Positive regulation of meiotic DNA double-strand break formation by activation of the DNA damage checkpoint kinase Mec1(ATR)

During meiosis, formation and repair of programmed DNA double-strand breaks (DSBs) create genetic exchange between homologous chromosomes—a process that is critical for reductional meiotic chromosome segregation and the production of genetically diverse sexually reproducing populations. Meiotic DSB formation is a complex process, requiring numerous proteins, of which Spo11 is the evolutionarily conserved catalytic subunit. Precisely how Spo11 and its accessory proteins function or are regulated is unclear. Here, we use Saccharomyces cerevisiae to reveal that meiotic DSB formation is modulated by the Mec1(ATR) branch of the DNA damage signalling cascade, promoting DSB formation when Spo11-mediated catalysis is compromised. Activation of the positive feedback pathway correlates with the formation of single-stranded DNA (ssDNA) recombination intermediates and activation of the downstream kinase, Mek1. We show that the requirement for checkpoint activation can be rescued by prolonging meiotic prophase by deleting the NDT80 transcription factor, and that even transient prophase arrest caused by Ndt80 depletion is sufficient to restore meiotic spore viability in checkpoint mutants. Our observations are unexpected given recent reports that the complementary kinase pathway Tel1(ATM) acts to inhibit DSB formation. We propose that such antagonistic regulation of DSB formation by Mec1 and Tel1 creates a regulatory mechanism, where the absolute frequency of DSBs is maintained at a level optimal for genetic exchange and efficient chromosome segregation.


Summary
During meiosis, formation and repair of programmed DNA double-strand breaks (DSBs) create genetic exchange between homologous chromosomes-a process that is critical for reductional meiotic chromosome segregation and the production of genetically diverse sexually reproducing populations. Meiotic DSB formation is a complex process, requiring numerous proteins, of which Spo11 is the evolutionarily conserved catalytic subunit. Precisely how Spo11 and its accessory proteins function or are regulated is unclear. Here, we use Saccharomyces cerevisiae to reveal that meiotic DSB formation is modulated by the Mec1(ATR) branch of the DNA damage signalling cascade, promoting DSB formation when Spo11-mediated catalysis is compromised. Activation of the positive feedback pathway correlates with the formation of single-stranded DNA (ssDNA) recombination intermediates and activation of the downstream kinase, Mek1. We show that the requirement for checkpoint activation can be rescued by prolonging meiotic prophase by deleting the NDT80 transcription factor, and that even transient prophase arrest caused by Ndt80 depletion is sufficient to restore meiotic spore viability in checkpoint mutants. Our observations are unexpected given recent reports that the complementary kinase pathway Tel1(ATM) acts to inhibit DSB formation. We propose that such antagonistic regulation of DSB formation by Mec1 and Tel1 creates a regulatory mechanism, where the absolute frequency of DSBs is maintained at a level optimal for genetic exchange and efficient chromosome segregation.

Introduction
To prevent the chromosome copy number doubling during each reproductive cycle, sexually reproducing organisms create gametes containing half the chromosome copy number of their parents. Halving the chromosome copy number occurs during meiosis-a specialized pair of sequential nuclear divisions that first segregates homologous chromosomes, then sister chromatids. In many organisms, accurate segregation of homologous chromosomes during meiosis I requires the formation and repair of numerous DNA double-strand breaks (DSBs) distributed across the genome. DSB repair uses homologous recombination to drive interaction and genetic exchange between homologous chromosomes, which facilitates accurate chromosome disjunction at anaphase I. Much of our detailed understanding of the mechanisms of meiotic recombination comes from studies performed in the budding yeast Saccharomyces cerevisiae [1].
In S. cerevisiae, meiotic DSB formation requires the function of 10 proteins, of which Spo11 is the catalytic subunit. The precise function of the nine accessory proteins is unclear, as are the detailed mechanisms that govern the frequency and distribution of Spo11-DSB formation across the genome. Spo11 shares similarity to a type-II topoisomerase (Top6A) from archaea and initiates DSB formation via a transesterification reaction that coordinately severs the DNA backbone on both strands, leaving Spo11 monomers covalently attached to the 5 0 ends of the DSB (reviewed in [2]).
Removing the function of checkpoint proteins causes entry into the meiotic divisions of mutants that accumulate unrepaired DSBs, leading to cell death [24]. Interestingly, budding yeast cells, which possess mutations in the checkpoint proteins, actually undergo a delayed DSB-dependent meiosis I division, suggesting that checkpoint proteins are directly involved in the DNA repair and recombination process itself [40]. Indeed, checkpoint mutants have increased rates of ectopic (also called, non-allelic) recombination [41], whereas mutation of Mek1 abolishes repair partner choice entirely, with most or all DSBs now competent to repair using the sister chromatid (rather than the homologous chromosome) as repair template [42 -44].
Recently, it has been shown that mutation of another PIKK kinase, ATM, results in increased DSB formation in mammals and in flies [45,46], and both Mec1(ATR) and Tel1(ATM) are necessary in budding yeast to reduce the likelihood that more than one DSB arises per chromatid quartet at a given chromosomal locus [47]. Here, we reveal a novel role for the Mec1 pathway in promoting DSB formation when the catalytic activity of Spo11 is compromised. Our results suggest that both positive and negative feedback loops influence the rate of DSB formation, creating a regulated system with optimal levels of recombination initiated.

Results
3.1. Loss of Rad24/Mec1(ATR) checkpoint activity causes a synergistic reduction in double-strand break signals in hypomorphic spo11-HA strains In each budding yeast meiotic cell, accurate reductional segregation of homologous chromosomes at the first nuclear division involves the formation and repair of approximately 160 Spo11-DSBs [48,49]. The first step in the repair of Spo11-DSBs is their nucleolytic processing by Mre11 and Sae2, creating covalent Spo11-oligonucleotide complexes (Spo11-oligos; figure 1a [11,18,48]). Spo11-oligos are detected by immunopurifying Spo11 and radiolabelling the ends of the covalently attached DNA [50]. The absolute abundance of Spo11-oligo complexes can be used to estimate total DSB formation. Spo11-DSBs trigger transient activation of the DNA damage checkpoint response machinery (DDR [24])-a state that persists in recombination-defective mutants such as dmc1D [22]. During our studies, we were intrigued to discover that while the abundance of Spo11-oligo complexes was increased in dmc1D strains (where DSB repair is abolished, and DSBs accumulate [22]), such an increase was absent when the DDR was inactivated (rad24D; figure 1b,c). Our observations suggest a link between checkpoint activity and Spo11-oligo abundance. However, one limitation of our method is that it makes use of a SPO11-HA3-His6 allele (henceforth referred to as spo11-HA) to enrich for Spo11-oligo complexes. This C-terminal epitope tag is reported to moderately reduce the catalytic activity of Spo11 (DSB formation drops by 11 -50% compared with the untagged wild-type [51]). Because the reduction in Spo11-oligo abundance in rad24D dmc1D could be due to fewer total DSBs or to their faster turnover-perhaps due to changes in meiotic cell cycle progression caused by rad24D-we were interested to determine the abundance of DSBs in our panel of strains.
We began by assessing DSB formation at the wellcharacterized HIS4::LEU2 reporter locus on chromosome III [52] in synchronized meiotic cultures, where Spo11 was either HA-tagged or untagged. At HIS4::LEU2, there are two prominent DSB sites flanking the LEU2 insertion (figure 2a). In the untagged strain (SPO11 þ ), DSB signals peaked at approximately 6% of the total lane signal (figure 2c), with more of the DSBs arising at site 1 (figure 2d). Contrary to our expectation, total DSB signals were not reduced by the Spo11-HA tag (figure 2c). However, the distribution was altered such that now a similar proportion of DSBs was detected at site 1 and site 2 (figure 2d; see also [52]).
Accurately measuring DSB abundance from a culture of wild-type cells is complicated by overlapping kinetics of DSB formation and repair. To work around this issue, we remeasured the DSB frequency in strains similar to those we had used for the Spo11-oligo analysis, where the Dmc1 repair protein was mutated, causing DSBs to accumulate (figure 2e).
In both the tagged and untagged Spo11 strains, DSBs accumulated to a similar extent, reaching approximately 15% and approximately 18% of total lane signal, respectively (figure 2f ). As before, the distribution of events was altered by the HA-tag, such that the fractions of DSBs forming at site 1 and site 2 were now similar (figure 2d). A second strong hotspot region, ARE1, showed the same trend of relatively unchanged DSB frequency when the spo11-HA allele was present (see electronic supplementary material, figure S1a,b).
Next, we assessed DSB formation in the checkpoint-defective rad24D strain (figure 2b-f). Reduced MEC1 checkpoint activity (as caused by mutations such as rad24D) results in rapid exonucleolytic hyper-resection of the DSB ends creating extensive 3 0 -ending ssDNA tails that cause DSB molecules to migrate much more rapidly and heterogeneously when separated by agarose electrophoresis [40]. At HIS4::LEU2, despite the signal arising in rad24D being more diffuse, DSBs remained readily detectable above lane background signals, peaking at 9% (figure 2c), slightly greater than wild-type cells, possibly due to a slight delay in DSB repair causing DSB signals to transiently accumulate to higher steady-state levels. In rad24D dmc1D cells, despite extensive hyper-resection, DSBs accumulated and persisted similar to RAD24 dmc1D, reaching a maximum level of approximately 16-20% of the total DNA (figure 2e,f). Notably, in the rad24D and rad24D dmc1D backgrounds, barely any signal was observed at DSB site 2, resulting in a significantly altered ratio compared with RAD24 þ cells (figure 2d). In addition, we observed increases in the frequency of ectopic recombination products in rad24D strains (see figure 2b,e; electronic supplementary material, figure S2a-c), as has been reported by others [41].
In contrast to the situation in RAD24 þ cells, the combination of spo11-HA and rad24D resulted in a substantial reduction in DSB signal (figure 2b-f). DSBs were reduced to 50% of the rad24D level in rad24D spo11-HA cells (figure 2b,c), and to less than 25% of the rad24D dmc1D level in the rad24D dmc1D spo11-HA strain, where DSB signals were barely detected despite being expected to accumulate (figure 2e,f). These reductions in DSB abundance correlated with reduced abundance of ectopic recombination products (see electronic supplementary material, figure S2). Similar effects were observed at the ARE1 locus (see electronic supplementary material, figure S1b). In rad24D dmc1D cells, despite extensive hyper-resection, DSBs accumulated to levels similar to both the dmc1D and dmc1D spo11-HA controls (see electronic supplementary material, figure S1b). By contrast, DSBs were undetectable in the dmc1D rad24D spo11-HA strain (see electronic supplementary material, figure S1b). This surprising reduction in DSB abundance was reproduced in two independently generated isolates of the same rad24D dmc1D spo11-HA genotype, indicating that the phenotype was rsob.royalsocietypublishing.org Open Biol 3: 130019 not due to an underlying defect in the genotype of the strain nor a culture problem (data not shown).
To determine whether the synergistic reduction in DSB abundance observed in rad24D spo11-HA strains was a general phenomenon affecting global DSB levels, we assessed total DSB signals in the dmc1D background across four chromosomes using pulsed-field gel electrophoresis (PFGE; figure 2g,h; electronic supplementary material, figure S3). We detected an (a)  rsob.royalsocietypublishing.org Open Biol 3: 130019 approximately 25-30% reduction in DSBs in the dmc1D spo11-HA strain relative to dmc1D (figure 2g,h). In rad24D dmc1D, DSBs were reduced by 10-20% relative to dmc1D (figure 2g,h). Finally, in the dmc1D rad24D spo11-HA strain, DSB frequency fell to just a third of the dmc1D level (figure 2g,h). Representative comparisons of lane profile plots are presented in the electronic supplementary material (figure S3). Collectively, these results indicate that the reduced abundance of Spo11-oligo complexes we had observed in rad24D dmc1D spo11-HA strains (figure 1b,c) is correlated with a reduced ability to detect DSBs by southern blotting.
In S. cerevisiae, as in mammals, meiotic recombination is necessary for accurate chromosome segregation-with errors in this process causing aneuploidy. To determine whether the molecular defects in DSBs and Spo11-oligo abundance correlate with defects in meiotic chromosome segregation, we measured the viability of the haploid spores formed upon completion of meiosis. By itself, the Spo11-HA tag is reported to have very little effect on spore viability [51,53]. We confirmed that spo11-HA alone causes no major reduction in spore viability (figure 2i). By contrast, in the rad24D strain, the spo11-HA allele caused spore viability to drop from 30% down to less than 5% (figure 2i).

Stimulation of double-strand break formation by a Rad24/Mec1(ATR)-dependent positive feedback loop
Mutation of the Rad24-dependent DDR was not expected to affect meiotic DSB formation. The spo11-HA allele, on the other hand, is reported to be a DSB formation hypomorph, variably reducing DSBs by 11-50% [51]. Contrary to these expectations, we have made two intriguing observations: (i) in the DMC1 and dmc1D backgrounds, DSB frequency at two recombination hotspots (HIS4::LEU2 and ARE1), and across three other chromosomes, is reduced less than we expected by spo11-HA; and (ii) DSB frequency, however, is more severely reduced when Rad24 checkpoint activity is abolished in spo11-HA; and spo11-HA dmc1D strains. Previous analysis of DSB abundance in spo11-HA made use of the rad50S mutant background [51]. In rad50S (as in the similar mutations, sae2D and mre11-'nuclease-dead'), DSB signals accumulate due to a failure to remove Spo11 from the DSB end [11,20,52]. We remeasured DSB abundance in the sae2D background at HIS4::LEU2 (figure 2j,k), ARE1 (see electronic supplementary material, figure S1c) and across the length of four chromosomes (figure 2g,h). In contrast to our previous results in SAE2 þ cells (see figure 2b-h; electronic supplementary material, figure S1b) and now more consistent with the published work [51], we observed DSB signals to be reduced by 50% in the spo11-HA sae2D strain compared with sae2D (see figure 2g,h,j,k; electronic supplementary material, figure S1c). Moreover, removal of checkpoint activity (rad24D) in sae2D strains had little impact on DSB formation in either SPO11 þ or spo11-HA (see figure 2g,h,j,k; electronic supplementary material, figure S1c).
Taken together, these observations suggest that while spo11-HA is deficient in DSB formation (as observed in sae2D), such a defect is substantially corrected in wild-type and dmc1D cells by a process that is dependent on the activity of the Rad24 (Mec1/ATR) checkpoint. We propose that these observations reveal the presence of a DDR-dependent positive feedback mechanism that is activated by ssDNA and promotes DSB formation under conditions of suboptimal catalysis (i.e. the spo11-HA hypomorph). Consistent with this idea, we note that the final frequency of DSBs that accumulates in rad24D dmc1D spo11-HA cells (as assessed by PFGE) is very similar to that observed in sae2D spo11-HA (figure 2 g,h), which, because of the absence of ssDNA intermediates, we expect will also lack efficient Rad24/Mec1 activation [34].
To determine whether the proposed feedback mechanism was a general feature of the DDR, we repeated experiments in strains mutated for components of either the checkpoint clamp (Rad17) or the Mec1(ATR) kinase itself (figure 3). Owing to problems with synchronizing mec1 cultures, we used an allele of MEC1 where expression is specifically repressed during meiosis ( pCLB2-MEC1; electronic supplementary material, figure S4). DSBs accumulated to high levels in both rad17D dmc1D and pCLB2-MEC1 dmc1D (approx. 16-20%; figure 3a-d). By contrast, in spo11-HA derivatives, DSB levels were reduced to approximately 30% of the SPO11 þ levels, as were ectopic recombination products (see electronic supplementary material, figure S2d,e), largely mimicking the synergistic DSB defects we observed with rad24D (figure 2). Furthermore, similar to the synergistic reduction in spore viability observed in rad24D spo11-HA strains (figure 2i), loss of the MEC1 or RAD17 activity in the spo11-HA background also caused a defect in spore viability (figure 3e,f). Overall, the effects conferred by the loss of Rad17 were less severe than Rad24 or Mec1, suggesting that there may be previously uncharacterized differences between the roles of these DDR components. Nevertheless, we tentatively conclude that the proposed positive feedback mechanism uses all major components of the canonical ssDNA-dependent DDR pathway.
3.3. The requirement for positive stimulation of doublestrand break formation by Rad24/Mec1 is not unique to the spo11-HA hypomorph The precise reason for the hypomorphic behaviour of the Spo11-HA protein is unknown. To determine whether the positive feedback mechanism we have uncovered is a general feature of meiotic recombination regulation (rather than uniquely required to rescue the spo11-HA allele), we investigated whether other Spo11 hypomorphs also benefit from Rad24/Mec1 checkpoint activation. We selected a weak hypomorphic allele of SPO11, spo11-D290A, which contains a point mutation in the putative Toprim domain thought to aid coordination of a metal ion necessary for DSB catalysis [53]. Under normal conditions, this allele is reported to have extremely mild defects in recombination, but suffer a severe defect at lower temperatures, or when HA-tagged [51,53]. For these reasons, we considered the spo11-D290A (untagged) allele to be a good candidate for testing the requirement for checkpoint activity. We first measured DSB formation at the HIS4::LEU2 locus (figure 4a-d). In the dmc1D background, the spo11-D290A allele resulted in a 30% reduction in DSB formation, and a dramatic redistribution of events away from DSB 1 and towards DSB 2 ( figure 4a,b,d). These effects were slightly more pronounced than the defects caused by spo11-HA (figure 2b-f), but nevertheless, in the DMC1 þ background, spo11-D290A supported almost wild-type levels of spore viability (figure 4e). In the sae2D background, spo11-D290A caused a more severe 50% rsob.royalsocietypublishing.org Open Biol 3: 130019 reduction in DSB signal at HIS4::LEU2 compared with SPO11 sae2D (figure 4a,c), and the same redistribution of events towards DSB 2 (figure 4d).
We next estimated DSB formation globally using PFGE and three chromosome probes (figure 4f,g). We detected no reduction in DSB formation in the dmc1D spo11-D290A background compared with dmc1D (on average it was 10% greater; figure 4g). By contrast, the spo11-D290A allele reduced DSB formation in the sae2D background by approximately 40% ( figure 4f,g), similar to what we had observed for spo11-HA (figure 2h). Loss of rad24D checkpoint activity caused DSB levels at HIS4::LEU2 to drop below our detection limit in the hours sae2D D290A sae2D spo11-HA rsob.royalsocietypublishing.org Open Biol 3: 130019 as a general mechanism to promote DSB formation when the efficiency of Spo11-DSB formation is compromised.

Involvement of Mek1 kinase in the double-strand break feedback mechanism
Activation of the DDR kinases (Mec1 and Tel1 in budding yeast) causes phosphorylation of Hop1 and subsequent activation of Mek1 [34], a meiotic kinase with roles related to Rad53/CHK2 [36 -39]. Mek1 activation is associated with phosphorylation on a number of residues [35]. To investigate the relationship between Mek1 activity and the proposed feedback process, we assessed the abundance of phosphorylated forms of Mek1 using phos-tag SDS-PAGE and western blotting ( figure 5a,b). In wild-type and spo11-HA cultures, slower migrating phosphorylated forms of Mek1 appear coincidently with the time of DSB recombination intermediates (3-6 h; figure 5a(i)). Phosphorylated Mek1 was slightly delayed in rad24D cells, suggesting a partial requirement for Rad24 to achieve Mek1 activation (figure 5a(ii)). Strikingly, phosphorylation of Mek1 was completely abolished in the rad24D spo11-HA strain (figure 5a(ii)). Because Mek1 phosphorylation signals are transient in DMC1 þ cells, we repeated the analysis in the dmc1D background, where DSB recombination intermediates and Mek1 signals would usually be expected to persist (figure 5b). In dmc1D and dmc1D spo11-HA, there was a steady accumulation in the abundance of multiple-shifted species of Mek1, indicating varying levels of phosphorylation (figure 5b). In the rad24D dmc1D strain, much of this hyper-phosphorylation was lost, confirming that hyper-activation of Mek1 is Rad24-dependent (figure 5b). In the rad24D spo11-HA dmc1D strain, all residual Mek1 phosphorylation was abolished (figure 5b), and not visible even when overexposing the film (figure 5b(ii)). Similar loss of Mek1 phosphorylation was observed in pCLB2-MEC1 dmc1D spo11-HA strains (data not shown). Thus, activated Mek1 correlates strongly with the ability to accumulate appreciable levels of meiotic DSBs.
3.5. Reduction of double-strand breaks in spo11-HA rad24D cells is not due to altered repair partner choice During meiosis, DSB repair reactions take place preferentially between homologous chromosomes rather than between sister chromatids [54,55]. The Mec1/Tel1 DDR checkpoint machinery is important for enforcing such 'repair partner choice' via activation of the Hop1 adapter protein and the Mek1 kinase [34,44]. It seemed plausible that the synergistic loss of DSB and Spo11-oligo signals, and drop in spore viability caused by the combination of spo11-HA and rad24D, could be explained by the observed failure to fully activate Mek1, and thus to appropriately control repair partner choice, resulting in increased occurrence of Rad51-dependent inter-sister recombination. To test this idea, we repeated our DSB analyses in rad51D dmc1D strains, thereby preventing all recombinationmediated pathways of DSB repair ( figure 6). As expected, in the rad24D rad51D dmc1D strain, DSB signals at HIS4::LEU2 accumulated, reaching a plateau of approximately 15% of total lane signal ( figure 6a,b). In the rad24D rad51D dmc1D spo11-HA strain, despite having removed the potential for both Rad51 and Dmc1-dependent repair, accumulated DSB signals were barely increased relative to rad24D dmc1D  rsob.royalsocietypublishing.org Open Biol 3: 130019 material, figure S3). We conclude that the loss of DSB signal observed in rad24D spo11-HA strains is not primarily due to changes in repair partner choice.
3.6. Reduction of double-strand break formation in spo11-HA rad24D cells is due to precocious Ndt80 activity DDR mutants initiate the first meiotic nuclear division about 1 h later than wild-type cells [40] (figure 6e). This delay is thought to be caused by errors in recombination. We noted that the spo11-HA allele alleviated the meiotic delay observed in the rad24D strain (figure 6e), consistent with the observed synergistic reduction in DSB signal alleviating recombination problems. However, it was possible that the reduced DSB signal in rad24D spo11-HA and rad24D dmc1D spo11-HA cells compared with rad24D SPO11 þ cells was itself due to the more rapid cell cycle progression beyond prophase I that arises in the rad24D spo11-HA double mutant compared with rad24D SPO11 þ (figure 6e). Exit from prophase I into anaphase I is controlled by expression of the Ndt80 transcription factor [56,57]. To distinguish between these two possibilities, DSB abundance was assessed in the ndt80D rad24D rad51D dmc1D spo11-HA background, where all DSB signals should accumulate, and where the cells will be forced to arrest at late prophase I in a checkpoint-independent manner owing to deletion of the NDT80 gene. To our surprise, the ndt80D-enforced prophase arrest was sufficient to restore DSB accumulation at HIS4::LEU2 to the otherwise DSB-deficient rad24D rad51D dmc1D spo11-HA strain ( figure 6a(ii)).
To confirm that the restoration of DSB formation by ndt80Denforced arrest was not unique to HIS4::LEU2, we measured DSB formation across four chromosomes using PFGE (figure 6d,f). Consistent with our results at HIS4::LEU2, the ndt80D arrest caused DSB signals to return to levels indistinguishable from the rad24D dmc1D rad51D control (figure 6d,f). Taken together, our results suggest that the apparent positive feedback on meiotic DSB formation is mediated by Mec1-dependent inhibition of Ndt80.

Transient ndt80-induced prophase arrest
improves spore viability in hypomorphic spo11 rad24D strains Our results suggest that the synergistic reduction in DSB signal observed in rad24D spo11-HA and rad24D spo11-D290A strains is due to precocious Ndt80 expression and/or activation causing more rapid passage through meiotic prophase, which we suppose prevents efficient DSB formation by the hypomorphic Spo11-HA and Spo11-D290A proteins. If this idea is correct, we reasoned that transiently extending meiotic prophase might be sufficient to restore DSB formation (and the subsequent crossover formation necessary for accurate chromosome segregation) to checkpoint-defective Spo11 hypomorphs. To test this idea, we placed the NDT80 transcription factor under the control of the GAL1-10 promoter, where expression can be induced by addition of oestradiol to the growth media [58]. As expected, without addition of oestradiol, cells failed to sporulate, arresting permanently in meiotic prophase, similar to what occurs upon deletion of the NDT80 gene (data not shown). By contrast, samples induced with oestradiol after 8 h in meiosis sporulated efficiently and permitted us to assess the viability of the haploid spores ( figure 7). Dramatically, we observed spore viability in both the rad24D spo11-HA and rad24D spo11-D290A strains to reach 73% and 68%, respectivelyincreases of approximately 20-fold-as a consequence of simply extending the period in meiotic prophase ( figure 7). Intriguingly, the transient ndt80-arrest improved spore viability significantly above that of the rad24D strain ( figure 7). This suggests that one of the reasons for low spore viability in rad24D cells is due to untimely onset of the meiotic divisions-perhaps before all DSB repair has finished. Collectively, these results provide compelling evidence for the activation of the Rad24/Mec1 checkpoint pathway being critical for the orchestration of meiotic recombination events with the meiotic nuclear divisions.

Cdc5 depletion does not permit double-strand
break formation in rad24D dmc1D spo11-HA strains To determine whether the DSB defects in rad24D strains is directly due to Ndt80 expression or instead to the onset of anaphase I, which ensues as a result of the Ndt80-dependent transcriptional cascade, we asked whether arresting cells at the metaphase-anaphase transition is sufficient to restore DSB formation in NDT80 þ rad24D dmc1D spo11-HA cells.
The budding yeast Polo-like kinase, Cdc5, regulates late prophase events including Holliday junction resolution and the dissociation of Rec8 cohesin from meiotic chromosomes to enable their segregation [59,60]. Because CDC5 is an essential gene, we took advantage of a pCLB2-CDC5 allele developed previously, which, similar to our pCLB2-MEC1 allele, is not expressed during meiosis [59,60]. Depletion of Cdc5 caused cells to arrest in prophase (data not shown), yet had little effect on measured DSB frequency at HIS4::LEU2 (see electronic supplementary material, figure  S5). In rad24D dmc1D pCLB2-CDC5 cells, DSB frequencies rose to 25% of total DNA at late time points (see electronic supplementary material, figure S5), a modest increase relative to CDC5 þ that may be due to prolonged prophase I arrest. Cdc5 depletion also had almost no effect on DSB accumulation in the rad24D dmc1D spo11-HA background, with measured signals barely rising above general lane background (see electronic supplementary material, figure S5). Our observation that DSB formation is restored by NDT80 deletion (figure 6), but not by Cdc5 depletion, points to a direct inhibitory role of Ndt80 itself, or of a protein other than Cdc5 that is upregulated by Ndt80 expression, rather than an indirect consequence of prophase arrest.

A model for regulation of double-strand break formation
Meiotic DSB formation is regulated at numerous levels by relatively poorly understood mechanisms. Although Spo11 is the evolutionarily conserved catalytic component, its activity depends on interactions made between itself and the accessory rsob.royalsocietypublishing.org Open Biol 3: 130019 DSB complex proteins, and with the chromosomal substrate. Spo11 activity is modulated temporally by the local timing of DNA replication [61] and by cell cycle-dependent phosphorylation events, such as phosphorylation of Mer2 by cyclin-dependent kinase (CDK) and Dbf4-dependent kinase (DDK) [62][63][64], and spatially by histone occupancy [48] and their post-translational modification [65,66], and by meiosisspecific chromosome structure components [48,67].
Here, by using an allele of Spo11, which is hypomorphic for DSB catalysis, we additionally reveal that DSB formation is modulated in a positive manner by activation of the DNA damage checkpoint machinery Mec1(ATR). Recent work in mice, flies and yeast [45][46][47], and additional observations from our own laboratory, suggest that the complementary Tel1(ATM) kinase pathway acts to inhibit DSB formation (S.G., R.A., V.G. & M.J.N. 2012, unpublished data). Collectively, these results point to the interaction of two regulatory loops that appear to work antagonistically to modulate DSB frequency, thereby creating a moderate level of recombination that is neither too high nor too low (figure 8a).
Our model contains a number of features ( figure 8a,b): the rate of Spo11-DSB formation is initially increased concomitantly with CDK/DDK-dependent phosphorylation of Mer2 [62][63][64], but reaches a plateau that we suggest is at least in part due to the activation of Tel1 and Ndt80 (see below). We propose that in the hypomorphic spo11-HA and spo11-D290A strains, DSB accumulation is initially slower, but that activation of the Rad24/Mec1(ATR) kinase by ssDNA formation at emergent DSBs causes transient inhibition of Ndt80, which enables the rate of DSB formation to increase until it reaches a plateau similar to that observed in wild-type cells. Therefore, activation of this positive feedback results in the similar efficiency of DSB formation that we observe in SPO11 þ relative to spo11-HA or spo11-D290A strains. By contrast, in the absence of the Mec1 checkpoint pathway, precocious Ndt80 activity results in a residual low rate of DSB formation in the hypomorphic spo11 strains. Our model is consistent with the increased Ndt80 transcription observed in the dmc1D rad17D checkpoint mutant relative to dmc1D [68]. We suggest that a comparable effect is observed in sae2D spo11-HA and sae2D spo11-D290A strains, which, owing to an inability to produce ssDNA, may also exhibit precocious Ndt80 activation.
Overall, our results are surprising given the proposed negative role of ATM in DSB formation in mouse and flies [45,46], and the negative role in DSB formation reported at the HIS4::LEU2 hotspot for Tel1 in S. cerevisiae [47]. We assume that over time Tel1 becomes hyper-activated by the presence of multiple ongoing recombination intermediates, and that this then downregulates DSB formation. As such, mutation of TEL1 increases the rate and/or prolongs the activity of the Spo11-DSB machinery (S.G., R.A., V.G. & M.J.N. 2012, unpublished data).
It is worth noting that we have revealed the requirement of the Mec1 checkpoint response in situations where Spo11 activity is downregulated due to defined genetic modification. Although we are unable to exclude that the mechanism we describe reflects a unique consequence of this mutant situation, we interpret our results to indicate that the transient prophase  Figure 7. Transient prophase arrest rescues rad24D spore viability defects. (a,b) Spore viability displayed as (a) total viability or (b) separated according to number of viable spores per tetrad for the indicated strains. Error bars are 95% confidence limits. NDT80 indicates spore viability of normally sporulated strains. pGAL-NDT80 derivatives were transiently arrested in prophase until 8 h, then induced to sporulate by addition of beta-oestradiol to 2 mM. delay caused by Mec1 activation will also promote meiotic DSB formation in other suboptimal circumstances. For example, mutation in a gene other than Spo11, defects in chromosome pairing and/or synapsis or perhaps slower Spo11-DSB formation for an environmental reason (see below).

Possible mechanisms of regulation
Mec1 and Tel1 are members of the PIKK-family, and preferentially phosphorylate substrate serine and threonine residues that precede a glutamine (SQ/TQ sites [69]). One of the many intriguing points of our study is how the two kinases are able to elicit both positive and negative effects. This could be explained if the rate of phosphorylating different substrates changes with kinase activity. For example, an initial low-level phosphorylation of an abundant (and easily phosphorylatable) target might promote DSB formation, but upon hyper-activation of Mec1/Tel1, additional substrates become phosphorylated that begin to negatively impact on DSB formation. Candidate targets are many, and could be histones, chromosomal structural proteins or members of the DSB-forming complex itself. The chromosomal protein, Hop1-a known target of Mec1 and Tel1 [34]-is involved in both DSB formation and DSB repair, and more recently Rec114 (a member of the Spo11-DSB complex) has been identified as a target of Mec1/Tel1 [70]. Our results also implicate the involvement of the downstream kinase, Mek1-whose substrates currently remain unclear-and of Ndt80, the transcription factor required to promote exit from meiotic prophase. Defining the relevant targets of Ndt80 and these three kinases, and of their mechanism of action, are important future goals if we are to fully understand how DSB formation is regulated.
The spatial and temporal distributions of meiotic DSB formation may also be modulated by meiotic chromosome morphogenesis. For example, chromosome axis-associated proteins involved in Spo11-DSB formation are excluded from synapsed chromosomes [71,72], suggesting that synapsis itself may promote their eviction. Consistent with this idea, in mouse spermatocytes, regions of chromosomes that fail to synapse display increased density of DSB markers [73]. Therefore, it remains possible that under conditions of suboptimal Spo11-DSB formation, concomitant reductions in chromosome pairing and synapsis will cause an increase in DSB formation at chromosomal regions that have yet to become paired or synapsed [70,73]. Such a homeostatic mechanism, which could arise via a reduction of cis-and/or trans-inhibition, is similar to that discussed by Zhang et al. [47]. We wish to emphasize that the transient prophase arrest activated by the Mec1(ATR)-or Tel1(ATM)-dependent damage response (as we observe in this work) could ensure that progression into late stages of meiotic prophase does not occur before sufficient numbers of recombination interactions have been established to ensure accurate meiotic chromosome segregation.

Differential effects on double-strand break
formation across the HIS4::LEU2 locus The HIS4::LEU2 DSB site 1 hotspot arose fortuitously via the ectopic insertion of a short bacterial DNA sequence during integration of the LEU2 marker downstream of the HIS4 gene [74]. DSB site 1 remains one of the strongest recombination sites characterized in the budding yeast genome [48]. The precise reason for becoming such a strong DSB site is not known, but if we assume DSB site 1 conforms to the general mechanisms regulating DSB formation at natural sites, it is likely to be due to the creation of a higher-order chromatin structure that makes the local DNA sequence particularly receptive to Spo11-DSB activity [48]. Given this idea, it is intriguing that the two Spo11 hypomorphs, spo11-HA and spo11-D290A, have such a severe effect on DSB site 1, but not at site 2 nor at DSB formation generally across the genome. DSB site 1 is particularly narrow relative to its heat (see electronic supplementary material, figure S6). When compared to all other mapped DSB hotspots [48], DSB site 1 is narrower than 60% of all hotspots (see electronic supplementary material, figure S6a), yet is in the top 3% of sites ranked according to the frequency of mapped Spo11-oligonucleotides (see electronic supplementary material, figure S6b)-with Spo11-oligo hits about 10 times greater than at other similarly narrow sites. Together, this makes DSB site 1 have the second highest density of mapped Spo11-oligonucleotides per bp (see electronic supplementary material, figure S6c), and a significant outlier relative to hotspots of similar width (see electronic supplementary material, figure S6a). We speculate that these unique characteristics render DSB site 1 particularly sensitive to reductions in Spo11's catalytic activity. By contrast, DSB site 2 maps to the promoter of the inserted LEU2 gene, and probably behaves in a way more similar to canonical DSB sites and provides a buffer for defects in DSB formation at site 1. DSB site preference was also modified by loss of Rad24 activity. However, in this case, rad24D caused an increase in the frequency of DSBs at site 1 (figure 2). Although we do not know the precise reason for this, one explanation is that while Rad24/Mec1 acts generally to promote DSB formation via inactivation of Ndt80, Rad24/Mec1 may also act locally to inhibit DSB formation at the equivalent locus on the sister chromatid or homologous chromosome [47]. The fact that such effects are observed in dmc1D recombination mutantswhich are thought to lack DNA strand exchange interactions between homologous chromosomes-suggests that such Rad24-dependent trans-inhibition may predominantly take place between sister chromatids.

Double-strand break homeostasis versus crossover homeostasis
During meiosis, only a portion of DSBs repair as interhomologue crossover events-the prerequisites for chiasma formation and the accurate segregation of homologous chromosomes. In budding yeast, about half of total DSB events resolve as a crossover [49,75], whereas in mammals the fraction is less than one-tenth [76]. An excess of precursor events (DSBs) may provide a large buffering pool from which to generate the essential, but smaller, number of crossovers.
Such a system has the potential to better tolerate cell-to-cell variation in DSB number. The name used to describe this phenomenon is crossover homeostasis, and it was revealed via the analysis of a hypomorphic series of Spo11 alleles that systematically reduce DSB frequency [51]. Martini et al. uncovered a nonlinear relationship between gene conversion and crossing over at a modified ARG4 locus, supporting the notion of crossover homeostasis [51]. However, it was not reported whether the hypomorphic SPO11 series affected DSB frequency and/or DSB distribution at this locus (such as we have witnessed at HIS4::LEU2). Changes in either parameter can alter measured gene conversion frequencies, and chromosomal estimates of DSB formation were made in a rad50S background, where, like sae2D, Spo11-DSBs accumulate without ssDNA resection [52].
Our work here suggests that the measured frequency of DSB formation obtained in such hypomorphic Spo11 strains may be substantially different in rad50S/sae2D when compared with resection-proficient wild-type or dmc1D strains due to the activation of Mec1 substantially improving DSB formation. Assuming our observations are an accurate reflection of DSB formation in repair-proficient (DMC1 þ ) meiosis, then the strength of crossover homeostasis may need to be reviewed. Alternatively, it is possible that the substantial increases in DSB formation we report for dmc1D spo11-HA and dmc1D spo11-D290A compared with sae2D spo11-HA and sae2D spo11-D290A are themselves overestimates due to checkpoint hyper-activation stimulating DSB formation in a manner that would not occur in repair-proficient cells. Developing accurate and quantitative measures of genome-wide recombination is needed to resolve such conundrums. 4.5. Why did checkpoint-dependent regulation of double-strand break formation arise?
The initiation of meiotic recombination by DSB formation is a potentially catastrophic event for the maintenance of genome stability: any unrepaired DSB can result in the loss of genetic information upon nuclear division. In this context, it makes sense for the DSB programme to be tightly temporally regulated: DSBs should form only in meiosis, and only after local DNA replication has occurred. DSB formation and repair must also occur efficiently across the genome prior to the onset of anaphase. We suggest that the positive and negative action of the meiotic checkpoint helps to make this system efficient and robust under what may often be suboptimal conditions. At this point, we note that budding yeast did not evolve to enter meiosis under the controlled laboratory conditions in which most experiments are performed. In the wild, variation in genotype and in the local environment may cause substantial deficiencies in these processes. Indeed, effects on recombination caused by polymorphism, temperature and nutritional deficiencies have all been reported [49,[77][78][79]. In this context, activation and utilization of the DNA damage response may be just one of many redundant mechanisms that help to modulate appropriate recombination outcome.

Double-strand break analysis
DSB signals were detected using standard techniques by indirect end-labelling of specific genomic loci after fractionation and transfer to nylon membranes [82]. For DSBs at HIS4:LEU2, genomic DNA was digested with PstI, fractionated on 0.7% agarose in 1Â tris-acetate-EDTA (TAE) for approximately 18 h at room rsob.royalsocietypublishing.org Open Biol 3: 130019 temperature, transferred to nylon membrane under denaturing conditions then hybridized with a probe that recognizes the MXR2 locus. For DSBs at ARE1, genomic DNA was digested with BglII, fractionated and transferred as above, and then hybridized with a probe that recognizes the RSC6 locus. To measure the total DSB formation along whole chromosomes, chromosomal-length DNA was prepared after first immobilizing cells in agarose plugs as described [82]. Chromosomes were fractionated using a CHEF-DRIII PFGE system (BioRad) using the following conditions: 1.3% agarose in 0.5Â TBE; 148C; 6 V cm -1 ; switch angle 1208; figures 2g and 6c, ramped switch time of 15-22 s over 25 h, then constant switch time of 45 s for 4 h; figures 4f and 6f, ramped switch time of 20-50 s over 28 h. After transfer to nylon membrane under denaturing conditions, genomic DNA was sequentially hybridized with DNA probes that localize close to the left telomere of four chromosomes. Probes used were CHA1 (Chr III), RMD6 (Chr V), CBP2 (Chr VIII) and JEN1 (Chr XI). Between each hybridization, radiolabelled probes were removed by sequential washes in 0.4 M NaOH (2 Â 5 min), 2Â SSC (2 Â 5 min) and six rinses in distilled water, and then air-dried on Whatman paper for 1 h and stored at 48C. Radioactive signals were collected on phosphor screens, scanned with a Fuji FLA5100 and quantified using IMAGEGAUGE software (FujiFilm). DSBs at HIS4::LEU2 are reported as a percentage of the total lane signal after background subtraction. For analysis of PFGE experiments, DSBs occurring far from the probed chromosome end will be underrepresented whenever they arise on a chromosome with additional DSBs closer to the probe. To correct for this, the measured frequency of DSBs (fraction of total lane signal) was estimated using the following standard formula: corrected DSB fraction ¼ -ln(1measured DSB fraction). This correction assumes that DSBs at different loci on the same chromosome occur independently of one another. Values plotted with standard deviation bars are the mean of at least two independent time-course experiments. For PFGE experiments, all data are the average of at least three independent time courses of each strain except for rad24D dmc1D (two experiments). Where the exact same sample has been analysed on different gels, the average measure for that sample is used before calculating the average and standard deviation of the experimental repeats. For PFGE data, DSB frequencies for all strains are expressed relative to the average frequency of DSBs forming in dmc1D strain.

Meiotic progression
Cells were fixed in 100% methanol and aliquots mixed with 1 mg ml 21 DAPI. Cells were scored for the presence of one, two or four nuclei and percentage at each time point calculated.

Spore viability
Fresh diploid colonies were incubated in 2% K-acetate liquid for 48 h at 308C, and then incubated with zymolyase 100 T at a final concentration of 10 mg ml 21 in a 150 mM sodium phosphate buffer at 378C for 10 min. Dissected spores were incubated for 2 days at 308C on YPD and scored for percentage viability per strain and viable spores per tetrad. For the pGAL-NDT80 experiments, synchronized cultures were split after 8 h in 2% K-acetate, and one fraction induced to sporulate by addition of beta-oestradiol to a final concentration of 2 mM. Cultures were then incubated for a further 40 h at 308C prior to dissection. Where shown, error bars are 95% confidence limits.

Spo11-oligonucleotide assay
Spo11-oligonucleotide complexes were detected by immunoprecipitation and end-labelling following established methods [11,18,50]. Briefly, cells were broken in 10% ice-cold TCA using zirconium beads and a BioSpec 24. Precipitated material was dissolved in SDS buffer, diluted with Triton X100, and Spo11 was immunoprecipitated from total soluble protein using anti-HA antibody (F-7; Santa Cruz Biotechnology) and protein-G-agarose (Roche). Oligonucleotide complexes were 3 0 -labelled with a-32 P dCTP using Terminal deoxynucleotidyl transferase (Fermentas) and fractionated on a 7.5% SDS-PAGE gel. Following transfer to polyvinylidene difluoride (PVDF) membrane, radioactive signals were collected on phosphor screens, scanned with a Fuji FLA5100 and quantified using IMAGEGAUGE software.