Identification of the molecular characteristics of intramucosal (IMCs) and submucosal cancers (SMCs) is essential to our understanding of early gastric carcinogenesis. However, little is known regarding the differences between the 2 lesions. One hundred and forty-eight patients with primary early gastric cancer [IMC, 106; SMC, 42] were characterized for expression of cell cycle-related proteins and loss of heterozygosity (LOH). We also examined microsatellite instability (MSI) and methylation status. For LOH and methylation studies, we used a panel of 17 microsatellite markers (3p, 4p, 5q, 9p. 13q, 17p, 18q and 22q) and promoter regions of 9 genes (MLH-1, RUNX3, p16, HPP1, RASSF2A, SFRP1, DKK-1, ZFP64 and SALL4) that are frequently altered or methylated in gastric cancers. Overexpression of p53 and cyclin D1 was observed in SMC. In addition, low expression of p27 was more frequent in SMC than in IMC. Frequencies of 4p, 9p, 13q and 22q were significantly higher in SMC than in IMC. The SALL4 gene was frequently methylated in SMC compared with IMC. However, other gene methylations were common in both IMC and SMC. The frequency of LOH-high status/methylation-low status was significantly higher in SMC than in IMC. However, LOH-low status/methylation-high status in SMC was more frequently found in IMC. Our data confirm that methylation of cancer-related genes plays a major role in the development of IMCs. Importantly, the results also show that gastric submucosal progression is characterized by the accumulation of specific genetic alterations. In addition, changes of cell cycle-related proteins are associated with cancer progression.
Gastric adenocarcinoma is one of the most commonly diagnosed malignancies worldwide and is a leading cause of cancer mortality in Japan, Korea and South America. Although human tumor development has been analyzed at the molecular level,1, 2 the underlying molecular alterations that drive the neoplastic process in gastric cancers are not understood. In human tumors, impairment of the cell cycle is likely a critical mechanism underlying tumor development. G1 cyclins (cyclins D1 and A) and cyclin-dependent kinase (CDK) complexes play important roles in the transition through the G1 phase of the cell cycle, and their overexpression is implicated in neoplasia.3–6 CDK inhibitors, including p27 and p21, negatively regulate G1 progression by binding to G1 cyclins/CDK complexes, thus inhibiting their activity and thereby preventing entry into the cell cycle.3–5 In addition, nuclear accumulation of β-catenin plays an essential role in cell cycle progression as it induces cyclin D1, c-myc and MMP7.7, 8, 9 Analysis of cell cycle-related protein expression is important for early gastric cancers (EGCs) to elucidate the mechanisms behind early gastric carcinogenesis.3, 4
Many laboratories have studied genomic instability and the resulting allelic imbalance in gastric carcinomas through analyses of loss of heterozygosity (LOH) and comparative genomic hybridization.10, 11 Alterations most commonly found in gastric carcinogenesis include LOHs at 3p, 5q, 9p, 13q, 17p, 18q and 22q.12–14 Thus, LOHs are useful markers for defining tumor aggressiveness. Accumulation of tumor LOHs (LOH-high status) is closely associated with tumor progression or tumor-invasive ability. Epigenetic aberrations are mechanistically important in human carcinogenesis.15 A number of tumor suppressor genes are silenced by promoter methylation during gastric cancer development. Some gastric cancers undergo promoter methylation, which is referred to as the CpG island methylator phenotype (CIMP, methylation status).15, 16 In general, methylation-high (CIMP-high) tumors have distinct features, such as favorable tumor location (proximal location), greater predilection for females and specific genetic alterations (BRAF mutation and low incidence of p53 alteration).17 In addition, microsatellite instability (MSI) defines a novel molecular subtype of tumors.17, 18 In previous studies, MSI detected different changes compared with those observed in LOH-high status tissue. MSI overlaps with CIMP status (methylation-high status) in terms of clinicopathological and molecular features.17–19 A molecular classification based on LOH status, methylation status and MSI is increasingly important in gastric carcinogenesis, because those alterations reflect global genomic or epigenetic aberrations.15, 17, 18
EGC is defined as a tumor that may invade into but is confined to the submucosa, irrespective of the presence of lymph node metastases.19 Analysis of molecular alterations in EGC is important for understanding initial events in early gastric carcinogenesis. EGC is subclassified into intramucosal (IMCs) and submucosal (SMCs) cancers.19 SMC could be regarded as an intermediate stage between IMC and advanced cancer. We hypothesize that submucosal tumors possess distinct molecular alterations compared with IMCs and that those alterations are critical for tumor progression. Thus, identifying the molecular differences between IMCs and SMCs is essential for understanding gastric carcinogenesis. In the present study, we attempted to identify these molecular differences.
Material and Methods
Detailed clinicopathological data derived from 106 IMCs and 42 SMCs are summarized in Table 1. The criteria used to diagnose IMC were based on Japanese histological criteria,20 which differ from those used by western pathologists.21, 22 The nuclear grade was determined according to published criteria.14
Table 1. Clinicopathological findings of early gastric cancers
For molecular investigations, tumor tissue was isolated from the resected stomach [samples from endoscopic submucosal dissection (ESD): 106 IMCs and 9 SMCs; samples from gastrectomy: 33 SMCs] using biopsy forceps within 30 min of resection. Most SMCs were obtained from gastrectomy, given that those SMCs were not an indication for ESD. The most distant normal gastric mucosa from the neoplasm was removed from the submucosa using scissors as a control for molecular analysis. Tissue for clinicopathological analysis was obtained from a region of the resected stomach adjacent to the region used for genetic analysis. All of the noncontrol tissue samples contained cancer tissue. Only tumor samples where the neoplastic cells accounted for at least 50% of the tissue cell population were selected (13 of 161 tumors were omitted due to failure to meet these criteria). In the SMCs, only tumor samples obtained from submucosal lesions were used for molecular analysis.
EGC specimens were fixed in buffered formalin and embedded in paraffin, according to routine procedures. For this study, 3-μm sections were prepared, dried, deparaffinized and rehydrated before microwave treatment (H2500, Microwave Processor, Bio Rad) in citrate buffer (pH 6.0) for 5 min. An automatic staining machine (DAKO Envision+ system) was used for the immunohistochemical procedure.14 The slides were counterstained in hematoxylin, dehydrated and mounted. The antibody sources used in this study are shown in Table 2.
Table 2. List of antibodies that we used
Nuclear immunostaining data for cyclin D1, cyclin A, p21, p27, p53, β-catenin, MLH-1 and ki-67 were expressed as the percentage of positive epithelial cells in relation to the total number of cells encountered in at least 5–10 representative high-power fields (500–1,000 epithelial cells). The immunoreactivity was measured by means of light microscopic examination and evaluated independently by 2 experienced pathologists. (T.S. and N.U., not included in this manuscript). Differences in interpretation were reconciled by reviewing the slides separately. No necrosis was found in the tissue examined. Only nuclear staining was regarded as significant, and tumor cells showed a range of staining intensities. Cytoplasmic staining was disregarded. Tumor cells with weak staining were not considered positive.
Cutoffs for positive expression of p21 and p27 were defined as >10%, while for cyclin D1 and cyclin A, positivity was >30% as conducted by previous studies.6, 23, 24 In the evaluation of p53 overexpression, samples showing >10% staining were considered positive.14 More than 20% of cells showing β-catenin nuclear expression were regarded as positive.25 Expression of MLH-1 was interpreted and compared with the surrounding tissue. If expression of MLH-1 in the tumor cell was considered to be low compared with its surrounding nonneoplastic tissue, the tumor tissue was regarded as low/negative staining. Finally, ki-67 staining >30% was judged as a highly proliferative tumor.
Mucin phenotypes of tumor cells were subclassified as follows: gastric, intestinal, mixed and unclassified, according to previous criteria.14 In brief, the gastric phenotype included tumors having predominantly intracytoplasmic expression of gastric mucin, as determined by immunostaining of human gastric mucin (MUC5AC) or (and) pyloric gland mucin (MUC6), but with no MUC-2 positive cells. The intestinal phenotype was constituted by MUC2-positive cells or (and) CD10-positive cells (along with a brush border). The mixed-type contained immunostaining features of both gastric type as well as intestinal type. Finally, tumors lacking gastric or intestinal staining patterns were termed “unclassified.” Immunopositivity of >5% of the tumor cells was regarded as positive according to the guidelines established by a previous report.26
PCR reactions were performed using a thermal cycler (GeneAmp PCR System 9600, Perkin-Elmer, CA) with 50–100 ng of genomic DNA as template, 25 pM of each primer, 0.2 mM deoxynucleotide triphosphate (dNTP), 1× reaction buffer containing 1.5 mM MgCl2, and 1.5 u Taq polymerase (Boehringer Mannheim Co., Germany) in a final reaction volume of 25 μL. Samples were processed for 25–30 cycles, with each cycle consisting of 30 sec at 94°C, 1 min at 55–58°C and 2 min at 72°C, followed by a final 10-min extension at 72°C. For quantitative detection of the allelic loss at each locus, PCR-LOH analysis and MSI were performed as described previously.27, 28 A 1-μL aliquot of the PCR product was added to 3 μL formamide and a 0.5 μL TAMRA 500 size standard (Applied Biosystems, CA), loaded on a 6% polyacrylamide-8 M urea gel, and run for 2–6 hr in a 373A Automated Sequencer (Applied Biosystems) at a constant power of 30 W.
Assessment of LOH by polymerase chain reaction
Allelic losses on chromosomes 3p, 4p, 5q, 9p, 13q, 17p, 18p and 22q were examined in paired tumor and normal DNA samples obtained from 148 EGC patients using 17 highly pleomorphic microsatellite markers (D3S2402, D3S1234, D4S2639, D4S1601, D5S107, D5S346, D5S299, D5S82, D9S171, D9S1118, D13S162, TP53, D18S487, D18S34, D22S274, D22S1140 and D22S1168). These microsatellite markers have been used frequently in studies of gastric carcinomas.10, 12–14 In addition, a variable number of tandem repeat polymorphisms at the DCC locus were tested. Microsatellite sequences were obtained from specific primers reported in the Genome Database (http://gdbwww.gdb.org/gdb/).
Determination of LOH
The peaks produced by the normal DNA sample were used to determine whether the cancerous sample was homozygous (1 peak) or heterozygous (2 peaks). The allelic ratio was calculated as described by Habano et al.27 A tumor was considered to have allelic loss if the allele peak ratio was less than or equal to 0.60, representing an allelic signal reduction of at least 40%. We interpreted this allelic imbalance as allelic loss (LOH) with the provision that, in some cases, the changes in the allele peak ratio may have resulted from allelic amplification. Tumors exhibiting MSI at a given locus were not evaluated for allelic loss. The data were collected automatically and analyzed using GeneScan software (Applied Biosystems) to determine the allele score and to assess the possibility of allelic loss.
Scoring of LOH
LOH status was scored according to the following criteria. A tumor sample was considered to be LOH-high if 3 or more of the markers showed allelic loss. When data showed that 1 or 2 markers were lost, the tumor was designated as LOH-low.19
Assessment of MSI
Six different loci were considered for MSI assessment, including all those recommended by the Bethesda panel for colon cancer (BAT25, BAT26, D5S346, D2S123 and D17S250).29 A tumor was defined as MSI-positive when a novel, abnormal-sized band occurred in the tumor sample compared with the corresponding normal DNA sample. MSI-positive colorectal carcinomas were used as controls in the study and were divided into 2 groups, those with high-level instability (i.e., MSI at [dbmtequ]33% of loci) and those with low-level instability (i.e., MSI at [dbltequ]17% of loci), as described previously. However, tumors with only 1 alteration of the marker examined using the above criteria and those previously categorized as MSI-low were considered MSI-negative tumors in this study.
Confirmation of methylation
Bisulfite treatment of genomic DNA was carried out as described previously.30 For examination of methylation status, we used a combined bisulphate restriction analysis (COBRA) as described previously.29, 30 The COBRA of MLH-1, RUNX3, p16, HPP1, RASSF2A, SFRP1, DKK-1, ZFP64 and SALL4 genes were determined using primers, restriction enzyme and conditions as previously described.39–37
The colon cancer cell lines RKO and SW48 (American Tissue Culture, Manassas, VA) and water were used as positive and negative controls, respectively. After amplification, the PCR products were digested with restriction enzymes and electrophoresed on 3% agarose gels. The gels were stained with ethidium bromide, and the proportion of methylated alleles was visually compared with unmethylated alleles. Cleavage fragments were quantified by densitometry.
Tumors were classified as methylation-negative/low (methylation-low) if 1 or 2 loci were methylated and methylation-high if 3 or more were methylated.
The data were analyzed using a chi-squared test with the aid of StatView-IV software (Abacus Concepts, Berkeley, CA). Samples were determined to be significantly different when the p value was less than 0.05.
Expression of cell cycle-related proteins in gastric IMCs and SMCs
The results of the survey are depicted in Figure 1. Overexpression of the p53 gene product was significantly more frequent in SMCs (40.5%) than in IMCs (11.3%; p < 0.01). Conversely, the frequency of low expression of the p27 product was statistically higher in SMCs (23.8%) than in IMCs (2.8%; p < 0.01). In addition, cyclin D1 overexpression was more frequently found in SMCs (26.5%) compared with IMCs (11.3%; p < 0.05). The Ki-67 positivity rate for SMCs (76.2%) was higher than that of IMCs (52.8%; p < 0.05). There were no significant differences in low expression of p21 and MLH-1 between IMCs (29.2 and 15.1%, respectively) and SMCs (35.7 and 9.5%, respectively). Overexpression of cyclin A and nuclear accumulation of β-catenin were commonly observed in IMCs (46.2 and 39.6%, respectively) and SMCs (61.9 and 26.5%, respectively). Although MMP7 was primarily expressed at the invasive front of the submucosal lesion, no significant difference of MMP7 was observed between IMCs (38.7%) and SMCs (47.6%). As for mucin phenotype, the gastric phenotype was found significantly more often in SMCs (33.3%) than in IMCs (14.2%) (p < 0.05). By contrast, the intestinal phenotype was expressed more in IMCs (41.5%) than in SMCs (19%).
Analysis of LOHs at multiple chromosomal loci in IMCs and SMCs
LOH data are summarized in Figure 2. Allelic loss of 5q was a common alteration in IMCs (38.5%) and SMCs (39.5%). Although allelic losses of 17p, 18q and 3p were more frequent in SMCs (31.4, 33.3 and 27.8%, respectively) than in IMCs (22, 19.1 and 20.8%, respectively), no significant differences were found. The frequencies of 4p, 9p and 22q allelic losses were significantly higher in SMCs (47.2, 30.6, and 34.2%, respectively) than in IMCs (15.1, 14.8 and 12.6%, respectively) (p < 0.01). Finally, allelic loss at 13q was more frequently found in SMCs (27.3%) than in IMCs (4.9%) (p < 0.05).
Analysis of methylation in IMCs and SMCs
The associations between the 2 lesions are shown in Figure 3. High frequencies of methylation of RUNX-3, RASSF2A, DKK1, SFRP1, HPP1 and ZFP64 were commonly observed in both IMCs (42.7, 57.5, 50, 66, 66 and 51.9%, respectively) and SMCs (50, 52.4, 47.6, 54.8, 54.8, and 45.2%, respectively). However, only SALL4 methylation was significantly more frequent in SMCs (40.5%) compared with IMCs (18.9%) (p < 0.01). Although p16 was frequently methylated in SMCs (11.9%), when compared with IMCs (3.8%), no significant difference was detected between them. The frequencies of MLH-1 methylation were low in IMCs (8.5%) and SMCs (7.1%).
Analysis of molecular status of IMCs and SMCs
The molecular status of intramucosal and submucosal tumors is displayed in Figure 4. The frequency of LOH-high/methylation-low status was significantly higher in SMCs (35.7%) than in IMCs (5.7%; p < 0.01). On the other hand, LOH-L/methylation-H status was more frequently found in IMCs (53.8%) than in SMCs (26.2%; p < 0.01) Although LOH-L/methylation-L status was found more frequently in IMCs (20.8%) compared with SMCs (9.5%), the difference did not reach a statistically significant level. There were no significant differences of LOH-H/methylation-H status between IMCs and SMCs. The frequencies of MSI in IMCs and SMCs were 7.6 and 9.5%, respectively. Finally, the frequency of p53 overexpression of LOH-high status cancers (47.6%) was significantly higher than that of LOH-L status (9.6%; p < 0.01).
Representative histological and molecular examples of SMC are shown in Figures 5 and 6.
This study was conducted to examine the differences in molecular alterations between IMCs and SMCs in early gastric carcinogenesis. Identifying the molecular differences between the 2 lesions is important in understanding carcinogenetic mechanisms involving SMCs. This study successfully documented a number of differences in the molecular status of IMCs and SMCs.
Recent studies have shown that mucin expression by tumor cells is closely associated with gastric tumorigenesis.38 In fact, the mucin phenotype of tumor cells correlates with clinicopathological findings and molecular alterations.14, 38 In the present study, although the intestinal phenotype was primarily found in IMCs, the gastric phenotype was frequently detected in SMCs. On the other hand, Wakatsuki et al. indicated that tumors with the intestinal phenotype had the worst prognosis.39 While this finding seems to conflict with our data, patient prognosis is often determined by a multitude of factors. Our finding suggests that tumor cells with a gastric phenotype are more likely to invade into the submucosa in gastric cancers.
In the present study, p53 overexpression was more frequently found in SMCs than in IMCs. This finding suggests that p53 overexpression plays a major role in submucosal invasion of IMC and represents a novel predictive marker for SMC. In addition, this finding implies that p53 overexpression is not associated with the mucosal onset of gastric carcinogenesis. In our data, the frequency of p53 overexpression by SMC was ∼40%, consistent with previous reports ranging from 35.7 to 57.1% in gastric SMCs.40, 41 On the other hand, it is well known that p53 overexpression is correlated with tumor nuclear grade or mucin phenotype.22 The present study showed that the frequencies of both gastric phenotype and tumor cells of high nuclear grade were greater in SMCs than in IMCs. These findings suggest that both gastric mucin phenotype and tumor nuclear grade reflect p53 overexpression in gastric SMCs.
Previous studies have shown that cyclins are useful immunohistochemical markers when evaluating the aggressiveness of tumor cells.4, 6 In the present study, cyclin D1 was upregulated in SMCs compared with IMCs. Cyclin D1, a target gene of Wnt signal transduction, is an essential molecule in human tumorigenesis. Cyclin D1 induces significant changes in gene expression in human tumors, such as phosphorylated Rb, myc and cell adhesion-related genes.8, 42 This finding suggests that cyclin D1 overexpression may play an important role in the progression from IMC to SMC. With regard to cyclin A, however, no difference was found between IMCs and SMCs, suggesting that upregulation of cyclin A is an early event in the development of differentiated-type gastric cancers.
Low or nonexistent expression of p27 was associated with patient outcome and prediction of tumor aggressiveness,43 although opposing data have been reported.44 However, the point at which low p27 expression occurs in the progression of gastric cancers remains unknown. We found that low expression of p27 was associated with submucosal invasion in differentiated-type gastric cancers. However, Oya et al. indicated that reduced expression of p27 occurred more frequently in carcinoma than in adenoma.45 Although the reason for these differences remains unknown, they could be due to the number of samples examined, differences in cutoff values for the examined proteins, the sources of antibodies, or the tumor grade examined in the study. In addition, our survey demonstrated that the reduction of p21 was a common alteration in both IMCs and SMCs. This finding suggests that p21 reduction is closely associated with early tumorigenesis of gastric cancers.
Numerous publications have reported regions of LOH that occurred frequently in gastric carcinogenesis, particularly those at 5q, 17p, 18q, 3p, 4p, 9p, 13q and 22q, including APC and p53 genes.10–14 One previous study showed that aggressive tumor cells tend to acquire LOH and thus LOH, in some cases, proves to be a novel marker for predicting tumor-invasive or aggressive activities.11 In the present study, 5q allelic loss was a common alteration in both IMCs and SMCs. In contrast, the frequencies of 4p, 9p, 13q and 22q allelic losses were significantly higher in SMCs than in IMCs. These findings suggest that although 5q allelic loss plays a major role in early gastric carcinogenesis, 4p, 9p, 13q and 22q allelic losses are associated with submucosal invasion from mucosal lesions. The frequencies of 17p, 18q and 3p allelic losses in IMCs were low in the present study. We suggest that those allelic losses also contribute to the early development of gastric carcinogenesis, because the frequencies of those allelic losses differed among mucin phenotypes of the differentiated EGCs. That is, the frequencies of those allelic losses were low in intestinal phenotype cancers, but high in gastric phenotype cancers.
We demonstrated that gastric cancer cells are heavily methylated in the early stages of gastric tumorigenesis. This finding suggests that a progressive increase in methylation of most of those genes does not occur during the progression of EGCs. Rather, the data indicate that methylation is an early and essential event in gastric carcinogenesis. In addition, we found that methylation of RUNX3, HPP1, RASSF2A, SFRP1, DKK-1 and ZFP64 gene promoters plays an important role in early gastric carcinogenesis. In particular, methylation of the ZFP64 gene was described in the first report identifying frequent alterations in gastric cancers, although its function remains unclear. The ZFP64 gene may be a key gene in gastrointestinal tumorigenesis.
The frequencies of p16 gene methylation were very low in both lesions. This suggests that inactivation of the p16 gene plays a minor role in EGC. According to our data, only SALL4 methylation frequency was higher in SMCs than in IMCs. SALL4 methylation may therefore be associated with submucosal invasion of IMC. Although the function of the SALL4 gene remains unknown,46 its expression is linked to Wnt signal transduction.47 Our finding implies that submucosal invasive ability may be activated by the methylation of the SALL4 gene, given that SALL4 methylation was significantly higher in aneuploid populations than in diploid populations.36 Further study will be needed regarding the role of SALL4 gene methylation in gastric carcinogenesis.
MSI is an early event in gastric carcinogenesis.46–48 In the present study, MSI showed similar frequencies in IMCs and SMCs, supporting previous studies.48, 49 However, our data indicate that MSI is involved in the progression of gastric carcinogenesis. Furthermore, in gastric carcinogenesis, MSI accelerates neoplastic progression.50 Here, LOH-high/methylation-low status was associated with SMCs. In contrast, a LOH-low/methylation-high status was frequently found in IMCs. These findings suggest that LOH-high status is associated with submucosal invasion of a mucosal lesion, while high methylation status plays an essential role in early development of gastric carcinogenesis. In addition, accumulation of LOHs (LOH-high status) in the mucosal lesion may be a predictive marker for submucosal invasion.
In conclusion, the present data suggest that SMCs acquire distinct molecular alterations in gastric carcinogenesis. These molecular characteristics are depicted in Figure 7. This illustration shows that although IMC is caused by methylation (epigenetic) dominant alterations, genetic alterations must accumulate for progression to SMC. In addition, alteration of some cell cycle-related proteins plays an important role in cancer evolution. Further studies will be required to clarify the origins of human early gastric carcinogenesis.
We gratefully acknowledge the technical assistance of Miss E. Sugawara and Mr. T. Kasai. We also thank the members of the Division of Molecular Diagnostic Pathology, Department of Pathology.