High Hepsin expression predicts poor prognosis in Gastric Cancer

Hepsin, a membrane-associated serine protease, is frequently upregulated in epithelial cancers and involved in cancer progression. Our study aims to describe the expression pattern and evaluate the clinical implication of hepsin in gastric cancer patients. The mRNA expression of hepsin was analyzed in 50 gastric cancer and matched non-tumor tissues, which was downregulated in 78% (39/50) of gastric cancer. By searching and analyzing four independent datasets from Oncomine, we obtained the similar results. Furthermore, we evaluated the hepsin expression by IHC in tissue microarray (TMA) containing 220 Gastric Cancer specimens. More importantly, Kaplan-Meier survival and Cox regression analyses were taken to access the prognosis of gastric cancer and predicted that hepsin protein expression was one of the significant and independent prognostic factors for overall survival of Gastric Cancer.

Scientific RepoRts | 6:36902 | DOI: 10.1038/srep36902 (Supplemental Figure 2). The tumor samples have similar tumor cell content (60% approximately) would be used to detect the mRNA or protein expression according to HE staining. The Hepsin mRNA expression level was downregulated in 78% (39/50) of gastric cancer patients (Fig. 1A). Furthermore, we searched and analyzed Hepsin mRNA expression in five independent microarray datasets from Oncomine database (Cho, Cui, Wang, Chen, and DErrico) (Fig. 1B, Supplemental Figure 1), which is consistent with our data. Hepsin protein expression levels were examined by western blot and immunohistochemical staining (IHC) methods (Figs 1C and 2). As shown in Fig. 1C, one protein bands, whose molecular weight is about 42 kDa, is observed in gastric cancer tissues and matched adjacent pericancer tissues in western blot analysis. A decrease in hepsin expression was observed in 55% (22/40) of the gastric cancer tumor tissues compared with the matched adjacent peritumor tissues (Fig. 1C,D). Further, the expression of hepsin protein was examined in a tissue microarray (TMA) containing 220 pairs of gastric tissues by IHC staining analysis. Evidence was presented that hepsin protein expression was mainly located in the nuclear and cytoplasm of gastric tumor cells, peritumoral tissue cells and normal tissue cells (Fig. 2). The IHC density of hepsin exhibits a significant difference in gastric tumor tissues and their matched adjacent non-tumor tissues (P < 0.001) (Fig. 2G).
Hepsin is a transmembrane serine protease and has one predicted N-glycosylation site in Asn-112. We found that N-glycosylation at Asn-112 is important for hepsin cell surface targeting and tumor invasion and migration. Wild hepsin was mainly localized in the cytoplasm and the cell membrane, while hepsin mutant N112Q was localized predominantly in the nucleus of the AGS cells (Supplemental Figure 3). So we speculated that nuclear hepsin in IHC should be de-glycosylated or not fully glycosylated. In addition, wild hepsin overexpression enhanced cell migration and invasion (Supplemental Figure 4), whereas hepsin mutant N112Q attenuated the invasion and metastatic potential compared to wild hepsin in MGC80-3 cells (Supplemental Figure 4). Similar roles of N-glycosylation in regulating cell surface expression and protease activity have been reported in other type II  transmembrane serine proteases, such as corin 24 , enteropeptidase 25 , matriptase 26 , and matriptase-2 27 , which are involved in blood pressure regulation, food digestion, epithelial function, and iron metabolism, respectively 24,28,29 .

Relationship between Hepsin expression and clinical parameters.
To determine the clinical significance of hepsin expression in gastric cancer, the Chi-square test was taken to assess the associations between hepsin protein expression and clinicopathological parameters (including age, gender, tumor location, Histological differentiation, Lauren classification, T classification, N classification, distant metastasis, clinical stage and intravascular cancer emboli). The results demonstrated that hepsin expression in gastric cancer tissues is closely associated with histological differentiation (P = 0.001), Lauren classification (P = 0.001), T classification (P = 0.011) N classification (P = 0.039) and clinical stage (P = 0.007). No significant associations were detected between hepsin expression and age, gender, tumor location, distant metastasis or intravascular cancer emboli ( Table 1). The results were confirmed in the validation set of patients. It was also demonstrated that hepsin expression was correlated to Lauren classification (P = 0.038), T classification (P = 0.033) N classification (P = 0.031) and clinical stage (P = 0.023) (Supplementary Table 1).

Table 1. Relation between intratumoral Hepsin expression and clinical characteristics of gastric cancer.
Abbreviation: TNM = tumour node metastasis. P-value < 0.05 marked in bold font shows statistical significant.

Correlation between Hepsin expression and prognosis in Gastric cancer patients.
To further evaluate the prognostic value of hepsin in gastric cancer, we explored the correlation between hepsin expression and clinical data by Kaplan-Meier analysis and log-rank test. As shown in Fig. 3, high hepsin expression was associated with poor overall survival (P = 0.0014). Furthermore, we explored the association between hepsin expression and overall survival in gastric cancer patients with early or advanced clinical stages and with or without lymphatic metastasis and vascular invasion. According to Kaplan-Meier analyses, we found the overall survival is shorter in gastric cancer patients with high hepsin expression in all stages ( Fig. 3A) or in III-IV stage (Fig. 3C). Similar result was confirmed in our validation set (Supplementary Figure 5). High expression of hepsin was also found to be associated with poor overall survival in gastric cancer patients without vascular invasion or with lymphatic metastasis (Supplemental Figure 6). Consistent with our results, the prognostic value of hepsin in gastric cancer was verified by online survival analysis software (http://www.kmplot.com/analysis/index. php?p= service&cancer= gastric), which integrated reported microarray datasets. The result demonstrated that high expression of hepsin correlated to poorer overall survival (Fig. 3D) as well as progression-free survival ( Fig. 3E) in gastric cancer patients. In order to obtain a more sensitive predictive model for outcomes of gastric cancer patients, we combined hepsin expression and TNM stage to create a prognostic score system. ROC analysis revealed that the combination of hepsin and TNM stage showed better prognostic value [area under curve (AUC) 0.785, 95% confidence interval (CI) 0.723-0.846] than TNM stage alone (AUC 0.755, 95% CI 0.689-0.820, P = 0.044) or hepsin expression alone (AUC 0.591, 95% CI 0.516-0.666, P < 0.001) (Fig. 3F). We also analyzed the recurrence free survival information of GSE26253 datebase (Supplemental Figure 7). In addition, univariate and multivariate analyses showed that hepsin could be useful as an independent risk factor for poor prognosis in the 220 cases of gastric cancer. The univariate Cox regression analyses showed that T classification (P < 0.001), N classification (P < 0.001), distant metastasis (P < 0.001), clinical TNM stage (P < 0.001), Intravascular cancer emboli (P = 0.0024) and hepsin expression (P = 0.0014) were significantly relevant with overall survival in gastric cancer. The multivariate Cox regression analyses, however, showed that T classification (P = 0.016), distant metastasis (P < 0.001), clinical TNM stage (P = 0.011) and hepsin expression  (Table 2). Then the Harrell's concordance index (C-index) analyses were examined to assess the predictive accuracies of TNM stage and hepsin protein expression (Table 3).
Nomogram and calibration plot analyses for gastric cancer patients. Based on obtained evidence, we used patients' data in the two cohorts to develop a nomogram to predict OS at 3 and 5 years after surgery (Fig. 4A). The predictors included tumor T stage, N stage, M stage and hepsin expression, all of which were independent prognostic indicators for OS. In the nomogram, a higher total point represents a worse survival. The calibration plot predicted 5-year overall survival were built to give the internal validation, which performed well compared with the ideal model (Fig. 4B). We next stratified the gastric cancer patients into 3 groups according to the score calculated using the nomogram: low-risk (< 25th percentile), intermediate-risk (25th-75th percentile), and high-risk (> 75th percentile) groups (Fig. 4C). As shown in Fig. 4C, the nomogram could effectively discriminate the risk of OS in gastric cancer patients.

Discussion
Although gastric cancer incidence has declined for decades, it remains the fifth most common cancer and the fifth leading cause of cancer-related mortality in USA 30 . The majority of gastric cancer patients are diagnosed at advanced stage, due to lacking early detecting methods 31 . To date, the outcomes for gastric cancer patients with similar TNM stage can be very different because of the heterogeneity of this tumor, and the prognostic models for gastric cancer patients are largely relied on the TNM stage 5 . Therefore, we need to identify some novel molecules associated with tumorigenesis of gastric cancer and better understand the tumor progression and predict the cancer clinical outcomes. It will be helpful to compare expression levels of mRNA and protein of hepsin in tumor tissues with matched normal tissues of gastric cancer patients to describe the physiological and pathophysiological importance of hepsin in gastric cancer.
To this end, we have first time described the hepsin expression pattern in gastric cancer tissues in both protein and mRNA levels in this study. Additionally, the hepsin protein expression and its relationship with the clinicopathological parameters and clinical prognosis values are illustrated.
In the study, hepsin mRNA and protein expression level is mainly downregulated in Gastric cancer tissues. High expression of hepsin is correlated with poorer overall survival, as well as progression-free survival in patients with gastric cancer. There is a significance between high hepsin expression and low hepsin expression in gastric cancer patients with advanced stage, III-IV, so we speculate that hepsin may contribute to gastric cancer in later stage. In Zuyan Luo's study, the OLFM4 expression pattern and correlation with gastric cancer patients' overall survival are similar with our results 32 . Our findings reflect distinct actions of gastric and gastric carcinoma cells in response to hepsin protein. Probably, gastric and gastric carcinoma cells use different signaling pathways in response to hepsin expression. For example, hepatocytes react differently than hepatoma cells to IL-6 stimulation in regulating HBV replication. In hepatoma cells, IL-6 stimulates HBV transcription by activating STAT-3, which interacts with HNF3 bound to the HBV enhancer 33 . However, in primary human hepatocytes, IL-6 suppresses HBV gene expression and replication through the down-regulation of HNF4a and HNF1a 34 .
Although a large quantity of membrane-associated proteinase has been found, their biological roles are still unknown. Associations between TMPRSS2 and TMPRSS4 and cancers have been reported. TMPRSS2 may be a potential diagnostic or therapeutic target for prostate cancer, which is considered to have a role in cell biology 35 . TMPRSS4 is usually overexpressed in pancreatic cancer, however, its functional significance remains to be illustrated 36 . Low expression levels of hepsin and TMPRSS3 are associated with poor breast cancer survival 37 . Because hepsin is upregulated in advanced stage of gastric cancer, it may contribute to expansion, growth, invasion and metastasis of these tumor cells.
As investigated, membrane-associated serine proteases play a vital role in tumor invasion and metastasis 38,39 . Hepsin expression may contribute to gastric cancer progression and metastasis by a few molecular mechanisms. For instance, hepsin could act as a growth factor, which plays a role in stimulating the proliferation and increase progression ability of cancer cells 40,41 . This hepsin activity for cultured hepatocytes has been determined 21 . Hepsin may also directly or indirectly degrade extracellular matrix proteins by activating matrix metalloproteinases (MMPs) 42 . As well known, proteolytic digestion of extracellular matrix proteins plays a crucial role in tumor invasion and metastasis. High hepsin expression in gastric cancer tissues may at some certain reflect the state of poorly differentiated gastric cancer cells. Testing gastric cancer models in hepsin-deficient mice should indicate the biological significance and the true molecular mechanism of hepsin in gastric cancer.
In conclusion, our study has demonstrated that increased hepsin expression is correlated with poor prognosis in gastric cancer patients, and hepsin may be identified as an independent prognostic factor and may be a potential target for the treatment of gastric cancer patients. In order to better understand hepsin's physiological functions, additional experimentation remains to be determined. Real-time PCR. Real-time PCR analyses were carried out as described previously 43 . The total RNA was isolated from the normal gastric tissues, gastric cancer tissues and matched peritumor tissues by using TRIzol reagent (Invitrogen, USA) according to the manufacturer's instruction. Hepsin mRNA expression levels were determined by using specific primers after normalization with glyceraldehyde 3 phosphate dehydrogenase (GAPDH). The primer sequence were: GAPDH, (Forwad) 5′ -GTCAAGGCTGAGAACGGGAA-3 and (Reverse) 5′ -AAATGAGCCCCAGCCTTCTC-3; Hepsin, (Forward) 5′ -GTCTGCAATGGCGCTGACTTCT-3′ and (Reverse) 5′ -TCCGAGAGATGCTGTCCTCACA-3′ .

Materials and Methods
Western blotting. Western blot analyses were performed as described previously 44 . Primary antibodies were rabbit anti-hepsin (Abcam, UK), mouse anti-actin (Cell Signal Technology, USA). And species-specific (mouse or rabbit) secondary antibody was purchased from Santa Cruz Biotechnology. Immunochemistry staining. Immunohistochemical staining protocol was constructed as previous described 43 . Depending on the staining extent, the score of hepsin expression was conducted: 0, 0-5%; 1, 5-25%; 2, 26-50%; 3, 51-75%; and 4, > 75%, and the staining intensity was categorized as follows: no staining scored 0, weakly staining scored 1, moderately staining scored 2 and strongly staining scored 3, respectively. The staining score was designated by multiplying staining area score by staining intensity score, yielding a series of results ranging from 0 to 12. High expression was considered as a total score > 4 and low expression with a total score ≤ 4 according to receiver operating characteristic (ROC) analysis. The immunochemistry staining scores were determined independently by two pathologists who were blinded to the patients' clinical data. Immunofluorescent staining. Immunofluorescent staining analyses were carried out as described previously 43 . Hepsin was detected with Anti-V5 (Invitrogen) and visualized with goat anti-mouse IgG-Alexa Fluor 488 (Jackson). 5 mg/ml DAPI (Beyotime) was used for nuclear staining. Images were taken by a Confocal Laser Scanning Microscope (Leica TCS SP5, Germany).
Transwell assay. Transwell assay was performed as described previously 45  respectively. The infiltrating cells were stained with crystal violet, and cell numbers were counted from five fields. Each experiment was repeated three times. Statistical analysis. SPSS19.0 (SPSS Inc.; Chicago, IL, USA) and GraphPad Prism 5 (San Diego, CA) software were used for statistical analyses and graphical representations. The χ 2 test was used to analyze the relations between hepsin expression and clinicopathological parameters in gastric cancer patients. Survival curves were evaluated using the Kaplan-Meier method, and differences between survival curves were tested by the log-rank test. Cox proportional hazards regression model was used to examine univariate and multivariate analyses. Only significantly different variables in univariate analysis were entered into the next multivariate analysis. Nomogram was generated by R software with "rms" package. Calibration plot for 5-year overall survival was constructed to examine the performance characteristics of the generated nomogram. The prognostic accuracy was measured by calculating the Harrell's concordance index (c-index). A two-sided P-value < 0.05 was considered statistically significant.