Requirement of the yeast MSH3 and MSH6 genes for MSH2-dependent genomic stability.

Defects in DNA mismatch repair result in instability of simple repetitive DNA sequences and elevated levels of spontaneous mutability. The human G/T mismatch binding protein, GTBP/p160, has been suggested to have a role in the repair of base-base and single nucleotide insertion-deletion mismatches. Here we examine the role of the yeast GTBP homolog, MSH6, in mismatch repair. We show that both MSH6 and MSH3 genes are essential for normal genomic stability. Interestingly, although mutations in either MSH3 or MSH6 do not cause the extreme microsatellite instability and spontaneous mutability observed in the msh2 mutant, yeast cells harboring null mutations in both the MSH3 and MSH6 genes exhibit microsatellite instability and mutability similar to that in the msh2 mutant. Results from epistasis analyses indicate that MSH2 functions in mismatch repair in conjunction with MSH3 or MSH6 and that MSH3 and MSH6 constitute alternate pathways of MSH2-dependent mismatch repair.

Defects in DNA mismatch repair result in instability of simple repetitive DNA sequences and elevated levels of spontaneous mutability. The human G/T mismatch binding protein, GTBP/p160, has been suggested to have a role in the repair of base-base and single nucleotide insertion-deletion mismatches. Here we examine the role of the yeast GTBP homolog, MSH6, in mismatch repair. We show that both MSH6 and MSH3 genes are essential for normal genomic stability. Interestingly, although mutations in either MSH3 or MSH6 do not cause the extreme microsatellite instability and spontaneous mutability observed in the msh2 mutant, yeast cells harboring null mutations in both the MSH3 and MSH6 genes exhibit microsatellite instability and mutability similar to that in the msh2 mutant. Results from epistasis analyses indicate that MSH2 functions in mismatch repair in conjunction with MSH3 or MSH6 and that MSH3 and MSH6 constitute alternate pathways of MSH2-dependent mismatch repair.
Mutations in the four human mismatch repair genes hMSH2, hMLH1, hPMS1, and hPMS2 are associated with hereditary nonpolyposis colorectal cancer (HNPCC) 1 as well as other cancers (1)(2)(3)(4)(5)(6)(7)(8). Cell lines derived from these tumors are defective in DNA mismatch repair and exhibit increased levels of spontaneous mutations and frequent alterations of microsatellite repeat sequences (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11). An MSH2 homolog, GTBP/P160, has recently been identified in humans. GTBP/P160 exists as a heterodimer with the hMSH2 protein, and this complex binds to a G/T mismatch or to heteroduplexes containing a (dT) 1 or (dT) 3 insertion on one DNA strand (12,13). Cell lines that are defective in GTBP exhibit elevated spontaneous mutability and are defective in the repair of base-base and single nucleotide insertion-deletion mismatches (12,14,15). However, the contribution of mutations in GTBP to hypermutability and cancer predisposition is not entirely clear at present, because two of the colorectal cancer cell lines, DLD-1 and HCT-15 that are mutated in GTBP, also harbor mutations in the 3Ј-5Ј-exonuclease "proofreading" domain of DNA polymerase ␦ (16). Thus, the observed phenotypes in these cell lines could arise from the mutations in GTBP or those in pol␦, or they could be due to the combined effects of the GTBP and pol␦ mutations.
In Saccharomyces cerevisiae, mutations in the mismatch repair genes MSH2, PMS1, or MLH1 result in marked increases in spontaneous mutation rates and microsatellite instability (17)(18)(19)(20). Purified MSH2 protein binds DNA containing mismatched base pairs and insertions (21,22), and PMS1 and MLH1 form a complex, which then binds the MSH2-DNA complex containing a mismatch (23). Mutations in MSH3, another MSH2 homolog, cause a moderate increase in microsatellite instability (24) and have little effect on spontaneous forward mutation rates (25). The yeast homolog of human GTBP has been identified in the yeast genome sequencing project; we refer to it as MSH6.
The manner by which MSH2, MSH3, and MSH6 proteins effect mismatch repair has not yet been determined. Studies with the human MSH2 and GTBP proteins could be interpreted to mean that while MSH2 can recognize loops larger than one nucleotide, MSH2-GTBP heterodimer is more efficient in the recognition of single base mispairs and one-base loops (12)(13)(14). By contrast, recent studies with the yeast MSH3 gene have been interpreted to suggest that MSH2 by itself could recognize single base mispairs (24). Overall, previously published results are in accord with the idea that MSH2 can function in mismatch repair independently of the other MSH proteins. To clarify the roles of the MSH2, MSH3, and MSH6 genes in mismatch repair and to identify the possible functional interaction among them, we have examined the effects of null mutations in these genes on microsatellite instability and spontaneous mutability and have carried out epistasis analyses. Our studies indicate that for its function in mismatch repair, MSH2 requires the MSH3 or the MSH6 gene and that MSH3 and MSH6 constitute alternate MSH2-dependent mismatch repair pathways.

Construction of Null Mutations of the MSH2, MSH3, and MSH6
Genes-Null mutations of MSH2, MSH3, and MSH6 were generated by transformation of yeast strain MS71 (MAT␣ ade5-1 his7-2 trp1-289 ura3-52 CAN1 S ) or its msh⌬ derivatives with the pUC18-derived plasmids pPM600, pEN33, or pBJ271, respectively. All deletion-generating plasmids utilize the URA3 "geneblaster" fragment, which carries the URA3 gene flanked by two identical 1.15-kilobase Salmonella typhimurium HisG sequences. pPM600 contains the MSH2 gene in which nucleotides (nt) ϩ11 to ϩ2279 of the 2898-nt open reading frame (ORF) have been replaced with the URA3 geneblaster fragment. The msh2⌬generating DNA fragment was obtained by digestion with AatII and XbaI. pEN33 contains the MSH3 gene in which nucleotides ϩ638 to ϩ3135 of the 3141-nt MSH3 ORF have been replaced with the URA3 geneblaster fragment. The linear msh3⌬-generating DNA fragment was obtained by digestion with EcoRI. pBJ271 contains PCR fragments of 692 and 751 nt corresponding to the 5Ј-and 3Ј-regions of the MSH6 gene, respectively, separated by the URA3 geneblaster fragment. This plasmid deleted nucleotides ϩ114 to ϩ3637 of the 3726-nt MSH6 ORF. The linear msh6⌬-generating DNA fragment was obtained by digestion with EcoRI. Genomic msh⌬ mutations were generated by the gene replacement method by transformation of yeast to Ura ϩ with the appropriate linear DNA fragments. Deletion mutations were confirmed by Southern (DNA) blot analyses. Loss of the URA3 gene following recom- bination between the HisG sequences was selected for by plating on media containing 5-FOA.
Rates of Repeat Tract Instability-Yeast strain MS71 and msh⌬ mutant strains were transformed with plasmid pSH91 or pSH31 (see "Results") to monitor alterations of plasmid-borne DNA repeats. To monitor alterations in the pSH91 repeat tract, for each strain, 19 independent 100-l cultures were grown in yeast extract/peptone/dextrose (YPD) media, each starting from ϳ10 5-FOA-sensitive cells. Cultures were grown to ϳ10 5 -10 6 cells before being plated onto medium containing 5-FOA. The rates of tract alterations were determined by the method of the median (26). To monitor tract alterations in pSH31, cultures were grown as above but only to ϳ10 3 -10 4 cells. Cells were then plated onto media containing 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-Gal). The rate of tract alterations was determined from the number of cultures with no blue colonies (26).
DNA Sequencing-Tract alterations were examined by two methods. First, plasmid DNA isolated from 5-FOA-resistant colonies was propagated in Escherichia coli and sequenced using double-stranded DNA and the "Ϫ40" primer as described (27). Alternatively, we used a PCR technique similar to that described by Strand et al. (24). Using plasmid DNA isolated from 5-FOA-resistant cells, a single-stranded ϳ130-nucleotide region encompassing the repeat tract was amplified by asymmetric PCR using 35 S-dATP (Amersham Corp.) and the primers 5Ј-CCATTCTAATGTCTGCCCC-3Ј and 5Ј-GTTTTCCCAGTCACGAC-3Ј.
To determine the size of the alterations, the products were compared with labeled PCR products of predetermined length on 6% polyacrylamide denaturing gels.
Rates of Spontaneous Forward Mutation-For each strain, 19 independent cultures were grown in 100 or 500 l of YPD, each starting from ϳ10 canavanine-sensitive cells. Rates of spontaneous forward mutation at the CAN1 S locus were determined from the number of canavanine-resistant colonies by the method of the median (26).

RESULTS
Identification of the Yeast GTBP Homolog-We have identified a yeast homolog (GenBank accession number Z47746) of human GTBP and refer to this gene as MSH6. The two proteins share 27% identical and 50% conserved residues (data not shown). As shown schematically in Fig. 1A, the two protein sequences are colinear without major gaps and the homology extends throughout the entire lengths of MSH6 and GTBP. As shown in Fig. 1B, the homology among MSH2, MSH3, and MSH6 and their E. coli counterpart MutS is particularly striking in the C-terminal portion of these proteins, which also contains the highly conserved Walker type A nucleotide-binding motif. In this region of 107 residues, the MSH proteins share a very high degree of identity among themselves and with the GTBP and MutS proteins.
Tract Instability in msh⌬ Strains-We used two plasmidbased assays (17,27) to examine the effect of mutations in MSH genes on the stability of simple repetitive DNA sequences. Centromeric plasmid pSH91 contains an in-frame 33-base pair (bp) insertion of poly(GT) 16 G in the coding sequence of a hybrid gene containing the yeast URA3 gene. The  Role of MSH3 and MSH6 in Mismatch Repair 7286 pSH91 repeat tract is in-frame with the URA3 gene, resulting in Ura ϩ cells. Alterations of the tract that produce an out-offrame mutation give rise to Ura Ϫ cells that are resistant to 5-FOA. Centromeric plasmid pSH31 contains an out-of-frame 29-bp poly(GT) 14 G insertion in the coding sequence of the E. coli ␤-galactosidase gene. Yeast cells carrying pSH31 do not produce ␤-galactosidase and form white colonies on medium containing X-Gal. Alterations in tract length that restore the correct reading frame result in blue colonies on X-Gal medium.
The msh6⌬ mutation resulted in a 7-fold increase in tract instability in pSH91 and a 6-fold increase in pSH31 (Table I).
The msh3⌬ mutation conferred a 32-fold elevation in tract instability in pSH91 and a 103-fold increase in pSH31, an effect similar to that reported recently (24). By contrast, the msh2⌬ mutation increases tract instability 233-fold in pSH91 and 897-fold in pSH31 (Table I) (17,20). To determine the functional relationship between MSH3 and MSH6, we examined tract instability in strains carrying null mutations in both these genes. Interestingly, in both plasmids pSH91 and pSH31, the rate of tract instability in the msh3⌬ msh6⌬ double mutant was nearly identical to the rate observed in the msh2⌬ strain (Table I). The synergistic increase in the rate of tract destabilization in the msh3⌬ msh6⌬ double mutant over that in the msh3⌬ or the msh6⌬ single mutant indicates that MSH3 and MSH6 provide for alternate pathways for maintaining tract stability. To determine the manner of interaction of MSH3 and MSH6 with MSH2, we examined tract instability in the msh2⌬ msh3⌬, msh2⌬ msh6⌬, and msh2⌬ msh3⌬ msh6⌬ mutant strains. For both plasmid systems, we found the rate of tract instability in these mutants to be the same as that in the msh2⌬ single mutant (Table I). These data indicate that the effect of the msh2⌬ mutation is epistatic to that of the msh3⌬ and msh6⌬ single mutations, as well as to that of the msh3⌬ msh6⌬ double mutation.
Types of Tract Alterations Generated in msh⌬ Strains-To determine the types of repeat alterations generated in various msh⌬ strains, we sequenced the tracts in plasmid pSH91 that had undergone alterations. As has been shown previously, in wild type cells there is a bias for additions of 2 bp with a ϳ15% occurrence of large (Ͼ8 bp) alterations, whereas in the msh2⌬ strain, there is a bias toward accumulation of 2-bp deletions (Table II) (17,20). The msh3⌬ mutant exhibits an even higher bias for 2-bp deletions than the msh2⌬ mutant (Table II). Our results with msh3 support the recently reported observations of Strand et al. (24), who found a 2-fold bias in favor of 2-bp deletions over 2-bp additions in the msh2⌬ strain but a 6-fold bias in favor of 2-bp deletions in the msh3⌬ strain. If our data for msh2⌬ and msh3⌬ mutants are pooled with those of Strand et al. (24), then the number of Ϫ2 and ϩ2 tracts is 73 and 29, respectively, in the msh2⌬ strains, and 87 and 14, respectively, in the msh3⌬ strain. The pooled data confirm that the msh3⌬ mutant accumulates a significantly higher proportion of 2-bp deletions than does the msh2⌬ mutant ( 2 for 1 degree of freedom ϭ 6.5; p ϳ 0.01). The msh6⌬ mutation does not appear to alter the pattern of insertions and deletions in comparison with wild type. The msh3⌬ msh6⌬ double mutant strain, however, exhibits a pattern of repeat length alterations that resembles the msh2⌬ pattern (Table II). The msh2⌬ msh3⌬, msh2⌬ msh6⌬, and msh2⌬ msh3⌬ msh6⌬ mutant strains exhibit a pattern of repeat length alterations similar to that observed in the msh2⌬ strain (Table II), providing further evidence for an epistatic effect of the msh2⌬ mutation on the msh3⌬ and msh6⌬ mutations.
Spontaneous Mutation Rates in msh⌬ Strains-To determine the role of MSH genes in spontaneous mutability, we examined the rates of spontaneous forward mutations at the CAN1 S locus in the various msh⌬ strains. In contrast to its modest effect on microsatellite instability, the msh6⌬ mutation caused a 12-fold increase in the rate of spontaneous forward mutation to canavanine resistance (can1 r ) ( Table III). The msh3⌬ mutation, however, resulted in only a 3-fold increase in the spontaneous forward mutation rate. The rate of can1 r mutation in the msh3⌬ msh6⌬ double mutant strain increased to a level similar to that in the msh2⌬ strain (Table III), indicating that, as for microsatellite stability, both genes have a role in maintaining normal levels of spontaneous mutability. However, while MSH3 appears to have a more predominant role than MSH6 in microsatellite stability, MSH6 contributes more significantly to normal spontaneous mutability. The rate of formation of can1 r mutations in the msh2⌬ msh3⌬, msh2⌬ msh6⌬, and msh2⌬ msh3⌬ msh6⌬ mutant strains was the same as that in the msh2⌬ mutant alone (Table III), indicating an epistatic effect of the msh2⌬ mutation on the msh3⌬ and the msh6⌬ mutations. DISCUSSION Our results indicate that MSH3 and MSH6 are both required for maintaining wild type levels of tract stability and spontaneous mutability in yeast cells; MSH3, however, has a more   (Fig. 2). The MSH2-MSH3 complex is more efficient at removing 2-4-bp insertions and deletions, while the MSH2-MSH6 complex is more efficient at removing single bp mismatches. Thus, even though purified MSH2 protein from both humans and yeast has been shown to bind DNA containing a G/T mismatch or insertion-deletion loop-type mismatches of up to 14 nucleotides (21,22), our results predict that complex formation with MSH3 or MSH6 is obligatory for the action of MSH2 in mismatch repair. Mutations in human MSH2, MLH1, PMS1, and PMS2 account for the majority of cancers in HNPCC kindreds. However, a large proportion of sporadic colon cancers and other types of cancer do not have mutations in these genes, and no germ line GTBP mutations have been identified in HNPCC kindreds that harbored no mutations in these four mismatch repair genes (14). Our results with the yeast MSH3 and MSH6 genes would suggest that mutations in the human MSH3 or the GTBP gene are unlikely to cause as severe a defect in mismatch repair as do mutations in MSH2. Mutational inactivation of both hMSH3 and GTBP, however, should result in increased microsatellite instability, hypermutability, and cancer predisposition characteristic of mutations in the hMSH2 gene. Role of MSH3 and MSH6 in Mismatch Repair 7288