The eukaryotic RAD51 gene family has seven ancient paralogs conserved between plants and animals. Among these, RAD51, DMC1, RAD51C and XRCC3 are important for homologous recombination and/or DNA repair, whereas single mutants in RAD51B, RAD51D or XRCC2 show normal meiosis, and the lineages they represent diverged from each other evolutionarily later than the other four paralogs, suggesting possible functional redundancy.
The function of Arabidopsis RAD51B, RAD51D and XRCC2 genes in mitotic DNA repair and meiosis was analyzed using molecular genetic, cytological and transcriptomic approaches.
The relevant double and triple mutants displayed normal vegetative and reproductive growth. However, the triple mutant showed greater sensitivity than single or double mutants to DNA damage by bleomycin. RNA-Seq transcriptome analysis supported the idea that the triple mutant showed DNA damage similar to that caused by bleomycin. On bleomycin treatment, many genes were altered in the wild-type but not in the triple mutant, suggesting that the RAD51 paralogs have roles in the regulation of gene transcription, providing an explanation for the hypersensitive phenotype of the triple mutant to bleomycin.
Our results provide strong evidence that Arabidopsis XRCC2, RAD51B and RAD51D have complex functions in somatic DNA repair and gene regulation, arguing for further studies of these ancient genes that have been maintained in both plants and animals during their long evolutionary history.
Genome stability is important for cellular homeostasis and an organism must be able to repair DNA damage. Among a variety of DNA damage, double-strand DNA breaks (DSBs) are caused by ionizing radiation, genotoxic chemicals or errors in DNA replication (Kuzminov, 2001; Tonami et al., 2005). Failure to correctly repair DSBs can cause genome instability, mutations, cell cycle arrest and even cell death (Glazer & Glazunov, 1999; Mills et al., 2003; Dudasova et al., 2004; Sasaki et al., 2004). DSBs are known to be repaired by two major pathways: homologous recombination (HR) and non-homologous end-joining (NHEJ). The NHEJ pathway involves the rejoining of two broken DNA ends without a template of similar sequence, often resulting in deletions or insertions. By contrast, HR is a relatively accurate pathway that depends on the homologous DNA sequence, thereby retaining the correct genetic information in the repair process (West et al., 2004; Bleuyard et al., 2006). In addition to its function in somatic DNA repair, HR is also required for normal meiosis to maintain the association of homologous chromosomes and contributes to the redistribution of genetic diversity among progeny.
The genes involved in HR were first identified in budding yeast and mainly belong to the RAD52 epistasis group, including RAD50, RAD51, RAD52, RAD54, RDH54/TID1, RAD55, RAD57, RAD59, MRE11 and XRS2 (Paques & Haber, 1999; Krogh & Symington, 2004). Further identification of their homologs in animals and plants suggests that the HR repair pathway is highly conserved (Krogh & Symington, 2004; Bleuyard et al., 2006). Among them, members of the RAD51 family, including DMC1, RAD51 and five RAD51 paralogs (RAD51B, RAD51C, RAD51D, XRCC2 and XRCC3) have crucial roles in HR or DNA repair in mammals. Mutations in several of these genes lead not only to elevated sensitivity to DNA damaging agents, but also to embryonic lethality (Tsuzuki et al., 1996; Shu et al., 1999; Deans et al., 2000; Pittman & Schimenti, 2000), suggesting that they are important for DNA repair during the mitotic cell cycle.
Homologs of DMC1 and RAD51 have been studied in many eukaryotes, including fungi, invertebrate animals and plants (Bishop et al., 1992; Habu et al., 1996; Klimyuk & Jones, 1997; Couteau et al., 1999). In Arabidopsis thaliana, they function in DNA repair via HR in somatic or meiotic cells (Couteau et al., 1999; Bleuyard & White, 2004; Li et al., 2004, 2005; Abe et al., 2005; Bleuyard et al., 2005; Osakabe et al., 2005). For simplicity, unless otherwise noted, the RAD51 paralogous genes and mutants refer to those of Arabidopsis. Both rad51c and xrcc3 knockout mutants are hypersensitive to DNA damaging agents and sterile with striking meiotic chromosome fragmentation, suggesting that RAD51C and XRCC3 are involved in mitotic DNA repair by somatic and meiotic recombination (Bleuyard & White, 2004; Abe et al., 2005; Bleuyard et al., 2005; Li et al., 2005). By contrast, the rad51b, rad51d and xrcc2 mutants show normal fertility without detectable meiotic defects, but are sensitive to various DNA damaging agents (Bleuyard et al., 2005), and RAD51B and XRCC2 seem to have a role in the suppression of meiotic recombination (Ines et al., 2013), suggesting that RAD51B, RAD51D and XRCC2 are involved in somatic and meiotic HR. Furthermore, the Arabidopsis RAD51B, RAD51C and RAD51 proteins also interact in yeast two-hybrid systems, similar to their mammalian counterparts, suggesting that they have conserved functions (Osakabe et al., 2005).
In mammals, the embryonic lethality of mutations in RAD51 and its paralogs makes it difficult to analyze their function in vivo. By contrast, none of the Arabidopsis RAD51 paralogs is required for survival in individual single mutants. It has been reported that the RAD51B, RAD51D and XRCC2 homologs form the three groups that occupy the last three branches in the RAD51 family tree (Lin et al., 2006). Therefore, these genes might have overlapping/redundant functions in plant mitotic cell cycle or meiotic cells. It is also possible that RAD51B, RAD51D and XRCC2 are not required for normal meiosis, even when their functions are lost simultaneously. Nevertheless, the Arabidopsis RAD51B, RAD51D and XRCC2 proteins might form a protein complex that interacts with RAD51C in the process of DNA repair by HR. To date, the genetic relationship among RAD51B, RAD51D and XRCC2 in Arabidopsis has not been studied. The study of their genetic interactions should provide clues to the understanding of the function and relationship of these genes during their long evolutionary history.
Materials and Methods
Plant materials and growth conditions
The xrcc2 (SALK_029106) and rad51b (SALK_024755) T-DNA insertional lines have been characterized previously by Bleuyard et al. (2005). rad51d (also named ssn1) was obtained from Professor Xinnian Dong's laboratory (Durrant et al., 2007). All plants were grown in growth chambers under a 16-h light : 8-h dark photoperiod at 22°C : 18°C, unless otherwise indicated.
Characterization of the double and triple mutants
F1 double heterozygous plants were generated by crosses between xrcc2, rad51b and rad51d homozygous single mutant plants. The triple heterozygous F1 plants were generated by crossing the rad51b xrcc2 double homozygous mutant with rad51d. The F2 or F3 progeny plants were genotyped with each of the gene-specific primers for RAD51B and XRCC2, combining with the T-DNA left board primer (Supporting Information Table S1). To genotype rad51d, PCR products were digested with SphI to produce two fragments of 233 and 86 bp for the wild-type and one fragment of 319 bp for the mutant allele.
Photographs of plants were taken with a Sony digital camera DSC-707 (Tokyo, Japan). The viability of mature pollen grains was examined after staining with Alexander's solution (Alexander, 1969). Mitosis was examined using root tips of 1-wk-old seedlings, as described previously (Li et al., 2004). Male meiosis was examined using chromosome spreading with 4′,6-diamidino-2-phenylindole (DAPI) staining, as described previously (Ross et al., 1996). Both pollen and meiotic cells were photographed using a Nikon dissecting microscope (Tokyo, Japan) with an Optronics digital camera (Goleta, CA, USA).
Treatment with DNA damaging agents
The eight genotypes of Col (wild-type), rad51b, rad51d (ssn1), xrcc2, rad51b rad51d, rad51d xrcc2, rad51b xrcc2 and rad51b rad51d xrcc2 were treated with either of two types of DNA damaging agent: the cross-linking agents cisplatin (Sigma P4394), methyl methanesulfonate (MMS; Sigma M4016) and mitomycin-C (MMC; Sigma M4287); the DSB-inducing agent bleomycin (Sigma B5507). Seeds were surface sterilized with 10% NaClO for 5 min and 75% ethanol for 5 min, and then sown on Murashige and Skoog (MS) plates containing different concentrations of MMC, bleomycin, cisplatin or MMS, as indicated in the text. The plates were placed at 4°C for 3 d, and then transferred to a growth chamber. The resistance or sensitivity was estimated by the average fresh weight of four plants after growth for 3 wk.
The comet assay
Fourteen-day-old plants grown on half-strength MS plates under normal conditions were incubated in 2 μg l−1 of bleomycin for 6 h and then harvested in liquid nitrogen. Comet assay for DNA damage was performed according to a previously described protocol (Menke et al., 2001; Zhu et al., 2006) with minor modifications. Comet slides were prepared and subjected to electrophoresis on ice for 2 min (2 V cm−1, 11 mA). Images of comets were captured under a Zeiss Axio Imager A2 fluorescence microscope with a high-resolution microscopy camera AxioCam MRc Rev. 3 FireWire (D). The comet data analysis was performed using CometScore software (http://autocomet.com). The DNA fragments in comet tails (% tail-DNA) were used to estimate the extent of DNA damage.
RNA extraction and real-time PCR
Three-week-old young plants grown without or with DNA damaging agent at a concentration of 23.5 μg ml−1 (in the same manner as for the phenotypic analyses) were collected and quickly frozen in liquid nitrogen. Total RNA was extracted using an RNeasy Plant Kit (Qiagen, Valencia, CA, USA) and its concentration was estimated on an Agilent 2100 Bioanalyzer (Agilent Technologies, Waldbronn, Germany). Approximately 1 μg of total RNA was used to synthesize cDNA according to the manufacturer's instruction (Promega, Madison, WI, USA). Real-time PCR was performed as described previously (Yang et al., 2011) in triplicate for each sample, using gene-specific primers (Table S1).
RNA-seq library preparation and sequencing
The ribosomal RNA was then removed by two purification steps with a PolyATtract® mRNA Isolation System (Promega) and a Poly (A) Purist™ Kit (Ambion, Austin, TX, USA), respectively. The removal of ribosomal RNAs was confirmed on an Agilent 2100 Bioanalyzer. Approximately 0.8 μg of mRNA was fragmented by RNase III at 37°C for 10 min and ligated with adaptor Mix A for reverse transcription. The 100–200 nucleotides of the first-strand cDNA were recovered by separation in 6% TBE (Tris-borate-EDTA)-Urea Gel (Invitrogen, Carlsbad, CA, USA). The fractionated cDNAs were amplified with 11–15 cycles of PCR and then purified to yield the SOLiD Fragment Library; 600 pg of the library was used for emulsion PCR; 50-base sequence reads were obtained using a SOLiD sequencer (ABI, Foster City, CA, USA).
Estimation of expression level and differential gene expression
Reads from each sample were aligned to The Arabidopsis Information Resource (TAIR) 10 Arabidopsis reference genome (http://www.arabidopsis.org) using SOLiD™ BioScope™ Software 1.3 (https://products.appliedbiosystems.com), a SOLiD data analysis package for transcriptome sequencing and other sequencing technologies. Afterwards, the aligned reads matching the annotated genes were used to estimate gene expression levels and to identify differentially expressed genes between treatments by Cufflinks (v1.2; Trapnell et al., 2010). To reduce the false-positive rate, a threshold for differential expression was set at a P value of 0.05 or less in the Cufflinks output, with a further requirement of a minimal gene expression level of at least 1.0 FPKM (fragments per kilobase of transcripts per millions of mapped reads).
Gene ontology (GO) enrichment analysis
We used the GO terms defined by the TAIR 10 GO Slim database (ftp.arabidopsis.org:/Ontologies/Gene_Ontology) for GO enrichment analysis employing the online tools AgriGO with Fisher's exact test and false discovery rate (FDR) correction (http://bioinfo.cau.edu.cn/agriGO/analysis.php). Transcription factor (TF) family annotations were downloaded from the PlantTFDB v2.0 database, containing 2023 TFs in 58 families for Arabidopsis thaliana (http://planttfdb.cbi.edu.cn/index.php; Zhang et al., 2011). The heat map of the expressed TFs was implemented by the pheatmap (Pretty Heatmaps) function in the pheatmap package (R version, 2.15, pheatmap version, 0.6.1; R Core Team, Vienna, Austria).
Expression patterns of the three RAD51 paralogs
Previous studies have shown that RAD51B is expressed widely and at a higher level in the floral buds than in other tissues (Osakabe et al., 2005). Similarly, RAD51D is expressed widely, but at very low levels (Durrant et al., 2007). The expression of XRCC2 has not been reported. To further examine the expression of these three Arabidopsis RAD51 paralogs (referred to hereafter as the RAD51 paralogs for convenience), we searched our microarray data and male meiocyte transcriptome by RNA-Seq (Zhang et al., 2005; Yang et al., 2011). Consistent with previous reports, our microarray data showed that RAD51B and RAD51D were widely expressed with relatively low levels, as was XRCC2 (Fig. S1a). Unlike DMC1, RAD51 and RAD51C, which showed the highest expression in stage-6 anthers containing meiocytes, RAD51B and RAD51D were expressed at lower levels in stage-6 anthers than in other tissues (Fig. S1a). Their low-level expression was further confirmed by our male meiocyte transcriptome data (Yang et al., 2011; Fig. S1b).
Mutants defective in the RAD51 paralogs showed normal vegetative and reproductive growth
Previous studies have shown that the rad51b, rad51d and xrcc2 single mutants display normal vegetative and reproductive growth (Bleuyard et al., 2005; Durrant et al., 2007). To test whether these genes were redundant for normal development, we generated the relevant double and triple mutants and found that they all exhibited normal development of the vegetative and floral organs and produced seedpods with normal seed numbers (Fig. 1). Furthermore, pollen grains from all single and multiple mutants were viable and indistinguishable from those of the wild-type (Fig. 1). Because mutations in RAD51 paralogs in mouse and chicken cause severe mitotic defects that are accompanied by chromosome fragmentation (Takata et al., 2000, 2001), it is possible that there are minor defects in mitotic cell division not observed from gross examination of the vegetative and reproductive growth of the Arabidopsis mutants. Mitotic chromosomes were further examined using DAPI staining of cells from 1-wk-old root tips of the eight genotypes. The results showed that mitotic chromosome features at metaphase, anaphase and telophase were indistinguishable among the eight phenotypes: wild-type (70 cells), rad51b (54 cells), rad51d (35 cells), xrcc2 (60 cells), rad51b rad51d (27 cells), rad51b xrcc2 (52 cells), rad51d xrcc2 (59 cells) and rad51b rad51d xrcc2 (74 cells; Fig. 2). Taken together, these results suggest that these three genes combined are not required for plant vegetative and reproductive development.
The single, double or triple mutants in the RAD51 paralogs appeared normal in male meiosis
In Arabidopsis, both RAD51C and XRCC3 are required for meiosis and DNA repair (Bleuyard & White, 2004; Abe et al., 2005; Li et al., 2005). Even though the mutants in the three RAD51 paralogs showed normal pollen phenotypes, we could not rule out the possibility that minor male meiotic defects might still be present, but not sufficiently severe to affect fertility, as was the case for the mus81 mutant (Berchowitz et al., 2007). We found that male meiosis in all mutant genotypes showed normal chromosome behavior, including typical diakinesis with five bivalents (Fig. 3). In addition, similar to the wild-type, the bivalents in all mutants were well aligned at the equatorial plane at metaphase I (Fig. 3) and then segregated to form two groups of chromosomes at anaphase I. The two groups of chromosomes further underwent decondensation and recondensation, and were again aligned at two division planes at metaphase II (Fig. 3). After anaphase II and telophase II, sister chromatids were separated to form four nuclei, which were packed into four microspores at the end of male meiosis (Fig. 3). We examined a total of > 300 meiocytes for the triple mutant, but found no obvious meiotic defects. Therefore, we conclude that these three genes are not essential for male meiosis, because mutations in the three genes together did not result in detectable abnormality.
RAD51B, RAD51D and XRCC2 were partially redundant for somatic DSB repair
Mutations in different RAD51 paralogs caused sensitivity to DNA damaging agents, such as γ-irradiation, MMC, cisplatin and bleomycin, in both animal cells and Arabidopsis (Liu et al., 1998; Takata et al., 2001; Abe et al., 2005; Osakabe et al., 2005), suggesting that the functions of the RAD51 paralogs in DNA damage repair are conserved between vertebrates and plants. To investigate the relationship between RAD51B, RAD51D and XRCC2 in DNA damage repair, we first examined whether the double and triple mutant plants conferred greater sensitivity than the single mutants to MMC, a DNA cross-linking agent (Warren et al., 1998). As shown in Fig. 4(c,d), the growth of the eight genotypes was indistinguishable at a low dose of MMC (30 μg ml−1). At 60 μg ml−1or higher concentrations, the growth of all mutants was inhibited significantly compared with that of the wild-type, to similar extents among the single, double and triple mutants (Fig. 4c,d). In a parallel experiment, the single, double and triple mutants exhibited slight or mild sensitivity in a similar manner to other DNA cross-linking agents, such as MMS and cisplatin (data not shown). These results indicate that these three genes play similar, but not redundant, roles in the repair of damage caused by DNA cross-linking agents.
Previous studies have demonstrated that RAD51B, RAD51D and XRCC2 have a role in DSB repair (Bleuyard et al., 2005; Osakabe et al., 2005; Durrant et al., 2007), but their genetic relationship in DSB repair is still lacking. Thus, we evaluated the sensitivity of the single, double and triple mutants to bleomycin, which causes DSBs in DNA (Favaudon, 1982). As shown in Fig. 4(e,f), all mutants grew normally on medium supplemented with 7.05 μg ml−1 of bleomycin. At 14.1 μg ml−1, all single, double and triple mutants were more severely affected than the wild-type, with the triple mutants being slightly more sensitive than the double mutants (Fig. 4e,f). The dose of 23.5 μg ml−1 bleomycin strongly inhibited the growth of all plants, with more dramatic effects on mutant genotypes, especially for the triple mutants, with significant differences compared with all other genotypes (Fig. 4e,f). The hypersensitivity of the triple mutant to the high dose of bleomycin was different from its response to the cross-linking agents, such as MMC, suggesting that the three RAD51 paralogs have partially redundant functions or are involved in different pathways for the repair of DNA damage induced by bleomycin.
To further evaluate the levels of DNA damage in the double and triple mutants exposed to bleomycin, the comet assay experiment, which reveals damaged DNA in a tail resembling that of a comet after electrophoresis, was performed to estimate the amount of DNA damage in 2-wk-old seedlings of the eight genotypes induced by bleomycin. The results showed that, under normal growth conditions, wild-type and triple mutant plants showed no significant difference in the levels of DNA damage (Fig. 5a,b). By contrast, with 2 μg ml−1 bleomycin induction for 6 h, the accumulation of DNA in the comet tail in the xrcc2 single mutants and all double and triple mutants was significantly higher than that in rad51b, rad51d and the wild-type (Fig. 5a,b). It is especially worth noting that the triple mutants showed the highest level of DNA damage with 92.22% of the nuclear DNA in the tail, consistent with the observation that the triple mutant displayed the highest sensitivity to bleomycin. These results further support the proposal that these three genes have partially redundant roles in DNA repair.
Expression of DNA repair genes was induced in the triple mutant or by bleomycin
Previous studies have revealed that mutants, such as rad51 and rad51d, which are defective in mitotic HR or DNA repair, are also affected in gene expression (Durrant et al., 2007; Tuteja et al., 2009; Wang et al., 2010; Liu & Gong, 2011). To investigate whether the hypersensitive phenotype of the triple mutant to bleomycin is related to the expression of HR or other DNA repair genes, we examined several representative genes in 3-wk-old seedlings with or without bleomycin treatment. In mammalian cells, RAD51B, RAD51D and XRCC2 form a complex with RAD51C (the BCDX2 complex), and RAD51C with XRCC3 (the CX3 complex; Wiese et al., 2002); furthermore, some of the plant members can also interact physically in a yeast two-hybrid assay (Osakabe et al., 2005). Therefore, we first examined the expression of RAD51 paralogs, including RAD51 and RAD51C. As shown in Fig. 6, RAD51 and RAD51C expression showed no significant difference between the wild-type and triple mutants with or without bleomycin treatment (Fig. 6). By contrast, the expression level of GAMMA RESPONSE1 (GR1) was increased dramatically by either the triple mutations or bleomycin induction (Fig. 6). We then examined the expression of BRCA1, which is important for HR and DNA repair in plants (Bundock & Hooykaas, 2002; Block-Schmidt et al., 2011), and found that BRCA1 expression in the wild-type and triple mutants resembled that of RAD51 under normal conditions (Fig. 6). By contrast, BRCA1 expression was sharply increased in the triple mutant compared with the wild-type under bleomycin treatment (Fig. 6). It is possible that DNA damage induced by bleomycin was repaired less effectively in the triple mutant, as suggested by its greater sensitivity to bleomycin relative to the single or double mutants, thereby triggering the elevated expression of some of the DNA repair genes.
RAD51B, RAD51D and XRCC2 affected normal gene expression
To identify additional genes with altered expression in the triple mutant, RNA-Seq was performed using mRNA isolated from wild-type and triple mutant seedlings by the SOLiD 3 platform, as described previously (Yang et al., 2011). We obtained a total of c. 378 million single-end reads of 50 bp (Table S2). Approximately 65.9% of reads were mapped to the Arabidopsis reference genome (TAIR 10), representing c. 70.8% of the annotated genes in TAIR 10 (Table S3) and providing high-quality data to explore the transcriptome.
We first identified genes with altered expression in the triple mutant compared with the wild-type (Fig. 7a), and found that 2111 genes were differentially expressed (FPKM ≥ 1 and P ≤0.05), including 1450 up-regulated (Table S4) and 661 down-regulated (Table S5) genes, although many of these genes might not be regulated directly by the RAD51 paralogs. The GO categorization for the 1450 up-regulated genes showed that molecular functions of DNA repair, transcriptional regulator activity, DNA binding, enzyme activity and developmental regulation were over-represented (P < 10−4, Fig. 7b). Specifically, among these were over 45 genes with known functions in somatic DNA repair or meiotic recombination, such as COM1/GR1, MND1 and RPA1A, as well as RAD3, RAD54, MutS homolog 2, 4, 7 and DNA damage repair 1 (DRT101; Table S6), suggesting that the triple mutant had more DNA damage even without bleomycin treatment, although this was not obvious using the comet assay.
To investigate possible effects of the DNA damaging agent bleomycin on gene expression in the wild-type, we compared the wild-type transcriptome with or without bleomycin treatment, and found that 4311 genes were differentially expressed (FPKM ≥ 1and P ≤0.05), with 2835 genes up-regulated (Table S7) and 1476 genes down-regulated (Table S8) in bleomycin-treated seedlings. The GO categorization indicated that DNA repair (53 genes; Table S6), response to stimulus, immune response, DNA binding and several kinds of enzymes were most enriched in the up-regulated genes (Fig. 7b). In the down-regulated genes, the most enriched categories were the same as those in the up-regulated genes of the triple mutant, such as transcriptional regulation, DNA binding and enzyme activity (Table S9), but the specific genes did not overlap between these sets.
The fact that several known genes involved in DNA repair were induced in both the triple mutant and by bleomycin treatment suggested that the respective sets of differentially expressed genes might be similar when compared with the untreated wild-type. To test this possibility, we compared the two sets of differentially expressed genes, and found that the number of differentially expressed genes in the bleomycin-treated wild-type was more than double the number of genes differentially expressed in the triple mutant, suggesting that bleomycin has more severe effects than the three mutations. In addition, among the up-regulated genes in the triple mutant, 739 genes (> 50%) were also found in the up-regulated genes caused by bleomycin treatment (Fig. 7a, Table S10), suggesting that a large number of genes were induced in the triple mutant probably as a result of the accumulation of DNA damage when the repair functions were reduced. Furthermore, functional annotation of these genes showed that the main molecular functions were related to DNA repair, chromatin structure and stimulus response (Fig. 7b), including 30 genes related to DNA repair as mentioned already, such as COM/GR1, MND1, RPA1A, DRT101 and MSH7. These results support the idea that bleomycin and mutations in the three RAD51 paralogs both cause the accumulation of DNA damage. Moreover, transcriptional regulation was also most enriched in a set of 704 genes induced in the triple mutant, but not by bleomycin (Fig. 7b), whereas the categories of stimulus and immune response, enzyme activity and cell death were most enriched in the 2047 genes induced by bleomycin, but unaffected in the triple mutant (Fig. 7b), indicating that bleomycin treatment and the triple mutations also induced distinct changes in gene expression.
The triple mutant exhibited a strong transcriptomic response to bleomycin
Although the triple mutant showed altered expression relative to the wild-type for many genes, it still responded to bleomycin in gene regulation. We therefore compared the transcriptomes between bleomycin-treated and untreated triple mutant seedlings, and found 2408 and 2076 genes to be up-regulated and down-regulated, respectively, in treated triple mutant seedlings (Fig. 7c, Tables S11, S12). GO annotation showed that most enriched categories in the up-regulated genes were the same as those in the bleomycin-treated wild-type (Table S13), whereas transcriptional regulation, DNA binding and oxidoreductase activity were enriched in down-regulated genes (Table S14).
We also found that bleomycin affected very different sets of genes relative to the triple mutations, as supported by the relatively large number of genes showing opposite effects for the triple mutations and bleomycin: 270 were expressed at higher levels in the wild-type than in the triple mutant, but were repressed by bleomycin, whereas 352 showed lower levels of expression in the wild-type than in the triple mutant, but were induced by bleomycin (Fig. 7c, Tables S15, S16). GO annotation analysis revealed that TF and regulation activity, DNA binding and enzyme inhibitor activity were enriched in the 352 genes induced by bleomycin in the triple mutant (Table S17), suggesting that the up-regulatory genes are important for the response to stress by the triple mutant.
Although the expression of similar numbers of genes was affected by bleomycin in both the wild-type and the triple mutant, the fact that the triple mutant was hypersensitive to bleomycin suggested that distinct sets of genes might be affected in these two genotypes. Indeed, 1428 of the genes up-regulated in the bleomycin-treated wild-type were not induced in the triple mutant by bleomycin, and 1001 of the genes induced in the triple mutant were not up-regulated in the treated wild-type (Fig. 8a, Table S18). Likewise, 624 of the genes that were down-regulated in the bleomycin-treated wild-type were not repressed by bleomycin in the triple mutant, whereas 1224 of the genes repressed by bleomycin in the triple mutant did not decrease in the bleomycin-treated wild-type (Fig. 8a). It is possible that the genes affected by bleomycin in the wild-type, but not in the triple mutant, require the function of the RAD51 paralogs for increased expression by bleomycin. Among the 1001 genes specifically induced by bleomycin in the triple mutant are those related to defense and immune responses, apoptosis, cell death and enzyme activities (Table S19). These genes might be sensitized by the defects of the RAD51 paralogs and might be more easily induced by bleomycin in the mutant compared with the wild-type.
To further investigate whether the hypersensitivity of the triple mutant to bleomycin is related to the expression of specific genes, we compared the expression of genes with known DNA repair functions, and found that most genes examined showed dramatically lower levels of expression in the bleomycin-treated triple mutant than in the bleomycin-treated wild-type (Table S6), similar to the above real-time PCR results (Fig. 6). The low-level expression of these DNA repair genes, in addition to the triple mutations, suggested that the DNA repair capacity was low and DNA damage probably accumulated abnormally in the triple mutant when treated with bleomycin.
To identify additional differentially affected genes, we compared the transcriptomes between the triple mutant and wild-type, both treated with bleomycin, and found 562 differentially expressed genes, with 238 up-regulated and 324 down-regulated in the triple mutant compared with the wild-type (FPKM ≥ 1 and P ≤0.05; Fig. 8b). Most of these genes (428) did not overlap with those differentially expressed between the triple mutant and wild-type when both were untreated with bleomycin (Fig. 8b), suggesting an interaction between the triple mutations and bleomycin. In other words, these 428 genes required the presence of bleomycin for differential expression by the triple mutations, with 189 and 239 genes up-regulated and down-regulated, respectively (Fig. 8b, Tables S20, S21). GO analysis of the 428 genes revealed gene function categories for metabolism, signaling, stresses, catalytic activity and transcriptional regulation (Fig. S2). A group for response to stimulus included 20 crucial genes for disease response, such as PR1, PRB1 and JAZ4 (Table S22), consistent with the previous finding that RAD51 and RAD51D regulate directly defense-related genes (Durrant et al., 2007; Wang et al., 2010).
To further examine the genes affected by the triple mutations in the presence of bleomycin, we focused on 84 TF genes belonging to 10 families (21 up-regulated and 63 down-regulated), and divided into five groups by hierarchical clustering, with additional expression data from the triple mutant and wild-type without bleomycin (Fig. 8c). The expression patterns of these genes suggested that transcriptional regulation was dramatically altered in the triple mutant, even when its morphology was normal. The expression of group I genes was induced by bleomycin in the wild-type, but generally similar in the triple mutant with or without bleomycin, indicating that their induction by bleomycin was dependent on the RAD51B–RAD51D–XRCC2 functions (Fig. 8c). By contrast, the expression of group II genes was induced by bleomycin in the triple mutant, but not in the wild-type, suggesting that these triple mutations might have sensitized these genes, making them more easily induced by bleomycin (Fig. 8c). Expression of the group III genes was reduced dramatically in the wild-type by bleomycin, but the difference in the triple mutant, with or without bleomycin treatment, was not dramatic (Fig. 8c), again suggesting a role of the RAD51 paralogs in the altered expression caused by bleomycin. The group IV genes were negatively regulated by the RAD51 paralogs, but bleomycin canceled the effect of the triple mutations (Fig. 8c). The group V genes were slightly repressed by bleomycin in the wild-type, and even more repressed by bleomycin in the triple mutant, suggesting an additive effect (Fig. 8c). Moreover, TF genes of the same family also showed distinct expression patterns in the four samples. For example, among five bHLH genes examined here, the expression pattern of At5G51780 and At1G74500 was similar to that of the group I genes, At5G15160 was similar to the group III genes, whereas At3G59060 and At5G56960 were similar to the group V genes (Fig. 8d). Similar phenomena were also found for MYB and ERF genes (Fig. 8e,f). The variety of expression patterns of the TF genes caused by the triple mutation or bleomycin treatment suggest that plants use many different regulators to achieve a comprehensive response to DNA damage from different causes.
The role of RAD51B, RAD51D and XRCC2 in somatic DNA repair
Similar to vertebrates, the Arabidopsis genome also contains seven RAD51 homologs, which can be divided into two ancient groups, the RADα and RADβ subfamilies (Lin et al., 2006). The RADα subfamily includes both RAD51 and DMC1, whereas the RADβ subfamily includes RAD51B, RAD51C, RAD51D, XRCC2 and XRCC3, which are also known as the five Arabidopsis RAD51 paralogs. Among the seven genes, RAD51, DMC1, RAD51C and XRCC3 have non-redundant roles in meiotic HR and are required for normal fertility (Li & Ma, 2006). Except for the meiosis-specific DMC1, the other three genes also function in somatic DNA repair (Bray & West, 2005). By contrast, single mutants defective in any of the RAD51B, RAD51D and XRCC2 genes exhibit increased sensitivity to DNA damaging agents, suggesting that they have a role in HR and/or DNA repair, but they show normal vegetative growth and fertility. The normal morphological phenotypes of these mutants are in dramatic contrast with the cellular phenotype and embryo lethality caused by the mutations in their corresponding homologs in humans and mice (Silva et al., 2010). In addition, yeast two-hybrid and immunoprecipitation studies have shown that animal XRCC2, RAD51B and RAD51D form a complex with RAD51C (the BCDX2 complex) and function as a complex in homologous recombinational DNA repair (Dosanjh et al., 1998; Liu et al., 1998, 2002; Schild et al., 2000; Masson et al., 2001; Wiese et al., 2002). This complex may facilitate the formation of RAD51 foci important for HR (Takata et al., 2000, 2001). Indeed, the survival of cell lines carrying mutations in the RAD51 paralogs is reduced significantly in response to γ-irradiation treatment (Takata et al., 2000, 2001). Recent studies have shown that the BCDX2 and CX3 complexes act upstream and downstream, respectively, of RAD51 recruitment on DNA damage in human cells (Chun et al., 2013).
A recent study has reported that Arabidopsis single mutants in the three genes do not affect the formation of radiation-induced RAD51 foci (Ines et al., 2013). Our study has demonstrated that the triple mutant exhibits greater sensitivity to bleomycin than the single and double mutants (Fig. 4), indicating that their functions are partially redundant. For example, they might interact with proteins in parallel pathways that have overlapping functions in somatic DNA repair. This idea is supported by recent high-throughput proteomic analyses of protein complexes containing mouse RAD51C, RAD51D and XRCC2, which identified > 100 candidates for interaction with each protein (Rajesh et al., 2009). More than 60% of these proteins were involved in DNA/RNA modification or metabolism, including DNA mismatch repair protein MSH2, DNA replication-licensing factor MCM2, SFPQ and NONO. Further studies have demonstrated that SFPQ–NONO form a heteromeric complex to repair DSB by rejoining DSB ends (Rajesh et al., 2009, 2010). In addition, plant cells may require different recombination factors in different DNA repair pathways; for example, mutations in the plant MRE11 and COM1 homologs do not affect either synthesis-dependent strand annealing (SDSA) or single-strand annealing (SSA), whereas mutations in RAD51, RAD51C and XRCC3, as well as RAD54, affect SDSA but not SSA (Roth et al., 2012). The idea that the three RAD51 paralogs together are involved in multiple pathways is supported by our transcriptomic results, which show that the expression of genes coding for factors involved in both SDSA and SSA pathways is affected in the triple mutant. It is also quite striking that the Arabidopsis RAD51B, RAD51D and XRCC2 genes have partially redundant functions because they diverged at least 1 billion yr ago.
Regulation of transcriptome by RAD51B, RAD51D and XRCC2
Previous studies have shown that RAD51D and RAD51 regulate pathogen-related genes on salicylic acid induction (Durrant et al., 2007; Wang et al., 2010), suggesting that other RAD51 paralogs might also have a role in transcriptional regulation. Our transcriptomic analysis detected 2111 differentially expressed genes in the triple mutant compared with the wild-type (Fig. 7a). In addition to several pathogen-related genes, other genes important for DNA repair, abiotic stress and transcriptional regulation were also enriched, suggesting that RAD51 paralogs have additional roles in gene regulation. It is possible that some of these genes might be regulated directly by the RAD51 paralogs, but many could be affected by the accumulation of DNA damage caused by the mutations in the RAD51 paralogs. However, this regulation does not seem to be vital for plant development under normal conditions, as the triple mutant showed normal vegetative and reproductive growth (Figs 1, 2). Genetic studies support a major function of these genes in somatic DNA repair. The normal development of the mutants suggests that little DNA damage, including DSBs, occurs under normal conditions, as supported by the comet assay (Fig. 5), consistent with the similar levels of RAD51 expression between the wild-type and triple mutant (Fig. 6) and the indistinguishable patterns of radiation-induced RAD51 foci between each single mutant and the wild-type (Ines et al., 2013), as well as the normal growth of the rad51 mutant (Li et al., 2004).
Bleomycin causes DNA DSBs (Favaudon, 1982) and inhibits the growth of both animal and plant cells, probably as a result of the accumulation of unrepaired DSBs. However, the molecular basis of the bleomycin-induced growth phenotype in plants is not clear. We found that of the > 4000 differentially expressed genes in the wild-type treated by bleomycin, over 58% of genes were also found to be altered in the same direction in the triple mutant, including 30 DNA repair genes (Table S6), suggesting that bleomycin and the triple mutations have similar effects on a large number of genes, including those for DNA repair. Although this result further supports the hypothesis that the main functions of the three RAD51 paralogs are in DNA repair, the fact that both bleomycin and the triple mutations also have specific sets of differentially expressed genes suggests that each also has a distinct role in gene regulation. The existence of genes that are specifically altered in the wild-type by bleomycin suggests that bleomycin might cause more severe damage or different type (s) of damage than that found in the triple mutant. However, the presence of genes specifically induced/repressed by the triple mutations raises the possibility that some genes are regulated by the RAD51 paralogs, but not by DNA damage, perhaps in a manner similar to the regulation of pathogen-related genes by RAD51 and RAD51D. The RAD51 paralogs are thought to be ATPases that associate with RAD51 and chromatin (Li & Ma, 2006; Lin et al., 2006); it is possible that, in addition to DNA repair, they affect chromatin structure and gene expression.
Although the triple mutant was clearly hypersensitive to bleomycin and probably deficient in DNA repair, > 4000 genes were still differentially expressed in the triple mutant in response to bleomycin (Fig. 8a), consistent with the above discussion that there is a substantial difference between the effects of mutations and bleomycin treatment. Over 60% of the genes altered in the triple mutant by bleomycin overlapped with those in the wild-type treated with bleomycin, indicating that these genes responded to bleomycin independent of the functions of the RAD51 paralogs. Nevertheless, the hypersensitivity of the triple mutant to bleomycin (Fig. 8a) can be explained by the observation that nearly one-half of the bleomycin-induced genes in the wild-type were not induced in the triple mutant, including some genes crucial for the repair of damage caused by bleomycin, such as COM1/GR1, MND1, RPA1A, RAD54, DRT101, ERCC1 and MSH2/7. We noted that a number of genes encoding TFs (groups I and III, Fig. 8c) were induced/repressed by bleomycin in the wild-type but not in the triple mutant; some of these TFs might be responsible for bleomycin-induced alteration in gene expression in the wild-type, and the failure of the triple mutant to regulate the expression of these TF genes provides an explanation for the lack of differential expression of many of the bleomycin-induced genes.
The other 40% of differentially regulated genes caused by bleomycin only in the mutant but not in the wild-type could be the result of regulatory pathways affected by the RAD51 paralogs. In particular, a number of TF genes (groups II and IV, Fig. 8c) were induced/repressed by bleomycin in the mutant but not in the wild-type; these could provide the needed regulatory function for mutant-specific changes in gene expression caused by bleomycin. Some of the genes might be important for DNA repair, and low-level accumulation of DNA damage in the triple mutant might have sensitized the plant cells for response to bleomycin. Another perspective is that the combination of the triple mutations and bleomycin might have a much greater effect (possible synergism) than either alone. In any case, the comparison of transcriptomes in both the triple mutant and wild-type, with or without bleomycin treatment, has revealed that there is an interaction between the triple mutations and bleomycin, suggestive of a complex function for the three RAD51 paralogs that was previously unappreciated. Our results support multiple hypotheses and highlight the importance of further studies with regard to the functions of these ancient genes that have been maintained in both animals and plants over the long history of eukaryotic evolution.
We thank X. N. Dong (Duke University, Durham, NC, USA) for providing the rad51d (ssn1) mutant, A. W. Dong's laboratory (Fudan University, Shanghai, China) for comet assay assistance and the Ohio State University Arabidopsis Stock Center for providing the SALK lines. This work was supported by grants from the Ministry of Sciences and Technology of China (2011CB944603), the National Natural Science Foundation of China (91131007), Rijk Zwaan, Fudan University (to H.M.) and Zhuoxue Plan of Fudan University, and the Shanghai Committee of Science and Technology Fund for 2013 Qimingxing Project (13QA1400200) (to Y.W.).