Identification of Gastric Cancer–Related Genes Using a cDNA Microarray Containing Novel Expressed Sequence Tags Expressed in Gastric Cancer Cells

Purpose: Gastric cancer is one of the most frequently diagnosed malignancies in the world, especially in Korea and Japan. To understand the molecular mechanism associated with gastric carcinogenesis, we attempted to identify novel gastric cancer–related genes using a novel 2K cDNA microarray.

Experimental Design: A 2K cDNA microarray was fabricated from 1,995 novel expressed sequence tags (ESTs) showing no hits or a low homology with ESTs in public databases from our 143,452 ESTs collected from gastric cancer cell lines and tissues. An analysis of the gene expression for human gastric cancer cell lines to a normal cell line was done using this cDNA microarray. Data for the different expressed genes were verified using semiquantitative reverse transcription-PCR, Western blotting, and immunohistochemical staining in the gastric cell lines and tissues.

Results: Forty genes were identified as either up-regulated or down-regulated genes in human gastric cancer cells. Among these, genes such as SKB1, NT5C3, ZNF9, p30, CDC20, and FEN1, were confirmed to be up-regulated genes in nine gastric cell lines and in 25 pairs of tissue samples from patients by semiquantitative reverse transcription-PCR. On the other hand, genes such as MT2A and CXX1 were identified as down-regulated genes. In particular, the SKB1, CDC20, and FEN1 genes were overexpressed in ≥68% of tissues and the MT2A gene was down-expressed in 72% of the tissues. Western blotting and immunohistochemical analyses for CDC20 and SKB1 showed overexpression and localization changes of the corresponding protein in human gastric cancer tissues.

Conclusions: Novel genes that are related to human gastric cancer were identified using cDNA microarray developed in our laboratory. In particular, CDC20 and MT2A represent a potential biomarker of human gastric cancer. These newly identified genes should provide a valuable resource for understanding the molecular mechanism associated with tumorigenesis of gastric carcinogenesis and for the discovery of potential diagnostic markers of gastric cancer.

Gastric cancer is one of the most frequently diagnosed malignancies in the world (1). It is particularly prevalent in Korea and Japan and is one of the leading causes of cancer death in these regions (2). Although the incidence and mortality have been decreasing during the last several years, gastric cancer still has a notorious position, with the first incidence and the second cause of mortality in Korea (3).

Advances in diagnostic and treatment technologies have enabled us to offer excellent long-term survival results for early gastric cancer, but the prognosis of advanced gastric cancer still remains poor (4). Recent molecular analyses revealed that gastric cancers are closely related to genetic alterations in several genes, such as p53, APC, E-cadherin, β-catenin, TGF-α, c-met, trefoil factor 1, and Runx3 (5–7). However, the common pathways of carcinogenesis and the subsequent progression of gastric cancer remained to be elucidated.

A cDNA microarray was used to simultaneously study the expression profiles of a number of genes at specific conditions in a single hybridization (8, 9). Many reports on gene expression profiles of various cancers and diseases using cDNA microarray techniques have been reported (10–14). Among them, changes in gene expression in gastric cancer cell lines and malignant tissues have been reported. In gastric adenocarcinomas, genes such as S100A4, CDK4, MMP1, and β-catenin genes have been reported as being up-regulated genes, the GIF gene was reported to be a down- regulated gene (15). Ji et al. (16) has also reported the first comprehensive review of gene expression patterns in gastric cancer cell lines on a genomic scale. In this study, they analyzed global gene expression patterns of 27 human cell lines, including 12 gastric carcinoma cell lines and compared heterogeneity between gastric cancer cell lines. In addition, a comparison of the gastric cancer–related genes using gastric cancer tissues and surrounding gastric mucosa tissues has been reported, as well as a connection between the clinical phenotypes of patients (17).

In a previous study, we collected an entire set of genes that are expressed in gastric cancer cell lines or tissues using full-length enriched cDNA libraries, subtracted cDNA libraries, and normalized cDNA libraries from gastric cancer cell lines and tissues from Korean patients and identified the genes associated with gastric cancer by examining their expression profiles (18). Using this process for identifying novel gastric cancer-related genes in which there were no hits or a low homology with known genes in public databases, we isolated 1,995 novel genes from the collected gastric expressed sequence tags (ESTs) and fabricated a cDNA microarray containing these genes. However, some of the ESTs were identified as known genes in recent updated public databases. Using the cDNA microarray, a gene expression analysis of these genes in gastric cancer cell lines and tissues was done. Here, we report on the identification of novel genes that are differentially expressed in gastric cancer cell lines and tissues.

Human gastric cancer cell lines, SNU-1, SNU-16, SNU-216, SNU-484, SNU-601, SNU-638, SNU-668, and SNU-719 were cultured in RPMI 1640 (Life Technologies, Grand Island, NY) and human normal gastric cell lines Hs 677.St (ATCC CRL-7407) in DMEM (Life Technologies) supplemented with 10% inactivated fetal bovine serum, 2 mg/mL sodium bicarbonate, and 1% antibiotic-antimycotic solution (Invitrogen Life Technologies, Carlsbad, CA). The Hs 677.St cell line was derived from normal fetal stomach tissue and had a morphology similar to a fibroblast. All cultured cells were incubated at 37°C in a humidified incubator maintained with a 5% CO₂ atmosphere (19, 20). When the cells were about 80% to 90% confluent, they were harvested and used for total RNA isolation. Fifty gastric tissues containing the tumor and normal regions of 25 gastric cancer patients were obtained from the College of Medicine, Chungnam National University, Korea with informed consent. The tumors were staged according to tumor-node-metastasis classification of Union Internationale Contre le Cancer. The obtained tissues were immediately frozen in liquid nitrogen. Total RNA was extracted from the cultured cells and tissues using a commercially available RNA isolation kit (Qiagen, Hilden, Germany) following the procedures recommended by the manufacturer.

The total 143,452 ESTs collected from human gastric cancer cell lines and gastric tissues were analyzed by a BLAST search against human mRNA (Genbank release 126, downloaded on Oct. 2001), UniGene (UniGene build 143, downloaded on Oct. 2001) and NR databases (downloaded on Oct. 2001). To isolate novel ESTs in which there were no hits or a low homology in public databases, ESTs having an identity of <90% for <50 bp with E ≤ 1 × 10⁻³ against the human mRNA and UniGene databases, and having an identity of <85% for <20 amino acids with E ≤ 1 × 10⁻⁵ against the NR database were selected. E ≤ 1 × 10⁻³ indicates that the probability that a query sequence have accidentally identity with a certain sequences in database under given condition is ≤1 × 10⁻³. These isolated ESTs were used to fabrication the cDNA microarray.

The novelty of these ESTs were reanalyzed by a BLAST search against human mRNA (Genbank release 138.0, downloaded on Dec. 2003), RefSeq (downloaded on Dec. 2003) under conditions of an identity of >90% for >50 bp with E ≤ 1 × 10⁻²⁰. The remaining ESTs were analyzed by a BLAT search against the human genome database (University of California Santa Cruz⁶

Golden Path genome database build 15) under the above conditions. Analysis of the ESTs that were not included in the above searches were done under conditions of an identity of >90% for >50 bp with E = 1 × 10⁻²⁰ to 1 × 10⁻³ against human mRNA and RefSeq databases and with E ≤ 1 × 10⁻¹ against the NR database (downloaded on Dec. 2003).

Clones containing the novel ESTs were grown in 96-well culture plates and plasmid DNAs were purified using a Millipore plasmid kit (Millipore Co., Bedford, MA). The inserts of cDNAs using purified plasmid DNAs were amplified by PCR with the sense primer 5′-GCAGAGCTCTCTGGCTAAC-3′, which is localized in the vector region and the antisense primer 5′-CGTGCGGCCGCT₂₁(G/A/C)-3′. After purifying the PCR products on Sephadex G-50 Superfine (Amersham Pharmacia Biotech AB, Uppsala, Sweden), they were suspended in a Microspotting solution (ArrayItTM Brand Products, TeleChem, Sunnyvale, CA) and spotted on CSS-100 Silyated Slides (Aldehyde; CEL Associates, Pearland, TX) using a Cartesian Prosys 5510 robot (Cartesian, Inc., Irvine, CA) with 32 printing tips. Our cDNA microarray contained a total of 6,912 spots in one slide including triplicates of 1,995 cDNA, control genes of GAPDH and β-actin, and empty spots for negative controls.

Twenty micrograms of total RNA from a normal cell line or cancer cell lines, respectively, were used in the cDNA microarray analysis. RNA of the normal cell line, labeled with Cy3, was used as a reference versus RNA with Cy5 from each of eight cancer cell lines as a sample. Probe labeling and hybridization were done using a 3DNA Array 50 kit (Genisphere, Inc., Hatfield, PA) according to the manufacturer's instructions. After the hybridization procedure, the slide was scanned at a wavelength of 532 nm for Cy3 and at 635 nm for Cy5 using a ScanArray 5000 scanner (Packard BioChip Technologies, Billerica, MA). To increase the accuracy of the experiment, each experiment was done in duplicate using two different cDNA microarrays.

The scanned images were analyzed using the GenePix Pro 4.0 program (Axon Instruments, Inc., Union City, CA) and the subsequent data were normalized using the scaled print-tip group Lowess method using the statistics for microarray analysis package of the R⁷

statistics software to remove intensity variances between spots themselves that originate from spotted locations. If the signal to background ratio was <1.4, the feature was processed as a null value to reduce bias. Using normalized M values [M = log₂(R/G)], we did a one class analysis using the significance analysis of microarrays⁸ program with a median false discovery rate of 0.10089 and Δ = 1.40 to select significantly expressed genes (21). Furthermore, to exclude spots having a low intensity, genes having an A > 6 [A = 0.5log₂(RG)] were selected. In addition, redundant clones were removed, because our cDNA microarray had triplicate spots. Finally, differentially expressed genes in gastric cancer cell lines were selected for further study based on the significance analysis of microarrays scores.

A homology search for the selected genes was done by a BLASTn analysis against the NR database with the National Center for Biotechnology Information⁹

default conditions. The search for the symbol and function of these genes were done by SOURCE¹⁰ and GeneCards.¹¹ In addition, an analysis for the chromosomal location of the selected genes was done using the University of California Santa Cruz Golden Path human genome database build 15 at conditions of 90% minimum identity.

The 1st cDNA was synthesized by the reverse transcription reaction with 5 μg of isolated RNA, 2 pmol/L of oligo (dT)₂₀, 1 μL of 10 μmol/L deoxynucleotide triphosphate, 4 μL of 5× buffer, 2 μL of 100 mmol/L DTT, 1 μL of RNaseOUT (40 units/μL, Invitrogen Life Technologies), and 1 μL of SuperScript II (200 units/μL, Invitrogen Life Technologies) at 42°C for 1 hour. The 1st cDNA was quantified using a human β-actin competitive PCR kit (TaKaRa Co., Tokyo, Japan) according to the manufacturer's instructions. The PCR conditions were 1 cycle of 2 minutes at 94°C, 25 cycles of 30 seconds at 94°C, and 1 minute at 68°C, and 1 cycle of 1 minute at 72°C with β-actin primer sets (Table 1). After electrophoresis in a 2% agarose gel, the DNA concentration of β-actin (275 bp) and actin competitor (340 bp) were analyzed using the TotalLab software program (Phoretix Co., Newcastle Upon Tyne, United Kingdom) and the amount of the 1st cDNA of each sample was adjusted based on the β-actin concentration. To quantify the expression level of the selected genes, the same volume of diluted 1st cDNAs synthesized from gastric cells was used as a template in a PCR reaction. Each gene was amplified by PCR which consisted of 27 cycles of 40 seconds at 94°C, 50 seconds at 55°C, and 1 minute at 72°C with specific primer sets (Table 1). The PCR products of each of the specific genes and β-actin (275 bp) were analyzed by 2% agarose gel electrophoresis and the expression ratio was calculated using the TotalLab software program (Phoretix).

When human gastric normal and cancer cell lines which were cultured in media, were about 80% to 90% confluent, they were rinsed with PBS, scrapped into 300 μL of cell lysis buffer containing 50 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 0.5% NP40, 1 mmol/L EDTA, 1 mmol/L phenylmethylsulfonyl fluoride, 1 μmol/L Pepstatin A, 1 μmol/L Leupeptin, 1 μmol/L Aprotinin, and placed on ice for 1 hour. The cells were then centrifuged at 15,000 × g for 15 minutes and the supernatant was harvested. Aliquots (50 μg) of soluble proteins were separated on SDS-polyacrylamide-gels and transferred to polyvinylidene difluoride membranes (Millipore). The membranes were incubated with the mouse monoclonal antibody against CDC20 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), a rabbit polyclonal antibody to SKB1 (Cell Signaling Technology, Inc., Beverly, MA) and a mouse monoclonal antibody to β-actin (Sigma, St. Louis, MO) at a dilution of 1:1,000, 1:1,000, and 1:50,000, respectively. After the blots were incubated with peroxidase-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch, WestGrove, PA) and horseradish peroxidase–conjugated goat anti-mouse antibody (Promega, Madison, WI), immunoreactive signals were detected using enhanced chemiluminescence kit (Amersham Pharmacia, Piscataway, NJ).

Paraffin sections of gastric cancer tissue from patients were deparaffinized with xylene and then rehydrated. Antigenic retrieval was processed by submerging in citrate buffer (pH 6.0) and microwaving. The sections were then treated with 3% hydrogen peroxide in methanol to quench endogenous peroxidase activity, followed by incubation with 1% bovine serum albumin to block nonspecific binding. The primary anti-CDC20 (1:100 dilution) and anti-SKB1 (1:100 dilution) antibodies that are used in Western blotting were incubated for 60 minutes at room temperature. After washing, the tissue section was then reacted with the biotinylated anti-mouse and anti-rabbit secondary antibodies, followed by incubation with streptavidin-horseradish-peroxidase complex. The tissue section was immersed in 3-amino-9-ethyl carbazole as a substrate, and counterstained with 10% Mayer's hematoxylin, dehydrated, and mounted by crystal mount. In the negative controls, the nonimmune mouse or rabbit IgG of the same isotype or the antibody dilution solution was replaced the primary antibody.

To isolate novel genes associated with stomach cancer, a cDNA microarray containing novel ESTs which have low homology or no hits in public databases was fabricated. A total of 1,995 ESTs contained in microarray were selected as novel ESTs from our 143,452 ESTs collected from human gastric cancer cell lines and gastric tissues by analysis of public databases (collected on Oct. 2001, see MATERIALS AND METHODS). A reanalysis of these selected genes against updated above databases, for novelty, (data collected on Dec. 2003) showed that 686 genes (34.4%) could be categorized into known human genes against the human mRNA and RefSeq databases with conditions of identity of >90% for >50 bp with E ≤ 1 × 10⁻²⁰, and 559 genes (28%) without known human genes were mapped only on the human genome against University of California Santa Cruz Golden Path genome database build 15 under above conditions (Table 2). In addition, 690 genes were categorized into ESTs of low homology and 60 genes (3.0%) showing “no hits”. Among known human genes, 289 genes (14.5%) were functionally classified by the Gene Ontology database.¹²

From these analyses, 1,309 ESTs excluding Known human genes were thought to be novel ESTs, although only 60 (3%) represented novel ESTs which were sacrificed by the first criteria, in which novel ESTs were defined as ESTs having an identity of <90% for <50 bp with E ≤ 1 × 10⁻³ against the human mRNA and UniGene databases.

We compared the gene expression profiles of eight gastric cancer cell lines with that of a normal gastric cell line using cDNA microarray. After cDNA microarray hybridization, normalization, and data analysis, we finally selected a total of 40 genes, 20 genes for up-regulation and 20 genes for down-regulation, based on significance analysis of microarrays scores, that showed significant expression changes in gastric cancer cells.

As shown in Table 3, the up-regulated genes in gastric cancer cells included known genes such as CKS1B, SCX, D1S155E, FKBP4, SKB1, NT5C3, p30, GPI, PRO2000, CDC20, FEN1, ZNF9, and RPS16 and functionally unknown genes such as FLJ31196, FLJ39478, and FLJ90345. In addition, four novel genes NSG-21-D10, NSG-18-A07, NSG-05-E12, and NSG-08-D09 were included. A search of the SOURCE and GeneCards database for the function of these selected genes indicated that their biological functions were diverse and included genes related to a cell cycle regulator (CKS1B, SKB1, and CDC20), transcription (SCX), development (D1S155E), protein folding (FKBP4), DNA repair (FEN1), and biosynthesis (ZNF9, RPS16). Furthermore, a search of chromosomal locations for the up-regulated genes was done by mapping the in University of California Santa Cruz Golden Path human genome database. The analysis showed that of the 20 genes, 13 were localized in chromosome 1 (CKS1B, D1S155E, and CDC20), chromosome 8 (SCX, FLJ39478, and PRO2000), chromosome 11 (FEN1 and NSG-18-A07), chromosome 17 (FLJ31196 and p30), and chromosome 19 (GPI, FLJ90345, and RPS16). The genes localized in chromosome 17 and chromosome 19 were clustered in 17p11.2 and 19q13, respectively.

Genes representing a down-regulated expression in gastric cancer cells included known genes such as LGALS1, OAZ1, PEA15, SEC61A1, LGP1, MT2A, MAGED2, NPDC1, CXX1, FKBP8, and PGR1 and functionally unknown genes such as DXS9879E, FLJ34386, FLJ20920, and FLJ30061 and five novel genes (Table 4). The functional analysis of these selected genes showed that the genes related with apoptosis (LGALS1), polyamine biosynthesis (OAZ1), protein targeting (SEC61A1), and protein folding (FKBP8) were included. In addition, many of the down-regulated genes were localized in chromosome 17, chromosome 19, and chromosome X. Among them, two genes LGP1 and FLJ20920 were clustered in 17q21.

To more quantitatively verify the data obtained from our DNA microarray, we randomly selected seven up-regulated genes (CKS1B, SKB1, NT5C3, ZNF9, p30, CDC20, and FEN1) and five down-regulated genes (LGALS1, OAZ1, DXS9879E, MT2A, and CXX1) in gastric cancer and did semiquantitative reverse transcription-PCR (RT-PCR) in nine normal and gastric cancer cell lines, and in 25 pairs of gastric normal and tumor tissues in the I to IV stages.

As shown in Fig. 1A, the expression of all the up-regulated genes were higher in most of the cancer cell lines than in normal cell lines, Hs 677.St. All of these genes were also highly expressed in most of the tumor tissues compared with their normal tissues (Fig. 1B). These genes were highly expressed in tumor tissues with a frequency of 40% to 88% in 25 tissue pairs that were classified as containing I to IV stages of gastric cancer. Among these genes, the CDC20 gene was the most highly expressed in 22 tumor tissues of the 25 tissue pairs with a high frequency of 88% which covered all stages of gastric cancer. The SKB1 and FEN1 genes were also detected at high levels in the IB and II stages of tumor tissues for SKB1 and in II and III A/B stages for FEN1 with a frequency of 68% (17 in 25 cases) and 72% (18 in 25 cases), respectively. All of the up-regulated genes detected in tissues were highly detected in most of the II stage gastric cancers. However, CKS1B was not detected in any of the gastric tissues, though it was detected in very low amounts in cancer cell lines. On the other hand, three down-regulated genes, except for OAZ1 and DXS9879E, were detected at low levels in many of the cancer cell lines compared with the normal cell line, Hs 677.St, as shown in Fig. 1A. When the expression levels of three genes, LGALS1, MT2A, and CXX1, were examined in gastric tissues, MT2A was found to be detected at low levels in the tumor tissues, but had high expression levels in normal tissues with a frequency of 72% (18 of 25 cases). Its higher expression was detected over a wide stage from IB to IV in normal gastric tissues. The other gene, CXX1, was highly expressed in normal tissues with frequencies of about 32% in various stages. However, LGALS1 was not detected in any of the gastric tissues, because of very low amounts in tissues. These results indicate that the mRNA levels of target genes in gastric tissues were largely consistent with those of the cell lines. Additionally, these results from semiquantitative RT-PCR are in relatively good agreement with the DNA microarray data.

We verified the protein levels for genes that had been confirmed by RT-PCR using Western blotting for nine gastric normal and cancer cell lines, and immunohistochemistry for six gastric tissues. Because antibodies for only CDC20 and SKB1 were available, these two proteins were selected as targets.

As shown in the Western blotting of Fig. 2A, high levels of protein for CDC20 were detected in the gastric cancer cell lines in comparison with the normal cell line, especially for SNU-601, SNU-638, and SNU-719. The immunohistochemistry also showed that CDC20 was highly detected in gastric tumor tissue, although it was present in normal tissue from the patient samples (Fig. 2B,, a-c). However, differently from normal tissue, it was localized in perinuclear region of the cell in tumor tissues and the localization change was more strongly detected in poorly differentiated gastric tumors. Otherwise, when the protein level for SKB1 was checked by Western blotting, it was also detected at high levels in gastric cancer cell lines. In particular, the amounts expressed were dramatically high in SNU-1, SNU-16, SNU-216, and SNU-638 (Fig. 2A). It is also noteworthy that a large band, higher than the 70 kDa band, corresponding to SKB1, was detected for SNU-216, which is thought to be the result of the post-modification of SKB1 or an alternative transcript. This band was also faintly detected in Hs 677.St. Figure 2B (d-f) shows immunohistochemical results for SKB1 in gastric normal and tumor tissues from the patient samples. As predicted, it was highly expressed in gastric tumors compared with normal tissue. As shown in CDC20, a change in localization for SKB1, mainly in the nuclear region, also detected in tumor tissues. These results indicate that an increase in the mRNA level for CDC20 and SKB1 in gastric tumor tissues coupled with that of the protein level and the change in the amount produced and their localization are associated with carcinogenesis in gastric tumors.

cDNA microarray technologies aid in analyses of the expression levels of several thousands of genes for multiple samples at the same time. Numerous attempts to identify genes related to carcinogenesis of various cancers including gastric cancer using a DNA microarray have been reported (10–17, 22). We selected ESTs having a low homology or no hits with ESTs in public databases from our Korean UniGene Information ESTs clone bank and used this as a DNA source for the fabrication of a microarray in order to identify novel genes that are associated with gastric cancer. All of the selected 1,995 ESTs were novel ESTs at the first stage. However, because a considerable amount of EST data has been recently submitted to public databases by rapid advances in high-throughput sequencing, of these ESTs, only 60 genes (3%), in a homology analysis against updated public databases represented novel ESTs which are sacrificed with the first criteria. However, as shown in Table 2, 1,309 ESTs excluding known human genes were classified as novel genes. When 2K microarray experiments using 1,995 cDNA were done, the signal intensities obtained were generally lower than those of a 14K cDNA microarray fabricated from our 143,452 ESTs (data not shown). In addition, the results of RT-PCR for the target genes indicated that the mRNA levels of many of the genes were very low or not detectable. These results indicate that the genes included in the 2K microarray were rarely expressed in cells and that the difference in expression of these genes also can be easily excluded, compared with those of abundantly expressed genes. Therefore, our 2K microarray might be potentially useful in identifying rare genes related to stomach cancer.

When the expression profiles of the gastric cancer cells and the normal cells were compared using our 2K microarray, 40 genes showing significant differences were found. Difference in the expression of these genes was also confirmed by semiquantitative RT-PCR data, collected from gastric cell lines and tissues from patients. Among the selected genes, several genes related to the cell cycle, CKS1B, CDC20, and SKB1, were identified as up-regulated genes. Interestingly, the CDC20 and SKB1 genes were highly represented with a very high frequency of 88% and 68% in gastric tumor tissues in comparison with normal tissues, although the CKS1B transcript was not detected in gastric tissues because of the low expression. Furthermore, a higher expression of two genes in gastric cancers was also detected by their protein level using Western blotting and immunohistochemistry. These results indicate that the up-regulation of two genes coupled transcription to translation. These results also showed changes in the localization of these proteins in tumor tissues, from the cytosol to the perinuclear region for CDC20 and to the nucleus for SKB1, respectively. These findings indicate that the amount of change of these genes that encoded transcript and protein as well as the change in localization is correlated with the oncogenesis of human gastric cancer.

CDC20 is known to directly bind to the anaphase-promoting complex with hCDH1 and activates anaphase-promoting complex by which anaphase is initiated and mitosis is terminated (23). The overexpression of CDC20 has previously been reported in human pancreatic cancer (24) and its alteration has also been detected in early-stage lung adenocarcinoma (25). The up-regulation of CDC20 in gastric cancer was confirmed by gene expression data linked to SOURCE in which CDC20 and CKS1B has been reported to be up-regulated in gastroesophageal adenocarcinomas (26). Meanwhile, the up-regulation of CDC20 has been reported to be related to apoptosis in Taxol-induced HeLa cells and NIH3T3 (27), myeloid cells (28, 29). Therefore, it is likely that function of the CDC20 in cells may depend on the stage, type and environments of the cells. CKS1B has been known to be a CDC28 protein kinase regulatory subunit 1B. The overexpression of CKS1B has been previously reported in gastric cancer (15, 22) and in pancreatic cancer (12). CKS1B has also been proposed to facilitate the transcription of the CDC20 gene through the remodeling of transcriptional complexes or chromatin that is associated with the CDC20 gene (30). These finding suggest that CDC20 and CKS1B may act sequentially in the tumorigenesis of gastric cancer, although it has not been reported that CDC20 is related to gastric cancer. It has previously been reported that SKB1 in fission yeast plays a role in the control of cell polarity (31), in the negative regulation of mitosis (32), and in the coordination of cell cycle progression (33). It has also been proposed to act as a mediator of the hyperosmotic stress response (34), but its relation to oncogenesis has not yet been reported.

Genes involving nucleotide metabolism, DNA repair, and cholesterol biosynthesis such as NT5C3, p30, FEN1, and ZNF9 are also up-regulated in gastric cancer cells, as evidenced from the microarray data as well as semiquantitative RT-PCR. In particular, FEN1 was highly expressed with a high frequency of 72% in gastric tumor tissues, compared with normal tissues. These observation are consistent with the finding that increased FEN1 expression leads to rapid tumor progression of mouse gastrointestinal tract cancer in a haplo-insufficient manner (35). The up-regulation of the gene has also been reported in human lung cancer cell lines (36). It has been reported that a deficiency in NT5C3 causes an autosomal recessive hemolytic anemia (37) and ZNF9 involve in myotonic dystrophy 2 (38). p30 has been identified as a component of a purified nucleoporin fraction from rat liver nuclei (39). Although these genes have not been reported to be related to human gastric cancer, they do, in fact, seem to be new candidates for gastric cancer, on based on the results herein, because the up-regulation of these genes was detected in gastric tumor tissues with a high frequency of 40% to 72%.

On the other hand, of the down-regulated genes in gastric cancer, MT2A was down-expressed with a high frequency of 72% in tumor tissues from the IB to IV stages. It is known to be involved in the regulation of carcinogenesis and apoptosis such as an activator of cell proliferation and an inhibitor of apoptosis, as well as various other physiologic processes (40, 41). Although this gene has been reported to be up-regulated in human breast cancers (42) and esophageal cancer (41), its expression is known to be down-regulated in gastroesophageal adenocarcinomas in gene expression data linked to SOURCE (26). Thus, it is likely that MT2A expression in tumor cells may depend on the developmental stage or the specific type of tumor. Genes such as LGALS1 and CXX1 were also down-regulated. LGALS1 is known to regulate cell apoptosis and to act as an autocrine-negative growth factor that regulates cell proliferation. Our data indicated that it represents a high priority candidate among the down-regulated genes in stomach cancer cell lines in comparison with normal cell lines, although it was not detected in stomach tissue because of its low abundance. However, contrary to our data, the up-regulation of LGALS1 has been reported in several tumors such as head and neck squamous carcinoma (43), human colon cancer (44, 45), and human pancreatic cancer (46). These observations imply that the mechanism of LGALS1 in human gastric cancer might be different from that reported for other cancers. Reports concerning CXX1 being down-regulated in tumor tissues with a frequency of about 32%, except having a CAAX box 1 have not yet appeared.

Some tumor suppressor genes and oncogenes under the control of genomic change were clustered in specific chromosomal regions. The data herein indicate that some of the up-regulated genes were clustered in chromosome 17p11.2 and chromosome 19q13, and some of the down-regulated genes in chromosome 17q21. These observations are supported by previous findings showing that the amplification and rearrangement of chromosome 17p11.2 occurred at a high frequency in Birt-Hogg-Dube syndrome (47), osteosarcoma (48, 49), and glioma (50), and the breakpoint of chromosomal abnormalities at band chromosome 19q13 is frequently found in primary gastric cancer (51). The presence of tumor suppressor genes on chromosome 17q21 is also supported by the proposal that chromosome 17q21, including the BRCA1 locus, may contain a candidate for tumor suppressor genes in gastric cancer (52). These reports and our data imply that the up-regulated genes clustered on chromosome 17p11.2 and chromosome 19q13 might be candidates for an oncogene, and the down-regulated genes on chromosome 17q21 candidates for a tumor suppressor.

Several groups have recently reported on the results of expression profile analyses in gastric cancers using high-density microarrays (15–17, 22, 53–59). The candidate genes reported by these groups were mostly abundantly or intermediately expressed genes in gastric cancers, whereas many of our candidate genes are rarely expressed genes or novel genes which were seldom selected by other groups. By combining these results, as mechanisms related to gastric cancer pathogenesis and progression, we propose that up-regulated CKS1B in gastric cancer cells might promote the expression of CDC20, the highly induced the CDC20 would also increase the activation of anaphase-promoting complex and the initiation of anaphase and the progression of the cell cycle then be accelerated. In addition, of hypoxia related proteins induced by HIF-1α, glycolytic enzymes such as GAPD, ENO1, PKM2, PGK1, and LDHA have been reported to up-regulated in gastric cancer (18, 60). GPI, an enzyme involved in glycolysis, is known to be a hypoxia-inducible factor in other forms of cancer (61). The findings here indicate that this gene is up-regulated in gastric cancer cell lines. From these results, one possibility is that the HIF-1α signaling pathway might be related to the pathogenesis and progression of gastric cancer. These newly identified genes should provide valuable resources for developing an understanding of the molecular mechanism associated with tumorigenesis of gastric cancer and for discovering potential diagnostic markers for gastric cancer.

We thank Dr. Young-il Yeom for spotting the DNAs on the slides.