Exo1-protected DNA nicks direct crossover formation in meiosis

In most sexually reproducing organisms crossing over between chromosome homologs during meiosis is critical for the viability of haploid gametes. Most crossovers that form in meiosis in budding yeast result from the biased resolution of double Holliday Junction (dHJ) intermediates. This dHJ resolution step involves the actions Rad2/XPG family nuclease Exo1 and the Mlh1-Mlh3 mismatch repair endonuclease. At present little is known about how these factors act in meiosis at the molecular level. Here we show that Exo1 promotes meiotic crossing over by protecting DNA nicks from ligation. We found that structural elements in Exo1 required for interactions with DNA, such as bending of DNA during nick/flap recognition, are critical for its role in crossing over. Consistent with these observations, meiotic expression of the Rad2/XPG family member Rad27 partially rescued the crossover defect in exo1 null mutants, and meiotic overexpression of Cdc9 ligase specifically reduced the crossover levels of exo1 DNA binding mutants to levels approaching the exo1 null. In addition, our work identified a role for Exo1 in crossover interference that appears independent of its resection activity. Together, these studies provide experimental evidence for Exo1-protected nicks being critical for the formation of meiotic crossovers and their distribution.


INTRODUCTION
Cells in meiosis undergo a single round of DNA replication followed by reductional and equational chromosomal divisions to produce haploid gametes. In most eukaryotes, including budding yeast and humans, the accurate segregation of homologous chromosomes during the first reductional division (Meiosis I) requires the formation of crossovers between homologs.
Physical linkages created by crossovers and sister chromosome cohesions distal to the crossover site are critical for proper segregation of chromosome pairs during Meiosis I (Maguire, 1974;Hunter, 2015;Zickler and Kleckner, 2015). The inability to establish these physical connections can lead to improper chromosome segregation and aneuploidy, and in humans is thought to be an important cause of birth defects and miscarriages (Hassold and Hunt, 2001;Nagaoka et al., 2012;Hunter, 2015).
In baker's yeast crossover formation in meiotic prophase is initiated through the genome-wide formation of roughly 150 to 200 Spo11-induced double-strand breaks (DSBs; Keeney et al., 1997;Pan et al., 2011). These breaks are resected in a 5' to 3' direction to form 3' single-stranded tails (Cao et al., 1990;Padmore et al., 1991). Strand exchange proteins coat the single stranded tails and promote their invasion into homologous sequences in the unbroken homolog (Hunter, 2015). In the major crossover pathway (Class I), the resulting D-loop intermediate is stabilized by ZMM proteins including Zip2-Zip4-Spo16 and Msh4-Msh5 to form a single end invasion intermediate (SEI; Figure 1A; Hunter and Kleckner, 2001;Fung et al., 2004;Borner et al., 2004;Lynn et al., 2007;De Muyt et al., 2018). This recombination intermediate forms concomitantly with the synaptonemal complex, a structure that is thought to remove chromosomal tangles and interlocks during the homology search process (Padmore et al., 1991;Sym et al., 1993;de Boer and Heyting, 2006). DNA synthesis from the SEI, followed by secondend capture, results in the formation of the double-Holliday junction intermediate (dHJ). The dHJ is thought to be stabilized by Msh4-Msh5 and resolved in a biased orientation to form ~90 crossovers (COs) in the yeast genome that are distributed so that they are evenly spaced (crossover interference) and every homolog pair receives at least one crossover ( Figure 1A; Gioia, Payero et al., 8/29/21 4 Szostak et al., 1983;Sym and Roeder, 1994;Kleckner, 1994, 1995;Borner et al., 2004;Hillers, 2004;Jones and Franklin, 2006;Mancera et al., 2008;Zakharyevich et al., 2012).
How dHJs are resolved in a biased manner to form crossovers is a major unanswered question. Investigators have suggested that the presence of nicks in dHJs ensures biased resolution by creating asymmetric structures that are resolved to form crossover-only products (reviewed in Machin et al., 2020). In support of such ideas, whole genome sequencing of hDNA tracts formed in meiosis inferred a model in which meiotic crossover resolution is biased towards DNA synthesis tracts (Martini et al., 2011;Marsolier-Kergoat et al., 2018). In this model nicks maintained at the ends of synthesis tracts could direct biased and asymmetric cleavage of the dHJ by recruiting a nick-binding protein that acts in the resolution mechanism. However, such a model is inconsistent with a denaturing gel analysis of dHJs that form at a meiotic hotspot in S. cerevisiae; this work showed that all single strands within the dHJs are continuous (Schwacha et al., 1994;. It is also inconsistent with recent work in S. cerevisiae showing that a vast majority of crossovers initiated at another hotspot displayed evidence of branch migration, with about half of the COs having formed from dHJs located on one side of the initiating double-strand break. In such a situation, nicks should not be present in positions that direct biased resolution (Ahuja et al., 2021). Thus, it remains unclear if nicks participate in meiotic crossover formation.
What factors act in the biased resolution of dHJs? The MMR endonuclease Mlh1-Mlh3 and the XPG/Rad2 family nuclease Exo1 have been shown to act in meiotic crossover resolution, with mlh3Δ and exo1Δ single and double mutant strains displaying similar crossover defects in crossing over (Khazanehdari and Borts, 2000;Zakharyevich et al., 2010;. Biochemical analyses of Mlh1-Mlh3 indicate that its endonuclease activity is required for its role in crossover formation, but not as a structure-specific nuclease that symmetrically cleaves Holliday junctions (Nishant et al., 2008;Rogacheva et al., 2014;Ranjha et al., 2014;. Exo1 acts in many steps in DNA metabolism, including creating 3' single-stranded ends for homologous recombination, telomere maintenance, DNA mismatch repair, DNA replication, and crossover-specific dHJ resolution in meiosis. Exo1 contains an N-terminal Rad2/XPG nuclease domain that is conserved in Rad2/XPG family members and an unstructured C-terminal tail that interacts with the mismatch repair factors Msh2 and Mlh1 (Tishkoff et al., 1997;Tran et al., 2001). In vitro studies demonstrated that Exo1 displays a robust and processive 5' to 3' exonuclease on the ends of a double-strand break, and on gapped and nicked duplex DNA. In addition, it displays 5' flap endonuclease activity (Kunkel and Erie, 2015;Goellner et al., 2015;Szankasi and Smith, 1992;Fiorentini et al., 1997;Lee and Wilson, 1999;Tran et al., 2002;Genschel and Modrich, 2003;Zakharyevich et al., 2010).
In meiosis exo1D strains display a defect in the 5' to 3' resection of Spo11-induced DSBs and a meiotic crossover defect. In fact, resection is reduced in exo1D to an average of 270 nt compared to 800 nt in wild-type. Despite showing these defects, exo1D mutants display wildtype timing and levels of meiotic recombination intermediates, including dHJs (Zakharyevich et al., 2010). Genetic analysis showed that disruption of a conserved Mlh1-Interaction Protein sequence (MIP box) in the Exo1 C-terminal domain conferred intermediate defects in meiotic crossing over, suggesting that Exo1 promotes meiotic crossovers through interactions with Mlh1 and possibly other factors (Amin et al., 2001;Argueso et al., 2003;Tran et al., 2004;Zakharyevich et al., 2010). Curiously, an exo1 mutation (D173A) that disrupts a metal binding site critical for nuclease function was shown to have only a minimal impact on meiotic crossing over. Together these analyses suggested that Exo1's interactions with Mlh1-Mlh3, but not its nuclease function, are critical for crossover formation (Abdullah et al., 2004;Zakharyevich et al., 2010;Keelagher et al., 2011).
The studies outlined above in addition to recent biochemical analyses have led to the proposal that Mlh1-Mlh3 interacts with Exo1, Msh4-Msh5 and the DNA polymerase processivity factor PCNA for biased resolution of double Holliday junctions (Cannavo et al., 2020;Sanchez et al., 2020;Kulkarni et al., 2020). This proposal suggests that DNA signals are present in dHJ intermediates that are critical for such resolution; however, these studies have not provided Gioia,Payero et al.,8/29/21 6 direct evidence for such signals. Here we provide genetic evidence that Exo1 acts to protect DNA from being ligated in recombination intermediates during the formation of crossover products. We also show that it plays a critical role in ensuring that meiotic crossovers are widely spaced for proper chromosome segregation in the Meiosis I division. These observations provide evidence for dynamic and distinct roles for Exo1 in both crossover placement and for maintaining a nicked recombination intermediate for the resolution of dHJs into crossovers.

Mutations in metal coordinating and active site residues in Exo1 do not disrupt meiotic crossing over.
The crystal structure of human Exo1 with 5' recessed DNA (PDB #3QE9) identified two metals in the catalytic site of the Exo1-DNA structure, with residue D171 assisting D173 in coordinating one metal, and residue D78 coordinating the other, to hydrolyze the phosphodiester backbone of DNA (Figures 1B and S1;Orans et al., 2011;Mueser et al.,1996;Hwang et al., 1998;Feng et al., 2004;Shi et al., 2017). While the exo1-D173A mutation in baker's yeast was shown to disrupt Exo1 nuclease activity (Tran et al., 2002), mutation of other amino acids that coordinate the catalytic metals was not performed. Mutation of other nucleases that act through a twometal catalysis mechanism suggested that altering a single metal binding residue does not fully ablate function and could create novel functions, perhaps because a water molecule can substitute as a ligand (Schiltz et al., 2019). For example, work by Lee et al. (2002) showed that the human exo1-D78A and exo1-D173A mutant proteins display nuclease activities, though at levels significantly lower than the wild-type protein.
In baker's yeast meiosis, mutation of a single metal binding residue (exo1-D173A) caused a disruption in the 5' to 3' resection steps of meiotically induced DSBs, but only minor, if any defects in meiotic crossing over, suggesting that Exo1's nuclease functions were not required in this step (Abdullah et al., 2004;Zakharyevich et al., 2010). We purified exo1-D173A from baculovirus infected Sf9 cells (Materials and Methods), but were unable to purify a full Gioia,Payero et al.,8/29/21 7 length variant (exo1-D78A,D173A) expected to disrupt both metal binding sites ( Figure S2). We tested the nuclease activity of exo1-D173A on a 2.7 kb pUC18 substrate containing four preexisting nicks ( Figure S2A) as well as supercoiled plasmids. As shown in Figure S2A and C, exo1-D173A was deficient for exonuclease activity on the substrate containing four pre-existing nicks. However, exo1-D173A displayed a weak DNA nicking activity on closed circular DNA similar to that seen for Mlh1-Mlh3 (~10% nicking of pUC18 at 20 nM exo1-D173A compared to ~20% nicking at 20 nM Mlh1-Mlh3), suggesting that a role for Exo1 nuclease activity in crossover resolution was not fully resolved . In contrast, wild-type Exo1 did not display such nicking activity, consistent with previous work showing that human Exo1 displayed little or no endonuclease activity on blocked-end DNA substrates ( Figure S2B; Lee et al., 2002). Interestingly, the addition of a mutation predicted to be critical for DNA binding, G236D (see below), decreased the nicking activity of the exo1-D173A protein by about two-fold, consistent with previous studies indicating that Exo1 nuclease activity was dependent on its DNA binding activity ( Figure S2D; Orans et al., 2011).
To test the effect of mutations in the Exo1 catalytic site we made D78A, D171A, and D173A mutations (Group I, Figure 1B) in combination to disrupt coordination of both metals.
We also mutated residues in Exo1 which interact with and position DNA in an orientation to be cleaved (Orans et al., 2011). These residues (H36, K85, R92, K121, Group II) contribute to the fraying of the duplex DNA bases away from its complement and reside within an α4-α5 helical arch microdomain that forms part of the Exo1 active site ( Figures 1B, S1). This microdomain is important for catalysis and also defines substrate specificity throughout the flap endonuclease (FEN) superfamily and consequently Exo1 5' flap binding (Ceska et al., 1996;Devos et al., 2007;Gloor et al., 2010;Orans et al., 2011). Within this region R92 has been shown to be a critical residue for Exo1 catalysis; it interacts with the scissile bond on the DNA to position it adjacent to the catalytic metal core, and the R92A mutation dramatically decreased nuclease activity of human Exo1 in vitro to similar levels of the D173A metal-coordinating mutation (Orans Gioia, Payero et al., 8/29/21 8 et al., 2011). K121 (R in human Exo1) is part of the α5 helix and coordinates passage of the DNA substrate through the active site.
We analyzed meiotic crossing over by tetrad analysis at four consecutive intervals on Chromosome XV (104.9 cM map distance in wild-type, 52 cM in exo1Δ) and at one interval (CEN8-THR1) on Chromosome VIII (~39% single crossovers in wild-type, 20% in exo1Δ; Figures 2A and 3A;Thacker et al., 2011). These two chromosomal regions showed defects in crossing over similar to those seen previously (exo1Δ, ~2-fold decreased; mlh3Δ, ~2-fold; msh5Δ, ~3-fold; exo1Δ mus81Δ, ~12-fold) and confirmed the epistatic relationship between exo1Δ and mlh3Δ ( Figure 2B; Argueso et al., 2004;Nishant et al., 2008;Zahkaryevich et al., 2012;Al-Sweel et al., 2017). As shown in Figures 2B and 3B and Tables S1 and S2, disruption of either one or both metal binding sites of Exo1 (Group I) had minor if any effects on meiotic crossing over. There was a small crossover (<10%) reduction in some of the catalytic mutants compared to wild-type; this reduction could result from defects in DNA binding that result from perturbation of the active site. In fact, the human exo1-D78A mutant protein showed defects in binding to DNA flap structures (Lee et al., 2002). In addition, the exo1-H36E, exo1-K85A/E, exo1-R92A and exo1-K121A/E mutations (Group II) had very modest, if any effect on meiotic crossing over compared to wild-type, suggesting that coordination of the scissile bond for catalysis within the active site is not critical for crossing over. The dramatic loss of nuclease activity seen with human Exo1 bearing K85A, R92A or K185A mutations (Orans et al., 2011;Li et al., 2019) further supports the dispensability of Exo1 catalytic activity for crossing over.
These observations indicate that the critical function(s) of Exo1 in meiotic crossover resolution are not catalytic in nature.
Mutation of DNA binding domains of Exo1 reveal a DNA binding role for Exo1 in meiotic crossing over.
The structure solved by Orans et al. (2011) revealed that Exo1 makes key contacts with DNA through several defined domains ( Figure 1B). For example, G236 (Group IV) is one of several 9 residues in a helix-two turn-helix motif that coordinates a metal ion and forms hydrogen bonds with DNA backbone oxygen residues to stabilize an interaction with Exo1 and the pre-nick duplex DNA. This conserved motif is only slightly modified from observed FEN-1 structures (Ceska et al., 1996;Feng et al., 2004) and is presumed to facilitate exonuclease processivity as the protein moves along the DNA backbone (Pelletier et al., 1996;Orans et al., 2011). K185 is part of a small hairpin loop between strands β6 and β7 and is also thought to be critical for recognition of duplex DNA (Orans et al., 2011;Li et al., 2019). The K185A mutation has been shown to diminish Exo1 nuclease activity several fold in vitro, and confer elevated sensitivity to DNA-damaging agents, likely due to a defect in binding duplex DNA (Li et al., 2019). A crucial component of Rad2/XPG members is the hydrophobic wedge ( Figure 1B, Group III), a structurally conserved domain which induces a sharp bend at a ds-ssDNA junction, and gives the enzyme family its specificity for gapped/nicked DNA substrates (Orans et al., 2011, Chapados et al., 2004. Several hydrophobic residues within the wedge motif displace the nonsubstrate strand, as well as two lysine residues which appear to coordinate this portion of the non-substrate strand ( Figure 1B). Figure 2B and 3B and Tables S1 and S2, the exo1-K185E and exo1-G236D mutations conferred significant decreases in crossover formation (68 cM, 29.1% tetratype in exo1-G236D and 73 cM, 24.5% tetratype in exo1-K185E) in the URA3-HIS3 and CEN8-THR1 intervals, respectively. Interestingly, the hydrophobic wedge mutations exo1-S41E (58.6 cM, 28.4% tetratype), and exo1-F58E (69.9 cM, 27.8% tetratype) also conferred crossover defects with double mutation combinations (exo1-K185E,G236D-24.2% tetratype; exo1-S41E,F58E-24.6% tetratype) conferring more severe phenotypes. We then made a series of double and triple mutants that included a catalytic, DNA binding, and Mlh1-interacting (MIP) mutations ( Figure 3B; Table S1). Combining groups did not confer crossover phenotypes equivalent to the exo1Δ, and including a catalytic mutation (-D171A, -D173A) with any single DNA binding mutation that conferred a crossover phenotype did not further impair crossover formation.
The data collected from assaying double and triple mutants validated the results of single catalytic and DNA binding mutations, identified DNA binding mutants that confer a near exo1Δ crossover phenotype, and showed that the Exo1 active site is relatively insensitive to mutation for crossover formation. These observations also indicated that the decrease in crossover frequency seen in single mutants is compounded in multiple mutant combinations ( Figure 3B).
We then examined the spore viability of exo1 mutant strains. The exo1Δ strain showed a tetrad spore viability pattern (74% spore viability; 4, 2, 0 viable tetrads > 3, 1) consistent with Meiosis I non-disjunction (Figures 2B;S3;Ross-Macdonald and Roeder, 2004;Abdullah et al., 2004). However, decreases in meiotic crossing over and spore viability did not correlate in the exo1 strains. For example, exo1 mutants with very similar defects in crossing over showed spore viabilities that ranged from 89% (exo1-G236D, exo1-MIP) to 71 to 73% (exo1-K185E, G236D,MIP). A plausible explanation for these differences is that the exo1 mutations display other phenotypes in addition to meiotic crossover phenotypes. In fact, some of the exo1 mutations analyzed above conferred defects in DNA repair, as measured by sensitivity to methyl-methane sulfonate (MMS). However, the MMS phenotypes did not correlate with defects in meiotic crossing over ( Figure S4). For example, the exo1-D78A, exo1-D171A, and exo1-D173A catalytic mutations conferred stronger MMS sensitivities compared to their nearly wild-type meiotic CO phenotypes. Similar disparities between DNA repair and CO phenotypes were seen for the active site mutations exo1-K85E and exo1-K121A, the DNA binding mutant exo1-K185E and the MLH interacting mutant exo1-MIP. This analysis suggested that the lack of correlation between spore viability and crossover phenotype seen in exo1 mutants was likely complicated by their defects in DNA repair. Further support for this idea was seen by the lack of a 4, 2, 0 viable tetrads > 3, 1 pattern in the exo1 mutant alleles, though this pattern was clearly displayed by exo1Δ ( Figure S3). One explanation for this lack of a pattern in exo1 mutants with strong crossover defects is that the DNA repair defects in these mutants conferred a pleiotropic decrease in spore viability, obscuring a Meiosis I non-disjunction phenotype. Another potential explanation (discussed below) is that exo1Δ strains show increased disjunction as the result of defects in crossover positioning (genetic interference, see below). Together, these observations provide evidence that Exo1 contains distinct DNA repair and meiotic CO functions and DNA binding by Exo1, but not its nuclease activity, is critical for meiotic CO resolution. %0-)0*1$% '.%&FSun et al., 2003, Ip et al., 2008Tomlinson et al., 2010). In yeast, RAD27 shares the highest sequence similarly with EXO1, suggesting functional overlap. In fact, previous studies have shown that EXO1 can complement some RAD27 functions, and the exo1D rad27D double mutant is inviable (Tishkoff et al., 1997, Xie et al., 2001Qiu et al., 1999). While the substrate preferences of Rad2 family proteins vary, all have been shown to bind nicked, gapped, and/or blunt end DNA, with a particular affinity for single-to double-stranded DNA junctions. They all appear to induce a sharp bend in the DNA substrate upon protein binding (Lee and Wilson, 1999;Genschel and Modrich, 2003;Orans et al., 2011). These observations structurally demonstrate how RAD2 family proteins can share redundant capacities for endoand exo-nucleolytic functions.

Expression of
We reasoned that a protein that mimicked the DNA binding affinity for similar DNA substrates could complement this function in cells lacking Exo1. We therefore tested the ability for Rad27 to complement the meiotic function of Exo1. We did not observe complementation by RAD27 expressed through its native promoter, but upon placing RAD27 under control of the EXO1 promoter (pEXO1-RAD27) we saw significant increases in crossing over on both Chromosomes VIII (from 21.5% to 29.9% tetratype; Figure 4A; Table S1B) and XV ( Figure 4B; 54 cM map distance in exo1D to 72 cM exo1D containing pEXO1-RAD27), likely due to the high levels of meiotic expression of the EXO1 promoter ( Figure S5; Brar et al., 2012). Efforts were made to improve exo1D complementation by fusing a MIP domain, or the entire C-terminus of Exo1 to Rad27 to create a functional Mlh1 interaction; however, they were unsuccessful.
We reasoned that if Rad27 complemented the meiotic role of Exo1 by binding a specific DNA substrate based on structural similarity divorced from catalytic activity, inactivating Rad27 through mutation of a metal-coordinating aspartic acid D179 (Shen et al., 1996;Gary et al., 1999) would not impact its ability to effect higher crossover frequencies. Indeed, exo1Δ cells expressing pEXO1-RAD27 or pEXO1-rad27-D179A showed similar levels of crossover complementation. This observation encouraged us to further test our hypothesis by making five additional rad27 mutations based on previous biochemical and structural characterization of the human homolog of Rad27, FEN-1. These included rad27-R101A; equivalent to FEN1-R100A, of which the mutant FEN-1 protein exhibited a strong catalytic defect but remained competent for flap binding and bending (Song et al., 2018), and rad27-R105A and rad27-K130A, equivalent to FEN-1-R104A and FEN-1-K132A, of which the mutant FEN-1 proteins exhibited 20-and 5fold reductions in flap cleavage but were not characterized for flap binding or bending (Tsutakawa et al., 2017). Two other mutations were analyzed based on Exo1 and Rad27 homology: rad27-A45E, which aligns to a mutation in the Exo1 hydrophobic wedge (exo1-S41E, Group III, Figure 1B), and rad27-H191E, which aligns to a mutation in the Exo1 DNA binding domain (exo1-K185E, Group IV). As shown in Figure 4A, rad27-R101A, rad27-R105A and rad27-K130A, which coordinate the scissile bond for catalysis, complemented the crossover defect in exo1D , consistent with the phenotypes exhibited by exo1 Group II mutations.
We also tested if RAD27 expression from the EXO1 promoter could improve meiotic crossover functions of exo1 strains bearing mutations within (exo1-K185E) or outside of the DNA binding domain (exo1-MIP). As shown in Figure 4C, meiotic crossing over in exo1-K185E, but not exo1-MIP, was increased in cells containing pEXO1-RAD27. These observations are consistent with Rad27 being able to substitute for Exo1 DNA binding functions because improved complementation by pEXO1-RAD27 was seen in a DNA binding mutant (exo1-K185E) but not in a mutant predicted to be functional for DNA binding (exo1-MIP), but defective in interacting with other crossover factors.
Finally, we saw no complementation of meiotic crossing over by pEXO1-RAD27 in strains lacking functional Mlh1-Mlh3 (mlh3Δ), indicating that Rad27 complementation was specific to Exo1 function. This observation differs from observations made by Arter et al. (2018), who found that expression of the Rad2/XPG nuclease Yen1 complemented crossover defects in both exo1D and mlh3D strains. One explanation for the Yen1 complementation phenotype is that Yen1 Holliday junction resolvase activity could bypass Mlh1-Mlh3-Exo1 dependent dHJ resolution steps.

Meiotic crossover phenotype of exo1 DNA binding mutants is significantly reduced when
Cdc9 ligase is overexpressed in meiosis. Reyes et al. (2021) et al. recently showed that overexpression of the budding yeast ligase Cdc9 disrupted DNA mismatch repair through the premature ligation of replication-associated nicks that act as critical repair signals. If the role of Exo1 in meiotic recombination involved nick binding/protection, then we reasoned that meiotic overexpression of CDC9, the budding yeast DNA ligase involved in DNA replication, could lead to premature ligation of DNA synthesisassociated nicks critical for maintaining biased resolution. We posited that some exo1 DNA binding mutants that maintained near wild-type levels of crossing over might be especially Gioia,Payero et al.,8/29/21 14 susceptible to Cdc9 overexpression. During meiosis CDC9 expression appears to be low relative to HOP1, whose expression increases dramatically in meiotic prophase and remains high through dHJ resolution (~6hrs in meiosis; Figure S5). We thus expressed CDC9 under control of the HOP1. As shown in Figure 4D we saw no disruption of crossing over in exo1 mutants that contained intact DNA binding domains (EXO1, exo1-MIP, exo1-D173A) or in a mutant (exo1-K85E) predicted to be defective in steps post-DNA bending (Orans et al., 2011).
However, we saw modest to severe losses of crossing over in exo1 DNA binding mutant hypomorphs. As shown in Figure 4D, pHOP1-CDC9 reduced single crossovers in exo1-K185A from 35.3 to 31.3% and in exo1-K61E from 35.1 to 25.2%. These data, in conjunction with the RAD27 complementation experiments, provide evidence for a nick protection role for Exo1 in crossover formation.

Interference analysis suggests a role for Exo1 prior to crossover resolution.
While expression of RAD27 under the EXO1 promoter (pEXO1-RAD27 plasmid) could partially complement CO defects in exo1Δ strains, it did not improve the meiotic spore viability or MMS resistance seen in exo1D strains (Figures 4B). We performed crossover interference analysis to determine if exo1D strains showed defects in addition to those seen in DSB resection and CO resolution. As described below, we found that exo1D strains displayed crossover interference defects that were not complemented by the pEXO1-RAD27 plasmid.
First, we analyzed exo1D strains bearing pEXO1-RAD27 for defects in crossover interference on chromosome XV using the Malkova method, which calculates genetic distances between intervals in the presence and absence of a neighboring crossover ( Figure 5; Table S3; Malkova et al., 2004;Martini et al., 2006). These measurements are presented as a ratio, wherein 0 indicates complete interference and 1 indicates no interference. Three pairs of intervals (URA3-LEU2-LYS2, LEU2-LYS2-ADE2, LYS2-ADE2-HIS3) were tested for interference. In all three interval pairs tested, exo1Δ displayed a loss of interference compared to wild-type. Most strikingly, two intervals that displayed strong interference in wild-type strains (Malova ratios of 0.48 at URA3-LEU2-LYS2 and 0.43 at LEU2-LYS2-ADE2) displayed a complete loss of interference in exo1Δ (1.28 and 0.84 respectively). These results are reminiscent of the interference defects observed previously in msh4D and msh5Δ (Ross-Macdonald and Roeder, 1994;Hollingsworth et al., 1995;Novak et al., 2001;Nishant et al., 2010; Figure 5). Interestingly, a lack of interference was observed in all three intervals in the exo1D strain containing pEXO1-RAD27 (Malkova ratios of 1.41, 0.90, and 0.81 in Intervals I, II, III, respectively; Figure 5), supporting the idea that RAD27 expression in meiosis could complement only Exo1's crossover functions.
The interference defect seen in exo1Δ (all three intervals showed a lack of interference) was stronger than that seen in the mlh3Δ strain (two intervals showed a lack of interference), suggesting a role for Exo1 in promoting interference independent from its association with Mlh1-Mlh3 in crossover resolution. To determine if the early resection role of Exo1 (Zahkaryevich et al., 2010) could account for this interference function, exo1-D171A,D173A and exo1-D78A,D173A catalytic mutants were analyzed for interference defects ( Figure 5). Strikingly, these mutants displayed interference similar to or stronger than wild-type. In fact, the interference defect observed in exo1Δ was not recapitulated in any of the exo1 alleles tested.
Interference was also measured using the COC (Coefficient of Coincidence) method (Papazian, 1952; Table S3A). COCs measure the double crossover rate compared to the expected rate in the absence of interference. The COC ratios were consistent with the Malkova ratio analysis, supporting the idea that loss of interference in exo1Δ was not recapitulated in any of the mutant alleles. Together the data indicate a previously uncharacterized role for Exo1 in establishing crossover interference and suggest that the pro-interference role of Exo1 is either more robust than the pro-crossover role or involves specific contact or interaction sites that were not examined in this study (see Discussion).

Genetic interactions involving Msh4-Msh5, Mlh1-Mlh3 and Exo1 also support roles for
Exo1 in crossover interference. Gioia,Payero et al.,8/29/21 16 The finding that exo1D showed defects in crossover interference encouraged us to determine if we could identify genetic interactions involving factors that interact with Exo1 and play roles in crossover interference. To initiate this work we analyzed exo1-F447A,F448A (referred to as exo1-MIP), which contains mutations in an Mlh1-interacting peptide box (MIP) that disrupt both Mlh1-Exo1 interactions and meiotic crossing over (Tran et al., 2007;Zakharyevich et al., 2010).
In the spore autonomous fluorescence assay we found that the exo1-MIP mutation conferred intermediate defects in CO formation (33.3% single crossovers (tetratype) compared to 37.5% in wild-type) when both this allele and MLH3 were present in two copies ( Figure S6; Table S4).
However, when both exo1-MIP and MLH3 were present in single copies, we observed a twofold reduction in CO levels (to 22.6% tetratype) that approached levels seen in 5$"BG ( Figure   S6). This observation confirmed interactions between Mlh1-Mlh3 and Exo1 and encouraged us to use gene dosage as an approach to identify additional genetic interactions involving Exo1 using mlh3 alleles, mlh3-42 and mlh3-54, that confer defects in Mlh3-mediated mismatch repair (MMR) but do not disrupt crossing over. Previous work showed that the mlh3 alleles disrupted Mlh1-Mlh3 interactions (Al-Sweel et al., 2017). We reduced the gene dosage of eleven meiotic genes from two to one and measured crossing over at the CEN8-THR1 interval on chromosome VIII ( Figure S6; Table S4). SGS1 and RMI1 were included because they encode components of a Sgs1-Top3-Rmi1 complex that acts as a pro-crossover factor in meiotic recombination (Jessop et al., 2006;Zakharyevich et al., 2012;Kaur et al., 2015).
As shown in Figure S6 and Table S4, we observed defects for both mlh3 alleles in crossing over when the gene dosage of EXO1, MSH4, or MSH5 was reduced to one copy. For MLH1, we observed such dosage effects with only the mlh3-54 allele, and for SGS1 and RMI1, with only the mlh3-42 allele ( Figure S6). Interestingly, the residues mutated in mlh3-54 mapped to the Mlh1-Mlh3 dimerization interface whereas residues mutated in mlh3-42 mapped to the distal periphery of the dimerization interface (Dai et al., 2021). While this observation might help explain the different effect of gene dosage for MLH1 in mlh3-42 and mlh3-54 backgrounds, it is unclear why the mlh3-42 allele disrupts the stability of Mlh1-Mlh3 or why it showed gene dosage interactions with SGS1 and RMI1. mlh3 allele-specific interactions were not observed when reducing dosage for a group of ZMM family genes (ZIP1, ZIP3, ZIP4, SPO16, MER3) which are thought to act upstream of Mlh1-Mlh3 to stabilize early recombination intermediates and promote CO outcomes (Agarwal and Roeder, 2000;Snowden et al., 2004;Borner et al., 2004;Kolas et al., 2005;Argueso et al., 2004;Shinohara et al., 2008;Hatkevich and Sekelsky, 2017). As shown in Figure S6, a reduction of gene dosage for ZIP1 and SPO16 did not alter crossing over in any MLH3 background, and a reduction of dosage for ZIP3 and MER3 led to CO decreases in MLH3, mlh3-42, and mlh3-54 backgrounds. ZIP4 fit a somewhat similar pattern to ZIP3 and MER3, but statistical significance was mixed, with significance for haploinsufficiency seen in only the mlh3-42 background. Together, these studies support a model in which Msh4-Msh5, Mlh1-Mlh3, and Exo1 form a group that participates in crossover interference (Santucci-Darmanin et al., 2002;Santucci-Darmanin et al., 2000;Zakharyevich et al., 2010;Krishnaprasad et al., 2021).

Msh5 DNA interactions and foci are not dependent on Exo1.
Crossover interference involves the recruitment of ZMM proteins which stabilize and identify a set of dHJs for Class I crossover resolution. Among this class of factors is Msh4-Msh5, which stabilizes SEIs after strand invasion (Boerner et al., 2004). During meiosis, the Msh4-Msh5 previous studies have shown that in exo1∆, joint molecule formation is normal, though there is a roughly 50% reduction in crossovers (Khazanehdari and Borts, 2000;Tsubouchi and Ogawa, 2000;Zakharyevich et al., 2010). Since interference and crossover formation is significantly reduced in msh5D, an explanation for the interference defect in exo1Δ is that Msh4-Msh5 recruitment to recombination intermediates is compromised due to reduced resection of DSBs (Zahkaryevich et al., 2010). To address this, we analysed Msh5 binding in an exo1∆ mutant using a combination of ChIP-qPCR and cytological methods.
Msh5 binding in exo1∆ was also analysed by cytological analysis of Msh5 foci ( Figure   6B). The average numbers of Msh5 foci per cell in exo1∆ at 3 hrs (34), 4 hrs (45) and 5 hrs (48) were comparable to the number of Msh5 foci in wild-type at the same time points (33, 42, and 48 respectively) ( Figure 6C). However, measurement of the foci intensity showed that the Msh5 foci appeared brighter in exo1∆ ( Figure 6C). These observations support the ChIP-qPCR data showing enhanced Msh5 binding in exo1∆ mutants, especially at DSB hotspots. Together the ChIP and Msh5 localization studies suggest that Msh4-Msh5 localization is not dependent on either the long-range resection activity of Exo1 or interaction with Exo1. This information, in conjunction with interference analysis of exo1 nuclease defective mutants supports a direct role for Exo1 in establishing interference.

DISCUSSION
In this study we identified a critical function for Exo1 in meiotic crossing over dependent on its ability to bind to nicked/flapped DNA structures. This conclusion is supported by the finding that meiotic expression of the structurally similar RAD2 family nuclease Rad27 can partially compensate for the loss of crossovers in the absence of Exo1, and that meiotic overexpression of the Cdc9 ligase conferred a significant crossover defect in exo1 DNA binding domain mutants. Based on these observations we propose that Exo1 acts in meiotic crossover formation by binding to nicks/flaps analogous to those created during lagging strand DNA synthesis ( Figure 7). In contrast to the functions of Rad27 and Exo1 during replication, which cleave 5' flaps in mechanisms that facilitates ligation of the resulting nick (Balakrishnan and Bambara, 2013), the Exo1/Rad27 meiotic crossover function occurs independently of nuclease activity. Such a nuclease-independent activity likely serves to protect nicks or flaps in recombination intermediates from premature ligation, ensuring their incorporation into a resolution mechanism. In addition, a nick/flap bound Exo1 could act to recruit Mlh1-Mlh3 to the dHJ. In support of this idea, work by Manhart et al. (2017) showed that the presence of Mlh1-Mlh3 polymer at a nicked strand can direct the endonuclease to cut the opposite strand, providing a possible mechanism for how biased resolution could occur.

Incorporating nick-protection with models of dynamic dHJs.
A role for a nicked recombination intermediate in forming meiotic crossovers has been proposed for many years, with a summary of a few studies provided below. 1. Electron microscopy studies of Holliday junction structures purified from yeast cultures in pachytene failed to reveal open centers expected of fully ligated junctions (Bell and Byers, 1983), though the structure of dHJs in vivo is not well understood, and so we cannot exclude the presence of factors that allow centers in fully ligated junctions to open. 2. Nicked HJs are favorable substrates for resolution by resolvase proteins in vitro (Fricke et al., 2005), and nicked HJs comprise a large proportion of Holliday junction structures observed in mutants defective in the structure-selective nucleases Yen1 and Mms4-Mus81, suggesting that they represent mitotic recombination intermediates (Garcia-Luis and Machin, 2014). 3. Whole genome sequencing of meiotic spore progeny inferred that the resolution of dHJs is biased towards new DNA synthesis tracts, implying that these tracts contain distinguishing features such as nicks (Marsolier-Kergoat et al., 2018). 4.
Biochemical studies have led to models in which nicks persisting during dHJ formation could provide a substrate for continued loading of MMR/replication factors implicated in dHJ resolution Gioia, Payero et al., 8/29/21 20 (e.g. RFC, PCNA, Msh4-Msh5;Kulkarni et al., 2020;Cannavo et al., 2020). Furthermore, Kulkarni et al. (2020) and Cannavo et al. (2020) showed that PCNA, which is loaded onto primer template junctions during DNA replication, promotes nicking by Msh4-Msh5 and Mlh1-Mlh3. The above observations, however, are challenging to reconcile with observations in S. cerevisiae indicating that single strands of DNA within dHJs appear to be continuous (at least at the resolution of denaturing alkaline gels; Kleckner, 1994, 1995) and dHJs are much How can nick protection be incorporated into crossover mechanisms that involve branch migration of HJs? One possibility is that nicks are translocated through "nick translation" (Marsolier-Kergoat et al., 2018). For certain types of branch migration, this mechanism would push the nicks to a new dHJ location, allowing bias to be maintained ( Figure 7B Alternatively, Mlh1-Mlh3 could nick at a distance from the Exo1-protected nick (Peterson et al., 2020, Figure 7B, lower panel), which could be reconciled based on previous studies showing that MLH proteins form polymers on DNA and can make multiple nicks on DNA (Hall and Kunkel, 2001;Manhart et al., 2017;Kim et al., 2019). In the Marsolier-Kergoat (2018) model, the synthesis of new DNA tracts has been hypothesized to be followed by processing of the resultant 5' flap to create a nick. Though appealing, this model needs to be balanced with our findings that the catalytic activity of Rad27 is not necessary to rescue crossing over in an exo1Δ strain.
A key aspect of extensive branch migration is that it should prevent DNA nicks from serving as substrates for biased resolution because they locate away from the resolution site.
To reconcile this observation with our analysis of Exo1, such nicks could act as substrates for the activation of an Mlh1-Mlh3 polymer ( Figure 7C). Previous work showed that Mlh1-Mlh3 requires a large DNA substrate for nuclease activation and that polymerization barriers impeded its nuclease activity . As such, branch migration may provide a way to move the dHJ from a constrained state that is occupied by factors that establish the dHJ such as Msh4-Msh5. In such a model, the signaling imposed by the binding of Exo1 to nicks could act across a distance, and through an initial Exo1-Mlh1-Mlh3 interaction, allowing the Mlh1-Mlh3 polymer to occupy the comparatively unconstrained DNA away from the invasion site ( Figure   7C). Thus, we may consider the Exo1-nick interaction site as a nucleation point for Mlh1-Mlh3.
This would add asymmetry to the polymer and ensure that Mlh1-Mlh3 nicks in a biased manner.
We illustrate this within the context of a model presented by Manhart et al. (2017), in which Mlh1-Mlh3 requires polymerization across multiple kilobases to be catalytically active to cleave Type II Holliday junctions. Variations of such a model have been presented by Kulkarni et al. (2020). These models would also provide an explanation for the importance of Exo1-Mlh1-Mlh3 interactions during meiotic crossing over (but see below). In this model, we see Exo1-nick interactions as a means of guarding essential nicks from premature ligation. This would ensure that the dHJ remains "flexible" if needed for Mlh1-Mlh3 polymerization and activation. These models are not mutually exclusive, and further work is required to understand how resolution factors interact with mobile and static dHJs.
An additional challenge with the models presented in Figure 7 is that while Exo1 and FEN-1 bind flap structures to coordinate tail removal and ligation steps, the endonuclease activities of these proteins do not appear to be required for crossover resolution. However, the finding that ligase overexpression can disrupt crossing over in exo1 DNA binding hypomorphs suggests that a ligatable nick serves as a critical recombination intermediate. One possibility is that there is a coordinated displacement of Exo1 by Mlh1-Mlh3 that induces Mlh1-Mlh3 nicking on the opposite strand. In such a model there could be other processing events that removal 5' tails such as one involving Msh2-Msh3 recognition of the flap, followed by endonuclease cleavage by Rad1-Rad10 (Sugawara et al., 1997). It is also worth noting that studies in which we observed complementation of the exo1D strain with the pEXO1-rad27-D179A plasmid contained native RAD27 that could act to remove 5' tails.
Does Exo1 direct Mlh1-Mlh3 nicking? A coordinated set of steps are required in meiotic recombination to promote Exo1 mediated resection of DSBs, D-loop formation, DNA polymerase mediated synthesis of the invading 3' strand, Exo1 protection of flaps/nicks, and ligation of cleaved dHJs. The transitions between these steps are likely to proceed through mechanisms that involve post-translational modifications (e.g. Bhagwat et al., 2021). Recent studies have shown that Exo1 has a key role in the activation of Mlh1-Mlh3 through Cdc5 Kinase , and a protein association/mass spectrometry study (Wild et al. Consistent with this, Mlh1-Mlh3 foci appear to form in meiotic prophase in the absence of Exo1  and RAD27 complementation of the exo1D crossover defect was not Gioia,Payero et al.,8/29/21 23 complete and did not improve crossover interference (Figure 4). One mechanism consistent with the above observations is that a DNA structure or protein barrier forms during meiotic recombination that activates the Mlh1-Mlh3 endonuclease, analogous to that seen for activation of Type I restriction enzymes through head-on collision of two translocating enzymes. (Szczelkun, 2002). Understanding how these transitions occur will require both in vitro reconstruction studies using purified proteins and novel in vivo approaches to identify nicks in dHJ intermediates.

A role for Exo1 in promoting genetic interference
In baker's yeast the ZMM factor Zip3 has been shown to be an early marker for crossover designation and interference, prior to the formation of physical crossovers, and previous work has suggested that crossover interference and crossover assurance are carried out as distinct functions by the ZMMs (Shinohara et al., 2008). These observations indicate that crossover interference is established prior to dHJ resolution (reviewed in Zhang et al., 2014). Interestingly, while mlh3∆ mutants lose dHJ resolution bias, residual interference in mlh3∆ mutants suggest that biased resolution is not required for interference. In contrast, a more severe loss of crossover interference in exo1∆ ( Figure 5) suggests a role beyond preserving resolution bias by protecting nicks, analogous to ZMM proteins which designate crossovers and assure interference on the maturing dHJ. The interference role for Exo1 was also reflected in spore viability patterning, as only the full exo1Δ displayed a viability pattern consistent with nondisjunction. While it is not possible to determine precisely how crossover patterning is disrupted in our exo1Δ data, the strong interference defect and clear non-disjunction pattern seen in exo1Δ strains is consistent with ZMM proteins that work early in imposing interference. The nature of this role remains unclear, as none of the exo1 alleles tested showed the interference defect seen in exo1Δ, and in fact some exo1 mutants showed increased interference. While Exo1 has been observed to interact with Msh2 through a Msh2-interacting-peptide (SHIP) box, direct interaction with Msh4-Msh5 has not been characterized (Goellner et al., 2018). A link Gioia,Payero et al.,8/29/21 24 between Exo1 and Msh4-Msh5 is also discouraged by the finding that Msh4-Msh5 localization is not dependent on Exo1 (Figure 6). This observation and previous work showing that joint molecule formation occurs at wild-type levels in exo1D mutants (Zakharyevich et al., 2010) suggest that the interference defect seen in exo1D mutants does not reflect the defective loading of Msh4-Msh5 to recombination intermediates.
Could the interference defect seen in exo1D mutants reflect a defect in resection of

MATERIALS AND METHODS
Exo1 homology model. The crystal structure of human Exo1 in complex with 5' recessed DNA (amino acids 2 to 356; Orans et. al., 2011) was used to map residues in yeast Exo1 critical for function. A homology model was constructed ( Figure 1B) using the Phyre2 software (http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id=index). The predicted structure was Gioia,Payero et al.,8/29/21 25 aligned to human Exo1 (PDB ID: 3QEB) using Pymol (https://pymol.org/2/). Metal binding residues mutated in this study were D78, D171, and D173. Active site residues mutated were H36, K85, R92, K121. Hydrophobic wedge residues mutated were S41, F58, and K61 and DNA binding residues mutated were K185 and G236. For Figure S1 the Exo1 protein sequence from S. cerevisiae was submitted to the BLASTP server at NCBI and run against the landmark database. Protein sequences of Exo1 homologs from different model organisms were analyzed and a multiple-sequence alignment was generated with MAFFT using default settings (Katoh et al., 2018).
Purification of Exo1. Exo1-FLAG variants (Exo1, exo1-D173A, exo1-G236D, exo1-D173A,G236D) were purified from pFastBac1 constructs (Table S6) in the baculovirus/Sf9 expression system as described by the manufacturer (Invitrogen) with the following modifications (Nicolette et al., 2010). Briefly, 250 ml of Sf9 cell pellet was resuspended in 7.5 mL of a buffer containing 50 mM Tris pH 7.9, 1 mM EDTA, 0.5 mM PMSF, 0.5 mM βmercaptoethanol, 20 μg/mL leupeptin, and 0.25x Halt protease inhibitor cocktail (Thermo). The suspension was incubated on ice for 15 min, after which NaCl was added to a final concentration of 100 mM and glycerol was added to a concentration of 18 % (v/v) and incubated on ice for 30 min. The cells were centrifuged at 30,000xg for 30 min. The cleared lysate was applied to a 2 mL SP Sepharose Fast Flow column at a rate of ~15 mL/hr. The column was washed with 10 mL of a buffer containing 50 mM Tris pH 7.9, 10 % glycerol, 100 mM NaCl, 0.5 mM PMSF, 5 mM β-mercaptoethanol, and 6.7 μg/mL leupeptin. Exo1 variant was eluted with the above buffer containing 700 mM NaCl. Fractions containing Exo1 protein variant were pooled and applied to 0.3 mL of M2 anti-FLAG agarose beads (Sigma) in batch, incubating with rotation for ~1.5 hours at 4 ºC. Unbound protein was isolated by centrifugation at 2,000 RPM for 5 min in a swinging bucket centrifuge at 4 ºC. The resin was resuspended in 7 mL of buffer containing 20 mM Tris pH 7.9, 150 mM NaCl, 10 % glycerol, 0.1 % NP40, 0.5 mM PMSF, 0.5 mM β-mercaptoethanol, 6.7 μg/mL leupeptin, and one-third of a Complete Protease Tablet (Roche) for every 100 mL of buffer and flowed into an empty column at ~15 ml/hr, allowing to pack. The column was then washed with 0.6 ml of the above buffer excluding the NP40 (wash buffer II). Exo1-FLAG variants were eluted using wash buffer II containing 0.1 mg/mL 3x-FLAG peptide (Sigma). After applying elution buffer, the flow was stopped after the first three fractions were collected and incubated for ~1 hr before resuming flow and collecting fractions. Fractions containing Exo1 variant were pooled, flash frozen in liquid nitrogen, and stored at -80 ºC. All purification steps were performed at 4 ºC. Protein concentration was determined by the method of Bradford (1976). Biolabs) and incubated at 37 ºC for 20 min. Products were resolved by 1.2% agarose gel containing 0.1 μg/mL ethidium bromide. Samples were prepared and gels were run as described previously . Gel quantifications were performed using GelEval (FrogDance Software, v1.37) using negative control reactions as background.
Strain constructions. Mutant alleles were transformed into S. cerevisiae with integration plasmids, geneXΔ::KANMX PCR fragments or on CEN6-ARSH4 and 2µ plasmids using standard techniques (Gietz et al., 1995;Rose et al., 1990). To confirm integration events, genomic DNA from transformants was isolated as described previously (Hoffman and Winston, 1987). Transformants bearing EXO1::KANMX and exo1::KANMX mutant derivatives were screened for integration by analyzing DNA fragments created by PCR using primers AO4061 and AO3838. Integration of exo1 alleles was confirmed by DNA sequencing of the DNA fragments created by PCR using primers AO3666 and AO3399 (Table S7). To confirm integration of geneXD::KANMX mutations, primers that map outside of the geneXΔ::KANMX PCR fragment were used (Table S7). At least two independent transformants for each genotype were made. exo1 integrating and EXO1, RAD27 and CDC9 expression plasmids.
Plasmids created in this study are shown in Table S6 and the oligonucleotide primers used to make plasmids are shown in Table S7. Genes expressed in plasmids are from the SK1 strain background (Kane and Roth, 1974 promoter region (400 bp immediately upstream ATG) amplified from the SK1 genome using AO4643 + AO4644. 3. The entire RAD27 ORF amplified from the SK1 genomic DNA using AO4645 + AO4637. 4. The EXO1 downstream region (400 bp immediately downstream of the stop codon) amplified from the SK1 genomic DNA using AO4638 + AO4636. rad27 mutant alleles were constructed with the Q5 mutagenesis kit (New England Biolabs) using pEAA720 as template The oligonucleotides used to make the alleles are shown in Table S7). All RAD27 plasmid constructs were confirmed by DNA sequencing. pEAM327 (9.3 KB), a CDC9, 2µ, URA3 plasmid, was constructed in two steps. First a fragment of the CDC9 ORF, containing 1000 bp upstream and 400 bp downstream sequence was created by PCR amplification of SK1 genomic DNA using primers AO4783 and AO4784. The resulting fragment was digested with HindIII and KpnI and then ligated to pRS426 (2µ, URA3) backbone also digested with HindIII and KpnI to create pEAM327. pEAM329 (8.8 KB) is a 2µ, URA3 plasmid that expresses CDC9 from the HOP1 promoter (pHOP1-CDC9). It was constructed through Hifi assembly using the following fragments: 1. A DNA backbone was created by PCR amplification of pEAM327 using primers AO4837 and AO4838; the resulting DNA fragment lacks the CDC9 promoter. 2. A 500 bp DNA fragment of the HOP1 promoter (up until the HOP1 start codon) was created by PCR amplification of SK1 genomic DNA using primers AO4839 and AO4840. The two fragments were then assembled using Hifi Assembly to create pEAM329, which was confirmed by DNA sequencing.
Tetrad analysis. Diploids derived from EAY1108/EAY1112 were sporulated using the zerogrowth mating protocol (Argueso et al., 2003). Briefly, haploid parental strains were patched together, allowed to mate overnight on complete minimal plates, and then struck onto selection Gioia,Payero et al.,8/29/21 30 plates to select for diploids. The resulting diploids were then transferred from single colonies to sporulation plates where they were incubated at 30 o C for 3 days. Tetrads were dissected on minimal complete plates and then incubated at 30 o C for 3-4 days. Spore clones were replicaplated onto relevant selective plates and assessed for growth after an overnight incubation.
Interference was measured by the Malkova method (Malkova et al., 2004). This method measures cM distances in the presence and absence of a neighboring crossover. The ratio of these two distances denotes the strength of interference, with a value closer to 1 indicating a loss of interference. Significance in the distribution of tetrads was measured using a G test Spore-autonomous fluorescence assay. We analyzed crossover events between sporeautonomous fluorescence reporter constructs at the CEN8-THR1 locus on Chromosome VIII (SKY3576, SKY3575; Thacker et al., 2011). To produce diploid strains for analysis in the spore autonomous fluorescence assay, haploid yeasts of opposite mating types were mated by patching together on YPD from freshly streaked colonies and allowed to mate for 4 hrs, and then transferred to tryptophan and leucine dropout minimal media plates to select for diploids.
Diploids grown from single colonies were patched onto sporulation plates and incubated at 30°C for approximately 72 hours. Diploid strains containing ARS-CEN or 2µ plasmids were also grown on selective media to maintain the plasmids until just prior to patching onto sporulation Gioia,Payero et al.,8/29/21 31 plates. Spores were treated with 0.5% NP40 and briefly sonicated before analysis using the Zeiss AxioImager.M2. At least 500 tetrads for each genotype were counted to determine the % tetratype. Two independent transformants were measured per allele. A statistically significant difference from wild-type and exo1Δ controls based on χ2 analysis was used to classify each allele as exhibiting a wild-type, intermediate, or null phenotype. We applied a Benjamini-Hochberg correction at a 5% false discovery rate to minimize α inflation due to multiple comparisons.
Sensitivity to methyl-methane sulfonate. Yeast strains were grown to saturation in YPD liquid media, after which they diluted in water and spotted in 10-fold serial dilutions (undiluted to 10 -5 ) onto YPD media containing 0.04% MMS (v/v; Sigma). Plates were photographed after a 2day incubation at 30 o C.
Haploinsufficiency screen. We created knockout transformation PCR fragments consisting of a KANMX4 antibiotic resistance marker flanked by 300 bp of upstream and downstream homology with respect to the open reading frame (ORF) of each gene of interest. These cassettes were amplified by PCR from genomic preps of the appropriate strains from the Saccharomyces genome deletion project (Giaever et al., 2014). In this collection, each ORF has been replaced with KANMX4.
EAY3486 (Table S5), a 5$"BG strain carrying a gene encoding a cyan fluorescent protein (CFP) on chromosome VIII, was transformed with the PCR amplified knockout cassette.
Cells were then plated on YPD-G418 plates and grown at 30ºC for three days. At least two independent transformants were verified by confirming resistance to G418 and PCR amplification of using genomic preps of G418 resistant transformants. For PCR verification, primers annealing 350 bp upstream and downstream of the ORF of the gene of interest were utilized to ensure integration at the proper locus. Haploids were then mated to four MLH3 strains each carrying a gene encoding a red fluorescent protein (RFP) on chromosome VIII. These four strains are as follows: EAY3252 (MLH3), EAY3255 (5$"BG), EAY3572 (mlh3-42), and EAY3596 (mlh3-54). Diploids were isolated by selecting on media lacking tryptophan and leucine and analyzed in the spore-autonomous fluorescence assay described below.
Our criteria for allele-specific interactions was one in which there was little to no change in percent tetratype in either an MLH3 and 5$"BG&background, but there was a significant drop of percent tetratype in either mlh3-42 or mlh3-54 backgrounds. Significance was assessed by χ 2 test between haplosufficient and haploinsufficient conditions. To minimize inflation due to multiple comparisons, we applied a Benjamini-Hochberg correction at a 5% false discovery rate (Benjamini and Hochberg, 1995). Chromosome spreads (3h, 4h and 5h) were prepared from synchronized meiotic cultures (3, 4 and 5hr) as described (Bishop, 1994;Shinohara et al., 2008;Challa et al., 2019). Msh5 staining was performed using primary antibody against Msh5 (Shinohara et al., 2008) at 1:500 dilution, followed by secondary antibody (Alexa fluor 488, Thermo Fisher Scientific) at 1:1500 dilution.
The Msh5 stained samples were imaged using an epi-fluorescence microscope (BX51, Olympus) with a 100X objective (NA,1.3). Images were captured by the CCD camera (CoolSNAP, Roper) processed using iVision (Sillicon) software. To quantify Msh5 focus intensity, the mean fluorescence of a whole nucleus was quantified with Fiji (ImageJ). The final fluorescence intensity of Msh5 was normalized with DAPI intensity for each nucleus.
Asterisks indicate the number of genetic intervals that are distinguishable from the exo1D containing the empty vector, as measured using standard error calculated through Stahl Laboratory Online Tools (https://elizabethhousworth.com/StahlLabOnlineTools/; Table S2). C. mlh3Δ and the indicated exo1 strains were transformed with pEXO1-RAD27 (pEAA720), pEXO1-rad27-D179A (pEAA724) and empty vector (pRS416), and examined for crossing over at the CEN8-THR1 locus. Significance (*p<0.05) compared to the exo1D strain containing an empty vector (panel A) was determined using a two-tailed Fisher's Exact Test. D. CDC9 Gioia,Payero et al.,8/29/21 46 overexpression in meiosis disrupts the crossover functions of exo1 DNA binding mutants.
Significance is shown between each empty vector-pHOP1-CDC9 pair using a two-tailed Fisher's Exact Test, with ** indicating p<0.01.   model, Exo1 protects nicks made by nick translation (resolution independent nicks) and recruits Mlh1-Mlh3 as in panel A. C. dHJ resolution through extended branch migration (Ahuja et al., 2021). Branch migration creates a substrate for Mlh1-Mlh3 polymerization . In such a model, the signaling imposed by the binding of Exo1 to nicks acts at a distance. Mlh1-Mlh3 is recruited by Exo1 and forms a polymer with a specific polarity that can displace other factors or be activated upon interaction with such factors. The polymer is activated to introduce a nick on one strand of the duplex DNA on Type II dHJs when it forms a critical length required for stability. See text for details.         decreases in crossover frequencies that were not mlh3 alleles-specific, and haploinsufficiency of SPO16 did not affect CO frequency. Crossing over was also measured in the 20 cM CEN8 to THR1 interval. Significance was assessed by χ 2 test between haplosufficient and haploinsufficient conditions. To minimize inflation due to multiple comparisons, we applied a Benjamini-Hochberg correction at a 5% false discovery rate.  YPD (   Homozygous mutations were made by crossing two independently constructed strains with the exo1 variants in the SKY3576 (containing cyan fluorescent protein; Table S5) and SKY3575 (containing red fluorescent protein) backgrounds. Heterozygous mutations were made by crossing two independently constructed strains with exo1 variants in the SKY3576 and EAY4151 (exo1Δ) backgrounds. Diploid strains were induced for meiosis and % tetratype in the CEN8-THR1 interval was measured, by determining the total tetratypes/sum of tetratypes and parental ditypes). At least 500 tetrads were counted for each allele, and unless indicated (*one transformant analyzed), at least two transformants were analyzed for each background. Significance was assessed by Fisher's exact test between mutant and wild-type EXO1 and exo1Δ tetratype values. To minimize inflation due to multiple comparisons, we applied a Benjamini-Hochberg correction at a 5% false discovery rate. +, indistinguishable from wild-type; -, indistinguishable from exo1D; INT, distinguishable from both wild-type and exo1D. Diploids of the indicated genotype that contain markers to measure crossing over in the CEN8-THR1 interval (Table S5) were transformed with the indicated plasmids (pEAA715-EXO1, URA3, CEN6-ARSH4; pRS416-URA3,CEN6-ARSH4; pEAA722-RAD27, URA3, CEN6-ARSH4; pEAA720-pEXO1-RAD27, URA3, CEN6-ARSH4; pEAA724-pEXO1-rad27-D179A, URA3, CEN6-ARSH4; pEAA727-rad27-A45E, URA3, CEN6-ARSH4; pEAA728-rad27-R101A, URA3, CEN6-ARSH4; pEAA729-rad27-R105A, URA3, CEN6-ARSH4; pEAA730-rad27-K130A, URA3, CEN6-ARSH4; pEAA731-rad27-H191E, URA3, CEN6-ARSH4) and selected for plasmid retention. The resulting strains were induced for meiosis and % tetratype (single crossovers) in the CEN8-THR1 interval was measured, by determining the total tetratypes/sum of tetratypes and parental ditypes. At least 500 tetrads were counted for each allele/plasmid combination, and at least two transformants were analyzed for each condition. Significance (presented in Figure 4A, C) was assessed by Fisher's Exact Test between exo1Δ strains containing pRS416 (empty vector) and test conditions with the indicated plasmids. To minimize inflation due to multiple comparisons, we applied a Benjamini-Hochberg correction at a 5% false discovery rate. The significance of % tetratype in exo1-K185E and exo1-F447A,F448A (MIP) strains containing pRS416 (empty vector) and pEAA720 (pEXO1-RAD27) was determined using Fisher's exact test. N/A, not applicable. Diploids of the annotated genotype were transformed with the indicated plasmid (pRS426-URA3, 2µ; pEAM329-pHOP1-CDC9, URA3, 2µ) and selected for diploidy and plasmid retention. Diploid strains were induced for meiosis and % Tetratype in the CEN8-THR1 interval was measured by determining the total tetratypes/sum of tetratypes and parental ditypes. At least 500 tetrads were counted for each allele/plasmid combination, and at least two transformants were analyzed for each condition. Significance was assessed by Fisher's exact test between pRS426 value and pEAM329 value and is shown in Figure 4D.
The Malkova ratio and coefficient of coincidence (COC, ratio of double crossovers observed/double crossovers expected) were performed for the indicated genotypes in the EAY1108/EAY1112 strain background (Materials and Methods, strains listed in Table S5). These methods were performed for intervals I (URA3-LEU2-LYS2), II (LEU2-LYS2-ADE2), and III (LYS2-ADE2-HIS3). 0 = Absolute Interference; 1= No interference. Significance of differences in tetrad distribution was assessed using a G test. Differences in distribution with p<0.05 were considered to be significant evidence of interference. Intervals with ratios significantly above 1 were observed and denoted with * to indicate potential negative interference. Detailed analysis of the Malkova ratio calculation is presented in Table S3B. Table S3B. Detailed calculations of Malkova ratios presented in Figure 5 and Table S3A.
Legend, Table S3B. Crossover interference was analyzed using the Malkova method (Malkova et al., 2004;Martini et al., 2006) for chromosome XV. For each genetic interval, tetrads were divided based on the presence or absence of a recombination event in a reference interval. For each reference interval, the map distance was measured in the adjacent intervals, thus obtaining two map distances for each interval. The significance of differences in tetrad distribution was assessed using a G test. Differences in distribution, with p<0.05, were considered to be evidence of interference. The data are presented as the average ratio of the two map distances in each neighboring interval, with a smaller ratio indicating stronger interference. An interval was considered to have a "loss of positive interference" phenotype when both adjacent intervals displayed no detectable positive interference. Ratios significantly greater than 1 are indicated with * to denote potential negative interference. TT, tetratype; NPD. nonparental ditype; PD, parental ditype.  (Table S5) containing the THR1::m-Cerulean-TRP1 and CEN8::tdTomato-LEU2 markers on chromosome VIII were induced for meiosis and % tetratype in the CEN8-THR1 interval was measured by determining the total tetratypes/sum of tetratypes and parental ditypes). At least two transformants were analyzed for each background. Significance was assessed by χ 2 test between mutant and wild-type EXO1 and exo1Δ tetratype values. To minimize inflation due to multiple comparisons, we applied a Benjamini-Hochberg correction at a 5% false discovery rate. +, indistinguishable from WT; -, indistinguishable from exo1D; +/-, distinguishable from both wild-type and exo1D. pEAE422 amp R , Gm R , exo1-G236D-FLAG exo1-G236D expression from pFastBac pEAE423 amp R , Gm R , exo1-D173A-G236D-FLAG exo1-D173A,G236D expression from pFastBac AO3888 ATTCCATTTGTATAGTCACAACC exo1 mutagenesis, G236D