level of Mec1 binding at an induced DSB, which is known to bind Mec1 (refs. 20,21,22,23). For these experiments, we used strains that were isogenic to the short telomere strain except that they contained a galactose-inducible HO endonuclease instead of a galactose-inducible FLP recombinase, a recognition site for the HO endonuclease about 13 kb from the VII-L telomere, and no chromosomal FLP sites24. We used two DSB strains (Fig. 3a): TG80-HO contained 80 bp of TG1–3 telomeric DNA adjacent to the HO site, whereas N80-HO had 80 bp of unrelated DNA adjacent to the HO site24.
The cells expressed the same Mec1-HA construct and were synchronized in the same manner as in the short telomere experiments with HO expression induced in G1-arrested cells and turned off before cells were allowed to progress through the cell cycle (Fig. 2 and Supplementary Fig. 2). As expected, Mec1-HA did not associate with the HO site in either strain in the absence of HO expression (Fig. 3b, (–gal) time point). However, in both strains, Mec1-HA associated strongly and to a similar extent and duration with the HO-induced breaks: binding was 20-fold greater than background in late G1 phase, increased to 80-fold at late S/G2 phase, and peaked at 140-fold by the end of the cell cycle. Thus, binding of Mec1-HA was similarly high whether the DSB was adjacent to telomeric or non-telomeric DNA, and binding was much higher than binding to short telomeres (compare Fig. 2a with Fig. 3b).
We also examined the binding of Cdc13 to the induced TG80 and N80 DSBs (Fig. 3). We could not detect binding of Cdc13 before HO cleavage in either strain. In the TG80 break, we could detect Cdc13 binding even in G1 phase (0 min time point), and it increased as cells progressed through the cell cycle, peaking at ~400-fold greater than binding to the ARO control sequence late in the cell cycle (Fig. 3e, open circles). This level of Cdc13 binding is similar to the level of Cdc13 binding detected at both short and wild-type length telomeres in late S/G2 phase in a congenic strain that has been synchronized and analyzed in the same manner8. By contrast, the binding of Cdc13 to the N80 DSB was much lower, peaking at about 12-fold higher than control (Fig. 3e,f). Using asynchronous cultures, Cdc13 has been shown to bind to the TG80 DSB12,25 and to a non-telomeric, non-repairable DSB26.
RPA and γH2AX do not bind preferentially to short telomeres
RPA, a heterotrimeric complex that binds in a non-sequence-specific manner to single-strand DNA, is essential for DNA replication, repair and recombination27,28. Because RPA is recruited to DSBs29,30,31 and has been suggested to have a role in telomerase recruitment32, we examined its association with DSBs in the induced HO break system and with telomeres in the short telomere assay using synchronized cells and the same epitope-tagged version of its largest subunit, Rfa1-Myc, in both experiments.
The signal in our telomere ChIPs is usually normalized to the amount of nontelomeric ARO1 DNA in the same immunoprecipitate. However, Rfa1-Myc should bind to every nuclear DNA sequence during its time of replication, including ARO1. As ARO1 and telomeres replicate at different times in the S phase, the data for Rfa1-Myc binding are presented as percentage immunoprecipitate in both the induced DSB (Fig. 3) and short telomere experiments (Fig. 4).
We could not detect Rfa1-Myc binding at the HO recognition site before HO expression in either the N80-HO (closed circles) or TG80-HO strains (open circles; Fig. 3c). However, Rfa1-Myc was associated with DSBs throughout the cell cycle, with low binding in G1 phase that increased as cells progressed through the cell cycle. Although we found robust Rfa1-Myc binding at both DSBs, at most time points the binding of Rfa1-Myc binding was about twice as high at the N80-HO break as at the TG80-HO break, which is adjacent to telomeric sequence. By contrast, binding of Rfa1-Myc to the internal ARO locus was low throughout the cell cycle except at the 45 min time point, when binding was similarly high in the two DSB strains but much lower than to the DSB in the same cells (Fig. 3d).
Using the induced telomere assay, we examined the association of Rfa1-Myc with three telomeres, VII-L (Fig. 4a), VI-R (Fig. 4b, open triangles) and XV-L (Fig. 4b, open circles) in both the experimental and control strains. Because the data for the VI-R and XV-L telomeres were identical in the two strains (data not shown), only the data from the experimental strain are shown for these telomeres. In both strains, Rfa1-Myc bound to the three wild-type length and short VII-L telomeres during a discrete interval in late S phase and peaked at 60 min, consistent with the expected time of telomere replication (Fig. 4a,b). Rfa1-Myc also bound to the ARO1 locus during a limited but earlier period in S phase (Fig. 4c), and its timing and extent of binding was similar to that seen in the DSB strains (Fig. 3d). The level of binding of Rfa1-Myc to the three wild-type length telomeres and to the ARO1 locus was similar, ranging from 0.25% (ARO1) to 0.37% (telomere VI-R). The level of binding to the short VII-L telomere was modestly higher late in the cell cycle compared to wild-type VII-L, but this difference was not significant. Moreover, binding of Rfa1-Myc was about five to nine times lower at telomeres than at DSBs (compare Fig. 4a with Fig. 3c; maximal binding to telomeres was 0.5% whereas binding to the N80-HO and TG80-HO breaks was 4.5% and 2.3%, respectively). Thus, preferential binding of RPA to short telomeres is unlikely to mark them for preferential lengthening by telomerase as has been suggested32.
An early response to DNA damage is the replacement of canonical H2A by an H2A variant called H2AX, which is then phosphorylated (referred to as γ-H2AX)33. As the sole version of H2A in yeast is analogous to the H2AX of other eukaryotes, yeast H2A is phosphorylated rather than replaced upon DNA damage34. Because telomeres have high levels of γ-H2AX35,36, we investigated whether this modification marks short telomeres for telomerase elongation. Using the inducible short telomere assay, we measured γ-H2AX at the short and control VII-L telomeres, at two native wild-type length telomeres (VI-R and XV-L) and at two nontelomeric loci (ARO1 and RPL11A) in both the control and experimental strains (Fig. 4d–f).
For all loci, the level of γ-H2AX was fairly constant from late G1 phase through the end of the cell cycle, with a modest decline after S phase (Fig. 4d–f). The level of γ-H2AX at the VII-L telomere was not affected by telomere length (Fig. 4d; open squares, short telomere; closed squares, wild-type length VII-L telomere; the only significant difference was at 75 min; P = 0.04). However, the levels of γ-H2AX at both VII-L telomeres were about twice as high as at either telomere VI-R or XV-L, which suggests that the action of FLP increases binding of γ-H2AX (Fig. 4e) as well as Mec1-HA (Fig. 2b). Nonetheless, H2A phosphorylation does not mark short telomeres for preferential elongation by telomerase.
Loss of Rif2, not Rif1, occurs as telomeres shorten
When telomeres are shortened by internal deletion as they are in the inducible short telomere system (Fig. 1a), Rif2 content is lower at short than at wild-type telomeres, whereas the level of Rif1 is similar at both8. We wished to test the theory that depletion of Rif2 marks short telomeres for preferential telomerase elongation. However, first it was important to determine whether telomeres shortened from their ends by incomplete replication, the normal mechanism of telomere shortening, show the same chromatin composition as telomeres shortened by FLP-mediated internal deletion. We constructed diploids that were homozygous for Rif1-Myc, Rif2-Myc or Yku80-Myc and heterozygous for a deletion of TLC1, the telomerase RNA gene (Fig. 5a). Diploids were sporulated, tetrads were dissected and individual tlc1Δ or TLC1 spore clones were grown for about 25–30 generations until the tlc1Δ telomeres had shortened to about 150 bp (Supplementary Fig. 3). The same samples were examined by ChIP to assess the level of Rap1 (using a polyclonal anti-Rap1 serum) or Myc-tagged proteins at the VI-R and XV-L telomeres (Fig. 5b–e).
At both the VI-R and XV-L telomeres, the levels of Yku80-Myc (Fig. 5b) and Rif1-Myc (Fig. 5c) were indistinguishable at wild-type length and shortened telomeres (Fig. 5). However, the binding of Rap1 at short telomeres was reduced (66%, telomere VI-R; 58%, XV-L; Fig. 5d). The amount of Rif2 was also lower at the short VI-R (65% of wild-type) and XV-L (21% of wild-type) telomeres (Fig. 5e).
Preferential Tel1 binding to short telomeres requires Rif2
If depletion of Rif2 is the signal that marks short telomeres for elongation, the ability of Tel1 to distinguish between short and wild-type length telomeres might be compromised in rif2Δ but not rif1Δ cells (Fig. 6). To test this possibility, we constructed diploid strains that were heterozygous for a deletion of both TLC1 and either RIF1 or RIF2. Both haploid parents also expressed Tel1-HA. Diploids were sporulated, tetrads were dissected and individual tlc1 spore clones that were wild-type, rif1 or rif2 were grown for about 25–30 generations to an average telomere length of around 150 bp (Supplementary Fig. 4). After immunoprecipitating the samples with anti-HA antibodies to precipitate DNA associated with Tel1-HA, we subjected both immunoprecipitate and input DNA to telomere PCR and then gel electrophoresis to determine telomere sizes (Fig. 6b). We examined the lengths of telomeres VI-R and XV-L in the anti-Tel1 immunoprecipitates. Because the sizes of individual telomeres change in a stochastic manner37, we compared telomere lengths before and after immunoprecipitation from a given spore clone. The data are presented as the mean percent decrease in length of the Tel1 immunoprecipitate sample compared to that of the input DNA for the individual spore clones examined (Fig. 6c).
As expected, in wild-type cells, Tel1 bound preferentially to short telomeres as the mean length of telomeres was shorter in the Tel1 immunoprecipitate than in the input sample (26% shorter for telomere VI-R; 45% shorter for telomere XV-L; Fig. 6c). We obtained similar results in rif1Δ cells where the DNA in the anti-Tel1 immunoprecipitate was 25% (telomere VI-R) and 44% (telomere XV-L) shorter than in the input samples (Fig. 6c). When the same experiment was done with rif2Δ cells (Fig. 6c), telomeres in the Tel1-HA immunoprecipitate were still shorter than in the input DNA, but the effect was greatly attenuated (11% shorter for telomere VI-R; 14% shorter for telomere XV-L). In rif2Δ cells, the difference in length between the input and immunoprecipitated DNA samples was not significant for either telomere (VI-R, P = 0.377; XV-L, P = 0.218). The average percent difference in length for rif2Δ cells was significantly different from both wild-type (VI-R, P = 0.002; XV-L, P < 10−4) and rif1Δ (VI-R, P = 0.008; XV-L, P = 0.0003) at both telomeres whereas the differences between rif1Δ and wild-type cells were not significant (VI-R, P = 0.699; XV-L, P = 0.814). These data support a model in which reduced Rif2 content is a signal that marks short telomeres for preferential Tel1 binding and telomerase elongation.
A short telomere generated by an internal deletion (Fig. 1a) is lengthened at a faster rate than a wild-type length telomere for multiple cell cycles after shortening4. Here we show that Mre11, Rad50 and Xrs2 each bound preferentially to these short telomeres (Fig. 1b). By contrast, RPA (Fig. 4a) and γ-H2AX (Fig. 4d) associated equally well with short and wild-type length telomeres and binding of Mec1 was low at all telomeres (Fig. 2a–c). These results differ from previous findings of robust Mec1 binding and low Tel1 binding to bulk telomeres in late S/G2 phase38,39. In a tel1Δ strain where Mec1 is essential for telomerase action13,14, we detected no binding of Mec1 to the telomeres (Fig. 2c). As a positive control, we monitored binding of Mec1 to a DSB in an identical experimental situation and found that it was at least 140-fold higher than to a short telomere in the first cell cycle after telomere shortening or break induction (compare Figs. 2c and 3b). As the binding of Cdc13 to the TG80 DSB (Fig. 3e) was similar to the binding of Cdc13 at short and wild-type length telomeres8, the ability to detect Mec1 at DSBs but not at telomeres was not due to the DSB being a better substrate for ChIP. Unlike Mec1, Tel1 binds robustly to short telomeres8. Mec1 must bind very transiently to telomeres, bind to only a small subset of telomeres or act by phosphorylating its targets when they are not associated with telomeres. The fact that telomeres are much shorter in tel1 cells than in wild-type cells indicates that Mec1 is less efficient than Tel1 at promoting telomerase-mediated telomere lengthening, which can be explained by its low telomere binding.
Est2 and Est1 do not bind preferentially to the inducible short VII-L telomere in the absence of Tel1 or in an xrs2-664 mutant that lacks the portion of the protein that interacts with Tel1 at DSBs8. Moreover, short telomeres are not processively lengthened in tel1Δ cells7. Therefore, the preferential binding of the MRX complex to short telomeres (Fig. 1b) is sufficient to explain how Tel1 and hence Est2 and Est1 act preferentially at short telomeres. MRX is not brought to short telomeres by differential Mec1, RPA or γ-H2aX levels, as these proteins were either absent (Mec1) or equally abundant at short and wild-type length telomeres (Figs. 2a and 4a–d).
As Rif2 was distributed differently on short and wild-type length telomeres (Fig. 5; see also Fig. 6d), we investigated whether Rif2 is important to direct Tel1 to short telomeres by determining the lengths of telomeres in anti-Tel1 immunoprecipitates from wild-type, rif1Δ and rif2Δ cells (Fig. 6). Telomeres in the anti-Tel1 immunoprecipitates were ~25% (telomere VI-R) or ~45% (telomere XV-L) shorter than bulk telomeres in both wild-type and rif1Δ cells. As Tel1 still bound preferentially to short telomeres in rif1Δ cells, Rif1 must inhibit telomerase at a step downstream of Tel1 binding. By contrast, in rif2Δ cells, neither the VI-R nor the XV-L telomere differed significantly in length in the anti-Tel1 immunoprecipitate compared to input DNA (Fig. 6c). Thus, in the absence of Rif2, Tel1 could not distinguish short from wild-type length telomeres. These data suggest that differential distribution of Rif2 on short and wild-type length telomeres is required to direct MRX, Tel1 and telomerase to short telomeres. In vitro, Rif2 (but not Rif1) interacts with the C terminus of Xrs2, and this interaction can prevent Xrs2 from interacting with Tel1 (ref. 12). Thus, as telomeres shorten and lose Rif2, MRX should be more effective at recruiting Tel1.
Rif1 and Rif2 must inhibit telomerase by different mechanisms, as Rif1 levels were the same at wild-type (~300 bp) and short (~150 bp) telomeres. Rif1 has 14 S/TQ sites, which are recognition sites for ATM kinases, whereas Rif2 has none. Rif1 is phosphorylated on at least one of these sites in vivo40. An appealing model is that telomere-associated Rif1 is phosphorylated by Tel1, and this phosphorylation reduces its inhibition of telomerase. In addition, the loss of Rif2 but not Rif1 as telomeres shorten from ~300 to ~150 bp (Fig. 5) suggests that the two proteins are distributed differently along the yeast telomere with Rif1 positioned closer to the centromere than Rif2 (Fig. 6d). Earlier studies are consistent with the idea that Rif1 and Rif2 act by different mechanisms and also suggest that Rif2 is more potent than Rif1 at inhibiting telomerase12,41.
Our results are relevant to an understanding of how cells distinguish telomeres from DSBs. The early events in DSB processing and telomerase-mediated lengthening are remarkably similar. At both, MRX binds and recruits Tel1, and at both, DNA resection occurs through the collaborative and partially overlapping actions of Sae2, Exo1 and Sgs1 (refs. 42,43,44). However, because DSBs occur throughout the genome, the resection-generated 3′ single-strand tails at these ends are not sequence-specific, whereas the ~50–100-bp telomeric 3′ tails that are generated by resection comprise exclusively TG1–3 DNA. We confirmed that DSBs were associated with RPA, even when the break was next to an internal tract of telomeric DNA (Fig. 3c). By contrast, at telomeres, single-strand tails are associated with Cdc13 (ref. 45). Although RPA was detected at telomeres late in the S phase, this binding was not higher at the short than at the control VII-L telomere (Fig. 4a). The most likely explanation for this telomeric RPA is that it occurs during semi-conservative replication of telomeric DNA. Although we cannot exclude the possibility that RPA binds to the resection-generated TG1–3 tails, there was about eight times more RPA at DSBs than at telomeres (Fig. 3c). Likewise, Mec1 binding was essentially undetectable at short telomeres whereas it bound robustly to DSBs. The fact that RPA recruits Mec1 to resected DSBs46,47 but Mec1 was not found at telomeres provides further evidence that the low level of RPA at telomeres was associated with conventional forked replication intermediates rather than bound to resection-generated TG1–3 tails.
Mec1 binding was also high at the TG80-HO and N80-HO DSBs (Fig. 3b); this result differs from those of earlier studies in which Mec1 binding was found to be lower when a break was adjacent to an internal tract of telomeric DNA21,25. However, these earlier studies used asynchronous cells and measured Mec1 binding 1–6 h after inducing the DSB. In our study, the DSB was introduced in G1-arrested cells, and the nuclease turned off and cells released into the cell cycle only after most of the cells had an HO-induced DSB (Supplementary Fig. 2). Thus, in our experiments, the binding of RPA and Mec1 to the DSB were measured during the first cell cycle after break formation, before checkpoint-mediated arrest (Fig. 3b). As the cell cycle arrest is not as lengthy when a DSB is adjacent to telomeric DNA48, the difference between the two experiments is probably explained by the different protocols used to detect Mec1 binding. Although we found that RPA was highly associated with the TG80-HO break, the level of RPA association was about 50% less at this break than at N80-HO breaks (Fig. 3c), perhaps because Cdc13 either competes with RPA for binding to the TG1–3 tails that are generated by resection of the TG80-HO break (Fig. 3) or inhibits resection of the DSB (or both).
In mammals49,50 and yeasts, the processes that occur at DSBs and replicating telomeres are similar. Although a single short yeast telomere can trigger at least one step in the DNA damage signaling cascade (phosphorylation of Rad53 (ref. 51)) it does not elicit a checkpoint-mediated cell cycle arrest as, by several criteria, the cell cycle after its induction is of normal length8. By contrast, a single DSB elicits a strong checkpoint response52. Our data suggest that the lack of cell cycle arrest in response to a short telomere is due to the fact that resected telomeres are coated by Cdc13, not RPA, and hence do not recruit Mec1, whose presence is necessary to trigger a full checkpoint response.
Detailed methods are available in Supplementary Methods. Yeast strains are listed in Supplementary Table 1. The inducible short telomere experiments were carried out in bar1Δ::KAN1 (ref. 54) derivatives of the W303 strain Lev220 (ref. 4). Tel1 and Mec1 were internally tagged with three HA epitopes as described18. Mre11, Rad50, Xrs2 and Rfa1 were tagged at their C termini with 13 Myc epitopes. Cdc13 was tagged at its C terminus with nine Myc epitopes45. TEL1 was deleted in the Mec1-HA strain and replaced with HIS3. The DSB experiments were performed in derivatives of W303 strains analogous to those in ref. 24 (YAB285 and YAB1083) that contained bar1Δ::NAT (and sml1Δ::HIS3 in the Mec1-HA strains). These strains contain a galactose-inducible HO gene at the leu2 locus on chromosome III and a modified chromosome VII-L with an HO endonuclease recognition site between the ADH4 and MNT2 loci24. TG80-HO strains contain 80 bp of TG1–3 repeats on the centromere-proximal side of the HO recognition site; in N80-HO strains, this 80-bp sequence is replaced with 80 bp of lambda DNA.
Experiments to determine protein composition of telomeres shortened from their ends used spore clones obtained from sporulation of tlc1::LEU2/TLC1 diploids that were homozygous for MYC-tagged RIF1, RIF2 or KU80, with tagging done as described8. tlc1Δ::LEU2 and TLC1 spore clones carrying the appropriate tagged genes were grown for 25–30 generations in log phase growth before determining telomere length by Southern blot analysis and protein content by ChIP. To determine the effects of RIF genes on Tel1 binding to telomeres of different lengths, diploids homozygous for Tel1-HA and heterozygous for tlc1Δ, rif1Δ and rif2Δ were sporulated and spore clones were grown for ~25–30 generations and processed for ChIP.
Galactose induction and cell synchronization in both the induced short telomere and induced DSB strains were done essentially as described8. Samples were taken at least every 15 min and processed for flow cytometry, Southern blot analysis to determine percent recombination or percent DSB formation, and ChIP.
Samples were processed for telomere PCR using minor modifications of the methods described in ref. 10. The primers used in this study are listed in Supplementary Table 2. The PCR products were resolved on a 3% (w/v) MetaPhor (Lonza) agarose gel using 1 kb Plus DNA Ladder as a marker (Invitrogen). The lengths of the telomere PCR products were determined by the AlphaImager 3400 Molecular Weight Analysis program. For all experiments, an aliquot of the telomere PCR products were gel-purified, cloned and sequenced to verify that they contained telomeric repeats. Of the 100 sequenced clones, 90% contained telomeric DNA and 10% had no insert.
All ChIPs were performed as described8,55,56. Anti-sera were anti-Myc (Clontech monoclonal antibody no. 631206), anti-HA (Santa Cruz monoclonal antibody no. SC7392X), anti-H2A phosphoS129 (Abcam polyclonal antibody no. ab15083) or an affinity-purified polyclonal anti-Rap1 serum57. The amount of DNA in ChIP and input samples was quantified using real-time PCR (BioRad iCycler). In most cases, percentage immunoprecipitated DNA was normalized to the amount of the non-telomeric ARO1 sequence in the immunoprecipitate and input samples. However, for the Rfa1-Myc and γ-H2AX experiments, results are presented as percent immunoprecipitate (amount of target sequence in immunoprecipitate/amount in input sample).
For all ChIP experiments, samples from each time point were amplified in duplicate or triplicate to obtain a mean value for each sample. Each synchrony was repeated at least three times. To determine protein content at telomeres shortened from their ends, ChIPs were carried out on three or more independent spore clones for each genotype. Data are presented as mean ± s.d. and a two-tailed Student's t-test was used to determine statistical significance. P ≤ 0.05 was considered significant.
We thank D. Shore and A. Bianchi (University of Geneva) for strains and advice on the DSB assay, K. Runge for advice on telomere PCR, T. Petes (Duke University) for strains, M. Jayaram for discussions about FLP-induced DSBs and C. Webb and Y. Wu for comments on the manuscript. This work was supported by grants from the US National Institutes of Health (NIH; GM43265 to V.A.Z.), postdoctoral fellowships from the NIH (M.S.), Deutsche Forschungsgemeinschaft (K.P.) and NJCCR (K.P.), and predoctoral fellowships from NJCCR (J.A.P., J.S.M.) and NIH (J.S.M.).
Supplementary Figures 1–4, Supplementary Tables 1 and 2 and Supplementary Methods