ABSTRACT
Meiotic recombination is initiated by the formation of programmed DNA double-strand breaks (DSBs) catalyzed by the Spo11 protein. DSBs are not randomly distributed along chromosomes. To better understand factors that control the distribution of DSBs in budding yeast, we have examined the genome-wide binding and cleavage properties of the Gal4 DNA binding domain (Gal4BD)-Spo11 fusion protein. We found that Gal4BD-Spo11 cleaves only a subset of its binding sites, indicating that the association of Spo11 with chromatin is not sufficient for DSB formation. In centromere-associated regions, the centromere itself prevents DSB cleavage by tethered Gal4BD-Spo11 since its displacement restores targeted DSB formation. In addition, we observed that new DSBs introduced by Gal4BD-Spo11 inhibit surrounding DSB formation over long distances (up to 60 kb), keeping constant the number of DSBs per chromosomal region. Together, these results demonstrate that the targeting of Spo11 to new chromosomal locations leads to both local stimulation and genome-wide redistribution of recombination initiation and that some chromosomal regions are inherently cold regardless of the presence of Spo11.
In most sexually reproducing organisms, homologous recombination is a prominent feature of meiosis, which creates the genetic diversity of the meiotic products by promoting a safe exchange of DNA information between the paternal and maternal chromosomes. As importantly, the crossovers maintain a physical connection between the homologs which ensures their proper disjunction at the first meiotic division (55). Therefore, understanding how the cells control the process of meiotic recombination is important because defects in the number or the localization of recombination events lead to failure in homolog disjunction or unviable gametes. For instance, errors in chromosome segregation are correlated with unusual crossover positions in many cases of human maternally derived trisomy 21 (27, 28). In particular, segregation defects are observed when a crossover occurs close to a telomere, probably because the structure is less efficient at providing a strong physical link between homologs (36, 43). Also, crossovers very close to the centromere lead to the premature loss of sister chromatid cohesion and nondisjunction at meiosis II (26, 42). Not surprisingly, then, the meiotic recombination process is tightly controlled from initiation to completion stages.
Once recombination is initiated, a decision is made to channel the early intermediates along a pathway ending in crossover formation or in noncrossover events. This step is influenced both by “crossover interference,” in which a crossover in one region makes it unlikely that another will occur nearby (18), and by “crossover homeostasis,” which regulates the crossover/noncrossover ratio to ensure a minimal amount of crossover per chromosome (34).
Before this stage, the frequency and the localization of the initiation events are the earliest determinants of how meiotic recombination events are distributed along the chromosomes. Extensive studies performed with Saccharomyces cerevisiae have shown the heterogeneous distribution of the initiating DNA double-strand breaks (DSBs) along the chromosomes and the >100-fold variation in cleavage frequency from site to site (39). These findings are consistent with the nonrandom distribution of recombination along the chromosomes observed in all organisms. The factors that determine whether a specific region is prone to DSB formation are not well understood at the molecular level. The primary sequence is not the main determinant of DSB formation since a reporter construct shows various DSB frequencies and recombination rates depending on where it is inserted into the genome (10, 52). Several rules regarding the distribution of DSBs have nevertheless been described. First, DSBs always occur in open chromatin regions, mainly in promoter-containing regions (4, 53). Second, DSBs are formed preferentially in the chromatin loops, as opposed to the loop basis in which cohesins are located (7, 13). Third, DSB frequencies are generally low in a 20-kb region around centromeres (9, 12) and the centromere itself has a strong inhibitory effect on meiotic recombination initiation (30). Fourth, the rate of meiotic recombination is also usually low close to the natural chromosome ends (3, 46) and DSB frequency is very low up to 40 kb from a telomere (9, 12). This effect may be exaggerated by the fact that these DSB measurements were made with rad50S or sae2Δ mutants, which accumulate unresected DSBs and in which DSB formation is specifically reduced in late-replicated regions (8). Finally, another factor influencing DSB formation at one site is the proximity to another DSB site (11, 19, 52, 54), although this factor may not be a general rule (15).
The catalytic activity for meiotic DSB formation is carried by the widely conserved Spo11 protein (5, 23). Besides Spo11, at least nine additional DSB proteins are absolutely required for DSB formation (reviewed in reference 24). Recent cytological and chromatin immunoprecipitation (ChIP) analyses have begun to dissect the chromatin association of these proteins on meiotic nucleus spreads and their requirement for Spo11 association with DSB hot-spot regions (24). However, how DSB sites are selected and how Spo11 is recruited to the chromatin of the DSB region remain to be understood.
To address these issues, we previously reported that a fusion between the Gal4 DNA binding domain and Spo11 (Gal4BD-Spo11) yields a protein that is able to introduce DSBs and stimulate meiotic recombination in the naturally cold GAL2 promoter, which contains Gal4 consensus binding sequences (CGGN11CCG, where N11 represents 11 various nucleotides) (38). Thus, the normal recruitment of Spo11 to chromatin can be bypassed by the Gal4BD moiety of the fusion protein. However, all the DSB proteins are still required for Gal4BD-Spo11 DSB formation at GAL2, showing that they are all important for Spo11's ability to create a DSB and not only for the recruitment of Spo11 to the DSB site. Importantly, it was shown that Gal4BD-Spo11 also cleaves the three natural Spo11 DSB sites examined.
Here, to determine to what extent the genome-wide Gal4BD-Spo11 DSB profile was modified and to investigate the mechanisms that regulate meiotic DSB distribution, we have further exploited the in vivo properties of the Gal4BD-Spo11 protein to probe the entire genome. We show that DSB distribution is profoundly remodeled upon the introduction of strong new Gal4BD-Spo11 DSBs and that DSB cleavage at a subset of Gal4BD-Spo11-bound sites remains subject to repressing position effects. Globally, the modification of Spo11 cleavage sites leads to a genome-wide redistribution of meiotic double-strand breaks without decreasing the meiotic product viability.
MATERIALS AND METHODS
Yeast strain constructions, culture, and sporulation.Details of plasmid and yeast strain constructions are described in the experimental procedures in the supplemental material. All yeast strains used are from the SK1 genetic background (20). Their genotypes are indicated in Table 1. Sporulation was performed as described previously (14).
Strains used in this studya
DNA electrophoresis and Southern blotting.Genomic DNA was prepared either by following a conventional method for electrophoresis after digestion with restriction enzymes (14) or by using agarose plugs for pulsed-field gel electrophoresis (10). After Southern blotting and hybridization with a radiolabeled probe, the DSB frequency was measured as described previously (9). Restriction digests and probes used for the various loci analyzed are described in the experimental procedures in the supplemental material.
Chromatin immunoprecipitation.For DSB mapping, 50 ml of rad50S mutant cells (about 109 cells) after 6 h in sporulation medium were processed for chromatin immunoprecipitation without cross-linking to select for covalently attached Spo11 (9). For Gal4BD-Spo11 binding, 40 ml of meiotic cells (8 × 108 cells) were treated with 1% freshly prepared formaldehyde for 15 min at room temperature and then 125 mM glycin for 5 min. In both cases, cells were then washed and chromatin immunoprecipitation and DNA purification were performed as described previously (40). Antibodies used were mouse anti-Myc (8 μg; clone 9E10) for Spo11-Myc13 or Gal4BD-Spo11-Myc13 covalent binding and mouse anti-Gal4BD (4 μl; clone 2GV3; Euromedex) for Gal4BD-Spo11 binding. Multiplex PCR was performed on 1/10 of the immunoprecipitate or 1/25,000 of the whole-cell extract, and results were analyzed as described previously (9). Primer sequences are available upon request. For hybridization on microarrays, two-thirds of the immunoprecipitated DNA or 1/4,000 of the DNA from the whole-cell extract was amplified by random primer extension followed by PCR amplification, incorporating amino-allyl dUTP for subsequent dye coupling (9).
Microarray hybridization.The S. cerevisiae open reading frames (ORFs) were reamplified using a pair of universal primers from 6,216 full-length ORFs amplified from genomic DNA by Research Genetics. The intergenic regions were amplified using yeast genomic DNA and 6,220 pairs of specific primers from Research Genetics. After purification by isopropanol precipitation, the sizes and concentrations of the PCR products were measured by agarose gel electrophoresis. DNA was resuspended in a Tris-EDTA-dimethyl sulfoxide (50/50) spotting buffer. DNA printings were performed on Corning Ultra Gaps II slides by using a Microgrid II arraying robot (Biorobotics). Amplified DNA was labeled by coupling with Cy3 or Cy5 and hybridization at 63°C, and washes were done under standard conditions (http://www.derisilab.ucsp.edu ).
Data analysis.Array slides were scanned using a 4000B scanner (Axon Instruments) and the images analyzed using the Genepix Pro 5.0 software. Extracted numeric data were transformed as described previously (9). We used the algorithm MANOR (37) to estimate and correct a spatial trend on each chip. Data for ORFs and intergenic regions were normalized separately by fitting the median to 0. Probes corresponding to repeated sequences present at least twice in the genome were removed from the analysis by using the algorithm BLAST (1). A list of these repeated elements is available upon request.
Each experiment was done from three independent cultures (except DSB measurements for the Spo11-Myc pADH1-Gal4BD strain, which were done twice). We computed a weighted average for the replicates, with the standard deviation of the chip as the weight. To determine the enriched probes, we computed the median percentile rank (MPR) among replicates, varying from 0 to 1 for each probe (32), and after examination of the MPR distributions, we chose two criteria such that the estimated numbers of true positives and false positives were optimized. The first defines a stringent category containing most of the true DSB (or binding) sites and very few false positives. We defined primary sites as those sites composed of at least two adjacent chromosomal features with an MPR greater than 0.9 (DSB experiments) or 0.97 (binding experiment). The second category defines secondary sites as those sites composed of all probes with an MPR of >0.8 (DSB experiments) or >0.9 (binding experiment) and not included in a primary site.
Two sites were considered common to two experiments if they shared at least one probe. Finally, sites were ranked according to the highest-log-ratio probe within the site.
Microarray data accession number.The microarray data described herein are available at Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo ) under the accession number GSE5884.
RESULTS
We have mapped both Gal4BD-Spo11 binding and DSB sites by ChIP followed by hybridization on a microarray (ChIP-chip), using an array that covers the whole budding yeast genome at 1- to 2-kb resolution. To compare results from different ChIP-chip experiments, we defined two categories of sites: robust sites were defined using a stringent threshold and secondary sites were defined using less stringent criteria (detailed in Materials and Methods).
Genome-wide mapping of Gal4BD-Spo11 binding sites.First we mapped the binding sites of the Gal4BD-Spo11 protein. In order to differentiate between the sites that are normally bound and cleaved by Spo11 and the new binding sites driven by Gal4BD, we took advantage of the constitutive Gal4BD-Spo11 expression behind the ADH1 promoter to map binding sites at the beginning of sporulation, about 4 h before DSB formation.
We identified 44 binding sites for Gal4BD-Spo11 (listed in Table S1 in the supplemental material; indicated along chromosomes depicted in Fig. 1A; see Fig. S1 in the supplemental material). The fusion protein is still able to recognize the natural Gal4 binding sites, since only 3 of the 22 strongest targets determined previously for the Gal4-Myc protein in galactose medium (P < 10−5) (41) are not included in our set. Then, using the recently released sequence of our S. cerevisiae background strain, SK1 (sequence data were produced by the Saccharomyces Genome Resequencing Project Sequencing Group at the Sanger Institute and obtained from ftp://ftp.sanger.ac.uk/pub/dmc/yeast ), we examined the presence of a Gal4 binding sequence in the Gal4BD-Spo11 binding regions. We found that half of these regions contain at least one occurrence of the Gal4 binding sequence, preferentially (54%) located in intergenic regions at a frequency that is significantly higher than the general distribution of the Gal4 binding sequence (chi-square P value, <0.036). This is most likely due to the better chromatin accessibility of intergenic regions.
Comparison of Gal4BD-Spo11 binding and cleavage sites. (A) Genome-wide map of binding and DSB sites of Gal4BD-Spo11 and Spo11 DSB sites. Each site is indicated as a vertical bar. Centromeres are indicated as blue circles. The black arrows point to the 16 Gal4BD-Spo11 DSB targeted sites, which correspond to Gal4BD-driven binding sites. Gal4BD-Spo11 binding, data from strain ORD7366 at 0 h; Gal4BD-Spo11 DSB, data from strain ORD7392 at 6 h without cross-linking; Spo11 DSB, data from strain ORD8127 at 6 h without cross-linking. Chr., chromosome. (B) Venn diagram representing the overlap among the sites determined under the different experimental conditions. Sites indicated in bold red are targeted sites (see the text). (C) Examples of Gal4BD-Spo11 binding regions. Data are from the same experiments as those described for panel A. ChIP log2 enrichment values are represented according to the chromosome coordinates. Consensus Gal4 binding sequences are indicated as blue (intergenic) or green (inside an ORF) dashed lines. For Gal4BD-Spo11 binding sites, red vertical lines indicate enriched probes included in a binding site. For the DSB sites, black vertical lines indicate enriched probes included in a DSB site and red vertical lines fill the same criteria and in addition coincide with a Gal4BD-Spo11 binding site. In all graphs, gray lines indicate unenriched probes.
The finding that half of the enriched regions had no Gal4 binding sequence was more surprising. Among them, 11 correspond to unique interstitial sequences, 8 are unique subtelomeric sequences (<25 kb from the end), and 3 others are repeated elements (two Ty1 long terminal repeats and one tRNA gene). Noteworthily, binding regions with no Gal4 binding sequence were also observed in the previous study of the Gal4-Myc protein by Ren et al. (41; supplemental material). Two explanations can be invoked for the enrichment of such regions. Gal4BD-Spo11 (like Gal4) may have binding affinity for other DNA sequences. However, we did not find any motif significantly represented among the sequences of the 22 sites with no Gal4 binding sequence by using the Meme software (2). An alternative possibility is that some of these sites might be recognized by Spo11 itself. We tested this hypothesis by mapping the premeiotic binding sites of pADH1-Spo11-Myc (unpublished results). We found that only 2 of the 22 Gal4BD-Spo11 binding sites with no upstream activation sequence correspond to the binding sites of Spo11 itself. Together, the above-cited results indicate that Gal4BD targets the Gal4BD-Spo11 fusion protein to a large number of chromatin sites within the yeast genome and thus provides a powerful method to modify the distribution of the DSB sites and monitor the effect of the chromosomal context, as described below.
Genome-wide mapping of DSB induced by Gal4BD-Spo11.Next, we used ChIP-chip to map DSBs from meiotic rad50S mutant cells expressing either Spo11-Myc or Gal4BD-Spo11-Myc without prior cross-linking since in this repair-deficient strain background, Spo11 remains covalently attached to the DSB ends (23). We first determined 239 DSB sites for the Spo11 protein (listed in Table S1 in the supplemental material; Fig. 1A; see Fig. S2 in the supplemental material). There was a good overlap with our previously determined Spo11-hemagglutinin DSB sites (9), since among the 100 strongest Spo11-hemagglutinin DSB sites, 71 are detected as DSB sites and 20 are detected as secondary DSB sites in the present study.
We next determined 191 DSB sites for the Gal4BD-Spo11 protein (see Table S1 and Fig. S3 in the supplemental material; Fig. 1A). As expected and as illustrated in Fig. 1B, many of the GalBD-Spo11 and Spo11 DSB sites are common (116 sites) and among the 66 Gal4BD-Spo11-specific DSB sites, we noted that 36 are secondary Spo11 DSB sites. Thus, 152/191 (80%) of Gal4BD-Spo11 DSB sites occur at natural Spo11 DSB sites.
The overlap between the Gal4BD-Spo11 binding sites and the DSB sites is summarized in Fig. 1B. Nine sites are specific to Gal4BD-Spo11, and seven are DSB sites shared by Spo11 and Gal4BD-Spo11. We will hereinafter refer to these 16 Gal4BD binding sites cleaved by Gal4BD-Spo11 as targeted sites. They include the well-known Gal4 targets the GAL1, GAL2, GAL7, GAL10, and GAL80 genes. When coinciding with a Gal4BD binding site, the natural DSB sites have a higher rank in the Gal4BD-Spo11 DSB experiment, suggesting that they are more frequently cleaved (Fig. 2A). The resolution of our data (1 to 2 kb) does not allow us to say if DSBs stimulated by Gal4BD-Spo11 at these sites occur exactly at the same nucleotide positions as those induced by natural Spo11.
Global long-distance effects of DSB targeting. (A) List of the 16 targeted sites. The extreme left and right chromosomal features are used to name each site. Numbers in the columns indicate the ranks of the sites in each experiment. +/− indicates a secondary site, and − indicates no site. For the analysis presented in panel B, the top 12 targeted sites were considered. The Gal4 binding sequence is CGGN11CCG. (B) Global effect of targeted DSBs on surrounding DSBs. The red line corresponds to the ratio of the cumulated number of DSB sites in the Gal4BD-Spo11 experiment to the cumulated number of DSB sites in the Spo11 experiment as a function of the distance around the 12 targeted sites shown in panel A, calculated every kilobase. The solid blue line represents the mean of results from 50 permutations computed as described above but for 12 nontargeted Gal4BD-Spo11 DSB sites picked at random (on the same chromosomes as the targeted sites). The dotted blue lines correspond to the confidence interval of 2 standard deviations computed for the 50 permutations.
Among the 30 Gal4BD-Spo11 DSB sites unrelated to Spo11 DSBs, 17 were secondary Gal4BD-Spo11 binding sites. One possible explanation for the 13 remaining sites would be that they become accessible to Gal4BD-Spo11 binding later in meiosis, for example, upon meiosis-specific gene induction. This might be the case for 5 out of these 13 binding sites, which occur in the promoters of meiosis-specific genes.
The expression of the Gal4BD-Spo11 construct under the constitutive ADH1 promoter raised the possibility that the variation in the DSB profile in the Gal4BD-Spo11 strain resulted from the construct's early expression and/or overexpression. We thus performed two additional control experiments. We first mapped genome-wide DSBs in a strain expressing Spo11-Myc13 from the ADH1 promoter. Among the 285 DSB sites detected, 167 corresponded to Spo11 DSB sites and 93 to secondary Spo11 DSB sites, and importantly, out of the 150 strongest pADH1-Spo11 DSB sites, only one was not detected at all in the endogenous Spo11 experiment. Thus, very few of the pADH1-Gal4BD-Spo11 DSB sites were due to the overexpression of Spo11. Second, we mapped DSBs made by Spo11-Myc in a strain expressing pADH1-GAL4BD alone. None of the 16 targeted Gal4BD-Spo11 DSB sites were stronger than Spo11 DSB sites in this experiment, and among the Gal4BD-Spo11 DSB sites corresponding to secondary Gal4BD binding sites, only one (YCR060W) had a much stronger enrichment in the experiment with Gal4BD than in that with Spo11 (Table S1 in the supplemental material). This indicates that Gal4BD overexpression has no indirect effect and that Gal4BD must be fused with Spo11 in order to induce DSBs at its binding sites.
Long-range cis effects of enhanced DSBs.Despite the substantial overlap between Spo11 and Gal4BD-Spo11 DSB sites, there is a relatively high proportion (24%) of the natural DSBs that are no longer cleaved by Gal4BD-Spo11. Indeed, among the 123 DSB sites cleaved only by Spo11 (Fig. 1B), 61 are not even among our 740 secondary Gal4BD-Spo11 DSB sites. Close examination of specific loci strongly suggests that this finding can be explained at least in part by cis inhibitory effects of Gal4BD-Spo11 DSBs on surrounding natural Spo11 DSBs. Indeed, strong Gal4BD-Spo11 DSBs at GAL2 and MLF3 were accompanied by reduced DSB formation over several kilobases in surrounding regions compared with DSB formation by Spo11 (Fig. 1C). To appreciate the generality of this cis effect, we examined the behavior of the regions surrounding the 12 strongest targeted sites (Fig. 2A). In these regions, DSB formation was significantly reduced compared with that in regions not affected by Gal4BD binding (Fig. 2B). The cumulative reduction effect around targeted sites increased up to 60 kb around the sites and then decreased with increasing distance from the targeted sites (Fig. 2B).
To deepen the analysis of this long-range effect on DSB distribution, we examined the chromosome XIII region around GAL80, which was a strong Gal4BD-Spo11 targeted site and a weaker Spo11 DSB site (Fig. 2A and 3A). In the spo11, pADH1-spo11, and spo11 pADH1-Gal4BD strains, DSBs occurred only in the promoter of the gene adjacent to GAL80 at a frequency of 1%, whereas in the pADH1-Gal4BD-spo11 strain, DSBs occurred exclusively in the GAL80 promoter at a frequency of 10%, near the Gal4 binding sequence (Fig. 3B). This difference in position may reflect local competition between these two adjacent regions. On a larger scale, the chromosome XIII ChIP-chip profile indicates that the number of Gal4BD-Spo11 DSBs around GAL80 was lower than that of Spo11 DSBs (Fig. 3A). Pulsed-field gel electrophoresis analyses confirm that the strongly stimulated GAL80 DSB was accompanied by a reduction in the number of surrounding DSBs, from the left end of chromosome XIII to the centromere on the right (Fig. 3C and D). In the Gal4BD-Spo11 strain, DSBs in the two regions of the left arm surrounding GAL80 occurred at a frequency 50 to 80% that of Spo11 DSBs. Consequently, the total DSB frequencies in the entire left arm were not significantly different in the Spo11 and Gal4BD-Spo11 strains (Fig. 3E). To determine if these differences in the distributions of the Gal4BD-Spo11 and Spo11 DSBs may be due to an impaired ability of Gal4BD-Spo11 to cleave some natural sites, we inactivated by site-directed mutagenesis the Gal4 binding sequence in the GAL80 promoter. This inactivation resulted in a decrease of GAL80 promoter DSB frequency, from 10 to 3%, and DSB formation occurred about 130 bp away in the same intergenic region (gel not shown). This lowered GAL80 promoter cleavage by Gal4BD-Spo11 was accompanied by an increase of DSB formation (between 1.1- and 1.9-fold) in the surrounding regions (Fig. 3F). This result shows that reduced DSB cleavage by Gal4BD-Spo11 away from GAL80 is due to this strongly targeted DSB. Finally, we confirmed the long-range inhibitory cis effect on DSBs surrounding the targeted GAL2 promoter DSB site on chromosome XII by pulsed-field gel electrophoresis with rad50S mutant cells (Fig. 3G, left panel). Interestingly, based on visual inspection of the pulsed-field gel, a DSB reduction was also observed around the targeted GAL2 DSB in another DSB repair-defective mutant, dmc1Δ, which accumulates hyperresected DSBs after the removal of Spo11 (6) (Fig. 3G, right panel).
Inhibitory cis effects of GAL80 and GAL2 DSBs on surrounding regions. (A) Binding and DSB profiles along chromosome XIII. Gal4 binding sequences and log2 ratios are represented as described for Fig. 1C. The red line corresponds to smoothed data with a 5-kb window. (B) DSBs at the GAL80 promoter detected by Southern blotting after genomic DNA extraction at the indicated times during meiosis and digestion with BamHI. The positions of ORFs are shown as vertical arrowheads indicating transcriptional sense. The star indicates the Gal4 binding sequence, and the arrow indicates the Spo11 DSB. Lanes 1 to 3, strain ORD8127; lanes 4 to 6, strain ORD8257; lanes 7 to 9, strain ORD7392; lanes 10 to 12, strain ORD8234; lanes 13 to 15, strain ORD8270. (C) Results from pulsed-field gel electrophoresis of undigested DNA analyzed with a chromosome XIII left-end probe. The dotted line indicates the approximate extent of DSB reduction surrounding the GAL80 DSB. Lanes 1 to 3, strain ORD7366 (GAL4BD-SPO11); lane 4, strain ORD7392 (GAL4BD-SPO11-MYC13); lanes 5 to 7, strain ORD8213 (SPO11); lane 8, strain ORD8127 (SPO11-MYC13). The blue circle indicates a centromere. (D) Graph showing the traces of phosphorimager signal from Gal4BD-Spo11 (red line, from lane 3 in panel C) and Spo11 (black line, from lane 7 in panel C). Sizes are according to the size standards (an XbaI digest of bacteriophage λ DNA and λ DNA concatemers) run along the samples shown in panel C. (E) Quantification of DSBs measured on the left arm and at YMR056C on the right arm at 6 h in two independent experiments. Values shown are mean percentages of DSB ± standard deviations. (F) Results from pulsed-field gel electrophoresis of undigested DNA analyzed with a chromosome XIII left-end probe. Lanes 1 to 3, strain ORD7392; lanes 4 to 6, strain ORD8283, which has a mutated Gal4 upstream activation sequence in the GAL80 promoter. The blue circle indicates a centromere. (G) Results from pulsed-field gel electrophoresis of undigested DNA extracted from cells at the indicated times during meiosis and analyzed with a probe specific for the left end of chromosome XII. The arrows indicate DSBs in the GAL2 promoter, and the dotted line indicates the approximate extent of surrounding DSB reduction. Lanes 1 to 3, strain ORD8213; lanes 4 to 6, strain ORD7366; lanes 7 to 8, strain ORD7354; lanes 9 to 10, strain ORD5854. Size markers used were an XbaI digest of bacteriophage λ DNA and λ DNA concatemers run along with the samples.
Factors determining DSB formation among the Gal4BD-Spo11 binding sites.Only 16 out of the 44 Gal4BD-Spo11 binding sites produced a DSB (Fig. 1B). This demonstrates for the first time that, in a strain without a mutation in any known DSB protein, Spo11 association with chromatin is not always sufficient for DSB formation. This finding suggests that additional factors locally control the cleavage step. We examined several factors that may influence cleavage at a Gal4BD-Spo11-bound site.
Ten out of the 13 targeted sites with a Gal4 consensus sequence had this sequence located in an intergenic region. DSB targeting thus follows the property of natural DSB sites, which in their vast majority are located in intergenic regions (4).
In agreement with results from previous studies (9, 12), our data show very weak average DSB levels around the centromeres (∼20 kb) in the presence of Spo11 as well as in the presence of Gal4BD-Spo11 (see Fig. S4 in the supplemental material). In addition, in the 20-kb regions around centromeres, there were significantly lower numbers of DSB sites induced by Spo11 (two sites; P < 0.0001) and Gal4BD-Spo11 (three sites; P < 0.001) than in the rest of the genome. However, the small number of Gal4BD binding sites within 20 kb from a centromere (YGL007W, CEN4, and GAL3) made it unlikely to see a global change between the two proteins when looking at all centromeric regions. These three sites behaved differently. YGL007W, located 14 kb to the left of CEN7, gave rise to a Gal4BD-Spo11 DSB (Fig. 2A), whereas CEN4 and GAL3, located 14 kb to the right of CEN4, did not (Fig. 1C). GAL3 was the second strongest Gal4BD-Spo11 binding site. The absence of DSB at GAL3 was verified both by ChIP-PCR (Fig. 4A) and by Southern blotting (Fig. 4B). DSB formation at GAL3 was also barely detectable in a dmc1Δ Gal4BD-Spo11 strain (Fig. 4C). Thus, we conclude that the binding of Gal4BD-Spo11 at GAL3 is not sufficient for DSB formation.
Binding of Gal4BD-Spo11 and absence of DSBs at GAL3. (A) ChIP analysis of pADH1-Gal4BD-Spo11 binding in formaldehyde-treated cells and ChIP analysis of covalent Spo11-Myc attachment to DSBs at the indicated times during meiosis. Lanes 1 to 4, strain ORD7366; lanes 5 to 6, strain ORD7392; lanes 7 to 8, strain ORD8257; lanes 9 to 10, strain ORD8270. Samples were analyzed by PCR with the indicated primer pairs. W, whole-cell extract; Ip, DNA purified after chromatin immunoprecipitation; C, control YCR013C site. (B) Southern blot analysis of DSBs at GAL3. The positions of ORFs are shown as vertical arrowheads indicating transcriptional sense. On the left is displayed a map of the region of the natural TRP1 locus (homozygous in the SPO11 strain, lanes 1 to 3, and heterozygous in the GAL4BD-SPO11 strain, lanes 10 to 12), and on the right is displayed a map of the region of the trp1::hisG locus where a TRP1 plasmid carrying the indicated gene has been integrated (lanes 4 to 15). The dotted line represents the plasmid sequences. The stars indicate the Gal4 binding sequence. Lanes 1 to 3, strain ORD8127; lanes 4 to 6, strain ORD8257; lanes 7 to 9, strain ORD7392; lanes 10 to 12, strain ORD8234; lanes 13 to 15, strain ORD8270. (C) Southern blot analysis of DSB in a GAL4BD-SPO11 dmc1Δ strain (ORD5864). In lanes 1 to 4, DNA was cut with BstEII and probed with a GAL3 probe. In lanes 5 to 8, DNA from the same samples was cut with XbaI and probed with a GAL2 probe. The positions of ORFs and Gal4 binding sequences are represented as in panel B. The black area indicates the hisG sequence.
In the subtelomeric regions (less than 45 kb from a telomere), a lower-than-average DSB level was observed, both for Spo11 and for Gal4BD-Spo11 (see Fig. S4 in the supplemental material). Consistently, none of the nine Gal4BD-Spo11 binding sites located within 45 kb from a telomere gave rise to a DSB (see Table S1 in the supplemental material).
Finally, the 25 binding sites that did not give rise to a DSB appear to be further away from a natural Spo11 DSB site than those where a Gal4BD-Spo11 DSB was detected (median values, 39 kb versus 11 kb; Wilcoxon signed-rank test; P < 0.014). This suggests that Gal4BD-Spo11 DSBs are better induced if binding occurs close to a natural DSB region. On the contrary, no DSB was formed at the binding site YGR183C, located 9 kb from one of the strongest endogenous Spo11 DSB sites, YGR176W (Fig. 1C). In this case, DSB formation at the Gal4BD binding site may be inhibited by competition with the nearby strong natural site.
In summary, it appears that Gal4BD-Spo11 generally follows the trend of the natural DSB sites with respect to chromosomal context but that cold but potentially permissive regions can be warmed up upon the assisted binding of Spo11.
Targeting of Gal4BD-Spo11 DSBs in a reporter cassette and position effect.Our ChIP-chip data show that Gal4BD-Spo11 association with chromatin is not always sufficient for DSB formation. It was previously shown that the chromosomal context has a strong influence on DSB formation, since a reporter cassette containing the URA3 gene and arg4 sequence shows various DSB levels depending on the place of insertion along the chromosome (10, 52); it showed very high levels of DSB formation at his4, located in a hot region of chromosome III, and very low levels when it was inserted at RVS161, in the cold centromere-associated chromosome III region (8). To determine whether the targeting of Gal4BD-Spo11 to binding sites would follow or overcome such position effects, we constructed a similar reporter cassette, URA3-gal2, containing four Gal4 binding sequences (Fig. 5A). This reporter cassette was introduced at the same loci (his4 and RVS161) as the URA3-arg4 cassette in strains expressing either Spo11 or Gal4BD-Spo11. In the Spo11 strains, frequent DSBs located in the plasmid sequence to the right of gal2 were observed upon insertion at his4 but none were observed upon insertion at RVS161 (Fig. 5B and D). In the strain expressing Gal4BD-Spo11, at his4, DSBs also frequently occurred in the junction plasmid sequence but, in addition, targeted cleavage occurred in the vicinity of the Gal4 binding sequences (Fig. 5B and D). Thus, Gal4BD-Spo11-specific DSBs can be targeted within an ectopic insert containing a Gal4 binding sequence. At RVS161, weak cleavage by Gal4BD-Spo11 occurred in the plasmid sequence, but none occurred in the Gal4 binding sequences, indicating a strong repression of target DSBs (Fig. 5B and D). Together, these results indicate that Gal4BD-Spo11 is subjected to DSB repression occurring in this naturally cold region of chromosome III.
Position effect on Gal4BD-Spo11 DSB targeting. (A) Structure of the URA3-gal2 reporter cassette. Thick horizontal lines indicate plasmid sequences and dotted lines chromosomal sequences. The horizontal arrow represents the transcription start point of gal2. Vertical arrows indicated the positions of DSBs in the GAL2 promoter and in the plasmid. The stars indicate the positions of Gal4 binding sequences; the bar under the URA3-gal2 cassette shows the position of the PCR product used to analyze the ChIP experiments described in the legend for Fig. 6. (B) Southern blot analysis of DSBs in the reporter construct. The integration locus is indicated above each panel. The structure of the reporter cassette is indicated on the right of each panel. Lanes 1 to 3, strain ORD8231; lanes 4 to 6, strain ORD8226; lanes 7 to 9, strain ORD8278; lanes 10 to 12, strain ORD8222. (C) Centromere effect on DSB formation at the RVS161 insert. DSB formation was examined in strains containing the RVS161::HphMX-gal2 construct and the indicated combination of the Spo11 source and the natural (CEN3) or repositioned (cen3Δ) chromosome III centromere. The structure of the HphMX-gal2 cassette is represented along the gel. Lanes 1 to 3, strain ORD8280; lanes 4 to 6, strain ORD8286; lanes 7 to 9, strain ORD8287. (D) Quantification of DSB in the reporter construct. Numbers indicate the DSB/chromosome ratio as mean percentages ± standard deviations, and the numbers of determinations are shown in parentheses.
Gal4BD-Spo11 binding in a centromere-associated cold region is not sufficient to induce cleavage.A simple interpretation of the position effect exerted on Gal4BD-Spo11 DSB formation would be that the RVS161 region is not accessible to GalBD-Spo11 (and, by extension, to Spo11) during meiosis. To test this hypothesis, we did a ChIP analysis of Gal4BD-Spo11 in formaldehyde-treated cells of strains that contain the URA3-gal2 cassette. Like the GAL3 promoter (Fig. 6A), the cassette-borne Gal4 binding sequences were bound throughout meiosis by Gal4BD-Spo11, whether the cassette was inserted at the his4 or the RVS161 locus (Fig. 6B and C). This shows that the absence of Gal4BD-Spo11 DSB formation at RVS161 is due not to a meiosis-specific exclusion of this region for Gal4BD-Spo11 binding but rather to a lack of cleavage activity. Interestingly, we also reproducibly observed stronger Gal4BD-Spo11 immunoprecipitation of GAL3 and URA3-gal2 regions as the cells progressed into meiosis (Fig. 6A and B, compare 0 h and later time points).
Binding of Gal4BD-Spo11 at the Gal4 binding sequences inserted in a hot or cold DSB region during meiosis. PCR was used after ChIP with an anti-Gal4BD antibody at the indicated times during meiosis. Strain with insertion at his4, ORD8276; strain with insertion at RVS161, ORD8278. (A) Binding at the GAL3 promoter. W, DNA from the whole-cell extract; Ip, DNA recovered after chromatin immunoprecipitation; C, control YCR013C site. (B) Binding in the URA3-gal2 cassette. The positions of primers used for the URA3-gal2 cassette are indicated in Fig. 5A. (C) Quantification of ChIP enrichment at the URA3-gal2 cassette shown in panel B versus enrichment at GAL3 shown in panel A. Each data point comes from results of two independent experiments. Error bars represent standard deviations.
Displacement of the centromere region restores Gal4BD-Spo11 targeted DSBs.The RVS161 locus is located in the vicinity of the chromosome III centromere. To find out if the centromere was involved in the repression of bound Gal4BD-Spo11 cleavage at the RVS161 insert, we used a strain in which 439 bp encompassing CEN3 were deleted and moved to the chromosome III left arm, between the LEU2 and NFS1 genes, 38 kb from RVS161 instead of 17 kb (T.-C. Wu and M. Lichten, unpublished results). We introduced the HphMX-gal2 reporter cassette at the RVS161 locus in both CEN3 and cen3Δ strains and examined meiotic DSB formation. Remarkably, Gal4BD-Spo11-specific DSBs occurring in the vicinity of the Gal4 binding sequences were restored (Fig. 5C and D). Therefore, the absence of DSB formation at the RVS161 locus by tethered Spo11 was due to centromere proximity. This finding further confirms that local factors control the cleavage step.
DISCUSSION
In this study of meiotic recombination in budding yeast, the ability to artificially tether the Spo11 protein to new chromosomal regions allowed us to answer several questions about the mechanisms and the control of DSB site selection, on both genome-wide and local scales.
Genome-wide modification of DSB distribution.In our previous study of DSB formation by the Gal4BD-Spo11 protein (38), we found that this fusion protein was able to introduce DSBs at a few natural Spo11 sites and in the Gal4 binding site-containing promoters of the GAL2, GAL1, GAL10, and GAL7 genes. We have now established the genome-wide distribution of the cleavage regions of Gal4BD-Spo11 in parallel with those of Spo11. The results confirm the dual and highly proficient properties of the fusion protein to cleave natural and novel sites. Indeed, 80% of the Gal4BD-Spo11 DSBs occur at Spo11 DSB sites, suggesting that although DSB frequency at some endogenous Spo11 sites might be affected, the choice of cleavage sites is extensively conserved between the two proteins. However, this is not a simple additive gain of novel targeted sites, since a large number of Spo11 sites (24%) are no longer cleaved by Gal4BD-Spo11 and the genome-wide distribution of DSBs is profoundly modified. Noteworthily, this redistribution has no consequence either on the sporulation efficiency or the spore viability (96%) of the diploid cells carrying the Gal4BD-Spo11 protein (38). We conclude that the natural profile of the meiotic recombination initiation sites in S. cerevisiae cells is flexible. However, as discussed below, control on potential target sites exists, since DSBs do not occur at all Gal4BD-bound sites.
Long-range cis-effect repression of DSB formation.Despite the overlap between Spo11 and Gal4BD-Spo11 DSB sites, some natural DSBs appeared as no longer cleaved by Gal4BD-Spo11. Part of this nonoverlap can be explained by the use of a stringent threshold criterion since 54% of the Gal4BD-Spo11-specific sites are secondary Spo11 DSB sites, indicating that most of these cases represent bona fide cleavage sites for both proteins but that cleavage occurred at different frequencies. DSB formation is reduced around the strongly targeted Gal4BD-Spo11 sites over long distances. We show that this is not due to an impaired ability of Gal4BD-Spo11 to cleave some natural sites since the mutation of the GAL80 binding site restores DSB formation in the surrounding regions.
The cumulated DSB frequency in a chromosomal region is thus kept constant. This DSB homeostasis occurring at the level of initiation over a chromosomal region likely proceeds through mechanisms very different from the classical crossover interference, in which a crossover in one region makes it less likely that another will be found nearby between the same chromosome pairs (reviewed in reference 24). Here, it is unlikely that the DSB reduction occurs on the same DNA molecule as the one that is cleaved. Indeed, such double cutting events are predicted to be very rare (for example 0.1% for two sites occurring each with a 10% frequency). Thus, the modification induced by the targeted DSB site may occur much more frequently than the cleavage itself. Alternatively, if one considers as one unit both sister chromatids or even all four duplexes of a bivalent, a site that shows up with breakage of 10% of the DNA may be considered as a single break at that locus in 40% of meiotic cells. Given the distances over which this phenomenon occurs, one can envisage that the suppression might occur in trans along all (or some) copies of a chromosome within a cell. Thus, it may be that suppression is linked directly to DSB formation, rather than reflecting competition for limiting factors independent of DNA cleavage.
As suggested before for short distances, cis-acting effects on DSB frequencies may be viewed as a competition phenomenon in which the DSB-forming factors available over a given chromosomal region would be attracted to the new and stronger site, rendering the surrounding sites less likely to be cleaved (52). Candidate factors are the DSB proteins Rec102, Rec104, and Mer2 which at early times of meiosis are chromatin associated but do not appear to preferentially localize at hot spots (17, 22). Perhaps, upon DSB site selection or cleavage, these proteins transiently diffuse to the chosen site and deplete other potential sites. Alternatively, preferential and perhaps earlier selection of the prominent DSB sites may induce a cis-acting inhibitory signal, for example, conformational changes or nucleosome modifications.
Position effects for cleavage of sites that are bound by Gal4BD-Spo11.This study allows us to distinguish three types of chromosomal regions with respect to DSB formation: the naturally proficient regions, the cold but permissive regions which can be warmed up by tethering Spo11 through a heterologous DNA binding domain, and the refractory cold regions that remain cold even in the presence of bound Gal4BD-Spo11 protein.
The most spectacular examples of a repressive position effect on Gal4BD-Spo11 potential cleavage sites occur at the natural Gal4BD binding site GAL3 and in the gal2 promoter-containing reporter cassette inserted at RVS161. Both are located in very cold DSB regions and close to a centromere (14 kb [or 23 kb with the Gal4BD-Spo11 construct] and 17 kb away, respectively). In both cases, ChIP analyses clearly demonstrate that Gal4BD-Spo11 is bound there throughout meiosis but that no detectable DSB formation is observed, whereas bringing Gal4BD-Spo11 to other regions like HIS4 is sufficient to produce DSBs. This position effect could reflect either a local inability of cleavage by Spo11 or a persistence of the “tight binding” intermediate, in which Spo11 is active and makes breaks but reseals them without engaging in the subsequent step, the irreversible cleavage and formation of stable DSBs (40). The molecular mechanisms underlying this position effect on DSBs remain to be identified, but most interestingly, we found that moving the CEN3 sequence 21 kb away was sufficient to restore a targeted Gal4BD-Spo11 DSB site in the URA3-gal2 reporter construct at RVS161.
One can envisage several nonexclusive ways for the centromere to exert this inhibitory effect. This inhibition might be through chromatin modifications, although no differences between cold centromere-associated and hot regions have been detected by measuring DNase I hypersensibility (10). No specific histone modifications have so far been identified at the proximity of budding yeast centromeres, except for a single nucleosome containing the centromere-specific H3-like protein, Cse4, on the core centromere sequence (35). Another possible mechanism is protection due to the binding of the cohesin complex over a 30- to 50-kb region centered on centromeres (13, 25, 31). This does not appear to be the case, since we examined genome-wide DSB formation in a rec8Δ rad50S strain and found that DSB formation was still low close to centromeres (data not shown). A third hypothesis would be that the centromere is located in a chromosome territory inaccessible to the DSB-forming machinery. Early in meiosis I, centromeres are paired in nonhomologous pairs (50). This pairing is then disrupted, and chromosome ends become clustered in a limited region of the nuclear envelope at the bouquet stage (44). This complex process may spatially exclude centromeric regions from DSB formation. Since DSB distribution is not affected in mutants suppressing the bouquet (48, 51) and DSB formation likely naturally occurs before the bouquet stage (45, 49), DSB formation may occur rather at the time of centromere pairing. Although we see a global reduction of DSB formation around centromeres when examining all chromosomes together, we have noticed that DSB repression close to an individual centromere is often asymmetrically focused on one side of the centromere. For instance, the left side of CEN3 is a hot region and the other side is cold. Similarly, some genes located near centromeres, like GAL3, are cold, whereas a DSB can be targeted at others (like YGL007W). Previous experiments have shown that the orientation of the centromere does not affect the directionality of recombination repression (29). One attractive explanation for these observations would be that centromeres may act as a boundary separating a DSB-permissive region from a DSB-refractory region, although they may not be a barrier for recombination intermediates since coconversion has been observed across the CEN3 region (47).
The centromere regions are not the only regions in the genome to be repressed while Gal4BD-Spo11 is bound. None of the Gal4BD-Spo11 binding sites in subtelomeric regions give rise to a DSB. However, we cannot rule out the possibility that for some of these sites, DSB frequency is underestimated due to the rad50S mutation (8).
To better understand these cleavage position effects, it will be interesting to determine which DSB factors bind the sites that are bound but not cleaved by Gal4BD-Spo11 and which proteins cannot be recruited to these sites.
Potential benefits of limiting DSB formation.From the data reported here and others before, we can identify two limitations of DSB formation: repression of DSB formation in certain regions of the chromosome (for example, at GAL3 and CEN3) and a balance among DSB sites such that the total number of DSBs in a chromosomal region is not changed (for example, on chromosome XIII upon DSB formation at GAL80). What could be the benefit of such DSB limitation? It has been suggested that too much recombination may be associated with missegregation (16, 33). Making about 150 DSBs in the genome is potentially dangerous for the cell as a single unrepaired DSB can be lethal. Thus, limiting the total amount of DSB formation might be an additional way to ensure that all the DSBs will be repaired properly. Restriction on the initiation position might facilitate the optimal function of chiasmata in chromosome segregation, for example, by avoiding crossing over too closely to a centromere. Our results obtained when we mutated the GAL80 promoter sequence also suggest that the reciprocal control exists which allows new DSBs to appear when DSB activity is lowered or inhibited in a chromosomal region. Similar effects on DSB homeostasis seem to exist in mammals, where the disappearance of one hot spot is often accompanied by the apparition of a new, nearby hot spot (latent hot spot) such that the global recombination frequency is unchanged (21).
Identifying the factors responsible for maintaining a constant level of recombination initiation over a particular chromosomal region and those repressing DSB formation in the subdomains of the chromosome remains an important challenge, but our ability to modulate the number and locations of sites bound by Spo11 provides new insights into the central issue of the nonrandom distribution of meiotic recombination events and a powerful method to modify meiotic recombination by a simple modification of the Spo11 protein.
ACKNOWLEDGMENTS
We thank Tzu-Chen Wu and Michael Lichten for providing the cen3Δ strain and communicating results before publication and Richard Durbin from the Sanger Institute and Ed Louis for allowing us to use the unpublished SK1 sequence information. We thank Michèle Vedel for constructing the pAP11 plasmid. We also thank Christine Mézard and Pierre-Antoine Defossez for comments that improved the manuscript and the anonymous reviewers for helpful suggestions.
This work was supported by grants from the Association pour la Recherche sur le Cancer (to A.N.). N.R. was supported by a predoctoral fellowship from Institut Curie. N.U. was supported by postdoctoral fellowships from the CNRS and the Institut Curie.
FOOTNOTES
- Received 3 November 2006.
- Returned for modification 30 November 2006.
- Accepted 12 December 2006.
- Copyright © 2007 American Society for Microbiology