OsRAD51 Plays a Vital Role in Promoting Homologous Recombination in Rice Meiosis

Meiotic recombination plays a pivotal role in achieving accurate chromosomal segregation and increasing genetic diversity. In the homologous recombination pathway, the detailed mechanisms of how OsRAD51 and OsDMC1 work in rice meiosis remain to be explored. Here, we obtained different types of mutants for Osrad51a1, Osrad51a2, Osdmc1a, and Osdmc1b through CRISPR/Cas9. Both Osrad51a1 and Osrad51a2 exhibited normal vegetative growth and fertility. Osrad51 (Osrad51a1 Osrad51a2) mutant plants show normal vegetative growth but exhibit complete sterility, indicating that OsRAD51A1 and OsRAD51A2 are functionally redundant in rice fertility. In contrast to the wild type, Osrad51 chromosomes are not paired perfectly at pachytene and synaptonemal complex (SC) formation is deficient. Moreover, univalents and multivalent associations were observed at metaphase I, chromosome fragments presented at anaphase I, and crossover formation is basically suppressed in Osrad51 pollen mother cells (PMCs). OsRAD51 foci emerge at leptotene and disappear from late pachytene and chromosome localization of OsRAD51 depends on the formation of double-strand breaks (DSBs). Most OsRAD51 foci can co-localize with OsDMC1 signals. OsRAD51 is essential for the loading of OsDMC1 onto chromosomes, and vice versa. In addition, both OsRAD51 and OsDMC1 can interact with OsFIGL1 and OsBRCA2, two important components in rice meiosis. Moreover, the Osrad51 Osdmc1 (Osrad51a1 Osrad51a2 Osdmc1a Osdmc1b) quadruple mutant PMCs exhibited similar defective phenotypes as Osrad51 in homologous pairing, synapsis, and DSB repair. Taken together, our results suggest that the recombinases DMC1 and RAD51 may functionally depend on each other and play important roles in meiotic recombination during meiosis in rice.


Introduction
Meiosis is a specialized eukaryotic cell division process in which two successive rounds of nuclear segregation and one round of DNA replication lead to the formation of haploid gametes. After fertilization, the fusion of gametes restores the diploid chromosome number of the species. This process guarantees that organisms that undergo sexual replication maintain the same number of chromosomes. Moreover, the reshuffle of the genetic information in homologous chromosomes occurred during meiosis, which creates genetic diversity among offspring and is a primary source of genetic variation in eukaryotes [1,2].
Meiotic homologous recombination (HR) is an important biological event that occurs during meiosis. HR is initiated by the formation of double-strand breaks (DSBs), which are catalyzed by the action of conserved DNA transesterase SPO11 proteins [3,4]. In Saccharomyces cerevisiae, SPO11 proteins are covalently bonded to the 5 end of a DSB and come off

Generation of Osrad51 and Osrad51 Osdmc1 Mutants by CRISPR/Cas9 Genome Editing
The Arabidopsis RAD51 amino acid sequence was used as a template in a BLAST search against the rice proteome [26]. We found two proteins that share a high level of similarity to AtRAD51 according to BLAST results. These two proteins, which are encoded by LOC_Os11g40150 and LOC_Os12g31370, were named OsRAD51A1 and OsRAD51A2, respectively. OsRAD51A1 is predicted to consist of 339 amino acid residues. OsRAD51A2 has 341 amino acid residues, and the two putative proteins share 95% identity in the amino acid sequence. Both OsRAD51A1 and OsRAD51A2 contain an N-terminal helixhairpin-helix DNA-binding motif class 1 domain (HhH1) and an ATPases associated with a variety of cellular activities domain (AAA) [27] (Supplementary Figure S1A). These two proteins were collectively named as OsRAD51 in this study. Multiple sequence alignment of OsRAD51 with orthologous proteins showed that it is highly evolutionarily conserved in eukaryotes (Supplementary Figure S1B). qRT-PCR assays revealed that OsRAD51 expressed the highest level in young panicles of wild-type rice, but is significantly low in leaves and roots (Supplementary Figure S1C).
Rice genomes contain two copies of DMC1 (OsDMC1A and OsDMC1B) [28]. To study the relationships between both two recombinases in-depth, the quadruple mutant is needed. It would be extremely difficult to obtain the quadruple mutant using routine mutagenesis and crossbreeding. Therefore, we constructed a CRISPR/Cas9 vector to edit these four genes ( Figure 1A). Through Agrobacterium-mediated transformation of the rice variety Yandao 8, seedlings of the first (T0) generation with specific mutations in OsRAD51A1, OsRAD51A2, OsDMC1A, and OsDMC1B were obtained. In the T1 generation, we obtained two Osrad51a1 Osrad51a2 alleles and individually designated them Osrad51-1 and Osrad51-2 (the Osrad51 mutants mentioned in this article are all Osrad51-1). One Osdmc1a Osdmc1b plant was designated Osdmc1, and one quadruple mutant was designated Osrad51 Osdmc1. We also isolated plants with single mutations in Osrad51a1, Osrad51a2, Osdmc1a, and Osdmc1b. To determine the genotype and heritability of every line, we sequenced the target genes to ascertain the genotypes in the subsequent generations. The results revealed that in the Osrad51-1 mutant there is a single 'A' insertion in the 8th exon of OsRAD51A1 and OsRAD51A2 ( Figure 1B OsRAD51A2 has 341 amino acid residues, and the two putative proteins share 95% identity in the amino acid sequence. Both OsRAD51A1 and OsRAD51A2 contain an N-terminal helix-hairpin-helix DNA-binding motif class 1 domain (HhH1) and an ATPases associated with a variety of cellular activities domain (AAA) [27] (Supplementary Figure S1A). These two proteins were collectively named as OsRAD51 in this study. Multiple sequence alignment of OsRAD51 with orthologous proteins showed that it is highly evolutionarily conserved in eukaryotes (Supplementary Figure S1B). qRT-PCR assays revealed that OsRAD51 expressed the highest level in young panicles of wild-type rice, but is significantly low in leaves and roots (Supplementary Figure S1C). Rice genomes contain two copies of DMC1 (OsDMC1A and OsDMC1B) [28]. To study the relationships between both two recombinases in-depth, the quadruple mutant is needed. It would be extremely difficult to obtain the quadruple mutant using routine mutagenesis and crossbreeding. Therefore, we constructed a CRISPR/Cas9 vector to edit these four genes ( Figure 1A). Through Agrobacterium-mediated transformation of the rice variety Yandao 8, seedlings of the first (T0) generation with specific mutations in OsRAD51A1, OsRAD51A2, OsDMC1A, and OsDMC1B were obtained. In the T1 generation, we obtained two Osrad51a1 Osrad51a2 alleles and individually designated them Osrad51-1 and Osrad51-2 (the Osrad51 mutants mentioned in this article are all Osrad51-1). One Osdmc1a Osdmc1b plant was designated Osdmc1, and one quadruple mutant was designated Osrad51 Osdmc1. We also isolated plants with single mutations in Osrad51a1, Osrad51a2, Osdmc1a, and Osdmc1b. To determine the genotype and heritability of every line, we sequenced the target genes to ascertain the genotypes in the subsequent generations. The results revealed that in the Osrad51-1 mutant there is a single 'A' insertion in the 8th exon of OsRAD51A1 and OsRAD51A2 ( Figure 1B

The Meiotic Process Is Disturbed in Osrad51
Both Osrad51a1 and Osrad51a2 exhibited normal vegetative growth and fertility (Supplementary Figure S4A-C). However, the Osrad51 plants were completely sterile, although they exhibited normal vegetative growth and development (Supplementary Figure S4D). Pollen grains from Osrad51a1 and Osrad51a2 were round (Supplementary Figure S4G,H), however, pollen grains from Osrad51 plants were empty and shrunken (Supplementary Figure S4I). Thus, OsRAD51A1 and OsRAD51A2 are functionally redundant in the rice reproductive process. In addition, the Osdmc1 mutant and Osrad51 Osdmc1 mutant showed normal vegetative growth but were sterile (Supplementary Figure S4E,F,K,L).
To explore the causes of the sterility in the Osrad51 mutant, we analyzed the meiotic chromosome behaviors of pollen mother cells (PMCs) from wild-type and Osrad51 mutant plants. In wild-type PMCs, meiotic chromosomes began to pair at zygotene (Figure 2A). Perfectly synapsed chromosomes were presented as thick threads at pachytene. At metaphase I, the further condensed bivalents aligned on the equatorial plate. Homologous chromosomes separated at anaphase I. In the second meiotic division, the equational segregation of sister chromatids occurred, and tetrads were formed ( Figure 2A).
In Osrad51 PMCs, a partially asynaptic phenotype was observed at the zygotenepachytene transition. Differences between Osrad51 and the wild type were evident at metaphase I. Contrary to the 12 well-defined bivalents observed in wild-type PMCs, univalents and multivalent associations were displayed in Osrad51 PMCs (n = 35 for PMCs). At anaphase I, chromosome bridges and chromosome fragments were observed as a result of the presence of univalents and multivalents at metaphase I (n = 40 for PMCs), and abnormal tetrads with scattered DNA eventually formed in the Osrad51 mutant (n = 30 for PMCs) (Figure 2A). Further cytological observations revealed that the meiotic defects in Osrad51-2 were similar to those observed in Osrad51-1 (Supplementary Figure S3B). These results suggest that the sterility of Osrad51 plants is caused by meiotic defects.

The Osrad51 Mutant Shows Defects in Homologous Pairing and Synapsis
During early prophase I, telomeres attach to the inner nuclear envelope and form a cluster. Telomere bouquet clustering is essential for homologous pairing [29,30]. To detect bouquet formation during meiosis, we adopted fluorescence in situ hybridization (FISH) analysis of wild-type and Osrad51 PMCs using the pAtT4 probe, which recognizes telomere-specific sequences [31]. In wild-type PMCs, telomeres were clustered within a confined region at early zygotene. There were no differences between Osrad51 PMCs and the wild-type (n = 30 for PMCs) ( Figure 2B). These results suggest that bouquet formation is independent of OsRAD51.
To examine chromosome pairing behavior, FISH was performed on Osrad51 PMCs employing 5S rDNA as a specific probe. 5S rDNA is a tandemly repeated sequence located on the short arm of chromosome 11 close to the centromere in rice. During early pachytene, two side-by-side 5S rDNA signals were observed in the wild-type (n = 25 for PMCs), indicating that homologous pairing was completed between homologous chromosomes. In the Osrad51 mutant, only 31.4% of the detected meiocytes showed paired signals for the 5S probe, and 68.6% of PMCs displayed unpaired signals (n = 70 for PMCs) ( Figure 2C). These results suggest that mutation of OsRAD51 leads to deficient homologous chromosomal pairing during rice meiosis.
To explore the causes of the sterility in the Osrad51 mutant, we analyzed the meiotic chromosome behaviors of pollen mother cells (PMCs) from wild-type and Osrad51 mutant plants. In wild-type PMCs, meiotic chromosomes began to pair at zygotene (Figure 2A). Perfectly synapsed chromosomes were presented as thick threads at pachytene. At metaphase I, the further condensed bivalents aligned on the equatorial plate. Homologous chromosomes separated at anaphase I. In the second meiotic division, the equational segregation of sister chromatids occurred, and tetrads were formed ( Figure 2A). We also detected the assembly status of the synaptonemal complex (SC) in Osrad51 PMCs via an immunolocalization assay employing antibodies against PAIR3, PAIR2, and ZEP1 protein. PAIR3, an axial protein, is essential for homologous pairing [32]. PAIR2 is associated with axial elements proteins, and is released quickly from the chromosomal axes when SC formation is completed. ZEP1 can be used to monitor synapsis in rice [33]. In wild-type PMCs, ZEP1 signals were fully elongated along the completely paired homologous chromosomes at pachytene ( Figure 3A). PAIR2 staining presented punctate signals at leptotene. The PAIR2 signals decreased gradually from synaptic chromosomes during zygotene and diminished from chromosomes after synapsis was completed at pachytene [34]. In Osrad51 PMCs, we detected that the PAIR2 signals vanished from the regions where ZEP1 was loaded (n = 30 for PMCs) ( Figure 3C, indicated by arrows). In order to detect the location pattern of PAIR2 on Osrad51 chromosomes in more detail, we used an immunolocalization assay combined with a super-resolution structured illumination microscope (SIM). The results showed that the PAIR2 signal disappeared from the synapsed regions (n = 35 for PMCs) (Supplementary Figure S5). We then analyzed 63 Osrad51 pachytene PMCs to measure the level of the SCs (the ratio of the length of ZEP1 to the length of PAIR3) [35,36]. The results showed that 77.8% of the PMCs presented punctuate or short liner ZEP1 signals. Of these, the SC level was <20%. In addition, 12.7% of the PMCs displayed slightly longer ZEP1 stretches, and the SC level was 20-50%. Some long ZEP1 tracks were present along the chromosomes in 9.5% of the PMCs, and the SC level was >50% ( Figure 3B). Taken all together, these results suggest that the function of OsRAD51 is pivotal for the full-length synapsis. to the length of PAIR3) [35,36]. The results showed that 77.8% of the PMCs presented punctuate or short liner ZEP1 signals. Of these, the SC level was <20%. In addition, 12.7% of the PMCs displayed slightly longer ZEP1 stretches, and the SC level was 20-50%. Some long ZEP1 tracks were present along the chromosomes in 9.5% of the PMCs, and the SC level was >50% ( Figure 3B). Taken all together, these results suggest that the function of OsRAD51 is pivotal for the full-length synapsis.

Crossover Formation Is Largely Inhibite in the Osrad51 Mutant
One of the main cytological defects observed in the Osrad51 mutant is the decreased frequency of bivalents ( Figure 2A). We inspected the generation of bivalents in Osrad51. Unlike the 12 bivalents in the wild type, the Osrad51 PMCs produced one to two bivalents at metaphase I. The mean number of bivalents was decreased in Osrad51 (1.9 ± 0.2, n = 35

Crossover Formation Is Largely Inhibite in the Osrad51 Mutant
One of the main cytological defects observed in the Osrad51 mutant is the decreased frequency of bivalents ( Figure 2A). We inspected the generation of bivalents in Osrad51.
Unlike the 12 bivalents in the wild type, the Osrad51 PMCs produced one to two bivalents at metaphase I. The mean number of bivalents was decreased in Osrad51 (1.9 ± 0.2, n = 35 for PMCs). Crossovers are essential for stable bivalents formation, so the results suggest that crossover formation is probably suppressed in the Osrad51 mutant.
We performed immunolocalization assays in Osrad51 employing antibodies against OsMER3, OsZIP4, and OsHEI10, three proteins that all participate in crossover formation [37][38][39]. OsREC8 was used to indicate meiotic chromosomes, which is a conserved component of the cohesion complex [40]. The number of OsMER3 foci in the wild type (269.4 ± 9.1; n = 17 for PMCs) was comparable to that of the Osrad51 mutant (267.9 ± 8.3, n = 17 for PMCs) at zygotene. The number of OsZIP4 foci in the Osrad51 mutant (297.6 ± 5.1, n = 18 for PMCs) was also almost equal to that in the wild type (296.9 ± 3.4, n = 18 for PMCs) ( Figure 4A,B). A previous study showed that HEI10 can indicate the Class I COs [39]. In the wild type, bright HEI10 foci were detected (25.4 ± 0.47, n = 25 for PMCs) at late pachytene, and the foci could also be detected at diplotene ( Figure 4B,C). Compared to wild-type, the emergence of the bright dot-like HEI10 foci was delayed in the Osrad51 mutant. At late pachytene, most OsHEI10 signals appeared as sparser foci, and this may be the result of delayed SC assembly. At diplotene, the prominent HEI10 foci were significantly decreased in Osrad51 (12.8 ± 0.43; n = 25 for PMCs) ( Figure 4B,C). Taken together, these results suggest that OsRAD51 can affect the crossover formation. for PMCs). Crossovers are essential for stable bivalents formation, so the results suggest that crossover formation is probably suppressed in the Osrad51 mutant. We performed immunolocalization assays in Osrad51 employing antibodies against OsMER3, OsZIP4, and OsHEI10, three proteins that all participate in crossover formation [37][38][39]. OsREC8 was used to indicate meiotic chromosomes, which is a conserved component of the cohesion complex [40]. The number of OsMER3 foci in the wild type (269.4 ± 9.1; n = 17 for PMCs) was comparable to that of the Osrad51 mutant (267.9 ± 8.3, n = 17 for PMCs) at zygotene. The number of OsZIP4 foci in the Osrad51 mutant (297.6 ± 5.1, n = 18 for PMCs) was also almost equal to that in the wild type (296.9 ± 3.4, n = 18 for PMCs) ( Figure 4A,B). A previous study showed that HEI10 can indicate the Class I COs [39]. In the wild type, bright HEI10 foci were detected (25.4 ± 0.47, n = 25 for PMCs) at late pachytene, and the foci could also be detected at diplotene ( Figure 4B,C). Compared to wildtype, the emergence of the bright dot-like HEI10 foci was delayed in the Osrad51 mutant. At late pachytene, most OsHEI10 signals appeared as sparser foci, and this may be the result of delayed SC assembly. At diplotene, the prominent HEI10 foci were significantly decreased in Osrad51 (12.8 ± 0.43; n = 25 for PMCs) ( Figure 4B,C). Taken together, these results suggest that OsRAD51 can affect the crossover formation. Quantification of OsMER3, OsZIP4, and HEI10 foci per meiocyte in wild type and Osrad51. Horizontal lines in the panels indicate the mean values. Statistical significance was calculated using the two-tailed Student's t-test. (C). Immunolocalization of HEI10 (green) at late pachytene and diplotene in wild type and Osrad51. Bars, 5 μm.

OsRAD51 Is Dynamically Localized on Chromosomes during Early Prophase I
To investigate the specificity of the OsRAD51 antibody, we conducted an immunolocalization assay in wild-type PMCs and Osrad51 mutant PMCs. No OsRAD51 signals were detected in Osrad51 mutant PMCs but were detected in the Osrad51a1 and Osrad51a2 Horizontal lines in the panels indicate the mean values. Statistical significance was calculated using the two-tailed Student's t-test. (C). Immunolocalization of HEI10 (green) at late pachytene and diplotene in wild type and Osrad51. Bars, 5 µm.

OsRAD51 Is Dynamically Localized on Chromosomes during Early Prophase I
To investigate the specificity of the OsRAD51 antibody, we conducted an immunolocalization assay in wild-type PMCs and Osrad51 mutant PMCs. No OsRAD51 signals were detected in Osrad51 mutant PMCs but were detected in the Osrad51a1 and Osrad51a2 mutants (Supplementary Figure S6A). In addition, results of the western-blot assay reveal that the OsRAD51 antibody could clearly recognize OsRAD51 in total extracts from the young panicles of the wild type, Osrad51a1, and Osrad51a2, but no signal was detected in the protein sample of Osrad51 (Supplementary Figure S6B). These results suggest that the antibody directed against RAD51 can recognize OsRAD51A1 and OsRAD51A2 simultaneously. To inspect the spatial and temporal localization of OsRAD51 during rice meiosis, we performed a dual immunolocalization assay in wild-type PMCs using specific antibodies directed individually against OsREC8 and OsRAD51. In the wild-type PMCs, OsRAD51 signals firstly emerged as punctate foci on chromosomes at leptotene, gradually appearing as high-intensity dot-like signals on zygotene chromosomes. During pachytene, the signals gradually became weak ( Figure 5A).  Figure S6A). In addition, results of the western-blot assay reveal that the OsRAD51 antibody could clearly recognize OsRAD51 in total extracts from the young panicles of the wild type, Osrad51a1, and Osrad51a2, but no signal was detected in the protein sample of Osrad51 (Supplementary Figure S6B). These results suggest that the antibody directed against RAD51 can recognize OsRAD51A1 and OsRAD51A2 simultaneously. To inspect the spatial and temporal localization of OsRAD51 during rice meiosis, we performed a dual immunolocalization assay in wild-type PMCs using specific antibodies directed individually against OsREC8 and OsRAD51. In the wild-type PMCs, OsRAD51 signals firstly emerged as punctate foci on chromosomes at leptotene, gradually appearing as high-intensity dot-like signals on zygotene chromosomes. During pachytene, the signals gradually became weak ( Figure 5A).  Meiotic recombination is induced by the programmed formation of DSBs that is catalyzed by Spo11, a functional homolog of topoisomerase-like protein. To determine whether the loading of OsRAD51 depends on the formation of DSBs, we carried out a dual immunolocalization assay in spo11-1 PMCs using antibodies against OsRAD51 and OsREC8. There were no obvious OsRAD51 signals in Osspo11-1 PMCs, which suggests that the loading of OsRAD51 depends on meiotic DSB formation (Supplementary Figure S7A). γH2AX is an isoform of histone H2A.γH2AX is a reliable cytological marker for detecting DSB in most eukaryotes [41]. Therefore, we performed dual immunolocalization assays in wild-type and Osrad51 PMCs employing polyclonal antibodies specifically against γH2AX and OsREC8 individually. In wild-type PMCs, γH2AX signals emerged as dots and patchy areas at zygotene. In the Osrad51 mutant, the pattern of γH2AX signals was comparable to that of the wild type (Supplementary Figure S7B), suggesting that DSB formation is not affected in Osrad51.

OsRAD51 and OsDMC1 Are Co-Localized and Functionally Interdependent
To further investigate the relationship between OsRAD51 and OsDMC1, we conducted a dual immunolocalization assay in wild-type PMCs using specific antibodies against OsRAD51 and OsDMC1 simultaneously. The results of super-resolution SIM on wild-type PMCs revealed that the OsRAD51 signals co-localized with OsDMC1 as bright yellow foci ( Figure 5B). Furthermore, quantitative co-localization analysis revealed that the co-located signals accounted for 84% of the OsDMC1 foci and 86% of OSRAD51 foci (n = 39 for PMCs) ( Figure 5B,D). These results show that OsRAD51 had considerable colocalization with OsDMC1. Immunostaining for OsRAD51 and OsDMC1 was also performed in the Osdmc1 and Osrad51 mutants, respectively. There were no obvious OsRAD51 signals on Osdmc1 meiotic chromosomes, and the OsDMC1 signal could not be detected in the Osrad51 mutant (n = 30 for PMCs) ( Figure 5C). These results indicate that the loading of OsRAD51 and OsDMC1 may be interdependent.
To determine whether OsRAD51 interacts with OsDMC1, we carried out yeast twohybrid (Y2H) and BiFC assays. The results revealed that OsRAD51 cannot interact with OsDMC1 ( Figure 6A,C). In order to verify this result, we implemented CoIP assays. Total proteins were extracted from rice panicles in the meiosis stage. The immunoprecipitates of anti-OsRAD51 were detected by anti-OsDMC1 and the immunoprecipitates of anti-OsDMC1 were detected by anti-OsRAD51. These results showed that OsRAD51 does not interact with OsDMC1 directly ( Figure 6B). We also found that OsRAD51A1 can interact with OsRAD51A2, and OsDMC1A can interact with OsDMC1B, but OsRAD51A1, OsRAD51A2, OsDMC1A, and OsDMC1B cannot interact with themselves ( Figure 6A,C). In addition, we found that both of these recombinases can interact with OsFIGL1 and OsBRCA2 ( Figure 6D). BRCA2 was previously reported to be a positive accelerator in controlling the dynamics of RAD51 and DMC1 focus formation, whereas FIGL1 is a negative regulator in Arabidopsis [42]. We further confirmed these interactions using BiFC assays ( Figure 6E). These results imply that BRCA2 and FIGL1 may mediate the functions of these two recombinases in rice meiotic homologous recombination.

Chromosomal Behavior in the Osrad51 Osdmc1 Mutant Is Similar to That in the Osrad51 Mutant
Meiosis is disrupted in both the Osrad51 and Osdmc1 mutants, but the former displays univalents and multivalent chromosome associations, whereas the latter shows ~24 univalents at metaphase I ( Figure 7A), and this result is consistent with previous articles [28]. To further investigate in depth the relationship between the two recombinases, we carried out a genetic analysis between them. The meiotic chromosomal phenotype of the Osrad51 Osdmc1 mutant is shown in Figure 7A. In Osrad51 Osdmc1 PMCs, univalents and multivalents were observed at metaphase I (n = 20 for PMCs). Chromosome bridges and chromosome fragmentation were detected at anaphase I (n = 20 for PMCs). At the end of meiosis, abnormal tetrads with scattered DNA were observed. Thus, chromosomal behavior in the Osrad51 Osdmc1 mutant is similar to that of Osrad51.

Chromosomal Behavior in the Osrad51 Osdmc1 Mutant Is Similar to That in the Osrad51 Mutant
Meiosis is disrupted in both the Osrad51 and Osdmc1 mutants, but the former displays univalents and multivalent chromosome associations, whereas the latter shows~24 univalents at metaphase I ( Figure 7A), and this result is consistent with previous articles [28]. To further investigate in depth the relationship between the two recombinases, we carried out a genetic analysis between them. The meiotic chromosomal phenotype of the Osrad51 Osdmc1 mutant is shown in Figure 7A. In Osrad51 Osdmc1 PMCs, univalents and multivalents were observed at metaphase I (n = 20 for PMCs). Chromosome bridges and chromosome fragmentation were detected at anaphase I (n = 20 for PMCs). At the end of meiosis, abnormal tetrads with scattered DNA were observed. Thus, chromosomal behavior in the Osrad51 Osdmc1 mutant is similar to that of Osrad51. In order to examine homologous pairing in the Osrad51 Osdmc1. FISH assays were performed in Osrad51 Osdmc1 PMCs using 5S rDNA as a specific probe. During early pachytene, 7.4% of the examined PMCs showed paired or partially paired 5S rDNA signals, and 92.6% of PMCs displayed unpaired signals (n = 54 for PMCs) ( Figure 7B). We further investigated SC assembly in Osrad51 Osdmc1 PMCs via immunolocalization assays employing antibodies specifically against ZEP1. ZEP1 signals emerged as only a few short lines on chromosomes at pachytene ( Figure 7C). A previous study reported that although homolog pairing occurs in the Osdmc1a Osdmc1b double mutant, synapsis is deficient [28]. In addition, we also found that the location of ZEP1 appears to be abnormal in the Osdmc1 mutant (n = 25 for PMCs) (Supplementary Figure S8). These results indicate that although OsRAD51 and OsDMC1 are essential for SC assembly, the individual functions of OsRAD51 and OsDMC1 are not identical in homologous recombination.
In addition to RAD51 and DMC1, there are five RAD51 paralogs, including RAD51B, RAD51C, RAD51D, XRCC2, and XRCC3 in Arabidopsis and rice. In Arabidopsis, mutations in AtRAD51, AtRAD51C, or AtXRCC3 cause the presence of entangled chromosomes [18,[46][47][48], while AtRAD51B, AtRAD51D, and AtXRCC2 seem to be unessential for meiotic DSB repair, because chromosomal behavior is normal in the triple mutant, and the In order to examine homologous pairing in the Osrad51 Osdmc1. FISH assays were performed in Osrad51 Osdmc1 PMCs using 5S rDNA as a specific probe. During early pachytene, 7.4% of the examined PMCs showed paired or partially paired 5S rDNA signals, and 92.6% of PMCs displayed unpaired signals (n = 54 for PMCs) ( Figure 7B). We further investigated SC assembly in Osrad51 Osdmc1 PMCs via immunolocalization assays employing antibodies specifically against ZEP1. ZEP1 signals emerged as only a few short lines on chromosomes at pachytene ( Figure 7C). A previous study reported that although homolog pairing occurs in the Osdmc1a Osdmc1b double mutant, synapsis is deficient [28]. In addition, we also found that the location of ZEP1 appears to be abnormal in the Osdmc1 mutant (n = 25 for PMCs) (Supplementary Figure S8). These results indicate that although OsRAD51 and OsDMC1 are essential for SC assembly, the individual functions of OsRAD51 and OsDMC1 are not identical in homologous recombination.
In addition to RAD51 and DMC1, there are five RAD51 paralogs, including RAD51B, RAD51C, RAD51D, XRCC2, and XRCC3 in Arabidopsis and rice. In Arabidopsis, mutations in AtRAD51, AtRAD51C, or AtXRCC3 cause the presence of entangled chromosomes [18,[46][47][48], while AtRAD51B, AtRAD51D, and AtXRCC2 seem to be unessential for meiotic DSB repair, because chromosomal behavior is normal in the triple mutant, and the triple mutant is fertility [49,50]. These proteins have partially functional redundancy in mitotic DNA repair [50]. Interestingly, in rice, the Osrad51d mutant displays abnormal homologous chromosome pairings and fragmented chromosomes [51,52], which indicates that the functions of RAD51 paralogs in meiotic HR have differentiated during evolution between rice and Arabidopsis.

The Role of RAD51 in Meiosis
In yeast, the rad51 mutant is hypersensitive to ionizing radiation and DNA-damaging chemicals [14]. Deletion of RAD51 in the nematode (Caenorhabditis elegans) leads to failure of chiasma formation and 12 properly condensed univalents at diakinesis, but pairing and synapsis are normal [53]. In the Drosophila melanogaster rad51 mutant, synapse formation was found to be appropriately initiated, but the resolution of synapsis was delayed [54]. Mutation of RAD51 in Tetrahymena thermophila causes an arrest early in meiosis I [55]. However, in mice and chicken, the rad51 null mutation is lethal [15,56]. Furthermore, the number of RAD51 foci is decreased in many maize mutants that are defective in pairing [23]. In Arabidopsis, RAD51 is not essential for vegetative and floral development, but the Atrad51 mutant displays severe meiotic defects, including partial homologous pairing and synapse, chromosome fragmentation, and multivalent chromosome associations [18,20]. Our data show that RAD51 is essential for normal meiosis in rice, but is not required for vegetative development. In contrast to the wild type, Osrad51 chromosomes are not paired perfectly at zygotene, SC formation is deficient. Moreover, univalents and multivalent associations are present at metaphase I, and chromosome fragments and unbalanced chromosome segregation are detected at anaphase I. Chromosome fragmentation may be the reason behind the abnormal segregation of chromosomes. The problems in homologous pairing and synapsis suggest that RAD51 is important for interhomolog recognition during meiosis. Marta et al. [57] claimed that DSBs were repaired in the RAD51-dependent manner, presumably using the sister chromatid as a template in meiotic recombination. Lambert et al. [58], reported that Rad51 is mostly responsible for homologous recombination between sister chromatids in somatic cells. Xu et al. [25] also revealed that RAD51 paralogs play essential roles in somatic homologous recombination for DNA repair. Taken together, RAD51 has different influences on meiosis in different organisms, and the function of RAD51 may be slightly different between mitosis and meiosis.

Functional Relationship between OsRAD51 and OsDMC1 in Meiosis
Once generated by Spo11, DSBs are processed by 5 nucleolytic resection to yield 3 single strands DNA (ssDNA) tails. Then, these DNA overhangs are coated by RAD51 and DMC1 [13]. The recombinases bind to the 3 ssDNA forming presynaptic filaments, which then invade double-stranded DNA to generate a heteroduplex [59]. The presence of meiosis-specific proteins that bond DNA strand exchange to homologous chromosomes makes meiotic HR different from mitotic HR. Previous studies have suggested that DMC1 and its cofactors specifically participated in the interhomolog bias [60]. Nevertheless, the exact way both recombinases work remains unclear. RAD51 plays a pivotal non-enzymatic role, facilitating DMC1 assembly and guiding it to invade a homologous chromosome rather than a sister chromosome [61,62]. Therefore, DMC1 is the dominant meiotic strandexchange enzyme [62,63].
In Arabidopsis, Kurzbauer reported that the number of DMC1 foci was sharply decreased in the Atrad51 mutant, whereas the number of RAD51 foci in Atdmc1 has no change compared to the wild type. They also confirmed that AtRAD51 and AtDMC1 cannot co-localize during any meiotic stage, and they declared that AtRAD51 and AtDMC1 occupy opposite DNA ends at a DSB [64]. Chen recently showed that AtRAD51 promotes AtDMC1 localization on meiotic chromosomes by inhibiting the SMC5/6 complex, while the SMC5/6 complex inhibits AtDMC1 during meiotic recombination. These authors confirmed that AtRAD51 interacts with AtDMC1 [24]. In rice, it is interesting that OsRAD51 cannot load onto Osdmc1 mutant chromosomes, and OsDMC1 cannot load onto Osrad51 mutant chromosomes. We also found that most OsRAD51 foci can co-localize with OsDMC1. Thus, we can draw the conclusion that the loading of OsRAD51 and OsDMC1 may be interdependent, and they may be functionally interdependent in rice meiotic recombination. Meiosis is abnormal in both the Osrad51 and Osdmc1 mutants, Osrad51 displays univalents, bivalents, and multivalents at metaphase I, and some chromosome fragmentation was observed at anaphase I, which could be the result of wrongly repaired DSBs, whereas Osdmc1 shows~24 univalents, the sister chromatid was used as the template to repair meiotic DSBs, and avoid the production of the chromosome fragmentation. Those results suggest that these two recombinases may play different roles in DNA repair. The meiotic chromosome behavior in the Osrad51 Osdmc1 mutant indicates that OsDMC1 may function in close collaboration with OsRAD51 in rice meiotic recombination. Yeast two-hybrid analysis, CoIP, and BiFC assays demonstrated that OsRAD51 cannot interact directly with OsDMC1. However, we found that both OsRAD51 and OsDMC1 can interact with OsFIGL1 and OsBRCA2. Kumar reported that BRCA2 is a positive accelerator that provides the stabilizing functions and FIGL1 acts as a negative regulator to disassemble the filaments in Arabidopsis [41]. Fu also confirmed that OsBRCA2 is indispensable for facilitating the loading of OsRAD51 and OsDMC1 onto 3 ssDNA of programmed DSBs during meiosis to facilitate single-end invasions of homologous chromosomes and precise recombination in rice [65]. Our data suggest that other proteins may mediate the functions of these two recombinases during rice meiosis. Although DMC1 and RAD51 are highly functional conserved in eukaryotes, there may be some differences between species. These results offer further insight into the inter-relationship between OsRAD51 and OsDMC1.

Vector Construction and Plant Cultivation
A multi-gene editing system for rice RAD51 and DMC1 was constructed. In this system, expression of the Cas9 gene was driven by the CaMV 35S promoter, and four sgRNA expression cassettes, which were driven by OsU3 promoters, were arranged in tandem. For target recognition (Supplementary Table S1 for the target sequences), 20-nt guide oligo-nucleotides were synthesized with appropriate adaptor sequences for seamless ligation with the promoter fragments. The tandem sgRNA expression cassettes were first constructed in the pUC19 intermediate vector, and the resultant DNA fragment was constructed into the pCAMBIA1300 backbone with the Cas9 expression cassette. The constructs were transformed into the calli of the japonica rice variety "Yandao 8" using Agrobacterium-mediated transformation. Other mutant materials were obtained from our previous study. All plant materials were grown in paddy fields. The genotypes were identified by PCR in the F 2 population derived from self-pollination of heterozygous F 1 plants. The sequences of the oligonucleotide primers used in this study are given in Supplementary Table S1.

Real-Time PCR for Gene Expression Analysis
Total RNA was separately extracted from the leaf, root, internode, and panicle of wild-type plants. The Bio-Rad CFX96 real-time PCR instrument was used to perform the real-time PCR assays. EvaGreen (Biotium) was used as the fluorescent dye for RT-PCR. The real-time PCR was performed with the specific primers RT-RAD51A1F, RT-RAD51A1R, RT-RAD51A2F, and RT-RAD51A2R for OsRAD51 as well as ActinF and ActinR for the rice ACTIN gene. The primer sequences are given in Supplementary Table S1. Each experiment was replicated three times. The experimental results were analyzed using Bio-Rad CFX Manager analysis software.

Immunofluorescence Analysis
Fresh young panicles were first fixed in 4% (w/v) paraformaldehyde for 30 min at room temperature. Chromosome preparation and immunofluorescence analysis were conducted as previously described [37]. The immunofluorescence analysis used the fol-lowing primary antibodies: rabbit antibodies directed against OsREC8 [39], PAIR2 [31], and OsRAD51 [25], and mouse antibodies directed against OsREC8 [40], PAIR3, ZEP1 [33], OsZIP4 [38], OsMER3 [37], OsDMC1 [28], and OsHEI10 [39] individually. All of the above antibodies were previously generated in our laboratory. Additionally, a 561-bp fragment of the PAIR3 cDNA encoding amino acids 248 to 393 was cloned into the vector pGEX4T-2. The anti-PAIR3 guinea pig antibody was generated with the PAIR3 fusion protein. The secondary antibodies, including rhodamine-conjugated goat anti-rabbit antibody, fluorescein isothiocyanate-conjugated sheep anti-mouse antibody, and AMCA-conjugated goat antiguinea pig antibody, were used for fluorescent detection. Images were captured with a Zeiss A2 fluorescence microscope equipped with a micro CCD camera. The super-resolution images were captured using a DeltaVision microscope (OMX TM V4; GE Healthcare, Chicago, IL, USA) and processed with S OFT W O R X (Applied Precision, Issaquah, WA, USA) to generate projected images. Co-localization analysis was performed using an automated graphic plugin running IMAGE J 1.37a software (http://wwwfacilities.uhnresearch.ca/wcif, accessed on 2 March 2021) as described by Li et al. [66].

Cytological Procedures and Data Analysis
Young panicles containing pollen mother cells (PMCs) entering meiosis were fixed in Carnoy's solution (ethanol:glacial acetic acid, 3:1) and stored at −20 • C to be used in chromosome spreads. PMCs at the meiotic stage were squashed on slides and stained with acetocarmine. The squashes were then covered with coverslips on the slides, which were then rapidly frozen in liquid nitrogen. The samples were dehydrated through an ethanol series (70, 90, and 100%) after the coverslips were removed. DAPI (4 ,6-diamidino-2-phenylindole) dissolved in an antifade solution (Vector Laboratories) was used to counterstain the chromosomes on the slides [37].

Yeast Two-Hybrid Assay
To perform Yeast two-hybrid assay, OsRAD51A1, OsRAD51A2, OsDMC1A, OsDMC1B, OsFIGL1, and OsBRCA2 were amplified using gene-specific primers (Supplementary Table S1) with KOD-plus polymerase and then ligated into the vectors pGBKT7 and pGADT7. Plasmid vectors were transformed into the Y2H Gold yeast strain using the LiAc/PEG method, and pGBKT7 and pGADT7 were used as the bait and prey vectors, respectively.
Yeast transformants were mated on YPDA medium for 48 h and selected on SD/-Trp-Leu plates for 36 h. Transformants were then selected on SD/-His-Ade-Trp-Leu medium containing X-α-Gal and AbA to test for positive interactions.

BiFC Assay
To conduct BiFC assays, OsRAD51A1, OsRAD51A2, OsDMC1A, OsDMC1B, OsFIGL1, and OsBRCA2 were amplified using gene-specific primers (Supplementary Table S1) with KOD-plus polymerase and then ligated into the BiFC vectors pSCYNE (SCN) and pSCYCE(R) (SCC) [67]. The plasmid constructs were transformed into protoplasts as described by Bart et al. [68]. After incubation at 28 • C in the dark for 18 h, the CFP signals were captured with a confocal laser scanning microscope at an excitation wavelength 405 nm (Leica TCS SP5).

Western-Blot Assay
Total proteins were separately extracted from wild type, Osra51a1 and Osrad51a2 rice panicles in meiosis stage with extraction buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% NP40, 1 mM phenyl methylsulfonyl fluoride, 1 × protease inhibitor cocktail). Protein samples were separated by SDS-PAGE on a 12% polyacrylamide gel and electroblotted onto polyvinylidene difluoride membrane (Amersham). Western blots were carried out to detect the OsRAD51 protein using OsRAD51 antibodies.

Fluorescence In Situ Hybridization
FISH analysis was conducted according to Zhang et al. [31]. The pAtT4 (telomeric repeats) and pTa794 (5S rRNA genes from wheat) [69] clones were used as probes. Chromosomes were counterstained with DAPI. The slides were observed under a Zeiss A2 fluorescence microscope and images were captured with a DVC1412 CCD camera using IPLab4.0 software.

Computational and Database Analysis
Protein sequence similarity searches were performed using BLASTp (www.ncbi.nlm. nih.gov/BLAST, accessed on 2 March 2021). Gene structure schematic diagrams were generated by GSDS (http://gsds.cbi.pku.edu.cn/index.php, accessed on 2 March 2021). The multiple sequence alignment diagram was generated using GenDoc software. The lengths of ZEP1 and REC8 on chromosomes were measured with Image J software.