|Home | About | Journals | Submit | Contact Us | Français|
In diverse organisms, telomerase preferentially elongates short telomeres. We generated a single short telomere in otherwise wild-type (WT) S. cerevisiae cells. The binding of the positive regulators Ku and Cdc13p was similar at short and WT-length telomeres. The negative regulators Rif1p and Rif2p were present at the short telomere, although Rif2p levels were reduced. Two telomerase holoenzyme components, Est1p and Est2p, were preferentially enriched at short telomeres in late S/G2 phase, the time of telomerase action. Tel1p, the yeast ATM-like checkpoint kinase, was highly enriched at short telomeres from early S through G2 phase and even into the next cell cycle. Nonetheless, induction of a single short telomere did not elicit a cell-cycle arrest. Tel1p binding was dependent on Xrs2p and required for preferential binding of telomerase to short telomeres. These data suggest that Tel1p targets telomerase to the DNA ends most in need of extension.
In most eukaryotes, telomeres are replicated by a specialized reverse transcriptase called telomerase. In both yeast and mammals, telomerase preferentially lengthens short telomeres (Hemann et al., 2001; Liu et al., 2001; Marcand et al., 1999; Ouellette et al., 2000; Steinert et al., 2000; Teixeira et al., 2004). In addition, yeast telomerase is cell-cycle regulated and able to lengthen telomeres only in late S/G2 phase (Diede and Gottschling, 1999; Marcand et al., 2000).
S. cerevisiae chromosomes end in ~300 bp of C1-3A/TG1-3 telomeric DNA. The G strand is maintained as a short (~13 bases) G tail throughout most of the cell cycle, whereas much longer G tails are detected in late S/G2 phase (Larrivee et al., 2004; Wellinger et al., 1993). The essential protein Cdc13p, which protects yeast telomeres from degradation and promotes their replication by telomerase, binds these G tails in vivo, with the highest level of binding in late S/G2 phase when G tails are long (Bourns et al., 1998; Garvik et al., 1995; Nugent et al., 1996; Taggart et al., 2002). The telomerase holoenzyme is a ribonucleoprotein complex comprised of a templating RNA (encoded by TLC1), a catalytic reverse transcriptase (Est2p), and accessory proteins (Est1p and Est3p). Elimination of any telomerase subunit results in the progressive loss of telomeric DNA and eventual death of most cells, the ever shorter telomere (est) phenotype (reviewed in Smogorzewska and de Lange, 2004; Vega et al., 2003).
Cells lacking both TEL1 and MEC1, the yeast homologs of, respectively, the checkpoint kinases ATM and ATR, also have an est phenotype (Ritchie et al., 1999). However, Tel1p has a much more critical role than Mec1p in telomere maintenance, as tel1Δ telomeres are very short, whereas mec1Δ telomeres are only modestly shorter than WT (Lustig and Petes, 1986; Ritchie et al., 1999). TEL1 is also required for de novo telomere addition at a double-strand break (DSB), but MEC1 is not (Frank et al., 2006). Because kinase-dead alleles of TEL1 have very short telomeres, Tel1p must act catalytically to promote telomere lengthening (Mallory and Petes, 2000).
Even though yeast telomerase acts only in late S/G2 phase (Diede and Gottschling, 1999; Marcand et al., 2000), Est2p is telomere associated throughout most of the cell cycle (Taggart et al., 2002), and telomerase activity in vitro is similar in extracts from G1 and G2/M phase cells (Diede and Gottschling, 1999). The G1/early S phase association of Est2p requires a specific interaction between the telomere-associated heterodimeric Ku complex and telomerase RNA (Fisher et al., 2004), whereas the late S/G2 phase Est2p binding is lost in the telomerase-deficient cdc13-2 strain (Taggart et al., 2002). Thus, Ku and Cdc13p are both positive regulators of telomerase. Est1p is telomere associated in late S/G2 phase when telomerase is active (Taggart et al., 2002). Tel1p facilitates recruitment of telomerase, as the telomere association of both Est2p and Est1p is severely reduced in late S/G2 phase in tel1Δ cells; in contrast, mec1Δ cells have near WT levels of Est2p and Est1p telomere binding (Goudsouzian et al., 2006).
The telomere binding proteins Rif1p and Rif2p (Bourns et al., 1998) associate with telomeres throughout the cell cycle (Smith et al., 2003) via their interaction with the duplex telomeric DNA binding protein Rap1p (Hardy et al., 1992; Wotton and Shore, 1997). Lack of either protein causes telomere lengthening, whereas deletion of both results in synergistic lengthening (Hardy et al., 1992; Wotton and Shore, 1997), and this lengthening is telomerase dependent (Teng et al., 2000). Thus, the Rif proteins are negative regulators of telomerase. Rif2p, but not Rif1p, also inhibits telomere addition to an induced DSB (Diede and Gottschling, 1999; Frank et al., 2006).
Telomerase action in yeast is regulated not only by the cell cycle but also by telomere length (Marcand et al., 1999; Teixeira et al., 2004). In a given cell cycle, telomerase lengthens only a subset of telomeres with preferential elongation of short telomeres: less than 10% of WT-length (~300 bp) telomeres are lengthened, whereas almost half of short (~100 bp) telomeres are elongated (Teixeira et al., 2004). In the absence of the negative regulators Rif1p or Rif2p, the frequency of elongation for all telomeres increases about 2-fold (Teixeira et al., 2004). In another study that used a strain with a single inducible short telomere (Figure 1), lengthening of the short telomere occurs in late S/G2 phase, and its rate of lengthening over subsequent generations is inversely proportional to its length (Marcand et al., 1999, 2000).
Here we use a strain with an inducibly short telomere (Marcand et al., 1999) to determine if the timing or level of binding of telomerase components or telomerase regulators is affected by telomere length. Two telomerase subunits, Est2p and Est1p, bound preferentially to the short telomere in late S/G2 phase, suggesting that telomerase access rather than its activity confers the preference for short telomeres. Ku, Cdc13p, Rif1p, and Rif2p bound robustly to both short and WT-length telomeres, suggesting that the presence or absence of these regulators does not mark short telomeres for elongation. Tel1p, which had low but significant binding to WT-length telomeres only in late S/G2 phase, bound to the short telomere even in G1 phase. This binding increased with cell-cycle progression, stayed high for at least two cell cycles, and was dependent on the carboxyl terminus of Xrs2p. However, there was no increase in the length of the cell cycle, indicating that a single short telomere did not elicit a checkpoint-mediated cell-cycle arrest and is thus not seen as a DSB. When Tel1p was lost from short telomeres, either by deletion or by elimination of its interaction with Xrs2p, preferential binding of Est1p and Est2p to short telomeres was lost. These data support a model in which Tel1p binds preferentially to short telomeres and modifies one or more telomeric substrates to facilitate access of telomerase to DNA ends.
To understand how short telomeres are preferentially elongated, we used the inducible short telomere strain (Marcand et al., 1999) in combination with chromatin immunoprecipitation (ChIP) to determine if the binding of telomerase itself or telomerase regulators is affected by telomere length. In this system, the left telomere on chromosome VII is modified such that galactose-induced expression of Flp1p, a site-specific recombinase that acts at FLP recombinase target (FRT) sites, excises a subtelomeric segment of ~1700 bp that remains in the cell as a nonreplicating circle (Figure 1A). As a result of the Flp1p-mediated recombination, the shortened VII-L telomere bears only ~100 bp of C1-3 A/TG1-3 telomeric DNA, making it ~200 bps shorter than the rest of the telomeres in the same cell (Figure S1 in the Supplemental Data available with this article online). This system is attractive because a short telomere can be studied in a strain that is WT for all telomere proteins, and its behavior can be compared to a WT-length telomere in the same cell.
The experimental strain with the inducible short telomere is hereafter referred to as strain A (Figure 1A). Seven different versions of strain A were generated, each expressing a different epitope-tagged protein. All tagged proteins were expressed from their endogenous promoters and supported WT or near WT telomere length (Fisher et al., 2004; Sabourin et al., 2007; Taggart et al., 2002; Tsukamoto et al., 2001)(Figures S2 and S5). We generated control versions of each of the epitope-tagged strains by allowing each experimental strain to grow for many generations after telomere shortening. After ~50-100 cell generations, telomerase re-extends the shortened VII-L telomere to a WT length. The control versions of the epitope-tagged strains in which the shortened VII-L telomere was re-extended to the WT length are hereafter referred to as control strain B (Figure 1A).
The experimental (A) and control (B) strains were arrested in late G1 phase with the yeast pheromone α factor and then galactose was added to both strains to induce shortening of the VII-L telomere in strain A (Figure 1B). After galactose addition, the only differences between the experimental and control strains are the length of the VII-L telomere and the presence of the excised circle (Figure 1A). After cells were released from α factor (time 0), both the control and the experimental strains proceeded synchronously and with indistinguishable kinetics through the subsequent cell cycle, indicating that telomere shortening did not delay the cell cycle (Figure S3). The association of the epitope-tagged protein with the short (strain A) or WT (strain B) VII-L telomere was determined throughout the synchronous cell cycle. In addition, in both strains, we determined the association of the tagged protein with unmodified telomere VI-R, which is WT in length in both strains. ChIP samples were quantitated by real-time PCR and normalized to both input DNA and to the amount of a nontelomeric sequence (ARO1) in the immunoprecipitate. Not only was telomere shortening efficient, occurring in ~75% of cells (Figure S1), but also, on the unrecombined VII-L telomere, the PCR primers were too far from a telomeric tract to be amplified efficiently (data not shown).
As shown previously (Taggart et al., 2002), Est2p was associated with the WT-length VII-L and VI-R telomeres throughout most of the cell cycle with peaks of association in both late G1 and late S/G2 phase (Figures 2A and 2B). Although the timing of association of Est2p with the short VII-L telomere (Figure 2A, open squares) was similar to WT (closed squares), its association in late S/G2 phase was ~3- to 4-fold higher than with WT-length VII-L in late S/G2 phase (see figure legends for p values). This increase was specific for the catalytically active late S/G2 peak of Est2p binding, as the level of binding of Est2p in G1 phase was not increased. In contrast, the level of Est2p binding to the WT-length VI-R telomere was indistinguishable in strains A and B (Figure 2B).
Similar results were found with Est1p (Figures 2D and 2E). Est1p binding was ~3-fold higher at 60 min at short telomere VII-L (Figure 2D, open squares) than at the WT-length VII-L telomere (closed squares), whereas Est1p bound equivalently to the VI-R telomere in the two strains (Figure 2E). These data suggest that short telomeres are preferentially elongated because they are preferentially bound by the telomerase holoenzyme in late S/G2 phase.
We examined the binding of four telomere binding proteins that are positive (Ku, Cdc13p) or negative (Rif1p, Rif2p) regulators of telomerase. As shown previously (Fisher et al., 2004; Schramke et al., 2004), the heterodimeric Ku complex was associated with WT-length telomeres throughout the cell cycle (VII-L, Figure 3A; VI-R, Figure 3D). Here we show that the amount and cell-cycle pattern of Ku80p binding was not affected by telomere length (Figure 3A, compare binding of WT-length VII-L, closed squares, to short VII-L, open squares). Thus, even though Ku recruits Est2p to telomeres in G1 phase and is required for WT levels of Est1p and Est2p telomere binding in late S/G2 phase (Fisher et al., 2004), its level and temporal pattern of telomere binding do not appear to contribute to the preference of telomerase for short telomeres.
Cdc13p is associated with WT-length telomeres throughout the cell cycle, but its binding increases greatly in late S/G2 phase when G tails are long (Taggart et al., 2002)(Figure 3B, VII-L; Figure 3E, VI-R). As with Ku80p, the abundance and timing of Cdc13p binding were similar at the short and WT-length VII-L telomeres (Figure 3B, compare binding to WT-length VII-L, closed squares, to short VII-L, open squares). Thus, although WT Cdc13p is required for the late S/G2 peak of Est2p telomere binding (Taggart et al., 2002), its level and timing of binding do not mark short telomeres for elongation.
As reported previously (Smith et al., 2003), the negative regulator Rif1p bound WT-length telomeres throughout the cell cycle (Figure 3C, VII-L; Figure 3F, VI-R). In our system, Rif1p binding was robust, with ~200-fold enrichment at both the VII-L and VI-R telomeres. This binding increased ~10-fold as cells reached the end of the cell cycle (Figures 3C and 3F). Although we anticipated that this negative regulator would be lost from short telomeres, the level of Rif1p binding to the short VII-L telomere (Figure 3C, open squares) was indistinguishable from its binding to the WT-length VII-L telomere (closed squares) at most time points. As cells progressed through the cell cycle, Rif1p binding to the short VII-L telomere increased just as it did at WT-length telomeres (Figure 3C).
As previously reported (Smith et al., 2003), Rif2p bound WT-length telomeres throughout the cell cycle (Figure 4A, VII-L; Figure 4B, VI-R). Telomeric DNA was enriched ~20-fold in the anti-Rif2p immunoprecipitate, and this enrichment increased modestly as cells proceeded through the cell cycle. Rif2p also bound robustly to the short VII-L telomere. However, at most time points, including at 0 min (the time when cells were released from α factor), its binding to the short VII-L telomere (Figure 4A, open squares) was ~50% of the level seen at the WT-length VII-L telomere (closed squares). Rif2p binding to the VI-R telomere was also higher in strain B than in strain A, but this difference was not significant (Figure 4B).
The Flp1p-mediated recombination that generates the short VII-L telomere also produces a small circle containing 540 bp of C1-3 A/TG1-3 DNA (Figure 1A). Rif1p and Rif2p do not bind DNA directly but rather are brought to telomeres by their interaction with the carboxyl terminus of Rap1p (Hardy et al., 1992; Wotton and Shore, 1997). Rap1p and hence the two Rif proteins also bind internal tracts of telomeric DNA (Bourns et al., 1998). Because Rap1p sites are eliminated by the recombination event that shortens the VII-L telomere (Figure 1A), Rap1p binding should be reduced at the short VII-L telomere. Consistent with this expectation, in G1 phase cells (0 time point), Rap1p binding to the short VII-L telomere (Figure 4C, VII-L, open bars) was ~50% of the level detected at control VII-L (Figure 4C, VII-L, closed bars), and using primers for the URA3 gene on the circle, Rap1p bound robustly to the excised circle (Figure 4C, URA3, open bars). Indeed, the sum of the amount of Rap1p on short VII-L with the amount on the circle was statistically indistinguishable from the amount on the control VII-L telomere (Figure 4C, VII-L, hatched bar, or Table 1). Likewise, Rif2p was also circle associated (Figure 4E, open triangles). As with Rap1p, the amount of Rif2p on the circle plus the amount on the short VII-L telomere was indistinguishable from the amount of Rif2p on the WT-length VII-L telomere (Figure 4D, compare closed squares to open diamonds).
Like Rif2p, Rif1p was readily detected on the excised circle (Figure 4E, open squares). However, at the 0 time point, the fraction of the Rif1p binding to the circle relative to WT telomere VII-L was lower (~26%) than the fraction of Rif2p (~55%) that was circle associated (Table 1). In addition, the amount of Rif1p on the short telomere VII-L throughout the cell cycle was not reduced compared to the amount of Rif1p on WT VII-L (Figure 3C), suggesting that Rif1p was loaded onto the circle and/or onto the shortened VII-L telomere after circle excision.
We also looked for the association of other proteins with the FLP-generated circle (Figure 4E and Table 1). Consistent with previous studies showing that Cdc13p does not bind internal tracts of telomeric DNA in vivo (Bourns et al., 1998), Cdc13p was not circle associated at any point (Table 1). Likewise, Est1p was not circle bound (Table 1), consistent with its being recruited to the telomere by its ability to interact with Cdc13p (Evans and Lundblad, 1999; Qi and Zakian, 2000). Ku80p immunoprecipitated the circle, but the level of binding was low, ~8% of the binding to WT-length telomere VII-L (Table 1). In late G1 phase (0 min time point), Est2p was circle associated (Figure 4E, open circles; Table 1). This binding was substantial, equivalent to ~60% of the amount of Est2p associated with WT-length VII-L telomere at this time (Table 1). However, Est2p binding to the circle was lost as cells progressed through the cell cycle (Figure 4E, open circles). These data suggest that a large fraction of the Est2p that is telomere associated in G1 phase is not bound to the terminal 100 bp of C1-3 A/TG1-3 DNA.
The kinase activity of Tel1p is needed to maintain telomere length (Mallory and Petes, 2000). In addition, the late S/G2 phase telomere association of Est2p and Est1p is markedly reduced in tel1Δ cells (Goudsouzian et al., 2006), a result repeated here in a different strain background (Est2p, Figure 2C; Est1p, Figure 2F). Thus, when all telomeres are made short by the absence of Tel1p, they bind less, not more, telomerase. In addition, we found that the preference of Est2p (Figure 2C) and Est1p binding (Figure 2F) for the short telomere was lost in the absence of TEL1. Although all telomeres in a tel1Δ strain are short, the recombined VII-L was still shorter than the control VII-L telomere (average difference in length between the shortened and control VII-L telomere was ~60 bp in the Est1p-Myc tel1Δ strain and ~90 bp in the Est2p-Myc tel1Δ strain (Figure S4). The loss of preferential telomerase binding to the short telomere did not correlate with telomere length: in the four experiments done in the Est2p-Myc tel1Δ strain, the experiment in which the short telomere had the largest difference in length compared to the control telomere (~70 versus ~195 bp, a difference of ~120 bp), there was no enrichment of Est2p-Myc at the short telomere (0.97-fold compared to control). In WT cells, telomeres of 70 bp are at least twice as frequently elongated compared to telomeres of 195 bp (Teixeira et al., 2004).
We also made a doxycycline-repressible version of Tel1p to examine telomerase binding in the absence of Tel1p. Because this strain had WT levels of Tel1p under repressive conditions, it was not useful for this experiment. However, under nonrepressed conditions, Tel1p was overexpressed 30- to 60-fold compared to WT levels of Tel1p and all telomeres were lengthened by 70–90 bp (Figure S6), providing additional support for a model in which Tel1p is a positive activator of telomerase.
We also examined the association of Tel1p with telomeres (Figure 5). Tel1p had very low but detectable binding to WT-length telomeres (2.8- to 3.5-fold above binding to nontelomeric ARO1 DNA at 60 min), and this binding was limited to late S/G2 phase (VII-L, Figure 5A, closed squares; VI-R, Figure 5B). The Tel1p binding profile to the short VII-L telomere (Figure 5A, open squares) was strikingly different from its binding to WT-length telomeres. Although Tel1p binding to WT-length telomeres was detectable only in late S/G2 phase, there was low but significant Tel1p binding to the short VII-L telomere even in G1 phase (Figure 5A, 0–30 min). In addition, Tel1p binding to the short telomere became more pronounced as cells proceeded through the cell cycle, and unlike binding of Est1p and Est2p, this binding remained high even at the end of the cell cycle (Figure 5A, 90 min). There was ~7-fold more Tel1p binding to short VII-L than to WT telomeres from late S through the end of the cell cycle (Figure 5A, 52.5-90 min), by far the highest binding differential for any of the proteins examined in this study. Tel1p did not bind the excised circle during this time period (Table 1).
Tel1p binding to the short VII-L telomere was high even at 90 min when cells were at the end of the cell cycle (Figure 5A, open squares). To determine if this high binding persists, we repeated the synchrony with 20 min time points and followed Tel1p telomere binding in both the first and second cell cycles after induction of VII-L telomere shortening (Figure 5C). We used expression of the G2/M cyclin Clb2p as an indicator of the transition from the first to second cell cycle (Cho et al., 1998) (Figure 5D). By this criterion, mitosis occurred in most cells at ~100 min. We also used RT-PCR to monitor levels of the cell-cycle-regulated RNR1 RNA, which is maximally expressed during late G1/S phase (Cho et al., 1998; Elledge and Davis, 1990)(Figure 5E). By this criterion, in the second cell cycle, cells were in S phase at 120 and 140 min. The time between the peaks of RNR1 RNA indicates that the cell cycle after telomere shortening was ~100 min, similar to the length of the cell cycle in the control strain. Thus, telomere shortening did not cause a cell-cycle arrest.
Tel1p binding to the short VII-L telomere not only remained high as cells moved into the second cell cycle after telomere shortening but also continued to increase as cells proceeded through a second S phase (Figure 5C, squares). In contrast, Tel1p binding to WT-length VI-R telomere in the same cells was low and limited to late S phase throughout the two cell cycles (Figure 5C, triangles).
The Mre11-Rad50-Xrs2 (MRX) complex and Tel1p bind sequentially, early, and transiently to DSBs (Lisby et al., 2004; Nakada et al., 2003; Shroff et al., 2004). In cells expressing carboxyl truncation alleles of Xrs2p, the MRX complex is intact and binds DSBs but Tel1p binding to breaks is not detected (Lisby et al., 2004; Nakada et al., 2003). The xrs2-664 allele, which lacks the terminal 190 amino acids from the 884 amino acid Xrs2p, has normal DSB repair but short telomeres (Shima et al., 2005). We introduced the xrs2-664 mutation into the inducible short telomere strains. Tel1p-HA binding to short telomere VII-L was dramatically reduced (Figure 6A, open triangles), and the preferential binding of Est2p-Myc (Figure 6B) and Est1p-Myc (Figure S7) was lost in xrs2-664 cells. In addition, the overelongation caused by overexpression of Tel1p (Figure S6) was lost in cells expressing xrs2-664 (data not shown). Together with the results from the tel1Δ strain (Figure 2C, F), these data argue that high levels of Tel1p binding to the short VII-L telomere are required for preferential lengthening by telomerase.
Although short telomeres are preferred substrates for telomerase from yeast to mammals, there is virtually no mechanistic understanding for how this preference is accomplished. Here we show that in yeast two telomerase subunits, Est2p and Est1p, show strong preferential binding to a short telomere (Figure 2). This preference was manifest specifically in late S/G2 phase when telomerase is active (Diede and Gottschling, 1999; Marcand et al., 2000). At WT-length telomeres, Est2p is telomere associated throughout much of the cell cycle, including in G1 and early S phase when telomerase is not active, whereas Est1p comes to the telomere only in late S/G2 phase. This pattern led us to propose that Est1p may activate the Est2p that is already telomere bound (Taggart et al., 2002). This model predicts preferential binding of Est1p, but not Est2p, to short telomeres, yet both proteins were enriched at the short VII-L telomere (Figures 2A and 2D). Our data are consistent with a model in which Est1p escorts a second Est2p to the telomere (Taggart and Zakian, 2003) promoted by a specific interaction between telomere-bound Cdc13p and Est1p (Bianchi et al., 2004; Evans and Lundblad, 1999; Qi and Zakian, 2000). This model suggests that the active form of telomerase is a dimer (Lin and Blackburn, 2004; Prescott and Blackburn, 1997). Alternatively, because over half of the Est2p that was telomere associated in G1 phase was not at the very end of the chromosome (Figure 4E and Table 1) and G1-bound Est2p did not increase at short telomeres (Figure 2A), G1-bound Est2p may have a noncatalytic role in telomere maintenance (Vega et al., 2007).
Ku and Cdc13p, two structural proteins with positive roles in recruiting telomerase to telomeres, were not enriched at the short telomere nor was their cell-cycle pattern of binding affected by telomere length (Figures 3A and 3B). Similar levels of Cdc13p binding to short telomeres suggest that short and WT-length telomeres have G tails of similar lengths arguing against a model in which only short telomeres acquire long G tails and bind Cdc13p in late S/G2 phase (Negrini et al., 2007). Thus, if these proteins promote the preferential recruitment of telomerase to short telomeres, they likely do so by being selectively modified at short telomeres. Indeed, Cdc13p is phosphorylated by Tel1p, and this phosphorylation is critical for its ability to promote telomerase-mediated telomere lengthening (Tseng et al., 2006).
The negative regulators Rif1p (Figure 3C) and Rif2p (Figure 4A) also bound robustly to the short telomere so their absence is not required for preferential binding of telomerase. In fact, as cells neared the end of the cell cycle, Rif1p and Rif2p binding to the short VII-L telomere increased, respectively, ~10- and 2-fold, just as they did at WT-length telomeres, even though the short VII-L telomere is a preferred substrate for telomerase for several cell cycles (Marcand et al., 1999). Nonetheless, the 50% reduction in Rif2p at the short telomere throughout the cell cycle (Figure 4A) may be sufficient to promote Tel1p binding. This possibility is supported by the finding that targeting a single Rif2p molecule to the VII-L telomere in a rap1ΔC strain prevents the telomerase-mediated hyper-elongation that is usually seen in this background, whereas Rif1p is not as effective at inhibiting telomerase by this assay (Levy and Blackburn, 2004). In addition, Rif2p, but not Rif1p, inhibits telomere addition to an induced DSB generated adjacent to a short tract of telomeric DNA (Diede and Gottschling, 1999; Frank et al., 2006).
Telomeres are very short in tel1Δ cells or in cells expressing a kinase-inactive Tel1p (Greenwell et al., 1995; Mallory and Petes, 2000). In fact, tel1Δ telomeres are similar in length to those telomeres that are preferentially elon-gated in a WT cell (Teixeira et al., 2004). Given that telomerase bound preferentially to short telomeres (Figures 2A and 2D), one might expect that tel1Δ telomeres would also have high levels of telomerase. However, Est2p and Est1p binding was low at tel1Δ telomeres (Goudsouzian et al., 2006)(Figures 2C and 2F), and the preferential binding of Est2p and Est1p to the short telomere was lost in this strain (Figures 2C and 2F). Taken together, these data suggest that Tel1p is needed to recruit, not activate, telomerase. Tel1p overexpression resulted in telomere lengthening (Figure S6) without any apparent effect on cell-cycle progression, consistent with the possibility that Tel1p directly promotes telomerase mediated telomere lengthening.
We also examined Tel1p binding to short and WT-length telomeres. We found very low Tel1p binding to WT-length telomeres, and this binding occurred in late S/G2 phase (Figure 5A). The low but significant binding of Tel1p in late S/G2 phase to WT-length telomeres (Figure 5B, inset) may occur at the ~7% of WT-length telomeres that are lengthened in a given cell cycle (Teixeira et al., 2004). These data disagree with an earlier study that found robust Tel1p binding to WT-length telomeres throughout the cell cycle except in late S phase (Takata et al., 2004). We obtained the epitope-tagged Tel1p used in this earlier study and found that its binding to WT telomeres was also low and limited to late S/G2 phase (data not shown). Another group also finds low Tel1p binding to WT-length telomeres (K. Runge, personal communication).
The temporal and quantitative binding pattern of Tel1p to short telomeres was dramatically different from its binding to WT-length telomeres. At its peak binding, Tel1p was enriched ~7-fold at the short VII-L telomere (Figure 5). Using a telomerase-deficient strain, another group also finds preferential binding of Tel1p to short telomeres (K. Runge, personal communication). In addition to strong Tel1p association with short telomeres late in the cell cycle, Tel1p was present at low but significant levels at the short VII-L telomere from G1 to early S phase. Moreover, high Tel1p binding persisted into a second cell cycle (Figure 5C), consistent with the preferential elongation of the short VII-L telomere for multiple cell cycles (Marcand et al., 1999). Preferential binding of Tel1p, Est1p, and Est2p was lost in cells expressing xrs2-664 (Figure 6 and Figure S7), a truncation that does not impair either MRX complex formation or DSB repair but does result in short telomeres (Shima et al., 2005). Tel1p must therefore act in cis to promote telomerase. In combination with our data on reduced telomerase binding to short telomeres in tel1Δ cells (Figures 2C and 2F), these results make a strong argument that Tel1p binding is critical to promote preferential lengthening of short telomeres. Indeed, in both tel1Δ and rad50Δ strains, the rate of lengthening of short telomeres is considerably slower that in WT cells (Brevet et al., 2003; Marcand et al., 1999).
We propose that Tel1p binding to short (and, more rarely, WT length) telomeres marks them for elongation by telomerase. Tel1p may be recruited preferentially to short telomeres by their relatively reduced Rif2p content (Figure 4A) and/or by other features of short telomeres. Because the kinase activity of Tel1p is needed for its role in maintenance of yeast telomeres (Mallory and Petes, 2000), telomere-bound Tel1p likely acts by phosphorylating one or more telomere binding proteins to make telomeric chromatin more accessible to telomerase. Although Rif2p has no ATM consensus phosphorylation sites, telomerase itself, Rif1p, Ku, and Cdc13p as well as other telomere binding proteins, such as Stn1p and Ten1p, are candidates for Tel1p phosphorylation. Indeed, Tel1p phosphorylates Cdc13p at multiple sites in vitro and in vivo, and mutation of some of these sites leads to an est phenotype (Tseng et al., 2006). Thus, telomere binding proteins with positive or negative effects on telomerase may be selectively modified rather than selectively removed or augmented at short telomeres.
Finally, the data presented here demonstrate that yeast do not see short telomeres as DSBs. First, the generation of a single DSB results in release of Ku from telomeres and Ku binding to the DSB (Martin et al., 1999), but the amount of Ku80p bound to short telomere VII-L did not increase, nor did the amount bound to WT VI-R decrease, in response to shortening of the VII-L telomere (Figures 3A and 3D). Second, Tel1p binding to a DSB is transient (Lisby et al., 2004), while its association at the short VII-L telomere persisted through at least two cell cycles (Figure 5C). Most importantly, there was no cell-cycle delay in entry into the second cell cycle in cells with the short VII-L telomere (Figures 5D and 5E). In contrast, either the induction of a single DSB immediately internal to the VII-L telomere (Sandell and Zakian, 1993) or the presence of multiple critically short telomeres as in an estΔ strain (Enomoto et al., 2002; IJpma and Greider, 2003) results in a checkpoint-mediated cell-cycle arrest.
See the Supplemental Data for additional experimental details.
All strains were bar1Δ::KAN1 PCR-mediated deletion (Lorenz et al., 1995) derivatives of the W303 strain Lev220 (Marcand et al., 1999) (Table S1). All gene disruptions were confirmed both by Southern analysis and by phenotype. The tel1Δ::HIS3 deletion has been described previously (Tsukamoto et al., 2001). The xrs2-664 mutation was generated by two-step gene replacement using pMS392 kindly provided by Dr. Shinohara (Shima et al., 2005). The Est1p, Est2p, Cdc13p, and Ku80p Myc epitope-tagged proteins were described previously (Fisher et al., 2004; Taggart et al., 2002; Tsukamoto et al., 2001). Tel1p was internally tagged with three HA epitopes as described (Mallory and Petes, 2000), and as reported therein, cells expressing Tel1p-HA had telomeres that were ~50 bp shorter than cells expressing untagged Tel1p (Figure S5). Cells expressing N-terminally tagged 5HA-Tel1p (Takata et al., 2004), either in the original strain background or as back-crossed to Lev220, had telomeres that were ~100 bp shorter than their no-tag counterparts (Figure S5). We used the more functional internally tagged Tel1p for all ChIP experiments. Rif1p and Rif2p were tagged with nine Myc epitopes separated from their carboxyl termini by eight glycine residues. Although the glycine linker improved the ability of both Rif1p and Rif2p to maintain telomere length as compared to nonlinkered Myc-tagged alleles (data not shown), tagged Rif1p was not fully WT as demonstrated by a modest increase in telomere length compared to the parental strain (Figure S2). For all strains, a control strain B was generated by exposing the experimental strain A to galactose and screening for cells that had lost the ability to grow in the absence of uracil. By the completion of this screen (~75 generations of cell growth), telomerase had re-extended telomere VII-L to the WT length of ~300 bp (Figure S1).
Paired experimental (strain A) and control (strain B) cells were grown at 30°C in rich media plus 2% raffinose to an OD660 of 0.3. α factor was added to a final concentration of 160 μM, and cells were cultured until at least 90% were unbudded. Dry galactose (Sigma) was added to a final concentration of 1%, and cells were returned to 30°C for an additional 3 hr. Cells were then transferred to rich medium plus glucose and α factor for 15 min before the α factor was removed by filtration and cells released into the cell cycle at 24°C by the addition of protease (Sigma; 150 μg/ml final concentration). Samples were taken at least every 15 min, except for the two-cell-cycle experiment with Tel1p (Figures 5C and 5D) when samples were taken at 20 min intervals. Samples were processed for flow cytometry, Southern blot analysis, and ChIP.
ChIP analysis was performed as described (Fisher et al., 2004; Goudsouzian et al., 2006), except for HA-tagged Tel1p, which was immuno-precipitated with monoclonal α-HA antibody (Santa Cruz), and Rap1p, which was immunoprecipitated as described in Alexander and Zakian (2003) with an affinity-purified polyclonal Rap1p antiserum prepared as in Conrad et al. (1990) and protein A Dynabeads (Invitrogen). PCR primer sequences are provided in Table S2. ChIP samples were quantitated by real-time PCR and normalized to input DNA; data are expressed as the fold enrichment over the amount of a nontelomeric sequence (ARO1) in the immunoprecipitate. In a given synchrony, samples from each time point were amplified in duplicate or triplicate to obtain an average value for each sample. In addition, each synchrony was repeated at least three times; the data are presented as the mean of the three or more synchronies plus or minus standard deviations. Where applicable, a two-tailed Student’s t test was used to determine statistical significance (p value ≤ 0.05). For Est1p (Figures 2D and 2E), the data displayed a nonnormal distribution; because we had a sufficient number of repeats (n = 8), error bars represent the standard deviation calculated separately for values above and below the mean.
The tetO2-TEL1-HA strain was constructed by first replacing the kanMX4 selectable marker of pCM224, the tetracycline operator-based promoter substitution cassette plasmid (Belli et al., 1998) deposited at EUROSCARF (Frankfurt, Germany), with the natMX4 selectable marker by EcoRI and PacI restriction enzyme digestion. Two independent strains were transformed with the PCR product to create yCTT226 and yCTT227, respectively, containing HA-tagged TEL1 controlled by the tetracycline-regulated (tetR) constitutive CYC1 promoter.
RNA was extracted from an equal number of cells from each time point by using the RNeasy Mini Kit (QIAGEN). RT-PCR was carried out by using the SuperScript One-Step RT-PCR with Platinum Taq Kit (Invitrogen). RNR1 RNA was amplified with the primers 5′-CATAGAC CAATTGCTTTGGGGTG-3′ and 5′-GACTCTTGAACGTTTAATTCTGCC-3′. For normalization (Spellman et al., 1998), TUB3 RNA was amplified with the primers described (Lin and Zakian, 1995).
We thank A. Chan for technical assistance, L. Kruglyak for advice on statistics, E. Gilson, A. Matsuura, T. Petes, and M. Shinohara for strains, L. Breeden for suggesting the RNR1 experiment, D. Kellogg for α-Clb2p antibody, C. DeCoste for help with flow cytometry, K. Runge for sharing data before publication, and J.-B. Boule, I. Cheung, K. Daumer, M. Mateyak, and C. J. Webb for their comments on the manuscript. This work was supported by National Institutes of Health grant GM43265 (VAZ) and postdoctoral fellowships to M.S. (NIH GM068218) and C.T.T. (NSF DBI-0610300).
Supplemental Data include Supplemental Experimental Procedures, Supplemental References, seven figures, and two tables and can be found with this article online at http://www.molecule.org/cgi/content/full/27/4/550/DC1/.