Bidirectional resection of DNA double-strand breaks by Mre11 and Exo1.

Repair of DNA double-strand breaks (DSBs) by homologous recombination requires resection of 5'-termini to generate 3'-single-strand DNA tails. Key components of this reaction are exonuclease 1 and the bifunctional endo/exonuclease, Mre11 (refs 2-4). Mre11 endonuclease activity is critical when DSB termini are blocked by bound protein--such as by the DNA end-joining complex, topoisomerases or the meiotic transesterase Spo11 (refs 7-13)--but a specific function for the Mre11 3'-5' exonuclease activity has remained elusive. Here we use Saccharomyces cerevisiae to reveal a role for the Mre11 exonuclease during the resection of Spo11-linked 5'-DNA termini in vivo. We show that the residual resection observed in Exo1-mutant cells is dependent on Mre11, and that both exonuclease activities are required for efficient DSB repair. Previous work has indicated that resection traverses unidirectionally. Using a combination of physical assays for 5'-end processing, our results indicate an alternative mechanism involving bidirectional resection. First, Mre11 nicks the strand to be resected up to 300 nucleotides from the 5'-terminus of the DSB--much further away than previously assumed. Second, this nick enables resection in a bidirectional manner, using Exo1 in the 5'-3' direction away from the DSB, and Mre11 in the 3'-5' direction towards the DSB end. Mre11 exonuclease activity also confers resistance to DNA damage in cycling cells, suggesting that Mre11-catalysed resection may be a general feature of various DNA repair pathways.

We sought to clarify Mre11-dependent processing reactions at DSBs with blocked termini. Blocked DSB ends are created in meiosis by the topoisomerase-like transesterase, Spo11, to initiate crossover recombination[14] (Fig. S1a). Aside from the importance of meiotic recombination to sexual reproduction and genetic diversity, we reasoned that the molecular reactions pertaining to repair of covalently blocked Spo11-DSBs would be informative to the repair of DNA lesions with other types of end blockage, such as failed topoisomerase reactions and DSB ends competitively bound by the nonhomologous end-joining (NHEJ) machinery.

Mre11-dependent endonucleolytic processing (nicking) of Spo11-DSBs generates two classes of Spo11-oligonucleotide fragments originating from the DSB end[15] (Fig. S1). Spo11-oligo complexes do not form in mre11 mutants completely abrogated for nuclease activity (mre11-D56N, and mre11-H125N; Fig. 1a) solidifying earlier conclusions that Mre11 directly processes Spo11-DSBs[7-13]. Recently, an allele of Mre11 that is exonuclease deficient, but mostly endonuclease proficient was described for the orthologous proteins in S. pombe and P. furiosus[16]. We generated the equivalent mutation in budding yeast Mre11 (Histidine 59 to Serine; Fig. S2a) and tested its function. During meiosis, Spo11-oligo complexes were readily detected in mre11-H59S cells (Fig. 1a), indicating that the allele is endonuclease proficient in vivo. Biochemical assays with recombinant ScMre11-H59S showed reduced, but not abolished, 3′-5′ exonuclease activity on linear duplex DNA, and retention of much of the ssDNA endonuclease activity (Fig. S2)—observations consistent with mre11-H59S partially separating the two nuclease functions.

Figure 1

Mre11 and Exo1-dependent resection and repair of meiotic DSBs

a, Autoradiograph of Spo11-oligo formation (arrowheads) in Mre11 nuclease-defective cells: mre11-D56N, -H125N, -H59S, and in control spo11-Y135F cells where Spo11-DSBs do not form. Immunoprecipitated Spo11-oligo complexes are 3′ end-labelled using terminal transferase and separated on SDS-PAGE[15] (see Fig. S1). Asterisk marks an unrelated labelling artifact; b-f, Timecourse of events during meiosis for the indicated genotypes: b-c, Quantitative analysis of DSB (b) and crossover (c) signals at HIS4::LEU2. DNA at each timepoint was digested with PstI (b) or XhoI (c), separated by agarose gel electrophoresis and blots hybridised with probes to LSB5 (b) or STE50 (c). DSB signals were quantified as a percentage of specific lane signal (see also Fig. S3); d, Relative extent of DSB resection at HIS4::LEU2 (at 4 hours in meiosis). DNA was digested with BglII and blots hybridised with a STE50 probe (see also Fig. S4); e, Western blot analysis of phosphorylated H2Ax and Hop1 (arrowhead) detected from TCA-precipitated whole cell lysates. Phosphorylated Hop1 indicated by dot. Asterisk marks cross-reacting band; f, Progression through anaphase I and II assessed by microscopic examination of DAPI-stained cells; g, Distribution of spore viabilities per tetrad (n; number of 4-spore tetrads dissected). The difference in distribution between mre11-H59S/exo1Δ and exo1Δ is highly significant (Chi-square test for goodness-of-fit; Chi value = 63.084, d.f. = 4, P < 0.0001). Absolute spore viabilities are: WT, 97%; mre11-H59S, 90%; exo1Δ, 76%, mre11-H59S/exo1Δ, 59%.

We assessed DSB repair kinetics at two meiotic recombination hotspots using southern blotting and probes for the relevant genomic loci (HIS4::LEU2 and ARE1; Fig. 1b; Fig. S3). In the mre11-H59S strain, DSBs formed at normal levels and repaired as crossovers with normal timing (Fig. 1b,c). Single-stranded DNA resection, which can be qualitatively assessed by the relative migration of the DSB band on native agarose gels, also appeared unaffected by the mre11-H59S mutation (Fig. 1d lanes 1 and 6; Fig. S4).

The lack of an obvious defect in ssDNA resection proficiency suggested that any potential contribution to ssDNA generation by Mre11 might be masked by the activity of another nuclease. During meiosis, the major resection pathway requires Exo1[17-21]. We tested this idea by combining the mre11-H59S allele with an EXO1 deletion (exo1Δ), which we found itself to have slightly delayed DSB repair kinetics, with fewer DSBs repairing as crossovers (Fig. 1b,c). Migration of the DSB band on agarose gels revealed that ssDNA resection was measurably reduced in exo1Δ—but not entirely abolished relative to a mre11-H125N control (where the failure to remove Spo11 prevents all resection; Fig. 1d, lane 3 and 4-5; Fig. S4). Our data agree with those of others investigating a meiotic role for Exo1[17-21].

The combination of mre11-H59S with exo1Δ caused DSBs to transiently accumulate for a longer period, with DSBs detectable for two hours longer than in matched controls (Fig. 1b), and with formation of crossover recombinants reduced and delayed (Fig. 1c). These defects in DSB repair are correlated with a reduction in the mobility of DSB DNA on agarose gels (Fig. 1d, S4). Specifically, in comparison to the mre11-H59S or exo1Δ single mutants, we observed DSB signals to migrate similarly to DSBs from the mre11-H125N control (Fig. 1d, lanes 2 and 3). This reduction in mobility is indicative of less single-stranded resection in mre11-H59S/exo1Δ than of either single mutant. We conclude that Exo1 and Mre11 collaborate to enable efficient ssDNA generation at meiotic DSBs.

Unrepaired DSBs cause Tel1/Mec1(ATM/ATR)-dependent phosphorylation of histone H2Ax on serine 129 and hyper-phosphorylation of Hop1 (a meiosis-specific adaptor of the DNA damage response[22]). Phosphorylated H2Ax and Hop1 were detected in all strains indicating that significant resection is not essential for activation of Tel1/Mec1 (Fig. 1e). However, phospho-H2Ax accumulated and persisted until late timepoints only in mre11-H59S/exo1Δ, and Hop1 phosphorylation persisted for at least two hours longer than in matched controls. These observations are consistent with a genome-wide defect in DSB repair in mre11-H59S/exo1Δ. The Tel1 branch of the signaling pathway is primarily activated by unresected DSBs[23]. We observed that Hop1 phosphorylation in the mre11-H59S/exo1Δ background is highly dependent on Tel1 (Fig. S5), consistent with less resection occurring genome-wide.

To investigate if DSB processing defects were affecting meiotic chromosome segregation, we determined the efficiency of progression through anaphase I and II (Fig. 1f). Meiotic progression of wildtype and mre11-H59S was essentially identical, while exo1Δ was delayed by about 1 hour, with slightly reduced overall efficiency. In contrast, nuclear division in the mre11-H59S/exo1Δ double mutant was poor, with more than half of the cells having failed to complete even the first meiotic nuclear division after 10 hours in meiosis. Analysis of sporulation efficiency after 24 hours revealed increased incidence of aberrant tetrad maturation and/or nuclear packaging where orphaned chromosome fragments were observed outside the maturing spore wall (Fig. S6). These defects in meiotic chromosome segregation manifested as reduced spore viability in the double-mutant compared to controls (Fig. 1g).

To characterise in greater detail the molecular defect caused by mre11-H59S, we looked carefully at the distribution of Spo11-oligonucleotide products generated in vivo by the 5′-DSB processing reaction (Fig. 2a,b). In wildtype cells, two major classes of Spo11-oligo are observed, which differ by the length of attached DNA[15]. By contrast, in mre11-H59S we observed a shift in this distribution towards higher molecular weight Spo11-oligo species (Fig. 2a and Fig. S7). Importantly, total Spo11-oligo formation was not itself delayed (Fig. S7), indicating that the Spo11-removal reaction initiates with normal timing in mre11-H59S cells, but is defective in forming shorter molecules. To clarify the precise size-distribution of the Spo11-oligo molecules, we fractionated deproteinised oligos using denaturing PAGE (Fig. 2b). In wild type cells, two peak areas of signal 10-17 nucleotides (nt) and 28-40 nt were apparent. These correspond to the shorter and longer Spo11-oligo classes detected on SDS-PAGE (Fig. 2a). We additionally detected signal (24 % of total) in molecules 18-27 and 41-300 nt long (Fig. 2b), indicating that the length-distribution of processed molecules is significantly more heterogeneous than previously assumed, and that nicks are made at up to 300 nt from the DSB end.

Figure 2

Mre11-exonucleolytic processing of DSB ends

a-b, Spo11-oligo detection in wildtype (WT) and mre11-H59S during meiosis. 3′ end-labelled Spo11 complexes are fractionated by SDS-PAGE (a) or by nucleotide resolution urea/PAGE after proteolytic removal of Spo11 peptide (b); c, Mre11-H59S displays reduced 3′-5′ exonuclease activity on a nicked duplex. Reactions were performed as for Fig. S2c. Stars indicate 5′ label, F3 3′-end has 5 nt extension and is refractory to Mre11-mediated resection[25]; d, Chromatin association of Spo11-oligo complexes. Wildtype cell extracts were fractionated and the abundance of Spo11-oligo complexes assessed in soluble versus chromatin-enriched material; e-f, Nuclease resistance of Spo11-oligo complexes. Chromatin-enriched material from (d) was treated with DNase I, and abundance of Spo11-oligo complexes compared to the simultaneous degradation of genomic DNA (e), or of a control 55 nt oligonucleotide (f).

In the mre11-H59S mutant, the distribution was shifted such that the long oligo molecules of 41-300 nt, made up a third of the material detected (Fig. 2b). Although it is possible that the altered distribution of oligo molecules is caused by a reduction in the endonuclease activity of Mre11-H59S, our physical and genetic observations in both WT and mre11-H59S can be readily explained if resection begins at relatively distant nicks (up to 300 nt from the DSB) and traverses bidirectionally both away from (using Exo1) and towards (using Mre11) the DSB end[21]. Such a model is consistent with the opposing polarities of the Mre11 and Exo1 exonuclease activities[24,25], and with the synergistic loss in resection we observe in mre11-H59S/exo1Δ. Moreover, the length distribution of Spo11-oligos is compatible with the extent of Exo1-independent resection reported recently by others[21], and as predicted, Spo11-oligo length is unchanged by loss of EXO1 (Fig. S8). Finally, only a low background of Spo11-oligo complexes are detected in endonuclease-defective mre11-D56N, ruling out the possibility that the long oligo molecules arise via an alternative nuclease nicking the 5′ strand (Fig. 2b). To test this mechanism in vitro, we incubated Mre11 protein with a nicked duplex substrate designed to mimic this proposed in vivo reaction, and found Mre11-H59S to resect from the nick with lower efficiency than wildtype Mre11 (Fig. 2c).

Together, these observations led us to consider that the steady-state length of Spo11-oligo complexes might arise via the relative processivity of the 3′-5′ Mre11-exonuclease and the relative sensitivity to nucleolytic degradation of DNA close to the DSB end. Spo11-DSB formation requires at least ten factors[14], suggesting that a large protein complex may reside at—and protect—the DSB end. If this model were correct, we expected Spo11-oligo complexes to be associated with chromatin and resistant to nucleolytic degradation. We tested this idea by incubating a chromatin-enriched nuclear pellet from meiotic yeast cells with DNase I, then assessed the amount of Spo11-oligo complexes remaining relative to both bulk DNA and to an exogenous protein-free oligonucleotide included in parallel reactions (Fig. 2d-f). Greater than 90 % of Spo11-oligos are found in the chromatin fraction and remarkably, no loss in signal was observed despite extensive nucleolytic degradation of both the chromosomal DNA and the control oligonucleotide. We conclude that Spo11-oligo complexes are occluded from degradation even by exogenous nucleases—a prediction of our model.

In cycling cells, Mre11 nuclease activity promotes the onset of resection—a requisite for repair of DSBs by homologous recombination[1]. To investigate a specific role for the Mre11 3′-5′ exonuclease during DNA repair in cycling cells, we challenged yeast cells with exposure to DNA damaging agents. Similar to complete abrogation of the endo/exonuclease activities (mre11-H125N), reduced Mre11 exonuclease activity (mre11-H59S) sensitised cells to the DNA alkylating agent methyl methanesulphonate (MMS), and to the topoisomerase poison camptothecin (CPT; Fig. 3). Compared to an MRE11 deletion however, mre11-H59S and mre11-H125N are themselves far less sensitive, consistent with physical interactions between the Mre11-complex being retained (Fig. S9). In agreement with Mre11 endonuclease activity being unaffected in mre11-H59S, and allowing redundant processing pathways, combining mre11-H59S with a deletion of EXO1 did not further sensitise cells to MMS (Fig. S10). Together these observations suggest that the exonuclease activity of Mre11 is involved in the repair of various classes of DNA lesion.

Figure 3

DNA damage sensitivity of exonuclease defective Mre11 cells

Ten-fold serial dilutions of the indicated strains were spotted onto solid media containing the indicated compounds and incubated at 30°C for 2 days (CPT) or 3 days (MMS).

Understanding the regulation of DSB repair is a complex issue involving multiple factors with overlapping roles. Here, we propose a biological function for the 3′-5′ exonuclease activity of the evolutionarily conserved Mre11 protein. Previous work has indicated DSB resection to traverse unidirectionally[1]. Here, we propose a refined model that involves the coordination of two resection activities of opposing polarity: Exo1 away from the DSB and Mre11 towards the DSB end (Fig. 4). We favour the view that this exonuclease reaction begins at nicks created by the Mre11 endonuclease (in conjunction with Sae2) and which are positioned at variable distance from the DSB end, perhaps due to locus-specific chromatin architecture[26]. Although our assay detects only the site of incision closest to the DSB end, Mre11 and Sae2 may create multiple nicks on the resecting strand[17,21,27] (A. S. H. Goldman, personal communication), which in combination with exonucleolytic processing might further enhance resection efficiency. Finally, the recent observation that the length and abundance of SPO11-oligonucleotide complexes is increased in Atm−/− mice (S. Keeney and M. Jasin personal communication), suggests that Mre11 exonuclease activity may be an evolutionarily conserved feature directly regulated by ATM.

Figure 4

Model for bidirectional processing of DSBs by Mre11 and Exo1

a, Following DSB formation with blocked ends (hatched squares), Mre11/Sae2-dependent nicks flanking the DSB ends create initiation sites for bidirectional resection by Exo1/Sgs1-Dna2 away from the DSB, and by Mre11 towards the DSB end. Such terminal blocks could arise after base damage, trapping of a topoisomerase, or by avid binding of the NHEJ complex. 3′ ends marked with triangles. Mre11/Sae2 may make multiple nicks on the 5′ strand (light grey arrows) facilitating resection; b, In meiosis, the DSB ends are terminally blocked by covalently bound Spo11 protein (grey ellipses), and may be protected from Mre11-dependent exonuclease degradation by a large metastable multisubunit complex (dashed outline), thereby generating the observed size-distribution of Spo11-oligonucleotide complexes.

During meiosis, bidirectional processing may help to reinforce subsequent steps of repair, which at least in some cases appear to occur differentially on either side of the DSB[26,28,29]. Our observation that liberated Spo11-oligo complexes remain chromatin-bound and relatively protected provides a clue to potential mechanisms of end differentiation. For example, retention of proteins on one or both of the DSB ends could influence subsequent steps of repair[15] (Fig. 4b). Finally, our observation that Mre11-dependent incision occurs at some distance from the DSB end suggests that DSB formation and processing reactions are coordinated over a considerable distance (300 bp of B-form DNA is ~100 nm). How these incision points are regulated—and restricted to only the 5′-ending strand—are fascinating questions for the future.

Although much of this work concerns specifics of meiotic DSB processing, we suggest that a similar pathway may occur whenever the DSB end is blocked by a lesion or protein complex that prevents direct loading of the 5′-3′ exonuclease machinery. Such DNA blockages may be crosslinked protein, damaged DNA ends, or simply the stable binding of high affinity proteins to the DSB. As such, the nucleolytic incision pathway may provide the key point of regulation that controls the balance between NHEJ and HR.

Methods summary

Yeast strains and culture methods

Meiotic cultures were prepared as described[15]. Strains were derived from SK1 using standard techniques. Spo11 protein is tagged by the HA3His6 epitope[15]. Mre11 mutations were introduced by pop-in/out using plasmids derived from pRS414-MRE11[10]. exo1Δ, tel1Δ, and sae2Δ are full replacements of the ORF with kanMX4, hphNT2, and kanMX6 respectively. A full strain list is available on request. For chronic DNA damage sensitivity, exponentially growing cultures were spotted in 10-fold dilutions onto freshly prepared media.

Molecular techniques

Spo11-oligonucleotide complexes were detected by two rounds of immunoprecipitation and end-labeling as described[15]. DSBs and crossover recombinants were detected by Southern blotting genomic DNA after fractionation on agarose gels using standard techniques[28]. Radioactive signals were collected on phosphor screens, scanned with a Fuji FLA5100, and quantified using ImageGauge software. Phosphorylated H2Ax and Hop1 protein were detected using Ab17353 (Abcam) and anti-Hop1 antisera (F. Klein) respectively. Subcellular fractionation of cell contents was performed by purifying hypotonically lysed spheroplasts through a sucrose cushion. Chromatin digests were performed at 25°C with DNase I (NEB). Reactions were split and processed for DNA extraction or Spo11-oligo purification. As a positive control, parallel reactions included a 5′-labeled 55 nt oligonucleotide. Recombinant GST-Mre11 alleles were purified and reacted with DNA substrates as described[10].