Real-time quantitative PCR assay was performed to analyze the copy number of thirty candidate genes in 30 pairs of gastric cancer and normal gastric tissues. With a gene copy number of 4 or more defined as gene amplification, we found that ERBB4, C-MET and CD44 genes were frequently amplified in gastric cancers, however, other genes were not or infrequently amplified in gastric cancers, ranging from 0 to 8% (data not shown). Subsequently, we used the same method to analyze the copy number of ERBB4, C-MET and CD44 genes in the 128 gastric cancers and 37 normal controls. The data showed that the prevalence of amplification of ERBB4, C-MET and CD44 was 67% (86/128), 30% (39/128) and 66% (84/128), respectively, but not in the normal gastric tissues. Copy number of each gene corresponding to each individual case of gastric cancers and normal gastric tissues was showed in Figure 1. Statistical analysis showed that the copy number of each gene in gastric cancers was significantly higher than normal gastric tissues (Figure 1).

Because highly frequent amplification of ERBB4, C-MET and CD44 was demonstrated in gastric cancer, their association with clinicopathological characteristics was analyzed in a cohort of clinically well-characterized gastric cancers. As shown in Table 1, there was a positive association of amplification of ERBB4 (OR = 2.62, 95% CI = 1.23–5.59, P < 0.05) and CD44 (OR = 2.28, 95% CI = 1.08–4.79, P < 0.05) with tumor differentiation. C-MET amplification was found to be significantly positively associated with tumor invasion (OR = 2.00, 95% CI = 1.03–3.89, P < 0.05). CD44 amplification was significantly positively associated with lymph node metastasis (OR = 2.23, 95% CI = 1.05–4.73, P < 0.05). Moreover, our data also showed that CD44 amplification was significantly positively associated with the number of lymph node metastasis (OR = 1.70, 95% CI = 1.08–2.69, P < 0.05). Notably, there was a significantly positive association of amplification of these genes with cancer-related death (Table 1). In order to assess the independent association of gene amplification with age, tumor differentiation, tumor stage, lymph node metastasis and survival status, we conducted multiple multivariable logistic regressions (Table 2). Similar to univariate analysis, after adjustment, amplification of ERBB4 (OR = 2.95, 95% CI = 1.27–6.86, P < 0.05) and CD44 (OR = 2.49, 95% CI = 1.08–5.79, P < 0.05) remained significantly associated with poor tumor differentiation (Table 2). Similarly, amplification of these genes remained significantly positively associated with cancer-related death (Table 2).

The Kaplan-Meier estimator of the survivorship function was used to evaluate the impact of amplification of these three genes on the survival of gastric cancer patients. The survival of gastric cancer patients with and without gene amplification was compared using the log-rank test. As shown in Figure 2, amplification of ERBB4, C-MET and CD44 significantly affected the poor survival of gastric cancer patients.

Numerous evidences showed that residual tumor after surgery is an independent risk factor for gastric cancer patients. We thus attempted to evaluate the effect of residual tumor after surgery on the survival of gastric cancer patients. As shown in Figure 3, the patients with residual tumor after surgery had significantly shorter survival times than the patients without residual tumor (343.2 months vs. 601.2 months on average, P = 0.03). Given the impact of residual tumor after surgery on poor survival in gastric cancer, we excluded the patients with residual tumor to explore the effect of gene amplification on poor prognosis of gastric cancer patients. Similar to the findings in Figure 2, the patients with gene amplification had shorter survival times than the patients without gene amplification (ERBB4: 508.8 months vs. 777.6 on average, P = 0.002; C-MET: 382.8 months vs. 690.0 months on average, P = 0.0005; CD44: 490.8 months vs. 798.0 months on average, P = 0.0002) (Figure 4). Multivariate Cox regression analysis indicated that gene amplification may be served as a predictor of poor prognosis for gastric cancer patients (ERBB4: HR = 2.00, 95% CI = 1.02–3.86, P = 0.04; C-MET: HR = 2.10, 95% CI = 1.20–3.69, P = 0.01; CD44: HR = 2.59, 95% CI = 1.34–5.01, P = 0.005) as an independently variable with respect to gender, age, differentiation, lymph node metastasis, and tumor stage (Table 3).

Gastric cancer is chronic proliferative disease characterized by genetic alterations of multiple genes, including gene amplification. In the present study, a high prevalence of concomitant amplification of 2 of 3 genes was found in gastric cancer, ranging from 28% (36/128) to 60% (77/128). As shown in Figure 5, concomitant amplification of 2 of 3 genes significantly shortened survival times (ERBB4/C-MET: 378.0 months vs. 795.6 on average, P = 0.001; ERBB4/CD44: 483.6 months vs. 817.2 months on average, P = 0.0004; C-MET/CD44: 372.0 months vs. 822.0 months on average, P < 0.0001), and might be more prognostic of poor survival than amplification of individual gene in gastric cancer.

With the approval of our institutional review board and human ethics committee, where required, we studied 128 patients with gastric cancer who underwent surgery at the First Affiliated Hospital of Xi’an Jiaotong University School of Medicine from 1999 to 2006. A total of 37 tissues from the patients with chronic gastritis were obtained from the First Affiliated Hospital of Xi’an Jiaotong University School of Medicine as normal controls. None of these patients received chemotherapy and radiotherapy before the surgery. Informed consent was obtained from each patient before the surgery. The histologic diagnosis of tumors was made and agreed upon by at least two senior pathologists at Department of Pathology of the Hospital based on World Health Organization (WHO) criteria. Relevant clinicopathologic characteristics were obtained from the patients’ files or by interview with the patients or their relatives, and the details were summarized in Table 4.

Serial sections from each tumor sample were cut. One section (5 μm) was stained by hematoxylin and eosin (H & E), and was marked as a tumor representative tissue by an expert surgical pathologist for gastric cancer. The next section (8 μm) was deparaffinized and stained with hematoxylin. Tumor tissues were isolated by manual microdissection under an inverted microscope using the marked H & E section for target tissue identification. Genomic DNA was extracted from isolated tissues as previously described [28]. Briefly, the tissues dissected were first treated with xylene for 12 hours at room temperature to remove the paraffin. All tissues were subsequently subjected to digestion with 1% sodium dodecyl sulfate (SDS) and proteinase K at 48 °C for 48 to 72 hours with addition of several spiking aliquots of concentrated proteinase K to facilitate digestion. Genomic DNA was isolated from the digested tissues followed by standard phenol-chloroform extraction and ethanol precipitation protocol, and stored at −80 °C until use.

We analyzed the copy number of candidate genes in gastric cancers and normal gastric tissues by real-time quantitative PCR technique on a CFX384 Thermal Cycler Dice^TM real-time PCR system (Bio-Rad Laboratories, Inc., Hercules, CA, USA) as described previously [29]. This method was well established and validated by florescence in situ hybridization (FISH) [29,30], which has been widely used in the various human cancers [28–32]. Specific primers and TaqMan probes were designed using Primer Express 3.0 (Applied Biosystems: Foster City, CA, USA, 2004) to amplify these genes and the internal reference gene β-actin. The primers and TaqMan probes for ERBB4, C-MET, CD44 and β-actin genes in the present study were presented in Table 5. The primers and TaqMan probes for other 27 genes need be requested, including SHFM1, CARD4, ELN, ARF5, SLC25A40, NRAS, CDK5, CREM, LMO2, DNMT1, FMR2, PSPHL, KRAS, PEG10, CDC2L5, HRAS, MGAM, ZP3, EPO, GUSB, ZPBP, TMEM60, PEPIN1, BRAF, AHR, DNMT3B and EZH2. All TaqMan probes were labeled with 5′-FAM (6-carboxyfluorescein, fluorescent reporter) and 3′-TAMRA (6-carboxy-tetramethylrhodamine, fluorescent quencher). Using a PCR protocol described previously [29], PCR amplification were carried out in buffer containing 16.6 mM ammonium sulfate, 67 mM Tris base, 2.5 mM MgCl₂, 10 mM 2-mercaptoethanol, 0.1% DMSO, 0.2 mM each of dATP, dCTP, dGTP and dTTP, 600 nM each of forward and reverse primers, 200 nM TaqMan probe, 0.6 unit Platinum Taq polymerase and 2% Rox reference dye. Each sample was run in triplicate, and β-actin was run in parallel to standardize the input DNA. Standard curves were established using serial dilutions of normal leukocyte DNA with a quantity range of 3.75 to 60 ng per 2 μL. The specificity of real-time quantitative PCR for these genes was confirmed by running thee PCR products on a 1.5% agarose gel to show single specific bands of the PCR products at the expected sizes (data not shown). The efficiency of real-time quantitative PCR assays for each target was shown in Table 5. Gene amplification was defined by a copy number ≥4.

The Mann–Whitney U test was used to compare copy number of these genes between gastric cancer and normal gastric tissues. Association of gene amplification with clinicopathological characteristics was assessed univariately using the SPSS statistical package (version 11.5; IBM Corporation: Chicago, IL, USA, 2003). Multivariate models were then developed that adjusted for the most important covariates, including age, differentiation, tumor stage, lymph node metastasis and survival status. Survival length was determined from the day of primary tumor surgery to the day of death or last clinical follow-up. The Kaplan–Meier method was used for survival analysis grouping with gene amplification. Differences between curves were analyzed using the log-rank test. Multivariate Cox regression analysis was used to evaluate the effect of gene amplification on survival of independently of gender, age, differentiation, lymph node metastasis, and tumor stage. All statistical analyses were performed using the SPSS statistical package. P values < 0.05 were considered significant.