Introduction

Most tumors arising within the context of the hereditary nonpolyposis colorectal cancer (HNPCC) syndrome, as well as about 15% of sporadic colorectal carcinomas, exhibit a type of genetic instability characterized by the accumulation of ubiquitous somatic alterations in the length of simple repeated sequences (Ionov et al, 1993). This genome-wide instability of simple repeat sequences, referred to as microsatellite instability (MSI), is seen in 14% to 39% of sporadic gastric carcinomas (SGC) (Fleisher et al, 1999; Halling et al, 1999; Kang et al, 1999; Leung et al, 1999; Oliveira et al, 1998; Santos et al, 1996; Yamamoto et al, 1999).

The MSI phenotype, as found in HNPCC, is associated with defective DNA mismatch repair (MMR) genes, such as hMLH1, hMSH2, hMSH3, and hMSH6, among others (Liu et al, 1995; Wu et al, 1997; 1999). At variance with this, mutations in MMR genes are rare in sporadic colorectal and gastric carcinomas with the MSI phenotype (Borresen et al, 1995; Liu et al, 1995; Moslein et al, 1996; Wu et al, 1997; Yamamoto et al, 1999).

An alternative mechanism to mutations for silencing gene expression is hypermethylation of the gene promoter (Costello et al, 2000; Jones et al, 1999). hMLH1 hypermethylation, with associated decreased protein expression, has been described in 44% to 100% of gastric carcinomas displaying a high level of MSI (MSI-H) (Fleisher et al, 1999; Leung et al, 1999; Suzuki et al, 1999; Toyota et al, 1999; Yamamoto et al, 1999). By contrast, this association has not been observed in microsatellite stable (MSS) tumors (Leung et al, 1999; Toyota et al, 1999; Yamamoto et al, 1999). The relationship between hMLH1 hypermethylation and MSI phenotype is less clear with regard to tumors displaying a low level of MSI (MSI-L): Leung et al (1999), Kang et al (1999), Yamamoto et al (1999) and Toyota et al (1999) reported the absence of hypermethylation of hMLH1, as well as normal protein expression, in this type of tumor. Fleisher et al (1999), however, reported the occurrence of hMLH1 hypermethylation in five out of six MSI-L tumors with associated diminished hMLH1 expression.

In a previous study, we reported that MSI SGC exhibit mutations in a series of target genes (TGFβRII, IGFIIR, and BAX) and display a particular clinicopathological profile (Oliveira et al, 1998): MSI-H tumors were found to be significantly associated with distal localization, Lauren’s intestinal and atypical histotypes, and Ming’s expansive pattern of growth. In the present study, we analyzed hMLH1 promoter hypermethylation in a series of 57 SGC, stratified into MSI-H (n = 28), MSI-L (n = 12), and MSS (n = 17). In a subset of cases, hMLH1 mutations and hMLH1 expression, as well as target gene mutations, were also evaluated. Our aims were twofold: (a) to characterize the molecular and clinicopathological features of SGC, with and without hMLH1 promoter hypermethylation, and (b) to compare molecular, namely, hMLH1 promoter methylation and target gene mutations, and clinicopathological features of MSI-L, MSI-H, and MSS tumors.

Results

MSI Status versus Clinicopathological Features

The 57 tumors were classified as MSS (n = 17), MSI-L (n = 12), and MSI-H (n = 28) using the criteria described in the “Materials and Methods” section. The comparison between MSS and MSI-H tumors showed a significant association between the MSI-H phenotype and Lauren’s intestinal and atypical histotypes (p = 0.05), Ming’s expansive pattern of growth (p = 0.04), lower pathological tumor, nodes, metastases (pTNM) stage of the tumor (p = 0.05), and the presence of mutations in target genes (p = 0.0001).

Table 1 summarizes the molecular and clinicopathological data and compares the MSI-L with MSS and MSI-H tumors. MSI-L tumors did not significantly differ from MSS tumors (Table1). By contrast, MSI-L tumors differ significantly from MSI-H tumors with regard to those same parameters that were found to be different in the comparison of MSS and MSI-H tumors (Table 1).

Table 1 Comparison of the Molecular and Clinicopathological Features of MSI-L and Those of MSS and MSI-H Sporadic Gastric Carcinomas
Full size table

hMLH1 Methylation Analysis

The hMLH1 promoter region of 21 of the 28 MSI-H cases (75%) could not be digested by HpaII, indicating that all four HpaII restriction sites were methylated. Six of the 12 (50%) MSI-L cases show the same methylation status. None of 17 MSS cases had methylation of the hMLH1 promoter region (Fig. 1). The DNA isolated from all these cases were sensitive to the digestion by MspI enzyme. The DNA from the normal mucosas of the stomach were all totally digested by HpaII. The association between the MSI phenotype and methylation status of the promoter region was statistically significant (p = 0.0001). Results are summarized in Table 2.

Figure 1
figure 1

Methylation of hMLH1 promoter in gastric carcinomas. (+) Cases with hypermethylation of hMLH1 promoter. The presence of PCR product in HpaII digestion indicates a hypermethylated promoter region of hMLH1. (−) Cases without hMLH1 promoter hypermethylation. The absence of PCR product in HpaII digestion indicates a nonmethylated promoter region of hMLH1. The absence of PCR product in MSPI digestion indicates a complete digestion of DNA and serves as control. U, undigested; H, HpaII-digested; M, MSPI-digested.

Full size image
Table 2 Summary of the Molecular and Clinicopathological Features of 57 Sporadic Gastric Carcinomas Regarding the Methylation Status of the hMLH1 Gene Promoter
Full size table

hMLH1 Methylation Status versus Clinicopathological Features

Data on the relationship between the clinicopathological features of the 57 SGC and the methylation status of hMLH1 are summarized in Table 2. A significant association was found between the methylation status and Ming’s classification (p = 0.01): the majority of expanding tumors showed hypermethylation (62%), whereas only a minority of infiltrative tumors presented this phenotype (27%).

hMLH1 Mutations

Mutations in all exons of hMLH1 were screened for 21 cases (17 methylation+ and 4 methylation−). No germline or somatic mutations of the hMLH1 gene were found in one MSS case, and no mutations were found in the 2 MSI-L and in the 18 MSI-H cases. Although we did not find any causative mutation in this gene for any of the cases analyzed, we did find two variants in the PCR amplicons of exons 8 (10/21, 48%) and 15 (9/21, 43%). Both these variants are common polymorphisms and were previously described by Liu et al (1995).

hMLH1 Gene Expression

hMLH1 mRNA expression was detected in all three MSS cases. In the 2 MSI-L cases and in 6 of the 10 MSI-H cases, where material was available for RNA isolation, there was no expression of hMLH1. In four MSI-H cases there was residual expression of hMLH1 when compared with GAPDH (Table 3 and Fig. 2).

Table 3 Summary of the Data Regarding the Methylation Status of the Promoter, cDNA Expression, Mutations of hMLH1 and the MSI Status in 15 of the 57 Sporadic Gastric Carcinomas
Full size table
Figure 2
figure 2

Expression of hMLH1 MSS, MSI-L, and MSI-H gastric carcinomas. (−) Cases with absence of expression of hMLH1. (+) Cases with hMLH1 expression. (↓) Cases with residual expression of hMLH1 in comparison with GADPH.

Full size image

Mutations in Target Genes (TGFβRII, IGFIIR, BAX, and TCF4)

Mutations in target genes were screened in 50 of the 57 cases. The presence of mutations in target genes (TGFβRII, IGFIIR, BAX, and TCF4) was significantly associated with hMLH1 promoter methylation (p = 0.0001) because 20 out of 23 (87%) cases positive for mutations in target genes showed hypermethylation of this MMR gene (Table 2). Six cases had no mutations in any of the target genes but had hMLH1 promoter methylation; five of these six tumors were MSI-L and the remaining case was a MSI-H carcinoma.

The association between mutations of target genes and the methylation status of hMLH1 promoter region was statistically significant for TGFβRII (p = 0.0001), BAX (p = 0.003), and IGFIIR (p = 0.03) and not significant for TCF4 (p = 0.34) (Fig. 3).

Figure 3
figure 3

Relationship between methylation of the promoter region of hMLH1 gene and the occurrence of mutations in the repetitive sequences of target genes: TGFβRII, IGFIIR, BAX, and TCF4.

Full size image

Methylation Status of MSI-L Carcinomas versus Molecular and Clinicopathological Features

The two MSI-L/Met+ cases analyzed by reverse transcription-polymerase chain reaction (RT-PCR) showed absence of expression of hMLH1 (Table 3). No mutations in target genes were observed in MSI-L cases (5 Met+ and 3 Met−) regardless of the presence or absence of hMLH1. The comparison between MSI-L/Met+ and MSI-L/Met− cases, regarding the clinicopathological features, did not yield any significant differences (data not shown).

Discussion

Hypermethylation of the hMLH1 promoter occurs with a very high frequency in SGC exhibiting MSI. We detected an aberrant hMLH1 promoter methylation in 75% of MSI-H gastric carcinoma cases leading to the absence or diminished expression of hMLH1. This percentage fits with those previously reported for SGC (Fleisher et al, 1999; Kang et al, 1999; Leung et al, 1999; Toyota et al, 1999; Wu et al, 2000; Yamamoto et al, 1999), sporadic colorectal carcinomas (Cunningham et al, 1998; Ghimenti et al, 1999; Kang et al, 1999), and endometrial carcinomas displaying MSI (Esteller et al, 1998; Simpkins et al, 1999). The RT-PCR analysis of hMLH1 expression showed that there is a significant association between hMLH1 hypermethylation and loss of hMLH1 expression as previously reported by Deng et al (1999). Using a polymorphism described in the promoter region of hMLH1, we were able to determine that in three MSI-H cases the hypermethylation occurs as a biallelic event as previously found by Veigl et al (1998) (data not shown). The high degree of instability associated with microsatellite markers in MSI tumors prevented us from accurately assessing the loss of heterozygosity (LOH) status of the hMLH1 gene in the tumors. Our results, together with data from literature, support the hypothesis that hMLH1 promoter hypermethylation is the most prevalent mechanism of MMR deficiency in SGC.

In 7 of the 28 MSI-H cases, no hMLH1 hypermethylation was observed, suggesting an alternative mechanism for hMLH1 inactivation in these cases. Somatic mutations in one of the MMR genes have been detected in up to 26% of sporadic MSI cancers at various sites, namely colon and endometrium (Borresen et al, 1995; Bubb et al, 1996; Katabuchi et al, 1995; Kowalski et al, 1997; Liu et al, 1995; Moslein et al, 1996; Thibodeau et al, 1996; Wu et al, 1997). Data on MMR gene mutations is scarce in SGC. Yamamoto et al (1999) showed the presence of hMSH2 and hMLH1 somatic mutations in, respectively, 6 and 1 out of 24 cases of MSI-H SGC. In our series we did not find hMLH1 mutations in any of the tumors, including the four MSI-H cases without hypermethylation that we were able to screen. Therefore, the mechanism of hMLH1 inactivation in hypermethylation/mutation-negative gastric carcinomas remains to be elucidated.

We observed a significant association between hMLH1 promoter methylation and mutations in target genes. We think that this finding may be ascribed, partly at least, to the close relationship between MSI-H and methylation status.

We found a significant association between hMLH1 hypermethylation and Ming’s classification: hMLH1 hypermethylation was detected in 62% of the expanding tumors and in 27% of the infiltrative tumors. If this finding is confirmed in a larger series, one might have to consider the possibility that target genes other than those we looked at are linked to the expanding growth pattern of the tumors. It is tempting to advance that the “new” target genes may be involved in cell-cell or cell-matrix adhesion. No other significant associations were found between hMLH1 hypermethylation and clinicopathological features. The discrepancy between the clinicopathological features of the MSI-H cases and those of cases with hMLH1 hypermethylation may be explained by two facts: the group of tumors with hMLH1 hypermethylation includes some MSI-L tumors and not all MSI-H tumors present hMLH1 hypermethylation.

In MSI-L gastric carcinomas, we found hMLH1 hypermethylation in 50% of the tumors. These results are in agreement with those reported by Fleisher et al (1999) in gastric carcinoma and by Esteller et al (1999) in endometrial carcinoma. There are no significant differences between MSS and MSI-L tumors as regards their clinicopathological profile. Toyota et al (1999) suggested that tumor development in cases presenting the MSI-L phenotype may be accompanied by random de novo hypermethylation. This hypermethylation, however, confers no selective advantage to the tumor cells. Curiously, if this hypothesis is true, MSS tumors should also have presented a proportion of cases with aberrant methylation of the hMLH1 promoter, but we did not observe this in our series. Furthermore, in two MSI-L cases in which hMLH1 expression was evaluated, we observed that hMLH1 promoter hypermethylation led to a loss of hMLH1 expression. The aforementioned facts suggest that hMLH1 hypermethylation does not occur merely by chance in MSI-L gastric carcinomas.

The possibility that MSI-L/Met+ cases represent an intermediate step in the MMR pathway has been advanced by Esteller et al (1999) regarding the occurrence of early endometrial carcinomas. This possibility does not fit with the results obtained in the present series: only two of the six MSI-L/Met+ cases were tumors restricted to the submucosa (early carcinomas), whereas the remaining four cases were advanced tumors. Because it is difficult to conceive that these MSI-L/Met+ advanced gastric carcinomas might be seen as cases not yet expressing a complete mutator phenotype, we would rather suggest that these tumors may have followed (an) unknown MMR-independent molecular pathway(s) of progression.

In biological terms we think that the distinction between MSI-L/Met+ gastric cancers and MSS tumors should be made because the aforementioned subset of MSI-L tumors share the same genetic alterations of MSI-H tumors: hypermethylation of the promoter region and absence of expression of hMLH1. For practical purposes, no benefit appears to be achieved by making a distinction between MSS and MSI-L tumors because they share the same clinicopathological profile. Our results show that mutations in target genes, more than hypermethylation or absence of expression of hMLH1, are the link between MSI status and most of the clinicopathological features of SGC.

Materials and Methods

Patients, Tissue Samples, and DNA Extraction

In this study we analyzed 57 tumors selected from a series of 152 gastric carcinomas consecutively resected at Hospital of S. João (Porto, Portugal) from 1988 to 1997, previously analyzed for MSI phenotype and, partly, for mutations in target genes (Oliveira et al, 1998). We analyzed all cases with MSI-H or MSI-L phenotype (see below) after excluding 16 cases because of the lack of good high-quality DNA and 10 cases because of insufficient material. We also analyzed a control group of 17 cases with MSS phenotype selected at random from the cases of MSS with good technical conditions. Hematoxylin and eosin-stained sections were used to classify the tumors according to Lauren’s classification. The pathological staging was achieved using the unified 1987 tumor, node, metastasis (TNM) system for gastric carcinoma. Orcein-stained sections were used for the detection of vascular invasion.

MSI Assays

The 57 gastric carcinomas were previously studied for MSI using a panel of at least five dinucleotide repeat sequences, as described by Santos et al (1996), and using a primer set localized on intron 5 of the hMSH2 gene, which amplifies an adenine quasi monomorphic mononucleotide repeat, BAT26, as described by Zhou et al (1998) and Oliveira et al (1998).

Cases were classified as having an MSI-H phenotype whenever they presented a high frequency of MSI (≥ 40%) and BAT26+. Cases were considered as MSI-L whenever they showed instability in less than 40% of the markers used. MSS cases showed no instability at any of the markers used.

Promoter Methylation Assays

The promoter methylation is analyzed as initially described by Kane et al (1997). In short, DNA is digested by HpaII or MspI, methylation sensitive and insensitive enzymes, respectively, after which PCR is performed amplifying the target region. Absence or presence of PCR product indicates nonmethylated or methylated promoter region, respectively. Tumor DNA samples were digested over 48 hours at 37° C in 100 μl total volume reactions, containing 250 ng of genomic DNA of each sample, no enzyme, 100 units of HpaII or 150 units of MspI. Five normal mucosas (distal to the tumor) of the stomach were analyzed as controls. To analyze the cleavage of the hMLH1 promoter region, 1 μl of each digest was amplified by PCR in 30 μl reactions containing 3 μl of 10× PCR reaction buffer (Pharmacia Biotech, Piscataway, New Jersey), 0.5 units of Taq polymerase enzyme (Pharmacia Biotech), 0.25 μm of each of the four deoxynucleotide triphosphates, and 0.4 μm of each primer (5′-CGCTCGTA-GTATTCGTGC-3′ and 5′- TCAGTGCCTCGTGCTC-AC-3′), designed to amplify nucleotides −670 to −65 of hMLH1 (Genbank Accession No. U83845). Thirty-five cycles of PCR were performed using the following conditions: 94° C, 30 seconds; 55° C, 30 seconds; 72° C, 120 seconds. The resulting amplification products of two independent experiments were analyzed by 2% agarose gel electrophoresis using standard conditions.

Mutation Analysis of hMLH1

Mutational analysis of hMLH1 gene was performed by denaturing gradient gel electrophoresis (DGGE). Multiplex PCR was carried out on 500 ng of DNA in a total volume of 30 μl for 35 cycles as follows: denaturing step at 94° C for 3 minutes; 5 cycles consisting of denaturation at 94° C for 1 minute, annealing at 56° C for 1 minute, and elongation at 72° C for 1 minute; 5 cycles consisting of denaturation at 94° C for 1 minute, annealing at 53° C for 1 minute, and elongation at 72° C for 1 minute; and 25 cycles each of denaturation at 94° C for 1 minute, annealing at 50° C for 1 minute, and elongation at 72° C for 1 minute; one final elongation cycle was performed at 72° C for 5 minutes. The PCR mixture contained 1× Taq reaction buffer, 1 mm MgCl2, 0.5 unit of Taq polymerase, 0.75 mm of deoxyribonucleoside triphosphate (dNTP), and 15 pmol of each primer. The primers used for amplification of hMLH1 gene are listed in the literature (Wu et al, 1997), along with fragment sizes, melting temperatures, and PCR-annealing temperatures. To increase the amount of heteroduplex molecules, a heteroduplex step was performed after PCR amplification, ie, the samples were denatured for 10 minutes at 96° C followed by renaturation for 45 minutes at 50° C. PCR products were loaded on a polyacrylamide gel containing a denaturing gradient of urea/formamide (UF) (10% UF to 70% UF). Electrophoresis was performed overnight in TAE buffer at 59° C. Gels were stained with ethidium bromide and photographed under a UV transilluminator.

Expression Analysis of hMLH1

RT-PCR studies were performed using frozen material from 15 cases (10 MSI-H, 2 MSI-L, 3 MSS). Total RNA was extracted using standard methods. First-strand synthesis was made by random 6-mer priming using M-MLV reverse transcriptase (Boehringer Mannheim, Lewes, United Kingdom) at 37° for 60 minutes. hMLH1 mRNA expression level analysis was performed by co-amplification of the target gene (hMLH1) and of the housekeeping gene GAPDH.

Amplification of the Target Genes (TGFβRII, IGFIIR, BAX, and TCF4)

Detection of mutations in target genes was performed in 50 cases (27 MSI-H, 8 MSI-L, and 15 MSS) in which constitutional and tumor DNA was available. Of the 50 cases analyzed for TGFβRII, IGFIIR, and BAX, 34 had been analyzed previously by Oliveira et al (1998). The repeat sequence (A)9 in the putative exon 10 of TCF4 could only be analyzed in 48 cases as described by Duval et al (1999). PCR products were analyzed for mutations in 6% single mutation detection enhancement/strand conformation polymorphism analysis (MDE/SSCP) gels. The results were confirmed with radioactive PCR, and the products were run in a 10% denaturing PAA sequencing gel. The tumors were classified as harboring a mutation whenever they showed band shifts relative to the normal tissue.

Statistical Analysis

The statistical analysis of the results was performed using the χ2 test with Yates correction, Fisher’s exact test, and Student’s t test. The parameters used for this analysis are listed in Tables 1 and 2. A p value less than 0.05 was considered statistically significant.