|Home | About | Journals | Submit | Contact Us | Français|
In the fission yeast Schizosaccharomyces pombe, we demonstrate that meiotic DNA joint molecules, recombination intermediates encompassing two DNA duplexes, arise preferentially between sister chromatids rather than homologs. These joint molecules predominantly contain single Holliday junctions, rather than the double Holliday junctions predicted by current models for crossing-over. The Mus81 protein is required for resolution of joint molecules and the production of crossover DNA, providing further evidence that Mus81 is a component of a meiotic Holliday junction resolvase. Our results contrast with observations from budding yeast, which shows a preference for interhomolog over intersister events and generates primarily double Holliday junction intermediates. Meiotic recombination in budding yeast has become a eukaryotic paradigm despite the absence of direct observation of DNA intermediates from organisms other than yeasts. We discuss a unifying recombination model in which single Holliday junctions arise from single- or double-strand breaks, lesions postulated by previous models to initiate recombination.
Homologous DNA recombination has two important roles in eukaryotes. In mitotically growing cells it acts as a general repair mechanism, faithfully correcting broken DNA molecules. This is particularly important during replication when DNA breaks are believed to arise frequently. Homologous recombination also plays a specific role in meiosis when it both promotes genetic diversity in gametes and helps ensure the correct segregation of homologous chromosomes during the first meiotic division (MI). Two distinct products of recombination are observed genetically: gene conversions and crossovers. Gene conversion is the non-reciprocal transfer of sequence information from one homolog to another. Crossing-over is the reciprocal exchange of both DNA strands between two homologous duplexes. Gene conversion and crossing-over often occur together in a single recombination event.
The current canonical model of crossing-over (Szostak et al., 1983; Sun et al., 1991) explains both gene conversions and crossovers as arising from an initiating DNA double-strand break (DSB) (Figure 1A; see Discussion for other models). The model predicts a DNA joint molecule intermediate containing two Holliday junctions, cleavage of which can produce a crossover. The predicted double Holliday junction (double HJ) intermediates have been observed in the budding yeast Saccharomyces cerevisiae by electron microscopy (Bell and Byers, 1983) and deduced from two-dimensional (2D) gel analysis of DNA (Schwacha and Kleckner, 1995), and it has been widely assumed that double HJs are a universal precursor of crossovers. However, their existence has not, to our knowledge, been reported in any other organism.
In meiotic recombination, joint molecules can form between sister chromatids or between homologous chromosomes, since either can provide the sequence homology needed to repair a DSB. However, only interactions with the homolog can result in crossovers that reassort genetic information and aid correct segregation of chromosomes at MI. Consistent with the importance of interhomolog events, the results of a study in Locusta migratoria suggested that interhomolog crossovers outnumber sister chromatid exchanges (visualized by differential BrdU staining) (Tease and Jones, 1979). More directly, in budding yeast it was observed that interhomolog joint molecules predominate over intersister joint molecules (Schwacha and Kleckner, 1994; Schwacha and Kleckner, 1997). To our knowledge, budding yeast is the only organism in which this question has been directly addressed, and therefore the universality of interhomolog bias is untested.
Resolution of Holliday junctions (HJs) is expected to be essential for the generation of crossovers. However, despite the isolation of eukaryotic nuclear protein fractions with resolvase activity (e.g., Constantinou, Davies and West, 2001), the identity of eukaryotic nuclear resolvases remained elusive for many years. However, in 2001 it was reported that, in the fission yeast Schizosaccharomyces pombe, mutations in the mus81 gene result in the phenotypes expected of a meiotic HJ resolvase (Boddy et al., 2001). In meiotic crosses of mus81 mutants very few viable spores are produced, but among these, crossovers are greatly reduced while there is little effect on gene conversion (Boddy et al., 2001; Smith et al., 2003; Osman et al., 2003). The phenotypes of mus81 mutants can be suppressed by expression of a bacterial HJ resolvase (Boddy et al., 2001; Osman et al., 2003; Smith et al., 2003). Mus81 with its partner protein Eme1, partially purified from fission yeast, can cleave HJs and closely related DNA structures; this cleavage is abolished by amino acid replacements in the putative nuclease active site, indicating that Mus81•Eme1 participates directly in the cleavage (Boddy et al., 2001; Gaillard et al., 2003; Osman et al., 2003). Together, these results implicated Mus81•Eme1 as an important meiotic HJ resolvase. It is possible, however, that the reduced crossover frequencies seen in the few viable spores are not representative of the whole population of meiotic events.
We examined the nature and processing of recombination intermediates in fission yeast and measured the frequency of crossovers in the whole meiotic population, not just among viable spores. To do so, we required a locus with high levels of recombination in a short interval, which would allow us to observe recombination intermediates and products at high frequency using physical techniques. Apart from budding yeast, fission yeast is the only organism in which meiotic DSB sites, which can initiate recombination, have been directly observed (Cervantes et al., 2000). We previously characterized one of the strongest sites of DSB formation, the mbs1 locus of chromosome 1, and showed it is also a hotspot for gene conversion and crossing-over (Young et al., 2002; Cromie et al., 2005).
Here we use physical assays of DNA, directed to the mbs1 locus, to address four questions. First, is Mus81 required for crossovers in the bulk meiotic population, as well as among viable spores? Second, can joint molecules be detected, e.g. in mus81 mutants, at a meiotic recombination hotspot? Third, what proportion of meiotic recombination intermediates are the result of intersister versus interhomolog recombination? Fourth, what is the structure of meiotic recombination intermediates in fission yeast: do they contain double Holliday junctions?
To measure meiotic crossovers physically, we constructed diploids heterozygous for restriction sites (“L” and “R”) flanking the mbs1 recombination hotspot (Figure 2A). Crossovers measured in meiotic tetrads occur at high frequency (~5% of chromatids) in the 4.8 kb interval between these two markers (Cromie et al., 2005). We expected to detect crossover-specific fragments by probing DNA for the mbs1 region after digestion with appropriate restriction enzymes (Figure 2A). We concentrated on the smaller crossover-specific fragment (“Recombinant 2”), as the larger fragment (“Recombinant 1”) could arise from partial digestion of parental molecules.
When we examined DNA from a meiotic time-course of a wild-type (rec12+ mus81+) strain, the Recombinant 2 fragment was absent at the beginning of the meiotic induction but began to appear 3 – 4 hr later and accumulated to a final maximum level of ~3.5% of total DNA (Figures 2B and 2C). The timing of appearance of this species was as expected for crossovers –after DNA replication and DSB formation but before MI (Figure 2D and Supplemental Figures 1 and 2A). Its accumulation, in contrast to the transient meiotic DSBs in the same strain, reflects crossovers being final products of the recombination pathway. The final frequency of the physical crossover products was comparable to the crossover level measured genetically (Cromie et al., 2005). Similar results of physical assays were seen using a diploid in which the L and R markers were coupled rather than oriented in repulsion (unpublished data).
We next examined the dependence of crossing-over on Mus81. Based on the quantitation of the Recombinant 2 fragment, crossovers appeared in mus81 mutant diploids with the same timing as in mus81+, but they accumulated to a much lower level, ~0.8% rather than ~3.5% (Figures 2B and 2C and see below). This demonstrates that crossovers are reduced in the whole population of meiotic cells in a mus81 mutant, not just in the ~0.1% that form viable spores (Boddy et al., 2000; Osman et al., 2003; Smith et al., 2003). Meiotic replication occurred in the mus81 mutant diploid at nearly the same time as replication in mus81+ (Supplemental Figure 1), but the high frequency of abnormal nuclei present at all time points in the mus81 mutant precluded measurement of the timing of the meiotic divisions. However, asci were observed after 24 hr in mus81 mutant meiosis just as in mus81+, demonstrating that progression through meiosis did occur.
To confirm that the recombinant fragments observed in both the mus81+ and mus81 mutant backgrounds represented bona fide meiotic recombination products, we tested their dependence on Rec12, the S. pombe ortholog of Spo11, which is the active site protein that makes DSBs in S. cerevisiae (Keeney, 2001). Rec12 is essential for meiotic DSB formation and recombination (DeVeaux et al., 1992; Cervantes et al., 2000). The rec12 mutation abolished recombinant DNA formation in both mus81+ and mus81 mutant backgrounds (Figure 2B and legend for 2C), while DNA replication was essentially unaffected (Supplemental Figure 1).
Although crossovers are reduced in the mus81 mutant, meiotic DSBs form at similar frequencies as in mus81+ cells and disappear with similar kinetics (Young, Hyppa and Smith, 2004; Supplemental Figure 2). We conclude that the recombination defect in a mus81 mutant occurs after DSB formation and disappearance, but before the formation of crossovers, i.e. in the processing of joint molecules (JMs).
To identify the JMs that give rise to crossovers, we used two-dimensional gel electrophoresis. In this assay, DNA molecules separate in the first dimension based primarily on their mass and in the second dimension based on both mass and structure. Replication forks and bubbles, linear DNA, and branched molecules containing HJs, all run in diagnostic positions on such gels (Brewer and Fangman, 1987; Figure 3A). Branched molecules containing X-shaped structures, such as HJs, run as a characteristic “spike.”
Analysis of two-dimensional gels probed for DNA containing the mbs1 locus revealed X-form molecules arising both from DNA replication and from recombination. Using DNA from a rec12+ mus81+ meiotic time-course, we observed replication intermediates at 2 – 3 hr after meiotic induction (Figure 3B), as expected from flow cytometry (Figure 2D and Supplemental Figure 1). These replication intermediates included Y-shaped species, but also a spike corresponding to X-form DNA. This X-form species has been seen during replication in previous studies (e.g. Segurado et al., 2003). However, X-form material was also seen at 4 – 5 hr (Figure 3B), when DNA replication was complete (Supplemental Figure 1) and replication forks had disappeared (Figure 3B). The lack of distinctive replication structures and the correlation with the expected timing of recombination suggested that the 4 and 5 hr X-form material consisted of recombination-related JMs, i.e. homologs or sisters held together by Holliday junctions. If so, we would expect this material, but not the replication intermediates, to depend on Rec12. The X-form species at 2 – 3 hr did not depend on Rec12, but those at 4 – 5 hr did (Figures 3B and 3C). Similarly only the 2 – 3 hr species were seen in a rad50S mutant, in which meiotic DSBs are not repaired (unpublished data). Therefore, we conclude that X-form molecules seen at 4 – 5 hr are, indeed, recombination intermediates.
The Rec12-dependent JMs reached a maximum of 0.8% at 4.5 hr and then disappeared (Figures 3B and 3C). The timing of maximum JM appearance suggests that formation and resolution of these intermediates occurs between DNA replication and MI (Figure 2D; Supplemental Figure 1), as expected for recombination intermediates.
We tested if Rec12-dependent JMs accumulated in a mus81 mutant to higher levels than in mus81+, as expected if Mus81 is a component of an HJ resolvase. In the mus81 mutant, X-form species were observed during replication (i.e. at 2.5 and 3 hr) and also after completion of replication at 4 hr and later (Figure 3B), similar to our observations with mus81+ cells. Only the X-form species observed at 4 hr and later were dependent on Rec12 in the mus81 mutant (Figure 3C), indicating that these molecules are recombination intermediates. As expected, these recombination JMs accumulated to a higher level in the mus81 mutant than in mus81+ (a maximum of 2.2% compared to 0.8% in mus81+ at 4.5 hr) and persisted, although a reduction in frequency was seen at later time points (Figure 3C). In contrast, no increase in the levels of replication Y- and X-form intermediates (at 3 hr) was seen in the mus81 mutant (Supplemental Figure 3 and Figure 3C). We conclude that mus81 mutants accumulate JMs and fail to produce crossovers, as predicted if JMs are resolved by Mus81•Eme1 into crossovers.
We expected that JMs would be seen at hotspots for breakage and crossing-over, but not at sites that had few meiotic DSBs. Consistent with this expectation, in the mus81 mutant the frequencies of JMs at 5 hr were higher at the mbs1 and mbs2 hotspots than at similarly sized regions with few, if any, DSBs located in the same cosmids (Figure 3D; Young et al., 2002). A similar pattern was seen in mus81+ (unpublished data).
The presence of the heterozygous L and R markers flanking mbs1 allowed us to determine the parental origin of JMs at mbs1. After digestion with PvuII, PmlI (specific for L) and XbaI (specific for R), three distinct mbs1 JM species should be observable, intersister 1 (“P1 X”), intersister 2 (“P2 X”), and interhomolog (“IH X”) (Figure 4A). These three species have different masses allowing separation in the first dimension of two-dimensional gels.
Two dimensional gel analysis of DNA from the 5 hr time-point of mus81 mutant and mus81+ meioses, triply digested and probed for mbs1, revealed multiple JM species (Figure 4B). At this time X-form DNA corresponds to recombination intermediates (see above). In both backgrounds two clear X-form spikes were present, with two different masses (Figure 4B). We confirmed that these represented the two types of intersister JMs by comparing their mobilities to those of JMs from haploid strains induced to initiate meiosis (unpublished data). In haploid cells only intersister 1 or intersister 2 intermediates can form. In diploids, in addition to the two intersister X-form spikes, we observed two prominent forms lying between the intersister species (Figure 4B). Their position corresponds to the expected mass of interhomolog JMs. We believe that these molecules appear as two forms rather than one because of the asymmetric structure of the interhomolog JMs (see below and Discussion). At higher exposures a weak “tail” of species is seen joining the two prominent interhomolog forms. This is similar to the distribution of the intersister species, with material concentrated at the top of a weaker spike. Both the two interhomolog forms and the intersister X-form spikes were observed at 5 hr in mus81 mutant and mus81+ inductions, while at 3 hr only the intersister X-form spikes were observed (Figure 4B). This supports our conclusion that the two forms between the intersister spikes represent recombination intermediates: X-form replication intermediates (at 3 hr) should be only intersister while recombination intermediates (at 5 hr) could also be interhomolog. As expected, Y-form replication intermediates from each parent (Figure 4A, “P1 Y” and “P2 Y”) were also seen at 3 hr (Figure 4B).
In both mus81+ and mus81 mutant cells the three X-form species at 4 hr and later depended on Rec12; i.e. they were all bona fide meiotic recombination intermediates (unpublished data). All of these recombination JMs accumulated to higher levels in the mus81 mutant than in mus81+ cells (Figure 4C). This suggests that Mus81 is required to resolve both intersister and interhomolog JMs.
In both mus81 mutant and mus81+ DNA the intersister JMs outnumbered the interhomolog molecules. In the mus81 mutant, at 5 hr the frequency of the intersister 1 species was 1.1%, the intersister 2 species 0.8%, and the combined interhomolog species 0.5% (Figures 4B and 4C). The total frequency of these three species (2.4%) is almost identical to the value for the combined JMs measured after PvuII digestion (2.3%; Figure 3C). However, together the two intersister species were approximately 4-fold more frequent than the interhomolog species. Due to low level DSBs and HJs between R and the rightward PvuII site (unpublished data), the actual ratio of intersister to interhomolog JMs in the interval L and R is closer to three to one. Therefore, at the mbs1 site of fission yeast it appears that there is a bias towards intersister recombination in contrast to the preference for interhomolog events seen in budding yeast (Schwacha and Kleckner, 1994, 1995).
To investigate the structure of the meiotic recombination intermediates observed at mbs1, we looked for evidence that they contained Holliday junctions. We did this in three ways: examining the sensitivity of the intermediates to a known HJ resolvase in vitro; testing the ability of high temperature to resolve the intermediates to linear forms by branch migration; and examining the intermediates directly by electron microscopy (EM).
In both the mus81+ and mus81 mutant backgrounds the recombination-derived JMs observed at mbs1 were sensitive to E. coli RuvC enzyme, a well-characterized HJ resolvase (Connolly et al., 1991), while linear DNA was not. This was true both for the combined JM population (after PvuII single digestion) and the distinct intersister and interhomolog forms (after PvuII, PmlI, XbaI triple digestion) (Figure 5A). While Holliday junctions are the preferred substrate of RuvC, it can cleave other branched DNA species, albeit with lower efficiency (Benson and West, 1994; Fogg et al., 1999). The preparation of RuvC used in Figure 5 showed a distinct preference for chemically synthesized X-shaped molecules, as expected (unpublished data). Therefore, these results indicate that the JMs are held together by HJs or perhaps other branched structures.
Incubation at high temperature causes HJs to branch-migrate and to be resolved into linear DNA when the HJs reach the ends of fully homologous DNA. We tested the ability of high-temperature incubation to resolve the intersister and interhomolog JMs formed in a mus81 mutant. As expected, the intersister forms were largely resolved to the corresponding linear fragments (31% unresolved) (Figure 5B). However, the interhomolog material was almost entirely resistant to heat treatment (98% unresolved). We believe that this resistance is explained by the asymmetric structure of the interhomolog JMs and the presence of single, rather than double, HJs. When branch migration of a single HJ reaches the left or the right end of one duplex present in the interhomolog JM, one or another stable Y-shaped structure is produced rather than resolution to linear DNA (Figure 5C). These two stable structures have the same mass but different shapes. We conclude that the two spots represent these two different forms: they ran at nearly the same position in the first (mass) dimension but ran differently in the second dimension (where shape is important) (Figure 4B). In contrast, interhomolog double HJs should be resolved into linear fragments by branch migration (Figure 5C).
To look more directly at the structure of total cellular JMs, we extracted DNA from a position above the arc of linear DNA, where JMs run, and examined the DNA by EM. In all preparations we saw branched molecules, including Y-shaped molecules and X-shaped molecules with unequal arm lengths; these are most likely replication intermediates or structures derived from them. HJ recombination intermediates are expected to have two short arms of identical lengths and two long arms of identical lengths. We used this criterion to designate molecules as HJs. As in the 2D gel analyses above, HJs were seen by EM in DNA prepared 5 hr after meiotic induction of mus81+ and mus81 mutant strains (Figure 6A and Supplemental Figures 4 – 7) but not in DNA from the rec12 mus81 double mutant. As expected from Southern blot analysis (Figure 3C), HJs appeared more abundant in DNA from mus81 mutants than from mus81+, comprising ~20% and ~1% of observed branched molecules, respectively (Supplemental Table 1).
In accord with the 2D gel analyses above, the great majority of the JMs that we observed by EM in DNA from fission yeast meiosis contained single, rather than double, HJs (Figure 6A, Supplemental Table 1 and Supplemental Figures 4 – 7). In the mus81 mutant 32 of 38 HJs, and in mus81+ 4 of 4 HJs, were single. Some of these HJs had an open center at the crossover position (Figure 6A and Supplemental Figure 5), unambiguously identifying them as single HJs. In contrast, budding yeast meiotic DNA prepared and examined in the same manner contained a majority of clear double HJ structures (21/26, Figures 6B); the remainder were single HJs. Another difference between the budding yeast and fission yeast HJs was the distance between the junctions in the double HJs (Supplemental Table 2). In 20 molecules of budding yeast DNA this distance ranged from 0.1 to 0.5 kb. In contrast, among the 6 double HJs observed in fission yeast three were separated by distances much larger than those observed in budding yeast (1.6, 2.1 and 2.6 kb). The other three fission yeast double HJs had 0.2 to 0.5 kb separating the individual HJs. This suggests that at least some of the fission yeast double HJs represent a different class than those seen in budding yeast, perhaps arising from two closely-spaced, independent recombination events (see Discussion). We conclude that, in contrast to budding yeast, meiotic recombination in fission yeast proceeds primarily through single HJs rather than double HJs.
To investigate the mechanism of homologous recombination in fission yeast, we examined the intermediates and products of meiotic recombination using physical methods. We concentrated on recombination at mbs1, a naturally occurring hotspot of DSBs and recombination. To analyze recombination at mbs1, we made only two single base-pair mutations so that the chromosomes were as close to wild type as possible. Our results demonstrate that the interhomolog recombination bias and the double HJ structure of recombination intermediates seen in budding yeast meiosis are not universal. Instead, we saw predominantly single HJs and a strong bias to intersister recombination. Our results also provide additional evidence that Mus81 is a component of a fission yeast meiotic HJ resolvase.
By two dimensional gel analysis we observed meiotic JMs with two characteristics expected of recombination intermediates: a mass greater than that of their linear parents and an “X” structure. Are these JMs genuine recombination intermediates or the products of a side pathway? Genuine intermediates of biochemical pathways, such as recombination, meet four criteria and the JMs observed in this study meet all four of these criteria. First, there should be mutations blocking product formation, which cause the molecules to accumulate. In this study, mus81 mutants are crossover-defective and caused accumulation of the JMs (Figures 3B and 3C; Smith et al., 2003; Osman et al., 2003). Second, there should be mutations acting earlier in the relevant pathway that prevent the molecules from forming. rec12 mutations were epistatic to mus81 (Boddy et al., 2001), and here we observe that rec12 mutants were recombination-defective and lack JMs (Figures 2C legend and 3C; DeVeaux et al., 1992). Third, the timing of appearance and disappearance of the molecules should be consistent with other features of the pathway. We saw that Rec12-dependent JMs appeared after DNA replication and DSB formation and before the first meiotic division, i.e. at the time of meiotic recombination (Figures 2D and 3C; Supplemental Figure 1). Finally, the molecules should be convertible into the final product of the relevant pathway. We cannot introduce the JMs into cells and follow their conversion to crossovers, but in vitro they are substrates of the E. coli HJ resolvase RuvC (Connolly et al., 1991; Figure 5A) and in vivo another E. coli HJ resolvase RusA can substitute for Mus81•Eme1 in production of crossovers (Boddy et al, 2001; Osman et al., 2003; Smith et al., 2003). Meeting these four criteria strongly suggests that the JMs we observe are genuine intermediates of meiotic homologous recombination.
Perhaps our most surprising conclusion is that, in fission yeast, the great majority of recombination intermediates contain unexpected meiotic recombination structures – single Holliday junctions. A preponderance of single HJs was inferred from the behavior of interhomolog JMs on two-dimensional gels (Figures 4B and and5B)5B) and was demonstrated directly by the observation of mostly single HJs, rather than double HJs, by EM (Figure 6A, Supplemental Table 1, and Supplemental Figures 4 –7). The junctions in these molecules may be fully intact HJs or may have contained a single-strand nick at their inception; in mus81+ cells such nicked HJs might be quickly resolved into products (Osman et al., 2003; Hollingsworth and Brill, 2004). A nick should pose no impediment to branch migration. Thus, a hypothetical nicked HJ, if unresolved, could quickly be converted into an intact HJ. Our results are consistent with the HJs observed here being either nicked or intact.
If the intersister bias observed at mbs1 extends throughout the genome, then most of the HJs observed by EM are likely to be intersister. To our knowledge there is no difference in the mechanism of intersister and interhomolog recombination. Consequently, it is simplest to assume that interhomolog as well as intersister JMs contain mostly single HJs, as indicated by the analysis of interhomolog JMs at mbs1 (Figure 5B).
The small number of double HJs observed in fission yeast by EM argues strongly against their being major recombination intermediates in this organism unless one HJ is almost immediately resolved or the junctions are so widely spaced (>5 kb) as to produce mostly single HJs after restriction enzyme digestion. Single HJs were seen by EM in wild type as well as the mus81 mutant, arguing against the possibility that closely spaced double HJs in wild type, similar to those seen in S. cerevisiae, become widely spaced in the absence of Mus81. We believe the possibility of very widely spaced HJs can be discounted for several reasons. All markers between the junctions of a double HJ could be co-converted, yet the gene conversion tracts observed in fission yeast, like those in budding yeast, generally are continuous and span <1 kb (Grimm, Bahler and Kohli, 1994; Cromie et al., 2005). Two interhomolog species appear to reflect heteroduplex DNA at R (Figure 4B); these species were rare, indicating that branch migration from mbs1 across R was uncommon. These data indicate that heteroduplex DNA and associated HJ(s) lie predominantly between L and R, consistent with most gene conversion events around mbs1 being between L and R (Cromie et al., 2005). Finally, the mean length (± SD) of the shorter arms of the single HJs observed by EM was 2.4 ± 1.0 kb, indicating that a second HJ was not located within that distance. Branch migration of HJs beyond this range during sample preparation is unlikely, since the DNA for EM analysis contained psoralen crosslinks ~1 kb apart (unpublished data); any branch migration during preparation would have been limited to about that distance.
The few double HJs seen in fission yeast might arise from two closely spaced, independent recombination events. Such events are expected in fission yeast due to the occurrence of DSBs in ~1 – 2 kb clusters at hotspot sites (Cromie et al., 2005; Steiner et al., 2002) and the absence of crossover interference (Munz, 1994), allowing two independent, closely spaced HJs to arise. The more variable, and sometimes much larger, separation of the individual junctions from fission yeast double HJs, compared to those from budding yeast (Supplemental Table 2), supports this notion. The rare fission yeast double HJ molecules with junction separations similar to those in budding yeast may also arise from two independent events or may represent a minor pathway that utilizes double HJs as intermediates in individual recombination events.
In contrast to fission yeast, in meiotic DNA from budding yeast we observed by EM a majority of double HJs (Figure 6B), along with a significant number of single HJs, as reported previously by Bell and Byers (1983). Single strands with a length indicative of double HJs, not single HJs, were seen in JMs from budding yeast (Schwacha and Kleckner, 1995). The detection threshold of this assay, however, does not preclude a significant fraction of single Holliday junctions. Budding yeast appears to have several pathways for generating meiotic crossovers (de los Santos et al., 2003; Argueso et al., 2004); the major pathway may involve double HJs and a minor pathway single HJs. In contrast, fission yeast may have only one pathway of crossing-over, dependent on Mus81•Eme1 and involving single HJs.
The model of Szostak et al. (1983) as modified by Sun et al. (1991) predicts the existence of double HJs because an HJ is formed by each end of the initiating DSB (Figure 1A). To produce a single HJ, the two ends would have to behave differently so that only one would produce an HJ. In fact, in the model of Szostak et al. (1983) the two ends of the DSB do not behave identically: only one end invades a duplex, while the other end is “captured” by annealing of single strands. One of these processes could generate an HJ and the other a different structure (see below). The difference in behavior of the two ends could be a simple matter of timing, e.g. the first end to find homology could carry out strand invasion and then the other end would get “captured.” Alternatively, recent evidence suggests that the two ends of the DSB are processed differently at an early stage, when Spo11 is removed (Neale, Pan and Keeney, 2005). This could actively direct the two ends towards different biochemical events.
In Figure 1B we propose a recombination model that is initiated by a DSB but has a single HJ as a recombination intermediate. As in the model of Szostak et al. (1983), the first DNA end generates a D-loop with an HJ at the right end in Figure 1B. Unlike the model of Szostak et al. (1983), the left end of the D-loop is cut before second end capture. This results in a JM containing a single HJ. We propose that Mus81 is required for resolution only of the single HJ and that the D-loop is cut by a different enzyme. In accord with this proposal, we observed accumulated HJs but not D-loops in a mus81 mutant. Our observation of single HJs by EM suggests that the putative D-loop cleavage is more rapid than HJ resolution. Interestingly, this pathway can accommodate recombination initiated by single-strand nicks, which have been proposed as recombinogenic lesions in many previous models (e.g. Holliday, 1964; Meselson and Radding, 1975; Radding 1982) and which may account for the frequent crossovers in intervals with few or no observed DSBs in fission yeast (Young et al., 2002; Cromie et al., 2005).
Our second surprising conclusion is that, in fission yeast meiosis, intersister recombination is preferred over interhomolog recombination. In fission yeast, meiotic intersister recombination occurs at a significant frequency (Schuchert and Kohli, 1988), but no direct comparison of intersister versus interhomolog frequencies has been made before now. Based on the relative frequency of JMs at the mbs1 recombination hotspot, intersister events outnumber interhomolog events by ~3 to 1 (Figure 4C and Results). This reverses the bias towards interhomolog JMs seen in budding yeast (Schwacha and Kleckner, 1994, 1997). Intersister bias at mbs1 is also supported by genetic data. Since 80% of interhomolog events (conversions) at mbs1 produce crossovers (Cromie et al., 2005), 3.5% crossing-over at mbs1 (Figure 2C) would require only 2.2% DSBs (2 × 2.2% × 0.8 = 3.5%). However, we observe 10 – 11% breakage at mbs1 (Supplemental Figure 2B; Young et al., 2002), implying that ~80% of mbs1 DSBs undergo sister chromatid repair, consistent with our physical analysis.
An interhomolog bias utilizes recombination more effectively with respect to chromosome segregation and reassortment of genetic information, as these are promoted only by interhomolog recombination. Hence, it has been widely assumed that interhomolog bias is universal, even though prior to this study budding yeast was the only organism to our knowledge in which the question had been directly addressed. However, despite this argument, intersister events may be favored by the close proximity of sister chromatids. The interhomolog bias seen in budding yeast could reflect an active system overcoming a mechanistic intersister bias, which may be identical to the barrier to intersister recombination seen in that organism (Schwacha and Kleckner, 1997; Niu et al., 2005). This active system is presumably absent from fission yeast. One well-known example of a regulation of recombination in budding yeast, but not in fission yeast, is crossover interference. Interference may result from a mechanism to ensure that even the small chromosomes of budding yeast, which has 16 chromosomes, receive at least one crossover and undergo proper MI segregation (Roeder, 1997; Hillers, 2004). In contrast, fission yeast, with three long chromosomes, achieves the same goal by simple random distribution of many (10 – 20) crossovers per chromosome. Interhomolog bias in budding yeast may be a further mechanism to ensure at least one interhomolog crossover per short chromosome.
Using a physical assay, we showed that in the absence of Mus81 crossovers are greatly reduced at the mbs1 recombination hotspot. A major advantage of physical assays of recombination is the ability to characterize the whole population of meiotic cells. This is helpful in backgrounds, such as a mus81 mutant, where viable spore yield is low and events in those viable spores may be atypical. Our physical assay showed that, in a mus81 mutant, crossovers are reduced in the whole population of meiotic cells (Figure 2C). In addition, we showed by two-dimensional gel analysis and EM that JMs containing HJs accumulate in a mus81 mutant during meiosis (Figure 3C and Supplemental Table 1). HJ-like structures also accumulate in mus81 mutants during mitotic replication of the highly repeated rDNA (Noguchi et al., 2004). Together with previous results (see Introduction), these data provide strong evidence that Mus81 is required for HJ resolution.
The frequency of interhomolog JMs in the mus81 mutant (0.5%) is too low to explain the ~3.5% crossing-over seen at mbs1 (Figure 2C). However, this assumes that in a mus81 mutant HJs accumulate without any loss. In fact, JMs decline at late time points in a mus81 mutant (Figure 3C). It is unclear if this represents an alternative, minor HJ resolution pathway, perhaps to non-crossovers, or simple deterioration of unresolved JMs in the cell. Therefore, it is likely that the true cumulative frequency of both intersister and interhomolog JMs is greater than that seen in a mus81 mutant.
The effect of a mus81 mutation on crossover frequency using our physical assay was somewhat less than that seen genetically among viable spores. Instead of a 20 – 90-fold reduction in crossover frequency (Osman et al., 2003; Smith et al., 2003) we observed only a 4-fold reduction (Figure 2C). The genetic data may overestimate the effect of the mus81 mutation in the total population: viable spores may have an unusually low number of DSBs and hence fewer potentially lethal, unresolved HJs. Alternatively, the physical assay may underestimate the effect of the mus81 mutation: genetic analysis showed that 14% of exchanges between the L and R markers involved conversions of L or R and were not simple crossovers (Cromie et al., 2005). Since mus81 mutations have little effect on gene conversion (Osman et al., 2003; Smith et al., 2003), the residual recombinant DNA observed in mus81 mutants using our physical assay may reflect gene conversions of L or R: crossovers may be nearly abolished.
Mus81 is conserved across a wide range of eukaryotes, suggesting that it has a conserved function. However, the meiotic phenotype of mus81 mutants is different in fission yeast than in other organisms examined, i.e. budding yeast and mice. In budding yeast, mus81 mutants display somewhat reduced spore viability but only a modest reduction in crossover frequency (Interthal and Heyer, 2000; de los Santos et al., 2003), and mus81−/− mice are fertile (McPherson et al., 2004; Dendouga et al., 2005). Differences between the two yeast species may reflect the presence of at least one other crossover pathway in budding yeast (requiring Msh4-Msh5) that is absent in fission yeast. Mutations in budding yeast msh4-msh5 and mus81-mms4 have additive effects on crossover frequency (de los Santos et al., 2003; Argueso et al., 2004), suggesting that Mus81 contributes to a specific subset of HJ resolution events. Msh4-Msh5 homologs are found in mice, and multiple pathways of HJ resolution could reconcile the fertility of mus81−/− mice with a hypothetical role for Mus81 in meiotic HJ resolution in mice. In contrast to the meiotic phenotypes, the mitotic phenotypes of mus81 mutants are very similar in budding and fission yeast, e.g. high sensitivity to agents such as camptothecin that are believed to cause DSBs at replication forks (Doe et al, 2002; Vance and Wilson, 2002). Broken replication forks are likely to have only a single duplex end and may have to be repaired using an HJ intermediate. In contrast, DSBs caused by ionizing radiation have two ends and can use the synthesis-dependent strand annealing (SDSA) recombination pathway to repair breaks without generating HJs. Budding yeast mus81 mutants are not hypersensitive to ionizing radiation (Interthal and Heyer, 2000). Therefore, the mitotic phenotypes of mus81 mutants in budding yeast and fission yeast are both consistent with a role for Mus81 in mitotic HJ resolution. The additional Msh4-Msh5-dependent crossover pathway in budding yeast and mice is meiosis-specific (Kunz and Schar, 2001; Her et al., 2001). Mus81 may be a universal eukaryotic mitotic HJ resolvase while additional meiosis-specific resolution pathways may exist in some organisms, but not in others, such as fission yeast.
In this study we show that budding and fission yeast differ with respect to two major features of meiotic recombination. The interhomolog bias in budding yeast contrasts with intersister bias in fission yeast. The predominantly double Holliday junction intermediates in budding yeast are mostly, or entirely, replaced by single Holliday junctions in fission yeast. The budding yeast features were assumed to represent a universal paradigm for meiotic recombination, a view that is no longer tenable. Given the different behavior of the only two organisms studied in these regards to date, it will be interesting to see whether other organisms resemble one of these or display their own novel features.
pat1-114 strains described in the Supplemental Material were thermally induced for meiosis and analyzed for DNA content by flow cytometry as described by Cervantes et al. (2000). Cells imbedded in agarose plugs were lysed with enzymes and treated with proteinase K and RNase A; the DNA was digested with appropriate restriction enzymes and analyzed by gel electrophoresis and Southern blot hybridization as described by Young et al. (2002) and detailed in the figure legends and Supplemental Material. Psoralen crosslinked DNA (Schwacha and Kleckner, 1994) was extracted from agarose gels and analyzed by electron microscopy as detailed in Supplemental Material. Inclusion of psoralen crosslinking had no discernable effect on DNA analyzed by gel electrophoresis (Supplemental Figure S8).
We are grateful to Ken Marians (Memorial Sloan-Kettering Cancer Center) for the kind gift of RuvC enzyme, Bonny Brewer (University of Washington) and Michael Lichten (National Cancer Institute) for helpful discussions about our results, and Sue Amundsen and Luther Davis for helpful comments on the manuscript. This work was supported by research grants GM031693 to G.R.S. and GM074223 to N.H. from the NIH.
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.