Role of mammalian RAD51L2 (RAD51C) in recombination and genetic stability.

The highly conserved RAD51 protein has a central role in homologous recombination. Five novel RAD51-like genes have been identified in mammalian cells, but little is known about their functions. A DNA damage-sensitive hamster cell line, irs3, was found to have a mutation in the RAD51L2 gene and an undetectable level of RAD51L2 protein. Resistance of irs3 to DNA-damaging agents was significantly increased by expression of the human RAD51L2 gene, but not by other RAD51-like genes or RAD51 itself. Consistent with a role for RAD51L2 in homologous recombination, irs3 cells show a reduction in sister chromatid exchange, an increase in isochromatid breaks, and a decrease in damage-dependent RAD51 focus formation compared with wild type cells. As recently demonstrated for human cells, we show that RAD51L2 forms part of two separate complexes of hamster RAD51-like proteins. Strikingly, neither complex of RAD51-like proteins is formed in irs3 cells. Our results demonstrate that RAD51L2 has a key role in mammalian RAD51-dependent processes, contingent on the formation of protein complexes involved in homologous recombination repair.

The use of cultured mammalian cell lines selected for sensitivity to DNA-damaging agents to identify genes and gene functions has had a major impact on our understanding of DNA repair pathways. In particular, components of the nucleotide excision repair and DNA end joining pathways were identified and their cellular responses characterized using DNA damagesensitive cell lines (1,2). Recently, the value of this approach has been reinforced by the identification of genes from the homologous recombination repair pathway, with the cloning of the XRCC2 and XRCC3 genes by their ability to complement sensitive mammalian cell lines (3,4).
The repair of DNA damage by homologous recombination is important for the maintenance of genetic stability in cells. The RAD51 protein is central to the recombination process, and this protein is highly conserved from yeast to humans (5). Using molecular recombination assays, yeast (Saccharomyces cerevisiae) and human RAD51 proteins have been shown to promote DNA strand exchange. In S. cerevisiae, two RAD51-like proteins, Rad55p and Rad57p, form a heterodimer and stimulate Rad51-mediated recombination reactions (6). Yeast mutants that lack either Rad51 or these Rad51-like recombination proteins are extremely sensitive to agents causing severe forms of damage to DNA, such as doublestrand breaks and interstrand cross-links (7). Rad55p and Rad57p also have counterparts in mammalian cells, including XRCC2, XRCC3, RAD51L1 1 (hREC2, RAD51B, R51H2), RAD51L2 (RAD51C), and RAD51L3 (R51H3, RAD51D) (8). The last three proteins were identified through data base searches using partial homologies to RAD51-like proteins (9)(10)(11)(12)(13), and have not as yet been found to be defective in DNA damage-sensitive mammalian cell lines. At present very little is known about the functions of these mammalian RAD51-like proteins, although specific protein-protein interactions have been described which suggest that they form heterodimers and larger complexes that may help recruit RAD51 to sites of DNA damage (14)(15)(16)(17)(18)(19)(20)(21)(22).
A series of DNA damage-sensitive hamster cell lines, termed the irs mutants, were previously isolated in this laboratory (23). We and others used the irs1 cell line to clone the XRCC2 gene by its complementing ability for sensitivity to the potent DNA cross-linking agent, mitomycin-C (24 -26). The irs1 line is also sensitive to other DNA-damaging agents, including ionizing radiation, ultraviolet light, and alkylating agents (23). It shows spontaneous genetic instability, with increased frequencies of mutations (27), chromosomal aberrations (28), and chromosome non-disjunction (29). It has also been shown that the repair of a site-specific double-strand break by homologous recombination is severely reduced in irs1 compared with the paternal V79 cells (30). We have recently established a functional link between XRCC2 and RAD51 by showing that the irs1 line is defective in the formation of damage-dependent RAD51 focus formation (31).
The irs3 cell line is another member of this series of damagesensitive cell lines. In cell fusions, irs3 was found to be able to complement damage sensitivity in several other radiation-sensitive mammalian cell lines, including irs1, showing that it has a unique genetic defect (32). However, the irs3 cell line has a similar damage sensitivity profile to irs1, being sensitive to x-rays (2-fold), ethyl methanesulfonate (2.5-fold) and especially MMC 2 (7-fold), and shows chromosomal instability (23,33). We have therefore considered that irs3 may have a similar functional defect to irs1, and have characterized the irs3 cell line further as well as identifying the defective gene responsible for this phenotype.

EXPERIMENTAL PROCEDURES
Gene Cloning and Vector Construction-The cloning of RAD51L1 and RAD51L3 was described previously (11). The human cDNAs were cloned into mammalian expression vectors; RAD51L1 in pEBS7 (34) and RAD51L3 in pZeoSV2(Ϫ) (Invitrogen). Human RAD51L2 (also called RAD51C (12)) was cloned from a human testis cDNA library (CLONTECH) into pZeoSV2(Ϫ), and was subsequently recloned into the bicistronic mammalian expression vector pIRESneo2 (CLON-TECH). The human RAD51 gene was cloned from the same cDNA library into pIRESneo2. The Chinese hamster RAD51L2 gene was cloned from a CHO-K1 cDNA library (Invitrogen); the central portion of the gene was cloned first using primers designed to conserved regions of the human (12) and mouse genes (mouse gene data came from sequencing a mouse IMAGE clone 3598583 found by data base search to have close homology to the human cDNA; data not shown). The remainder of the hamster gene was cloned from the same library using nested primers designed to the central portion and to the library vector, pCDNA1. The accession number of the Chinese hamster RAD51L2 sequence is AJ413202. All cloned genes were verified by sequencing.
Isolation and Chromosomal Mapping of a Genomic Clone Carrying RAD51L2-Gridded filters and clones from the human genomic PAC library RPCI1 were supplied by the UK HGMP Resource Centre. Filters were probed with full-length human RAD51L2 cDNA, and DNA from positive clones was isolated using a Qiagen large-construct kit. One PAC clone (259-O5) was mapped to human metaphase chromosomes using a biotinylated probe, detected with Texas Red avidin. Well spread metaphases were captured and karyotyped using the Genus Applied Imaging system.
Cell Culture and DNA Transfection-Parental V79 and mutant irs3 cells were cultured at 37°C as monolayers in minimal essential medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine and antibiotics. 10 g of the RAD51L1, RAD51L2, RAD51L3, or RAD51 constructs were electroporated into 10 6 irs3 cells (BioRad Gene Pulser set at 500 F, 400 V) and transfected clones were selected in the appropriate drugs. Following clone isolation, resistance to mitomycin-C (MMC) was initially tested in 24-well plates by seeding 100 -300 cells/ well in graded concentrations of MMC (31). In control experiments we did not see any reversion of the irs3 phenotype (i.e. no background clonal growth under selective conditions; data not shown). PAC 259-O5 was chosen to test complementation with genomic RAD51L2 DNA. 15 g of PAC DNA was co-transfected with 2 g of the vector pEGFP-N2 (CLONTECH) into irs3 cells as described above, followed by selection in 500 g/ml G418. Colonies of G418-resistant cells were pooled and respread in 300 g/ml G418 and 4 ϫ 10 Ϫ8 M MMC. The presence of the appropriate RAD51L gene (cDNA or PAC) was tested in drug-resistant irs3 clones by PCR using cellular DNA, and gene expression was checked by reverse transcriptase (RT-) PCR (Superscript II, Invitrogen) of cellular RNA with oligo(dT) as the first-strand primer.
Mitomycin-C and X-ray Survival-Subconfluent cells were harvested as a single-celled suspension. Appropriate numbers of cells were reseeded into 9-cm dishes for MMC treatment, which lasted for the duration of the experiment. Compared with acute treatment (23), chronic MMC treatment increased the survival difference between irs3 and V79 cells to about 20-fold. X-ray treatment (250 kV) was given to 2 ϫ 10 5 cells in suspension, followed by reseeding appropriate numbers into 9-cm dishes. Three dishes were used for each dose point.
Immunofluorescence-Cells were seeded onto coverslips and grown for 48 h to subconfluent levels before being mock-treated or exposed to 10 Gray X-rays, a dose previously shown to induce a measurable frequency of foci (31,35). After a further 5-h incubation at 37°C, the cells were washed with phosphate-buffered saline and fixed in 1% paraformaldehyde for 2 h at 4°C. Fixed cells were permeabilized for 10 min at room temperature in 0.1% Triton X-100, 0.1 ϫ SSC and then blocked for 1 h at room temperature in phosphate-buffered saline, 5% normal horse serum. Cells were incubated overnight at 4°C with 1:100 dilution primary antibody, rabbit polyclonal anti-RAD51 (Ab-1) antibody (Oncogene Research), then incubated with 1:100 dilution secondary antibody, Cy5-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch) for 1 h at room temperature in the dark. The coverslips were mounted in Vectashield anti-fade medium and the cells analyzed by confocal microscopy at 633 nm (BioRad MRC600). Images were selected at random, and were scored from coded slides by two independent scorers, recording the number of discrete strongly fluorescing nuclear foci present in each cell. At least 50 cells were scored for each data point, and data were compared statistically to assess homogeneity using the Mann-Whitney U test.
Chromosome Analysis-To measure sister chromatid exchanges (SCE), cells were incubated in the dark with bromodeoxyuridine (10 g/ml) for 2 cell cycles to differentially label sister chromatids. MMC was added for 2 h at the start of the experiment. Colcemid (0.05 g/ml) was added for 2 h prior to harvesting the cells. Metaphase preparations were made using standard techniques. Differential staining of chromatids was achieved with the fluorescence plus Giemsa (FPG, "harlequin") staining method (36). Chromatid and chromosome aberrations were scored and classified (37) from coded slides. Chromatid breaks were scored where a physical displacement of the broken fragment was seen.
Hamster RAD51L2 Gene Analysis-RNA was isolated from parental V79 and from the damage-sensitive lines irs1, irs2, and irs3 using an RNeasy Mini kit (Qiagen). This RNA was used in an RT-PCR reaction (as above). The resulting cDNA was used as a template for PCR primers designed to amplify the full-length RAD51L2 hamster gene. Individual PCR products were sliced out of gels and cloned into pCR2.1-TOPO (Invitrogen) before sequencing. In analyzing PCR products for splice variants, intron/exon boundaries were assumed to be the same as in human genomic sequence (9 exons found using data from accession numbers AC021455 and AC025521).
Immunoprecipitation Analysis-V79 or irs3 cell pellets were resuspended in lysis buffer (50 mM Tris-HCl, pH 7.5, 0.5 M NaCl, 0.5% Nonidet P-40) containing protease inhibitors, incubated for 30 min on ice, dounced 10 times and sonicated twice for 30 s. Insoluble material was removed by high-speed centrifugation. Protein complexes in the supernatant (equivalent to ϳ15 mg of extract) were pulled down for 1.5 h at 4°C using preimmune serum or pAbs raised against RAD51L2 or XRCC2 cross-linked to Aminolink beads (Pierce). Complexes were washed four times in lysis buffer and visualized by Western blotting using monoclonal antibodies.

RESULTS
The RAD51L2 Gene of irs3 has a G 3 T Mutation in Exon 6 -To establish a link between RAD51L2 and the irs3 phenotype, we sought to clone the hamster RAD51L2 gene and check its integrity in irs3. We used sequence homologies between mouse and human RAD51L2 genes to design primers for the isolation of the hamster gene (see "Experimental Procedures"). Cloning of the full-length hamster cDNA showed that its predicted protein sequence is highly conserved in comparison to the human and mouse RAD51L2 proteins. The mouse and hamster proteins show 84% identity to each other, and each shows 77% identity to the human protein with only one amino acid difference in length (Fig. 1). Reverse-transcribed RNA from the V79 parent and irs3 lines was used to make RAD51L2 cDNA. Using full-length gene primers in PCR, both cell lines showed a number of different sized cDNA products ( Fig. 2A). To check that the altered profile of cDNA products in irs3 was related to the RAD51L2 gene defect, we assayed the RAD51L2 cDNA profiles in two other radiation-sensitive mutants isolated in the same screen from mutagenized V79 cells. The irs1 line (XRCC2-deficient) and the irs2 line (shown by complementation testing to be defective in a different gene from irs1 or irs3 (32)) both showed RAD51L2 profiles that were essentially the same as that for V79 ( Fig. 2A).
The largest cDNA product in V79 was cloned and sequenced to show that it was the full-length RAD51L2 gene; other products were found to represent splice variants with different exons missing. In repeated amplifications of cDNA from irs3 the full-length gene product was either of very low abundance or not present, but there were also visible levels of some splice variants ( Fig. 2A). The largest RAD51L2 product that could be seen readily from irs3 preparations was found to lack exon 6. Exon 6 could not be cloned from irs3 using primers to cDNA sequence either side of the exon (Fig. 2B). However, we were able to obtain sufficient product for cloning by "forcing" the amplification of exon 6 in two pieces, using a combination of overlapping primers internal and external to the exon. The sequence of exon 6 from irs3 consistently showed a single nucleotide change (G 3 T) at position ϩ1 (Fig. 2C). This base substitution is consistent with the mutation spectrum of ethyl nitrosourea, which was used to derive the irs3 cell line, although G 3 T is one of the less common types of substitution found (38,39). This change alters the consensus for the 3Ј acceptor splice site (see "Discussion"), and gives an amino acid change (Val 3 Phe). While it is difficult to comment on the importance of this residue for mammalian species, since it is one of many highly conserved amino acids (Fig. 1), comparisons to putative RAD51L2 orthologues in plant species show that this valine is widely conserved (Fig. 2C, which shows comparisons to the cress Arabidopsis thaliana and rice Oryza sativa where the overall identity to the human protein is reduced to about 40%).
The effect of this mutation on RAD51L2 protein levels is seen in Fig. 2D. We found that polyclonal (not shown) or monoclonal FIG. 1. Conservation of mammalian RAD51L2 protein sequences. RAD51L2 aligned from human, mouse, and hamster (human sequence from accession number O43502 and mouse data 3 and accession number AAK58420 (61)). Note that the human sequence is shown starting from a different codon (9 amino acids downstream) than previously published (12), since neither the mouse nor hamster sequence show conservation of that putative start codon. Predicted alignments from PILEUP (GCG version 10), shaded using BOXSHADE (www.ch.embnet.org/).

FIG. 2. RAD51L2 is mutated in irs3 cells.
A, RAD51L2 cDNA profiles amplified from V79 parental and the derived radiation-sensitive lines irs1, irs2, and irs3, using primers designed to give the full-length gene product. M, molecular size marker (Invitrogen), with sizes shown to the right. B, amplification of cDNA from V79 and irs3 lines with primers placed either side of exon 6 of hamster RAD51L2 cDNA, to give full-length product (514 bp) or product missing exon 6 (447 bp). C, position of the mutation in the RAD51L2 gene of irs3 (upper diagram, arrow) and conservation of protein sequence in the region of the mutation, in RAD51L2-like proteins from mammalian and plant species (lower diagram, arrow). D, Western blots of V79 and irs3 whole cell extracts probed with human RAD51L2 mAb (lanes 2 and 3) and human RAD51 mAb (lanes 5 and 6). Lanes 1 and 4 contain purified human RAD51L2 his10 , migrating more slowly than the endogenous hamster protein, and human RAD51, respectively. antibodies raised against human RAD51L2 recognized the hamster protein (V79 extract, lane 2), whereas RAD51L2 was not seen in the irs3 extract (lane 3). The levels of RAD51 were the same in wild type V79 and the irs3 cell lines (lanes 5 and 6).
The Human RAD51L2 Gene Restores DNA Damage Resistance to irs3-To show that the phenotype of irs3 is related to the observed RAD51L2 defect, we transfected the human RAD51L2 cDNA into irs3 cells. To ensure that any responses found were specific to this gene, we also transfected separately the human RAD51L1 and RAD51L3 cDNAs into irs3. Cell clones carrying stably integrated cDNAs were selected using dominant drug-resistant markers present in the vectors, and were subsequently tested for resistance to MMC. None of the clones transfected with RAD51L1 or RAD51L3 showed an increase in MMC resistance, despite the presence of transcripts from the full-length gene as seen by RT-PCR (data not shown). However, with RAD51L2, a fraction of the selected clones showed significant MMC resistance. To verify this result under conditions where the RAD51L2 cDNA is expressed coordinately with the dominant selectable marker, we recloned RAD51L2 into an expression vector with an internal ribosome entry (IRES) site. Following transfection of this construct, every clone showing the presence of transcripts from the RAD51L2 gene product was found to be resistant to MMC (Fig. 3A).
In addition to complementation tests with the human RAD51L2 cDNA, we identified human genomic (PAC) DNA carrying the RAD51L2 gene (see "Experimental Procedures"). In Fig. 3B we show that the PAC DNA mapped to the expected chromosomal site of the gene, chromosome 17q (12) (now known to be at 17q23; see genome.ucsc.edu/). Stable PAC-transfected clones of irs3 cells gave a similar level of MMC resistance to the cDNA carrying clones (Fig. 3A). We conclude that there is a direct correlation between the presence of the RAD51L2 gene and the restoration of MMC resistance to irs3 cells.
Cells deficient in homologous recombination genes show sensitivity to a number of DNA-damaging agents, in addition to MMC. We have previously shown an ϳ2-fold increase in sensitivity to x-rays in irs3 (23). In the present series of experiments, we confirmed this radiation response and in agreement with the MMC data showed that the RAD51L2 cDNA was able to partially complement the irs3 X-ray sensitivity (Fig. 3C).
Elevated Isochromatid Breaks and Reduced Sister Chromatid Exchange in irs3-Importantly, recombination-deficient cells show spontaneous genetic instability, and irs3 is no exception to this finding. The frequency of chromosome damage in irs3 cells was substantially higher than in the V79 parent line, and we found again that the presence of RAD51L2 largely complemented this defect (Table I). It is notable that that the frequency of isochromatid breaks was especially high in irs3 (77% of total breaks; see "Discussion").
SCE are thought to result from molecular crossing over of DNA strands, especially arising during replication when recombination may rescue forks stalled at sites of damage. A small reduction in spontaneous SCE frequency was found for irs3 (Table II). However, using the same MMC concentration for both V79 and irs3 cells (4 ϫ 10 Ϫ7 M; i.e. not allowing for their survival difference) irs3 cells gave a highly significant 30% reduction in SCE frequency. If differences in the MMC sensitivity of V79 and irs3 are taken into account, using approxi- mately equitoxic MMC concentrations, much larger differences in SCE frequency were found. Again, stable clones transfected with the RAD51L2 cDNA showed a substantial increase in SCE frequencies compared with the irs3 line (Table II).
Defective Formation of RAD51-like Protein Complexes in irs3-It has recently been shown that two distinct complexes of RAD51-like proteins occur in human cells; a heterodimer of RAD51L2 with XRCC3 (16,17), and a heterotetramer of RAD51L1, RAD51L2, RAD51L3, and XRCC2 (18,20,22). To show that hamster V79 cells also contain these complexes, we immunoprecipitated the RAD51-like proteins from extracts using appropriate polyclonal antibodies. As seen in Fig. 4A, anti-RAD51L2 antibodies pulled down XRCC3 from V79, representing the heterodimer (Fig. 4B). Use of anti-XRCC2 antibodies pulled down RAD51L1, RAD51L2, and RAD51L3, but not XRCC3 (Fig. 4C), demonstrating the presence of the heterotetramer in V79 cells (Fig. 4D). In irs3 extracts, neither of these complexes was formed (Fig. 4, A and C), although RAD51L3 was pulled down by anti-XRCC2 antibodies indicating that the RAD51L3/XRCC2 heterodimer is formed (Fig. 4C).
Damage-dependent RAD51 Focus Formation Is Reduced in irs3-The RAD51L2 gene has been suggested to be involved in RAD51-dependent repair processes in mammalian cells (15), although definitive proof of this is lacking. An important indicator of RAD51 response is the formation of nuclear foci following x-ray damage (35,40,41). Measurement of RAD51 foci in irs3 cells showed that this response was significantly reduced when compared with V79 cells. In irs3 clones transfected with RAD51L2 cDNA the RAD51 focus formation was partially restored, in agreement with the other responses measured (Fig.  5, A and B). A low frequency of RAD51 foci forms in unirradiated cells, primarily in S-phase (42), suggesting that differences in cell-cycle distribution may influence the ability to detect foci in irs3 after irradiation. However, cell-cycle profiles for irradiated irs3 and V79 cells were very similar, with most cells blocked in G 2 phase at 5 h after a dose of 10 Gray (data not shown). The formation of RAD51 foci may also be affected by the levels of RAD51 protein in irs3 relative to V79, but we found that RAD51 levels were similar in irs3, in irs3 lines transfected with RAD51L2, and in V79, both before and after irradiation ( Fig. 2D and data not shown).
Finally we wished to know whether, as in both yeast (43) and chick (44) cells, the RAD51 gene itself can partially correct DNA-damage sensitivity when expressed in cell lines defective in RAD51-like genes. To check this possibility, we cloned the human RAD51 cDNA into the IRES vector and transfected it into irs3. However, clones expressing RAD51 transcript were found to survive MMC treatment no better than irs3 itself (Fig. 3A). DISCUSSION Our data show a direct correlation between the presence of the RAD51L2 cDNA and resistance to MMC in the irs3 cells, while the RAD51L1 and RAD51L3 cDNAs showed no ability to complement. In support of these results, we also found that human genomic DNA carrying the RAD51L2 gene significantly complemented the MMC sensitivity of irs3. RAD51L2 cDNA or genomic DNA did not fully complement this phenotype, suggesting that the human RAD51L2 gene is not fully functional in hamster cells. In addition to the complementation of sensitivity to MMC, we found a similar degree of complementation for x-ray sensitivity, spontaneous chromosomal damage, SCE formation, and damage-dependent RAD51 focus formation. These data reinforce the conclusion that mutation of the RAD51L2 gene causes the irs3 phenotype.
The mutation detected in exon 6 of RAD51L2 in irs3 (G 3 T at position ϩ1) is not typical of splicing defects, which commonly alter intronic consensus sites. However, the first and last base of exons do form part of the consensus splice junction sequences, where the base G is generally conserved and the base T is found in less than 10% of splice sites (45). Loss-offunction mutants arising from changes at these sites have been documented in well studied genes, such as human HPRT and TP53 (46,47). This mutation may reduce "exon definition"; exon 6 is relatively short (67 bp) for mammalian exons (average size 137 bp) and it is known that short exons have an increased chance of differential splicing (48). The loss of exon 6 ( Fig. 2) from the full-length transcript provides a possible explanation for the irs3 phenotype. While other splice variants occur in both irs3 and wild-type cells, these may not contribute to the irs3 phenotype (e.g. sequencing of different sized transcripts shows that all have loss of exon 6; data not shown). We detected no other type of RAD51L2 transcript in irs3, consistent with the hemizygous nature of many genes in established hamster lines (49). We presume that these exon-deleted transcripts are not acting as dominant negatives, since the phenotype of irs3 is recessive in hybrids with wild-type cells (32) and can be complemented by introduction of the wild type RAD51L2 gene. The exon 6 mutation in irs3 also gives rise to an amino acid change at position 271 (Val 3 Phe) which may play a part in the loss of gene function. However, the mutation has clear consequences for the level of RAD51L2 protein; none could be measured by Western blot analysis of irs3 extracts (Fig. 2D). Our RAD51-like protein interaction data support inferences initially developed from two-hybrid systems (15) suggesting that the association of RAD51L2 in complexes with
FIG. 5. Damage-dependent RAD51 focus formation is reduced in irs3. A, foci in wild type V79, irs3, and irs3 cells transfected with RAD51L2 cDNA, with and without X-ray exposure (measured at 5 h after 10 Gray). Representative fields are shown. B, quantitation of focus formation at 5 h following 10 Gray X-rays. Data shown for two independent scorers, with standard errors calculated from individual cell counts. All differences between irradiated cell lines were significant (p Ͻ 10 Ϫ6 ) except for the two clones transfected with RAD51L2 (p ϭ 0.63).
other RAD51-like proteins may be functionally important. The interaction of human RAD51L2 with both XRCC3 (16,17) and RAD51L1 (19,21) has recently been confirmed in biochemical studies. Additionally it has been shown that there are two protein complexes containing RAD51L2 in human cells: one containing RAD51L2 and XRCC3, and the other consisting of four proteins, RAD51L1, RAD51L2, RAD51L3, and XRCC2 (18, 20 -22). In the present study we have shown that these two RAD51-like protein complexes exist similarly in hamster cells, generalizing previous data. Strikingly, neither of these complexes is formed in irs3; the only complex remaining is the RAD51L3-XRCC2 heterodimer (14). Even this heterodimer may be less stable in the absence of RAD51L2, since slightly lower levels of these two proteins were consistently found in irs3 (Fig. 4C). This is the first demonstration that the native proteins require RAD51L2 for complex formation.
The RAD51-like protein complexes bind preferentially to single-stranded DNA, suggesting that they may assist in the recruitment of RAD51 to sites of damaged DNA. For example, RAD51 filament formation occurs close to the RAD51L1-L2-L3-XRCC2 complex when it is bound to gapped duplex DNA (18). In chronic myeloid leukemia cells, lower levels of the RAD51L2-XRCC3 complex and higher levels of RAD51L1, RAD51L2, and XRCC2 (as well as RAD51) are expressed compared with normal cells (50). Furthermore, in brain tissue, where the key recombination protein RAD51 is in low abundance, it is suggested that the RAD51L2-XRCC3 complex carries out some of the functions of RAD51 (16). These studies suggest that the RAD51L2-XRCC3 complex may have a different role in recombination repair than the RAD51L1-L2-L3-XRCC2 complex. It is clear from our studies that RAD51L2 is a key component of both complexes, and therefore has a major role in homologous recombination processes in mammalian cells.
In eukaryotes there are several different pathways involving homologous recombination that can lead to the repair of DNA damage, and only some of these depend on RAD51 (51). In a number of different cell types it has been shown that RAD51 protein accumulates at nuclear foci following DNA damage, but we found that this process was much reduced in irs3 cells. This finding links RAD51L2 to RAD51-dependent recombination pathways, joining a growing list of mammalian proteins, including XRCC2 (31), XRCC3 (35), BRCA1 (52), and BRCA2 (53) implicated by lack of RAD51 focus formation.
The irs3 cells have a phenotype consistent with loss of homologous recombination repair, including their profile of sensitivity to DNA-damaging agents, spontaneous genetic instability, and reduction in SCE frequency. Very recently chick DT40 cells with a disruption of a homologue of the RAD51L2 gene have also been shown to have these defects, although to different degrees (44). For example, chick RAD51L2 homologue-deficient cells show similar survival differences to irs3, compared with wild type, in response to MMC or X-rays. However, the chick mutant cells have a larger relative reduction in SCE frequency than irs3, possibly because of a greater reliance of chick lymphoblastoid cells on homologous recombination for repair (54). The validity of these comparisons will depend on whether the chick gene is an orthologue of the mammalian RAD51L2 genes (the full coding sequence of the chick gene has not been identified). Significantly, we found that the human RAD51 cDNA could not complement the defect in irs3, while human RAD51 can partially complement the RAD51L2-homologue gene knockout in chick cells (44). Other differences between the functioning of homologous recombination proteins in chick and mammalian cells have previously been observed: for example, RAD54 is required for RAD51 focus formation in mouse cells (55), but not in chick cells (56). Taken together these data emphasize the need for caution in extrapolating functional  a, b, and d) or fork regression (a, c, and e). a, stalled replicating DNA molecule with a strand break (thick and thin lines with large arrows: newly synthesized leading and lagging strands, respectively). b and d, strand exchange to form a recombination intermediate, followed by repair synthesis (small arrows). c and e, regression so that damage now behind fork, while newly synthesized strands can anneal to form a 4-way (Holliday) junction. At d and e the junctions may be resolved to give no crossover (cut at open arrowheads) or crossover (cut at closed arrowheads); however, the intermediate is very vulnerable, having 3 strands with "free" ends, and the fourth strand is broken by cutting (junction resolution) at the open arrowheads. f, RAD51L2 present, SCE can occur in a proportion of chromosomes if correct repair of damage is followed by crossing over; g, RAD51L2 absent, attempted repair in d or lack of repair (e) of the damage can lead to free ends in all strands and separation to give isochromatid breaks (ISO). Note that there would not necessarily be a direct numerical relationship between SCE and ISO formation. data for homologous recombination from one species to another.
A further feature of the present data that supports the proposed defect in homologous recombination in irs3 is the reduction in SCE frequency together with an increase in chromosome aberrations (in particular, isochromatid breaks). While SCE can arise from crossing-over during homologous recombination repair, isochromatid breaks (Table I) can be viewed as one outcome of failure to properly carry out homologous recombination during replication. When a replication fork stalls due to a break on one DNA strand, the break may be resolved by RAD51-dependent recombination, including strand transfer and Holliday junction formation (Fig. 6, a, b,  and d). Alternatively the fork may regress (57,58) to allow damage repair against an intact strand, but this allows the annealing of newly synthesized strands and the formation of a Holliday junction at the fork (Fig. 6, a, c, and e). In either case the replicating DNA is in a highly vulnerable state, where incomplete resolution of the junctions could lead to disintegration of the intermediate structure (as shown in Fig.  6g). We suggest that in irs3, lack of the RAD51L2 protein increases the probability of disintegration at this stage, yielding isochromatid breaks. Such a failure would also lead to a reduction in the frequency of SCE. Support for these models is seen when a transfected human RAD51 cDNA is downregulated in chick RAD51 knockout cells: a massive increase in chromosomal damage occurs, especially isochromatid breaks (59), associated with an ϳ50% reduction in SCE frequency (60). The first model (Fig. 6, a, b, and d), however, implies that RAD51-dependent processes are still (at least partially) active in irs3. This is supported by our finding that some damage-dependent RAD51 focus formation occurs in irs3. RAD51 in these circumstances may not form stable recombination intermediates, leaving these vulnerable to disintegration. In the second model ( Fig. 6, a, c, and e) RAD51 is not required for isochromatid formation, consistent with the findings in RAD51-deficient chick cells (see above); in the absence of RAD51L2 the ability of RAD51-dependent recombination to rescue the fork is severely compromised, and isochromatid breaks may result. While the RAD51-like proteins are thought to be involved primarily in early stages of homologous recombination repair, to promote strand exchange by RAD51 (8), the first model suggests that RAD51L2 may normally have an additional role in later stages of recombination repair. This finding is not general for other RAD51-like proteins; for example, the XRCC2-deficient cell line irs1 does not show a striking increase in isochromatid breaks (31). 3 The high degree of conservation of RAD51L2 between mammalian species suggests that the gene has an important function in mammals, although this has yet to be clearly defined. We believe that the availability of a mammalian cell line lacking RAD51L2 activity will be of considerable value in identifying this function.