Binding of Multiple Rap1 Proteins Stimulates Chromosome Breakage Induction during DNA Replication

Summary

Telomere length is maintained primarily through equilibrium between telomerase-mediated lengthening and the loss of telomeric sequence through the end-replication problem. In budding yeast Rap1 protein binds to telomeric TG repeat and negatively regulates telomerase recruitment in a dosage-dependent manner. In this paper we provide evidence suggesting an alternative Rap1-dependent telomere shortening mechanism in which binding of multiple Rap1 proteins mediates DNA break induction during DNA replication. This process does not involve recombination events; therefore, it is distinct from loop-mediated telomere trimming.

Introduction

Telomeres are specialized nucleoprotein complexes at the ends of linear eukaryotic chromosomes. The DNA component of telomeres typically comprises a double-stranded DNA (dsDNA) region of a tandem repeat and a 3’ protruding single-stranded DNA (ssDNA) region of the G-rich strand [1,2]. Both the dsDNA and ssDNA regions are covered with sequence-specific binding proteins. Telomeres protect chromosome ends from degradation or fusion [2,3]. Telomeres also promote DNA replication at the chromosome ends. Since conventional DNA polymerases cannot complete DNA synthesis at telomeres, linear chromosomes shorten progressively with every round of cell division. In most eukaryotes, continuous telomere shortening can be counteracted by telomerase [4].

The length of the duplex telomeric repeat is kept within a relatively narrow range in a cell-type specific manner [1]. In cells that express telomerase, telomere length homeostasis results from a balance between telomerase-dependent telomere addition and telomere shortening. The average telomere length varies between 5 and 15 kb in human, whereas much shorter telomeres (~300 bp) are maintained in the budding yeast Saccharomyces cerevisiae. Telomere shortening results from incomplete DNA end replication and 5’-resection (end-replication problem) [2,5]. For example, in the absence of telomerase, telomeres of human and budding yeast cells lose 50–200 and 3–5 nucleotides per round of DNA replication, respectively [6–9]. Telomere shortening can also occur in a more accelerated manner called telomeric rapid deletion (TRD), some of which involves intrachromosomal recombination events between telomeric repeats [10–13].

The structure of telomeres in budding yeast is typical of most eukaryotes, except that the repeat unit (abbreviated TG_1-3) is heterogeneous [2]. As in other eukaryotes, telomeres consist of double-stranded and single-stranded units in budding yeast. The G-rich strand extends to form a single-strand tail, which is covered with Cdc13 in complex with Stn1 and Ten1 [14–16]. The Cdc13-Stn1-Ten1 complex acts as a telomere cap to protect telomeres from degradation [2]. Telomerase comprises the catalytic subunit Est2, two accessary subunits (Est1 and Est3) and the RNA component Tlc1 in budding yeast [2,4]. Cdc13 interacts with Est1 and contributes to telomerase recruitment to chromosome ends [2,4]. Double-stranded telomeric DNA repeats are bound by the sequence-specific binding protein Rap1.

The budding yeast Rap1 protein is an essential protein involved in many diverse processes including transcription and telomere maintenance [17]. Rap1 consists of three conserved domains: a BRCT domain in the N-terminal region, a centrally located DNA-binding domain with two Myb-like folds, a transcription activation (TA) domain, and an independent C-terminal domain called RCT (Rap1 C-terminal) [17]. The central region, containing two Myb domains and the TA domain, plays an essential role in cell viability by regulating transcriptional activation of various genes [18]. The N-terminus has been shown to potentiate DNA bending of the Rap1-DNA complex [19]. The RCT domain regulates localization of Sir3 and Sir4 and promotes transcriptional silencing [20]. This domain also forms a complex with and recruits Rif1 and Rif2 to telomeres [21–24]. Targeting of the RCT, Rif1 or Rif2 to a telomere induces telomere shortening to an extent that is proportional to the number of targeted molecules [25,26]. This observation has led to the “protein counting model” of telomere length control, in which increasing numbers of telomere-bound Rap1 molecules negatively regulates telomerase-mediated telomere extension. The “protein counting model” has been defined by recent studies. As in other eukaryotes, the checkpoint protein kinases Mec1 and Tel1 control telomere length in budding yeast [2]. Although Mec1 is a central checkpoint kinase in the response to DNA replication block and DNA damage [27], it has only minor functions in telomere homeostasis. In contrast, Tel1 plays a major role in telomere length maintenance in budding yeast [28]. Tel1 localizes to short telomeres as well as DNA double-strand breaks (DSBs) [29–32]. In turn, Tel1 recruits telomerase to short telomeres and promotes telomere addition [32–34]. Telomere extension increases Rap1 binding sites at telomeres. Rap1 collaborates with Rif1 and Rif2 and inhibits localization of Tel1 to DNA ends [35,36]. Thus, Rap1 cooperates with Rif1 and Rif2 to impede telomerase recruitment, therefore negatively regulating telomere length.

The G-rich strand of telomeric TG repeats can fold into G-quadruplex (G4) structures, which could cause replication fork blocks and induce DNA breaks [37]. Indeed, internal tracts of TG repeats are unstable and spontaneously converted to telomeres [38]. Rap1 binds internal tracts of TG repeat sequence as well as telomeres [21,22,35]. Since Rap1 binds to duplex DNA, binding of Rap1 to TG repeat sequence could inhibit G4 structure formation. However, there is evidence suggesting that Rap1 also binds to single-stranded DNA and stimulate G4 formation [39]. Alternatively, it is formally possible that DNA binding of Rap1 promotes DNA break induction independently of the G4 structure.

In this study we provide evidence supporting that binding of multiple Rap1 proteins is involved in chromosome breakage during S phase. We devised systems to determine whether DNA binding of Rap1 stimulates DSB induction. We found that Rap1 mediates DSB induction at both non-telomeric and telomeric regions. While Rap1 operates together with Rif1 and Rif2 to inhibit telomerase recruitment, Rap1 acts independently of Rif1 or Rif2 function to induce chromosome breakage. Rap1 tethering prompts DSB generation in a copy number-dependent manner during DNA replication. While Rap1-mediated DSB induction shortens artificially extended telomeres, this telomere shortening does not involve homologous recombination events. Our results suggest that Rap1 contributes to telomere homeostasis by inducing DNA breaks.

Results

Rap1-dependent chromosome truncation at TG repeats

We examined whether Rap1 is involved in DSB induction, thereby converting internal tracts of TG sequence to telomeres. To this end, we placed an 81 bp (TG₈₁) or a 250 bp telomeric TG (TG₂₅₀) repeat sequence between the KanMX marker and the K. lactis URA3^Kl gene at the ADH4 locus (Fig 1A). The TG₈₁ or TG₂₅₀ sequence, derived from endogenous telomeres, contains four or twelve Rap1 binding motifs, respectively [36,40]. There is no essential gene from the ADH4 locus to the chromosome end. Eukaryotes utilize two major pathways for DSB repair; homologous recombination (HR) and non-homologous end joining (NHEJ) [27]. In budding yeast, HR is the central DSB repair pathway. However, HR cannot efficiently repair centromere-proximal DSB ends generated between KanMX and URA3^Kl because there is no homologous donor sequence available (see below). Instead, telomerase-dependent telomere addition occurs at DNA ends with telomeric TG repeat sequence nearby [25,40], a phenomenon which is referred to as “telomere healing” (Fig 1B). Although URA3 cells cannot proliferate on medium containing 5-fluoroorotic acid (5-FOA), ura3 mutant cells can [41]. Therefore, if a DNA break is induced within or near the TG sequence, telomere formation can eliminate the distal portion including the URA3^Kl marker gene, therefore generating 5-FOA resistant colonies. Cells were first maintained in medium selective for URA3 cells and then transferred to non-selective medium. Saturated cultures were diluted and spread on 5-FOA plates to monitor the rate of URA3^Kl marker loss (Fig 1C). The URA3^Kl marker was stably maintained if there was no TG repeat (TG₀) sequence. Introduction of the TG₂₅₀ repeat sequence, however, stimulated loss of the URA3^Kl marker very efficiently (19,000-fold). Placement of the TG₈₁ repeat increased URA3^Kl marker loss (430-fold) but much less efficiently compared with the TG₂₅₀ sequence. All of the twenty Ura^- cells examined possessed a telomere near the TG₈₁ or TG₂₅₀ sequence (S1 Fig)[38]. Consistent with telomerase-dependent telomere addition at TG sequences, inactivation of the RAD52-dependent HR pathway did not affect URA3^Kl marker loss [40] (Fig 1C). We examined whether chromosome breakage occurs near the TG₂₅₀ repeat sequence by Southern blotting analysis. Cells carrying the TG₀ or TG₂₅₀ cassette were first cultured in medium selective for URA3 cells and then transferred to non-selective medium. Introduction of the TG₂₅₀ repeat accumulated cells containing a DNA end nearby (Fig 1D). Several lines of evidence have established the model in which persistent DNA fork stalling leads eventually to DSB induction [42]. Replication forks slow during their passage through telomeric TG tracts [43]. We confirmed that replication forks paused at the TG₂₅₀ repeat sequence by two-dimensional gel electrophoresis analysis (Fig 1E and S2 Fig).

We investigated whether Rap1 is required for URA3^Kl marker loss. Since the RAP1 gene is essential for cell proliferation, we examined the effect of partial Rap1 depletion using a copper-inducible rap1 degron, rap1-(Δ) [44]. Consistent with an essential role of Rap1 in cell proliferation, rap1-(Δ) mutants grew very poorly in the presence of 0.5 mM CuSO₄ (Fig 2A) as the expression of Rap1 degron protein was decreased (Fig 2B). In contrast, incubation with 0.05 mM CuSO₄ did not significantly affect cell proliferation (Fig 2A) although the Rap1 expression level was decreased (Fig 2B). We thus investigated the effect of rap1-(Δ) mutation on URA3^Kl marker loss in the presence of 0.05 mM CuSO₄ (Fig 2C). Partial Rap1 depletion was found to decrease URA3^Kl marker loss. These results are consistent with the hypothesis that Rap1 promotes DSB induction in a copy number-dependent manner.

Rap1 comprises three conserved domains: a BRCT domain in the N-terminal region, a central region with DNA-binding Myb and TA domains, and a C-terminal RCT domain [17] (Fig 1C). Although the central region is essential for cell viability, the BRCT or RCT domain is not [17,45,46]. To determine the role of the BRCT or the RCT domain, we examined the effect of rap1-ΔN or rap1-ΔC mutation on loss of the URA3^Kl marker. However, neither the N-terminal nor the C-terminal deletion significantly affected the generation of Ura^- cells (Fig 1C), raising a possibility that the central region of Rap1 protein is involved in DSB induction.

Stimulation of homologous recombination nearby by tethering of Rap1 on chromatin

The above results support the model in which binding of multiple Rap1 proteins results in DSB induction. However, it remains possible that Rap1 binding mediates the formation of G-quadruplex (G4) structures at TG repeats, which could cause replication fork blocks and induce DNA breaks [37]. The TG₈₁ and the TG₂₅₀ repeat sequence can potentially form two and eight G4 structures, respectively [47]. Moreover, Rap1 binding could stimulate telomerase activity at DNA ends. To exclude these possibilities, we set up a system that recruits Rap1 to non-TG sequences (Fig 3A). We constructed a LacI-Rap1 fusion (Fig 3B), which restores proliferation to rap1Δ mutants (S3 Fig) and functions as a negative regulator of telomere length (see below). Expression of LacI-Rap1 fusion was driven from the GAL1 promoter in medium containing 2% galactose and 0.5% glucose (galactose hereafter), thereby maintaining the expression level of LacI-Rap1 protein similar to that of endogenous Rap1 protein (S4 Fig). To target LacI-Rap1 to non-TG sequences, we inserted cassettes with different copy numbers of the LacI-binding sequence (lacO) (Fig 3A). No putative G4 forming sequence was found on the lacO sequences [47]. One LacI homodimer can bind to each lacO operator [48]; therefore, one lacO copy can be covered with two LacI-Rap1 molecules. However, the LacO₄ cassette behaved like ~80 bp of telomeric TG sequence containing four Rap1 binding sites after LacI-Rap1 expression (see below). The average telomere length is ~300 bp and the Rap1 binding motif appears every 20 bp at telomeres [2]. Thus, the LacO₁₆ repeat in the presence of LacI-Rap1 could correspond to wild-type length telomere (~320 bp of TG repeat sequence) in terms of Rap1 binding. We introduced a 3’-terminal truncation of URA3^Kl adjacent to the lacO sequences at the ADH4 locus and a 5’-terminal truncation of URA3^Kl on a different chromosome (Fig 3A). In this system DNA breaks at the lacO sequences can be repaired by homologous recombination, generating the full-length URA3^Kl marker gene. We thus determined the homologous recombination frequency between the truncated ura3^Kl genes to estimate LacI-Rap1-mediated DSB induction. Cells carrying pGAL-LacI-RAP1 or the control vector were initially grown in sucrose and then incubated with galactose to express LacI-Rap1. Saturated cultures were diluted and plated to score Ura⁺ cells (Fig 3A). LacI-Rap1 expression stimulated interchromosomal recombination more efficiently when cells contained longer lacO sequences (Fig 3B). For example, LacI-Rap1 expression led to 2,000-fold higher recombination events in cells containing the LacO₁₆ cassette compared to cells containing no lacO (the LacO₀ cassette). LacI alone stimulated recombination near the LacO₁₆ cassette but much less efficiently than LacI-Rap1 fusion (Fig 3C). The observed effect is not specific to LacI-fusion, since TetR-Rap1 fusion also stimulated recombination near the tetO repeat (S5 Fig).

Since the N-terminus or C-terminus of Rap1 was dispensable for TG-mediated chromosome truncation (see Fig 1C), we examined whether the central region of Rap1 stimulates interchromosomal recombination (Fig 3C). We constructed a LacI-Rap1 fusion lacking both the N- and C-termini of Rap1, named LacI-Rap1 (224–663) (Fig 3B). The LacI-Rap1 (224–663) fusion stimulated recombination strongly compared with LacI alone but weakly compared with LacI-Rap1 (Fig 3C). The expression level of LacI alone and LacI-Rap1 (224–663) was not significantly differently from that of LacI-Rap1 (S4 Fig). Although the central domain of Rap1 triggered recombination less efficiently than the full-length Rap1 protein, deletion of the central region abolished recombination stimulation; LacI-Rap1 (ΔM1-M2-TA) behaved similarly to LacI alone (S6 Fig). These results support the idea that the central region of Rap1 plays a key role in DSB induction. The central region of Rap1 consists of two Myb domains and a TA domain. Neither Myb nor TA domain was specifically involved in recombination stimulation, suggesting that the overall structure of the central region is critical for DSB induction (S6 Fig). The C-terminal region of Rap1 mediates interaction with Rif1 or Rif2 for telomere homeostasis and Sir3 or Sir4 for transcriptional repression [2,17]. In agreement with the finding that the C-terminus is dispensable for Rap1-mediated DSB induction, neither rif1Δ rif2Δ nor sir3Δ sir4Δ double mutation affected recombination (Fig 3D). It is possible that any transcriptional activation protein stimulates DSB induction similar to Rap1. Gal4 is a potent transcription activator and the calculated molecular mass of Gal4 is similar to that of Rap1. We examined the effect of LacI-Gal4 expression on DSB induction (Fig 3C). LacI-Gal4 expression increased recombination compared with the control vector but behaved like LacI alone. The expression level of LacI-Gal4 was similar to that of LacI-Rap1 (S4 Fig). Both Rap1 and LacI bind to the respective consensus sequence with a high affinity (Kd = ~1x10^-11M) [49,50]. One explanation could be that DSB induction involves tight DNA binding. We addressed this possibility by using the LacI** variant that has weak affinity for the binding sequence but does not impair the accumulation at lacO repeat sequences [51]. We found that LacI**-Rap1 protein was defective in DSB induction (Fig 3C) although the expression of LacI-Rap1** was similar to that of LacI-Rap1 (S4 Fig). Thus, anchoring of Rap1 appears to promote DSB induction rather specifically.

Induction of DNA breakage by tethering of Rap1 on chromatin

We investigated whether LacI-Rap1 tethering induces DNA breaks at the LacO₁₆ repeat by Southern blot analysis. We used a strain carrying the LacO₁₆ cassette between KanMX and URA3^Kl at the ADH4 locus (Fig 4A). Cells transformed with pGAL-LacI-RAP1 or pGAL-LacI were grown in sucrose and then transferred to galactose medium for 4 hr. LacI-Rap1 expression induced DNA breakage near the LacO₁₆ repeat whereas no apparent cleavage was detected with LacI expression (Fig 4B). Neither Ku-dependent NHEJ nor Rad52-dependent HR significantly affected DNA breakage detection (Fig 4B). This observation is consistent with the view that NHEJ is a minor pathway in budding yeast [27] and there is no homologous donor sequence available for DNA breaks generated between KanMX and URA3^Kl. It was estimated that 3% of cells received a DNA break at the LacO₁₆ locus 4 hr after LacI-Rap1 expression (S7 Fig).

The terminal deoxynucleotidyl transferase (TdT) has been used to end-label DNA ends including DSBs and telomeres [52]. To confirm that LacI-Rap1 tethering generates DNA breakage, we used an assay by combining TdT-mediated end-labeling and PCR (Fig 4C and S8 Fig). In this assay PCR amplification detects the addition of G-tracts at 3’-DNA ends near the LacO₁₆ repeat. PCR amplified DNA fragments that correspond to DSB induction at the LacO₁₆ repeat in both directions after LacI-Rap1 expression whereas no discrete band was observed after LacI expression, indicating that LacI-Rap1 expression results in DSB induction near the LacO₁₆ repeat.

We next addressed in which cell cycle stage LacI-Rap1 expression induces DNA breaks by the TdT-based PCR assay. Cells were grown in sucrose and treated with α-factor or nocodazole to synchronize in G1 or G2/M phase, respectively (Fig 5A). Synchronized cells were then incubated with galactose to induce LacI-Rap1 expression. No DSB induction was detected in G1 or G2/M-arrested cells. We next examined whether DNA breakage occurs in S phase (Fig 5B). Cells arrested with α-factor in G1 were incubated with galactose to express LacI-Rap1 fusion. Cells were then released from α-factor arrest or remained arrested. DNA flow cytometry analysis confirmed that cells underwent S phase after α-factor release (Fig 5B). DSB induction was detected in S phase after α-factor release but not in G1-arrested cells. Thus, LacI-Rap1 expression leads to DSB induction at the LacO₁₆ repeat during S phase. To address whether DSB induction is coupled with DNA replication, we examined the effect of temperature-sensitive cdc17-1 mutation on break induction during S phase (Fig 5C). CDC17 encodes a catalytic subunit of DNA polymerase α. Wild-type or cdc17-1 mutants carrying pGAL-LacI-RAP1 were arrested with α-factor in G1 and incubated with galactose at the restrictive temperature and then released from α-factor arrest. We then monitored DSB induction by Southern blotting analysis (Fig 5C). While wild-type cells underwent DNA replication, cdc17-1 mutants arrested in early S phase (Fig 5C). The cdc17-1 mutation suppressed DSB induction. These results show that replication fork progression in S phase is required for DSB induction.

We analyzed replication fork progression at the LacO₁₆ locus after LacI-Rap1 or LacI expression by two-dimensional gel electrophoresis (Fig 6A and S9 Fig). Cells carrying pGAL-LacI-RAP1, pGAL-LacI, or the control vector were initially grown in sucrose and then incubated with galactose for 4 hr to induce LacI-Rap1 or LacI expression. There was no replication fork pausing in cells carrying the control vector. Experiments using two different sets of restriction enzymes confirmed that replication forks pause at the LacO₁₆ repeat. Replication fork pausing was seen after both LacI expression and LacI-Rap1 expression but corresponding signals were only two-fold more intense in cells expressing LacI-Rap1 than in those expressing LacI alone. As described above, however, LacI-Rap1 expression resulted in DNA breakage much more robustly than LacI expression (Fig 4B and S7 Fig); LacI-Rap1 was estimated to promote DSB induction ~100-fold more strongly than LacI alone (Fig 3C). Thus, fork stalling per se did not fully explain the mechanism of DSB induction after LacI-Rap1 expression.

The ATR/Mec1 checkpoint pathway has been proposed to facilitate replication progression and prevent chromosome breakage during replication stress [42,53]. We tested the possibility that Rap1 impairs Mec1 function that prevents chromosome breakage (Fig 6B). If this were the case, LacI expression could generate DNA breaks as efficiently as LacI-Rap1 expression in mec1Δ mutants. However, DSB induction was still hardly detectable in mec1Δ mutants after LacI expression. Moreover, the introduction of mec1Δ mutation did not elevate DSB induction after LacI-Rap1 expression (Fig 6B). Mrc1 and Tof1 stabilize DNA replication forks and contribute to the activation of the Mec1 pathway during replication stress [54–57]. Previous studies have suggested that Mrc1 and Tof1 have a role in telomere stability in addition to fork progression [58–60]. We examined the effect of mrc1Δ or tof1Δ mutation on DNA break induction at LacO₁₆ after LacI-Rap1 expression (Fig 6C). Neither mrc1Δ nor tof1Δ mutation increased the frequency of DSB induction. Thus, inactivation of Mec1 checkpoint function does not affect Rap1-induced DSB induction.

Rap1-mediated DSB induction at artificially extended telomeres

We addressed whether Rap1 mediates DSB induction at telomeric regions. If the above model were applied to telomeres, Rap1 binding could generate DNA breakage at extended telomeres and truncate them. In parallel, however, Rap1 inhibits telomerase recruitment at extended telomeres and negatively regulates their length. To distinguish between these two different types of telomere shortening, we developed a system that generates an artificially elongated telomere. In this system the LacO₁₆ sequence is integrated between a 33 bp TG sequence (TG₃₃) and the VII-L telomere (Fig 7A). Non-telomeric sequences can be counted as telomeric sequences if Rap1 is anchored [25]. Therefore, the TG₃₃-LacO₁₆-telomere becomes an extended telomere mimic after LacI-Rap1 expression (Fig 7B). Once converted to an extended telomere, telomeres become shorter gradually (3–5 nucleotides per generation) because of the end-replication problem. While telomeric TG sequence is retained, Cdc13-telomere capping blocks DNA degradation [2]. However, once telomere shortening reaches the LacO₁₆ sequence after 60–90 generations, the DNA end loses Cdc13-dependent protection and exonuclease activities start degrading the LacO₁₆ repeat. It is estimated that DNA degradation occurs at the rate of 4 kb/hour [61]. Previous studies showed that 33 bp TG repeat sequences act as a telomere seed in vivo [25,62]. We have shown that DNA ends with 22 bp of TG repeat nearby are efficiently converted to telomeres [63]. If DNA degradation reaches to TG₃₃ repeat sequences, telomere extension occurs using TG₃₃, generating TG-telomeres. However, DSB induction near or within LacO₁₆ could skip gradual telomere shortening that results from the end-replication problem. Thus, DSB induction can be detected as swift conversion from the TG₃₃-LacO₁₆-telomere to TG-telomere. As mentioned above, there is no putative G4 forming sequence on the LacO repeat sequences. This system therefore enables us to examine the effect of Rap1 binding on DSB induction without increasing the length of TG repeat sequence that potentially generates G4 structures.

We first confirmed that targeted LacI-Rap1 to the LacO₄ repeat behaves in a manner identical to Rap1 with respect to telomere length control. We used a strain with the LacO₄ repeat adjacent to VII-L telomere (LacO₄-telomere). Previous studies showed that the GAL4 binding site adjacent to telomeres behaves in a manner similar to a telomere sequence after expression of a Gal4-Rap1 fusion protein [25]. Cells containing the LacO₄-telomere were transformed with pGAL-LacI-RAP1 or pGAL-LacI and cultured in sucrose to stationary phase. The cultures were diluted 1,000-fold in galactose and grown for 24 hr (eight generations) for successive serial dilutions. Cells were collected at each dilution point and the length of LacO₄-telomere was monitored by Southern blot analysis (Fig 7C). When LacI-Rap1 was expressed, the length of the telomere repeat tract was gradually decreased as telomerase activity is down regulated at longer telomeres [25]. In contrast, the expression of LacI alone did not have any effect on telomere length. LacO₄-telomeres became about 80 bp shorter after the serial dilution culture (Fig 7C). Thus, the lacO array sequence is counted as a part of the telomere after LacI-Rap1 expression, supporting the hypothesis that the TG₃₃-LacO₁₆-telomere with LacI-Rap1 coated on the lacO sequence behaves like an extended telomere.

We then examined the effect of LacI-Rap1 or LacI expression on the length of TG₃₃-LacO₁₆-telomeres. Cells containing TG₃₃-LacO₁₆-telomere carrying pGAL-LacI-RAP1 or pGAL-LacI were cultured as above and telomere length was monitored by Southern blot analysis (Fig 7D). As LacI expression had no impact on the length of LacO₄-telomeres (Fig 7C), it did not affect the length of TG₃₃-LacO₁₆-telomere. In contrast, if LacI-Rap1 was expressed, two different types of telomere shortening were observed. First, TG₃₃-LacO₁₆ telomeres (PRE) became gradually shorter consistent with the above finding that LacI-Rap1-covered lacO sequence is counted as a telomere sequence. This observation indicates that telomerase activity is inhibited because TG₃₃-LacO₁₆ telomeres behave like an extended telomere in the presence of LacI-Rap1. Second, LacI-Rap1 expression generated shorter telomeres (TG-telomeres) corresponding to those extended from the TG₃₃ repeat, consistent with the hypothesis that DSB induction occurs near or within the LacO₁₆ repeat. As Rap1-mediated DSB induction occurs during S phase, continuous culturing increased the population size of cells containing TG-telomeres. 50% of the cells converted from the TG₃₃-LacO₁₆ telomere to the TG-telomere after three successive dilutions (24 generations); it was estimated that 2% of cells received a DNA break within or adjacent to the LacO₁₆ repeat per generation.

To confirm that the rapid generation of TG-telomeres does not result from the inhibition of telomerase activity, we compared telomere shortening after LacI-Rap1 expression and telomerase depletion (Fig 7E). We introduced a deletion mutation of EST1, which encodes a regulatory protein for telomerase, in cells containing the TG₃₃-LacO₁₆ telomere. Growth of the est1Δ strains was maintained by wild-type EST1 on a URA3-marked plasmid. Wild-type cells and est1Δ mutant cells were transformed with pGAL-LacI-RAP1 and pGAL-LacI, respectively. To select for loss of the complementing EST1 plasmid, cells were grown on sucrose plates containing 5-FOA. The resulting single colonies were inoculated into galactose medium (1st dilution). After 24 hr culture, it was diluted 1000-fold for successive serial dilutions. Southern blotting analysis confirmed that the inhibition of telomerase activity caused slow telomere shortening (~30 bp during each dilution) but did not promote TG-telomere formation (Fig 7E). As discussed above, the generation of TG-telomeres was dependent on telomerase activity (S10 Fig). To exclude the possibility that the emergence of TG-telomeres results from homologous recombination among telomeric sequences, we tested the effect of rad52Δ mutation on shortening of the TG₃₃-LacO₁₆-telomere after LacI-Rap1 expression (Fig 7F). Cells were first grown in sucrose and cultured continuously in galactose to express LacI-Rap1 after serial dilutions. The rad52Δ mutation did not affect shortening of TG₃₃-LacO₁₆ telomeres after LacI-Rap1 expression. Thus, it is unlikely that Rap1 binding stimulates homologous recombination between telomeric TG repeat sequences. Collectively, our results support the model in which Rap1 binding is involved in telomere shortening by introducing DNA breaks.

Discussion

Rap1 binds to double-stranded telomeric TG-repeat sequences and recruits Rif1 and Rif2 proteins via its C-terminal domain [21–24]. Previous studies have uncovered a negative-feedback mechanism that counts telomere-binding Rap1 protein to control telomere length [25,26]. In this feedback loop, Rap1 collaborates with Rif1 and Rif2 and inhibits the localization of the protein kinase Tel1 to adjacent DNA ends, thereby attenuating the recruitment of telomerase to long telomeres [30–32,34–36]. In this report we have provided evidence suggesting that an alternative Rap1-dependent mechanism operates to trim elongated telomeres. In this telomere shortening mechanism, Rap1 binding coupled with DNA replication promotes DSB induction independently of Rif1 or Rif2. These observations support a model in which Rap1 negatively regulates telomere length through two distinctive mechanisms.

DNA breaks adjacent to telomeric TG repeats can be readily converted to telomeres by telomerase [38,40]. It is therefore complicated to detect DNA breaks at telomeres. We have developed several experimental systems to monitor DSB induction and shown that binding of multiple Rap1 proteins induces DNA breaks at both telomeric and non-telomeric regions. Rap1-mediated DSB induction appears to operate in a dosage-dependent manner. First, longer TG repeats trigger chromosome truncation more efficiently than shorter TG repeats. Second, LacI-Rap1 expression promotes nearby recombination at longer lacO arrays more frequently than at shorter ones. As discussed above, the LacO₁₆ repeat in the presence of LacI-Rap1 could correspond to wild-type length telomere in terms of Rap1 binding. To detect DSB induction at telomeres, we used TG₃₃-LacO₁₆-telomeres, which correspond to ~600 bp long telomeres in the presence of LacI-Rap1. Two percent of TG₃₃-LacO₁₆-telomeres were estimated to receive DNA breaks near or within the LacO₁₆ repeat every cell division in the presence of LacI-Rap1. Although the frequency is low (2% per generation), approximately 50% of cells would receive DNA breaks after 32 generations. DNA ends adjacent to TG repeats, once generated, are protected from DNA degradation by Cdc13-mediated telomere capping [2,35]. Thus, Rap1-mediated DSB induction at telomeres seems to be a physiological phenomenon that keeps telomeres in normal-length ranges, although it remains possible that telomere-specific activity suppresses DSB induction at native telomeres. Rap1 collaborates with Rif1 and Rif2 to inhibit telomerase recruitment [64]. In contrast, Rif1 and Rif2 are dispensable for Rap1-mediated DSB induction. Telomerase inhibition steadily shortens telomeres 3–5 bp per cell division because of the end-replication problem [6,9]. In contrast, DSB induction can delete longer telomere sequence although it may not constantly occur. The extent of telomerase-dependent telomere extension varies at each telomere and is independent of telomere length [65]. Cells take full advantage of different telomere shortening modes to cope with the heterogeneous nature of telomeres.

Replication forks in yeast and other organisms move more slowly through telomeric DNA than non-telomeric regions [43,66–68]. This replication problem is thought to result from the G-rich nature of telomeric DNA, which allows it to form G4 DNA. G4 DNA interferes with DNA replication and generates DNA breaks [37]. In addition to G4 structures, we have shown that binding of multiple Rap1 proteins generates a barrier that stimulates DSB induction during DNA replication. Persistent fork stalling could lead to fork collapse and subsequent DSB induction [42]. Consistent with this model, tight binding of LacI was critical for LacI-Rap1-mediated DSB induction. However, fork stalling by itself does not appear to result in DSB induction at Rap1 bound regions. We found that LacI-Rap1 expression induced DSB formation about 100-fold more efficiently than LacI expression alone whereas the rate of fork stalling after LacI-Rap1 expression was only two-fold higher compared with LacI expression alone. Moreover, larger complex formation with Rif1, Rif2, Sir3 and Sir4 did not increase the frequency of DSB induction. It seems less likely that DSB induction results from DNA bending because the N-terminus is dispensable [19]. Several lines of evidence have shown that paused forks by themselves are relatively stable [69]. Indeed, 40% of the replicating intermediates contained pausing forks at the LacO₁₆ repeat after LacI expression but 0.02% of cells were expected to receive DNA breaks at the LacO₁₆ repeat per generation. The ATR/Mec1 checkpoint pathway has been proposed to prevent chromosome breakage during replication stress [42,53]. However, inactivation of the Mec1 checkpoint pathway did not affect DSB induction after LacI or LacI-Rap1 expression. Thus, Rap1 appears to impair other mechanisms than Mec1 checkpoint function at DNA replication forks. We found that the overall structure of the central Rap1 region plays an important role in DSB induction. It has been shown that the central region of Rap1 inhibits NHEJ at telomeres [70]. This Rap1 region, once arrayed, might disrupt replication complexes as well as repair machinery along the DNA tract. We note that not all telomere-binding proteins have a negative impact on DNA replication. Taz1 in fission yeast and TRF1 in human have been shown to promote DNA replication at telomeres [67,68]. Interestingly, Taz1 and TRF1 possess a Myb domain as does the Rap1 central region. The central region of Rap1 also stimulates DSB induction for meiotic recombination [46]; however, meiotic DSB induction occurs through a different mechanism. Transcription factors including Rap1 generate nucleosome free regions where Spo11 binds to chromatin and catalyzes DSB [71].

Over-elongated telomeres can be shortened to normal length through a mechanism termed telomeric rapid deletion (TRD) [10]. TRD appears to result from an intra-chromosomal recombination event between telomeric repeats because TRD depends on the major recombination protein, Rad52 [10]. In addition, hpr1 mutation, which increases recombination between direct repeats, elevates the rate of TRD [10]. Similar telomere shortening has been observed in other systems including human cell lines and most likely involves homologous recombination-mediated removal of telomere loops [11–13]. While TRD largely depends on Rad52 function, some fraction of TRD was found to occur in a Rad52-independent manner [10]. Since DSBs near telomeric TG sequences are healed by telomere addition, DSB generation at over-elongated telomeres would lead to Rad52-independent TRD. G4 structure formation could promote DSB induction at telomeres [37], thereby trimming telomeres independently of Rad52 function. However, our results suggest that Rap1-mediated DSB induction contribute to Rad52-independent TRD as well. Indeed, TG₃₃-LacO₁₆-telomeres became shorter independently of Rad52 function or the end-replication problem after LacI-Rap1 expression.

In summary, we have provided evidence indicating that binding of multiple Rap1 proteins promotes DNA break induction during DNA replication. Since Rap1 binds extensively at telomeric DNA regions [72], it seems likely that Rap1 binding promotes DSB induction at telomeres. Rap1 binds to telomeres and controls their function in other eukaryotes [73–75]. Given that telomeres consist of repetitive sequences and sequence-specific binding proteins, a similar system may function in other organisms as well.

Materials and Methods

Strains

The strain carrying the TG₀-URA3, TG₈₁-URA3, TG₂₅₀-URA3 or LacO₁₆-URA3 cassette was generated by the pNO-URA3, pT81-URA3, pT250-URA3 or pO16-URA3 plasmid after digestion with NotI and SalI, respectively. The strain carrying the LacO₀-ura3-ΔC, LacO₄-ura3-ΔC, LacO₈-ura3-ΔC or LacO₁₆-ura3-ΔC or TetO₈-ura3-ΔC cassette was generated by the pUN-O0, pUN-O4, pUN-O8, pUN-O16 or pUN-tetO8 plasmid after digestion with EcoRI and SalI, respectively. The strain containing the LacO₄-telomere, TG₃₃-LacO₁₆-telomere, or TG-telomere was generated by the pO4-TG-HO, pTG33-O16-TG-HO or pTG₈₁-HO plasmid after digestion with EcoRI and SalI, respectively. The ura3-ΔN-Hph cassette was introduced into the YER186 locus by a PCR-based method [76]. The copper-inducible RAP1 degron (rap1-(Δ)) construct has been described [44]. The N-terminal deletion of RAP1 (rap1-ΔN) has been described [46]. TG-telomere cells were generated from TG₈₁-HO cells after HO expression [35](See S1 Fig). The mec1Δ, rad52Δ or sml1Δ mutation has been described [77]. The mec1-81 or tel1Δ mutation has been published [29]. The est1Δ mutation has been described [78]. The C-terminal rap1 truncation (rap1-ΔC), rif1Δ and rif2Δ mutation have been described [35]. The rap1-ΔC allele encodes the same truncated Rap1 proteins as the rap1-17 mutation does [26,45]. Disruption of SIR3 was performed as described [79]. The copper inducible protein degradation system was described [80]. The SIR4 disruption plasmid was obtained from Masayasu Nomura. The strains used in this study are listed in S1 Table.

DNA preparation and hybridization probes for Southern blot analysis

The probe that detects the KanMX gene was obtained by NotI-digestion of the pFA6-kanMX4 plasmid [79]. The probe that detects DSB induction was a PCR fusion product of the KanMX coding sequence and the LTE1 locus, which were amplified by the primer pair KSX050/89 and KS3004/3005, respectively. The probe that detects endogenous telomeres or telomere addition at the ADH4 locus has been described [35,78]. DNA probes were DIG-dUTP—or ³²P-labeled by using the DIG prime (Roche) or the Random Primer DNA Labeling Kit (Clontech), respectively. Genomic DNA was purified using a MasterPure yeast DNA purification kit (Epicentre).

TG-mediated URA3 marker loss

Cells from a single colony were fully grown in uracil dropout medium. The culture was then diluted 1000-fold and grown in rich medium for 24 hr. Aliquots of the cultures were diluted and plated on plates containing 5-FOA or non-selectable rich medium. Rates per generation were calculated using the FALCOR program based on the Luria-Delbruck fluctuation analysis [81]. We confirmed that essentially all 5-FOA resistant cells derived from TG₈₁-URA3 or TG₂₅₀-URA3 are Ura^- cells. The promoter of URA3^Kl was replaced with the strong ADH1 promoter (see S1 Text). More than 500 colonies of each TG₈₁-URA3, TG₂₅₀-URA3, TG₂₅₀-URA3 rad52Δ or TG₂₅₀-URA3 rap1-(Δ) cells on 5-FOA plates were tested by replica-plating to uracil drop-out plates. None of them grew on uracil-drop out plates.

URA3 recombination assay

Cells were transformed with TRP1-marked plasmids and grown in tryptophan-dropout medium containing 2% sucrose overnight. The culture was then diluted and grown in tryptophan-dropout medium containing 2% galactose or 2% sucrose and 0.5% glucose. Aliquots of the cultures were diluted and plated on uracil-dropout or rich medium to estimate the URA3 recombination frequency after expression of LacI or LacI-fusion protein. Rates per generation were calculated using the FALCOR program [81].

DNA end detection by dC-tailing

Purified genomic DNA (100 ng) was incubated with one unit of terminal deoxynucleotidyl-transferase (TdT; New England BioLabs) in a supplied reaction buffer supplemented with 0.2 mM dCTP at 37°C for 60 min, followed by PCR using a poly(dG)-oligonucleotide with either the TdT forward or reverse primer. The PCR condition was 33 cycles of denaturation at 94°C for 30 s, annealing at 62°C for 30 s, and elongation at 72°C for 60 s. Sequences of PCR primers are described in S2 Table.

Other methods

Immunoblotting or DNA flow cytometric analysis was performed as described [35,78]. Rap1 was detected with affinity-purified antibody (gift from V. Zakian, Princeton University) or antibody against the C-terminus (yC-19, Santa Cruz biotechnology). LacI fusions were detected with anti-LacI antibodies (clone 9A5, Millipore). Two-dimensional gel electrophoresis was performed in Tris-borate-EDTA as previously described [43]. Details of plasmid construction are described in S1 Text.

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5

Attachment 6

Attachment 7

Attachment 8

Attachment 9

Attachment 10

Attachment 11

Attachment 12

Attachment 13