Meiotic DNA break resection and recombination rely on chromatin remodeler Fun30

Abstract

DNA double-strand breaks (DSBs) are nucleolytically processed to generate single-stranded DNA tails for homologous recombination. In Saccharomyces cerevisiae meiosis, this 5’-to-3’ resection involves initial nicking by the Mre11–Rad50–Xrs2 complex (MRX) plus Sae2, then exonucleolytic digestion by Exo1. Chromatin remodeling adjacent to meiotic DSBs is thought to be necessary for resection, but the relevant remodeling activity was unknown. Here we show that the SWI/SNF-like ATPase Fun30 plays a major, non-redundant role in resecting meiotic DSBs. A fun30 null mutation shortened resection tract lengths almost as severely as an exo1-nd (nuclease-dead) mutation, and resection was further shortened in the fun30 exo1-nd double mutant. Fun30 associates with chromatin in response to meiotic DSBs, and the constitutive positioning of nucleosomes governs resection endpoint locations in the absence of Fun30. We infer that Fun30 directly promotes both the MRX- and Exo1-dependent steps in resection, possibly by removing nucleosomes from broken chromatids. Moreover, we found that the extremely short resection in the fun30 exo1-nd double mutant is accompanied by compromised interhomolog recombination bias, leading to defects in recombination and chromosome segregation. Thus, this study also provides insight about the minimal resection lengths needed for robust recombination.

Introduction

During meiosis, recombination yields crossovers between homologous chromosomes, providing physical connections important for proper segregation in the first division (Hunter 2015). Meiotic recombination is initiated by DSBs catalyzed by Spo11 in a topoisomerase-like reaction that leaves Spo11 covalently linked to the newly created 5’ DNA termini (Bergerat et al. 1997; Keeney et al. 1997; Neale et al. 2005) (Figure 1A).

Figure 1: Testing chromatin remodeler mutants for shortened meiotic resection.

(A) Left, overview of meiotic DSB formation and resection within the context of local chromatin structure. Right, schematic of S1-Southern blotting and S1-seq methods. RE: restriction enzyme.

(B) Resection endpoint distributions detected by S1-Southern blotting at the GAT1 (top) and CCT6 (bottom) hotspots in the indicated mutants. All samples were collected at 4 h in meiosis. Vertical black lines to the left of the gene maps indicate probe positions. P, parental length restriction fragments.

DSB ends are then resected in two steps (Mimitou and Symington 2009) (Figure 1A). Endonucleolytic cleavage of each Spo11-linked DNA strand by MRX/Sae2 provides entry sites for two exonucleases with opposite polarities: 3’-to-5’ Mre11 exonuclease activity (digesting toward Spo11) and 5’-to-3’ Exo1 exonuclease activity (digesting away from the DSB) (Neale et al. 2005; Zakharyevich et al. 2010; Garcia et al. 2011; Cannavo and Cejka 2014; Mimitou et al. 2017). Consequently, Spo11 proteins at DSB ends are released with covalently linked oligonucleotides (Spo11 oligos) (Neale et al. 2005). The resulting 3’ single-stranded DNA (ssDNA) tails, which average ~800 nt in length, are bound by strand-exchange proteins Rad51 and Dmc1, setting recombination in motion (Brown and Bishop 2014).

Most DSBs in S. cerevisiae occur within hotspots that usually span < 200 bp and mostly correspond to the nucleosome-depleted regions (NDRs) at transcription promoters (Ohta et al. 1994; Wu and Lichten 1994; Baudat and Nicolas 1997; Pan et al. 2011). Because these NDRs are typically flanked by nucleosome arrays, the resection nucleases must traverse several nucleosomes’ worth of DNA (Figure 1A). However, while Exo1 is a processive enzyme in vitro with an average run length of ~6 kb on naked DNA (Myler et al. 2016; Myler and Finkelstein 2017), it is strongly blocked by nucleosomes (Adkins et al. 2013). Moreover, MRX/Sae2 preferentially nicks near but not within nucleosomes in vitro (Wang et al. 2017) and nucleosomes impede MRX/Sae2 incision in vegetative cells (Gnugge et al. 2023). Therefore, the resection nucleases may require chromatin remodeling to be able to digest nucleosomal DNA in vivo (Figure 1A). Multiple chromatin remodelers appear to play partially overlapping roles in resection of DSBs in vegetative cells, including RSC, INO80, and Fun30 (Karl et al. 2021), but whether the same is true in meiosis has not yet been established.

We previously mapped resection endpoints genome wide by digesting ssDNA with S1 endonuclease followed by adapter ligation and deep sequencing (S1-seq) (Mimitou et al. 2017; Mimitou and Keeney 2018; Yamada et al. 2020). We found that population averages of resection endpoints are modestly enriched at the preferred positions of linkers between nucleosomes and that altered local chromatin structure (caused by elimination of transcription factors Bas1 or Ino4) was associated with changes in resection tracts (Mimitou et al. 2017). Moreover, computational simulations suggested that Exo1 digestion rates in vivo are comparable to rates in vitro on naked DNA (Mimitou et al. 2017). These findings suggested a scenario in which Exo1 runs on DNA duplexes that are effectively nucleosome-free, then terminates digestion upon encountering the first intact nucleosome (Figure 1A). This model reinforced the idea that nucleosomes may be actively removed or destabilized from broken chromatids before Exo1-mediated resection.

Here, we investigate roles of chromatin remodelers in meiotic resection. We demonstrate that Fun30 is locally recruited in response to meiotic DSB formation and strongly influences the spatial patterns of both MRX/Sae2 nicking and further Exo1 processing. We further leverage the greatly shortened resection tracts in the fun30 exo1-nd double mutant to explore the functional importance of normal ssDNA tail lengths for recombination partner choice, crossover formation, and production of viable spores.

Results

Efficient resection requires Fun30

To identify chromatin remodelers with roles in meiotic resection, we screened candidate factors that have known effects on resection in vegetative cells. We tested mutations eliminating Fun30, a SWI/SNF-like ATPase that promotes resection at long distances (> 5 kb away from the break site) (Chen et al. 2012; Costelloe et al. 2012; Eapen et al. 2012); Arp8, a non-essential subunit of the INO80 chromatin remodeler (Morrison et al. 2004; Tsukuda et al. 2005; Gospodinov et al. 2011); and Swr1, the ATPase subunit of the SWR-C complex (van Attikum et al. 2004; Papamichos-Chronakis et al. 2006). We also tested a mutant lacking histone variant H2A.Z (htz1Δ) because of reported effects of H2A.Z on resection in vivo (van Attikum et al. 2007; Lademann et al. 2017) and on Exo1 digestion of nucleosomes in vitro (Adkins et al. 2013).

We examined resection at two strong DSB hotspots (CCT6 and GAT1) using S1-Southern blots, in which genomic DNA from meiotic cultures was digested with S1 endonuclease to remove ssDNA then visualized by Southern blotting and indirect end labeling after agarose gel electrophoresis (Figures 1A,B). Wild-type cells displayed a ladder of blunted DSB fragments that migrated considerably faster on the gel than the unresected DSBs seen in a sae2Δ mutant (Figure 1B, compare lanes 1 and 15 to lane 19). The banding pattern in wild type reflects at least in part the effect of chromatin structure on the positions of resection endpoints (Mimitou et al. 2017). As previously shown (Mimitou et al. 2017), resection lengths are shorter in nuclease-dead exo1-nd mutants (exo1-D173A; (Tran et al. 2002)), so S1-digested DSB fragments migrated more slowly than in wild type (Figure 1B, lanes 6,14,17).

The fun30Δ mutant similarly exhibited slower-migrating S1-treated fragments than wild type, with the magnitude of the change comparable to but distinct from that seen in exo1-nd (Figure 1B, lanes 5 and 16). In contrast, none of the other mutations tested gave shortened resection lengths, either alone or in combination with one another, and none of them caused further changes in resection in a fun30Δ background (Figure 1B, panels i and ii). We did not detect any difference in DSB distributions between FUN30 and fun30Δ in the sae2Δ background (Figure 1B.iv), ruling out the possibility that fun30Δ affects resection endpoints only indirectly by altering DSB locations. We conclude that Fun30 is required to generate resection tracts of normal length.

DSB fragments in a fun30Δ exo1-nd double mutant migrated even more slowly than either the fun30Δ or exo1-nd single mutants (Figure 1B.iii). The resection seen in an exo1-nd background reflects the distribution of the MRX/Sae2 nicks that are the furthest from each DSB. Therefore, seeing shorter resection tracts in the fun30Δ exo1-nd double mutant compared to exo1-nd alone indicates that Fun30 also influences where MRX/Sae2 can act.

Fun30 shapes the global resection landscape

To interrogate the genome-wide distribution of resection endpoints, we performed S1-seq on cells collected at 4 h after meiotic induction in wild type, exo1-nd, fun30Δ, and fun30Δ exo1-nd. Biological replicates correlated well (Figure S1A), so they were averaged for further analyses. As previously shown (Mimitou et al. 2017), S1-seq reads at hotspots mapped to the top strand for resection moving rightward and to the bottom strand for leftward resection, and these S1-seq signals shifted closer to hotspot centers in exo1-nd (Figure 2A). Consistent with the S1-Southern results, S1-seq signals were also closer to the hotspot center in fun30Δ and even closer in fun30Δ exo1-nd (Figure 2A).

Figure 2: The meiotic resection landscape in fun30Δ mutants.

(A) S1-seq signals around the CCT6 hotspot in reads per million mapped (RPM). Reads mapping to the top strand are shown in blue; bottom strand reads are in red. All S1-seq samples were collected at 4 h in meiosis. Spo11-oligo data are from (Pan et al. 2011).

(B) Genome average of S1-seq signal around 3908 hotspots. Bottom strand reads were reoriented and combined with the top strand to calculate the average. Data were smoothed with a 100-bp Hann window.

(C) Histograms of resection tract lengths calculated for “loner” hotspots that had no other hotspot within 3 kb (n = 405). Lighter colored bars indicate tracks that were omitted to calculate the censored median estimates shown in parentheses. Censoring had little effect, indicating that the measurements are not strongly influenced by outliers.

(D) Schematic comparing FUN30 and fun30Δ for the distance to the most distal MRX/Sae2 nicking positions (as measured in exo1-nd mutants) and the inferred lengths of Exo1 digestion.

Global resection patterns were displayed by co-orienting and combining top- and bottom-strand reads around 3908 previously defined hotspots (Mohibullah and Keeney 2017) and plotting the average relative to hotspot centers (Figure 2B). Both single mutants had distributions shifted towards the hotspot midpoints, but with notable differences: exo1-nd showed substantially more signal than fun30Δ within 200 nt of hotspot midpoints, while fun30Δ showed a modestly higher frequency of longer resection tracts (~700–1200 nt) (Figure 2B). These differences indicate that, in the absence of Fun30, the combined action of MRX/Sae2 and Exo1 resects nearly all DSBs at least 200 nt and occasionally resects almost as extensively as the longer tracts in wild type. In contrast, MRX/Sae2 aided by Fun30 but without Exo1 nuclease often resects for only very short distances (< 200 nt). The fun30Δ exo1-nd double mutant had even shorter resection tracts, with a modal endpoint within 100 nt of hotspot midpoints (Figure 2B).

We estimated median resection tract lengths by focusing on 405 narrow (<400 bp wide) “loner” hotspots that had no other hotpots within 3 kb. This analysis gave median resection lengths of 755 nt in wild type and 354 nt in exo1-nd (Figure 2C), comparable to previous results (Mimitou et al. 2017). Median resection was shortened to 439 nt in the fun30Δ mutant, and to 130 nt in the fun30Δ exo1-nd double mutant (Figure 2C).

The difference between exo1-nd and fun30Δ exo1-nd reinforces the inference from S1-Southern data that the fun30Δ mutation alters MRX/Sae2 nicking positions. Importantly, however, the S1-seq data also suggest that Fun30 promotes Exo1-dependent resection as well. If we estimate the contribution of Exo1 to total resection as the difference between EXO1+ and exo1-nd backgrounds, then Exo1 resects further in the presence of Fun30 (median^{wild type} – median^exo1-nd = 401 nt) than when Fun30 is missing (median^fun30Δ – median^{fun30Δ exo1-nd} = 309 nt) (Figure 2D). We conclude that Fun30 affects both resection steps.

S1-seq in a sae2Δ background, which maps DSB locations (Mimitou et al. 2017; Mimitou and Keeney 2018), confirmed that the fun30Δ mutation had little if any effect on either the frequency or distribution of DSBs within hotspots (Figures S1B,C). Thus, Fun30 does not affect where Spo11 cleaves but instead influences resection per se. S1-seq further confirmed that Swr1 is dispensable for resection genome-wide (Figures S1D-F). We also performed S1-seq on cells lacking Rad9, which has an inhibitory effect on resection in vegetatively growing yeast (Chen et al. 2012), but detected little or no difference from wild type (Figures S1D-F).

Fun30 is recruited to chromatin in response to meiotic DSBs

To test if Fun30 is recruited to sites of meiotic DSBs, we performed chromatin immunoprecipitation followed by deep sequencing (ChIP-seq) of Fun30-myc in SPO11 and spo11-yf strains (Y135F; catalytically inactive mutant). The methods we used would not generate sequencing reads from the ssDNA left after resection. Moreover, we reasoned that any direct role of Fun30 in remodeling the chromatin of broken chromatids should occur prior to the onset of resection and that resection might remove proteins that were bound to the DNA at the time of nuclease action. Therefore, we conducted these experiments in a sae2Δ background to prevent resection. To quantitatively compare ChIP-seq signals between datasets, we used Saccharomyces mikatae cells expressing myc-tagged Rec114 as a spike-in control.

Both SPO11 and spo11-yf strains showed pan-genomic Fun30 ChIP-seq signals that were on average more than 16-fold above an untagged control (Figures 3A and S2A). In both strains, Fun30 was enriched in genomic regions where its preferential binding has been previously reported, including tRNA genes, replication origins (ARS), and centromeres (Neves-Costa et al. 2009; Durand-Dubief et al. 2012) (Figures S2B,C). The DSB-independent component of this basal chromatin association presumably reflects constitutive roles of Fun30 in chromatin modulation throughout the genome.

Figure 3: DSB-dependent recruitment of Fun30 to chromatin.

(A–C) Fun30-myc ChIP-seq samples were collected at 4 h in meiosis. The DSB map (S1-seq in a sae2Δ mutant) is the same dataset shown in Figures S1A,B as sae2Δ #2. The Rec114-myc ChIP-seq sample collected at 4 h in meiosis is from a previous study (Murakami and Keeney 2014). Fun30 ChIP-seq datasets were normalized using a spike-in control. All data were smoothed using a 1 kb Parzen (triangular) sliding window.

(A) Fun30 ChIP-seq signals across a representative region of chromosome III. The upper graph shows the normalized ChIP-seq coverage profiles for wild type, the DSB-defective spo11-yf mutant, or a wild-type strain carrying untagged Fun30. The lower graph shows the DSB-dependent component of the Fun30 ChIP-seq signal, obtained by subtracting the normalized values from the spo11-yf strain from the values from the tagged wild-type strain. DSBs and Rec114 ChIP-seq profiles are shown for comparison.

(B) Average Fun30 ChIP-seq signals around previously defined DSB hotspots (Mohibullah and Keeney 2017) and Rec114 peaks (Murakami and Keeney 2014). Profiles around a set of random genomic positions are shown for comparison.

(C) Correlations (Pearson’s r) between the DSB-dependent Fun30 ChIP-seq signals (summed in 1-kb windows) and both DSBs (left graph) or Rec114 ChIP-seq (right graph).

Importantly, there was also pronounced DSB-stimulated Fun30 enrichment at hotspots and other genomic regions. To investigate this DSB-dependent component of Fun30 binding, we plotted the difference in calibrated ChIP-seq signal between the SPO11 and spo11-yf data sets (Figure 3A bottom). We observed peaks that coincided with DSB sites (i.e., peaks in the sae2Δ S1-seq map), but also peaks that lined up with peaks in ChIP-seq maps for the Spo11-accessory factor Rec114. The Rec114 peaks are thought to be preferred sites of assembly of DSB-promoting machinery along chromosome axes (Panizza et al. 2011; Murakami and Keeney 2014).

The average Fun30 ChIP-seq signal around DSB hotspots showed a peak in SPO11 but a modest depression in spo11-yf, so the averaged difference map had a broad peak extending ~500 bp to each side of hotspot centers (Figure 3B). No such enrichment was observed for a random sample of genomic loci (Figures 3B and S2C). Furthermore, the local DSB-dependent ChIP-seq signal was correlated with hotspot strength (Figure 3C), supporting the conclusion that Fun30 is recruited in cis to sites where DSBs have occurred.

Unlike around hotspots, the Fun30 ChIP-seq signal was already enriched in the absence of DSBs (spo11-yf) at preferred Rec114 binding sites (n = 2010, (Murakami and Keeney 2014)), and became further enriched in the SPO11 strain (Figures 3B and S2C). The DSB-dependent component of the Fun30 ChIP-seq signal was correlated with the Rec114 ChIP-seq signal (Figure 3C). Similarly, the other genomic features that showed basal enrichment of Fun30 (tRNA genes, ARSs, centromeres) also showed further increases in ChIP-seq signal in response to DSBs (Figures S2B,C), which appears to be at least partly explained by the proximity of these features to Rec114 peaks (Figure S2D). Implications of DSB-stimulated Fun30 association with these sites away from hotspots are addressed in the Discussion.

Fun30 remodels chromatin to facilitate efficient resection

In vitro, Fun30 slides nucleosomes and can exchange histone proteins in reconstituted nucleosomes (Awad et al. 2010; Byeon et al. 2013; Markert et al. 2021). In vivo, the fun30Δ mutation modestly alters nucleosome occupancies at centromeres and around promoters (Durand-Dubief et al. 2012; Byeon et al. 2013). Because Fun30 has chromatin remodeling activities, we sought evidence that its function in meiosis involves promoting the ability of the resection nucleases to act on nucleosomal substrates.

The NDRs of many yeast promoters are flanked by positioned nucleosomes, with the transcription start site (TSS) in the first (+1) nucleosome (Jansen and Verstrepen 2011). To test if resection endpoints correlate spatially with preferred positions of nucleosomes, we averaged S1-seq signals around hotspots, using as a reference point the midpoints of +3 nucleosomes for wild type or +1 nucleosomes for the shorter resection tracts in the mutants (Figure 4A). We then compared these population-average resection profiles to population-average chromatin structure as measured by sequencing of mononucleosomes released by digestion with micrococcal nuclease (MNase-seq; (Pan et al. 2011)).

Figure 4: S1-seq distribution relative to nucleosomes.

(A) Average S1-seq resection signal (colored lines; 41-bp smoothed) and MNase-seq (gray filled; 0 h in meiosis (Pan et al. 2011)) centered on midpoints of +3 or +1 nucleosomes with no other hotspots ≤3 kb downstream (n = 1815 for +3 nucleosomes and 1832 for +1 nucleosomes).

(B) Tighter view of the exo1-nd fun30Δ resection profile from panel A. The boxplot below shows that substantial S1-seq signals overlap with the MNase signals within the +1 nucleosome. The total S1-seq reads in the +1 nucleosome (−73 to +73 bp relative to the midpoint) and NDR (−219 to −74) regions were taken as 100%, and the percentages of S1-seq in the four regions are shown in light blue numerals. Boxplots are shown as in Figure S2C.

(C,D) Resection tract lengths in the absence of Fun30 reflect basal nucleosome occupancy. Hotspots were grouped according to resection patterns using k-means clustering applied to spatial distributions of S1-seq signals in fun30Δ (panel C) or fun30Δ exo1-nd (panel D). We used hotspots that were separated from their nearest neighboring hotspot by >1 kb and considered each side of each hotspot separately (2778 hotspots and 5556 hotspot sides). The S1-seq and premeiotic (t = 0 h) MNase-seq maps were averaged within each cluster and plotted as a function of distance from the hotspot midpoint. The number of hotspot sides in each cluster (n) and median resection lengths are indicated.

As previously shown (Mimitou et al. 2017), the averaged S1-seq signal in wild type was broadly distributed with a maximum between the +3 and +4 nucleosomes, while exo1-nd showed two peaks overlapping the +1 nucleosome and the linker between the +1 and +2 nucleosomes (Figure 4A). Both profiles also showed modest scalloping in register with the edges of preferred nucleosome positions.

In contrast, the fun30Δ S1-seq showed a comparatively narrower distribution, with a sharp, scalloped peak between the +1 and +2 nucleosomes (Figure 4A). The fun30Δ exo1-nd distribution was narrower still, forming a single smooth peak centered to the left of the MNase-seq peak for the +1 nucleosome and straddling the right edge of the NDR and the left edge of the +1 nucleosome (Figures 4A,B). The S1-seq signal overlapped substantially with the +1 nucleosome position but tapered off sharply about halfway through (Figures 4B and S3). This pattern implies either that Mre11 can nick readily within the first half of the nucleosome, or that the +1 nucleosome is moved or destabilized in a Fun30-independent fashion after DSB formation. The results also show that nicking by Mre11 cannot spread readily beyond the first nucleosome in the absence of Fun30. Thus, in the absence of Fun30 there appears to be an even stronger correlation between chromatin structure and the distribution of resection endpoints.

To test this correlation further, we used k-means clustering to divide hotspots into groups that differed according to resection tract lengths in fun30Δ (Figure 4C) or fun30Δ exo1-nd (Figure 4D). We reasoned that, if Fun30 normally mitigates nucleosomal barriers to resection, then these clusters should also have systematic differences in their average MNase-seq maps. Specifically, places where resection tends to go further even in the absence of Fun30 should be enriched for locations that intrinsically have a lower basal nucleosome occupancy, while places that tend to have shorter resection lengths should be those with higher nucleosome occupancy.

This analysis gave the predicted patterns. The cluster with the shortest resection tracts in fun30Δ had an average MNase-seq profile with pronounced peaks and high sequence coverage, indicating a tendency toward there being positioned nucleosomes of comparatively high occupancy (Figure 4C, left). Conversely, the cluster with the longest resection tracts had markedly lower MNase-seq coverage that formed less well-defined peaks, indicating a tendency toward a more open chromatin structure with less regularly positioned nucleosomes and lower occupancy (Figure 4C, right). Interestingly, the intermediate cluster had its highest MNase-seq coverage at 501 bp from the hotspot centers, just downstream of where most resection tracts in this cluster terminated (Figure 4C, middle). These findings reinforce the correlation between Fun30-independent resection and constitutively open chromatin.

Similar results were observed for MRX/Sae2-mediated resection in the fun30Δ exo1-nd mutant. The cluster with the shortest resection tracts had well defined peaks of high MNase-seq signal immediately flanking the hotspot NDR (Figure 4D, left), whereas the cluster with longer resection tracts had more poorly defined and lower coverage MNase-seq peaks, consistent with a more open and less regular chromatin structure (Figure 4D, right). These observations further support that Fun30 helps the resection machinery overcome the inhibitory effects of nucleosomes.

This exercise also provided insight into the ability of Exo1 to resect nucleosomal DNA in the absence of Fun30. The narrow distribution of resection endpoints in clusters 1 and 2 in the fun30Δ exo1-nd mutant (Figure 4D, left and middle) implies that nucleosomes are maintained in the absence of Fun30 and constrain the spread of nicking by MRX/Sae2. Nevertheless, when S1-seq profiles from fun30Δ were overlaid to show what the combined action of MRX/Sae2 and Exo1 could accomplish for these same hotspots, median resection lengths were ~300 nt longer than for MRX/Sae2 alone (blue traces in Figure 4D). This implies that the same nucleosomes that constrain MRX/Sae2 can be overcome by Exo1.

The fun30Δ exo1-nd mutant has reduced spore viability

The fun30Δ exo1-nd double mutant had decreased spore viability (58.8%), with an increased proportion of tetrads with two dead spores as well as smaller increases in tetrads with one, three, or four dead spores (Figure 5A,B). By contrast, the exo1-nd single mutant had a more modest decrease in viability (87.3%) that was mostly attributable to an increase in three-spore viable tetrads, while the fun30Δ single mutant showed normal spore viability.

Figure 5: Decreased spore viability and recombination defects in fun30Δ exo1-nd.

(A) Spore viability of the indicated strains (n = 110 tetrads for each genotype).

(B) Spore viability patterns for the tetrads shown in panel A.

(C) Estimates of contributions from random spore death and MI non-disjunction to total spore death. Tetrad dissection data from panel B were subjected to the TetFit algorithm (Chu and Burgess 2016; Chu et al. 2017). Data from iml3Δ and msh5Δ (Chu and Burgess 2016) are shown for comparison.

(D) Parental configuration of spore-autonomous fluorescent markers (left) (Thacker et al. 2011) used to measure genetic distance (middle; error bars indicate standard errors) and MI non-disjunction (NDJ) frequency (right).

Tetrads with either two or four dead spores can arise from meiotic recombination defects that cause chromosome missegregation during the first meiotic division (MI nondisjunction: MI-NDJ) (Chu and Burgess 2016). Tetrads with one or three dead spores are generally ascribed to random spore death (RSD), which can arise, for example, by defects in sister chromatid segregation. We used TetFit (Chu and Burgess 2016; Chu et al. 2017) to estimate the relative contributions of RSD (24.0%) and MI-NDJ (23.7%) to spore death in the fun30Δ exo1-nd double mutant (Figure 5C). This gives an estimated 4.0% MI-NDJ per chromosome. The increased frequency of MI-NDJ is consistent with a meiotic recombination defect, but the apparent mixture of RSD and MI-NDJ is in contrast to mutants with more unitary modes of spore death, such as iml3Δ (mostly RSD from defects in sister chromatid disjunction) or msh5Δ (mostly MI-NDJ from defects in homologous recombination) (Chu and Burgess 2016).

To further interrogate meiotic recombination in fun30Δ exo1-nd, we measured genetic distances in two intervals and the frequency of MI-NDJ of chromosome VIII using a spore-autonomous fluorescence assay (Figure 5D and Table S2) (Thacker et al. 2011). MI-NDJ was increased 7.3-fold by this assay to 2.26%, consistent with the TetFit analysis. The double mutant also had slightly decreased crossovers as compared to wild type in the CEN8-ARG4 interval but not ARG4-THR1 (Figure 5D and Table S2).

fun30Δ exo1-nd double mutants have compromised interhomolog recombination bias

To more rigorously evaluate recombination timing and efficiency, we used direct physical analysis of recombination intermediates and products at the strong HIS4LEU2 hotspot. Genomic DNA was isolated from synchronized meiotic cultures, digested with appropriate restriction enzymes, separated by either one- or two-dimensional agarose gel electrophoresis, and analyzed by Southern blotting and indirect end labeling. Restriction site polymorphisms between the homologous chromosomes flanking the hotspot allow detection and quantification of DSBs; branched recombination intermediates (single-end invasions (SEIs) and double Holliday junctions (dHJs)) between sister chromatids or between homologs; and both crossover and noncrossover recombination products (Figure S4A-D) (Hunter and Kleckner 2001; Kim et al. 2010).

In wild type, signals from DSBs, SEIs, and dHJs appeared and disappeared, crossover and noncrossover recombination products accumulated, and nuclear divisions occurred with the expected kinetics (Figures 6A-E and S4E,F). More signal was observed for interhomolog dHJs (IH-dHJs) than for intersister (IS-dHJs) throughout meiotic prophase I (Figures 6D and 6E bottom), reflecting the normal bias in recombination partner choice that favors using homologous chromosomes rather than sister chromatids (Schwacha and Kleckner 1997; Kim et al. 2010). The exo1-nd mutation increased the time-averaged amount of DSB signal (1.8-fold, estimated from comparative areas under the time-course curves) and recombination intermediates (1.7-fold for SEIs and 1.9-fold for total dHJs) but did not affect crossover levels and reduced the amount of noncrossovers slightly (Figures 6A,B,D,E and S4E,F). Similar to previous findings (Zakharyevich et al. 2010), exo1-nd maintained interhomolog recombination bias, albeit somewhat weakened (Figures 6D and 6E bottom).

Figure 6: Diminished interhomolog bias at HIS4LEU2 in fun30Δ exo1-nd mutants.

(A) One-dimensional (1D) gel analysis of DSBs and crossovers (COs) at the HIS4LEU2 hotspot.

(B) Quantification of DSBs and crossovers from 1D gel analysis. DSB and crossover levels are shown as a percentage of total hybridization signal per lane.

(C) Meiotic division time courses. The graph shows the percentage of cells that have completed one or both divisions.

(D) Representative two-dimensional (2D) gels of SEIs and dHJs at HIS4LEU2. Arrows indicate interhomolog joint molecules (green) or Mom-Mom (red) and Dad-Dad (blue) intersister molecules (color coding as in Figure S4A).

(D) Quantification of SEIs and dHJs from 2D gel analyses.

In all graphs, the data are the mean ± SD for three-independent meiotic cultures.

Unexpectedly, the fun30Δ mutation delayed the onset of DSB formation, appearance of crossovers, and completion of meiotic divisions, with each affected to a similar extent (~1 h) (Figures 6A,B,C,E and S4E,F). These delays are likely attributable to slower meiotic entry, as FACS analysis showed similarly delayed meiotic DNA synthesis (~1 h) in fun30Δ compared to wild type (Figure S5). IH-dHJs exceeded IS-dHJs in fun30Δ, indicating that strong IH bias was maintained, and final crossover levels were normal (Figures 6B,E and S4F).

The fun30Δ exo1-nd double mutant combined several of the phenotypes of the single mutants. For example, meiotic DNA synthesis and DSB onset were delayed comparably to the fun30Δ mutant (Figures 6A,B and S5), and peak levels of both DSBs and recombination intermediates were elevated similarly to exo1-nd (Figure 6A,B,D,E). However, the double mutant also showed additional defects not seen with either single mutant, including prolonged presence of DSBs, which may indicate continued DSB formation and/or longer DSB lifespan (Figure 6A,B); defective interhomolog bias (more IS-dHJs than IH-dHJs; Figure 6D,E); substantially delayed (2.5–3 h) and reduced formation of both crossovers (decreased ~2-fold) and noncrossovers (decreased ~3-fold) (Figures 6A,B and S4E,F); and greatly delayed (~3 h) meiotic divisions (Figure 6C).

We verified the changes in DSB timing and/or levels for the single and double mutants at four natural hotspots (ERG1, CYS3, BUD23 and ARG4), observing similar results as at HIS4LEU2: DSBs were delayed in fun30Δ; elevated in exo1-nd; and delayed, elevated, and persistent in fun30Δ exo1-nd (Figure S6A-H). We also confirmed using two-dimensional gel analysis at the ERG1 hotspot that the interhomolog bias of dHJs was maintained in fun30Δ, weakened in exo1-nd, and greatly reduced in fun30Δ exo1-nd double mutants (Figure 7A-C).

Figure 7. Interhomolog bias at the ERG1 hotspot.

(A) Physical map of the ERG1 hotspot on chromosome VII showing diagnostic SacII restriction enzyme sites and the position of the Southern blot probe. Parental chromosomes were distinguished by SacII restriction enzyme site polymorphisms.

(B) Two-dimensional (2D) gel Southern blot analysis of SEIs and dHJs at ERG1. Arrows indicate interhomolog dHJs (green) and either Mom-Mom (red) or Dad-Dad (blue) intersister dHJs.

(C) Quantification of dHJs from 2D gel analysis. The data are the mean ± SD for three-independent meiotic cell cultures.

(D) Progression of interhomolog bias at HIS4LEU2 throughout meiotic prophase. The strength of interhomolog bias was estimated by calculating the log₂-transformed ratio of IS-dHJ to IH-dHJ from the dataset in Figure 6E (mean ± SD of three biological replicates). Smaller values indicate stronger interhomolog bias.

(E) Different impacts of shortening resection tract length on interhomolog bias and crossover formation. Blue points show the log₂-transformed ratio of IS-dHJs to IH-dHJs calculated at the time point with maximal total dHJ signal for each individual replicate (wild type: 4, 5, 4 h; fun30Δ: 5, 6, 5 h; exo1-nd: 4, 5, 5 h; fun30Δ exo1-nd: 7, 7, 7 h). Gray points show the average of the maximum crossover frequency for each time course. Values were calculated from the datasets in Figures 6B,E. Error bars indicate mean ± SD.

We consider it likely that the pronounced defect in interhomolog bias in the fun30Δ exo1-nd double mutant is at least one cause of the persistent DSB signal, reduced number of crossovers, and strong delay in crossover and noncrossover formation. It is also likely that these recombination defects in turn cause delayed prophase I exit (resulting in greatly delayed meiotic divisions), increased MI-NDJ, and decreased spore viability (see Discussion).

Discussion

We previously proposed that one or more chromatin remodelers contributes to meiotic resection by removing nucleosomes in the vicinity of DSBs (Mimitou et al. 2017). Our findings here implicate Fun30 in this role in yeast. Chromatin remodeling contributes to resection in vegetative yeast and mammalian cells as well (Peritore et al. 2021; Fowler et al. 2022). However, unlike in vegetative cells, where Fun30 acts redundantly with other remodelers to promote resection specifically over relatively long distances (Chen et al. 2012; Costelloe et al. 2012; Eapen et al. 2012), during meiosis Fun30 by itself strongly affects both initial (MRX/Sae2) and secondary (Exo1) resection steps. Fun30 thus plays a predominant role in DSB processing uniquely in the context of meiotic recombination.

We found no effects of the other chromatin remodeler mutations tested, either alone or in combination with fun30Δ. Because we used swr1Δ and htz1Δ null mutants, we can exclude significant contributions of Swr1 and Swr1-dependent deposition of Htz1. Fun30 thus may be the only relevant remodeler in meiosis, but it is important to note that we do not formally exclude contributions of other remodelers. Normal mitotic resection involves the nonessential INO80 subunit Arp8 (Tsukuda et al. 2005); although we found that Arp8 was dispensable for meiotic resection, we cannot rule out that INO80 has an Arp8-independent role in meiotic resection. Moreover, we were unable to test the involvement of RSC because deletion of its nonessential subunit Rsc2 or meiotic depletion of its essential subunit Sth1 (Peritore et al. 2021) interfered with meiotic entry in our strains (unpublished observations), consistent with RSC’s role in IME2 expression (Inai et al. 2007).

Spatiotemporal coordination between DSB formation and resection

Meiotic chromosomes are organized into arrays of chromatin loops that are anchored at their bases by proteinaceous axial elements that include (among other things) cohesins and meiosis-specific axis proteins such as Red1 and Hop1 (Kleckner 1996; Smith and Roeder 1997; Klein et al. 1999; Zickler and Kleckner 1999; Kleckner 2006; Panizza et al. 2011). The DSB-forming machinery appears to localize to the axes but most DSBs occur in DNA segments that are usually in the loops. This paradox has led to a model in which tethered loop-axis complexes (TLACs) are formed by recruitment of loop DNA to axes to allow DSB formation and subsequent recombination (Blat and Kleckner 1999; Kleckner 2006; Panizza et al. 2011) (Figure S7).

We therefore interpret the DSB-dependent enrichment of Fun30 at both hotspots and Rec114 peaks (which are thought to represent axis-associated assemblies of DSB-promoting factors) as reflecting the spatial coordination between hotspots and axes within the context of TLACs (Figure S7). We previously showed that histone H3 is phosphorylated on threonine 11 by the DSB-responsive kinase Mek1, and that phospho-H3 is enriched at both axis attachment sites and around hotspots (Kniewel et al. 2017). Fun30 thus provides another example of a protein whose chromatin localization in response to DSBs appears to be shaped by the higher order organization of meiotic chromosomes.

Our ChIP-seq data indicate that meiotic DSBs provoke the recruitment of Fun30 nearby, likely on the broken chromatid and/or its sister chromatid. DSBs also activate a feedback control where local activation of the DSB-response kinase Tel1 (ortholog of mammalian ATM) inhibits Spo11 cleaving the same hotspot again (Lange et al. 2011; Johnson et al. 2021; Prieler et al. 2021). An interesting implication of post-DSB recruitment of Fun30 in cis is that the resulting chromatin remodeling occurs exclusively on the homolog that has lost DSB competence because of Tel1-mediated feedback control. By restricting chromatin changes to hotspots that are unlikely to experience more DSBs, Spo11 may thus be constrained to cut only where allowed by chromatin that has not been acted on by Fun30. This constraint would inhibit DSBs from forming within gene bodies, possibly affecting the mutagenic potential of recombination. Fun30 recruitment and activation at DSBs are cell cycle regulated in vegetative cells (Chen et al. 2016; Bantele and Pfander 2019), suggesting that multiple regulatory strategies may have evolved to limit Fun30 remodeling in different contexts.

Nucleosomes as a resection barrier in vivo

In vitro, resection by Exo1 is almost completely blocked by a nucleosome (Adkins et al. 2013), and MRX/Sae2 incision can occur directly adjacent to but not apparently within a nucleosome (Wang et al. 2017). Genomic studies also suggest inhibition of MRX/Sae2 incision by nucleosomes in vegetative cells (Gnugge et al. 2023). We found here that constitutive chromatin structure strongly affected the residual Fun30-independent resection that could be carried out by MRX/Sae2 alone or in combination with Exo1, supporting the interpretation that nucleosomes pose barriers to both resection nucleases during meiosis as well (Mimitou et al. 2017). Interestingly, however, our findings also indicate that nucleosomes are unlikely to be an absolute block to the nucleases.

First, we found that MRX/Sae2 by itself (fun30Δ exo1-nd) frequently cleaves within the position usually occupied by the +1 nucleosome rather than being strictly constrained to the NDR and the linker between nucleosomes. The DNA cleavage positions in this case mostly fall within the first half of the nucleosome position and rarely spread up to the next nucleosome, so we can rule out that these cleavage events are solely due to the +1 nucleosome having been removed entirely in all or a subset of cells.

Second, we found that Exo1 still carries out substantial resection in the absence of Fun30 (comparing fun30Δ to fun30Δ exo1-nd). This means that Exo1 is able to digest DNA within the same nucleosomes that are constraining MRX/Sae2 and causing resection tracts to shorten in fun30Δ exo1-nd compared to exo1-nd alone.

We therefore conclude that both MRX/Sae2 and Exo1 have some limited ability to digest nucleosomal DNA in vivo without intervention by Fun30, even though they appear incapable of doing so in vitro (Adkins et al. 2013; Wang et al. 2017). One possibility is that both enzymes can act on nucleosomal DNA directly. We note that biochemical studies to date have relied on the strong artificial 601 nucleosome positioning sequence, which provides an exceptionally stable nucleosome. Perhaps both nucleases are better able to act on nucleosomes of more physiological stability. Alternatively, there may be Fun30-independent processes that destabilize nucleosomes. We can rule out Swr1 or Htz1, but remaining nonexclusive candidates for such processes could include transcription, action of remodelers other than Fun30 or Swr1, histone posttranslational modifications, and/or DSB-provoked changes in chromatin structure (e.g., caused by topological changes or DSB response signaling).

Minimum resection length required for meiotic recombination

In vegetative cells, recombination can occur with homology lengths less than 40 bp, and normal recombination efficiency is estimated to need only ~100 to 250 bp on each side of the DSB (Jinks-Robertson et al. 1993; Manivasakam et al. 1995; Hua et al. 1997; Inbar et al. 2000; Ira and Haber 2002). It has been suggested that the reason meiotic resection in wild type is considerably longer than these minima is to ensure that all DSBs are resected enough to allow recombination to proceed efficiently and accurately (Zakharyevich et al. 2010). It was further noted that most DSBs are resected further than these minima in an exo1-nd mutant, and only a small fraction of DSBs had resection tracts short enough to potentially impair recombination (Zakharyevich et al. 2010).

Our findings refine these ideas and suggest important roles of resection aside from simply providing enough ssDNA for homology searching. First, we note that the fun30Δ single mutation reduced median resection lengths almost as much as exo1-nd alone, but gave better spore viability. One important difference between these single mutants is that fun30Δ cells had a longer minimum resection distance than exo1-nd, such that fewer resection tracts were very short (e.g., only ~4.6% were <200 nt long in fun30Δ, vs. 8.4% in exo1-nd). We surmise that the relative paucity of extremely short resection tracts is the reason for better preservation of successful meiosis in fun30Δ mutants than in exo1-nd, measured as better spore viability.

Second, the fun30Δ exo1-nd double mutant had extremely shortened resection as well as recombination defects leading to increased rates of homolog missegregation and spore death. We do not exclude the possibility that there are resection-independent defects contributing to these phenotypes, e.g., altered gene expression (Durand-Dubief et al. 2012; Byeon et al. 2013). (Altered gene regulation is a good candidate to explain the delayed meiotic entry in the fun30Δ background.) However, it is plausible that the meiotic recombination defects in the double mutant are mostly (or even entirely) the consequence of the shortened resection itself. If so, and if the principal importance of resection length is to provide enough ssDNA for efficient homology searching, we initially expected to observe fewer and/or less stable recombination intermediates. Unexpectedly, though, both SEIs and dHJs were abundant, suggesting that strand exchange and its prerequisite homology search were both still effective despite the greatly reduced ssDNA length. Instead, we observed a pronounced decrease in interhomolog bias that was probably sufficient to explain the reduced recombination and increased chromosome missegregation.

Third, as reported previously (Joshi et al. 2015), wild type progressively establishes interhomolog bias during meiotic prophase (Figure 7D). In contrast, exo1-nd showed delays in the establishment of interhomolog bias in addition to reducing overall bias, and fun30Δ exo1-nd failed almost entirely to establish the bias. The median resection length negatively correlated with the degrees of interhomolog bias (blue points in Figure 7E). Interestingly, however, crossover formation showed a more pronounced threshold effect, with significant defects only apparent with the most extreme reduction in resection length (gray points in Figure 7E). We suggest that homeostatic mechanisms that control crossover outcomes (e.g., crossover homeostasis (Martini et al. 2006)) provide robustness in the face of relatively modest defects in interhomolog bias. There appears to be a threshold between 17 and 47% of wild-type ssDNA content, below which overt crossover defects begin to materialize.

We envision two reasons for the observed decrease in interhomolog bias when resection is very short. One possibility is that there is a minimal requirement for ssDNA-provoked DNA damage signaling. The establishment of interhomolog bias involves activation of Mec1 and Tel1 kinases in response to DSB formation, which in turn activates Mek1 kinase (Hollingsworth and Gaglione 2019). Because Mec1 activation requires ssDNA-bound RPA (Zou and Elledge 2003), reducing the total ssDNA content by shortening resection without decreasing DSB numbers could reduce interhomolog bias similar to the effect of having fewer DSBs (Joshi et al. 2015). A second, non-exclusive possibility is that very short resection tracts might compromise the necessary loading of both Dmc1 and Rad51. Interhomolog bias is defective in the absence of Rad51, which plays a strand-exchange-independent role in promoting normal partner choice (Schwacha and Kleckner 1997; Cloud et al. 2012; Hong et al. 2013; Lao et al. 2013). Perhaps extremely short resection tracts sometimes fail to load sufficient Rad51, thereby compromising interhomolog bias. Regardless of which of these scenarios is correct, our results strongly indicate that a major constraint shaping how long meiotic resection needs to be is the minimal amount of ssDNA necessary to achieve appropriate regulation of recombination partner choice.

Materials and Methods

Yeast strain and plasmid construction

Unless otherwise noted, all yeast strains used in this study (Table S1) were of the SK1 background (Kane and Roth 1974). We used a standard lithium acetate method (Gietz and Schiestl 2007) and verified transformants by PCR and Southern blotting. The fun30Δ, arp8Δ, htz1Δ, and swr1Δ mutants were generated by replacing the coding sequences with a G418 resistance cassette (KanMX4). Appropriate crosses and tetrad dissection were then used to generated single and double mutants. The exo1-nd (D173A) mutation creates a DrdI restriction enzyme site that was used to follow the exo1-nd allele in crosses. Spore autonomous fluorescent markers (THR1::m-Cerulean-TRP1 and CEN8::tdTomato-LEU2, ARG4::GFP*-URA3) (Thacker et al. 2011) were introduced to the fun30Δ exo1-nd mutant by crossing and tetrad dissection.

For ChIP-seq, Fun30 was C-terminally tagged by integrating a DNA fragment containing 13 copies of the Myc epitope and the KanMX6 cassette amplified from pFA6a-13Myc-KanMX6 (Longtine et al. 1998) before the stop codon of the FUN30 open reading frame. The primer sets (uppercase: 50 bp homology sequences, lowercase: annealing sequences) are as follows: 5′-TGGAGGATATAATTTATGATGAAAACTCGAAACCGAAGGGAACCAAAGAAggtggtggtggtggtggtggtggtCGGATCCCCGGGTTAATTAA; 5′-TTTATTTTCTGCTTATCTATTTACTTTTTTACTATATTTTTATTTATTTActggatggcggcgttagtatcgaatcgacagcagtatagcgacc.

Sporulation

Meiotic cultures for S1-Southern and S1-seq.

Yeast strains were sporulated as previously described (Mimitou and Keeney 2018). We performed pre-sporulation culture in YPA (1% yeast extract, 2% Bacto peptone, 2% potassium acetate, 0.001% antifoam 204 (Sigma)) for 14 h, then transferred to sporulation medium (2% potassium acetate with amino acids and 0.001% polypropylene glycol). Meiotic cells from 66 ml sporulation medium harvested at 4 h after transfer to sporulation medium were washed with 50 ml of water and 50 mM EDTA pH 8.0 and stored at −80 °C.

Tetrad dissection.

To avoid accumulation of lethal mutations, we avoided prolonged culture of diploid strains where possible. Instead, haploid strains were freshly mated on YPD plates for 6 h and immediately sporulated in sporulation medium for >48 h. Tetrads were treated with 100 μg/ml zymolyase 20T (US Biological) at 30 °C for 20 min and dissected on YPD plates. Spore viability was scored after 2 days of incubation at 30 °C.

Meiotic time courses for physical analysis of recombination.

Procedures were as described previously (Kim et al., 2010; Hong et al., 2013; Yoon et al., 2016; Hong et al., 2019). Yeast diploid cells were patched onto a YPG plate (1% yeast extract, 2% peptone, 2% agar, and 3% glycerol) and incubated at 30 °C for 12 h. Cells were then streaked on a YPD plate (1% yeast extract, 2% peptone, 2% agar and 2% glucose) and incubated at 30 °C for two days. Single colonies were picked and inoculated in YPD liquid medium (1% yeast extract, 2% peptone and 2% glucose) and cultured at 30 °C for 24 h. To synchronize yeast cell cultures, the YPD-cultured cells were diluted in 200 ml of pre-warmed SPS medium (0.5% yeast extract, 1% peptone, 0.17% yeast nitrogen base without amino acids, 0.5% ammonium sulfate, 1% potassium acetate and 50 mM potassium biphthalate; pH was adjusted to 5.5 with 10 N KOH) and grown at 30 °C for 17–18 h. Meiosis was induced by culturing cells in sporulation medium (SPM; 1% potassium acetate, 0.02% raffinose and 2 drops of antifoam per liter) at 30 °C. SPM-cultured cells were harvested at 0, 2.5, 3.5, 4, 5, 6, 7, 8, 10 and 24 h and then cross-linked with 0.1 mg/mL trioxsalen (Sigma, T1637) under 365-nm ultraviolet light for 15 min.

S1-Southern and S1-seq

Procedures for the agarose plug preparation, S1 nuclease treatment and the subsequent Southern blotting or S1-seq library preparation were described previously (Mimitou et al. 2017; Mimitou and Keeney 2018). These procedures were followed here with the following modifications; section numbers refer to (Mimitou and Keeney 2018): Section 4.4.2, steps 3–5: Each cell pellet was resuspended in 700 μl 50 mM EDTA pH 8.0. Seven hundred microliter of the cell suspension was mixed with 238 μl Solution 1 (SCE: 1 M sorbitol, 0.1 M sodium citrate, 60 mM EDTA pH 7.0 containing 5% β-mercaptoethanol and 1 mg/ml zymolyase 100T) at 40 °C and aliquoted to plug molds (BioRad) to prepare 20 agarose plugs. Section 5.2.2, steps 1–3: GELase was replaced by β-agarase I (NEB). The reaction (1 U per plug) was carried out at 42 °C. Section 5.2.2, step 23: The sonication step was carried out using Covaris E220 evolution with the settings PIP = 175 w, DF = 10%, CPB = 200, time: 180 sec at 4 °C. Less than 130 μl sample per microtube was loaded in microTUBE AFA Fiber Crimp-Cap on Rack E220e 8 microTUBE Strip. Section 5.4.2, step 6: For the post-PCR step, instead of ethanol precipitation with ammonium acetate, we purified library DNA using QIAquick PCR Purification Kit (Qiagen) with a final elution volume of 30 μl warm 10 mM Tris-HCl pH 8.0. Section 5.4.2, step 11: For the size selection step, instead of 5X TBE PROTEAN gel, samples were run on 1.5% agarose at 100 volts for 2 h. The gel region from 200 to 700 bp was excised and DNA was purified by QIAquick Gel Extraction Kit (Qiagen) following the manufacturer’s instructions (we included step 6 in the Quick-Start Protocol for QIAquick: add 500 μl Buffer QG to the QIAquick column and centrifuge for 1 min). Library DNA was eluted in 25 μl warm 10 mM Tris-HCl pH 8.0.

Chromatin immunoprecipitation for Fun30-Myc

ChIP experiments were performed as described previously (Murakami and Keeney 2014), with modifications in shearing chromatin and calibrating datasets. Similar to the previously described calibrated ChIP method (Hu et al. 2015), we used S. mikatae cells as a spike-in control to compare ChIP-seq signals between datasets. For each S. cerevisiae strain, cells were sporulated using the YPA presporulation protocol as described above. Fifty ml (2 × 10⁹ cells) of culture at 4 h in meiosis was fixed with 1% formaldehyde for 15 min at room temperature, with mixing at 50 rpm. Crosslinking was quenched by adding glycine to 131 mM for 5 min. Cells were washed twice with 20 ml cold TBS buffer, frozen with liquid nitrogen, and stored at −80 °C until further steps. S. mikatae cells were sporulated using the SPS presporulation protocol (Murakami et al. 2020). Cells harvested at 4 h in meiosis were fixed and washed with the same condition described above. An aliquot of 2 × 10⁸ S. mikatae cells (10% of the number of S. cerevisiae cells) was added to each sample.

Cells were resuspended in FA lysis buffer (50 mM HEPES-NaOH pH 7.5, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 7 μg/ml aprotinin, 10 mg/ml each of leupeptin, pepstatin A, and chymostatin, 1 mM PMSF, and 1× each of Roche phosphatase and 1% protease inhibitor cocktails [Sigma]) (Vale-Silva et al. 2019; de Jonge et al. 2020) in 2-ml screw-cap Eppendorf tubes, and disrupted using zirconia/silica beads (0.5 mm, Biospec Products; ~900 μl per sample) and a FastPrep-24 (MP Biomedicals) with ten rounds of vigorous shaking at 6.5 m/sec for 60 sec. Lysates were pelleted by centrifugation at 15,000 rpm for 5 min at 4 °C. Chromatin in the whole-cell extracts was sheared by sonication using Covaris E220 evolution. To yield an average DNA size of around 350 bp (range 100–500 bp), 1 ml whole cell extract with SDS added to 0.1% final concentration (Pchelintsev et al. 2016) was loaded in miliTUBE with AFA fiber and assembled on miliTUBE holder. Sonication conditions were 140 w, DF = 5%, CPB = 200, time = 8 min at 4 °C. Sonicated chromatin was centrifuged at 15,000 rpm for 5 min at 4 °C, and the supernatant (input) was collected for immunoprecipitation steps as described (Murakami and Keeney 2014). DNA purified from input and immunoprecipitate samples was further sonicated prior to library preparation at the genomics core facility (Integrated Genomics Operation, Memorial Sloan Kettering Cancer Center)

Physical analysis of recombination intermediates

Sporulation conditions and subsequent molecular biology procedures to detect recombination intermediates were as described previously (Oh et al. 2009; Kim et al. 2010). Genomic DNA preparation has been described (Kim et al. 2010; Hong et al. 2013; Yoon et al. 2016; Hong et al. 2019). Cells were treated with Zymolyase (US Biological) and then lysed in guanidine-HCl solution (4.5 M guanidine-HCl, 0.1 M EDTA, 0.15 M NaCl, and 0.05% sodium lauroyl sarcosinate) at 65 °C for 15 min. Genomic DNA was extracted twice with phenol/chloroform/isoamyl alcohol (25:24:1) and precipitated with ethanol. The DNA pellets were washed with 70% ethanol and dried at 4 °C overnight. Meiotic recombination was analyzed by gel electrophoresis at HIS4LEU2 and ERG1 loci similarly. Physical DNA analysis at the HIS4LUE2 locus on chromosome III was performed as described previously (Kim et al., 2010; Hong et al., 2013; Yoon et al., 2016; Hong et al., 2019; Lee et al., 2021). Parental homologs are distinguished via XhoI restriction site polymorphism. Genomic DNA (2 μg) was digested with XhoI or XhoI plus NgoMIV for one-dimensional (1D) gel analysis. The ERG1 locus on chromosome VII was detected to obtain the DNA species of DSBs, and joint molecules by SacII restriction site polymorphisms (Thacker et al., 2014; Lao et al., 2013; Lee et al., 2021). The DNA samples were loaded onto 1D gels (0.6% UltraKem LE agarose (Young Science) in 1× Tris-borate-EDTA buffer), and electrophoresis was carried out in a 1× Tris-borate-EDTA buffer for 24 h. For two-dimensional (2D) gel analysis, genomic DNA (2.5 μg) was digested with XhoI for the HIS4LEU2 locus and SacII for the ERG1 locus. The DNA samples were loaded onto 1D gels (0.4% Seakem Gold agarose (Lonza) in 1× Tris-borate-EDTA buffer), and electrophoresis was carried out in 1× Tris-borate-EDTA buffer for 21 h. The 1D gel was stained with 0.5 μg/ml ethidium bromide (EtBr), and the gel strips of interest were cut and placed in a 2D gel tray. The gel electrophoresis was carried out in 1× Tris-borate-EDTA buffer in 4 °C cold room. The gels were transferred to Biodyne B membrane (Pall). For Southern blot analysis, hybridization was carried out using probes labelled with [α-³²P]-dCTP (Thacker et al., 2014; Lao et al., 2013; Lee et al., 2021). Hybridization signals were visualized using a phosphoimager and quantified using Quantity One software (BioRad).

Bioinformatic analysis

S1-seq mapping and analysis.

Sequencing (50-bp paired-end reads; Illumina HiSeq2000) was performed in the MSK Integrated Genomic Operation. In silico clipping of library adapters and mapping to the genome was performed by the MSK Bioinformatics Core Facility using a custom pipeline as described (Mimitou and Keeney 2018) with modifications. The code for read processing and mapping is available online at https://github.com/soccin/S1Seq. After mapping, the reads were separated into unique and multiple-mapping sets, but only uniquely mapping reads were analyzed in this study.

All downstream analyses were performed using R (RStudio version 1.0.143, R version 4.0.3 GUI 1.73 Catalina build). Map curation before analysis was performed by masking DNA ends and meiotic DSB-independent reads, with mitochondrial DNA and the 2-micron plasmid excluded as described (Mimitou et al. 2017). A global view of meiotic double-strand break end resection and the mask coordinates can be found at https://github.com/soccin/S1Seq. Each map was normalized to reads per million remaining mapped reads and the biological replicates were averaged.

We used a hotspot list compiled from a combination of multiple independent wild-type Spo11-oligo maps (Mohibullah and Keeney 2017). Different from (Mimitou et al. 2017), the left arm of chromosome 3 was not censored because all strains used for S1-seq are without HIS4LEU2 and leu2::hisG artificial hotspots on this chromosome arm. Published maps were used of nucleosome occupancy (Pan et al. 2011), nucleosome midpoints (Zhang et al. 2011)DNA</keyword></keywords><dates><year>2011</year><pub-dates><date>Jun</date></pub-dates></dates><isbn>1549-5469 (Electronic, and S1-seq in a spo11 mutant with altered DSB locations (Claeys Bouuaert et al. 2021).

Calibrated Fun30 ChIP-seq.

Paired-end 50-bp reads were filtered and end trimmed, followed by removal of the reads containing tag sequences, and then mapped to a custom reference genome consists of the S. cerevisiae (sacCer2) and S. mikatae (IFO1815) reference genomes. Each S. cerevisiae coverage map was normalized according to the S. mikatae read density for the same antigen from the same culture. Coverage maps generated from S. cerevisiae reads as previously described (Murakami and Keeney 2014) were divided by the total number of reads that were uniquely mapped to S. mikatae chromosomes to create input and immunoprecipitate maps normalized to spike-in control. Each normalized immunoprecipitate map was divided by the corresponding normalized input map to generate a ChIP coverage map normalized to the spike-in control. The S. mikatae spike-in control minimizes the effects of sample-to-sample variation during lysis, immunoprecipitation, and library preparation. Also, the fixed ratio of S. cerevisiae to S. mikatae is an excellent scaling factor for comparing the amount of Fun30 enrichment between samples for the same antigen (Hu et al. 2015). We used ARS and tRNA coordinates (sacCer2) downloaded from Saccharomyces Genome Database (yeastgenome.org).

K-means clustering (Figures 4C,D).

To investigate the relationship between pre-existing chromatin structure (i.e., the chromatin structure before DSB formation or DSB-provoked remodeling at hotspots) and the resection patterns in fun30Δ mutants, we used k-means clustering to group hotspots by their resection endpoint distributions. To do so, we selected hotspots with no other hotspots located within 1 kb (n = 2778) to avoid confounding effects of resection from neighboring DSBs. For each hotspot, we took S1-seq signals within 1 kb to the right (top strand reads) and left (bottom strand reads) of the hotspot midpoint, resulting in a total of 5556 resection profiles. We subtracted the background from each profile (defined as the lowest S1-seq signal within each profile), then normalized each profile to its total signal. This normalization was done to remove differences in DSB frequency and focus the clustering exercise on the spatial distribution rather than signal strength. The collection of S1-seq profiles from each dataset (fun30Δ single mutant or fun30Δ exo1-nd double mutant) was then clustered into three groups using the kmeanspp function in the LICORS package in R (https://CRAN.R-project.org/package=LICORS). For each group, the average of the resection profiles was overlayed with the averaged nucleosome data for the corresponding genomic locations (Pan et al. 2011).

TetFit.

We estimated the frequencies of MI nondisjunction death and random spore death based on finding the best-fit distribution of tetrads with 4, 3, 2, 1 and 0 viable spores to an observed distribution using the R algorithm TetFit (Chu and Burgess 2016) with the default parameters ndint = 500, rsdint = 500, chr = 16, anid = 0.035, ndm = 10, minrsd = 0.0, maxrsd = 0.8, minnd = 0.0, maxnd = 0.017.

Data availability

Raw and processed S1-seq and ChIP-seq sequence reads for new maps generated in this study Table S3) are deposited in the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo; accession no. GSE221033 and GSE221377).