Rad51-mediated interhomolog recombination during budding yeast meiosis is promoted by the meiotic recombination checkpoint and the conserved Pif1 helicase

During meiosis, recombination between homologous chromosomes (homologs) generates crossovers that promote proper segregation at the first meiotic division. Recombination is initiated by Spo11-catalyzed DNA double strand breaks (DSBs). 5’ end resection of the DSBs creates 3’ single strand tails that two recombinases, Rad51 and Dmc1, bind to form presynaptic filaments that search for homology, mediate strand invasion and generate displacement loops (D-loops). D-loop processing then forms crossover and non-crossover recombinants. Meiotic recombination occurs in two temporally distinct phases. During Phase 1, Rad51 is inhibited and Dmc1 mediates the interhomolog recombination that promotes homolog synapsis. In Phase 2, Rad51 becomes active and functions with Rad54 to repair residual DSBs, making increasing use of sister chromatids. The transition from Phase 1 to Phase 2 is controlled by the meiotic recombination checkpoint through the meiosis-specific effector kinase Mek1. This work shows that constitutive activation of Rad51 in Phase 1 results in a subset of DSBs being repaired by a Rad51-mediated interhomolog recombination pathway that is distinct from that of Dmc1. Strand invasion intermediates generated by Rad51 require more time to be processed into recombinants, resulting in a meiotic recombination checkpoint delay in prophase I. Without the checkpoint, Rad51-generated intermediates are more likely to involve a sister chromatid, thereby increasing Meiosis I chromosome nondisjunction. This Rad51 interhomolog recombination pathway is specifically promoted by the conserved 5’-3’ helicase PIF1 and its paralog, RRM3 and requires Pif1 helicase activity and its interaction with PCNA. This work demonstrates that (1) inhibition of Rad51 during Phase 1 is important to prevent competition with Dmc1 for DSB repair, (2) Rad51-mediated meiotic recombination intermediates are initially processed differently than those made by Dmc1, and (3) the meiotic recombination checkpoint provides time during prophase 1 for processing of Rad51-generated recombination intermediates.


Introduction
Sexually reproducing diploid organisms create haploid gametes through the specialized cell division of meiosis so that the fusion of two gametes maintains a constant chromosome number in each generation. Meiosis divides the chromosome number in half by having one round of DNA replication followed by two rounds of chromosome segregation. At Meiosis I (MI), homologs segregate to opposite poles, while sister chromatids are disjoined at Meiosis II (MII). Accurately sorting the chromosomes such that each gamete contains exactly one copy requires that homologs be physically connected by a combination of crossovers and sister chromatid cohesion [1]. The process of meiosis is evolutionarily conserved and defects in crossover formation result in increased MI nondisjunction and inviable progeny in organisms ranging from yeast to mammals [2,3].
The process of crossover formation begins after DNA replication when the highly conserved, meiosis-specific, topoisomerase-like Spo11 protein catalyzes programmed DNA double strand breaks (DSBs) at preferred sites in the genome called hotspots [4][5][6][7]. The 5' ends are then resected, producing 3' single strand (ss) tails [8][9][10]. In many organisms, including budding yeast and mammals, these 3' ends are bound by two recombinases: Rad51, which is also used for mitotic recombination, and the meiosis-specific Dmc1 [11][12][13][14][15][16][17][18]. In yeast and mammals, Rad51 and Dmc1 assemble on the ssDNA in adjacent patches as a right-handed helix to make a presynaptic filament [19][20][21][22][23][24]. The presence of Rad51, but not its strand exchange activity, is required for normal loading of Dmc1 onto the presynaptic filament and wild-type (WT) levels of meiotic recombination [25][26][27][28][29][30]. In mice, Dmc1 is localized at the 3' ends of each filament [22]. The function of recombinases is to mediate a homology search of the genome followed by strand invasion of the donor duplex to form a three-strand structure called a displacement (D) loop [17,31,32]. D-loop formation begins when a presynaptic filament interacts with the donor duplex to form a nascent D-loop with an unstable paranemic joint [33][34][35][36]. Paranemic joints are subsequently converted to stable plectonemic joints (or mature D-loops) when the recombinase is removed from the invading ssDNA in concert with formation of heteroduplex DNA between the invading strand and the donor strand of opposite polarity, thereby displacing the strand of like-polarity [34,36,37]. The bacterial recombinase, RecA, can mediate both the homology search and conversion of nascent to mature D-loops [38]. In contrast, Rad51 can create nascent D-loops, but requires a DNA translocase called Rad54 to make mature D-loops [32,35,39,40].
Rad54 is related to the Snf2 family of helicases and functions as a motor protein that translocates along double-stranded DNA [41,42]. Rad54 has several functions involved in making D-loops: it directly interacts with Rad51 and stabilizes the Rad51-ssDNA filament, facilitates binding of the presynaptic filament to the donor DNA, and has been proposed to promote a translocation-based homology search along the DNA [39,[43][44][45][46][47][48][49]. In addition, Rad54 can remove Rad51 from nascent D-loops leading to a model in which Rad54 acts as a "pump" that couples removal of Rad51 with heteroduplex formation to generate a mature D-loop [37,50,51].
Once a mature interhomolog D-loop is made, the 3' end is extended by DNA Polymerase δ (Pol δ) [52][53][54][55]. Some of the extended strands are disassembled from D-loops by the STR (Sgs1, Top3, Rmi1) complex and annealed to the 3' tails on the other side of the break, leading to noncrossovers through a process called synthesis-dependent strand annealing [56][57][58][59][60][61]. Alternatively, extension of the invading strand can result in the displaced single strand annealing to the other side of the DSB to create a double Holliday junction [62][63][64]. During meiosis, most D-loops that form double Holliday junctions are processed by a functionally diverse set of proteins called the "ZMMs" that protect them from disassembly and are resolved almost exclusively as crossovers [56][57][58]65,66] In addition to promoting interhomolog crossovers, recombination intermediates generated by the ZMM pathway are necessary for chromosome synapsis [66][67][68]. At the beginning of prophase I, sister chromatids condense along protein cores made of meiosis-specific proteins to create structures called axial elements [69]. In budding yeast, these proteins are Hop1, Red1 and Rec8-containing cohesin [70][71][72]. ZMM-mediated recombination between homologs creates synaptic initiation sites from which the meiosis-specific transverse filament protein, Zip1, and the central element proteins, Ecm11 and Gmc2, polymerize to "zip" homologous axial elements together to form synaptonemal complexes (SCs) [65][66][67]73].
Mek1 is also the effector kinase for the meiotic recombination checkpoint (MRC) that couples prophase I exit with DSB repair [91]. Mek1 phosphorylation controls the timing of prophase I exit by inhibiting the meiosis-specific transcription factor Ndt80 [77,91,92]. Since synapsis only occurs when ZMM crossover-designated recombination intermediates have been generated between homologs, complete chromosome synapsis serves to indicate that additional recombination is no longer necessary [66,93]. Synapsis reduces Mek1 activity both by decreasing DSB formation and by removing the bulk of Mek1 from chromosomes [94][95][96][97]. As a result, Ndt80 is activated and genes required for crossover formation, SC disassembly and meiotic progression are expressed [91,94,[98][99][100][101][102]. Complete inactivation of Mek1 allows Rad51-Rad54 to repair any remaining DSBs prior to entry into the meiotic divisions [18,94,103]. Rad51 activity during meiotic prophase I can therefore be divided into two phases that are delineated by the Mek1-controlled activation state of Ndt80. In Phase 1, Rad51 is inactive and Dmc1 mediates recombination primarily between homologs; in Phase 2, Rad51-Rad54 mediates repair of residual DSBs through strand invasion of sister chromatids. A similar transition from interhomolog to intersister recombination late in prophase I has been observed during nematode meiosis and DSB repair in late meiosis I prophase in mammals also appears to switch to a Rad51-mediated pathway [104][105][106].
The fact that two independent mechanisms inhibit Rad51 strand invasion in Phase 1 suggests that having Dmc1 and Rad51 active at the same time is disadvantageous to the cell. However, when Rad51 is constitutively activated in the presence of DMC1 by hed1Δ or hed1Δ RAD54-T132A (for simplicity, henceforth hedΔR), the phenotypes are modest: a two-fold increase in intersister recombination and a delay in meiotic progression with little to no effect on spore viability [88,89,107]. Rad51 inhibition is therefore only partially responsible for interhomolog bias, indicating there are other Mek1 targets yet to be discovered. In contrast, a hed1Δ dmc1Δ mutant exhibits an approximately 30-fold decrease in interhomolog bias [88,89]. The milder phenotypes observed in hed1Δ compared to hed1Δ dmc1Δ led to the hypothesis that Dmc1 itself partially inhibits Rad51-mediated strand exchange [88,89]. An alternative hypothesis is that Dmc1 outcompetes activated Rad51 in strand invasion of DSBs [107]. However, neither of these hypotheses explains why constitutive activation of Rad51 by hedΔR results in a meiotic progression delay [88,107].
In vegetative cells, break-induced replication (BIR) is a DNA repair pathway that occurs when, because homology on the other side of the DSB is poor or absent, the invading strand of a Rad51-mediated D-loop is extended to the end of the chromosome by Polδ [108,109]. BIR requires the conserved 5'-3' helicase Pif1 that interacts with PCNA and stimulates Polδ-mediated extension, creating a migrating D-loop [109][110][111]. Recent work has shown that PIF1 also has a role in meiotic recombination in yeast [112]. Pif1 localizes to meiotic DSBs, where it interacts with PCNA to promote extension of the invading strand in D-loops formed by Dmc1 [112]. In wild-type (WT) cells, however, the contribution of PIF1 to strand extension is relatively minor, as its activity is inhibited by the presence of the Mer3-MutLβ complex [112].
To better understand why the strand exchange activity of Rad51 is normally repressed during Phase 1, we have investigated the basis for the hedΔR meiotic progression delay and discovered that allowing Rad51 to interact with Rad54 in Phase 1 results in Rad51 mediated interhomolog strand invasion at a subset of DSBs that is less robust than invasion mediated by Dmc1. The resulting interhomolog recombination intermediates take longer to repair, thereby keeping the MRC active for longer. In the absence of the MRC, constitutive activation of Rad51 results in increased intersister recombination, decreased crossovers, increased MI nondisjunction and reduced spore viability. Furthermore, our work suggests that processing of these intermediates is promoted by PIF1 and its paralog, RRM3. We propose that Rad51 is normally inhibited in Phase 1 to prevent it from competing with Dmc1 for strand invasion and that multiple mechanisms have evolved to promote processing of interhomolog Rad51-mediated D-loops should they occur. In addition, our findings also reveal a potential mechanism for the Rad51-dependent DSB repair events that occur during Phase 2.

Constitutive activation of Rad51 during meiosis delays DSB disappearance and prophase I exit
When Rad51 is active throughout prophase I, entry into the meiotic divisions is delayed [88,107]. This result was reproduced using the hedΔR combination that removes both Mek1-dependent mechanisms that inhibit Rad54 from interacting with Rad51 ( Fig 1A). There was significant variability in the kinetics of meiotic progression between different biological replicates of both WT and hedΔR, but the amount of sporulation was similar in both strains ( Fig  1A-1C). To compare the timing of meiotic progression between different strains, a T 50 value was calculated by determining the time it took for each culture to reach half of its maximum %MI + MII value determined by fluorescent staining of nuclei (MI = binucleate, MII = tetranucleate). The average T 50 value for hedΔR was 6.4 hours, significantly longer than the WT average of 5.6 hours ( Fig 1C).
One explanation for the hedΔR meiotic progression delay is that constitutive activation of Rad51 delays DSB repair, thereby keeping the meiotic recombination checkpoint active for longer. DSBs and nascent D-loops occurring throughout the genome can be cytologically detected in single cells by staining chromosome spreads with antibodies against Rad51 and Dmc1 [27] (Fig 1D). As controls for our antibodies, Rad51 and Dmc1 foci were only observed in meiotic cells, and were dependent on RAD51 and DMC1, respectively (S1A Fig). In addition, few Dmc1 foci were observed in rad51Δ, as expected given that RAD51 is needed to promote assembly of Dmc1 onto the presynaptic filament (S1A Fig) [25,27]. Finally, no foci were observed in a spo11Δ diploid, confirming that the Dmc1/Rad51 foci were dependent on Spo11-generated DNA DSBs ( Fig 1D).
In hedΔR, the fraction of cells containing both Rad51 and Dmc1 foci accumulated to higher levels and persisted at later timepoints compared to WT (Fig 1E). The lifespan of the period when nuclei have a substantial number of DSBs (� 3 Rad51 and Dmc1 foci per nucleus) was determined by graphing the percent of Dmc1 and Rad51 positive spreads at each time point for each replicate, then dividing the area under the curve (AUC) in each graph by the maximum %MI+MII cells (Fig 1F) [113]. The average lifespan of cells exhibiting both Rad51 and Dmc1 foci was significantly longer than WT (4.4 vs 2.7 hours) (Note that this analysis underestimates this lifespan in hedΔR since the frequency of Rad51 + Dmc1 + cells did not reach zero by eight hours).
Unrepaired DSBs cause extended activation of the MRC and thus delay exit from prophase I [77,114,115]. Measuring meiotic progression by counting the number of nuclei does not provide information as to which part of meiosis is delayed. To determine whether prophase I length is specifically increased by hedΔR, whole cell immunofluorescence monitoring the presence of the axial element protein Red1 was used (Fig 1G). Red1 is degraded upon Ndt80 activation due to the appearance of the polo-like kinase, Cdc5, making it an excellent marker for prophase I [94,98,101,103]. The cytological specificity of the Red1 antibody was confirmed by the lack of signal in cells from an ndt80Δ red1Δ control (S1B Fig).
In WT meiotic timecourses, the frequency of Red1 + cells peaked at three hours and gradually decreased until approaching zero by~8 hours (Fig 1H). In hedΔR, Red1 + cells appeared with WT kinetics but disappeared more slowly than WT. A quantitative value for prophase I lifespan was determined for each replicate by calculating the AUC for % Red1 + cells and normalizing to the maximum percentage of MI+MII cells [Prophase lifespan (hr) = AUC/ Maximum # cells completing either MI or MII]. The average prophase I lifespan for hedΔR was 3.8 'hours, significantly longer than WT (3.0 hours) (Fig 1I). The meiotic progression delay observed for hedΔR can be accounted for by the increase in prophase I length, as the difference between hedΔR and WT for T 50 and prophase I lifespan was the same (~0.8 hours). Rad51 strand exchange activity is responsible for the hedΔR delay in prophase I exit Rad51:Dmc1 stoichiometry in the presynaptic filament is important for normal interhomolog recombination [116]. Hed1 and Rad54 can stabilize Rad51 on the filament, raising the possibility that the problems in DNA repair observed in hedΔR are due to improperly assembled presynaptic filaments resulting from the absence of HED1 and/or the presence of Rad54 [49,117,118]. An alternative possibility is that presynaptic filament assembly is normal, but Rad51-mediated strand invasion events take longer to process than those created by Dmc1. The latter hypothesis was tested using the rad51-II3A mutant, which forms normal presynaptic filaments but is defective in strand exchange [25].
The rad51-II3A diploid exhibited a meiotic progression delay and a small, but significant, decrease in spore viability, likely due to the failure of Rad51 to repair residual DSBs in Phase 2 (Fig 2A-2C) [25,94,103]. Consistent with this idea, the distribution of viable spores in rad51-II3A tetrads was consistent with random spore viability and MI nondisjunction was not increased (Fig 3F and 3H). Furthermore, while the T 50 value for meiotic progression in rad51-II3A was significantly greater than WT, prophase I length was not significantly different (Fig 2C and 2E). One explanation for the rad51-II3A meiotic progression delay is that some DSBs persist past MI, thereby triggering the DNA damage checkpoint prior to MII [119].
To see if constitutive Rad51-mediated strand invasion delays prophase I exit, rad51-II3A was combined with hedΔR, which should allow the strand exchange defective Rad51-II3A Sporulation. Cells were examined by light microscopy for the presence of asci containing one to four spores. For each biological replicate, 200 cells were counted. (C) Quantification of meiotic progression timing. T 50 represents the amount of time (hr) it took for a sporulation culture from Panel A to reach one-half its maximum %MI+MII value. (D) Cytological assay for the presence of DSBs. Representative images of chromosome spreads generated from WT (NH716) and spo11Δ (NH1055) cells taken after four hours in Spo medium. DNA was visualized with DAPI and antibodies against Rad51 and Dmc1 were used to detect the presence of DSBs. Scale bar is 2 μm. (E) Quantification of percent cells exhibiting at least three Dmc1 and Rad51 foci. Chromosome spreads of WT (n = 3), and hedΔR (n = 3) diploids from a subset of the replicates shown in Panel A were analyzed for the presence of Rad51 and Dmc1 foci at various timepoints. Nuclei containing a minimum of three Rad51 and three Dmc1 foci were scored as positive. At least 120 nuclei were examined for each timepoint. (F) Lifespan of cells exhibiting both Rad51 and Dmc1 foci. Each time course from Panel E was graphed and the lifespan of Rad51/Dmc1 positive nuclei (hr) was calculated by dividing the area under each curve by the maximum number of cells that completed either MI or MII for the corresponding culture [113].    Fig 1. (E) Spore viability. Sporulated cells from both liquid and solid sporulation media were dissected to determine the percent of protein to interact with Rad54. The hedΔR rad51-II3A diploid exhibited lower spore viability than hedΔR, as expected since the constitutively active Rad51-II3A protein is unable to repair residual DSBs during Phase 2 (Fig 2A). In contrast, preventing Rad51 mediated strand invasion in the hedΔR rad51-II3A diploid suppressed the hedΔR meiotic progression delay by decreasing the length of prophase I to be the same as WT (Fig 2B-2E). The average T 50 value for hedΔR rad51-II3A was 5.7 hours, significantly less than the 6.4 hours observed for hedΔR. In addition, prophase I lifespan was decreased significantly from 3.8 hours in hedΔR to 3.1 hours in hedΔR rad51-II3A. These results show that constitutively allowing Rad54 to interact with Rad51 during meiosis results in strand exchange events that trigger a delay during prophase I.

The prophase I delay caused by constitutive activation of Rad51 depends on the MRC
The delayed exit from prophase I and persistence of DSBs in hedΔR suggests that MRC activity is extended in response to activated Rad51. If true, then eliminating this checkpoint should rescue the hedΔR prophase I delay. Mek1 controls the MRC by binding to an RPSKR motif in Ndt80 and phosphorylating the Ndt80 DNA binding domain [91]. An NDT80 allele lacking the RPSKR motif (called NDT80-mid, for Mek1 Interaction Defective) makes Ndt80 insensitive to the checkpoint without affecting its ability to activate transcription [91].
The NDT80-mid diploid exhibited similar meiotic progression timing as WT (Fig 3A  and 3B). However, abrogating the checkpoint in the presence of activated Rad51 (hedΔR NDT80-mid) did not eliminate the meiotic progression delay seen in hedΔR (Fig 3A and 3B). One explanation for the failure of NDT80-mid to suppress the hedΔR meiotic progression delay is that this mutant combination lengthens meiotic progression for a different reason than hedΔR alone: e.g., a delay in the metaphase I to anaphase I transition due to activation of the spindle assembly checkpoint in response to nonexchange chromosomes [120]. Since the MRC specifically controls the timing of prophase I exit, prophase I lifespan was determined for hedΔR NDT80-mid. For reasons that are unclear, hedΔR NDT80-mid was delayed in entering prophase I compared to WT, NDT80-mid and hedΔR (Fig 3C). Despite entering prophase I later, hedΔR NDT80-mid cells exited prophase I significantly earlier than hedΔR, with an average lifespan of 2.9 hours compared to 3.8 hours for hedΔR ( Fig 3D). These data indicate that constitutive activation of Rad51 during meiosis results in strand exchange intermediates that take longer to process and trigger the MRC to delay exit from prophase I.

The MRC is required for proper Meiosis I disjunction and spore viability when Rad51 is constitutively active during prophase I
If the MRC is providing time for cells to process strand invasion intermediates generated by Rad51, there should be phenotypic consequences when the checkpoint is absent. Diploids in which either Rad51 was constitutively activated (hedΔR), or the MRC was eliminated (NDT80-mid), exhibited very small, but significant, decreases in spore viability ( Fig 3E). In viable spores. At least 20 tetrads were dissected for each replicate. In addition to the diploids listed for Panel A, hedΔR NDT80-mid rad51-II3A (NH2705 RCEN) and rad51-II3A were included. (F) The distribution of viable spores among tetrads dissected in Panel E. (G) Schematic of the chromosome VIII homologs used to monitor nondisjunction. The tetrad on the left is an example of normal chromosome VIII segregation while MI nondisjunction (NDJ) occurred in the tetrad on the right. B = blue, R = red, Bl = black. (H) Frequency of chromosome VIII MI nondisjunction. Numbers above each bar indicate the total number of tetrads assayed. Panels A-E for WT, hedΔR, and rad51-II3A used the same data shown in Figs 1 and 2. The statistical significance of differences between strains in Panels B and D was determined using an unpaired, twotailed Student's T-test, while the Mann-Whitney test was used for Panels E and H. ( � = p < .05; �� = p < .01; ��� = p < .001, ���� = p < .0001) https://doi.org/10.1371/journal.pgen.1010407.g003 contrast, activating Rad51 in the absence of the checkpoint in the hedΔR NDT80-mid diploid dramatically reduced spore viability to an average of~60% ( Fig 3E). This value is less than that predicted from the product of the individual spore viabilities of hedΔR and NDT80-mid (86.5%), indicating a synergistic effect. Furthermore, there was increased variability in the spore viability observed between individual colonies in the hedΔR NDT80-mid strain compared to hedΔR alone ( Fig 3E).
The tetrads from hedΔR NDT80-mid exhibited a pattern of spore inviability consistent with MI nondisjunction, where four viable spore tetrads are reduced in favor of two and zero viable spore tetrads (Fig 3F) [121]. To monitor MI disjunction directly, genes encoding either a spore autonomously-expressed red fluorescent protein (RFP) or cyan fluorescent protein (CFP) were introduced at allelic positions adjacent to the centromere of Chromosome VIII (Fig 3G) [122]. Proper segregation of Chromosome VIII produces two red and two blue spores in each tetrad, while MI non-disjunction results in two pink and two black spores (Fig 3G). No MI nondisjunction was observed out of 2075 WT tetrads (Fig 3H). Like WT, chromosome segregation was highly accurate in the hedΔR diploid (0.1%, 943 tetrads), while a small, but significant, increase in MI nondisjunction was detected for NDT80-mid (0.2%, 10,988 tetrads) ( Fig 3H). In contrast, MI nondisjunction was dramatically increased in the hedΔR NDT80-mid strain (9.4%. 5322 tetrads), consistent with the decrease in spore viability ( Fig 3H).
The hedΔR NDT80-mid disjunction and spore viability defects were due to Rad51 mediated strand invasion because disrupting Rad51's strand invasion activity in the hedΔR NDT80-mid rad51-II3A diploid restored spore viability to the same level as rad51-II3A alone, as well as significantly reducing MI nondisjunction (Fig 3E and 3H). The incomplete suppression of MI nondisjunction by rad51-II3A in hedΔR NDT80-mid may reflect a low level of strand invasion by Rad51-II3A when activated during Phase 1. The MRC therefore has a vital role in promoting proper chromosome segregation when Rad51 is constitutively activated.

The MRC promotes crossover formation in hedΔR diploids
One explanation for the large increase in MI non-disjunction in hedΔR NDT80-mid is that crossover formation is reduced. Crossovers were measured in two ways that did not require viable spores. The first method used physical analysis of crossover and noncrossover products resulting from DSB repair of the HIS4LEU2 hotspot [92,123]. The DSB site of this meiotic hotspot is flanked with XhoI restriction sites that are differentially located on each homolog, with an NgoMIV site very near the DSB site on one of the homologs ( Fig 4A) [123]. Double digestion of the genomic DNA with XhoI and NgoMIV produces unique fragments indicative of a subset of crossovers (CO2) and noncrossovers (NCO1). Meiotic timecourse analysis revealed that similar to previous results, both NCO1 and CO2 were delayed and reduced~2 fold in hedΔR (S2C and S2B Fig) [88,107]. However no further decrease in the amount of either CO2 or NCO1 was observed in the hedΔR NDT80-mid diploid when the average values were plotted for two different timecourses (S2C and S2B Fig).
Because of the high variability in spore viability exhibited by different biological replicates of hedΔR NDT80-mid, more than two replicates are necessary to determine statistical significance. End point analysis was therefore performed by comparing CO2 and NCO1 frequencies in several biological replicates at the 10 hour timepoint, by which time all of the strains had completed either MI or MII (Figs 2B and 3A). CO2 and NCO1 fragments were quantified as a percent of total DNA to allow statistical significance to be determined. With the larger sample size, a statistically significant reduction in the number of crossovers was observed when hedΔR was combined with NDT80-mid, correlating well with the spore inviability and MI nondisjunction phenotypes (Fig 4B and 4C). Noncrossovers were also reduced two-fold relative to , hedΔR NDT80-mid (NH2505 and NH2610), hedΔR NDT80-mid rad51-II3A (NH2705 RCEN) and rad51-II3A (NH2666 RCEN) were incubated for 10 hours in Spo medium. DNA was isolated, digested with XhoI and NgoMIV and crossovers and noncrossovers were analyzed by Southern blot as described in [123]. A representative gel of an individual culture from each strain is shown. P1 and P2 represent the parental homologs. NCO1 and CO2 represent unambiguous non-crossover and crossover bands. The numbers for each lane are colored coded with the indicated strains in the legend. (C) Quantification of CO2 after 10 hours in Spo medium. (D) Quantification of NCO1 after 10 hours in Spo medium. The statistical significance of differences between strains in Panels C and D was determined using an unpaired, two-tailed Student's T-test. ( � = p < .05; �� = p < .01; ��� = p < .001, ���� = p < .0001). (E) Genetic analysis of crossover formation. Single colonies from WT (NH2598 RGC) (n = 7; tetrads = 3374), hedΔR (NH2616 RGC) (n = 7; tetrads = 3812), NDT80-mid (NH2634 RGC) (n = 7; tetrads = 3592), hedΔR NDT80-mid (NH2610 RGC) (n = 8; tetrads = 3899), pif1-md (NH2695 RGC)(n = 7; tetrads scored = 2974), hedΔR pif1-md (NH2702 RGC)(n = 6; hedΔR in the hedΔR NDT80-mid diploid, but this reduction was not statistically significant, perhaps due to the limited sample size (Fig 4B and 4D).
The rad51-II3A mutant alone exhibited CO2 and NCO1 levels equivalent to that of WT as expected (Fig 4B-4D) [25]. Addition of rad51-II3A to the hedΔR NDT80-mid diploid increased NCO1 levels, but a statistically significant increase in CO2 was not observed compared to hedΔR NDT80-mid (Fig 4C and 4D). However, given the suppression of both MI non-disjunction and spore inviability phenotypes in hedΔR NDT80-mid rad51-II3A diploids, the failure to see suppression of crossover formation could be due to the limited sample size and the variability in the rate of meiotic progression observed for hedΔR NDT80-mid.
Crossover frequencies were also determined genetically using a fluorescent spore assay [122]. One chromosome VIII contained RFP adjacent to CEN8. On the homolog, a GFP � (green fluorescent protein) gene was integrated at ARG4 and CFP was integrated at THR1. Crossovers in different intervals were distinguished by distinctive fluorescent spore patterns (S3 Fig). In contrast to the HIS4LEU2 hotspot analysis, the map distance in the CEN8-ARG4 interval was higher than WT for both hedΔR and NDT80-mid ( Fig 4E). This increase was specific to the centromere adjacent region, as hedΔR and NDT80-mid crossovers in the ARG4-THR1 interval occurred at frequencies that were either equal to or lower than WT ( Fig  4E). Interestingly a similar pattern was observed for map distances in different intervals of chromosome III in hed1Δ and WT diploids [88]. In that experiment, crossover formation was higher in the CEN3-MAT interval in the hed1Δ mutant but equal to or lower than WT in intervals further out on the arms [88]. Eliminating the MRC in the Rad51-activated strain (hedΔR NDT80-mid) significantly decreased crossovers for both the CEN8-ARG4 and ARG4-THR1 intervals compared to hedΔR ( Fig 4E). These results indicate that a subset of interhomolog recombination intermediates created when Rad51 is activated in Phase 1 require the MRC to be successfully processed into crossovers.

Constitutive activation of Rad51 results in an accumulation of interhomolog and intersister joint molecules that is dependent upon the MRC
Crossovers are the end point of recombination formed by the resolution of double Holliday junction intermediates (dHJs). Previous work using hed1Δ showed a modest decrease in interhomolog (IH) dHJs with little to no effect on intersister (IS) dHJs and a reduction in interhomolog bias from 5.7 IH:IS ratio to 2.1 [88]. In our hedΔR diploid, the peak level of IH dHJs was delayed relative to WT, indicating a delay either in strand invasion or second end capture ( Fig  5A and 5B). IS-dHJs were increased in hedΔR and the IH:IS ratio bias was reduced from~8:1 in WT to 2:1 in hedΔR (Fig 5A and 5B, and S1 Data). In cells lacking the MRC (NDT80-mid), the kinetics of IH and IS dHJs was similar to WT and there was a specific reduction in IH dHJs, decreasing the IH:IS ratio to 5:1 (Fig 5A and 5B). The overall reduction in IH dHJs for tetrads = 2799) and hedΔR NDT80-mid pif1-md (NH2687 RGC)(n = 8; tetrads = 2605) were sporulated on solid medium and analyzed by fluorescence microscopy for crossovers in the CEN8-ARG4 and ARG4-THR1 intervals (S3 Fig). Error bars show standard error. Since MI nondisjunction and nonparental ditype tetrads are indistinguishable in the CEN8-ARG4 interval (S3 Fig), the frequency of nonparental ditypes was corrected using the MI nondisjunction frequency for hedΔR and NDT80-mid. No correction was necessary for WT because MI nondisjunction was not observed. All CEN8-ARG4 NPDs were assumed to be due to MI nondisjunction in strains with Chromosome VIII MI nondisjunction frequencies � 9% (Fig 3H) [122]. Map distances (cM) and standard errors were calculated using the Stahl lab online tools (https:// elizabethhousworth.com/StahlLabOnlineTools/). (F) Crossover interference ratios between the CEN8-ARG4 and ARG4-THR1 intervals were determined using the method described by Malkova, Swanson [124]. Arrows above and below the chromosome VIII diagram indicate ARG4-THR1 and CEN8-ARG4 as the reference interval, respectively. Test interval/Reference interval map distance ratios significantly less than one are indicative of interference and p values are indicated by asterisks ( � <0.05; ���� <0.0001) (G-test).
https://doi.org/10.1371/journal.pgen.1010407.g004 NDT80-mid could be caused by an inability to capture these species due to more rapid resolution of dHJs, making them more transient when transcription of CDC5 is no longer regulated by the MRC. The lack of reduced crossovers and noncrossovers at the HIS4LEU2 hotspot in NDT80-mid is consistent with this idea (Fig 4C and 4D).
Activating Rad51 in the absence of the MRC (hedΔR NDT80-mid) greatly reduced both IH and IS dHJs compared to hedΔR (Fig 5A and 5B). The IH:IS ratio for hedΔR NDT80-mid was also decreased to 1:1 (Fig 5A and 5B, and S1 Data). Although there were lower steady state levels of IS dHJs in hedΔR ndt80-mid compared to hedΔR, this amount was increased relative to NDT80-mid, potentially explaining the decrease in crossovers observed when Rad51 is activated in the absence of the MRC.

Extended time in prophase I promotes chromosome disjunction when Rad51 is constitutively activated
An unexpected finding was the amount of variability in prophase I length between cultures derived from different single colonies of the same diploid, including WT ( Fig 3D). Despite this variability, there was no correlation between spore viability and prophase I length in either WT, hedΔR, or hedΔR rad51-II3A (Fig 6A). In addition, specifically eliminating the checkpoint in WT cells using NDT80-mid did not change this lack of correlation. In contrast, prophase I length was positively correlated with spore viability and inversely correlated with chromosome VIII MI nondisjunction in hedΔR NDT80-mid cultures (Fig 6A and 6B). The fact that these correlations were lost in the hedΔR NDT80-mid rad51-II3A strain indicates that Rad51 strand invasion activity is required for this phenotype (Fig 6A and 6B). We conclude that the longer prophase I lengths observed for hedΔR are due to the creation of recombination intermediates that take longer to be productively processed into crossovers than those mediated by Dmc1. In the absence of the checkpoint, the length of prophase I is no longer controlled, and therefore cells without a sufficient number of crossovers prematurely enter MI, resulting in homolog nondisjunction.

Strand invasion is less robust when Rad51 is constitutively active
The persistence of Rad51/Dmc1 foci in strains where Rad51 is constitutively activated, as well as the need for a checkpoint-mediated delay to successfully complete meiotic recombination in hedΔR strains, suggests that strand invasion might be less robust than in WT. Previous studies have suggested that meiotic recombination frequently involves multiple rounds of strand invasion and D-loop intermediate disassembly before stable recombination intermediates or products form by strand annealing [125][126][127][128]. Less robust strand invasion, by shifting the balance away from D-loop formation and towards disassembly/annealing, would decrease the number of strand invasion events that are detectable by DNA tract synthesis before stable intermediates or products form.
To test this idea, the parental strand contributions to recombinants in a highly polymorphic test interval where recombination is initiated by a tightly-focused DSB at the URA3-tel-ARG4 hotspot ( Fig 7A) were determined using high throughput sequencing to score segregation of the polymorphic markers in 159 tetrads from an hedΔR msh2Δ diploid [127]. While crossing-fraction of the maximum IH-dHJ levels in the parallel WT culture (black; here 1.7% of total hybridization signal). (C) Quantitative analysis of effects of pif1-md in the WT, hedΔR and hedΔR NDT80-mid backgrounds on IH-HJs (left) and IS-dHJs (right; tc14). JM levels are expressed as fraction of the maximum IH-dHJ levels in the parallel WT culture (black), here 1.0% of total hybridization signal.   interhomolog strand invasion. The similarity in map distance for the natMX-hphMX interval between WT and hedΔR is consistent with previous observations that Rad51 activation does not dramatically decrease map distances ( Fig 4E) [88]. In addition, while the majority of events in WT contained heteroduplex DNA on both sides of the DSB, only a minority of events in hedΔR were two-sided ( Fig 7C and S4 Fig, and S2 Data Sheet 5). Since two-sided events can be produced by events where both DSB ends invade and prime synthesis, the reduced number of such events provides additional indication of reduced interhomolog strand invasion in hedΔR.
Meiotic recombination frequently involves template switching, where strand invasion switches between homolog and sister chromatids, thus undergoing multiple rounds of invasion, end-primed synthesis, and disassembly [126][127][128][129][130][131]. Template switching was also reduced in hedΔR relative to WT (Fig 7D and 7E, and S2 Data Sheet 7). Comparing hedΔR and WT, the average number of template switches per conversion tract (0.25 versus 0.57; p = 0.02, chisquare test) and the fraction of events displaying template switching (0.22 versus 0.51; p = 0.003, chi-square test) were substantially reduced when Rad51 was constitutively activated. In principle, template switching can be reduced either by decreasing either strand invasion or D-loop disassembly; the latter would be expected to increase the time available for end extension and thus increase the length of gene conversion segments. However, gene conversion segments in hedΔR were, if anything, shorter than in WT ( Fig 7E, and S2 Data, Sheet 6).
In summary, both overall gene conversion frequencies and the fraction of events with multiple strand invasions were reduced in hedΔR cells, consistent with the suggestion that meiotic interhomolog strand invasion is reduced when Rad51 is constitutively activated.

PIF1 helicase activity and PCNA interaction promote spore viability and chromosome disjunction when Rad51 is constitutively active
The delay in both exit from prophase I and appearance of IH dHJs observed when Rad51 is constitutively active suggests that processing of Rad51-generated intermediates is different than that of Dmc1-generated D-loops. One possible difference is the use of the Pif1 helicase which plays a key role in extension of Rad51-mediated D-loops in BIR and the formation of a fraction of vegetative crossovers but has only a minor role in Dmc1-mediated recombination [109,112]. To see if PIF1 has a function when Rad51 is constitutively activated during prophase I, a meiotic depletion allele (pif1-md) was created by placing PIF1 under the control of the CLB2 promoter [112,132]. CLB2 is expressed in vegetative cells but not during meiosis [133]. The pif1-md mutant alone exhibited~85% spore viability, a significant decrease from WT, while Chromosome VIII MI non-disjunction increased only slightly (0.4% vs 0%) (Fig 8A and  8C) [112]. Activation of Rad51 in the absence of PIF1 (hedΔR pif1-md) had no obvious effect nonmendelian segregation (NMS) /total tetrads) for each polymorphic marker. Conversion frequencies in hedΔR are significantly less than in wild type (p < 0.0001, Wilcoxon matched-pairs signed rank test). Underlying data are in S2 Data, Sheet 4 and in [127]. (C) One-sided and two-sided hybrid DNA among noncrossovers (NCO) and crossovers (CO). Schematic-in one-sided events, one DSB end invades a homolog and primes DNA synthesis before disassembly and annealing, leading to hybrid DNA to one side of the DSB; in two-sided events, both ends invade a homolog and extend, leading to hybrid DNA on both sides of the DSB. Table-number of tetrads with one-sided and two-sided events, with fraction of tetrads of each type in parentheses. Individual events are illustrated in S4 Fig. More events are one-sided in hedΔR than in WT (55% and 37%, respectively; p = 0.02, Fisher's exact test). Underlying data are in S2 Data, Sheet 5 and in [127]. (D) Template switching, expressed as fraction of non-mendelian segregation (NMS) half-tracts (interval from DSB to last converted marker) with the indicated number of template switches. hedΔR (green) has fewer template switches than WT (black; p = 0.02, chi-square test). Inset-fraction of tetrads with (filled) or without (clear) template switching (p = 0.003, chi-square test); Ccrossovers, N-noncrossovers. Underlying data are in S2 Data, Sheets 6 and 7 and in [127]. (E) Length of gene conversion segments, which represent the stretch of DNA between the DSB and the end of a gene conversion tract if no template switching occurs, or, if template switching occurs, the stretch of DNA either between the DSB and a template switch or between two template switches. Median segment lengths are indicated below the X axis; red lines indicate lower, median, and upper quartiles. p values (Mann-Whitney test, two-tailed) for comparisons between hedΔR (green) and WT (black) are at the top of the plot. Underlying data are in S2 Data, Sheet 6 and in [127].
https://doi.org/10.1371/journal.pgen.1010407.g007 on spore viability but significantly increased MI nondisjunction to~3%, indicating a synergistic effect (Fig 8A and 8C). Furthermore, depleting PIF1 in the hedΔR NDT80-mid diploid dramatically reduced the average spore viability from~60% to~25% and increased chromosome VIII MI nondisjunction three-fold compared to hedΔR NDT80-mid (Fig 8A and 8C). The effect of pif1-md on chromosome segregation was not limited to Chromosome VIII, as the MI non-disjunction pattern of the viable spores in tetrads was exacerbated in the hedΔR NDT80-mid pif1-md diploid compared to hedΔR NDT80-mid (S5B Fig). Both the spore inviability and MI non-disjunction phenotypes of hedΔR NDT80-mid pif1-md were complemented by integration of a PIF1 plasmid, confirming that depletion of PIF1 is responsible for these phenotypes (S5A and S5B Fig). In addition, these phenotypes were suppressed by rad51-II3A, demonstrating that the more extreme mutant phenotypes observed in the absence of PIF1 are due to constitutive activation of Rad51 and the absence of the MRC (Fig 8B and 8D).
https://doi.org/10.1371/journal.pgen.1010407.g008 that Pif1 helicase activity and its ability to interact with PCNA are required for processing of Rad51-mediated D-loops during meiotic prophase I.

PIF1 promotes meiotic progression and crossover formation in strains with constitutively active Rad51
While similar meiotic progression timing was observed in WT and pif1-md, depleting Pif1 from the hedΔR background significantly delayed meiotic progression compared to hedΔR alone (S6A and S6B Fig). This delay was not due to increased prophase I lifespan (S6C Fig). Even so, eliminating the MRC from the hedΔR pif1-md diploid using NDT80-mid suppressed the meiotic progression delay to the same T 50 value as hedΔR NDT80-mid and restored prophase lifespan to that of WT (S6 Fig). These results suggest activating Rad51 in the absence of PIF1 may create intermediates that are not repaired before MI which may then trigger a DNA damage checkpoint delay in MII [119].
pif1-md exhibited WT kinetics for both CO2 and NCO1 formation at the HIS4LEU2 hotspot, but the levels of both were slightly reduced, perhaps due to the reduced number of cells that entered meiosis (S2 and S6A Figs). IH dHJs were decreased, but there was no corresponding increase in IS dHJs (Fig 5C). CO2 formation was delayed in hedΔR pif1-md but reached similar levels compared to hedΔR and hedΔR NDT80-mid (S2 Fig). IS dHJs were increased in hedΔR pif1-md compared to either pif1-md or hedΔR, perhaps explaining the small increase in MI non-disjunction observed in this diploid (Figs 5C and 8C). CO2 and NCO1 levels in hedΔR NDT80-mid pif1-mid were reduced more than hedΔR NDT80-mid (S2 Fig). Crossover frequencies determined using the orthogonal fluorescent spore system were consistent with the physical analysis. The pif1-md mutant alone exhibited little to no reduction in map distance in the CEN8-ARG4 and ARG4-THR1 intervals on chromosome VIII (Fig 4E). In contrast depleting PIF1 from the hedΔR diploid significantly reduced crossovers in the CEN8-ARG4 interval, and further reductions were observed in both intervals with the hedΔR NDT80-mid pif1-md diploid (Fig 4E).
Crossover interference is a genetic phenomenon in which a crossover in one interval lowers the probability of a crossover in an adjacent interval [137]. To analyze interference occurring between crossovers in the CEN8-ARG4 and ARG4-THR1 intervals, the method of Malkova, Swanson (124) was used. A crossover interference ratio <1 indicates that interference is occurring. Within each strain the crossover interference ratios calculated using either CEN8-ARG4 or ARG4-THR1 as the reference interval were similar (Fig 4F). Crossover interference was observed in WT, hedΔR, NDT80-mid, pif1-md, and hedΔR pif1-md at levels consistent with the literature (~0.2) [122]. In contrast, interference was absent in hedΔR NDT80-mid pif1-md. One explanation for the loss of interference is that inefficient Rad51-mediated DSB repair in Phase 1 (due to the absence of both PIF1 and the MRC) results in numerous DSBs, including those that were designated to be interfering crossovers, being repaired using a sister chromatid. Loss of these potential crossover sites could therefore randomize the crossover distribution.
The dmc1-II3A mutant is specifically defective in strand exchange activity [25]. Substitution of dmc1Δ with dmc1-II3A in the hedΔR background significantly improved spore viability (S5C Fig). This observation is consistent with the idea that Rad51 and Dmc1 together contribute to making presynaptic filaments with the correct stoichiometry necessary to promote both Dmc1 and Rad51-mediated interhomolog recombination [26,27,83,116]. Depletion of PIF1 from either hed1Δ dmc1Δ or hedΔR dmc1Δ significantly decreased spore viability (S5C Fig). Since Rad51 is the only recombinase in these strains, these results confirm that PIF1 is required to processing of Rad51-generated recombination intermediates.
The PIF1 paralog, RRM3, promotes chromosome segregation when Rad51 is constitutively active during meiosis RRM3 and PIF1 have distinct functions in some processes, e.g., RRM3 is not required for BIR, and Rrm3 travels with the replisome to remove non-nucleosomal proteins while Pif1 is recruited to the replisome only in specific circumstances [109,[138][139][140][141]. However, there are also examples where PIF1 and RRM3 appear to function redundantly [142,143]. A role for RRM3 in meiosis was therefore examined. No effect on spore viability was observed in rrm3Δ or NDT80-mid rrm3Δ cells (Fig 8A). In contrast, a small, but significant, decrease in spore viability was observed in hedΔR rrm3Δ (Fig 8A). In addition, there was a corresponding increase in Chromosome VIII MI non-disjunction (Fig 8C). These phenotypes were exacerbated when the meiotic recombination checkpoint was removed (hedΔR NDT80-mid rrm3Δ) (Fig 8A and  8C). Both the spore inviability and MI nondisjunction phenotypes were complemented by addition of RRM3 on a plasmid (S5A and S5D Fig). The spore viability defect was not the result of mutations that accumulated during vegetative growth in the absence of RRM3, since the spore inviability phenotype of the hedΔR NDT80-mid rrm3Δ was complemented by expressing RRM3 specifically during meiosis using the REC8 promoter (S5A Fig).
If PIF1 and RRM3 function redundantly in processing Rad51 generated intermediates during meiosis, a synergistic defect in spore viability when the two mutants were combined is expected. Alternatively, if they function independently of each other, the predicted spore viability of pif1-md rrm3Δ in various backgrounds would be the product of their individual spore viabilities. A diploid containing pif1-md and rrm3Δ exhibited the same level of viable spores as pif1-md alone, consistent with the lack of any phenotype for rrm3Δ in an otherwise WT diploid (Fig 8A). In contrast, in the hedΔR and hedΔR NDT80-mid backgrounds, the average spore viabilities for the pif1-md rrm3Δ containing diploids were close to the product of the individual values (hedΔR pif1-md x hedΔR rrm3Δ, expected: 82% � 80% = 66%; observed for hedΔR pif1-md rrm3Δ 59%; hedΔR NDT80-mid, expected: 25% � 47% = 12%; hedΔR NDT80-mid pif1-md rrm3Δ observed 16%). We conclude that PIF1 and RRM3 play different roles in processing Rad51-mediated intermediates.

The meiotic recombination checkpoint promotes proper chromosome segregation and spore viability in WT and hedΔR diploids
It has often been suggested that the MRC functions to delay prophase I exit to allow time for DSB repair to occur. One difficulty in proving this idea, however, has been the pleiotropic nature of the Mek1 kinase. MEK1 is required not only to regulate Ndt80 and therefore prophase I exit, but also for interhomolog bias [75]. Inactivation of Mek1, through mutation or kinase inhibition, results in rapid repair of DSBs via the sister chromatids, thereby removing the signal to the checkpoint and allowing efficient meiotic progression [77][78][79]. In addition, mutants that disrupt the MRC such as RAD24, RAD17, and MEC1/TEL1 all play a role in Mek1 activation and therefore their effects could be indirect due to affecting other MEK1 processes as well [92,[144][145][146]. Artificially prolonging the length of prophase I using an inducible allele of NDT80 was found to improve the spore viability of mec1 and rad24 mutants, and this condition was suggested to mimic the prophase I delay caused by the meiotic recombination checkpoint [147]. A problem with this interpretation, however, is that cells behave aberrantly during the extended prophase I arrest that occurs when NDT80 is inactive. For example, in a mek1Δ ndt80Δ diploid, intersister joint molecules predominate early in prophase I but this bias is switched to interhomolog JMs with increasing time in the arrested cells [148]. Consistent with this result, mek1Δ spore inviability can be rescued by allowing increased time in prophase I prior to NDT80 induction [94]. Our work instead utilized the NDT80-mid mutant, which abolishes the MRC without affecting Mek1 kinase activity, to examine the role the MRC plays in WT cells, as well as in the hedΔR mutant that exhibits a modest defect in DSB repair where meiotic progression is delayed but not arrested.
Prophase I length in the NDT80-mid mutant was not significantly shorter than WT. However, spore viability was decreased by a small, but significant amount, as were crossovers in the ARG4-THR1 interval, and MI non-disjunction of Chromosome VIII was slightly elevated. These results indicate that there are DSB repair problems in a small number of otherwise WT cells that require time provided by the MRC to resolve.
Unexpectedly, cultures derived from different single colonies from a WT diploid displayed variability in the length of prophase I ranging from 2-5 hours (Fig 6A). However, high levels of viable spores were observed in all cultures, whether they went fast or slow, suggesting that some cultures completed recombination earlier and exited prophase I, while others needed more time for DSB repair. Variability was also observed in hedΔR, although in this case prophase I length was longer, ranging from 3-6 hours. Nevertheless, spore viability remained high (Fig 6A). In contrast, there was a striking, statistically significant, positive correlation between prophase I lifespan and spore viability and a negative correlation between prophase I lifespan and MI non-disjunction when the meiotic recombination checkpoint was abolished in hedΔR NDT80-mid (Fig 6A and 6B). These correlations were eliminated in hedΔR NDT80-mid rad51-II3A consistent with the fact that Rad51 strand exchange activity was responsible for the reduced spore viability. These experiments demonstrate that a key function of the MRC is to provide time for processing recombination intermediates during meiotic DSB repair.

Interaction between Rad51 and Rad54 in Phase I allows Rad51-mediated interhomolog strand invasion in the presence of Dmc1
Two major differences between mitotic and meiotic recombination in yeast are that during meiosis, the presynaptic filament contains patches of both Rad51 and Dmc1 and strand invasion is biased toward the homolog and not sister chromatids. The interplay between these recombinases is complex, however: while the presence of the Rad51 protein is required for proper Dmc1 loading and interhomolog bias, Rad51 strand exchange activity is not required for interhomolog recombination and is actively suppressed [25,30,84,86,87]. In addition, although deletion of DMC1 abolishes interhomolog strand invasion, activation of Rad51 in a dmc1Δ mutant results in significantly higher levels of interhomolog recombination than mek1Δ, indicating that activated Rad51 can generate interhomolog intermediates when Mek1 is active [88,89]. The modest phenotypes exhibited by hed1Δ and hedΔR diploids have been attributed to the presence of Dmc1 itself contributing to the inhibition of Rad51 strand exchange activity [88,89]. However, the idea that activating Rad51 should decrease interhomolog recombination in the presence of Dmc1 assumes that Rad51 would preferentially direct strand invasion to sister chromatids, as it does in vegetative cells. This work shows that Rad51 strand exchange activity is not fully suppressed by Dmc1 in hed1Δ or hedΔR yeast mutants. Instead, the near WT levels of crossovers and spore viability exhibited by hed1Δ and hedΔR diploids are due to a combination of different factors that promote the creation of Rad51-mediated interhomolog crossovers during Phase 1.
The fact that specific elimination of the MRC in a hedΔR diploid decreased interhomolog events, increased MI non-disjunction and decreased spore viability indicates that Rad51 is generating intermediates that the MRC is providing time to process. Proof of this idea is that all these mutant phenotypes were suppressed by preventing Rad51 exchange activity using the rad51-II3A mutant. Therefore, in these cells, Rad51 is active and generating crossovers during Phase 1 when interhomolog bias is mediated by Mek1. The high levels of crossovers and spore viability observed in hed1Δ diploids compared to hed1Δ dmc1Δ likely result from the presence of Dmc1 helping the presynaptic filament search for homology on homologs rather than sister chromatids through accessory factors like Hop2/ Mnd1 [149].
There are some organisms such as the nematode, Caenorhabditis elegans, the fruit fly, Drosophila melanogaster, and the fungus, Sordaria macrospora, that use only Rad51 for meiotic recombination [3,150]. This work raises the interesting question of whether Rad51 mediated meiotic recombination in these organisms occurs similarly to the pathway we have discovered in yeast. This idea could be tested by determining whether Pif1 orthologs are required for meiotic recombination in these species.

A model for Rad51-mediated recombination during meiosis
We propose that processing of nascent meiotic D-loops generated by Rad51 is different from D-loops made by Dmc1 for several reasons. First, activation of Rad51 creates a delay in prophase I exit, indicating that the cell requires more time to satisfy the MRC when there is Rad51-strand exchange activity than when Dmc1 is the sole functioning recombinase. Second, the MRC plays only a minor role in promoting chromosome disjunction and spore viability in WT cells when interhomolog recombination is mediated by Dmc1 but is very important when Rad51 also is active during Phase 1. Third, PIF1 activity is normally inhibited such that its contribution to Dmc1-mediated repair is negligible, but promotes repair of Rad51-generated intermediates [112].
The existing model for Dmc1-mediated recombination is based on the assumption that, similar to mice, Rad51 and Dmc1 form adjacent patches in the presynaptic filament with Dmc1 closest to the 3' end where it is responsible for strand invasion (Fig 9 WT) [19,22]. After the homology search results in formation of a nascent D-loop (Fig 9B), Dmc1 is removed from the 3' end by a DNA translocase to form heteroduplex between the invading strand and the complementary donor strand, resulting in the displacement of the strand of like polarity to make a mature D-loop (Fig 9C). It is unclear what DNA translocase is responsible for this step. While Rdh54 seems the obvious choice given in vivo and in vitro evidence that it preferentially works with Dmc1, this activity has not yet been demonstrated for Rdh54 [32,151]. The 3' end of the invading strand then acts as a primer for DNA synthesis using PCNA and Polδ to extend the invading strand (Fig 9D) [52][53][54][55]. Disassembly of the D-loop followed by annealing to the other side of the break creates noncrossovers, while second end capture of the extended end generates double Holliday junctions that can be resolved as crossovers (Fig 9E). Note that our model shows a simplified view of the recombination process. Recent work has demonstrated that meiotic DSB repair is very dynamic, with multiple rounds of strand invasion and disassembly, and template switching between homologs and sister chromatids [126][127][128]. Fig 9 (hedΔR) illustrates a mechanistic model for DSB repair mediated by Rad51 when it is artificially activated during Phase 1. Rad54 binds to gaps between patches of Rad51 on ssDNA in vitro [118,152]. We propose that Rad54 similarly binds in gaps present between Rad51 and Dmc1 on the presynaptic filament (Fig 9F). The resulting presynaptic filaments then form nascent D-loops similarly to WT (Fig 9B and 9G). However, in a subset of DSBs, Rad54 sequentially removes Rad51 monomers, thereby creating heteroduplex DNA between the invading strand and donor strand of opposite polarity to generate an internal D-loop ( Fig  9H). The formation of internal D-loops by Rad51-Rad54 has previously been demonstrated in vitro [37].
Formation of an internal D-loop by Rad51 creates a problem in that the 3' end remains bound by Dmc1 and therefore is prevented from serving as a primer for DNA synthesis by DNA Polδ (Fig 9H). For a D-loop to be extended, DNA polymerase must be able to access the 3' hydroxyl to initiate DNA synthesis [50]. For this 3' hydroxyl to be available to DNA polymerase, the nascent D-loop created by Dmc1 must be converted to a mature D-loop by a translocase removing Dmc1 from the presynaptic filament and creating heteroduplex DNA. In addition, the absence of a free 3' end may make this intermediate more susceptible to disassembly by the STR complex which would explain why there are fewer two-sided events and template switches in hedΔR. Having to remove the Dmc1 to free up the 3' end ( Fig 9I) is an extra step that would delay strand extension, thereby taking longer for second end capture to occur (Fig 9J) which could explain the delay in dHJ formation observed for hedΔR. In addition, the increased time it takes to convert the break into an intermediate that is not recognized by the MRC would result in an MRC-dependent prophase I delay.
Pif1 is known to move in the 5'-3' direction to remove proteins from DNA, so its function in processing Rad51-generated intermediates could be the removal of Dmc1 after internal Dloop formation. However, this activity was not detected for Pif1 using in vitro experiments with DNA curtains comprised of Dmc1-bound ssDNA, making this possibility less likely (Eric Greene, personal communication). Another possibility is that the end containing Dmc1 is cleaved off to create a free 3' end located upstream of the DSB site. Extension of this end is predicted to produce 3:1 gene conversions around the DSB site which were not observed. The simplest explanation is that Dmc1 is removed by the same mechanism shown in Fig 9C. Pif1 activity is normally inhibited during Dmc1-mediated recombination by Mer3 and MutLβ [112]. It is possible that these proteins do not recognize a Rad51-generated internal Dloop and therefore are not present to inhibit Pif1, allowing Pif1 to promote synthesis of the invading strand by binding to PCNA and unwinding the DNA (Fig 9J). The resulting extended strand could then be processed in a similar way used for Dmc1-generated D-loops to make crossovers and noncrossovers (Fig 9K). A key test of the model will be to see if conversion track length is decreased in a hedΔR pif1-md diploid.

Is downregulation of Rad51 activity conserved in mammals?
No ortholog of Mek1 has been reported in mammals, raising the question of whether Rad51 activity is inhibited during prophase I in mammalian cells. Several observations suggest that the answer is yes. First of all, repair of exogenous DNA damage in mouse oocytes exhibits two temporally distinct processes during pachytene-an early one that occurs in the presence of both Dmc1 and Rad51 and a later stage where Rad51, but not Dmc1, is associated with the DSBs [106]. In addition, depletion of Rad51 specifically in mouse spermatocytes results in apoptosis early in prophase I [153]. This phenotype is very different from the zygotene arrest exhibited by Dmc1 depletion or Dmc1 -/ spermatocytes [14,153,154]. These results show that (1) the presence of Rad51 is required for Dmc1-mediated recombination and (2) that Rad51 is not mediating strand invasion when Dmc1 is absent during Phase I, suggesting that its strand exchange is being inhibited similar to yeast. The mechanism of this inhibition is unclear, but it requires a functional chromosome axis as mutants in genes encoding the mouse axial element proteins Sycp2 and Sycp3 allow Rad54 +/+ -dependent intersister repair of DSBs in Dmc1 -/oocytes [155].
In summary, it has long been established that the presence of Rad51 on the presynaptic filament is necessary for Dmc1 to efficiently mediate interhomolog recombination in both yeast and plants [26,30,83,116]. However, yeast cells go to great lengths to prevent Rad51 from engaging in strand invasion during Phase 1 of meiotic recombination. Previous studies in which Rad51 was constitutively activated during meiosis revealed little to no phenotypes with regard to spore viability and interhomolog bias [88,89,107]. This work shows that when Rad51 is allowed to interact with Rad54, Rad51 will compete with Dmc1 to mediate interhomolog strand invasion. However, processing of these intermediates involves additional constraints such as time provided by the MRC and factors such as Pif1, making Rad51-mediated Phase 1 recombination less desirable compared to Dmc1. Once sufficient DSB repair has occurred to inactivate the MRC, Rad51 is naturally activated to repair residual DSBs using sister chromatids. Since this repair also begins with filaments that contain both Rad51 and Dmc1, it will be interesting to see if PIF1 promotes Rad51-mediated repair in Phase 2 as well. Pif1 is conserved in mammals where it is also required for BIR [156]. It would be interesting to see if Pif1 has a role in the repair of DSBs in Dmc1 -/-Sycp2 -/mouse oocytes.

Strains
All strains were derived from the SK1 background and their genotypes are listed in S1 Table. Liquid and solid media are described in [157]. Polymerase chain reaction (PCR) mediated gene deletions were generated using the drug resistance markers kanMX6, natMX4 and hphMX4 which confer resistance to the antibiotics G418, nourseothricin (NAT) and Hygromicin B (HygB), respectively [158,159]. Deletions were confirmed by PCR using a forward primer upstream of the gene's open reading frame (ORF) and a reverse primer located within the drug marker, while the absence of the WT allele was assayed using the same forward primer and a reverse primer located within the open reading frame (ORF). The pif1-md allele was constructed by fusing the WT PIF1 ORF to the CLB2 promoter. A kanMX6-P CLB2 -3HA PCR fragment with 50 bp homology from -100 to -50 from the PIF1 ATG on the upstream end and 50 bp homology from +1 to +50 from the ATG on the downstream end was amplified using pMJ787 (also called 455) [132] and transformed into haploid yeast strains selecting for G418 resistance. The gene fusion was confirmed by colony PCR using a forward primer in the CLB2 promoter and a reverse primer in the PIF1 ORF.
The NDT80-mid, RAD54-T132A and rad51-II3A alleles were introduced into haploid strains by two-step gene replacement [160]. The NDT80-mid URA3 integrating plasmid pNH317 was digested with SnaBI to target integration of the plasmid upstream of ndt80Δ:: kanMX6. Ura + transformants were grown in YPD and then plated on 5-fluoro-orotic acid (5-FOA) to select for cells that had lost the URA3 plasmid [161]. Foa R colonies that retained the NDT80-mid allele were identified by loss of the ndt80Δ::kanMX6 marker, rendering the cells sensitive to G418.
The RAD54-T132A URA3 integrating plasmid, pHN104, was digested with BsiWI to target integration upstream of RAD54. Ura + transformants were grown in YPD and then plated on 5-FOA. A fragment of the RAD54 gene from each Foa R colony was amplified by PCR and sequenced to detect the presence of the T132A mutation. With the exception of the PacBio sequencing used for the DSB hotspot analysis (see below), all DNA sequencing was performed by the Stony Brook University DNA Sequencing Facility.
The rad51-II3A URA3 integrating plasmid, pAM1-II3A, was digested with SpeI to target integration upstream of RAD51. Ura + transformants were grown in YPD then plated on 5-FOA. Foa R colonies were patched onto YPD and replica plated to YPD+0.04% methyl methanesulfonate (MMS) to screen for MMS sensitivity. The rad51-II3A allele contains three mutations, R188A, K361A and K371A. To confirm that the Mms S Foa R cells contained all three mutations, the RAD51 gene was amplified from genomic DNA and sequenced.
All of the strains used for fluorescent spore analysis are isogenic with NH716, which is created by crossing the NHY1210 and NHY1215 haploids. To allow integration of the appropriate plasmids, the LEU2 and TRP1 genes were deleted from NHY1210 and NHY1215 (and their mutant derivatives), respectively. All but the first and last 100 bp of the LEU2 ORF ectopically located in the NHY1210 haploid were replaced with kanMXM6 (designated leu2ΔI::kanMX6). Additionally, the 5' end of the TRP1 gene was replaced with hphMX4 in the NHY1215 haploid [80].
To measure the frequency of chromosome VIII nondisjunction [122], the plasmids, pSK693 (pYKL050c-RFP LEU2) and pSK694 (pYKL050c-CFP TRP1) were digested with AflII to target integration to the right arm of chromosome VIII adjacent to the centromere in NHY1210 and NHY1215 (and their mutant derivatives), respectively. Diploids with the RFP and CFP genes in allelic positions are indicated by "RCEN". To measure genetic map distances between CEN8-ARG4 and ARG4-THR1, pSK729 (pYKL050c-GFP � URA3) and pSK695 (pYKL050c-CFP TRP1) were digested with AflII and sequentially integrated into NHY1215 haploids at ARG4 and THR1, respectively. These haploids were crossed to the appropriate NHY1210 CEN8::RFP haploids to create diploids containing the marker configuration indicated by "RGC".
The hedΔR msh2Δ haploids used for the DSB hotspot experiment were created in several steps. First the HED1 gene was deleted from S5495 and S5497 using natMX4. To inactivate natMX4 so this gene could be used in a later step, the hed1Δ::natMX4 haploids were then transformed with pSZ5, a 2μ plasmid containing LEU2 and CAS9 fused to 5'TTCGTCGCGTACGGGGACGA3', a guide DNA sequence that, once transcribed, targets DSB formation within natMX4. Because the broken chromosome is lethal, there is a strong selection for cells in which error-prone non-homologous end joining has occurred that destroys the Cas9 cut site and disrupts the natMX4 ORF. Leu + transformants were patched onto SD-leu plates and then replica-plated to YPD + NAT to look for patches that failed to grow. Nat S (NS) patches were streaked for single colonies on YPD that were then replica plated to SD-leu to find cells that had lost pSZ5. The RAD54-T132A allele was introduced into S5495 hed1NS and S5497 hed1NS as described above. The resulting strains were designated S5495 hΔR and S5497 hΔR. The hed1Δ::NS RAD54-T132A mutants were then introduced into haploid strains containing msh2Δ and the URA3-tel-ARG4 hotspot with or without SNPs. First, S5495 hΔR was crossed to S4955 to make NH2694. This diploid was heterozygous for hed1Δ::NS, RAD54-T132A, as well as leu2-R, his4Δ::kanMX, his4::URA3rev-tel1-ARG4 +SNPs-RNQ-hphMX-FUS1 and msh2Δ:kanMX6. NH2694 was sporulated and 100 tetrads dissected. Strains that were Ura + , Arg + , HygB R and G418 R were then screened for mating type and MATa strains were selected. (The his4Δ::kanMX gene is on chromosome III, as is his4::URA3rev-tel1-ARG4 +SNPs-RNQ-hphMX-FUS1, so it should not be present in this group). The presence of msh2Δ::kanMX6 and hed1Δ::NS was confirmed by PCR, while spore colonies containing RAD54-T132A were detected by DNA sequencing. The resulting haploid, NH2694-a2 was frozen down. To construct the other parent, S5497 hΔR was crossed to S5085 to generate NH2685. This diploid was heterozygous for hed1Δ::NS RAD54-T132A, as well as his4Δ::kanMX, STE50-natMX-RRP7-his4::URA3rev-tel1-ARG4 and msh2Δ: kanMX6. 100 tetrads from NH2685 were dissected and the spore colonies screened for Nat R Ura + Arg + and G418 R . From this group, MATα strains were selected and tested for msh2Δ:: kanMX6 and hed1Δ::NS by PCR. The presence of RAD54-T132A was determined by DNA sequencing. The haploid NH2695-f8 was frozen down.
Prior to transformation, plasmids containing PIF1 in the URA3 integrating vector, pRS306, were digested with BmgBI to target integration downstream of the P CLB2 -PIF1 (pif1-md) allele.
Similarly, URA3 integrating plasmids containing various alleles of RRM3 were digested with SnaBI to target integration downstream of rrm3Δ. The presence of the introduced allele was determined by PCR using a locus-specific primer and a gene-specific primer. For the vector control, pRS306 was digested with NsiI and integrated downstream of ura3.

Plasmids
The genotypes of the plasmids used in this study are listed in S2 Table. The pAM1-II3A plasmid containing the rad51-II3A (rad51-R188A-K361A-K371A) allele from Cloud, Chan (25) was created by PCR site-directed mutagenesis. First a 3.7 kb BamHI fragment containing RAD51 was cloned into the YIp5 plasmid backbone digested with BamHI to create pAM1. Site-directed mutagenesis was used on pAM1 to subsequently introduce R188A, then K621A and finally K371A to create pAM1-II3A. All mutations introduced into pAM1-II3A were confirmed by DNA sequencing.
The pNH257 plasmid contains the REC8 promoter in a 266 bp fragment located immediately upstream of the REC8 ORF. This plasmid was created by amplifying this fragment from a plasmid containing REC8 with Not1 and BamHI restriction sites engineered on the ends. The BamHI site is located immediately upstream of where the REC8 start codon would be. This fragment was cloned into NotI/BamHI digested pRS306.
All of the plasmids created by Gibson assembly used the NEBuilder HiFi DNA Assembly kit (New England Biolabs, #E2621). To construct the URA3 RRM3 plasmid, pBG22, a 3.5 kb fragment containing the RRM3 ORF and 900 bp 5' flanking sequence and 400 bp of 3' flanking sequence was amplified from pBG18 and cloned by DNA assembly into pRS306 digested with EcoRI and ClaI. pJW7 was constructed by cloning a fragment containing the RRM3 ORF and 378 bp of downstream sequence 18 bp downstream of the REC8 promoter in pNH257 vector digested with EcoRI and ClaI using the Gibson Assembly method. pJW5 was constructed by first amplifying the entire PIF1 gene with 904 bp and 401 bp of upstream and downstream flanking sequences, respectively, from SK1 genomic DNA. This fragment was cloned into pRS306 digested with EcoRI and XhoI by Gibson Assembly. Mutant derivatives of pJW5 were also generated by Gibson Assembly by introducing two fragments that partitioned the insert employed for the pJW5 construction: one stretching from 904 bp upstream of the PIF1 open reading frame (ORF) to the mutation and another stretching from the mutation to 401 bp downstream of the PIF1 ORF, into pRS306 digested with EcoRI and XhoI. pJW14 (pif1-m1-3FLAG) was constructed by Gibson Assembly cloning of three fragments into EcoRI-HindIII-digested pRS306: fragment 1 stretches from 904 bp upstream of the PIF1 ORF to the pif1-m1 mutation (M1A). Fragment 2 stretches from the pif1-m1 mutation to the 3xFLAG tag sequence and fragment 3 stretches from the 3xFLAG tag sequence to 401 bp downstream of the PIF1 ORF. Mutant derivatives of pJW14 were constructed by a similar workflow to pJW14 except that Fragment 2 was amplified from derivatives of pJW5 containing either the K264A (AAA to GCG) or R3E (I817R, M820R, L821R, R823E). pJW11 (URA3 PIF1-3FLAG) was constructed by amplifying two fragments, one of which contains the PIF1 gene and the other the 3FLAG sequence and cloning them by Gibson assembly into pRS306 digested with EcoRI and XhoI. All plasmid constructions were confirmed by DNA sequencing at the Stony Brook University DNA sequencing facility.

Immunoblots and antibodies
Protein extracts were generated using the trichloroacetic acid method described in [162]. The Rad51 antibody was generated by Covance Research Products (now Labcorp) in guinea pigs using the peptide QKDGNDFKFNYQGDEC. The specificity of the antibody for immunofluorescence of chromosome spreads was demonstrated by the lack of foci observed at the 4.5 hr meiotic timepoint in a rad51Δ diploid, compared to numerous foci that were present at the same timepoint in WT (S1A Fig). For the Pif1-m1-3FLAG immunoblots, mouse α-FLAG antibodies (Sigma F-1804) were used at a 1:5000 dilution at room temperature for two hours. The secondary antibody was goat α-mouse (Invitrogen, #PI32430), used at 1:10,000 dilution for one hour at room temperature. Arp7 protein was used a loading control and detected using α-Arp7 primary antibodies (Santa Cruz Biotechnology, #SC8961) with mouse α-goat secondary antibodies (Santa Cruz Biotechnology, #SC2354) used at 1:10,000 dilution at room temperature for 1 hr. Proteins were detected by a chemiluminescent assay using the Advansta Westernbright ECL kit from Fisher (#K-12045-D20).

Measuring meiotic progression and whole cell immunofluorescence
1.5 ml of 4.5x10 7 sporulating cells from meiotic timecourses were fixed overnight at 4˚C with a final concentration of 3.7% formaldehyde at various timepoints. The next day, fixed cells were washed once with 1 ml phosphate buffered saline (PBS) and stored in 1.5 ml PBS at 4˚C. For analysis of meiotic progression using DAPI, 30 μl of fixed cells were placed onto each well of coated slides (Carlson Scientific cat# 101206). Cells were allowed to settle onto the slides for 15 min at room temperature (RT). After incubation, the remaining liquid was aspirated away, and slides were air dried. 3 μl of DAPI plus mounting medium (Vectashield H-1200) was added to each well and the slide was covered with a coverslip and sealed with nail polish. DAPI stained cells were observed using a Zeiss Axio Imager.Z2 microscope with a Zeiss Plan-Apochromat 63X objective. 200 cells were counted at each timepoint and categorized as either mononucleate, binucleate (meiosis I) or tetranucleate (meiosis II). T 50 values, or the amount of time in hr for a particular sporulation culture to reach ½ its maximum %MI+MII value, were determined graphically for each culture.
For whole cell immunofluorescence analysis, coated slides containing 12 wells (Carlson Scientific cat# 101206) were treated by pipetting 30 μl 0.01% poly-L-lysine (Sigma Cat# p4707) into each well and incubating at RT for 15 min. Residual poly-L-lysine was removed from the wells using an aspirator. The slide was then washed twice by pipetting 30 μl of water into each well, followed by aspiration.
Permeabilized cells were resuspended in 40 μl sorbitol/MES buffer. 30 μl of resuspended cells was added onto each well of a printed microscope slide. After allowing the cells to settle onto the slide for 15 min, the remaining liquid was aspirated away. The slides were placed into a coplin jar with methanol at -20˚C for 6 min and then transferred to a coplin jar with acetone at -20˚C for 30 seconds. Afterwards, slides were air dried until all the residual acetone evaporated. 30 μl of 1% Bovine serum albumin (BSA) in Tris buffered saline (TBS) (140 mM NaCl, 2.7 mM KCl, 25 mM Tris pH8.0) was added to each well and incubated for 30 min at room temperature in a slide box with moist paper towels. After incubation, slides were washed in TBS buffer for 30 sec. Excess TBS was removed by leaning the slides on an edge with a paper towel. The rabbit α-Red1 polyclonal antibody (rabbit 16440) (Wan et al. 2004) was diluted 1:300 in 1% BSA. 80 μl of primary antibody dilution was added to the entire slide, then a 25x50 mm coverslip was placed onto the slide. Primary antibodies were incubated overnight (~14-18 hrs) at 4˚C in slide box with moist paper towels. The next day, coverslips were removed in TBS and slides washed 2 X in TBS for 10 min. The goat α-rabbit Alexa 488 antibody (Fisher Cat# A11008) was diluted 1:1000 in 1% BSA. 80 μl of secondary antibody dilution was added to each slide and covered with a coverslip. Secondary antibodies were incubated for 2 hr at 4˚C in slide box with moist paper towels. After the secondary antibody incubation, coverslips were removed in TBS and slides were washed 2X in TBS for 10 min. Slides were completely air dried in the dark and 3 μL DAPI plus mounting medium was added to each well. A coverslip was added to the slide and sealed with nail polish. Microscopy images were taken using a Zeiss Axio Imager.Z2 microscope with a Zeiss Plan-Apochromat 63X objective.
To analyze the length of prophase I in different timecourses, 200 cells were counted at each timepoint and classified as either being Red1 + or Red1 -. The %-Red1 + cells were graphed over time and the area under the curve (AUC) was calculated using GraphPad Prism software. The AUC for each timecourse was divided by the maximum %MI+MII value to normalize for the number of cells that entered meiosis for that timecourse. This calculation yields the "prophase I lifespan" (hr) and is analogous to the method previously used to calculate the length of premeiotic S phase [113].

Nuclear spreads
Nuclear spreads were prepared as described in Grubb, Brown [163]. For Rad51 and Dmc1 costaining immunofluorescence, polyclonal guinea pig α-Rad51 antibody serum and polyclonal goat α-Dmc1 antibody serum (a generous gift from D. Bishop, U. of Chicago) were diluted together 1:100 or 1:200 respectively into 1% BSA. 80 μl of diluted primary antibodies were added to each slide and covered with a 25 x 50 mm coverslip. Slides were incubated in a moist slide box at 4˚C overnight (~14-18 hrs). The next day, coverslips were removed in TBS and slides washed 2 X in TBS for 10 min. Donkey α-guinea pig Alexa 488 (Jackson ImmunoResearch 706-545-148) and donkey α-goat Alexa 594 (Fisher Cat#A11058) secondary antibodies were diluted 1:800 and 1:1000 respectively into 1% BSA. 80 μl of diluted secondary antibodies were added to each slide and covered with a coverslip. Slides were incubated in a moist slide box at 4˚C for 2 hrs. Afterwards, slides were washed 2X in 1X TBS buffer for 10 min. Slides were air dried in the dark for 1 hr before 30 μL of DAPI was added to the slide. The slide was covered with a coverslip and sealed with nail polish. For nuclear spreads, microscopy images were taken using a Zeiss Axio Imager.Z2 microscope with a Zeiss Plan-Apochromat 100X objective.

Fluorescent spore assays
Images of the fluorescent spores in tetrads used to score chromosome VIII nondisjunction and crossovers were taken using a Zeiss Axio Imager.Z2 microscope with a Zeiss Plan-Apochromat 63X objective. Cells were sporulated on solid Spo at 30˚C for at least 3 days or in liquid Spo at 30˚C for at least 24 hours before scoring. The genetic configurations of fluorescent markers for each system and patterns of fluorescent spore predicted are shown in Fig 3G and  S3 Fig. In the RGC system, parental ditypes (PD), tetratypes (T), and nonparental ditypes (NPD) can be distinguished between the CEN8-ARG4 and ARG4-THR1 intervals based on resulting fluorescent spore patterns (S3 Fig). However, an NPD in the CEN8-ARG4 interval results in the same fluorescent spore pattern as Chr VIII nondisjunction (S3 Fig), making these two events indistinguishable. Therefore, the number of NPD tetrads in the CEN8-ARG4 interval was estimated using the frequency of chromosome VIII nondisjunction observed in the RCEN system. This MI nondisjunction frequency was multiplied by the total number of RGC tetrads and that number was subtracted from the white:white;black;black tetrads to get the NPDs. For WT, no correction was done because no Chr VIII nondisjunction was observed in this strain. For pif1-md, hedΔR and NDT80-mid, the number of CEN8-ARG4 NPD tetrads that could be explained by the frequency of chr VIII nondisjunction in the RCEN system was subtracted during map distance calculations. For hedΔR NDT80-mid, hedΔR pif1-md and hed1Δ NDT80-mid pfi1-md, all of CEN8-ARG4 non-parental ditype tetrads were assumed to be due to nondisjunction [122]. Genetic map distance was calculated using the Perkins equation [164] using the Stahl lab online tool (https://elizabethhousworth.com/StahlLabOnlineTools/). Genetic interference ratios between CEN8-ARG4 and ARG4-THR1 intervals were calculated as in Malkova, Swanson [124] using the Stahl lab online tool. One interval was designated as the reference and tetrads from this interval were divided into two groups: one with parental ditypes (PDs) that have no crossovers and the other with tetratypes (Ts) or nonparental ditypes (NPDs) that contain at least one crossover. Map distances in the adjacent test interval were then calculated using tetrads from the PD or T+NPD groups from the reference interval. The map distance in the test interval derived from the T+NPD group was divided by that obtained from the PD group to give the crossover interference ratio. If crossovers in the reference interval interfered with crossovers in the test interval, this ratio should be less than one, while the absence of interference gives a ratio equal to one. Significance of crossover interference between the CEN8-ARG4 and ARG4-THR1 intervals was done using a G-test of independence.

Physical Analysis of crossovers, noncrossovers and dHJs
To analyze crossover and noncrossover formation at the HIS4-LEU2 DSB hotspot in meiotic timecourses [123]. Genomic DNA was isolated from sporulation cultures using MasterPure Yeast DNA Purification kit (Lucigen MPY80200). Processing of genomic DNA samples and Southern blot analysis used to visualize crossovers and noncrossovers at the HIS4LEU2 hotspot were done as in [165]. For detection of joint molecules including SEIs, IH-dHJs and IS-dHJs, cells from meiotic time courses were subjected to psoralen-crosslinking, followed by extraction of genomic DNA, restriction digest with XhoI and 2D gel Southern blot analysis as previously described in Sandhu, Monge Neria (128).

URA3-tel-ARG4 hotspot analysis
To prevent mutations from accumulating during vegetative growth due to a lack of mismatch repair in the NH2700 hedΔR msh2Δ diploid, the haploid parents were first streaked out from freezer stocks onto YPDcom and allowed to grow for 51 hours at 30˚C. Five single colonies from each haploid were combined in 200 μL YPD in a microfuge tube and the cells thoroughly mixed by vortexing. The entire volume was pipetted onto a YPD plate and the cells incubated at 30˚C for 7 hours to allow mating. Cells were scraped from the plate, patched onto a plate containing sporulation medium (1% potassium acetate, 2% agar and 0.0004% histidine) and incubated for two days at 30˚C. The resulting tetrads were dissected onto YPDcom.
Spore colonies from tetrads that exhibited four viable spores and 2:2 segregation for hphMX and natMX4 were inoculated into 15 ml glass test tubes containing 2 ml YPD and incubated overnight at 30˚C with shaking. Care was taken to ensure that the entire spore colony was used for each inoculation. For storage, 0.4 ml from each culture was transferred to a well in a sterile 1 ml 96 well plate containing 0.4 ml 50% glycerol and frozen at -80˚C. In addition, 0.1 ml cell culture was transferred to the corresponding positions in a 0.45 ml 96 well plate with pointy bottoms (Fisher Scientific, #249946). Cells were pelleted in a Beckman tabletop centrifuge at 3000 rpm for 10 minutes and the supernatants were removed by aspiration. Each plate was sealed using a Microseal "F" PCR plate seal (Bio-Rad, #MSF1001) and stored at -80˚C. This process was repeated to generate nine plates containing spore colonies from a total of 216 tetrads (9 � 24 tetrad/plate).
DNA was isolated from the 0.1 ml cell pellets as follows. Two 96 well plates were processed at a time. First, 5.5 ml Buffer 1 (100 ml 1 M sorbitol, 0.1 M EDTA) was placed in a 100 ml sterile reservoir (Stellar Scientific, #1930-1030) to which 1430 μL 10 mg/ml Zymolyase 100T (100 units/mg)(US Biological, #Z1004) was added. To spheroplast the cells, each cell pellet was resuspended in 20 μL of Buffer 1 + Zymolyase using a multichannel pipettor. The plates were covered and incubated with shaking for 30 min at 37˚C. Six ml of Buffer 2 (50 mM Tris-HCl, pH 8, 20 mM EDTA, 0.35 M SDS) was placed in a fresh reservoir and 20 μl was added to each well and mixed well with a pipettor. The plates were covered and incubated for 30 min at 65˚C with rotation. Six ml 5 M postassium acetate in a reservoir was used to distribute 16 μL to each well. After mixing well, the plates were covered and incubated at 4˚C for 10 min. Nine ml 1 X TE (10 mM Tris-HCl, pH8, 1 mM EDTA) was used to add 80 μL TE to each well, after which the plates were covered and spun in a Beckman tabletop centrifuge for 60 min at 3500 rpm at 4˚C. To precipitate the DNA, 50 μL supernatant from was transferred to a new 96 well plate containing 50 ul isopropanol in each well and mixed. The plates were covered with a plastic lid and incubated at room temperature for five minutes. DNA pellets were obtained by spinning the plates using a Beckman tabletop centrifuge for 10 min at 3500 rpm, after which the supernatants were carefully removed by aspiration. The DNA pellets were washed with 100 μl 70% ethanol and centrifuged as before. After removing the supernatants by aspiration, uncovered plates were placed at 42˚C for 30 minutes to allow the pellets to dry. After the addition of 30 μL TE each well and the plates were sealed and incubated at 4˚C overnight. The next day the DNA was resuspended by mixing with the pipettor. The DNA concentration of 1 μL of each sample was determined using the Qubit dsDNA HS Assay kit (Fisher Scientific, #33231) in black, round bottom non-treated 96 well plates (Fisher Scientific 07200762) per the manufacturer's instructions using a BioTek Synergy 2 fluorimeter.
To facilitate the DNA sequencing analysis PCR amplicons containing the 6,865 bp recombination interval were amplified using the following primers: ACGGCACCACTATAAACCCG and GTGGGCTAAAGAACGCGAAC with spore specific barcodes on the 5' end (S2 Data). All reactions were performed in 96 well plates as indicated in S2 Data. Each reaction contained 20 μL and used Q5 polymerase (New England Biolabs, #M0491). The reagents were distributed into each well of a 96 well plate as follows. A "Column Master Mix" was created by adding (13 X 4.0 μL 5X Q5 reaction buffer) + (13 X 2.6 μL dNTPs) + (13 X 0.75 μL water) + (13 X 0.4 μL 50 mM MgCl 2 ) + (13 X 0.5 μL Q5 polymerase) and 82 μL was added to each of 12 strips containing eight 0.2 ml tubes. Ten μL of 10 uM Forward primer for the appropriate column was then aliquoted into each tube of a given strip. After mixing, 9.25 μL from each tube in a single strip was aliquoted into the appropriate column of a 96 well PCR plate. A "Row Master Mix" was generated by adding 14 X 8.75 μL water and then aliquoting 122.5 μL water into 8 strips of 8 tubes. Each strip represents a row, so 14 μL of the 10 μM Reverse primer designated for that row was added to each tube within a strip. 9.75 μL of the Row Master mix was then added to the 12 wells in each row of the PCR plate already containing the Column mix. Finally, one μL of DNA was added to the appropriate well of each plate. The plate was covered with an optically clear PCR plate plastic seal and spun for 2 min at 2000 rpm in a Beckman tabletop centrifuge. PCR reactions were performed using a Roche LightCycler 96 with the following program: 98˚C, 40 sec; 98˚C 20 sec, 67˚C, 20 sec, 72˚C, 7 min (30 cycles); 72˚C, 10 min; 20˚C hold.
Once the PCRs were complete, the samples were centrifuged as before to collect all of the liquid at the bottom of each well. Two μL from each reaction were run on a 0.9% agarose gel in 0.5X TBE buffer at 100 V for one hour. All reactions were examined for the presence of añ 6 kb fragment. There were 159 tetrads in which all four spores exhibited a prominent PCR band and were included for the sequencing analysis. Four μL from each successful reaction were combined into one 1.5 ml microfuge tube and the primers were removed using Agencourt AMPure XP beads (Beckman Coulter #A63881).

DNA sequencing and analysis
The pooled amplicons were sequenced on a Pacbio RS II instrument. Subsequently, barcodes were identified among PacBio reads (2.53 million) using the locate function of seqkit (Shen et al., 2016). Based on the barcode information the Pacbio-fastq file (https://www.ncbi.nlm. nih.gov/sra/PRJNA873686) was demultiplexed into sample specific fastq files (n = 794) using a custom script. Each of the sample specific fastq files was aligned to the hotspot (Ahuja et al., 2021) using minimap2 [166], Base composition per nucleotide per read was determined using the sam2tsv function of the jvarkit [167].
Subsequently, phase was determined by a custom script that functions as follows: (1) determine the base(s) present at each polymorphic position (n = 48) on each read (n = 2,384,252); (2) assign W (WT) or S (SNPs) for each polymorphic position per read depending on the nucleotide(s) present at the position; (3) select reads that have information on all polymorphic loci; and (4) group reads by identical marker configuration at all polymorphic locations and plot them as haplotype and number of reads that support the haplotype. Next, these plots were manually analyzed, and strands were placed into the following categories: (1) only one haplotype, signifying homoduplex at all markers across the interval i.e., the two strands are identical; (2) two haplotypes across the interval signifies two strands of the spore colony; (3) more than two haplotypes in these plots. In these cases, two haplotypes are supported by a majority of reads and some other reads may be promiscuous rearrangement due to PCR issues and these need to be manually inspected and removed. The final haplotypes used in the current project are derived from 1,179,721 reads, where per haplotype reads have a mean of 1584 reads with standard deviation of 1286 reads. Once, a consensus haplotype for each strand of each spore colony is determined and is scored in S2 Data. Schematics of the marker configurations in each tetrad are in S3 Data. Single colonies of the hedΔR NDT80-mid rrm3Δ diploid (NH2596) containing two copies of the vector (v) (pRS306), RRM3 (pBG22) or P REC8 -RRM3 (pJW7) and the hedΔR NDT80-mid pif1-md diploid, NH2661, containing two copies of pRS306, PIF1 (pJW5), pif1-m1 (pJW5-m1), PIF1-3FLAG (pJW11) or pif1-m1-3FLAG (pJW14) were sporulated on solid medium and at least 20 tetrads were dissected per transformant. The statistical significance of differences between strains was determined using the Mann-Whitney test ( � = p < .05; �� = p < .01; ��� = p < .001, ���� = p < .0001). (B) Distribution of viable spores in tetrads for a subset of the hedΔR NDT80-mid pif1-md dissections shown in Panel A. (C) Single colonies of hed1Δ dmc1Δ (NH942), hed1Δ dmc1Δ pif1-md (NH2716), hedΔR dmc1Δ (NH2701), hedΔR dmc1-II3A (NH2714), and hedΔR dmc1Δ pif1-md (NH2693) were sporulated and dissected and analyzed as described for Panel A. (D) Distribution of viable spores in tetrads for the hedΔR NDT80-mid rrm3Δ dissections shown in Panel A. (TIF) S6 Fig. PIF1 promotes meiotic progression when Rad51 is constitutively active during meiotic prophase I. (A) Meiotic progression. Timecourse analysis to determine the percentage of bi and tetranucleate cells was performed as in Fig 1A and the average %MI + MII values plotted for each timepoint. The strains and data for WT, hedΔR, and hedΔR NDT80-mid are the same as in Fig 3A with