O6-methylguanine is responsible for homologous recombination induced by N-methyl-N′-nitro-N-nitrosoguanidine (MNNG) [H.Zhang et al. (1996) Carcinogenesis, 17, 2229]. To test the hypothesis that mismatch repair is causally involved in this process, we generated mismatch repair-deficient strains from a human fibroblast line containing a substrate for detecting intrachromosomal homologous recombination. The four strains selected for study exhibited greatly increased resistance to the cytotoxic effects of MNNG, which was not affected by depletion of O6-alkylguanine-DNA alkyltransferase, and greatly increased sensitivity to the mutagenic effect of MNNG, suggesting that the mutagenic base modifications induced in these four cell strains by MNNG persist in their genomic DNA. Tests showed that their extracts are deficient in the repair of G:T mismatches. The frequency of homologous recombination induced by MNNG in three of these strains was significantly (5–7-fold) lower than that induced in the parental cell strain. This was not the result of a generalized defect in recombination, because when (±)-7β,8α-dihydroxy-9α,10α-epoxy-7,8,9,10-tetrahydrobenzo[a]pyrene was used to induce recombination, all three lines responded with a normal or even a somewhat higher frequency than that observed in the parental strain. The lack of recombination displayed by the fourth strain was shown to result from the loss of part of the recombination substrate. The results strongly suggest that functional mismatch repair is required for MNNG-induced homologous recombination.
Mitotic recombination between homologous genes is one of the mechanisms responsible for cells becoming homozygous for a particular recessive allele (1–5). Spontaneous and carcinogen-induced homologous recombination is also implicated in other types of genetic alterations required for malignant transformation, such as chromosomal rearrangements, translocations, deletions and gene amplifications (6). Maher and colleagues compared the ability of various carcinogens to induce intrachromosomal homologous recombination in murine (7,8) and human (9–11) cells. Their results suggested that in human cells deficient in nucleotide excision repair, recombination induced by UV radiation (10,11) or 1-nitropyrene (10) is triggered by the presence in the genomic DNA of unexcised bulky lesions that block DNA replication, such as UV photoproducts (12) or adducts formed by (±)-7β,8α-dihydroxy-9α,10α-epoxy-7,8,9,10-tetrahydrobenzo[a]pyrene(BPDE) (13) or 1-nitropyrene. They also showed that low doses of the methylating agent N-methyl-N′-nitro-N-nitrosoguanidine (MNNG), which do not cause inhibition of DNA replication, can induce intrachromosomal homologous recombination (7). To determine whether the DNA lesion responsible for MNNG-induced homologous recombination was O6-methylguanine (O6-MeG), Zhang et al. (14) compared the frequency of such recombination events induced in a series of human cell lines that differ significantly in their level of O6-alkylguanine-DNA alkyltransferase (AGT), which reverts O6-MeG back to guanine (15). Their results showed that cell lines lacking AGT exhibited MNNG-induced recombination at significantly lower doses than cells expressing high levels of this repair enzyme. In one part of that study, the target cells being compared were the identical cell strain, expressing AGT or depleted of AGT by pre-treatment with O6-benzylguanine (O6-BzG), an inhibitor of AGT. These data indicated that O6-MeG is the lesion principally responsible for MNNG-induced recombination.
During DNA replication, O6-MeG readily base pairs with both thymine and cytosine, giving rise to O6-MeG:T or O6-MeG:C mismatches (16). However, in contrast to UV photoproducts or BPDE adducts, these structures do not block progression of the replication fork. (At the doses used, i.e. doses that lower the survival of colony-forming ability to between 60 and 25% of the untreated control, the treated populations undergo the first doubling at the same rate as the control, i.e. in ~22 h; data not shown.) Zhang et al. (14) therefore hypothesized that the recombination induced by very low levels of O6-MeG is triggered by the cells processing an O6-MeG:T or O6-MeG:C. It is known that such mismatches are recognized and repaired by post-replicative mismatch repair (16,17). Mismatch repair is initiated by the recognition of mismatches in newly replicated DNA. This step, which is mediated primarily by the heterodimer of hMSH2 and hMSH6, triggers the assembly of the repairosome, which leads, in turn, to the exonucleolytic degradation of the error-containing newly synthesized strand up to and past the mismatch. The process is completed by the resynthesis of the degraded tract and by ligation of the remaining nick (for reviews, see refs 18–21). During replication of methylated DNA, the presence of O6-MeG in the template strand will give rise to the above-mentioned O6-MeG:T or O6-MeG:C mismatches. Should these trigger mismatch repair, the exonucleolytic degradation process will be directed to the strand containing the mismatched thymine or cytosine. As the O6-MeG residue persists in the template strand, the resynthesis step will once again generate O6-MeG:T or O6-MeG:C, which will again be addressed by the repair system. This cyclic, `futile' repair (22) has one important consequence: the methylated, template strand will be single-stranded for much of the time and the newly synthesized strand will contain persisting DNA termini, which are potentially recombinogenic.
Inherited defects in genes encoding mismatch repair proteins are linked to hereditary non-polyposis colorectal cancer, a genetic predisposition to cancer of the colon, endometrium, ovary and other organs (18,23–29). Mismatch-repair-deficient epithelial cell lines, derived from colon or endometrial tumours, not only exhibit a high spontaneous mutation rate (30–34), often accompanied by microsatellite instability (22,32,34–37), but also a high tolerance to the cytotoxic effects of methylating agents (22,32,35–39). The tolerance has been attributed to their failure to carry out the futile mismatch repair discussed above (22,40). This explanation is supported by the fact that selection of mammalian cells in culture for resistance to the cytotoxic effect of MNNG or N-methyl-N-nitrosourea has led to the isolation of infinite life span epithelial or lymphoblastoid cell lines that are defective in mismatch repair and exhibit abnormally high frequencies of spontaneous mutations (40–44), as well as mutations induced by methylating agents (40,45). Mismatch repair is known to prevent recombination between homologous, but non-identical DNA sequences (19,20,46). Correspondingly, mammalian cells that have lost mismatch repair have been shown to exhibit increased rates of spontaneous recombination (44). Hays and colleagues (47,48) showed that in Escherichia coli, UV-induced recombination of non-replicating λ phage depended on functional bacterial mismatch repair capability. However, the role mismatch repair plays in homologous recombination in mammalian cells has not been widely studied. To test the hypothesis that MNNG-induced homologous recombination results from the processing of DNA regions containing O6-MeG:T and/or O6-Me:C mismatches by mismatch repair, we attempted to generate cell strains that had lost the capacity for mismatch repair. We treated MSU-1.2-E.7.2 cells, a repair-proficient, human fibroblast cell strain that contains a substrate for detecting intrachromosomal homologous recombination (14), with MNNG and selected for cells that were significantly resistant to killing by this agent. A large number of resistant candidate clones were then screened to identify those that were also hypermutable by MNNG. Four unequivocally independent, clonally derived candidate cell strains were selected for further study. Cytoplasmic cell-free extracts from these four cell strains and their parental strain were compared for the ability to repair G:T mismatches. None of the four candidate strains could do so. When compared with their mismatch-repair-proficient parental cell strain for the frequency of MNNG-induced homologous recombination, one produced no recombinants, and three strains showed very significantly reduced frequencies. The inability of the latter three strains to carry out homologous recombination was not a generalized defect because they retained the ability to carry out homologous recombination induced by BPDE and also by UV radiation. These results strongly suggest that mismatch repair is required for MNNG-induced homologous recombination.
Materials and methods
Cell strains and culture conditions
MSU-1.2-E7.2 is a human fibroblast cell strain containing a substrate for intrachromosomal recombination, i.e. duplicate copies of the gene coding for hygromycin phosphotransferase (hyg), each mutated by a 10 bp HindIII linker inserted at a different site (14). It has normal DNA repair capacity and a normal level of AGT activity. A clonal derivative of MSU-1.1-E7.2, designated MSU-1.2-10A, containing a transfected gene coding for a tetracycline-responsive fusion transcriptional activator and a hisD gene, was used as the parental strain for the present study. Cells were routinely cultured in McM medium (49) containing 10% (v/v) supplemented calf serum (Hyclone), 100 U/ml penicillin,100 μg/ml streptomycin and, for increased attachment, 1 μg/ml hydrocortisone (culture medium).
Treatment with MNNG or BPDE
Cells were plated at 104 cells per cm2. After 14–16 h, cells were treated with MNNG (Pfaltz and Bauer, Flushing, NY), using the conditions described by Zhang et al. (14). For treatment with BPDE (Chemsyn Science Laboratories, Lenexa, KS), the conditions used were those described by Yang et al. (50).
Assay for cytotoxicity
The cytotoxic effect of the carcinogens was determined as described (14,51) from the relative colony-forming ability of the treated cells compared with the untreated control cells. Briefly, following treatment of the cells with carcinogen at the indicated density, cells were rinsed with PBS, detached with trypsin, plated at cloning densities in culture medium, and allowed 14 days to form colonies, with one refeeding with culture medium.
Depletion of AGT activity
Two hours of pre-treatment of the cells with 25 μM O6-BzG as described (14) was used to deplete them of AGT activity.
Assay of the frequency of hypoxanthine phosphoribosyltransferase (HPRT) gene mutants
Cells in exponential growth were plated into a series of 150 mm diameter dishes at a density of 104 cells per cm2. A sufficient number of dishes was used for each dose to ensure at least 1–2×106 surviving target cells. After treatment with carcinogen, the cells were rinsed with PBS, refed with culture medium and maintained in exponential growth. After 8 days, at least 1×106 cells from the control and from each treated population were selected for resistance to 6-thioguanine at a density of 500 cells per cm2 (40 dishes per 106 cells selected) as described (51). The cloning efficiency of the cells at the time of selection was also determined and was used to correct the observed frequency of drug-resistant clones.
Mismatch repair assays with cell-free extracts
The efficiency of repairing G:T mismatches by the cell-free extracts was tested as described previously (52). The cytoplasmic cell-free extracts were prepared as described (53). The protein concentrations ranged from 5 to 10 μg/μl. Briefly, M13mp2 heteroduplexes containing a G:T mispair, and with a single-strand break in the strand containing the T, were incubated with 50 μg of cell-free extract (54). The double-stranded viral DNA was then purified and electroporated into a mutS mismatch repair-deficient strain of E.coli, and these cells were plated along with the α-complementation E.coli strain CSH50 onto agar containing isopropyl β-d-thiogalactoside and 5-bromo-4-chloro-3-indolyl β-d-galactoside, to score plaques as described (54). If no repair occurs, a high percentage of mixed plaques, composed of approximately equal numbers of wild-type or mutant phage and thus producing both blue and colorless areas, is observed. Reduction in the percentage of mixed plaques and a concomitant increase in single-color plaques is indicative of mismatch repair. Repair efficiency (%) is presented as 1 – (percentage of mixed plaques in an extract-treated sample/percentage of mixed plaques in an untreated control sample). To complement the mismatch repair activity in the various extracts, hMutSα, a heterodimer containing proteins hMSH2 and hMSH6, and/or hMutLα, a heterodimer containing proteins hMLH1 and hPMS2, were added to the extracts as described by Marra et al. (52). Cell-free extracts from HeLa cells were used as a positive control. Extracts from MT1 cells, which lack wild-type hMSH6 protein (45,55), and from HCT116 cells, which lack wild-type hMLH1 protein (27), were used as negative controls for the assay.
Assay for homologous recombination
For each dose, sufficient numbers of target cells, plated at 104 cells per cm2, were used to ensure at least 2×106 surviving cells. Extra sets of dishes containing cells plated at the same density were used to determine the number of attached cells at the time of treatment and the cytotoxic effect of the treatment. After ~14 h, the number of attached cells was determined, and the cells in the rest of the dishes were mock-treated or treated with MNNG or BPDE as described. Immediately following treatment, the medium was removed, and the cells were rinsed and refed with culture medium. Those to be used to determine cell survival were assayed for cytotoxicity as described above. After ~48 h, the medium on the cells to be assayed for the frequency of homologous recombination was exchanged for culture medium containing 100 U/ml hygromycin-B (Calbiochem). The frequency of recombination was determined from the number of hygromycin resistant (hygr) colonies, divided by the number of surviving cells in each population. The latter was determined from the cytotoxicity assay and the number of cells attached in the dishes at the time of treatment. The frequency of induced recombination was determined by subtracting the background frequency observed in the untreated control population.
Assay for the types of recombination events
The hyg genes from hygromycin-resistant clones were amplified using PCR as described (14). The amplified PCR products were digested by HindIII, and the types of recombination events were determined as described (11,14) from the HindIII digestion pattern on gels.
Isolation of candidate mismatch-repair-deficient cell strains
Zhang et al. (14) showed that the frequency of MNNG-induced intrachromosomal homologous recombination in the MSU-1.2-E7.2 strain is highly correlated with the level of O6-MeG in the cells' DNA: when this strain was pre-treated with O6-BzG to inhibit the endogenous AGT and then exposed to MNNG, the cells exhibited greatly increased sensitivity to both the cytotoxic and the recombinogenic effect of MNNG (14). In order to show that this effect was causally linked with mismatch repair, we needed to generate cell clones carrying the recombination reporter system (14) in a mismatch-repair-deficient background. Previously, Maher and colleagues (56) showed that elimination of AGT by O6-BzG pre-treatment also renders human cells extremely sensitive to the mutagenic effect of MNNG. We decided to make use of this phenomenon in generating mismatch-repair-deficient clones from MSU-1.2-E7.2 clone MSU-1.2-10A. If successful, the mismatch-repair-deficient cell strains thus generated could be identified by virtue of their being abnormally resistant to the cytotoxic effect of methylating agents (22,38,42,45), but abnormally sensitive to their mutagenic effect. If, on the other hand, cells were to acquire resistance to the cytotoxic effect of MNNG as a result of increased expression of AGT (43), they would also acquire resistance to MNNG-induced mutations. We plated 6×106 target cells into four 150 mm diameter dishes. After 24 h, they were pre-treated with O6-BzG for 2 h to inactivate AGT, and then exposed to 1 μM MNNG for 1 h, as described. [Under these conditions, the MSU-1.2-E7.2 cells, the parent cell strain for the MSU-1.2–10A cells, would exhibit a survival of <1% of the untreated control (14).] Four days later, the cells remaining in the dishes were exposed to 1.5 μM MNNG for 1 h under the same conditions, including O6-BzG pre-treatment. Four days later, the cells remaining in the dishes were similarly treated, but using 4 μM MNNG. After another 4 days, when the majority of the cells had detached from the dishes, this latter step was repeated.
The surviving cells, which were abnormally resistant to MNNG, were allowed to form large-sized colonies, and 43 of these colonies were expanded into cell strains and screened for the desired phenotype, i.e. extreme resistance to the cytotoxic effect of MNNG and sensitivity to its mutagenic effect, an indication of a mismatch repair defect (16). For these latter studies, the cell strains were pre-treated with O6-BzG, exposed to 6 μM MNNG for 1 h and, after an 8 day expression period, assayed (2×105 per strain) for 6-thioguanine resistance. Four unequivocally independent strains that yielded a high frequency of mutants in this prescreen were chosen for further study and were subcloned. The four subcloned strains were designated MSU-1.2-A12.2, MSU-1.2-B7.7, MSU-1.2-C14.6 and MSU-1.2-D1.1. In this paper, they and their parental strain, MSU-1.2-10A, are referred to as A12, B7, C14, D1 and 10A, respectively.
Evidence of increased resistance to the cytotoxic effect of MNNG, but sensitivity to its mutagenic effect
The four candidate cell strains were tested for their response to MNNG. In these studies, O6-BzG pre-treatment was not used. As shown in Figure 1A, all four strains were significantly more resistant than the parental cell strain to the cytotoxic effect of MNNG. (To reduce survival to 90% of the untreated control required a dose of 0.5 μM MNNG for parental cell strain 10A, 1.5 μM for B7, 2.5 μM for C14 and 4 μM for A12 and D1.) The response of the parental strain 10A was identical to that found earlier for its parental cell strain, MSU-1.2-E7.2 (14). The sensitivity of the cell strains to the mutagenic effect of MNNG is shown in Figure 1B. All four were significantly (4–13-fold steeper slope of the curve) more sensitive to the mutagenic effect of MNNG than their parental cell strain 10A. A dose of 4 μM MNNG yielded 310–990 mutants per 106 cells, compared with 80 per 106 cells for the parental strain. The abnormal resistance of the four candidate strains to cell killing by MNNG cannot be the result of increased expression of AGT because, as we showed previously in human fibroblasts (56), such expression would also decrease the mutant frequency, not increase it.
Three of these candidate strains were compared with their parental strain, for sensitivity to the cytotoxic effect of MNNG after depletion of AGT by pre-treatment with O6-BzG (Figure 2). Although this pre-treatment drastically increased the sensitivity of the parental strain to the cytotoxic effect of MNNG, it did not affect that of the three candidate cell strains. Their response was virtually identical to that found when pre-treatment with O6-BzG was omitted (Figure 1A). This indicates that their resistance to the cytotoxic effect of MNNG was not the result of increased AGT levels, but resulted from their inability to process mismatches involving O6-MeG.
Mismatch repair capacity of the candidate cell strains and their parental strain
The mismatch repair capacity of the cell strains was analyzed using a cell-free repair assay involving a heteroduplex containing a G:T mismatch and a nick 5′ from the mismatch in the T-containing strand (52,54). Two independently isolated cytoplasmic extracts of the strains were prepared as described (52–54) and tested for their capacity for mismatch repair. Extracts from HeLa cells (epithelial) were used as a positive control, while extracts from MT1 cells (lymphoblastoid), mutated in both copies of the hMSH6 gene (45,55), and from HCT116 cells (epithelial), a colon-cancer-derived strain carrying homozygous mutations in the hMLH1 gene (27), were used as negative controls. As shown in Figure 3, extracts from the parental 10A cells and HeLa cells were proficient in G:T repair. Extracts from strains A12, B7, C14 and D1, like those from MT1 and HCT116 cells, were unable to repair the mismatch.
The ability of purified recombinant hMutSα (a heterodimer of hMSH2/hMSH6), and/or purified recombinant hMutLα (a heterodimer of hMLH1 and hPMS2) (52) to complement the mismatch repair defect in the various cell extracts was compared. As shown in Figure 3, the mismatch repair defect in extracts from strains A12 and B7 was fully complemented by hMutSα, suggesting that these strains are deficient in MSH2 and/or MSH6. In contrast, the defect in repair of G:T mismatches in extracts from strains C14 and D1 could not be fully complemented by the addition of hMutSα. Their respective G:T mismatch repair activity reached only 35 and 75% of that shown by their parental cell strain A12. Nor could it be fully complemented by the hMutLα protein complex. Addition of purified hMutLα to these extracts increased their mismatch repair activity to only 50 and 60%, respectively, of that seen with HCT116 cells, which lack MLH1 (27). Full complementation required both heterodimers, suggesting that the C14 and D1 strains have a complex phenotype. Interestingly, analysis of the hMSH2, hMSH6, hMLH1 and hPMS2 proteins in extracts of these cells by western blotting revealed that all were expressed at a normal level and were full-length (data not shown). This suggests that the mismatch repair defects in these cells strains are caused by inactivation of the function of hMutSα and hMutLα through missense mutations, such as would be expected following mutagenesis with MNNG.
Frequency of MNNG-induced homologous recombination in the mismatch-repair-deficient cell strains
Once we had shown that cell-free extracts from the four candidate cell strains lacked the ability to repair G:T mismatches, we compared them with their parental repair-proficient cell strain for the frequency of MNNG-induced homologous recombination in order to determine if repair of O6-MeG:T mismatches could be responsible for the O6-MeG-induced recombination seen by Zhang et al. (14). The cells were exposed to various doses of MNNG and assayed for the induced frequency of hygromycin-resistant clones arising as a result of homologous recombination between the two inactivated copies of the hyg gene. [Generation of a wild-type hyg gene in the substrate used for the recombination assay has been shown to be completely dependent upon the presence of a duplicate copy in the chromosome (57).] As shown in Figure 4, MNNG induced a dose-dependent increase in the frequency of homologous recombination in the mismatch repair-proficient parental strain, 10A, above a background frequency of ~13 per 106 cells. This frequency is virtually identical to that obtained previously by Zhang et al. (14) using strain MSU-1.2-E7.2, from which the 10A cell strain was derived. In contrast, the frequency of MNNG-induced recombination in the mismatch repair-deficient cell strains A12, C14 and D1 was significantly lower (slopes of the curves were at least 6-fold less steep). Cell strain B7 yielded no recombinant hygr clones at all.
Ability of the strains to carry out homologous recombination induced by BPDE
To address the possibility that the significantly lower frequency of MNNG-induced recombination resulted from a generalized defect in the recombination process, we tested the mismatch repair-deficient strains for their frequency of BPDE-induced homologous recombination. BPDE forms bulky adducts that strongly inhibit fork progression during DNA replication (13,58). As shown in Figure 5, the frequency of recombination induced by BPDE in strains A12, C14 and D1 was not lower than that found with their parental strain. In fact, the slopes of their curves were slightly higher than that of the parental strain. Strain B7 once again failed to produce any recombinants. Subsequent PCR analysis of the recombination substrate in cell strain B7 revealed that one of the two hyg genes, the SacII mutant, is no longer intact (data not shown). This accounts for the observed total lack of recombination in that strain. As shown in Figure 5A, the mismatch repair-deficient strains A12, B7 and D1 were somewhat more resistant to cell killing by BPDE than the parental strain.
PCR analysis of spontaneous, MNNG- or BPDE-induced recombinants derived from the mismatch repair-deficient cell strains and their parental cell strain, followed by HindIII digestion and gel electrophoresis, revealed that each was the product of a gene conversion event, rather than of a single reciprocal exchange. The total number of independent recombinant clones analyzed for cell strains A12, C14, D1 and 10A were 16, 22, 32 and 53, respectively. Finding a very high proportion of gene conversions agrees with what was found previously using this recombination substrate to detect recombination induced in human cells by UV (11) or MNNG (14).
In this study, we used derivatives of cell strain MSU-1.2-E7.2 (14), which contain an integrated substrate for detecting genetic recombination between duplicated copies of the hyg gene, each of which has been interrupted by a HindIII restriction site at a different position (11). Productive recombination, i.e. gene conversion or reciprocal exchange between the two hyg genes, yields a wild-type hyg gene that can be detected by selection for hygromycin resistance (11,14). The purpose was to test the hypothesis that the homologous recombination induced in human cells by O6-MeG after exposure to methylating agents depends on mismatch repair to generate the necessary intermediate DNA structures. That hypothesis was based on the results of two earlier studies by Maher and colleagues (10,11), who showed that human cells deficient in nucleotide excision repair are much more sensitive than repair-proficient cells to the induction of recombination by DNA damage that blocks replication fork progression, such as UV photoproducts or bulky chemical adducts. Those data strongly suggested that recombination between the duplicated mutant hyg genes results from the generation of potentially recombinogenic single-strand DNA termini at the site of the replication block. Later, Zhang et al. (14) showed that human cells lacking AGT and, therefore, unable to remove methyl groups from the O6 position of guanine, are much more sensitive than AGT-containing cells to MNNG-induced recombination. Since O6-MeG lesions readily mispair during DNA replication and do not result in a strong replication fork block, Zhang et al. (14) hypothesized that such recombination was being stimulated by postreplicative mismatch repair acting on the strand containing a mismatched thymine or cytosine opposite the O6-MeG (59). This single-stranded DNA could provide a signal for recombination.
To test the hypothesis that MNNG-induced recombination is linked to the processing of such mismatches by post-replicative mismatch repair, we sought to generate, from our human fibroblast cell strain containing the substrate for intrachromosomal homologous recombination, a series of cell strains that would lack the mismatch repair proteins necessary to repair mismatches involving O6-MeG. The four candidate strains were chosen because they exhibited attributes of the alkylation-tolerant phenotype associated with defective mismatch repair (22,38,41,43). They showed significantly increased resistance to the cytotoxic effects of MNNG (Figure 1A), even when pre-treated with O6-MeG to deplete AGT activity (Figure 2), and increased sensitivity to MNNG-induced mutations (Figure 1B). Subsequent analysis showed that extracts prepared from the four candidate cell strains were, indeed, incapable of repairing a G:T mismatch (Figure 3). It is true that these four cell strains were derived from exposure to a mutagen and, therefore, they can be expected to contain multiple mutations. However, the chance that four independently derived cell strains that no longer repair mismatches would exhibit the attributes of mismatch-repair-deficient cells as a result of a mutation in some random genes, rather than from a defect in mismatch repair itself, is very low.
Since the defect in extracts from strains A12 and B7 could be fully complemented in vitro by hMutSα, we conclude that they have a defective hMutSα complex. Full complementation of the extracts from strains C14 and D1, which were unable to repair mismatches, required addition of both recombinant heterodimers, hMutSα and hMutLα, suggesting that both these factors are defective in the latter two strains. Given that all four constituent proteins of the two mismatch repair factors were present in the extracts in amounts comparable with normal controls (data not shown), the defect is likely to be due to missense mutations, rather than to truncations or promoter inactivations. Support for this hypothesis is provided by the finding that purified hMutSα or hMutLα could partially complement the mismatch repair defect of the extracts from cell strains C14 and D1. This implies that both wild-type recombinant factors could interact with their respective mutated cellular counterparts to an extent sufficient to mediate mismatch correction. However, further experiments are needed to determine the exact nature of the defects in the mismatch repair genes of these cell strains.
The present study shows that these cell strains display an abnormally low frequency of MNNG-induced recombination (Figure 4). The total lack of recombinants in strain B7 proved to be the result of a loss of part of the recombination substrate, rather than loss of mismatch repair capacity. Because the sensitivity of the other three mismatch repair-deficient strains to the cytotoxic and recombinogenic effects of BPDE was similar to that of their mismatch-repair-proficient parental cell strain (Figure 5), we conclude that these three mismatch-repair-deficient human fibroblast cell strains are still able to carry out the steps needed to produce hyg gene recombinants induced by agents that strongly block replication forks. Because these same strains are significantly less able than mismatch-repair-proficient cells to produce recombinants induced by MNNG, we conclude that recombination induced by O6-MeG lesions in DNA in human fibroblasts is initiated by the action of mismatch repair on O6-MeG:T and/or O6-MeG:C mismatches. Future experiments designed to measure the effect of restoration of mismatch repair capacity on the frequency of recombination induced by MNNG in cell strains A12 and C14 or D2 will test this conclusion.
Recently, Wu et al. (60) reported that both hMSH6-defective MT1 human lymphoblastoid cells and hMLH1-defective HCT116 human epithelial tumor-derived cells are more resistant to the cytotoxic effect of BPDE than their mismatch-repair-proficient counterparts and that the BPDE-induced cell death in mismatch repair-proficient cells is the result of apoptosis, both p53-dependent and p53-independent. In our study with G:T mismatch-repair-deficient fibroblast cell strains, we observed only a slightly increased resistance to BPDE compared with the parental cells (Figure 5A). [All five of our fibroblast cell strains exhibit wild-type p53 transactivating ability (data not shown).] The difference in sensitivity to the cytotoxic effect of BPDE between the cell lines studied by Wu et al. (60), and our cell strains may reflect differences in the conditions used for the assay, or the fact that our strains are fibroblasts. It has been reported that fibroblasts do not undergo apoptosis as readily as lymphoblasts or epithelial cells (61,62). Further experiments are needed to understand the difference between our cell strains and those of Wu et al. (60) in their response to damage by bulky chemicals.
In conclusion, it is known that mismatch repair processes can lower the frequency of recombination by disrupting attempted recombination between strands of DNA that are homologous, i.e. strands that contain a number of mismatches at separate places in the sequence and, conversely, that loss of mismatch repair activity increases the frequency of such homologous recombination (20,44,46). In contrast, the results of our study show that human fibroblast cell strains that have lost the ability to repair O6-MeG:T or O6-MeG:C mismatches have greatly decreased frequencies of MNNG-induced homologous recombination. This implies that in mismatch repair-proficient cells, processing of O6MeG:T mismatches can trigger recombination between homologous DNA sequences. If such homologous recombination were to occur in a segment of DNA containing a mutated allele of a gene critical for malignant transformation, the recombination event could cause both alleles of the gene to exhibit the same mutation. This event would result in loss of heterozygosity and could lead ultimately to tumor formation. Examples of such recombination induced by carcinogen treatment have been reported (63).
We thank Dr Thomas A.Kunkel, National Institute of Environmental Health Sciences, for many helpful discussions on mismatch repair. We also thank Beatrice Tung, Carcinogenesis Laboratory of Michigan State University, for her helpful advice and expert assistance carrying out related studies in support of this research. We thank Dr Richard Iggo, Institut Suisse de Recherches sur Cancer, Lausanne, Switzerland, for generously providing the materials for the yeast assay of the transactivating ability of the p53 gene in the five fibroblast cell strains used in this study. This research was supported by DHHS grants CA21253 and CA48066 to V.M.M. from the National Cancer Institute. J.J. and G.M. gratefully acknowledge the generous support of the Swiss National Science Foundation.