Telomere Recombination Preferentially Occurs at Short Telomeres in Telomerase-Null Type II Survivors

In telomerase negative yeast cells, Rad52-dependent recombination is activated to maintain telomeres. This recombination-mediated telomere elongation usually involves two independent pathways, type I and type II, and leads to generation of type I and type II survivors. It remains elusive whether the recombination-mediated telomere elongation prefers to take place on shorter or longer telomeres. In this study, we exploited the de novo telomere addition system to examine the telomere recombination event in telomerase negative cells. We show that recombination preferentially occurs on shorter rather than longer telomeres in both pre-survivors and established type II survivors. In type II survivors, the short VII–L telomeres could invade either terminal TG1–3 sequence or short tracts of TG1–3 sequence in subtelomeric Y′-X and Y′-Y′ junction to initiate recombination. Unexpectedly, short VII–L telomere recombination still takes place in type II survivors lacking either Rad50 or Rad59, which are required for type II survivor generation in senescing telomerase-null cells. Our results support the notion that Rad50 and Rad59 are not essential for the maintenance of type II survivors once established.


Introduction
Telomeres are the physical ends of eukaryotic chromosomes, and consist of telomeric DNA of repetitive sequence and protein complexes. They prevent chromosomal ends from degradation or fusion, and are essential for chromosomal integrity and genome stability in eukaryotes [1]. In budding yeast Saccharomyces cerevisiae, telomeric DNA consists of ,300675 bps of TG 1-3 /C 1-3 A repeats with a protruding G-rich single-stranded 39 overhang (Goverhang) [2,3]. Immediately internal to the telomeric TG 1-3 sequence, there are telomere associated sequences/elements, called X and Y9 elements [4]. Y9 elements are either 6.7 or 5.2 kb long [5]. Some telomeres have no Y9 element, and some have as many as four copies of Y9 elements. Short tracts of TG [1][2][3] sequence are often found between two Y9 or X and Y9 elements. Like most of the eukaryotes, budding yeast elongates its telomeric DNA through a specialized reverse-transcriptase, telomerase, which adds telomeric DNA sequences to the 39 G-overhang using its intrinsic RNA template [6,7]. The telomerase holoenzyme in S. cerevisiae is composed of the catalytic subunit Est2, the RNA template subunit Tlc1 and two accessary subunits, namely Est1 and Est3 [8][9][10][11][12]. Deletion of any of the telomerase subunits causes gradual telomere shortening and cellular senescence [10,13]. Most of the telomerase-null senescing cells eventually die, while a very small portion of cells can overcome the crisis and remain viable. These cells are called ''survivors'', whose chromosome ends are lengthened by homologous recombination [14,15]. According to different telomeric DNA structures, survivor cells are roughly divided into two categories, type I and type II survivors [16,17]. The recombination-mediated telomere elongation (namely ''alternative lengthening of telomeres'' or ALT) is also found in mammalian cells. For example, approximately 15% of immortalized human tumor cells apply the ALT mechanism to maintain telomeres [18,19].
The generation of telomerase-null survivors appears to involve multiple steps of repairing shortened telomeres through homologous recombination. The type I survivors amplify subtelomeric Y9 elements and short terminal telomeric TG tracts. The process of amplifying Y9 elements is dependent on the canonical homologous recombination proteins Rad51, Rad52, Rad54, Rad55 and Rad57 [20]. The type II survivors generate long heterogeneous telomeric TG 1-3 tracts, and this process requires a subset of DNA repair proteins, such as Rad52, Rad59, Sgs1 and the Mre11-Rad50-Xrs2 (MRX) complex, as well as Mec1 and Tel1 kinases [20][21][22][23]. The type II survivors of yeast are thought to resemble the ALT cells in mammals because both have long and heterogeneous telomeric repeats [15]. The type I survivors have extremely short TG tracts which affect the cell cycle progression and make the cells grow poorly [14,24]. In contrast, the type II survivors grow at a rate similar to telomerase proficient cells. Therefore, type II survivors eventually overtake type I survivors in liquid-grown cultures [15,20].
Previous studies have shown that in wild-type yeast cells, telomerase usually acts on and preferentially elongates short telomeres [25][26][27][28]. In the absence of telomerase, Rad52-depen-dent homologous recombination is believed to be the major pathway that repairs telomeres. However, it remains elusive whether the recombination machinery also prefers shorter to longer telomeres. Recently Chang et al reported that long telomeres are the more preferable targets of recombination during survivor emergence [29], while other two independent studies suggested that only short telomeres engage in recombination in either pre-survivors or survivors [21,30]. In this study, we examined the recombination-mediated telomere replication events in telomerase-null cells, and observed a preference for the extension of short telomeres in both pre-survivors and established type II survivors. In type II survivors, the short VII-L telomeres could invade either terminal TG 1-3 sequence or short tracts of TG 1-3 sequence in subtelomeric Y9-X and Y9-Y9 junction to initiate recombination. Surprisingly, unlike Rad52, Mre11/ Rad50/Xrx2 complex and Rad59, which are required for type II survivor generation, greatly affect, but are not absolutely required for the de novo telomere elongation through recombination in type II survivor cells.

Telomere Sequencing
Telomere PCR and following sequencing were performed as described in [32].

HO Induction Assay
Cells were grown in yeast complete media lacking both uracil and lysine (YC/Ura 2 Lys 2 ) plus 2% raffinose (Sigma), diluted into YC/Ura 2 plus 2% raffinose and cultured to logarithmic phase. Galactose (Sigma) was added to a final concentration of 3%, and cells were cultured for an additional 24 hrs. Samples were processed for plating or dotting assays. Both liquid and solid cultures were grown at 30uC.

Dotting, Plating and Viability Assays
Cell cultures were diluted with sterile water to a concentration of OD 600 ,0.3, and 3 ml of each 5-fold serial dilutions was dotted on YC plates. Plates were incubated at 30uC for 2 to 3 days and then photographed. 100 ml of the 5 th dilution sample (in dotting assay) was plated. YC/Ura 2 Lys 2 (uninduced total) was used for uninduced samples, and YC/Ura 2 (induced total) and YC/ Ura 2 Lys 2 (induced uncut) was used for induced samples. Plates were incubated at 30uC for 2 to 3 days and colony numbers were counted. The viability was calculated using the formula: Viability = (Induced total -Induced uncut)/Uninduced total.

Shorter Telomeres are Subjected to Elongation in the Senescing Telomerase-null Cells
In budding yeast Saccharomyces cerevisiae, inactivation of telomerase causes telomere shortening and cellular senescence [9,10,13]. In order to know the change of telomere sequence in the senescing telomerase-null cells, we obtained an est2D spore by dissecting a tetrad derived from the heterozygous est2D/EST2 strain, and passaged the spore for about 60 generations, a time point that survivor had not emerged yet. Then we analyzed the telomere sequences of chromosome I-L. 131 independent clones were subjected to telomere sequencing (Fig. 1A, Seq. S1). The sequencing results showed that telomere I-L of 14 clones contains divergent sequence (marked in red, Fig. 1A) [33]. These sequence divergences presumably resulted from recombination events because telomerase had been inactivated in the est2D cells. Interestingly, the recombination events appeared to take place only on some of the telomeres that were shorter than 100 bp, and similar results were obtained in est3D cells (Fig. 1B, Seq. S2). In the wild-type cells, telomere elongation presumably mediated by telomerase occurs on most of the telomeres [32]. This observation indicated that telomere recombination is in action in the senescing telomerase-null cells, but not efficient enough to repair all the short telomeres to prevent the cells from entering crisis and senescence. The extremely low emerging frequency of survivors could also be attributed to the insufficient recombination activity. We also noticed that the extent of new sequence added onto short ends seemed not to be greater than the lengths of non-recombined telomeres present in the cells. This may result from simply copying from another eroded short telomere rather than copying from a tcircle by a rolling circle mechanism, which could generate very long telomeres [34]. Furthermore, it seems that in est3D cells, recombination takes place on much shorter telomeres than that in est2D cells (Figs. 1A and 1B), possibly suggesting a protection role of Est2/Tlc1 in senescing cells.
To determine whether recombination-mediated telomere elongation prefers shorter telomeres, we exploited the de novo telomere addition system that was reported previously [31]. In our experiments, we chose 81-bp and 300-bp TG-tract to represent ''short'' and ''long'' telomeres. In the strain of TG81 or TG300, a TRP1 selection marker, 81 bp or ,300 bp (wild-type) telomeric ''seed'' and an HO endonuclease recognition site were integrated into the left arm of chromosome VII at the ADH4 locus (Fig. 1C). When galactose is added into the culture, the HO gene under the control of a galactose-inducible promoter will be expressed. The HO recognition sequence is recognized and cut by the HO endonuclease, resulting in a 16 bp non-telomeric chromosome end including a TGTT 39 overhang, which is adjacent to TG ''seed'' of either 81 or 300 bp (Fig. 1C). The fragment with native chromosome end is released, and the integrated telomeric ''seed'' is exposed (Fig. 1C). The exposed ''seed'' could be subjected to recombination upon resection of the remaining HO recognition sequence if telomerase is absent. Following these thoughts, we first deleted TLC1 in the TG81 or TG300 strain using the plasmid integration approach [35]. Right after tlc1D colonies were formed on a selective plate, the tlc1D cells of three independent colonies were transferred into galactose-containing liquid medium to induce HO cutting and passaged to allow senescence to take place and type II survivors to arise (Fig. S1). The recombination Figure 1. Recombination occurs on shorter telomeres in the senescing telomerase negative cells. Telomere I-L sequencing analysis of (A) est2D and (B) est3D cells. Each bar represents one sequenced telomere. Constant sequence is marked in blue and divergent sequence is marked in red. (C) Schematic representation of HO induction assay. The LYS2 and MNT2 genes serve as genetic markers to monitor the HO cleavage. The HO endonuclease recognition site and the TRP1 probe used to detect the TG 1-3 /HO end are shown in purple and red, respectively. The EcoRV and SpeI sites indicated are used for Southern digestion, which generates a fragment of ,2.9 kb (TG81 strain) or ,3.1 kb (TG300 strain). The EcoRV site within the TRP1 gene is located 600 bp away from the inserted TG seed. The fragment between the two EcoRV sites (,1.6 kb) serves as a loading control. This figure is not precisely drawn to scale. (D) Southern blot examination of the HO induction assay in senescing tlc1D cells. HO induction was performed immediately after TLC1 deletion. Three single clones of each strain (TG81 or TG300) were inoculated and passaged for 13 times (labeled on the top of each panel) in liquid. EcoRV and SpeI were used for Southern digestion and the TRP1 probe was used for hybridization. (''un'', uninduced control). doi:10.1371/journal.pone.0090644.g001 events of telomere VII-L in the senescing tlc1D cells of TG81 and TG300 were examined by Southern blot analysis using a TRP1 probe. The genomic DNA was digested with EcoRV and SpeI. The HO-uncut fragments recognized by the TRP1 probe should be ,2.9 kb and ,3.1 kb in length in the TG81 and TG300 strains, respectively (Fig. 1D). After HO induction, the chromosome fragments containing LYS2 would be lost, and the fragments recognized by the TRP1 probe should be shifted to ,0.7 kb and ,0.9 kb in the TG81 and TG300 strains, if there are no resection and recombination, which respectively shortens and lengthens the TG-tracts exposed (Fig. 1D). All the clones in both TG81 and TG300 strains did not grow on lysine minus plates (data not shown), indicating that they were auxotrophic for lysine, and that the HO digestion worked well. In the three clones of TG81 strain, the smearing hybridization signals of longer products indicated that the recombination occurred on telomere VII-L at the 5 th , 5 th and 7 th passage respectively, whereas in the TG300 clones, the recombination took place on telomere VII-L at the 9 th passage (Fig. 1D). It seems that once the telomeres reach a critical length, recombination happens, so it takes more cell divisions for the telomeres to be trimmed to trigger recombination in the TG300 strain. During the process of survivor formation in the TG81 strain (TG81-2 and TG81-3), the bulk telomere recombination, compared with the recombination on telomere VII-L, occurred simultaneously with the lengthening of the short HO-cut telomeres, which is earlier than that in the TG300 strain (Figs. 1D and S1). These results suggest that the short VII-L telomeres in TG81 cells are not only extended sooner, but also induce a more rapid generation of type II survivors.

Recombination Occurs on Shorter Telomeres in Type II Survivors
In addition to the distinct bands, the smear signals were detected in the Southern blot membrane (Fig. 1D), suggesting that telomerase-null progenies have chosen different templates as donors to accomplish one round of telomere recombination. In order to further dissect the recombination-mediated telomere extension events, we decided to use the well-established type II survivors to perform HO induction assay. For this experiment, TLC1 was deleted and then passaged in liquid culture. Cell density was measured and the cultures were diluted to a density of OD 600 ,0.05 in fresh medium every 24 hrs. The procedures were repeated 14 times and the well-established type II survivors were obtained. After ,25 generations of plate growth, a single colony of the type II survivors was inoculated into liquid medium and grown to logarithmic phase, galactose was then added, and the HO endonuclease digestion lasted for 24 hrs. The cells were plated on solid medium until individual colonies arose. Ten independent colonies were randomly picked, their genomic DNA was digested with a pair of restriction enzymes, SpeI and EcoRV, and the telomere length of chromosome VII-L was examined by Southern blot assay using a TRP1 probe (Fig. 1C). In Figure 2A and 2B, both the ''uninduced'' and ''induced'' lanes served as controls. Notably, the telomeres in all TG300 clones were shortened compared to the one shown in the induced lane (Fig. 2B). In contrast, the telomeres in all TG81 clones were longer than the induced products ( Fig. 2A). Interestingly, the elongated bands in many TG81 clones resembled the size in the uninduced lane (,2.9 kb), so we wondered if these were HO-uncut or re-ligation products via non-homologous-end-joining (NHEJ) pathway. To solve this doubt, we employed a probe of MNT2 gene, which is located to the same side of the HO recognition site as LYS2, to detect the existence of the LYS2 arm (Fig. S2A). The Southern blot showed that the ''lengthened telomeric fragments'' in all the 10 clones of TG81 tlc1D strain are recombination products (Fig. S2B). Thus, these results indicate that shorter telomeres are more susceptible targets of recombination in type II survivors, and argue against the notion that survivors have acquired a mutation permitting telomere uncapping (and hence telomere recombination) at long telomeres [36].
Previous study has suggested that the recombination event occurring on one single telomere does not trigger the abrupt lengthening of other telomeres once type II survivors are established [21]. We also tested this model in the de novo telomere addition system. Three independent clones of each strain shown in Figure 2A-B were passaged in liquid medium, and their VII-L telomere length of passages 2, 8 and 14 was examined by Southern hybridization using the TRP1 probe (Fig. 2C). As expected, both the recombined telomeres (in TG81 clones except No. 9) and the eroded telomeres (in TG300 clones) underwent gradual shortening, and a new round of recombination abruptly took place when a critical telomere length was reached (Fig. 2C). We also detected the total telomere signals between passages 2 and 14 using the TG 1-3 probe, and found that most of the telomeres were experiencing gradual shortening, which was not keeping pace with VII-L telomere (Fig. 2D). These data strongly supported the notion that in type II survivors, recombination occurring on a shorter telomere does not provoke massive recombination on other telomeres.
We noticed that in TG81 tlc1D strain, 8 out of 10 clones generated the recombination products which were uniformly around 3 kb when the genomic DNA was digested with SpeI and EcoRV ( Fig. 2A). The 3 kb band in clone 9 of TG81 tlc1D strain appeared not to be shortened during sequential passages (Fig. 2C), suggesting that these ,3 kb restriction fragments might not be the results of stochastic recombinations between the short VII-L telomere and the terminal TG 1-3 tracts in other telomeres (see the schematic model in Fig. 3B). Because there are short TG 1-3 tracts between subtelomeric X and Y9 elements, as well as two Y9 elements (Fig. 3B), it is possible that the ,3 kb bands seen in TG81 tlc1D strain are resulted from copying Y9 element (see the schematic model in Fig. 3C). To test this model, we randomly picked 10 clones of the TG81 tlc1D cells, extracted their genomic DNA, and used two groups of restriction enzymes respectively, SpeI/EcoRV and SpeI/EcoRV/NspI/ScaI, to digest the DNA. The combination of SpeI/EcoRV/NspI/ScaI restriction enzymes can cut Y9 sequence at multiple sites, generating fragments with the size of shorter than 1 kb, while leave the TRP1-TG 1-3 fragment on the induced end intact. As a result, the recombination products resulting from terminal TG 1-3 tracts may appear at various lengths and give rise to identical bands upon SpeI/EcoRV and SpeI/EcoRV/NspI/ScaI treatments (Fig. 3B), whereas the Y9-containing recombination products display ,3-4 kb bands after SpeI/EcoRV digestion but migrate to ,0.8 kb after SpeI/ EcoRV/NspI/ScaI digestion (Fig. 3C). Then we used the TRP1 probe to perform Southern blot assays and found that SpeI/ EcoRV digested DNA fragments with the size of ,3-4 kb (Fig. 3D, lanes marked in red in the left panel) disappeared, and the fragments with the size of 0.7-0.8 kb showed up upon additional NspI and ScaI digestion (Fig. 3D, right panel), suggesting that the clones exhibiting ,3-4 kb signals have copied Y9 sequence onto the induced short telomere ends (Fig. 3C). The recombinational bands shown with SpeI/EcoRV digestion in clone 1 and 5 were not sensitive to the digestion of NspI and ScaI (Fig. 3D, compare left and right panels), and they were likely produced by copying different lengths of TG 1-3 tracts (Fig. 3B). To further validate that the ,3-4 kb signals on the SpeI/EcoRV membrane contain Y9 sequence(s), we carried out a PCR experiment using both Y9 and TRP1 specific primers (the positions of which are labeled by blue arrows in Fig. 3C), and detected very specific signals in the clones which had ,3-4 kb bands in the SpeI/EcoRV treated DNA (Fig. 3E, upper panel). Consistently, the PCR analysis using MNT2 primers illustrated that all the ,3-4 kb bands were not HO-uncut or NHEJ products (Fig. 3E, lower  panel). Taken together, our results suggest that the short VII-L telomeres can initiate recombination by invading both TG 1-3 tracts being internal to Y9 sequence (Fig. 3C) and the very terminal TG 1-3 sequence of other telomeres (Fig. 3B).

Efficient Telomere Recombination Requires Rad50, Rad59 and Cgi121 in Type II Survivors
It is now widely accepted that in S. cerevisiae, Rad50 and Rad59 are essential for the type II survivor formation, Rad51 is essential for the type I survivor formation, while Rad52 is required for the generation of both types of survivors in telomerase deficient cells [13,14,17]. Cgi121 is a subunit of yeast KEOPS complex, which is indispensable for type II survivor generation [22]. Therefore, we suspected that recombination could not occur in the type II survivors lacking RAD52, RAD50, RAD59 or CGI121. We knocked out RAD52, RAD50, RAD59 or CGI121 in the tlc1D-derived type II survivors, and performed HO induction assay followed by dotting assay to examine the cell viability in these mutants. If there was no 81-bp short telomere induced by HO endonuclease, all the strains showed regular growth except that the tlc1D rad52D mutant exhibited much lower viability than others (Fig. 4A, uninduced panel). When an 81-bp short telomere was created, the tlc1D rad52D mutant died immediately, and the tlc1D rad50D, tlc1D rad59D and tlc1D cgi121D mutants displayed considerably reduced viability (Fig. 4A, induced panel). These results indicated that Rad52 is absolutely required for repairing shortened telomeres [16], while Rad50, Rad59 and Cgi121 are not essential for, but facilitate telomere recombination.
Since critical short telomere(s) triggers cell cycle arrest and cell death if such short telomere(s) is not repaired by recombination, we used a plating assay to examine the cell viability, which is able to quantify the recombination efficiency on the short telomere of VII-L induced by HO endonuclease. In the TG300 strain, nearly 100% of the cells were viable upon HO induction, providing a control for other strains (Figs. 4A and 4B). The viability of the TG81 strain was 36% (Fig. 4B), because only the cells that recombine their short telomeres could be selected to survive. The recombination efficiency of the rad50D, rad59D and cgi121D mutants were 16%, 15% and 12%, respectively, which were significantly lower than that of the TG81 control (Fig. 4B).
Previous studies demonstrated that in telomerase positive cells, Tel1 preferentially associates with short telomeres to stimulate their elongation by telomerase [26][27][28], and Tel1 participates in   . Viability assays. All the mutants were derived from the TG81 tlc1D type II survivors (TG81 for short), and TG300 tlc1D type II survivors (TG300 for short) served as a control. (A) Dotting assay. Cell cultures were diluted with sterile water to a concentration of OD 600 ,0.3, and 3 ml of each 5-fold serial dilutions was dotted on YC plates. Plates were incubated at 30uC for 2 to 3 days and then photographed. rad50D and cgi121D are mutants of slow growth, which were dotted earlier than other strains. (B) Quantification of viability assays. 100 ml of the 5 th dilution sample in (A) was plated. Plates were incubated at 30uC for 2 to 3 days and colony numbers were counted. The viability was calculated using the formula: Viability = (Induced total -Induced uncut)/Uninduced total. Error bars represent standard deviations of three independent experiments. Compared to TG81 control, P values of the rad50D, rad59D, cgi121D, tel1D and chk1D mutants are 0.030, 0.025, 0.015, 0.295 and 0.317, respectively. Significance (*) is defined by P value ,0.05. doi:10.1371/journal.pone.0090644.g004 the generation of type II survivor [23]. In the well-established type II survivor cells, deletion of TEL1 did not reduce the recombination efficiency (Figs. 4A and 4B), suggesting that Tel1 functions in the steps of generating type II survivors, but not in the period of type II survivor maintenance. It will be of interest to perform an extensive study on whether the genes required for type II survivor generation are still indispensable in type II survivor maintenance [22]. We also explored the telomere recombination efficiency in the mutant of another DNA damage checkpoint protein, Chk1, and obtained similar results as seen in tel1D mutants (Figs. 4A and 4B).
Rad50, Rad59 and Cgi121 are not Essential for Telomere Recombination in Type II Survivors Since Rad50, Rad59 and Cgi121 are required for type II survivor generation, it remains possible that upon the inactivation of either Rad50, Rad59 or Cgi121, the tlc1D type II survivors might be forced to activate type I recombination, and switched to type I survivors. To test this possibility, we passaged the tlc1D rad50D, tlc1D rad59D and tlc1D cgi121D type II survivors on both solid and liquid media for 250 to 300 generations, and examined their telomere structures by Southern blot analysis. The results showed that the telomeres in all these mutants were of typical type II, and no significant Y9 amplification was detected (Figs. 5A and 5B).
To further verify the existence of recombination-mediated elongation of short telomeres in the tlc1D rad50D, tlc1D rad59D and tlc1D cgi121D type II survivors, we performed HO induction assay and examined the VII-L telomeres in ten independent clones. The Southern blot results with both a TRP1 and a MNT2 hybridization probe revealed that recombination took place on the HO-induced VII-L telomeres in the rad50D, rad59D and cgi121D mutants, except that rad50D-1 clone was not successfully cut by HO (Figs. 6A-C and S3). We then performed VII-L telomere PCR in the samples of rad59D-1 and cgi121D-8 (Figs. 6B and 6C, indicated with arrows), and the sequencing data confirmed that the telomere recombination products resulted from TG 1-3 amplification (Fig. 6F). Thus, we concluded that Rad50, Rad59 and Cgi121 are not essential for telomere recombination in type II survivors [17]. In addition, the telomere recombination events were also observed in the tel1D and chk1D mutants (Figs. 6D, 6E and S3).
In addition to the fragments that are shorter than 3 kb, some larger fragments of ,3-4 kb were also detected in the rad50D, rad59D, cgi121D, tel1D and chk1D mutants of TG81 tlc1D type II survivors (Figs. 6A-E). To evaluate whether those recombination events involved copying Y9 element (Fig. 3C), we randomly chose ten more clones of TG81 tlc1D rad59D, TG81 tlc1D cgi121D, TG81 tlc1D chk1D mutants, in which HO-induced short VII-L telomere had experienced recombination-mediated elongation. The Southern blot experiments showed that most of the ,3-4 kb bands were sensitive to the digestion of NspI and ScaI, while all the shorter bands (,3kb) were not sensitive to the digestion of NspI and ScaI (Fig. 7, panels labeled SpeI/EcoRV/NspI/ScaI). The PCR analyses revealed that most of the long telomere VII-L had copied Y9 sequence (Fig. 7, lanes marked in red in the panels labeled Y9-PCR). Notably, the clone 10 of the TG81 tlc1D rad59D mutant (Fig. 7A) and the clone 4 of the TG81 tlc1D cgi121D mutant (Fig. 7B) had long teminal TG 1-3 sequence because their telomere VII-L were not sensitive to SpeI/EcoRV/NspI/ScaI digestion. These results support both models shown in Figure 3B and 3C: a critical short telomere in a type II survivor can invade into the TG 1-3 homologous sequence either at subtelomeric region to copy Y9-containing sequence, or at the terminal telomeric region to copy TG 1-3 sequence.

Discussion
It is well known that short telomeres are preferred substrates for telomerase [25]. We wonder whether this preference also exists in the process of recombination-mediated telomere elongation. In this study, we have shown that in Saccharomyces cerevisiae, short telomeres are preferentially elongated by recombination in the senescing telomerase-null pre-survivors (Figs. 1A and 1B). Our result is consistent with that shown in a previous study by Chang et al, who also found that in S. cerevisiae, short telomeres in presurvivor cells are more likely to be extended by recombination [29]. Furthermore, McEachern and Topcu studied the recombinational telomere elongation events in another kind of yeast, Kluyveromyces lactis, and demonstrated that recombination is greatly induced near shortened telomeres [37,38]. Interestingly, Chang et al also followed the growth of the est2D cultures to identify the senescence point, and showed a recombination preference for long telomeres during the emerging stage of survivors [29]. However, their approach could not tell whether long or short telomeres are the preferred targets of recombination during the maintenance stage of type II survivors, because telomeres of several kb in length, which are common in type II survivors, can hardly be recovered in telomere sequencing assay. We employed the HO induction system and assayed telomere length (both gross telomere signals and single telomere of VII-L) by Southern blot analysis. In the established type II survivors, the short VII-L telomeres in the TG81 clones have recombined to generate longer ones, whereas the 300-bp long VII-L telomeres in TG300 clones have experienced shortening before any recombination event occurs ( Figs. 2A and 2B). Previously Teng et al reported that in the established type II survivors, the long VII-L telomeres did not undergo recombination-mediated lengthening, but rather short-  . Characterization of VII-L telomere recombination in the rad59D, cgi121D and chk1D mutants. All the mutants were derived from the TG81 tlc1D type II survivors. HO induction assay, Southern blot (left panels) and PCR analyses (right panels) were performed in (A) rad59D, (B) cgi121D and (C) chk1D mutants. SpeI/EcoRV or SpeI/EcoRV/NspI/ScaI digestion and the TRP1 probe were used for Southern blot. The 1.6 kb (SpeI/ EcoRV panels) and 0.83 kb (SpeI/EcoRV/NspI/ScaI panels) fragments serve as the loading control. Y9-PCR was performed as described in Figure 3E. doi:10.1371/journal.pone.0090644.g007 ened during ,350 cell divisions, and an abrupt elongation took place once the telomeres shortened to a critical length [21]. Taken together, these observations indicate that short telomeres are the preferential substrates for recombination in both pre-survivors and established type II survivors. Due to the survival pressure cells preferentially lengthen short telomeres because senescence would be induced if the critically short telomeres are not repaired. Thus, even in the emerging survivors that prefer long telomeres, the very short telomeres were also elongated by recombination [29]. In presurvivors, the newly added sequences seem to be very short (Figs. 1A and 1B), while in type II survivors, telomeres elongated by recombination could reach several kb (Fig. 2). These results suggest that in order to maintain certain telomere length, presurvivors and type II survivors might use different recombination mechanisms, i.e. copying an eroded short telomere and replicating from a template like t-circle to generate a long telomeric sequence respectively, although they all exhibit the preference for short telomeres. However, we cannot exclude the possibility that the long recombination products in pre-survivors (Figs. 1A and 1B) were not retrievable in telomere sequencing assay due to technical difficulties. To this end, we propose the model that recombination machinery favors different telomere substrates at different stages: in pre-survivors and established type II survivors, recombination occurs on short telomeres (this study), while long telomeres are preferentially elongated during the type II survivor formation [29]. We do not know why shorter telomeres are more susceptible to homologous recombination at the stages of pre-survivors and established type II survivors, while longer ones are more favorable at the emerging stage of type II survivors. One possibility is that in pre-survivors and established type II survivors, the shorter telomeres are less protected than the longer ones; but in in emerging survivors, the longer telomeres are less protected than the shorter ones. Less protected telomeres might facilitate singlestrand invasion, a critical step of homologous recombination. The other possibility is that t-circles are only generated in the emerging stage, and they are more accessible by longer telomeres. A third possibility is that during the critical phase of survivor formation, gross telomere recombination is activated, and longer telomeres are preferred because they provide more and better substrates for recombination. These speculations need experimental evidence to support.
In the established type II survivors, recombination occurring on a shorter telomere does not provoke massive recombination on other telomeres (Figs. 2C and 2D) [21]. In contrast, the HO-cut short VII-L telomere in the TG81 tlc1D cells seems to be able to trigger more rapid gross telomere recombination because the formation of type II survivors in the TG81 tlc1D cells likely emerged earlier than that in the TG300 tlc1D cells (Figs. 1D and S1). These results suggest that longer telomeres might be involved in recombination during survivor formation, and support the notion raised by Chang et al that recombination prefers long telomeres during the emerging stage of survivors [29].
It is generally accepted that both Rad50 and Rad59 are required for type II survivor generation [17]. In the established type II survivors, inactivation of either Rad50 or Rad59 did not prevent recombination from taking place on the HO-induced VII-L telomeres (Figs. 6A and 6B). It is surprising because our results seem to be contradictory to the well-documented role of Rad50 and Rad59 in type II survivor generation [17]. We thought that the discrepant conclusions might be attributed to the differences between the two assays. In the telomerase-null senescing cells, the 32 telomeres all undergo shortening until to the critical length, resulting in cell cycle arrest [39,40]. In order to generate type II survivor(s), presumably multiple (if not all of the 32) critical short telomeres must experience the repair process via recombination. That's why even under the conditions that all the factors involved in recombination are present, the probability of generating a type II survivor is still very low (,10 6 ) [20]. Therefore, the high recombination efficiency, granted by Rad52 group factors (including Rad50 and Rad59), is crucial for the survivor generation, and it is not surprising to see that both Rad50 and Rad59 are required for type II survivor formation. However, in an established type II survivor, most of the telomeres are long. To maintain the viability of a type II survivor, only the critical short telomere(s), which account for a few, need to be elongated. Inactivation of Rad50 or Rad59 may cause a problem, but it might not result in catastrophic defect to the viability of a type II survivor. Figure S1 Examination of massive telomere recombination in senescing tlc1D cells. The same samples of genomic DNA as in Figure 1D were digested by XhoI and Southern blot was performed using a TG 1-3 probe.