Abstract
Background/Aim: Erythropoietin-producing hepatocellular carcinoma receptor A (EphA) is associated with angiogenesis and invasive tumor progression. In this study, we evaluated the EphA1-4 expression levels in advanced gastric cancer. Patients and Methods: Tumor tissues obtained from 114 patients with advanced gastric adenocarcinoma who underwent gastrectomy were analyzed. In addition, the impact of EPHA 1-4 mRNA expression on survival was analyzed using the Kaplan-Meier plotter database on the website. Results: High EphA 1, 2, and 4 expression levels were significantly related to recurrence (p<0.01, p=0.04, and p<0.01). Both high EphA 1 and 4 expression levels were independent predictors of relapse-free interval (hazard ratio [HR]=2.0, p=0.03; HR=2.4, p=0.03) and disease-specific survival (HR=2.0, 95% p=0.03; HR=2.5, p=0.02) on multivariate analysis. In the Kaplan-Meier plotter database, high EPHA2 mRNA expression was significantly associated with poor survival in patients with gastric cancer (p=0.0098), and high expression levels of EPHA1 and 4 tended to be associated with poor survival (p=0.050, p=0.052). Conclusion: EphA 1, 2, and 4 may play key roles in recurrence and survival in patients with advanced gastric cancer.
Gastric cancer (GC) is a common cause of cancer-related death worldwide (1). Surgery followed by adjuvant chemotherapy is standard treatment for locally advanced GC. However, many patients with advanced disease have recurrence after complete macroscopic removal of the tumor and metastatic lymph nodes. Many candidate prognostic factors or promising biomarkers have been proposed in gastric cancer, although few molecular targets are clinically valuable.
Erythropoietin-producing hepatocellular (Eph) receptors are the largest family of receptor tyrosine kinases and are activated by interacting with cell-surface ligands, ephrins. Eph receptors are classified into A-type (EphA 1-8 and EphA 10) and B-type (EphB1-4 and EphB6) according to their interactions with ephrin ligands (2). Eph receptors and ephrin ligands modify the organization of the actin cytoskeleton and influence the activities of integrins and intercellular adhesion molecules, thereby regulating cell morphology, adhesion, or migration (3). The up-regulation of Eph receptors and ephrin has been reported in various types of cancer. Eph and ephrin can promote tumor progression by activating downstream signaling pathways, while combinations of Eph receptors and ephrin ligands are thought to depend on the specific type of cancer (4).
We previously reported the immunohistochemical expression levels of individual EphA1, 2, 3, and 4 receptors in GC. Individual overexpression of EphA1, 2, 3, or 4 receptor was significantly associated with progression and survival in GC, although about half of the GCs in our previous studies were early stage (5-7). Therefore, the present study was designed to globally evaluate the clinical significance of EphA 1 to 4 receptor expression in locally advanced GC treated surgically with no residual tumor. In addition, we analyzed the effects of mRNA expression levels of EPHA1, 2, 3, and 4 on survival using a Kaplan-Meier plotter, which is an online analysis tool for investigating the effects of genes on survival in patients with GC (8).
Patients and Methods
Patients. The study group comprised a total of 114 patients who were given a histopathological diagnosis of advanced gastric adenocarcinoma and underwent gastrectomy without residual tumors from January 2003 through December 2007 in the Department of Gastrointestinal Surgery, Tokyo Medical and Dental University. No patient received neoadjuvant therapy. Each tumor was classified according to the TNM classification (9). All tumors were pathologically classified according to the World Health Organization (WHO) pathological classification. All patients were provided a sufficient explanation of the study and provided written informed consent. This study was approved by the Institutional Review Board of Tokyo Medical and Dental University (No. M2000-831). All patients were evaluated for recurrent disease by undergoing diagnostic imaging examinations (computed tomography, ultrasonography, magnetic resonance imaging, or endoscopy) every 3 to 6 months. Positron emission tomography and bone scintigraphy were additionally performed, if necessary. The median follow-up was 60 months (range=5-109 months). Twenty-one patients (18%) received postoperative adjuvant chemotherapy with S-1 (an oral fluoropyrimidine preparation consisting of tegafur, gimeracil, and oteracil potassium; Taiho Co., Ltd., Tokyo, Japan) alone. A total of 48 (44%) patients had recurrent disease. Finally, 46 (40%) patients died of recurrent disease, and 3 (3%) died of other causes. HER2 status was previously evaluated and scored according to standardized assessment criteria by one of the co-authors (YT).
Immunohistochemistry. The detailed method used to perform immunohistochemical analysis has been described in our previous reports (5-7). The antibody of EphA1 (mouse monoclonal antibody, MAB8303) was purchased from Abnova Inc. (Taipei, Taiwan), and antibodies of EphA2 (rabbit polyclonal antibody, sc-924), EphA3 (rabbit polyclonal antibody, sc-920), and EphA4 (rabbit polyclonal antibody, sc-921) were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA). Representative formalin-fixed, paraffin-embedded tissue blocks were sliced into 4-μm-thick sections. After deparaffinization and rehydration, antigen retrieval was performed. Subsequently, endogenous peroxidase and non-specific binding were blocked. The slides were incubated with each antibody against EphA1 (dilution 1:500), EphA2 (dilution 1:100), EphA3 (dilution 1:500), and EphA4 (dilution 1:150) at 4°C overnight. The primary antibodies were purchased from Abcam (Cambridge, UK). The sections were incubated with peroxidase-labeled antirabbit or anti-mouse antibody (Histofine Simplestain Max PO; Nichirei Co., Tokyo, Japan) for 30 min at room temperature. Peroxidase activity was detected with diaminobenzidine (DAB; Nichirei Co., Tokyo, Japan). Subsequently, the sections were counterstained with Mayer's hematoxylin (Wako, Tokyo, Japan). Negative control sections were treated similarly, except that the antibodies were replaced by normal rabbit IgG (Santa Cruz Biotechnology, Inc.). Strongly stained gastric specimens obtained from the same block were used as positive controls to reduce bias arising from the staining conditions in each immunohistochemical assay.
Interpretation of immunohistochemical staining. The stained slides were evaluated by at least two separate investigators (YT, MN, and NB) who were blinded to the patients' outcomes. The investigators counted wholly stained cancer cells of representative cross-sectional slices. To evaluate heterogeneous staining of cancer cells, we counted at least three fields per section of the invasive front of each tumor. The scoring system of EphA staining was somewhat modified in this study, because different scoring systems were used in previous studies (5-7). The staining intensity was scored into the following three grades: 0, no staining; 1, weakly positive; and 2, moderately or strongly positive. The staining extent (positive frequency) was also scored into three grades according to the percentage of stained tumor cells as follows: 0, <10%; 1, 10% to 50%; and 2, >50% stained cells. For statistical analysis, composite scores were calculated by adding the intensity and extent scores. Composite scores of ≥3 were defined as high expression, and scores of <3 were defined as low expression.
Statistical analysis. Categorical data were compared with the use of the chi-square test or Fisher's exact test, as appropriate. Kaplan-Meier curves were plotted to assess the effect of EphA expression levels on relapse-free interval (RFI) and disease-free survival (DSS), and different survival curves were compared using the log-rank test. Before multivariable analysis it was confirmed that the prognostic factors did not strongly correlate with each other. Multivariable Cox proportional hazards regression models were used to assess the prognostic significance of EphA expression levels and of several clinicopathological factors associated with RFI and DSS. Values of p<0.05 were considered to indicate statistical significance. All statistical analyses were performed with the statistical software package SPSS 24 (SPSS Japan Inc., Tokyo, Japan).
Survival analysis using the Kaplan-Meier plotter database. The Kaplan-Meier plotter datasets were obtained from Gene Expression Omnibus, Cancer Biomedical Informatics Grid, and The Cancer Genome Atlas. In the present study, the Kaplan-Meier plotter database http://kmplot.com/analysis/index.php?p=service&cancer=gastric, accessed November 17th, 2017) was used to show the distinct prognostic values of the EPHA 1, 2, 3, and 4 genes in GC. We excluded the GSE52254 dataset from the present study because of markedly different characteristics as compared with the other datasets of GC in compliance with the website suggestion and then included the remaining five datasets in this study (GSE14210, GSE15459, GSE22377, GSE29272, GSE51105) (8). The JetSet best probe sets were selected for each gene (EPHA1, 205977_s_at; EPHA2, 203499_at; EPHA3, 206070_s_at; EPHA4, 227449_at) in compliance with the website recommendation. The cut-off values of high or low mRNA expression were also selected automatically on the website. Thus, the overall survival curves associated with high or low expression of EPHA1 - 4 mRNA levels were available on the website. The survival curves were compared using the log-rank test.
Results
High EphA1, 2, 3, and 4 expressions were observed in 63 (55%), 83 (72%), 63 (55%), and 69 (61%) tumors, respectively (Figure 1). The relations between EphA expressions and clinicopathological factors are shown in Table I. High EphA1 and EphA3 expression levels were significantly associated with undifferentiated type (p<0.01) and differentiated type (p<0.01), respectively. Serosal invasion was significantly associated with high EphA4 expression (p<0.01). No expression level was related to lymphatic invasion by the primary tumor, whereas only high EphA2 expression was significantly related to venous invasion by the primary tumor (p=0.01). Lymphatic metastasis was significantly associated with high EphA2 and 4 expression levels (p=0.02, p<0.01). Advanced stage (stage III vs. I/II) was significantly associated with high EphA3 and 4 expression levels (p=0.03, p<0.01).
Immunostaining for EPHAs in representative primary gastric cancers, EPHA1 (A), EPHA2 (B), EPHA3 (C), and EPHA4 (D). Magnification ×400.
Recurrence was significantly more frequent in patients with high EphA1, 2, and 4 expression levels than in patients with low EphA1, 2, and 4 expression levels (p<0.01, p=0.04, and p<0.01). In particular, lymphatic recurrence was significantly associated with high EphA1 and 4 expression levels (p=0.03, p=0.01), and peritoneal recurrence was slightly but not significantly associated with high EphA1 and 4 expression levels (p=0.08, p=0.07). Hematogenous recurrence was not significantly associated with any expression level. The presence of only high EphA3 expression was not associated with recurrence, but was significantly associated with male sex (p=0.03), higher age (>65 years) (p=0.03), and high HER2 expression (p=0.02).
Patients with high EphA1, 2, and 4 expression levels had significantly shorter RFIs than patients with low expression levels (p=0.02, p=0.04, and p<0.01; Figure 2). High EphA1 and 4 expression levels were both independent predictors of RFI on multivariate analysis of clinicopathological features adjusted for the following variables: serosal invasion, lymph-node metastasis, pathological type, and high EphA expression levels (EphA1, hazard ratio [HR]=2.0, 95% confidence interval (CI)=1.1-3.9, p=0.03; EphA4, HR=2.4, 95%CI=1.1-5.3, p=0.03; Table II). As for RFI, high EphA1 and 4 expression levels were both independent prognostic factors of DSS on multivariate analysis (HR=2.0, 95%CI=1.1-3.9, p=0.03; HR=2.5, 95%CI=1.1-5.5, p=0.02; Table III). As for combinations of EphA expression, the subgroup of patients with 3 or more high expression levels of EphA1-4 had the worst RFI as compared with the subgroup of patients with no high expression and the subgroup of patients with only 1 or 2 high expression levels. RFI differed significantly among the 3 groups (p=0.001, Figure 3).
Relations between the expression levels of EphA1 to 4 and clinicopathological factors.
In the analysis of overall survival using the Kaplan-Meier plotter database, EPHA1, 2, and 3 expression levels were analyzed in 593 patients, and EPHA4 expression levels were analyzed in 348 patients. High mRNA levels of the EPHA1, 2, and 3 genes were present in 444 patients (75%), 427 patients (72%), and 395 patients (67%), respectively. High EPHA4 expression was found in 116 patients (33%). The survival curves are shown in Figure 4. High EPHA2 expression was significantly associated with poor survival (p=0.0098). High EPHA1 and 4 expression levels were associated with slightly but not significantly poorer survival (p=0.050, p=0.052, respectively). Conversely, high EPHA3 expression was significantly associated with better survival (p=0.0013).
Survival curves of our samples. The Kaplan-Meier curve for the relapse-free interval of patients with EphA1 (A), EphA2 (B), EphA3 (C), and EphA4 protein expression (D).
Discussion
Our results showed that high EphA1 and 4 protein expression levels were independent predictors of RFI in patients with locally advanced GC and were significantly associated with recurrent disease, especially lymph-node recurrence. EphA4 expression was also significantly related to serosa invasion and lymph-node metastasis in the present study, whereas EphA1 was significantly related to these variables only in undifferentiated-type GC.
Several studies have previously assessed EphA4 expression in human gastrointestinal cancers. In another study of GC, EphA4 overexpression was significantly related to tumor growth factors and recurrence (10). EphA4 was significantly associated with the invasion depth of the primary tumor, lymph-node metastasis, and survival in patients with rectal cancer. More interestingly, high EphA4 expression was a significant predictor of a poor response to chemoradiotherapy in rectal cancer (11). Another study showed that tumors with higher EPHA4 mRNA levels were significantly associated with liver metastasis from colorectal cancer (12), whereas our study did not show a positive relationship between EphA4 protein expression and hematogenous recurrence in GC. This discrepancy might be ascribed to methodological differences or different recurrence patterns between GC and colorectal cancer. Recurrence of GC is mainly associated with lymphatic, peritoneal, and hematogenous metastasis, while recurrence of colorectal cancer is mainly associated with hematological metastasis and local recurrence.
Survival curve based on the number of EphA with high expression levels.
Prognostic factors for relapse-free interval in univariate and multivariate Cox proportional-hazards regression models.
In another study of EphA1 in GC, high EphA1 protein expression was associated with advanced disease, lymph-node metastasis, and poorer survival of patients, although EphA protein expression was down-regulated in most tumors with methylation of the EPHA1 gene (13). We previously showed that no tumor with high EphA1 protein expression has the gene amplification on fluorescence in situ hybridization (6). Several in vitro studies have demonstrated the tumor-progressive function of EphA1. Knockdown of the EPHA1 gene inhibited proliferation, invasion, and migration via the regulation of multiple signaling pathways, such as matrix metalloproteinase (MMP)-2, extracellular signal-regulated kinase 2 (ERK2), and proto-oncogene c-MYC in ovarian cancer cell-lines (14). EphA1-activated SDF-1 expression enhanced the recruitment of endothelial progenitor cells, which can promote angiogenesis, to hepatocellular cancer cells (15). On the other hand, EphA1 expression is down-regulated in advanced colorectal cancer (16).
Prognostic factors for disease-specific survival in univariate and multivariate Cox proportional-hazards regression models.
Survival curves in the Kaplan-Meier plotter database. The Kaplan-Meier curve for the overall survival of patients with EPHA1 (A), EPHA2 (B), EPHA3 (C), and EPHA4 gene expression (D).
EphA2 is up-regulated in many cancers and its expression has been linked to increased malignancy and poor clinical outcomes (4). Some functions of EphA2 have been reported in various malignancies. EphA2 signaling in malignant melanoma can promote vascular mimicry (17). The consequent activation of EphA2 by ephrin-A1 enhances the angiogenic effects of VEGF in endothelial cells (18). In colorectal cancer, EphA2 overexpression was associated with high density of microvessels (19). EphA2 activated Akt in glioma stem cells (20). Conversely, however, EphA2 has been reported to have tumor suppressive activity in non-small cell lung carcinoma (21). Activated Akt decreases EphA2 phosphorylation through a negative regulatory loop and can thus inhibit cell migration and invasion (22).
EphA3 was highly expressed in various cancers (23). In another study of GC including early disease, high EphA3 protein expression was significantly associated with tumor depth, lymph-node metastasis, distant metastasis, poor survival of patients, and VEGF expression (24). EphA3 was related to both differentiated-type and HER2 expression and tended to be associated with hematogenous recurrence in the present study. EphA3 overexpression was related to poor outcomes in patients with hepatocellular carcinoma and colorectal cancer (25, 26).
We could not analyze the gene expression levels of EPHA 1 and 4 in our GC samples. We therefore used the Kaplan-Meier plotter database to analyze the impact of these levels on survival. Overexpression of EPHA 1, 2, and 4 was associated with poor survival in GC, as protein overexpression was associated with recurrence and poor survival in our immunohistochemical study. EPHA3 overexpression was conversely associated with outcomes, while the protein expression was not associated with recurrence in our immunohistochemical study. Nearly all GC tumors in the database were advanced, although a detailed TNM classification was unavailable because of blank data.
In conclusion, immunohistochemical high expression levels of EphA1, 2, and 4 may play key roles in recurrence and outcomes of advanced GC. In particular, EphA1 or 4 was an independent predictor of survival in advanced GC treated by surgery. The inhibition of EphA1, 2, and 4 might serve as a target for preventing tumor metastasis in GC.
Footnotes
Conflicts of Interest
The Authors have no conflicts of interest or financial ties to disclosure.
- Received December 17, 2017.
- Revision received January 12, 2018.
- Accepted January 15, 2018.
- Copyright© 2018, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
