Biglycan Overexpression Predicts Poor Prognosis of Gastric Cancer


 Background: Biglycan (BGN) encodes an extracellular matrix (ECM) proteoglycan. However, the potential diagnostic and prognostic value of BGN in gastric cancer (GC) have not yet been reported. In this analysis, BGN expression in GC was evaluated across the Gene Expression Omnibus (GEO), The Cancer Genome Atlas (TCGA), and Oncomine databases, and verified using immunohistochemistry (IHC). The relationship between BGN expression and clinicopathological parameters was assessed by chi-square test and logistic regression. We analyzed the prognostic value of BGN. Then, Gene set enrichment analysis (GSEA) was used to screen the signaling pathways involved in high BGN expression datasets in GC. Finally, CIBERSORT was used to evaluate the infiltration of immune cells in GC tissues, and the correlation between BGN and infiltrating immune cells was analyzed.Results: The results showed that the mRNA levels of BGN were significantly up-regulated in GC compared with normal tissues (all P <0.001). The Kaplan-Meier plotter online database suggested that patients with high BGN expression had a poor prognosis (P=1.3e-10). In addition, using gene sets analysis, we found that pathways of bladder cancer, Wnt-signaling, TGF-beta signaling, and ECM-receptor interaction were differentially activated in high-expression BGN tissues. Furthermore, CIBERSORT analysis for the proportion of TICs revealed that macrophages M2 was positively correlated with BGN expression. Conclusions: In conclusion, BGN can be used as potential diagnostic markers of GC, and immune cell infiltration plays an important role in the occurrence and progression of GC. The finding may have significant implication for the diagnosis, prognosis and treatment of GC.


Background
Gastric cancer (GC) is a high mortality disease and is often found more frequently in men all over the world (1). Although the overall incidence of GC is declining, when the GC tissue in ltrates into submucosa, enters into muscle layer or has passed through muscle layer to serosa, the overall survival rate of advanced GC is less than one year (2). Even after various treatments, the 5-year survival rate was only 36%-47% (3). Poor survival may be closely related to late detection. At present, the diagnosis of GC mainly depends on gastroscopy and nuclear magnetic resonance imaging. Moreover, researchers have recently proposed oncolytic virus-mediated uorescence imaging combined with endoscopic imaging, which may improve the accurate imaging and treatment of gastrointestinal tumors (4). However, even these imaging modalities are ineffective in some patients with GC (5,6). Therefore, it has clinical signi cance to discover effective biomarkers or target molecules for the diagnosis and treatment of GC. Biglycan (BGN) encodes an extracellular matrix (ECM) proteoglycan that binds effectively to TGF-beta (7) and has carcinogenic effect (8). BGN not only participates in the long-distance migration of cells, but also has the effect of adhesion (9). In previous reports, the expression level of BGN has been studied as a potential biomarker for esophageal and bladder cancer (10,11). It has been shown that BGN can inhibit the proliferation of pancreatic cancer (12). BGN has the function of regulating the size of collagen bers and may be deposited in blood vessels when combined with collagen (13). A large number of studies have pointed out that BGN plays an important and positive role in promoting angiogenesis (14). There is strong evidence for the involvement of BGN in neovascularization of tumors. In addition, BGN regulates osteocyte differentiation through BMP signaling pathway (15). However, the potential diagnostic and prognostic value of BGN in GC have not yet been determined.
Here, we provided the rst evidence that BGN overexpression predicts poor prognosis of GC and BGN is a potential biomarker for GC prognosis. The higher expression of BGN in GC tissues compared to normal tissues was further validated using immunohistochemistry. We identi ed the correlation between BGN and clinical features of GC. Furthermore, BGN regulation signaling pathways in GC were explored and enriched by Gene Set Enrichment Analysis (GSEA). The prognostic role of BGN in GC and its relationship with immune in ltration were also studied. The nding may have signi cant implication for the diagnosis, prognosis and treatment of GC.

Results
The mRNA expression of BGN in GC The expression of BGN in various tumor types and corresponding normal tissues was analyzed by oncomine database (Fig. 1A). The result shows that there are 421 unique analyses for BGN, respectively. In GC, BGN was signi cantly increased in 73 datasets and 5 datasets showed a reduced level. we performed a meta-analysis of BGN expression in 6 analyses with a threshold set as p-Value ≤ 1E-4, fold change ≥ 2 and top 10% gene rank in the Oncomine database ( Fig. 1B, P < 1.24E-6). This analysis showed that high expression of BGN in GC tissues, and its gene rank was higher, with statistical signi cance. The details of the GEO series were summarized in Table 1. Table 1. Details of GEO series included in this analysis.
As illustrated in Fig. 2, the expression of BGN in GC samples was all signi cantly higher than non-tumor samples in GSE29272, GSE29998, GSE54129 and GSE66229 ( Fig. 2A-D, all P < 0.05).
Interestingly, the expression level of BGN was also found to be signi cantly higher in GC that in normal tissues in TCGA datasets (Fig. 2E, P < 1.159e-16). The paired difference analysis also showed that BGN expression was higher in GC than that in normal tissues (Fig. 2F, P < 1.527e-9).

Immunohistochemical analysis of BGN in GC
We have collected 15 clinical GC cases and determined the BGNs' protein expression in GC tissues and normal tissues using immunohistochemistry (IHC). The positive expression of BGN was shown as brown granules in cytoplasm (Fig. 3A). We found that BGN are highly expressed in the GC tissues than that in the normal tissues ( Fig. 3B, P < 0.001).

GC patient characteristics
A total of 443 GC samples were downloaded from the TCGA online website, which included clinical information and gene expression datasets ( Table 2). The average age of GC patients was 67 years old, 47.63% of the patients were younger than 67 years old, 52.37% of the patients were older than or equal to 67 years old. The gender analysis showed 158 females (35.67%) and 285 males (64.33%). Table 2. Clinical characteristics of the gastric cancer patients.

BGN overexpression in GC
As shown in Fig. 4A-E BGN expression was closely related to clinical stage (p = 0.002), histological grade (p = 0.004), and T classi cation (P = 6.605e-6). As an independent variable, BGN has a certain correlation with other clinical features (Table 3). Relevance was judged by logistic regressive univariate analysis. Increased BGN expression in GC was statistically signi cantly correlated with higher stages (OR = 3.40 for I vs. II, and OR = 3.18 for I vs. III) and higher T classes (OR = 12 for T1 vs. T2, OR = 21.27 for T1 vs. T3, and OR = 24.86 for T1 vs. T4). The results suggest that BGN overexpression in GC was more likely to develop into advanced cancer and worse prognosis. Table 3. Biglycan expression associated with clinical characteristics (logistic regression).

Survival outcomes, univariate and multivariate Cox analyses
By using Kaplan Meier plotter analysis, we found that BGN overexpression in GC had a worse overall survival rate (P = 1.3e-10, Fig. 5). The median OS of the low expression of BGN group was 89.43 months, while the median OS of the high expression of BGN group was 23.6 months.
Because survival was signi cantly correlated with BGN expression in GC patients. According to Cox proportional hazard regression model, 375 patients with GC were analyzed by univariate and multivariate analysis to evaluate the effect of BGN expression and other clinicopathological factors on survival rate.

Immune Cell In ltration Results
To further validate the correlation of BGN expression with the immune microenvironment, the proportion of tumor-in ltrating immune subsets was analyzed using CIBERSORT algorithm, and 21 different kinds of immune cell pro les in GC samples were constructed ( Fig. 7A-B).

Correlation Analysis between BGN and In ltrating Immune Cells
The violin plot of the immune cell in ltration difference showed that, plasma cells decreased in the high expression of BGN group, while macrophages M2 showed an increase in the high BGN expression group (Fig. 8A). Correlation analysis showed that BGN was positively correlated with macrophages M2 (r = 0.26, p = 0.0031), and negatively correlated with activated dendritic cells (r = − 0.2, p = 0.023) and Plasma cells (Fig. 8B, r = − 0.3, p = 0.00057). The results from the difference and correlation analyses showed that a total of two kinds of TICs (M2 macrophage and plasma cells) were correlated with the expression of BGN (Fig. 8C).

Discussion
We found that BGN overexpression is positively correlated with the progression and worse survival of GC patients. BGN encodes a proteoglycan, which belongs to an ECM protein (16). ECM proteins play an important role in the tumor microenvironment of GC (17). Interestingly, BGN may in uence the progression of GC by promoting neovascularization (18). Although Wang et al. (19) considered that BGN was related to the prognosis of GC, but did not provide speci c survival analysis, and did not elaborate on the content of immune cell in ltration in the tumor microenvironment of GC. Based on the literature search, the impact of BGN expression on GC prognosis has not been fully studied yet. For the rst time, our study comprehensively analyzed the transcriptional level and prognostic value of BGN in GC. Our results showed that the mRNA level of BGN in GC tissue was increased compared with normal tissue, and the protein expression of BGN was also higher than normal tissue. Therefore, this study provided the rst evidence that BGN overexpression predicts poor prognosis of GC and BGN is a potential biomarker for GC prognosis. The nding may have signi cant implication for the diagnosis and treatment of GC and the evaluation of patients' prognosis.
At present, the diagnosis and prognosis of GC are based on clinical classi cation, pathological staging, and histological grading. We found that the expression of BGN statistically signi cantly increased gradually with histological grades T1/T2/T3/T4 and G1/G2/G3. The expression of BGN in stage I/II was also relatively high. GC. The univariate and multivariate Cox analysis provides evidence that BGN may be an independent biomarker for the prognosis of GC patients. Sun et al. (20) and Jacobsen et al. (21) revealed that BGN expression was up-regulated in endometrial and prostate cancer respectively. Li et al. demonstrated that over-expression of BGN activated the TGF-beta/Snail signaling pathway leading to poor survival in patients with colon cancer (22). These reports are consistent with our ndings.
To understand the role and mechanisms of BGN in GC, we use GSEA to enrich and analyze the signaling pathways. We found that Wnt signaling pathway, TGF-beta signaling pathway, ECM-receptor interaction, and the bladder cancer pathways were signi cantly enriched in the BGN over-expression tissues. Previous studies suggest that BGN was involved in the transformation, progression and metastasis of many tumors. Sun (24). This report is consistent with our identi cation of TGF-beta signaling pathway in GC. Recent studies have shown that EREG gene can be used as a marker of independent prognosis of GC (25). There is no report of Wnt signaling pathway in relation to BGN in any cancers. Further mechanistic studies are needed.
Nowadays, an increasing interest in tumor immune in ltrates was paid in tumor immunotherapy. To further explore the role of immune cell in ltration in GC, we used CIBERSORT to conduct a comprehensive evaluation of GC immune in ltration. In this study, the CIBERSORT analysis for the proportion of tumor-in ltrating immune cell revealed that macrophages M2 was positively correlated with BGN expression in GC patients. In addition, there was a negative correlation between plasma cells and BGN expression in GC patients. We suggest that an increased in ltration of macrophages M2 and a decreased in ltration of plasma cells may be related to the occurrence and development

Conclusion
It can be speculated that understanding the expression of macrophages M2 in GC tissues has a certain clinical value in predicting the prognosis of GC. This study found that the positive correlation between the amounts of macrophages M2 and BGN expression in GC patients. These results revealed the possibility that BGN may be closely involved in the tumor immunity in GC. BGN may promote the immune escape of cancer cells of GC by promoting M2 polarization of macrophages, thus affecting tumor growth and prognosis.

Materials And Methods
Microarray data Oncomine analysis The Oncomine database (http://www.oncomine.org) is an expression database for most cancer gene chips (33,34). The database is a whole-genome expression analysis aimed at formulating new therapeutic goals. We investigated the transcriptional levels of BGN in GC and corresponding normal tissues by using Oncomine.
GEO microarray series (GSE29272, GSE29998, GSE54129, GSE66229) containing GC and non-tumor samples were obtained from (http://www.ncbi.nlm.nih.gov/geo/). In addition, the gene expression data of GC were also downloaded from the TCGA database (https://cancergenome.nih.gov/). The data were standardized, and log2 transformed. Gene expression of BGN was determined by GC and non-tumor samples. Meanwhile, relevant clinical information was obtained from the TCGA website. All analysis operating procedures were run by Perl and R software. All analysis operating procedures were run by Perl and R software (35).

Immunohistochemistry and evaluation of immunostaining intensity
The study consisted of 30 samples from 15 patients diagnosed with GC. All participants signed informed consent forms. Samples were collected and approved by the Ethics Committee of the Jingzhou First People's Hospital, Yangtze University. The para n section of GC tissue was dewaxed, and the endogenous peroxidase was blocked by 3% H 2 O 2 , the surface antigen was repaired by microwave oven, and goat serum was blocked. The appropriate amount of primary antibody BGN (1: 300, ab209234; Abcam) was added, overnight at 4°C. After rewarming, the second antibody was added and incubated in 37°C incubator for 60 min, then 3, 3-diaminobenzidine tetrahydrochloride (DAB) chromogenic solution was added. Then add hematoxylin, dehydrate step by step, seal the lm and observe under microscope. Image Pro Plus 6.0 was used to analyze and count the results. The relative expression of BGN was measured by the integrated optical density (IOD).

Statistical analysis
The signi cance between characteristic clinical information in GC and genes, such as BGN, were determined by Kruskal-Wallisl test, Wilcoxon test and logistic regression method. The expression of BGN in normal tissues and GC tissues was shown by scatter plot and paired difference analysis plot. The Kaplan-Meier plotter database (www.kmplot.com) can be used to analyze the survival trends of 54675 genes, including 1440 GC samples. Kaplan-Meier plotter online database was used to analyze the overall survival rate of BGN in GC patients. The clinical information data of GC were analyzed by univariate Cox, and potential positive variables were selected. The correlation between BGN expression and survival along with other clinical factors was compared by multivariate Cox analysis (P < 0.05).

Gene Set Enrichment Analysis
GSEA is a method of enrichment analysis and calculation based on existing gene sets (36). GSEA could determine which group of prede ned genes was closer to the high BGN group and the low BGN group to nd signi cant differential expression. Genome replacement tests were set up by 1000 times. The BGN gene expression was used as phenotype label. P < 0.05 and q-value < 0.25 were set as the cut-off criteria.     The survival prognostic value of biglycan expression on overall survival in gastric cancer patients. Red lines represent high expression of genes in gastric cancer, and black lines represent low expression of genes in gastric cancer.