Numerical and spatial patterning of yeast meiotic DNA breaks by Tel1

Increased Spo11-oligo numbers in tel1Δ

In yeast, global DSB levels are often estimated by Southern blotting of chromosomes separated on pulsed-field gels. Such studies of tel1Δ mutants gave conflicting results (Argunhan et al. 2013; Blitzblau and Hochwagen 2013; Carballo et al. 2013; Garcia et al. 2015). One issue is the difficulty in controlling for DSB lifespan, since in wild-type cells, DSBs disappear as they are repaired by recombination (Hunter and Kleckner 2001). Because Tel1 promotes normal bias in favor of using the homolog rather than the sister chromatid for DSB repair (Carballo et al. 2008; Joshi et al. 2015), absence of Tel1 may alter average DSB lifespan and affect the ability to detect ongoing de novo DSB formation as meiosis proceeds. The use of DSB repair-defective backgrounds was once thought to obviate the lifespan issue, but recent studies have shown that these mutants introduce confounding effects by severing or hyperactivating negative feedback pathways that regulate DSB numbers (Cooper et al. 2014; Keeney et al. 2014; Thacker et al. 2014). Finally, assays based on indirect end-labeling of electrophoretically separated genomic DNA are often unable to distinguish lone DSBs from the clustered DSBs that form in the absence of Tel1, leading to underestimated DSB levels (Garcia et al. 2015).

To circumvent limitations of gel-based quantification of broken chromosomes, we assayed Spo11-oligo levels instead. Spo11 makes a DSB by forming covalent protein-DNA complexes that are then endonucleolytically cleaved to release Spo11 bound to a DNA oligo, with two Spo11-oligo complexes generated from each DSB (Supplemental Fig. S1A; Neale et al. 2005). Because Spo11-oligo complexes are a direct by-product of DSB formation and because their lifespan appears to be tied to progression out of the first meiotic prophase rather than being tied to progression of the recombination reaction, they can be used to quantify relative DSB levels genome-wide in an otherwise wild-type background, mitigating confounding effects of repair-defective mutations (Lange et al. 2011; Thacker et al. 2014).

Whole-cell extracts were prepared from equivalent numbers of wild-type and tel1Δ cells harvested at different times after meiotic induction. Phenotypically normal protein-A-tagged Spo11 was affinity-purified, 3′ ends of Spo11 oligos were radiolabeled using terminal deoxynucleotidyl transferase and [α-³²P] dCTP, and Spo11-oligo complexes were separated on SDS-PAGE gels and detected by phosphorimager (Fig. 1A, top). A Western blot for total Spo11 confirms that Spo11 protein expression is induced to a similar extent in wild-type and tel1Δ cultures (Fig. 1A, bottom). Wild-type cells displayed the typical kinetics of appearance and disappearance of Spo11-oligo complexes (Fig. 1A,B), previously shown to coincide with DSB formation (Neale et al. 2005; Thacker et al. 2014). In the tel1Δ mutant, Spo11-oligo complexes first appeared with similar timing and levels as wild type but continued to accumulate, reaching a maximum at 5 h that was 2.2-fold higher on average than in wild type (Fig. 1A,B). Increased numbers of Spo11-oligo complexes were also apparent in the Western blot detecting total Spo11: A slower-migrating band was now apparent that comigrated with the upper radioactive signal. (The species corresponding to the lower radioactive signal was masked on the Western blot by the signal for free Spo11.)

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 1.

Elevated levels of Spo11-oligo complexes in tel1Δ mutants. (A) Representative time course. Autoradiograph of SDS-PAGE gel (top) shows radiolabeled Spo11-oligo complexes at the indicated times in sporulation medium. Anti-protein A Western blot (bottom) monitors total Spo11. Note that most signal in the Western blot is free Spo11, i.e., not bound to oligos. The arrow indicates the small amount of Spo11 protein that comigrates with slower migrating Spo11-oligo species, visible in the tel1Δ sample because of increased DSB numbers. (B) Quantification of radiolabeled Spo11-oligo complexes in tel1Δ cells, relative to wild-type cultures collected in parallel. Mean ± SD for three pairs of cultures is shown. (C) Meiotic progression (the percentage of cells completing one or both divisions). Mean ± SD for the same cultures as in B; ≥100 cells counted per time point.

DSB formation is inhibited when cells exit from the pachytene stage of prophase I (for review, see Keeney et al. 2014). Thus, DSBs can continue to form under conditions of prophase I arrest, e.g., in response to unrepaired DSBs or in the absence of the transcription factor Ndt80. However, tel1Δ cells divided with similar kinetics as wild-type cells (Fig. 1C), indicating that the increased Spo11-oligo levels are not a consequence of either increased DSBs or altered Spo11-oligo-complex lifespan caused by delayed meiotic progression. Thus, quantification of Spo11-oligo complexes, combined with prior analyses of recombination products (Zhang et al. 2011; Anderson et al. 2015) and some physical analyses of broken genomic DNA (Carballo et al. 2013; Garcia et al. 2015), supports the conclusion that loss of Tel1 increases DSBs globally. Furthermore, the kinetics of accumulation of Spo11-oligo complexes (Fig. 1B) indicate that DSB formation continues over a longer period in tel1Δ than in wild type.

Roles of Tel1 kinase activity and Tel1 phosphorylation targets

An unresolved question is whether negative regulation of DSB numbers by ATM/Tel1 depends on its kinase activity and, if so, what the relevant phosphorylation targets are. To address these issues, we first asked if a known tel1 kinase-dead mutation (“tel1-kd”) (Ma and Greider 2009) increased DSBs. Indeed, Spo11-oligo complexes were elevated in tel1-kd cultures to an extent similar to that in tel1Δ mutants (Fig. 2A). These results suggest that Tel1 can suppress DSB formation via the phosphorylation of one or more target proteins.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 2.

Spo11-oligo complexes in tel1-kd, rec114-8A, and hop1-scd mutants. (A–D) Representative time courses and quantification of (A) two, (B) four, and (C) six independent mutant cultures processed in parallel with (A) three, (B) four, and (C) four independent wild-type cultures are shown as in Figure 1. To aid comparison, tel1Δ data are duplicated from Figure 1. (D) Meiotic division profiles for the same cultures.

A candidate target is Rec114, which is required for DSB formation and bears matches to the S/TQ phosphorylation motif for Tel1 and its related kinase Mec1 (Carballo et al. 2013). Mec1 phosphorylates Rec114 in vitro (Tel1 was not tested); Mec1 and Tel1 are redundantly required for DSB-dependent phosphorylation of Rec114 in vivo; and mutation of eight S/TQ sequences in Rec114 to AQ (“rec114-8A”) yielded indirect evidence of increased and/or earlier DSB formation (Carballo et al. 2013). We therefore asked if rec114-8A phenocopies absence of Tel1. Strikingly, although the rec114-8A mutant showed a slight increase in Spo11-oligo complexes at late time points, this was substantially less than in the tel1Δ or tel1-kd mutants (Fig. 2B). We also combined rec114-8A with mutation of three S/TQ residues of Hop1, another known Tel1 and Mec1 target (known as the hop1-scd allele) (Carballo et al. 2008). However, even the small increase seen at late times with rec114-8A alone was not apparent in the rec114-8A hop1-scd double mutant (Fig. 2C). These mutations had little or no effect on division timing (Fig. 2D). We infer that Rec114 and Hop1 phosphorylation are largely dispensable for Tel1-mediated DSB suppression.

Altered Spo11-oligo sizes in tel1 mutants

Most Spo11-oligo complexes are in two size classes that differ in the lengths of the attached oligos (Fig. 1A; Neale et al. 2005). The ratio between these classes appeared altered relative to wild type in the tel1Δ mutant, with more of the higher molecular weight species (Fig. 1A). To examine this change, Spo11 oligos were separated on a sequencing gel (Fig. 3A). Consistent with the SDS-PAGE, tel1Δ showed an increase in oligos >25 nt long relative to those <20 nt, including increased abundance of very large oligos near the top of the gel (Fig. 3A,B). The distribution of larger oligos was also altered: A modest shoulder at ∼35 nt in wild type became more prominent in tel1Δ, and even longer oligos (from ∼40 to ∼70 nt) displayed a more pronounced banding pattern in the mutant (Fig. 3A,B).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 3.

Sizes of Spo11 oligos in Tel1 pathway mutants. (A) Radiolabeled, de-proteinized Spo11 oligos separated on a denaturing 15% polyacrylamide sequencing gel. Lanes are from a single gel with an intervening lane omitted. (M) 10-nt ladder. Note that oligos from lower-molecular-weight Spo11-oligo complexes are underrepresented because most of the shortest oligos (less than ∼12 nt) are lost upon ethanol precipitation. (B,C) Lane profiles from panel A.

The tel1-kd mutant was similar with respect to increased abundance of longer Spo11 oligos (those >25 nt relative to those <20 nt), the more prominent population at ∼35 nt, and the banding at ∼40–70 nt (Fig. 3A,B). Interestingly, these alterations were even more extreme in tel1-kd than in tel1Δ, indicating that presence of a kinase-dead protein does not precisely phenocopy absence of Tel1.

The distribution of oligo sizes was also altered in the rec114-8A mutant, in similar ways but to a lesser degree than in tel1Δ (Fig. 3A,C). This finding is consistent with Rec114 phosphorylation contributing to regulation of the formation and/or processing of DSBs, but the results also reinforce the conclusion that Rec114 is not the sole Tel1 substrate important for DSB control.

The DSB landscape responds differently to Tel1 loss at different stages of meiosis

Having confirmed that Tel1 regulates DSB numbers, we investigated how it shapes the DSB distribution. Analysis of recombination products (Anderson et al. 2015) supported an earlier hypothesis that Tel1 shapes the DSB landscape (Lange et al. 2011), but recombination maps are indirect and low-spatial-resolution indicators of DSB positions. We deep-sequenced Spo11 oligos purified from wild-type and tel1Δ cultures at 4 h in sporulation conditions, when DSBs are maximal in wild type and the tel1Δ mutant has only slightly elevated total Spo11-oligo levels, and at 5 h and 6 h, when the tel1Δ mutant has substantially elevated Spo11 oligos (Fig. 1B; Supplemental Table S1). As expected, the map at 5 h consistently showed patterns intermediate between the 4-h and 6-h time points. Two tel1Δ biological replicates were highly similar (Fig. 4A; Supplemental Fig. S2) and were averaged together. Wild-type samples displayed little systematic variation over this time window, so maps from six samples (two each from 4, 5, and 6 h) were averaged together to generate a consensus wild-type map used for all comparisons to mutant maps.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 4.

Spo11-oligo landscapes. (A) Reproducibility of Spo11-oligo maps. A representative segment from Chr III is shown at the top and a close-up view of averaged biological replicates in the yellow-highlighted region is shown below. Both Spo11-oligo maps were smoothed with a 201-bp Hann window. (B) Subchromosomal domains. Boxplots show the log-fold change in Spo11-oligo counts in each mutant relative to wild type for hotspots in the indicated genomic regions. (ST) Subtelomeric, 20 kb at each chromosome end (62 hotspots), (P) pericentric, 20 kb centered at each centromere (82 hotspots), (R) rDNA-proximal region (Chr XII: 451,577–467,570; 20 hotpots), (IS) interstitial, i.e., all other genomic regions. Thick horizontal lines are the medians, box edges are the 25th and 75th percentiles, whiskers indicate lowest and highest values within 1.5-fold of the interquartile range, and open circles are outliers. Colored dashed lines show the genome-average fold change for each mutant (1.5 for tel1Δ 4 h, 2.3 for tel1Δ 5 h, 3.5 for tel1Δ 6 h, 1.8 for zip3Δ 5 h, and 1.3 for tel1Δ zip3Δ 4 h). Black dashed lines at zero denote no change. Wild-type data for these and related comparisons are the averaged data for 4-, 5-, and 6-h samples. (C) Regional variation in the response of hotspots to the absence of Tel1. Chr IX is shown as an example. Each point is the log-fold change in Spo11-oligo count relative to wild type. Solid lines are local regression curves; dashed lines indicate genome-wide average fold change or no change. (D) Local domains of correlated behavior. Each point compares the log-fold change in hotspots with the change in their neighbors located in a 5-kb window the indicated distance away. Shaded regions show 95% confidence intervals for hotspots randomized within-chromosome; randomized data at 4 and 5 h show higher correlation because of the residual chromosome size effect (see Fig. 5A and Supplemental Fig. S3D). Subtelomeric, pericentric, and rDNA-proximal regions were excluded.

At short (sub-kb) size scales, Spo11 oligos from tel1Δ cells mapped preferentially to hotspots highly similar in location and width to those from wild type (Fig. 4A; Supplemental Fig. S3A,B). We conclude that the influence of local chromatin structure on hotspot position and shape is largely unaffected by tel1Δ and that the increase in Spo11-oligo levels is not principally from DSBs in regions that were previously inaccessible to Spo11. Interestingly, these findings contrast with those in mouse, where absence of ATM caused more hotspots to become detectable and caused hotspots to become wider (Lange et al. 2016) (see Discussion).

When assessed on larger size scales, however, tel1Δ showed substantial alterations that changed over time in meiosis. At 4 h, Spo11-oligo frequencies had increased less than average in hotspots within 20-kb zones at telomeres and centromeres and had increased even less near the ribosomal DNA (rDNA) (Fig. 4Bi). Accumulation of Spo11 oligos in these regions continued to lag behind the genome average at 5 h (Fig. 4Bii), but by 6 h, hotspots in these domains were similar to or higher than the genome average, i.e., were all about equally elevated (Fig. 4Biii). Thus, these domains, which are suppressed for DSBs in wild type (Pan et al. 2011), are initially somewhat insulated from the elevation in DSB levels caused by Tel1 deficiency but over time continue to accumulate additional DSBs to reach a similar extent as elsewhere in the genome. It is possible that the behavior of subtelomeric regions is influenced by telomere changes that occur during vegetative growth in the tel1Δ mutant (Lustig and Petes 1986).

The remaining interstitial hotspots showed variable responses to the tel1Δ mutation at all time points (Fig. 4Bi,Bii,Biii,C) but with only a modest correlation between the 4-h and 6-h samples (Pearson's r = 0.34) (Supplemental Fig. S3C). The 5-h sample showed intermediate correlation but was more similar to 6 h (Supplemental Fig. S3C). Local regression of the fold change in Spo11-oligo signal along chromosomes displayed peaks and valleys, suggesting that groups of hotspots respond to the tel1Δ mutation in a similar manner (Fig. 4C). To test this conclusion, we measured the correlation in response between hotspots and their neighbors. At 4 h, the change in each hotspot's Spo11-oligo count was positively correlated with the change in its neighbors, with correlation strength decaying with distance but still above randomized controls up to ∼125 kb away (Fig. 4D). In contrast, the 6-h time point had a weaker starting correlation for close hotspots (≤5 kb), and this correlation decayed to background levels by 15 kb (Fig. 4D). The 5-h time point had intermediate behavior: Similar to 6 h, correlation strength decayed to background within a short distance, but background (randomized) correlation levels were higher as in the 4-h time point, reflecting chromosome size effects discussed below (Supplemental Fig. S3D). Thus, while hotspots appear to respond to absence of Tel1 in a “domainal” fashion early in meiosis, this organization is largely erased as meiosis progresses. The convergence of the tel1Δ Spo11-oligo map with the wild-type map is coincident with the period when additional DSBs are forming in the tel1Δ mutant (Fig. 1B). Therefore, we conclude that loss of Tel1 leads to substantial alterations in the DSB landscape, and these alterations change over time as cells continue to make DSBs. The “normalization” of the tel1Δ Spo11-oligo map at later times indicates that the remaining (Tel1-independent) DSB regulatory systems have the net effect of homeostatic maintenance of the DSB distribution despite changes in DSB numbers.

Relationship between Tel1 and the homolog engagement pathway for DSB control

DSB numbers are also elevated in mutants lacking “ZMM” proteins (Thacker et al. 2014), factors needed to complete recombination and synaptonemal complex formation (Lynn et al. 2007). This and findings from other organisms (Wojtasz et al. 2009; Hayashi et al. 2010; Kauppi et al. 2013) have led to the hypothesis that homologous chromosomes that successfully engage one another largely stop making DSBs (Kauppi et al. 2013; Keeney et al. 2014; Thacker et al. 2014). This negative feedback regulation, which contributes to DSB spatial patterning (Thacker et al. 2014), is proposed to be distinct from Tel1-mediated DSB control (Keeney et al. 2014). To test this proposal, we compared how the DSB landscape changes in the absence of Tel1 to how it changes in the absence of the ZMM protein Cst9 (referred to hereafter by its more commonly used alias, Zip3). Spo11-oligo maps from a zip3Δ mutant are currently available only for 5 h in sporulation (Thacker et al. 2014), so the degree to which DSB patterns change over time is unknown. Nonetheless, comparisons with patterns in the tel1Δ mutant proved informative.

In wild type, smaller chromosomes generate more DSBs per kb than larger ones, revealed as an inverse correlation between Spo11-oligo density and chromosome size (Pan et al. 2011). This anti-correlation is absent in zip3Δ at 5 h, indicating a role for ZMM proteins and thus homolog engagement (Fig. 5A; Thacker et al. 2014). The tel1Δ 4-h time point displayed an attenuated version of the zip3Δ 5-h pattern, i.e., a reduced but not absent anti-correlation (Fig. 5A). Similarly, the responses of subtelomeric, pericentric, and rDNA-proximal regions to the absence of Tel1 at 4 h were qualitatively similar to those in the zip3Δ mutant but in a milder form (Fig. 4Bi,Biv), and there was a positive correlation between the tel1Δ 4-h and zip3Δ 5-h samples for changes at interstitial hotspots (r = 0.53) (Supplemental Fig. S3C). The two data sets also displayed similar correlation strengths and length scales for the local correlation in hotspot behavior (Fig. 5B).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 5.

Relationship between Tel1- and Zip3-dependent control pathways. (A) Negative correlation between Spo11-oligo density and chromosome size requires Zip3 but not Tel1. zip3Δ data are from Thacker et al (2014). (B) Local domains of correlated behavior (as in Fig. 4D). Data for tel1Δ 4 h are reproduced from Figure 4D. See Supplemental Figure S4C for randomized controls. Data from a mutant lacking the transcription factor Bas1 serve as a negative control for this analysis, because this transcription factor affects DSB formation within a subset of individual hotspots rather than coordinately influencing local domains containing multiple hotspots (Zhu and Keeney 2015). (C–E) Spo11-oligo time courses as in Figure 1. Mean ± SD for four pairs of cultures is shown. Data from zip3Δ are reproduced from Thacker et al (2014) for comparison (see also Supplemental Fig. S4B).

However, essentially all similarity with the zip3Δ mutant was erased by 6 h in tel1Δ, whether compared in specific subchromosomal domains (Fig. 4B), in local domains of correlated behavior (Fig. 4D), or across all interstitial hotspots (r = −0.01) (Supplemental Fig. S3C). Furthermore, at 6 h, the negative correlation between Spo11-oligo density and chromosome size mirrored the pattern in wild type (Fig. 5A). The 5-h tel1Δ time point was intermediate for all comparisons but generally more similar to the 6-h time point (Figs. 4B,D, 5A; Supplemental Fig. S3C,D). Thus, Tel1 is dispensable for chromosome size-related control of DSB frequency, although it is needed for timely establishment.

To further characterize the apparent similarities between the tel1Δ 4-h and zip3Δ 5-h DSB landscapes, we examined a tel1Δ zip3Δ double mutant. At time points up to ∼5 h, the double mutant yielded levels of Spo11-oligo complexes that were indistinguishable from the tel1Δ single mutant, i.e., with maximum elevated ∼2.4-fold relative to wild type (Fig. 5C,D). For comparison, we previously found that the peak was elevated by 1.8-fold in the zip3Δ mutant (Fig. 5D; Supplemental Fig. S4B; Thacker et al. 2014). These findings indicate that loss of both Tel1 and ZMM function does not cause an additive increase in DSB numbers early in meiosis, suggesting at least some degree of overlap between these mechanisms of DSB control. At later time points, the tel1Δ zip3Δ double mutant maintained Spo11-oligo complexes at high levels (Fig. 5C,D). This is unlike the tel1Δ single mutant in which Spo11-oligo complexes decreased over these time points but is more reminiscent of the pattern in the zip3Δ single mutant (Supplemental Fig. S4B). Absence of Tel1 does not alleviate the arrest caused by the zip3Δ mutation (Fig. 5E); the block to meiotic progression may explain why amounts of Spo11-oligo complexes remain elevated at later times in the tel1Δ zip3Δ double mutant.

We next generated Spo11-oligo maps from the tel1Δ zip3Δ double mutant (Supplemental Fig. S4A), focusing on the 4-h time point when Spo11-oligo complexes are elevated to similar degrees in the double mutant as in each single mutant (Fig. 5D). Analysis of these maps indicated that the epistasis relationships between the tel1Δ and zip3Δ mutations are complex. There was no longer a residual anti-correlation between Spo11-oligo density and chromosome size in the double mutant, indistinguishable from the zip3Δ single mutant at 5 h (Fig. 5A). Similarly, the double mutant more closely mirrored the stronger patterns seen in the zip3Δ single mutant for hotspots in the subtelomeric and rDNA-proximal regions but less so in pericentric regions (Fig. 4Bv). Thus, zip3Δ is epistatic to tel1Δ for these features of the DSB landscape.

For other features, however, Spo11-oligo maps revealed largely nonepistatic relations instead. For interstitial hotspots, the log-fold change of Spo11-oligo counts in the double mutant correlated best with the zip3Δ single mutant map (r = 0.79) (Supplemental Fig. S3C), indicating that most of the change in the double mutant can be accounted for by the zip3Δ mutation. However, there was also a stronger correlation of the tel1Δ 4-h and 5-h data sets with tel1Δ zip3Δ than with the zip3Δ single mutant (r = 0.62 vs. 0.53 for comparisons involving tel1Δ at 4 h; r = 0.51 vs. 0.30 for comparisons involving tel1Δ at 5 h) (Supplemental Fig. S3C), and the tel1Δ 6-h map now displayed a modest correlation with the tel1Δ zip3Δ map that was not seen with the zip3Δ map (r = 0.18 vs. −0.01) (Supplemental Fig. S3C). These observations suggest that the DSB landscape in the double mutant combines changes that are shared in common between the two single mutants with features that are unique to either tel1Δ or zip3Δ. Furthermore, neighboring hotspots displayed more strongly correlated behavior that spread over longer distances (up to ∼200 kb) in the tel1Δ zip3Δ double mutant than in either single mutant (Fig. 5B). This suggests that effects of tel1Δ or zip3Δ mutations are additive (nonepistatic) for this feature of the DSB landscape. Possible explanations for this complex behavior are provided in the Discussion.

Interactions between DSB sites in cis along chromosomes

Inserting a strong artificial hotspot suppresses DSB formation nearby when measured on a population average (Wu and Lichten 1995; Xu and Kleckner 1995). This suppressive effect decays with distance but has been documented to extend ≥60 kb (Wu and Lichten 1995; Jessop et al. 2005; Robine et al. 2007). Tel1-dependent DSB interference operates over a similar size scale (see Introduction) (Garcia et al. 2015), so interference can, in principle, explain the local effect of strong artificial hotspots, assuming that interference includes both sister chromatids. Site-specific endonucleolytic DSBs (i.e., not made by Spo11) also suppress Spo11-generated DSBs nearby (Fukuda et al. 2008), consistent with an inhibitory signal that spreads after DSB formation. Importantly, however, it remains unknown whether the effect of artificial Spo11 hotspot insertion is the same as DSB interference: To meet the formal definition of interference, occurrence of multiple events on the same chromosome must be shown to occur less often than expected from the product of the single-event frequencies, but this has not been tested. An alternative possibility is that artificial hotspot insertion diminishes the intrinsic DSB potential of nearby hotspots, for example, by altering large-scale chromosome structures.

We reasoned that, if artificial hotspot insertion works solely via DSB interference, then the DSB suppressive effect should require Tel1. To test this prediction, we inserted the arg4-URA3 artificial hotspot (Wu and Lichten 1995) at either of two positions in the genomes of otherwise wild-type or tel1Δ cells: at HIS4 on Chr III or POR2 on Chr IX. We then examined the effects of the hotspot insertions in cis by compiling Spo11-oligo maps from 4-h meiotic cultures.

As previously demonstrated (Wu and Lichten 1995), hotspot insertion on Chr III in a TEL1⁺ background caused a substantial drop in DSB formation nearby: Spo11-oligo density was decreased by fourfold or more in hotspots flanking the insertion site (Fig. 6A,B). Suppression decayed with distance, extending ∼60 kb to the left (nearly to the telomere) and ∼130 kb to the right, extending past the centromere and even spanning, albeit more weakly, the DSB-cold domain from CEN3 to MAT. In tel1Δ, DSBs were suppressed to a similar degree from the insertion site to the left and from the insertion site to the centromere, but the CEN3–MAT region was no longer suppressed (Fig. 6A,B). We conclude that, for this insertion site, Tel1 is dispensable for DSB suppression over distances up to ∼60 kb. We further conclude that the suppressive effect of this hotspot insertion is not simply a consequence of DSB interference, because interference is fully Tel1-dependent when measured in this same region of Chr III (Garcia et al. 2015). We will use the term “hotspot competition” to distinguish the Tel1-independent phenomenon from Tel1-dependent DSB interference (Keeney et al. 2014). Further studies are needed to test whether DSB interference explains the Tel1 dependence of DSB inhibition over longer distances in the CEN3–MAT interval.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 6.

Communication between hotspots in cis along chromosomes. (A–C) Effect of artificial hotspot insertions. A hotspot construct containing arg4 and URA3 on a bacterial plasmid backbone (Wu and Lichten 1995) was inserted within the HIS4 gene on Chr III or the POR2 gene on Chr IX (green triangles and vertical dotted lines). (A,C) Spo11-oligo maps (smoothed with a 201-bp sliding Hann window) are shown for wild-type and tel1Δ strains with and without the insertion. (B) Each point shows, for each 5-kb segment moving outward from the insertion points, the log-fold change in the Spo11-oligo count caused by the presence of the artificial hotspot insertion (position of insertions denoted by red triangles). Curves were drawn manually to highlight the trends in the data. Gray shading encloses ± 1 SD for all 5-kb bins genome-wide (excepting Chr III and IX) to illustrate the range of sampling noise in this analysis. (D,E) Signature consistent with Tel1-dependent DSB interference around the strongest natural hotspots. In panel D, each point displays the log-fold change in Spo11-oligo count (4-h tel1Δ vs. wild type) for hotspots located the indicated distance away from either the GAT1 (top) or CCT6 (bottom) promoter hotspots. Horizontal dashed lines indicate genome-average change or no change. In panel E, boxplots summarize the log-fold change for hotspots located within 10 kb of the next 28 hottest hotspots (rank numbers 3–30) or within 10 kb of the middle third of hotspots (rank numbers 1304–2608). (*) P = 0.0011 for a one-sided Wilcoxon rank-sum comparison to all hotspots genome-wide, (n.s.) not significant (P = 0.61). Horizontal dotted line indicates genome-average change. Boxplots are as defined in Figure 4B.

Chr IX gave a strikingly different result. In the TEL1⁺ background, this hotspot insertion caused a substantial drop in Spo11-oligo counts for hotspots immediately flanking the insertion site, yielding a similar magnitude of suppression as on Chr III (Fig. 6B,C). However, the suppressive effect covered a much smaller zone, only ∼20 kb to the left and ∼15 kb to the right. The more limited distance of suppression may be, in part, a consequence of the artificial hotspot being ∼4.5-fold weaker at the Chr IX location, based on normalized Spo11-oligo counts in TEL1⁺ (Supplemental Fig. S5; Supplemental Table S4). Context dependence for DSB frequency within this insert was documented previously (Borde et al. 1999). In the tel1Δ background, the hotspots closest to the insertion (<2 kb away) were still suppressed, but the next hotspots further away (≥5 kb) were unaffected by the insertion (Fig. 6B,C). Thus, the zone of DSB suppression is much smaller at this location, and Tel1 is essential beyond the most immediate vicinity of the new hotspot. More broadly, these results show that the local effects of a hotspot insertion can be context-dependent in terms of both the distance over which Tel1-independent hotspot competition operates and the degree to which Tel1-dependent processes (presumably DSB interference) contribute.

A natural hotspot in the HIS4 promoter can exert a suppressive effect on an artificial one inserted nearby (Fan et al. 1997), and adjacent natural hotspots can be suppressed when DSBs are artificially targeted to novel sites by fusing Spo11 to the DNA binding domain of Gal4, illustrating functional interactions between natural yeast DNA sequences (Robine et al. 2007). However, other studies found no evidence of such interactions between two natural hotspots near HIS2 (Bullard et al. 1996) or between Bas1- or Ino4-regulated hotspots and their neighboring hotspots (Zhu and Keeney 2015). It has thus been unclear to what extent suppressive interactions occurring between natural hotspots can contribute to population-average DSB frequencies. To address this question, we asked if absence of Tel1 causes a greater than average increase in DSBs for hotspots located near strong natural hotspots, because weak hotspots should be more highly suppressed by Tel1 signals emanating from their stronger neighbors.

To test this prediction, we examined GAT1, the strongest natural hotspot in the SK1 strain background (Pan et al. 2011). At 4 h in the tel1Δ mutant, this hotspot displayed a Spo11-oligo count that was slightly below the genome-average increase (Fig. 6D; Supplemental Fig. S6A). In contrast, nearly all of its neighbors within ∼10–20 kb on either side experienced above-average increases in Spo11-oligo levels; hotspots further away were not consistently increased above the genome average (Fig. 6D). Similarly, most hotspots in the vicinity of another exceptionally strong natural hotspot, CCT6, also increased more than average (Fig. 6D; Supplemental Fig. S6B), as did hotspots located up to ∼10 kb away from the next 28 hottest hotspots considered as a group (Fig. 6E). This pattern meets the prediction for Tel1-dependent inhibitory interactions between strong hotspots and their weaker neighbors and is thus consistent with being a consequence of DSB interference. Notably, however, we detected no reliable hotspot-suppression signature for hotspots that are less active than this elite group (Fig. 6E).

Taken together, these findings support the hypothesis that Tel1-dependent DSB interference can shape the population-average DSB landscape but that it provides a quantitatively modest contribution that is readily detected only at relatively strong hotspots and only over short distances. Our findings also demonstrate a Tel1-independent component to interactions between artificial hotspots and their surroundings. This hotspot competition is context-dependent, as insertions on Chr IX and Chr III behave differently.