RAD51B plays an essential role during somatic and meiotic recombination in Physcomitrella

The eukaryotic RecA homologue Rad51 is a key factor in homologous recombination and recombinational repair. Rad51-like proteins have been identified in yeast (Rad55, Rad57 and Dmc1), plants and vertebrates (RAD51B, RAD51C, RAD51D, XRCC2, XRCC3 and DMC1). RAD51 and DMC1 are the strand-exchange proteins forming a nucleofilament for strand invasion, however, the function of the paralogues in the process of homologous recombination is less clear. In yeast the two Rad51 paralogues, Rad55 and Rad57, have been shown to be involved in somatic and meiotic HR and they are essential to the formation of the Rad51/DNA nucleofilament counterbalancing the anti-recombinase activity of the SRS2 helicase. Here, we examined the role of RAD51B in the model bryophyte Physcomitrella patens. Mutant analysis shows that RAD51B is essential for the maintenance of genome integrity, for resistance to DNA damaging agents and for gene targeting. Furthermore, we set up methods to investigate meiosis in Physcomitrella and we demonstrate that the RAD51B protein is essential for meiotic homologous recombination. Finally, we show that all these functions are independent of the SRS2 anti-recombinase protein, which is in striking contrast to what is found in budding yeast where the RAD51 paralogues are fully dependent on the SRS2 anti-recombinase function.


INTRODUCTION
DNA double-strand breaks (DSBs) are a major threat for genome integrity and the accumulation of such damage leads to oncogenesis, cell death and severe dysfunction of cells. The DNA DSB repair system is crucially important for survival of living cells and highly conserved in eukaryotes. DSB repair can be achieved by multiple pathways, which can be grouped into two classes based on the involvement of sequence homology: non-homologous end-joining (NHEJ) and homologous recombination (HR). Repair of DSBs through HR generally preserves the integrity of the genomic material. In addition, HR events play an essential role in assuring proper meiotic chromosomal disjunction and represent an essential mechanism to create genetic diversity.
The RECA/RAD51 family proteins are involved in HR mechanisms in prokaryotic and eukaryotic cells and have been the object of considerable interest (for review see (1)). In the budding yeast Saccharomyces cerevisiae, 4 RecA-like homologues, Rad51, Rad55, Rad57 and Dmc1, have been identified. Rad51, Rad55 and Rad57 are expressed ubiquitously while Dmc1 is a meiosis-specific gene. Rad51 and Dmc1 genes are conserved in many eukaryotes, while Rad55 and Rad57 seem to be specific to yeasts. All these Rad51like proteins play roles in DSB repair including HR (2) and yeast mutants impaired in any of the Rad51-like proteins are defective in meiotic recombination and exhibit reduced spore viability (3). Even though Rad51 and Dmc1 have been demonstrated to be the core protein for the formation of the nucleoprotein filament for strand invasion, the function of the Rad55 and Rad57 paralogues in the process of HR is less clear. However, a recent study has demonstrated that the Rad55/Rad57 complex is involved in the formation of the Rad51/DNA nucleofilament, showing that this complex acts to counterbalance the anti-recombinase activity of SRS2 helicase (4).
In addition to RAD51 and DMC1, five RAD51 paralogues (XRCC2, XRCC3, RAD51B/RAD51L1, RAD51C/RAD51L2 and RAD51D/RAD51L3) have been identified in mammals, with 20-30% protein sequence identity with RAD51 and with each other (5). Unlike RAD51, no self-assembly of individual paralogues has been detected. Biochemical studies have unravelled different complexes: RAD51B-RAD51C-RAD51D-XRCC2, RAD51C-XRCC3, RAD51C-RAD51B, RAD51C-RAD51D and RAD51C-RAD51D-XRCC2 (reviewed in (6)). The functions of these various subcomplexes are still unclear mostly because knockouts of the RAD51 paralogue genes are embryo-lethal in animal model systems and hardly any in vivo data are available (7)(8)(9)(10)(11). Nevertheless, RAD51 paralogue functions investigated using immortalized vertebrate cell lines, such as chicken DT-40 or hamster CHO cells (12)(13)(14)(15), showed that RAD51 paralogues all play roles in somatic recombination, DNA repair and chromosome stability (for reviews see (5)). One of the complexes, the RAD51B/RAD51C complex, has been shown to promote strand exchange activity of RAD51-ssDNA filaments (16,17). Based on in vitro studies it has been recently proposed that this function could be mediated by the direct formation of the RAD51B/RAD51C complex on ssDNA that would partially stabilize the RAD51 nucleoprotein filament against the anti-recombinogenic activity of the Bloom's syndrome (BLM) helicase (18). Consistent with a major role of RAD51B in HR, chicken DT40 rad51b mutant cell lines show spontaneous chromosomal aberrations, high sensitivity to cross-linking agents, mild sensitivity to gamma rays and significantly attenuated RAD51 focus formation after gamma ray exposure. In addition, RAD51B deficiency abolishes targeted integration in the Chicken DT40 cell line mutant (14). Finally, the high expression levels of RAD51B observed in testis (19) suggest that this protein could also be involved in animal meiosis and possibly in meiotic HR.
Due to their tolerance, in term of development, to mutation of the RAD51 paralogues, plants represent a very attractive model to study the role of these proteins during different phases of development, and particularly meiosis. In the plant kingdom, the moss Physcomitrella patens (Physcomitrella) exhibits rates of gene targeting (GT) comparable to S. cerevisiae and has permitted advances in the comprehension of HR-mediated DNA repair and transgene integration via GT in plants (36)(37)(38)(39)(40)(41)(42). To gain further knowledge on somatic and meiotic HR in plants, we investigated the role of RAD51B in somatic and meiotic recombina-tion. The results show that RAD51B is indispensable for DNA damage repair and HR in somatic cells and that this role is independent of the SRS2 anti-recombination helicase. Moreover, we show that if some RAD51 functions are clearly dependent on RAD51B, others, like its role in development of Physcomitrella, are RAD51B independent. Finally, our data show for the first time an essential role for RAD51B during meiosis.

Plant material
Physcomitrella patens (Hedw.) B.S.G. 'Gransden2004' was vegetatively propagated as previously described (43). Individual plants were cultured as 'spot inocula' on BCD agar medium supplemented with 1 mM CaCl2 and 5 mM ammonium tartrate (BCDAT medium), or as lawns of protonemal filaments by subculture of homogenized tissue on BC-DAT agar medium overlaid with cellophane for the isolation of protoplasts. Transformation experiments were performed as previously described (36) using linear fragments of DNA generated by digestion of transforming vectors with restriction enzymes (40). Growth conditions for the generation of deletion strains were performed as described (41). The rad51-1-2 double mutant has been described previously (41).

Gene identification and isolation
Genomic DNA and total RNA was isolated from Physcomitrella as previously described . For verification of gene models, RNA was extracted as described in (44). Physcomitrella genomic sequences encoding the RAD51B and SRS2 genes were identified by Basic Local Alignment Search Tool (http://www.phytozome.net/Physcomitrella er.php). The available gene models were used for the design of polymerase chain reaction (PCR) primers to amplify cognate genomic sequences, which were cloned in the TOPO-TA (Life Technologies, USA) or pBluescript (Stratagene, USA) plasmids. PCR primers used are listed in Supplementary Table S1. In order to obtain a correct gene model for each sequence, full-length cDNA sequences were amplified from Physcomitrella polyribosome-derived RNA by rt-PCR (44). Predicted polypeptide sequences were aligned with the orthologous genes from Arabidopsis and Homo sapiens using CLUSTALW.
Moss protoplasts were transformed with pRAD51Bdelta digested with AvrII and BsrGI, or with a 3.2 kb pSRS2delta targeting fragment obtained by PCR amplification. Stable disruptants were selected by successive subculture on selective and non-selective medium and PCR analysed as described in (44). Clean deletions in the PpRAD51B (exons 3-10) and PpSRS2 (exons 2-24) genes were obtained by transient Cre recombinase expression (40). Deletions in the recombinant loci were confirmed by PCR amplification for five independent clones using gene-specific external primers PpRAD51B#1 and PpRAD51B#2, and PpSRS2#10 and PpSRS2#11, respectively. Primers APT#16 and APT#19 were used as positive controls (Supplemental Table S1). The five rad51bΔ deletion mutants show similar phenotypes when tested. Two out of the five independent clones are described in this study.

Targeted gene knockin
In order to replace the PpRAD51B gene from ATG to stop by the AtRAD51B cds, we produced the pPpAtRAD51B-KanR plasmid. First, we amplified a PCR fragment covering the AtRAD51B cDNA region (21) using Arabidopsis cDNA as the starting template (coordinate 47-1037 on Genbank: AB194809) and cloned it into a pUC18 plasmid to obtain pAtRAD51B. To generate the hybrid Physcomitrella-Arabidopsis RAD51B expression cassette pPpAtRAD51B, a 575 bp 5 targeting fragment (partial promoter/5 UTR of PpRAD51B, coordinates 11047187-11046613 on chromosome 25) and a 560 bp 3 targeting fragment (partial 3 UTR of PpRAD51B, coordinate 11042647-11042088 on chromosome 25) were amplified from Physcomitrella genomic DNA and cloned in pPpA-tRAD51B upstream of the ATG and downstream of the stop codon of the AtRAD51B cDNA, respectively. To obtain pPpAtRAD51B-KanR, the 35S::neoR cassette flanked by two LoxP sites was recovered from pBNRF (41) and inserted in pPpAtRAD51-2 between the AtRAD51B cDNA stop codon and the 3 UTR of PpRAD51B.
Moss protoplasts were transformed with pPpAtRAD51B-KanR digested with BsrGI. Stable disruptants were selected by successive subculture on selective and non-selective medium and PCR analysed as described in (44). Elimination of the 35S::neoR cassette was obtained by transient Cre recombinase expression (40). Replacement of the PpRAD51B gene by the AtRAD51B cds at the recombinant loci was confirmed by PCR amplification using locus-specific external primers PpRAD51B#1 and PpRAD51B#2. Primers APT#16 and APT#19 were used as positive controls (Supplementary Table S1).

Analysis of gene expression in mutants
Transcript abundance in selected knockout and knockin lines was determined by rt-PCR of cDNA as described in (44). Detection of RAD51B and SRS2 mRNA in mutant lines was by rt-PCR using primers indicated in Supplementary Table S1.

Bleomycin and ultraviolet (UV)-B sensitivity assays
Protoplasts of wild-type (WT), rad51B and rad51-1-2 mutants in BCDAT liquid medium supplemented with mannitol were treated with bleomycin (Bleocin inj., Euro Nippon Kayaku GmbH, Germany) at concentrations indicated in the text for 1 h. Protoplasts were washed two times and then resuspended in liquid mannitol medium. After 20 h in the dark, the protoplasts were spread on BCDAT agar medium supplemented with mannitol (ca. 10 5 protoplasts per Petri dish). After 6 days regeneration, the number of survivors was counted. We repeated these experiments three times.
Protoplasts of WT, rad51B and rad51-1-2 mutants were spread (ca. 50 000/plate) on protoplast agar medium (PpNH4 + 0.5% glucose + 8.5% mannitol). Plates were immediately irradiated with UV-B light (308 nm) in a Stratagene Stratalinker. The intensity of the irradiation was controlled using the internal probe of the Stratalinker and one plate of each strain was treated simultaneously. The experiment was repeated three times. Plates were immediately transferred to darkness for 24 h after treatment then to standard growth conditions for protoplast regeneration. Survival was determined after 1 week by microscopic observation.

Evaluation of spontaneous mutation frequency
Mutations in the PpAPT gene confer resistance to 2fluoroadenine (2-FA), a toxic compound for cells. The number of 2-FA resistant colonies that grow following protoplast regeneration reflects the frequency of spontaneous mutations. Protoplasts of WT and rad51b mutant were regenerated for 6 days on BCD agar medium supplemented with mannitol (ca. 10 5 protoplasts per Petri dish) and then transferred on to BCDAT agar medium supplemented with 10 M 2-FA (Fluorochem). After 2 weeks, the number of resistant colonies was counted. Experiments were repeated three times and statistically analysed using the Fisher's exact test. The adenine phosphoribosyl transferase (APT) genomic sequence from 2-FA resistant clones was subsequently PCR amplified using primers PpAPT#2 and PpAPT#20 (Supplementary Table S1) and sequenced to identify the mutation responsible for the resistance.

GT assays
Transformation efficiency and APT targeting frequency were measured as previously described (41). Moss protoplasts (4.8 × 10 5 ) were transformed with the nonhomologous pBHRF plasmid (41) digested with HindIII to produce a linear fragment containing the 35S::hygroR marker, or with the PpAPT-KO2 plasmid digested with BsaAI/HindIII to produce the targeting APT fragment containing the 35S::hygroR cassette (from pBHRF) flanked by genomic PpAPT sequences. Targeted integration of PpAPT-KO2 at the APT genomic locus confers resistance to 2-FA. We selected primary transformants (unstable + stable) with 25 mg/l hygromycin B (Duchefa). Integrative transformants were isolated following a second round of selection. Small pieces of protonemal tissue from these transformants were then transferred onto medium containing 5 M of 2-FA to detect APT GT events. Experiments were repeated three times and statistically analysed using the χ 2 test.

Cytology
Binocular observations were made with a Nikon SMZ1000.
For chromosome preparation by squashing one immature sporogonium was placed on a dry microscope slide in a drop of aceto-carmine stain (SIGMA, C6152) and crushed with a dissecting needle to remove the meiocytes and expose them to the stain. A coverslip was applied and meiocytes were squashed by tapping the coverslip gently with the flat back of a pencil. Slides were checked under a phase contrast microscope and slides showing meiocytes of the appropriate stages were immersed in liquid nitrogen until frozen (about 30 s), and their coverslips flicked with a razor blade. The air-dried preparation was dehydrated through an increasing ethanol series (70%-90%-100%, 2 min each). Airdried slides were mounted in Vectashield antifade mounting medium (Vector laboratories H-1000) containing a mix of DAPI (Roche, 10236276001) and propidium iodide (Sigma, P4170), 2 g/ml each.
For chromosome preparation by spreading fixed leafyshoots bearing immature sporogonia were transferred to an embryo dish and washed in 10 mM citrate buffer pH 4.5 at room temperature and the buffer changed once before incubating with an enzyme mixture comprising 0.3% w/v cytohelicase (C1794), 0.3% w/v pectolyase (C8274) and 0.3% w/v cellulase (P5936) (all Sigma) in citrate buffer for 3.5 h at 37 • C. Replacing the enzyme mixture with ice cold water stopped the reaction. A single sporogonium was transferred to a clean slide. The digested sporogonium was tapped out in 5 l of water, using a fine needle. Note that 15 l 60% acetic acid was added to the preparation and after gentle stirring with a hook, the slide was then placed on a 45 • C heating block for 2 min. Air-dried slides were mounted in Vectashield antifade mounting medium (Vector laboratories H-1000) containing DAPI (Roche, 10236276001) at 2 g/ml. All observations were made using a Zeiss Axio Im-ager2 microscope. All images were further processed with ZEN 2011 (Zeiss) or Adobe Photoshop 7.0.

Molecular and phenotypic characterization of the PpRAD51B deletion mutant
Sequence homology searches of the Physcomitrella genome identified a single putative homologue for all the RAD51 paralogues except XRCC3 (Supplementary Figure S1). PpRAD51B is located on chromosome 25 (Phpat.025G051900) and our reverse transcriptase-PCR (RT-PCR) and Rapid Amplification of cDNA Ends (RACE) analyses of this gene, demonstrated that it consists of 12 exons and 11 introns ( Figure 1A) and contains an open reading frame (ORF) of 362 amino acids ( Figure 2). Phylogenetic analysis revealed that the protein encoded by PpRAD51B belongs effectively to the RAD51B family of proteins ( Figure 2). The PpRAD51B protein has 46% and 36% sequence identity and 63% and 57% sequence similarity to AtRAD51B and HsRAD51B, respectively. There is extensive similarity among all the RAD51B polypeptides and PpRAD51B contained the characteristic features for RecA/Rad51 family proteins ( Figure 2).
To investigate PpRAD51B function we generated a rad51b deletion mutant, named rad51bΔ, using targeted gene disruption followed by Cre/lox mediated elimination of the resistance cassette ( Figure 1A and Materials and Methods). We confirmed that a portion of the coding region was removed by PCR using primers that flanked the deletion ( Figure 1A) and sequence analysis confirmed the presence of a single LoxP site at the deletion site (not shown). RT-PCR analysis established that the full-length transcript was no longer produced in the mutant ( Figure 1B). For all further experiments, we used two independent rad51bΔ strains, which displayed similar phenotypes.

The PpRAD51B deletion mutants show subtle developmental defects
Like all mosses, the life cycle of Physcomitrella is characterized by an alternation of two generations, a haploid gametophyte that produces gametes and a diploid sporophyte where haploid spores are produced. A spore develops into a filamentous structure called protonema. Protonema filaments grow exclusively by tip growth of their apical cells and can originate side branches from subapical cells. Some side branch initial cells can differentiate into buds rather than side branches. These buds give rise to gametophores, more complex structures bearing leaf-like structures, rhizoids and the sexual organs: female archegonia and male antheridia. Sperm cells can swim from the antheridia to an archegonium and fertilize the egg within. The resulting diploid zygote originates a sporophyte, where thousands of haploid spores are produced by meiosis.
The phenotype of the rad51bΔ mutants was characterized throughout this entire life cycle. By contrast with the double rad51 mutant already described (41,47), that shows a strong developmental phenotype on rich medium ( Figure  3C and Supplementary Figure S2), the rad51bΔ mutants display only a very weak defect in gametophytic development ( Figure 3E and 3F), showing a normal protonema development but fewer gametophores than WT (81% of the WT, see Figure 3A and Supplementary Figure S2). However, after multiple subcultures we occasionally observed some individuals showing strong developmental phenotypes (see one example in Figure 3G and 3H) suggesting the possibility of a genetic instability in mutant lines.
Gametophores produced by the rad51bΔ mutants are normal. However, compared to WT, rad51bΔ plants produce fewer mature sporophytes ( Figure 4A) and these sporophytes produce capsules containing fewer spores (on average, 4000 spores/sporogonium for the WT and 900 spores/sporogonium for the mutant, n = 20). Spores produced by the rad51bΔ plants were smaller than the WT, irregular in shape, often remained as tetrads and could not germinate ( Figure 4B and Supplementary Figure S3). Thus, if PpRAD51B is not essential for vegetative development, it is essential for fertility.

The rad51bΔ mutants display a mutator phenotype
In order to confirm the genetic instability phenotype of the rad51bΔ mutant (see Figure 3G) and to evaluate its nature, we measured the intensity of its mutator phenotype. For this purpose, we investigated the effect of the loss of the PpRAD51B gene on mutation frequency of the APT reporter gene. Spontaneous mutations in this gene leading to loss of APT activity confer resistance to adenine analogues, such as 2-FA and can be selected on standard medium supplemented with 10 M 2-FA. We tested a total of 5.3 × 10 6 rad51bΔ protoplast derived colonies and 107.2 × 10 6 WT colonies for resistance to 2-FA. We observed two resistant clones in the WT and a total of 145 2-FA resistant clones in the rad51bΔ mutant (Table 1). Thus, our analysis revealed that mutation rates were increased at least 1350 times in the rad51bΔ mutant as compared to WT, a value three times greater than the mutation rate observed in the rad51-1-2Δ double mutant (41) and 20-fold lower than that observed in the mismatch repair mutant PpMsh2Δ (40). In order to characterize the type of mutations found in the rad51bΔ mutant context we partially sequenced the APT gene of 13 randomly chosen 2-FA resistant clones (Supplementary Table S2). Mutations resulting from transitions, transversions and deletions were recovered. These data support a direct involvement of PpRAD51B in repairing naturally occurring DNA damage to prevent the accumulation of mutations in the genome.
The rad51bΔ mutants are hypersensitive to UV-B and bleomycin C WT, rad51-1-2Δ and rad51bΔ mutant were also analysed for their sensitivity to DNA damaging agents. Sensitivity of mutants and WT strains to UV-B (308 nm) was investigated using a protoplast survival assay (40). The rad51bΔ mutant displayed increased sensitivity to UV-B compared to the WT; this increase is of the same order of magnitude as that observed for the rad51-1-2Δ mutant ( Figure 5A). We further investigated sensitivity of the mutants to the DSBinducing agent bleomycin (48). We tested the acute toxicity of bleomycin in WT and in the different mutants at the cellular level. Following incubation for 1 h with increasing concentrations of bleomycin, the ability of protoplasts to divide and regenerate into colonies was assessed by subculture on drug-free medium. Survival was calculated as the ratio    of protoplasts surviving after 15 days of regeneration following treatment to the number of protoplasts undergoing normal regeneration without treatment. The LD50 for the WT was about 500 ng/ml bleomycin, whereas the rad51-1-2Δ and rad51bΔ cells were more sensitive, with an LD50 of 100 ng/ml ( Figure 5B). The rad51bΔ cells show a sensitivity similar to the cells depleted of the RAD51 genes, previously described as hypersensitive to bleomycin (44).

GT is abolished in the rad51bΔ mutants
In order to investigate the involvement of the PpRAD51B protein in transgenesis, we quantified GT efficiency in WT and rad51bΔ mutant after transformation with linear pPpAPT-KO2 which confers hygromycin resistance (selective marker) and 2-FA resistance upon targeted integration at PpAPT locus, as described in (41). GT efficiency was determined as the frequency of 2-FA resistant plants among hygromycin-resistant integrative transformants. In the rad51bΔ mutant, GT was completely abolished, even though relative transformation frequencies (RTF for Hyg R around 1.6 × 10 −3 ) was similar to that observed in the WT strain with a homologous vector ( Table 2). This experiment shows that PpRAD51B is essential for GT. Moreover, the fact that RTF is not altered in the rad51bΔ mutant suggests that the capacity of random integration of the linear vector is enhanced in the absence of PpRAD51B.
To further assess the specific role played by RAD51B in the mechanism of random DNA integration in Physcomitrella, we determined transformation efficiencies with a vector sharing no homology with the moss genome, pBHRF, in WT and rad51bΔ plants. We then compared these values with Hyg R RTF achieved with the PpAPT-KO2 homologous vector. The absence of sequence homology on the transforming DNA reduces transformation frequencies ∼8-fold in WT (Table 2), which is consistent with previous reports (36). By contrast, Hyg R RTF was similarly high with both vectors in rad51bΔ  mutant, demonstrating that PpRAD51B directly represses random integration independently of the presence of moss homologous sequence on the transforming DNA. These data demonstrate that loss of function of RAD51B in Physcomitrella de-repressed the random integration that probably takes place via an illegitimate recombination pathway upon transformation as previously observed in the absence of RAD51 (41).

A flowering plant RAD51B cannot complement the Physcomitrella rad51bΔ mutant defaults
GT efficiency is low in vascular plants compared to Physcomitrella but can be detected (49), implying that the recombination machinery necessary for GT is present. The Physcomitrella RAD51B is in the plant group of RAD51B proteins and is thus clearly of plant origin (Supplementary Figure S1). Furthermore, the Arabidopsis AtRAD51B protein shows a high level of identity (45%) with the PpRAD51B protein ( Figure 2). In order to determine whether the AtRAD51B protein could fulfil the RAD51B function of Physcomitrella we replaced the PpRAD51B gene (ATG to stop) by the AtRAD51B cds (for details see Materials and Methods). Complete loss of the PpRAD51B transcripts and expression of the AtRAD51B cDNA were confirmed by RT-PCR in the Pp::AtRAD51B lines (Figure 1). For all experiments, we used two independent Pp::AtRAD51B lines; results were comparable in the two lines. The Pp::AtRAD51B lines show the same developmental phenotypes as the rad51bΔ mutant (not shown) including sterility. GT efficiency was tested at the PpAPT locus as previously described. We observed that GT efficiency is completely abolished in both Pp::AtRAD51B lines compared to the WT (Table 1). These results indicate that expression of the Arabidopsis RAD51B cDNA under the control of the Physcomitrella RAD51B promoter is not able to restore a correct fertility or high GT efficiency.

Absence of SRS2 does not restore GT in the rad51bΔ mutant
In S. cerevisiae, Srs2 encodes a helicase which is a key regulator of Rad51 filament formation and disassembly and is thus considered as an 'anti-recombinase'. It has been shown that deletion of Srs2 can suppress the sensitivity to ionizing radiation of the mutants of the S. cerevisiae Rad51 par-Nucleic Acids Research, 2014, Vol. 42, No. 19 11973 alogues, Rad55 and Rad57 (4,50). Moreover, the Rad55-Rad57 heterodimer has been shown to counteract the antirecombination activity of the Srs2 helicase and a model has been proposed whereby Rad51 pre-synaptic filament formation is modulated by a balance between the stabilizing function of Rad55-Rad57 and the destabilizing function of Srs2 anti-recombinase (4). In order to test the possibility that the RAD51 pre-synaptic filament formation could be modulated by such a balance and antagonistic effects between the RAD51B and SRS2 proteins in Physcomitrella, we have produced a mutant in the PpSRS2 gene and analysed its effect on GT.
By performing a sequence homology search with the yeast SRS2 protein sequence, we were able to identify only one candidate gene homologue in Physcomitrella, which is located on chromosome 1 (Phpat.001G116600). These results were in agreement with the previously described presence of an SRS2 homologue in bryophytes (51). Our RT-PCR and RACE analyses of this gene, named PpSRS2, demonstrated that it consists of 29 exons and 28 introns ( Figure 1A), and contains an ORF of 1301 amino acids. Sequence analysis and domain determination revealed that the PpSRS2 protein has 50% and 26% sequence identity and 64% and 43% sequence similarity to AtSRS2 and ScSRS2, respectively, and that the conserved UVRD/REP domain of the proteins is clearly aligned (Supplementary Figure S4).
To investigate PpSRS2 function and its potential interaction with PpRAD51B we generated a srs2 deletion mutant, named srs2Δ, and we retransformed this mutant to produce a double srs2Δ/rad51b mutant ( Figure 1A). RT-PCR analysis established that the full-length transcripts were no longer produced in the mutants ( Figure 1B). In terms of gametophytic development srs2Δ could not be distinguished from WT, nor srs2Δ/rad51b from rad51bΔ mutant ( Figure  3). The srs2Δ mutant has normal sporogonia and produces normal and viable spores, the double srs2Δ/rad51b mutant, like the rad51bΔ, is completely sterile ( Supplementary Figure S3).
We quantified GT efficiencies in the srs2Δ and srs2Δ/rad51b mutants after transformation with linear pPpAPT-KO2. In the srs2Δ mutant, GT efficiency was similar to the WT but in the double srs2Δ/rad51b mutant, as already observed in the simple rad51bΔ mutant, GT was completely abolished (Table 2). These data demonstrate that SRS2 is not involved in GT and that its loss of function does not restore GT in the rad51bΔ mutant context.

Meiotic recombination is altered in the rad51bΔ mutant
In order to investigate the origin of the sterility of the rad51bΔ mutant we compared the spore germination efficiency between WT and the different mutants. The rad51bΔ mutant, Pp::AtRAD51B line and srs2Δ/rad51b double mutant produce no viable spores, compared to a mean germination efficiency of 97.6% and 96.2% for the WT and srs2Δ mutant plants, respectively (Supplementary Figure  S3). Thus, PpRAD51B is essential for spore formation and loss of function of SRS2 or addition of the AtRAD51B function does not restore spore viability in the rad51bΔ mutant context. Spores are the direct product of meiosis. In order to explore the potential role of RAD51B in meiosis in Physcomitrella we have compared meiotic progression between the WT and the rad51bΔ mutant.
Young sporangia were first examined by fluorescence microscopy after cell squashing and DAPI staining of the chromosomes. In the WT (Figure 6), the Physcomitrella chromosomes condense during meiotic prophase I ( Figure  6A and B) and can be seen as 27 pairs of homologous chromosomes (bivalents) in metaphase I ( Figure 6C and Supplementary Figure S5). Homologous chromosomes then separate from each other and migrate to the opposite poles of the cell in anaphase I ( Figure 6D). The second meiotic division starts with the alignment of chromosomes in metaphase II ( Figure 6E) and is followed by separation of sister chromatids in anaphase II ( Figure 6F). Telophase II ensues, chromosomes decondense ( Figure 6G) and cytoplasm is partitioned to produce a tetrad containing four haploid microspores ( Figure 6H), which will differentiate into mature spores ( Figure 4B). In young sporangia of the rad51bΔ mutant, prophase I proceeds ( Figure 6I and J) but, at the transition between metaphase I and anaphase I, a mass of entangled chromosomes can be seen in place of the 27 expected bivalents ( Figure 6K). Hardly any chromosome segregation could be observed, but cytokinesis nevertheless takes place, always generating four daughter cells with aberrant quantities of DNA ( Figure 6L-N) and eventually producing unbalanced tetrads ( Figure 6O and P). Maturation of these aberrant tetrads of microspores results in aggregated spores with abnormal shape ( Figure 4B).
In order to better characterize meiotic prophase, we established a chromosome spreading technique (see Materials and Methods) that allowed the visualization of the close association of chromosomes along their length (Figure 7A, C and D) (n = 25). These close associations are very likely to reflect the formation of mature synaptonemal complexes between homologous chromosomes during WT meiosis (pachytene stage). In the rad51bΔ mutant, such close association was never observed ( Figure 7B) (n = 19), except on short portion of the nucleus ( Figure 7E and F).
Taken together, these results show that RAD51B in Physcomitrella is required for early meiotic events and for the formation of normal bivalents.

DISCUSSION
We report here the identification and characterization of the Physcomitrella mutant for the RAD51 paralogue RAD51B. The first important result of this study is that Pprad51b mutants display only a weak developmental phenotype, characterized by a slight reduction of gametophore production. This is in strong contrast with the situation in vertebrates where mutation in RAD51B impairs viability (5), and also different from vascular plants were the rad51b mutant does not show any developmental phenotype (22).
Although Pprad51B mutants are viable and show no strong phenotypic defect under normal conditions, these mutants show an increase in sensitivity to UV-B compared with the WT. A role for RAD51B in UV-C-induced apoptosis of hamster CHO cells has been proposed (52), and more recently, a stronger increase of UV-C-induced HR frequency has been observed in the Arabidopsis rad51b mutant compared to the WT (35). However, the role of the RAD51B protein in mid-wavelength UV (UV-B) sensitivity, that is likely to be more biologically relevant than UV-C, has not been studied and we show here for the first time the involvement of RAD51B in this stress response. Deciphering the UV-B resistance pathway in plants is of particular interest in light of the potential influence of the solar UV-B radiation on the genomic stability of plant populations (53).
The Pprad51B mutant also shows a strong increase in sensitivity to the DSB inducing agent bleomycin. Hypersensitivity to bleomycin is a hallmark of mutants affected in RAD51 and the RAD51 paralogues in S. cerevisiae (54) and can be correlated to a default in HR-dependent DNA DSB repair. A Rad51b mutant of DT40 chicken cells shows a slight increase in sensitivity to gamma rays, another DSB inducing agent (14). In vascular plants, the rad51b mutant of Arabidopsis shows a normal or only a slight increase in sensitivity to gamma rays (21,22) and is slightly more sensitive to bleomycin than the WT (23). The strong increase of sensitivity to bleomycin of the Physcomitrella mutant is in good agreement with an essential role of the HR-dependent DNA DSB repair pathway in this moss (41), and confirms, in regard of the difference of sensitivity to bleomycin be- tween the Physcomitrella and Arabidopsis rad51b mutants, the differences in the use of recombination pathways between the two plants (47).
We have shown previously that the loss of RAD51, that is essential for HR function, induces genetic instability in Physcomitrella (41). Here, we show that the rad51b mutant also exhibits a mutator phenotype and thus that the RAD51B protein is also essential for genetic stability. The higher frequency of mutations in the rad51 and rad51b mutants compared to the WT is probably due to the role of the HR machinery in error-free-mediated damage repair during DNA replication blocks (2,55). However, it must be noted that, in addition to its general role in fork-associated DNA damage repair via HR, RAD51 has also an early role during replication blocks, facilitating replication fork restart when forks are still viable (56,57). In animal cells, this second RAD51 function consists, in part, of a direct interaction between polEta and RAD51 that is essential for the restart of the replication fork (58). If this second role of RAD51 is conserved in Physcomitrella, failure to restore synthesis and replication in the absence of RAD51 (15,55,58) could explain the decrease in growth rate ob-served in the Physcomitrella rad51 mutant and its associated developmental phenotype (41,47). This second function of RAD51 would be independent of RAD51B and in the rad51b mutant, the presence of RAD51 and polEta could permit the restart of the replication fork and thus cell division would be only slightly affected in this context. Nevertheless, in addition to the mutations resulting from the failure to repair DNA damage via HR, the rad51b mutant would accumulate mutations due to the restart of replication via an error-prone polymerase (59). Supporting this hypothesis, mutations, like deletions, that were not found in the Physcomitrella msh2Δ mutant context for example (40), and could result from the translesion synthesis-mediated bypass of damages (60) can be found in the rad51bΔ mutant. Taken together, these results could account for the weaker developmental phenotype and stronger mutator phenotype observed for the Physcomitrella rad51b mutant compared to the rad51 mutant.
The lethality of the RAD51B mutations in animals has made the study of the role of RAD51B in HR difficult. Moreover, due to the absence of an efficient test for GT in most animals and vascular plants, no data are available on the putative role of RAD51B in HR-mediated GT in these organisms. However, studies in human cells recently showed that RAD51-dependent HR is reduced, but not abolished, in the absence of RAD51B (61). Similarly, based on the use of an in planta intrachromosomal recombination assay, it has been shown in Arabidopsis that somatic recombination is decreased but not abolished, in a rad51b mutant context suggesting that RAD51B is involved but not essential for RAD51 nucleofilament formation in Arabidopsis (24). By contrast, RAD51B has been shown to be essential for HRmediated GT in chicken DT-40 cells (14). The situation is also different in unicellular eukaryotes: while RAD51B is essential for GT in trypanosomatidae (62), the Rad55 and Rad57 paralogues are not essential for GT in S. cerevisiae (63). We show here that the RAD51B protein is absolutely required for HR-mediated GT in Physcomitrella. The essential role of this RAD51 paralogue in HR for GT competent eukaryotes, such as Physcomitrella, trypanosomatidae and DT-40 cell lines contrasts with the less prominent role of the RAD51 paralogues for GT observed in S. cerevisiae (63). The importance of RAD51B for GT in Physcomitrella clearly distinguishes these two species and reflects the more important role of RAD51 in GT in Physcomitrella compared to S. cerevisiae (41).
The HR machinery is essential for targeted integration of a homologous vector. Does this mechanism affect untargeted integration of a non-homologous vector? In S. cerevisiae, illegitimate recombination of non-homologous DNA fragments occurs at low frequencies (64), and the absence of RAD51 leads to a further decrease in the number of Illegitimate Recombination (IR) events compared to WT (63). By contrast, in A. nidulans, S. pombe and Physcomitrella (41,(65)(66) the absence of RAD51 increases IR rates. Our data show that in Physcomitrella, deletion of RAD51B, like deletion of RAD51, leads to an increase in IR. This implies that whereas in S. cerevisiae the HR machinery is necessary for the integration of exogenous DNA with or without homology, in moss there is clearly competition for DNA integration between HR and IR (probably NHEJ). Thus, the idea that the outcome of transgenesis reflects the balance between HR-and IR-mediated DNA repair is strongly supported in moss.
The Rad55/Rad57 complex promotes Rad51 nucleoprotein filament stabilization and facilitates Rad51-dependent strand invasion reactions in yeast (2). Because the mammalian RAD51 paralogue complexes share biochemical features and genetic functions with the Rad55/Rad57 heterodimeric complex it has been proposed that early role of the RAD51 paralogues in HR is to promote RAD51 nucleoprotein filament assembly. Recently, in vitro studies have shown that the RAD51B-RAD51C complex can regulate the stability of the RAD51 nucleoprotein filament by acting as a pro-recombinase against the BLM helicase that acts as an anti-recombinogenic factor (18). However, evidence of direct in vivo interaction between RAD51 and the RAD51 paralogues is still lacking and the respective roles of the different RAD51 paralogues in the formation of the RAD51 nucleofilament is still unclear (6). In S. cerevisiae, the Rad55/Rad57 paralogue-mediated Rad51 filament assembly was shown to be counterbalanced by the Srs2 antirecombinase (4,50) which led us to test the potential inter-play between SRS2 and RAD51B in Physcomitrella. For this purpose we isolated the first plant srs2 mutant. Our results show that the srs2 mutant has no obvious developmental phenotype, is fully fertile, is not affected in GT and shows a meiotic behaviour similar to WT which is very different from the situation in yeast where Srs2 is involved in somatic DNA repair and in meiosis (67). We also observed that the absence of SRS2 does not restore the GT capacity of the rad51b mutant. It should be noted, however, that even in yeast, rad55 or rad57 recombination defects are only partially restored by the srs2 mutation (4). The diversification of the RAD51 paralogues in multicellular eukaryotes probably reflects a specialization of these paralogues in different aspects of recombination. One possibility is that RAD51B is not involved in the recombination steps potentially controlled by SRS2 and that another RAD51 paralogue might interact with SRS2 to control RAD51 loading in multicellular eukaryotes, a hypothesis that we are currently testing.
Yeast mutants impaired in any of the RAD51 paralogues are defective in meiotic recombination and exhibit reduced spore viability (3). The situation in multicellular eukaryotes is more complex. Indeed, while the XRRC3 and RAD51C paralogues have been shown to be indispensable for meiosis in Drosophila, Arabidopsis and rice (25)(26)(27)(68)(69), the RAD51D paralogue has been shown to be essential for meiosis only in rice (28). Moreover, it was recently shown in Arabidopsis that the absence of XRCC2, and to a lesser extent of RAD51B, increases meiotic crossing over rates (24). By contrast, we show here that RAD51B is essential for meiosis in Physcomitrella and that in its absence, homologous chromosomes do not form bivalents at metaphase I but instead appear as a mass of entangled chromosomes that cannot subsequently segregate. This phenotype is associated with evidence of synapsis defects during prophase and is reminiscent of the meiotic phenotype of the flowering plant mutants affected in meiotic DSB repair, such as rad51, xrcc3, mnd1 or hop2 (29). Therefore, it is very likely that RAD51B is involved in meiotic recombination in Physcomitrella, although more studies will be necessary to define precisely at which steps of meiotic DSB repair. Further studies will also be necessary to determine whether the essential role of RAD51B during meiosis can be extrapolated to other organisms. The high expression levels of RAD51B observed in mouse testis (19) suggests that this could be the case.
Our results and the recent demonstration that RAD51D is also essential for rice meiosis (28) contrast strikingly with the role of these genes during Arabidopsis meiosis (24) and reinforce the idea that the function of the RAD51 paralogues during meiosis can vary from one species to another. More generally, the differences observed between the Physcomitrella and Arabidopsis rad51b or rad51 mutants and the fact that the Arabidopsis RAD51B or RAD51 genes cannot complement the defects observed in the corresponding Physcomitrella mutants (this study and unpublished data), leave the open question of whether there are distinctly different somatic and meiotic HR machineries in bryophytes and vascular plants.