Germ-line msh6 Mutations in Colorectal Cancer Families¹

CRC³ is one of the most prevalent cancers in the Western world. Inherited susceptibility appears to account for approximately 2–5% of CRC cases, with HNPCC being the most striking manifestation of hereditary CRC (1). Among unselected Amsterdam Criteria HNPCC cases, 50–60% have been reported to carry inherited mutations in two mismatch repair genes, MSH2 and MLH1. Moreover, 70–100% of HNPCC cases whose tumors manifest MSI (MSI+) reportedly have germ-line mutations in these two genes. Few germ-line mutations in MSH6, PMS1, and PMS2 have been reported in HNPCC patients, indicating that inherited mutations in these mismatch repair genes do not play a major role in HNPCC (2, 3, 4, 5, 6). Among CRC cases in general, approximately 11–38% are MSI+, consistent with a somatic (acquired) mismatch repair defect (7, 8). Recent studies have shown that up to 90% of sporadic MSI+ cases have transcriptional silencing of MLH1, and the remainder are consistent with inactivation of MSH2 or MLH1 by somatic mutation (9, 10).

An interesting component of CRC is the familial class, defined herein as CRC cases who also have one or more relatives with CRC but do not fulfill the Amsterdam or Modified Amsterdam criteria for HNPCC. Because of the high prevalence of CRC, many familial cases may simply represent chance clustering. Dietary and other shared environmental influences might also account for some family aggregates of these cancers (11). In addition, inheritance of mutations in genes for late-onset disease might produce small familial clusters of CRCs that are not HNPCC.

MSH6 is a MutS homologue that was initially described as a subunit of the MSH2-MSH6 mispaired base recognition factor required for mismatch repair (12). Because MSH2 and MSH6 are subunits of the same protein complex, it was surprising that germ-line mutations in MSH6 were not found in HNPCC families, whereas germ-line mutations in MSH2 are common (5). Genetic and protein-protein interaction studies performed in Saccharomyces cerevisiae subsequently demonstrated two mispaired base recognition complexes, MSH2-MSH3 and MSH2-MSH6 (13), a result supported by studies in human and murine systems (14, 15, 16, 17). These complexes are partially redundant with each other, such that MSH2-MSH6 appears to recognize both base:base mispairs and insertion/deletion mispairs, whereas MSH2-MSH3 appears to only recognize insertion/deletion mispairs. Because of this redundancy, mutations in MSH2 result in high rates of accumulation of base substitution mutations and frameshift mutations in microsatellite sequences, whereas mutations in MSH6 result in high rates of accumulation of base substitution mutations (13, 18). This observation provides two explanations for the rarity of germ-line msh6 mutations in HNPCC families. First, virtually all initially examined HNPCC families were MSI+ at dinucleotide repeat loci, a phenotype not caused by loss of MSH6 function, leading to the exclusion of potential msh6 mutant families from study (5, 13, 18). Second, many of the target tumor suppressor genes inactivated in HNPCC are inactivated by frameshift mutations in mononucleotide runs, the major mutable target sequences when mismatch repair is completely inactivated (13, 19, 20). Because loss of MSH6 function does not result in highly increased rates of such frameshift mutations (13), these target genes might not be hypermutable when MSH6 function is lost.

Inactivation of MSH6 causes a mutator phenotype that is primarily confined to the accumulation of base substitution mutations, raising the possibility that inactivation of MSH6 could cause increased cancer susceptibility. The weaker mutator phenotype caused by msh6 mutations as compared to that caused by msh2 or mlh1 mutations might cause a later age at onset and lower penetrance than observed with germ-line msh2 or mlh1 mutations. Evidence of msh6 mutations conferring cancer includes msh6 mutations in mice causing increased susceptibility to the spectrum of tumors found in msh2 and mlh1 mutant mice, but with a later age at onset (17, 21, 22). In addition, case studies have reported germ-line msh6 mutations in three colon cancer patients (3, 4, 6). In the present report, we describe results of germ-line MSH6 analysis of colorectal cancer cases with diverse family histories.

Initially, a total of 127 blood samples from unrelated CRC patients were collected, and the DNAs were examined for msh6 mutations. Ninety cases were enrolled through a population-based study of colorectal cancer patients ascertained by the cancer registries of Orange, Imperial, and San Diego Counties, California. These population-based cases were selected partly on the basis of family history (45 familial non-HNPCC colorectal cancer cases and 45 sporadic cases), and all were demonstrated to not have a germ-line msh2 or mlh1 mutation. The other 37 cases were obtained from a subset of 58 cases from a clinic-based series of patients at the Dana-Farber Cancer Institute who fulfill the Amsterdam (13 cases) or Modified Amsterdam (9 cases) Criteria or have familial non-HNPCC (15 cases). These 37 cases were demonstrated to not have a germ-line msh2 or mlh1 mutation (23). The remaining 21 HNPCC cases at the Dana-Farber Cancer Institute were not analyzed further because they had already been found to have a germ-line msh2 or mlh1 mutation (23) and would not be expected to also have a germ-line msh6 mutation. In all cases, family history information was as reported by the proband; thus far, it has been possible to verify this information from medical or death records or by interviewing family members in only some cases. When possible, tumor samples from cases with msh6 mutations were obtained for MSI analyses. After the analysis of the above-described cases was completed, an additional 46 familial non-HNPCC cases, which were not tested for msh2 or mlh1 mutations, and 4 HNPCC cases that did not have msh2 or mlh1 mutations were recruited from the cancer registries of Orange, Imperial, and San Diego Counties, California and analyzed for msh6 mutations. In total, 177 of the 198 cases were examined for msh6 mutations.

Genomic DNA was isolated for blood samples using Qiagen kits and instructions provided by the manufacturer. Tumor blocks were microdissected to enrich for tumor cells, and then DNA was isolated from these paraffin-embedded tumor samples as described previously (9). Standard PCR primers were from either Cybersyn Corp. or the Dana-Farber Cancer Institute Molecular Biology Core Facility. Fluorescence-labeled PCR primers were obtained from Life Technologies, Inc.

Mutations in MSH6 were detected by amplifying individual exons and flanking intron sequences by PCR using primers (Table 1) devised from the sequence of the MSH6 genomic locus (GenBank accession numbers U73732 and U73737). PCR set up was performed with a Tecan Genesis 100 Robotic Workstation, and PCR was performed on Perkin-Elmer 9600 and 9700 PCR instruments. PCR was performed in 25-μl volumes using three different PCR buffer and cycling conditions, depending on the region of MSH6 analyzed: (a) condition A (PC2 buffer), 50 mm Tris-HCl (pH 9.1), 3.5 mm MgCl₂, 16 mm NH₃ (SO)₄, 150 μg/ml BSA, 100 μm each nucleotide triphosphate, 2 m GC-Melt (Clontech), 0.4 μm each primer, 25 ng of genomic DNA, and 0.5 unit of Klentaq DNA polymerase (Ab Peptides) for 1 cycle of 96°C for 4 min; 15 cycles of 96°C for 20 s, 70°C for 20 s (−1°C/cycle), and 68°C for 20 s; 25 cycles of 96°C for 20 s, 55°C for 20 s, and 68°C for 20 s; and 1 cycle of 68°C for 7 min; (b) condition B (PE Buffer II, 10 mm Tris-HCl (pH 8.3), 50 mm KCl, 1.5 mm MgCl₂, 100 μm each deoxynucleotide triphosphate, 0.4 μm each primer, 25 ng of genomic DNA, and 0.63 unit of AmpliTaq DNA polymerase (Perkin-Elmer/Applied Biosystems) for 1 cycle of 95°C for 10 min; 10 cycles of 95°C for 40 s, 65°C for 20 s (−1°C/cycle), and 72°C for 20 s; 25 cycles of 94°C for 20 s, 55°C for 20 s, and 72°C for 20 s; and 1 cycle of 72°C for 7 min; (c) condition C, the same as condition B, except that the time of incubation at 72°C was increased to 40 s. To purify the resulting PCR products, 20 μl of PCR product were incubated with 2–8 units of shrimp alkaline phosphatase and 40 units of Escherichia coli exonuclease I (Pharmacia/Amersham) for 15 min at 37°C, followed by incubation for 15 min at 80°C. The PCR products were sequenced on an ABI377 or ABI3700 sequencer (Perkin-Elmer/Applied Biosystems) with standard M13 forward and reverse primers and Big Dye terminator chemistry using reagents obtained from Perkin-Elmer/Applied Biosystems. In some cases, the initial sequencing strategy resulted in sequencing across a site that was heterozygous for an insertion/deletion, resulting in an unreadable sequence downstream from the heterozygous site. In these cases, the PCR products were resequenced using internal primers, or the regions were reamplified using different PCR primers that allowed sequencing of the relevant region using standard M13 forward and reverse primers (see Table 1). Analysis of the sequence chromatograms was carried out using Sequencher (Gene Codes Corp.) and Phred/Phrap/Consed software (University of Washington) as well as by visually examining printed chromatograms to detect sequence changes and heterozygous nucleotides.

MSI was evaluated at five different microsatellite loci [BAT26 (mononucleotide), D2S123 (dinucleotide), D5S346/APC (dinucleotide), D17S250/Mfd15 (dinucleotide), and BAT25 (mononucleotide)] proposed as a reference set by the National Cancer Institute Workshop on MSI as well as at three additional microsatellite loci [D18S69 (dinucleotide), D10S197 (dinucleotide), and BAT40 (mononucleotide); Refs. 7 and 19]. Each locus was amplified from tumor and normal DNA samples with fluorescence-labeled PCR primers and analyzed using an Perkin-Elmer/Applied Biosystems 377 sequencer and GeneScan software, essentially as described by published methods (7). Primer sequences were from the Genome Database and as published by others. Microsatellite stable was defined as no observed instability, MSI MSI-L was defined as one unstable locus of the eight tested, and MSI MSI-H was defined as two unstable loci of the five reference loci or at least three of the eight total loci analyzed (19).

The plasmid pRDK439 contains the wild-type S. cerevisiae MSH6 gene cloned as a 5.4-kb BamHI to HindIII fragment into the single copy (CEN) vector pRS315. Mutations were introduced into the S. cerevisiae MSH6 gene by PCR mutagenesis, and then the MSH6 gene was sequenced to verify that the introduced mutation was the only sequence change present (24). The mutant plasmids, as well as wild-type and vector control plasmids, were transformed into the S. cerevisiae strain RDKY 3523 (trp1 ura3-52 ade2-1 hom3-10 leu2-3,112 msh3::hisG msh6::hisG), and the rate of reversion of the hom3-10 frameshift mutation was determined by fluctuation analysis. All of the genetic methods, media, and strains were essentially as described previously (13).

Germ-line msh6 mutations were detected by direct sequencing of the gene using genomic DNA from an initial group of 127 blood samples (Fig. 1). All of these samples had been previously shown to not have germ-line msh2 and mlh1 mutations (23). Four different potentially pathogenic msh6 mutations were found. Three of these (cases 1742, 2929, and 2949) were found among the 45 non-HNPCC familial cases in the population-based series studied, and one case (case 261) was found among the subset of 15 clinic-based non-HNPCC families studied (Tables 2 and 3). The clinical histories of these probands show that all were diagnosed with CRC in their 60s (ages 62, 63, 69, and 69 years). All four of them (cases 261, 1742, 2929, and 2949) had only one affected first-degree relative who developed CRC at age 54, 74, 73, and 48 years. Case 2949 also had an affected grandfather. One of the cases had a family history of other cancers, including a relative with both breast and ovarian cancers and a relative who had developed lung cancer.

Genetic analyses of the four cases revealed one splice site mutation and three missense variants (Table 3). The splice site mutation in case 2949 changed the invariant nucleotide at the −2 position of the splice acceptor site of intron 5. Two of the potential missense mutations (case 1742, S285I; case 2929, G566R) resulted in significant amino acid substitutions, whereas the third (case 261, L396V) resulted in a relatively conservative amino acid substitution. Two of these three missense variants (cases 1742 and 2929) were not detected in DNAs from anonymized blood samples from 188–200 cancer-free controls, suggesting that these two variants are loss of function mutations rather than polymorphisms. The missense variant found in the clinic-based non-HNPCC familial case (case 261, L396V) was seen in 3 of 200 normal controls, suggesting it might represent a rare polymorphism (Table 3). A large number of other polymorphisms were found in both coding and intronic sequences (Table 4).

The missense mutations found in case 2929 with familial non-HNPCC (G566R) and the rare polymorphism in case 261 (L396V) changed highly conserved amino acids where the normal amino acid was the same in the human, mouse, and S. cerevisiae MSH6 proteins. To further evaluate the significance of these changes, mutations creating the same amino acid substitutions were made in the S. cerevisiae MSH6 gene by site-directed mutagenesis. The mutant gene present on a CEN vector was tested for the ability to complement the mismatch repair defect of an msh3 msh6 double mutant strain. In the case of the G566R change (case 2929), the mutant plasmid failed to fully complement the increased rates of accumulation of base substitution and frameshift mutations caused by the mismatch repair defect in the msh3 msh6 mutant strain, whereas the wild-type plasmid fully complemented this defect (Table 5). These data support the view that this missense mutation is a loss of function mutation in the human MSH6 gene. In contrast, the plasmid with the L396V change found in case 261 complemented the mismatch repair defect of the msh3 msh6 mutant strain, supporting the view that this change is a polymorphism that does not affect MSH6 function.

In three of the cases, we were able to obtain tumor samples for further analysis (Fig. 1; Table 3). In case 1742 with the S285I (g.854G>T, exon 4) change, sequencing of the mutation-containing region of the tumor sample showed loss of the wild-type allele and loss of heterozygosity at a closely linked heterozygous polymorphic site (g.642C>T, exon 4; Table 3 and Fig. 1). Further analysis of heterozygous polymorphic sites in this sample showed that heterozygosity was maintained at a polymorphic site in exon 2 (g.276A>G), heterozygosity was lost at a polymorphic site in intron 5 (IVS5+14A>T), and heterozygosity was maintained at a polymorphic site in intron 6 (IVS6+145A>G); these sites are located approximately 8 kb upstream and approximately 5.2 and 6.8 kb downstream from the mutant site, respectively. Thus, the wild-type allele in this tumor was inactivated by an intragenic deletion covering exons 4 and 5 and possibly exon 3. MSI analysis of case 1742 with the S285I alteration (Table 3) showed that it was MSI-L. In case 2929 with the G566R change (g.1696G>A, exon 4), heterozygosity was present in the tumor at heterozygous polymorphic sites in exon 2 (d.276A>G), intron 6 (IVS6+145A>G), and intron 7 (IVS7+31-32insATCT) and at the site of the germ-line mutation. However, heterozygosity was lost at a polymorphic site in intron 5 (IVS5+14A>T; data not shown). These data are consistent with the presence of an intragenic deletion, possibly covering exon 5, in the tumor from case 2929. MSI analysis indicated this case was MSI-H (Table 3). Limited analysis of MSH3 showed that the MSH3 A8 tract was not mutated in the tumor from case 2929 (4, 20). In case 261 with the L396V change, the tumor was found to be MSI-L (Table 3). We could not sequence all of MSH6 from the tumor DNA in this case, but we did observe that heterozygosity at the site of the L396V change was maintained in the tumor.

To extend the observation that msh6 mutations could be found in familial non-HNPCC CRC cases, an additional 46 population-based familial non-HNPCC and 4 population-based HNPCC cases were analyzed. The analysis revealed one frameshift mutation and three missense variants among the familial non-HNPCC cases and no mutations among the HNPCC cases (Table 3). The frameshift mutation (case 1980) was found in a 47-year-old patient with colon cancer, and his brother was affected at age 48 years. This frameshift mutation was clearly a loss of function mutation because it was predicted to result in the production of a truncated protein missing 644 COOH-terminal amino acids. Cases 3260, 3959, and 2936 with missense variants each had one first-degree relative with colon cancer. Two of the missense variants (case 3959, D803G; case 2939, P1087T) resulted in significant amino acid substitutions, whereas the third (case 3260, V800L) resulted in a relatively conservative amino acid substitution. Two of these three missense variants (cases 3959 and 2939) were not detected in DNAs from 189 anonymized blood samples from cancer-free controls, suggesting that these two variants are loss of function mutations rather than polymorphisms. The missense variant (case 3260, V800L) was seen in 1 of 189 normal controls, essentially the same frequency as found in the CRC cases, suggesting that it might represent a rare polymorphism (Table 3). None of the missense mutations were amenable to functional testing in S. cerevisiae, and thus far, we have been unable to obtain tumor material to extend the analysis of these cases. The clinical histories of these probands (Table 2) show that all were diagnosed with CRC in their late 40s to 60s (ages 47, 49, 57, and 64 years). All of them had only one affected first-degree relative who developed CRC (at age 48, 76, 90, and 56 years, respectively), and all of them had a family history of other cancers, including relatives who had developed lung, breast, cervix, stomach, or prostate cancer.

Our initial study identified potential germ-line msh6 mutations in four familial non-HNPCC cases. In three population-based familial cancer cases, data strongly suggest loss of function mutations that cause mismatch repair defects. The splice site mutation changed the invariant nucleotide at the −2 position of the intron 5 splice acceptor site; this would be predicted to cause skipping of exon 6 and result in the synthesis of a truncated protein containing exons 1–5 and an additional eight amino acids from the −1 reading frame of exon 7. This mutant protein would be missing a domain required for interaction with MSH2 (25). In S. cerevisiae, we have shown that deletion of this region of MSH6 results in a complete loss of function, consistent with this region of MSH6 containing a critical domain (see Table 3).⁴ The two other population-based cases had missense mutations that caused nonconservative changes absent in 188–200 normal controls. These findings indicate that they are unlikely to represent common polymorphisms. In case 1742 with the S285I change, the wild-type allele was inactivated by a deletion in the patient’s tumor, which is consistent with loss of MSH6 function in this tumor. In case 2929 with the G566R change, partial loss of MSH6 function was demonstrated in a S. cerevisiae-based functional assay, and the tumor from this case also appeared to contain an intragenic deletion. These results confirm and extend the results of others showing that, contrary to initial reports (5), it is possible to identify germ-line msh6 mutations in CRC cases with diverse family histories (3, 4, 6). In contrast, the L396V alteration in clinic-based case 261 may be a rare polymorphism. The change was found in 3 of 200 controls and did not cause functional loss in our S. cerevisiae-based functional assay; because the tumor from this case was MSI-L, additional studies of this variant are probably warranted.

A follow-up study of 46 additional population-based familial non-HNPCC cases resulted in the identification of additional msh6 mutations at essentially the frequency predicted from the initial analysis of familial non-HNPCC cases. One (case 1980) was a frameshift mutation that was clearly a loss of function mutation and two cases (case 3959, D803G; case 2939, P1087T) were missense variants that caused nonconservative amino acid substitutions and were not found in 189 normal controls, consistent with their being loss of function mutations. The fourth (case 3260, V800L) was a missense change causing a relatively conservative amino acid substitution that was found in 1 of 189 normal controls. This latter case seemed more likely to represent a rare polymorphism, due to the conservative nature of the amino acid change caused and because it was found in normal controls. Because we have not yet been able to analyze tumor material from these latter three cases or perform functional studies, the data are suggestive but not definitive that at least two of these missense changes cause loss of function.

We were able to perform MSI analysis on three tumor samples where MSH6 function was likely inactivated. In two cases (case 1742, S285I; case 261, L396V), the tumors were found to be MSI-L, due to the observation of instability at a mononucleotide repeat and a dinucleotide repeat, respectively. In the third case (case 2929, G566R), the tumor showed MSI at both mononucleotide and dinucleotide repeat loci and was MSI-H. One explanation for the finding of a MSI-H tumor in a patient with a germ-line msh6 mutation and no mutations in MSH2 or MLH1 is that the tumor from this patient had also lost MSH3 function, leading to a MSI-H phenotype (4, 13, 16, 18). Unfortunately, we did not have enough tumor DNA from case 2929 to allow complete analysis of MSH3; however, the highly mutable MSH3 A8 tract was not mutated in the tumor from this case (4, 20). Previous studies of three CRCs associated with germ-line msh6 mutations have identified the following: (a) tumor MSI at mononucleotide repeat loci but not at dinucleotide repeats (3); (b) tumor MSI (MSI-H) where both MSH3 and MSH6 were mutated (4); and (c) tumor MSI (MSI-H) where MSH3 status was not analyzed (6). The combined results, although based on a small number of samples, suggest that there may be no clear-cut MSI criteria for identifying CRC cases with germ-line msh6 mutations. This is likely because germ-line msh6 mutations cause only small increases in the rate of accumulating frameshift mutations but can be associated in some cases with MSI-H tumors because a secondary mutation of MSH3 can occur, leading to a MSI-H phenotype; this view is consistent with results in S. cerevisiae, mouse, and human cell line systems (13, 16, 17, 18).

We found only one rare MSH6 polymorphism among 15 familial, non-HNPCC families and none in 22 HNPCC families ascertained from a clinic-based referral population. These 37 kindreds were part of a larger series of 58 primarily HNPCC families in which 21 msh2 or mlh1 mutations had previously been found (23). In addition, no msh6 mutations were found in four HNPCC kindreds from the population-based cases that did not have msh2 or mlh1 mutations. In the literature, there are two case reports from Japan of inherited msh6 mutations in CRC cases diagnosed at ages 52 and 66 years from suggested HNPCC families; other family members with CRC were diagnosed at ages 72 and 75 years (4, 6). There is also a report of a germ-line msh6 mutation in an early-onset CRC case (age 48 years) who did not have any other relatives with CRC (3). Thus, germ-line msh6 mutations are rare in HNPCC families, whereas our study found six familial CRC cases with germ-line msh6 mutations (cases 1742, 2929, 2949, 1980, 3959, and 2936) among a total of 85 population-based familial non-HNPCC cases (7.1% of probands). Because familial non-HNPCC accounts for approximately 18% of CRCs in our population database, after correcting for the age distribution of the probands analyzed, we calculate that germ-line msh6 mutations might be present in 1.4% of all United States CRC cases; the corresponding number of United States CRC cases with germ-line msh6 mutations would be approximately 1,800 (1.4% of 132,000) each year. This figure approaches the 1,500–2,500 CRC cases annually in the United States that may be attributable to germ-line msh2 and mlh1 mutations combined (26). However, both sets of estimates are based on the small sample size and could change.

The limited amount of clinical information on the six families with potential msh6 mutations reveals that the median age at cancer diagnosis in the probands was 61 years (Table 2). Median age at colorectal cancer diagnosis in their affected relatives was higher (74 years), but the msh6 mutation status of these relatives could not be determined. The ages at CRC diagnosis in the probands are only slightly lower than the corresponding ages for all cases in the United States (68 years) but are 1–2 decades older than the ages at diagnosis in HNPCC families (27). Other cancers in our six families include isolated cases of carcinomas of the breast, ovary, lung, cervix, and prostate, but additional studies are needed to clarify the tumor spectrum associated with germ-line msh6 mutations.

In aggregate, our observations suggest that msh6 mutations account for a substantial fraction of hereditary CRCs, possibly with later ages at onset as compared with CRCs due to msh2 and mlh1 germ-line mutations. These results parallel the observations of later-onset cancer susceptibility seen in msh6 mutant mouse models compared with either msh2 or mlh1 mutant mouse models (17, 21, 22). If confirmed by a larger study, our data might indicate that msh6 germ-line mutations are usually expressed as later-onset familial CRC cases that fail to meet HNPCC criteria. This class of genes almost certainly exists for cancers at other organ sites. However, these genes are more difficult to identify, because their phenotypes are not as striking as those for high-penetrance genes such as MLH1, MSH2, and BRCA1 and BRCA2. Innovative new strategies may be needed to identify other novel intermediate penetrance genes.

We thank Drs. Webster Cavenee, Karen Arden, and Jean Wang for helpful discussions and Jill Green and Steve Hourmazian for technical help.

Note Added in Proof

We have identified an additional germline msh6 mutation (g.431G→T, Ser144Ile) in the 91 population-based familial, non-HNPCC samples. This change was not found in 199 normal controls and caused loss of function in the S. cerevisiae-based functional assay, indicating that this change is most likely a loss of function missense mutation. This mutation has also been described in a Dutch CRC case (Y. Wu, PhD Thesis, University of Groningen). Based on this result, the revised values for the percentage of familial non-HNPCC (adjusted for the age distribution of samples analyzed) and percentage of total CRC attributable to germ-line mutations in MSH6 are 8.1 and 1.5%, respectively, and the mean age of diagnosis for msh6 mutant CRC cases is 61 years.