Recombination at subtelomeres is regulated by physical distance, double‐strand break resection and chromatin status

Abstract

Homologous recombination (HR) is a conserved mechanism that repairs broken chromosomes via intact homologous sequences. How different genomic, chromatin and subnuclear contexts influence HR efficiency and outcome is poorly understood. We developed an assay to assess HR outcome by gene conversion (GC) and break‐induced replication (BIR), and discovered that subtelomeric double‐stranded breaks (DSBs) are preferentially repaired by BIR despite the presence of flanking homologous sequences. Overexpression of a silencing‐deficient SIR3 mutant led to active grouping of telomeres and specifically increased the GC efficiency between subtelomeres. Thus, physical distance limits GC at subtelomeres. However, the repair efficiency between reciprocal intrachromosomal and subtelomeric sequences varies up to 15‐fold, depending on the location of the DSB, indicating that spatial proximity is not the only limiting factor for HR. EXO1 deletion limited the resection at subtelomeric DSBs and improved GC efficiency. The presence of repressive chromatin at subtelomeric DSBs also favoured recombination, by counteracting EXO1‐mediated resection. Thus, repressive chromatin promotes HR at subtelomeric DSBs by limiting DSB resection and protecting against genetic information loss.

Introduction

DNA lesions, arising from either environmental stress or endogenous events, challenge genomic integrity. Double‐stranded breaks (DSBs) are a highly genotoxic form of DNA damage, and their improper repair leads to genomic instability or cell death. DSB repair can occur through two different mechanisms: non‐homologous end joining (NHEJ) and homologous recombination (HR).

HR is engaged at DSBs following resection and the formation of 3′ single strands, required for homology search and strand invasion. MRX/Sae2, Exo1 and Sgs1/Dna2 act together to resect DSBs (Gravel et al, 2008; Mimitou & Symington, 2008; Zhu et al, 2008) and generate 3′ ssDNA overhangs that are rapidly stabilized by RPA (Alani et al, 1992). Subsequent coating by Rad51 engages the broken 3′ ssDNA ends in genome probing, likely through repeated interactions with nearby sequences to find a homologous sequence (Renkawitz et al, 2013). One of the 3′ ssDNA ends invades a homologous template and copies the DNA sequences needed to seal the break. The second 3′ end of the DSB can then anneal with this extended and displaced 3′ end that initiated strand invasion, in repair by the “synthesis‐dependent strand annealing” (SDSA) pathway. Alternatively, it can form a double Holliday junction by annealing with the single‐stranded DNA in the displacement loop (D‐loop) generated by strand invasion. In this case, D‐loop resolution yields either non‐crossover (NCO) or crossover (CO) products (Heyer et al, 2010). SDSA, NCO and CO repair are usually collectively referred to as gene conversion repair (GC).

GC is typically an error‐free mechanism that limits loss of heterozygosity (LOH) to a small region surrounding the DSB. However, HR can produce break‐induced replication events (BIR) when the formation of the initial displacement loop (D‐loop) is followed by conservative replication. In BIR, the D‐loop migrates to the end of the chromosome, resulting in kilobase‐long tracks of LOH (Llorente et al, 2008; Donnianni & Symington, 2013; Saini et al, 2013; Wilson et al, 2013). BIR is the predominant repair pathway employed when only one DSB end is available for strand invasion. It is also used to restart collapsed replication forks and to elongate telomeres when telomerase is absent or telomeres are uncapped (McEachern & Haber, 2006; Llorente et al, 2008). GC and BIR pathways are well defined at the molecular level, but the factors limiting their usage and relative efficiency in the context of the nucleus are far from being fully understood.

The nuclear organization of Saccharomyces cerevisiae is well defined through microscopy and chromosome conformation capture experiments (Taddei & Gasser, 2012). During the exponential growth phase, interphase budding yeast chromosomes assume a Rabl‐like conformation, with the 16 centromeres held together by the spindle pole body (SPB) at one nuclear pole. The 32 telomeres are found at the nuclear periphery, forming 3–4 foci where the yeast silent chromatin or heterochromatin factors (SIRs—silent information regulators) concentrate. Although budding yeast lacks the molecular factors associated with heterochromatin found in most other eukaryotes, silent chromatin generated by the SIR complex shares most of the functional features of heterochromatin. These features include late replication timing, heritable repression of transcription, a preferential association with the nuclear envelope and the formation of foci. Foci arise from the trans‐association of “heterochromatic” loci (Meister & Taddei, 2013). Although these foci do not contain reproducible subsets of chromosomal ends, telomeres at the extremities of chromosomes with comparable arm lengths interact more frequently (Duan et al, 2010; Therizols et al, 2010; Guidi et al, 2015).

Because of this genome organization, some loci are in contact more frequently and may be more prone to recombination. Consistently, HR efficiency generally negatively correlates with the spatial distance between a DSB and its homologous targets (Wilson et al, 1994; Burgess & Kleckner, 1999; Agmon et al, 2013; Lee et al, 2016). However, these studies also revealed outliers, for which spatial distances and HR efficiencies were uncorrelated. These outliers suggest that additional factors, other than spatial distance, limit HR efficiency. Indeed, DSB resection, along with the quality of homology between recipient and donor sequences, was shown to limit HR in S. cerevisiae (Lee et al, 2016). Furthermore, sequence context and chromatin structure have recently emerged as regulatory elements for DNA repair efficiency. Studies in Drosophila and mammals suggest that irradiation‐induced DNA damage in heterochromatin is preferentially repaired by HR rather than NHEJ, but whether HR efficiency differs in heterochromatic compared to euchromatic regions remains unclear (Goodarzi et al, 2008; Chiolo et al, 2011; Goodarzi & Jeggo, 2013). In addition, a preference for HR may be specific to IR lesions because, at least in Drosophila tissues, an I‐SceI break induced at heterochromatic loci is repaired through both HR and NHEJ (Janssen et al, 2016).

In budding yeast, the impact of repressive chromatin structures on HR regulation has not been fully addressed, probably due to the under‐representation of loci embedded in heterochromatin‐like structures. Finally, how the different limiting factors affect HR efficiency and the balance between GC and BIR remains to be determined. To address these questions, we developed an assay to score DSB‐induced recombination events at subtelomeric loci. This assay allows us to measure the competition between GC and BIR. We coupled this assay with genetic modifications that modulate the physical distances between telomeres, and/or spreading of repressive chromatin in subtelomeric regions, to determine how these parameters regulate HR efficiency and outcome (BIR versus GC). We discovered that subtelomeric DSBs are predominantly repaired by BIR, despite the presence of homology with donor sequences on both sides of the break. Further, we found that spatial distance, one of the limiting factors for HR efficiency, mainly impacts GC rate between subtelomeres. However, when recombination is between subtelomeric and intrachromosomal regions, DSB chromosomal location and chromatin structure show a stronger impact than spatial distance on HR efficiency. Our data suggest that DNA loss from the telomere proximal fragment limits GC at subtelomeric DSBs. Moreover, our findings are compatible with a model where silent chromatin favours GC by inhibiting DSB processing and the ensuing loss of sequences surrounding the DSB.

Results

Subtelomeric DSBs are efficiently repaired through BIR‐mediated non‐reciprocal translocations

We developed a recombination assay that scores DSB‐induced repair events between URA3 alleles inserted at different chromosomal positions (Fig 1A and B). This system includes two recombination cassettes of about 1.3‐kb: a “recipient cassette” bearing a ura3 allele with a single 30 bp I‐SceI cleavage site (Colleaux et al, 1988), and a “donor cassette” with a ura3‐1 allele. The 30 bp I‐SceI recognition site was efficiently cleaved after induction of I‐SceI endonuclease, which is driven by the inducible GAL promoter (Fig 1C). We estimated DSB repair by comparing the colony‐forming abilities of cells with I‐SceI‐induced DSB breaks (grown on galactose‐containing medium) to those without I‐SceI‐induced DSB breaks (grown on glucose‐containing medium). Cells that formed colonies on galactose‐containing medium lost the I‐SceI restriction site on the recipient cassette, as inferred from in vitro digestion of a PCR‐amplified cassette. Thus, cleavage was efficient and DSBs were subsequently repaired through inaccurate NHEJ or recombination‐based mechanisms (Fig 1D). Plating efficiencies of donor‐less strains containing a DSB in the recipient cassette were < 0.5%, indicating that repair through inaccurate NHEJ rarely occurred (Fig 1G). Indeed, the presence of the donor ura3‐1 allele at the URA3 locus increased cell survival on galactose plates 17‐ to 300‐fold, indicating that HR is the most frequent repair pathway when a homologous sequence is present (Fig 1G and H).

Figure 1. An assay to score recombination efficiency reveals prominent BIR repair of subtelomeric DSB.

1. Schematic representation of the two ura3 alleles used to test recombination efficiencies and outcome: i, the recipient ura3‐I‐SceI has a 30 bp I‐SceI sequence inserted out of frame ii, the donor ura3‐1 bears a missense mutation.
2. The two ura3 alleles are introduced at chosen loci by PCR gene targeting.
3. DSB cleavage efficiency measured in donor‐less strains by qPCR using primers flanking the DSB site. Error bars represent the standard deviation (SD) of at least three independent experiments.
4. Disappearance of the I‐SceI cleavage site in survivors on galactose medium assessed by in vitro digestion by I‐SceI of PCR products amplified with primers flanking the DSB site.
5. Schematic of the primers used to test GC, BIR or NRT by PCR.
6. Representative PCR obtained for the TEL6R TEL4R and TEL6R ura3‐1i strains.
7. Survival frequencies observed after induction of a DSB with or without recombination substrate. Error bars represent the survival standard error (SEM) of at least three independent experiments.
8. Survival frequencies and GC (dark) or Pol32‐dependent BIR (light grey) repair events observed after induction of a DSB at a subtelomeric position. Error bars represent the survival standard error (SEM) of at least three independent experiments. Asterisks indicate statistical differences for survival (****P < 0.001). See Table EV1 for statistical analysis of the GC and BIR repair events and Table EV2 for SEM values.
9. BIR and GC events are distinguished on high‐resolution pulsed‐field gel electrophoresis. Repair by BIR leads to a shift of the chromosome size as indicated. Numbers refer to chromosomes.

Subtelomeric DSBs repaired by either GC or BIR will ensure cell viability. In contrast, BIR at intrachromosomal loci would be lethal. To distinguish between the repair pathways used at subtelomeric DSBs, we performed PCR with primers flanking the recombination cassettes (Fig 1E and F). We discovered that up to 75% of subtelomeric DSBs exhibited typical BIR non‐reciprocal translocations (NRTs), despite both sides of the break having about 700 bp of homology with the subtelomeric donor (Fig 1A and F). High‐resolution pulsed‐field gel electrophoresis (PFGE) confirmed NRTs of the expected size for BIR events and revealed neither a change of the donor chromosome size nor additional gross chromosomal rearrangements (Fig 1I). Subtelomeric DSBs also engaged in BIR rather than GC when the donor was inserted at an intrachromosomal site, as long as the orientation of the sequence allowed viable repair events. Indeed, 50% of the survivors exhibited typical BIR rearrangements following induction of a DSB in the TEL6R subtelomeric region, in the presence of a ura3‐1 allele in the proper orientation (TEL6R ura3‐1i, Fig 1B, H and I). NRTs depended strongly on the DNA Polδ subunit Pol32, a factor required for BIR (Fig 1H; Lydeard et al, 2007; Deem et al, 2008). BIR events between TEL6R and ura3‐1i or TEL4R, which required polymerization of 117 kb or 10 kb, respectively, relied entirely on POL32 (Fig 1H). However, in the absence of Pol32, recombination between TEL6R and TEL9R subtelomeres still led to NRTs that account for 5% of survival (Fig 1H). These could correspond to recombination events in which the D‐loop extended over 3.2 kb to the end of chromosome 9R. It is noteworthy that although Pol32 has been described as specifically required for BIR, we also observe a twofold to fourfold decrease in GC repair events in the absence of Pol32, suggesting that Pol32 also participates in GC (Fig 1H, Tables EV1 and EV2). In summary, subtelomeric DSBs show limited GC efficiency compared to intrachromosomal DSBs, but are repaired efficiently through BIR, which accounts for 50–75% of repair events. Together, these results show that BIR is favoured at subtelomeric DSBs even in the presence of homologous sequences sufficient to allow significant levels of GC (TEL6R ura3‐1i; Fig 1H).

Telomere clustering upon sir3A2Q overexpression favours gene conversion but not BIR

Strikingly, the survival rate of cells with a subtelomeric DSB varies from 19 to 75%, depending on chromosomal location of the donor sequence (Fig 1H). These differences in HR efficiency may reflect a “barrier” that limits HR between subtelomeric and intrachromosomal loci (Pryde & Louis, 1997), and/or the spatial distance between homologous sequences at different loci (Agmon et al, 2013). In agreement with the second hypothesis, the highest HR efficiency was observed between TEL6R and TEL9R subtelomeric loci. These loci are at similar distances from their centromeres and thus spatially close, as shown both by microscopy and HiC (Duan et al, 2010; Therizols et al, 2010; Guidi et al, 2015). In contrast, the recombination rate between TEL6R and the more distant TEL4R, which lies at the end of a long chromosome arm, was almost two times lower (Fig 1H). These results reinforce the previously observed correlations between HR efficiencies and spatial distances, inferred either from the overlap of positions occupied by loci in the nucleus (Agmon et al, 2013) or from HiC contact maps (Lee et al, 2016). However, in these studies, the relative contribution of BIR versus GC events on survival rate was not monitored.

To directly test the impact of the physical distance on recombination rate and choice of repair pathway, we modulated the clustering of telomeres in the nucleus. Specifically, we overexpressed Sir3, which leads to the formation of a “hypercluster” of telomeres at the centre of the nucleus (Ruault et al, 2011; Fig 2A and B). Importantly, this increased clustering can be uncoupled from silencing by overexpressing the sir3A2Q silencing‐defective allele (Ruault et al, 2011). To overexpress sir3A2Q, the A2Q mutation was inserted by gene targeting in the endogenous SIR3 gene, along with a strong GDP promoter to drive overexpression. As overexpression of sir3A2Q also led to loss of subtelomeric silencing (Ruault et al, 2011), we used the sir3Δ mutant as a control. As expected, the absence of SIR3 or overexpression of sir3A2Q did not significantly affect HR efficiency between two intrachromosomal cassettes (LYS2 ura3‐1; Fig 2C and Table EV1). In addition, deletion of SIR3 did not affect HR levels between TEL6R and TEL4R or TEL9R, indicating that loss of SIR3 and subtelomeric silencing did not have a global impact on recombination efficiency (Fig 2D and Table EV1). However, sir3A2Q overexpression that increased physical proximity between subtelomeric cassettes (TEL6R TEL4R and TEL6R TEL9R) significantly increased HR efficiencies (Fig 2D). Interestingly, our molecular analysis revealed that this effect stemmed from a twofold to threefold increase in GC efficiency, whereas BIR was not affected in these conditions for the two telomere pairs tested (Fig 2E and F, Table EV1). Similarly, GC events increased fourfold in response to a DSB at TEL4R upon sir3A2Q‐mediated telomere clustering, showing that this effect is not specific to TEL6R (Fig EV1). These data argue that telomere clustering, which increases spatial proximity, favours recombination between homologous sequences. This posits that homology searching is a limiting factor for GC efficiency. However, BIR efficiency did not increase when the proximity between homologous sequences increased, suggesting that BIR efficiency is limited by another step, such as initiation or progression of new DNA synthesis (Jain et al, 2009; Donnianni & Symington, 2013).

Figure 2. Telomeres clustering upon sir3A2Q overexpression specifically favours gene conversion.

• A
  Rap1 foci grouping upon sir3A2Q overexpression. The sir3A2Q allele is silencing defective and its overexpression leads to telomere clustering at the centre of the nucleus. Representative fluorescent images of the telomere‐associated protein GFP‐Rap1 and of the nucleolus visualized through Sik1‐mRFP in WT, sir3∆ and sir3A2Q‐overexpressing cells. Scale bar is 500 nm.
• B
  Schematic representation of the experimental design showing telomere organization, chromatin status and DSB localization.
• C, D
  Survival frequencies observed after induction of a DSB in WT, sir3∆ and sir3A2Q‐overexpressing cells as in Fig 1G and H.
• E, F
  Repair events (GC and BIR) after induction of a DSB at TEL6R with TEL4R (E) orTEL9R (F) as a donor in WT and in cells overexpressing the sir3A2Q mutant protein.

Data information: Error bars represent the standard error (SEM) of at least three independent experiments. Asterisks indicate statistical differences (*P < 0.05; **P < 0.01; ***P < 0.005; ****P < 0.001). See Table EV1 for statistical analysis of the GC and BIR repair events and Table EV2 for SEM values.

Figure EV1. Telomeres clustering upon sir3A2Q overexpression specifically favours GC at TEL4R.

1. Survival frequencies observed after induction of a DSB at TEL4R in WT‐ and sir3A2Q‐overexpressing cells as in Fig 1G and H.
2. Repair events (GC and BIR) after induction of a DSB at TEL4R with TEL6R as a donor in WT and in cells overexpressing the sir3A2Q mutant protein.

Data information: Error bars represent the standard error (SEM) of at least three independent experiments. Asterisks indicate statistical differences (***P < 0.005).

DSBs in subtelomeric regions are repaired with low efficiency using intrachromosomal donors

As discussed above, the existence of a barrier to recombination has been proposed to explain the low HR rate between subtelomeric and internal loci (Pryde & Louis, 1997). In agreement with this hypothesis and with previous reports (Marvin et al, 2009a,b; Agmon et al, 2013), we observed a lower rate of recombination between a subtelomeric DSB at TEL6R and the URA3 internal locus (10% survival), than between two internal loci (30% survival between LYS2 and URA3) or two subtelomeric loci (survival ranging from 40 to 80% for TEL6R TEL4R and TEL6R TEL9R; Fig 1H). However, these pairs also show different physical distances that could account for the different HR efficiency. To distinguish between the contributions of physical distance and chromosomal location of the DSB, we compared the recombination efficiency between reciprocal pairs of intrachromosomal and subtelomeric loci (Fig 3A). Unexpectedly, we observed a very strong asymmetry in repair efficiency within pairs of intrachromosomal and subtelomeric loci. Indeed, a DSB induced at the intrachromosomal URA3 locus was repaired efficiently with donor sequences inserted at the TEL6R or TEL4R subtelomeres, leading to 54 and 45% survival, respectively (Fig 3B). In contrast, DSBs induced at TEL6R or TEL4R with the URA3 locus as a donor only led to 20 and 3% survival, respectively (Fig 3B). Subtelomeric loci thus appear to be efficient recombination donors but poor acceptors. It is noteworthy that subtelomeric DSBs recombine with intrachromosomal donors with low efficiency, even though BIR repair led to viable progenies and accounted for one‐third to half of the survivors for TEL4R and TEL6R, respectively. The low HR efficiency of subtelomeric DSBs with intrachromosomal sequences recapitulated the results of spontaneous recombination analyses (Marvin et al, 2009a,b). However, the efficient HR between intrachromosomal DSBs and subtelomeric donors that we observed is not consistent with a recombination barrier between these two loci. Furthermore, the HR rate varied up to 15‐fold for reciprocal pairs of loci—thus at the same spatial distance—depending on which locus is damaged. This demonstrates that the pre‐existing physical distance between homologous sequences is not the only limiting factor for recombination efficiency.

Figure 3. Subtelomeric loci are good recombination donors but poor acceptors for HR.

1. Schematic representation of the assay showing DSB localization and telomere at the nuclear periphery in WT cells.
2. Survival frequencies and GC and BIR repair events after DSB induction as in Fig 1H. Error bars represent the standard error (SEM) of at least three independent experiments. Asterisks indicate statistical differences for survival (****P < 0.001). See Table EV1 for statistical analysis of the GC and BIR repair events and Table EV2 for SEM values.

Recombination efficiency between subtelomeric and intrachromosomal loci is independent of telomere perinuclear anchoring

The nuclear position of subtelomeric and intrachromosomal loci differs, as telomeres are anchored to the nuclear periphery (Palladino et al, 1993). We previously showed that DSBs relocate to the nuclear periphery, and proposed that this change in position favours repair through a still unknown mechanism (Nagai et al, 2008). We hypothesized that relocalization of an intrachromosomal DSB to the nuclear periphery, close to telomeric foci, would favour its encounter with a subtelomeric donor, thus improving HR efficiency. Conversely, a subtelomeric DSB, restrained at the nuclear periphery, would less frequently encounter and recombine with an intrachromosomal donor.

To test this hypothesis, we evaluated recombination efficiencies in cells overexpressing the sir3A2Q variant, which triggers telomere clustering in the nuclear interior, independently of silent chromatin formation (Ruault et al, 2011 and Fig 2A). We first scored the localization of TEL6R in three concentric zones of equal area (Hediger et al, 2002 and Fig 4A). Although more than 60% of WT cells displayed TEL6R in the outermost zone, only 40% of the sir3A2Q‐overexpressing cells showed TEL6R at this position. This loss of peripheral localization was not simply a consequence of silencing disruption, since TEL6R is found in the outermost zone in 60% of sir3∆ cells (Fig 4B). Therefore, overexpression of sir3A2Q is indeed sufficient to drive TEL6R from the nuclear periphery to the nuclear interior.

Figure 4. Loss of telomere perinuclear anchoring has no effect on recombination efficiency.

1. Schematic representation of Lac^op‐tagged ARS609 on chromosome VI. Position of the GFP‐tagged locus was scored relative to the NE (Nup49‐mCherry). Ratios of distance from NE and diameter in focal plane are binned into three equal surfaces.
2. Position of ARS609 in WT, sir3∆ cells or cells overexpressing sir3A2Q or SIR3. Asterisks indicate statistical differences compared to WT measured by proportion test (*P < 0.05; ****P < 0.001).
3. Schematic representation of the assay showing DSB localization and telomere at the nuclear centre in cells overexpressing sir3A2Q.
4. Survival frequencies after DSB induction at the intrachromosomal locus LYS2 in WT, sir3∆‐ and sir3A2Q‐overexpressing cells as in Fig 1G and H.
5. Survival frequencies and GC and BIR repair events observed after DSB induction at a subtelomeric position in WT‐, sir3∆‐ and sir3A2Q‐overexpressing cells as in Fig 1G and H.

Data information: Error bars represent the standard error (SEM) of at least three independent experiments. See Table EV1 for statistical analysis of the GC and BIR repair events and Table EV2 for SEM values.

We then examined the recombination efficiency between intrachromosomal and subtelomeric sequences upon sir3A2Q overexpression, using the sir3∆ mutant to control for loss of silencing (Fig 4C). Overexpression of sir3A2Q or deletion of SIR3 had no significant effect on the repair of an intrachromosomal DSB at the LYS2 locus, with a TEL4R subtelomeric donor cassette (Fig 4D). Thus, subtelomeres remain efficient recombination donors for intrachromosomal DSBs, even when located at the nuclear interior. Further, a subtelomeric DSB induced at TEL6R was inefficiently repaired with an intrachromosomal cassette in sir3A2Q‐overexpressing cells, similar to WT cells (Fig 4E). We obtained similar results with DSBs induced at TEL4R (Fig EV2). Therefore, perinuclear anchoring of telomeres does not account for the asymmetry in repair efficiency between intrachromosomal and subtelomeric loci. Indeed, it is neither favouring the use of subtelomeric sequences as donors for intrachromosomal DSB nor preventing subtelomeric DSBs from recombination with intrachromosomal donors.

Figure EV2. Loss of telomere perinuclear anchoring has no effect on recombination efficiency at TEL4R.

• Survival frequencies and GC BIR repair events observed after induction of a DSB in WT‐ and sir3A2Q‐overexpressing cells as in Fig 1G and H. Error bars represent the standard error (SEM) for survival of at least three independent experiments.

Loss of DNA from the telomeric fragment correlates with limited gene conversion efficiency

We sought to decipher the molecular mechanisms that limit GC between subtelomeric DSBs and internal loci. We first compared the kinetics of RPA foci formation in individual cells, as a marker of ssDNA formation following I‐SceI DSB induction. The number of cells forming RPA foci was similar in subtelomere TEL6R and intrachromosomal LYS2 donor‐less strains, indicating that both loci convert to ssDNA with the same kinetics (Fig 5A and B).

Figure 5. Loss of DNA from the telomeric fragment limits gene conversion and favours BIR.

• A
  Representative image of Rfa1‐YFP foci in response to an I‐SceI‐induced DSB at TEL6R or at LYS2 in WT donor‐less cells. Scale bars are 2 μm.
• B
  Quantification of cells with Rfa1‐YFP foci after DSB induction in donor‐less strains. Mean of two independent experiments are shown. Error bars indicate standard deviation (SD).
• C
  Quantification of Rfa1‐YFP foci intensity using Q‐foci after induction of I‐SceI in donor‐less strains. Data plotted represent the pool of two independent experiments. Asterisks indicate statistical differences using a Mann–Whitney test (****P < 0.001).
• D
  Schematic representation of TEL6R and LYS2 loci with primer location indicated for DSB cleavage efficiency (blue) and flanking DNA measurement (red and grey). The telomeres are represented as wavy lines.
• E
  DNA levels measured at 0.9 kb from the I‐SceI cut site over time by qPCR normalized to DNA levels at the OGG1 locus in donor‐less strains. Error bars represent the standard error (SEM) of at least three independent experiments.
• F
  Schematic representation of the quantitative PCR assay to monitor I‐SceI‐induced DSB end resection. Red arrows show primers used for real‐time PCR. RE: restriction enzyme cut site.
• G
  Quantification of ssDNA among cut DNA at 0.9 kb from the DSB at each time point relative to t0. The mean values for three independent experiments are plotted and error bars show standard error (SEM).
• H–K
  Survival frequencies and distribution of the repair events after DSB induction in the indicated strains as in Fig 1G and H. Error bars represent the standard error (SEM) of at least three independent experiments. Asterisks indicate statistical differences for survival (****P < 0.001). See Table EV1 for statistical analysis of the GC and BIR repair events and Table EV2 for SEM values.

In contrast, we noticed that the intensity of RPA foci was significantly lower in cells with a subtelomeric DSB compared to cells with an intrachromosomal DSB, suggesting a lower amount of ssDNA (Fig 5C). This difference increased with time, reaching a maximum at 6 h after DSB induction. As the difference in RPA foci intensity could also reflect differential recruitment or retention to DSBs, we performed qPCR to quantify DNA amounts near TEL6R or LYS2, 0.9 kb away from I‐SceI cleavage sites, in donor‐less strains (Fig 5D and E). Disappearance of the sequence below 50% reflects loss of both strands of DNA, as previously reported (Zierhut & Diffley, 2008; Chen et al, 2013; Toledo et al, 2013), and irreversibly prevents recombination. Although the TEL6R centromere proximal fragment and the equivalent site at the LYS2 locus showed a similar rate of disappearance, the telomere proximal fragment of TEL6R showed a faster and more substantial disappearance (Fig 5E). Only 5% of the TEL6R telomere proximal fragment remained after 6 h, compared to 22% for the equally distant centromere proximal fragment (Fig 5E). A similar trend was observed following the induction of a DSB at TEL4R (Fig EV3A). We also monitored DSB resection and loss of ssDNA using a restriction digest/qPCR‐based method (Zierhut & Diffley, 2008). We designed primers flanking a HaeIII restriction site located 0.9 kb from the DSBs, and compared the yield of PCR products from HaeIII‐digested and mock‐treated genomic DNA. DSB resection results in HaeIII‐resistant ssDNA and PCR amplification (Fig 5F). We found that the PCR signal associated with resected, ssDNA decreased faster for the TEL6R subtelomeric flanking sequence than for the LYS2 flanking sequence, confirming that ssDNA was lost faster on the telomere proximal fragment (Fig 5G).

Figure EV3. Loss of DNA from the telomeric fragment limits gene conversion and favours BIR at TEL4R.

• A
  DNA levels measured at 0.9 kb on both side of the I‐SceI cut site at TEL4R in WT donor‐less cells. qPCR signals were normalized to DNA level at the OGG1 locus and corrected for differences in cleavage efficiency.
• B, C
  Survival frequencies and GC (dark) or BIR (light grey) repair events observed after induction of a DSB at TEL4R in the indicated strains as in Fig 1G and H.

Data information: Error bars represent the standard error (SEM) of at least three independent experiments. Asterisks indicate statistical differences for survival (****P < 0.001). See Table EV1 for statistical analysis of the GC and BIR repair events and Table EV2 for SEM values.

To determine whether this loss of ssDNA underlies the lower GC efficiency at subtelomeric DSBs, we evaluated whether moving the DSB away from the telomere would delay processing and increase recombination efficiency. Indeed, a DSB cassette inserted 10 kb away from TEL6R displayed a significantly increased GC efficiency compared to the DSB 1.4 kb from TEL6R (Fig 5H and Table EV1). This distance‐dependent effect of DSB position on repair efficiency suggests that extensive resection of the telomere proximal fragment limits GC. Together these data argue that, upon DSB resection, the TEL6R telomere proximal ssDNA is particularly unstable and reduces GC efficiency.

Exo1‐mediated resection limits gene conversion and favours BIR at subtelomeric DSBs

Next, we tested whether inhibiting resection affects recombination efficiency and outcome at TEL6R subtelomeric DSB by deleting the EXO1 gene, which encodes a nuclease involved in long‐range resection of DSB extremities (Gravel et al, 2008; Mimitou & Symington, 2008; Zhu et al, 2008). We induced a DSB at 1.4 kb from TEL6R in WT and exo1∆ strains, to be repaired with the intrachromosomal ura3‐1i donor. The exo1∆ strain displayed an almost threefold increase in survival compared to WT cells, upon induction of the DSB (Fig 5I). Interestingly, the survival rate of the exo1∆ strain reached the level of the reciprocal pair in WT cells (compare TEL6R‐I‐SceI ura3‐1i exo1∆ and URA3‐I‐SceI TEL6R; Figs 5I and 3B). This increase in survival reflected a fivefold increase in GC events, whereas BIR displayed an insignificant decrease (Fig 5I, Table EV1). Reintroduction of EXO1 on a high‐copy number (2μ) plasmid in exo1∆ cells restored low levels of GC (Fig 5I). Deletion of EXO1 also improved survival and increased GC at TEL4R (Fig EV3B). In contrast, limiting ssDNA formation via EXO1 deletion did not have a significant effect on GC efficiency at the LYS2 intrachromosomal DSB, as shown previously (Fig 5J; Lydeard et al, 2007; Zhu et al, 2008). Together, these results show that Exo1‐mediated resection causes lethality after subtelomeric DSB by limiting GC repair. Deletion of EXO1 in strains where the repair of DSBs at TEL6R occurs between subtelomeric loci also increased GC and significantly decreased BIR events (Fig 5K and Table EV1). GC was also increased at TEL4R in exo1∆ cells (Fig EV3C). Thus, Exo1‐dependent resection appears to be a critical event limiting GC efficiency and favouring BIR repair events in subtelomeric sequences.

Silent chromatin spreading over DSB sites counteracts Exo1p‐induced lethality

Subtelomeric sequences assemble in repressive chromatin, mediated by the SIR complex in budding yeast. We wondered if silent chromatin regulates repair of DSBs induced at TEL6R. Although we detected low levels of Sir3 binding at the insertion site of the recombination cassette on TEL6R by ChIP‐chip, we found that the URA3 gene positioned 1.4 kb from TEL6R was not silenced, reflected by a lack of growth on 5‐FOA plates (Fig 6A and B). Consistently, deletion of SIR3 had no impact on the repair of DSBs at TEL6R (Fig 6D).

Figure 6. Silent chromatin spreading over DSB site impairs resection.

• A
  Sir3‐binding at TEL6R in WT cells or in cells overexpressing SIR3 (oeSIR3). Binding is probed by ChIP‐chip using antibodies directed against untagged Sir3p. The mean of two independent biological replicates is shown and error bars correspond to the variation between duplicates. The red inserts mark the insertion site of the URA3 recombination cassette.
• B
  Telomeric silencing assay at TEL6R in WT cells, cells overexpressing SIR3 (oeSIR3) or sir3A2Q (oesir3A2Q). Increased growth on 5‐FOA plates reflects an increase in telomeric silencing.
• C–E
  Survival frequencies observed after induction of a DSB in the indicated strains as in Fig 1G and H. Survival frequencies for sir3∆, oesir3A2Q and WT from Figs 2C and 4E are plotted for comparison with strains overexpressing SIR3. Error bars represent the standard error (SEM) of at least three independent experiments. Asterisks indicate statistical differences for survival (**P < 0.01; ****P < 0.001). See Table EV1 for statistical analysis of the GC and BIR repair events and Table EV2 for SEM values.
• F, G
  DNA levels measured at 0.9 kb from the I‐SceI cut site over time by qPCR in donor‐less WT‐, exo1Δ‐ and SIR3‐overexpressing cells after induction of I‐SceI cleavage at TEL6R (F) or LYS2 (G). DNA levels were normalized to DNA levels at the OGG1 locus and corrected for differences in DSB cleavage efficiency (see Materials and Methods for details). Error bars represent the standard error (SEM) of three to five independent experiments.
• H, I
  Quantification of ssDNA among cut DNA at 0.9 kb from the DSB at each time point relative to t0 at TEL6R (H) and LYS2 (I). The mean values for three independent experiments are plotted, and error bars show standard error (SEM) of at least three independent experiments.

Next, we overexpressed SIR3 to determine whether heterochromatin formation might alter DSB repair. Overexpression of SIR3 increases the spread of silent chromatin and transcriptional repression in subtelomeric regions (Hecht et al, 1996; Strahl‐Bolsinger et al, 1997; Katan‐Khaykovich & Struhl, 2005). In our conditions, overexpressed SIR3 spread over 15 kb on TEL6R subtelomeric regions (Fig 6A) and repressed transcription, as cells bearing the URA3 gene inserted at this locus grew on 5‐FOA plates (Fig 6B). Importantly, we did not observe a change in the recombination efficiency between intrachromosomal loci in SIR3‐overexpressing cells compared to WT cells, suggesting that Sir3 spreading does not indirectly modulate the repair process by altering gene expression (Fig 6C). In contrast, SIR3 overexpression increased survival by favouring GC, and to a lesser extent BIR, at a DSB within TEL6R with the intrachromosomal ura3‐1i donor (Fig 6D; Table EV1). This effect was confirmed with a DSB at TEL4R (Fig 7A–C). These increases likely arise from silent chromatin spreading at the DSB site. Indeed, telomere clustering by sir3A2Q overexpression, thus in absence of silent chromatin spreading, did not affect BIR and had a significantly weaker effect on GC (Fig 6D).

Figure 7. Silent chromatin spreading over TEL4R DSB site impairs resection.

• A
  Sir3‐binding at TEL4R in WT cells or in cells overexpressing SIR3 (oeSIR3). Binding is probed by ChIP‐chip using antibodies directed against untagged Sir3. The mean of two independent biological replicates is shown, and error bars correspond to the variation between duplicates. The red inserts mark the insertion site of the URA3 recombination cassette.
• B
  Telomeric silencing assay at TEL4R in WT cells, cells overexpressing SIR3 (oeSIR3) or sir3A2Q (oesir3A2Q). Increased growth on 5‐FOA plates reflects an increase in telomeric silencing.
• C
  Survival frequencies observed after induction of a DSB in the indicated strains as in Fig 1G and H. Error bars represent the standard error (SEM) of at least three independent experiments. Asterisks indicate statistical differences for survival (****P < 0.001). See Table EV1 for statistical analysis of the GC and BIR repair events and Table EV2 for SEM values.
• D
  Schematic of TEL4R locus with primer location indicated for DSB cleavage efficiency (blue) and flanking DNA measurement (red and grey). The telomeres are represented as wavy lines.
• E, F
  DNA levels measured at 0.9 kb on the telomeric proximal (E) and on the centromeric proximal (F) sides of the I‐SceI cut site at TEL4R in donor‐less WT‐, exo1Δ‐ and SIR3‐overexpressing cells. DNA levels were normalized to DNA level at the OGG1 locus and corrected for differences in DSB cleavage efficiency (see Materials and Methods for details). Error bars represent the standard error (SEM) of three independent experiments.

Since limiting Exo1‐dependent resection favoured GC repair at subtelomeric DSB, we hypothesized that silent chromatin at DSB could regulate resection. In this case, EXO1 overexpression should suppress the effects of SIR3 overexpression by restoring DSB resection. As predicted, overexpression of EXO1 along with SIR3 partially restored HR between TEL6R or TEL4R and ura3‐1i to WT levels (Figs 6E and 7C). Thus, increased resection activity counteracts the effects of Sir3‐mediated silent chromatin. Collectively, these data suggest that silent chromatin inhibits resection.

Silent chromatin spreading over DSB sites limits resection

To directly determine how silent chromatin spreading at the DSB site affects resection, we compared DSB processing in SIR3‐overexpressing and WT cells. We monitored DNA levels on the telomere proximal side of TEL6R DSBs or around LYS2 DSBs in donor‐less strains by qPCR. Since DSB induction was slightly delayed at silent TEL6R (Fig EV4), we corrected DSB‐flanking DNA amounts for differences in I‐SceI cleavage efficiency (See experimental procedures for calculation details). The fraction of cleaved DNA measured for the TEL6R subtelomeric fragment was twofold to 10‐fold higher in SIR3‐overexpressing than in WT cells 2 and 6 h after cleavage induction, respectively (Fig 6F). ssDNA, measured by the restriction digest/qPCR‐based method, was also higher in SIR3‐overexpressing cells than in WT cells and comparable to ssDNA detected in exo1∆ cells (Fig 6H). As expected, SIR3 overexpression did not impact resection kinetics at the LYS2 control locus, ruling out indirect effects (Fig 6G and I). The retention of the subtelomeric fragment upon SIR3 overexpression strongly suggests that resection is impaired at DSBs in silent chromatin. Accordingly, limiting resection by deleting EXO1 also led to the retention of the TEL6R telomeric proximal fragment, although to a lesser extent than heterochromatin spreading (Fig 6F). Similarly increasing silent chromatin spreading or deleting EXO1 protects the telomeric fragment upon DSB induction at TEL4R, and increases GC efficiency with the ura3‐1i allele (Fig 7). We conclude that increasing silent chromatin in subtelomeric regions limits resection at subtelomeric DSBs, avoiding loss of genetic information.

Figure EV4. Delay in DSB cleavage efficiency at silent chromatin DSB site.

• DSB cleavage efficiency at the two loci in WT‐, exo1∆‐ and SIR3‐overexpressing cells measured as in Fig 1C. Error bars represent the standard deviation (SD) of at least three independent experiments.

Discussion

Here, we show that gene conversion efficiency is limited by the physical distance separating homologous sequences, the chromosomal location of the DSB, DSB resection rate and chromatin status.

DSB proximity to the telomere has a strong impact on HR efficiency, independently of perinuclear anchoring and spatial distance to the donor sequence

We show that the active grouping of two homologous sequences improves recombination efficiency, demonstrating directly that the physical distance between two homologous sequences in the nucleus is a limiting factor for homologous recombination. This is compatible with a stochastic homology search, likely governed by random walk‐like kinetics, that explores first the immediate surrounding and takes more time to reach distant sites. In that instance, searching for distant homologous DNA might be facilitated by increased global chromatin and DSB dynamics induced in response to DNA damage (Dion et al, 2012; Miné‐Hattab & Rothstein, 2012; Strecker et al, 2016). That physical distance negatively correlates with the capacity of two sequences to recombine (Agmon et al, 2013; Lee et al, 2016 and our data) suggests that either a time limit and/or a spatial restriction reduces the likelihood of some sequences to meet. The time limit could be defined by DSB processing that is required to unmask homologous sequences but that ultimately ends with loss of the DSB proximal sequences, and spatial restriction could be imposed by anchoring of telomeres to the nuclear envelope. Finally, the existence of a recombination barrier impairing HR between subtelomeric and intrachromosomal loci has been proposed to explain the low level of spontaneous HR observed between these sites (Pryde & Louis, 1997; Marvin et al, 2009a,b). However, we show that intrachromosomal DSBs are efficiently repaired with subtelomeric sequences, demonstrating that a barrier does not impair contact and recombination between intrachromosomal and subtelomeric sequences. Interestingly, the low level of HR between an induced subtelomeric DSB and an intrachromosomal donor reproduced results from spontaneous recombination assays (Pryde & Louis, 1997; Marvin et al, 2009a,b). Therefore, the previously suggested barrier may simply result from a higher level of spontaneous damages at subtelomeric loci (possibly during replication) and from the lower ability to repair these damages through HR with intrachromosomal templates. The fragility of subtelomeric regions may be conserved in humans, where subtelomeres exhibit high levels of sister chromatid exchanges (Cornforth & Eberle, 2001).

Telomere tethering to the nuclear periphery was also previously proposed as essential for BIR repair of subtelomeric DSBs (Therizols et al, 2010; Chung et al, 2015). In contrast, we observed that displacing the DSB‐bearing site to the nuclear interior, by overexpressing the silencing dead sir3A2Q allele, did not decrease BIR nor GC repair of a subtelomeric DSB with an intrachromosomal donor sharing perfect homology. BIR using a subtelomeric recombination donor also remained unchanged in this context, and increased when displacing telomeres in the nuclear interior in the context of increased silent chromatin formation by overexpressing Sir3. Thus, in our conditions the initial perinuclear position of telomeres does not favour BIR events, in contrast with previous studies (Therizols et al, 2010; Chung et al, 2015). In these studies, viable cells result from rare and complex events due to the absence of homologous sequence on both sides of the break. These possibly require events favoured at the nuclear periphery, which seems dispensable for efficient HR with DSB‐flanking homologies. However, in this set‐up, limiting resection by SAE2 deletion also favours BIR at subtelomeric DSBs (Chung et al, 2015). Finally, telomere anchoring could also prevent HR events with internal sites by sequestering telomeres at the nuclear periphery. However, releasing telomeres from the nuclear membrane by deleting SIR4 (Hediger et al, 2002; Taddei et al, 2004) did not affect HR efficiency or outcome (Fig EV5).

Figure EV5. Telomere anchoring is not limiting for recombination.

1. Survival frequencies and GC (dark) or BIR (light grey) repair events observed after induction of a DSB at a subtelomeric position in WT and sir4∆ cells as in Fig 1H.
2. Survival frequencies observed after induction of a DSB at an intrachromosomal position in WT and sir4∆ cells as in Fig 1H.

Data information: Error bars represent the standard error (SEM) of at least three independent experiments. See Table EV1 for statistical analysis of the GC and BIR repair events and Table EV2 for SEM values.

In conclusion, neither the spatial constraint nor perinuclear anchoring explains the low repair of subtelomeric DSBs by intrachromosomal loci.

Subtelomeric DSBs are efficiently repaired by BIR despite the presence of homologous sequences on both sides of the break

The high frequency of BIR at a subtelomeric DSB flanked by homologous sequences has escaped the attention of previous studies that did not analyse the molecular nature of the repair events (Agmon et al, 2013). Our analysis reveals that, upon subtelomeric DSBs, BIR‐mediated events account for 50–75% of the survivors depending on the configuration tested, despite the presence of a two‐ended homologous template in WT cells. Favouring BIR at subtelomeric DSBs has implications for genome stability and evolution. On the one hand, BIR heals damaged ends with telomeric sequences ensuring chromosome stability, but on the other hand, BIR causes loss of heterozygosity and is highly mutagenic (Deem et al, 2008). Yet, BIR that results in duplication and exchange between subtelomeres could also be beneficial and serve as a nursery for new genes where diversity evolves faster than in single copy genomic regions (Linardopoulou et al, 2005; Brown et al, 2010).

As BIR preferentially engages when only one end of the DSB shares homology with the donor site, this suggests that the telomeric fragment freed by cleavage of the subtelomeric cassette is not available for GC. Our molecular analysis shows that the telomeric proximal fragment rapidly disappears upon resection at TEL6R and to a lesser extent at TEL4R. This alone would favour BIR at the expense of GC. Interestingly, limiting the loss of the telomeric fragment increased HR to levels comparable to those of the reciprocal strain. However, we also observe that a DSB 10 kb from telomeres at TEL4R recombines 10‐fold less than a DSB at TEL6R only 1.4 kb from telomeres, with the same intrachromosomal donor. A likely explanation is that the numerous repeated sequences (PAU, Y’ and core X sequences) at TEL4R interfere with HR repair at URA3 by engaging subtelomeric DSBs into competing recombination events, leading to unviable outcomes, as shown for multi‐genes family or TY elements (Jain et al, 2016). These events are likely to accelerate the loss of the telomeric fragment by generating 3′end flap structures that could be cleaved by the Rad1/Rad10 endonuclease activity. Limiting resection would limit unmasking of the subtelomeric repeated elements and, as a consequence, competing HR events. Indeed, our molecular analysis shows that EXO1 deletion limits the loss of telomeric proximal sequences at both TEL6R and TEL4R. These results argue that resection‐mediated loss of the telomeric proximal DSB extremity limits GC repair of subtelomeric DSBs but favours BIR.

Resection limits HR efficiency at subtelomeric double‐stranded breaks

The impact of resection on HR efficiency may not be limited to subtelomeric loci. Indeed, it was recently shown that resection limits recombinational repair at some intrachromosomal sites (Lee et al, 2016). This could occur following the loss of homology at the 3′ end. Indeed, we show that DNA amount decreases below 50% 0.9 kb away from the DSB accordingly with previous studies showing that 3′ ssDNA overhangs are unstable and their loss likely coupled to 5′ strand resection (Zierhut & Diffley, 2008; Chen et al, 2013; Toledo et al, 2013). Loss of genetic information on the 3′ overhang concomitantly with 5′ resection could explain how limiting 5′ strand resection at an intrachromosomal DSB rescues HR efficiency. This model also accounts for improved recombination by increasing the size of sequence homology (Lee et al, 2016). Thus, while the HR process requires sequence unmasking and formation of a ssDNA recombination filament, resection can also be limiting when too extensive, notably if it causes loss of homologous sequences at the 3′ end before finding the homologous donor sequence.

We show that subtelomeric DSBs are particularly sensitive to Exo1‐mediated resection. Indeed, we observe a strong recombination increase associated with resection delay at two subtelomeric DSBs in the exo1∆ strain, and no significant effect at the LYS2 intrachromosomal position. Thus, resection possibly proceeds differently depending on the DSB‐flanking sequences or chromatin contexts.

Silent chromatin spreading improves subtelomeric DSB repair by limiting resection

We report here, for the first time, the molecular impact of repressive chromatin spreading at a DSB in yeast. Heterochromatic DSBs show improved HR, correlating with decreased resection of sequences surrounding the DSB. This result supports a regulatory role for chromatin structure on resection.

Although in vitro experiments have shown that Sir3 loading on the donor molecule impairs joint molecule formation (Sinha et al, 2009), its impact on the broken molecule has not been addressed. We show here that SIR complex spreading at DSB sites favours HR in vivo; it increases both GC and BIR repair events at subtelomeric DSBs. The mechanism through which repressive chromatin favours BIR at subtelomeres remains to be deciphered. However, our data clearly demonstrate that increased GC arises from a resection defect in repressive chromatin, that is alleviated by overexpressing EXO1. Because EXO1 overexpression is sufficient to bypass the requirement for a functional MRX complex in the resection of a HO‐induced DSB (Tsubouchi & Ogawa, 2000; Moreau et al, 2001; Lewis et al, 2002), Sir3 spreading could also restrain resection mediated by other nucleases. How Sir3‐mediated heterochromatin limits resection remains to be explored. It could act as a physical block to delay the progression of one or several nucleases mediating resection. Alternatively, histone modifications associated with repressive chromatin could, directly or indirectly, impair nuclease recruitment or activation. Repressive chromatin also seems to repress resection to some extent in mammalian cells (Lemaître et al, 2014; Tsouroula et al, 2016). In contrast, in Drosophila, ATRIP/TopBP1 foci form earlier and appear brighter in heterochromatin, suggesting either an increased resection or persistent binding (Chiolo et al, 2011). These results highlight the impact of chromatin structure on the early response to DSB and HR; however, the molecular details of this regulation are still unclear.

In conclusion, our work directly demonstrates that spatial distance limits HR efficiency, but chromosomal location of the DSB has an even more prominent effect. Altogether, our data support a model in which HR efficiency reflects the rates of the resected DSB finding its recombination partner, determined by the physical distance between the DSB and the donor sequence, versus losing its homologous sequence, determined by the resection rate. We reveal that not all genomic loci are equal in this race, as subtelomeric DSBs are more sensitive to exonuclease activities. The vulnerability of subtelomeric DSBs might be due to their proximity to the ends of chromosomes, and/or to repeated elements that, once unmasked by resection, could commit the broken chromosome to lethal events. Importantly, repressive chromatin can limit resection and thus the loss of genetic information at subtelomeric DSB. Interestingly, the silencing factor Sir3 can promote GC between telomeres by two means: reducing the distance between telomeres, by promoting telomere clustering (Ruault et al, 2011) and preventing the loss of the telomeric fragment, by inhibiting extensive resection. These functions of Sir3 could be specifically relevant in yeast quiescent cells, where telomeres form a hypercluster in the nuclear centre, an organization that contributes to the long‐term viability in this state (Guidi et al, 2015). This connection of heterochromatin with DSB processing and repair may yield new insights into how cells maintain genome stability to avoid tumorigenesis.

Materials and Methods

Plasmids

Plasmid pAT274 was constructed by inserting the URA3 gene amplified from pRS316 with primers containing restriction sites suitable for cloning, and the digested PCR product was ligated into pUG6 SpeI and SacII sites (from EUROSCARF). pAT275 was made by directed mutagenesis from pAT274 using the Pfu Turbo polymerase from Stratagene with primers inserting the 30 bp I‐SceI recognition sequence. The ura3‐1 mutation was introduced into pAT274 with primers changing the GGA to a GAA codon at position 701 giving the plasmid pAT276. pAT277 was made by introducing the C207T mutation into the pAT276 with primers changing the CCA to a CTA codon leading to the elimination of the NcoI restriction site.

To construct the 2μ plasmid carrying the EXO1 gene (pKD232), the EXO1 gene was amplified from pHL546‐EXO1 (Mantiero et al, 2007) and inserted into pRS424 digested by PciI and PsiI by SLIC (Li & Elledge, 2007).

Yeast strains

All strains used in this study are isogenic to W303 (MATa RAD5 ADE2 leu2‐3,112 his3‐11,15 trp1‐1 ura3‐1) and are listed in Table EV3. Strains were constructed by integrating a 1.6‐kb lox‐KanMx‐lox‐ura3‐I‐SceI recipient cassette from pAT275 at different loci in the yeast genome. All strains are deleted for the HML locus to avoid indirect recombination effects caused by pseudo‐diploidization following the derepression of the cryptic mating type loci in strains with SIR3 deletion or sir3A2Q overexpression (Aström et al, 1999; Lee et al, 1999). For recombination assays, a lox‐KanMx‐lox‐ura3‐1 homologous sequence from pAT277 or pAT276 was inserted as a donor cassette in different regions. Loss of the KanMx marker flanked by loxP sequences was selected on appropriate medium after transformation with pSH62 (pGAL1‐CRE‐HIS3 from EUROSCARF) that was further eliminated by successive restreak on non‐selective medium.

The cassettes were inserted at the following coordinates: 268813 (TELVI‐R, subtelomeric), 470815 (+3108 LYS2, intrachromosomal), 1522383 (TELIV‐R, subtelomeric), 436650 (TELIX‐R, subtelomeric), 116167 (URA3, intrachromosomal). All insertions were verified by PCR, and all PCR primers are listed in Table EV4.

Gene deletions (sir3, ura3, exo1, pol32, dnl4) and insertions of a strong constitutive promoter (pGPD) were performed by PCR‐based gene targeting (Longtine et al, 1998).

For DSB induction or EXO1 overexpression, cells were transformed with pKD89 (pRS413‐pGAL1‐I‐SceI) or pKD232 (pHL546 EXO1 TRP1) and selected on glucose‐containing synthetic medium lacking histidine or tryptophan, respectively.

For DSB end resection and chromatin immunoprecipitation, the I‐SceI gene was introduced in the yeast genome by transformation of the cells with Pbp2 (pRS404‐pGAL1‐I‐SceI; gift from S. Marcand) digested by PmlI to target it to TRP1.

For the silencing test, a wild‐type URA3 gene was amplified from pAT274 and integrated at the same position than the recombination cassettes in subtelomeric regions.

Media and growth conditions

Yeast strains were grown in rich medium (yeast extract–peptone–dextrose, YPD) or synthetic complete (SC) medium lacking the appropriate amino acid at 30°C. Rich or synthetic medium containing 2% lactate, 3% glycerol, 0.05% glucose and lacking the appropriate amino acids was used to grow the cells overnight prior the induction of I‐SceI by plating onto 2% galactose plates or addition of 2% galactose to liquid culture.

Silencing test

For telomeric silencing assay, strains were grown overnight in YPGAL and then plated in fivefold serial dilutions starting at OD_600 nm = 1 (corresponding to 10⁷ cells/ml) on YPGAL medium and galactose synthetic complete (SC) medium containing 0.1% 5‐fluoroorotic acid (5‐FOA).

Recombination efficiency measurement

Each strain was freshly transformed with pKD89, and a single transformant was grown overnight in 2 ml of SLGg lacking histidine to select for the plasmid. The day after each culture was appropriately diluted and plated on SC‐HIS plates supplemented with 2% glucose (Glc‐HIS) to repress I‐SceI expression or 2% galactose (Gal‐HIS) to induce I‐SceI expression. Colonies were counted after 2–3 days of incubation at 30°C. From Gal‐HIS plates, 48 isolated recombinants were analysed by PCR as exemplified in Fig 3A and B. For each strain, at least three independent experiments were performed with the corresponding controls.

Pulsed‐field gel electrophoresis

Yeast DNA embedded in agarose plugs was prepared as follows: cells were cultured in rich medium overnight and about 15 OD₆₀₀ of cells were washed twice in 1 ml of 50 mM EDTA, 10 mM Tris (pH 7.5) and resuspended in 150 μl of 50 mM EDTA, 10 mM Tris (pH7.5). The suspension was quickly warmed to 42°C with 0.6 μl of Zymolyase (20 mg/ml), mixed with 250 μl of pre‐warmed 1% agarose LMP (Low Melting Point) and distributed into 80 μl wells placed into a cool surface. The plugs were extruded and incubated for 24 h at 37°C in 1.4 ml of 500 mM EDTA, 10 mM Tris–HCl (pH7.5) followed by 24 h at 55°C in 1.25 ml of 500 mM EDTA, 10 mM Tris (pH8), 1% N‐Laurylsarcosyl and 0.4 mg/ml Proteinase K. Plugs were washed for 1 h three times in 1.5 ml of 50 mM EDTA, 10 mM Tris (pH7.5). Pulsed‐field electrophoresis was carried out in a 0.9% agarose gel in 0.5× TBE at 14°C with a CHEF DRII from Bio‐Rad for 22 h (initial time = 10 s, final time = 25 s). After electrophoresis, the gel was stained with EtBr and photographed.

Monitoring of DSB‐flanking DNA and resection by real‐time PCR

Yeast cells were grown in 2 ml of YPD overnight. Cultures were then diluted in YPLGg and grown OD₆₀₀ = 0.3–0.8. I‐SceI was induced by addition of galactose to a final concentration of 2%. Cells samples were collected before and after induction at different time points. DNA were extracted and purified on EconoSpin mini spin columns (Epoch Life Science). For total DNA measurements, quantitative PCRs were performed using primers located 0.9 kb from the I‐SceI cut site and primers flanking the I‐SceI restriction site. A control primer pair was used to amplify a region of the OGG1 locus. For ssDNA assessment, gDNA were digested by HaeIII and primers flanking the HaeIII site 0.9 kb away from the I‐SceI cut site and a control primer pair amplifying a region of OGG1 locus devoid of HaeIII site were used (see Table EV4 for primer details). Quantitative PCRs were performed using Power SYBR Green Master Mix (Applied Biosystems) with the 7900HT Fast Real‐Time PCR System and corresponding software (Applied Biosystems). To correct for differences in DSB cleavage efficiency, the fraction of uncut DNA (Fu) was subtracted from fraction of total DNA at 0.9 kb (Ft) at each time point and normalized to the fraction of cleaved DNA (Fc). Thus, cleaved remaining DNA at 0.9 kb = (Ft‐Fu)/Fc. The per cent of ssDNA at 0.9 kb among I‐SceI cut DNA was given by x = [100/(2^∆Ct−1)]/Fc, where ∆Ct = C_t,digestion ‐ C_t0,mock and Fc is the fraction cut by I‐SceI determined by qPCR using primers flanking the I‐SceI cut site.

Sir3 chromatin immunoprecipitation

Yeast cells grown overnight after inoculation in YP Raffinose at 0.01 OD₆₀₀. After dilution in 50 ml of YP Raffinose 3% at an OD₆₀₀ = 0.2, cells were allowed to grow for an additional 2 h, before addition of galactose to a final concentration of 2%. ChIP was carried out as previously described with minor modifications (Ruault et al, 2011). Samples were incubated with 1 μl of polyclonal antibody anti‐Sir3 (Agro‐bio), and 50 μl of Magnetic Dynabeads Protein A (NEB) was added to each sample. After washes, elution of the proteins and reversal of crosslinks, samples were treated with RNase A (50 μg/ml) followed by purification of the DNA on QIAquick PCR purification columns (Qiagen).

ChIP‐chip preparation and hybridization

Samples used for ChIP‐chip were analysed by qPCR prior to microarray hybridization. For microarray hybridization, four of five of the immunoprecipitated DNA and the DNA from the input were ethanol precipitated and resuspended in 10 μl of water (Gibco). Purified material was amplified, incorporating amino‐allyl‐dUTP. The size of the amplified fragments (~500 bp) was assessed by gel electrophoresis. For each sample, 1.5 μg of amplified DNA was coupled either with Cy5 (immunoprecipitated sample) or Cy3 (input sample) and hybridized on 44k yeast whole genome tiling array (Agilent) as described in Borde et al (2009).

Microarray data acquisition, analysis and visualization

The microarray was imaged using an Agilent DNA microarray scanner and quantified using GenePix Pro6.1 as described in Borde et al (2009). Data were normalized so that the mean of the ratio of the median fluorescence of the probes corresponding to mitochondrial DNA is equal to 1. Data visualization was done using the R package ggplot2. All scripts used are available upon request. The microarray data from this publication have been deposited to the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) and assigned the identifier GSE99117.

Statistical analyses

To compare survival of two different strains, we used a χ² test. Asterisks on graphs indicate statistical differences (*P < 0.05; **P < 0.01; ***P < 0.005; ****P < 0.001).

To determine change in repair events, we applied a proportional analysis with a confidence limit of 95%. All p‐values are listed in Table EV1.

To compare Rfa1 foci intensity, a Mann–Whitney test was applied. Asterisks on graphs indicate statistical differences (*P < 0.05; **P < 0.01; ***P < 0.005; ****P < 0.001).

Microscopy

Live cell images were acquired using a wide‐field microscope based on an inverted microscope (Leica DMI‐6000B) equipped with Adaptive Focus Control to eliminate Z drift, a 100×/1.4 NA immersion objective with a Prior NanoScanZ Nanopositioning Piezo Z Stage System, a CMOS camera (ORCA‐Flash4.0; Hamamatsu) and a solid‐state light source (SpectraX, Lumencore). The system is piloted by MetaMorph software (Molecular Device).

For GFP‐mCherry two‐colour images, 21 focal steps of 0.25 μm were acquired sequentially for GFP and mRFP with an exposure time of 200 ms using solid‐state 475‐ and 575‐nm diodes and appropriate filters (GFP‐mRFP filter; excitation: double BP, 450–490/550–590 nm and dichroic double BP 500–550/600–665 nm; Chroma Technology Corp.). Three‐dimensional data sets were deconvolved using the blind deconvolution algorithm of AutoQuant (Media Cybernetics, Inc.) with the Point Spread Function appropriate to our microscope at each emission wavelength. Further processing was done using ImageJ software (National Institutes of Health).

YFP images were acquired at indicated time points before and after induction; 21 focal steps of 0.25 μm were acquired with an exposure time of 300 ms using a solid‐state 500‐nm diode and a YFP filter (excitation 470–510 nm and dichroic 495 nm; Chroma Technology Corp.) Quantification of foci intensity has been performed as previously described using Q‐foci (Ruault et al, 2011). All the images shown are a z projection of z‐stack images.

Author contributions

AB, CB and HB generated strains and performed recombination assays and molecular analysis. AB, MR and KD performed microscopy experiments and image analyses. CB and HB performed and analysed the measure of dsDNA and ss DNA levels and PFGE experiments. AH performed and analysed SIR3 ChIP on ChIP experiments. AA and CB constructed plasmids. KD, AT and AB designed and interpreted experiments, drafted the figures and wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Table EV3

Table EV4

Review Process File

The EMBO Journal (2017) 36: 2609–2625