Top3-Rmi1 DNA single-strand decatenase is integral to the formation and resolution of meiotic recombination intermediates

Summary

The topoisomerase III (Top3)-Rmi1 heterodimer, which catalyzes DNA single strand passage, forms a conserved complex with the Bloom's helicase (BLM, Sgs1 in budding yeast). This complex has been proposed to regulate recombination by disassembling double-Holliday junctions in a process called dissolution. Top3-Rmi1 has been suggested to act at the end of this process, resolving hemicatenanes produced by earlier BLM/Sgs1 activity. We here show that, to the contrary, Top3-Rmi1 acts in all meiotic recombination functions previously associated with Sgs1, most notably as an early recombination intermediate chaperone, promoting regulated crossover and noncrossover recombination, and preventing aberrant recombination intermediate accumulation. In addition, we show that Top3-Rmi1 has important Sgs1-independent functions that ensure complete recombination intermediate resolution and chromosome segregation. These findings indicate that Top3-Rmi1 activity is important throughout recombination, to resolve strand crossings that would otherwise impede progression through both early steps of pathway choice and late steps of intermediate resolution.

Introduction

The repair of DNA double-strand breaks (DSBs) by homologous recombination is central to the maintenance of genome integrity. Recombination acts during the S- and G2-phases of the mitotic cell cycle to repair DSBs that arise from genotoxic DNA damage or as a consequence of dysfunctional chromosome replication (Heyer et al., 2010). During meiosis, programmed DSB formation by the Spo11 protein (Keeney, 2008) and repair by recombination promotes progressive association and alignment of homologs (homologous chromosomes of different parental origin) during the prophase of the first meiotic division (meiosis I), and the crossover (CO) products of homologous recombination provides stable inter-homolog connections that promote homolog segregation at meiosis I (Bhalla and Dernburg, 2008; Nicklas, 1997).

DSB repair by homologous recombination can proceed via several different mechanisms. During S and G2 of the mitotic cell cycle and during meiosis I prophase, DSB ends are resected to produce 3’ single-strand overhangs, which invade homologous duplex sequences and form early strand invasion displacement-loop (D-loop) intermediates (Symington and Gautier, 2011). If a D-loop is disassembled after the invading strand primes DNA synthesis, the two DSB ends can anneal and form a noncrossover (NCO) product in a process called synthesis dependent strand annealing (SDSA; Nassif et al., 1994). Alternatively, if an D-loop is not disassembled, the other end of the DSB can be captured to form a double Holliday junction (dHJ) joint molecule (JM) (Szostak et al., 1983). This intermediate can be resolved to form NCOs in a process called dissolution, hypothesized to involve helicase-driven inward migration of Holliday junctions followed by topoisomerase-mediated decatenation (Bizard and Hickson, 2014). Alternatively, as described in the canonical double-strand break-repair model, a dHJ intermediate can be resolved by cleavage of the two Holliday junctions to produce either COs or NCOs, depending on the relative orientation of the cleavage at the two junctions (Szostak et al., 1983). Mus81-Mms4/Eme1, Yen1/Gen1, and Slx1-Slx4, structure selective nucleases that hereafter will be called SSNs, have been identified as candidate resolvases (Schwartz and Heyer, 2011), especially during the mitotic cell cycle (Blanco et al., 2014; Matos et al., 2011).

RecQ family helicases, in particular the BLM helicases (Sgs1 in budding yeast), regulate the balance between different recombination mechanisms (Bernstein et al., 2010). BLM/Sgs1 is part of a complex containing topoisomerase III (Top3) and Rmi1 (BLAP75 in mammals), two other conserved proteins (Chen and Brill, 2007; Gangloff et al., 1994; Mullen et al., 2005). This BTR/STR complex has important roles in replication and recombination intermediate metabolism. It prevents the accumulation of potentially toxic branched DNA structures that, if not resolved by SSNs, would prevent chromosome segregation. Biochemical studies have identified activities, of Sgs1/BLM alone or of Sgs1/BLM-Top3-Rmi1, that could promote NCO formation; these include D-loop disassembly and dHJ dissolution (Bachrati et al., 2006; Cejka et al., 2010; Plank et al., 2006; van Brabant et al., 2000; Wu et al., 2006). The in vivo relevance of these activities is supported by observations of elevated levels of mitotic crossing-over in sgs1 and BLM^−/− mutants (Chaganti et al., 1974; Ira et al., 2003; Watt et al., 1996), and from studies indicating a major role for Sgs1 in regulating recombination pathway choice during meiosis (Klein and Symington, 2012). Top3 and Rmi1 also form a heterodimer independent of Sgs1 (Mullen et al., 2005). Top3-Rmi1 has a weak relaxation activity but robustly catalyzes reaction involving passage of one DNA single strand through another (Cejka et al., 2012), an activity that is necessary during the last steps of dHJ dissolution (Bizard and Hickson, 2014). In mitotic cells, top3Δ and rmi1Δ mutants, like sgs1Δ mutants, display a hyper-recombination phenotype, are sensitive to DNA damaging agents, and show synthetic lethality with mus81, mms4, slx1 or slx4 mutants, which is suppressed by abolishing homologous recombination, suggesting that the absence of any STR component results in persistent recombination intermediates that require SSNs for resolution (Mullen et al., 2001; Mullen et al., 2005; Shor et al., 2002; Wallis et al., 1989). However, unlike sgs1 mutants, top3 and rmi1 mutants have a slow-growth phenotype that is suppressed by preventing the strand invasion step of recombination, or by eliminating Sgs1 helicase (Gangloff et al., 1994; Mullen et al., 2005; Oakley et al., 2002). This suggests that, in the absence of Top3 or Rmi1, Sgs1 creates structures that cannot be resolved by SSNs or by other mechanisms.

During budding yeast meiosis, COs and NCOs products form by mechanistically and temporally distinct pathways. NCOs and JM intermediates form at the same time, but most COs do not appear until JMs resolve (Allers and Lichten, 2001). Formation of stable JMs, but not NCOs, requires the ZMM proteins, a set of meiosis-specific proteins associated with the synaptonemal complex (Borner et al., 2004; Lynn et al., 2007). JM resolution and CO formation require the yeast polo kinase, Cdc5, whose meiotic expression is driven by the Ndt80 transcription factor (Chu and Herskowitz, 1998), while most NCOs are independent of Cdc5 and Ndt80 (Allers and Lichten, 2001; Clyne et al., 2003; Sourirajan and Lichten, 2008). In addition, mutants in the Mlh1-Mlh3-Exo1 (MutLγ-Exo1) complex reduce meiotic COs without affecting JM formation (Argueso et al., 2004; Zakharyevich et al., 2010; Zakharyevich et al., 2012). This indicates that the MutLγ-Exo1 complex mediates biased, CO-only resolution of JMs.

These findings, and limited findings in other organisms (Kohl and Sekelsky, 2013), suggest that most meiotic recombination proceeds through two major pathways: D-loop disassembly to form NCOs by SDSA; and stabilization of early intermediates by ZMM proteins to form JMs between homologs that later resolve as COs in a Cdc5-triggered, MutLγ-Exo1-dependent mechanism. A third pathway, here called the alternative (ALT) pathway, normally plays a minor role. It is ZMM-independent and forms inter-homolog JMs, inter-sister JMs and multichromatid JMs that link three or four chromatids. These JMs are not resolved by MutLγ-Exo1, but rather are resolved by SSNs in a Cdc5-dependent process that produces both COs and NCOs (Argueso et al., 2004; Blanco et al., 2014; de los Santos et al., 2003; De Muyt et al., 2012; Jessop and Lichten, 2008; Matos et al., 2011; Oh et al., 2007; Oh et al., 2008; Zakharyevich et al., 2012). Sgs1 plays a central role in controlling flux through these three pathways. sgs1 mutants do not form early NCOs, form JMs independent of ZMM proteins, display elevated levels of multichromatid and inter-sister JMs relative to inter-homolog JMs, and resolve JMs in an unbiased SSN-dependent and MutLγ-Exo1-independent manner, producing both COs and NCOs (De Muyt et al., 2012; Jessop and Lichten, 2008; Jessop et al., 2006; Oh et al., 2007; Oh et al., 2008; Rockmill et al., 2003; Zakharyevich et al., 2012). Thus, Sgs1 promotes both the early NCO-SDSA and the ZMM-dependent CO-only pathways, and prevents events from proceeding through the ALT pathway. We suggested that Sgs1 accomplishes this by acting as a recombination intermediate chaperone, disassembling early recombination intermediates that would otherwise populate the ALT pathway, thus allowing further rounds of strand invasion and, in effect directing most events towards either NCO formation by SDSA or ZMM-and MutLγ-Exo1-dependent CO formation (De Muyt et al., 2012).

Considerably less is known about the meiotic roles of the other two STR complex proteins, Top3 and Rmi1. sgs1 mutants progress through meiosis with only slightly reduced spore viability (Jessop et al., 2006; Rockmill et al., 2003), but top3Δ mutants do not complete meiotic nuclear divisions or form spores (Gangloff et al., 1999). Although sgs1Δ restores sporulation to a small fraction of top3Δ cells, the extent of suppression is much less than is observed in mitotic cells, and most top3Δ sgs1Δ cells fail to divide nuclei or form spores. These progression defects are suppressed by eliminating meiotic recombination, but it cannot be excluded that, in top3 null mutants, lesions formed during premeiotic growth or meiotic S phase contribute to the mutant phenotypes.

Here, we examine the roles of budding yeast Top3 and Rmi1 in meiosis, using conditional expression alleles that allow a specific focus on Top3-Rmi1 function during meiotic recombination. We show that Top3 and Rmi1, like Sgs1, are required for early, Cdc5-independent NCO formation, and prevent flux through the ALT recombination pathway. These findings suggest that the entire STR complex promotes normal recombination pathway choice. However unlike wild type or sgs1 mutants, cells lacking Top3 or Rmi1 fail to completely resolve joint molecules, and as a consequence display defects in meiotic nuclear division. We propose that Top3-Rmi1 strand passage activity has two important function during meiosis: early, acting with Sgs1, to promote proper recombination pathway choice; and later, independent of Sgs1, to prevent the persistence of unresolvable strand entanglements in recombination intermediates.

Results

Top3 and Rmi1 are required for normal noncrossover formation

To examine the contribution of Top3 and Rmi1 to meiotic recombination intermediate metabolism, we used a recombination reporter system and synchronized meiotic cultures to score recombination intermediates and products in real time (Figures 1A, 2A, S1A, Jessop et al., 2005). To avoid the mitotic growth defects of null mutants, we used meiotic depletion alleles (top3-md and rmi1-md), where the CLB2 promoter, which is active during mitosis and inactive during meiosis, replaces the native promoters. In these mutants, Top3 and Rmi1 protein levels decreased by 69% and 92%, respectively, by three hours after induction of meiosis, the time when DSBs are first detected (Figure S1B). Depletion of Top3 or Rmi1 was not accompanied by Sgs1 depletion during the first 6 hours of meiosis, when most meiotic recombination occurs (Figure S1C).

Figure 1. Top3 and Rmi1 are required for normal NCO formation.

(A) Recombination reporter used, showing EcoRI-XhoI digests used to detect COs, NCOs and DSBs. (B) Southern blots of DNA from NDT80 strains. (C) CO and NCO frequencies normalized to those at 9 hr. Arrows indicate times of half maximum values. (D) Southern blots of samples from ndt80Δ strains. (E) NCO and (F) CO frequencies, plotted as percent of total lane signal. Data are represented as mean ± SEM from two independent experiments. sgs1-md ndt80Δ values are from De Muyt et al. (2012). See also Figure S1 and Table S1.

Figure 2. Top3 and Rmi1 prevent aberrant joint molecule accumulation.

(A) Recombination reporter, showing XmnI digests to detect inter-sister (IS) and interhomolog (IH) double Holliday junctions (dHJ), and multichromatid joint molecules containing 3 or 4 chromosomes (mcJM). (B) Southern blot showing joint molecule (JM) formation and resolution in NDT80 strains. (C) Southern blot showing joint molecule (JM) accumulation in ndt80Δ strains. (D) Fraction of JMs that are mcJMs in NDT80 strains. Data are represented as mean ± SEM of 5 and 6 hr samples from three independent experiments. (E) Total JM accumulation in ndt80Δ strains, plotted as percent of total lane signal. (F) fraction of JMs in ndt80Δ strains that are mcJMs. Data in (E) and (F) are represented as mean ± SEM from two independent experiments. Values for sgs1-md ndt80Δ are from De Muyt et al. (2012). See also Figures S1, S2 and Table S1.

In normal meiosis, NCOs form before COs, and are independent of stable JM formation and of Cdc5-triggered JM resolution (Allers and Lichten, 2001; Borner et al., 2004; Clyne et al., 2003; Sourirajan and Lichten, 2008). In contrast, in sgs1-md cells, NCOs and COs are formed at the same time by Cdc5- and SSN-dependent resolution of JM intermediates (De Muyt et al., 2012). top3-md and rmi1-md strains formed and repaired DSBs (Figure S1D, E), and final levels of NCOs and COs were similar in wild type, sgs1-md, top3-md and rmi1-md (Figures 1B, S1D). However, unlike in wild type, NCO formation was delayed in top3-md, rmi1-md and sgs1-md, and NCOs and COs formed at same time (Figure 1C). Furthermore, whereas NCO formation is Ndt80-independent in wild type (Allers and Lichten, 2001), NCOs were greatly reduced in top3-md ndt80Δ, rmi1-md ndt80Δ and sgs1-md ndt80Δ (Figure 1 D-F). Thus, all members of the STR complex are required for early, Ndt80-independent NCO formation.

Top3 and Rmi1 limit aberrant joint molecule accumulation

In wild type, most JMs detected are inter-homolog double Holliday junction, with a minor contribution from inter-sister JMs (Schwacha and Kleckner, 1994). In sgs1 mutants, levels of inter-sister and multichromatid JMs are elevated, indicating that Sgs1 promotes appropriate partner choice during JM formation (Jessop and Lichten, 2008; Oh et al., 2007). top3-md and rmi1-md strains displayed similarly elevated levels of multichromatid and inter-sister JMs, with the majority being resolved by the end of meiosis (Figures 2A, B, D, S1D). These JMs are products of meiotic recombination, rather than events occurring during premeiotic replication, as none were detected in top3-md spo11-Y135F and rmi1-md spo11-Y135F mutants (Figure S2A), which do not form meiotic DSBs (Bergerat et al., 1997).

Because transient JMs are difficult to quantify and characterize, we also examined the JMs that accumulate in resolution-defective ndt80Δ mutants (Allers and Lichten, 2001). top3-md ndt80Δ and rmi1-md ndt80Δ accumulated aberrant JMs similar to those seen in sgs1-md ndt80Δ (Figures 2C, S2B-D). Overall JM levels in all three STR-deficient strains were about twice those seen in otherwise wild type ndt80Δ strains (Figure 2E), and multichromatid JM levels in all three STR-deficient ndt80Δ strains were 5 to 10-fold greater than those in ndt80Δ (Figures 2F, S2C). These findings further support the conclusion that Sgs1, Top3 and Rmi1 act together to regulate JM formation.

Mus81-Mms4 and Yen1 resolve JMs formed in the absence of Top3 or Rmi1

Meiotic JMs formed in the absence of Sgs1 are resolved by the combined activities of the SSNs Mus81-Mms4, Yen1 and Slx1-Slx4, with Mus81-Mms4 and Yen1 playing the most important roles (De Muyt et al., 2012; Jessop and Lichten, 2008; Oh et al., 2008; Zakharyevich et al., 2012). In contrast, resolution of most JMs formed in wild type is SSN-independent (de los Santos et al., 2003; De Muyt et al., 2012; Oh et al., 2008; Zakharyevich et al., 2012). We asked if SSNs also resolve JMs that form in top3-md and rmi1-md strains (Figure 3). Unlike STR-competent mms4-md yen1Δ cells, but similar to sgs1-md mms4-md yen1Δ, top3-md mms4-md yen1Δ and rmi1-md mms4-md yen1Δ strains accumulated unresolved JMs and displayed substantial reductions in COs and NCOs. Thus, Mus81-Mms4 and Yen1 resolve the JMs that form in the absence of Top3 or Rmi1.

Figure 3. Mus81-Mms4 and Yen1 resolve JMs formed in the absence of Top3 or Rmi1.

JMs, COs and NCOs in mms4-md yen1Δ strains. (A) Southern blots to detect JMs (top, XmnI digest) and COs and NCOs (bottom, EcoRI/XhoI digest). Labels are as in Figure 2. (B) Frequencies of the total JMs (left), COs (middle) and NCOs (right) plotted as percent of the total lane signal. Values for sgs1-md mms4-md yen1Δ are from De Muyt et al. (2012). Data are represented as mean ± SEM from two independent experiments. See also Table S1.

Incomplete JM resolution in the absence of Top3 or Rmi1

In wild type meiosis, Cdc5-triggered JM resolution produces mostly COs (Clyne et al., 2003; Sourirajan and Lichten, 2008), while Cdc5-triggered resolution of JMs that form in sgs1-md produces both COs and NCOs (De Muyt et al., 2012). To determine if the JMs that form in the absence of Top3 or Rmi1 are similarly resolved, we used an estradiol-regulated version of CDC5 (Sourirajan and Lichten, 2008) to induce expression of Cdc5 in ndt80Δ mutants (Figure S3A).

NCO formation was Cdc5-independent when the STR complex was present, and JMs were resolved as COs upon Cdc5 induction (Figure 4A,E, Sourirajan and Lichten, 2008). In contrast, both NCOs and COs were Cdc5-dependent in strains lacking any STR complex member (Figure 4B-D, F-H). However, unlike wild type and sgs1-md, in which Cdc5 induction results in near-complete JM resolution within 4 h (Figure 3E, F, De Muyt et al., 2012; Sourirajan and Lichten, 2008), JM resolution in top3-md and rmi1-md was delayed by about 1 h and was incomplete, with 1/3 - 1/2 of JMs unresolved after 5 h of CDC5 expression (Figures 4G, H). This was accompanied by a corresponding delay and reduction in CO and NCO formation.

Figure 4. Top3 and Rmi1 are required for CO-specific and complete JM resolution.

JM formation and resolution, and CO and NCO formation in ndt80Δ CDC5-IN strains. At 7 hr, cells were divided into two portions: uninduced (no β-estradiol; –CDC5, black) and induced (β-estradiol added to 1μM; +CDC5, red). (A-D) Southern blots to detect JMs (top, XmnI digest) and recombination products (bottom, EcoRI/XhoI digest). A superfluous lane in the upper gel in panel A has been deleted. (E-H) frequencies of total JMs (top), COs (middle) and NCOs (lower) plotted as percent of total lane signal. Data for ndt80Δ CDC5-IN are from De Muyt et al. (2012). Data are represented as mean ± SEM from two independent experiments. See also Figure S3 and Table S1.

Loss of Sgs1 suppresses the mitotic slow-growth and extreme genome instability of top3 and rmi1 mutants (Chang et al., 2005; Gangloff et al., 1994; Mullen et al., 2005). However, sgs1-md did not similarly suppress the meiotic JM-resolution defect of top3-md and rmi1-md (Figure S3). In sgs1-md top3-md and sgs1-md rmi1-md double mutants, multichromatid JM levels and JM dynamics were similar to those in top3-md or rmi1-md single mutants (Figures S1D, S3B). In top3-md sgs1-md ndt80Δ CDC5-IN or rmi1-md sgs1-md ndt80Δ CDC5-IN strains, about 1/3 of JMs remained unresolved after induced Cdc5 expression (Figure S3C, D), similar to the levels of unresolved JMs seen in top3-md ndt80Δ CDC5-IN and rmi1-md ndt80Δ CDC5-IN strains (compare Figure 4G, H with Figure S3C). Thus, the JM resolution defects of top3-md and rmi1-md mutants are not due to STR-independent Sgs1 activity. These findings point to a unique function for Top3-Rmi1, to ensure JM resolution in a timely and complete manner.

The marked JM resolution defects seen in ndt80Δ CDC5-IN strains that were top3-md and rmi1-md (Figure 4) stands in contrast to the relatively low levels of unresolved JMs seen at the end of meiosis in corresponding NDT80 strains (Figures 2B and S1D). This may be due to ongoing strand exchange in pachytene-arrested ndt80Δ cells producing unresolvable JMs at higher levels than during normal meiotic progression. Because chromosome segregation can be blocked by unresolved JMs that are below the limits of molecular detection, it provides a more sensitive assay for JM resolution. We therefore analyzed nuclear division and chromosome segregation in otherwise wild type top3-md and rmi1-md cells (Figure 5).

Figure 5. Top3 and Rmi1 are required for normal meiotic chromosome segregation.

(A) Percent of cells having progressed through meiosis, detected as cells with visible spore walls. (B) Percent of cells having undergone at least one nuclear division, detected as cells with ≥2 nuclei. (C) Percent of cells having segregated chromosomes to all four spores at 9 h. Data are represented as mean ± SEM from two independent experiments. (D) Representative micrographs of the nuclear segregation and sporulation phenotypes reported in panels A-C, all from 9 h samples. Top— DNA detected by DAPI staining; lower--DIC. See also Figure S4 and Table S1.

Wild type and sgs1-md efficiently underwent meiotic nuclear division, with more than 90% of nuclei dividing at least once by 9 h (Figure 5B). In contrast, only 64% oftop3-md and 27% of rmi1-md cells contained 2 or more DNA masses by 9 h. top3-md and rmi1-md, but not wild type or sgs1-md, frequently displayed stretched nuclei diagnostic of homolog separation defects (Figure 5D and data not shown). Only a fraction of the top3-md or rmi1-md cells that formed spore walls (see below) contained DNA in all four spores (52% and 37% respectively), a further indication of a chromosome segregation defect (Figure 5C). In addition, spore viability was reduced in the limited number of mature ascii that did form (Figure S4A). Nuclear division defects were suppressed by spo11-Y135F (Figure S4B), consistent with them being caused by unresolved JMs.

Nuclear division failure also can be caused if the meiotic recombination checkpoint limits Ndt80 activity and thus prevents meiotic progression (Pak and Segall, 2002). To test this possibility, we monitored spore wall formation, which requires Ndt80-dependent gene expression (Chu and Herskowitz, 1998). Using this criterion, 83% of top3-md and 69% of rmi1-md strains cells progress through meiosis (Figure 5A). We therefore conclude that the nuclear division defect in these mutants is caused by a failure to resolve recombination intermediates.

As was seen in ndt80Δ CDC5-IN strains, sgs1 mutation did not suppress the defects of top3-md and rmi1-md (Figure 5). Only 11% of top3-md sgs1-md and 18% of rmi1-md sgs1-md strains divided nuclei by 9 hr, and only 7% of top3-md sgs1-md and 4% of rmi1-md sgs1-md cells that formed spore walls contained DNA in all four spores. However, the rare four-spore tetrads that did form were as viable as those from top3-md and rmi1-md single mutants (Figure S4A).

In summary, sgs1-md, top3-md and rmi1-md strains display similar accumulation of aberrant recombination intermediates, but the nuclear division and spore viability defects diagnostic of JM resolution failure are seen only in top3-md and rmi1-md. sgs1 mutation does not suppress, but instead enhances the sporulation and spore viability defects of top3 and rmi1 mutants, indicating that these defects are not the result of Sgs1 activity, and reinforcing the conclusion that Top3-Rmi1 activity promotes JM resolution.

Top3 catalytic activity is required for normal meiotic recombination

The above results suggest that Top3-Rmi1 strand passage activity is required for normal meiotic recombination, but could also be explained if Top3-Rmi1 acts as a non-catalytic STR component, either to maintain Sgs1 in a stable, active conformation, or to increase the affinity of Sgs1 for its substrates (Cejka et al., 2010). To test this, we expressed a catalytically inactive Top3-Y356F protein (Mankouri and Hickson, 2006) specifically during meiosis, using a top3-md ndt80Δ strains that contained a top3-Y356F mutant gene (or a control TOP3 gene) expressed from a meiosis-specific HOP1 promoter (Figure 6). In these strains, active Top3 was gradually replaced by inactive Top3-Y356F, so that by the time JMs appeared (5 hr), at least 95% of Top3 was the inactive form (Figures 6, S5). In controls expressing wild-type TOP3 throughout, interhomolog dHJs were the predominant JM, COs were reduced, and NCOs formed normally. In contrast, expressing top3-Y356F during meiosis resulted in phenotypes similar to top3-md: JMs accumulated to higher levels, with a high fraction of multichromatid JMs, and neither COs nor NCOs were formed (Figure 6C-H). Thus, catalytic activity is important for Top3 function during meiotic recombination.

Figure 6. Top3 catalytic activity is required for normal JM metabolism.

(A) Top: construct to express TOP3 in mitosis (CLB2 promoter) and during meiosis (HOP1 promoter). White boxes—selectable markers; grey boxes—TOP3 coding sequences; magenta boxes—epitope tags; arrows—promoters; black boxes—transcription terminators. Bottom: Western blots, probed with anti-Top3, to detect epitope-tagged Top3 (expressed in mitotic cells) and untagged Top3 (expressed in meiotic cells). Blots were stripped and reprobed for Arp7 as a loading control. (B) Top: construct to express TOP3 in mitosis and top3-Y356F in meiosis: red box—top3-Y356F coding sequences; all other features as in (A). Bottom: Western blots, as in (A), to detect expression of epitope-tagged Top3 and untagged Top3-Y356F. Quantification for both blots is in Figure S5. (C) Southern blots to detect JMs (digests and figure labels as in Figure 2). (D) Total JMs, plotted as percent of total lane signal. (E) Percent of JMs that are multichromatid JMs. (F) Southern blots of EcoR1-Xho1 digests to detect COs and NCOs. Frequencies of NCOs (G) and COs (H), plotted as a percentage of total lane signal. Data are represented as mean ± SEM from two independent experiments. See also Figure S5 and Table S1.

Discussion

Previous studies identified the Sgs1 helicase as integral to meiotic recombination intermediate metabolism (De Muyt et al., 2012; Jessop and Lichten, 2008; Jessop et al., 2006; Oh et al., 2007; Oh et al., 2008; Rockmill et al., 2003; Zakharyevich et al., 2012). Sgs1 promotes early NCO formation by SDSA and regulated CO formation by the meiosis-specific ZMM, MutLγ-Exo1-dependent pathway. It does so by preventing flux through the ALT-recombination pathway, which forms unregulated JMs and resolves them as both COs and NCOs, using SSNs that also function during the mitotic cell cycle. These findings can be understood if Sgs1 acts as a recombination intermediate chaperone, disassembling early intermediates that would otherwise populate the ALT-recombination pathway, but leaving intact unbranched SDSA intermediates that are NCO precursors, and ZMM protein-protected JMs that will be resolved as COs (Figure 7A, De Muyt et al., 2012).

Figure 7. Roles of Top3 and Rmi1 in meiotic recombination.

(A) Recombination intermediate chaperone model for Sgs1-Top3-Rmi1 activity during meiosis. Combined Sgs1 and Top3-Rmi1 activity disassembles branched recombination intermediates. Disassembly of early strand invasion intermediates facilitates NCO formation by SDSA or return of events to the original DSB state, facilitating capture and stabilization by ZMM proteins and dHJ formation by second-end capture. Cdc5 triggers MutLγ-Exo1-dependent resolution of these dHJs specifically as COs. A few events escape STR-disassembly and populate the ALT pathway, forming unregulated JMs. STR can dissolve these JMs to form NCOs, or they can undergo Cdc5-triggered resolution by SSNs to form both COs and NCOs. Top3-Rmi1 may also act during JM formation or resolution to prevent accumulation of links that these nucleases cannot resolve. (B) In top3-md and rmi1-md cells, D-loop intermediates are not disassembled, and form ZMM-independent JMs, populating the ALT pathway that forms both COs and NCOs by SSN-mediated resolution. Some JMs contain links that cannot be resolved, and these remain at the end of meiosis. (C) Top3-Rmi1 single-strand passage activity is required for dissolution. Convergent migration of two HJs results in hemicatenanes (circled) that require single-strand passage for resolution. (D) Top3-Rmi1 single-strand passage activity can facilitate D-loop disassembly. If both ends being unwound are topologically constrained, unwinding creates strand crossings (circled) that must be resolved before unwinding can proceed further. (E) An example of a JM intermediate that would require Top3-Rmi1 activity before it can be resolved. This structure could be formed if the dark blue strand invaded the red duplex, disengaged and re-engaged with the other blue strand, and then underwent a second round of strand invasion. The outer HJs will be substrates for SSN-mediated resolution, while resolving the central hemicatenated strands (circled) will require strand passage.

We show here that Top3 and Rmi1, the other two members of the STR complex, similarly contribute to recombination chaperone activity. In top3-md and rmi1-md mutants, or in the absence of Top3 catalytic activity, early NCOs are not formed (Figures 1, 6), JMs include a high fraction of inter-sister and multichromatid JMs (Figures 2, 6), and JM resolution requires mitotic resolvases, producing both COs and NCOs (Figure 3). Thus, both Sgs1 helicase and Top3-Rmi1 single-strand passage activity are required for regulated NCO and CO recombination during meiosis. However, Top3-Rmi1 has a second function that is independent of Sgs1, to prevent accumulation of JMs that cannot be resolved. Similar findings have been obtained in an independent study (Tang et al., 2014).

The primary activity of Top3-Rmi1 is DNA single strand-passage (Cejka et al., 2012). Our findings identify, for the first time, important roles for this activity during meiotic recombination. Below, we discuss the implications of these findings for the regulation of meiotic recombination and for the biochemical function of this conserved helicase-topoisomerase complex.

The STR complex promotes regulated meiotic recombination through early intermediate disassembly

During normal budding yeast meiosis, NCOs form at pachytene onset, before COs, and independent of the regulatory factors (Ndt80 and Cdc5) that drive exit from pachytene and JM resolution as COs (Allers and Lichten, 2001; Borner et al., 2004; Clyne et al., 2003; Hunter and Kleckner, 2001; Sourirajan and Lichten, 2008). The lack of early NCOs, increased levels of aberrant JMs, and unbiased, SSN-dependent JM resolution seen in sgs1-md, top3-md and rmi1-md can, in theory, be explained by either of two distinct STR-dependent mechanisms acting on recombination intermediates (Figure 7A-D). In one, JM dissolution (Figure 7C), combined Sgs1 helicase and Top3-Rmi1 decatenase activities dissolve mature JMs, producing NCOs and removing the substrate for SSN-mediated JM resolution. In other words, JM dissolution limits flux through the ALT pathway by terminating events that have already entered the pathway. In the second mechanism, D-loop disassembly, the STR complex can either produce NCOs by SDSA or return intermediates to the initial DSB state, where they can engage in another round of recombination. Thus, D-loop disassembly limits flux through the ALT pathway by preventing events from entering it.

While both mechanisms are formally possible, current data indicate that the STR promotes the ZMM pathway for CO formation, in that the ALT pathway quantitatively replaces the ZMM pathway in STR-deficient mutants (Figure 7, De Muyt et al., 2012; Tang et al., 2014; Zakharyevich et al., 2012). This finding is inconsistent with JM dissolution being the only activity of the STR, since dissolution can only produce NCOs (Figure 7A,C). We therefore favor a model for STR action in which D-loop disassembly, catalyzed by both the Sgs1 helicase and Top3-Rmi1 single strand passage activity, drives early intermediates towards two alternate fates: SDSA and early NCO formation; or capture by ZMM proteins leading to regulated JM formation and subsequent biased resolution as COs.

D-loop disassembly can, in theory, be driven by helicase activity alone, as has been observed in studies of BLM action on artificial D-loop substrates (Bachrati et al., 2006; van Brabant et al., 2000). However, these in vitro studies of D-loop disassembly used protein-free model substrates where the invading strand was relatively short (21-54 nt). D-loop intermediates that form in vivo are expected to contain substantially longer invading strands (Jinks-Robertson et al., 1993) and are in complex with recombination and other chromosomal proteins that could limit DNA rotation and create topological constraints not present in model substrates. We suggest that, in vivo, D-loop disassembly is likely to form intermediates in which the displaced strand blocks further unwinding of the invading strand (Figure 7D). Passage of one single strand through another by Top3-Rmi1 would be required to overcome this barrier. Consistent with this, Top3-Rmi1 recently has been shown to be required for the in vitro disassembly of D-loops that are in complex with Rad51 protein (Fasching et al., 2014).

Top3-Rmi1 is required in vivo for Sgs1-mediated end resection (Zhu et al., 2008), and in vitro, it increases Sgs1 affinity for branched substrates when RPA is present (Cejka et al., 2010). In theory, Top3-Rmi1 could play a non-catalytic role in STR-mediated D-loop disassembly, by increasing affinity of the complex for substrates. This is not likely, as expression of top3-Y365F during meiosis does not suppress the multichromatid JM-accumulation phenotype of top3-md mutants (Figure 6), and the cognate mutation does not alter the affinity of human Top3α for DNA (Goulaouic et al., 1999). Therefore, the requirement for Top3-Rmi1 in STR-mediated recombination chaperone activity most likely reflects a requirement for single-strand passage during D-loop disassembly.

Top3-Rmi1 has STR-independent function that promotes JM resolution

STR-defective mutants display elevated levels of aberrant JMs, and resolution of these JMs requires the SSNs that resolve only a minor fraction of JMs during normal meiosis. However, unlike mutants lacking Sgs1 activity, top3-md and rmi1-md mutants display marked JM resolution defects that are not suppressed by sgs1-md. These defects are most prominent upon Cdc5 expression in pachytene-arrested cells (Figure 4), but are also revealed by recombination-dependent defects in nuclear division and segregation seen in unarrested cells (Figure 5, Tang et al., 2014). We suggest that these unresolvable JMs are produced by strand switching during the unchaperoned strand invasion that occurs in the absence of the STR complex. Strand switches during invasion can form junctions that are poor substrates for Holliday junction-directed resolvases, and that require single strand passage for complete resolution (Figure 7E). The more prominent presence of resolution-refractory intermediates in ndt80Δ mutants (Figure 4) may be due to ongoing strand exchange that occurs in these strains during pachytene arrest; evidence for such ongoing strand exchange is provided by the continual increase, over time, in the fraction of total JMs that are multichromatid JMs (Figure 2F).

The resolution-refractory links present in JMs formed in the absence of Top3 or Rmi1 could reflect a requirement for Top3-Rmi1 activity at two different steps in recombination. Top3-Rmi1 could act during JM formation, to remove single-strand links that would later lead to JMs with hemicatenanes, or Top3-Rmi1 could act during JM resolution, unlinking hemicatenanes concurrent with HJ cleavage.

STR complex targets may differ in meiosis and in the mitotic cell cycle

The mitotic growth defects and DNA damage sensitivity of top3 and rmi1 mutants are largely suppressed by sgs1 mutation (Gangloff et al., 1994; Mullen et al., 2005). In contrast, we find that sgs1-md does not suppress the meiotic defects of top3-md and rmi1-md mutants, and even can exacerbate them (Figures 5, S3). To account for this, we suggest that some types of substrates for Top3-Rmi1 are formed during the mitotic cell cycle, but not during meiotic recombination. The most likely candidates for such substrates are branched structures associated with stalled or broken replication forks. In the absence of Top3-Rmi1 strand passage, Sgs1 helicase could act with strand invasion activities, to convert these substrates into unresolvable intertwined structures that would interfere with nuclear division. The presence of these substrates during premeiotic replication, but not during meiotic DSB repair, also can account for differences between the mutant phenotypes seen in our studies with the more severe phenotypes seen in a previous study of top3Δ mutants (Gangloff et al., 1999). In top3Δ mutants, Top3 is absent throughout meiosis, while in our study, Top3 and Rmi1 are present at substantial levels (at least 45% and 20% of initial levels, respectively, Figure S1B) during the first 2 h of meiosis, when most replication occurs (Murakami and Keeney, 2014).

Concluding remarks

While function of RecQ-related helicases during homologous recombination has been the subject of considerable attention (Bernstein et al., 2010), somewhat less is known about their conserved partner, topoisomerase III. We have shown here that Top3-Rmi1 plays a role equal to Sgs1 in recombination intermediate chaperon activity during meiosis, and that Top3-Rmi1 has an Sgs1-independent role that ensures full resolution of JMs formed in the meiotic recombination process that most closely corresponds to events that occur in mitotic cells. Our findings thus indicate that single-strand decatenation is as important as DNA unwinding in meiotic recombination. Limited data from studies of Top3-Rmi1 function during meiosis in multicellular organisms suggest that this conclusion is broadly applicable (Chelysheva et al., 2008; Hartung et al., 2008; Kim et al., 2000; Wicky et al., 2004).

In closing, we note that the Holliday junction and related branched structures have long been a primary focus in the study of homologous recombination mechanisms (Holliday, 1964). The findings reported here indicate that other modes of DNA-DNA interaction in recombination intermediates are likely to be of similar importance, and call for future exploration of the nature of these links and the mechanisms for their formation and resolution.

Experimental Procedures

Yeast strains

All strains are of the SK1 background. Details are in Table S1.

Sporulation and sample preparation

Liquid culture sporulation and sample processing was as described (Allers and Lichten, 2001; Goyon and Lichten, 1993; Jessop et al., 2005; Jessop et al., 2006). For CDC5 induction (De Muyt et al., 2012; Sourirajan and Lichten, 2008), cells were aerated in sporulation medium for 7h, cultures were split, and either β-estradiol or carrier was added. DNA preps used CTAB extraction to stabilize JMs (Allers and Lichten, 2000). Protein extracts were made by TCA precipitation (Foiani et al., 1994). Nuclear division was monitored by DAPI staining (Goyon and Lichten, 1993). Cells with stretched nuclei or two or more nuclei were scored as having initiated meiosis I.

DNA analysis

Southern blots to detect recombination intermediates and products were as described (De Muyt et al., 2012; Jessop et al., 2006). JMs were detected using XmnI digests probed with +156 to +1413 of ARG4 coding sequences. DSBs, NCOs and COs were detected using EcoRI-XhoI digests probed with +539 to +719 of HIS4 coding sequences.

Protein analysis

Western blots were as described (Jessop and Lichten, 2008) with the following changes: 6%, 10% or 8-16% polyacrylamide gels (Invitrogen) were used; incubation with primary antibody was either 3 h at room temperature or 16 h at 4°C; washes were 6 times for 5 min each; blots were stripped for reprobing for Arp7 using OneMinute stripping buffer (GM Biosciences). Primary antisera were: anti-HA (clone 12CA5, Roche), 1:10,000; anti-Top3 (Santa Cruz Biotechnology, sc-98274), 1:2000; anti-Cdc5 (Santa Cruz Biotechnology, sc-6733), 1:2000; anti-ARP7 (Santa Cruz Biotechnology, sc-8961), 1:1000. Secondary antibodies were alkaline phosphatase conjugates of goat anti-mouse (Sigma, A3562); rabbit anti-goat (Sigma, A4187) and goat anti-rabbit (Sigma, A3687), all 1:5,000.