The DNA Mismatch Repair Genes Msh3 and Msh6 Cooperate in Intestinal Tumor Suppression¹ Free

HNPCC³is an autosomal dominant disorder, and individuals with HNPCC have a predisposition to develop tumors of a number of organ systems,including the colon. Tumors from HNPCC patients have a DNA MMR defect,and this defect is manifested by an instability of simple nucleotide repeat sequences (microsatellite instability; Ref. 1). Germline mutations in mammalian homologues of bacterial methyl-directed MMR genes MutS and MutL (MSH2 and MLH1,respectively; reviewed in Ref. 2) have been shown to be responsible for a large proportion of HNPCC cases. Similarly,MMR-defective sporadic colorectal cancer is primarily due to somatic inactivation of MLH1 or MSH2.

In human cells, the two protein complexes consisting of MSH2-MSH3 and MSH2-MSH6 appear to be responsible for the recognition of mispaired bases during MMR (3, 4). The MSH2-MSH6 complex recognizes single nucleotide and small insertion/deletion mismatches, and the MSH2-MSH3 complex recognizes small insertion/deletion mismatches (3, 4, 5, 6). Both of these complexes interact with a complex of two MutL-related proteins, MLH1-PMS2, as well as other MMR proteins. These observations suggest that mutations in MSH3 or MSH6 alone or in combination would result in MMR defects and that, therefore, these genes could be involved in cancer predisposition (7). To test these predictions, we generated mice with mutations in the Msh6 and Msh3 genes and examined their phenotypes.

We have shown that mice carrying an Msh6 mutation have a cancer predisposition phenotype, and it is now established that certain HNPCC cases are the result of germline mutations in this gene (8, 9, 10). We now show that Msh3−/− mice have defects in the repair of insertion/deletion mismatches, develop tumors late in life, and when combined with Msh6−/− mice, show a phenotype that is indistinguishable from MLH1 or MSH2 deficiency. These results provide an explanation as to why germline mutations in MSH3 have not been detected in HNPCC families. On the basis of these results, we suggest that MSH3 cooperates with MSH6 in tumor suppression.

A genomic Msh3 fragment containing exons 4, 5, 6, and 7 was obtained by screening a mouse genomic Charon 35, 129/Ola phage library with the pDHH8 probe corresponding to the Msh3 promoter region (see Ref. 11). The 6.5-kb genomic fragment was subcloned into the EcoRI site of pcDNA II (Invitrogen). A 1.7-kb AatII-SacI PGKneo fragment was blunted with T4 polymerase and cloned into the ClaI restriction site(nucleotide position 342 of the coding region) also blunted with T4 polymerase at codon 314 in exon 7 of the 1091-amino acid MSH3 coding region. Finally, a 1.1-kb XhoI/SalI fragment from MC1tk was treated with T4 polymerase and cloned into the T4 polymerase-blunted SacI restriction site located 3′ to exon 7.

The targeting vector pMsh3NTK (50 μg) was linearized at the single NotI site, electroporated into 5.5 × 10⁷ WW6 ES cells, and selected with G418 (150μg/ml) and ganciclovir (2 μm) as described previously. Colonies were picked after 9 days, and their DNA was screened by PCR using forward primer A (5′-CATCTCAGTAGCATCTCACC-3′) and reverse primer B (5′-TGGAAGGATTGGAGCTACGG-3′). The reaction was performed using the Long Range PCR kit (Boehringer Mannheim) according to the manufacturer’s specification. Positive ES cell clones were identified by an 8.8-kb PCR fragment specific for the targeting event. Five positive cell lines were identified among 232 doubly positive colonies, and the correct targeting event was shown by BglII digestion of high molecular weight DNA and Southern blot analysis using a 1.6 kb KpnI probe directed at the intron 8 region of the Msh3 gene. A total of five colonies, designated MSH3-48,-127, -130, -148, and -153, had the correct targeting event.

Two colonies, MSH3-127 and MSH3-130 were injected into C57Bl/6 blastocysts, and the chimeric offspring were tested for germline transmission. All chimeras tested transmitted the mutation through the germline. F₁ heterozygotes were interbred to obtain homozygous Msh3 mutant animals. To obtain Msh3;Msh6 double-mutant animals, Msh3and Msh6 mutant mice were intercrossed. All of the mice used for this study were of a mixed genetic background and had an estimated genetic composition of 60% C57Bl/6, 37.5% 129/Sv, and 2.5% SJL/J.

For Northern blot analysis, Poly(A)⁺ RNA (5 μg)from adult liver was separated on 1.0% agarose formaldehyde gels,transferred to nitrocellulose, and hybridized with an Msh3probe corresponding to nucleotides 940-2234 (including exons 7–17) of the coding region and a human Actb (β-actin) probe. For RT-PCR analysis, poly(A) RNA (1 μg) was treated with DNase I and subjected to RT-PCR analysis using the Titan RT-PCR kit (Boehringer Mannheim) according to the manufacturer’s specification. For the Dhfr control reaction, primer A (5′-TGGTTCGACCATTGAACTGC-3′)and primer B (5′-CTTGCCAATTCCGGTTGTTC-3′), spanning nucleotides 2–322 of the coding region, were used. Dhfr expression was indicated by a 321-bp fragment. For the Msh3 reaction,primer A (5′-AAGAAGGGGAACCTTTCCGT-3′) and primer B(5′-TTGGTGGCTCTTTGGATGAG-3′) spanning exons 7 and 8, were used. Wild-type Msh3 mRNA was indicated by a 322-bp fragment.

DNA was extracted from tumor tissue and subjected to PCR. Four end-labeled primer pairs were used to amplify sequences containing dinucleotide repeats (D1 Mit36, D7 Mit91, D10 Mit2, and D14 Mit15; Ref. 12), and four others were used to amplify sequences containing mononucleotide repeats [JH 102, JH103, and JH104; primer sequences were as follows: JH102F, 5′-CATTTCTCTGGGATCGCCTT-3′; JH102R,5′-CCCGCCTTTGATTCCTTTGT-3′; JH103F,5′-TTCATCAGTCTTCTGGCTCC-3′; JH103R,5′-AGTGGTGATAGCAGCTTAGC-3′; JH104F, 5′-AGGTGATTGTAACGGGCATC-3′; and JH104R, 5′-TATCCTCTCAGTGGTGAGTG-3′ (8). Primer sequences for locus U12235 were as follows: P1, 5′-GCTCATCTTCGTTCCCTGTC-3′; and P3, 5′-CATTCGGTGGAAAGCTCTGA-3′ (13)]. Amplified PCR products were separated on a denaturing polyacrylamide gel and autoradiographed for analysis. Instability was expressed as the percentage of reactions that yielded an abnormal size product.

Procedures for extract and heteroduplex preparation and for measuring repair activity were as described (14, 15, 16).

Tumors from sacrificed mice were removed and fixed in 10%neutral-buffered formalin. The GI tract was opened longitudinally and examined under a dissecting microscope for the presence of tumors. All tumors of the GI tract and other tissues were processed for paraffin embedding, and sections were prepared for staining with H&E. Statistical analysis of tumor incidence were performed with the Mann-Whitney test.

The mouse homologue of human MSH3, Msh3, originally designated as Rep-1(Rep-3), was cloned in 1989 (11). Different Msh3 transcripts that result from alternative splicing have been reported, although the functional significance of these different transcripts has not been established. All of the known transcripts of this gene contain exon 7 sequences (17). To obtain a null mutation at this locus, we constructed a gene-targeting vector designated as pMsh3ex7 (Fig. 1,a) that was transfected into WW6 mouse ES cells. Five of 232 clones examined yielded the 8.3-kb PCR product that was expected from appropriately modified cells. The targeting events were confirmed by Southern blot analysis, and cells from two of these colonies,designated MSH3-127 and MSH3-130, were used to generate mice carrying the mutation (Fig. 1 b). We were able to obtain homozygous Msh3 mutant mice that were viable and reproduced normally.

To examine the effect of the genetic modification on gene expression, we examined RNA from mutant mice by Northern blot analysis. Poly(A)⁺ RNA from wild-type and heterozygous mutant mice contained a detectable Msh3transcript that was absent in tissue from homozygous mice (Fig. 1,c). The absence of a wild-type Msh3 transcript without the neomycin insertion in exon 7 in Msh3−/− mice was further confirmed by RT-PCR analysis of RNA with a pair of primers that detect a transcript spanning exons 7 and 8 (Fig. 1 d). In this assay, wild-type transcripts were detected in RNA from heterozygous and wild-type animals but were not found in homozygous mutant animals, indicating inactivation of the Msh3 gene.

Microsatellite instability analysis and biochemical MMR assays performed on MSH2- and MLH1-deficient mice indicated that they were incapable of repairing single-nucleotide as well as insertion/deletion mismatches.⁴In contrast, cells from Msh6−/− mice were incapable of repairing single nucleotide mismatches but have a robust repair activity of insertion/deletion mismatches (8). To ascertain the effect of the Msh3 mutation on repair activity, cellular extracts from different Msh3 mutant genotypes were examined for nick-directed MMR. Extracts from wild-type cells repaired all mismatches tested (Fig. 2 and Refs. 8, 18). Extracts of Msh3−/− cells showed only a partial repair defect. Five different base-base mismatches and two single-base insertion/deletion mismatches were repaired, whereas a third single-base insertion/deletion mismatch and three different substrates containing two extra nucleotides were not efficiently repaired (Fig. 2). To ascertain whether loss of both MSH3 as well as MSH6 would lead to a MMR deficiency that is comparable to that observed in Msh2−/− or Mlh1−/− cells, we mated the Msh3 mutant mice with Msh6 mutant mice. Results from analysis of extracts prepared from Msh3−/−;Msh6−/− as well as Msh6−/− cells are shown in Fig. 2. Compared with the wild type, the MSH3/MSH6 double-mutant cell extracts failed to correct any mismatch examined. These results are compatible with the proposed roles of MSH3 and MSH6 in the DNA MMR process (3, 4, 5, 6, 7, 19, 20).

To study the effects of the absence of MSH3, we examined the survival of Msh3+/+ and Msh3−/− mice (Fig. 3). Msh3−/− mice had a 50% survival time of 22 months, and their survival was not significantly different from that of the wild-type control animals (P = 0.0879,log-rank test). When Msh3−/− mice were compared with Msh6−/− mice, we noted significant differences. The Msh6−/− mice had a 50% survival time of 11 months, which is significantly different from that of Msh3−/− mice(P < 0.0001, log-rank test; Fig. 3). These observations suggest that the presence of MSH6 protein is more critical to the survival of mice than MSH3. To definitively address the issue of whether MSH3 has any role in survival, we examined Msh3−/−;Msh6−/− double-mutant mice. The survival of the double mutants was significantly lower than the Msh3−/− or Msh6−/− single mutants (50%survival age of 6 months: P < 0.0001,log-rank test). These results show that although MSH3 deficiency alone does not have a significant effect, it does play a critical role in survival in combination with an MSH6 deficiency.

To determine whether Msh3−/− mice show tumor susceptibility, we examined 2-year-old mice for tumors (Table 1). Msh3+/+ as well as Msh3−/− mice developed tumors at this old age with a similar incidence. However, the Msh3−/− mice developed a few GI tumors that were also seen in Msh2−/−, Mlh1−/−, and Msh6−/−mice (8, 18, 21, 22, 23) but not in wild-type mice. Because of the small number of tumors detected in Msh3−/− mice, it was not possible to definitely conclude that the absence of MSH3 causes tumor susceptibility.

Because the combination of Msh3 and Msh6mutations severely reduces survival and Msh3−/−;Msh6−/− double-mutant mice have survival rates comparable to Mlh1−/− or Msh2−/− mice, we examined the double mutants for tumor susceptibility. As shown in Fig. 3, 50% of the double-mutant mice become moribund or morbid by 5–6 months of age. Among the 16 mice examined, all had at least one tumor. The tumors were predominantly GI tumors or non-Hodgkin’s lymphomas (Table 1 and Fig. 4). This tumor spectrum is very similar to that seen in Msh6−/− or Mlh1−/− mice examined in our laboratory or published for Msh2−/− mice (21, 22). When the number of GI tumors per mouse was ascertained,wild-type mice had no tumors, Msh3−/− mice had 0.13 ± 0.07 tumors, Msh6−/− mice had 0.62 ± 0.24 tumors, and the double-mutant Msh3−/−;Msh6−/− mice had 2.06 ± 0.45 tumors. The tumors in the double mutants presented at a much younger age, and the number was significantly greater than those seen in Msh3−/− (P < 0.001) and Msh6−/− (P < 0.05) mice.

We examined tumors from Msh3−/− and Msh3−/−;Msh6−/− mutant mice for microsatellite instability. Markers corresponding to mononucleotide as well as dinucleotide repeats were used for the analysis. In Msh3−/− tumors, 8 of 35 reactions (22.9%) revealed novel size bands. The frequency of mononucleotide and dinucleotide instability in these tumors was 2 of 15 (13.3%) and 6 of 20 (30.0%),respectively. In tumors derived from Msh3−/−;Msh6−/− mice, 13 of 33 reactions(39.4%) showed microsatellite instability, and 4 of 15 mononucleotide repeats (26.7%) and 9 of 18 dinucleotide repeats (50.0%) showed novel-sized bands. It should be noted that the Msh3−/−tumors were derived from mice that were >18 months, whereas the Msh3−/−;Msh6−/− tumors were from mice no older than 7–8 months, and the difference in the ages of the tumors might complicate the direct comparison of the microsatellite instability frequencies.

Single-nucleotide mismatches and small insertion/deletion mismatches are repaired by a complex of proteins that contains MSH2,MLH1, PMS2, and either MSH3 or MSH6. Germline mutations in all of the genes that encode for these proteins except that encoding for MSH3 were detected in HNPCC patients. To ascertain the role of MSH3, we generated mice with a mutation in this gene.

The Msh3 mutant mice contain an insertion into exon 7 of the gene. The insertion introduced multiple stop codons in all three possible reading frames. Because exon 7 is present in all known alternately spliced forms of Msh3 mRNA, no functional mRNAs for Msh3 should be produced in these mice. Consistent with this, no functional MSH3 mRNA was expressed in Msh3−/− cells. Although no direct protein data are available, our data are consistent with the absence of a functional MSH3 protein in Msh3−/− mice. Cell extracts from the Msh3−/− mice had the ability to repair single-nucleotide mismatches and single-nucleotide insertion/deletion mismatches but were defective in repair of larger insertion/deletion mismatches. Earlier results with extracts of Msh6−/− cells showed that these extracts could not repair base-base mismatches but could efficiently repair insertion/deletion mismatches (8). Extracts of the Msh3−/−;Msh6−/− double-mutant cells were defective in repair of all single-nucleotide mispairs and insertion/deletion mispairs tested. Collectively, these data are consistent with a model [reviewed by Kolodner (7)] in which base-base mismatches are primarily repaired by the MSH2-MSH6 complex, whereas mismatches involving two or more unpaired nucleotides are repaired by the MSH2-MSH3 complex, and each of the two complexes can function in the repair of single-base insertion/deletion mismatches. The data generally are consistent with many observations on the repair activity of the yeast and human MutS complexes and are consistent with the view that the Msh3 mutation constructed here is a null allele.

Our results show that MSH3 does play a role in suppression of tumorigenesis. Msh3−/− mice developed a few GI tumors that were similar to those seen in other MMR-deficient mice. The development of tumors did not result in increased morbidity, with 50%of the Msh3−/− mice being alive at 22 months of age compared with 11 months for Msh6−/− mice and 6–7 months for Mlh1−/− mice (Fig. 3 and Refs. 23, 24).

While this manuscript was under review, another study by de Wind et al. (25) described the results for Msh3 and Msh6 mutant mice on a different genetic background. It is of interest to note that the survival time of these Msh6 mutant mice was shorter compared with the survival of our Msh6 mutant mouse line used in this study and that was reported earlier (8). The shorter survival seems to be caused by the development of predominantly lymphoid tumors in these animals. Another difference in the tumor spectrum is the presence of gynecological tumors and a relatively low number of intestinal tumors in the Msh6 and Msh3/Msh6 mutant mice. Although an increase in the intestinal tumor number in the double-mutant Msh3/Msh6animals was observed by these authors, the small number of intestinal tumors in these mouse lines makes the assessment of the roles of Msh3 and Msh6 in intestinal tumorigenesis difficult. These differences between the mouse lines also indicate the influence of genetic modifiers on cancer susceptibility caused by MMR deficiency.

No germline mutations in MSH3 were discovered in HNPCC families. The observations presented in this report are consistent with this observation and may provide an explanation for the lack of germline MSH3 mutations in HNPCC families. If the age of onset of tumors in Msh3−/− mice was translated to human age, it would correspond to a relatively old age. If families with MSH3 mutations exist, the predicted late onset of tumorigenesis and the low tumor incidence would make the identification of such a familial cancer syndrome difficult. The tumor susceptibility profile would likely not satisfy the Amsterdam criteria for HNPCC, at least with regard to age of onset. Why would a Msh3 mutation lead to a much later time of onset of tumorigenesis compared with Msh6 mutant mice? A simple explanation is that the dinucleotide and larger unit repeat sequences that are the major target of mutagenesis in Msh3 mutants are relatively infrequent in coding sequences. This is in contrast to the unique sequences and mononucleotide runs that are the targets of mutagenesis in Msh6 mutants. Thus,defects in MSH3 would produce lower rates of mutations in tumor suppressor genes and proto-oncogenes than the defects caused by a mutation in Msh6. The differences in mutagen sensitivity between Msh3−/− and Msh6−/− cells (25) might also provide an explanation for these differences.

More definitive evidence for the role of MSH3 in tumorigenesis was obtained by combining Msh3 and Msh6mutations. The combination of Msh3−/− and Msh6−/− mutations resulted in a much greater decrease in survival rate compared with each of the individual mutations and was comparable to that seen in Msh2−/− and Mlh1−/− mutant mice (21, 22, 23, 24). Tumors from the double-mutant mice also showed a high degree of microsatellite instability, making these double-mutant mice truly analogous to MSH2-or MLH1-deficient mice. The types of tumors seen in these double-mutant mice were very similar to those seen in Msh2−/− and Mlh1−/− mice. Although the Msh3 and Msh6 mutant animals used in this study have a different genetic background than those of the published Msh2 mutant animals, the overall survival rate and the types of GI cancers and lymphomas appear to be the same. These results suggest that MSH3 and MSH6 cooperate in tumor suppression and that inactivation of MSH3 together with MSH6 would cause a susceptibility to intestinal tumors that is indistinguishable from those seen in classic HNPCC patients.

The results from several different laboratories now provide explanations for the roles of all of the mismatch recognition proteins. Because MSH2 is the common subunit of the mismatch recognition complexes, its absence leads to a severe MMR deficiency that affects repair of both single-nucleotide and small insertion/deletion mismatches. The role of MLH1 parallels that of MSH2 in that it is also the common subunit of two different MutL-related complexes and the phenotypes of individuals or mice with mutations in MSH2 and MLH1 are indistinguishable. MSH6 is also a tumor suppressor gene, although inactivation of MSH6 yields a slightly less severe phenotype than that caused by MSH2 or MLH1 mutations because in MSH6-deficient cells, one class of mismatches, insertions/deletions,apparently are repaired normally. PMS2 is a partner for MLH1, and its deficiency in mice leads to tumor susceptibility, although no GI tumors were observed (23, 26). Combining Pms2 with an Apc mutation (27) does indeed result in increased tumor susceptibility in the GI tract. Results presented in this report show that MSH3 also plays a cooperating role, especially with MSH6, in tumorigenesis.

We thank Harry Hou, Jr. for blastocyst injections.