Functional Validation of Rare Human Genetic Variants Involved in Homologous Recombination Using Saccharomyces cerevisiae

Systems for the repair of DNA double-strand breaks (DSBs) are necessary to maintain genome integrity and normal functionality of cells in all organisms. Homologous recombination (HR) plays an important role in repairing accidental and programmed DSBs in mitotic and meiotic cells, respectively. Failure to repair these DSBs causes genome instability and can induce tumorigenesis. Rad51 and Rad52 are two key proteins in homologous pairing and strand exchange during DSB-induced HR; both are highly conserved in eukaryotes. In this study, we analyzed pathogenic single nucleotide polymorphisms (SNPs) in human RAD51 and RAD52 using the Polymorphism Phenotyping (PolyPhen) and Sorting Intolerant from Tolerant (SIFT) algorithms and observed the effect of mutations in highly conserved domains of RAD51 and RAD52 on DNA damage repair in a Saccharomyces cerevisiae-based system. We identified a number of rad51 and rad52 alleles that exhibited severe DNA repair defects. The functionally inactive SNPs were located near ATPase active site of Rad51 and the DNA binding domain of Rad52. The rad51-F317I, rad52-R52W, and rad52-G107C mutations conferred hypersensitivity to methyl methane sulfonate (MMS)-induced DNA damage and were defective in HR-mediated DSB repair. Our study provides a new approach for detecting functional and loss-of-function genetic polymorphisms and for identifying causal variants in human DNA repair genes that contribute to the initiation or progression of cancer.


Introduction
Genomic instability due to defects with DNA repair proteins causes various inherited genetic disorders in humans and also leads a generalized cancer predisposition. DNA damage, including double-strand breaks (DSBs) and cross-linkages, occurs spontaneously during the normal cell cycle and upon exposure to ionizing radiation or mutagenic chemicals [1,2]. Defects in the DNA repair process lead to DNA damage, which in turn can cause cell cycle arrest, apoptosis, and tumorigenesis [1,2]. Homologous recombination (HR) is a crucial metabolic pathway found in all organisms, and is involved in the maintenance of somatic genome integrity in the presence of DNA DSBs, DNA inter-strand crosslinks, or stalled DNA replication forks [3]. In meiosis, a highly regulated HR process mediates the exchange of genetic information between the maternal and paternal chromosomes to produce non-identical haploid germ cells [4][5][6].
Rad51 and Rad52 are key proteins involved in DSB repair by HR and are conserved between humans and yeast [7][8][9]. Rad51 is a eukaryotic homolog of the bacterial RecA protein, which mediates homology searching and DNA strand exchange activities that result in joint molecule (JM) formation mediated by presynaptic filaments between homologous chromatids [9,10]. Rad52 plays a role in DSB repair (DSBR) as well as in single-strand annealing, a Rad51-independent DSBR pathway [11]. During DSBR, Rad52 stimulates HR, which facilitates the formation of Rad51-ssDNA nucleofilaments in the presence of replication protein A complex, composed of the subunits Rfa1, Rfa2, and Rfa3 [12,13]. Because Rad52 functions to anneal homologous single strand DNA in second-end capture, i.e., synthesis-dependent strand annealing and single-strand annealing [14,15], cells deficient for Rad52 exhibit a defect in HR. In turn, the importance of Rad51 in the HR pathway is highlighted by the tumor suppressor protein BRCA2, which is involved in breast and ovarian cancers, as well as other types of cancers [16,17]. The localization of human Rad51 to the DNA DSB requires the formation of a BRCA1--PALB2-BRCA2 complex [18], wherein BRCA2 interacts with Rad51 to initiate the strand-invasion step [19][20][21][22][23]. The BRC repeat domain of BRCA2 stabilizes the Rad51-ssDNA complex by inhibiting DNA-dependent Rad51 ATPase activity [24,25]. Loss of this control owing to BRCA2 or Rad51 mutations may lead to gross chromosomal rearrangements and increased susceptibility to cancers [26,27].
Many human diseases have a strong genetic component, and human genetic studies have successfully determined the causes of numerous rare Mendelian disorders [28]. Missense mutations can result in fatal or serious Mendelian disorders, or can be slightly deleterious, effectively neutral, or beneficial [29]. However, the molecular basis underlying the function of causal variants that result in the manifestation of these diseases are still largely unexplained.
Although many mutations in DNA repair genes have been reported, it is difficult to determine their clinical significance. Here, we describe functional analysis of causal variants in human Rad51 and Rad52 to determine if these missense mutations also result in loss-of-function or lethality in Saccharomyces cerevisiae. Specifically, we analyzed the JMs and recombination products of rad51 and rad52 mutants with causal variants that are associated with deficient strand exchange activity of the HR pathway. We validated the findings for deleterious mutations associated with human RAD51 and RAD52 and assessed their possible complementary roles as DNA damage repair proteins using a model system.

Results and Discussion
Identification of Deleterious rad51 and rad52 Alleles in Humans and Yeast We selected 12 single nucleotide polymorphisms (SNPs) within RAD51 (8 SNPs) and RAD52 (4 SNPs) from the NCBI SNP database (S2 Table). We compared the homologs of these two genes in humans and yeast, but excluded several SNPs that were synonymous. The PolyPhen (http://genetics.bwh.harvard.edu/pph) and SIFT (http://sift.jcvi.org) programs were used to predict the effects of amino acid substitutions on protein function and structure, respectively ( Fig 1A). From these analyses, we predicted that the RAD51 SNPs, rad51-F259I and rad51-K313Q, and the RAD52 SNPs, rad52-G59R, rad52-R70W, and rad52-G125C lead to the formation of non-functional proteins that would likely be non-functional in physiological conditions ( Fig 1B). Specifically, these missense SNPs were located in the ATPase domains of Rad51 and DNA-binding domains of Rad52 that affect HR activity; these functional domains are highly conserved between humans and yeast (Fig 1C and 1D).

Similarity of Functional Domains in Human and Yeast Genes
Rad51 and Rad52 protein sequences are highly conserved between human and S. cerevisiae. We aligned the Rad51 and Rad52 sequences in various species, including Homo sapiens and S. cerevisiae, Pan troglodytes, Canislupus, Bos taurus, Rattus norvegicus, Mus musculus, Gallus gallus, Danio rerio, Kluyveromyces lactis, and Eremothecium gossypii using the ClustalW program for multiple sequence alignment of proteins (S2 Fig). We evaluated deleterious polymorphisms in human RAD51 and RAD52 and compared equivalent mutations at the analogous positions in the corresponding yeast genes (Fig 1A and 1B). Human Rad51 and yeast Rad51 exhibit functional similarity and the Rad51 ATPase domains containing Walker A and Walker B motifs were identical ( Fig 1C) [30]. For example, human rad51-F259 and K313 mutations, located in the ATPase activity region, corresponded to S. cerevisiae rad51-F317 and K371, respectively ( Fig 1C). Similarly, three RAD52 SNPs (rad52-G59R, rad52-R70W, and rad52-G125C), which were selected using the NCBI SNP database, were located at the DNA binding site in both human and S. cerevisiae Rad52 proteins ( Fig 1D). It has been reported that the DNA-binding, RPA, and Rad51 interaction regions are conserved in both the human and S. cerevisiae Rad52 domains ( Fig 1D) [31,32]. Thus, based on the human and S. cerevisiae Rad51 and Rad52 domains, we believe that the rad51-F317I and rad51-K371Q mutations affect the ATPase activity of Rad51 and that the rad52-G41R, rad52-R52W, and rad52-G107C mutations affect the DNAbinding activity of Rad52, respectively.
To determine whether the variants affected the structures of the functional domains, the WT and variant Rad51 and Rad52 proteins were analyzed with SWISS-MODEL (Fig 2A). Non-synonymous SNPs did not dramatically change the overall structure of the protein, but the RMSD changed for the regions of the amino acid replacements [33]. The impact of SNPs on the function of variant proteins can be predicted by measuring local RMSD, where the area under the curve (AUC) predicts functional variation ( Fig 2B). The AUC of human rad51-K313Q was lower than that of human rad51-F259I. The AUC of human rad52-G59R was much lower than those of human rad52-R70W and rad52-G125C. However, functional assays for five mutant genes were not sufficient to determine the overall validity of this method for predicting functional loss due to point mutations. We are currently confirming the validity of the method using additional DNA repair genes in humans.

Identification of the DNA Damage-sensitive Alleles of RAD51 and RAD52
To analyze whether the mutagenesis of human RAD51 and RAD52 had deleterious effects in the budding yeast system, we developed an S. cerevisiae-based assay system for the analysis of SNPs identified in human genes that are mutated at the same site in budding yeast (Figs 1 and  2). It has been reported that the absence of Rad51 or Rad52 leads to defects in DNA damage repair induced by MMS [34][35][36][37][38]. MMS-sensitive variants were selected to identify missense mutations that caused defects in HR as well as in other types of DNA damage responses (Fig 3). We observed that the variants rad51 and rad52 alleles were normally localized in the nucleus in Comparison of the functional domains of Rad51 in humans and yeast [30]. * regions containing rad51-F317I and rad51-K371Q mutations. (D) Comparison of the functional domains of Rad52 in humans and yeast [31,32]. * regions containing rad52-G41R, rad52-R52W, and rad52-G107C mutations.
doi:10.1371/journal.pone.0124152.g001  Because the rad51-F317I mutation site was located in the ATPase activity domain at the C-terminus, ATPase may not be functional in the presence of this mutation (Fig 3). These results are consistent with the conclusion that the ATPase activity of Rad51 is essential for its roles in DNA repair and HR [34,35,39]. Furthermore, the rad52-R52W and rad52-G107C strains were extremely sensitive to MMS, similar to rad52Δ (Fig 3). This result implied that the rad52-R52W and rad52-G107C mutations impaired the Rad52-dependent HR pathway. Because the mutation sites were in the DNA-binding domain, these mutants might have reduced DNA-binding activity. Moreover, we created the heterozygous strains expressing reduced levels of various rad51 and rad52 mutants to analyze MMS sensitivity of the variants accurately. The heterozygous diploid cells containing RAD51/rad51-F317I, RAD52/rad52-R52W, and RAD52/rad52-G107C could function normally in the repair of MMS-induced DNA damage (S4 Fig). In particular, we found that the heterozygous strains with rad52-G41R and other deleterious rad52 alleles exhibited mild sensitivity to the same concentration of MMS (S4 Fig). Thus, we concluded that the rad51-F317I, rad52-R52W, and rad52-G107C sites have critical roles in DNA damage repair related to HR, and the complete loss-of-function mutations could not overcome DNA damage-sensitivity. In contrast to the prediction of PolyPhen and SIFT, rad51-K371Q and rad52-G41R activities related to DNA repair and damage responses were not completely suppressed. We further characterized the chromosome repair functions in physical analysis of DSB repair. Sensitivity of rad51 and rad52 alleles to DNA damage. The spots from left to right represent 10-fold serial dilutions of the yeast cultures for WT, rad51Δ, rad51-F317I, rad51-K371G, rad52Δ, rad52-G41R, rad52-R52W, and rad52-G107C cells. Sensitivity to DNA damage was induced by methyl methane sulfonate (MMS). Cells were cultured in YPD liquid for 24 h, and then spotted onto YPD plates containing MMS (0.02% or 0.03%). These plates were incubated at 30°C.

Physical Analysis of Double-Strand Break Repair
To investigate mitotic DSB repair mediated by HR, we evaluated recombination intermediates and recombinants by one-dimensional (1D) and two-dimensional (2D) gel electrophoresis of the HIS4LEU2 locus (Fig 4) [40]. Yeast cells were synchronized in the G1 phase and then transferred to YP media containing 2% galactose to induce I-SceI enzyme expression. After I-SceI induction, DSBs occurred at the I-SceI cleavage site of the maternal chromosome at the HIS4LEU2 locus (Fig 4A). I-SceI induction of the Gal1 promoter was repressed by the addition of 3% glucose after a 45-min culture. Cell samples harvested at various time points were treated with psoralen-UV crosslinking and genomic DNA was purified as described previously [40][41][42]. XhoI-restriction fragments were separated by gel electrophoresis and hybridization signals identified by Southern blot analysis. DSBs and recombinants at individual time points were identified in 1D gel analysis and the levels were quantified (Fig 4B). DSBs within 3.3 kb of the site of I-SceI cleavage in the HIS4LEU2 locus of chromosome III were detected as a single band ( Fig 4B). Furthermore, recombinants R1 (5.6 kb) and R2 (4.6 kb) were also detected in 1D gel analysis. To monitor the progress of recombination, 2D gel analysis of recombination intermediates was carried out to identify branched JMs, which are inter-strand DNA structures stabilized by the DNA crosslinking procedure (Fig 4C). JMs from native/native 2D gel analysis exhibited a distinct molecular weight and shape [40][41][42]. Because XhoI restriction polymorphisms were different for the maternal ("Mom") HIS4::LEU2 and paternal ("Dad") his4x::LEU2 loci, interhomolog JMs (IH-JMs), representing interhomolog interactions, could be distinguished from intersister JMs (IS-JMs), representing "Mom + Mom" interactions. We could identify the transient formation of IH-JMs and IS-JMs, in addition to replication Y-arc signals, in the 2D-gel analysis (Fig 4C).
Recombinants Did Not Occur in Cells with rad51-F317I, rad52-R52W, and rad52-G107C Alleles In the mitotic cell cycle of diploid cells, IH-JMs are formed and then resolved into IH-recombinants [40]. To investigate the function of rad51 and rad52 alleles in DSB repair, we prepared a galactose inducible I-SceI system as described above. In the WT strain, DSBs were first detected 30 min after the addition of galactose to the YP medium; they peaked at 1h, and were processed efficiently to produce recombinants (Fig 5A and 5B). The timing of DSBs occurrence in rad51-F317I, rad51-K371Q, rad52-G41R, rad52-R52W, and rad52-G107C mutants were similar to those in the WT strain (Figs 5 and S5); DSBs were detected at 15 min and then gradually decreased, but small quantities of DSBs remained up to 5 h. Thus, all strains produced proper DSBs and efficiently processed DSB repair. However, the DSB life spans differed slightly among strains owing to factors such as among-culture differences in cell synchronization, Southern hybridization, and quantification of signals (Figs 5 and S5). In the WT strain, recombinants also appeared to form, and plateaued at approximately 7.5% (Fig 5B and 5D). As shown in WT, recombinants were observed in rad51-K371Q and rad52-G41R, suggesting that these variants process DSB repair via the HR pathway in a normal manner. However, recombinants were not detected in the rad51-F317I, rad52-R52W, and rad52-G107C mutant strains, as observed in rad51Δ and rad52Δ mutant cells. Consistent with the results from MMS analysis of the mutant strains, rad51-F317I, rad52-R52W, and rad52-G107C cells exhibited inefficient DSB repair-mediated recombination (Figs 3 and 5).

JM Formation Was Abolished in Deleterious rad51 and rad52 Mutations
The structure of mitotic recombination intermediates was further investigated with native/native 2D gel analysis in which the DNA fragments were separated based on molecular weight in the first dimension and shape in the second dimension. In all experiments, cellular DNA was subjected to psoralen-UV crosslinking prior to DNA isolation immediately after harvesting cells from each time point. Psoralen-UV crosslinking stabilizes recombination-associated JMs such as single-end invasions (SEIs) and double-Holliday junctions (double-HJs). After DNA replication, "Mom + Mom" IS-JM species, representing sister chromatid interactions established for the repair of DSBs, were expected to form a JM of 11.8 kb (Fig 4A and 4C). "Mom + Dad" IH-JM species, which represented recombinant intermediates, were expected to form a JM of 10.2 kb (Fig 4A and 4C). Thus, high levels of IH-JM and recombinant intermediate formation represented efficient interhomolog interactions between "Mom" and "Dad" chromosomes. The double-HJ was presupposed to be the major intermediate that has been identified in the progress of meiotic recombination [40]. Moreover, using native/denaturing 2D gel analysis, Hunter and colleagues found that double-HJs are intermediates of DSB-promoted recombination in mitotic cells and that a portion of IH-JMs could also be recombinant single-HJs [40]. Thus, the IH-JMs observed in this study could also include both double-HJs and single-HJs. DNA replication arcs in all strains appeared at a similar time, indicating that the mutant alleles did not affect the progression of DNA replication. The JM species were analyzed sequentially on Southern blots with a specific probe that bound to either the "Mom" or "Dad" homolog ( Fig 6A). WT cells formed approximately 0.07% IS-JM ("Mom + Mom") and approximately 0.03% IH-JM ("Mom + Dad") at 1.75 h after addition of galactose to the cultures (Fig 6B). In this assay system, "Dad + Dad" IS-JMs could not form because DSBs only occurred at the I-SceI site of the "Mom" chromosome. In rad51-K371Q cells, the spectrum of JM is similar to that of WT cells. Interestingly, rad52-G41R cells appear to form slightly lower JM levels than WT, which suggests that the variant was partially defective in JM formation, and additional evidence confirmed this, i.e., the MMS-sensitivity test for heterozygous strains with rad52Δ or rad52 deleterious alleles (Figs 6B and S4 and S6). However, both IH-JMs and IS-JMs were not detected in 2D gel analysis of rad51-F317I, rad52-R52W, and rad52-G107C cells (Figs 6 and S6). The rad51-F317I, rad52-R52W, and rad52-G107C mutations resulted in cells that were severely deficient in JM formation and could not form recombinant products (Figs 5 and  6). These results suggested that rad51-F317I, rad52-R52W, and rad52-G107C mutations impaired the strand exchange activity of the HR pathway and led us to conclude that the three mutations play a crucial role in the DSBR pathway. However, DSBs in the mutant strains were largely disappeared over a 5-h period. We inferred that DSBs in the strains carrying mutant alleles were most likely processed by a synthesis-dependent strand annealing (SDSA) mechanism, which yields non-crossovers or disappears by DSB-end hyperresection.

Deleterious rad51 and rad52 Alleles Affect the Efficiency of Sporulation
During meiosis, yeast cells with unsegregated DNA fail to form mature spores. We investigated whether deleterious rad51 and rad52 alleles affect the efficiency of spore formation. The rad51 and rad52 alleles were cultured to induce sporulation in SPM media for 24 h at 30°C, and spore viability was analyzed after tetrad dissection. In WT cell, 94% of tetrads underwent normal sporulation based on the levels of viable spores, however, the spores of rad51Δ, rad51-F317I, rad52-R52W, and rad52-G107C cells were not viable, in accordance with the physical DNA analysis results (Fig 7). The rad51-K371Q and rad52-G41R mutations gave rise to 88 and 94% viable spores, respectively (Fig 7). This result further suggested that the deleterious SNPs (rad51-F317I, rad52-R52W, and rad52-G107C) inactivated the DSBR pathway and prevented formation of viable spores.

Conclusion
HR is a fundamental mechanism for the repair of DNA DSBs caused by ionizing radiation, ultraviolet light, chemical agents, and replication errors. Mitotic HR supports genomic stability and suppresses tumorigenesis via the proper repair of DSBs [43]. Mitotic crossing-over is a rare but important phenomenon. In human cancer, it is thought to play a major role in enabling the expression of recessive cancer-causing mutations [44]. Previous studies have shown that Rad51 can interact with the single-stranded DNA-binding proteins RPA and Rad52, and catalyzes the ATP-dependent strand exchange between homologous sequences [45]. The Rad52 DNA-binding site is critical for the establishment of heteroduplex joints required in the formation of the D-loop by the Rad51-Rad52 complex, and this complex function may be important for homologous recombination as well [46]. Rad54 is also a key player in HR and physically interacts with Rad51 to perform the homology search and strand invasion steps in HR [47]. The mammalian homolog to yeast Rad54, Rad54B, has been identified in human cells and encodes a protein that consists of 747 amino acids and is 52% identical to yeast Rad54 [48,49]. However, in this study, we could not identify rad54 alleles as causal variants because the sequence similarity of the functional domain was low between human and yeast.
Systematic approaches using S. cerevisiae alleles to examine mammalian genes have been well established and successfully used for screening mutants affecting DNA damage repair. In this study, the approach included bioinformatic and structure-based predictions were used to identify missense mutations in human RAD51 and RAD52 genes. We identified a number of deleterious alleles of the human Rad51 and Rad52 proteins that were defective in the strand exchange activity of HR using a yeast-based system. Based on the in vivo analysis, we estimated that the accuracy of functional variation prediction by SIFT and PolyPhen was 60%, which is impressive compared to previous estimates [50,51]. Furthermore, protein structural analysis of Rad51 and Rad52 variants showed that the mutation sites of the deleterious alleles were functionally important for the mediation of DSB repair (Fig 2). Through the physical analysis of recombination intermediates, we observed that the formation of JMs and recombinants was abolished in rad51-F317I, rad52-R52W, and rad52-G107C mutants, similar to rad51 or rad52 deletion mutants. These results suggested that the ATPase activity domain of Rad51 and the DNA-binding domain of Rad52 are crucial for DNA repair and that mutations in these regions might be associated with various cancers.
Model systems for the analysis of DNA damage repair have the potential to be used for identifying and understanding novel genes and pathways relevant to human cancer. Further studies in human patients with various cancer types may support these new insights into potential disease mechanisms and therapeutic strategies. We anticipate that the deleterious mutations of human HR genes verified herein could contribute to genomic instability and be related to cancer initiation and progression.

Single Nucleotide Polymorphism Selection
Single nucleotide polymorphisms (SNPs) in RAD51 and RAD52 were selected from the National Center for Biotechnology Information Single Nucleotide Polymorphism (NCBI SNP) database using the following methods. (i) Human RAD51 and RAD52 gene sequences were compared to corresponding sequences in other species, including yeast. (ii) SNPs in the functional domain were selected. (iii) The selected human and yeast SNPs were matched by position. (iv) Finally, the impact of the amino acid substitutions was predicted using Sorting Intolerant From Tolerant (SIFT) and Polymorphism Phenotyping (PolyPhen) to select "damaging" and "deleterious" SNPs. The open reading frames (ORFs) of RAD51 and RAD52 were aligned by EBI ClustalW (http://www.ebi.ac.uk/Tools/msa/clustalw2/). SNPs were selected from the NCBI SNP database (http://www.ncbi.nlm.nih.gov/snp/) and confirmed to be located in exons. The RAD51 and RAD52 SNPs were processed using SIFT and PolyPhen. SNPs located in exons were examined using SIFT and PolyPhen (http://asia.ensembl.org/Homo_sapiens/ Gene/) to select deleterious and damaging SNPs.

Yeast Strains
SK1 strains were used in this study and are described in S1 Table. The HIS4LEU2 locus and I-SceI cleavage site have been described previously [40].

Mitotic Time Course
Diploid cells were cultured in YPD (1% bacto yeast extract, 2% bacto peptone, and 2% glucose) for 24 hr. To synchronize cells in the G1 phase, the YPD cultures were diluted 1:500 and grown in SPS (0.5% bacto yeast extract, 1% bacto peptone, 0.67% yeast nitrogen base, 1% potassium acetate, and 50 mM potassium biphthalate, pH 5.5) for 18 hr at 30°C. Synchronized cells were then washed once and resuspended with fresh YP including 2% galactose to induce I-SceI expression. After 45 min, 3% glucose was added to repress I-SceI expression. Cells were harvested at the time points.

DNA Physical Analysis
Genomic DNA preparation and DNA physical analysis were performed as described previously [42,[52][53][54]. Cells were cross-linked prior to spheroplasting by exposing them to 360 nm UV light in the presence of 0.1 mg/mL trioxsalen (Sigma, St. Louis, MO, USA). Cells were then spheroplasted, and the DNA was extracted as described [42]. For the physical analysis of recombination in 2-dimensional (2D) gels, 4 μg of genomic DNA was digested with XhoI and precipitated in > 99% ethanol. Digested DNA samples were allowed to separate in 0.6% agarose gel in TBE buffer (89 mM Tris-borate and 2 mM EDTA, pH 8.3) without ethidium bromide at 2 V/cm for 24 hr. Gels were stained with ethidium bromide, and the lanes were cut out and placed in a 2D gel tray. Electrophoresis in the second dimension was carried out at 5 V/cm for 6 hr. Gels were subjected to Southern hybridization analysis using Probe A (Fig 3). Quantification of DNA species was performed using a phosphoimage analyzer and the Quantity One software (Bio-Rad, Hercules, CA, USA). The cell cycle was monitored by fluorescent-activated cell-sorting using Sytox Green as described previously [55].

Immunofluorescence Analysis
Yeast cells were fixed with 3% paraformaldehyde in PBS for 20 min. Then, the cells were treated with 0.3% Triton X-100 in PBS, and rinsed with PBS. An anti-Rad51 (Cat. No. sc-33626) or anti-Rad52 (Cat. No. sc-50445) polyclonal antibody were used for immunostaing. The following day, the cells were washed five times for 5 min each time with 2% BSA in PBS, then incubated with FITC-conjugated anti-rabbit IgG (H+L) secondary antibody (Jackson ImmunoResearch, West Grove, PA, USA) at a dilution of 1:200 for 1 h at room temperature. Cells were washed three times with blocking buffer, and then rinsed with PBS at room temperature. The nuclei were incubated with DAPI and mounted using Dako fluorescent mounting medium (Dako Corporation, Carpenteria, CA, USA). Immunofluorescence was visualized on an Olympus BX53 Fluorescence Microscopy System (Olympus, Tokyo, Japan).

Prediction of Structural Changes in the Variants
Structural alterations of Rad51 and Rad52 at the protein level induced by SNPs were determined by SWISS-MODEL (http://swissmodel.expasy.org/) and visualized by PyMOL (The PyMOL Molecular Graphics System, Version 0.99 rc6 Schrödinger, LLC., New York, NY, USA). Overall structural changes induced by side chain variations could be measured by comparisons of the root-mean-square deviation (RMSD) of C-alpha. The overall RMSD reflects the average distance between all of the backbone atoms of wild type (WT) and variant proteins. The RMSDs of Calphas of consecutive 10-amino acid-long fragments (RMSD 10aa ) were measured using PyMOL.