Chromatin modifiers and recombination factors promote a telomere fold-back structure, that is lost during replicative senescence

Summary

Telomeres have the ability to adopt a lariat conformation and hence, engage in long and short distance intra-chromosome interactions. Budding yeast telomeres were proposed to fold back into subtelomeric regions, but a robust assay to quantitatively characterize this structure has been lacking. Therefore, it is not well understood how the interactions between telomeres and non-telomeric regions are established and regulated. We employ a telomere chromosome conformation capture (Telo-3C) approach to directly analyze telomere folding and its maintenance in S. cerevisiae. We identify the histone modifiers Sir2, Sin3 and Set2 as critical regulators for telomere folding, which suggests that a distinct telomeric chromatin environment is a major requirement for the folding of yeast telomeres. We demonstrate that telomeres are not folded when cells enter replicative senescence, which occurs independently of short telomere length. Indeed, Sir2, Sin3 and Set2 protein levels are decreased during senescence and their absence may thereby prevent telomere folding. Additionally, we show that the homologous recombination machinery, including the Rad51 and Rad52 proteins, as well as the checkpoint component Rad53 are essential for establishing the telomere fold-back structure. This study outlines a method to interrogate telomere-subtelomere interactions at a single unmodified yeast telomere. Using this method, we provide insights into how the spatial arrangement of the chromosome end structure is established and demonstrate that telomere folding is compromised throughout replicative senescence.

Keywords:

DNA damage – Histones – Chromatin – Polymerase chain reaction – Protein folding – Saccharomyces cerevisiae – Telomeres – Yeast

Introduction

Telomeres are essential nucleoprotein structures at the physical ends of eukaryotic chromosomes consisting of non-protein coding DNA repeats and telomere bound protein complexes. In human cells, telomere structure and function are controlled by a six-protein complex called Shelterin, that binds to telomeres in a sequence specific manner [1]. In yeast, these functions are executed by the CST (Cdc13-Stn1-Ten1) complex [2] together with Rap1, Rif1 and Rif2 [3–5]. Although telomeres resemble a one-ended DNA double-strand break (DSB) they are refractory to being acted upon by the DNA damage response (DDR) [6]. Hence, telomeres prevent illegitimate repair events that would cause chromosome end-to-end fusions and lead to genome instability [6]. This “end protection” property of telomeres has largely been attributed to the associated protein complexes, but a telomeric loop structure (t-loop) also appears to be critical. Telomeric lariat structures have been demonstrated in species ranging from yeast to human [7–13], however, their properties may vary considerably between organisms. In mammalian cells, the t-loop forms within the telomeric repeat tract, via strand-invasion of the 3’ overhang into the double-stranded region of the telomere [13,14]. In contrast, telomeres in S. cerevisiae are proposed to fold back into the subtelomeric region, where they are maintained by an, as of yet, unknown mechanism [12,15,16].

Little information exists about the regulation of telomere fold-back structures. TRF2, a component of the Shelterin complex, is required for t-loop formation in human cells [13,14]. In addition to promoting t-loops, TRF2 also regulates their dissolution in S phase via recruitment of the helicase RTEL1 [17,18]. In S. pombe the telomeric protein Taz1 has been demonstrated to remodel telomeric DNA into loops [8] and in S. cerevisiae histone deacetylases have been shown to be required for the fold-back into the subtelomere [15,19]. In addition, a genome-wide screen using a fold-back reporter gene on a modified telomere in mutants of the non-essential yeast deletion collection indicates, that other protein families, including chromatin remodelers and transcription regulators, are also involved in the maintenance of telomere structure [15].

Human and mouse t-loops can be visualized via electron- and super resolution-microscopy [13,14,20]. The existence of looped structures could also be shown microscopically in K. lactis harboring overelongated telomeres [11]. The short length and base composition of S. cerevisiae telomeres prevent such microscopy-based approaches, hence only genetic- and chromatin immunoprecipitation (ChIP)—based experiments have been employed to study telomere folding in yeast [12,15,16,19]. A major concern of the genetic approach is that the subtelomere is deleted and modified with a reporter [12,15,19], which may strongly alter protein composition and chromatin status in the vicinity of the telomere. On the other hand, ChIP of telomeric proteins to the subtelomere may indeed represent telomere folding, however, the spreading of telomere binding proteins into subtelomeric regions cannot be completely ruled out [15,16].

Telomeres become shortened with each cell division due to the end replication problem [21–23]. The loss of telomeric repeats leads to the progressive loss of chromosome end protection and replicative senescence, a permanent cell cycle arrest. To counteract telomere shortening, certain mammalian cell types, yeast cells and 73% of cancer cells express the reverse transcriptase telomerase, which adds telomeric sequences to the 3’ end of the shortest telomeres [24–27]. Yeast constitutively express telomerase but deletion of the catalytic subunit (EST2) or the RNA component (TLC1) of telomerase renders them susceptible to the end replication problem and they eventually undergo replicative senescence [26]. As a rare event, some cells (survivors) can overcome senescence and crisis by elongating their telomeres via homologous recombination-based mechanisms [28].

Critically short telomeres arising during replicative senescence activate the DNA damage checkpoint and are recognized by DNA repair factors, predominantly by the protein kinase Mec1 (ATR) but also by Tel1 (ATM), both master regulators of the DSB response [29–32]. The activation of either kinase triggers the phosphorylation of the effector kinases Rad53 and Chk1, which leads to a cell cycle arrest. Additionally, the homology directed repair (HDR) proteins Rad51 and Rad52 colocalize with critically short telomeres in the absence of telomerase to promote their extension (repair) [33,34]. Surprisingly, it was shown that proteins involved in DDR, such as the MRX (Mre11, Rad50, Xrs2) complex and Tel1 also localize to telomeres in telomerase positive cells and play a general role for telomere stability and telomere length regulation [35]. Cells that lack both Mec1 and Tel1 show progressive telomere shortening and undergo replicative senescence [36,37] in addition to harboring increased telomere-telomere fusions (T-TFs) [38–41]. At mammalian telomeres, the DNA damage response and the homologous recombination machinery also play a role at functional telomeres. It was reported for human cells, that ATM and ATR are transiently recruited to telomeres during telomere replication where they activate a local DDR. [42,43]. After telomere processing is complete, Rad51 and Rad52 are recruited and likely promote strand invasion of the 3’ overhang into the double-stranded telomeric repeats to generate a t-loop. This transient localization does not result in unscheduled recombination and sister chromatid exchange between telomeres. Recent studies have indicated that the t-loop structure plays an important role in the regulation of the DNA damage response at telomeres [20,44]. When the t-loop is not formed, telomeres aberrantly activate an ATM response, however they remain resistant to T-TFs.

Here, we developed telomere chromosome conformation capture (Telo-3C) at a natural yeast telomere, which provides a direct readout for the telomere folding structure into the subtelomere. Wild type (wt) length telomeres are able to fold-back (and form a “closed” state). However, when the HDR proteins Rad51 or Rad52 are not present, telomere folding occurs less frequently, suggesting a strand invasion mechanism similar to what has been described for the t-loop in human cells. Consistent with the DDR being involved in fold-back establishment, telomere folding is also defective when the checkpoint kinase Rad53 is depleted. As cells enter replicative senescence, we show that telomere folding is not maintained. We conclude that this “open” state is not a result of telomere shortening per se, but seems to be linked to the senescence program. We identified the histone modifiers Sir2, Sin3 and Set2 as potential regulators of telomere folding during replicative senescence. Indeed, cells have a decreased capacity to establish telomere folding when lacking any of these factors. We hypothesize that the reduced expression of these factors likely contributes to ensuring an open telomere state during senescence. Taken together, out data offer new insights into understanding how telomeres physically interact with other regions of the genome.

Results

Telo-3C detects telomere folding in S. cerevisiae

To gain insights into the dynamic regulation of telomere folding in S. cerevisiae, we developed a telomere chromosome conformation capture (3C) approach [45] (Telo-3C). This provides a direct readout to quantitatively assess the interaction of one telomere (left telomere of chromosome I) with its adjacent subtelomeric region. As illustrated in Fig 1A, chromosome contacts were crosslinked in vivo using formaldehyde and the extracted chromatin was digested with a suitable restriction enzyme (NcoI for telomere 1L) (see Fig 1B for precise restriction sites). The compatible DNA ends were ligated under very dilute conditions, resulting in the ligation of only those loci that were in close proximity at the time of the crosslink. After reversing the crosslink, a specific 3C ligation product can be amplified from primers that were initially co-directionally oriented on genomic DNA (no PCR product) but get converted through the 3C reaction to being convergent (giving a PCR product, see blue arrows Fig 1A). The abundance of this specific PCR product can be quantified by quantitative PCR (qPCR).

To ensure the specificity of the Telo-3C procedure we performed the protocol without ligating or digesting the chromatin. In these conditions, we did not detect a Telo-3C product of the same intensity as compared to the complete Telo-3C protocol (Fig 1C). Similarly, when the chromatin was not crosslinked with formaldehyde, we detected only weak Telo-3C products as compared to the crosslinked sample, despite increasing amounts of chromatin input (Fig 1D). This suggests that a specific Telo-3C qPCR product could only be amplified when the loci of interest were located in close proximity. We were interested in measuring the interaction of the telomeric repeats with several loci along the subtelomere and designed primers according to the NcoI restriction sites that were available (Fig 1B). Primer P5, directly juxtaposed to the telomere repeat sequences, was used as an anchor primer in combination with P6, P7 and P9 thereby comparing the interaction of the telomeric repeats to a subtelomeric region with a distance of ~2 kb (P6) and ~9 kb (P7). With P9 we chose a primer outside the region which we would define as subtelomere with a distance of ~47 kb from the telomeric repeats. The interaction frequency was reduced to half when moving from a ~2 kb (P6) to ~ 9 kb (P7) distance from the telomeric repeats (Fig 1E), similar to what has been reported previously [45]. We also detected a Telo-3C product between primers P5 and P9 (~47 kb) which suggests that the telomere may also interact with loci of greater distance to the telomeric repeats, although with a reduced frequency (Fig 1E). The sequences of the subtelomeric Telo-3C qPCR products (P5 + P6 and P5 + P7) were confirmed by Sanger sequencing (S1A and S1B Fig). In order to understand if telomere folding occurs more frequently than a random interaction of two loci on the same chromosome, we analyzed chromosome interactions at internal NcoI restriction sites on chromosome I (Fig 1F). The respective 3C primers (P10 (anchor), P11 and P12) have similar distances as the ones used on subtelomere 1L. Although we could detect a weak interaction between these loci using the internal primer sets, the interactions between telomere 1L and its subtelomere were more robust (Fig 1G). These results demonstrate that the interactions at the end of the chromosome occur more frequently than a random encounter of two loci on the same chromosome.

In summary, we established and validated a 3C method that detects the interaction of telomeric and subtelomeric regions on a single chromosome, which provides further evidence for the presence of a folded structure at chromosome ends in budding yeast. It appears, that telomere folding occurs more frequently, or more stably, than internal interactions on the same chromosome. Moreover, the telomeric repeats might touch down at multiple regions, and the interaction frequency diminishes with increasing distance from the telomeric tract.

Telomere 1L opens during replicative senescence

To understand how telomere structure may be regulated throughout the process of replicative senescence, we passaged budding yeast lacking the RNA component of telomerase (TLC1) over multiple generations to induce gradual telomere shortening [46]. As expected, the tlc1 cells progressively experienced decreased growth potential (Fig 2A). The loss of growth potential was caused by bulk telomere shortening due to the lack of telomerase activity, as shown in a telomere restriction fragment analysis (S2A Fig, representative example of one clone). At crisis, the point of lowest growth potential, some cells eventually overcome the permanent cell cycle arrest by acquiring the capacity to elongate their telomeres based on a recombination-mediated pathway [28]. These survivors showed the amplification of TG repeats by the appearance of multiple additional bands on the Southern blot [47] (S2A Fig). We employed Telo-3C during this time course and measured the interaction of the telomere 1L with its subtelomere at the indicated time points (+ in Fig 2A). Telo-3C analysis at PD (population doubling) 9 revealed that telomere folding of tlc1 mutants was comparable to wt telomeres, despite their telomeres being much shorter (Figs 2B and S2A). However, when tlc1 cells had critically short telomeres and approached the point of telomeric crisis, telomere folding decreased significantly (PD 47, 56 and 64 in Fig 2B). Similar results were obtained when using the P5 + P7 primers (S2B Fig). In survivors, however (PD 101 in Figs 2B and S2B), telomeres re-acquired the ability to establish a fold-back structure. Using a probe specific for telomere 1L (the telomere where we measure the fold-back), we observed that at PD 9, telomeres were shorter in tlc1 cells than in wild type cells but were still fold-back proficient (Fig 2C). At PD 47 and 56, telomere 1L became even shorter and as senescent cells accumulated the fold-back structure was lost (Figs 2B, 2C and S2B). At PD 64, we observed some recombination events but the bulk of telomere 1L was still critically short (and cells remained fold-back defective) (Figs 2B, 2C and S2B). At PD 73 and 101, telomere length was recovered by HDR and this was when the fold-back began to be restored (Figs 2B, 2C and S2B). Flow cytometry analysis for DNA content revealed that the analyzed cell populations did not demonstrate drastic changes in terms of cell cycle distribution (S2C Fig). To rule out that extensive resection at short telomeres may result in loss of the NcoI restriction site, and hence loss of the Telo-3C signal, we performed the same senescence assays in an EXO1 deficient strain, where extensive telomere resection is defective [33]. In the exo1 background, we observed a similar loss of telomere folding, indicating that extensive resection via Exo1 was not responsible for the decreased Telo-3C signal (S2D and S2E Fig).

The fact that telomeres remain folded at PD 9 suggests that telomere shortening alone is not responsible for the loss of the folded structure (Figs 2B and S2B). To address this further, we employed Telo-3C in several yeast mutants with very short, albeit stable, telomeres. All of the mutants analyzed harbored telomeres (both bulk telomeres and 1L specifically) that were comparable in length to telomerase negative tlc1 cells but showed neither progressive shortening nor a senescence phenotype (Fig 2D and 2E). We found that the telomere 1L of these mutants established a fold-back structure that varied in strength between mutants, but was consistently more frequent than in tlc1 mutants (Fig 2F). This suggests that short telomeres per se are still able to fold back onto the subtelomere. We cannot, however, exclude that critically short telomeres, which accumulate in telomerase negative cells, contribute to telomere unfolding during senescence. Additionally, ku70, mre11, tel1 (Fig 2F) and hrs1 (S2F Fig) mutants have all been demonstrated to be defective in telomere clustering [48,49], suggesting that clustering is not promoting the fold-back structure nor influencing the Telo-3C signal.

As replicative senescence is induced by a cell cycle arrest due to a DNA damage response at telomeres, we investigated whether other forms of DNA damage would also result in an open telomere state. Upon treatment of wild type cells with the DNA damaging agent methyl methanesulfonate (MMS), we observed decreased telomere folding, that became more pronounced over time (Figs 2G and S2G). Since both MMS treatment as well as replicative senescence lead to a G2/M delay, we addressed whether the cell cycle arrest itself may contribute to the opening of the fold-back structure. To arrest cells in G2/M in a DNA damage-independent manner, we constructed an auxin-inducible degron (AID*) version of Cdc20 (S2H Fig). This system allowed us to study the impact of a G2/M arrest on telomere folding by degrading Cdc20 upon addition of indole-3 acetic acid (IAA) to the culture media (S2H and S2I Fig). We observed reduced telomere folding in a time-dependent manner following Cdc20 depletion (S2J Fig), suggesting that the prolonged G2/M arrest of senescent cells may contribute to telomere opening.

In conclusion, we demonstrate that critically short telomeres open their folded structure during replicative senescence and prolonged exposure to MMS. The opening of the fold-back structure is not solely due to the short nature of these telomeres during senescence. Instead, the decreased telomere folding ability during crisis might rather be due to other features of the “senescence program”, including the checkpoint mediated delay at the G2/M border.

Rad51, Rad52 and Rad53 are required for telomere folding

In order to identify regulators of telomere folding, we first employed a candidate approach of factors that could potentially influence telomere fold-back formation and maintenance. First, we chose to investigate the impact of the structural maintenance of chromosomes (SMC) complexes on telomere folding. Cohesin and condensin have been demonstrated to play important roles in terms of chromosomal architecture, including chromatin loop formation and chromatin condensation, respectively [50]. We also wanted to test if the SMC5/6 complex might be involved in telomere folding due to its described roles at telomeres [51–55]. We tagged one subunit of each SMC complex with an auxin-inducible degron [56] (Fig 3A). We treated the culture with auxin for 1 h, which resulted in a nearly complete depletion of the respective proteins (Fig 3A), while not impacting the viability of the cells (S3A Fig). Telo-3C results demonstrated that telomere folding was not affected by the lack of any of the SMC complexes under these conditions (Fig 3B). Therefore, although the SMC complexes are implicated in chromosome compaction [57–59] and the establishment of chromosome domain boundaries and chromatin loops [60,61], they are not a requirement for telomere folding.

We tested the telomeric proteins Rif1 and Rif2 for their contribution to establishing telomere folding. Both Rif1 and Rif2 were previously reported to regulate telomere structure as measured by the fold-back reporter gene [15]. However, by Telo-3C we did not observe a telomere folding defect in either rif1 or rif2 deletion mutants arguing that these two telomeric proteins might not contribute to the establishment of telomere folding (Fig 3C). This discrepancy between the two methods might arise from the two different readouts employed, or due to the introduction of an artificial subtelomere with the reporter gene. Additionally, the chromatin structure at telomeres of rif1 and rif2 mutants is altered which affects natural and modified subtelomeres in different ways [4,5].

In human cells, the t-loop consists of a D-loop recombination intermediate, which is suggestive of a strand invasion event. To test whether recombination may also be involved in yeast telomere folding, we performed Telo-3C experiments in cells lacking the central HDR factors RAD51 and RAD52. We found that telomere folding was decreased in both rad51 and rad52 mutants (Figs 3D and S3B). This strongly suggests that in yeast, like in human cells, a strand invasion event contributes to the establishment of a closed telomere state. Importantly, the gene loop (between the terminator and promoter) that forms at the HEM3 locus [62] was not influenced in either rad51 or rad52 mutants (S3C Fig), suggesting that the HDR machinery may be specific for folding at chromosome ends.

As Rad51 and Rad52 promote repair in response to a DDR, we tested whether the most upstream factors of the DDR, the checkpoint kinases Mec1 (ATR) and Rad53 (CHK2), may also contribute to telomere folding. We depleted cells of Rad53 for 2 h using an auxin-inducible degron (Fig 3E), a treatment which did not result in decreased cell viability (S3D Fig). Under these conditions, we observed a significant reduction of telomere folding (Fig 3F), suggesting that an impaired DNA damage checkpoint does affect the structure of the telomere. Surprisingly, the deletion of either MEC1 (Fig 3G) or TEL1 (Fig 2F) alone did not affect the folding behavior of telomere 1L. We propose, that Mec1 and Tel1 might have overlapping functions in terms of telomere folding which could potentially compensate for each other. Testing the tel1 mec1 double mutant is complicated by the fact that it induces replicative senescence [36] and a has a high incidence of telomere fusions [38,40]. Alternatively, Rad53 may promote telomere folding in a Mec1- and Tel1-independent manner.

Loss of Sir2, Sin3 and Set2 contribute to telomere opening in senescence

In order to understand the underlying mechanisms behind the structural changes that occur at telomeres throughout senescence, we analyzed the whole proteome of tlc1 cells as they approached telomeric crisis and compared it to wt cells. Label-free quantitative proteomics analysis revealed 751 significantly downregulated and 678 upregulated proteins during senescence (Fig 4A, see S4 Fig for a comprehensive list, see S5A and S5B Fig for gene ontology analysis). To identify candidates that potentially regulate telomere folding during senescence, we compared the results from a previous screen for regulators of telomere folding in S. cerevisiae [15] with the downregulated proteins from our proteomics data set. This genetic screen by Poschke et al., identified 113 genes and we manually added sir2 to this list. Sir2 has been demonstrated to promote telomere folding, however, its deletion cannot be detected in the screening procedure due to technical reasons [12,19,63]. We found 21 overlapping candidates between the fold-back defective mutants and proteins that become downregulated during senescence (Fig 4B). Downregulation of these candidate proteins during senescence may therefore contribute to the loss of a closed telomere state. Compared to the complete set of downregulated proteins during senescence, the 21 candidates (Fig 4C) were significantly enriched for the gene ontology terms of protein and histone deacetylation (S5C Fig), suggesting an epigenetic mechanism underlying telomere folding regulation during senescence. The proteins Sir2, Sin3 and Set2 were of particular interest as they antagonize histone acetylation and were reported to be involved in telomere silencing or the regulation of telomeric chromatin [64–71]. We tested the identified chromatin modifiers in a Telo-3C assay and found that indeed the deletion mutants of SIR2, SIN3 and SET2 were telomere folding defective (Figs 4D and S6A). We further confirmed the downregulation of Sir2, Sin3 and Set2 during senescence (tlc1) as observed by mass spectrometry (Fig 4A) through western blotting (Fig 4E). In pre-senescent cells and post-senescent survivors, protein levels did not differ from those of telomerase positive wild type cells (Fig 4E). Of note, we observed a modified Sir2 species in senescent cells suggesting a specific regulation of Sir2’s activity, localization or stability during the senescence program (Fig 4E). Telomere length of these mutants is similar to wt length or in the case of sin3, only slightly shorter (Figs 4F and S6B). This further supports the notion that telomere length and the state of telomere folding are uncoupled from each other.

We estimated the telomere silencing status by measuring the levels of the long non-coding RNA (lncRNA) TERRA from telomere 1L. TERRA is a lncRNA transcribed from subtelomeric promoters which are under the influence of the local chromatin state [72]. As expected, subtelomeres of cells lacking SIR2 were desilenced and TERRA levels were approx. 10x higher than in wt cells (Fig 4G). As previously reported, the increase in TERRA was heterogeneous in tlc1 cells, due to the stochastic appearance of critically short telomeres (Fig 4G) [73]. TERRA levels were also slightly increased in sin3 and set2 mutants (Fig 4G). Therefore, the state of telomere folding may have effects on TERRA levels in the cell.

We show here, that Sir2, Sin3 and Set2 are involved in the regulation of telomere folding. The levels of these histone modifiers are downregulated during senescence, which may contribute to the open state during this process. Interestingly, Sir2, Sin3 and Set2 protein levels were not downregulated in a folding-proficient tel1 mutant with very short telomeres (Figs 2D, 2E, 2F and S6C). This observation supports our hypothesis that telomere folding is reduced due to a feature of the senescence program, which may include chromatin alterations through downregulation of chromatin modifiers, but is not caused solely by short telomeres per se.

We were wondering what the functional consequences of reduced telomere folding are, and if the opening of the telomere structure might lead to a higher incidence of non-homologous end joining (NHEJ) between chromosome ends. We addressed how frequently telomere-telomere fusions (T-TFs) occur in sir2, sin3 and set2 mutants by quantitative PCR for T-TFs between chromosomes V and XV [40,41] (Fig 4H). As expected mec1 tel1 sml1 cells accumulated T-TFs as previously reported [38–41] (Fig 4H). In contrast, the telomeres of sir2, sin3 and set2 mutants, as well as those of telomerase negative cells (tlc1), did not present an increased incidence of T-TFs. Also the double mutants of sir2, sin3 and set2 in combination with loss of TLC1 showed similar T-TFs frequencies as wt cells (Fig 4H). Therefore, decreased folding does not render telomeres more susceptible to NHEJ.

The open state of telomeres may also influence the recombination capacity of telomeres during the senescence process. We observed that in the absence of telomerase both sir2 and set2 mutants displayed a delayed senescence phenotype (S6D Fig). This delay was completely dependent on Rad52, as its deletion led to an increased rate of senescence (S6D Fig). This suggests that an unfolded telomere state may allow increased recombination, however in the absence of Rad52, the open state telomeres may be rendered more vulnerable to becoming dysfunctional (e.g. through resection). It was not possible to make such speculations based on the sin3 mutant as it showed an enhanced senescence rate that was independent of RAD52 (S6E Fig).

Discussion

In this study we adapted the 3C methodology to detect and quantify the interaction of an unmodified yeast telomere with adjacent regions on the juxtaposed subtelomere (Telo-3C). Interestingly, interactions between telomere 1L and its subtelomere occurred much more frequently (or stably) than internal interactions (of non-telomeric regions) on the same chromosome. We can envisage different explanations for this result, the most straightforward being that DNA ends have a greater mobility as compared to constrained internal loci and hence experience more frequent interactions with nearby loci. Indeed, the interaction frequency of telomere 1L gradually decreases along the chromosome arm towards the centromere, clearly indicating a proximity effect. Alternatively, but not mutually exclusive, the existence of a specific heterochromatin domain at the end of the chromosome may favor the formation of a fold-back structure. This is supported by the genetic dependence of telomere folding on chromatin modifiers such as Sir2, Sin3 and Set2, all of which promote histone deacetylation and hence a more heterochromatic environment. Although Telo-3C can confirm that such telomere-subtelomere interactions are occurring, we cannot infer the stability of the interactions. Indeed, a recent study employing ChEC (Chromatin Endogenous Cleavage) did not find evidence of a constitutively folded-back telomere [74]. This may be due to the transient nature of the interaction that is not detectable without a crosslinking step. In our study, we can also not gain insight into the structural features of the fold-back. It may resemble a lariat or loop with a single touch down point or it may fully fold back onto itself and have multiple touching points. We also cannot rule out that the heterochromatic domain at telomeres may exist in the shape of a condensed “knot” rather than the classical fold-back model, whereby interactions within the “knot” occur more frequently and more stably than outside of this structure. It is tempting to speculate which factors or chromatin landscape might define the borders of this heterochromatin domain. One possibility is that the borders are defined by a histone mark transition zone, as suggested for an extended silencing domain at subtelomeres [75], mainly concerning H3K79 methylation which is set by the histone methyltransferase Dot1. Finally, we cannot exclude that a portion of the Telo-3C signal arises from inter-telomere ligations. We believe this to be unlikely however, as the Telo-3C signals were unaltered in clustering-defective mutants (ku70, mre11, tel1 and hrs1). Furthermore, Sanger sequencing of the qPCR products revealed the expected sequences of the telomere 1L ligation products. The P5 and P6 primer set could potentially also detect an internal 3C interaction upstream of the Y' element on subtelomere 4R. However, the combination of the sequencing results, the use of clustering-defective mutants, and the fact that P5 and P6 bind ca. 7 kb away from the telomere at 4R, make it highly unlikely. Furthermore, P7 is telomere 1L-specific and the Telo-3C results with the primers P5 and P7 show the same trend as P5 and P6 through all experiments.

Surprisingly, telomere folding is not affected when any of the SMC complexes are depleted. This suggests, that the telomere folding mechanism is independent from the loop extrusion mechanism of cohesin, as well as being distinct from condensin-mediated chromatin condensation. We also observed, that the deletions of the telomeric proteins Rif1 or Rif2, which both result in telomere overelongation [4,5], do not reduce the frequency of telomere folding, as detected by Telo-3C. Moreover, the loss of Ku70, Tel1, Upf1 or Mre11, which leads to extensive telomere shortening, does not affect the ability of telomeres to fold back. Therefore, telomere length is not a determining factor with respect to telomere structure. The successful telomere folding in rif1 mutants also suggests that telomere folding is likely different from the chromatin architecture forming around a double-strand break. It was recently described in human cells that 53BP1 and RIF1 form distinct nanodomains around a double-strand break [76]. In this case, RIF1 –together with cohesin—was proposed to localize to the boundaries of these domains and stabilize chromatin topology.

Interestingly, we observed decreased telomere folding in cells depleted for the DNA damage checkpoint kinase Rad53. The absence of either one of the upstream kinases, Mec1 (ATR) or Tel1 (ATM), however, did not restrain the proper establishment of telomere folding, suggesting a redundant role for these kinases upstream of Rad53. Alternatively, Rad53 might promote telomere folding in a Mec1- and Tel1-independent manner. That the DNA damage checkpoint factors are important for proper telomere structure is congruent with the finding that in human cells, both ATM and ATR transiently localize to telomeres during the late S and G2 phases of the cell cycle, and are activated there [42,43]. This transient activation of the checkpoint at telomeres might not only be required for completing replication of telomeres but also for processing telomeric overhangs. Additionally, Verdun and Karlseder showed that the checkpoint activation then in turn transiently recruits the homology directed repair machinery, which likely assists in t-loop formation [42]. Accordingly, we found that by deleting both of the HDR factors, RAD51 and RAD52, yeast cells were not able to establish a closed state. Taken together, these genetic data strongly suggest that a certain heterochromatic state, established by histone deacetylases, together with DDR and HDR factors, promote a fold-back structure (Fig 5A, top) that may be analogous to the “closed” state telomere proposed in human cells. The fact that Rad51 and Rad52 are implicated suggests a strand invasion event into a homologous sequence, which would further the resemblance to mammalian t-loops.

To gain insight into telomere structure during replicative senescence, we used telomerase negative yeast cells (tlc1) as a model for this process. When cells enter replicative senescence and crisis, we found that telomeres transition from a closed, to an open state (unfolded). Since the opened telomere structure is not due to short telomere length there is likely another feature of the senescence program which is responsible for the opening. This notion is also supported by the observation that telomere folding is decreased after MMS treatment, which was shown to elicit a similar gene expression response as replicative senescence in budding yeast [77]. In both cases, cells activate a DNA damage response and a stress response in addition to an altered metabolic program [77]. We identified the histone modifiers Sir2, Sin3 and Set2 as potential regulators of telomere folding during senescence. We propose that their reduced protein levels during senescence contribute to ensuring an open state telomere. Sir2, Sin3 and Set2 all antagonize histone acetylation and were reported to be involved in telomere silencing or the regulation of the telomeric chromatin state. The integral role of the SIR complex at telomeres in terms of gene silencing through histone deacetylation, referred to as the telomere position effect (TPE), is well characterized [64,66,78]. Sin3 is part of the Rpd3 histone deacetylase complex and is involved in the regulation of silencing on a genome-wide level but also in TPE [67–70]. The SET domain-containing protein Set2 correlates with histone deacetylation because it methylates histone H3K36 which leads to recruitment of the Rpd3 histone deacetylase complex [71]. Sir2 levels did not drastically decrease during replicative senescence, however, we observed an uncharacterized modification of the protein. Sir2 was previously shown to be SUMOylated during mother cell aging in yeast, which leads to its relocalization from telomeres to the nucleolus [79]. Therefore, we hypothesize that Sir2 may also get redistributed from telomeres to the rDNA locus during senescence, contributing to the decreased telomere folding. Accordingly, Sir2 was shown to relocalize from telomeres to other genomic loci under various types of DNA damage [80], which could account for the MMS-induced telomere opening. Interestingly, Rap1, the direct telomere binding protein, was also shown to be redistributed during senescence [81]. This might potentially contribute to decreased SIR binding at telomeres during senescence, since the recruitment of the SIR proteins to telomeres depends on the presence of Rap1. Additionally, it is evident that histone levels themselves drastically decrease during senescence in yeast and human cells [81,82] and that histone synthesis is sensitive to DNA damage signaling [82]. This might also point towards a mechanism by which reduced nucleosome occupancy at subtelomeres contributes to the opening of telomere folding during senescence.

In mutants defective for telomere folding we did not observe an increased incidence of telomere-telomere fusions. Also when telomerase was deleted in addition to the histone modifiers Sir2, Sin3 and Set2, telomere-telomere fusion did not increase. This suggests that the non-homologous end joining repair pathway cannot act on unfolded telomeres and that they remain protected from fusions (Fig 5B). Open, but non-fusogenic telomeres in yeast may be similar to intermediate state telomeres that have been described in human cells as part of the three-state model [83,84]. In this model, telomeres exist in a “closed”, “intermediate” or “uncapped” state. In the fully capped, “closed” state, the presence of a t-loop was suggested to prevent a DNA damage response that would elicit repair activities at telomeres (as in Fig 5A, top). In the “intermediate” state, telomeres activate a DNA damage response likely due to the loss of the protective loop structure at the chromosome end. This state, however, still retains the capacity to suppress fusions which would correspond to the observed yeast telomeres lacking a folded structure but being resistant to NHEJ (Fig 5B). In the “uncapped” state, telomeres are fully dysfunctional, open, and permissive to repair events (Fig 5A, bottom).

In summary, we propose that the presence of the histone modifiers Sir2, Sin3 and Set2, as well as the DDR and HDR factors, is required for the successful establishment of telomere folding (Fig 5). In wt yeast cells, DNA damage checkpoint factors, such as Rad53, and the HDR machinery (including Rad51 and Rad52) might only transiently associate to telomeres. Here they may assist establishing a telomere fold-back structure, with potential strand invasion of the 3’ overhang. In this scenario, telomere folding might prevent sister chromatid exchange between telomeres and/or limit excessive resection, which corresponds to a “closed” telomere state. However, when either Sir2, Sin3 or Set2 are absent, telomeres do not fold-back efficiently and the telomeric chromatin state is likely altered (Fig 5B). We believe, that this situation might represent the “intermediate” state telomeres which have lost the folded structure but still suppress telomere fusions.

When telomerase negative cells enter crisis, we observed that the levels of Sin3 and Set2 are decreased and that Sir2 becomes modified. We hypothesize that the loss of these histone modifiers contributes to the maintenance of the open state (Fig 5A). It is difficult, however, to know if the loss of the histone modifiers is the key event in triggering the opening or if it plays a supportive role in ensuring that the open state is maintained. Additionally, the accumulation of cells at the G2/M border during senescence might also contribute to achieve an unfolded state. Nonetheless, the open telomeres in combination with uncapping are subject to processing and associate with Rad51 and Rad52 [33]. In this context, the HDR factors are not able to promote telomere folding due to the absence of the required chromatin modifiers, but rather drive telomere repair through HDR (sister chromatid exchanges and break-induced replication).

Here, we have identified key regulatory factors for establishing telomere folding in budding yeast. Our results point towards the distinct telomeric chromatin environment and the homologous recombination machinery as being major requirements for the folding of yeast telomeres. We provide insight into how the topological reorganization of telomeres integrates into the senescence program.

Materials and methods

Yeast strains and plasmids

Strains used in this study are listed in S1 Table. Strains were grown under standard conditions in YPD (Yeast Peptone Dextrose) at 30°C if not indicated otherwise.

Measurement of telomere length by PCR

Genomic DNA was extracted from 10 ml of exponentially growing cells with the Gentra Puregene Yeast/Bact. Kit (QIAGEN) according to the manufacturer’s protocol. For the C-tailing reaction, 100 ng of genomic DNA were mixed with 0.9 μl of NEB buffer 4 in a final volume of 9 μl. To denature the DNA, the mixture was incubated at 96°C for 10 min and cooled to 4°C. 4 units of terminal transferase (NEB) were added together with dCTP in a final concentration of 0.1 mM. The C-tailing reaction was carried out as follows: 37°C for 30 min; 65°C for 10 min; 96°C for 5 min; hold at 65°C. 30 μl of pre-heated PCR mix were added to the DNA. The PCR mix contained 4 μl of 2 mM dNTPs, 4 μl of PCR buffer (670 mM Tris-HCl pH 8.8, 160 mM (NH₄)₂SO₄, 70% glycerol, 0.1% Tween-20), 0.4 μl of Q5 Hot Start DNA Polymerase (2 u/μl, NEB) and the primers oBL358 and oBL359. The following cycling conditions were used for PCR: 95°C for 3 min; 45 cycles of 95°C for 30 s, 63°C for 15 s and 72°C for 20 s; 72°C for 5 min; hold at 4°C. PCR products were separated on an 1.8% agarose gel containing 1X SYBR Safe DNA gel stain and imaged on the Bio-Rad ChemiDoc Touch Imaging System.

Flow cytometry analysis for DNA content

0.18 OD₆₀₀ units of cells were collected and the cell pellet was washed with 1 ml of water. Cells were fixed in 70% ethanol overnight at 4°C. After fixation, cells were centrifuged at 13,000 rpm for 5 min and washed in 800 μl 50 mM Tris pH 7.5. RNA was digested in a 50 mM Tris pH 7.5 solution containing 0.25 mg/ml RNase A for 2 h at 37°C. 25 μl of 20 mg/ml proteinase K were added and incubated at 50°C for 2 h. Samples were sonicated for 10 s in a BRANSON sonifier 450 with the following settings: constant mode, 10 s, power 1. After sonication SYTOX Green was added to a final concentration of 2 μM in 50 mM Tris pH 7.5 and analyzed on a BD FACSVerse flow cytometer. 20,000 events were recorded per sample and the data was analyzed with BD FACSuite.

Flow cytometry analysis for cell viability

0.68 OD₆₀₀ units of cells were collected and the cell pellet was washed with 1 ml 50 mM Tris pH 7.5. After spinning down the cells for 1 min at 7000 rpm, the pellet was resuspended in 1 ml 50 mM Tris pH 7.5 containing 0.5 μM SYTOX Green. Cells were analyzed on a BD FACSVerse flow cytometer recording 20,000 events. The data was analyzed with FlowJo 10.6.1. As a control sample for dead cells, cells were incubated at 95°C for 15 min and subjected to the described protocol.

RNA extraction and qRT-PCR to measure TERRA levels

RNA extraction, DNase treatment and reverse transcription were performed as described in [73]. qPCR analysis was performed by the CFX384 Touch Real-Rime PCR Detection System (Bio-Rad) using DyNAmo Flash SYBR Green (Thermo Scientific).

Telomere chromosome conformation capture (Telo-3C)

25 ml of exponentially growing cells were crosslinked with 1.2% formaldehyde at OD₆₀₀ 0.4 for 20 min at room temperature. The reaction was quenched with 360 mM glycine for 5 min at RT and cells were placed on ice afterwards. The cells were washed twice with cold PBS and the pellet was resuspended in 200 μl FA-lysis buffer (50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA pH 8.0, 1% Triton X-100). Cell lysis was performed in Lysematrix C tubes (MP Biomedicals) using the FastPrep (MP Biomedicals; 2x 30 s with 1 min on ice in between runs; 6.5 M/s). Cell extracts were recovered in FA-lysis buffer containing 0.1% sodium deoxycholate (SOD) and centrifuged for 15 min at 4°C, 13,000 rpm. The pellet was washed with FA-lysis buffer and resuspended in 200 ml of 20 mM Tris-HCl pH 7.5. The chromatin concentration of this 3C extract was estimated by determining its protein concentration by Bradford. Subsequently, the amount of chromatin corresponding to 60 μg of protein was digested with 100 units NcoI-HF (NEB) overnight at 37°C. The digestion reaction was stopped by incubating the samples at 65°C for 20 min in the presence of 1% SDS. The SDS was sequestered with 75 μl of 10% Triton X-100 and 570 μl of 20 mM Tris-HCl pH 7.5 were added. The chromatin was pelleted for 15 min at 4°C, 13,000 rpm, and resuspended in 750 μl ligation mixture containing 200 units T4 DNA ligase (NEB/IMB core facility protein production), 1X T4-Ligation buffer (NEB) and Tris-HCl pH 7.5. After ligation at RT for 5 h, RNA was digested with 20 μg RNase A and crosslinks were reversed with 7.5 μl proteinase K and 3.5 μl of 20% SDS at 65°C overnight. To isolate the DNA, the 3C reaction was mixed with 750 μl Phenol:Chloroform:Isoamylalcohol and the DNA was precipitated from the aqueous phase by adding 2 μl glycogen, 75 μl of 3 M KAc and 2 ml of 100% Ethanol. The DNA pellet was washed with 70% Ethanol and resuspended in 40 μl ddH₂O. The DNA concentration was determined with Qubit HS DNA reagent (Thermo Scientific). 1.5 ng of DNA were used per qPCR reaction, performed with CFX384 Touch Real-Rime PCR Detection System (Bio-Rad) and Dynamo Flash SYBR GREEN (Thermo Scientific). Primers for Telo-3C were used in a final concentration of 0.5 μM in the qPCR reaction. To account for the amount of input DNA that is used for the qPCR reaction, the Telo-3C-specific qPCR signal was normalized to the intensity of a control qPCR product (primers oRS39 and oRS40), which is robustly amplified from genomic DNA. The relative interaction frequencies of telomere 1L with its corresponding subtelomere in the investigated conditions were then determined relative to the respective wt or untreated sample, as follows:

The adjusted p-values were obtained from the statistical tests indicated in the figure legends. The type of post-hoc test was employed according to GraphPad Prism 8. The Tukey test was employed when comparing every mean with every other mean, such as in Fig 3B or 3G. The Dunnett’s test was used when comparing a control mean (e.g. the wt sample) to the other means, such as in Figs 2F, 3C, 3D and 4D. The Sidak method was employed when a set of means was selected to compare, such as in Fig 2B (wt and tlc1 Telo-3C interaction frequencies were compared for each time point) and Fig 2G (Telo-3C interaction frequencies +/- MMS were compared for each time point).

Protein extraction, SDS-PAGE and western blot

Proteins were extracted from 2 OD₆₀₀ units of yeast cells as described in [73]. Protein extracts were loaded on 4–15% precast SDS-PAGE gels (Bio-Rad). Proteins were blotted on a nitrocellulose membrane with the Trans-Blot Turbo Transfer System (Bio-Rad) using the high molecular weight program. The membrane was blocked for 1 h with 5% skim milk in PBST and primary antibodies were used in the following dilutions in 5% skim milk in PBST: PAP (Sigma Aldrich) at 1:3000; anti-Sir2 (Santa Cruz) at 1:1000; anti-Actin (Millipore) at 1:2000; anti-MYC-tag (Cell Signaling Technology/NEB) at 1:1000; anti-Pgk1 (Invitrogen) at 1:20,000. The membranes were incubated with the primary antibodies overnight at 4°C. After four washing steps with PBST, the respective secondary antibodies were added for 1 h at RT: goat anti-mouse-HRP conjugate (Bio-Rad) at 1:3000; goat anti-rabbit-HRP conjugate (Bio-Rad) at 1:3000; IRDye 680RD goat anti-mouse IgG (LI-COR) at 1:10,000; IRDye 800CW goat anti-rabbit IgG (LI-COR) at 1:10,000. Before detection, membranes were washed three times in PBST and once in PBS.

Membranes with HRP-coupled secondary antibodies were developed using the Super Signal West Pico Chemiluminescent Substrate (Thermo Scientific) or the Super Signal West Dura Extended Duration Substrate (Thermo Scientific) on the Bio-Rad ChemiDoc Touch Imaging System.

Membranes with secondary antibodies coupled to IRDyes were detected on the Odyssey CLx 9140 Imaging System (LI-COR) and processed via the Image studio software (LI-COR).

Southern blot

Genomic DNA (5 μg) was digested with the respective restriction enzyme for 5 h at 37°C (XhoI for all telomeres; SalI for telomere 1L). The digested DNA was separated on a 0.8% agarose gel overnight at 50 V. DNA in the gel was denatured for 1 h (0.4 M NaOH, 0.6 M NaCl) and neutralized for 1 h (1 M Trizma Base, 1.5 M NaCl, pH 7.4). DNA was transferred to a nylon membrane (Hybond NX, GE Healthcare) via capillary transfer in 10X SSC overnight and crosslinked to the membrane with UV light (auto X-link, UV Stratalinker 2400, Stratagene). The membrane was pre-hybridized for 5 h at 55°C in hybridization solution (PerfectHyb Plus Hybridization Buffer, Sigma Aldrich). Hybridization of the respective radio-labeled probe was carried out overnight at 55°C. The membrane was washed twice in 2X SSC with 0.1% SDS for 5 min and twice in 0.5X SSC with 0.1% SDS for 20 min. All washing steps were performed at 55°C. The membrane was exposed to a phosphoimager screen and the signal was detected via Typhoon FLA 9500 (GE Healthcare).

Probes for Southern blot

The telomere-specific probe was obtained by digestion of pBL423 with EcoRI followed by gel extraction using the QIAquick Gel Extraction Kit (Qiagen). Plasmid pBL423 was a kind gift from M.P. Longhese (pSP100). The probe for subtelomere 1L was generated by PCR with oligos oSM1 and oSM2 on genomic DNA of wt S. cerevisiae cells. The PCR product was separated on a 1% agarose gel and purified using the QIAquick Gel Extraction Kit (Qiagen). 60 ng of each probe were labeled with dATP [α-32P] (DECAprime kit II; Thermo Scientific) by random priming according to the manufacturer’s instructions.

Senescence curve

Liquid senescence assays were performed as described in [85]. Briefly, the OD₆₀₀ of every culture was measured every 24 h and cultures were diluted back to OD₆₀₀ 0.01 subsequently. All analysis during the senescence curve were performed on exponentially growing cultures, starting from the saturated culture after 24 h.

MS/MS

1 OD₆₀₀ unit of exponentially growing cells was resuspended in 1X LDS buffer supplemented with 100 mM DTT. Cells were heated for 10 min at 95°C and separated on a 4–12% NuPAGE Bis–Tris precasted PAGE gel (Thermo Scientific). Samples were run at 180 V for 10 min and processed by in-gel digestion [86]. Briefly, samples were reduced in reduction buffer (10 mM DTT in 50 mM ammonium bicarbonate (ABC) buffer) for 1 h at 56°C and alkylated in alkylation buffer (50 mM iodoacetamide in 50 mM ABC buffer) for 45 min in the dark. Proteins were digested with 2 μg Protease LysC (Wako Chemicals) overnight at 37°C in 50 mM ABC buffer. Digested peptides were desalted on a C18 StageTip as described [87] and analyzed by nanoflow liquid chromatography on an EASY-nLC 1000 system (Thermo Scientific) coupled to a Q Exactive Plus mass spectrometer (Thermo Scientific). The peptides were separated on a self-packed reverse phase capillary (75 μm diameter, 25 cm length packed with C18 beads of 1.9 μm (Dr Maisch GmbH). The capillary was clamped on an electrospray ion source (Nanospray Flex, Thermo Scientific). A 90 min gradient starting from 2% - 60% gradient acetonitrile in 0.1% formic acid was used at a flow of 225 nl/min. Data was collected in data-dependent acquisition mode with one MS full scan followed by up to 10 MS/MS scan with HCD fragmentation.

MS data processing and bioinformatics analysis

MS raw files were processed using the MaxQuant software (version 1.5.2.8) and the ENSEMBL S. cerevisiae protein database (Saccharomyces_cerevisiae.R64-1-1.24). LFQ quantitation and match between run options were activated. MaxQuant output files were analyzed using an in-house R script. Known contaminants, reverse hits and protein groups only identified by site modification were excluded. Identified protein groups (minimum 2 peptides, 1 of them unique) were further filtered to a minimum of 2 quantification events per experiment. Missing values were imputed using a downshifted and compressed beta distribution within the 0.001 and 0.015 percentile of the measured values for each replicate individually. The LFQ intensities were log₂ transformed and a two sample Welch t-test was performed. Volcano plots were generated by plotting -log₁₀(p-values) and fold changes. The threshold line for enriched proteins is defined with p-value = 0.05. Gene ontology analysis were performed with the PantherDB.org [88] overrepresentation Test (Release 20190711) with the annotation database released on 20190703. Fisher’s exact test followed by FDR correction was applied to calculate the p-values. Heatmap for enriched proteins were generated using the “pheatmap” (version 1.0.12) or “ggplot2” (version 3.2.1) package in R. Additionally the Venn diagram was generated using the “eulerr” (version 5.1.0) package in R (version 3.5.1).

Telomere-telomere fusions analysis

Telomere-telomere fusions (T-TFs) were analyzed by quantitative PCR analyses as reported [40], with the modifications described in [41]. Briefly, ~100 ng of Sau3A-treated genomic DNA extracted by standard protocols from asynchronous cultures was PCR-amplified for quantitative analyses using a primer from the X element of chromosome XV-L and a primer from the Y’ element of chromosome V-R. A DNA fragment from HIS4 was PCR-amplified as input control. The oligonucleotides used for quantitative analyses and the PCR conditions described previously [40]. The frequency of T-TFs per genome was calculated with the formula: T-TFs/genome = 2^–N / N = Ct (T-TFs)–Ct (HIS4). Prior to applying this formula, the curves representing the increasing amounts of DNA for the two products as a function of the number of PCR cycles were confirmed to be parallel (i.e., the slope of the curve representing the log of the input amount versus ΔCt was < 0.1).

Construction of strains with auxin-inducible degrons

Strains carrying auxin-inducible degrons (AID*) for SMC complexes were created as described before [56]. Primers for tagging are listed in S2 Table.

The RAD53-AID*-9MYC construct was transferred from the strain published by Morawska and Ulrich into the S288C background [56]. For this purpose, the C-terminal sequence of RAD53 including the tag was amplified by PCR and integrated into a strain carrying leu2::AFB2::LEU2 (yKB244).

Materials

Materials such as antibodies, enzymes, chemicals and software are listed in S3 Table.

Numerical data

Underlying numerical data for all graphs are provided in S4 Table.

Supporting information

S1 Fig [pdf]

S2 Fig [tg]

S3 Fig [pdf]

S4 Fig [pdf]

S5 Fig [go]

S6 Fig [pdf]

S1 Table [xlsx]
Yeast strains used in this study.

S2 Table [xlsx]
Primers used in this study.

S3 Table [xlsx]
Materials used in this study.

S4 Table [xlsx]
Underlying numerical data.