The Role of hPMS1 and hPMS2 in Predisposing to Colorectal Cancer¹

HNPCC³ is a genetically heterogeneous disorder that is believed to account for 5–8% of all CRC cases in the Western world (1). Characterization of HNPCC families was performed previously from a population-based series of cases (2). An early onset of CRC and an increased frequency of other cancers, including cancers of the endometrium, stomach, ovary, pancreas, and the hepatobiliary and urinary tract, characterize this syndrome. HNPCC is caused by germline mutations in at least three genes of the human DNA MMR system, including hMLH1, hMSH2, and hMSH6 (3, 4, 5, 6, 7). MSI is a specific feature of HNPCC and occurs as a result of failing DNA MMR (8). Of the tumors from HNPCC families, 75–86% show MSI (9, 10). To date, mutations in the MMR genes hMSH2 and hMLH1 account for a majority of families with HNPCC.⁴ Germline hMSH6 mutations seem to be responsible for the disease in a small number of atypical as well as typical HNPCC families (6, 7, 11, 12), with an excess of endometrial cancer (11) and often late onset (11, 12). Mutated hMSH3 genes have been found only in somatic tissues of HNPCC patients, and no germline changes have yet been identified in this gene (13, 14).

The two DNA MMR genes hPMS1 and hPMS2 have also been suggested to be associated with HNPCC (15). In that report, probands from 40 HNPCC families were studied for mutations in hMLH1, hMSH2, hPMS1, and hPMS2. One germline in-frame deletion of hPMS2 and one nonsense hPMS1 mutation were identified (15). A subsequent study from the same group, and including 34 additional HNPCC kindred, revealed additional mutations in hMLH1 and hMSH2 but no further mutations in hPMS1 or hPMS2 (16). Subsequently, Hamilton et al. (17) screened five colon cancer susceptibility genes in 14 Turcot’s syndrome families. In addition to 10 APC mutations and 1 hMLH1 mutation, 1 nonsense hPMS2 mutation was identified in a Turcot’s patient (17). Another study reported a missense mutation in hPMS2 in a Turcot’s syndrome patient without a family history of this condition (18). Recently, two compound missense mutations were found in a patient with Turcot’s syndrome without a family history of cancer (19). Viel et al. (20) screened 22 unrelated HNPCC families, in which hMLH1 and hMSH2 mutations were already excluded, for mutations in hPMS2. No disease-causing mutation was detected. A mutation screening of 75 HNPCC and HNPCC-like families was also undertaken for the hMLH1, hMSH2, hMSH6, hPMS1, and hPMS2 genes (21). No germline hPMS2 mutation was identified. Two undefined DNA variants were detected in one patient from an incomplete HNPCC family, and both of them affected the same allele. Finally, two somatic hPMS2 mutations have been identified in a tumor cell line that derived from a colon cancer (22) and in an endometrial cancer cell line (23).

In the present study, patients from 84 HNPCC or HNPCC-like families were studied for mutations in hPMS1 and hPMS2. No germline mutations in hMLH1, hMSH2, hMSH3, and hMSH6 had been identified in these families in previous analyses (24, 25, 26, 27, 28, 29).

HNPCC or HNPCC-like families were recruited from hereditary cancer registries of Sweden (50 families), Finland (26 families), Italy (6 families), and Denmark (2 families; Table 1). In total, 24 families fulfilled stringent Amsterdam Criteria I (30). Twenty-one families contained at least three close relatives affected by CRC. In 28 families, two close relatives were affected by CRC. Eight families had only one case of CRC at a very young age (before 35 years). Three families had endometrial cancer alone segregating as an autosomal dominant trait.

Blood samples and specimens of fresh-frozen or archival tumor tissue were collected after informed consent according to institution guidelines. MSI analysis was performed (27, 28, 31) according to the Bethesda guidelines (8) or, in some cases, using mononucleotide repeat markers BAT 25 and BAT 26 (32) only.

RNA was extracted from all available EBV cell lines using the Ultraspec-II RNA PCR kit (Perkin-Elmer, Foster City, CA). To synthesize cDNA, the random hexamer priming method was used with the GeneAmp RNA PCR kit (Perkin-Elmer). The PTT was carried out according to the manufacturer’s instructions (Promega, Madison, WI). A set of primers was used to amplify the cDNA in two or three overlapping fragments for hPMS2 and hPMS1, respectively. Primer sequences and conditions for PCR have been described previously (15). As positive control in the hPMS1 analysis, we used the sample with the first identified hPMS1 mutation.

Alterations identified by RT-PCR/PTT were sequenced from cDNA. Products of RT-PCR were purified with OIAEXII gel extraction kit (Qiagen, Germany) before sequencing. The Thermo Sequenase Radiolabeled Terminator Cycle Sequencing Kit was used according to instructions from the manufacturer (Amersham, Cleveland, OH).

Genomic DNA was studied from families F5, F32, F33, F70, F75, F80, and I9, using XL PCR kit (Perkin-Elmer) and Ampliwax PCR Gem 100 (Perkin-Elmer) for hot start to increase specificity (33). Primers and exact conditions are available from the authors on request.

Southern-blot analysis used 7–10 μg of genomic DNA digested to completion with TaqI, HindIII, NsiI, EcoRI, RsaII, and HeaIII. The digested DNA was electrophoresed on a 0.8% agarose gel, and DNA was transferred to a nylon-nitrocellulose membrane. The membranes were hybridized in polyethylene glycol-SDS at 65°C with randomly labeled probes generated by regular PCR. Primer sequences and PCR conditions for preparing cDNA probes of hPMS1 have been described previously (15). The restriction enzymes used were HindIII, NsiI, TaqI, RsaI, and EcoRI, separate and in combinations. Three probes used for hybridization were generated with PCR, representing codons 1–500, 270–755, and 488–933. RT-PCR products were purified with an OIAEXII gel extraction kit (Qiagen). The filters were washed in 0.1 × SSC-0.1% SDS at room temperature and 65°C. Autoradiography was carried out at −70°C for 24–96 h.

The conversion technique used in the present study has been described previously (34).

Using a RNA-based mutation detection technique with RT-PCR coupled to PTT as a starting point, we screened 84 unrelated families for germline mutations in hPMS1 and hPMS2. No germline mutations were detected in hPMS2. Five aberrant bands were identified in hPMS1 in seven families (six from Finland and one from Italy). cDNA sequencing revealed a deletion in all of these samples. These alterations were del1937–2422 (families F5, F33, and F75), del1779–2447 (family F32), del1937-2553 (family F70), del1927–2422 (family F80), and del2306–2418 (family I9). The alterations could be reproduced using the same cDNA in a new PCR experiment; thus, the variants were not likely to be PCR artifacts.

We conducted additional experiments in an attempt to confirm these variants at a genomic DNA level. Genomic DNA, extracted from lymphocytes, was available from the six Finnish families. From the Italian family, the only available source of DNA was a paraffin-embedded tumor block. Long-range PCR was conducted on all seven genomic DNA samples, using primer pairs that covered the deleted regions in the cDNA. No products were obtained in the expected size range (data not shown). Because the PCR approach was hampered by the unavailability of information about exon-intron border sequences of hPMS1, we next used Southern-blot hybridization analyses for the six Finnish samples for which genomic DNA was available. During the experimental procedure, five restriction enzymes and three different probes were used. No abnormal bands could be detected.

For segregation studies, it was possible to obtain RNA samples from one additional affected member from two Finnish families each (F5 and F80); from the Italian family (I9), a new cDNA sample from the same patient was obtained. No alterations could be detected in these new samples by the same experimental procedure. Thus, the RNA-based alterations did not segregate and did not cause predisposition to the disease in these three families. The mutation in Finnish families F33 and F75 was the same mutation as in family F5. Therefore, we concluded that the alterations first detected in families F5, F33, F75, F80, and I9 were unlikely to be disease related and could result from alternative splicing, e.g., in analogy to variants previously reported for the mismatch repair genes hMSH2 and hMLH1 (35). For the remaining two families (F32 and F70) with RNA variants, no conclusive results could be obtained. The deletion found in family F70 started at the same position as the deletion in the three families, F5, F33, and F75, and the deletion found in family F80 ended at the same position as the deletion in these three families. This supports the idea that a splice variant was picked up in the original setting of the experiment.

The development of new methods for mutational analysis and the ascertainment of additional family members allowed a reexamination of the original kindred with the hPMS1 mutation (Fig. 1). Using conversion technology to separate the alleles of chromosomes 2, we found that the proband, who developed a colon cancer at age 51 and ovarian cancer at age 43, contained a large deletion encompassing exons 1–7 of hMSH2 (Fig. 2). The hMSH2 mutation was found on the same chromosome as the hPMS1 mutation (short arm and long arm, respectively). We were also able to examine other affected members of this family, including a nephew of the proband who developed colon cancer at age 42. Using converted templates, we showed that this individual harbored the hMSH2 deletion but not the hPMS1 mutation.

The present study failed to detect pathogenic germline mutations in hPMS1 and hPMS2 in a series of 84 HNPCC and HNPCC-like families in which mutations in other DNA MMR genes could not be found. The fact that we did not find any clear-cut mutations in hPMS1 and hPMS2 could reflect possible shortcomings of the mutation-screening technique, RT-PCR/PTT, used for the present study. For example, missense variants and alterations outside the coding region would escape detection by this technique. However, the first published mutations in hPMS1 and hPMS2 in two HNPCC patients were originally detected by PTT (15, 17), whereas reported polymorphisms and DNA variants with an undefined role for cancer risk were identified by other techniques (20, 21).

From the literature, the only clear association of germline mutations in hPMS2 with CRC involves the Turcot’s variant. One nonsense mutation in hPMS2 was in a patient with Turcot’s syndrome and a sister with CRC (17). The colon cancer from the index patient in this family showed MSI. Mutations in hMSH2 and hMLH1 were excluded by sequencing (17). Another case with Turcot’s syndrome was reported to have two germline mutations, one on each chromosome (1221delG and 2361delCTTC; Ref. 19). The tumor was MSI positive. These mutations were inherited from the unaffected parents, which is why it is possible that compound heterozygosity was necessary for the pathogenic phenotype. An additional case of a patient with Turcot’s syndrome with no family history of the syndrome has been reported (18). One astrocytoma, three colon carcinomas, and two colonic adenomas from this patient all demonstrated MSI. Moreover, even normal fibroblast and brain tissue from this patient showed MSI. This patient had a germline missense mutation in hPMS2, GAG(705)AAG. The mutation was inherited from his unaffected mother; therefore, even in this case, there might have been an additional unknown germline or somatic mutation that contributed to the severe phenotype. Mutations in the hMLH1, hMSH2, hPMS1, and JTV1 genes were excluded by various methods in the patient and his two parents. The first mutation in hPMS2 identified in an HNPCC patient was an in-frame deletion of codons 268–669. The tumor was MSI positive, and the wild-type allele was found to have a somatic inactivating alteration, suggesting that the hPMS2 gene was involved in tumorigenesis. There was no information on segregation in the family. Germline mutations in hMLH1, hMSH2, and hPMS1 were excluded in the patient.

Recent mouse studies have indicated that Msh2, Pms2, and Pms1 each contributes differently to MSI and, therefore, must vary in their roles in predisposing to neoplasia through deficient DNA MMR (36, 37, 38). The mouse results predict that the fraction of HNPCC patients with hPMS2 mutations will be low and that the fraction with hPMS1 mutations will be even lower. Pms2-deficient mice show MSI in normal tissue, but they do not develop gastrointestinal tumors (39). However, mice deficient for both Pms2 and APC develop MSI-positive gastrointestinal tumors (39). Thus, in combination with an increased tumor initiation because of a deficiency in the APC gene, Pms2 is likely to contribute to tumor progression via an increased mutation rate associated with MMR deficiency. In summary, the fact that no mutation in hPMS2 has been found to segregate with CRC in families, and the lack of gastrointestinal tumors in Pms2-deficient mice, suggests that germline hPMS2 mutations do not confer as high a risk for CRC as do analogous mutations in hMLH1 or hMSH2. However, hPMS2 mutations have been repeatedly observed in CRC patients, primarily with Turcot’s syndrome, and this implies a likely role for hPMS2 in carcinogenesis, perhaps in combination with other germline or somatic variations.

The first published hPMS1 mutation was a nonsense mutation resulting in exon skipping in a patient with a MSI-positive CRC (15). No other members of the family were available for testing at the time. Given that Pms1-deficient mice do not show MSI in colonic tissue and that hPMS1 has been shown not be involved in postreplicative MMR (40), MSI positivity in the tumor from the hPMS1 germline mutation carrier suggested that a mutation in another predisposing gene contributed or caused the increased CRC risk. Using the newly established conversion technique, we found a germline mutation in hMSH2 consistent with the MSI-positive tumor from the proband in this family. Another relative, who recently developed CRC, shared the hMSH2 mutation but not the hPMS1 mutation. This clearly demonstrates that hMSH2 predisposes to CRC in this family. Interestingly, the patient with this hPMS1 mutation belonged to a HNPCC family that was unusual in that it had a large number of lung cancer cases. Pms1-deficient mice do show a tissue-specific increased mutation rate in mononucleotide repeats in fibroblasts, but not in colonic epithelium (34). Although hPMS1 and the yeast counterpart, MLH2, have been shown not to be involved in DNA MMR (40, 41), a postmeiotic non-Mendelian segregation occurs in MLH2 mutants (42), and those mutants also exhibit a weak resistance to some DNA-damaging agents (43). It is possible that the initiation of cancer is not always dependent on MSI and a deficient MMR, and why a deficient function in hPMS1 could potentially still be mutagenic and cause cancer. Thus, although the hMSH2 mutation predisposes to CRC, it is possible that the hPMS1 mutation could have contributed to the phenotypic finding of lung cancer cases in this family. The only other screening study done for hPMS1 mutations identified two missense mutations, Met 394 Thr and Gly 501 Arg, which affected the same allele in one patient from a colon cancer family (20). There were no other members of the family available to test for segregation, and MSI was not tested in the tumor. Mutations in the genes hMLH1, hMSH2, hPMS2, and hMSH6 were not found by various methods. It is likely that this family also has an additional germline alteration, which could explain the predisposition for CRC.

In at least 17 of the presently studied families, involvement of MMR genes was expected because tumors from these families demonstrated MSI. Because Pms2 is associated with MSI, hPMS2 was the most likely gene to be involved. However, our failure to detect any such mutation implies that other known and unknown MMR genes, including hMLH3 (44), might be involved. In all MSI-negative tumors, any other known or unknown gene predisposing to CRC could cause or contribute to disease. Although mutations in hMLH1, hMSH2, and hMSH6 were previously excluded by various screening methods in these families, it is still possible that refractory mutations in these genes could have been missed. Newer technologies may help identify these mutations in the future (34). Until there is better evidence for an increased CRC risk associated with germline mutations in the Pms1 and Pms2 genes, a conservative interpretation of the role of mutations in these genes is advised.

We are indebted to the families included in this study and to K. Romans for assistance with clinical coordination.