Germline and somatic mutations in hMSH6 and hMSH3 in gastrointestinal cancers of the microsatellite mutator phenotype

Gene 272 (2001) 301±313

www.elsevier.com/locate/gene

Germline and somatic mutations in hMSH6 and hMSH3 in gastrointestinal cancers of the microsatellite mutator phenotype Naoki Ohmiya, Sandra Matsumoto, Hiroyuki Yamamoto, Svetlana Baranovskaya, Sergei R. Malkhosyan, Manuel Perucho* The Burnham Institute, 10901 North Torrey Pines Road, La Jolla, CA 92037, USA Received 5 February 2001; received in revised form 19 April 2001; accepted 4 May 2001 Received by T. Sekiya

Abstract Hereditary and sporadic gastrointestinal cancer of the microsatellite mutator phenotype (MMP) is characterized by a remarkable genomic instability at simple repeated sequences. The genomic instability is often caused by germline and somatic mutations in DNA mismatch repair (MMR) genes hMSH2 and hMLH1. The MMP can be also caused by epigenetic inactivation of hMLH1. The MMP generates many somatic frameshift mutations in genes containing mononucleotide repeats. We previously reported that in MMP tumors the hMSH6 and hMSH3 MMR genes often carry frameshift mutations in their (C)8 and (A)8 tracks, respectively. We proposed that these `secondary mutator mutations' contribute to a gradual manifestation of the MMP. Here we report the detection of other frameshift, nonsense, and missense mutations in these genes in colon and gastric cancers of the MMP. A germline frameshift mutation was found in hMSH6 in a colon tumor harboring another somatic frameshift mutation. Several germline sequence variants and somatic missense mutations at conserved residues were detected in hMSH6 and only one was detected in hMSH3. Of the three hMSH6 germline variants in conserved residues, one coexisted with a somatic mutation at the (C)8 track and another had a somatic missense mutation. We suggest that some of these germline and somatic missense variants are pathogenic. While biallelic hMSH6 and hMSH3 frameshift mutations were found in some tumors, many tumors seemed to contain only monoallelic mutations. In some tumors, these somatic monoallelic frameshift mutations at the (C)8 and (A)8 tracks were found to coexist with other somatic mutations in the other allele, supporting their functionality during tumorigenesis. However, the low incidence of these additional somatic mutations in hMSH6 and hMSH3 leaves many tumors with only monoallelic mutations. The impact of the frameshift mutations in gene expression was studied by comparative analysis of RNA and protein expression in different tumor cell clones with different genotypes. The results show that the hMSH6 (C)8 frameshift mutation abolishes protein expression, ruling out a dominant negative effect by a truncated protein. We suggest the functionality of these secondary monoallelic mutator mutations in the context of an accumulative haploinsuf®ciency model. q 2001 Elsevier Science B.V. All rights reserved. Keywords: Mismatch repair; Microsatellite instability; Polymorphisms; Frameshifts

1. Introduction A genome-wide instability at simple repeat sequences characterizes gastrointestinal cancers of the MMP (Perucho, 1996). By failing to repair the spontaneous errors of replicaAbbreviations: ATCC, American Type Culture Collection; ATPase, adenosine triphosphatase; cDNA, DNA complementary to RNA; DNase, deoxyribonuclease; HNPCC, hereditary non-polyposis colon cancer; HRP, horseradish peroxidase; kDa, kilodalton; LOH, loss of heterozygosity; MMP, microsatellite mutator phenotype; MMR, mismatch repair; MSI-L, microsatellite instability low; PCR, polymerase chain reaction; RFLP, restriction fragment length polymorphism; RT-PCR, reverse transcription PCR; SDS-PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis; SSCP, single strand conformation polymorphism * Corresponding author. Tel.: 11-858-646-3112; fax: 11-858-646-3190. E-mail address: [email protected] (M. Perucho).

tion of these unstable sequences, these tumors accumulate hundreds of thousands of insertion and deletion mutations in microsatellites (Ionov et al., 1993). MMP tumors differ from those without enhanced microsatellite genomic instability in many biological, clinical, and molecular parameters (Ionov et al., 1993; Thibodeau et al., 1993; Aaltonen et al., 1993; Kinzler and Vogelstein, 1996; Perucho, 1996, 1999; Boland et al., 1998). These differences support the concept that the MMP underlies a distinct molecular pathway for carcinogenesis (Ionov et al., 1993; Perucho et al., 1994; Perucho, 1996; Olschwang et al., 1997). The MMP and resulting enhanced genomic microsatellite instability are caused by `mutator mutations' such as those inactivating DNA MMR gene products (Kinzler and Vogelstein, 1996; Kolodner, 1996). Germline mutations (hMSH2,

0378-1119/01/$ - see front matter q 2001 Elsevier Science B.V. All rights reserved. PII: S 0378-111 9(01)00517-0

302

N. Ohmiya et al. / Gene 272 (2001) 301±313

hMSH6, hMLH1 and hPMS2) have been described in HNPCC families. However, there are some families that do not show mutations in any of these genes, despite the majority of HNPCC tumors exhibiting the MMP (Kinzler and Vogelstein, 1996). Inactivation of hMLH1 by epigenetic silencing (hypermethylation) has been also shown as a mechanism for the manifestation of the mutator phenotype (Kane et al., 1997). We recently reported that non-selected colorectal and gastric cancers of the MMP contain somatic slippagerelated frameshift mutations in the hMSH3 and hMSH6 genes, which we de®ned as `secondary' mutators (Malkhosyan et al., 1996; Yamamoto et al., 1997). We reasoned that frameshift mutations of the secondary mutators must be induced by mutations of `primary' mutators, such as hMLH1, hMSH2, or other unidenti®ed equivalents. Based on these ®ndings, we proposed a `mutator that mutates the other mutator' model to describe the stepwise nature of the unfolding of the MMP (Perucho, 1996). According to the model, secondary mutator mutations enhance the depth and/ or width of genomic instability of tumor cells, accelerating the accumulation of oncogenic mutations in target cancer genes of the MMP pathway for carcinogenesis. The model has gained support by the con®rmation of the presence of common somatic mutations in these genes (Boland et al., 1998), but the functional signi®cance of these secondary mutator mutations has not been demonstrated. The mutations described in the (A)8 and (C)8 hotspots are often heterozygous (Malkhosyan et al., 1996; Yamamoto et al., 1997, 2000; Schwartz et al., 1999). While biallelic (homozygous) or hemizygous mutations in these genes are obligatorily functional, as has been shown in yeast and mice model systems, the role in tumorigenesis of monoallelic (heterozygous) mutations in the hMSH6 and hMSH3 genes is not clear because of the recessive nature of the MMP (Casares et al., 1995). In analogy with other target genes of the MMP, with frameshifts in mononucleotide tracts, some tumors may be heterozygous for these mutations (i.e. the hotspot frameshift occurs in only one allele) but functionally homozygous for the gene inac-

tivating mutation, because the other allele may carry other mutations (Yamamoto et al., 1997). The objective of this study was to determine whether twohit inactivation of hMSH3 and hMSH6 is frequent in carcinogenesis of the MMP pathway. We tested the hypothesis that tumors with monoallelic mutations in the mononucleotide tracts of these genes may in fact be homozygous, by the presence of additional somatic (or germline) mutations in the other allele. Here we report a low incidence of LOH and other mutations in the hMSH3 and hMSH6 genes in gastrointestinal tumors of the MMP. 2. Materials and methods 2.1. Tumor samples The tumors utilized in this study represent a consecutive series of unselected tumors without any obvious bias other than the availability after surgery of matching normal and tumor tissues from the same cancer patients (Malkhosyan et al., 1996; Rampino et al., 1997; Yamamoto et al., 1997; Schwartz et al., 1999). Cell lines were obtained from the ATCC. Genomic DNA was extracted with phenol-chloroform and diluted to a concentration of 20 ng/ml before PCR ampli®cation. Total RNA was isolated using RNA STAT-60 (TEL-TEST Inc., Friendswood, TX) and further puri®ed by DNase I (Boehringer Mannheim, Indianapolis, IL) digestion. 2.2. Microsatellite instability Somatic microsatellite alterations were analyzed by PCR as reported previously (Ionov et al., 1993). MMP 1 tumors were de®ned as those with somatic deletion mutations in mononucleotide repeats of (A)18 in APD3 and/or (A)26 in intron 5 of hMSH2 (BAT 26) and deletions or insertions of more than one repeated unit in dinucleotide microsatellite sequences (D1S158 and D8S199) using MapPairs primers (Research Genetics, Huntsville, AL). Tumors exhibiting sporadic dinucleotide microsatellite instability shifts of only one repeated unit were not considered MMP positive

Table 1 Frequency of hMSH3 and hMSH6 mutations in tumors of the MMP No. of tumor samples analyzed

Colon 32 Stomach 28 All 60 a b c d e

hMSH3 (%)

hMSH6 (%)

Germline a Somatic b Frameshift c Biallelic mutation

Monoallelic Germline a Somatic b Frameshift c Biallelic mutation mutation

1 (3.5) d 0 (0) 1 (1.6)

9 (28) 11 (39.3) 20 (33.3)

15 (47) 15 (47) 13 (46) 13 (46) 28 (46.6) 28 (46.6)

6 (19) 2 (7.1) 8 (13.3)

4 (12.5) e 0 (0) 4 (6.6)

12 (34) 12 (43) 24 (40)

Number of tumors with germline mutations. Number of tumors with somatic mutations. Number of tumors with frameshifts in the repeated tracts. Missense sequence variant. Three tumors with missense sequence variants and one tumor with a frameshift. See Table 3 for details.

12 (34) 12 (43) 24 (40)

5 (15.6) 2 (7.1) 7 (11.6)

Monoallelic mutation 7 (23.8) 10 (35.7) 17 (28.3)

N. Ohmiya et al. / Gene 272 (2001) 301±313

303

but MMP 1 /2 (Rampino et al., 1997; Perucho, 1999), also known as MSI-L (Boland et al., 1998). These tumors were analyzed with three additional dinucleotide repeats (D9S120, D5S394 and D5S399), two trinucleotide repeats (DM and ARI), and two tetranucleotide repeats (CSF1R-T and D7S1802).

of exon 23 was ampli®ed by PCR, digested with Hha I for 12 h at 378C, and separated by electrophoresis on 10% polyacrylamide gels.

2.3. Mutational analysis of hMSH6 and hMSH3

cDNA was synthesized using the SuperScripte preampli®cation system (Gibco BRL), and ampli®ed by PCR with primers 5 0 -AGAGTGTATCGCAGTGTT-3 0 and 5 0 -TCGTAATGCAAGGATGGCGT-3 0 for hMSH6, and 5 0 -TCTGTGCATCTCTGAAAATAAGG-3 0 and 5 0 -TGTGGCAGGCTGCACTCC-3 0 for hMSH3. These two sets of primers were designed to encompass introns 4 of hMSH6 and 7 of hMSH3 to exclude ampli®cation of contaminated DNA. Primers for DNA ampli®cation of the (C)8 and (A)8 tracks of hMSH6 and hMSH3, respectively, were described previously (Malkhosyan et al., 1996).

According to the exon/intron boundary sequences (GenBank accession numbers U73732±U73737), 21 sets of primers were designed to amplify the entire coding region, including each splicing site of hMSH6. Similarly, using the exon/intron boundary sequences (GenBank accession numbers D61398±D61419 and J04810), 24 sets of primers were designed to amplify the entire coding region, including each splicing site of hMSH3 except exons 12 and 14. Those two exons were not ampli®ed by PCR with primers designed in introns but with primers containing the 5 0 and 3 0 ends of exons. PCR was carried out using Taq DNA polymerase (Perkin Elmer, Foster City, CA) for one cycle of 948C for 4 min followed by 35 cycles of 948C for 30 s, 46±618C for 30 s and 728C for 30 s with a ®nal 7 min extension at 728C in the presence of 0.2 mCi of [ 32P]dCTP. PCR-SSCP was performed using both MDE (FMC BioProducts, Rockland, ME) gels and 5.5% non-denaturing polyacrylamide gels containing 7.5% glycerol. ELONGASEe Enzyme Mix (Gibco BRL, Gaithersburg, MD) was used for ampli®cation of long DNA templates. Regions encompassing the (C)8 and (A)8 tracks of hMSH6 and hMSH3, respectively, were ampli®ed by PCR with Vent DNA polymerase (New England Biolabs Inc., Beverly, MA). 2.4. DNA sequencing Sequencing was performed as described previously (Yamamoto et al., 1997; Rampino et al., 1997). The PCR product was eluted from the gels, reampli®ed, and subcloned into pCRTM2.1 (Invitrogen, San Diego, CA). Recombinant plasmids were sequenced by the dideoxy chain termination method using a Sequenase DNA sequencing kit (Amersham Life Science Inc., Arlington Heights, IL). DNA was also reampli®ed, puri®ed using the QIAquick PCR puri®cation kit (Qiagen, Santa Clarita, CA), and subjected to direct sequencing using the dsDNA cycle sequencing system (Gibco BRL). Mutations were con®rmed by the two sequencing methods. 2.5. hMSH6 and hMSH3 LOH For hMSH6, allele loss was evaluated by PCR-SSCP of polymorphisms at exons 1 and 3. For hMSH3, allele loss was evaluated on a SSCP gel of exons 4 and 21 (Benachenhou et al., 1998), PCR Sequagel-6 (National Diagnostics Inc., Atlanta, GA) of exon 1 (Benachenhou et al., 1998), and PCR-RFLP of exon 23 (Benachenhou et al., 1998) with polymorphisms. A 264 bp region encompassing codon 1045

2.6. Analysis of mutant and wild-type hMSH6 and hMSH3 alleles by PCR and RT-PCR

2.7. Protein expression Western blot analysis was performed on MSH3 and MSH6 proteins isolated from cell lines based on a method previously described (Ausubel et al., 1987). Brie¯y, proteins were extracted from cultured cells by cell lysis followed by centrifugation to eliminate cellular membrane and debris. The protein concentration was determined using a colorimetric assay (Pierce, Rockford, IL) and measuring the OD at 562 nm. Proteins were size fractionated by SDS-PAGE and human epidermoid carcinoma A431 nuclear extract (Santa Cruz Biotechnology Inc., Santa Cruz, CA) was used as a positive control. Proteins were transferred to a nylon PVDF membrane for blotting. Non-speci®c protein binding was blocked with 10% non-fat milk incubation for 1 h, followed by washing with a 1 £ TBST (20 mM Tris base, 150 mM sodium chloride, 0.1% Tween 20) solution. The blot was next incubated with diluted (1:100) primary antibody, rabbit anti-hMSH6 (Santa Cruz Biotechnology Inc., Santa Cruz, CA), at 48C overnight. Any unbound primary antibody was removed with a 1 £ TBST wash and the blot was incubated with diluted secondary antibody, 1:1000 anti-rabbit IgG HRP conjugate (Santa Cruz Biotechnology Inc., Santa Cruz, CA), for 1 h. Unbound secondary antibody was removed and MSH6 protein was detected with chemiluminescent substrate speci®c for HRP and X-ray ®lm. 3. Results 3.1. Somatic mutations at hMSH3 About half the MMP tumors contain frameshift mutations in a (A)8 track in codons 381±383 of hMSH3 (Malkhosyan et al., 1996; Yamamoto et al., 1997) (Table 1). We investigated the presence of other mutations by PCR-SSCP and sequencing. Colon tumor 442 harbored a somatic 2 bp deletion frameshift mutation in codon 626. This tumor lacked

304

N. Ohmiya et al. / Gene 272 (2001) 301±313

the frameshift mutation in the (A)8 tract. Gastric tumor J29 harbored a somatic 1 bp deletion (A)6 ! (A)5 in codons 737±739 creating a new termination codon 20 bp downstream. This tumor contained a 1 bp deletion in the (A)8 track (Fig. 1). LOH was observed in tumor J100 in addition to the 1 bp deletion frameshift in the (A)8 track (Fig. 2). LOH within hMSH3 occurred in 4/19 (21%) informative MMP1 gastric tumors. No LOH was detected in any of the MMP1 colon tumors.

There were no signi®cant differences in allele frequency between cancer patients and cancer-free individuals (data not shown). Colon tumor 440 harbored a germline missense variant (Pro ! Ser at codon 681), a residue conserved in the mouse MSH3 homologue (Fig. 3). This variant was not detected in 190 individuals, including 66 other MMP 1 , 15 MMP1/2 and 108 MMP negative tumors.

3.2. Germline variants of hMSH3

Somatic frameshift mutations in a (C)8 track of the hMSH6 gene (codons 1116±1118 of exon 5) are frequent in gastrointestinal cancers of the MMP (Malkhosyan et al., 1996; Yamamoto et al., 1997) (Table 1). We also screened our panel of MMP positive tumors for mutations in all exons of hMSH6 by PCR-SSCP and sequencing to determine the extent of two-hit inactivation (Table 2). Colon tumor 377 contained a C ! T transition at residue 240 generating a stop codon, in addition to a 1 bp somatic deletion frameshift in the (C)8 track (Fig. 4). Colon tumor 426 harbored a somatic missense mutation Cys ! Arg at codon 1158, conserved among the mouse and yeast homologues. Gastric tumor A85 contained an Arg ! Gln transition at codon 772. In gastric tumor J32 a somatic transversion Asp ! Val was identi®ed at codon 1031, conserved only in the mouse homologue. A somatic Val ! Ala transition was detected at codon 800 in gastric tumor J7, and was not conserved in either the mouse or yeast homologues. However, this residue has been found to be mutated in a HNPCC family and has been concluded to be functional (Obmolova et al.,

The sequencing data permitted estimation of the frequency of hMSH3 germline mutations and sequence variants (polymorphisms) among these cancer patients. This information was compared with additional data from other cancer patients without microsatellite instability and from cancer-free individuals. The gene has a variable region at its 5 0 end with three, four, ®ve, six, and seven copies of a 9 bp repeating unit in exon 1 (Benachenhou et al., 1998). The frequencies of alleles with three, four, ®ve, six, and seven copies were 58/238 (24.4%), 2/238 (0.8%), 3/238 (1.3%), 156/238 (65.5%), and 19/238 (8.0%), respectively. We detected other previously described polymorphisms. Exons 21 and 23 contained germline amino acid changes (Arg ! Gln at 949 and Thr ! Ala at 1045) (Benachenhou et al., 1998). We also detected a silent polymorphism at exon 4 (G ! A at 723) that has not been described. The frequencies of polymorphic alleles in exons 4, 21, and 23 were 25/192 (13.0%), 35/372 (9.4%), and 69/370 (18.6%), respectively.

3.3. Somatic mutations at hMSH6

Fig. 1. Somatic hMSH3 mutations in gastric tumor J29. (A) SSCP and sequencing analysis of a somatic frameshift deletion mutation in exon 15 of hMSH3. (B) PCR analysis of somatic frameshift mutation at the (A)8 hotspot in exon 5. N, normal tissue; T, tumor tissue.

N. Ohmiya et al. / Gene 272 (2001) 301±313

305

Fig. 2. Somatic hMSH3 mutations and LOH in gastric tumor J100. (A) PCR-RFLP analysis of exon 23. The 264 bp DNA fragment ampli®ed from the polymorphic allele is cleaved to form fragments of 42 and 222 bp by Hha I digestion, whereas that from the normal allele is not. Loss of the polymorphic allele is shown in tumor J100. A 600 bp fragment of the human homeobox HOX 4A gene which is cleaved to various fragments of 96±11 bp by Hha I digestion was ampli®ed with primers 5 0 -CACTTCAACCGCTACTTGTG-3 0 and 5 0 -GGCCGAGAGATCTGTGTAGGT-3 0 , and added to the solution to con®rm complete digestion. M, DNA molecular weight marker. (B) PCR analysis of a somatic frameshift mutation at the (A)8 hotspot. N, normal tissue; T, tumor tissue.

2000). Colon tumor 211 harbored a silent (T ! C) somatic mutation at codon 1248. 3.4. Germline mutations at hMSH6 Exon 1 contained a polymorphism (A or G at position 116) that resulted in two different amino acids (Glu or Gly at codon 39) and a silent change (C or A at position 186). The frequencies of alleles G116/C186, G116/A186, and A116/ C186 were 154/222 (69.4%), 43/222 (19.4%), and 25/222 (11.3%), respectively. Exon 3 contained a silent polymorphism (T ! C at position 540) in 9/55 cases (16%). We found three other silent polymorphisms (A ! G at 276) in 1/79 (1.2%), C ! T at 642 in 2/79 (2.5%), and T ! A at 3306 in 6/79 (7.6%) cases. We also found a polymorphism (TTGA insertion) at codon 1356 in 8/155 (5.1%) as described previously (Miyaki et al., 1997). Colon tumor 441 harbored both a germline 4 bp deletion frameshift mutation at codons 602±603 and a somatic 5 bp deletion frameshift mutation at codons 1010±1011 (Fig. 5). Cloning and sequencing of PCR fragments encompassing codons 602±1011 and 1010±1118 revealed that the germline 4 bp deletion frameshift mutation and the somatic 5 bp deletion frameshift mutation were on different alleles. A germline missense Gly ! Ala mutation in colon tumor 453 was located in codon 685, conserved in mouse and yeast MSH6 (Fig. 3). This tumor also contained a 1 bp insertion at the (C)8 tract. Colon tumor 435 without the (C)8 frameshift harbored both a germline Val ! Ala transition at codon 878, conserved in the mouse homologue, and a somatic

Arg ! Gln transition at codon 772, conserved among both the mouse and yeast homologues (Fig. 3). Cloning and sequencing of PCR fragments encompassing codons 772± 878 revealed that these two transitions were not on the same allele (data not shown). Colon tumor 397 harbored a germline Lys ! Met transversion in codon 854, conserved only in the mouse homologue. The germline sequence variants mentioned above were not observed in 119 additional individuals analyzed. LOH was observed in gastric tumor A57 in addition to a 1 bp insertion frameshift in the (C)8 track (data not shown). LOH within hMSH6 occurred in 2/8 (25%) informative MMP1 gastric tumors. However, no LOH was observed in MMP1 colon tumors. 3.5. Microsatellite instability in tumors with hMSH6 mutations The role of hMSH6 in MMR appears to be speci®c for single-base mismatches and single-base loops, but not for larger loops. Therefore, the detection of hMSH6 mutations in tumors initially classi®ed as MMP positive raised the question of whether these mutations were responsible for their microsatellite instability. Some of the tumors with hMSH6 mutations (Table 1) also contained genetic or epigenetic inactivation of other members of the MMR gene family (Table 2). Thus, colon tumor 453, with a germline variant and a somatic frameshift mutation at hMSH6, also contained germline and somatic mutations at hMSH2, LOH at hMLH1, and a somatic frameshift at hMSH3. Tumor 397, with a germline variant at hMSH6, also contained a mono-

306

N. Ohmiya et al. / Gene 272 (2001) 301±313

Fig. 3. Localization of germline (G) and somatic (S) hMSH3 and hMSH6 mutations within the sequence alignments of DNA MMR proteins MSH2, MSH3 and MSH6 from yeast (y), mouse (m) and human (h). The ®gure is adapted from Obmolova et al. (2000) on a 3D structure of the Termophilus aquaticus (TAQ) MutS protein (domains depicted on top of the aligned sequences). Residues in white with a black background represent residues altered in colon and gastric tumors of the MMP (Table 2). Germline sequence variants and somatic mutations are shown in white with a black background below the sequence alignments. Numbers immediately below the mutations represent codons at which alteration occurs. White residues with a dark gray background represent functional missense mutations as determined by yeast assays for MMR or by detection in HNPCC families. Light gray areas represent conserved domains for structural integrity; dark gray, interdomain recognition. Sequences of MSH6 conserved among the three species are highlighted in medium gray.

Fig. 4. Somatic hMSH6 mutations in colon tumor 377. (A) SSCP and sequencing analysis of a somatic missense (Arg ! Stop) mutation in exon 4. (B) PCR analysis of a somatic frameshift mutation at the (C)8 hotspot in exon 5. Symbols as in Fig. 1. Arrowheads indicate mutant bands.

N. Ohmiya et al. / Gene 272 (2001) 301±313

307

Fig. 5. Germline and somatic hMSH6 mutations in colon tumor 441. (A) SSCP and sequencing analysis of a germline frameshift 4 bp (AAAG) deletion mutation in exon 4. (B) SSCP and sequencing analysis of a somatic frameshift 5 bp (TATCA) deletion mutation in exon 4. N, normal tissue; T, tumor tissue. Arrowheads indicate mutant bands.

Table 2 hMSH6 mutations in tumors of the MMP Case

Tumor

Mutation

Exon

Nucleotide

Codon

bp alteration

Amino acid

Domain a

Function b

377 441

Colon Colon

ET153 453 J7 435

Endometrium Colon Stomach Colon

A85 397 J32 426

Stomach Colon Stomach Colon

Somatic Germline Somatic Somatic Germline Somatic Germline Somatic Somatic Germline Somatic Somatic

4 4 4 4 4 4 4 4 4 4 4 6

718 1806 3028 2389 2054 2399 2633 2315 2315 2561 3092 3472

240 602 1010 797 685 800 878 772 772 854 1031 1158

CGA ! TGA TCAAAG ! TC ACTATT ! T GAC ! AC GGT ! GCT GTT ! GCT GTT ! GCT CGG ! CAG CGG ! CAG AAG ! ATG GAC ! GTC TGT ! CGT

Arg ! Stop Frameshift Frameshift Frameshift Gly ! Ala Val ! Ala Val ! Ala Arg ! Gln Arg ! Gln Lys ! Met Asp ! Val Cys ! Arg

N-term II4-IIb IVb IIId IId IIId IIIg-h IIId IIId IIIg IVb Va-V5

Connector/dimerization Lever/interdomain interaction Core/structural integrity Connector/dimerization Core/structural integrity Core/structural integrity Core/structural integrity Core/structural integrity Core/structural integrity Lever/interdomain interaction ATPase

a According to amino acid sequence homology with the bacterial MutS gene (see Fig. 3) and based on its 3D structure (Obmolova et al., 2000; Lamers et al., 2000). b According to the proposed function of the bacterial MutS gene based on its structural features (Obmolova et al., 2000; Lamers et al., 2000).

308

N. Ohmiya et al. / Gene 272 (2001) 301±313

Table 3 hMSH3 and hMSH6 mutations in tumors of the MMP a

with mutant bands (i.e. (A)7 for hMSH3, (C)9 or (C)7 for hMSH6) stronger than the wild-type bands ((A)8 for hMSH3 and (C)8 for hMSH6) were considered to carry biallelic mutations (data not shown, and see Yamamoto et al., 1997, 2000; Schwartz et al., 1999). When the mutant band was of an intensity similar to normal, and the tumor had little contaminating normal tissue, the mutation was considered monoallelic (see example J29 in Figs. 1B, 2B, 4B and 6). When the mutant band was fainter than the normal and the tumor had signi®cant contaminating normal tissue, the mutations were also considered monoallelic. Ambiguous cases could be resolved sometimes by contrasting the mutation pattern of hMSH3 or hMSH6 (or MBD4) with the patterns of frameshifts in other genes such as TGFb RII or Bax (Yamamoto et al., 1997).

allelic somatic mutation at hMLH1. However, three other tumors with hMSH6 mutations contained no other detectable MMR mutations or epigenetic alterations. We analyzed the extent of microsatellite instability in these three tumors: colon tumor 441, which harbored both germline and somatic frameshifts, colon tumor 435, with germline and somatic missense mutations, and gastric tumor J32, which contained a somatic frameshift and LOH (Table 2). All three tumors exhibited microsatellite alterations in the markers analyzed, including mononucleotide and also dinucleotide, trinucleotide, and tetranucleotide repeats. Therefore, tumors with hMSH6 inactivation but without other detectable MMR alterations display enhanced microsatellite instability not only at mononucleotide repeats. 3.6. Absence of hMSH6 mutations in tumors with low instability (MMP 1 /2 or MSI-L)

a

Column 1: colon (C), stomach (S) or endometrium (E) tumors of the MMP positive for hMSH3 and/or hMSH6 frameshift mutations at the (A)8 and (C)8 hotspots, which were analyzed also for mutations in the other allele by SSCP. Column 3: hMSH3 mutations. The table is sorted into tumors with biallelic mutations, monoallelic mutations and no mutations. Column 4: hMSH6 mutations (see Table 1). Column 5: hMSH2 mutations (data from Malkhosyan et al., unpublished data). Includes a germline splicing mutation and a somatic frameshift of colon cancer case 453 and somatic isoleucine deletion of case 440, a frameshift mutation of stomach cancer case J100T and a missense mutation of stomach cancer case J93T. Column 6: hMLH1 mutations (all somatic). Includes splicing mutations of colon cancer cases 397 and 405, and a somatic nonsense mutation of stomach cancer case J2. Column 7: somatic MBD4 frameshift mutations at its (A)10 hotspot (Malkhosyan et al., unpublished data). Black box, biallelic mutation (includes (A)8 and (C)8 frameshifts); gray box, monoallelic mutation; clear box, no mutation. *Missense mutation; **LOH; ***missense mutation plus LOH. G, germline mutation; G*, germline sequence variant resulting in an amino acid change; f, frameshift not at hotspots; m, gene silencing by promoter methylation; n, nonsense mutation; NA, not analyzed. The allelic status of the mutations was determined after estimating the extent of contaminating normal tissue. This was done by comparing the relative intensity of the bands corresponding to the wild-type and mutant alleles of BAT-26 and APD3 mononucleotide repeats (Yamamoto et al., 1997). Cases

We analyzed 15 colon tumors with occasional alterations in dinucleotide, but not in mononucleotide microsatellites (MMP1/2 or MSI-L). We tested the possibility that such low microsatellite instability could be due to hMSH6 inactivation. However, no somatic mutations in hMSH6 were found in any of the 15 tumors. These results are in accordance with Parc et al. (2000), who also did not ®nd an association between hMSH6 mutations and low instability. One tumor contained a germline C ! T transition resulting in a missense Pro ! Ser at codon 1087 that is only conserved in the mouse gene (data not shown). No microsatellite alterations were found at the additional markers examined in this tumor (data not shown). Altogether, these results add strong support to the concept that the low instability of MMP1/2 (or MSI-L) tumors does not represent true instability, but spontaneous errors of replication in these unstable repetitive sequences that are detected by the clonality of the tumors, without a functional link with cancer pathogenesis (Perucho, 1999). 3.7. Relative expression of hMH3 and hMSH6 wild-type and mutant alleles The results of the mutation analysis indicated that the incidence of biallelic mutations in both hMSH3 and hMSH6 was very low in MMP positive tumors (Table 3).

N. Ohmiya et al. / Gene 272 (2001) 301±313

This raised the issue of the functional signi®cance of these mutations. They could have a dominant negative effect, by interfering with the stoichiometry of MMR in vivo. Frame-

309

shifts in genes often destabilize the mRNA. If this were the case, the possible dominant negative effect of such frameshift mutants could be ruled out. To determine whether the hMSH6 or hMSH3 frameshift mutations in¯uenced RNA stability, we compared the relative proportion of mutant and normal alleles at the hotspots by PCR ampli®cation from DNA and mRNA (Fig. 6A). Colon cancer cell line HCT116 and prostate tumor cell line DU145 harbored homozygous and heterozygous 1 bp deletion frameshift mutation in hMSH3, respectively. HCT8 and AN3CA cell lines contained only wild-type alleles. The mutant allele of HCT116 cells was transcribed. Although the mutant transcript appeared to be underrepresented relative to the wild-type, DU145 cells produced both mutant and wild-type RNA transcripts. Cell line HCT116 harbored a heterozygous 1 bp insertion frameshift mutation in hMSH6 while endometrial cell line AN3CA harbored a heterozygous 1 bp deletion frameshift mutation. Cell lines HCT8 and DU145 were wild-type for hMSH6. Cell lines HCT116 and AN3CA produced both mutant and wild-type RNA transcripts, although the mutant allele transcript was less abundant than the wild-type. Therefore, the frameshift mutations in both hMSH3 and hMSH6 did not signi®cantly destabilize the mRNA. 3.8. Frameshift mutations abolish hMSH6 protein expression

Fig. 6. Relative expression of wild-type and mutant hMSH3 and hMSH6 alleles. (A) RT-PCR and PCR analysis of the indicated human tumor cell lines. All are derived from colon cancers except AN3CA which is from endometrium and DU145 which is from prostate. HCT8 and AN3CA are wild-type for hMSH3 (A)8 and HCT8 and DU145 for hMSH6 (C)8. HCT116 is homozygous, while DU145 and LS180 are heterozygous for mutant (A)7 hMSH3. LS180 and AN3CA are heterozygous for mutant (C)7 hMSH6 while HCT116 is heterozygous for mutant (C)9 hMSH6. Both wild-type and mutant alleles are expressed although the mutant allele appears underrepresented in RNA compared with DNA in both genes (compare the relative intensities of mutant and wild-type alleles in the patterns obtained from DNA and from RNA). (B) PCR and Western analysis for hMSH6 of human colon tumor cell lines LS180, SW620, and SW48. Several single cell clones, wild-type (8:8), homozygous mutant (7:9), and heterozygous mutant (8:9), for hMSH6 of the SW48 cell line are compared. Protein expression analysis in the three SW48 clones reveals an absence of wildtype protein in the homozygous mutant (7:9) clones. The heterozygous clones express less protein than the wild-type clones. (Bottom) Western blot with a longer exposure time. WT, wild-type allele; MUT, mutant allele; M, protein molecular weight marker with bands at 132, 90, and 55 kDa; 160 kDa, molecular weight of wild-type hMSH6 protein; 127 kDa, theoretical size of truncated hMSH6 protein. No traces of this putative truncated polypeptide are detectable.

Colon tumor cell line SW48 harbors a heterozygous 1 bp insertion frameshift mutation in hMSH6. We isolated two subclones of the SW48 cell line. One clone harbored wildtype hMSH6 alleles, while a second clone harbored a homozygous mutation involving a 1 bp insertion in the microsatellite region on one allele, in addition to a 1 bp deletion on the other (Fig. 6B). The existence of these three SW48 clones, wild-type (8:8), heterozygous mutant (8:9), and homozygous mutant (7:9), for hMSH6 was useful to study the effect of hMSH6 heterozygous versus homozygous mutations at both the level of RNA and protein expression. Western blots performed on hMSH6 protein from these cell line subclones revealed that the frameshift mutations abolished protein expression (Fig. 6B). Only wild-type hMSH6 protein was visible in all SW48 clones with wildtype alleles. The heterozygous mutant SW48 clone (8:9) produced about half the amount of protein, and the homozygous mutant (7:9) produced no visible protein. Full-size hMSH6 has a molecular weight of 160 kDa while its mutant, truncated product (1 bp deletion or insertion at codon 1116) has a molecular weight of 127 kDa. This truncated polypeptide was not visible even after longer exposure (Fig. 6B, bottom). Therefore, the possible dominant negative effect of hMSH6 frameshift mutations at the (C)8 hotspot can be ruled out.

310

N. Ohmiya et al. / Gene 272 (2001) 301±313

4. Discussion Frameshift mutations in the (C)8 track of the hMSH6 gene and the (A)8 track of the hMSH3 gene are common in cancer of the MMP (ranging from 15 to 65% depending on the gene and tumor type). This relatively high frequency of hMSH3 and hMSH6 mutations and the absence (or lower incidence) of mutations in other genes with similar repeated sequences (Malkhosyan et al., 1996; Yamamoto et al., 1997; Rampino et al., 1997) suggests that these frameshift mutations are functional during tumor progression. This interpretation is reinforced by the ®nding that mutations other than those in the (C)8 and (A)8 tracks also occurred in tumors of the MMP, albeit with low frequency. To explain these frameshift mutations we proposed a `mutator that mutates another mutator' model (Malkhosyan et al., 1996; Perucho, 1996) for the gradual manifestation of the mutator phenotype by these tumors. The concept that a mutator phenotype may target the same mutator gene that generates the mutator phenotype is not new, and was proposed long ago on theoretical grounds (Loeb et al., 1974). However, because the mutator genes are recessive, the auto-targeting of the mutator gene by itself does not seem to occur. According to our model, primary mutators (i.e. hMSH2 and hMLH1) generate a mutator phenotype which often targets other MMR genes such as hMSH3 and hMSH6 (secondary mutators). These secondary mutator mutations may contribute to tumor progression by increasing the width or depth of the mutator phenotype. In support of this model, we have obtained evidence that tumor cells with inactivated hMLH1 and homozygous hMSH6 (C)8 frameshift mutations exhibit a higher mutation frequency (the depth of the mutator phenotype) and a spectrum of mutations signi®cantly different (the width of the mutator phenotype) relative to sister cell clones with only hMLH1 inactivation and wild-type hMSH6 (Baranovskaya et al., unpublished data). Some tumors had no identi®able MMR gene mutations other than the frameshifts at the (A)8 and (C)8 hotspots (colon tumors 205, 334, 328, and 367, and gastric tumors J100, J7, J64 and J32, see Table 2). The role of hMSH3 and hMSH6 as primary mutators in these tumors is supported by the detection of LOH in some tumors, especially because MMP positive tumors are known to display a very low incidence of LOH (Aaltonen et al., 1993; Ionov et al., 1993; Thibodeau et al., 1993). This interpretation is based on the postulate that the primary mutator mutation occurs in the absence of a mutator phenotype. Consequently, the mechanism of inactivation of a primary mutator gene must be identical to the mechanisms of inactivation of other cancer genes (i.e. tumor suppressors) in tumors without the mutator phenotype. Thus, hMLH1 can be inactivated by mutation, LOH or epigenetic silencing, in all possible combinations, in tumors of the microsatellite mutator phenotype, irrespective of their hereditary or sporadic origins (Table 3 and unpublished data).

The relationship between hereditary and sporadic MMP tumors has been controversial. We had proposed that hereditary and non-hereditary tumors of the MMP are essentially indistinguishable in genotype and phenotype, with the exception of the presence or absence of inactivating germline mutator mutations (Perucho et al., 1994; Perucho, 1996, 1999). In contrast, others have stressed some divergent features between these two kinds of tumors (Aaltonen et al., 1993; Kinzler and Vogelstein, 1996; Boland et al., 1998; Kuismanen et al., 2000). But in a more recent publication, some of these authors acknowledged that microsatellite unstable colorectal tumors from HNPCC patients and those from sporadic cases are `largely similar but not identical' (Peltomaki et al., 2000). Hypermethylation of hMLH1 in sporadic and hereditary cancers of the MMP is well documented, with incidences in general higher in sporadic than in hereditary tumors (Kane et al., 1997). This is not surprising since hereditary cancers already carry one mutated allele and somatic biallelic inactivation may be more easily achieved by epigenetic than by genetic means. The preferential distribution of MMP tumors in the proximal colon could be explained by the predominance of tumors with methylated genes, such as hMLH1 in this part of the colon. This has led to the proposal that a hypermethylator phenotype underlies the mutator phenotype which in turn underlies tumorigenesis (Toyota et al., 1999). However, the concept of a methylator phenotype awaits rati®cation in the absence of a clear boundary in genotype or phenotype separating tumors with and without this hypermethylator phenotype. This is in contrast with colon cancers with and without the MMP, which segregate into two groups with a well demarcated bimodal distribution (Perucho et al., 1994). In addition to these somatic frameshift mutations, several germline frameshift mutations and missense variants were detected in hMSH6 and hMSH3 (Tables 2 and 3). The inactivating germline mutation in hMSH6 of patient 441, with no obvious family history of cancer, nor mutations of hMSH2, hMSH3, or hMLH1 (Table 2), con®rms the pathogenic role of hMSH6 for some HNPCC kindreds (Akiyama et al., 1997; Miyaki et al., 1997). Analysis of microsatellite instability detected a few alterations in di-, tri-, and tetranucleotide repeats in tumor 441. MSH6 forms a complex with MSH2. Generally, this MSH2±MSH6 protein complex ef®ciently binds single-base substitution and single-base insertion/deletion mispairs, whereas the MSH2±MSH3 protein complex also binds larger base insertion/deletion mispairs (Kolodner, 1996). Other studies using cell extracts have shown that the MSH2±MSH6 protein complex ef®ciently binds insertion/deletion heterologues with one, but also two and three unpaired nucleotides (Drummond et al., 1995). This may be the cause of the low dinucleotide and trinucleotide repeat instability in case 441. However, other unidenti®ed components for MMR may be also involved in this instability, particularly in cases 435 and J32, which contained frequent instability of insertion/deletion mispairs

N. Ohmiya et al. / Gene 272 (2001) 301±313

longer than one nucleotide, but had no mutated MMR genes other than hMSH6 (Table 3). Recently the crystal structure of E. coli DNA MMR protein MutS was reported (Obmolova et al., 2000; Lamers et al., 2000). In addition to conserved domains of MutS that interact with substrate DNA and affect structural integrity, other theoretical functions were also disclosed such as DNA recognition, protein dimerization, ATPase activity, and interdomain interactions. This information has important implications for understanding how mutations in these MMR genes may alter their binding abilities within heterodimeric complexes and play a pathogenic role in HNPCC. In their paper, Obmolova et al. (2000) provided a structurebased sequence alignment of MutS homologues (MSH2, MSH3, MSH4, MSH5, and MSH6) from yeast and humans. Placement of the germline variants and somatic missense mutations described in this paper (Table 2) within the published sequence alignments for these MMR genes reveals some clues to their possible functionality (Fig. 3). Some of these hMSH6 somatic missense mutations may be functional. For instance, the same somatic missense mutation at codon 800 found in gastric tumor J7 has been previously found as a germline variant in a HNPCC family (Fig. 3). Other somatic missense mutations also fell into very well conserved protein regions that are structurally important. Thus, the same Arg ! Gln mutation at codon 772 was identi®ed in two unrelated tumors, one of which also carried a germline missense variant (case 435, Tables 2 and 3). Changes at the same Arg 772 residue have been shown to be functional in the bacterial MutS protein (Fig. 3). In the 3D structure of TAQ MutS protein, the homologue amino acid (Arg 286 in domain IIIc) interacts through two hydrogen bonds with the Asp at position 331 in domain IIIf, located at about 3 Amstrongs, and the substitution by Gln will certainly disrupt this interaction. In gastric cancer J32 an Asp ! Val somatic change occurred at codon 1031 which is located in the IVb domain, close to the point of interaction with DNA. This tumor also had undergone LOH at the MSH6 gene region, and had no other detectable MMR gene mutations, although it harbored a monoallelic frameshift mutation at the MBD4 gene (Malkhosyan et al., unpublished data and Table 3). Finally, the 1158 Cys ! Arg mutation of case 426 occurs in an extremely conserved region at the ATPase domain, and the adjacent Gly has been shown to be mutated in the hMSH2 gene in a HNPCC kindred (Fig. 3). In the 3D structure of the bacterial protein, the 1158 residue is nested inside several aliphatic amino acids (L288, L270, V529, and L287) and the mutant Arg would be buried and not able to fold properly. It is dif®cult to determine whether the germline sequence variants at hMSH6 and hMSH3 are mutations or neutral polymorphisms (Table 2). Some missense variants in MSH2 have been shown to affect DNA repair activity representing therefore important but subtle differences in cancer susceptibility (Drotschmann et al., 1999). The nucleotide differences detected in hMSH3 and hMSH6

311

have not been published, and occur at frequencies of less than 1%. A transversion leading to a germline change (Gly ! Ala) at codon 685 in hMSH6 lies within structural domain IId, in between conserved residues for structural integrity (Fig. 3a). This amino acid is conserved among the yeast homologues of MSH3 and MSH6. However, it is unclear whether this substitution of glicine for alanine affects protein structure or function, because alanine is found at this position in the hMSH3 gene. Another germline variant (Lys ! Met) at codon 854 in hMSH6, which is conserved among the three genes of known sequence, lies within structural domain IIIg, adjacent to a conserved residue important for structural integrity (Fig. 3b). The close localization of this amino acid to conserved residues that impact structural integrity suggests that it may be a functional mutation. The Val ! Ala germline variant at codon 878 does not fall into a conserved region in MMR genes and its possible functionality remains obscure. However, this region represents the `hinge' between the ATPase domain and the domain interacting with other MMR proteins on the one hand, and the domains interacting with DNA on the other. Mutations in this region may affect the alosteric bendability of the protein required to allow domain IV to bind the mismatched nucleotides in the DNA molecule. The detection of another somatic hMSH6 missense mutation in the same tumor also supports the functionality of this germline variant. We only identi®ed a sequence variant at hMSH3. This Pro ! Ser germline variant resides in a non-conserved region, and because the amino acid present at this position in the yeast gene is threonine, which is structurally similar to the variant serine (Fig. 3), it is likely that this variant represents a rare polymorphism that lacks functional signi®cance. While homozygous mutations in secondary mutators are functional, by increasing the mutator phenotype of the tumor cells, in many of these tumors, the hMSH3 and hMSH6 frameshift mutations appeared to be present in only one allele (Table 3). The ®rst explanation for these data is that heterozygous somatic frameshift mutations may be inconsequential to tumorigenesis, and simply the result of the increased mutation rates of these tumor cells. Another possibility is that these mutations may create dominant negative proteins. We showed that these mutations do not abolish gene expression, although the frameshift mutations at the hotspots in both hMSH3 and hMSH6 slightly diminished RNA stability. Nevertheless, only wildtype gene products were detected in the hMSH6 mutants by Western blotting techniques and the amount of product correlated with genotype. This means that the truncated protein is unstable and therefore cannot act as a dominant negative. Similar experiments could not be performed for hMSH3, because of the lack of antibodies, and it remains to be established whether truncated hMSH3 is stable or not. The prospect that apparent monoallelic mutations are indeed biallelic, because of epigenetic silencing of the other allele, is interesting but requires supporting experimen-

312

N. Ohmiya et al. / Gene 272 (2001) 301±313

tal evidence. It is also possible that heterozygous mutations in the hMSH6 and hMSH3 genes may contribute to tumor genomic instability, especially in conjunction with the concomitant mutations in other MMR genes. We have proposed an accumulative haploinsuf®ciency model in an attempt to account for the large number of monoallelic mutations found in genes involved in cell growth and survival (Schwartz et al., 1999; Yamamoto et al., 2000). Similarly, heterozygous MMR mutations likely reduce the amount of gene products involved in the repair process. This alone may be suf®cient to decrease replication ®delity. In cell lines where the dihydrofolate reductase locus (localized adjacent and divergent to the MSH3 locus) was ampli®ed by methotrexate, concomitant overexpression of the hMSH3 gene signi®cantly altered the relative concentration of hMSH2± hMSH6 and hMSH2±hMSH3 complexes by sequestering hMSH2, preferentially in the latter heterodimer. As the hMSH2±hMSH3 complex does not bind to base±base mispairs, one would predict that these cell lines would be similar to those lacking hMSH6 in demonstrating microsatellite instability (Marra et al., 1998). Thus, decreased MMR ef®ciency may occur as a result of imbalances in the relative amounts of hMSH3 and hMSH6 proteins. Moreover, several tumors only harbored monoallelic mutations in hMSH2 or hMLH1 (colon cancers 61, 211, 405, 442, and 397, and gastric cancers J36, J73, and J93, Table 2; Malkhosyan et al., unpublished data). While mutations in some of these tumors might have been missed by our SSCP approach, it seems likely that some tumors are altogether lacking biallelic mutations. Indeed, a review of the literature also reveals that the biallelic nature of MMR gene mutations in cancer of the MMP might not be the norm, but rather the exception (Liu et al., 1995; Thibodeau et al., 1996; Wu et al., 1997; Percesepe et al., 1998). We propose that haploinsuf®ciency in hMMR may lead to the manifestation of a transient and weak mutator phenotype which is eventually stabilized by the accumulation of additional mutator mutations in the hMMR gene family or in other unknown equivalents. The hypothesis particularly applies to tumors without biallelic mutations in the primary mutators, to tumors with mutations in primary mutators which do not completely abolish protein activity (such as splicing mutations), and to tumors where the manifestation of the mutator phenotype by the primary mutator mutation must be by itself a gradual process (such as epigenetic silencing). Acknowledgements We thank Drs Teruhiko Yoshida and Masaaki Terada (National Cancer Center Research Institute), Drs Fumio Itoh and Kohzoh Imai (Sapporo Medical University) and Drs Chanil Park and Hoguen Kim (Yonsei University) for providing DNA and tissue samples. We also appreciate the advice on protein structure by Drs Russell Doolittle and Ana

Rojas (University California, San Diego, CA). This work was supported by NIH grants CA63585 and CA38579 and the Cancer Research Fund under Interagency Agreement #9712013 (University of California Contract #98-00924V) with the Department of Health Services, Cancer Research Program.

References Aaltonen, L.A., Peltomaki, P., Leach, F.S., Sistonen, P., Pylkkanen, L., Mecklin, J.P., Javinen, H., Powell, S.M., Jen, J., Hamilton, S.R., Petersen, G.M., Kinzler, K.W., Vogelstein, B., de la Chapelle, A., 1993. Clues to the pathogenesis of familial colorectal cancer. Science 260, 812±816. Akiyama, Y., Sato, H., Yamada, T., Nagasaki, H., Tsuchiya, A., Abe, R., Yuasa, Y., 1997. Germ-line mutation of the hMSH6/GTBP gene in an atypical hereditary nonpolyposis colorectal cancer kindred. Cancer Res. 57, 3920±3923. Ausubel, F.M., Brent, R., Kingston, R.E., Moore, D.D., Smith, J.A., Seidman, J.G., Struhl, K., 1987. Current Protocols in Molecular Biology. Green Publishing Associates and Wiley-Interscience, New York. Benachenhou, N., Guiral, S., Flipot, I.G., Labuda, D., Sinnett, D., 1998. High resolution deletion mapping reveals frequent allelic losses at the DNA mismatch repair loci hMLH1 and hMSH3 in non-small cell lung cancer. Int. J. Cancer 77, 173±180. Boland, C.R., Thibodeau, S.N., Hamilton, S.R., Sidransky, D., Eshleman, J.R., Burt, R., Meltzer, S.J., Rodriguez-Bigas, M.A., Fodde, R., Ranzani, N., Srivastava, S.A., 1998. National Cancer Institute workshop on microsatellite instability for cancer detection and familial predisposition: development of international criteria for the determination of microsatellite instability in colorectal cancer. Cancer Res. 58, 5248±5257. Casares, S., Ionov, Y., Ge, H.Y., Stanbridge, E., Perucho, M., 1995. The microsatellite mutator phenotype of colon cancer cells is often recessive. Oncogene 11, 2303±2310. Drotschmann, K., Clark, A.B., Kunkel, T.A., 1999. Mutator phenotypes of common polymorphisms and missense mutations in MSH2. Curr. Biol. 9, 907±910. Drummond, J.T., Li, G.M., Longley, M.J., Modrich, P., 1995. Isolation of an hMSH2-p160 heterodimer that restores DNA mismatch repair to tumor cells. Science 268, 1909±1912. Ionov, Y., Peinado, M.A., Malkhosyan, S., Shibata, D., Perucho, M., 1993. Ubiquitous somatic mutations in simple repeated sequences reveal a new mechanism for colonic carcinogenesis. Nature 363, 558±561. Kane, M.F., Loda, M., Gaida, G.M., Lipman, J., Mishra, R., Goldman, H., Jessup, J.M., Kolodner, R., 1997. Methylation of the hMLH1 promoter correlates with lack of expression of hMLH1 in sporadic colon tumors and mismatch repair-defective human tumor cell lines. Cancer Res. 57, 808±811. Kinzler, K.W., Vogelstein, B., 1996. Lessons from hereditary colorectal cancer. Cell 87, 159±170. Kolodner, R., 1996. Biochemistry and genetics of eukaryotic mismatch repair. Genes Dev. 10, 1433±1442. Kuismanen, S.A., Holmberg, M.T., Salovaara, R., Schweizer, P., Aaltonen, L.A., de la Chapelle, A., Nystrom-Lahti, M., Peltomaki, P., 2000. Epigenetic phenotypes distinguish microsatellite-stable and -unstable colorectal cancers. Proc. Natl. Acad. Sci. USA 96, 12661±12666. Lamers, M.H., Perrakis, A., Enzlin, J.H., Winterwerp, H.H., de Wind, N., Sixma, T.K., 2000. The crystal structure of DNA mismatch repair protein MutS binding to a G x T mismatch. Nature 407, 711±717. Liu, B., Nicolaides, N.C., Markowitz, S., Willson, J.K., Parsons, R.E., Jen, J., Papadopolous, N., Peltomaki, P., de la Chapelle, A., Hamilton, S.R., Kinzler, K.W., Vogelstein, B., 1995. Mismatch repair gene defects in

N. Ohmiya et al. / Gene 272 (2001) 301±313 sporadic colorectal cancers with microsatellite instability. Nat. Genet. 9, 48±55. Loeb, L.A., Springgate, C.F., Battula, N., 1974. Errors in DNA replication as a basis of malignant changes. Cancer Res. 34, 2311±2321. Malkhosyan, S., Rampino, N., Yamamoto, H., Perucho, M., 1996. Frameshift mutator mutations. Nature 382, 499±500. Marra, G., Iaccarino, I., Lettieri, T., Roscilli, G., Delmastro, P., Jiricny, J., 1998. Mismatch repair de®ciency associated with overexpression of the MSH3 gene. Proc. Natl. Acad. Sci. USA 95, 8568±8573. Miyaki, M., Konishi, M., Tanaka, K., Kikuchi-Yanoshita, R., Muraoka, M., Yasuno, M., Igari, T., Koike, M., Chiba, M., Mori, T., 1997. Germline mutation of MSH6 as the cause of hereditary nonpolyposis colorectal cancer. Nat. Genet. 17, 271±272. Obmolova, G., Ban, C., Hsieh, P., Yang, W., 2000. Crystal structures of mismatch repair protein MutS and its complex with a substrate DNA. Nature 407, 703±710. Olschwang, S., Hamelin, R., Laurent-Puig, P., Thuille, B., De Rycke, Y., Li, Y.J., Muzeau, F., Girodet, J., Salmon, R.J., Thomas, G., 1997. Alternative genetic pathways in colorectal carcinogenesis. Proc. Natl. Acad. Sci. USA 94, 12122±12127. Parc, Y.R., Halling, K.C., Wang, L., Christensen, E.R., Cunningham, J.M., French, A.J., Burgart, L.J., Price-Troska, T.L., Roche, P.C., Thibodeau, S.N., 2000. hMSH6 alterations in patients with microsatellite instability-low colorectal cancer. Cancer Res. 60, 2225±2231. Peltomaki, P., de la Chapelle, A., Kuismanen, S.A., Holmberg, M.T., Salovaara, R., 2000. Genetic and epigenetic modi®cation of MLH1. Am. J. Pathol. 157, 1052±1053. Percesepe, A., Kristo, P., Aaltonen, L.A., Ponz de Leon, M., de la Chapelle, A., Peltomaki, P., 1998. Mismatch repair genes and mononucleotide tracts as mutation targets in colorectal tumors with different degrees of microsatellite instability. Oncogene 17, 157±163. Perucho, M., 1996. Cancer of the microsatellite mutator phenotype. Biol. Chem. 377, 675±684. Perucho, M., 1999. Correspondence (re: a National Cancer Institute workshop on microsatellite instability for cancer detection and familial predisposition: development of international criteria for the determina-

313

tion of microsatellite instability in colorectal cancer, Cancer Res., 1998, 58, 5248±5257). Cancer Res. 59, 249±253. Perucho, M., Peinado, M.A., Ionov, Y., Casares, S., Malkhosyan, S., Stanbridge, E., 1994. Defects in replication ®delity of simple repeated sequences reveal a new mutator mechanism for oncogenesis. Cold Spring Harbor Symp. Quant. Biol. 59, 339±348. Rampino, N., Yamamoto, H., Ionov, Y., Li, Y., Sawai, H., Reed, J.C., Perucho, M., 1997. Somatic frameshift mutations in the BAX gene in colon cancers of the microsatellite mutator phenotype. Science 275, 967±969. Schwartz Jr., S., Yamamoto, H., Navarro, M., Maestro, M., Reventos, J., Perucho, M., 1999. Frameshift mutation at mononucleotide repeats in caspase-5 and other target genes in endometrial and gastrointestinal cancer of the microsatellite mutator phenotype. Cancer Res. 59, 2995±3002. Thibodeau, S., Bren, G., Schaid, D., 1993. Microsatellite instability in cancer of the proximal colon. Science 260, 816±819. Thibodeau, S.N., French, A.J., Roche, P.C., Cunningham, J.M., Tester, D.J., Lindor, N.M., Moslein, G., Baker, S.M., Liskay, R.M., Burgart, L.J., Honchel, R., Halling, K.C., 1996. Altered expression of hMSH2 and hMLH1 in tumors with microsatellite instability and genetic alterations in mismatch repair genes. Cancer Res. 56, 4836±4840. Toyota, M., Ahuja, N., Ohe-Toyota, M., Herman, J.G., Baylin, S.B., Issa, J.P., 1999. CpG island methylator phenotype in colorectal cancer. Proc. Natl. Acad. Sci. USA 96, 8681±8686. Wu, Y., Nystrom-Lahti, M., Osinga, J., Looman, M.W., Peltomaki, P., Aaltonen, L.A., de la Chapelle, A., Hofstra, R.M., Buys, C.H., 1997. MSH2 and MLH1 mutations in sporadic replication error-positive colorectal carcinoma as assessed by two-dimensional DNA electrophoresis. Genes Chromosomes Cancer 18, 269±278. Yamamoto, H., Sawai, H., Perucho, M., 1997. Frameshift somatic mutations in gastrointestinal cancer of the microsatellite mutator phenotype. Cancer Res. 57, 4420±4426. Yamamoto, H., Gil, J., Schwartz Jr., S., Perucho, M., 2000. Frameshift mutations in Fas, Apaf-1, and Bcl-10 in gastro-intestinal cancer of the microsatellite mutator phenotype. Cell Death Diff. 7, 238±239.