Germ line DNA mismatch repair mutations in MLH1 and MSH2 underlie the vast majority of hereditary non-polyposis colon cancer. Four mammalian homologues of Escherichia coli MutL heterodimerize to form three distinct complexes: MLH1/PMS2, MLH1/MLH3, and MLH1/PMS1. Although MLH1/PMS2 is generally thought to have the major MutL activity, the precise contributions of each MutL heterodimer to mismatch repair functions are poorly understood. Here, we show that Mlh3 contributes to mechanisms of tumor suppression in the mouse. Mlh3 deficiency alone causes microsatellite instability, impaired DNA-damage response, and increased gastrointestinal tumor susceptibility. Furthermore, Mlh3;Pms2 double-deficient mice have tumor susceptibility, shorter life span, microsatellite instability, and DNA-damage response phenotypes that are indistinguishable from Mlh1-deficient mice. Our data support previous results from budding yeast that show partial functional redundancy between MLH3 and PMS2 orthologues for mutation avoidance and show a role for Mlh3 in gastrointestinal and extragastrointestinal tumor suppression. The data also suggest a mechanistic basis for the more severe mismatch repair–related phenotypes and cancer susceptibility in Mlh1- versus Mlh3- or Pms2-deficient mice. Contributions by both MLH1/MLH3 and MLH1/PMS2 complexes to mechanisms of mismatch repair–mediated tumor suppression, therefore, provide an explanation why, among MutL homologues, only germ line mutations in MLH1 are common in hereditary non-polyposis colon cancer.
- DNA mismatch
- Resistant to Alkylating Agent
- gastrointestinal tumor susceptibility
- Microsatellite instability
- Mlh3 and Pms2
- DNA damage, DNA repair, and mutagenesis
- Alkylating agents
- Animal models of cancer
DNA mismatch repair contributes to tumor suppression by reducing mutations and promoting apoptosis in response to certain types of DNA damage ( 1– 5). In humans, the majority of hereditary nonpolyposis colon cancer (HNPCC) families and sporadic colorectal cancer with microsatellite instability have germ line MLH1 or MSH2 mutations or somatic inactivation of MLH1 ( 6). Whereas a recent study suggests that PMS2 mutations might be underestimated because of paralogous genes that complicate mutation detection ( 7), it is unlikely that this alone accounts for the >40:1 ratio of identifiable MLH1 versus PMS2 mutations in HNPCC ( 6). The preponderance of MLH1 and MSH2 mutations is at first glance paradoxical because multiple MutS and MutL homologues exist in humans. Studies in yeast and mammals have shown that MutS proteins form MSH2/MSH6 and MSH2/MSH3 heterodimers, which have partially overlapping specificities for binding to DNA mismatches ( 4, 8, 9). These same studies also show that deficiency in MSH2 or double deficiencies in MSH3 and MSH6 results in a mismatch repair “null” phenotype, whereas a single deficiency in either MSH6 or MSH3 produces attenuated mismatch repair defects. This partial functional overlap between MSH3 and MSH6 likely explains the preponderance of MSH2 mutations in HNPCC compared with either MSH6 or MSH3 mutations ( 10– 12).
A similar situation exists for the heterodimeric MutL homologues, which associate with the MutS heterodimers following initial mismatch binding. MLH1 can heterodimerize with PMS2 (orthologue of yeast Pms1p; refs. 1, 4, 5), MLH3 (orthologue of yeast Mlh3p), or PMS1 ( 8). However, our understanding of the relative contributions of each mammalian MutL heterodimers to mismatch repair and tumor suppression is incomplete.
Our previous studies have shown that Mlh1, but not Pms2 or Pms1, deficiency predisposed to gastrointestinal malignancy in the mouse ( 13). We also showed that Mlh1-deficient mice have significantly higher levels of microsatellite instability, most notably in mononucleotide repeat tracts, than either Pms2- or Pms1-deficient mice ( 14). Of particular relevance are studies in budding yeast, which showed partial functional overlap between pms1 (orthologue of mammalian PMS2) and mlh3 and suggested that a similar overlap may exist for the mammalian MutL orthologues ( 4, 8, 15). Based on the yeast and mouse studies, we and others ( 1, 4, 13– 16) have proposed that the lower mutation levels and gastrointestinal tumor susceptibility in Pms2- versus Mlh1-deficient mice was due to functional overlap between Pms2, Pms1, and/or Mlh3.
MLH3 is the most recently characterized eukaryotic MutL homologue ( 15, 17). Studies of colorectal cancer patients have indicated that MLH3 mutations are very rare in HNPCC families ( 18– 21). However, the extent to which mammalian MLH3 participates in tumor suppression, and the precise mechanisms involved, has not been addressed. Here, we use mouse models to show that Mlh3 deficiency alone causes increased microsatellite instability, defective DNA damage–induced response, increased cancer susceptibility in gastrointestinal and extragastrointestinal tissues, and early mortality. We also find that mice deficient for both Mlh3 and Pms2 show phenotypes that are more severe than those in either single-deficient mouse. In fact, all phenotypes studied in Mlh3;Pms2 double-deficient mice are indistinguishable from Mlh1-deficient mice. In particular, the effect on mutation in mononucleotide repeats is more than additive in the double-deficient mouse, suggesting partial redundancy between Mlh3 and Pms2 for mismatch repair-mediated mutation avoidance. The apparent requirement for both Mlh3 and Pms2 to achieve full mismatch repair activity in the mouse (presumably each functioning as a heterodimer with Mlh1) has important implications for MutL activity in mammals. In summary, our findings show that budding yeast mismatch repair is a reliable paradigm for mammalian mismatch repair, provide a basis for the higher mutation frequencies observed in Mlh1 versus Pms2 deficiency, and help to explain why Mlh1-deficient but not Pms2-deficient mice develop gastrointestinal tumors. Most importantly, our results showing contributions by both Mlh1/Pms2 and Mlh1/Mlh3 complexes to DNA-damage response and tumor suppression provide insight into why, among the MutL homologues, only germ line MLH1 mutations are common in HNPCC.
Materials and Methods
Mouse strains and survival analyses. Wild-type (Wt), Mlh3+/−, Pms2+/−, and Mlh1+/− mice were all generated and maintained on the 129 Sv/Ev genetic background and intercrossed to generate Mlh3−/−, Pms2−/−, and Mlh1−/− mice, respectively. Mlh3−/−;Pms2−/− mice were generated by intercrossing Mlh3+/−;Pms2+/− mice. In addition to the mice used in the present study, for comparison, previously published cancer susceptibility studies done with Pms2−/− and Mlh1−/− mice on other strain backgrounds are also included in Table 1 . Mlh1−/− mice previously described by Edelmann et al. ( 22) were on a mixed 129/Ola and C57B/L6 background, whereas the Pms2−/− and Mlh1−/− mice previously generated by Prolla et al. ( 13) were on a mixed 129Sv/Ev and C57B/L6 background. Kaplan-Meier survival curves were generated as previously done ( 10, 22– 27). The numbers of mice included in survival analyses are Wt (n = 34), Mlh3+/− (n = 27), Mlh3−/− (n = 34), Mlh3−/−;Pms2−/− (n = 20), Pms2−/− (n = 16), and Mlh1−/−, (n = 36). Statistical significance between genotypes was determined using the log-rank test as previously done ( 10, 22– 27).
Analysis of tumors. All lines of mice were observed and not necropsied until they became morbid or moribund. Sacrificed mice were surveyed for tumors and suspicious masses were analyzed histologically as previously done ( 10, 22– 30). Tumors were pathologically confirmed by Dr. Kan Yang, an experienced mouse cancer pathologist who has analyzed many different mismatch repair–defective strains for cancer susceptibility ( 10, 22– 30). Statistical analyses of tumor onset and incidence among the different mouse lines were done by using the Mann-Whitney test as previously done ( 10, 22– 30).
Immunohistochemistry. Formalin-fixed tissues were deparaffinized and rehydrated as previously described ( 22– 30). Antigen retrieving was done by autoclaving slides in 0.01 mol/L citric acid solution at 121°C for 20 minutes. The sections were stained overnight with antibody for mammary gland (anti-Neu/erbB2 antibody, 1:2,000 at 4°C; DAKO Corp., Carpinteria, CA) or antibody directly recognizing Mlh1 (clone Ab-2, 1:10 at room temperature; Calbiochem, San Diego, CA). The secondary antibodies were added and signal-enhancing steps were done following the protocol described on the Vectastain Elite ABC kit (Vector Laboratories, Inc., Burlingame, CA). Slides were developed using VIP staining (Vector Laboratories) and counterstained with methyl green.
Cell Lines and 6-thioguanine Treatment
Intercrosses of Mlh3+/−;Pms2+/− and Mlh3+/−;Pms2+/− mice were used to generate Mlh3−/−, Pms2−/−, and Mlh3−/−;Pms2−/− and Wt mouse embryonic fibroblasts. Generation of Mlh1−/− mouse embryonic fibroblasts has been previously described ( 31). Mlh1−/− mouse embryonic fibroblasts were derived from embryos generated by intercrosses of Mlh1−/− mice on a mixed 129 Sv/Ev and C57B/L6 background. Mouse embryonic fibroblasts were established from day 12.5 postcoitus embryos isolated from the uteri of pregnant mice as previously described ( 32). To analyze the cytotoxicity in response to 6-thioguanine–induced DNA damage, spontaneously immortalized mouse embryonic fibroblasts were sparsely plated (1,000 cells/plate) into 100 mm plates. The following day, cells were exposed to different concentrations (0-5.0 μmol/L) of 6-thioguanine (Sigma, St. Louis, MO) in DMEM complete media containing 15% defined calf serum (Hyclone, Logan, UT) at 37°C for 24 hours. Cells were rinsed twice with 5 mL PBS and were incubated in 15% complete medium. Seven to 10 days later, cells were fixed in 30% ethanol and stained with 0.25% methylene blue. Only colonies containing >100 cells were counted. For each dose, four to six 100 mm plates were analyzed. For each genotype, two to three independently derived mouse embryonic fibroblasts were used and four experiments were tested in each cell line. Our treatment protocol is essentially identical to our previous studies of 6-thioguanine–induced death in Mlh1−/− mouse embryonic fibroblasts ( 31).
Flow cytometry analysis. Mouse embryonic fibroblasts were synchronized by incubating with 0.5% complete medium. After 24 hours, 2 × 104 to 1 × 105 cells were seeded into 100 mm plates depending on the planned day of harvest and incubated with 15% complete medium. After 12 hours, cells were rinsed with PBS twice, treated with 5 μmol/L 6-thioguanine in 15% complete medium, and maintained in the 6-thioguanine–containing medium for 72 or 96 hours. When harvesting cells, both floating cells and attached cells were collected and washed with PBS twice. Cells were pelleted and resuspended in sample buffer containing 0.1% glucose in PBS. Cells were repelleted and fixed by adding cold 70% ethanol dropwise and then stored in 4°C overnight for complete fixation. The next day, samples were stained with propidium iodine (Sigma) for cell cycle analysis. The percentage of sub-G1 cells from each genotype was quantified from Flow Cytometry plots with FLOWJO software (Tree Star, Inc., Ashland, OR). For each genotype, two independently derived mouse embryonic fibroblasts were used and three experiments were tested for each cell line. Statistical significance of differences in the percentage of sub-G1 cells was analyzed using Student's paired t test.
Immunoblotting. Mouse embryonic fibroblast cell extracts were prepared as previously described ( 32) and 60 μg protein from each lysate were separated on 4% to 12% gradient SDS-PAGE gels followed by electrotransfer onto nitrocellulose membranes. The membrane was first incubated with rabbit anti-Mlh1 antibody (Ab-2; Calbiochem) and then with secondary horseradish peroxidase–conjugated anti-rabbit IgG antibody (Jackson Immunoresearch Laboratories, West Grove, PA). Immunoblotting of β-actin (Santa Cruz Biotechnology, Santa Cruz, CA) was done as a loading reference.
Microsatellite instability analysis. For microsatellite instability analysis, >1,000 individual microsatellite templates were analyzed for insertion/deletion mutations by single-molecule PCR using the same protocols as previously done by our group and others ( 14, 26, 33). Briefly, for mononucleotide microsatellite marker JH117, equal amounts of tail DNA were isolated from five mice of each genotype, pooled, and diluted to 0.5 to 1.5 genome equivalents ( 14). Four mice each were studied for Mlh3−/− and Mlh3−/−;Pms2−/− for U12235. Statistically significant differences in the mutation frequencies between genomic DNA from different genotypes were tested by the method of equality of two binomial proportions ( 14). We computed a confidence interval for each mutation frequency and calculated a P value to test the statistical hypothesis that the frequencies in the two genotypes were the same (two-sided test). To test additivity of mutation frequencies in Mlh3−/−;Pms2−/− DNA compared with Mlh3−/− or Pms2−/− DNA, we used the equation X12 = X1 + X2 − (X1X2), where X1, X2, and X12 are the mutation probabilities for Mlh3−/−,Pms2−/−, and Mlh3−/−;Pms2−/− double-knockout mice. The probabilities were corrected for the fact that the observed mutation frequencies include the background frequency observed in wild-type genomic DNA.
Survival and cancer susceptibility in mice with MutL homologue defects. Because yeast studies predicted some degree of functional overlap between mammalian MLH3 and PMS2 ( 15), which could account for different phenotypes of Mlh1−/− and Pms2−/− mice ( 14), we compared life span and cancer susceptibility in wild-type (Wt), Mlh3−/−, Pms2−/−, Mlh3−/−;Pms2−/−, and Mlh1−/− mice on the same 129 Sv/Ev strain background. For these studies, we used mice that were functionally null for the genes under study based on lack of detectable protein by Western analysis ( 14, 17). Consistent with previous studies ( 13, 22), Mlh1-deficient mice have early mortality (mean 6.7 months) due to gastrointestinal and extragastrointestinal cancers, and Pms2-deficient mice have less pronounced early mortality mean (9.7 months) associated with only extragastrointestinal cancers ( Fig. 1 ; Table 1). Interestingly, Mlh3−/− mice have a significantly shorter life span compared with either Wt or Mlh3+/− mice (P < 0.0001 for both comparisons, log-rank test; ( Fig. 1) but live significantly longer than Pms2−/− or Mlh1−/− animals (P < 0.0001 for both; Fig. 1; Table 1).
Next, we analyzed the tumor spectra in a cohort of 27 morbid or moribund Mlh3−/− mice. More than 50% of Mlh3−/− mice developed primary malignancies in the lower gastrointestinal tract, including stomach, small intestine, colon, and rectum. Both adenomas (n = 9) and adenocarcinomas (n = 4) were observed ( Fig. 2A-B ; Supplementary Fig. S1A; Supplementary Table S1). In parallel, no gastrointestinal cancers were observed in >30 age/gender-matched Wt littermate controls ( Fig. 1; Table 1). Therefore, the gastrointestinal cancer susceptibility of Mlh3−/− mice is highly significant (P < 0.0001, log-rank test; Fig. 1; Table 1). Tissue surveys showed that Mlh3−/− mice also have significantly higher risk of extragastrointestinal cancer, including lymphomas, basal cell carcinoma of the skin, mammary gland carcinomas, osteosarcomas, testicular cancer, and hepatic adenomas compared with wild-type mice (P < 0.0001, log-rank test; Fig. 3 ; Table 1; Supplementary Table S1). These data suggest that Mlh3 contributes to suppression of both gastrointestinal and extragastrointestinal tumors. Consistent with previous reports, gastrointestinal tumors were not seen in the Pms2−/− animals ( Table 1; ref. 13).
Of particular interest, survival of Mlh3−/−;Pms2−/− mice was indistinguishable from Mlh1−/− mice (P = 0.21, Mann-Whitney test; Fig. 1) and 84% of Mlh3−/−;Pms2−/− mice examined developed gastrointestinal or extragastrointestinal cancers ( Table 1; Supplementary Fig. S1B-C). In Mlh3−/−;Pms2−/− mice, gastrointestinal cancers and lymphomas were detected significantly earlier than in Mlh3−/− mice (P < 0.003 and P < 0.001, both Mann-Whitney test; Table 2 ), and lymphomas significantly earlier than in Pms2−/−mice (P < 0.01, Mann-Whitney test; Table 2). In fact, susceptibility to gastrointestinal cancer and lymphoma in Mlh3−/−;Pms2−/− mice was indistinguishable from Mlh1−/− mice on the same 129 Sv/Ev strain background (P = 0.57 for gastrointestinal cancer, P = 0.4 for lymphoma, Mann-Whitney test; Table 2; refs. 13, 22). In contrast, significant differences were seen in mean gastrointestinal tumor occurrence for comparisons of Mlh3−/− versus Mlh3−/−;Pms2−/− (P < 0.003) or Mlh1−/− (P < 0.003), respectively ( Table 2). For lymphomas, significant differences were seen for comparisons of Mlh3−/− versus Mlh3−/−;Pms2−/− (P < 0.001), Mlh1−/− (P < 0.001), and Pms2−/− mice (P < 0.01; Table 2). Significant differences were also seen for comparisons of Mlh1−/− versus Pms2−/− (P < 0.005) and Mlh3−/−;Pms2−/− versus Pms2−/− mice (P < 0.01, all Mann-Whitney test; Table 2). In summary, our data indicate a tumor suppressor role for mouse Mlh3, suggest that Mlh3 and Pms2 both contribute to mismatch repair–mediated tumor suppression, and show that the combination of Mlh3 and Pms2 deficiencies is essentially indistinguishable from Mlh1 deficiency for both tumor spectra and onset.
Mlh3 and Pms2 roles in mismatch repair–mediated mutation avoidance. The association between MLH3 deficiency and microsatellite instability is poorly defined ( 18, 34, 35). Our previous analyses on pooled Mlh3−/− mouse embryonic fibroblasts using “short” (≤14 bp) mononucleotide microsatellite repeat sequences did not detect significant microsatellite instability ( 35). However, short microsatellite repeat sequences in mammalian cells are less susceptible to microsatellite instability mutations than longer repeat sequences ( 14, 36). To assess whether Mlh3 deficiency causes an attenuated mutator phenotype in mononucleotide repeats similar to yeast Mlh3 deletion ( 15), we did more sensitive microsatellite instability assays by making two changes in our experimental procedures: First, we used the more quantitative technique of single-genome microsatellite instability analysis ( 14) and, second, we analyzed longer mononucleotide repeat tracts, A24 and A33, respectively. For Mlh3−/− tail DNA, the microsatellite instability rate was significantly elevated compared with DNA from Wt mice at the A24 (U12235A) and A33 (JH117) mononucleotide repeat tracts (P = 0.015 and P = 0.008, respectively, test of equality of two binomial proportions; Table 3 ).
To test the functional relationship of Mlh3 and Pms2 for mismatch repair–mediated mutation avoidance, we compared microsatellite instability in the double- and single-knockout animals. We found that mutation levels at both An markers in Mlh3−/−;Pms2−/− mice were significantly higher than either Mlh3−/− (P < 0.001) or Pms2−/− (P ≤ 0.003) animals, and were indistinguishable from Mlh1−/− mice (P ≥ 0.19; Table 3). In the Mlh3−/−;Pms2−/− mice, the mutation levels seemed to be greater than additive based on a binomial proportions test (P < 0.02). In all three genotypes, most mutations in the mononucleotide repeat tracts were 1 to 2 bp deletions. In summary, Mlh3 deficiency significantly increases mononucleotide repeat instability but to an extent less than that in Pms2−/− mice. Furthermore, results with the Mlh3−/−;Pms2−/− mice suggest partial overlap of function for mutation avoidance, similar to the results seen in budding yeast ( 15).
Mlh3 and Pms2 roles in mismatch repair–mediated response to DNA damage. Human and/or mouse cell lines deficient in MLH1, PMS2, MSH2, or MSH6 show increased resistance and decreased levels of apoptosis in response to certain DNA-damaging agents ( 37– 43), including the DNA alkylation mimetic drug 6-thioguanine ( 44). To determine the relative importance of mouse Mlh3 and Pms2 in response to alkylation damage, we compared survival of spontaneously immortalized mouse embryonic fibroblasts after 6-thioguanine treatment using a colony formation assay. We found that Mlh3−/− mouse embryonic fibroblasts displayed increased resistance to 6-thioguanine compared with Wt cells (P = 0.007, Student's paired t test; Fig. 4A ). The level of resistance for Mlh3−/− mouse embryonic fibroblasts was not significantly different from Pms2−/− mouse embryonic fibroblasts (P = 0.2) but was less than that observed for Mlh1−/− cells (P < 0.001, Student's paired t test; Fig. 4A). Under the same conditions, Mlh3−/−;Pms2−/− mouse embryonic fibroblasts were more resistant than either Mlh3−/− or Pms2−/− mouse embryonic fibroblasts (P = 0.003 in both comparisons) and indistinguishable from Mlh1−/− mouse embryonic fibroblasts (P = 0.62, Student's paired t test; Fig. 4A).
To characterize the DNA-damage responses of the different genotypes further, we used flow cytometry to analyze cell cycle profile of these mouse embryonic fibroblasts with treatment of 5 μmol/L 6-thioguanine for 72 or 96 hours. Consistent with previous studies, treatment of this high concentration of 6-thioguanine caused a significantly increased sub-G1 cell population (indicative of cellular apoptosis or necrosis) in Wt mouse embryonic fibroblasts by 72 hours ( Fig. 4C; data not shown). After 96 hours of treatment, all mouse embryonic fibroblast lines have significantly increased numbers of sub-G1 cells in response to 6-thioguanine–induced DNA damage. However, mismatch repair–deficient mouse embryonic fibroblasts showed a reduced fraction of sub-G1 cells compared with Wt mouse embryonic fibroblasts indicative of an impaired 6-thioguanine DNA-damage response in mismatch repair–deficient cells ( Fig. 4C and D). Consistent with the colony formation studies, the fraction of sub-G1 cells in Mlh3−/− and Pms2−/− mouse embryonic fibroblasts were similar (P = 0.87), and Mlh3−/−;Pms2−/− mouse embryonic fibroblasts have a sub-G1 population that is indistinguishable from Mlh1−/− mouse embryonic fibroblasts (P = 0.16, t test; Fig. 4C and D). Previous studies in mouse embryonic fibroblast cells have shown that Mlh1 is required for 6-thioguanine–induced G2 cell cycle arrest ( 31). We observed significantly increased G2 arrest in wild-type mouse embryonic fibroblasts treated with 5 μmol/L 6-thioguanine for 72 hours, but this arrest is impaired in mismatch repair–deficient cells ( Fig. 4C and D). Consistent with our other DNA-damage studies, the G2 arrest is less pronounced in Mlh3−/−, Pms2−/−, Mlh3−/−;Pms2−/−, and Mlh1−/− mouse embryonic fibroblasts. Therefore, both Pms2 and Mlh3 contribute to 6-thioguanine–induced G2 cell cycle arrest in mouse embryonic fibroblasts ( Fig. 4C and D). Similar results were seen with treatment of methylating agent N-methyl-N′-nitro-N-nitrosoguanidine (data not shown). In summary, these results show impaired DNA-damage response in the context of either Mlh3 or Pms2 mutations. Comparisons with Mlh3;Pms2 double-deficient or Mlh1-deficient animals indicate that both Mlh3 and Pms2 contribute to mismatch repair-mediated response to DNA alkylation damage, presumably by each functioning as heterodimers with Mlh1 protein.
Mlh1 protein expression in Mlh3−/− and Mlh3−/−;Pms2−/− tumors and cell lines. Previous studies have shown that Pms2 levels are reduced in Mlh1-deficient cells ( 16). To evaluate whether the increased mismatch repair–related phenotypes observed in the Mlh3−/− and Mlh3−/−;Pms2−/− mice and cells might be due to decreased Mlh1 protein abundance, we determined Mlh1 levels in mouse embryonic fibroblast lines by Western blot ( Fig. 4B), and in gastrointestinal tumors and normal bowel by immunohistochemistry ( Fig. 2C-D; Supplementary Fig. S1D). As expected, Mlh1 protein expression was not detectably reduced in Pms2−/− mouse embryonic fibroblasts ( Fig. 4B). Mlh1 protein expression was also not detectably reduced in Mlh3−/− or Mlh3−/−;Pms2−/− mouse embryonic fibroblasts ( Fig. 4B). Similarly, intestinal tumors from Mlh3−/− and Mlh3−/−;Pms2−/− mice showed strong nuclear Mlh1 immunostaining in both tumor and surrounding normal gastrointestinal tissue ( Fig. 2C-D; data not shown), whereas only background immunostaining was present in Mlh1−/− normal gastrointestinal tissue (Supplementary Fig. S1D). These results suggest that the phenotypes determined here for Mlh3 or Mlh3;Pms2 deficiency in the mouse were not secondary effects of decreased Mlh1 protein levels. Additionally, we looked at Pms2 protein levels in the same cell lines. As expected, Pms2 protein levels were much lower in Mlh1−/− cells. However, Pms2 protein levels seem not to be notably different in Mlh3−/− cells ( Fig. 4B).
Among MutL homologues, MLH1 mutations are by far the most common cause of HNPCC, whereas PMS2, PMS1, and MLH3 mutations are rare ( 1, 8, 18, 34, 45, 46). Analyses of the knockout mouse models for all the MutL homologues show that Mlh1-deficient mice develop gastrointestinal tumors, whereas Pms2- or Pms1-deficient animals do not ( 13). These findings in HNPCC patients and the MutL mouse models are somewhat surprising because the MLH1-PMS2 heterodimer is generally accepted as the only significant MutL activity functioning in mismatch repair–related mutation avoidance and DNA-damage responses (for reviews see, refs. 4, 5, 47). A potential explanation for the differences between MLH1 and PMS2 phenotypes is some degree of functional redundancy between the MutL homologues PMS2, MLH3, and PMS1, each of which partner with MLH1 ( 1, 4, 9, 13– 16). Studies in yeast show that pms1 and mlh3 (orthologues of human PMS2 and MLH3, respectively) have partially overlapping functions in mutation avoidance ( 15), which led to the suggestion that MLH3 might perform similar roles in mammalian mismatch repair mediated tumor suppression. To test this possibility and to determine more precisely the consequence of Mlh3 deficiency in the mouse, we compared the phenotypes of animals deficient in Mlh3, Pms2, or both (Mlh3−/−;Pms2−/−) with Mlh1−/− mice (which most likely represent the “mismatch repair-null” phenotype). In brief, we found that Mlh3 deficiency alone has significant consequences for mutation avoidance, DNA-damage response, gastrointestinal and extragastrointestinal tumor susceptibility, and life span. Our findings suggest that mouse Mlh3 and Pms2 proteins both contribute to mismatch repair mechanisms relevant to tumor suppression, presumably as Mlh1/Mlh3 and Mlh1/Pms2 complexes, and that both proteins must be present for complete mismatch repair function in somatic cells.
Mlh3−/− mice showed increased incidence of cancer at later mean age than either Pms2- or Mlh1-deficient mice. Mlh3−/− mice had significantly higher occurrence of both gastrointestinal and extragastrointestinal (largely lymphomas and basal cell carcinoma of skin) cancers compared with Wt mice. The extragastrointestinal tumors were of later onset and with reduced multiplicity in Mlh3-deficient mice compared with either Mlh1- or Pms2-deficient mice. In gastrointestinal epithelium, tumors occurred later in Mlh3−/− compared with Mlh1−/− mice, and as previously reported ( 14) gastrointestinal tumors were not seen in Pms2−/− mice. We found it somewhat surprising that gastrointestinal cancers were seen in Mlh3- but not in Pms2-deficient mice, especially given their respective mutator phenotypes. However, because Pms2-deficient mice died (mainly from lymphomas) earlier than Mlh3-deficient mice, gastrointestinal tumors might not have had adequate time to develop.
Consistent with increased tumor susceptibility, Mlh3−/− mice showed two additional hallmarks of a mismatch repair defect. First, DNA from Mlh3−/− mice had increased levels of mutations in mononucleotide repeat sequences. By using an assay with increased sensitivity, we found that mutation of mononucleotide repeats in Mlh3−/− mice was on average ∼50% of the level seen in Pms2−/− mice, which, in turn, was on average ∼50% of that seen in Mlh1−/− mice. An ∼2-fold difference in mutation avoidance between Mlh1−/− and Pms2−/− mice was well documented in our previous reports ( 13, 14, 31, 48, 49). Second, using colony formation and flow cytometry analyses to assess the DNA-damage response, mouse embryonic fibroblast cells from Mlh3−/− mice displayed increased survival relative to wild-type cells in response to the DNA-damaging agent 6-thioguanine. Therefore, Mlh3 knockout mice show multiple phenotypes, including increased tumor susceptibility, increased microsatellite instability, and decreased responses to induced DNA damage, each of which are characteristic of mismatch repair defects relevant to carcinogenesis. Overall, our results clearly show that Mlh3 deficiency in the mouse causes gastrointestinal and extragastrointestinal cancer susceptibility and defects in mechanisms of mismatch repair tumor suppression. However, the phenotypes of Mlh3-deficient mice were less severe than those of either Pms2- or Mlh1-deficient mice.
Studies in budding yeast suggested that partial functional redundancy between mouse Pms2 and Mlh3 could explain the difference between Mlh1 and Pms2 knockout mice ( 15). Therefore, we compared Mlh3;Pms2 double-deficient mice with Mlh3, Pms2, and Mlh1-deficient mice. Deficiency in both Mlh3 and Pms2 caused a reduced life span compared with the Mlh3 or Pms2 single-deficient mice. Furthermore, the spectrum and timing of gastrointestinal and extragastrointestinal cancer occurrences in the double-deficient mice was similar to that of Mlh1−/− animals ( 13, 22). Consistent with increased tumor susceptibility, the absence of both Mlh3 and Pms2 resulted in mutation levels that were higher than either single-deficient animals and, importantly, similar to Mlh1−/− animals ( 14). When the double-deficient mice were compared with the single-deficient mice, the effect on mutation frequency within mononucleotide repeats is more than additive. These results are consistent with Mlh3 and Pms2 performing partially overlapping roles, presumably when complexed with Mlh1, during repair of single base loop mispairs in mononucleotide repeats. Previous studies have shown that defects in MLH1 or PMS2 can compromise normal DNA-damage responses ( 37– 43). Our data here show that mouse embryonic fibroblast cells from Mlh3−/− and Pms2−/− animals are both more resistant to 6-thioguanine–induced cell death and G2 cell cycle arrest than Wt cells, but less resistant than Mlh1−/− mouse embryonic fibroblasts. Importantly, the Mlh3;Pms2 double-deficient cells are more resistant to killing than either single-mutant, showing a level of resistance indistinguishable from Mlh1-deficient cells, and suggesting that both Mlh3 and Pms2 are important for the 6-thioguanine–induced DNA-damage response. Western and immunohistochemical analyses provided evidence that the observed increased mutator phenotype, impaired DNA-damage response, and cancer susceptibility associated with Mlh3 or Mlh3;Pms2 deficiency was not simply due to decreased protein levels of Mlh1.
Overall, we find that the phenotypes of the Mlh3−/−;Pms2−/− mice are more severe than either Mlh3−/− or Pms2−/− mice and indistinguishable from Mlh1−/− mice. Extending the paradigm developed for the yeast MutL homologues MLH3 and PMS1 ( 15), the stronger phenotype of the Mlh3−/−;Pms2−/− mice suggests partial functional redundancy between Mlh3 and Pms2 for mismatch repair function. Whereas further studies are clearly required, our initial findings suggest that DNA mismatch repair functions that help prevent tumorgenesis require both Pms2 and Mlh3, presumably with each acting as a heterodimer with Mlh1 ( Fig. 5 ). Contributions by both MLH1/MLH3 and MLH1/PMS2 complexes to mismatch repair functions underlying tumor suppression provide insight and an explanation as to why among HNPCC kindreds, only MLH1 mutations are vastly more common than mutations in the other MutL homologues. Finally, our results show that deficiency for Mlh3 alone in the mouse can impair mismatch repair–mediated processes and result in the acceleration of both gastrointestinal and extragastrointestinal cancers. Perhaps, further and more targeted genetic testing will reveal mutation of human MLH3 in certain familial or sporadic cancers.
Grant support: USPHS grants CA87588 and R01CA98626, and American Cancer Society (S.M. Lipkin), R37GM032741 (R.M. Liskay), R01GM045413 (R.M. Liskay), and RO1GM36745 (N. Arnheim).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Shaheen Sikander and Jaquelyn Nguyen for technical assistance and Peter Calabrese for testing the microsatellite instability data for additivity.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received March 7, 2005.
- Revision received June 15, 2005.
- Accepted July 28, 2005.
- ©2005 American Association for Cancer Research.