|Home | About | Journals | Submit | Contact Us | Français|
Budding yeast lacking the Sgs1 helicase and the Mus81/Mms4 endonuclease are inviable, and indirect studies implicate homologous recombination gone awry as the cause of death. We show that mutants lacking both enzymes have profound defects in meiotic recombination intermediate metabolism and crossover (CO) formation. Recombination intermediates (joint molecules; JMs) accumulate in these cells, many with structures that are infrequent in wild type cells. These JMs persist, preventing nuclear division. Using an inducible expression system, we restored Mus81 or Sgs1 to sgs1 mus81 cells at a time when JMs are forming. Mus81 expression did not prevent JM formation, but restored JM resolution, CO formation, and nuclear division. In contrast, Sgs1 expression reduced the extent of JM accumulation. These results indicate that Sgs1 and Mus81/Mms4 collaborate to direct meiotic recombination towards interhomolog interactions that promote proper chromosome segregation, and also indicate that Mus81/Mms4 promotes JM resolution in vivo.
Meiosis is the process by which diploid cells produce haploid gametes. At the first meiotic division, homologous parental chromosomes (homologs) segregate, while at the second meiotic division sister chromatids segregate. Interhomolog recombination helps to ensure proper homolog alignment at meiosis I metaphase and thereby proper segregation at the first division (reviewed in Cromie and Smith, 2007). Recombination is initiated by DNA double strand breaks (DSBs) that are formed by the Spo11 nuclease (Keeney et al., 1997). Repair of these DSBs produces heteroduplex DNA-containing recombinant products: crossovers (COs), with flanking marker exchange; and non-crossovers (NCOs), with flanking markers in parental configuration.
Joint molecules (JMs) that contain DNA molecules linked by Holliday junctions are a central intermediate in CO formation. In wild-type budding yeast meiosis, most JMs contain two Holliday junctions (Schwacha and Kleckner, 1995) and are called double-Holliday junction (dHJ) intermediates. The proteins that resolve JMs as COs are not yet known. Cells lacking the mid-meiosis transcription factor Ndt80 or the polo-like kinase Cdc5 show JM accumulation, reduced COs and normal NCO levels, indicating that meiotic JM-resolution, with COs as the primary product is regulated by these proteins (Allers and Lichten, 2001b; Clyne et al., 2003). A second set of proteins, the ZMM proteins, are required for normal levels of both JMs and CO products, but not for NCO recombination ([reviewed in Lynn et al., 2007). Some ZMM proteins have biochemical activities that could stabilize nascent JMs, and all are associated with the synaptonemal complex, a structure that assembles between fully co-aligned homologous chromosomes during meiotic prophase.
One important ZMM protein function is to protect early pre-CO intermediates from Sgs1 (Jessop et al., 2006; Oh et al., 2007). Sgs1 is a RecQ-family DNA helicase, a homolog of the mammalian BLM helicase, and both are found in complex with a heterodimer of topoisomerase III (Top3) and Rmi1 (BLAP75 in mammals; Chang et al., 2005; Mullen et al., 2005; Yin et al., 2005). Loss of BLM activity in Bloom’s syndrome results in cancer predisposition and a marked increase in sister chromatid exchange (reviewed in Mankouri and Hickson, 2007). It has been suggested that the BLM complex suppresses exchange by disassembling dHJs, and in vitro studies have shown that reconstituted BLM/TopIIIα/Rmi1 can resolve synthetic dHJs to NCOs (Mankouri and Hickson, 2007). Loss of Sgs1 in S. cerevisiae also causes increased chromosome instability, elevated recombination, and DNA-damage sensitivity, suggesting a role for the Sgs1/Top3/Rmi1 complex in regulating mitotic recombination (Gangloff et al., 1994; Mullen et al., 2000; Watt et al., 1996). During meiosis, sgs1 mutants show a modest increase in COs and transient accumulation of inter-homolog, inter-sister, and multichromatid JMs (JMs with more than two chromatids), but NCO levels are unaltered (Jessop et al., 2006; Oh et al., 2007; Rockmill et al., 2003). Loss of Sgs1 partially restores COs to zmm mutants, indicating that the Sgs1 complex disassembles early CO-specific intermediates that are normally stabilized by the ZMM proteins (Jessop et al., 2006; Oh et al., 2007).
Mus81 is an XPF-family nuclease, and functions in complex with a non-catalytic partner, Mms4 (Eme1 in S. pombe and mammals; Kaliraman et al., 2001). Purified Mus81/Mms4 and Mus81/Eme1 resolve HJs in vitro, although they are more active on nicked HJs (Gaskell et al., 2007 and references therein). During vegetative growth, sgs1Δ is synthetically lethal with mus81Δ or mms4Δ, and this lethality is suppressed by loss of the RecA homolog Rad51 (Bastin-Shanower et al., 2003; Fabre et al., 2002). This suggests that Sgs1 and Mus81 complexes act in alternate processes that keep lethal recombination intermediates from accumulating, either by preventing their formation or by catalyzing their resolution. During S. pombe meiosis, CO recombinant recovery is largely dependent upon Mus81/Eme1 (Smith et al., 2003), and mus81 cells fail to undergo meiotic divisions and accumulate JMs with single HJs (Cromie et al., 2006). In S. cerevisiae, meiotic COs are much less dependent upon Mus81/Mms4, and viable progeny from single mutants show only a modest CO reduction (Argueso et al., 2004; de los Santos et al., 2003; de los Santos et al., 2001). Thus, uncertainty remains regarding the in vivo activity and targets of Mus81/Mms4 during budding yeast meiosis.
To learn more about the roles that Sgs1 and Mus81/Mms4 play during meiotic recombination, we examined the meiotic defects of double mutants, using meiotic null (mn) alleles of SGS1 and MMS4 to bypass the vegetative inviability of sgs1 mus81 and sgs1 mms4 mutants. We also created inducible (IN) alleles of both SGS1 and MUS81, allowing restoration of either protein to sgs1-mn mus81Δ double mutants late in meiosis I prophase.
We find that Sgs1 and Mus81 play significant roles in JM formation and disassembly during meiosis. In cells lacking both Sgs1 and Mus81/Mms4, COs are reduced, JMs accumulate and cells are unable to divide nuclei; similar observations are reported by Oh et al. (S. D. Oh, J. Lao, A. F. Taylor, G. R. Smith and N. Hunter, personal communication). Inducing expression of MUS81 in these cells causes JMs to be resolved, while SGS1 induction prevents further JM accumulation. In both cases, COs and meiotic divisions are partially restored. These findings illuminate the complementary functions of Sgs1 and Mus81/Mms4 in meiotic recombination intermediate metabolism, and in particular indicate that Mus81/Mms4 can resolve Holliday junctions in vivo.
We used a meiotic null allele of SGS1 (sgs1-mn), where the normal SGS1 promoter is replace by the CLB2 promoter, which is inactive during meiosis. The sgs1-mn strain shows normal meiotic nuclear divisions (Jessop et al., 2006; Figure 1A). Using the same method, we constructed mms4-mnΔ Single mus81Δ or mms4-mn mutants show only modest nuclear division defects at 30°C. In contrast, sgs1-mn mus81Δ and sgs1-mn mms4-mn cells fail to accomplish even a single nuclear division by 8 hours, the time at which nuclear divisions are complete in wild type and single mutants (Figure 1A and Supplementary Figure 1A).
During meiosis, sgs1-mn cells lack the entire Sgs1 protein (Jessop et al., 2006). Many sgs1Δ phenotypes are moderated by expressing sgs1-ΔC795, a mutant that produces the N-terminal Top3/Rmi1-interaction domain of Sgs1 but not the C-terminal helicase and Holliday junction-binding domains (Mullen et al., 2000). However, expressing the N-terminal domain in sgs1-mn/sgs1-ΔC795 mms4-mn cells suppresses neither the nuclear division defect nor any of the other defects described below (Supplementary Figure 1), suggesting that the absence of Sgs1 helicase and/or Holliday junction-binding activities are the main cause of these meiotic defects.
If cells fail to repair DSBs or complete meiotic recombination, they arrest at the pachytene stage of meiosis I prophase, a stage that in wild-type is characterized by intact synaptonemal complex (SC), unseparated spindle pole bodies, and unresolved JMs (Roeder and Bailis, 2000). This arrest is due to a failure to activate Ndt80, a transcription factor required for exit from pachytene (Chu and Herskowitz, 1998; Xu et al., 1995). The nuclear division failure of sgs1-mn mus81Δ and sgs1-mn mms4-mn cells is not due to a DSB repair defect, as DSB formation and disappearance are similar in wild-type, single, and double mutants (Figure 1B, Supplementary Figure 1B). In addition, sgs1-mn mus81Δ cells accumulate and degrade the SC component Zip1 with normal kinetics (Figure 1C). Thus, the nuclear division defect in sgs1-mn mus81Δ or sgs1-mn mms4-mn mutants does not reflect a pachytene arrest. Consistent with this conclusion, sgs1-mn mus81Δ cells form normal meiotic spindles. Immunocytology, using anti-beta tubulin sera, showed that linear meiosis I spindles form to about the same extent and with similar timing in wild type and sgs1-mn mus81Δ (Figure 1D). Meiosis II spindles can be seen at later time points, and spore walls form in some cells, though the spores do not contain DNA (data not shown). In contrast, pachytene-arrested ndt80. cells fail to form meiotic spindles, and contain beta tubulin as a single focus in all cells [(Xu et al., 1995), Figure 1D].
Meiotic nuclear division might fail in sgs1-mn mus81Δ cells if chromosomes are held together by unresolved recombination intermediates. To test this suggestion, we examined meiotic recombination at the molecular level, using a 3.5 kb recombination reporter insert that contains divergently transcribed URA3 and ARG4 genes, along with a 65 nt telomere-repeat insert that creates a single strong DSB hotspot (Jessop et al., 2005; White et al., 1993). This reporter is inserted at HIS4 on one parental chromosome III and at LEU2 on the other, and one insert contains a mismatch correction-resistant palindromic mutation. Different restriction enzyme and probe combinations allow detection of DSBs, CO and NCO recombinants, and both interhomolog and intersister recombination intermediates on Southern blots (Jessop et al., 2005, Figure 2A, Supplementary Figure 2).
As described above, DSB repair is completed in all mutants examined (Fig 1B, Supplementary Figure 1B), and NCO recombinants form at levels similar to or slightly greater than those seen in wild-type (Fig 1E, Supplementary Figure 1D). In sgs1-mn, mus81Δ and mms4-mn single mutants, CO recombinants form with near wild-type levels and timing. In contrast, COs are present in sgs1-mn mus81Δ and sgs1-mn mms4-mn at about 45% of wild type levels (Fig 1E and Supplementary Figure 1D), indicating that Mus81/Mms4 and Sgs1 are redundantly required for normal CO formation.
To determine why COs are reduced, we examined JM formation and resolution. Single sgs1-mn, mus81Δ or mms4-mn mutants formed and resolved JMs. In contrast, unresolved JMs persisted in sgs1-mn mus81Δ and sgs1-mn mms4-mn (Figure 2B, 2C and Supplementary Figure 1C), reaching a peak at ~6hr before slightly declining as cells attempted nuclear divisions. In sgs1-mn mus81Δ, the maximum JM level (6.5% of chromosomes) is about 4-fold greater than the transient maximum seen in wild type (1.5%) and about 80% of the cumulative level in ndt80Δ. (8%; Figure 2B and data not shown). A similar JM accumulation occurs in sgs1-mn mms4-mn at URA3-ARG4 (Supplementary Figure 1C), and is also observed at an independent recombination reporter locus (S. Oh, J. Lao, A. F. Taylor, G. R. Smith and N. Hunter, personal communication). JMs also accumulate in sgs1-mn sgs1-ΔC795 mms4-mn (Supplementary Figure 1C), showing that Sgs1 helicase and/or HJ-binding domains are needed to prevent JM accumulation. JMs also accumulate in sgs1-mn mus81Δ at a native DSB hotspot, (YGR175C, Supplementary Figure 3a), and do not accumulate in DSB-defective sgs1-mn mus81Δ spo11-Y135F cells (Supplementary Figure 3B and C). Thus, the JMs that accumulate in the absence of Sgs1 and Mus81/Mms4 are products of Spo11-initiated meiotic recombination, as opposed to recombination initiated by other processes (such as pre-meiotic replication).
While interhomolog JMs normally predominate in wild-type, Oh et al. have shown that sgs1-ΔC795 mutants accumulate substantial amounts of intersister and multichromatid JMs, which are subsequently resolved in an Ndt80-dependent manner (Oh et al., 2007). We observe a similar increase in intersister and multichromatid JMs among the intermediates that form transiently in sgs1-mn and that accumulate in sgs1-mn mus81Δ and sgs1-mn mms4-mn (Figure 2 and Supplementary Figure 1C; similar observations by S. D. Oh, J. Lao, A. F. Taylor, G. R. Smith and N. Hunter, personal communication). While ndt80Δ mutants show an interhomolog/intersister JM ratio of ~7.5, consistent with the ratio seen in wild-type (Oh et al., 2007; Schwacha and Kleckner, 1994), intersister JMs are formed at greater levels in sgs1-mn mus81Δ, (interhomolog/intersister JM ratio of ~1.7; Figure 2D and Supplementary Figure 4B; similar differences also reported in Oh et al., 2007). In addition, a substantial fraction of the JMs that form transiently in sgs1-mn and that accumulate in sgs1-mn mus81Δ cells contain 3 and 4 parental-length molecules, identified on the basis of electrophoretic migration and hybridization patterns with HIS4- and LEU2-specific probes [Figure 2C and D, Supplementary Figure 4B, (Oh et al., 2007) and data not shown]. These multichromatid JMs do form in wild-type, but are considerably more frequent in sgs1-mn and sgs1-mn mus81Δ. 25–30% of JMs formed in sgs1-mn and sgs1-mn mus81Δ are multi-chromatid, as opposed to 9–10% of JMs formed in wild-type and ndt80. [Figure 2c and data not shown; similar differences reported by (Oh et al., 2007)].
In summary, cells lacking both the Sgs1 helicase and the Mus81/Mms4 endonuclease form NCOs at near-normal levels, COs at substantially reduced levels, and accumulate unresolved JMs, even though these cells continue to progress through other stages of meiosis. The JMs that form contain reduced levels of interhomolog JMs and elevated levels of intersister and multichromatid JMs relative to wild-type.
The persistence of unresolved JMs in sgs1-mn mus81Δ and sgs1-mn mms4-mn, and not in sgs1-mn alone, implies that Mus81/Mms4 can disassemble these structures in vivo. To test this, we constructed strains where MUS81 expression is rapidly induced by the addition of estradiol (MUS81-IN). While meiotic divisions are completely blocked in sgs1-mn mus81Δ, basal MUS81 expression in sgs1-mn mus81Δ/MUS81-IN (~25% of wild-type) allows some cells to complete meiotic divisions (12%; Figure 3C). When MUS81 expression is induced at 4 hours (to at least 4 times wild type, figure 3A), many cells (64%) complete at least one meiotic nuclear division (Figure 3C). This restoration of nuclear division is associated with increased and accelerated JM resolution. In the absence of induced MUS81 expression, about 60% of accumulated JMs are resolved; when MUS81 expression is induced at 4 hours, almost all JMs disappear (Figure 3B). JMs are not resolved when MUS81 expression is induced in sgs1-mn mms4-mn mus81Δ/MUS81-IN (data not shown), indicating that JM resolution in sgs1-mn mus81Δ/MUS81-IN cells is due to Mus81/Mms4 endonuclease activity, and not due to Mus81 acting in the absence of Mms4.
Mus81 induction has no effect on NCO levels, indicating that Mus81/Mms4 does not resolve inter-homolog JMs to NCOs (Figure 3D). In contrast, MUS81 expression causes a 1.4-fold increase in COs, which is approximately that expected from the increased JM resolution that occurs. At the end of meiosis, 5.5% of chromosomes are present in JMs in sgs1-mn mus81Δ, in contrast to only 0.6% when MUS81 is induced (Figure 3B). Based on the ratio of interhomolog JMs to intersister JMs in sgs1-mn mus81Δ (Figure 2D), we estimate about 60% of JMs, or about 3.3% of chromosomes, could potentially be resolved to form inter-homolog COs. This agrees well with the increase in COs (3% of chromosomes) observed upon MUS81 induction. While the final level of COs (5.7%) is less than in wild type (8%), these results indicate that Mus81/Mms4 can resolve most of the JMs that are present in sgs1-mn mus81Δ mutants, with COs being the predominant product when interhomolog JMs are resolved.
Although in vitro studies show that human BLM/TopIIIα/Rmi1 resolves synthetic dHJs to NCOs (Wu et al., 2006), studies of Sgs1 meiotic function have not revealed a role in NCO formation (Jessop et al., 2006). To further address this issue, we constructed a strain where SGS1 expression was estradiol-inducible (SGS1-IN; Experimental Procedures). When estradiol is absent, Sgs1 cannot be detected (Figure 4A), less than 5% of cells divide nuclei (Figure 4C), and JMs accumulate to levels (5.5% at 8 hr) similar to those seen in sgs1-mn mus81Δ (7%; Figure 4B). When SGS1 expression is induced at 4 hours (4-fold greater than the endogenous level; figure 4A), 40% of cells complete at least one division, (Figure 4C) and JM accumulation is prevented (figure 4B). Maximum JM levels upon SGS1 induction at 4 hours (1.4%) are only slightly greater than the maximum in wild type (1.5%), and are substantially less than the maximum seen without SGS1 induction (5.5%). Although SGS1 expression in sgs1-mn mus81Δ cells led to substantially fewer JMs accumulating, it did not cause extensive NCO or CO production, with modest increases in both products upon SGS1 induction (induced vs. uninduced: NCOs, 2.5% vs. 1.9%; COs, 4.7% vs. 2.9%). These reduced JM levels suggest that Sgs1, either alone or in complex with Top3 and Rmi1, prevents JMs from forming or promotes their rapid disassembly in vivo. If the latter occurs, it does not appear to produce exclusively NCOs or COs.
During meiosis, S. pombe mus81− mutants accumulate unresolved JMs containing single HJs (Cromie et al., 2006). We used two-dimensional native/denaturing electrophoresis (Allers and Lichten, 2001b) to ask if this occurs in S. cerevisiae sgs1-mn mus81Δ mutants (Supplementary Figure 4A). In this assay, interhomolog JMs with an even number of HJs will produce only parental-length single strands, while interhomolog JMs with an odd number of HJs will produce parental- and crossover-length strands in equal amounts. Interhomolog JMs from ndt80Δ cells contain primarily parental length strands, but ~13% of component strands are CO length. CO-length strands are present at similar levels (15%) in JMs from sgs1-mn mus81Δ cells, suggesting that approximately ¼ of JMs in these mutants contain single HJs. Consistent with this suggestion, four-arm JMs with junction structures expected for single HJ intermediates are frequently observed in electron micrographs of meiotic DNA from sgs1-mn mms4-mn mutant cells (S. D. Oh, J. Lao, A. F. Taylor, G. R. Smith and N. Hunter, personal communication). Because MUS81 expression results in the resolution of more than 90% of the JMs that form in sgs1-mn mus81Δ, these results indicate that Mus81/Mms4 resolves HJs in intermediates that contain either even or odd numbers of HJs.
In this paper, we show that the mitotic lethality seen in cells lacking Sgs1 and Mus81 or Mms4 (hereafter referred to as sgs1 mus81/mms4 mutants) also occurs in meiosis. During meiosis, sgs1 mus81/mms4 mutants accumulate unresolved JMs and fail to divide nuclei, although they form MI and MII spindles on schedule and even form spore walls; similar observations have been made by S. D. Oh, J. Lao, A. F. Taylor, G. R. Smith and N. Hunter (personal communication). sgs1 mus81/mms4 mutants do not arrest at a meiotic checkpoint prior to JM resolution and nuclear division. Instead, as in mus81 or eme1 single mutants of S. pombe, the failure to undergo meiotic nuclear division appears to be to the result of persistent inter-homolog links, caused by unresolved recombination intermediates (Boddy et al., 2001; Cromie et al., 2006). This mutant phenotype is suppressed to a substantial extent by restoration of Mus81/Mms4 at a time after JM formation, and to a lesser extent by restoration of Sgs1. As will be discussed below, these results are consistent with the suggestion of distinct roles for the Sgs1/Top3/Rmi1 complex (hereafter called the Sgs1 complex) and for Mus81/Mms4 in preventing the accumulation of unresolved (and thus lethal) recombination intermediates.
During the mitotic cell cycle, sgs1 mus81 mutants die with a G2 DNA content, consistent with a nuclear division failure caused by unresolved JMs (Ii et al., 2007). This lethality is suppressed by preventing homologous recombination, a finding that is generally taken to indicate that Sgs1 and Mus81/Mms4 act after the initial steps of strand invasion. However, it remains possible that they act earlier, to prevent formation of lesions or abnormal structures that are then processed by homologous recombination to create lethal intermediates. We show here that, in sgs1 mus81, meiotic JM accumulation/persistence and blocked nuclear division both require Spo11 activity. This indicates that the primary lethal defect in the sgs1 mus81/mms4 mutants is a failure to eliminate JMs formed by Spo11-initiated meiotic recombination. It remains possible, and even likely, that recombinogenic lesions form in sgs1 mus81/mms4 during premeiotic S phase, as they do during mitosis (Ii et al., 2007). However, these lesions must be sufficiently rare to allow nuclear division in the majority of cells and to escape detection on Southern blots in the absence of meiotic recombination initiation. In summary, our data indicate that Sgs1 and Mus81/Mms4 have critical meiotic functions that involve action during late steps in recombination.
The vast majority of JMs formed during wild-type meiosis contain two DNA molecules, with contributions from both homologs (80–90%) rather than from sister chromatids (Oh et al., 2007; Schwacha and Kleckner, 1994). The JMs that we observe in mus81 single mutants are similar in composition to those formed in wild type and are efficiently resolved. Both COs and NCOs are produced at near normal levels, and the majority of cells progress through meiotic divisions. A previous study reported altered JM metabolism and reduced COs in mms4Δ mutants (de los Santos et al., 2003). The basis for these differences remains to be determined, but it may reflect different sporulation conditions and auxotrophic strain requirements used in the two studies (T. Goldfarb and ML, unpublished data). Nevertheless, our data clearly show that, under the appropriate conditions, mus81Δ cells can produce a relatively normal spectrum of recombination intermediates and products.
By contrast, in cells lacking Sgs1, JMs accumulate during meiosis to substantially greater levels than are seen in wild type, and intersister and multichromatid JMs are prominent (Oh et al., 2007). All JMs are eventually resolved, and recombinants are recovered at normal levels (Jessop et al., 2006; Oh et al., 2007). In the current work, we show that sgs1 mus81/mms4 mutants persistently accumulate a similar JM spectrum as is seen transiently in sgs1, and also display normal NCO but reduced CO levels. Similar findings are also reported by S. D. Oh, J. Lao, A. F. Taylor, G. R. Smith and N. Hunter (personal communication). Our finding of JM persistence in sgs1 mus81/mms4 mutants, and their ultimate resolution upon MUS81 expression in sgs1-mn mus81Δ/MUS81-IN, indicates that the Mus81/Mms4 endonuclease can resolve HJs that form in the absence of Sgs1.
While there is considerable biochemical and indirect genetic evidence suggesting that Mus81/Mms4 (Mus81/Eme1) can resolve HJs (Gaskell et al., 2007 and references within), direct in vivo evidence has not been obtained. Our findings are most readily interpreted as showing that Mus81/Mms4 has such an activity. It is of interest to note that about one-quarter to one-third of the interhomolog JMs that accumulate in sgs1-mn mus81Δ and in ndt80Δ mutants have the strand composition expected for a JM with a single HJ (Supplementary Figure 4; (Allers and Lichten, 2001b). Single-HJ JMs may be primary products of DSB repair (Cromie et al., 2006), or they may result when one of the two HJs in a dHJ intermediate is resolved. The near-complete JM resolution seen upon MUS81 expression in sgs1-mn mus81Δ cells indicates that Mus81/Mms4 can catalyze JM resolution and CO production, regardless of the number of HJs that JMs contain.
While we favor the suggestion that Mus81/Mms4 resolves HJs, other explanations are possible. For example, MUS81 induction could lead to the degradation of JM-containing structures, rather than their resolution. We believe this to be unlikely for two reasons. First, no increase in DNA molecules of lessthan-parental length is detected upon MUS81 induction (data not shown). Second, the increase in CO recombinants upon JM disappearance is that expected if most inter-homolog JMs were resolved as COs.
It should be kept in mind that there is at least one other HJ resolving activity in budding yeast meiosis, as JMs are resolved and COs are produced at wild-type levels in mus81Δ cells, and sgs1-mn mus81Δ cells produce CO recombinants at 45% of the level seen in wild-type. The decline in JM levels seen late in meiosis in sgs1-mn mus81Δ cells may also reflect the presence of other HJ-resolving activities. JMs that persist in sgs1 mus81/mms4 cells may form outside of the normal chromosome context in which other resolvases act, or they may contain structures that are poorly resolved by these enzymes. The same may be true in S. pombe, where loss of Mus81/Eme1 causes a 75% CO reduction, the persistence of JMs with single HJs, and blocked nuclear division (Cromie et al., 2006 and references within). While Mus81 nuclease may be more important for JM resolution in fission yeast than in budding yeast, other HJ-resolving proteins may be active in S. pombe, and it remains possible that some of the JMs formed in this organism contain two HJs.
In vitro studies have shown that BLM/TopIIIα/Rmi1 disassemble synthetic dHJs to form NCO products. However, it appears that this is not the primary activity of the Sgs1 complex during meiosis. sgs1 mutants show a transient JM hyper-accumulation (Oh et al., 2007), which is not accompanied by a decrease or delay in NCO recombinant formation (Jessop et al., 2006). Instead, sgs1 mutants show altered meiotic CO distributions and partially restore JMs, COs and tight homolog association to mutants lacking ZMM proteins (Jessop et al., 2006; Oh et al., 2007; Rockmill et al., 2003), again without altering NCO levels. This suggests that the primary meiotic function of Sgs1 is to prevent unregulated JM assembly, and that CO-designated intermediates are protected from this activity by SC components (Jessop et al., 2006; Oh et al., 2007).
We show here that SGS1 expression in sgs1-mn mus81Δ mutants substantially reduces JM accumulation, causing only modest increases in both NCOs and COs. This indicates that the Sgs1 complex prevents JM accumulation, perhaps by disassembling early strand invasion intermediates that are precursors to the stable JMs detected here; alternatively Sgs1 might disassemble JMs after they form. Our current data do not distinguish unambiguously between these two possibilities. However, when SGS1 is expressed late in meiosis I prophase in an sgs1 ndt80 mutant, JMs remain unresolved (L.J. and M.L., unpublished data), consistent with Sgs1 acting primarily to prevent JM formation.
These results are consistent with the suggestion that, by eliminating precursors to inter-sister and multi-chromatid JMs, Sgs1 redirects recombination events towards a second round of interhomolog recombination. Alternatively, Sgs1 may drive pre-existing JMs into conformations that can be resolved by other structure-specific endonucleases, such as Slx1/Slx4 or Rad1/Rad10, perhaps in a way that is not biased towards CO formation. It is also possible that Sgs1 drives some nascent recombination intermediates away from JM/CO formation and towards NCO formation, but in such a way that heteroduplex tracts in products are less than 600 nt long. These would not be detected as NCO recombinants in our assay, as they would not include the palindromic marker used to score NCO recombinants (Jessop et al., 2005). Finally, it should be noted that Sgs1 is overproduced 4-fold in our induction experiments. Because the Top3/Rmi1 heterodimer is relatively stable in the absence of Sgs1 (Chang et al., 2005; Mullen et al., 2005), it is likely that at least some of the re-expressed Sgs1 protein reconstitutes the normal Sgs1/Top3/Rmi1 complex, but it is also likely substantial Sgs1 is present without its usual partners. Whether the JM reduction we see is due to Sgs1/Top3/Rmi1 or due to free Sgs1 remains to be determined.
Conventional pathway analysis of data from various loss of function mutants has led to a simply-branched unidirectional pathway model for meiotic recombination (reviewed by Cromie and Smith, 2007). In this model, most COs form via a spatially and temporally regulated process, promoted by SC-associated proteins. This process produces single-end invasion and dHJ JMs that are, in turn, resolved in an Ndt80-, Cdc5-regulated manner. The remaining COs, hypothesized to be “unregulated”, are suggested as forming via a ZMM-independent, Mus81/Mms4-dependent process that is otherwise poorly defined. Features of our data and of other recent findings are difficult to reconcile with this view. First, the ZMM-dependence of interhomolog JM and CO formation is partially relieved by loss of Sgs1, which indicates that the Sgs1 complex has a meiotic anti-crossover activity (Jessop et al., 2006; Oh et al., 2007). Second, loss of Mus81/Mms4 has a relatively modest impact on interhomolog JMs and COs, but the further loss of Sgs1 causes reduced COs and the accumulation of unresolved JMs (Figure 1 and Figure 2; S. Oh, J. Lao, A. F. Taylor, G. R. Smith and N. Hunter, personal communication). This would indicate that Sgs1 has a meiotic pro-crossover activity, at least in the absence of Mus81/Mms4.
To reconcile this apparent contradiction, we suggest that the initial steps of meiotic DSB repair populate multiple sets of early recombination intermediates, and that these intermediates undergo multiple rounds of disassembly and reassembly. A similar suggestion has recently been put forth to explain mitotic recombination by break-induced replication (Smith et al., 2007) and for meiotic NCO recombination (McMahill et al., 2007). The Sgs1 complex, and in particular its helicase activity, prevents all types of meiotic JM formation by disassembling nascent strand invasion structures before they can be stabilized by second end-capture and ligation (Figure 5; Oh et al., 2007). This, combined with the protection of CO-designated precursors by ZMM proteins (Jessop et al., 2006; Oh et al., 2007), has the net effect of inhibiting the formation of JMs that cannot produce crossovers by simple resolution (hereafter called “off-pathway” JMs), by dynamically redirecting early strand exchange intermediates toward interhomolog, bi-chromatid recombination. We further suggest that “off-pathway” intermediates that escape Sgs1-mediated disassembly in wild-type (see signal in wild-type and ndt80Δ, Figure 2), or that form in the absence of Sgs1, are frequently resolved by Mus81/Mms4 in a ZMM-independent manner. Whether or not JM resolution is biased towards CO formation, the net effect would be that a minor fraction of COs would be Mus81/Mms4-dependent and ZMM-independent in cells where Sgs1 is active, as has been inferred in several genetic studies (Argueso et al., 2004; de los Santos et al., 2003; de los Santos et al., 2001). It is also possible that, instead of being resolved to full-length products, JMs are cleaved by Mus81/Mms4 and re-enter the recombination intermediate pool. This is unlikely to be the only way that Mus81/Mms4 redirects recombination intermediates, as CO formation by such a process would be expected to remain ZMM-dependent.
In conclusion, we show here that efficient JM metabolism and CO formation during budding yeast meiosis requires the combined activity of the Sgs1 helicase and the Mus81/Mms4 structure-specific endonuclease. We suggest that the former disassembles “off-pathway” precursors, redirecting them towards processes compatible with ordered homolog pairing and regulated crossing-over, while the latter resolves JMs that are formed by events that escape Sgs1 surveillance, and whose structure or context is incompatible with regulated resolution. It will be of considerable interest to determine whether or not Sgs1 and Mus81/Mms4 perform similar functions during the mitotic cell cycle, and to determine the meiotic phenotypes of mutants in genes that display synthetic mitotic lethality with sgs1 or with mus81/mms4 mutants (Mullen et al., 2001; St Onge et al., 2007).
Strains are listed in Supplementary Table 1 and are SK1 derivatives (Kane and Roth, 1974). For MJL2984 and derivatives, the recombination interval used contains a single DSB hotspot (Figure 2; Jessop et al., 2006). For MJL2442 and derivatives, the recombination interval contains two DSB hotspots and is less active than the interval in MJL2984 (Allers and Lichten, 2001b; Jessop et al., 2005).
The sgs1-mn (Jessop et al., 2006) and mms4-mn alleles were created by replacing the relevant promoter regions (+1 to −89 of SGS1, +1 to −90 of MMS4) with a kanMX-CLB2 promoter construct (Lee and Amon, 2003). Estrogen-inducible SGS1-IN and MUS81-IN alleles were made by replacing the same promoter regions with a GAL1 promoter fragment from pFA6a-kanMX6-PGAL1-3HA (Longtine et al., 1998) in strains that contained a GPD1 promoter-driven Gal4-estrogen receptor fusion inserted at URA3 [ura3::pGPD1-GAL4(848).ER::URA3 (Benjamin et al., 2003)]. SGS1-IN includes a 3xHA tag at the Sgs1 N-terminus; MUS81-IN does not include the 3xHA tag, as it abrogates MUS81 function (data not shown). Components used in constructing these strains were a gift from A. Amon.
3xFLAG C-terminal epitope-tagged Mus81 constructs, used to measure protein levels, were constructed using a modified p3FLAG-kanMX, (Gelbart et al., 2001) in which the kanMX6 cassette was replaced by a hygromycin-resistance cassette (hphMX, Goldstein and McCusker, 1999), integrated at the C-termini of the MUS81 or MUS81-IN open reading frames.
Details of strain construction and primer sequences are available upon request.
Liquid media, presporulation and sporulation conditions were as described (Jessop et al., 2006).
DNA isolation and Southern blot analysis were performed as described (Allers and Lichten, 2000; Allers and Lichten, 2001a). JMs at YGR174C were detected using Southern blots of XmnI-digested DNA, probed with the entire CPB4 open reading (chromosome VII coordinates 845915-846396). In graphs, values plotted with error bars reflect the mean value ± standard error of the mean for multiple, independent experiments.
Protein extracts from 2 mL of sporulating culture were prepared by TCA precipitation (Foiani et al., 1994). The equivalent of 10µL of a sporulating culture was displayed on 4–12% polyacrylamide gradient Tris-Glycine mini gels (Invitrogen) and electroblotted to PVDF membrane (Invitrogen) using the XCell II Blot module (Invitrogen) as recommended by the manufacturer. Blots were washed for at least 1 hr in 0.2% I-block (Tropix) in 1X PBS, on an orbital shaker at RT. Primary antibody, diluted in blocking buffer, was added to the blot, and incubated on an orbital shaker at RT for at least 1 hour. Blots were then washed 3 × 10 minutes with blocking buffer, and then incubated with secondary antibody for 1 hour with shaking at RT, and wash steps were repeated. Signal was developed using the chemiluminescent CPD-star substrate (Tropix) and detected using a Fuji LAS3000 CCD camera. All blots were stripped by the 2-mercaptoethanol/SDS method in the Tropix Western-Light (Applied Biosystems) technical manual, and reprobed for Arp7 to normalize for loading. Primary antisera were: Zip1 polyclonal (a gift from G.S. Roeder) at 1:1000; Arp7 polyclonal (Santa Cruz Biotechnology, Inc; sc-8961) at 1:500; anti-HA monoclonal (clone 12CA5, Roche; 11583816001) at 1:10000; FLAG M2 monoclonal (Sigma, F1804) at 1:10000. Secondary antibodies were alkaline phosphatase conjugates of goat-anti-mouse (Sigma, A3562) and goat-anti-rabbit (Sigma, A3687), both used at 1:10000.
For microtubule detection, 4 mL aliquots of sporulating cultures were fixed and stained as described (Nasmyth et al., 1990). Anti-Tub2 (a gift from M. Basrai) was used as a primary antibody at a 1:500 dilution in 1% BSA, 1X PBS, and AlexaFluor 488 goat anti-rabbit IgG antisera (Molecular Probes; A11034) was used as a secondary antibody at a 1:100 dilution. Vectashield with DAPI (Vector Laboratories, Inc, H1200) was used to simultaneously stain DNA.
Nuclear divisions were monitored by staining alcohol-fixed whole cells with DAPI (Goyon and Lichten, 1993).
We thank Angelika Amon and Gaku Mizuguchi for strains and plasmids, Shirleen Roeder and Munira Basrai for antisera, Martin Xaver for assistance with immunocytology, Neil Hunter for communicating data in advance of publication, and Cyril Buhler, Robert Shroff and Tamara Goldfarb for other assistance. We thank members of the Lichten group, Yikang Rong and Dhruba Chattoraj for discussions. This research was supported by the Intramural Research Program at the Center for Cancer Research, National Cancer Institute, National Institutes of Health.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.