Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Systematic search for gastric cancer-specific genes based on SAGE data: melanoma inhibitory activity and matrix metalloproteinase-10 are novel prognostic factors in patients with gastric cancer

Abstract

Gastric cancer (GC) is one of the most common malignancies worldwide. Genes expressed only in cancer tissue will be useful molecular markers for diagnosis and may also be good therapeutic targets. However, little is known about cancer-specific genes, at least in GC. In this study, we searched for GC-specific genes by serial analysis of gene expression (SAGE) data analysis and quantitative reverse transcription (RT)–PCR. Comparing GC SAGE libraries with those of various normal tissues in the SAGEmap database, we identified 54 candidate GC-specific genes. Quantitative RT–PCR analysis of these candidates revealed that APin protein (APIN), taxol resistance-associated gene 3 (TRAG3), cytochrome P450, family 2, subfamily W, polypeptide 1 (CYP2W1), melanoma inhibitory activity (MIA), matrix metalloproteinase-10 (MMP-10), dickkopf homolog 4 (DKK4), GW112, regenerating islet-derived family, member 4 (REGIV), and HORMA domain-containing 1 (HORMAD1) were expressed much more highly in GC than in 14 kinds of normal tissues. Immunohistochemical staining for MIA, MMP-10, and DKK4 was found in 47 (31.1%), 68 (45.0%), and two (1.3%) of 151 GCs, respectively, and staining for both MIA and MMP-10 was correlated with poor prognosis in advanced GC (P=0.0001 and 0.0141, respectively). Moreover, enzyme-linked immunosorbent assay showed high levels of MMP-10 (65/69, 94.2%) in serum samples from patients with GC. Levels of MIA were raised in a small proportion of serum samples from patients with GC (4/69, 5.8%). In Boyden chamber invasion assays, MIA-transfected GC cells were up to three times more invasive than cells transfected with empty vector. Taken together, these results suggest that MMP-10 is a good marker for the detection of GC and that MIA and MMP-10 are prognostic factors for GC. As expression of MIA and MMP-10 is narrowly restricted in cancer, these two molecules may be good therapeutic targets for GC.

Introduction

According to the World Health Organization, gastric cancer (GC) is the fourth most common malignancy worldwide, with approximately 870 000 new cases occurring yearly. Mortality due to GC is second only to that due to lung cancer (Ohgaki and Matsukura, 2003). Cancer develops as a result of multiple genetic and epigenetic alterations (Yasui et al., 2000; Ushijima and Sasako, 2004). Better knowledge of changes in gene expression that occur during gastric carcinogenesis may lead to improvements in diagnosis, treatment, and prevention. Identification of novel biomarkers for cancer diagnosis and novel targets for treatment is a major goal in this field (Yasui et al., 2004). Genes encoding transmembrane/secretory proteins expressed specifically in cancers may be ideal biomarkers for cancer diagnosis (Buckhaults et al., 2001). If the function of the gene product is involved in the neoplastic process, this gene may constitute a therapeutic target.

We previously performed serial analysis of gene expression (SAGE) on four primary GC samples (Gene Expression Omnibus accession number GSE 545; SAGE Hiroshima GC tissue) and identified several genes and tags that are potentially involved in invasion, metastasis, and carcinogenesis (Oue et al., 2004; Yasui et al., 2004). In this study, to identify potential molecular markers for diagnosis of GC and molecular therapeutic targets, we systematically searched for GC-specific genes in SAGE libraries. Comparing GC SAGE libraries with those of various normal tissues in the SAGEmap database (Lal et al., 1999), we identified 54 candidate GC-specific genes in GC libraries but not in libraries from 14 normal tissues, including brain, lung, and heart. We also performed quantitative reverse transcription (RT)–PCR to investigate the specificity of these candidate GC-specific genes. We show that APin protein (APIN), taxol resistance-associated gene 3 (TRAG3), cytochrome P450, family 2, subfamily W, polypeptide 1 (CYP2W1), melanoma inhibitory activity (MIA), matrix metalloproteinase-10 (MMP-10), dickkopf homolog 4 (DKK4), GW112, regenerating islet-derived family, member 4 (REGIV), and HORMA domain-containing 1 (HORMAD1) were much more highly expressed in GC than in normal tissues. Among these genes, overexpression of REGIV and GW112 in GC has been reported (Oue et al., 2004, 2005; Zhang et al., 2004). Immunohistochemical analysis of MIA, MMP-10, and DKK4 in 151 GC samples revealed that MIA and MMP-10 are frequently overexpressed in GC. We also measured MIA and MMP-10 in serum and peritoneal wash fluid from patients with GC to investigate the potential utility of these measurements in cancer diagnosis.

Results

Identification of genes expressed more highly in GC than in normal tissues

To identify genes expressed specifically in GC, we compared tags from each GC SAGE library to the normal SAGE libraries (white matter, cerebellum, thalamus, heart, lung, stomach, colon, liver, kidney, leukocyte, peritoneum, skeletal muscle, spinal cord, and lymph node) as described in Materials and methods. We obtained 24 candidates from W226T, 15 candidates from W246T, 27 candidates from S219T, and 13 candidates from P208T. In total, we identified 54 individual candidate genes in our GC libraries but not in the normal libraries (Table 1). To confirm that these candidates were GC-specific, quantitative RT–PCR was performed to measure the expression of these candidates in nine GC samples and in 14 normal tissues (heart, lung, stomach, small intestine, colon, liver, pancreas, kidney, bone marrow, peripheral leukocytes, spleen, skeletal muscle, brain, and spinal cord). Representative results are shown in Figure 1. Expression of the 54 candidate genes was not necessarily specific for GC. However, several genes showed much higher expression in GC than in normal tissues. We then focused on cancer specificity. We calculated the specificity index for each gene. First, we identified the normal tissue in which the target gene expression was highest (mRNA expression levels are shown as A, Table 1). We then identified the GC among nine in which the target gene expression was highest (mRNA expression levels are shown as B, Table 1). The specificity index (B/A ratio) for each gene is shown in Table 1. Of the 54 candidates, nine genes: APIN, TRAG3, CYP2W1, MIA, MMP-10, DKK4, GW112, REGIV, and HORMAD1 were found to show high specificity for GC.

Table 1 Summary of quantitative RT–PCR analysis of candidate genes specifically expressed in gastric cancer
Figure 1
figure1

Quantitative RT–PCR analysis of candidate cancer (at least GC)-specific genes in 14 normal tissues and nine GC samples. Definitions of high specificity, low specificity, and no specificity are as described in Materials and methods. mRNA expression levels of MIA, MMP-10, and DKK4 were much higher in GC samples than in normal tissues. In contrast, mRNA expression levels of MAPK13, JUN, and IL16 were not significantly different between GC and normal tissues.

mRNA expression of high-specificity genes for GC

Expression of the nine high-specificity genes for GC was analysed by quantitative RT–PCR in an additional 44 GC samples and corresponding non-neoplastic mucosa samples. We calculated the ratio of target gene mRNA expression levels between GC tissue (T) and corresponding non-neoplastic mucosa (N). T/N ratios >2-fold were considered to represent overexpression. Genes showing overexpression in >40% of the samples included GW112 (25/44, 56.8%), MMP-10 (23/44, 52.3%), CYP2W1 (22/44, 50.0%), HORMAD1 (20/44, 45.5%), and TRAG3 (18/44, 40.9%). Other genes were overexpressed in <30% of the samples examined (MIA, 13/44, 29.5%; APIN, 12/44, 27.3%; DKK4, 11/44, 25.0%). mRNA expression levels of GW112 were correlated with T grade (depth of tumor invasion, P<0.0001), N grade (degree of lymph node metastasis, P=0.0089), and tumor stage (P=0.0019; Table 2). Those of MIA were also correlated with T grade (P=0.0007), N grade (P=0.0335), and tumor stage (P=0.0068; Table 2).

Table 2 Relation between mRNA expression and clinicopathologic characteristics in gastric cancer

Immunohistochemical staining for MIA, MMP-10, and DKK4 in GC and noncancerous tissues

To confirm overexpression of genes whose expression by RT–PCR was much higher in GC than in normal tissues, we performed immunohistochemical analysis of MIA, MMP-10, and DKK4 because antibodies against these three proteins are available. Immunohistochemical analysis was performed in noncancerous tissues with the highest mRNA expression to serve as positive controls. Immunostaining of MIA and MMP-10 in the lung, in which mRNA expression of both 2 genes was the highest, showed staining of chondrocytes in peribronchial cartilage, but not of epithelial components (Figure 2a, b). Both MIA and MMP-10 are reported to be expressed in cartilage (Dietz and Sandell, 1996; Kevorkian et al., 2004). Little is known about DKK4 expression. DKK4 mRNA is reported to be undetectable in all human adult and fetal tissues examined by Northern blotting (Krupnik et al., 1999). As our quantitative RT–PCR showed the highest expression of DKK4 in the duodenum, immunostaining for DKK4 was performed in the duodenum, and staining for DKK4 was observed in a small number of epithelial cells (Figure 2c). Staining was absent with antibody preincubated with DKK4 protein (data not shown).

Figure 2
figure2

Immunohistochemical analysis of MIA, MMP-10, and DKK4 in noncancerous tissue and GC tissue. Staining for MIA (a) and MMP-10 (b) was found in peribronchial cartilage of the lung. DKK4 staining (c) was detected in a small number of epithelial cells in the duodenum. In GC samples, staining for MIA (d), MMP-10 (e), and DKK4 (f) was found in GC cells. Original magnifications, a, b, and df, × 400; c, × 200. (g) Prognostic value of MIA and MMP-10 staining. The prognosis of patients with MIA- or MMP-10-positive tumors was significantly worse in the group of 58 advanced GC patients (P=0.0001 and 0.0141, respectively, log-rank test).

Immunohistochemistry was then performed on 151 GC samples (Figure 2d–f). Of these, 47 (31.1%) were positive for MIA staining, 68 (45.0%) were positive for MMP-10 staining, and two (1.3%) were positive for DKK4 staining. We analysed the relation between staining for each of these three proteins and clinicopathologic characteristics. Staining for MIA was correlated with T grade (P=0.0002), N grade (P=0.0015), and tumor stage (P<0.0001) (Supplementary Table 1). Staining for MMP-10 was correlated with T grade (P=0.0306) (Supplementary Table 2). There was no clear correlation between DKK4 staining and clinical characteristics (data not shown). We also analysed the prognostic value of MIA and MMP-10 staining. The prognosis of patients with MIA- or MMP-10-positive tumors was significantly worse in the group of 58 advanced GC (invading through the muscularis propria into the serosa) patients (P=0.0001 and 0.0141, respectively, log-rank test) (Figure 2g). In corresponding non-neoplastic gastric mucosa from GC patients, staining for MIA and MMP-10 was weak or negative, whereas DKK4-positive cells were detected in intestinal metaplasia of the stomach (data not shown).

MIA and MMP-10 levels in serum and peritoneal wash fluid from patients with GC

MIA and MMP-10 are reported to be secreted (Blesch et al., 1994; Ramos et al., 2004). Therefore, we determined whether these proteins can be detected in sera from patients with GC by enzyme-linked immunosorbent assay (ELISA). Among the 151 GC cases analysed by immunohistochemistry, serum samples were available for ELISA from 69 GC cases. MIA is known to be a tumor marker to detect metastatic disease in patients with malignant melanomas (Bosserhoff et al., 1997), but MIA levels in serum from patients with GC have not been investigated. MIA levels were significantly higher in four of 17 serum samples from patients with stage IV GC than in those of healthy individuals (Figure 3a). MIA serum levels were not significantly different between patients with stage I, II, or III GC and healthy individuals. Of 20 serum samples from patients with gastritis, one showed a high MIA level. But, levels of MIA in all 20 serum samples were below 15 ng/ml. When the cutoff level for MIA was set at 20 ng/ml, the sensitivity for detection of GC was only 5.8% (4/69), but specificity was 100.0% (60/60). We compared the protein expression status obtained by immunostaining with serum levels of the MIA measured by ELISA in 69 GC cases. Levels of MIA in serum samples from the patients with GC showing MIA-positive immunostaining (n=29, mean±s.e. 7.7±2.1 ng/ml) did not differ significantly from those with GC showing MIA-negative immunostaining (n=40, 4.1±0.3 ng/ml) (P=0.7656, Mann–Whitney U-test). We also measured MIA levels in peritoneal wash fluid from patients with GC (Figure 3a). Of two peritoneal wash cytology-positive samples, one showed a very high MIA level.

Figure 3
figure3

Analysis of serum and peritoneal wash fluid samples from patients with GC by ELISA. (a) High levels of MIA were detected in serum samples from four patients with stage IV GC. Yellow bars indicate the cutoff levels defined in this study. Red bars indicate the mean of protein levels. In peritoneal wash fluid samples, one and two cytology-positive samples showed high levels of MIA. (b) High levels of MMP-10 were detected in 65 serum samples from patients with GC, including stage I GC. Among 60 serum samples from healthy individuals, high levels of MMP-10 were detected in 15. Yellow bars indicate the cutoff levels defined in this study. Red bars indicate mean of protein levels. Error bars indicate s.e. from the mean. In peritoneal wash fluid samples, one and two cytology-positive samples showed high levels of MMP-10. *Mann–Whitney U-test.

To our knowledge, although some MMPs are good serum markers for cancer detection (Zucker et al., 1999), there are no reports regarding MMP-10 levels in serum from patients with cancer including GC. MMP-10 was also detected in serum samples. In contrast to levels of the MIA, high levels of MMP-10 were detected in serum samples from most of the patients with GC (mean±s.e.; stage I, 455.8±38.1 pg/ml; stage II, 526.5±68.5 pg/ml; stage III, 574.1±61.1 pg/ml; stage IV, 546.0±51.0 pg/ml), even at stage I (Figure 3b). Levels of MMP-10 in serum samples from the patients with GC showing MMP-10-positive immunostaining (n=34, 553.0±38.3 pg/ml) were higher than those with GC showing MMP-10-negative immunostaining (n=35, 451.0±33.7 pg/ml), but not statistically significant (P=0.1770, Mann–Whitney U-test). High levels of MMP-10 were also detected in serum samples from some healthy individuals (81.4±25.5 pg/ml) and some patients with gastritis (47.8±23.0 pg/ml). When the cutoff level for MMP-10 was set at 200 pg/ml, the sensitivity and specificity for detection of GC was 94.2% (65/69) and 85.0% (51/60), respectively. Sensitivity for patients with stage II–IV GC was 100%. Levels of MMP-10 in all 27 peritoneal wash cytology-negative samples were below 50 pg/ml (Figure 3b). Two peritoneal wash cytology-positive samples showed levels of MMP-10 that were significantly higher than those in peritoneal wash cytology-negative samples.

Effect of MIA on cell growth and invasive activity of MKN-28 cells

High levels of MIA mRNA expression were correlated with T grade, N grade, and tumor stage in GC tissues. In addition, immunostaining for MIA protein was correlated with T grade, N grade, tumor stage, and poor prognosis. MIA acts as a potent tumor cell growth inhibitor for malignant melanoma cells (Blesch et al., 1994) but not for pancreatic cancer cells (El Fitori et al., 2005), whereas overexpression of MIA enhances the invasiveness of both melanoma cells and pancreatic cancer cells (Bosserhoff et al., 2001; El Fitori et al., 2005). To investigate the biologic significance of MIA in GC, the MKN-28 GC cell line was stably transfected with vector expressing MIA. MKN-28 cells were selected for low MIA expression (data not shown). Cells were transfected with plasmid vectors capable of expressing MIA constitutively. Clones were selected in G418 and examined for MIA expression by MIA ELISA (Figure 4a). Clones that expressed MIA at significantly increased levels relative to the parent are designated as MKN-28-1, MKN-28-2, and MKN-28-3. To determine the effect of MIA on cell growth, MTT assays were performed. Cell growth of MKN-28 cells expressing higher levels of MIA did not differ from that of cells transfected with empty vector up to days 2 (Figure 4b). We then performed Boyden chamber invasion assays. MIA-transfected MKN-28 cells were up to three times more invasive than cells transfected with empty vector on day 2 (MKN-28-1, P=0.014; MKN-28-2, P=0.046; MKN-28-3, P=0.025) (Figure 4c).

Figure 4
figure4

Expression of MIA in MKN-28 cells. (a) MIA protein levels in the culture media of MKN-28 cells transfected with pcDNA-MIA or pcDNA 3.1 constructs were measured by ELISA. In MKN-28 cells stably transfected with pcDNA-MIA, increased MIA levels were detected. (b) Effect of MIA expression on cell growth of MKN28 cells. Cell growth was assessed by MTT assay at 1 and 2 days after seeding on 96-well plates. (c) Effect of MIA on cell invasion. MKN-28 cells transfected with pcDNA-MIA or pcDNA 3.1 constructs were incubated in Boyden chambers. After 1 and 2 days, invading cells were counted. Bars and error bars, mean and s.d. of three different experiments. On day 2, the clones were up to three-fold more invasive than the empty vector-transfected cells. NS=not significant.

Discussion

Several tumor (breast cancer, lung cancer, and renal cell cancer)-specific genes have been identified by a combination of subtractive hybridization and cDNA microarray technology (Amatschek et al., 2004). In this study, we searched for GC-specific genes by SAGE data analysis and quantitative RT–PCR. True cancer-specific genes were not found, but APIN, TRAG3, CYP2W1, MIA, MMP-10, DKK4, GW112, REGIV, and HORMAD1 were expressed much more highly in GC than in 14 types of normal tissues. As these genes were identified by SAGE and quantitative RT–PCR analysis of bulk GC tissues, immunohistochemistry was required to determine which cells expressed these genes. Antibodies against MIA, MMP-10, and DKK4 were available, and staining for all three proteins was confirmed in GC cells. But, DKK4 expression was present in only two out of 151 GC cases and generally absent.

MIA was first isolated as an 11-kDa protein secreted by malignant melanoma cell lines (Blesch et al., 1994). MIA is a potent inhibitor of proliferation of malignant melanoma cells and other neuroectodermal tumor cells (Blesch et al., 1994). Overexpression of MIA has been reported in breast cancer (Bosserhoff et al., 1999), colorectal cancer (Hau et al., 2002), glioma (Hau et al., 2004), and pancreatic cancer (El Fitori et al., 2005), and association of MIA with tumor progression has been reported. We found that expression of MIA was correlated with T grade, N grade, tumor stage, and patient prognosis, indicating that it plays an important role in GC progression and may serve as a good marker of GC progression. In contrast to malignant melanoma, high MIA levels were detected in only four of 17 serum samples from patients with stage IV GC, although overexpression of MIA was frequently found in primary GC tissues by immunohistochemistry. Therefore, MIA is not a suitable serum marker for early detection for GC, but it is a good indicator of a poor prognosis. We cannot explain the discrepancy between MIA expression level in primary GC and in serum. We did confirm that MIA was present in the culture medium of GC cells stably transfected with MIA, suggesting that MIA is secreted by GC cells. The origin of MIA in serum samples may be circulating GC cells but not primary GC cells. Levels of MIA in serum from patients with pancreatic cancer are reported to be low, despite MIA mRNA and protein overexpression in pancreatic cancer tissues (El Fitori et al., 2005).

In addition to the usefulness of MIA as an indicator of poor prognosis, it has been reported that MIA enhances migration and invasion ability and inhibits apoptosis of melanocytic cells (Bosserhoff et al., 2001; Poser et al., 2004). Here, we showed that transfection of MIA enhanced invasive activity of MKN-28 cells. As expression of MIA was highly specific to cancer cells, it may be a good therapeutic target with less adverse effects for various types of cancers, including GC.

Among the nine genes overexpressed in GC, MMP-10 (also known as stromelysin 2) was frequently overexpressed in GC. MMPs induce extracellular matrix breakdown associated with normal tissue remodeling and are associated with tissue destruction in arthritis, cancer invasion, and metastasis (Nelson et al., 2000; Visse and Nagase, 2003). Overexpression of MMP-10 has been reported in cancers of the lung, head, and neck (Muller et al., 1991), esophagus (Mathew et al., 2002; Sharma et al., 2004), brain (Thorns et al., 2003), and liver (Bodey et al., 2000). The present immunohistochemical study showed MMP-10 to be correlated with a poor prognosis in patients with GC. Importantly, high levels of MMP-10 protein were detected in serum samples from most of the patients with GC, even at stage I. An available tumor marker for GC is carcinoembryonic antigen (CEA) (Molnar et al., 1976). Despite the reliability of CEA as a marker for detection of GC, the preoperative rate of serum CEA positivity in GC is 20–40% (Koga et al., 1987; Shimizu et al., 1987). In the present study, of 36 serum samples from patients with stage I GC, 32 (88.9%) showed high levels of MMP-10 protein, indicating that MMP-10 is a serum tumor marker with high sensitivity. In contrast, immunohistochemical staining of MMP-10 was detected in 35 (42.2%) of 83 stage I GC samples. This discrepancy between immunohistochemical and ELISA results may be due to methodological differences. MMP-10 immunohistochemical results were evaluated with reference to the percentage of stained cancer cells; the intensity of immunostaining was not evaluated because of lack of internal control in immunohistochemistry. More detailed quantitative methods for the measurement of MMP-10 protein will be necessary to determine the relation between MMP-10 protein levels in serum samples and primary GC samples. In addition, because high levels of MMP-10 protein were detected in nine (15.0%) of 60 serum samples from healthy individuals, characterization of these individuals is necessary to determine if MMP-10 is a valid serum tumor marker. At least, gastritis may not be a reason for high levels of MMP-10 protein because it was not detected in 15 of the 20 serum samples of patients with gastritis, and only three samples showed low levels (200–400 pg/ml). Overexpression of MMP-10 was reported in human diabetic corneas (Saghizadeh et al., 2001), but no expression was observed in synovial samples from patients with rheumatoid arthritis (Hembry et al., 1995).

Clinical trials using MMP inhibitors as cancer therapeutics have been reported (Rudek et al., 2001; Rizvi et al., 2004). MMP inhibitors inhibit a broad spectrum of MMPs, and several adverse effects have been reported. MMP-10 activates proMMP-7 and proMMP-9 (Nakamura et al., 1998), which are thought to be particularly important for the malignant behavior of GC cells (Nomura et al., 1996; Yamashita et al., 1998). As expression of MMP-10 was narrowly restricted in cancer, MMP-10-specific inhibitors may provide antitumor drugs with less adverse effects for the treatment of various types of cancers, including GC.

Of two peritoneal wash cytology-positive samples, one showed a high MIA level. The levels of MMP-10 in peritoneal wash fluid were higher in two cytology-positive samples than those in cytology-negative samples. Peritoneal wash cytology is a frequently performed and important technique in the diagnosis of peritoneal dissemination. However, it often fails to detect malignant cells. It has been reported that peritoneal wash cytological examination is the most significant factor predicting peritoneal recurrence, with a sensitivity of 56% (Bando et al., 1999). As pellets from peritoneal wash fluid are used for cytology, it may be useful to investigate the levels of MIA or MMP-10 in the supernatants of cytology-negative peritoneal wash fluid to detect micrometastases of GC cells in the peritoneal cavity. As we studied here a small number of peritoneal wash fluids, additional investigation with more samples will clarify whether measurement of the levels of MIA or MMP-10 is useful to detect occult cancer cells in the supernatants of cytology-negative peritoneal wash fluid.

DKK4 immunostaining was identified in only 1.3% of GC samples, whereas overexpression of DKK4 mRNA was observed in 25.0% of GC samples by quantitative RT–PCR. As bulk GC tissues were used for quantitative RT–PCR analysis, the resulting data may not reflect the expression levels of DKK4 in cancer cells alone. DKK4-positive cells were identified in the intestinal metaplasia of the stomach by immunostaining. Expression of DKK4 mRNA observed in quantitative RT–PCR may be due to non-neoplastic tissue contamination, such as intestinal metaplasia. DKK4 suppresses the Wnt signaling pathway (Mao and Niehrs, 2003), which is thought to participate in carcinogenesis (Beachy et al., 2004). Thus, DKK4 may not be a good therapeutic target, at least for GC.

In addition to MIA, MMP-10, and DKK4, APin protein, GW112, and Reg IV are reported to be secretory proteins (Clark et al., 2003; Solomon et al., 2003). These secreted molecules may constitute good serum tumor markers. Creation of antibodies and immunohistochemical and functional analyses of APin protein, GW112, and Reg IV should be performed. In addition, recent data have shown the antiapoptotic activity of GW112 (Zhang et al., 2004).

Although we identified several genes that were overexpressed in GC SAGE libraries, there were many genes that were overexpressed according to SAGE but not quantitative RT–PCR. The inconsistent results between SAGE and quantitative RT–PCR may be due to the small number of SAGE libraries. Among 54 candidates, the expression of many genes was highest in the normal pancreas or duodenum by quantitative RT–PCR. To further identify cancer-specific genes, detailed SAGE libraries of normal tissues, such as pancreas and duodenum, are needed.

In conclusion, our present study yielded a list of genes that are potential tumor markers of GC. We identified MMP-10 as a serum tumor marker for diagnosis of GC and identified MIA and MMP-10 as prognostic indicators of GC. We identified several genes by quantitative RT–PCR that have not previously been implicated in GC. Although the functions of these genes, APin, GW112, and Reg IV, are not well understood in cancer, they may provide novel therapeutic targets for GC. Our current data also provide information with respect to the expression of these genes throughout the body. As both MIA and MMP-10 play important roles in tumor cell invasion, specific inhibitors against MIA or MMP-10 may constitute good anticancer drugs with less adverse effect. As the number of samples from normal organs studied here was small, additional examination will certify the specificity of nine GC-specific genes identified in this study.

Materials and methods

Tissue samples

In all, 195 primary tumors, 69 serum samples, and 29 peritoneal wash samples were collected from patients diagnosed with GC. Patients were treated at the Hiroshima University Hospital or an affiliated hospital.

For quantitative RT–PCR, 44 GC samples and corresponding non-neoplastic mucosa samples were used. The samples were obtained during surgery at the Hiroshima University Hospital or an affiliated hospital. We confirmed microscopically that the tumor specimens were predominantly (>80%) cancer tissue. Samples were frozen immediately in liquid nitrogen and stored at −80°C until use. Noncancerous samples of heart, lung, stomach, small intestine, colon, liver, pancreas, kidney, bone marrow, peripheral leukocytes, spleen, skeletal muscle, brain, and spinal cord were purchased from Clontech (Palo Alto, CA, USA).

For immunohistochemical analysis, we used archival formalin-fixed, paraffin-embedded tissues from 151 patients who had undergone surgical excision for GC. Of the 151 patients , 59 had early GC and 92 had advanced GC. Early GC is limited to the mucosa, or the mucosa and submucosa, regardless of nodal status. Advanced GC is a tumor whose invasion is beyond muscularis propria (Hohenberger and Gretschel, 2003). Information on patient prognosis was available for 58 of the 92 advanced GC cases.

Among 151 GC cases used for immunohistochemical analysis, serum samples were available for ELISA from 69 GC cases (44 men and 25 women; age range, 35–88 years; mean, 68.7 years). Serum samples were collected presurgically, before initiation of therapy, and stored at −80°C until analysis. Serum samples from 20 patients with chronic-active gastritis with Helicobacter pylori infection (13 men and seven women; age range, 57–85 years; mean, 68.8 years) were also collected. The presence of H. pylori was determined with the following tests: the H&E staining and Giemsa staining of biopsied tissues, rapid urease test, and urea breath test. Control serum samples were obtained from 60 healthy individuals (32 men and 28 women; age range, 30–83 years; mean, 63.4 years). Peritoneal wash samples from 29 patients with GC were obtained at the time of surgery at the Hiroshima University Hospital or an affiliated hospital.

Histologic classification (intestinal-type or diffuse-type) was according to the Lauren classification system (Lauren, 1965). Tumor staging was according to the TNM staging system (Sobin and Wittekind, 2002). As written informed consent was not obtained, for strict privacy protection, identifying information for all samples was removed before analysis; this procedure was in accordance with the Ethical Guidelines for Human Genome/Gene Research of the Japanese Government.

Identification of candidate GC-specific genes

To identify GC-specific genes, we used four GC SAGE libraries (W226T, GSM8867; W246T, GSM8505; S219T, GSM7800; P208T, GSM9103) published by us and 14 normal SAGE libraries (white matter, GSM676; cerebellum, GSM695; thalamus, GSM713; heart, GSM1499; lung, GSM762; stomach, GSM784; colon, GSM728; liver, GSM785; kidney, GSM708; leukocyte, GSM709; peritoneum, GSM738; skeletal muscle, GSM819; spinal cord, GSM2386; lymph node, GSM14785) available from the SAGEmap online database (http://www.ncbi.nlm.nih.gov/SAGE/; Lal et al., 1999). We compared tags from each GC SAGE library with those of normal SAGE libraries. To exclude tags generated by sequencing errors, we selected only tags that occurred at least three times in each GC SAGE library. In addition, we selected tags that were not found in any of the 14 normal SAGE libraries.

Evaluation of the specificity of gene expression

To evaluate the specificity of expression of each gene, a specificity index was calculated as follows: first, we identified the normal tissue in which the target gene expression was the highest of the 14 normal tissues analysed by quantitative RT–PCR (the mRNA expression level in this tissue was denoted as A). We then identified the GC among the nine GC samples in which the target gene expression was highest by quantitative RT–PCR (the mRNA expression level in this tissue was denoted as B). The ratio B/A was defined as the specificity index. When the specificity index of the target gene was >10, the gene was considered to show high specificity for GC. When the specificity index of the target gene was <10 and >2, the gene was considered to show low specificity for GC. When the specificity index of the target gene was <2, the gene was considered to show no specificity for GC.

Quantitative RT–PCR analysis

Total RNA was extracted with an RNeasy Mini Kit (Qiagen, Valencia, CA, USA), and 1 μg of total RNA was converted to cDNA with an First Strand cDNA Synthesis Kit (Amersham Biosciences Corp., Piscataway, NJ, USA). PCR was performed with a SYBR Green PCR Core Reagents Kit (Applied Biosystems, Foster City, CA, USA). Real-time detection of the emission intensity of SYBR green bound to double-stranded DNA was performed with an ABI PRISM 7700 Sequence Detection System (Applied Biosystems) as described previously (Kondo et al., 2004). ACTB-specific PCR products were amplified from the same RNA samples and served as internal controls (Kondo et al., 2004). Primer sequences and additional PCR conditions are available upon request.

Immunohistochemistry

A Dako LSAB Kit (Dako, Carpinteria, CA, USA) was used for immunohistochemical analysis. In brief, microwave pretreatment in citrate buffer was performed for 15 min to retrieve antigenicity. After peroxidase activity was blocked with 3% H2O2–methanol for 10 min, sections were incubated with normal goat serum (Dako) for 20 min to block nonspecific antibody-binding sites. Sections were incubated with the following antibody dilutions: mouse monoclonal anti-MIA (2F7), 1:50; mouse monoclonal anti-MMP-10, 1:100 (Novocastra, Newcastle, UK), and mouse monoclonal anti-DKK4, 1:100 (R&D Systems, Abingdon, UK). The specificity of the MIA antibody has been characterized in detail (Bosserhoff et al., 1999). Specificity of DKK4 staining was confirmed by preabsorption of the anti-DKK4 antibody with excess DKK4 protein (R&D Systems). Sections were incubated with primary antibody for 8 h at 4 C, followed by incubations with biotinylated anti-mouse IgG and peroxidase-labeled streptavidin for 10 min each. Staining was completed with a 10-min incubation with the substrate-chromogen solution. The sections were counterstained with 0.1% hematoxylin. The percentage of stained cancer cells was evaluated for each antibody. A result was considered positive if at least 50% of the cells were stained. When fewer than 50% of cancer cells were stained, the immunostaining was considered negative.

ELISA

Serum and peritoneal wash fluid levels of MIA and MMP-10 were measured with an MIA ELISA Kit (Roche Diagnostics Co., Indianapolis, IN, USA) and a Quantikine Human MMP-10 Immunoassay Kit (R&D Systems) according to the manufacturer's instructions.

Cell lines, expression vector, and stable transfection

A human GC-derived cell line, MKN-28, was kindly provided by Dr Toshimitsu Suzuki. MKN-28 cells were maintained in RPMI 1640 medium (Nissui Pharmaceutical Co., Ltd, Tokyo, Japan) containing 10% fetal bovine serum (BioWhittaker, Walkersville, MA, USA) in a humidified atmosphere of 5% CO2 and 95% air at 37°C.

For constitutive expression of the MIA gene, cDNA was PCR amplified and subcloned into pcDNA 3.1 (Invitrogen Corp., Carlsbad, CA, USA). The pcDNA-MIA expression vector was transfected into MKN-28 cells with FuGENE6 (Roche Diagnostics) according to the manufacturer's instructions. Stable transfectants were selected after 2 weeks of culture with 80 μg/ml G418 (Invitrogen). The amount of secreted MIA protein in cell culture supernatants was determined by MIA ELISA.

Cell growth and in vitro invasion assays

Cultured cells were harvested from 80% confluent monolayer cultures by brief treatment with 0.1% trypsin and 0.l% EDTA. The cells were seeded at a density of 2000 cells per well in 96-well plates. Cell growth was monitored after 1 and 2 days by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (Alley et al., 1988).

Modified Boyden chamber assays were performed to examine invasiveness. Stably transfected cells were plated at 10 000 cells per well in RPMI 1640 medium plus 1% serum in the upper chamber of a Transwell insert (8 μm pore diameter; Chemicon, Temecula, CA, USA) coated with Matrigel. Medium containing 10% serum was added in the bottom chamber. After 1 and 2 days, cells in the upper chamber were removed by scraping, and the cells remaining on the lower surface of the insert were stained with CyQuant GR dye to assess the number of cells.

Statistical methods

Statistical analyses were carried out with Fisher's exact test and the Mann–Whitney U-test. Kaplan–Meier survival curves were constructed for MIA or MMP-10-positive and MIA or MMP-10-negative patients. Survival rates were compared between MIA or MMP-10-positive and MIA or MMP-10-negative groups. Differences between survival curves were tested for statistical significance by log-rank test (Mantel, 1966). P-value less than 0.05 was considered statistically significant.

Accession codes

Accessions

GenBank/EMBL/DDBJ

References

  1. Alley MC, Scudiero DA, Monks A, Hursey ML, Czerwinski MJ, Fine DL et al. (1988). Cancer Res 48: 589–601.

  2. Amatschek S, Koenig U, Auer H, Steinlein P, Pacher M, Gruenfelder A et al. (2004). Cancer Res 64: 844–856.

  3. Bando E, Yonemura Y, Takeshita Y, Taniguchi K, Yasui T, Yoshimitsu Y et al. (1999). Am J Surg 178: 256–262.

  4. Beachy PA, Karhadkar SS, Berman DM . (2004). Nature 432: 324–331.

  5. Blesch A, Bosserhoff AK, Apfel R, Behl C, Hessdoerfer B, Schmitt A et al. (1994). Cancer Res 54: 5695–5701.

  6. Bodey B, Bodey Jr B, Siegel SE, Kaiser HE . (2000). Anticancer Res 20: 4585–4590.

  7. Bosserhoff AK, Echtenacher B, Hein R, Buettner R . (2001). Melanoma Res 11: 417–421.

  8. Bosserhoff AK, Kaufmann M, Kaluza B, Bartke I, Zirngibl H, Hein R et al. (1997). Cancer Res 57: 3149–3153.

  9. Bosserhoff AK, Moser M, Hein R, Landthaler M, Buettner R . (1999). J Pathol 187: 446–454.

  10. Buckhaults P, Rago C, St Croix B, Romans KE, Saha S, Zhang L et al. (2001). Cancer Res 61: 6996–7001.

  11. Clark HF, Gurney AL, Abaya E, Baker K, Baldwin D, Brush J et al. (2003). Genome Res 13: 2265–2270.

  12. Dietz UH, Sandell LJ . (1996). J Biol Chem 271: 3311–3316.

  13. El Fitori J, Kleeff J, Giese NA, Guweidhi A, Bosserhoff AK, Buchler MW et al. (2005). Cancer Cell Int 5: 3.

  14. Hau P, Apfel R, Wiese P, Tschertner I, Blesch A, Bogdahn U . (2002). Anticancer Res 22: 577–583.

  15. Hau P, Ruemmele P, Kunz-Schughart LA, Doerfelt A, Hirschmann B, Lohmeier A et al. (2004). Oncol Rep 12: 1355–1364.

  16. Hembry RM, Bagga MR, Reynolds JJ, Hamblen DL . (1995). Ann Rheum Dis 54: 25–32.

  17. Hohenberger P, Gretschel S . (2003). Lancet 362: 305–315.

  18. Kevorkian L, Young DA, Darrah C, Donell ST, Shepstone L, Porter S et al. (2004). Arthritis Rheum 50: 131–141.

  19. Koga T, Kano T, Souda K, Oka N, Inokuchi K . (1987). Jpn J Surg 17: 342–347.

  20. Kondo T, Oue N, Yoshida K, Mitani Y, Naka K, Nakayama H et al. (2004). Cancer Res 64: 523–529.

  21. Krupnik VE, Sharp JD, Jiang C, Robison K, Chickering TW, Amaravadi L et al. (1999). Gene 238: 301–313.

  22. Lal A, Lash AE, Altschul SF, Velculescu V, Zhang L, McLendon RE et al. (1999). Cancer Res 59: 5403–5407.

  23. Lauren P . (1965). Acta Pathol Microbiol Scand 64: 31–49.

  24. Mantel N . (1966). Cancer Chemother Rep 50: 163–170.

  25. Mao B, Niehrs C . (2003). Gene 302: 179–183.

  26. Mathew R, Khanna R, Kumar R, Mathur M, Shukla NK, Ralhan R . (2002). Cancer Detect Prev 26: 222–228.

  27. Molnar IG, Vandevoorde JP, Gitnick GL . (1976). Gastroenterology 70: 513–515.

  28. Muller D, Breathnach R, Engelmann A, Millon R, Bronner G, Flesch H et al. (1991). Int J Cancer 48: 550–556.

  29. Nakamura H, Fujii Y, Ohuchi E, Yamamoto E, Okada Y . (1998). Eur J Biochem 253: 67–75.

  30. Nelson AR, Fingleton B, Rothenberg ML, Matrisian LM . (2000). J Clin Oncol 18: 1135–1149.

  31. Nomura H, Fujimoto N, Seiki M, Mai M, Okada Y . (1996). Int J Cancer 69: 9–16.

  32. Ohgaki H, Matsukura N . (2003). Stomach cancer. In: Stewart BW, Kleihues P (eds). World Cancer Report. IARC Press: Lyon, p 197.

    Google Scholar 

  33. Oue N, Hamai Y, Mitani Y, Matsumura S, Oshimo Y, Aung PP et al. (2004). Cancer Res 64: 2397–2405.

  34. Oue N, Mitani Y, Aung PP, Sakakura C, Takeshima Y, Kaneko M et al. (2005). J Pathol 207: 185–198.

  35. Poser I, Tatzel J, Kuphal S, Bosserhoff AK . (2004). Oncogene 23: 6115–6124.

  36. Ramos MC, Steinbrenner H, Stuhlmann D, Sies H, Brenneisen P . (2004). Biol Chem 385: 75–86.

  37. Rizvi NA, Humphrey JS, Ness EA, Johnson MD, Gupta E, Williams K et al. (2004). Clin Cancer Res 10: 1963–1970.

  38. Rudek MA, Figg WD, Dyer V, Dahut W, Turner ML, Steinberg SM et al. (2001). J Clin Oncol 19: 584–592.

  39. Saghizadeh M, Brown DJ, Castellon R, Chwa M, Huang GH, Ljubimova JY et al. (2001). Am J Pathol 158: 723–734.

  40. Sharma R, Chattopadhyay TK, Mathur M, Ralhan R . (2004). Oncology 67: 300–309.

  41. Shimizu N, Wakatsuki T, Murakami A, Yoshioka H, Hamazoe R, Kanayama H et al. (1987). Oncology 44: 240–244.

  42. Sobin LH, Wittekind CH (ed). (2002). TNM Classification of Malignant Tumors, 6th edn. Wiley-Liss, Inc.: New York, pp 65–68.

    Google Scholar 

  43. Solomon A, Murphy CL, Weaver K, Weiss DT, Hrncic R, Eulitz M et al. (2003). J Lab Clin Med 142: 348–355.

  44. Thorns V, Walter GF, Thorns C . (2003). Anticancer Res 23: 3937–3944.

  45. Ushijima T, Sasako M . (2004). Cancer Cell 5: 121–125.

  46. Visse R, Nagase H . (2003). Circ Res 92: 827–839.

  47. Yamashita K, Azumano I, Mai M, Okada Y . (1998). Int J Cancer 79: 187–194.

  48. Yasui W, Oue N, Ito R, Kuraoka K, Nakayama H . (2004). Cancer Sci 95: 385–392.

  49. Yasui W, Yokozaki H, Fujimoto J, Naka K, Kuniyasu H, Tahara E . (2000). J Gastroenterol 35: 111–115.

  50. Zhang X, Huang Q, Yang Z, Li Y, Li CY . (2004). Cancer Res 64: 2474–2481.

  51. Zucker S, Hymowitz M, Conner C, Zarrabi HM, Hurewitz AN, Matrisian L et al. (1999). Ann N Y Acad Sci 878: 212–227.

Download references

Acknowledgements

We thank M Takatani for excellent technical assistance and advice. This work was carried out with the kind cooperation of the Research Center for Molecular Medicine, Faculty of Medicine, Hiroshima University. We thank the Analysis Center of Life Science, Hiroshima University for the use of their facilities. This work was supported, in part, by Grants-in-Aid for Cancer Research from the Ministry of Education, Culture, Science, Sports, and Technology of Japan, and from the Ministry of Health, Labor, and Welfare of Japan.

Author information

Affiliations

Authors

Corresponding author

Correspondence to W Yasui.

Additional information

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Aung, P., Oue, N., Mitani, Y. et al. Systematic search for gastric cancer-specific genes based on SAGE data: melanoma inhibitory activity and matrix metalloproteinase-10 are novel prognostic factors in patients with gastric cancer. Oncogene 25, 2546–2557 (2006). https://doi.org/10.1038/sj.onc.1209279

Download citation

Keywords

  • gastric cancer
  • MIA
  • MMP10
  • DKK4
  • SAGE
  • tumor serum marker

Further reading

Search

Quick links