Differing requirements for the Arabidopsis Rad51 paralogs in meiosis and DNA repair

Authors


(fax +33 473 407 777; e-mail chwhite@univ-bpclermont.fr).

Summary

In addition to the recombinase Rad51, vertebrates have five paralogs of Rad51, all members of the Rad51-dependent recombination pathway. These paralogs form two complexes (Rad51C/Xrcc3 and Rad51B/C/D/Xrcc2), which play roles in somatic recombination, DNA repair and chromosome stability. However, little is known of their possible involvement in meiosis, due to the inviability of the corresponding knockout mice. We have recently reported that the Arabidopsis homolog of one of these Rad51 paralogs (AtXrcc3) is involved in DNA repair and meiotic recombination and present here Arabidopsis lines carrying mutations in three other Rad51 paralogs (AtRad51B, AtRad51C and AtXrcc2). Disruption of any one of these paralogs confers hypersensitivity to the DNA cross-linking agent Mitomycin C, but not to γ-irradiation. Moreover, the atrad51c-1 mutant is the only one of these to show meiotic defects similar to those of the atxrcc3 mutant, and thus only the Rad51C/Xrcc3 complex is required to achieve meiosis. These results support conservation of functions of the Rad51 paralogs between vertebrates and plants and differing requirements for the Rad51 paralogs in meiosis and DNA repair.

Introduction

Homologous recombination (HR) and non-homologous end-joining (NHEJ) are the two major pathways for DNA double-strand break (DSB) repair (reviews by Dudas and Chovanec, 2004; Lees-Miller and Meek, 2003). While NHEJ is an error-prone pathway, frequently leading to deletions and/or insertions, repair of DNA 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 proteins involved in HR mechanisms in eukaryotic cells have been extensively studied and mainly belong to the RAD52 epistasis group, first identified in budding yeast (for review see Dudas and Chovanec, 2004). Within this group, the sub-family of the Rad51-like proteins has been the object of a considerable interest as the finding that Rad51 is the homolog of the bacterial recombinase RecA (Aboussekhra et al., 1992; Shinohara et al., 1992). In addition to the highly conserved recombinase Rad51, three Rad51-like proteins have been identified in budding yeast: Dmc1 is a meiosis-specific Rad51-like protein and is conserved in many eukaryotes, while Rad55 and Rad57 are expressed ubiquitously and seem to be specific to yeasts. All these Rad51-like proteins play roles in DSBs repair and/or HR (for review see Dudas and Chovanec, 2004).

As well as Rad51 and Dmc1, the genomes of vertebrates encode five additional Rad51-like proteins, referred to as the Rad51 paralogs: Rad51B (Albala et al., 1997; Cartwright et al., 1998b; Rice et al., 1997), Rad51C (Dosanjh et al., 1998), Rad51D (Cartwright et al., 1998b; Pittman et al., 1998), Xrcc2 (Cartwright et al., 1998a; Liu et al., 1998) and Xrcc3 (Liu et al., 1998; Tebbs et al., 1995). Two-hybrid and immunoprecipitation experiments have shown that these five Rad51 paralogs form two complexes: an heterodimer composed of Rad51C and Xrcc3 (CX3) and an heterotetramer composed of Rad51B, Rad51C, Rad51D and Xrcc2 (BCDX2) (Liu et al., 2002; Masson et al., 2001; Miller et al., 2002, 2004; Schild et al., 2000; Wiese et al., 2002). The embryonic lethality observed in individual knockouts of mouse RAD51B, RAD51D and XRCC2 shows that, as is the case for Rad51, the Rad51 paralogs are required for animals viability (Deans et al., 2000; Pittman and Schimenti, 2000; Shu et al., 1999; Tsuzuki et al., 1996). However, in contrast to Rad51, knockouts of the Rad51 paralogs in chicken DT-40 cells do not lead to cell death, indicating that these proteins are not required for vertebrate cells viability (Sonoda et al., 1998; Takata et al., 2000, 2001). The numerous experiments performed with mutant cells defective in the different Rad51 paralogs have shown that they all play roles in somatic recombination, DNA repair and chromosome stability (Brenneman et al., 2000; Cui et al., 1999; Deans et al., 2000, 2003; French et al., 2002; Godthelp et al., 2002; Johnson et al., 1999; Liu et al., 1998; Mohindra et al., 2004; Pierce et al., 1999; Takata et al., 2000, 2001; Tebbs et al., 1995). However, the embryonic lethality of mutants defective in any one of the Rad51 paralogs has complicated investigation of their possible involvement in meiosis (Deans et al., 2000; Pittman and Schimenti, 2000; Shu et al., 1999).

The absence of Rad51 foci following DNA damaging treatment in cells defective in any one of the Rad51 paralogs suggests a role in the initial steps of HR process (Bishop et al., 1998; Godthelp et al., 2002; Liu, 2002; O'Regan et al., 2001; Takata et al., 2000, 2001; Tarsounas et al., 2004). This hypothesis is supported by several biochemical studies. In vitro Rad51-dependent strand exchange reactions have shown that a sub-complex composed of Rad51B and Rad51C facilitates the assembly of the Rad51-ssDNA nucleoprotein filament in the presence of RPA and thus has a mediator role similar to that of the Rad55-Rad57 heterodimer in yeast (Lio et al., 2003; Sigurdsson et al., 2001; Sung, 1997). The Rad51C protein, the CX3 complex and a Rad51D/Xrcc2 sub-complex possess homologous pairing activities similar to that of Rad51 (Kurumizaka et al., 2001, 2002; Lio et al., 2003).

Several recent studies however support a later role for the Rad51 paralogs. In the absence of the Xrcc3 protein, both conversion tract lengths and the frequency of discontinuous tracts are increased (Brenneman et al., 2002). In vitro, the Rad51B protein and BCDX2 complex preferentially bind branched DNA strands, such as Holliday junctions (HJ) (Yokoyama et al., 2003, 2004), and Rad51C and Xrcc3 play roles in HJ resolution (Liu et al., 2004). Finally, the meiotic defects observed in an Arabidopsis xrcc3 mutant suggest that the Xrcc3 protein plays a post-synaptic role (Bleuyard and White, 2004).

The genome of Arabidopsis thaliana codes for seven Rad51-like proteins. In addition to the previously identified Rad51, Dmc1, Rad51C and Xrcc3 Arabidopsis homologs (Doutriaux et al., 1998; Klimyuk and Jones, 1997; Osakabe et al., 2002; Sato et al., 1995; Urban et al., 1996), the construction of a phylogenetic tree has allowed us to identify the Arabidopsis homologs of Rad51B, Rad51D and Xrcc2, showing that Arabidopsis has the same family of Rad51-like proteins as vertebrates. To investigate the conservation of functions of the Rad51 paralogs between vertebrates and plants, we identified mutants defective for three of the four other Arabidopsis Rad51 paralogs. Absence of any one of the Rad51B (AtRad51B), Rad51C (AtRad51C) and Xrcc2 (AtXrcc2) Arabidopsis homologs confers hypersensitivity to the DNA cross-linking agent Mitomycin C (MMC), but not to ionizing radiation. Furthermore, only mutants impaired for the AtRAD51C gene show meiotic defects similar to those of atxrcc3 mutants. These results clearly show that the role of the Rad51 paralogs in DNA repair is conserved between vertebrates and plants and that only AtRad51C and AtXrcc3 (Bleuyard and White, 2004), which together form the CX3 complex, play essential roles in meiosis in Arabidopsis, and very probably in other higher eukaryotes.

Results

Identification and molecular characterization of Arabidopsis mutants defective for the Rad51 paralogs

The genome of the model plant A. thaliana codes for seven proteins of the Rad51 recombinase family. Previous studies have reported the identification of Rad51 (AT5G20850), Dmc1 (AT3G22880), Rad51C (AT2G45280) and Xrcc3 (AT5G57450) homologs (Doutriaux et al., 1998; Klimyuk and Jones, 1997; Osakabe et al., 2002; Sato et al., 1995; Urban et al., 1996). To define the relationships existing between Arabidopsis Rad51-like proteins and Rad51-like proteins from other model species, we performed a phylogenetic analysis with 23 Rad51-like proteins from Arabidopsis, Drosophila, human and Saccharomyces (Figure 1a). The resulting phylogenetic tree clearly shows that Arabidopsis has a single homolog for each of the five Rad51 paralogs first identified in vertebrates (AT2G28560, AT1G07745 and AT5G64520 products correspond respectively to the AtRad51B, AtRad51D and AtXrcc2 proteins). The A-type nucleotide binding consensus amino acid sequence [G/A]XXXXGK[S/T] (Walker A motif) is conserved in the Arabidopsis Rad51 paralogs, except for AtXrcc2, while the B-type binding consensus amino acids sequence hhhhD (Walker B motif) is conserved in the five Arabidopsis Rad51 paralogs (Figure 1b) (Higgins et al., 1985; Walker et al., 1982).

Figure 1.

Arabidopsis genome encodes homologs of the five Rad51 paralogs.
(a) The dendrogram illustrates the sequence relationships among 23 Rad51-like proteins in Saccharomyces cerevisiae (Sc), Drosophila melanogaster (Dm), Human (Hs) and Arabidopsis (At). The branch lengths are proportional to the sequence divergence. Numbers along branches are bootstrap values (1000 replicates) calculated using the phylip package. The scale represents 0.1 substitutions per site.
(b) The core-conserved sequences were aligned using the ClustalX program. The amino acid sequences of the proteins are shown in the single-letter code. Gaps are indicated by dashes. Conserved amino acids are black shaded and similar amino acids are gray shaded. The positions of the amino acids in each protein are shown at left and right. Black boxes indicate the positions of Walker motifs A and B.
Accession numbers for 23 deduced amino acids sequences used in this analysis are as follows: AtDmc1 (AAC49617), AtRad51 (CAA04529), AtRad51B (NP_180423), AtRad51Cα (BAB64343), AtRad51D (NP_172254), AtXrcc2 (NP_851268), AtXrcc3 (BAB64342), DmCG2412 (NP_610466), DmCG6318 (NP_573302), DmRad51 (BAA04580), DmSpn-B (NP_476740), DmSpn-D (AAP13056), HsDmc1(BAA10970), HsRad51 (BAA02962), HsRad51B (AAC39723), HsRad51C (AAC39604), HsRad51D (AAC39719), HsXrcc2 (CAA70065), HsXrcc3 (AAC05368), ScDmc1 (AAA34571), ScRad51 (BAA00913), ScRad55 (AAA19688), ScRad57 (AAA34950).

In order to investigate the roles of the Arabidopsis Rad51 paralogs in meiosis and the cellular responses to DNA damage, we searched for mutants in public T-DNA insertion line collections. Three lines carrying T-DNA insertions in the ATRAD51B (Salk_024755), ATRAD51C (Salk_021960) and ATXRCC2 (Salk_029106) coding sequences were found in the SIGnAL T-DNA express database (Alonso et al., 2003) and we have named the corresponding alleles atrad51b-1, atrad51c-1 and atxrcc2-1 respectively. Plants homozygous for atrad51b-1, atrad51c-1 and atxrcc2-1 T-DNA insertions were identified by PCR in the T3 seeds provided by the Nottingham Arabidopsis Stock Centre. A sterility phenotype was observed in plants homozygous for atrad51c-1, thus the atrad51c-1 T-DNA insertion was kept at the heterozygous state and atrad51c-1 mutant plants were identified by PCR in the progeny of ATRAD51C-1+/− plants.

To characterize the insertions molecularly, the T-DNA junctions were amplified and the PCR products sequenced (Figure 2). The atrad51b-1 allele T-DNA is inserted in intron 4 and is associated with a deletion of 14 bp, including the first 4 bp of exon 5 (Figure 2a). The atrad51c-1 allele T-DNA is inserted in exon 3 and is associated with a deletion of 47 bp, containing the end of exon 3 and the beginning of intron 3 (Figure 2b). The atxrcc2-1 allele T-DNA is inserted in intron 5, 3 bp after the end of exon 5, and is associated with a deletion of 1 bp (Figure 2c). These three T-DNA insertions are surrounded by two left borders in opposite orientations, designated as LB1 and LB2 (Figure 2 diagrams). In the case of atrad51b-1, sequencing showed that the T-DNA insertion is followed by an insertion of 70 bp of filler DNA (Figure 2b).

Figure 2.

Molecular characterization of Arabidopsis Rad51 paralog T-DNA insertion mutants.
The diagrams show the genomic structure of ATRAD51B (a), ATRAD51C (b) and ATXRCC2 (c) loci, with the sites of T-DNA integrations. The gray boxes represent the exons and the white boxes indicate 5′ and 3′ UTRs. The position of the T-DNA insertions is indicated by triangles. Sequences of the T-DNA junctions and RT-PCR detection of the ATRAD51B (a), ATRAD51C (b) and ATXRCC2 (c) transcripts are presented at the right of each panel. The white box indicates the T-DNA insertion. LB1 and LB2 indicate the two left borders surrounding the T-DNA insertions, and arrows indicate their respective orientation. Micro-homologies between the T-DNA ends and the genomic sequence are indicated in gray. The positions of the primers used for the RT-PCR detection of ATRAD51B (a), ATRAD51C (b) and ATXRCC2 (c) transcripts are represented on the diagrams. Amplification of the adenosin phosphoribosyl transferase (APT1) transcript has been used as a control for reverse transcription. The numbers represent positions relative to the start codon.

Plants homozygous for the T-DNA insertions were selected by PCR, and semiquantitative RT-PCR analysis performed to assess the presence of ATRAD51B, ATRAD51C and ATXRCC2 transcripts in total RNA isolated from wild type and mutant flower buds. ATRAD51B, ATRAD51C and ATXRCC2 transcripts were detected in the wild type. In contrast, the ATRAD51C mRNA was not detectable in atrad51C-1 plants, and only truncated ATRAD51B and ATXRCC2 mRNAs were detected in atrad51b-1 and atxrcc2-1 plants respectively (Figure 2). The T-DNA insertions in the atrad51b-1, atrad51c-1 and atxrcc2-1 mutants thus prevent the production of the full-length mRNAs of ATRAD51B, ATRAD51C and ATXRCC2 respectively. Furthermore, the complete absence of ATRAD51C transcript in atrad51c-1 plants shows that it is a null allele, while atrad51b-1 and atxrcc2-1 may potentially encode truncated proteins.

atrad51b-1, atrad51c-1 and atxrcc2-1 mutant plants are hypersensitive to Mitomycin C, but not to γ-irradiation

Studies performed with Chinese Hamster Ovary and DT40 (Chicken B-Lymphocyte) cell lines have shown that mutations of the different Rad51 paralogs confer moderate sensitivity to DNA DSB inducing agents such as γ-rays and hypersensitivity to DNA cross-linking agents such as MMC (Godthelp et al., 2002; Liu et al., 1998; Takata et al., 2001). Our recent work with an Arabidopsis atxrcc3 mutant showed that plants lacking the AtXrcc3 protein were slightly sensitive to bleomycin, a γ-ray mimetic agent, and much more sensitive to MMC, suggesting a conservation of the role of Arabidopsis Rad51 paralogs in response to DNA damage (Bleuyard and White, 2004).

To confirm the involvement of AtRad51B, AtRad51C and AtXrcc2 proteins in the cellular response to DNA damage, seeds from wild type, atrad51b-1 and atxrcc2-1 plants and self-fertilized heterozygous ATRAD51C-1+/− plants were either irradiated with γ-rays and sown on germination medium, or sown on plates containing germination medium and increasing doses of MMC. After 2 weeks, plants were scored for γ-irradiation or MMC sensitivity. In the absence of treatment, most plants developed at least four true leaves (in addition to the cotyledons), thus plants with three true leaves or less were considered to be sensitive to γ-rays or MMC. Previous studies on DNA repair defective mutants have shown that a dose of 100 Grays (Gy) allow discrimination between atku80, atlig4 and atatm radiosensitive mutants and wild-type plants (Friesner and Britt, 2003; Garcia et al., 2003). In our experiments, no significant difference was found between wild type and atrad51b-1, atrad51c-1 and atxrcc2-1 mutant plants, even at a dose of 200 Gy (data not shown). Similar γ-irradiation assays were performed on the seedlings of an ATXRCC3+/− plant and, as for atrad51b-1, atrad51c-1 and atxrcc2-1 mutant plants, no significant difference was found when compared with the wild type (unpublished data).

Examples of non-sensitive and sensitive plants and dose–response curves for the percentage of sensitive plants are shown in Figure 3. atrad51b-1 and atxrcc2-1 mutant plants clearly show hypersensitivity to MMC. Due to the sterility of the atrad51c-1 mutant plants, the MMC hypersensitivity of atrad51c-1 plants was assayed on the progeny of ATRAD51C-1+/− plants, one quarter of which are mutants (Figure 3c). That the sensitive plants were the atrad51c-1 mutants was verified by PCR genotyping in one experiment and all 27 sensitive plants scored were mutants. The AtRad51B, AtRad51C and AtXrcc2 proteins are thus required to repair DNA cross-links but are not essential for the repair of DSBs, presumably due to the repair of DSBs by NHEJ.

Figure 3.

Mutations in Arabidopsis Rad51 paralogs confer hypersensitivity to the DNA cross-linking agent, Mitomycin C (MMC).
(a) Comparison of non-sensitive and sensitive plants at 40 μm MMC. In the absence of MMC, all plants develop at least four true leaves (excluding the cotyledons), thus plants with three leaves or less were considered as sensitive. Scale bars = 1 cm.
(b, c) The percentage of sensitive plants was used to produce an MMC dose–response curve. Values represent three replicates, each replicate containing an average of 100 plants per dose. Error bars are ± 1 standard deviation.

atrad51c-1 mutants, but not atrad51b-1 or atxrcc2-1, are sterile

In the progeny of self-fertilized heterozygous ATRAD51C-1+/− plants (41 plants screened): 10 (24.4%) were sterile and 31 (75.6%) were fertile, corresponding well to the 3:1 segregation expected for a single Mendelian locus (chi-squared, 1 d.f. = 0.008). Genotyping confirmed that the sterile plants were exclusively homozygous atrad51c-1 mutants. atrad51c-1 plants produce atrophied siliques, which are devoid of any seed, while heterozygotes for atrad51c-1 and homozygotes for atrad51b-1 or atxrcc2-1 do not show any fertility defects and all mutant plants grew normally with normal vegetative development (data not shown).

We investigated the origin of the sterility of atrad51c-1 plants and confirmed the absence of such defects in atrad51b-1 and atxrcc2-1 mutant plants (Figure 4). Anthers were dissected from wild type and atrad51b-1, atrad51c-1 and atxrcc2-1 mutant flower buds and stained as described by Alexander (1969) to assess pollen grain viability (Figure 4a–d). While none of the observed atrad51c-1 anthers contained any viable (red-purple) pollen grains, anthers from atrad51b-1 and atxrcc2-1 mutant plants could not be differentiated from those of the wild type in terms of the number of viable pollen grains produced. To assess female gametophytic defects, we monitored post-meiotic nuclear divisions during embryo sac development. In the wild type, a single megaspore mother cell differentiates in each ovule and undergoes meiosis (Figure 4e). Meiosis in female tissues is followed by degeneration of three of the four meiotic products, to preserve a single functional megaspore (Figure 4f). The functional megaspore nucleus then undergoes three divisions to produce the eight-nuclei embryo sac, which is the mature female gametophyte (Figure 4g,h). In atrad51c-1 ovules, the megaspore mother cell could not be differentiated from the wild type (Figure 4i), but gametophytic development is blocked after meiosis. The presence of a single degenerative cell, which persists throughout embryo sac development, suggests that the megaspore mother cell is unable to properly achieve meiosis in atrad51c-1 ovules (Figure 4j–l). In some cases, one of the meiotic products is preserved, but such products were never able to proceed further than the first post-meiotic division (data not shown). The AtRad51C protein is thus required to complete both male and female gametogenesis, as is the case for AtXrcc3. In contrast, AtRad51B and AtXrcc2 are not required to achieve gametogenesis.

Figure 4.

AtRad51C, but not AtRad51B and AtXrcc2, is required for gametophytic development.
(a–d) Alexander staining was applied to anthers from wild type (a), atrad51b-1 (b), atrad51c-1 (c) and atxrcc2-1 (d) plants to discriminate viable (stained in red-purple) and dead (stained in green) pollen grains.
Wild type (e–h) and atrad51c-1 (i, j) ovules were cleared to observed embryo sac development. (e, i) megaspore mother cell; (f) functional megaspore; (g) two-nuclei stage; (h) four-nuclei stage; (j–l) degenerative cell observed at different developmental stages.

The AtRad51C protein is required to ensure chromosome stability during meiosis

Meiotic progression in wild type, atrad51b-1, atrad51c-1 and atxrcc2-1 pollen mother cells (PMCs) was examined by fluorescence microscopy after DAPI staining of chromosomes. In the wild type, the 10 Arabidopsis chromosomes condense during meiotic prophase I (Figure 5a) and can be seen as five bivalents (corresponding to paired homologous chromosomes) in metaphase I (Figure 5b). Homologous chromosomes then separate from each other and migrate to the opposite poles of the cell in anaphase I (Figure 5c). The second meiotic division starts with the alignment of chromosomes in metaphase II (Figure 5d), followed by separation of sister chromatids in anaphase II (Figure 5e). Telophase II ensues, chromosomes decondense (Figure 5f) and cytoplasm is partitioned to produce a tetrad containing four haploid microspores, which will differentiate into mature pollen grains.

Figure 5.

DAPI staining of wild-type meiotic chromosomes.
(a) Prophase I, (b) metaphase I, (c) anaphase I, (d) metaphase II, (e) anaphase II and (f) telophase II. Arrowheads indicate bivalents (b) or chromosomes (c–e). Scale bars = 10 μm. White dotted lines have been added to clearly indicate the separate sets of chromosomes.

In atrad51c-1 mutant PMCs, prophase I proceeds normally up to pachytene (Figure 6a), but in place of five expected bivalents, a varying number of entangled chromosome fragments can be seen in metaphase I figures (Figure 6b,c). This chromosome fragmentation becomes more apparent in anaphase I, with random segregation of chromosome fragments and the presence of bridges, indicating chromosome fusion events and the presence of dicentric chromosomes (Figure 6d,e). In metaphase II, most of the visible chromosome fragments are aligned on the spindle, with some fragments scattered throughout the cytoplasm (Figure 6f,g). Anaphase II separates several groups of chromosome fragments (Figure 6h,i) and is followed by chromosome decondensation in telophase II (Figure 6j). The observation of bridges in anaphase II figures suggests that fused chromosome fragments are still present at this stage and that chromosome fragmentation continues to the end of anaphase II. Meiosis in atrad51c-1 PMCs finally gives rise to ‘polyads’, containing variable numbers of products with variable DNA contents. In contrast, meiotic progression in the atrad51b-1 and atxrcc2-1 mutants is normal. These two Rad51 paralogs are thus not required to ensure chromosome stability during meiosis (data not shown). Taken together, these results indicate that, as previously shown for AtXrcc3, the AtRad51C protein is required to achieve meiosis. The chromosome fragmentation observed in atrad51c-1 meiosis presumably resulting from mis- or un-repaired meiotic double-strand breaks, in agreement with a role of AtRad51C in HJ resolution (Liu et al., 2004; Symington and Holloman, 2004).

Figure 6.

Meiosis is severely disturbed in atrad51c-1 pollen mother cells.
(a) Prophase I, (b, c) metaphase I, (d, e) anaphase I, (f, g) metaphase II, (h, i) anaphase II and (j) telophase II. Arrowheads indicate bridges. Scale bars = 10 μm.

Discussion

We report here the identification and characterization of three Arabidopsis mutants, defective for the Rad51 paralogs AtRad51B, AtRad51C and AtXrcc2 respectively. The first striking result from this study is that, in contrast to vertebrates, mutations in any one of these Arabidopsis Rad51 paralogs do not impair plant viability (Deans et al., 2000; Pittman and Schimenti, 2000; Shu et al., 1999). Studies carried out with vertebrate cell lines defective for the Rad51 paralogs have shown that these proteins are involved in DNA repair, with mutant cell lines showing relatively moderate sensitivity to DSB inducing agents such as γ-rays and high sensitivity to chemicals inducing the formation of interstrand cross-links (ICLs) (Godthelp et al., 2002; Liu et al., 1998; Takata et al., 2000, 2001). However, Drosophila xrcc3 (spn-B) and rad51c (spn-D) mutants do not present DNA repair defects and this role of the Rad51 paralogs is thus not self-evident (Abdu et al., 2003). In previous work, we have shown that cultured cells defective for the Arabidopsis XRCC3 homologue have increased sensitivity to the radiomimetic agent Bleomycin (Bleuyard and White, 2004). Here we report that mutations in either ATRAD51B, ATRAD51C or ATXRCC2 genes did not increase sensitivity of plants to γ-rays. Similar results were also obtained with ATXRCC3-defective plants (unpublished data), and we ascribe this difference to the different effects of chronic exposure of cultured cells to Bleomycin compared with the DNA breakage produced by the acute γ-irradiation. Our data thus indicate that mutations in four different Arabidopsis Rad51 paralogs confer little or no sensitivity to DSB-inducing agents. In contrast, the hypersensitivity to MMC observed in atrad51b-1, atrad51c-1 and atrad51-1 mutant plants confirms the important role of Arabidopsis Rad51 paralogs in the repair of ICLs. Taken together, our data strongly support functional conservation of the Rad51 paralogs in DNA repair between vertebrates and plants (this study; Bleuyard and White, 2004).

Meiotic function of the CX3 complex

Yeast rad51 mutants are unable to repair meiosis-specific DSBs and their ability to produce viable spore is dramatically reduced (Cao et al., 1990; Shinohara et al., 1992). Similarly, Drosophila and Caenorhabditis mutants defective for Rad51 show defects in chromosome morphology during oogenesis and thus, reduced fertility (Rinaldo et al., 2002; Staeva-Vieira et al., 2003). In yeast and Drosophila, the Rad51 paralogs share the meiotic defects observed in rad51 mutants. Yeast rad55 and rad57 mutants are unable to repair meiotic DSBs and have reduced spore viability and oogenesis and fertility are severely disturbed in Drosophila spnB (xrcc3) and spnD (rad51c) mutants (Abdu et al., 2003; Game and Mortimer, 1974; Ghabrial and Schupbach, 1999; Ghabrial et al., 1998; Schwacha and Kleckner, 1997).

Similarly, Arabidopsis AtRad51 and AtXrcc3 proteins are both required to achieve meiosis and repair meiotic DSBs (Bleuyard and White, 2004; Li et al., 2004). In contrast to atxrcc3 mutants, the absence of meiotic homologous chromosome synapsis in atrad51-1 mutants shows that AtRad51 and AtXrcc3 proteins have distinct roles in meiosis, AtRad51 acting prior to AtXrcc3. Osakabe et al. (2002) have shown that AtXrcc3 and AtRad51C can interact together and thus presumably form a heterodimer in vivo, as is the case in vertebrates. In this study, we show that atrad51c-1 mutant plants present meiotic defects similar to those observed in the atxrcc3 mutant plants (Figure 6), confirming the meiotic role of the Arabidopsis CX3 complex. Mutations in the ATSPO11-1 gene, the Arabidopsis SPO11 homolog, dramatically reduce meiotic HR and homologous chromosome synapsis (Grelon et al., 2001). Absence of Spo11 activity in the atxrcc3 mutant suppresses the chromosome fragmentation in half the meiotic cells and delays fragmentation to the second meiotic division in the other half (Bleuyard et al., 2004), raising the possibility that the meiosis II defects derive from unresolved sister chromatid HR events. In addition, Liu et al. (2004) reported that Rad51C plays a major role in HJ branch migration and resolution activities, while Xrcc3 is involved in HJ resolution. Taken together, these findings strongly suggest that the CX3 complex is involved in the Spo11 meiotic recombination pathway, presumably in the HJ resolution.

Meiotic requirement for the Rad51 paralog proteins

Neither atrad51b-1 nor atxrcc2-1 mutants present visible meiotic defects. A trivial explanation for this would be that putative truncated proteins produced from incomplete transcripts of these alleles are able to carry out the meiotic functions of the native proteins. However, the atrad51b-1 and atxrcc2-1 mutant plants are hypersensitive to DNA cross-linking agents, indicating defects in homologous recombinational repair of ICLs (Figure 3). Furthermore, the studies performed to identify interaction domains within the Rad51 paralogs have shown that any deletion in either the N-terminal or the C-terminal parts of the proteins eliminate protein–protein interactions (Dosanjh et al., 1998; Kurumizaka et al., 2003; Miller et al., 2004). This finding led the authors to suggest that even a very short deletion can severely disturb the folding of the Rad51 paralogs (Miller et al., 2004). It thus appears very unlikely that putative truncated proteins produced in either atrad51b-1 or atxrcc2-1 would be functional.

Our finding that the AtRad51B and AtXrcc2 proteins are not required to achieve meiosis shows that only the CX3 complex plays an essential role for the repair of AtSpo11-1 induced DSBs. A recent study by Liu et al. (2004) has shown that the mammalian Rad51C and Xrcc3 proteins are both involved in HJ resolution, while the other Rad51 paralogs are implicated in branch migration processes. These results strongly support the idea that the CX3 and BCDX2 (or at least the Rad51B, Rad51C and Xrcc2 proteins) complexes have distinct roles in HR mechanisms and hence in meiotic recombination.

In the absence of the AtXrcc3 protein, meiosis is severely disturbed (Bleuyard and White, 2004), indicating that the CX3 complex has an essential function during meiosis and that this function cannot be complemented by the BCDX2 complex. In addition, one might expect that mutations in the ATRAD51C gene lead to more critical defects, due to the disruption of both CX3 and BCDX2 complexes. However, atrad51c-1 and atxrcc3 mutants present very similar defects (this study; Bleuyard and White, 2004), supporting the existence of an essential role for the CX3 complex during meiosis, while the BCDX2 complex is dispensable. At this point we cannot however exclude the possibility that the BCDX2 complex plays a non-essential role in meiotic recombination processes in contrast to the essential role of the CX3 complex (resolvase activity?), absence of which leads to chromosome fragmentation in the first meiotic prophase. We note that, although very probable, we cannot be certain that the Arabidopsis AtRad51B, AtRad51C, AtRad51D and AtXrcc2 proteins form a BCDX2 complex in vivo, as this has not yet been formally tested. Our results show the absence of essential meiotic roles for the AtRad51B and AtXrcc2 proteins in Arabidopsis, but that this conclusion also applies to the BCDX2 complex must remain tentative until formal demonstration of the existence of the complex in this plant.

In vertebrates, the embryonic lethality of knockout animals has greatly complicated studies of the meiotic roles of Rad51-like proteins (Deans et al., 2000; Pittman and Schimenti, 2000; Shu et al., 1999; Tsuzuki et al., 1996). In contrast to other model organisms, Arabidopsis carries the same range of Rad51-like proteins as vertebrates and mutants defective for Rad51 or any of the Rad51 paralogs are viable (this study; Bleuyard and White, 2004; Li et al., 2004). With the recent availability of public, sequence-tagged mutant collections, Arabidopsis thus shows great promise as a model to study the meiotic functions of proteins involved in recombination.

Experimental procedures

Phylogenetic analysis

Sequence alignments were carried out using the ClustalX software package (Version 1.83, Thompson et al., 1997). Evolutionary distances were calculated using the Henifoff/Tillier PMB (Probability Matrix from Blocks, Veerassamy et al., 2003) distance method of the Protdist program (phylip package version 3.6, Felsenstein, 1989). The coefficient of variation of the γ-distribution (to incorporate rate heterogeneity) was obtained by pre-analyzing the data with the Tree-Puzzle program (Version 5.0, Strimmer and von Haeseler, 1997). The phylogenetic tree was inferred using the unweighted pair group method with arithmetic mean method in the neighbor program (phylip package version 3.6, Felsenstein, 1989). The tree was displayed using TreeView program (version 1.6.6). Consensus trees were inferred using the Consense program (phylip package version 3.6, Felsenstein, 1989) and the significance of the various phylogenetic lineages was assessed by bootstrap analyses (Hedges, 1992).

Plant material, growth conditions and mutant screening

All Arabidopsis plants used in this work were of ecotype Columbia (Col0). A. thaliana seeds were sown directly into damp compost or solid germination medium and under white light (16 h light/8 h dark) as previously described by Gallego et al. (2001).

The atrad51b-1 (Salk_024755), atrad51c-1 (Salk_021960) and atxrcc2-1 (Salk_029106) T-DNA insertion lines were found in the public T-DNA Express database established by the Salk Institute Genomic Analysis Laboratory accessible from the SIGnAL website at http://signal.salk.edu (Alonso et al., 2003).

Plants heterozygous and/or homozygous for the atrad51b-1, atrad51c-1 or atxrcc2-1 T-DNA insertion loci were identified by a PCR genotyping assay. The following primer combinations were used to amplify the different loci: the wild type ATRAD51B locus, o519 (5′-GAGTTAGTTGGTCCTCCTGG-3′) and o520 (5′-AAATTCAGCAAGCGATCTGG-3′); the atrad51b-1 mutant locus, o519 and o405 (5′-TGGTTCACGTAGTGGGCCATCG-3′); the wild type ATRAD51C locus, o527 (5′-TTTTGTGACTAAACAAAGGAGC-3′) and o528 (5′-ACCTCCACTTAAGCTAGTCAAGG-3′); the atrad51c-1 mutant locus, o527 and o405; the wild type ATXRCC2 locus, o523 (5′-TAGTCCAATGTAACTTTCGCAG-3′) and o524 (5′-GTCACGAGACAATGACAATACC-3′); the atxrcc2-1 mutant locus, o523 and o405. atrad51c-1 mutant plants identification was confirmed based on their sterility phenotype.

Sequencing of T-DNA insertion sites

The following primer combinations were used to amplify DNA flanking the T-DNA: atrad51b-1 LB1 left border, o519 and o405; atrad51b-1 LB2 left border, o520 and o405; atrad51c-1 LB1 left border, o527 and o405; atrad51c-1 LB2 left border, o528 and o405; the atxrcc2-1 LB1 left border, o523 and o405; the atxrcc2-1 LB2 left border, o524 and o405.

The PCR products were then purified on a QIAquick column (Qiagen, Courtaboeuf, France) and directly sequenced. Sequence reaction were performed using one of the primers used for amplification and the CEQ DTCS Quick Start Kit (Beckman Coulter, Fullerton, CA, USA), and analyzed on a CEQ 2000 DNA Analysis System (Beckman Coulter).

Semiquantitative RT-PCR

For semiquantitative RT-PCR, total RNAs extracted from flower buds were treated with RNase-free DNase I (Roche, Meylan, France). One microgram of DNA-free total RNA was reverse transcribed in 20 μl of reaction mixture containing 50 units of Expand Reverse Transcriptase (Roche), 1X random hexanucleotide mix (Roche), 1 mm of each deoxyribonucleotide triphosphate, and 20 units of RNasin ribonuclease inhibitor (Promega, Charbonnieres, France). PCR was performed in 25 μl reaction mixtures containing 2 μl of RT reaction mixture, 1 unit of HotStarTaq DNA polymerase (Qiagen), 2.5 mm MgCl2, 100 μm of each deoxyribonucleotide triphosphate, and 0.4 μm of gene-specific primers.

The gene-specific primers were: o548 (5′-TTTCCAGTAGCTTATGGAGG-3′) and o549 (5′-ATATGCCAACCCAACTGAGG-3′), or o546 (5′-AGTGAAGCTACTTCTCCACC-3′) and o547 (5′-CCGGAAAGCTTTCCAGTCCC-3′) for ATRAD51B; o453 (5′-CTTGATAACATTTTGGGCGG-3′) and o454 (5′-CAAGATGATTGACCAATGCG-3′), or o450 (5′-ATGATTTCATTTGGGCGGCG-3′) and o554 (5′-TAATACGCGGCAAAGACTCC-3′) for ATRAD51C; o552 (5′-GCATTGGTGCTTTTCACTGG-3′) and o553 (5′-ATTCACGAAATGGAGGTTGC-3′), or o550 (5′-GAAGCAGATGTTATCAAGGG-3′) and o551 (5′-CCATGCTCCATTTCCTAACC-3′) for ATXRCC2. The initial denaturation was performed at 95°C for 15 min, then amplification was performed for 45 cycles with a denaturation time of 30 secec at 94°C, followed by annealing for 30 sec at 58°C and extension for 45 sec at 72°C. The APT1 (adenine phosphorybosyl transferase) transcript has been used as a control for reverse transcription (Moffatt et al., 1994). The gene-specific primers were apt1 (5′-TCCCAGAATCGCTAAGATTGC-3′) and apt2 (5′-CCTTTCCCTTAA-GCTCTG-3′). The initial denaturation was performed at 95°C for 15 min, then amplification was performed for 35 cycles with a denaturation time of 30 sec at 94°C, followed by annealing for 30 sec at 52°C and extension for 45 sec at 72°C.

Mitomycin C and γ-irradiation assays

Col0, atrad51b-1, atrad51c-1 and atxrcc2-1 seeds were surface-sterilized with 7% calcium hypochlorite solution (w/v).

For MMC assays, seeds were sown on plates containing fresh solid germination medium with different concentrations of MMC (Sigma no. M-0503, Sigma, Lyon, France). The plates were then incubated for 2 weeks (23°C, 16 h light), and resistance or sensitivity was scored by the number of true leaves (excluding the cotyledons) per plant.

For γ-irradiation, surface-sterilized seeds were kept in sterile water at 4°C for approximately 24 h. Then seeds were exposed to 50, 100 or 200 Gy (9.12 Gy min−1) from a 137Cs source (CIS Bio International, Gif sur Yvette, France) and sown on plates containing fresh solid germination medium. The plates were then incubated for 2 weeks (23°C, 16 h light), and resistance or sensitivity was scored as for the MMS treatment.

Light and fluorescence microscopy

Cytological observations of Alexander-stained anthers, embryo sac development and meiotic chromosomes were conducted as previously described (Bleuyard and White, 2004). Images were captured on a Zeiss Axioplan 2 Imaging microscope with a Zeiss Axiocam HRc video camera (Zeiss, Le Pecq, France) and enhanced using Adobe Photoshop 6 software.

Acknowledgements

We thank members of BIOMOVE for their help and discussions and the Salk Institute Genomic Analysis Laboratory for providing the sequence-indexed Arabidopsis T-DNA insertion mutants. We also thank Hong Ma and Bernd Reiss for communicating their data to us before publication.

This work was partly financed by a European Union research grant (QLG2-CT-2001-01397), the Centre National de la Recherche Scientifique and the Université Blaise Pascal.

Ancillary