Antagonism between BRCA2 and FIGL1 regulates homologous recombination

Abstract Homologous recombination (HR) maintains genome stability by promoting accurate DNA repair. Two recombinases, RAD51 and DMC1, are central to HR repair and form dynamic nucleoprotein filaments in vivo under tight regulation. However, the interplay between positive and negative regulators to control the dynamic assembly/disassembly of RAD51/DMC1 filaments in multicellular eukaryotes remains poorly characterized. Here, we report an antagonism between BRCA2, a well-studied positive mediator of RAD51/DMC1, and FIDGETIN-LIKE-1 (FIGL1), which we previously proposed as a negative regulator of RAD51/DMC1. Through forward genetic screen, we identified a mutation in one of the two Arabidopsis BRCA2 paralogs that suppresses the meiotic phenotypes of figl1. Consistent with the antagonistic roles of BRCA2 and FIGL1, the figl1 mutation in the brca2 background restores RAD51/DMC1 focus formation and homologous chromosome interaction at meiosis, and RAD51 focus formation in somatic cells. This study shows that BRCA2 and FIGL1 have antagonistic effects on the dynamics of RAD51/DMC1-dependent DNA transactions to promote accurate HR repair.


INTRODUCTION
Homologous recombination (HR) is critical for preserving genome integrity because it facilitates error-free DNA repair (1). In somatic cells, a dysfunctional HR process can lead to greater genomic instability and has been linked to the development of various types of cancer and genetic diseases in humans (2). In meiotic cells, HR is essential for generating crossovers (COs)--the reciprocal exchange of genetic material among homologous chromosomes. COs among homologs are crucial to establish the physical link required for the accurate segregation of homologous chromosomes (3). Therefore, errors or a lack of meiotic HR can cause sterility or chromosomal mis-segregation leading to aneuploidy diseases such as Down's syndrome in humans (4). In addition, COs enhance genetic diversity in offspring by reshuffling the parental genomes in sexually reproducing organisms.
During meiosis, the HR repair process is initiated by the developmentally programmed formation of DNA doublestrand breaks (DSBs) (5). Nucleolytic processing at break sites generates 3 single-stranded DNA (ssDNA) overhangs. These ssDNA overhangs are used for homology searches, and subsequent strand invasion of the intact homologous template forms recombination intermediates wherein the invading strand serves as a primer for DNA synthesis (6). These intermediates can give rise to two types of repair products: CO with reciprocal exchanges among homologous chromosomes and non-crossovers (NCO) without reciprocal exchange resulting from synthesis-dependent strand annealing. In most eukaryotes, meiotic COs result from two pathways characterized as class I and class II (7). Class I COs depend on a group of proteins called ZMM (Zip1-4, Msh4-5, Mer3; first identified in budding yeast), which stabilizes recombination intermediates to promote class I CO formation (8). The class II CO pathway requires activity of structure-specific endonucleases including MUS81. In Arabidopsis, class I contributes to 85-90% of COs, whereas MUS81-dependent COs constitute a minor fraction (9,10).
Central to HR repair is the homologous recognition and DNA strand exchange of ssDNA overhangs catalyzed by two recombinases: (i) RAD51, which acts during mitosis and meiosis and (ii) DMC1, a meiosis-specific paralog present in most eukaryotes (11). The reason why many organisms have two recombinases and others only one remains unclear. Both recombinases can polymerize at the ssDNA overhangs to produce nucleoprotein filaments. RAD51-and DMC1-filaments forming nuclear foci localized on the chromosome are indicative of HR repair events in somatic and in meiotic cells (11,12). However, differences may lie in their relative contributions that appear to influence the fate of the repair event during meiosis. DMC1 is the main catalytic recombinase that promotes meiotic DSB repair on the homologous chromosome, likely to ensure CO formation among homologs. In contrast, RAD51 acts as an essential co-factor for DMC1-mediated repair (13,14). In somatic cells, RAD51, but not DMC1, is indispensable for HR repair. RAD51-ssDNA and DMC1-ssDNA filaments can be highly toxic intermediates if they remain unresolved in the cell. The DNA transactions mediated through these filaments appear dynamic and may be lost with defective filament assembly or impaired filament disassembly. To promote efficient HR repair and to avoid deadend nucleoprotein complexes, a dynamic balance between their loading and unloading at break sites is achieved with pathway reversibility, implemented via positive and negative regulators (1). However, the interplay between multiple positive and negative regulators to control the dynamic assembly/disassembly of RAD51/DMC1 filaments in multicellular eukaryotes remains poorly characterized.
BRCA2 (breast cancer susceptibility protein 2), a highly conserved protein in multicellular eukaryotes, is a positive mediator protein of RAD51 and DMC1 nucleation onto ss-DNA overhangs as shown in in vitro studies (15,16). BRCA2 interacts directly with RAD51 and DMC1 (16)(17)(18)(19). The current model supports the notion of BRCA2 being a cargo to recruit RAD51 and DMC1 molecules on ssDNA by virtue of its binding affinity for ssDNA and its capacity to displace the replication protein A (RPA2) coated on ssDNA overhangs. These biochemical studies are well supported, with in vivo roles of BRCA2 regulating RAD51-and DMC1mediated HR. BRCA2-deficient cells show hypersensitivity to DNA damaging agents and exhibit gross genomic instability (20)(21)(22). Further, the disruption of BRCA2 function leads to aberrant meiosis as well. In Arabidopsis, there are two BRCA2 homologs: BRCA2A and BRCA2B. The Arabidopsis brca2a brca2b double mutant is sterile with defective meiosis showing no RAD51 and DMC1 foci, no synapsis and unrepaired meiotic breaks resulting in chromosome fragmentation (21,23). Arabidopsis brca2a and brca2b single mutants are however fertile with normal meiosis, suggesting that the two paralogs have redundant roles during meiosis. However, brca2a, but not brca2b, is hypersensitive to DNA damage, revealing functional differences between BRCA2A and BRCA2B for somatic DNA repair (20). These BRCA2 deficiency phenotypes raise a question as to why RAD51 and DMC1 fail to form filaments in vivo, although they are capable of doing so in vitro in the absence of BRCA2 (24). It is therefore conceivable that active mechanisms prevent RAD51 and DMC1 from forming filaments in vivo.
We recently identified a novel FIGL1 complex comprising FIDGETIN-LIKE-1 (FIGL1) and FIDGETIN-LIKE-1 INTERACTING PROTEIN (FLIP) that likely acts as a negative regulator of RAD51 and DMC1 during meiosis. Mutations in FIGL1 and FLIP in Arabidopsis thaliana increase meiotic CO frequency and modulate the number and/or dynamics of RAD51/DMC1 foci (25,26). The widely conserved FIGL1 complex physically interacts with RAD51 and DMC1. FIGL1 is also required for efficient HR in human somatic cells, attributed to its interaction with RAD51 (27). Overall, the FIGL1 complex appears to be a conserved regulator of recombinase activity at the strand invasion step of HR, in somatic and in meiotic cells. FIGL1 belongs to the large family of AAA-ATPase proteins that are implicated in structural remodeling, unfolding and disassembly of proteins or oligomer complexes (28,29).
In this study, we report a genetic interaction between BRCA2 and the anti-crossover factor FIGL1 and show that their antagonistic functions regulate RAD51 and DMC1 focus formation in both mitotic and meiotic cells, thereby promoting HR repair. Our data reveal a new regulatory step controlling the dynamics of RAD51 and DMC1 focus formation with the antagonistic functions of BRCA2 being a positive accelerator and FIGL1 hindering the reaction as negative regulator.

Genetic material
The Arabidopsis plants were cultivated in a greenhouse as previously described (30). The T-DNA insertion and ethyl methane sulfonate (EMS) mutant lines used in this study were brca2a-1 (N427769), brca2b (N537617) (21), mus81 (N607515) (31), figl1, figl1 mus81 (25), zip4 (N568052) (32). The suppressor figl1mus81(s)12/brca2a-2 line was sequenced using IIlumina technology at the Genome Analysis Centre, Norwich, UK. Mutations were identified in the MutDetect pipeline (33). Genetic mapping of causal mutations was performed in F2 fertile plants generated after backcross with figl1 mus81. Using polymerase chain reaction, we followed the segregation of homozygous EMSgenerated mutations to identify a recessive mutation associated with partial restoration of fertility. The brca2a-2 causal mutation was a C-to-T substitution at the TAI10 chr4:5656 position in At4g00020. The primers used for genotyping are listed in the supplemental dataset file. Fertility of plants was examined by counting seeds per silique (fruit) on the scanned image of siliques fixed in 70% ethanol using Zeiss Zen software.
For RAD51 immunostaining, root tip nuclei were treated as previously described (38). Briefly, 5-day-old seedlings of each genotypes were incubated with or without 4 g/ml mitomycin C (MMC) for 2 h to induce DNA damage. After briefly rinsing, root tips were fixed for 45 min in 4% paraformaldehyde in PME [50 mM piperazine-N,N'-bis(2ethanesulphonic acid) (PIPES), pH 6.9; 5 mM MgSO4; 1 mM EGTA] followed by three washes in PME for 5 min. Then, root tips were digested for 30 min in 1% (w/v) cellulase, 0.5% (w/v) cytohelicase, 1% (w/v) pectolyase in PME and were washed three times for 5 min in PME. Digested cells were then squashed gently onto slides with a coverslip and immersed in liquid nitrogen. Slides were air-dried and stored at −80 • C. RAD51 immunostaining with rat anti-RAD51 (36) antibodies, along with DAPI on the root-tipsquashed cells, was performed according to Da Ines et al. (12). Images were obtained using the z-stack option with a Zeiss AxioObserver microscope and were analyzed using Zeiss Zen software. RAD51 foci were counted manually on maximum intensity projection images generated from the z-stack. Histograms and statistical analysis were performed using GraphPad Prism 7 software.

Root growth assays
To perform the root growth analysis, seeds from wild type, figl1, brca2a-2 brca2b and figl1 brca2a-2 brca2b lines were surface-sterilized and sown on solid medium containing 0.5 X Murashige and Skoog salts, 1% sucrose, 0.8% agar and 0, 0.5, 1.0, 2.0 or 4.0 g/ml mitomycin C (Duchefa). After stratification for 2 days at 4 • C, plants were grown in longday conditions by keeping plates in a vertical position for 14 days. Plates were scanned and the primary root length of seedlings was measured using Fiji with the smart root plugin. Mean values calculated from the root length value from 10 to 20 seedlings for each genotype were plotted to examine MMC sensitivity. Reduced mean root length compared to wild-type was considered as sensitivity to MMC. Statistical significance was computed using a one-way ANOVA and Dunnett's multiple-comparisons test.

A genetic screen identifies a mutation in Arabidopsis BRC2A that suppresses figl1 phenotypes
We identified FIGL1 as an anti-CO gene in a previous suppressor screen using Arabidopsis zmm mutants (25). Here, we set out to find antagonists of FIGL1 to better understand its function and the regulation of recombination. The Arabidopsis figl1 mutant is however fertile with no macroscopic phenotypes that could be easily screened to identify suppressors. Thus, we performed a suppressor screen using a figl1 mus81 double mutant that displayed reduced fertility with short fruit length and low seed set. The reduced fertility in figl1 mus81 was attributed to defective DSB repair due to persistent unrepaired breaks and chromosome fragmentation at metaphase I (25). Given that the loss of FIGL1 and MUS81 activity results in the accumulation of aberrant recombination intermediates, we sought to identify mutations in pro-DSB repair factors, which may antagonize FIGL1, MUS81 or both and could enhance seed set of figl1 mus81. Thus, we applied EMS mutagenesis on figl1 mus81 mutants, and screened the obtained population for enhanced fertil-ity (longer fruits) as a proxy for an attenuated chromosome fragmentation defect at meiosis.
We isolated a figl1mus81(S)12 suppressor showing higher seed set compared with figl1 mus81 ( Figure 1B). This seed set however was still low compared with the wild-type, suggesting only partial restoration of fertility in the suppressor line. We next examined if the partial restoration of fertility could be due to attenuation of chromosome fragmentation at metaphase I, which would suggest restored DSB repair efficiency. Based on DAPI (4 ,6-diamidino-2-phenylindole dihydrochloride) staining, we classified metaphases into three categories: (i) cells with five distinguishable DAPI bodies, suggesting nonfragmented chromosomes with almost wild-type-like DSB repair; (ii) fragmented metaphases with more than five DAPI bodies, considered as partially defective DSB repair; (iii) aberrant metaphases with entangled chromosome structures indicative of severe defective DSB repair. figl1mus81(S)12 exhibited attenuated defects compared with figl1 mus81 ( Figure 1C and D), suggesting partial restoration of DSB repair. Whole genome sequencing and genetic mapping of figl1mus81(S)12 identified the causal mutation at the splice donor site in exon 15 (chromosome 4: 5656, g to a) in the AT4G00020 gene encoding for BRCA2A (brca2a-2) ( Figure 1A). This mis-splicing results in the loss of 37 bp from exon 15 (Supplementary Figure S1A), leading to out-of-reading-frame translation after the E861 residue (BRCA2A-E861SS) that likely produces a BRCA2A protein truncated at the C-terminal end ( Supplementary Figure S1B). We tested another mutated allele of BRAC2A (T-DNA insertion; GABI 290C04/ brca2a-1) ( Figure 1A) that was also able to enhance fertility of figl1 mus81 ( Figure 1B). Further, figl1 mus81 brca2a-2/brca2a-1 exhibited a higher seed set rate per fruit compared with figl1 mus81 ( Figure  1B), demonstrating that brca2a-2 and brca2a-1 were allelic. Mutations in BRCA2A can thus restore fertility and meiotic DNA repair of figl1 mus81.
We then examined if mutations in BRCA2A can suppress figl1 phenotypes, also in the presence of MUS81 activity. One feature of the figl1 mutation is to restore bivalent/CO formation and fertility in zip4 mutants (25). We therefore predicted that, if BRCA2A antagonizes FIGL1 activity, the brca2a-2 mutation would specifically reduce bivalent/CO formation and fertility of figl1 zip4. Indeed, zip4 figl1 brca2a-2 showed a clear decrease in fertility compared with zip4 figl1 ( Figure 1E). Further, there was a significant decrease in bivalent formation in zip4 figl1 brca2a-2 compared with zip4 figl1 ( Figure 1F and G), showing that mutation in BRCA2A reduces bivalent formation in the zip4 figl1 background. The bivalent formation in zip4 was not different in zip4 brca2a-2, suggesting that brca2a-2 does not affect zip4 per se. Taken together; our data demonstrate that BRCA2A can antagonize FIGL1 activity.

Epistasis analysis of BRCA2A, BRCA2B and FIGL1
BRCA2B is the second BRCA2 homolog in Arabidopsis (Supplementary Figure S1C). A brca2b mutant was previously described as a likely null allele (21). The single brca2a-1 and brca2b mutants are fertile without meiotic defects (Supplementary Figure S1D (21). We then tested if BRCA2B is also able to antagonize FIGL1 functions as did BRCA2A. The fertility in zip4 figl1 brca2b was not reduced compared with zip4 figl1 (Figure 2A). Further, the number of bivalent formations at metaphase was not reduced in zip4 figl1 brca2b compared with zip4 figl1 (Figure 2B and C). zip4 brca2b plants did not show any reduction in fertility either, or in bivalent formation compared with zip4, suggesting no effect of brac2b mutation in zip4 (Figure 2A and  B). We concluded that mutations in BRCA2B do not suppress figl1 phenotypes, implying a difference in either nature or importance of the BRCA2A and BRCA2B functions at meiosis.
We then investigated the functional consequences of the loss of antagonism between functions of FIGL1 and the two Arabidopsis homologs BRCA2A and BRAC2B. We first compared the point mutation brca2a-2 with the T-DNA insertion allele brca2a-1 (21) for fertility and meiosis. The single brca2a-2 and brca2a-1 mutants showed normal fertility and no differences in bivalent formation at meiosis (Supplementary Figure S1D). In contrast to single mutants, brca2a-1 brca2b had almost no seeds and brca2a-2 brca2b produced only a few seeds ( Figure 2D). The slight difference in fertility between the two double mutants suggests that brca2a-1 is a null allele whereas brca2a-2 retains some residual activity and that the C-terminus (after residue E861) of BRCA2A, which contains an OB-fold domain (21,39) (Supplementary Figure S1B and C), is important but not absolutely essential for BRCA2 activity.
Next, we tested the ability of the figl1 mutation to restore fertility in brca2a brca2b double mutants. figl1 brca2a-2 brca2b mutants revealed a higher seed set compared with that of brca2a-2 brca2b. This confirms the antagonism between BRCA2A and FIGL1 functions. Intriguingly, FIGL1/ figl1 brca2a-2 brca2b also displayed a higher seed set compared with that of the brca2a-2 brca2b mutant (Figure 2D), suggesting that DNA repair efficiency is very sensitive to FIGL1/BRCA2 dosage. In contrast, introducing the figl1 mutation (either homozygous or heterozygous) in brca2a-1 brca2b did not improve its fertility, further suggesting the brca2a-1 is a stronger allele than brca2a-2, and that the figl1 mutation is not able to re-establish DNA repair in the complete absence of BRCA2A/B ( Figure 2D).
We then explored the causes behind the restoration of fertility in figl1 braca2a-2 braca2b. brca2a-2 brca2b showed a complete lack of bivalent formation and severe chromosome fragmentation at metaphase I, as in brca2a-1 brca2b (21). Further, analysis of metaphase chromosomes revealed fragmentation in figl1 brca2a-2 brca2b and FIGL1/ figl1 brca2a-2 brca2b, suggesting that the loss of FIGL1 does not restore complete DSB repair and bivalent formation in brca2a-2 brca2b mutants ( Figure 2E). However, the increased fertility of figl1 brca2a-2 brca2b compared with brca2a-2 brca2b suggests that the repair defect is reduced, but not enough to be detected by our analysis of chromosome shapes.
Mutation in FIGL1 restores RAD51/DMC1 focus formation and synapsis in the brca2a braca2b double mutant at meiosis RAD51 and DMC1 foci are localized on meiotic chromosomes as early recombination markers in the wild-type. Neither recombinase forms foci in brca2a-1 brca2b mutants, demonstrating the in vivo requirement for BRCA2 as a positive mediator (21). On the other hand, Arabidopsis figl1 mutants showed an accumulation and/or delayed kinetics of RAD51 and DMC1 foci, suggesting that FIGL1 is a negative regulator of the recombinases (26). We hypothesized that RAD51 and DMC1 repair foci can be restored in triple mutants upon loss of BRCA2 and FIGL1 owing to their antagonistic properties. We thus performed immunolocalization of RAD51 and DMC1 on double (brca2a-1 brca2b, brca2a-2 brca2b) and triple (figl1 brca2a-1 brca2b, figl1 brca2a-2 brca2b) mutants in combination with staining of chromosome axis proteins, either REC8 or ASY1, to mark meiocytes in prophase I. We did not find any abnormalities in the localization of REC8 and ASY1 (Figure 3 and Supplementary Figure S2), suggesting no implication of BRCA2 and FIGL1 in forming chromosome axes. Meiocytes at prophase I from both brca2a-1 brca2b and brca2a-2 brca2b double mutants lacked RAD51 and DMC1 focus formation compared with wild-type meiocytes, consistent with the previous results obtained in the brca2a-1 brca2b background. In contrast, figl1 brca2a-1 brca2b and figl1 brca2a-2 brca2b triple-mutant meiocytes showed restored RAD51 and DMC1 focus formation in comparison with double-mutant meiocytes (Figure 3 and Supplementary Figure S2), demonstrating antagonistic functions of BRCA2 and FIGL1 in the regulation of RAD51/DMC1 focus formation. This result also suggests that BRACA2 is not essential per se for DMC1/RAD51 focus formation.

FIGL1 antagonizes BRCA2 activity to regulate RAD51 focus formation upon DNA damage in somatic cells
We next examined functional consequences of the loss of antagonism between BRCA2 and FIGL1 in somatic cells upon induction of DNA damage. Here, we tested whether figl1, brca2a-2 brca2b and figl1 brca2a-2 brca2b mutants are sensitive to MMC treatment. We thus performed root growth assays on media containing different MMC concentrations for each genotype and measured root length after 14 days of growth ( Figure 4A). As shown in Figure 4B, we plotted mean values of root length (mm) ± SD to quantify the growth levels of each genotype. Wildtype plants showed no differences, with almost similar root growth on all tested MMC concentrations. The root growth of brca2a-2 brca2b compared with that of the wild-type at 2.0 and 4.0 g/ml MMC was strongly affected, demonstrating the hypersensitivity of brca2a-2 brca2b ( Figure 4A and B), as observed previously for brca2a-1 brca2b. Further, we noticed a significant reduction in root growth for figl1 (Sidak's multiple-comparison test, P < 0.000001) in comparison with the wild-type at concentrations of 2.0 and 4.0 g/ml MMC ( Figure 4A and B). This reduction in root growth suggests that FIGL1 is required for normal HR repair in mitotic cells upon DNA damage. Comparison of root growth between figl1 and brca2a-2 brca2b plants also revealed that figl1 mutants were less affected ( Figure 4A and B). At those MMC concentrations, figl1 brca2a-2 brca2b plants did not show root growth different from brca2a-2 brca2b, but both exhibited yet stronger decreases in root growth in comparison with figl1 plants (at 4 g/ml MMC, Sidak's multiple-comparison test, P < 0.00001) ( Figure 4A and B). Altogether, our results demonstrate that figl1 and brca2a-2 brca2b are hypersensitive to MMC treatment and that FIGL1 and BRCA2 play essential roles for repairing breaks arising from treatment of this DNA-crosslinking agent. However, loss of FIGL1 in brca2a-2 brca2b does not help rescue root growth upon DNA damage, which contrasts with the ability of figl1 to increase the fertility of the same genotype.
We then tested whether the MMC hypersensitivity of figl1 and brca2a-2 brca2b mutants is due to a defective HR process leading to the presence of unrepaired breaks, monitored as RAD51 foci in root cells upon treatment. Using anti-RAD51 antibodies, we immunolocalized RAD51 in wild-type, figl1, brca2a-2 brca2b and figl1 brca2a-2 brca2b root cells without or with MMC treatment (4 g/ml) for 2 h (Figure 4C and D; Supplementary Figure S3). To quantify RAD51 foci, we distributed root cells into four classes: (i) nuclei without any foci, (ii) nuclei with 1-10 foci, (iii) nuclei with 11-50 foci, (iv) nuclei with more than 50 foci. We recorded significantly lower numbers of total RAD51 positive nuclei in brca2a-2 brca2b compared with that of the wild-type with or without MMC treatment ( Figure 4C and D). This strong reduction in RAD51 positive nuclei in brca2a-2 brca2b confirms that BRCA2 activity is important for RAD51 focus formation in somatic cells. Interestingly, FIGL1-deficient root cells exhibited a large increase in RAD51 foci in the absence and presence of MMC (Figure 4C and D). This suggests that FIGL1 negatively regulates RAD51 focus formation in somatic cells. Importantly, we found that figl1 mutation restores formation of RAD51 foci in brca2a-2 brca2b nuclei with or without MMC treatment ( Figure 4C and D). This restoration demonstrates that the antagonism between FIGL1 and BRCA2 also regulates RAD51 focus formation in somatic cells. Overall, these data support the positive function of BRCA2 and the negative function of FIGL1 in the regulation of RAD51 in somatic cells and their interplay is required to facilitate HR repair.

DISCUSSION
Here, we identified a genetic interaction between Arabidopsis BRCA2 and FIGL1 that is essential for regulating HR repair in both mitotic and meiotic cells. We established that BRCA2 and FIGL1 antagonize each other for RAD51 focus formation in somatic cells, and for RAD51 and DMC1 focus formation and synapsis in meiotic cells.
We previously identified Arabidopsis FIGL1 as an anti-CO factor that limits meiotic crossovers. Our current hy-pothesis is that FIGL1 prevents formation of aberrant recombination intermediates in wild-type meiosis. We propose that FIGL1 negatively regulates the strand invasion step of HR through direct interaction with RAD51 and DMC1 (26). Absence of FIGL1 would thus result in exacerbated RAD51/DMC1 activity leading to the formation of aberrant recombination intermediates requiring the structure-specific endonuclease MUS81 to be repaired. We found that a mutation in BRCA2A attenuates meiotic DSB repair defects in figl1 mus81. We therefore suggest that the mutation of BRCA2A reduces BRCA2 dosage, presumably without affecting the levels of the second homolog BRCA2B. This reduction in BRCA2 dosage attenuates meiotic defects, suggesting that there may indeed be less aberrant recombination intermediates. BRCA2A and BRCA2B appear to have redundant activities as judged from normal fertility and meiosis in brca2a-2 and brca2b single mutants. However, only the mutation of BRCA2A, but not of BRCA2B, restored bivalent formation in the zip4 figl1 background. This observation suggests that BRCA2A functions are slightly different from those of BRCA2B, either in nature or dosage, for bivalent formation. Such differences in functionality between BRCA2A and BRCA2B has also been reported for somatic HR and may be due to differential expression levels (20). Importantly, we also noticed dosage effects for FIGL1 activity in FIGL1/figl1 brca2a-2 brca2b plants showing significantly higher numbers of seeds per fruit compared with brca2a-2 brca2b. Therefore, partial loss of FIGL1 activity in brca2a-2 brca2b may compensate for residual/reduced levels of BRCA2 to facilitate DSB repair. Altogether, dosage effects suggest that some critical threshold levels of both BRAC2 and FIGL1 are essential to promote accurate HR repair.
Our data indicate that antagonism between BRCA2 and FIGL1, operating in both mitotic and meiotic cells, regulates RAD51-and DMC1-dependent homology recognition and the strand invasion step during HR repair. Thus, how does antagonism between BRCA2 and FIGL1 modulate RAD51 and DMC1 focus formation? A key to maintaining genome stability via HR is to promote legitimate repair and, at the same time, restrict inappropriate repair. Activity of RAD51-and DMC1-filaments generates DNA transactions that lie at the intersection of multiple pathways to yield CO/NCO involving different molecular mechanisms within HR. These DNA transactions, at least during meiosis, appear to be highly dynamic and can generate a variety of recombination intermediates (40,41). We propose that the antagonistic regulation of RAD51-and DMC1-focus formation acts as a 'stop-and-go' mechanism, whereby FIGL1 applies the brakes by dismantling RAD51/DMC1 filaments and BRCA2 act as an accelerator that stimulates filament assembly. In this scenario, in the wild-type, BRCA2 promotes HR to push the reaction forward through nucleation and stabilization of filaments, and FIGL1 allows the pathway to reverse and restart, restricting HR that may originate from repair-and non-repairassociated DNA-bound forms of RAD51/DMC1. In turn, this mechanism may contribute to the dynamic nature of DNA transactions during HR.
The orthologs of both BRCA2 and FIGL1 are absent in yeast, suggesting that this antagonism is functionally mediated through other proteins. Our data are consistent with a proposed model in budding yeast wherein antagonism between the stabilizing function of RAD51 paralogs Rad55-Rad57 and the destabilizing function of Srs2 antirecombinase modulates RAD51-filament/focus formation in somatic HR repair (42). In analogy to this model in Arabidopsis, BRCA2 provides the stabilizing functions and FIGL1 acts as an anti-recombinase to destabilize the filaments. SRS2 also exists in Arabidopsis (43) and may also participate in this equilibrium. This model predicts that deletion of stabilizing or destabilizing activity shifts the balance in a backward or forward direction, respectively. Likewise, we observed that the equilibrium of RAD51/DMC1 foci was shifted toward low numbers of foci in brca2a-2 brca2b, whereas figl1 mutants showed an increase in RAD51/DMC1 focus formation. Importantly, we also observed the restoration of RAD51/DMC1 focus formation in figl1 brca2a-2 brca2b. Based on these results, we propose that BRCA2 protects filaments from the likely antirecombinase activity of FIGL1, in addition to its other known roles including in vivo nucleation of RAD51/DMC1 filaments. Whether BRCA2 does so by directly interacting with FIGL1, as does the Rad55-Rad57 complex to counteract Srs2 activity, remains to be determined. Alternatively, BRCA2 and FIGL1 act competitively to reach a dynamic equilibrium between assembly and disassembly of filaments. Further, the restoration of RAD51/DMC1 foci in figl1 brca2a-2 brca2b is insufficient to restore either wild-type level of fertility or normal root growth upon DNA damage induced by MMC treatment. This insufficiency suggests that the execution of complete HR repair requires the presence of both BRCA2 and FIGL1 activity. A number of observations are compatible with this interpretation. BRCA2 deficiency is hallmarked by gross genome instability in both somatic and meiotic cells (21,23,44). In somatic cells, BRCA2 prevents gross genomic instability by counteracting non-homologous end-joining pathways (45), which does not seem to be the case during meiosis (46). It is tempting to propose that FIGL1 inhibits HR repair in the absence of BRCA2 by dismantling RAD51/DMC1 filaments, which may result in unrepaired breaks or break repair via other pathways. Loss of FIGL1 activity can also induce genome instability. Although not observed in Arabidopsis, rice figl1 mutants display severe meiotic chromosome fragmentation showing unrepaired breaks (47).
In conclusion, we revealed an interplay between BRCA2 and FIGL1 that appears to ensure accurate HR and inhibit illegitimate HR when required. This interplay is important for completing HR repair under different cellular contexts and may be of particular interest to advance our understanding on genomic instability mechanisms associated with human pathologies such as cancer.

DATA AVAILABILITY
Supplemental Dataset: A spreadsheet file with nine sheets named according to the corresponding figures contains the raw data along with the analysis and the calculation of statistical significance tests used. In an additional sheet, the list of oligonucleotides used in this study is given.