Dna2 removes toxic ssDNA-RPA filaments generated from meiotic recombination-associated DNA synthesis

Abstract During the repair of DNA double-strand breaks (DSBs), de novo synthesized DNA strands can displace the parental strand to generate single-strand DNAs (ssDNAs). Many programmed DSBs and thus many ssDNAs occur during meiosis. However, it is unclear how these ssDNAs are removed for the complete repair of meiotic DSBs. Here, we show that meiosis-specific depletion of Dna2 (dna2-md) results in an abundant accumulation of RPA and an expansion of RPA from DSBs to broader regions in Saccharomyces cerevisiae. As a result, DSB repair is defective and spores are inviable, although the levels of crossovers/non-crossovers seem to be unaffected. Furthermore, Dna2 induction at pachytene is highly effective in removing accumulated RPA and restoring spore viability. Moreover, the depletion of Pif1, an activator of polymerase δ required for meiotic recombination-associated DNA synthesis, and Pif1 inhibitor Mlh2 decreases and increases RPA accumulation in dna2-md, respectively. In addition, blocking DNA synthesis during meiotic recombination dramatically decreases RPA accumulation in dna2-md. Together, our findings show that meiotic DSB repair requires Dna2 to remove ssDNA-RPA filaments generated from meiotic recombination-associated DNA synthesis. Additionally, we showed that Dna2 also regulates DSB-independent RPA distribution.


INTRODUCTION
Meiosis is a specialized cellular program to produce haploid gametes. During meiosis, one round of DNA replication is followed by two successi v e rounds of chromosome segregation: homolo gous chromosomes (homolo gs) segregation in meiosis I, and sister chromatids segregation in meiosis II. The hallmark of meiosis is using the homologs as templates to repair the programmed DNA double-strand breaks (DSBs) and generate crossovers ( 1 , 2 ).
The topoisomerase-like Spo11 works with cofactors to catalyze the formation of DSBs to initiate meiotic recombination (3)(4)(5). DSBs are then immediately resected from their 5 -terminal strands to produce 3 single-stranded tails, which are coated by RPA and later substituted by the RecA family proteins, Rad51 and Dmc1, to form nucleoprotein filaments (6)(7)(8)(9). The nucleoprotein filaments mediate DNA strand invasion, which results in the formation of displacement loop (D-loop) intermediates ( 1 , 2 , 10 ). Subsequently, de novo DNA synthesis begins from the 3 end of the invaded strand. If this invasion strand is displaced and then anneals with the other resected end, the DSB is finally r epair ed to be a non-crossover (NCO) via the synthesis-dependent strand annealing pathway (SDSA) ( 11 ). If this invasion strand is stabilized, the DSB would be r epair ed via the double-strand br eak r epair (DSBR) pathway. The second end would be captured during this process to form a double Holliday junction (dHJ), which would be resolved as a crossover (CO) ( 12 ) (Supplementary Figure S1).
Both SDSA and DSBR pathways r equir e de novo DNA synthesis from the invasion end and also the other end to fill the gap ( 13 ). Extensi v e studies show that DNA polymerase ␦ (Pol ␦) catalyzes de novo DNA synthesis in both mitotic and meiotic nuclei (14)(15)(16)(17). The newly synthesized DNA could displace the parental strand to generate singlestrand DN A (ssDN A) fla ps via the strand displacement activity of Pol ␦ ( 16 , 17 ). These ssDN A fla ps could be coated by RPA and visualized as ssDNA-RPA filaments. The simple ssDN A and ssDN A-RPA filaments need to be efficiently and timely removed to restore genome integrity. For meiotic n uclei, n umer ous pr ogrammed DSBs have the potential to gener ate a consider able number of ssDNA-RPA filaments. Thus, the timely removal of these filaments is a big challenge. Howe v er, the molecules involved in removing these ssDNA-RPA filaments are still unknown (Supplementary Figure S1).
During the lagging-strand DNA replication, each Okazaki fragment is initiated by the formation of a primer composed of RNA and initiator DNA. When Pol ␦ encounters the 5 -end of a preceding Okazaki fragment, this fragment may be dissociated from the template strand by Pol ␦ through its strand displacement activity ( 18 , 19 ).
Extensi v e studies show that Rad27 (Fen1) is the leading nuclease removing a majority of 5 -ssDN A fla ps in S. cerevisiae ( 20 , 21 ). The rest, a small portion of 5 -ssDNA flaps that are not timely removed by Rad27, are extended to form long ssDN A fla ps by continuous DN A synthesis and are coated by RPA. In vitro studies show that ssDNA-RPA flaps are suitable substrates for Dna2 helicase / nuclease ( 22 , 23 ). Ther efor e, Dna2 probably provides the final safeguard in removing the ssDN A fla ps during the lagging strand DNA synthesis ( 19 , 22 , 24 ). One recent study directly analyzing Okazaki fragments synthesized in vivo found that Rad27 processes the majority of the 5 -ssDN A fla ps of lagging strands and Exo1 makes a significant additional contribution, howe v er, the role of Dna2 in this process is extremely limited ( 25 ). Therefore, the role of Dna2 in removing long ssDN A fla ps generated in DN A replication is still not so clear.
Her e, we systematically explor ed the role of Dna2 in meiosis. The meiosis-specific depletion of Dna2 ( dna2-md ) resulted in the accumulation of RPA as well as a high frequency of chromosome mis-segregation and thus inviable spor es. The r esult of ChIP-seq r e v ealed that RPA was enriched at DSBs, probably at resected single DSB ends, in both wild type (WT) and dna2-md in a Spo11-dependent manner. Along with the progression of meiotic recombination, RPA enrichment gradually diminished in WT while RPA expanded to broader regions around DSBs in dna2md. As a result, DNA integrity is compromised, howe v er, une xpectedly, the le v els of crossov ers and non-crossov ers seem to be unaffected in dna2-md . Pachytene induced expression of Dna2 could efficiently remove the accumulated RPA and rescue spore viability in dna2-md . Moreover, the removal of Pol ␦ activator Pif1 and Pif1 inhibitor Mlh2 in dna2-md pre v ented and stim ulated RPA accum ulation, respecti v ely. In addition, b locking ne w DNA synthesis during meiotic recombination by adding hydroxyurea (HU) resulted in a dramatic decrease in RPA accumulation in dna2-md . These results support the proposal that Dna2 is r equir ed to r emove ssDNA-RPA filaments generated during meiotic recombina tion-associa ted DNA synthesis. Additionally, Dna2 regulates RPA distribution in a Spo11independent manner.

Yeast strains
The Sacchar om y ces cer evisiae strains used in this stud y are deri vati v es of the SK1 background and listed in Supplementary Table S1. The p CLB2-DNA2 strain was constructed by replacing its nati v e promoter with the CLB2 promoter using the polymerase chain reaction (PCR) method (26)(27)(28)(29). Similar ly, the pCLB2-PIF1 str ain was also constructed by the PCR method. The p GAL1-DNA2 strain was constructed by replacing the nati v e promoter of the DNA2 gene with the GAL1 promoter. The pSCC1-CDC6 strain was constructed by replacing its nati v e promoter with the SCC1 promoter using the PCR method (27)(28)(29)(30). Site-directed mutagenesis of DNA2E675A and DNA2R1253Q was introduced by PCR. The DNA2-AID strain was constructed by inserting an AID (auxin inducible degron) tag into the Cterminal of DNA2. The AID tag was amplified by PCR from a pNAT-AID 71-114 -6HA plasmid ( 31 , 32 ). The DNA2-3HA strain was constructed by inserting a triple-HA epitope into the C-terminal of DNA2 via the PCR method (27)(28)(29). The triple-HA tag was obtained from a pKT221 plasmid ( 33 ).

Meiotic time course
The meiotic time course was performed as previously described ( 34 , 35 ). Briefly, yeast cells were patched onto YPG plates (3% glycerol, 1% y east extr act, 2% Bacto peptone, and 2% Bacto agar) and incubated for 14 h at 30 • C. Then cells were steaked onto YPD plates (1% yeast extract, 2% Bacto peptone, 2% glucose, and 2% Bacto agar) and incubated for 56 h at 30 • C. A single healthy colony was inoculated into 4 ml of YPD liquid medium and incubated at 30 • C for 24 h with shaking. An appropriate amount of cultur e was transferr ed into the SPS medium (0.5% yeast extract, 1% Bacto peptone, 0.67% yeast nitrogen base without amino acids, 1% potassium acetate and 50 mM potassium biphthalate, pH 5.5) and cultured at 30 • C for 16 h with shaking. Synchronized cells were harvested and transferred into the sporulation medium (SPM; 1% potassium acetate, 0.02% raffinose) to induce meiosis. DAPI staining was used to analyze meiotic nuclear divisions under a microscope. To induce pGAL1-DNA2 expression, 1 M ␤-estradiol was added at the desired time points during meiosis. To induce the degradation of Dna2, 25 M CuSO 4 and 2 mM auxin (3-indoleacetic acid) (or DMSO as a control) were added to the cultures of DNA2-AID at desired time points.

Sporulation efficiency and spore viability assays
After the cells were transferred to SPM liquid medium for 24 h, sporulation efficiency was analyzed by examining the frequency of cells with asci under a light microscope. Tetrads were dissected onto YPD plates under a dissection microscope and incubated at 30 • C for 3 days. Spore viability was determined as the frequency of viable spores over the expected total spores (the number of dissected tetrads times 4).

Spor e-specific fluor escence assa y of chr omosome missegregation
The spore-specific promoter-dri v en YFP ( P YKL050C -YFP ) and RFP ( P YKL050C -RFP ) were separately integrated into the allelic position near the centromere of chromosome 9 for the two parental haploids ( 36 ). Yeast cells were patched onto YPD plates and incubated at 30 • C for 14 h. Then the fr esh cells wer e transferr ed into 4 ml of SPM and cultur ed at 30 • C for 2-3 days with shaking. Spore patterns with distinct fluorescence in tetrads were determined using a Zeiss fluorescence microscope (AxioImager.Z2). The four spores in one tetrad with accurate chromosome segregation produce two spor es expr essing YFP and the other two spores expressing RFP, while tetrads with mis-segregated chromosome 9 exhibit abnormal patterns of yellow / red fluorescence. The chromosome 9 mis-segr egation fr equency was calculated as the number of tetrads with mis-segregated chromosome 9 divided by the total number of tetrads with YFP / RFP spores.

Chromosome spreading and immunofluorescence
For chromosome spreading, synchronized yeast cells were collected at appropriate time points, processed into spheroplasts with Zymolyase 100T, spread on a clean glass slide with 1% Lipsol, and fixed using 3% paraformaldehyde containing 3.4% sucrose ( 37 ). For immunostaining, slides with spread nuclei were first dipped into 0.2% Photo-Flo for 30 seconds, then transferred to Tris-Buffered Saline (TBS) (136 mM NaCl, 3 mM KCl and 25 mM Tris-HCl, pH 8.0), and subsequently incubated at room temperature for 15 min. The slides were blocked with 1 × TBS with 1% BSA for 10 min and incubated with the appropriate primary (4 • C overnight) and secondary (37 • C for 2 h) antibodies. The primary antibodies used in this study included r abbit polyclonal anti-RPA (Agriser a, Cat# AS07214), rat monoclonal anti-HA (Roche, Cat#11867423001), goat polyclonal anti-Zip1 (Santa Cruz Biotechnology, Cat# sc-48716) and mouse monoclonal anti-Myc (Santa Cruz Biotechnology, Cat# sc-40). The secondary antibodies used in this study were Alexa 488-conjugated donkey anti-mouse / goat / rabbit (Thermo Fisher Scientific, Cat#A-21202 / A-11055 / A-21206) and Alexa 594-conjugated donkey anti-ra t / goa t (Thermo Fisher Scientific, Ca t# A-21209 / A-11058). Chromosomal DNA was stained with DAPI. Stained samples were visualized and imaged under a Zeiss fluorescence microscope (AxioImager.Z2) with an EMCCD camera and appropriate filters.

Quantification of RPA foci and immunofluorescence intensity
About 100 properly spread nuclei were analyzed at each time point using Ima geJ ( https://ima gej.nih.gov/ij ), spread nuclei that were overstretched or only partially spread were excluded in this analysis ( 38 ). The following criteria were applied to determine an RPA focus (similarly for a focus of other proteins): (i) located within DAPI stained nuclear ar ea, (ii) wer e punctate in appearance (and thus well above the background) and (iii) were separated from adjacent foci by at least two pixels ( ∼0.16 m; otherwise it was counted as a single focus). The focus number was manually determined with the help of the 'multi-point' tool in ImageJ. Samples were counted blindly by two observers. Quantification of RPA immunofluorescence intensity was performed as previously described ( 34 ). To accurately measure the intensity of RPA in different cells, the same buffers, the same concentration of antibodies, and the same incubation time were used to immunostaining RPA. The same parameters (e.g. exposure time and laser power) were used for image acquisition. The fluorescence intensity of RPA was quantified with Im-ageJ software. The target cell was circled according to the DAPI signal and the fluorescence intensity was measured as the raw intensity. The same circle was drawn in a region nearby this cell and the intensity was measured as the background intensity. The RPA fluorescence intensity was calculated as the raw intensity minus the background intensity.

Western blot
Synchronized yeast cells were collected a t appropria te time points in SPM and lysed in 20% trichloroacetic acid (TCA) using glass beads. After centrifuged at 12000 g for 1 min, the resultant pellet was extracted with Laemmli buffer and denatured in boiling water for 5 min. The proteins were separated on 10% SDS-PAGE and transferred to nitrocellulose filter membranes (Millipore, Cat# HATF00010). Primary antibodies used were mouse monoclonal anti-HA (Sigma, Cat# H9658), mouse monoclonal anti-PGK1 (Abcam, Cat# a b113687), ra bbit polyclonal anti-Rad53 (Abcam, Cat# ab104232), guinea pig anti-Red1 polyclonal antibody against amino acids 492-709 of budding yeast Red1 protein was pr epar ed by Dai-an Biological Technology In-Nucleic Acids Research, 2023, Vol. 51, No. 15 7917 corporation (Wuhan, China). The quantification of signal intensity was performed using ImageJ software.

DNA physical assay
DSBs , dHJs , crossovers , and non-crossovers were analyzed by DNA physical assays in combination with Southern blot as described previously (39)(40)(41). Briefly, the genomic DNA was extracted from synchronized yeast cells at desired time points in SPM and digested with Xho I. Onedimension (1D) gel electrophoresis was carried out, followed by a Southern blot to detect DSBs , crossovers , and non-crosso vers. Tw o-dimension (2D) gel electrophoresis in combination with Southern blot was performed to detect IH-dHJs and IS-dHJs. For both crossovers and noncrossovers analysis, the genomic DNA was digested with Xho I and NgoM IV, separated by 1D gel electrophoresis, and detected by Southern blot. The probe for the Southern blot was labeled with ␣-32 P-dCTP by a random labeling system (Thermo Fisher, Cat#18187-013). All DNA species in Southern blot were visualized using a phosphorimager (the Cyclone Plus Storage Phosphor System, PerkinElmer) and quantified using Quantity One software (Bio-rad).

Native-denaturing 2D gel electrophoresis
The length of DSB end resection was analyzed by nati v edenaturing 2D gel electrophoresis followed by the Southern blot as described previously ( 40 , 42 ). Briefly, the genomic DNA of synchronized yeast cells was isolated and digested with Xho I. Subsequently, the DNA was separated on a 0.6% Seakem LE agarose gel in 1 x TBE for 24 h at room temperature for the first-dimension gel electrophoresis, and then the gel was stained with ethidium bromide for 30 min at room temperature. Then the gel slices containing targeted DNA were excised and used for the second-dimension denaturing gel electrophoresis with 1 × alkaline running buffer (50 mM NaOH, 1 mM EDTA) for 30 h in a cold room. The DNA species were detected by Southern blot and visualized using a phosphorimager (the Cyclone Plus Storage Phosphor System, PerkinElmer).

Pulse field gel electrophoresis (PFGE) analysis
The synchronized yeast cells were collected at appropriate time points in SPM, and then the genomic DNA was prepared in plugs of low-melting agarose to avoid shearing as previously described ( 43 , 44 ). For PFGE, chromosomes were separated with a CHEF-DRIII PFGE system (Bio-Rad) with the following conditions: 1% agarose in 0.5 × TBE; 15.1 s initial switch time; 25.1 s final switch time; switch angle 120 • ; 6 V / cm; 14 • C; running time 28 h. DNA was detected by Southern blot and signals were visualized using a phosphorimager (the Cyclone Plus Storage Phosphor System, PerkinElmer) and quantified by ImageJ software ( https://imagej.nih.gov/ij ).

Yeast alkaline comet assay
The yeast alkaline comet assay was carried out as previously described with minor modifications ( 45 ). Synchronized yeast cells were collected and processed into spheroplasts b y Zymoly ase 100T in S buffer (1 M sorbitol, 25 mM KH 2 PO 4 , pH 6.5). The spheroplasts were collected by centrifugation and washed twice with ice-cold S buffer. The r esultant pellets wer e r esuspended in 1.5% (wt / v ol) lo wmelting agarose and spread over a slide coated with 0.5% (wt / vol) normal-melting agarose. The slides were solidified in a cold room. Subsequently, the slides were submerged in lysis buffer (30 mM NaOH, 1 M NaCl, 0.05% laurylsarcosine, 50 mM EDTA, and 10 mM Tris-HCl, pH 10.0) for 20 min at 4 • C and then in electrophoresis buffer (30 mM NaOH, 10 mM EDTA, and 10 mM Tris-HCl, pH 10.0) for 20 min at 4 • C. The samples were electrophoresed using the same electrophoresis buffer for 8 min at 0.5 V / cm a t 4 • C . Subsequently, the slides were incubated in neutraliza tion buf fer (10 mM Tris-HCl, pH 7.4) for 10 min, 76% ethanol for 5 min, and 96% ethanol for 5 min at room temperatur e. Samples wer e stained with DAPI and images were acquired using a Zeiss fluorescence microscope (AxioImager.Z2) with an EMCCD camera.

Flow cytometry
One hundred microliters of synchronized cells were collected and fixed in 1 ml of 60% ethanol with 0.1 M sorbitol. The fixed cells were washed twice with 0.5 ml of 50 mM Tris-HCl (pH 7.5) and resuspended in 0.5 ml of 50 mM Tris-HCl with 0.1% Tween 20 (Sigma, Cat#P9416), 0.1 mg / ml RNase A (Sigma, Cat#R6513), and 0.5 l of SYTOX ™ Green dead cell stain (Thermo Fisher, Cat#S34860). The cells were incubated at 25 • C for 12 h in a dark room with 150 rpm shaking. For each sample, 20 000 cells were sorted by BD LSRFortessa and the result was analyzed by FlowJo (Version 10.8.1).

Processing of illumina sequence data
The raw reads were filtered based on the quality value (q25) by Trim-galore (version 0.6.6). BOWTIE2 (Version 2.4.2 ( 47 )) was used to align the filter ed r eads to the S. cerevisiae r efer ence genome (SacCer3) and gener ate B AM files. To analyze the RPA distribution, the reads of BAM files were normalized to counts per million (CPM) by using DEEPTOOLS (Version 2.30.0 ( 48 )). The enrichment of RPA reads around Spo11-oligo hotspots was detected by computeMatrix (Version 3.5.0 ( 48 )). To analyze the correlation between RPA reads and the counts of Spo11-oligo, the genome was divided into 1kb / bin by MultiBam Summary (version 3.5.0 ( 48 )) and then the Pearson correlation analysis was performed. MACS2 (Version 2.2.7.1 ( 49 )) was used for peak calling with the following parameters: Nomodel mode, 200 bp extension size, the cutoff of FDR (q-value) at 0.05.

Dna2 is essential for normal meiosis
Dna2 is a highly conserved nuclease and helicase in eukaryotic cells and plays multiple roles in DNA metabolism during mitotic growth, including DNA replication, DSB end r esection, stalled r eplication fork processing, and mitochondrial genome integrity maintenance (50)(51)(52)(53). Howe v er, the function of Dna2 in meiotic cells is unknown. First, the expression of Dna2 during meiosis was examined. For this purpose, the endogenous Dna2 was tagged with a 3 × HA epitope at its C-terminus. For WT, in a well-synchronized meiotic culture in the sporulation medium (SPM), cells go through each step of meiosis at fixed time points ( Figure  1 A) ( 34 , 39 , 40 , 54 ). According to previous reports, most cells complete pre-meiotic DNA replication and undergo DSBs at ∼3 h in SPM (leptotene) in a standard meiotic timecourse ( 39 , 55 ). Accompanied by DSB repair, the synaptonemal complex (SC) begins to assemble between homologs and cells gradually reach zygotene at ∼4 h in SPM ( 39 , 55 ). When SC extends to the full homolog axes at ∼5 h in SPM, cells enter pachytene. Most, if not all, DSBs are repaired, and homologous recombination is completed during pachytene ( 2 , 40 , 41 ). Afterward, SC begins to disassemble from chromosomes and nuclei undergo the first division to segregate homologs at ∼6h in SPM ( 39 , 55 ). Western blot with synchronized meiotic samples showed that Dna2 was constituti v ely e xpr essed during meiosis (Figur e 1 B, top). Subsequentl y, we anal yzed the dynamics of Dna2 foci in WT thr ough chr omosome spread and immunofluorescence staining (Figure 1 C, D). The results showed that the number of Dna2 foci significantly increased at SPM 5 h and was maintained at a high le v el at SPM 6h, then dramatically decreased, implying that Dna2 is chromosome bound around pachytene.
To further investigate the functions of Dna2 in meiosis, we replaced its nati v e promoter with the CLB2 promoter, which is acti v e only in mitosis but not in meiosis ( 26 , 31 ). This meiosis-specific depletion mutant, pCLB2-HA3-DNA2 , was simply named dna2-md . In dna2-md , Dna2 protein was quickly depleted to an almost undetectable le v el after 1h upon meiosis induction in SPM (Figure 1 B, bottom). DAPI staining of synchronized meiotic samples showed that dna2-md timely underwent nuclear division (time course) and the frequency of nuclear division also reached the WT le v el (92% versus 95%) (Figure 1 A). Moreover, dna2-md sporulated as efficiently as WT (93% in both WT and dna2-md ) (Figure 1 E). Howe v er, when the spore viability was analyzed through tetrad dissection, it dramatically decreased to 4.6% in dna2-md (95.9% in WT), indicating that Dna2 is r equir ed for normal meiosis (Figure 1 F).
Inviable spores are mainly caused by aberrant chromosome segregation ( 56 ). We wondered whether dna2md showed an increased frequency of chromosome missegregation. To this end, we integrated the red fluorescent protein (RFP) and the yellow fluorescent protein (YFP) dri v en by a spore-specific promoter into a pair of chromosomes 9, respecti v ely (Supplementary Figure S2A; Materials and Methods). After meiosis, two spores show yellow fluorescence and the other two spores show red fluorescence in a tetrad. Howe v er, aberrant yellow / red fluor escence patterns r epr esent mis-segr egated chromosome 9. The frequency of chromosome 9 mis-segregation was defined as the number of tetrads with mis-segregated chromosome 9 divided by the number of total tetrads examined. In WT, the frequencies of chromosome segregation errors were very low in both meiosis I (MI) and meiosis II (MII) (a total of 1.7%). Howe v er, in dna2-md , the frequencies significantly increased to 3.2% and 24.6% in MI and MII, respecti v ely, with the total mis-segregation frequency reaching 27.8% ( Figure 1 G and Supplementary Figure S2B). The high frequency of chromosome mis-segregation contributes to the extremely low spore viability in dna2-md.

Both the nuclease and helicase activities of dna2 are required during meiosis I
The spore-specific assay showed that chromosome missegregation mostly occurred at MII but not MI in dna2-md . This may suggest that Dna2 is r equir ed for MII. Howe v er, it is also possible that the accumulated errors in MI lead to the high chromosome mis-segregation frequency in MII.   We tried to determine the time window for Dna2 r equir ed for meiosis through two different systems.
First, Dna2 was degraded at different time points during meiosis. An auxin-inducible degron (AID) was fused to the C-terminus of Dna2 by the PCR-mediated method, hereafter Dna2-AID ( 31 , 32 ) (Figure 1 H). When induced by copper, Dna2-AID expressed OsTIR1, an E3 ligase. Dna2-AID was ubiquitinated by an OsTIR1-containing E3 ubiquitin ligase complex and degraded by proteasomes in 15-30 min in the presence of auxin (IAA) and copper ( 32 ) (Supplementary Figure S3A). In synchronized cultures, Dna2-AID was efficiently removed by adding 2 mM IAA and 25 M CuSO 4 in SPM (Figure 1 I). A low le v el of spore viability was observed when IAA and CuSO 4 were added at 0 to 4 h in SPM (Figure 1 J). Howe v er, the spore viability was significantly increased to 58.6% and 68.5% when IAA and CuSO 4 were added at 5 and 6 h post SPM incubation, respecti v ely (Figure 1 J). Most cells have already reached or passed pachytene during this period ( 39 , 55 , 57 ). These results suggest that Dna2 is r equir ed at / befor e pachytene (5-6 h in SPM) to ensure successful meiosis and spore viability.
Second, one DNA2 nati v e promoter was replaced by the CLB2 promoter (to express Dna2 during v egetati v e growth but not meiosis) and the other DNA2 promoter was replaced by the GAL1 promoter, hereafter pCLB2-DN A2 / pGAL1-DN A2 (Figure 1 K). This strain also contained an estradiol-induced transcriptional cascade ( 58 ).  (Figure 1 M). These results further confirm that Dna2 is r equir ed during prophase I (most likely ∼5 h, i.e. around pachytene, and not late than 6 h in SPM) to ensure successful meiosis.

Abundant accumulation of RPA in dna2-md through spo11dependent and -independent manners
Gi v en that the nuclease activity of Dna2 is stimulated by RPA in in vitro analysis ( 22 , 64 ), subsequently, we examined RPA localization in WT and dna2-md by chromosome spread of meiotic nuclei in combination with immunostaining using an anti-Rfa1 antibody (Figure 2 A). The RPA signal was detected as individual foci in WT from 2 h in SPM, peaked at 4h ( ∼37 foci per nucleus), and then gradually decreased to a low le v el at 7 h ( ∼12 foci) when the majority of nuclei completed MI (Figure 2 A, B). The dynamics of RPA foci reflect the dynamics of the formation and repair of meiotic DSBs ( 65 ).
In dna2-md , two types of morpholo gicall y different RPA signals were detected: individual RPA foci as in WT and long RPA lines rarely seen in WT (Figure 2 A). These RPA lines were often observed at late time and it seemed that they resulted from many RPA molecules assembled on long ss-DNA strands in dna2-md . In dna2-md , RPA foci were also detected at 2 h in SPM and the number gradually increased to a comparable level at 4h as in WT (38 versus 37 foci).
Howe v er, in contrast to WT, the number of RPA foci continuously increased to a higher le v el at 6 h ( ∼55 foci) and was maintained at this le v el in dna2-md (Figure 2 B). The number of RPA lines was hard to be quantified since they had different lengths and were usually tangled together. Alternati v ely, the frequency of nuclei bearing RPA lines was examined. The RPA lines were observed in several nuclei as early as 2 h in SPM, and the frequency of nuclei having RPA lines gradually increased to ∼70% at 6 h in SPM (Figure 2 C).
Subsequently, to examine whether the accumulated RPA foci and / or lines in dna2-md were associated with meiotic r ecombination, they wer e further examined in dna2md spo11Y135F and spo11Y135F m utants, w here the DN A cleavage activity of Spo11 was abolished and thus the formation of meiotic DSBs did not occur ( 3 ). As expected, the RPA signal was rarely observed in the spo11Y135F mutant (Figure 2 A, B). Howe v er, RPA foci were observed from 2 h in SPM and their number was increased to a high le v el at 4 h in dna2-md spo11Y135F , although it was lower than in dna2-md (18 versus 55 foci; Figure 2 A, B). This suggests ∼1 / 3 RPA foci in the dna2-md form in a Spo11-independent manner. Moreover, nuclei with RPA lines were observed in dna2-md spo11Y135F as in dna2-md with a similar frequency (Figure 2 C). These results suggest that RPA accumulates in dna2-md in Spo11-dependent and -independent ways.
The existence of RPA lines was confirmed by further analysis under a structured illumination microscope (SIM). In WT, only RPA foci were observed on Rec8 labelled chromosome axes. Howe v er, there were lots of RPA lines and it seemed that they extruded from Rec8-axes outward in dna2md (Figure 2 D). Further, RPA signal intensity was quantified at a per-nucleus le v el (Figure 2 E). As expected, the intensity of the RPA signal in dna2-md was much higher than in WT. Interestingly, the RPA intensity in dna2-md spo11Y135F was also significantly higher than that in WT (Figure 2 E). This is probably due to each RPA line containing a large number of RPA molecules.
Previous studies suggest that Dna2 is required for DNA replication in mitosis although one recent study showed the role of Dna2 in DNA replication is very limited ( 25 , 50 , 66 , 67 ). To characterize whether the accumulated RPA result from impaired pre-meiotic DNA replication in dna2-md , the nati v e promoter of CDC6 was replaced by SCC1 to express Cdc6 only in mitosis but not in meiosis, named cdc6-md . In this strain, DNA replicates normally during mitotic growth but the pre-meiotic DNA replication is almost eliminated ( 30 ). The kinetics and efficiency of DSB formation and meiotic recombination in cdc6-md are also comparable with that in WT ( 30 , 68 , 69 ). Subsequently, we used chromosome spread of meiotic nuclei in combination with immunostaining to characterize the RPA signal in cdc6-md and dna2-md cdc6-md (Figure 2 F). The number of RPA foci reached 55 in dna2-md cdc6-md as in dna2-md , w hich was significantl y higher than WT (37 RPA foci) and cdc6-md (26 RPA foci) (Figure 2 G). Furthermore, the percentage of cells with RPA lines in dna2-md cdc6-md was comparable to that in dna2-md , which was dramatically higher than that in cdc6-md (Figure 2 H). These findings suggest that the abundantl y accum ulated RPA in dna2-md is not due to impair ed pr e-meiotic DNA r eplication. Consistently, the dynamics of pre-meiotic DNA replication in WT and dna2-md strains were comparable as revealed by flow cytometry analysis (Supplementary Figure S4).

Accumulated RPA expands from DSBs to broader regions in dna2-md
The result of immunostaining showed that abundant RPA accumulated in dna2-md . To explore the effect of Dna2 on RPA distribution, we performed a genome-wide ChIPseq in WT, dna2-md and dna2-md spo11Y135F . In WT, along with meiotic r ecombination progr ession, the number of RPA foci reaches the peak at 4h and dramatically decreases to a low le v el at 7h in SPM (Figure 2 A, B; ( 65 ). Consistentl y, RPA was highl y enriched around Spo11 oligoenriched regions at 4h, and its enrichment was moderately decreased at 7h in WT when most DSBs were repaired. In dna2-md , a similar enrichment of RPA at Spo11-oligo sites was observed at 4h as in WT (Figure 3 A, B, and Supplementary Figure S5). Intriguingly, in contrast to WT, the distribution of RPA was greatly expanded to broader regions from DSB hotspots at 7h in dna2-md , as shown at ERG25 and BAT1 DSB hotspots (Figure 3 A, B, and Supplementary Figure S5). This result suggests that programmed DSB repair is defecti v e in dna2-md . In dna2-md spo11Y135F , no obvious enrichment for RPA at Spo11 oligo sites was observed (Figure 3 A, B, and Supplementary Figure S5), which is consistent with the absence of DSBs in this strain and further confirms that RPA enrichment at Spo11-oligo sites in WT and dna2-md depends on DSBs.
RPA enrichment at DSB flanks was further analyzed in detail at another four DSB hotspots in dna2-md (Figure 3 C). Consistent with the preceding analysis, (i) RPA was highly enriched around all four DSB hotspots at 4 h in WT and dna2-md but not in dna2-md spo11Y135F , (ii) RPA enrichment around DSB hotspots was gr eatly r educed in WT but only slightly decreased in dna2-md at 7 h and (iii) RPA enrichment extended to broader regions flanking DSBs at 7 h in dna2-md but not in WT.
Subsequently, the correlation between RPA reads and Spo11 oligo counts was analyzed by dividing the genome of S. cerevisiae into 1 kb bins (Figure 3 D-I). As expected, RPA reads and Spo11 oligo counts exhibited stronger correlations in WT than in dna2-md ( r = 0.62 and 0.47 at 4h and 7h, respecti v ely, in WT; r = 0.56 and 0.33 at 4 and 7 h, respecti v ely, in dna2-md ) since RPA expanded to broader regions from DSB sites in dna2-md (Figure 3 D-G). Howe v er, no obvious correlation was detected between the RPA reads and Spo11 oligo counts in dna2-md spo11Y135F ( r = 0.16 and 0.10 at 4 and 7 h, respecti v ely) (Figure 3 H, I). This is consistent with the Spo11-independent accumulation of RPA in dna2-md spo11Y135F .

The amount of DSB formation, the length of DSB end resection, and the levels of cr ossov er / non-cr ossov er are comparable to WT, but DSB repair is defective in dna2-md
Abundant RPA accumulation detected cytologically and by ChIP-seq implies that the process of meiotic recombination and DSB repair is defecti v e in dna2-md . Since meiotic recombination is initiated by programmed DSBs, we first examined the DSB formation at the well-characterized HIS4-LEU2 hotspot by a standard southern hybridization analysis in WT and dna2-md (Figure 4 A) ( 39 ). To avoid the DSB turnover issue and accurately measure the DSB le v el, the experiment was performed in a rad50S (K81I) background, which blocks the DSB end resection and accumulates DSBs. At HIS4LEU2 , the accumulated DSB le v el reached ∼24% of total DNA in both WT and dna2-md (Figure 4 B, C). DSBs were also examined with PFGE (Plus field gel electrophoresis) followed by southern hybridization at the chromosomal scale. Again, similar DSB le v els and patterns were detected between rad50S and dna2-md rad50S on chromosomes III, V and XI, respecti v ely (Figure 4 D). These results suggest that Dna2 is not r equir ed for efficient DSB formation.
Soon after DSB formation, their 5 ends are resected to generate 3 ssDNA ends. The length of these 3 ssDNA ends and thus resection le v el can be observ ed and e valuated on a nati v e-denatur ed 2D gel (Figur e 4 A, gr een box, and Figure 4 E) ( 42 ). For this purpose, the genomic DNA of WT and dna2-md was extracted and digested with Xho I. The digested DNA was analyzed by nati v e electrophoresis in the first dimension and followed by denatured electrophoresis thr ough sodium hydr oxide in the second dimension ( 40 , 42 ). The resection length was analyzed after the southern blot. No obvious difference in DSB end resection was observed between WT and dna2-md (Figure 4 F). This result suggests that Dna2 has no or little role in the 5 end resection of meiotic DSBs. This is consistent with the previous finding that Exo1 is the main nuclease for DSB end resection, but Sgs1 has no contribution to this process in S. cerevisiae meiosis, although the Sgs1 / Dna2 complex has an important role in mitotic DSB end resection ( 42 , 51 , 70 , 71 ).
Resected DSB ends invade homologous templates and are finally repaired as crossovers or non-crossovers. The recombina tion intermedia te dHJ (double Holliday junction) and final pr oducts, cr ossovers and non-cr ossovers, can also be examined at the HIS4LEU2 hotspot (Figure 4 A) ( 39 , 40 ). Both the inter-homolog and inter-sister double Holliday junctions (IH and IS-dHJ) were detected on 2D gels in combination with southern hybridization. Quantification analysis showed the ratio of IH-dHJ to IS-dHJ was ∼6:1 in both WT and dna2-md (Figure 4 G). The 1D gels in combination with southern hybridization showed ∼20% crossover DNA and ∼1:1 ratio of crossover to non-crossover in both WT and dna2-md (Figure 4 H-K). These results suggest that Dna2 is not r equir ed for the formation of crossovers or noncrossovers, at least at the HIS4LEU2 hotspot.
Synapsis is another hallmark of meiosis. The possible role of Dna2 in synapsis was also examined. To avoid the possible asynchronization issue, an ndt80 Δ was introduced into the dna2-md strain to arrest cells at pachytene ( 75 ). The synapsis was evaluated by the morphology of the SC central element Zip1 ( 35 , 65 ). During meiosis, the morphology of Zip1 can be divided into four classes: no signal (Class I), dotty (Class II), partially elongated (Class III), and fully elongated (Class IV). In ndt80 Δ, when Zip1 was visualized in spread nuclei collected at 8 h in SPM, 98% of the nuclei had a detectable Zip1 signal. The Class I-Class IV nuclei took up 2%, 9%, 19% and 70%, respecti v ely (Supplementary Figure S6E, F). Similar frequencies of nuclei for each class were observed in dna2-md ndt80 Δ like in WT (4%, 6%, 23% and 66%, respecti v ely) (Supplementary Figure S6E, F). Ther efor e, these r esults indica te tha t Dna2 has no or little role in synapsis.
Taken together, our findings suggest that accumulated RPA have a limited effect on the amount of DSB formation, the length of DSB end resection, the le v els of cr ossover / non-cr ossover, and synapsis in dna2-md . Howe v er, the high frequency of chromosome mis-segregation and extremely low spore viability imply that accumulated RPA likely impair the meiotic DSBs repair in dna2-md . To confirm this idea, PFGE in combination with southern hybridization was performed to examine the extent of DSB r epair on thr ee r epr esentati v e chromosomes with different lengths, III, V, and XI. At 0 h in SPM, almost no DNA fragment was detected ( Figure 5 A-C, lane 1, '0 h'). Abundant DNA fragments were observed at 4 h in SPM due to meiotic DSBs in WT ( Figure 5 A-C, lane 2). The abundance of DNA fragments was gradually decreased, accompanied by DSB repair, and they were barely detectable at 8 h in SPM ( Figure 5 A-C, lanes 3-6). A similar le v el and pattern of DNA fragments were observed at 4h in dna2-md as in WT ( Figure 5 A-C, lane 8, red versus green curves), suggesting DSB formation is not affected in dna2-md . Howe v er, there was still a high le v el of DNA fragments at 8h in dna2-md ( Figure 5 A-C, lane12, purple versus blue curves), which indicates defecti v e DSB repair. This observation is consistent with the Spo11-dependent accumulation of RPA in dna2md . To further confirm the defecti v e DSB repair in dna2md , we performed the comet assay (Supplementary Figure  S7A-C). At 4h in SPM, most nuclei in both WT and dna2md (86.6% versus 84.0%) showed obvious comet tails, indicating se v er e DNA damage. At 7 h in SPM, comet tails wer e only observed in a few WT nuclei (6.9%), implying efficient DSB repair. Howe v er, a high frequency of nuclei with obvious comet tails was still observed at 7 h as at 4 h in dna2-md (86.7% versus 84.0%). This is consistent with the accumulation of RPA and thus impaired DSB repair in dna2-md .
As expected, no / little DNA fragments were observed in either dna2-md spo11Y135F or spo11Y135F m utant, w here the occurrence of programmed DSBs is abolished (Figure 5 D-F). This data further supports the idea that RPA accumulation in dna2-md spo11Y135F is independent of Spo11-mediated DSBs. In addition, when the nuclear division (time course), spore viability, and the frequency of chromosome mis-segregation were analyzed in dna2-md spo11Y135F and spo11Y135F , no significant differences w ere found betw een these two mutant strains, implying that no Spo11-independent DSBs occur in dna2-md spo11Y135F (Supplementary Figure S8A-C). Exogenous DNA lesions trigger Rad53 phosphorylation, which can be seen as alter ed electrophor etic mobility in spo11 Δ meiotic cells (Figure 5 G, top panel) ( 76 ). Consistent with our above idea that ther e ar e no Spo11-independent DSBs in dna2-md and dna2-md spo11Y135F , no Rad53 phosphorylation signal was detected in these strains (Figure 5 G, bottom panel). Our results collecti v ely re v eal that defecti v e meiotic DSB repair occurs in dna2-md .

Dna2 r emo ves RPA filaments gener ated during meiotic recombination-associated DNA synthesis to maintain genome integrity
Our above results suggest a pparentl y normal crossover and non-crossover formation in dna2-md . However, the PFGE result showed most DSBs were unrepaired. This paradox can be reconciled if new DNA is synthesized from DSB ends and extended longer than the restriction cleavage sites used for 1D and 2D gel analysis. This would gi v e seemingly intact crossovers and non-crossovers at a local scale, howe v er, sim ultaneousl y produce long DN A fla ps w hich would be visualized as long RPA lines and unr epair ed ssDNA breaks at a large scale if they are not timely removed (see Discussion) ( 17 , 22 ). If so, induced expression of Dna2 at pachytene may remove these accumulated RPA in dna2-md . To test this idea, ndt80 was introduced into the pCLB2-DN A2 / pGAL1-DN A2 strain to arrest cells at pachytene (Figure 6 A). As expected, a considerable number of RPA foci was observed in ndt80 Δ, and abundant RPA foci and lines were observed and maintained in pachytene-arrested nuclei in pCLB2-DN A2 / pGAL1-DN A2 ndt80 Δ (Figure 6 A,  B). When ␤-estradiol was added to induce Dna2 expression in the latter strain, almost all RPA signal disappeared quickly (Figure 6 A, B). Further, to examine whether the removal of accumulated RPA at pachytene could rescue spore viability, we sim ultaneousl y induced the expression of Dna2 and Ndt80 after pachytene arrest, and thus the strain could complete meiosis (Figure 6 C, D). In this case, RPA was dramatically decreased and the spore viability reached up to 67% (Figure 6 E). These results suggest that Dna2 induction at pachytene can largely remove accumulated RPA and rescue spore viability. The lower spore viability in this strain compared with WT (67% versus 96%) may be due to the late expression of Dna2, which cannot timely remove RPA and thus pre v ent the timely repair of DSBs before pachytene exits. These results support the proposal that Dna2 removes ssDNA-RPA filaments at / before pachytene for DSB repair and thus spore viability.  Subsequently, we further explored whether the accumulated RPA in dna2-md result from meiotic recombinationassociated DNA synthesis. Pif1 is known to stimulate Pol ␦ strand displacement activity during homologous recombination in mitosis, and a recent study discovered that Pif1 is also involved in producing long DNA synthesis tracts during meiotic recombination in mlh2 Δ ( 16 , 17 , 77 , 78 ). The meiosis-specific depletion of Pif1 ( pif1-md ) would decrease the activity of Pol ␦, resulting in less-extended D-loops and fewer ssDN A-RPA filaments, w hile, the removal of Pif1 in-hibitor Mlh2 w ould o ver-stimulate Pol ␦, resulting in overextended D-loops and more ssDNA-RPA filaments in dna2md. These pr edictions wer e confirmed by the fact that fewer RPA accumulation in pif1-md dna2-md but more RPA accumulation in dna2-md mlh2 Δ were observed (Figure 7 A,  B). Mor eover, spor e viability was 84% in pif1-md dna2-md and 86% in pif1-md dna2-md mlh2 Δ (85% in pif1-md ) but only 0.8% in dna2-md mlh2 Δ (95% in mlh2 Δ) (Figure 7 C). Consistent with the above proposal, the mean length of heteroduplex DN A (hDN A) tracts is decreased although  dna2-md, pif1-md , pif1-md dna2-md , mlh2 Δ, mlh2 Δ dna2-md and pif1-md dna2-md mlh2 Δ. Sample size, n = 143, 62, 58, 50, 113, 94, 165, 175,  179, 160, 168, 91, 119, 169, 192, 148, 140, 113, 85, 158, 238, 153, 120, 100, 55, 176, 208, 161, 106, 112, 54, 140, 151, 153, 106 and 107 , n = 132, 136, 148,  137, 128, 144, 123, 154, 132, 141, 93, 106, 151, 128, 109, 136, 120, 128, 80 and 131 nuclei, respecti v ely. Error bar, SD (B, D, F), SEM (G) or 95% confidence interval (C). Scale bar, 5 m (A, E). Two-tailed Student's t -test (B, D, F) or Two proportion Z-test (C); ns (not significant), * ( P < 0.05), ** ( P < 0.01) and *** ( P < 0.001). slightly in a pif1 mutant pif1m2 (1.2 kb) and pif1m2 mlh2 Δ (1.2 kb), while it is dramatically increased in mlh2 Δ (3.0 kb) compared with that in WT (1.3 kb) when they were analyzed with publicly available re-sequenced spores (Figure 7 D).
Further, we anal yzed w hether the accum ulation of RPA r equir es new DNA synthesis by adding a ribonucleotide reductase inhibitor, hydroxyurea (HU), at 4 h in SPM to deplete dNTPs and thus block DNA synthesis for DSB repair. The accumulation of RPA reached its peak at ∼4 h in SPM and then gradually decreased along with DSB repair in WT, howe v er, the RPA signal reached a much higher le v el in dna2-md at late time (Figures 2 A, B and 7E-G). In WT cells, the addition of HU at 4 h resulted in a faster decrease of the RPA signal. In dna2-md cells, a quick decrease in RPA signal (the number of RPA foci and the frequency of cells with RPA lines) was observed in the presence of HU although it was further accumulated in the absence of HU (Figure 7 E-G). The different RPA dynamics caused by HU probably result from the combined effects of (i) the depletion of dNTPs and block of DN A synthesis, w hich prevents further accumulation of RPA, and (ii) RPA displacement by Rad51 / Dmc1. Taken together, our findings support the idea that Dna2 is r equir ed for removal of ssDNA-RPA filaments formed during meiotic recombina tion-associa ted DNA synthesis.

Spo11-independent RPA distribution is regulated by dna2
Our cytological observations re v ealed that a considerable le v el of RPA signal was Spo11-independent in dna2-md . Detailed analysis of our ChIP-seq results showed that ∼20% of RPA peaks were distributed outside of Spo11-oligo regions and at centromeres and tRNA regions in WT (Supplementary Figure S9A). Interestingly, a large fraction of RPA was enriched in type I and II (but not III-V) retrotransposons in dna2-md (Supplementary Figure S9A, B). Moreover, the enrichment of RPA at retrotransposons was detected more clearly at 7h than at 4h (Supplementary Figure S9A, B). Howe v er, almost no RPA enrichment was observed at retrotransposons in WT at either 4 or 7 h (Supplementary Figure  S9A, B). These results suggest that Dna2 pre v ents RPA accumula tion a t types I and II retr otransposons. This r ole of Dna2 is independent of Spo11-mediated DSBs since similar RPA enrichment was observed in the dna2-md spo11Y135F double mutant. Compared with WT, RPA also tended to enrich at telomeres in dna2-md and dna2-md spo11Y135F (Supplementary Figure S9C, left panel). These findings imply that, as in mitosis, Dna2 plays a critical role at telomeres in meiosis ( 79 , 80 ). We also noted that RPA was enriched at rDNA in dna2-md and dna2-md spo11Y135F but not in WT (Supplementary Figure S9C, right panel). We observed a number of RPA peaks in dna2-md spo11Y135F enriched a t regions tha t are Spo11 oligo hotspots in WT , especially at 4h in SPM. These results suggest that Dna2 also regulates Spo11-independent RPA distribution.

Dna2 r emo ves ssDNA-RPA filaments f ormed during meiotic recombination-associated DNA synthesis
We showed that Dna2 is essential for meiosis. When Dna2 is meiosis-specificall y depleted ( dna2-md ), RPA dramaticall y extends from DSB sites to broad r egions, r esulting in unpaired DSBs as re v ealed by PFGE and abundant accumulation of RPA, ultimately chromosomes show a high frequency of mis-segregation and spores are inviable. Consistent with a putati v e role of Dna2 in cleaving ssDN A fla ps during mitosis, the Dna2 nuclease-dead mutant showed the same defects as the dna2-md , and pachytene induction of Dna2 can effecti v ely remov e accumulated RPA and r estor e spor e viability. Ther efor e, we propose that Dna2 pre v ents / remov es ssDNA-RPA filaments formed during meiotic recombina tion-associa ted DNA synthesis ( Figure  8 ). The role of Dna2 in this model is also consistent with its role in mitosis to cleave ssDN A fla ps during lagging strand synthesis ( 50 ). Our model is further supported by the following evidence. (1) The accumulation of RPA r equir es new DNA synthesis since the addition of HU to deplete dNTPs pre v ents the abundant accumulation of RPA signal.
(2) Depletion of the Pol ␦ activator Pif1 in dna2-md prevents RPA accumulation and r estor es spor e viability. Mor eover, removing a Pif1 inhibitor, Mlh2, which is predicted to cause D-loop ov er-e xtension, results in RPA ov er-accumulation in dna2-md . Additionally, the mean hDNA tract length in pif1m2 (1.2 kb) and pif1m2 mlh2 Δ (1.2 kb) is comparable to WT (1.3 kb), while it is dramatically increased in mlh2 Δ (3.0 kb). Howe v er, we also found that dna2-md shows the normal le v el of DSBs and meiotic recombination progresses normally to generate WT le v els of recombination products, i.e. COs and NCOs, when examined by Southern blot at HIS4LEU2 and ERG1 hotspots. This result seems to be contradicted with our above proposal.
Howe v er, this paradox can be easily reconciled as discussed below.
During meiotic recombination, Pol ␦ catal yzes DN A synthesis ( 16 ). When this newly synthesized strand encounters the second DSB end, the strand displacement activity in combination the polymerase activity of Pol ␦ generates fla pped ssDN A w hich is coated by RPA to form ssDN A-RPA filaments. Dna2 r emoves / pr events ssDNA-RPA filaments for timely nick ligation to r estor e genome integrity.
Howe v er, in the absence of Dna2, continuousl y DN A synthesis causes the newly synthesized strands continuously grow and displace the parental strands, which ultimately results in the generation of longer ssDNA-RPA filaments and the migration of DNA nicks far away from the original DSB sites (Figure 8 ). In Southern blot, a small region spanning DSBs ( ∼3 kb on each side of HIS4LEU2 and ∼5 kb at ERG1 ) is analyzed to detect DSBs and recombination products (Figure 4 A) ( 72 ). When newly synthesized DNA strands extend out of the analyzed regions, seemingly repaired DSBs (and recombination products) are detected on 1D and 2D gels (Figure 4 H-K, and Supplementary Figure  S6A, B). Consistently, CO-associated Zip3 foci, assumed to mark dHJ, are also observed at the WT level (Supplementary Figure S6C, D). Howe v er, these persistent ssDNA-RPA filaments pre v ent nick ligation and thus chromosome fragments are observed by PFGE (Figures 5 A-C).
Although Dna2 plays an important role in DNA replication during mitosis, our study does not show an observable defect in pre-meiotic DNA in dna2-md . (i) The dynamics of pre-meiotic DNA replication in WT and dna2-md are comparable. (ii) When pre-meiotic DNA replication is abolished in dna2-md cdc6-md , abundant RPA accumulates  Figure S1 and the text for a detailed description of the DSB repair model. In this model, Pol ␦ probably catalyzes new DNA synthesis and thus the formation of ssDNA through its strand displacement activity. Pif1 functions as an activator of Pol ␦ and is also involved in this process. In WT, ssDNA-RPA filaments are removed by Dna2 and then the nick is ligated for complete meiotic DSB repair. Howe v er, in dna2-md , the persistent ssDNA-RPA filaments block nick ligation and (1) result in extended DNA synthesis and thus 'seemingly normal' crossovers and non-crossovers when examined at a short scale (e.g. on 1D and 2D gels at hotspots), and (2) produce DNA fragments seen on PFGE due to closely spaced DNA nicks on opposing strands, newly synthesized DNA strand encounters a nick on the template strand, or two synthesized DNA strands collide. The green arrows at the bottom indicate the restriction enzyme cleavage sites used to examine DSBs and recombination at a hotspot. as in dna2-md . This suggests the recombination defects in dna2-md are not resulted from defects in pre-meiotic DNA replication. (iii) The abundant accumulation of RPA signal r equir es de novo DNA synthesis during meiotic recombination as re v ealed by the HU e xperiments. Howe v er, our results do not mean that Dna2 has no role in pre-meiotic DNA replication since a low le v el of Dna2 in dna2-md may be enough for this process.
In summary, our findings suggest that Dna2 mainly r emoves / pr events ssDNA-RPA filaments generated from newly synthesized DNA during meiotic recombination and accumulated filaments are a roadblock that leads to the DSB repair defect. Howe v er, there are still important issues that remain elusi v e and are worth further investigation.
(1) Our findings in combination with pr evious r eports suggest that Pif1 stimulates Pol ␦ to generate DN A fla ps, which are coated by RPA, and then these RPA filaments have to be removed by Dna2 ( 16 ). In the absence of Pif1, abundant RPA accumulation is inhibited and the spore viability is rescued to nearly WT le v el in dna2-md pif1md . Since pif1 depletion can also rescue the lethality of dna2 Δ in mitosis ( 22 , 62 ), one possible interpretation for our result is that inviable spores in dna2-md are due to the mitotic defects after meiosis. Howe v er, this possibility is less likely because (1) Dna2 is expressed in mitosis in dna2-md and (2) dna2-md does show se v ere meiotic defects (above discussion). Ther efor e, we pr efer another interpreta tion tha t in the absence of Pif1, the activity of Pol ␦ is partially inhibited, and thus Dna2 is not necessary during meiosis. Howe v er, this raises the interesting question of why meiosis r equir es Pif1 to generate long DN A fla ps and then r equir es Dna2 to r emove them. (2) Our findings suggest that Pif1, Pol ␦, and Dna2 ar e r esponsible for the formation and removal of ssDNA-RPA filaments during meiotic recombinationassociated DNA synthesis. Howe v er, it is unclear w hether ssDN A-RPA filaments form at the first or the second DSB end, or both. This is most likely related to whether Pol ␦ mediates first or second strand synthesis, or both, though evidence suggests that Pif1 may only be involved in the first-end synthesis ( 16 ). Another critical question is how Pol ␦ collaborates with Pif1, RPA, Dna2, and other factors to timely resolve ssDNA-RPA filaments and maintain genome integrity. . This is consistent with our proposal that Dna2 removes ssDNA-RPA filaments via its nuclease activity. Interestingly, the helicase-dead mutation of Dna2 also showed a moderate le v el of accumulation of RPA and a lower le v el of spore viability ( Figure 1 P and Supplementary Figure S10B-D). This raises the issue that why the helicase activity of Dna2 is also r equir ed for normal meiosis in S. cerevisiae . The physiological function of Dna2 helicase activity remains also elusi v e in mitosis. It seems that the Dna2 helicase activity is required at a small subset of troubled replication forks although it is not r equir ed for bulk DNA replication in mitosis ( 19 , 63 , 81 ). In Dna2 helicase-dead cells in mi-tosis, a low le v el of Rad53 phosphorylation was detected probably due to trouble DNA replication caused DNA damages and this phosphorylation is suppressed by deletion of the DNA damage checkpoint mediator Rad9 ( 63 ). Ther efor e, one possibility is that dna2-hd has a pre-meiotic DNA replication defect as that in mitosis. Howe v er, this is less likely since our results suggest Dna2 is not r equir ed for pr e-meiotic DNA r eplication (see abo ve discussion). Moreo ver, we failed to detect the phosphorylation of Rad53 in dna2-md and dna2-md spo11Y135F (Figure 5 G). Another possibility is that the full activity of nuclease r equir es its helicase activity for an unknown reason. This is also less likely since the helicase-dead mutant used in this study shows full nuclease activity in in vitro studies ( 63 ). It is also possible that Dna2 may use its helicase activity to resolve a short piece of the double-stranded DNA at the base of the RPA filaments for more efficient cleavage. A similar process likely occurs during mitotic recombination, wher e Dna2 r equir es the helicase activity of Sgs1 for efficient DSB end resection ( 51 , 71 ). (4) During meiosis, defects in homologous recombination / synapsis trigger the meiotic prophase checkpoint, resulting in cell cycle arrest for repair and rescue or cell apoptosis (82)(83)(84)(85). The meiotic checkpoint machinery shares many components with the canonical DNA damage response pathway, the most important components among which are the e volutionarily conserv ed sensor kinases, ATM / Tel1 and ATR / Mec1 ( 83 , 85-88 ). Activated ATM / Tel1 and ATR / Mec1 phosphorylate a large set of substrates to either directly implement the checkpoint response or transmit the signal to downstream effectors depending on the nature of the defect and the local environment ( 83 , 84 ). In dna2-md , no defect in synapsis is observed, but RPA-coated ssDNA is accumulated to a high le v el in meiosis. RPA-coated ssDNA is a classical signal triggering DNA damage response in both mitosis and meiosis ( 7 , 89-91 ). Howe v er, it does not acti vate the meiotic checkpoint and nuclei divide timely resulting in inviable spores in dna2-md . How is this possible? Different from other recombination / synapsis-defecti v e mutants (e.g. dmc1 Δ, zip1 Δ, hop2 Δ, mus81 Δ) which show aberrant recombination at various steps and / or synapsis (92)(93)(94)(95), dna2-md shows timely and normal synapsis and recombination at each step (at least at a local scale, also see above discussion): DSB formation, end resection, homolog bias, importantly normal le v els of recombination products when examined at HIS4-LEU2 by Southern blot. This means that the key e v ents monitored by the meiotic recombination checkpoint progress normally in dna2-md . Consistently, the dynamics of Mek1 phosphorylation and dephosphorylation in dna2-md are comparable to that in WT, which is totally different from the persistent phosphorylation of Mek1 as seen in dmc1 Δ (Supplementary Figure  S11A, 76) . In addition, the Mek1-dependent phosphorylation of Red1 is r equir ed to activate the pachytene checkpoint and its dephosphorylation is necessary for pachytene exit ( 96 , 97 ). Similar dynamics of Red1 phosphorylation were also observed between WT and dna2-md , howe v er, a high le v el of phosphorylated Red1 was persistent in dmc1 Δ (Supplementary Figure  S11B). This suggests the accumulated RPA in dna2-md is different from that in dmc1 Δ and probably also other recombination / synapsis-defecti v e mutants . Ther efor e, the accumulated RPA filaments do not activate the checkpoint in dna2-md .

Does dna2 function in other processes in meiotic cells?
Dna2 is r equir ed for DNA r eplication and DSB end r esection in mitosis ( 51 , 98 ). Howe v er, one recent study shows that the function of Dna2 in DNA replication is extremely limited ( 25 ). Ther efor e, the role of Dna2 in mitotic DNA replication is still controversial. In this study, it seems that Dna2 is not r equir ed for pr e-meiotic DNA r eplication as r evealed by the results of flow cytometry and dynamics of nuclear di vision. Howe v er, we cannot rule out the possibility that the residual Dna2 performs these functions efficiently in dna2-md . Meanwhile, the results of nati v e-denatured 2D gel in combination with Southern hybridization indicate that Dna2 is not r equir ed for meiotic DSB resection, which is consistent with previous reports that Exo1 is the core nuclease for long-term DSB resection in S. cerevisiae meiotic cells ( 42 , 70 ). Taken together, we re v eal that unlike in mitosis, Dna2 has no or only very limited roles in DNA replication and DSB end resection during meiosis in S. cerevisiae .
In the absence of both meiotic DSBs and Dna2 ( dna2-md spo11Y135F ), a considerable level of RPA is also observed by imm unostaining. Interestingl y, the ChIP-seq anal ysis reveals tha t RPA accumula tes a t types I and II retrotransposons in dna2-md and dna2-md spo11Y135F, suggesting that Dna2 acti v ely pre v ents RPA accumulation in these regions in a Spo11-independent manner. S. cerevisiae contains fiv e families of LTR (long terminal r epeat) r etrotransposons, including Ty1, Ty2, Ty3, Ty4, and Ty5 elements, which contribute to approximately 3% of the genome (99)(100)(101). These r etrotransposons shar e the basic structur e: two dir ect terminal r epeats flank the TYA and TYB encoding genes, which are analogous to the gag and pol genes of retroviruses ( 100 , 102 ). It is currently unknown why and how RPA onl y accum ula tes a t types I and II but not a t other types of retrotransposons regardless of the presence or absence of meiotic-DSBs. We can imagine se v eral possibilities. (i) For the S. cerevisiae S288C genome, there are 32 copies of Ty1 and 13 copies of Ty2, but only 2 copies of Ty3, 3 copies of Ty4, and 1 copy of Ty5 ( 100 ). The high copy numbers of Ty1 and Ty2 may contribute to the preferential RPA accumulation. (ii) Although there are no meiotic DSBs and the integrity of genomic DNA is not compromised in dna2-md spo11Y135F , it is still possible that a very low level of un-programmed DSBs occurs at types I and II retrotransposons since they are unstable, especially in the absence of Dna2. (iii) Dna2 depletion improves the cDNA stability of retrotransposons in mitotic nuclei ( 62 ). In mitosis, se v eral m utants associated with DN A replication and repair, including Fen1, Sgs1, R ad57, R ad2, R ad50, Mre11, and Tel1, have increased stability of retrotransposon cDNA but not RNA (103)(104)(105). It is very likely that increased cDNA le v els also occur in meiotic nuclei to promote the formation of ssDN A-RPA fla ps. (iv) It is also pos-sible that in Dna2 depleted meiotic nuclei, there is increased transcription at these retrotransposons and the ssDNA is directly coated by RPA to generate ssDNA-RPA filaments, or that increased transcription causes DN A damage, w hich further recruits RPA binding. (v) Another possibility is that in Dna2-depleted meiotic nuclei, RNA transcripts bind to DN A templates (especiall y w hen transcriptional activity is high) and generate R-loops, leaving displaced ssDNA that is then coated by RPA ( 106 ). Further investigations are required to figure out how Dna2 regulates RPA enrichment at Ty1 and Ty2 retrotransposons.
In the depletion of Dna2, RPA also tends to be enriched at telomeres and rDNA. As a result, Dna2 is r equir ed in both meiosis and mitosis to pre v ent ssDN A accum ulation and thus maintain genome stability ( 79 , 80 ). Additionally, a fraction of RPA in dna2-md spo11Y135F locates in regions that are Spo11-oligo hotspots in WT. These regions are most likely acti v ely transcribed and easily form ssDNA. Ther efor e, Dna2 may also be involved in other processes to pre v ent Spo11-independent RPA accumulation.

DA T A A V AILABILITY
Data used in the paper ar e pr esent in the paper and / or the Supplementary Materials. The RPA ChIP-seq data of WT, dna2-md and dna2-md spo11Y135F are deposited in the NCBI Gene Expression Omnibus (GEO) database under the accession number GSE217628. Data used for the analysis of hDNA tract lengths are obtained from NCBI under SRA accession number SRP075437 ( 16 ).