|Home | About | Journals | Submit | Contact Us | Français|
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact firstname.lastname@example.org
Homologous recombinational repair (HRR) restores chromatid breaks arising during DNA replication and prevents chromosomal rearrangements that can occur from the misrepair of such breaks. In vertebrates, five Rad51 paralogs are identified that contribute in a nonessential but critical manner to HRR proficiency. We constructed and characterized a knockout of the paralog Rad51D in widely studied CHO cells. The rad51d mutant (clone 51D1) displays sensitivity to a diverse spectrum of induced DNA damage including γ-rays, ultraviolet (UV)-C radiation, and methyl methanesulfonate (MMS), indicating the broad relevance of HRR to genotoxicity. Spontaneous chromatid breaks/gaps and isochromatid breaks are elevated 3- to 12-fold, but the chromosome number distribution remains unchanged. Most importantly, 51D1 cells exhibit a 12-fold-increased rate of hprt mutation, as well as 4- to 10-fold increased rates of gene amplification at the dhfr and CAD loci, respectively. Xrcc3 irs1SF cells from the same parental CHO line show similarly elevated mutagenesis at these three loci. Collectively, these results confirm the a priori expectation that HRR acts in an error-free manner to repress three classes of genetic alterations (chromosomal aberrations, loss of gene function and increased gene expression), all of which are associated with carcinogenesis.
A major goal in cancer biology is to understand the mechanisms of origin of genomic alterations that promote the progressive conversion of normal cells to a fully malignant phenotype. Double-strand break (DSB) repair pathways play a pivotal role in carcinogenesis, as chromosomal rearrangements are a fundamental feature of cancer cells. In model systems, mutations in DSB repair pathways consistently manifest phenotypes of spontaneous chromosomal instability as elevated aneuploidy, chromosome breaks and exchanges and micronuclei. However, there is little information on the contributions of the DSB repair pathways to other crucial types of spontaneous genetic alterations that are integral to carcinogenesis, such as gene mutation or amplification.
The DNA repair pathways of nonhomologous end joining (NHEJ) and homologous recombinational repair (HRR) are responsible for eliminating DSBs arising from endogenous processes or DNA damage caused by exogenous agents. In mammalian cells HRR is critical for restarting broken replication forks that encounter single-strand breaks or other lesions (1,2) (www.landesbioscience.com), and for the error-free repair of DSBs occurring in chromosomal regions that have already replicated during S phase. HRR is essentially inactive during G1 phase since DSB-mediated recombination between homologous chromosomes occurs at a very low frequency (10−5 to 10−6) (3). HRR activity is responsible for the classical S-phase resistance of cells to ionizing radiation (IR) and may decline in G2 phase (4,5).
In vertebrate cells, HRR is mediated by the Rad51 strand transferase acting with other proteins that include the five Rad51 paralogs (XRCC2-3, Rad51B-C-D), as reviewed in (6,7). Although cycling cells die rapidly without Rad51 (8), the ancillary function(s) provided by the Rad51 paralogs are not essential, as mutants can grow, but with impaired viability. Mutant cell lines defective in XRCC2, XRCC3 or RAD51C were produced by random mutagenesis in Chinese hamster cells, isolated on the basis of IR sensitivity (9,10), and used to clone the complementing human cDNAs (11–14). These mutants consistently show high levels of chromosomal aberrations, extreme sensitivity to crosslinking agents, and modest sensitivity to IR. They also show defective IR-induced Rad51 focus formation and defective HRR measured by the repair of an enzymatically induced DSB in a direct repeat substrate (6). In general, the rodent Rad51 paralog mutants are phenotypically similar to BRCA2-defective cells as reviewed previously (6), but it is noteworthy that only BRCA2 is essential for cell viability (15). Although the hamster cell Rad51 paralog mutants have proved valuable in studying HRR, none is isogenic with its parental cell line because of induced mutagenesis, and they are often not fully complemented by transfected human cDNAs or genes. In chicken DT40 cells, gene targeting has produced mutants for all five paralogs (16,17). Although they share some common phenotypic traits and resemble the hamster mutants in many respects, among them they have differential sensitivity to the crosslinker cisplatin and the DSB-induced agent camptothecin (18).
Gene knockouts of Xrcc2, Rad51b and Rad51d in mice cause embryonic lethality, usually early in development (19–21). In the Xrcc2 mouse knockout study (22), early growth arrest of MEFs from 13.5-day-old mutant embryos in culture was observed. However, it was possible to obtain immortalized MEF cultures at a very low frequency. Both primary and immortalized MEFs displayed substantial gain and loss of chromosomes in addition to elevated chromosomal aberrations (22). Using the rad51d knockout mouse, MEF cell lines could be established in a Trp53-deficient background, and these cells exhibited chromosomal instability, aneuploidy and centrosome fragmentation, but no reduction in spontaneous sister-chromatid exchange (SCE) (23).
In this study we describe a new isogenic rad51d mutant of CHO cells and characterize its phenotype, with emphasis on spontaneous genetic instability. As expected, we find greatly enhanced spontaneous chromosomal breakage and exchange although, paradoxically, SCE is unchanged. Most importantly, we demonstrate increased spontaneous rates of mutagenesis in the form of gene amplification of the CAD and dhfr loci, as well as a greatly increased mutation rate at the hprt locus. We confirm these findings with another Rad51 paralog CHO mutant, xrcc3 irs1SF. These studies provide the first determination of the quantitative contribution of Rad51 paralogs in preventing these two classes of gene-specific alterations that are intrinsically relevant to carcinogenesis.
CHO AA8 cells (24) were grown in monolayer or suspension culture in αMEM supplemented with 10% fetal bovine serum, 100 µg/ml streptomycin and 100 U/ml penicillin. Cells were counted and analyzed on a Coulter® Multisizer II. The plating efficiency of AA8 and other repair-proficient cell lines was ~90%, and that of 51D1 was ~70%; the doubling times for AA8 and 51D1 were ~13 and ~16 h, respectively.
To determine the cell cycle distribution of each cell line 5 × 105 cells were treated with 10 µg/ml BrdUrd for 20 min at 37°, fixed with 70% ethanol and stained with fluorescein isothiocyanate (FITC)-conjugated anti-BrdUrd antibody (BD Biosciences) and propidium iodine to determine the proportion of cells in each phase and DNA content, respectively. Fluorescence measurements of each sample were made on a FACscan (Becton Dickinson) and the data analyzed using Cell Quest software.
Mutagen sensitivity was determined by colony formation in 10 cm dishes. When most colonies were clearly visible by eye, dishes were rinsed with phosphate-buffered saline (PBS), fixed with 95% ethanol and stained with Gram Crystal Violet (Becton Dickinson). Exposure to genotoxic agents was as follows: UV radiation, as described (24); 137Cs γ-irradiation, at 5 × 105 cells in 15 ml tubes kept on ice; methyl methanesulfonate (MMS) and mitomycin C (MMC), at 1 × 106 cells in 10 ml suspension cultures were exposed to drug at 37°C for 60 min, chilled on ice, centrifuged, resuspended in fresh medium.
Mitotic cells were collected by a shake-off procedure, obviating the need for colcemid collection, and centrifuged at 200 × g for 3 min. The cell pellet was gently broken up, resuspended in 10 ml of 37°C 75 mM KCl hypotonic buffer, and incubated for 7 min in a 37°C water bath. Two ml of fresh 3:1 methanol:acetic acid (Carnoy's) fixative was added directly to the cell suspension in hypotonic buffer and gently mixed. The suspensions were centrifuged at 200 × g for 4 min, the supernatant was removed and the cell pellet was gently broken up and fixed dropwise with 4 ml of fresh fixative. This procedure was repeated two more times, and cell suspensions were dropped on to cold, wet slides, air-dried and desiccated for 24 h at 37°C. The next day, slides were stained in a 10% Giemsa solution (Gurr), dried and mounted with CytoSeal™ 60 mounting medium (Microm International) and a coverslip. Non-polyploid metaphase chromosome spreads of good quality were examined under a 100× objective and 2× optivar using a Nikon Microphot microscope. Chromosome numbers and chromosome aberration frequencies (mainly chromatid-type) were scored. Chromatid gaps were defined as fully achromatic lesions less than the width of a chromatid arm, chromatid breaks being separated at a width greater than the chromatid arm or displaced from the main chromatid axis (25). Sister chromatid exchange measurements were performed as described previously (26) by measuring 50 cells for each cell line in each of two experiments.
Gene-targeting vectors pTnT-neo.LARA and pTnT-puro.LARA were derived from a universal backbone targeting vector designated pTnT.neo (details available on request). pTnT-neo.LARA was constructed by first inserting a 285 bp PCR fragment containing RAD51D exon 4 between XhoI and AflII of pTnT-neo. A 2.8 kb HincII/HindIII blunted-end fragment of the left arm containing the sequences of Rad51D intron 3 region was ligated into pTnT-neo at the BstEII site to create a 9.1 kb pTnT-neo.LA. The right arm containing a 2.7 kb KpnI/XmnI blunted-end fragment from regions of exons 5 and 6 and introns 4–6 was inserted into pTnT-neo.LA at the filled-in AscI site to create the 11.8 kb pTnT-neo.LARA construct used to target the first RAD51D allele. The left arm of the gene-targeting vector containing the puromycin gene was created by first ligating a 4.2 kb BamHI/PacI fragment of pTnT-puro and a 5.0 kb BamHI/PacI fragment from pTnT-neo.LA to produce a 9.2 kb pTnT-puro.LA. A 335 bp AscI/XhoI fragment containing exon 4 derived from pTnT-neo.LARA was cloned into the AscI/XhoI sites of pTnT-puro.LA, and the 2.7 kb AscI/ClaI right arm fragment was inserted into the modified pTnT-puro.LA at the AscI/ClaI restriction sites containing exon 4 to create the 12.1 kb pTnT-puro.LARA, which was used to disrupt the second RAD51D allele.
For gene targeting, 3 × 107 cells were washed and resuspended in 1 ml cold electroporation buffer [20 mM HEPES (pH 7), 137 mM NaCl, 5 mM KCl, 0.7 mM Na2HPO4, 6 mM glucose], mixed with 10 µg linearized (pTNT.neo.LARA or pTNT.puro.LARA) DNA, electroporated at 260 V/1600 µF, incubated for 5 min on ice and plated in T150 flasks for 24 h to allow for selective marker expression (see below).
Cells transfected with pTNT.neo.LARA were plated into 10 cm dishes at ~2 × 106 cell/dish in 20 ml medium containing 1.7 mg/ml G418 (Gibco Invitrogen) and incubated for 5 days at 5% CO2 and 37°C, after which the medium was replaced with fresh medium (supplemented with 10% dialyzed serum) containing 0.1 µM 2′-fluoro-2′-deoxy-1-beta-D-arabinofuranosyl-5-iodouracil (FIAU); cells were incubated an additional 5 days. Each dish contained a pool of ~150 drug-resistant colonies, which were harvested for freezing and DNA isolation (QIAamp® DNA Blood Mini Kit, Qiagen Inc.). The frequency of G418 resistant colonies averaged 2 × 10−4, and the FIAU enrichment was ~6.3-fold. A single pool of clones containing a targeting event was processed through two successive rounds of PCR screening, first in 30 sub-pools of 15 clones each, and then as single clones. Transfectant pools were screened by PCR in three steps. First, we tested for HR of the left arm of the targeting vector by PCR amplification from neo to upstream Rad51D sequence. Second, pools were tested by PCR across the right arm from neo to downstream Rad51D sequence, and finally across the entire targeted (extended) gene. The experiment yielded at least three independent clones (one electroporation yielded three positive dishes). One positive clone was then used to target the second RAD51D allele by transfecting with pTNT.puro.LARA and selecting in 10.5 µg/ml puromycin for 2 days. Selection media was changed to G418/FIAU medium at 1.7 mg/ml and 0.1 µM, respectively (10% dialyzed serum) and grown for 5 days. Selection in G418 ensured elimination of clones that retargeted the allele already targeted in the first round of targeting. Puromycin (10.5 µg/ml) was then added back to the dishes and incubated an additional 4 to 5 days to ensure all killing of cells without puromycin integration. Each dish contained a pool of ~150 drug-resistant colonies. The frequency of puro-resistant colonies averaged ~5 × 10−4, and the FIAU enrichment was ~6.5-fold. Gene-targeted cells were identified and cloned as detailed above. Three clones independently targeted for the second allele were identified.
Two positive clones were treated with Cre recombinase plasmid, pBS185. We identified single clones that underwent Cre-mediated recombination between the LoxP sites flanking the targeted 51D exon 4 and selectable marker sequences. The identification was based on sensitivity to both selectable markers, as well as MMC.
A gene-complemented clone of 51D1, called 51D1.3, was created by transfecting 51D1 cells with a bacterial artificial chromosome (BAC) containing the complete hamster RAD51D gene, followed by continuous selection in 15 nM MMC. Single colonies were picked, and Rad51D expression was verified by western blot.
Nuclear extracts were prepared from CHO cells using the NE-PER kit (Pierce Biotechnology) and separated on a 12% Bis–Tris gel (NuPage system, Invitrogen) after normalization for loading by Bradford analysis. After blotting to PVDF membrane, the blot was blocked overnight in PBS-T containing 5% milk at 4°C. Anti-Rad51D 5A8/4 (Novus Biologicals) was used at 1:300 for 2 h at room temperature in PBS-T/5% milk and washed with PBS-T before incubation with secondary antibody, anti-mouse-horseradish peroxidase (HRP) (Santa Cruz Biotechnology Inc.) in PBS-T/5%milk. Final development was achieved using a chemiluminescent HRP substrate (BioRad). Antibody against Lamin A/C (H-110) (Santa Cruz Biotechnology Inc.) was used as a loading control, as described above at 1:300 with an anti-rabbit-HRP secondary (Santa Cruz Biotechnology Inc.).
A total of 10 ml of cell suspension at 1 × 105 cell/ml was treated with 5 µM MMC or 75 µg/ml MMS for 1 h, washed once with medium, and resuspended in 10 ml fresh medium. Cells were incubated for 4 h and then centrifuged on to glass slides at 2000 r.p.m. for 5 min using a Cytospin® 4 cytocentrifuge (Thermo Shandon). Cells were fixed in 2% paraformaldehyde for 15 min, permeabilized in cold 0.2% Triton X-100 for 5 min, and blocked in 1% BSA for 1 h. The slides were incubated with anti-Rad51 antibody (clone H-92, Santa Cruz Biotechnology Inc.) at 4°C overnight (1:1000 dilution in 1% BSA), and Alexa Fluor® 488 goat anti-rabbit secondary antibody (A-11008; Molecular Probes) at room temperature for 1 h. Glass slides were mounted using Vectashield mounting medium with DAPI (H-1200; Vector Laboratories). Fluorescence images were captured on Quips PathVysion using an Axiophot II fluorescence microscope and Rad51 foci were counted visually.
Hprt mutation rate and gene (dhfr and CAD) amplification rates were determined by fluctuation analysis (27). Replica cultures (12–24 per experiment) were seeded with 100 cells and grown in suspension to 1–2 × 106 cells/replica, plated and incubated under 6S-Gua selection for hprt mutant recovery (24). To recover cells having amplified dhfr or CAD genes, selection was done in 300 nM methotrexate (28) or in 360 µM PALA [N-(phosphonacetyl)-L-aspartate] and 1 µM dipyridimole (29), respectively. Hprt, dhfr and CAD mutation rates were calculated using the Poisson P0 term (27) or the method of the mean (30).
The strategy for inactivating RAD51D was to delete exon 4 (amino acids 88–115), which contains the GKT Walker A-box for ATPase activity. This deletion also changes the reading frame and results in a highly truncated polypeptide (S88→R…114X). In order to disrupt the two alleles of RAD51D [as determined by fluorescence in situ hybridization (FISH) analysis; data not shown], alleles were targeted one at a time using two different selectable markers (Figure 1A). Targeting vectors were designed to maintain functional RAD51D alleles such that LoxP sites flank both exon 4 and the associated selectable marker. After transfection with the targeting vector containing the neo gene and selecting G418 resistant clones, pools of clones were screened by PCR analysis, using primers specific for DNA sequence in neo and genomic DNA sequence outside of the homologous arms in the vector, to determine the presence of correctly targeted cells. Positive pools were reduced to pure clones as detailed in Materials and Methods, and such a clone was then transfected with the vector containing the puro selectable marker. Transfectants were selected in puromycin and G418 to insure the initial allele was not re-targeted. Pools of clones were screened by PCR analysis in the same manner as in the first-allele targeting. Two clones, from independent pools and having both Rad51D alleles targeted, were isolated and designated 51D1Lox and 51D2Lox. Simultaneous deletion of exon 4 in both alleles was accomplished by transfection with Cre recombinase, resulting in mutant clones 51D1 and 51D2, respectively. Disruption was verified by the inability of Cre-treated cells to grow in the presence of puromycin, G418, and MMC (Figure 1B). Western analysis was performed to verify absence of the Rad51D protein in the knockout cells. Figure 1C shows the absence of the 39 kDa band in the 51D1 cells. The level of Rad51D appears higher in the AA8 parental cells than in the 51D1Lox and the 51D1.3 gene-complemented cells. RAD51D expression may be reduced in 51D1Lox cells because of the embedded neo and puro gene promoters, which may retard transcription. The 51D1.3 cells may possess only one copy of the gene, whose expression could also be influenced by its ectopic location.
Exponentially growing 51D1 cells show clear abnormalities in their cell cycle distribution (Figure 1D). There is a much higher proportion of cells in both S and G2/M phases compared with the AA8 and 51D1.3 control lines. Moreover, there is a measurable increase in the proportion of tetraploid cells (5%) in 51D1 cultures. Each of these features is qualitatively similar to what was seen in rad51d trp53 knockout mouse embryonic fibroblasts (23).
We used colony formation assays to determine survival of rad51d cells after exposure to four commonly used and distinctly different DNA-damaging agents (Figure 2). The rad51d 51D1 cells show exquisite sensitivity (~80-fold) to the interstrand crosslinking agent MMC relative to parental AA8 and 51D1Lox, and the gene-complemented clone 51D1.3 (Figure 2A). (Fold sensitivity is measured as the dose-reduction factor at 37% cell survival.) With γ-rays, both the 51D1 and 51D2 mutant clones show a ~1.5-fold sensitivity (Figure 2B), similar to that of the xrcc3 irs1SF mutant. 51D1 cells show substantial sensitivity (~5-fold) to the alkylating agent MMS (Figure 2C), as well as ~2-fold sensitivity to UV-C (Figure 2D). For each agent the mutant cells are fully complemented when they express the hamster RAD51D gene (clone 51D1.3). These results show that the impairment of HRR through loss of Rad51D compromises the ability of CHO cells to deal with very diverse types of DNA damages.
Lack of Rad51 nuclear focus formation after exposure to DNA-damaging agents is generally associated with HR deficiency in Rad51 paralog mutants of rodent and chicken cells [reviewed in (6)], including mouse rad51d knockout cells (23). As expected, 51D1 cells exposed to 8 Gy γ-rays or 5 µM MMS were grossly defective in Rad51 focus formation (Figure 3A). Whereas the control cultures treated with γ-rays or MMS showed 62–85% of cells with >5 foci per cell, 51D1 showed only 7 and 1% of cells with foci, respectively (Figure 3B). In untreated cultures the 51D1 cells showed a less severe focus defect, i.e. an average of 2 foci per cell versus 3 foci per cell in the two control cell lines (data not shown).
Chromosomal instability is a hallmark of HR-deficient cells. To determine the role of Rad51D in maintaining chromosomal integrity, we measured spontaneous chromosomal aberrations in a large population of cells. Relative to the Rad51D-proficient cell lines tested, the 51D1 cells showed a significant increase in the levels multiple types of aberrations (Table 1): chromatid breaks (5- to 12-fold), chromatid gaps (~3-fold), and isochromatid breaks (5- to 6-fold). Also seen in the 51D1 cells was a low level of chromatid exchanges, as illustrated in Figure 3C, which were not detected in the Rad51D-proficient cells. Though dicentric chromosomes were no more prevalent in the 51D1 cells than in the parental 51D1Lox cells, ringed chromosomes were only detected in the 51D1 cells. Aneuploidy (chromosome gain or loss) has been associated with CHO and MEF cells deficient in the Rad51 paralogs XRCC2 and XRCC3 (22,31). We wished to determine if the Rad51D deficiency caused a similar problem with maintenance of chromosome number. CHO AA8 cells have a modal chromosome number of 21 (32), which was maintained in both the 51DLox and 51D1 lines (Table 2). The percentage of cells having a gain or loss of one or two chromosomes from the modal value also remained constant. Furthermore, the Rad51D-complemented 51D1.3 cells maintained the same distribution of chromosome numbers. These findings suggest that aneuploidy is not associated with the rad51d mutation in CHO cells.
As a classical, cytological manifestation of crossing over between sister chromatids, SCE could be expected to be reduced in 51D1 mutant cells. The average value from two experiments was 0.30 ± 0.1 SCE per chromosome in each of the AA8, 51D1 and 51D1.3 cell lines. Thus, somewhat surprisingly, there was no detectable reduction in the rad51d cells.
An increased mutation rate at the hprt locus is a measure of genomic instability at the single-gene level. Using fluctuation analysis to measure mutation rates (27), we found that rad51d cells have a greatly increased rate (~12-fold) of hprt mutation calculated by both the P0 method and the method of the mean (27,30) (Figure 4). Importantly, the gene-corrected 51D1.3 cells and the pre-Cre-treated 51D1Lox cells had mutation rates like that of the parental AA8 cells. Hypermutability was also seen with the xrcc3 irs1SF cells, which had an even higher hprt mutation rate (~20-fold increased over AA8) than the 51D1 cells. 1SFwt8 cells, which express human XRCC3 cDNA, were partially corrected for hprt mutagenesis (Figure 4), consistent with the incomplete complementation previously measured for cell survival after γ-irradiation and MMC treatment (33). The aprt locus in AA8 cells is heterozygous due to a point mutation at one of the two alleles (24,34), and aprt is known to have ~10-fold higher mutation rate than hprt because of a high rate of deletion (35). There was no significant increase in mutation rate at the aprt locus in either rad51d or xrcc3 cells. The aprt rate for AA8 was (10 ± 3) × 10−6 and for 51D1 was (13 ± 4) × 10−6, based on the method of the mean. An increase in mutability at the aprt locus in the 51D1 cells could be difficult to detect, as the spontaneous rate is so high in AA8 cells that a Rad51D-dependant increase comparable to that at hprt would be less than a doubling in the 51D1 cells.
Gene amplification is another form of mutagenesis associated with tumor cells, in which megabase regions of DNA are replicated in excess and maintained chromosomally or as extrachromosomal elements. The role of HRR in preventing such events is not known, but CHO cells deficient in the DSB repair pathway NHEJ, because of mutant DNA-PKcs, have increased rates of amplification (29). We determined that the rate of gene amplification was increased 3- to 10-fold in HRR-deficient 51D1 and irs1SF cells at both the dhfr and CAD loci (Table 3). The gene-corrected mutant cell lines showed rates of amplification similar to those of the parental cells. To see if this amplification phenotype could be seen in another paralog mutant, we tested the V79-derived xrcc2 cell line, irs1. It also displayed an increased amplification rate compared to the corresponding XRCC2-complemented cells (GT619) (13), consistent with a general role for the paralogs in preventing gene amplification.
In this study, we created in CHO AA8 cells a knockout mutant of RAD51D (clone 51D1) that is isogenic with respect to AA8 and 51D1Lox parental lines, and to the BAC gene-complemented control (clone 51D1.3). To further ensure an isogenic relationship, the 51D1 mutant was complemented with the hamster RAD51D gene, rather than a human homolog as has often been done historically. Although the level of Rad51D expression in 51D1.3 cells appears to be lower than in AA8 cells (Figure 1C), the level is adequate for full complementation of all aspects of the mutant phenotype. 51D1 is the first isogenic mutant in DSB repair to be constructed in CHO cells.
The rad51d CHO cells resemble rad51d knockout mouse MEFs in showing very high sensitivity to killing by MMC and lesser sensitivity to IR, UV-C and MMS (23). Irs1SF xrcc3 CHO cells have also been shown to be very sensitive to DNA-replication inhibitors (camptothecin and hydroxyurea) that result in broken replication forks (36,37). Thus, HRR allows cells to cope with a broad range of genotoxic agents, all of which result in one-sided DSBs when replication forks stall and then break (2). Such DSBs will also arise when replication forks encounter single-strand breaks produced either directly by the DNA-damaging agent (e.g. IR) or as an intermediate during repair (e.g. MMS and MMC). Another theoretical possibility is a role for HRR in bypassing damaged bases in an error-free manner through a process of fork regression/reversal followed by restart of the replication fork (i.e. ‘chickenfoot’ intermediate) (38).
The profile of sensitivity of the rad51d CHO cells to various DNA-damaging agents is also similar to that seen previously for hamster cells deficient in other Rad51 paralogs: XRCC2 (irs1), XRCC3 (irs1SF) and RAD51C (irs3) cells (11–13,39,40). The 80-fold sensitivity of 51D1 cells to MMC is dramatically higher than the ~3-fold sensitivity of the DT40 rad51d mutant to MMC (16). Similar differences have been seen between systems for other Rad51 paralog mutants, emphasizing that there are inherent differences between these model systems. The similarities among these Rad51 paralog mutants within each system suggest that the paralogs have a common role in HRR.
Gene amplification, a form of mutagenesis in which large regions of genomic DNA are multiply replicated and maintained in the genome, results in tumor cells having increased copies of oncogenes and multi-drug resistance phenotypes [reviewed in (41)]. DNA damage appears to play an important role in promoting gene amplification, as it is enhanced when cells are exposed to agents that break DNA (i.e. γ-rays and hydrogen peroxide) (42). Defects in DNA repair systems have previously been associated with increased rates of gene amplification, including NHEJ deficiency in CHO cells (29) and mismatch repair deficiencies in human tumor cells (43,44). We have also found increased gene amplification rates in a fancg CHO knockout mutant defective in the Fanconi anemia chromosome stability pathway (J.M. Hinz and L.H. Thompson, unpublished data). As shown in Table 3, we have made the novel observation that HRR also plays an important role in preventing gene amplification as evidenced by the highly elevated rates in rad51d cells, and confirmed in the xrcc3 irs1SF and xrcc2 irs1 mutants. However, the highest reported rates of CAD amplification occur in dna-pkcs cells, in which the elevation was 20- to 150-fold (29). From these collective findings it appears that multiple DNA repair pathways, including HRR, contribute to the prevention of gene amplification.
Sister chromatid exchange, which is thought to be a cytological manifestation of HRR, occurs only a few times during each S phase. SCE is greatly increased by a broad range of genotoxic agents including inhibitors of DNA replication (45,46). Evidence that SCE is caused at least in part by HRR was presented in chicken cells (47). However, models have also been proposed in which fork breakage and rejoining by NHEJ could be responsible (45,48). SCE can presumably occur when broken replication forks are restarted by HRR (2). A priori, one might expect that the rad51d deletion would partially suppress SCE, but we found that 51D1 cells had no change compared with control cell lines. Although the rad51d mutant and other Rad51 paralog mutants in chicken DT40 cells have 2- to 3-fold reduced rates of SCE (16), mouse rad51d trp53 knockout MEFs also show no reduction compared with the trp53 control cells (23).
These findings of normal SCE rates in mammalian rad51d mutants have several possible explanations. First, Rad51D may have no role in the putative HRR exchange event associated with restarting a broken replication fork through the processing of a Holliday junction (HJ) intermediate (2). This interpretation was put forth to explain the mouse rad51d MEF data (23). However, given the very high levels of spontaneous chromatid breaks in the mouse and hamster rad51d cells, which most likely derive from unrepaired broken replication forks, this explanation seems unlikely. Second, SCE in wild-type cells might arise primarily from NHEJ by a process in which both parental strands are broken and forks are restored by rejoining parental strands with daughter strands. However, the normal frequency of SCE in xrcc5/ku80 CHO mutants argues against this possibility (49). Third, the restart of broken forks by HRR may normally occur primarily through the ‘crossover’ mode of HJ resolution, which would not produce SCE as cytologically identifiable events (2). The visible SCE could arise in a different manner, e.g. ‘non-crossover’ mode of HR resolution, perhaps independently of Rad51D. In any event, other studies show that the loss of Rad51 paralogs in mammalian cells has only a modest or no influence on spontaneous SCE. Neither xrcc2 irs1 nor xrcc3 irs1SF shows a significant reduction in SCE (10,50), and xrcc2 knockout mouse cells (both primary and immortalized) show only a 30% reduction (22). Two rad51c V79 mutants have a slightly reduced rate of spontaneous SCE (11,12). Thus, mammalian Rad51 paralogs differ from the chicken homologs in having a lesser quantitative contribution to the rate of SCE. However, since the rad51d mutant MEF cells have a clear deficiency in MMC-induced SCEs (23), the mechanistic details of induced SCEs must differ from spontaneous events.
We observed substantially elevated chromosomal aberrations in rad51d cells, particularly chromatid and chromosome breaks, which presumably arise from broken replication forks that remain unrepaired. An essential role for Rad51D in telomere maintenance also has been reported for telomerase-deficient cells. Rad51D was shown to localize at the sites of telomeres in HeLa cells, and rad51d trp53 MEFs have a high level of telomere end-to-end fusions as well as chromatid breaks and other chromosomal abnormalities possibly associated with telomere dysfunction (23,51). The association of Rad51D with telomeric sequences appeared to be specific compared with other Rad51 paralogs. In addition, Rad51D-deficient human cells exhibited telomeric DNA repeat shortening (51). CHO and other immortalized Chinese hamster cells lack cytologically visible telomeres (52), but they express telomerase activity (53) and exhibit interstitial telomeric bands that are subject to amplification (54). The absence of a requirement for Rad51D in telomere maintenance in CHO cells may explain why our rad51d mutant cells grow relatively robustly compared with the rad51d trp53 mouse cells. Importantly, 51D1 cells show only a mild growth retardation, having a doubling time of ~16 h compared with ~13 h for 51D1Lox and AA8 cells.
Genomic alterations in nucleotide sequence that are undetectable by cytogenetics are a crucial aspect of cancer progression, as reviewed by (41). Mutation rate measurements at the hprt gene have been widely used in human, Chinese hamster, and other mammalian cells for quantifying mutagenesis in a locus that is responsive to both point mutations and deletions (24,55–57). Due to the functionally hemizygous nature of the hprt locus [and physical hemizygosity in CHO cells (32)], point mutations, insertions and gene-size deletions are detectable, but large-scale multigenic deletions and interchromosomal rearrangements are not recovered (58). High levels of spontaneous hprt mutagenesis are often associated with defects in mismatch repair due to increased tolerance for mis-incorporated nucleotides (59,60), or with overexpression of translesion polymerases, such as DINB1, Pol κ and Pol β (61–63), as such polymerases synthesize DNA in an error-prone manner.
In the rad51d cells constructed in this study, and in the commonly used CHO model cell line for HRR-deficiency (xrcc3 irs1SF), we find highly elevated rates of mutagenesis, showing a clear role for these Rad51 paralogs in suppressing spontaneous mutagenesis. HRR-defective brca2 V-C8 cells were reported to have ~4-fold elevated hprt mutation rate compared with the non-isogenic parental hamster V79 line, and the mutant spectrum contained an increased proportion of deletions (64). These results, along with our data, highlight the importance of HRR in preventing loss-of-function mutations that are presumed to arise during DNA replication when broken forks are inaccurately repaired by NHEJ.
In conclusion, the tumorigenic progression of somatic cells toward malignancy depends predominantly on two kinds of altered gene expression: loss of tumor suppressor gene function and gain of inappropriate gene expression (e.g. oncogenes and multi-drug resistance). In this study, we determined the contribution of HRR to two classes of spontaneous gene-level mutagenesis: (i) loss of hprt gene function measured as 6-thioguanine resistance, and (ii) dhfr and CAD gene amplification, which confer drug resistance via increased gene dosage (overexpression). The rad51d cells have >10-fold higher hprt mutagenesis and 4- to 10-fold elevated gene amplification rates. For each of the three marker genes, we confirmed the instability phenotype in non-isogenic xrcc3 cells from the same parental CHO line. These genetic instabilities measured at the single-gene level should be viewed in concert with the high, classically HRR-associated cytological measurements of chromosomal breakage and exchange, which are elevated by a similar magnitude in rad51d cells, but only represent a minor facet of the complex genomic alterations required for tumor progression.
The authors thank Lynn Carr and Tricia Allen, teachers from Burroughs High School, Ridgecrest, CA, for their assistance with fluctuation analyses. The authors also thank Angela Hinz for her technical expertise. Deserving recognition is the Drug Synthesis and Chemistry Branch, Division of Cancer Treatment, National Cancer Institute, for providing PALA. This work was performed under the auspices of the U.S. Department of Energy by the University of California, Lawrence Livermore National Laboratory under Contract No. W-7405-Eng-48. The DOE Low-Dose Program and NCI/NIH grant CA89405 funded this work. Funding to pay the Open Access publication charges for this article was provided by NIH grant 1 R01 CA112566-01A1.
Conflict of interest statement. None declared.