Dbl2 Regulates Rad51 and DNA Joint Molecule Metabolism to Ensure Proper Meiotic Chromosome Segregation

To identify new proteins required for faithful meiotic chromosome segregation, we screened a Schizosaccharomyces pombe deletion mutant library and found that deletion of the dbl2 gene led to missegregation of chromosomes during meiosis. Analyses of both live and fixed cells showed that dbl2Δ mutant cells frequently failed to segregate homologous chromosomes to opposite poles during meiosis I. Removing Rec12 (Spo11 homolog) to eliminate meiotic DNA double-strand breaks (DSBs) suppressed the segregation defect in dbl2Δ cells, indicating that Dbl2 acts after the initiation of meiotic recombination. Analyses of DSBs and Holliday junctions revealed no significant defect in their formation or processing in dbl2Δ mutant cells, although some Rec12-dependent DNA joint molecules persisted late in meiosis. Failure to segregate chromosomes in the absence of Dbl2 correlated with persistent Rad51 foci, and deletion of rad51 or genes encoding Rad51 mediators also suppressed the segregation defect of dbl2Δ. Formation of foci of Fbh1, an F-box helicase that efficiently dismantles Rad51-DNA filaments, was impaired in dbl2Δ cells. Our results suggest that Dbl2 is a novel regulator of Fbh1 and thereby Rad51-dependent DSB repair required for proper meiotic chromosome segregation and viable sex cell formation. The wide conservation of these proteins suggests that our results apply to many species.


Introduction
During meiosis, haploid gametes are produced from diploid precursor cells. The reduction of chromosome number is achieved by a single round of DNA replication followed by two rounds of chromosome segregation, termed meiosis I and meiosis II. While meiosis II is similar to mitosis in that sister centromeres segregate from each other, centromeres of homologous chromosomes (homologs) segregate to opposite poles in meiosis I [1,2].
Three meiosis-specific features are essential for proper segregation of chromosomes during meiosis I-formation of crossovers that connect homologs, mono-orientation of sister kinetochores, and a stepwise loss of sister chromatid cohesion. The formation of crossovers, as a result of meiotic recombination, and the attachment of sister kinetochores to microtubules emanating from the same spindle pole (mono-orientation) ensure that homologous centromeres are pulled in opposite directions on meiosis I spindles [2,3].
Crossovers and cohesion between sister chromatids distal to crossovers are responsible for holding homologs together until the onset of anaphase I, when a protease called separase cleaves cohesin along chromosome arms [4][5][6]. This allows segregation of recombined homologs to opposite poles of the meiosis I spindle. During meiosis I, cleavage of centromeric cohesin is blocked by Sgo1 (called Mei-S332 in Drosophila) complexed with the protein phosphatase 2A (PP2A) [7][8][9]. Deprotection of centromeric cohesin and a second round of separase activation allow cleavage of the centromeric cohesin at the onset of anaphase II, which is followed by segregation of sister centromeres in meiosis II [10].
Homologous recombination involves programmed formation of DNA double-strand breaks (DSBs) and an evolutionarily conserved pathway for DSB repair, which operates in both mitotic and meiotic cells. To promote high-level recombination (including crossovers) in meiosis, programmed DSBs are made by the highly conserved topoisomerase-like protein Spo11 (called Rec12 in the fission yeast Schizosaccharomyces pombe) and several essential accessory factors [11][12][13]. These meiotic DSBs are repaired using the intact sister chromatid or the homolog as a template [14]. Only recombination between homologs can lead to formation of the crossovers required for proper segregation of chromosomes during meiosis I. During DSB formation, Rec12 becomes covalently attached to DNA 5' ends and is subsequently removed by an endonuclease, Mre11-Rad50-Nbs1 (MRN) complexed with Ctp1 [15][16][17]. The DNA 5 0 ends are further resected to generate long 3 0 single-stranded DNA (ssDNA) overhangs [18]. These ssDNA ends are then coated by Rad51 and (in some species) meiosis-specific Dmc1, both of which are homologs of the bacterial DNA strand-exchange protein RecA [19]. Rad51 promotes the formation of DNA joint molecules (JMs) between Rad51-ssDNA filaments and homologous double-stranded (ds) DNA [20]. Auxiliary proteins, called "recombination mediators", such as Rad52, Rad54, Rad55, Rad57, Sfr1, Swi5, and Rdh54, promote the formation and/or stabilization of Rad51-ssDNA filaments [20,21]. Other proteins, such as the F-box DNA helicase Fbh1 and (in the budding yeast Saccharomyces cerevisiae) the Srs2 helicase, are negative regulators of Rad51. Members of the Swi2/Snf2 family of DNA motor proteins, Rad54 and Rdh54, enhance Rad51-mediated formation of JMs but are also involved in the removal of Rad51 from DNA, suggesting that JM metabolism needs to be carefully regulated [22][23][24][25][26]. Rad51-ssDNA filaments invade homologous dsDNA to form a displacement loop (D-loop). Subsequent DNA synthesis primed by the invading 3' DNA end extends the invading strand [20]. The recombination reaction can then take one of two different paths. If the extended invading strand is displaced and anneals with the other DSB end, a non-crossover is produced in a hypothetical process called synthesis-dependent strand-annealing (SDSA). Alternatively, the strand invasion intermediate is stabilized, and capture of the second DSB end leads to formation of a Holliday junction (HJ) [14]. HJs can be resolved by endonucleolytic activities such as Mus81-Eme1, which is critical in S. pombe, where it is the only known complex involved in meiotic HJ resolution [27][28][29].
Regulation of the formation and processing of meiotic JMs, the subject of this report, is complex and incompletely understood. We report here a new role for Dbl2, which was first identified in a screen for S. pombe mutants defective in chromosome segregation during meiosis [30]. It was later identified in a screen for proteins forming microscopic foci at HO endonuclease-induced DSBs [31]. Dbl2 is required for normal DSB targeting of the DNA-repair helicase Fml1 [31]. Here, we show that Dbl2 is required for proper segregation of chromosomes during meiosis by regulating Rad51 function and JM metabolism, apparently by promoting formation of the helicase Fbh1 foci at the sites of DSB repair to dissociate a minor class of JMs. We discuss molecular mechanisms by which JM formation and processing are properly regulated for successful meiosis and the conservation of these proteins among species.

Dbl2 is required for proper segregation of chromosomes during meiosis and mitosis
To identify novel proteins required for faithful meiotic chromosome segregation, we screened a library of about 3200 S. pombe deletion mutants purchased from Bioneer. We found that deletion of the dbl2 gene frequently led to missegregation of chromosomes during meiosis. The dbl2 gene was also identified in our previous screening in which we deleted 180 functionally uncharacterized genes whose expression is upregulated during meiosis and screened for mutants defective in meiotic chromosome segregation [30]. To confirm that this phenotype is due to deletion of dbl2, we deleted the dbl2 gene in a haploid homothallic (h 90 ) strain in which the centromere of chromosome 2 was marked with GFP (cen2-GFP [32]). We sporulated mutant cells, stained nuclei with DAPI and scored segregation of GFP dots in asci with four nuclei. Indeed, we found that dbl2Δ mutant cells frequently missegregated chromosomes during meiosis (Fig 1A).
To investigate chromosome segregation directly in anaphase cells, we fixed cells and stained with antibodies against tubulin and GFP. In wild-type cells, homologous centromeres segregate to opposite poles during anaphase I. However, we frequently observed non-disjunction of homologous centromeres and lagging chromosomes in dbl2Δ anaphase I cells (Fig 1B and 1C). In the majority of anaphase I cells with lagging chromosomes, telomeres of chromosome 1 (sod2-GFP signals) lagged, while centromeres of chromosome 2 (cen2-GFP signals) segregated to the poles (S1 Fig). This indicates that in dbl2Δ mutant cells microtubules are frequently able to attach to kinetochores and pull the sister kinetochores to opposite poles, while the chromosomal arms and telomeres lag behind. Although we have analyzed only telomeres of chromosome 1 and centromeres of chromosome 2, we suppose that this is representative also for other chromosomes and thus it is unlikely that chromosomes in dbl2Δ mutant cells lag because they do not attach to microtubules.
In addition, in dbl2Δ mutant cells we frequently observed mononucleate cells containing one spindle and stretched but undivided chromatin ( Fig 1C) as well as mononucleate cells containing more than one spindle or four spindle pole bodies (SPBs) (Fig 1D and 1E). This result suggests that despite the failure to segregate chromosomes at meiosis I, dbl2Δ cells proceeded to form metaphase II spindles within a single nucleus. Live-cell imaging of dbl2Δ cells confirmed chromosome missegregation and failure to segregate chromosomes to opposite poles during meiosis I (Fig 2A). Remarkably, deletion of rec12 (spo11 homolog) suppressed the meiosis I chromosome segregation failure in dbl2Δ cells (Fig 2). In 13 out of 26 dbl2Δ zygotes observed, chromosomes failed to segregate to opposite poles during meiosis I, whereas in all 17 dbl2Δ rec12Δ zygotes observed, chromosomes segregated during meiosis I (Fig 2A and 2B). The suppression of the failure of meiosis I chromosome segregation in dbl2Δ rec12Δ zygotes was observed in both live and fixed cells (Fig 2B). Because Rec12 is required for meiotic DSB The images show spindle and chromatin morphology at metaphase II or anaphase II in wild type (JG15456 x JG11318) and dbl2Δ mutant (JG17208 x JG17207) meiosis. Cells were fixed and immunostained for tubulin; DNA was visualized by Hoechst staining. Mononucleate dbl2Δ zygotes with two or three spindles are shown. (E) Spindle pole bodies (SPBs) and chromosomes were visualized by endogenously tagged Pcp1-GFP and Hht1-mRFP, respectively. In a wild-type strain (JG16917), four SPBs are present at the completion of meiosis II, and each is associated with one of the four nuclei. In the dbl2Δ mutant (JG17116), mononucleate zygotes containing up to four SPBs are observed.
formation and recombination [11,33,34], these results imply that Dbl2 acts after the initiation of meiotic recombination to allow faithful chromosome disjunction.
Failure to segregate chromosomes in meiosis I could also be caused by an inability to remove cohesin from chromosome arms, which physically links two homologs that have recombined until the onset of anaphase I [6,35,36]. However, we found no evidence for defective removal of cohesin in dbl2Δ mutant cells when we analyzed the Rec8-GFP cohesin subunit, which is cleaved to allow cohesin removal from chromosomes [35] (S2 Fig).
Failure of chromosome segregation in meiosis I has also been observed in mutants defective in mono-orientation of sister kinetochores. In this case, centromeric sister chromatid cohesion,  Fig 1D). In live-cell imaging, 20 wild-type, 26 dbl2Δ, 10 rec12Δ and 17 dbl2Δ rec12Δ zygotes were scored. For the analysis of fixed cells, the strains (JG12618, JG17130, JG17351, JG17353) were immunostained for tubulin, DNA was visualized by Hoechst staining, and 100 zygotes from each strain were scored. which persists throughout the first meiotic division, prevents segregation of bi-oriented sister chromatids to opposite poles [37][38][39][40][41]. Mutants defective in mono-orientation attempt but fail to divide nuclei during the first meiotic division; elimination of centromeric cohesion allows them to undergo an equational meiosis I division [37][38][39][40][41]. If meiosis I nuclear division failure in dbl2Δ mutant cells were due to a defect in mono-orientation of sister kinetochores, then elimination of centromeric cohesion at the onset of anaphase I should allow dbl2Δ mutant cells to undergo an equational meiosis I division. Elimination of the centromeric cohesin protector Sgo1 in a dbl2Δ strain with only one copy of chromosome 2 marked with GFP (cen2-GFP) did not increase equational segregation of sister centromeres during meiosis I ( Fig 3A). Moreover, we frequently observed mononucleate cells containing more than one spindle in dbl2Δ sgo1Δ mutant cells, indicating that the meiosis I nuclear division failure was not suppressed by elimination of Sgo1. These results are consistent with the idea that in dbl2Δ mutant cells the failure to segregate chromosomes during meiosis I is not caused by a defect in mono-orientation of sister kinetochores. In addition to defects in meiosis I, we observed a high frequency of lagging chromosomes and missegregation of sister centromeres in dbl2Δ cells during anaphase II and mitosis ( Fig  3B-3D). Thus, in addition to segregation of homologs during meiosis I, Dbl2 is required for proper segregation of sister chromatids during both meiosis II and mitosis.
Meiotic DSBs and Holliday junctions are formed and processed normally in dbl2Δ mutant cells Our observation that the meiosis I chromosome segregation failure in dbl2Δ is suppressed by rec12Δ as well as a previous report that Dbl2 binds to DSBs to facilitate targeting of DNA repair helicase Fml1 [31] prompted us to analyze the role of Dbl2 in the formation of asci and viable spores and in recombination.
The dbl2Δ mutant produced abnormal asci, which often contained fewer than four spores or unequal-size spores (S3 Fig), in accord with the frequent chromosome missegregation noted above. The overall yield of viable spores per cell in the mating mixture was about 13 times less in the dbl2Δ mutant than in wild type (Table 1), indicating a severe defect in meiosis. Viability of the few spores produced was about 57% of that of wild type (Table 2). However, the frequency of both intergenic (ade6 -arg1) and intragenic (ade6) recombination in viable spores was similar to that in wild-type ( Table 1). As expected from the nearly wild-type levels of recombination, Southern blot hybridizations of DNA extracted from dbl2Δ mutant cells induced for meiosis revealed no defect in the formation and disappearance of DSBs at six DSB hotspots spanning a 0.50 Mb interval, mbs1 and mbs2 being the most prominent (Figs 4A and S4). In rad50 + strains the timing of appearance and disappearance of DSBs was similar to that in wild type; in rad50S strains, in which DSB repair is blocked, DSBs accumulated to similar levels in wild type and the dbl2 mutant [27,42]. Similarly, Holliday junction intermediates of meiotic recombination were formed and resolved with nearly wild-type kinetics (Figs 4B, 4C and S4). These results are consistent with the nearly wild-type recombination levels in the dbl2 mutant.
We noted that a low level of joint molecules (JMs) persisted in the dbl2 mutant longer than in wild type (Fig 4B and 4C). JMs persisted for at least 8 hr after meiotic induction in dbl2Δ but were scarcely detectable at or after 6 hr in dbl2 + . Accumulation of these species was Rec12-dependent, indicating that these JMs arise from meiotic DSBs and might be related to recombination intermediates. Since many of these JMs migrated on the arc that includes Y-shaped molecules, such as replication forks, we suspect they are D-loops, or closely related structures, which are postulated to be precursors to Holliday junctions (see Discussion) [27]. These persistent JMs could account for the failure of chromosome segregation and lagging chromosomes noted above (Figs 1-3). These persistent JMs were seen at two different DSB hotspots: mbs1 ( Fig 4B) and ade6-3049 (S4 Fig).
Although defects in processing Holliday junctions can lead to chromosome segregation defects [27,28,43,44], we think it is unlikely that this is the case in dbl2Δ mutant cells. Consistent with the idea that Dbl2 has a function independent of the Mus81-Eme1 Holliday junction resolvase [27,28,43,44] is the synthetic mitotic growth defect of a dbl2Δ eme1Δ double mutant strain: the double mutant grew much more poorly than either single mutant (S5 Fig). Moreover, expression of a wild-type E. coli Holliday junction resolvase RusA but not the nucleasedead RusA-D70N mutant suppressed camptothecin-sensitivity of eme1Δ mutant cells, as Table 2. Genetic interaction between dbl2Δ and mutations in homologous recombination-related genes.

Strain
Zygotes with Rad51 foci Mononucleate zygotes with more than one spindle (%) Spore viability (%) Data are means ± SEM of three independent experiments. Cells were fixed at 10-17 hr after meiotic induction and immunostained for tubulin and Rad51; DNA was visualized by Hoechst staining. At least 100 cells and 500 spores were scored. We cannot exclude the possibility that spore viability in fbh1Δ strains was affected by suppressor mutations which may occur in fbh1Δ mutant cells [23].  Strains GP6656 (dbl2 + ), GP8664 (dbl2Δ) and GP8450 (dbl2Δ rad50S) were induced for meiosis; DNA was prepared at the indicated times, digested with NotI, and analyzed by pulsed-field gel electrophoresis and Southern blot hybridization using a probe at the left end of the 501 kb NotI fragment J (band at the top of the gel) [42]. GP3718 (dbl2 + rad50S) is from [94]. The fraction of total DNA broken at mbs1 and at mbs2 (indicated by arrows on the right) at the indicated time for each strain is shown below each blot. The timing of appearance and disappearance of DSBs and their frequencies in both rad50 + and rad50S strains are similar to those previously reported in wild-type cells [42,95]. (B) Formation and resolution of DNA joint molecules at the mbs1 hotspot. Strains GP6656 (dbl2 + ), previously reported, but no suppression was seen in fbh1Δ, rqh1Δ (lacking the Rqh1 DNA repair helicase [45]) or dbl2Δ mutant cells (Fig 5A and 5B) [28,43]. RusA expression failed to suppress the meiosis I chromosome segregation defect in dbl2Δ and fbh1Δ mutant zygotes but partially suppressed the eme1Δ mutant phenotype as assessed by scoring mononucleate zygotes containing more than one spindle ( Fig 5C). These data indicate that Dbl2 does not have a major role in DSB formation or disappearance, Holliday junction formation or resolution, or GP8664 (dbl2Δ), and GP8836 (dbl2Δ rec12Δ) were induced for meiosis; DNA was extracted at the indicated times, digested with BsrGI, and analyzed by two-dimensional gel electrophoresis and Southern blot hybridization using a probe (~1 kb long) near the mbs1 DSB hotspot on the 10.5 kb BsrGI fragment (major spot at the bottom right) [96]. Black and white arrowheads, respectively, indicate Yarc and X-spike species (replication and recombination intermediates) found transiently in wild-type cells. Black and white arrows indicate Y-arc and X-spike species, respectively, that persist only in dbl2Δ rec12 + cells (at 6, 7, and 8 hr). (C) Quantification of branched DNA (JMs) in panel B. ImageQuant analysis was used to quantify the amount of branched DNA (Y-arc plus X-spike) relative to the total DNA. Data are the mean ± SEM from n assays (dbl2 + and dbl2Δ) or individual data from duplicate assays (rec12Δ dbl2Δ). See S2 Table for   . Cells were grown on EMM2 liquid medium without leucine for one day, diluted in 5-fold steps, and spotted onto EMM2 plates lacking thiamine and containing the indicated amounts of camptothecin (CPT). RusA and RusA-D70N were expressed under the control of a thiaminerepressible nmt1-promotor using a pREP41 vector [97]. pREP41-RusA-D70N, expressing an inactive version (D70N) of RusA [98], and an empty vector were used as negative controls. Plates were incubated at 32°C for 4 days and photographed. (C) Strains described in panels A and B and strains JG11355, JG17146, JG17544, JG17827, JG17949 and JG17951 were grown with or without thiamine (15 μM), fixed and immunostained for tubulin and GFP. DNA was visualized by Hoechst staining. 100 zygotes with more than one spindle on a single DNA mass were scored in three independent experiments. Data are means ± SEM of three independent experiments.

Dbl2 is required for timely removal of Rad51 nucleofilaments
Homologous recombination is promoted by Rad51, which forms filaments with ssDNA that undergo strand exchange with homologous dsDNA molecules [19,20]. The formation and timely disassembly of Rad51-ssDNA filaments is essential for successful DNA repair by homologous recombination. In wild-type S. pombe cells, Rad51 foci are present during meiotic prophase, when DSBs are formed and repaired, but are no longer present in metaphase I cells [46]. In the absence of proteins required for timely removal of Rad51-ssDNA filaments, such as the F-box DNA helicase Fbh1, Rad51 foci persist throughout both meiotic divisions [23]. We used an antibody raised against fission yeast Rad51 to visualize formation of Rad51 foci in meiotic cells. Interestingly, in dbl2Δ mutant cells Rad51 foci formed during meiotic prophase and, as in fbh1Δ mutant cells, remained visible during meiosis I and meiosis II and in mononucleate cells containing more than one spindle (Fig 6, Table 2). Thus, the failure to segregate chromosomes in the absence of Dbl2 correlates with persistent Rad51 foci.
Deletion of rad51 suppresses the chromosome segregation defect in dbl2Δ mutant cells If the failure to segregate chromosomes in the dbl2Δ mutant were due to a defect in removal of Rad51-ssDNA filaments or Rad51-dependent JMs, then elimination of Rad51 should allow dbl2Δ mutant cells to segregate chromosomes during meiosis I, though perhaps improperly due to unrepaired DSBs or reduced crossover numbers. Indeed, no mononucleate zygotes containing more than one spindle were observed in a dbl2Δ rad51Δ double mutant (Table 2). A similar suppression of the dbl2Δ mutation was observed upon deletion of any one of the genes encoding mediator proteins Rad52, Rad55, Rad57, Sfr1, and Rad54 that promote Rad51-ssDNA filament formation [47,48]; such suppression was not observed upon deletion of negative regulators of Rad51 function (Fml1 or Fml2), the meiosis-specific Rad51 paralog Dmc1, or Rdh54 (Table 2). Consistent with the notion that the chromosome segregation defect in dbl2Δ mutant cells is Rad51-dependent, the synthetic lethality of the rqh1Δ dbl2Δ double mutant was suppressed by deletion of rad51 in mitotic cells. These data suggest that Rad51 prevents proper segregation of chromosomes in the absence of Dbl2.

Dbl2 is required for efficient formation of Fbh1 foci in CPT-induced DNA lesions
Our observation that Rad51 foci persist in dbl2Δ mutant cells beyond meiotic prophase raises the possibility that Dbl2 regulates proteins responsible for disassembly of Rad51-ssDNA filaments. Fbh1, an F-box helicase related to bacterial UvrD, negatively regulates Rad51-mediated homologous recombination by disrupting Rad51 nucleoprotein filaments [22]. Interestingly, the fbh1Δ mutant phenotype resembles that of dbl2Δ-meiotic DSB formation and repair as well as recombination are close to wild-type levels, chromosomes frequently fail to segregate during meiosis I, Rad51 foci persist beyond meiotic prophase and the fbh1Δ mutant phenotype is suppressed by rad51Δ, rad52Δ or rad57Δ [23,[49][50][51].
We therefore tested whether Fbh1 foci, which co-localize with Rad51 foci after meiotic DSB formation [23], are affected in dbl2Δ mutant cells. Because Fbh1 is not visible when tagged at the endogenous locus, we used a strain in which the endogenous fbh1 gene was deleted and the yellow fluorescent protein (YFP)-tagged Fbh1 was expressed from a strong nmt promoter [23].
Fbh1-YFP forms foci in response to Rec12-dependent DSBs that co-localize with Rad51 foci, suggesting that this construct is functional [23]. We used camptothecin (CPT), a topoisomerase I inhibitor, to induce DNA lesions in vegetative cells [52]. Fbh1-YFP foci were visible in dbl2 + cells but were strongly reduced in dbl2Δ mutant cells (Fig 7A and 7B). This reduction was not due to a higher frequency of plasmid loss in dbl2Δ mutant cells because the stability of the Fbh1-YFP plasmid was similar in both dbl2 + cells (5.8% plasmid loss per generation) and dbl2Δ cells (6.7% plasmid loss per generation). We obtained similar results when we induced DNA lesions in vegetative cells using methyl methanesulfonate (MMS) [53,54] (S6 Fig). Fbh1-YFP foci were also visible in rad51Δ, rad52Δ, rad55Δ, rad57Δ, sfr1Δ and rad54Δ mutant strains, but the frequency of Fbh1-YFP foci was decreased by the dbl2Δ mutation in each of Fig 6. Failure to segregate chromosomes in the absence of Dbl2 correlates with persistent Rad51 foci. Cells were mated on SPA sporulation agar and at 10-17 hr fixed and immunostained for tubulin and Rad51; DNA was visualized by Hoechst staining. Representative images show that Rad51 foci persisted until anaphase II in fbh1Δ (JG17544) and dbl2Δ (JG17146) mutant cells but not in wild type (JG11355). Deletion of rad52, rad57, sfr1 or rad54 restored nearly wild-type frequencies of nuclei with levels of Rad51 foci in the dbl2Δ mutant cells. The strains were rad57Δ dbl2Δ (JG17751 x JG17752), rad52Δ dbl2Δ (JG17747 x JG17748), sfr1Δ dbl2Δ (JG17811), rad54Δ dbl2Δ (JG17821 x JG17819). No Rad51-GFP foci were detected in anaphase I or anaphase II zygotes of single mutants rad57Δ (JG17749 x JG17750), rad52Δ (JG17823 x JG17824), sfr1Δ (JG17746), rad54Δ (JG17815 x JG17817) and wild-type (JG11355) (see Table 2).   Dbl2 and Its Role in Recombination these strains, both in the absence and presence of CPT (Fig 7C). The reduction of Fbh1-YFP foci in dbl2Δ mutant cells was not due to reduced levels of DNA lesions because Rad52-m-Cherry foci, which represent sites of active DNA repair [55], were not reduced in dbl2Δ mutant cells (S7 Fig). To investigate whether localization of Dbl2 and Fbh1 to DSBs are interdependent, we analyzed Dbl2-YFP expressed from a strong nmt promoter [56]. A previous report showed that Dbl2-YFP formed a focus at an HO endonuclease-induced DSB [31]. We observed that Dbl2-YFP formed nuclear foci in wild-type cells when DNA lesions were induced by MMS or CPT (Fig 7D, S6 Table). Deletion of fbh1 had no effect on the formation of Dbl2 foci (Fig 7D,  S6 Table). This suggests that Dbl2-YFP expressed from a strong nmt promoter is able to form foci in the absence of Fbh1. However, we cannot exclude the possibility that this is due to overexpression of Dbl2-YFP or the presence of spontaneous suppressor mutations which may occur in fbh1Δ mutant cells [23].
If the role of Dbl2 were to promote accumulation of Fbh1 at DNA lesions, we would expect overexpression of Fbh1 to suppress the dbl2Δ mutant phenotype. Indeed, both CPT-sensitivity and the meiotic chromosome segregation defect in dbl2Δ mutant cells were nearly fully suppressed by expression of Fbh1-YFP from a strong nmt promoter (Fig 8A and 8B).

A critical role for Dbl2 in chromosome segregation
Dbl2 was identified in screenings for mutants defective in chromosome segregation during meiosis [30] and for proteins that localize to DNA double-strand breaks (DSBs) [31]. Here, we report that in the absence of Dbl2, Rad51 foci and Rad51-dependent DSB-repair intermediates (DNA joint molecules, or JMs) persist and frequently prevent proper segregation of chromosomes during meiosis I. Also in the absence of Dbl2, foci of the F-box helicase Fbh1 are less abundant than in the presence of Dbl2. We propose that a subset of JMs requires Fbh1 for their reversal or processing into Holliday junctions (HJs) resolvable by Mus81-Eme1 and that Dbl2 is required to promote formation of Fbh1 foci at these JMs. As predicted, the phenotypes of dbl2Δ and fbh1Δ are similar, although not identical, as discussed below. These proteins, like the formation of JMs during meiosis, are widely conserved, suggesting that diverse species require Dbl2 for successful meiosis and reproduction. Dbl2 promotes formation of Fbh1 foci to process a rare class of DNA joint molecules The data reported here can be accounted for by the following proposal. During DSB repair, a subset of JMs is not converted into HJs. Some of these JMs might be D-loops, since they migrate on the "Y-arc" characteristic of such JMs (Figs 4 and S4). Others might be related to HJs, such as hemicatenanes, since they migrate on the "X-arc" characteristic of such JMs. (Figs  4 and S4). The helicase Fbh1, recruited by Dbl2, either reverses these JMs to allow synthesisdependent strand annealing (SDSA) or enables their extension and conversion into HJs, which can be resolved by the Mus81-Eme1 HJ resolvase. This proposal accounts for our observations as follows.
We observed a rare class of JMs that persisted late in meiosis in dbl2Δ cells (Figs 4B, 4C and  S4). These JMs account for about 0.3% of the total DNA in a 12 kb interval containing the mbs1 or the ade6-3049 DSB hotspot (Figs 4C and S4), which corresponds to about 3 JMs per cell. This estimate is uncertain, because the low level of these JMs is not far above the background level and because the density (number per kb) of these JMs might be less in DSB-cold regions. Furthermore, some of the persistent JMs detected might not block segregation. Nevertheless, this frequency of persistent JMs is compatible with the frequency of missegregating or lagging chromosomes or cells with one nucleus but two spindles, which are seen in 20-50% of cells (Figs 1, 2 and 3; Table 2); if Poisson distributed, an average of one persistent JM per cell would result in 63% of cells with failed chromosome segregation. Thus, these persistent JMs are likely the cause of the failure of chromosomes to segregate properly during both meiosis I and II, consistent with rec12Δ eliminating both the persistent JMs (Figs 4 and S4) and chromosome segregation failure (Fig 2). These JMs might be between sisters or between homologs, but for the reasons below, we suspect they are primarily intersister JMs. During meiosis I, when homologous centromeres segregate, a persistent intersister JM distal to a crossover would prevent segregation in a similar way that non-cleaved cohesin mediating sister chromatid cohesion prevents segregation [35,57,58]. A persistent interhomolog JM would also prevent segregation. During meiosis II, when sister centromeres segregate, an intersister JM would directly prevent segregation if not aided by a DNA helicase, such as Fbh1.
Other meiotic phenotypes of dbl2Δ are accounted for by this proposal. The suppression, by elimination of DSBs (rec12Δ), of the dbl2Δ chromosome segregation failure (Fig 2) and of persistent JM formation (Figs 4B, 4C and S4) is accounted for by the requirement for Rec12 and DSBs to form meiotic JMs. Similarly, the suppression of both Rad51-focus accumulation and chromosome segregation failure in dbl2Δ mutants by rad52Δ, rad55Δ, rad57Δ, sfr1Δ, and rad54Δ is accounted for by the enhancement of Rad51 strand exchange by the corresponding proteins (Table 2). It is particularly noteworthy that dmc1Δ does not suppress dbl2Δ but rad52Δ does (Table 2). Dmc1 is not required for formation of intersister JMs at the loci tested in S. pombe [59] or in S. cerevisiae [60], and genetic data indicate that S. pombe Rad52 is required for intersister but not interhomolog JM formation [61]. Furthermore, purified S. pombe Rad52 stimulates Rad51 strand exchange, although it inhibits Dmc1 strand exchange [62,63]. Failure to convert intersister JMs into HJs would not reduce recombinant frequencies but would prevent proper chromosome segregation and decrease viable spore yield, as observed in dbl2Δ (Table 1). Failure to recruit Fbh1 to JMs in dbl2Δ cells would allow accumulation of Rad51 foci (Fig 6; Table 2) and rare JMs (Figs 4B and S4).
The mitotic phenotypes of dbl2Δ are similarly accounted for. After DNA damage, the appearance of Fbh1 foci is Dbl2-dependent (Fig 7). dbl2Δ mutants are sensitive to camptothecin (CPT) (Fig 5), which inhibits topoisomerase 1 and leaves DNA lesions that must be repaired [52]. Mitotic DNA repair is thought to be primarily with the sister chromatid [64]. If some intersister JMs other than HJs, such as D-loops and hemicatenanes, persist in dbl2Δ mutants, expression of the bacterial RusA HJ resolvase would not alleviate the problem, as observed (Fig 5). In contrast, the CPT-sensitivity and presumed accumulation of HJs in eme1Δ mutants, which lack the S. pombe HJ resolvase Mus81-Eme1 [28], is suppressed by expression of catalytically active RusA (Fig 5).
The phenotypes of dbl2Δ and fbh1Δ are remarkably similar. In both mutants during meiosis, DSB formation and repair as well as recombination are close to wild-type levels, chromosomes frequently fail to segregate during meiosis I, Rad51 foci persist beyond meiotic prophase, and chromosome missegregation is suppressed by rad51Δ, rad52Δ or rad57Δ (Figs 1, 2, 3, 4, 6 and S1; Table 2) [23,51]. The slightly higher spore viability and milder chromosome segregation defects in dbl2Δ than in fbh1Δ may result from residual Dbl2-independent binding of Fbh1 to DSBs or JMs. Moreover, during mitotic growth, both dbl2Δ and fbh1Δ mutant cells are sensitive to camptothecin [65] and show negative genetic interaction (phenotype of double mutant is stronger than single mutants) with mutants defective in DNA repair such as srs2Δ and rsc4Δ and positive genetic interaction (suppression of the mutant phenotype) with rad55Δ and rad57Δ [66].
It is interesting that Dbl2 promotes the formation of Fbh1 foci in the absence of Rad51 after DNA damage (Fig 7C). This result suggests that Dbl2 recruits Fbh1 to DNA lesions before Rad51-ssDNA filament and JM formation. This feature may reflect a type of "fail-safe" mechanism, in which the repair machinery is recruited even before it is needed. This appears to be the case for some other events in meiosis. For example, in some species, such as S. cerevisiae, the Mre11-Rad50-Nbs1 (MRN) complex is required for meiotic DSB formation even though its catalytic activity is apparently needed only after DSB formation, i.e., for the processing of DNA ends to allow Rad51-ssDNA filament formation [67]. If Fbh1 is recruited to most meiotic DSBs before JM formation, it may play a more prominent role in meiotic DSB repair than discussed above: it may aid metabolism of JMs other than the rare (few per cell) persistent JMs detected in dbl2Δ mutants (Figs 4B, 4C and S4).
Purified Fbh1 can remove Rad51 from Rad51-ssDNA filaments and has DNA helicase activity [22]. Therefore, both Rad51 foci and JMs should accumulate in dbl2Δ and fbh1Δ mutants, as observed (Table 2; Figs 4B, 4C and S4). This is not the case for other S. pombe helicase mutants, fml1Δ and fml2Δ, or a chromatin remodeling mutant, rdh54Δ (Table 2). Thus, Fbh1 has a special role, which may be related to its having a second activity, ubiquitin ligase as part of the SCF complex [68]. We cannot distinguish whether the phenotypes observed here are due to loss of the helicase or the ligase activity, or both, but the accumulation of JMs (Figs 4B, 4C and S4) suggests that the Fbh1 helicase is important for meiotic joint molecule metabolism. The ligase activity may also be important to remove Rad51, but this action might be a result of the helicase alone.

Multiple means for meiotic DNA joint molecule processing
Previous work has shown that during meiosis JMs can be processed in a variety of ways into chromosomes suitable for segregation. In S. cerevisiae at least six mechanisms have been shown or inferred to convert JMs into recombinant chromosomes [69,70] (and references therein). In contrast, in S. pombe only one mechanism has been described-the formation of HJs and their resolution nearly exclusively by the Mus81-Eme1 resolvase [27][28][29]44,71]. Here, we propose an additional mechanism-Fbh1, recruited by Dbl2, acts on D-loops to reverse them or to convert them into HJs suitable for Mus81-Eme1 resolution. This action is important, for in the absence of Dbl2 viable spore yields are reduced by a factor of 13 (Table 1) and chromosomes missegregate in up to half of the cells (Figs 1, 2 and 3). There is little effect on recombinant frequencies, however, because these JMs are rare (Figs 4B and S4) and, we propose, many of these JMs are intersister, which cannot produce recombinants. Thus, studies of recombination have overlooked the important function of Dbl2 and Fbh1 for meiotic chromosome segregation.

Conservation of Dbl2 function
Dbl2 is predicted to encode a 78 kDa protein with a domain of unknown function (DUF2439) at the N-terminus. Our bioinformatic searches revealed that within the DUF2439 region Dbl2 is highly conserved in fungi, animals and plants. Orthologs can be detected in Saccharomyces cerevisiae (Mte1, YGR042W), Homo sapiens (ZGRF1) and Arabidopsis thaliana (AT4G10890) (S8 Fig). The only other sequence family that is significantly related to the DUF2439 domain is the one including Rdh54 (S. cerevisiae, S. pombe) and RAD54B (H. sapiens) (S8 Fig). Interestingly, some proteins of both sequence families contain an AAA+ ATPase domain and/or a helicase domain (ZGRF1, Rdh54, RAD54B) at their C-termini. The C-terminal part of Dbl2 is conserved only within the Schizosaccharomyces species; it is enriched for polar residues and is predicted to be highly disordered and thus unlikely to form an active helicase. However, Dbl2 is required for targeting helicases Fml1 and Fbh1 to DSBs (this work and [31]). Similarly, recent studies in S. cerevisiae showed that Mte1 is required for localization of Mph1 helicase at sites of DNA damage and regulates Mph1 activity [72][73][74]. Thus, we speculate that Dbl2 acts as an adaptor or recruiter for helicases Fml1 and Fbh1 and perhaps for other proteins. The mechanism may be similar to that of other adaptor proteins containing disordered C-terminal regions such as endocytic adaptor proteins Epsin1 and AP180 and Atg13 adaptor, which controls the initiation of autophagy [75,76].
What is the role of the DUF2439 domain? Surprisingly, the N-terminal truncation that removes the DUF2439 domain had no effect on either Dbl2 focus formation at DSBs or on its ability to confer camptothecin-resistance [31]. Similarly, deletion of the budding yeast rdh54 confers sensitivity to MMS but N-terminal truncations of Rdh54 that remove or truncate the DUF2439 domain do not affect MMS sensitivity [77]. Interestingly, these Rdh54 truncation mutants lost their ability to interact with Rad51 and had impaired ability to dissociate Rad51-DNA complexes [77][78][79]. The reason why the DUF2439 domain is important for Rad51 interaction but not for resistance to camptothecin or MMS is not known, but one possible explanation is that the role of Dbl2 and Rdh54 in repair of MMS-or camptothecin-induced DNA damage may not be through Rad51. The N-terminal domain of human RAD54B which includes the DUF2439 domain binds to branched DNA substrates and interacts with both RAD51 and DMC1 [80]. Similarly as in Rdh54 and RAD54B, the DUF2439 domain of Dbl2 may mediate interaction with Rad51. This would allow Dbl2 to bring Fml1 and Fbh1 helicases directly to Rad51 nucleoprotein filaments. Consistent with this notion is our finding that Dbl2 interacts with Rad51 and weakly with Fml1 in yeast two-hybrid assays (S9 Fig). Interestingly, our observation that the Fbh1 foci are formed in the absence of Rad51 (Fig 7C) indicates that, to some extent, Dbl2 is able to promote Fbh1 focusformation at meiotic DSBs independently of Rad51. We observed no interaction between Dbl2 and Fbh1 in yeast two-hybrid assays (S9 Fig). These results do not, however, exclude the possibility that Dbl2 interacts directly with Fbh1.
Both Dbl2 and Fbh1 are evolutionarily conserved proteins present from yeast to humans, suggesting that our results may apply widely. A notable exception is the absence of an Fbh1 homolog in the budding yeast S. cerevisiae. However, it has been proposed that S. cerevisiae Srs2 helicase is a functional counterpart of Fbh1 [81]. The Dbl2 and Fbh1 proteins, like other proteins central to DNA repair, are widely conserved and play crucial roles in maintaining cell viability, genome integrity, and high fertility, making their further study important.

Strains, growth media and general methods
The genotypes of the yeast strains used in this study are listed in the S1 Table. Standard media (rich YES and appropriately supplemented minimal EMM2 and sporulation SPA) were used to maintain, grow and mate S. pombe strains [39,[82][83][84]. To induce mating and meiosis (Figs 1, 2, 3, 6 and 7 and Table 2), cells were grown in liquid YES to mid-log phase at 32°C, washed three times with water, mixed (for heterothallic matings), transferred to EMM2-NH 4 Cl plates, and incubated at 25°C for 10-17 hr before examination [85]. DNA was extracted from meiotically induced cells and analyzed as described [86]. S. pombe was transformed using the lithium acetate method and genes deleted as described [87]. The immunostaining and microscopy used to analyze chromosome segregation and subcellular localization of Rec8 and Rad51 in S. pombe cells were performed as described [7]. Subcellular localization of Rad51 was determined using anti-Rhp51 polyclonal antibody (Cosmo Bio) diluted 1:500. No foci were detected in rad51Δ mutant cells, indicating that in our assays anti-Rhp51 antibody specifically detected Rad51 (S10 Fig and Table 2). Live-cell imaging and spore viability determinations were performed as described [88]. Viable spore yields and recombinant frequencies (Table 1) were determined as described [89].
Plasmid loss assay fbh1Δ (JG17775) and fbh1Δ dbl2Δ (JG17777) cells harboring plasmid pMW651 (expressing Fbh1-YFP and containing the LEU2 + marker, which complements the leu1-32 mutation in the parental strains) were grown in selective EMM medium lacking leucine at 32°C to mid-log phase. The cells were washed with water and resuspended in nonselective YES liquid medium, incubated at 32°C for 18 hr (7 generations) and then plated on YES plates. Colonies were formed after 3 days at 32°C and replica-plated onto selective plates (EMM lacking leucine). The percentage of plasmid loss per generation was determined as described by Osman et al. [90].

Bioinformatics analysis
The DUF2439 domain is conserved within the Dbl2/ZGRF1 protein family in fungi, metazoa and plants. In a series of NCBI-BLAST searches [91] within the NCBI non-redundant protein database using S. pombe Dbl2 as starting query and applying significant e-value thresholds (<0.001), we identified an N-terminally located region (Dbl2 amino acids 7 to 77) conserved in fungi, animals and plants, that is also described as the domain of unknown function DUF2439 in the Pfam database [92]. Orthologues include S. cerevisiae YGR042W (= Mte1) and human ZGRF1, and reciprocal NCBI-Blast searches in the S. pombe proteome with the respective orthologous regions significantly yielded only Dbl2. The animal ZGRF1 protein family is extended, relative to yeast proteins, by one GRF zinc finger region and two AAA+ ATPase domains at the C-terminal half.
The Rdh54/RAD54B protein family shares the DUF2439 domain. To find distantly related DUF2439-containing proteins, we performed iterative NCBI-PSIBLAST searches. For 16 out of 17 queries, the NCBI-PSIBLAST search converged before it reached the specified threshold of 20 iterations. Other than the Dbl2 and ZGRF1 proteins, all significant hits were to an unidentified conserved domain in the Rdh54/RAD54b family. For example, S. pombe Dbl2 (residues 3-85) found Cryptococcus gattii Rad54b in iteration 3 (e-value of 2 x 10 −4 ), fission yeast Rdh54 in iteration 4 (e-value of 3 x 10 −7 ), and human RAD54b in iteration 6 (e-value of 1 x 10 −6 ). The relationship was confirmed in a reciprocal search with S. pombe Rdh54 (residues 49-128), which identified Dbl2 in iteration 6 (e-value of 6 x 10 −4 ) and human ZGRF1 in iteration 4 (e-value of 8 x 10 −5 ). In an independent approach, using hmmsearch with the Pfam DUF2439 hidden Markov model [93], Rdh54 and Rad54b proteins were among the significant hits, including S. pombe Rdh54 (e-value of 2.2 x 10 −5 ) and Dictyostelium fasciculatum RAD54b (e-value of 3.1 x 10 −5 ). Whereas Rad54 and RAD54b protein families share the same overall domain architecture, including their DEAD-like helicase domain, the N-terminal region is specific to each protein. In human RAD54b, the N-terminal region has been reported to bind to Rad51 and Dmc1 and branched DNA structures [80].  Fig 4A, DSBs are formed and repaired in a similar manner in strains GP6656 (dbl2 + ) and GP8664 (dbl2Δ). The results are from two independent experiments from which extracted DNA was digested with NotI and analyzed by pulsed-field gel electrophoresis and Southern blot hybridization using a probe at the left end of the 501 kb NotI fragment J (band at the top of the gel) [42]. The fraction of total DNA broken at mbs1 and at mbs2 (indicated by arrows on the right) at the indicated time for each strain is shown below each blot. (B) Strains GP6656 (dbl2 + ), GP8664 (dbl2Δ), and GP8836 (rec12Δ dbl2Δ) were induced and their DNA analyzed as in Fig 4B and 4C for joint DNA molecules. Southern blots of DNA extracted at the indicated times after meiotic induction were hybridized with a radioactive probe (~1 kb long) near the ade6-3049 DSB hotspot on the 11.8 kb BsrGI fragment or near the mbs1 hotspot on the 10.5 kb BsrGI fragment [1]. Black arrowheads indicate Y-shaped replication (0-3 hr) and recombination intermediates (4-5 hr); white arrowsheads indicate Holliday junctions at 4 and 5 hr. Persistent joint molecules seen at 6, 7, and 8 hr are X-shaped (white arrows) or Y-shaped (black arrows). These joint molecules at the ade6-3049 hotspot persist in dbl2Δ but not in dbl2 + and are Rec12-dependent.  [2] and visualized in Jalview [3], using the ClustalX colouring profile. The sequence identifiers from the NCBI protein database are given in parentheses. The numbers of the first and last residues flank the region aligned. (TIF) S9 Fig. Dbl2 interacts with Rad51 and Fml1 in yeast two-hybrid assay. Strains expressing Dbl2 fused to the GAL4 transcription activation domain and Rad51, Fbh1 or Fml1 fused to the GAL4 DNA-binding domain were grown on SD plates lacking tryptophan and leucine (SD-L, W) and then spotted at 5-fold serial dilutions on SD plates lacking tryptophan and leucine (SD-L,W) or SD plates lacking tryptophan, leucine and histidine (SD-L,W,H) or SD plates lacking tryptophan, leucine and adenine (SD-L,W,A). The empty vectors pGADT7 and pGBKT7 containing GAL4 transcription activation domain and GAL4 DNA-binding domain, respectively were used as negative controls. Growth on plates without histidine or without adenine indicates interaction between the fusion proteins [4]. (TIF) S10 Fig. Anti-Rhp51 antibody detects foci in wild-type but not rad51Δ zygotes. To test the specificity of anti-Rhp51 antibody, we analyzed subcellular localization of Rad51 using anti-Rhp51 polyclonal antibody (Cosmo Bio) diluted 1:500 in wild-type (JG11355) and rad51Δ (JG17993, JG17540) prophase I cells. Cells were mated on SPA sporulation agar and at 10-17 hr fixed and immunostained for tubulin and Rad51, and examined by fluorescence microscopy; DNA was visualized by Hoechst staining. (TIF) S1 Table. S. pombe strains. (DOCX) S2 Table. Holliday junctions are formed and repaired similarly in wild-type and dbl2Δ mutant, but Rec12-dependent joint molecules persist in dbl2Δ late meiosis-analysis at the mbs1 DSB hotspot. (DOCX) S3 Table. Holliday junctions are formed and repaired similarly in wild-type and dbl2Δ mutant, but Rec12-dependent joint molecules persist in dbl2Δ late meiosis-analysis at the ade6-3049 DSB hotspot.