Copyright  1998 by the Genetics Society of America

Functional Overlap in Mismatch Repair by Human MSH3 and MSH6

Asad Umar,* John I. Risinger,†,§ Warren E. Glaab,**,‡ Kenneth R. Tindall,‡ J. Carl Barrett†,§ and Thomas A. Kunkel* *Laboratory of Molecular Genetics, †Laboratory of Molecular Carcinogenesis, ‡Laboratory of Environmental Carcinogenesis and Mutagenesis, National Institute of Environmental Health Sciences, Research Triangle Park, North Carolina 27709 and §Curriculum in Genetics and Molecular Biology, **Curriculum in Toxicology, University of North Carolina, Chapel Hill, North Carolina 27599

ABSTRACT Three human genes, hMSH2, hMSH3, and hMSH6, are homologues of the bacterial MutS gene whose products bind DNA mismatches to initiate strand-specific repair of DNA replication errors. Several studies suggest that a complex of hMSH2·hMSH6 (hMutS␣) functions primarily in repair of base·base mismatches or single extra bases, whereas a hMSH2·hMSH3 complex (hMutS␤) functions chiefly in repair of hetero- duplexes containing two to four extra bases. In the present study, we compare results with a tumor cell line (HHUA) that is mutant in both hMSH3 and hMSH6 to results with derivative clones containing either wild-type hMSH3 or wild-type hMSH6, introduced by microcell-mediated transfer of 5 or 2, respectively. HHUA cells exhibit marked instability at 12 different loci composed of repeat units of 1 to 4 base pairs. Compared to normal cells, HHUA cells have mutation rates at the HPRT locus that are elevated 500-fold for base substitutions and 2400-fold for single-base frameshifts. Extracts of HHUA cells are defective in strand-specific repair of substrates containing base·base mismatches or 1–4 extra bases. Transfer of either (hMSH3)or2(hMSH6) into HHUA cells partially corrects instability at the microsatellite loci and also the substitution and frameshift mutator phenotypes at the HPRT locus. Extracts of these lines can repair some, but not all, heteroduplexes. The combined mutation rate and mismatch repair specificity data suggest that both hMSH3 and hMSH6 can independently partici- pate in repair of replication errors containing base·base mismatches or 1–4 extra bases. Thus, these two gene products share redundant roles in controlling mutation rates in human cells.

HE pace of studies to identify and determine the in the strand-specific repair of substrates containing mis- Tfunctions of DNA mismatch repair (MMR) genes matches or extra bases (Umar et al. 1994a,b, 1997; in humans accelerated when mutations in the hMSH2 Boyer et al. 1995; Drummond et al. 1995; Li et al. 1995; gene were first linked to hereditary colon cancer Risinger et al. 1995). hMSH2-mutant cells are also more (Fishel et al. 1993; Leach et al. 1993). This gene is one resistant to killing by certainDNA-damaging agents than of several (reviewed in Fishel and Wilson 1997) that are MMR-proficient cells (Risinger et al. 1995; Umar et share with the bacterial mismatch al. 1997). Transfer of chromosome 2 from a repair- repair gene MutS, whose product binds to mismatches proficient cell to an hMSH2-mutant cell lowers the muta- and initiates their repair. Human MSH2 is located on tion rate, stabilizes , and sensitizes the the short arm of chromosome 2 and encodes a protein cells to MNNG (Umar et al. 1997) and cisplatin (Fink et of 105 kD. The human (Fishel et al. 1994a,b) and yeast al. 1996). MMR activity is restored to extracts of hMSH2- (Alani et al. 1995) MSH2 proteins bind to DNA con- mutant cells both by chromosome 2 transfer (Umar et al. taining base·base mispairs and to substrates containing 1997) and by addition of the hMutS␣ protein complex from one to 14 extra bases. Tumor cell lines with mutant (Drummond et al. 1995). hMutS␣ is a heterodimer of hMSH2 genes have elevated spontaneous mutation rates hMSH2 and the 160-kD product of the gene located on in endogenous genes (Branch et al. 1995; Malkhosyan chromosome 2p that is homologous to the yeast MSH6 et al. 1996; Richards et al. 1997; Umar et al. 1997) and gene.1 Efficient binding of hMutS␣ to a G·T mismatch exhibit microsatellite instability (Orth et al. 1994; Umar requires both hMSH2 and hMSH6 proteins (Palombo et al. 1994a; Boyer et al. 1995; Risinger et al. 1995; et al. 1995), and yeast MutS␣ likewise binds to this mis- Shibata et al. 1995). Similarly, mutant yeast strains match (Iaccarino et al. 1996). Human and yeast MutS␣ also have strongly elevated mutation rates (see below bind to substrates containing extra nucleotides in one and, for review, see Sia et al. 1997b). Studies in vitro strand (Drummond et al. 1995; Alani 1996; Palombo reveal that extracts of hMSH2-mutant cells are defective

1 The human gene is also known as p160 (Drummond et al. 1995) and GTBP (Palombo et al. 1995; Papadopolous et al. 1995), the latter Corresponding author: Thomas A. Kunkel, Laboratory of Molecular for G ·T mismatch binding protein, the property upon which the gene Genetics, E3-01, National Institute of Environmental Health Sciences, product was first identified (Hughes and Jiricny 1992). For simplicity P.O. Box 12233, Research Triangle Park, NC 27709. and given its homology to the yeast MSH6 gene, we refer to the human E-mail: [email protected] gene as hMSH6 and its gene product as hMSH6 protein.

Genetics 148: 1637–1646 (April, 1998) 1638 A. Umar et al. et al. 1996; Alani et al. 1997) and to a variety of damaged in vitro (Marsischky et al. 1996). Recombinant yeast DNA substrates (Duckett et al. 1996; Li et al. 1996; MSH2 and MSH3 proteins form a stable heterodimer Mello et al. 1996; Mu et al. 1997). that has low affinity for binding to a G·T mismatch These data strongly suggest that MSH2 is an essential but has higher binding affinity for insertion/deletion protein for the initial mismatch recognition step in the mismatches involving 2–14 extra bases (Habraken et al. strand-specific repair of replication errors in eukaryotic 1996). These data support the hypothesis that at least cells (for review, see Jiricny 1996; Kolodner 1996; two different MSH2-dependent, strand-specific MMR Modrich and Lahue 1996; Umar and Kunkel 1996). processes exist in yeast. The presence of hMSH6 in the hMutS␣ complex sug- Evidence that MSH3 functions in strand-specific MMR gests that it too functions in mismatch repair. In support in human cells comes from studies of a human endome- of this hypothesis, hMSH6-mutant human tumor cells trial tumor cell line, HHUA (Risinger et al. 1996). The are mutators (Bhattacharyya et al. 1994; Branch et hMSH3 gene is on chromosome 5 and encodes a protein al. 1995; Malkhosyan et al. 1996; Glaab and Tindall of 1137 amino acids (reviewed in Fishel and Wilson, 1997; Ohzeki et al. 1997) and are resistant to killing by 1997). HHUA cells harbor a single base deletion muta- alkylating agents (Kat et al. 1993; Branch et al. 1995; tion in hMSH3 and the MSH3 protein is missing 723 Umar et al. 1997) and 6-TG (W. E. Glaab, J. I. Risinger, C-terminal amino acids, many of which are conserved A. Umar, J. C. Barrett, T. A. Kunkel, unpublished among eukaryotic MutS homologues. Microsatellites are results). Extracts of these cells are also deficient in MMR highly unstable in HHUA cells, and an extract of HHUA activity (Kat et al. 1993; Umar et al. 1994a, 1997; Drum- cells is unable to repair base·base or insertion/deletion mond et al. 1995; Risinger et al. 1996). All three of these mismatches (Risinger et al. 1996). Furthermore, trans- phenotypes are reversed by transferring chromosome 2 fer of chromosome 5 from a repair-proficient cell to from a repair-proficient cell into an hMSH6-mutant cell HHUA cells stabilizes microsatellites, and extracts of (Umar et al. 1997; W. E. Glaab, J. I. Risinger, A. Umar, these cells exhibit strand-specific repair of one- and J. C. Barrett, T. A. Kunkel and K. R. Tindall, unpub- four-base deletion mismatches (Risinger et al. 1996). lished results). However, the degree of microsatellite Consistent with these results, recombinant human MSH2 instability in hMSH6-mutant cells is not as high as in and MSH3 form a stable heterodimer that binds to one- hMSH2-mutantcells (Bhattacharyya et al. 1995; Papa- to four-base insertion/deletion mismatches (Palombo dopoulos et al. 1995; Shibata et al. 1995). Also, studies et al. 1996). Collectively, the data strongly suggest a in yeast (Strand et al. 1993; Johnson et al. 1996; Mar- functional role for human hMSH3 in MMR repair. sischky et al. 1996; Sia et al. 1997a) clearly show that, In our initial study (Risinger et al. 1996), HHUA cells although a null mutation in MSH2 strongly elevates the containing wild-type chromosome 5 still exhibited a de- rate of a variety of mutations, a null mutation in MSH6 tectable level of microsatellite instability, and extracts has a smaller effect on the rate of some frameshift muta- of these cells still lacked MMR activity for base·base tions and a negligible effect on others. Moreover, ex- mismatches and some insertion mismatches. These re- tracts of MSH6-mutant human tumor cells are only par- sults suggested that HHUA cells were mutant in another tially repair-deficient. Although they fail to repair single mismatch repair gene. Subsequent analysis revealed a base mismatches, they do retain the ability to repair mutation in hMSH6 that was consistent with loss of func- some insertion/deletion mismatches (Drummond et al. tion (see below). The fact that HHUA cells are mutant 1995; Risinger et al. 1996; Umar et al. 1997). Thus, in both hMSH3 and hMSH6 provides the opportunity repair of some mismatches may be initiated by hMSH2 to investigate the roles of these two genes in human alone or in complex with another protein. cells by studying each in the absence of the other. To A candidate for this other protein is MSH3, a third do so, we have introduced chromosome 2 into HHUA eukaryotic MutS homologue. Yeast strains containing a cells to allow direct comparison of a hMSH3/hMSH6 null mutation in MSH3 are mutators (Johnson et al. double mutant (HHUA) to derivatives corrected for one 1996; Marsischky et al. 1996; Sia et al. 1997a), indicat- or the other MMR gene (HHUA plus chromosome 2 ing a role for this gene in MMR. Several studies of yeast or chromosome 5 ). Here we describe three properties strains containing MSH2, MSH3, and MSH6 mutations of these three cell lines: microsatellite instability and (Strand et al. 1995; Johnson et al. 1996; Marsischky HPRT gene mutation rates in vivo and MMR capacity et al. 1996; Greene and Jinks-Robertson 1997; Sia et in vitro. The results indicate that hMSH3 and hMSH6 al. 1997a) suggest that MSH2 and MSH6 are primarily share redundant roles in human cells. responsible for repairing base·base mismatches. MSH2 can participate with either MSH6 or MSH3 to repair one- and two-base insertion/deletion mispairs, whereas MATERIALS AND METHODS repair of insertion/deletions involving repeating units Cell lines, analysis of microsatellite instability, and sequence of four to 16 bases requires MSH2 and MSH3 but not analysis of hMSH6: The source and growth of cell lines is MSH6. Yeast MSH2 protein has been shown to physically given in Risinger et al. (1996). The method used to analyze interact with both MSH6 and MSH3 proteins translated microsatellite stability in single cell clones has been described Mismatch Repair Specificity of hMSH3 and hMSH6 1639

(Risinger et al. 1996). Sequence analysis of hMSH6 was per- mismatched substrates and procedures for measuring mis- formed as described previously (Risinger et al. 1996), using match repair activity have been described in Thomas et al. primers described in Papadopoulos et al. (1995). (1995). Repair reactions (25 ␮l) contained 30 mm 4-(2- Chromosome transfer: The method used for chromosome hydroxyethyl)-1-piperazine-ethanesulfonic acid (pH 7.8), transfer is described in Koi et al. (1989). The introduced 7 mm MgCl2, 200 ␮mol each CTP, GTP, UTP, 4 mm ATP, contain the gene conferring resistance to G418, 100 ␮mol each dCTP, dATP, dGTP, dTTP, 40 mm creatine such that cells receiving that chromosome are resistant to phosphate, 100 mg/ml creatine phosphokinase, 15 mm so- this drug. The presence of a transferred chromosome 2 was dium phosphate (pH 7.5), 1 fmol of substrate DNA, and 50 confirmed by detecting the allele of the donor cell line in ␮g of extract proteins. Following incubation at 37Њ for 30 min, early passage DNAs using the tetranucleotide repeats D2S405, the DNA substrates were processed and introduced into E. coli D2S407, D2S410, D2S423, and D2S433. When genomic DNAs NR9162 (mutS) via electroporation. Cells were plated, plaque from the chromosome-transferred cell lines described were colors were scored, and repair efficiency calculated as de- amplified using PCR primers specific for mouse B2 repetitive scribed (Thomas et al. 1995). sequence elements, no detectable mouse DNA was observed. HPRT mutation frequency and rate determinations: Muta- tion frequencies were obtained by plating 106 cells in 40 ␮mol RESULTS 6-TG at a density of 5 ϫ 104 cells per 10-cm dish. Cells were r incubated 12–14 days and 6-TG colonies were visualized by Introduction of hMSH6 into HHUA cells: HHUA cells staining with 0.5% crystal violet (in 50% methanol, v/v). Muta- contain a single base deletion mutation in hMSH3, such tion rate determinations were performed using cell popula- tions cleansed of preexisting HPRT mutants by culture in HAT that the MSH3 protein is missing 723 C-terminal amino (100 ␮mol hypoxanthine, 0.4 ␮mol aminopterin, and 16 ␮mol acids (Figure 1A). Many of these residues are conserved thymidine) medium. HAT medium was then removed and among eukaryotic MutS homologues, consistent with the initial HPRT mutant frequency was determined, and 2–3 ϫ the loss of hMSH3 function observed in an earlier study 6 10 cells were subcultured in nonselective medium. Additional (Risinger et al. 1996). However, HHUA cells containing mutant frequencies were obtained at 2- to 3-day intervals, while maintaining the cells in logarithmic growth. At each 6-TG wild-type chromosome 5 still exhibited a detectable level selection, a subculture of 2–3 ϫ 106 cells was plated in nonse- of microsatellite instability, and extracts of these cells lective medium for use in subsequent mutant frequency deter- still lacked MMR activity for base·base mismatches and minations. Additionally, population doublings were deter- some insertion mismatches. These results suggested that mined between each 6-TG selection. After obtaining 5–6 HHUA cells were mutant in another mismatch repair mutant frequencies, mutation rates were obtained by plotting Risinger the observed mutant frequency as a function of population gene. Subsequent analysis of hMSH6 ( et al. 1996) revealed a homozygous C T mutation (Figure doubling and calculating the slope by linear regression. The → slope of the curve yields the mutation rate, expressed as muta- 1B) that changed codon 1219 from threonine (ACT) tions per cell per generation (Glaab and Tindall 1997). to isoleucine (ATT). Threonine is present at this posi- HPRT mutant selection and sequencing: Independent mu- tion in 12 of 14 MutS homologues (Figure 1C), sug- tants resistant to 6-TG were obtained by first cleansing cell cultures of preexisting HPRT mutants in HAT medium. Fol- gesting that the change to isoleucine may impair lowing removal of HAT medium, 100 cells were plated in hMSH6 function. nonselective medium. These independent 100-cell cultures We began this study by transferring chromosome 2 -ϫ 106 cells and then selected in 6-TG as (encoding hMSH6 2) into HHUA cells by microcell fu 2.5ف were grown to described above. Colonies were allowed to grow for 14–18 sion (Koi et al. 1989). Single cell clones receiving chro- days, and individual 6-TGr clones were isolated. These 6-TGr clones were transferred to 24-well dishes and grown to conflu- mosome 2 were selected for resistance to G418, and ence. Independent mutants were defined as those spontane- the presence of the wild-type hMSH6 gene sequence in ous mutants arising in different 100-cell inocula. Amplification G418-resistant clones was analyzed by DNA sequence of HPRT mRNA from mutant clones was then performed by analysis of the region of hMSH6 known to be mutant Yang a procedure modified from et al. (1989). The cDNA of in HHUA cells. Both the HHUA parent cells (left se- the HPRT gene was then analyzed by automated sequencing. quence in Figure 2) and a clone (HHUA-5.5) containing Assay for mismatch repair activity: A circular M13mp2 DNA substrate is used, containing a covalently closed (ϩ) strand chromosome 5 (middle sequence in Figure 2) had only and a (Ϫ) strand with a nick (to direct repair to this strand) the mutant ATT sequence at codon 1219 reported ear- located several hundred base pairs away from the mispair lier (Risinger et al. 1996). In contrast, a clone receiving located in the lacZ ␣-complementation coding sequence. The chromosome 2 contained both the mutant ATT codon (ϩ) strand encodes one plaque phenotype (either colorless or and a wild-type ACT codon 1219. blue) and the (Ϫ) strand encodes the other plaque phenotype. When the unrepaired heteroduplex is introduced into an Analysis of microsatellite instability: As is typical of Escherichia coli strain deficient inmethyl-directed heteroduplex many cell lines with mutations in mismatch repair genes, repair, plaques have a mixed plaque phenotype on selective hMSH3/hMSH6-mutant HHUA cells exhibited micro- plates, because both strands of the heteroduplex are ex- satellite instability (Risinger et al. 1996). This includes pressed. However, repair occurring during incubation of the the parental HHUA clone used here, where, for any substrate in a repair-proficient human cell extract will reduce the percentage of mixed plaques and increase the ratio of the (ϩ) strand phenotype relative to that of the (Ϫ) strand phenotype, because the nick directs repair to the (Ϫ) strand. 2 For clarity, we have in several places provided the name of the The method for preparing extracts for repair reactions is relevant MutS homologue in parentheses after the chromosome that described in Roberts and Kunkel (1993). Preparation of was transferred. 1640 A. Umar et al.

Figure 2.—MSH6 gene sequence in HHUA and its deriva- tives. The sequence analysis was performed as described in materials and methods.

The extent of microsatellite instability was reduced in single cell clones derived from two clones (2.5 and 2.10 in Table 1) into which chromosome 2 (hMSH6) was introduced. For example, although 19 of 36 parental HHUA clones were unstable at the D17S791 dinucleo- tide repeat locus, none of 24 clones of the 2.5 derivative and only two of 36 clones of the 2.10 derivative were observed to be unstable. Reduced instability was observed Figure 1.—Mutations in hMSH3 and hMSH6 in HHUA cells. at all loci examined (Table 1). These results suggest that (A) A schematic drawing of hMSH3 and corresponding muta- tion in HHUA cell line. The total number of amino acid the human MSH6 gene participates in the repair of residues is indicated in parentheses. Shaded areas represent addition and deletion replication errors at all of these conserved regions of the protein. The position of highly con- loci. This includes errors in which the inferred muta- served Walker type A nucleotide binding sequence (GKS) is tional intermediates could involve 1, 2, 3, or 4 extra bases indicated as a dark bar. ⌬A indicates the loss af an A·T base residing in the primer strand (additions) or the template pair, resulting in protein truncation. (B) A schematic drawing of hMSH6 and the corresponding mutation in the HHUA cell strand (deletions). Single cell clones derived from line. The total number of amino acid residues is indicated in HHUA cells into which chromosome 5 (hMSH3) was parentheses. Shaded areas represent conserved regions of the introduced (5.5 and 5.10 in Table 1) also exhibited re- protein. The position of highly conserved Walker type A nucle- duced instability. Thus, among 71 total clones examined otide binding sequence (GKS) is indicated as a dark bar. (C) (35 for the 5.5 derivative and 36 for the 5.10 derivative), Alignment of the conserved region of MutS homologues at the region where the mutation in HHUA cells is found in few loci were observed to be unstable. However, the stabi- hMSH6 protein. The top line represents the consensus se- lizing effects of chromosome 5 transfer were confined quence; identical sequences are indicated by a (·) and differ- to di-, tri-, and tetranucleotide microsatellites. Increased ences from consensus for each MutS homologue are indicated stability was not apparent for the four homopolymeric by the single letter code for amino acids. The bottom line microsatellites examined in clone 5.5. These data sug- shows the sequence for hMSH6 in the HHUA cell line. Eco, E. coli; Spell 1, Spellchecker 1 (the MutS homologue in Dro- gest that the human MSH3 gene also participates in sophila melanogaster); x, Xenopus;y,Saccharomyces cerevisiae;m, the repair of addition and deletion replication errors mouse; r, rat; h, human. involving 2, 3, or 4 extra bases. However, there is no indication of hMSH3-dependent repair of single-base deletions at the four homopolymeric loci examined. given microsatellite, instability was observed in from Analysis of mutation rates at the HPRT locus: To four to 19 of the 36 single cell clones examined (Table explore further the mutator phenotype of HHUA cells 1, parent line). Instability was also observed in from five and the effects of introduction of chromosomes 2 and to 11 of 36 clones derived from a clone (17.3 in Table 5, we determined mutation rates at the HPRT locus 1) into which chromosome 17 was introduced, used in the HHUA parent clone and its derivatives. Consis- here as a negative control. In all, 12 microsatellite loci tent with results obtained with hMSH2 and hMSH6 sin- were examined, including four homopolymeric se- gle-mutant tumor cell lines (Bhattacharyya et al. quence loci, five loci comprised of dinucleotide repeats, 1994; Branch et al. 1995; Malkhosyan et al. 1996; W. E. two trinucleotide repeat loci, and one tetranucleotide Glaab, J. I. Risinger, A. Umar, J. C. Barrett, T. A. microsatellite. The instabilities observed included both Kunkel and K. R. Tindall, unpublished results; Ohzeki additions and deletions of one or more repeat units of et al. 1997), the double-mutant HHUA parent clone one to four bases. had an elevated HPRT mutation rate of 2.1 ϫ 10Ϫ5, Mismatch Repair Specificity of hMSH3 and hMSH6 1641

TABLE 1 Microsatellite instability in single cell clones of HHUA derivatives

Repeat Parent 17.3 2.5 2.10 5.5 5.10 BAT25 (A)n U U — S U — BAT26 (A)n U U — S U — BAT40 (A)n U U — S U — MSH2EX5 (A)n 7a 5a —0 6a — D2S123 (CA)n 12a 7a —1 1a — D14S78 (CA)n 13a 11a 002a 0 D2S119 (CA)n 4 5 — 0 0 — D17S791 (CA)n 19a 11a 021a 0 D5S107 (CA)n 9 8 — 1 2 — D7S1794 (CTT)n 10a 6a 110b 0 D17S1330 (CTT)n 14 7 — 0 0 — vWA (TCTA)n 5a 5a 000a 0 Total clones analyzed 36 36 24 36 35 24 PCR analysis of the stability of the BAT loci, containing long runs of A·T base pairs, generates a ladder of many bands, making assignment of instability in individual clones problematic. A dash indicates that this analysis was not performed. The designation “U” or “S” indicates general instability or stability among all the clones when analyzed side-by-side on the same autoradiogram, as judged by the overall banding pattern. The banding patterns for analysis of the other loci are simpler (e.g., see Risinger et al. 1995, 1996). The number of population doublings during growth of the single cell clones was 26 (parent), 12 (17.3), 22 (2.5), 25 (2.10), 23 (5.5), and 27 (5.10). Clone 17.3 is used as a negative control because chromosome 17 does not encode the hMSH3 or hMSH6 gene. At several of these loci, no instability has been observed in single cell clones of cell lines proficient in mismatch repair (Umar et al. 1994a; Risinger et al. 1995; Shibata et al. 1995). a From Risinger et al. (1996). b The six clones reported to be unstable earlier (Risinger et al. 1996) actually represented loss of one of the two alleles rather than apparent insertion or deletion of one or more repeat units. or 450-fold higher than the rate for normal human normal fibroblasts, with the rate of transition mutations fibroblasts (Table 2). Clone 17.3, which showed no re- elevated to a twofold greater extent than that of the duction in microsatellite instability (Table 1), also had transversion rate. These hMSH3/hMSH6-mutant cells a high mutation rate (1.7 ϫ 10Ϫ5). The rate for HHUA have an even more strongly elevated rate of single-base cells into which the hMSH6 gene on chromosome 2 was frameshifts, which are increased by 2400-fold relative to introduced (clone 2.10) was 2.9 ϫ 10Ϫ7 (Table 2). This NHF-1 cells. There is an approximately fourfold greater value is 72-fold lower than that of the double-mutant increase in rate observed for additions than for deletions. HHUA cells, indicating a functional role for the hMSH6 Using the overall low HPRT mutation rate of 2.9 ϫ gene. HHUA cells into which the hMSH3 gene on chro- 10Ϫ7 in HHUA clone 2.10 cells, we calculated the mini- mosome 5 was introduced (clone 5.10) also had a lower mum extent to which introduction of chromosome 2 mutation rate (3.2 ϫ 10Ϫ6; Table 2). This value is 6.6- containing the hMSH6 gene lowered the mutation rate fold lower than the rate for HHUA cells, indicating that relative to parental HHUA cells. The results (Table 2) the hMSH3 gene also functions in MMR. However, the show that the rates for transitions, transversions, addi- mutation rate in clone 5.10 remains 68-fold higher than tions, and deletions were reduced by Ն96, Ն91, Ն92, that of normal human fibroblasts (Table 2) known to and Ն94%, respectively. The spectrum of spontaneous be MMR proficient (Boyer et al. 1993). Thus, introduc- mutations observed in clone 5.10 was similar to that tion of hMSH3 yields only partial correction of the muta- observed in the parental HHUA cells. There were tion rate. slightly more base substitutions (six of nine indepen- The specificity of spontaneous mutagenesis in mis- dent mutants) than frameshift mutations (three of nine match repair-proficient human cells is known, allowing independent mutants). Calculations based on these calculation of the rates for single-base substitution and data (Table 2) show that the rates for transitions, trans- frameshift mutations at the HPRT locus (Table 2). Spon- versions, additions, and deletions were reduced by 90, taneous mutations in HHUA included 57% single-base 67, 81, and 92%, respectively, in cells receiving chromo- substitutions (25 of 44 independent mutants) and 43% some 5 encoding the hMSH3 gene. single-base frameshifts (19 of 44 independent mutants), Restoration of MMR activity in extracts: Extracts of 11 of which were in a homopolymeric run of six gua- HHUA cells were deficient in strand-specific mismatch nines (base 207–212) in exon 3. The calculated muta- repair in vitro of substrates containing base·base and tion rates (Table 2) reveal that the overall base substitu- insertion/deletion mismatches. As previously reported tion rate in HHUA cells was elevated 500-fold relative to (Risinger et al. 1996), extracts of clone 5.5 cells repaired 1642 A. Umar et al.

TABLE 2 Mutation rates at the HPRT locus in HHUA derivatives

NHF-1 HHUA-P HHUA-2.10 HHUA-5.10 Mutant frequency (ϫ 10Ϫ6) 180 6.4 13 Mutation rate (ϫ 10Ϫ8)a 4.7 2100 29 320b Relative ratec 1 450 6.2 68 Percent Percent Rate Rate Relative Rate correction Rate correction (ϫ 10Ϫ8)(ϫ10Ϫ8) ratec (ϫ 10Ϫ8) efficiencyd (ϫ 10Ϫ8)e efficiencyd

Base substitutions 2.4 1200 500 Յ29 Ն98 210 83 Transitions 1.1 720 650 Յ29 Ն96 110 85 Transversions 1.0 330 330 Յ29 Ն91 110 67 Frameshifts 0.38 910 2400 Յ29 Ն97 110 88 Plus-one 0.07 380 5400 Յ29 Ն92 71 81 Minus-one 0.31 330 1100 Յ29 Ն91 36 89 a Mutations at HPRT locus/cell/generation, for duplicate determinations. The value for NHF-1 cells is from Glabb and Tindall (1997). The rates for individual classes of errors in NHF-1 cells were calculated using specificity data from a database of spontaneous HPRT mutations in normal human cells (Cariello 1996). The rates for individual classes of errors in HHUA cells were calculated using sequencing data mentioned in results. Among the 44 mutants sequenced were three base substitutions at splice junctions that were not defined as transitions or transversions and two mutants containing two deletions. To obtain the mutation rate for each class of mutations, the fraction of each class was multiplied by the mutation rate at the HPRT locus in HHUA cells. b The HPRT mutation rate for a second clone of HHUA cells into which chromosome 5 was introduced (clone 5.5) was 94 ϫ 10Ϫ8. c Relative to NHF-1 cells. d The percent correction efficiency is 100 ϫ [1 Ϫ (rate in corrected clone/rate in parental clone)]. The correction efficiencies in HHUA-2.10 cells are minimum estimates that are calculated using the overall mutation rate, because no mutants were sequenced. e The rates for individual classes of errors in HHUA-5.10 cells were calculated as described above for HHUA cells, using the sequencing data described in results. some substrates containing extra nucleotides but were mismatch, indicating that this repair activity that de- deficient in repair of those containing base·base mis- pends on the newly introduced chromosome 2 has the matches or certain other extra nucleotides. To deter- characteristic bidirectional repair capacity of the gen- mine the effect on MMR activity of introducing chromo- eral MMR system (Fang and Modrich 1993). Thus, some 2 into HHUA cells, we examined the ability of the data in Figure 3 demonstrate that introduction of extracts of the clones receiving chromosome 2 (hMSH6) chromosome 2 into HHUA cells restored to extracts the to repair nine heteroduplex substrates in vitro. An ex- ability to perform strand-specific MMR to a variety of tract of MMR-proficient HeLa cells repaired all of these heteroduplexes. substrates in a strand-specific manner, whereas an ex- tract of parental HHUA cells failed to repair any of them (Figure 3). Extracts of HHUA clone 2.5 cells re- DISCUSSION paired seven of the nine substrates. Of the two substrates The fact that introduction of chromosome 2 (hMSH6) that were not repaired, one contained a single extra from an MMR-proficient cell into the hMSH3/hMSH6- nucleotide and the other contained two extra bases mutant, MMR-deficient HHUA human endometrial tu- (fourth and fifth substrates shown in Figure 3). Note mor cell line reduces microsatellite instability at 12 loci that these substrates are repaired in an extract of MMR- (Table 1), reduces the mutation rate at the HPRT locus proficient HeLa cells. The seven substrates that were (Table 2), and restores MMR activity to extracts in vitro repaired in extracts of HHUA clone 2.5 cells included (Figure 3) suggests a functional role for the hMSH6 a G·G and a G·T mismatch and heteroduplexes con- gene in postreplication mismatch repair. The data are taining one to four extra bases in one strand (Figure consistent with many other observations in yeast and 3). In all cases where repair was observed, the change human cells, including the fact that introduction of in the ratio of blue to colorless plaques indicates that chromosome 2 also corrects an independently derived repair occurred in the strand containing the nick, which colon tumor cell line containing other hMSH6 muta- is known to serve as a strand-discrimination signal in tions (Umar et al. 1997). A functional role in MMR is vitro (Holmes et al. 1990; Thomas et al. 1991). Repair inferred for hMSH3 as well, on the basis of the fact that was observed when the nick was either 3Ј or 5Ј to the introduction of chromosome 5 encoding the hMSH3 Mismatch Repair Specificity of hMSH3 and hMSH6 1643

Figure 3.—Mismatch repair activity in HHUA and its derivative cell lines. The analysis was performed as described in materials and methods using the substrates indicated (described in Risinger et al. 1996) and reactions were incubated for 15 min. Substrates designated with a (⍀) contain the number of extra nucleotides that accompany the symbol. Substrates are designated with a 3Ј when the nick in the (Ϫ) strand is at the AvaII site (position Ϫ264, where position ϩ1 is the first transcribed base of the lacZ ␣-complementation gene) or with a 5Ј when the nick is at the Bsu36 I site (position ϩ276). The nucleotide position of the mismatch or unpaired bases in the lacZ ␣-complementation gene is indicated after the symbol @. In all substrates, the nick is in the (Ϫ) strand. The (ϩ)or(Ϫ) sign designates the strand containing the extra base(s). Results are expressed as percent repair determined from counting several hundred plaques per variable. Repair values of Յ1% are represented as 1%. All three extracts were competent for SV40 origin-dependent DNA replication activity. gene into hMSH3/hMSH6-mutant HHUA cells also re- of redundancy in human cells and compares the results duces microsatellite instability and restores MMR activ- to those obtained in yeast. ity (Risinger et al. 1996). That earlier study of hMSH3 With respect to MSH6 function, the reduced substitu- gene function is extended here by examination of sev- tion rate at the HPRT locus (Table 2) and the ability eral additional microsatellite loci (Table 1) and by the to repair base·base mismatches in vitro (Figure 3) upon demonstration that introducing chromosome 5 also re- introduction of chromosome 2 into HHUA cells implies duces the HPRT mutation rate (Table 2). that hMSH6 participates in repair of base·base mis- Human tumor cell lines with mutations in hMSH2 matches. This is consistent with the increased base sub- exhibit robust microsatellite instability, highly elevated stitution rate at the canavanine locus in a yeast mutation rates in endogenous genes, and deficiency in mutant (Marsischky et al. 1996) and with the ability repair of a variety of base·base and insertion/deletion of human and yeast MutS␣ to bind to base·base mis- mismatches (see citations in Introduction). These data matches (cited above). Introduction of chromosome 2 and studies (cited above) of the substrate binding speci- into HHUA cells results in increased stability of micro- ficity of MSH2·MSH6 and of MSH2·MSH3 heterodimers satellites comprised of one-, two-, three-, or four-base suggest that MSH2 is involved in the initial recognition repeats (Table 1), a reduction in mutation rates for step of strand-specific repair of a wide variety of replica- single-base deletions and additions at the HPRT locus tion errors in eukaryotic cells. The hMSH3Ϫ/hMSH6Ϫ (Table 2), and an ability of extracts to repair some (but double mutant HHUA cell line has similar genetic insta- not all) substrates containing one, two, three, or four bility and MMR deficiencies as MSH2-mutant cells, and extra bases (Figure 3). This implies that human MSH6 partial correction is observed upon introduction of ei- also participates in the repair of insertion/deletion mis- ther wild-type hMSH3 or hMSH6, suggesting that hMSH3 matches involving one, two, three, or four extra bases. and hMSH6 share partially redundant functions in MMR Mononucleotide and dinucleotide repeats are likewise in human cells. Table 3 summarizes the inferred extent unstable in a yeast msh6 mutant (Table 3), consistent 1644 A. Umar et al.

TABLE 3 Inferred substrate specificity of MSH6- and MSH3-dependent MMR

MSH6-dependent MSH3-dependent Human Human Substrate MINa HPRTb MMRc Yeast MINa HPRTb MMRd Yeast Base·base mismatch NA ϩϩϩeNA ϩϪϪe One extra nucleotide ϩϩϩand Ϫϩe,f,g Ϫϩϩand Ϫϩe,f,g Two extra nucleotides ϩ NA ϩ and Ϫϩf,h ϩ NA ND ϩf,h,i Three extra nucleotides ϩ NA ϩ ND ϩ NA ND ND Four extra nucleotides ϩ NA ϩϪf ϩNA ϩϩf NA, not applicable; ND, not determined; MIN, microsatellite instability. a From Table 1. b From Table 2. c From Figure 3. d From Figure 3 in Risinger et al. 1996. e Marsischky et al. 1996. f Sia et al. 1997a. g Greene and Jinks-Robertson, 1997. h Johnson et al. 1996. i Strand et al. 1995. with MSH6-dependent repair of insertion/deletion mis- in yeast (Marsischky et al. 1996), again providing at matches containing one or two extra bases. However, least the possibility of differences between the yeast and Sia et al. (1997a) observed no effect of an msh6 mutation human mismatch repair systems. on the stability of repetitive sequences containing repeat Introducing hMSH3 on chromosome 5 into HHUA units of four or more bases. Thus, MSH6-dependent cells increases the stability of microsatellites comprised repair specificity could possibly differ in yeast and hu- of two-, three-, or four-base repeats (Table 1), decreases man cells. Alternatively, the difference could reflect the the rate of single-base deletions and additions at the use of different repeat sequences or any of several other HPRT locus (Table 2), and restores to extracts the capac- variables (discussed below). ity to repair some (but not all) substrates containing With respect to MSH3 function, introduction of chro- one or four extra bases (Risinger et al. 1996). This mosome 5 (hMSH3) into HHUA cells reduced the base implies that human MSH3 also participates in repairing substitution rate at the HPRT locus (Table 2). This im- some insertion/deletion mismatches involving one, two, plies that hMSH3 can participate in repair of base·base three, or four extra bases. This conclusion is consistent mismatches, at least under circumstances where hMSH6 with five studies in mutant yeast strains (Table 3). is mutant. A limited role for MSH3 in repair of base·base These inferences on the role of MSH3 and MSH6 in mismatches is consistent with the slight but detectable repair of various types of mismatches are derived from binding of human (Figure 4a in Palombo et al. 1996) studies in which one gene is mutant and the other wild and yeast (Figure 3b in Habraken et al. 1996) type. Unknown is the relative contribution of MSH3 MSH2·MSH3 complexes to a G·T mismatch. Moreover, and MSH6 to repair of different mismatches in normal binding of the yeast MSH2·MSH3 complex is enhanced cells containing two wild-type copies of both genes. Fur- by the yeast MLH1·PMS1 complex (Habraken et al. ther studies will be required to understand why, despite 1997). A lower apparent affinity of MSH2·MSH3 for the apparent ability to repair some insertion/deletion base·base mispairs as compared to MSH2·MSH6 may mismatches, extracts of HHUA-2.5 cells (Figure 3) and explain why the effect on base substitution rates at the HHUA-5.5 cells (Figure 3 in Risinger et al. 1996) fail HPRT locus was less for introduction of chromosome 5 to repair certain others and why, even after introducing than for introduction of chromosome 2 into HHUA chromosome 5, HHUA cells continue to exhibit instabil- cells (Table 2). A lower binding affinity of MSH2·MSH3 ity at four homopolymeric microsatellites (Table 2). to base·base mismatches may also explain why introduc- These data indicate that MSH6- and MSH3-dependent tion of chromosome 5 failed to restore base·base mis- insertion/deletion mismatch repair capacity may de- match repair activity to extracts (Table 3; Risinger et pend on variables other than simply the number of al. 1996). It is also noteworthy that the degree to which extra bases. These variables may include the identity of the base substitution rate at the canavanine locus inyeast the extra nucleotides, whether they are present in the is elevated in msh2, msh3, and msh6 mutants suggests that template strand (as for deletions) or the primer strand MSH3 may not participate in base·base mismatch repair (as for additions) or within repetitive or nonrepetitive Mismatch Repair Specificity of hMSH3 and hMSH6 1645 sequences. Also, repair capacity may be influenced by association with hereditary nonpolyposis colon cancer. Cell 75: 1027–1038. the location of the mismatch relative to the (presently Fishel, R., A. Ewel and M. K. Lescoe, 1994a Purified human MSH2 unknown) strand-discrimination signal, relative to the protein binds to DNA containing mismatched nucleotides. Can- replication fork (e.g., leading or lagging strand replica- cer Res. 54: 5539–5542. Fishel, R., A. Ewel, S. Lee, M. K. Lescoe and J. Griffith, 1994b tion errors), or relative to transcription. The cell lines Binding of mismatched microsatellite DNA sequences by the described here also provide the opportunity to examine human MSH2 protein. Science 266: 1403–1405. the effects of mutations in hMSH3 and hMSH6 on in- Glaab, W. E., and K. R. Tindall, 1997 Mutation rate at the hprt W. E. locus in human cancer cell lines with specific mismatch repair- duced cytotoxicity and mutagenesis (e.g., see gene defects. Carcinogenesis 18: 1–8. Glaab, J. I. Risinger, A. Umar, J. C. Barrett, T. A. Greene, C. N., and S. Jinks-Robertson, 1997 Frameshift intermedi- Kunkel and K. R. Tindall, unpublished results). ates in homopolymer runs are removed efficiently by yeast mis- match repair proteins. Mol. Cell. Biol. 17: 2844–2850. We thank Samuel E. Bennett and Karin Drotschmann for criti- Habraken, Y., P. Sung, L. Prakash and S. Prakash, 1996 Binding cally evaluating the manuscript. T.A.K. dedicates this article to Dr. of insertion/deletion DNA mismatches by the heterodimer of John W. Drake, in gratitude for 15 years of insightful scientific com- yeast mismatch repair proteins MSH2 and MSH3. Curr. Biol. 6: ments, strong administrative support and sage advice as a mentor, 1185–1187. Habraken, Y., P. Sung, L. Prakash S. Prakash, and as a friend. and 1997 Enhance- ment of MSH3-MSH3-mediated mismatch recognition by the yeast MLH1-PMS1 complex. Curr. Biol. 7: 790–793. Holmes, J., S. J. Clark and P. Modrich, 1990 Strand-specific mis- match correction in nuclear extracts of human and Drosophila LITERATURE CITED melanogaster cell lines. Proc. Natl. Acad. Sci. USA 87: 5837–5841. Hughes, M. J., and J. Jiricny, 1992 The purification of a human Alani, E., 1996 The Saccharomyces cerevisiae Msh2p and Msh6p mismatch-binding protein and association of its associated form a complex that specifically binds to duplex oligonucleotides ATPase and helicase activities. J. Biol. Chem. 267: 23876–23882. containing mismatched DNA base pairs. Mol. Cell. Biol. 16: 5604– Iaccarino, I., F. Palombo, J. Drummond, N. F. Totty, J. J. Hsuan 5615. Alani, E., N.-W. Chi and R. D. Kolodner, 1995 The Saccharomyces et al., 1996 MSH6, a Saccharomyces cerevisiae protein that binds cerevisiae Msh2 protein specifically binds to duplex oligonucleo- to mismatches as a heterodimer with MSH2. Curr. Biol. 6: 484– 486. tides containing mismatched DNA base pairs and insertions. Jiricny, J., Genes Dev. 9: 234–239. 1996 Mismatch repair and cancer. Cancer Surv. 28: Alani, E., T. Sokolsky, B. Studamire, J. J. Miret R. S. Lahue, 47–68. and Johnson, R. E., G. K. Kovvali, L. Prakash S. Prakash, 1997 Genetic and biochemical analysis of Msh2p–Msh6p: role and 1996 of ATP hydrolysis and Msh2p–Msh6p subunit interactions in mis- Requirement of the yeast MSH3 and MSH6 genes for MSH2- dependent genomic stability. J. Biol. Chem. 271: 7285–7288. match base pair recognition. Mol. Cell. Biol. 17: 2436–2447. Kat, A., W. G. Thilly, W. Fang, M. J. Longley, G. M. Li Bhattacharyya, N. P., A. Skandalis, A. Ganesh, J. Groden and et al., 1993 M. Meuth, 1994 Mutator phenotypes in human colorectal carci- An alkylation-tolerant, mutator human cell line is deficient in noma cell lines. Proc. Natl. Acad. Sci. USA 91: 6319–6323. strand-specific mismatch repair. Proc. Natl. Acad. Sci. USA 90: Bhattacharyya, N. P., A. Ganesh, G. Phear, B. Richards, A. Skan- 6424–6428. Koi, M., H. Morita, H. Yamada, H. Satoh, J. C. Barrett dalis et al., 1995 Molecular analysis of mutations in mutator et al., 1989 colorectal carcinoma cell lines. Hum. Mol. Genet. 4: 2057–2064. Normal human chromosome 11 suppresses tumorigenicity of Boyer, J. C., D. C. Thomas, V. M. Maher, J. J. McCormick and human cervical tumor cell line SiHa. Mol. Carcinogenesis 2: T. A. Kunkel, 1993 Fidelity of DNA replication by extracts of 12–21. Kolodner, R., normal and malignantly transformed human cells. Cancer Res. 1996 Biochemistry and genetics of eukaryotic mis- 53: 1–6. match repair. Genes Dev. 10: 1433–1442. Leach, F. S., N. C. Nicolaides, N. Papadopoulos, B. Liu, J. Jen Boyer, J. C., A. Umar, J. I. Risinger, J. R. Lipford, M. Kane et al., et al., 1995 Microsatellite instability, mismatch repair deficiency, and 1993 Mutations of a mutS homolog in hereditary nonpolyposis genetic defects in human cancer cell lines. Cancer Res. 55: 6063– . Cell 75: 1215–1225. 6070. Li, G.-M., and P. Modrich, 1995 Restoration of mismatch repair Branch, P., R. Hampson and P. Karran, 1995 DNA mismatch bind- to nuclear extracts of H6 colorectal tumor cells by a heterodimer ing defects, DNA damage tolerance, and mutator phenotypes in of human MutL homologs. Proc. Natl. Acad. Sci. USA 92: 1950– human colorectal carcinoma cell lines. Cancer Res. 55: 2304– 1954. 2309. Li, G. M., H. Wang and L. J. Romano, 1996 Human MutSalpha Cariello, N., 1996 Databases and software for the analysis of muta- specifically binds to DNA containing aminofluorene and acetyl- tions in the human p53 gene, the human HPRT gene and the aminofluorene adducts. J. Biol. Chem. 271: 24084–24088. LacI gene in transgenic rodents. Nucleic Acids Res. 24: 119–120. Malkhosyan, S., A. McCarty, H. Sawai and M. Perucho, 1996 Drummond, J. T., G. M. Li, M. J. Longley and P. Modrich, 1995 Differences in the spectrum of spontaneous mutations in the Isolation of an hMSH2-p160 heterodimer that restores DNA mis- hprt gene between tumor cells of the microsatellite mutator phe- match repair to tumor cells. Science 268: 1909–1912. notype. Mutat. Res. 316: 249–259. Duckett, D. R., J. T. Drummond, A. I. Murchie, J. T. Reardon, A. Marsischky, G. T., N. Filosi, M. F. Kane and R. Kolodner, 1996 Sancar et al., 1996 Human MutSalpha recognizes damaged Redundancy of Saccharomyces cerevisiae MSH3 and MSH6 in DNA base pairs containing O6-methylguanine, O4-methylthy- MSH2-dependent mismatch repair. Genes Dev. 10: 407–420. mine, or the cisplatin-d(GpG) adduct. Proc. Natl. Acad. Sci. USA Mello, J. A., S. Acharya, R. Fishel and J. M. Essigmann, 1996 The 93: 6443–6447. mismatch-repairprotein hMSH2 binds selectively to DNA adducts Fang, W.-H., and P. Modrich, 1993 Human strand-specific mis- of the anticancer drug cisplatin. Chem. Biol. 3: 579–589. match repair occurs by a bidirectional mechanism similar to that Modrich, P., and R. Lahue, 1996 Mismatch repair in replication of the bacterial reaction. J. Biol. Chem. 268: 11838–11844. fidelity, genetic recombination, and cancer biology. Annu. Rev. Fink, D., S. Nebel, S. Aebi, H. Zhang, B. Cenni et al., 1996 The Biochem. 65: 101–133. role of mismatch repair in platinum drug resistance. Cancer Res. Mu, D., M. Tursun, D. R. Duckett, J. T. Drummond, P. Modrich 56: 4881–4886. et al., 1997 Recognition and repair of compound DNA lesions Fishel, R., and T. Wilson, 1997 MutS homologs in mammalian (base damage and mismatch) by human mismatch repair and cells. Curr. Opin. Genet. Dev. 7: 105–113. excision repair systems. Mol. Cell. Biol. 17: 760–769. Fishel, R., M. K. Lescoe, M. R. Rao, N. G. Copeland, N. A. Jenkins Ohzeki, S., A. Tachibana, K. Tatsumi and T. Kato, 1997 Spectra et al., 1993 The human mutator gene homolog MSH2 and its of spontaneous mutations at the hprt locus in colorectal carci- 1646 A. Umar et al.

noma cell lines defective in mismatch repair. Carcinogenesis 18: repeat unit size and DNA mismatch repair genes. Mol. Cell. Biol. 1127–1133. 17: 2851–2858. Orth, K., J. Hung, A. Gazdar, A. Bowcodk, J. M. Mathis et al., Sia, E. A., S. Jinks-Robertson and T. D. Petes, 1997b Genetic 1994 Genetic instability in human ovarian cancer cell lines. Proc. control of microsatellite instability. Mutation. Res. 381: 61–70. Natl. Acad. Sci. USA 91: 9495–9499. Strand, M., T. A. Prolla, R. M. Liskay and T. D. Petes, 1993 Desta- Palombo, F., P. Gallinari, I. Iaccarino, T. Lettieri, M. Hughes et bilization of tracts of simple repetitive DNA in yeast by mutations al., 1995 GTBP, a 160-kilodalton protein essential for mismatch- affecting DNA mismatch repair. Nature 365: 274–276. binding activity in human cells. Science 268: 1912–1914. Strand, M., M. C. Earley, G. F. Crouse and T. D. Petes, 1995 Muta- Palombo, F., I. Iaccarino, E. Nakajima, M. Ikejima, T. Shimada et tions in the MSH3 gene preferentially lead to deletions within al., 1996 hMutSbeta, a heterodimer of hMSH2 and hMSH3, tracts of simple repetitive DNA in Saccharomyces cerevisiae. Proc. binds to insertion/deletion loops in DNA. Curr. Biol. 6: 1181– Natl. Acad. Sci. USA 92: 10418–10421. Thomas, D. C., J. D. Roberts T. A. Kunkel, 1184. and 1991 Heteroduplex Papadopoulos, N., N. C. Nicolaides, B. Liu, R. Parsons, C. Len- repair in extracts of human HeLa cells. J. Biol. Chem. 266: 3744– gauer 3751. et al., 1995 Mutations of GTBP in genetically unstable Thomas, D. C., A. Umar T. A. Kunkel, cells. Science 268: 1915–1917. and 1995 Measurement Richards, B., H. Zhang, G. Phear M. Meuth, of heteroduplex repair in human cell extracts, pp. 187–197 in and 1997 Condi- E. C. tional mutator phenotypes in hMSH2-deficient tumor cell lines. Methods: A Companion to Methods in Enzymology, edited by Friedberg. Academic Press, Orlando, FL. Science 277: 1523–1526. Umar, A., and T. A. Kunkel, 1996 DNA-replication fidelity, mis- Risinger, J. I., A. Umar, J. C. Boyer, A. C. Evans, A. Berchuck et match repair and genome instability in cancer cells. Eur. J. Bio- al., 1995 Microsatellite instability in gynecological sarcomas and chem. 238: 297–307. in hMSH2 mutant uterine sarcoma cell lines defective in mis- Umar, A., J. C. Boyer, D. C. Thomas, D. C. Nguyen, J. I. Risinger match repair activity. Cancer Res. 55: 5664–5669. Risinger, J. I., A. Umar, J. Boyd, A. Berchuck, T. A. Kunkel et al., 1994a Defective mismatch repair in extracts of colorectal et al., and endometrial cancer cell lines exhibiting microsatellite insta- 1996 Mutation of MSH3 in endometrial cancer and evidence bility. J. Biol. Chem. 269: 14367–14370. for its functional role in heteroduplex repair. Nature Genet. 14: Umar, A., J. C. Boyer and T. A. Kunkel, 1994b DNA loop repair 102–105. by human cell extracts. Science 266: 814–816. Roberts, J. D., T. A. Kunkel, and 1993 Fidelity of DNA replication Umar, A., M. Koi, J. I. Risinger, W. Glaab, K. R. Tindall et al., 1997 in human cells. Meth. Molec. Genetics 2: 295–313. Correction of hypermutability, N-methyl-NЈ-nitro-N-nitrosogua- Shibata, D., M. A. Peinado, Y. Ionov, S. Malkhosyan and M. Peru- nidine-resistance and defective DNA mismatch repair by intro- cho, 1995 Genomic instability in repeated sequences is an early ducing chromosome 2 into human tumor cells with mutations somatic event in colorectal tumorigenesis that persists after trans- in MSH2 and MSH6. Cancer Res. 57: 3949–3955. formation. Nature Genet. 6: 273–281. Yang, J. L., V. M. Maher and J. J. McCormick, 1989 Amplification Sia, E. A., R. J. Kokoska, M. Dominska, P. Greenwell and T. D. and direct sequencing of cDNA from the lysate of low numbers Petes, 1997a Microsatellite instability in yeast: dependence on of diploid human cells. Gene 83: 347–354.