The prognostic significance of BGN for gastric cancer
In the GSE65801 dataset, we came to the same conclusion that BGN mRNA level was higher in tumor tissues than that in normal tissues (Figure 1A). Kaplan-Meier analysis showed that higher expression of BGN was significantly associated with poorer recurrence-free survival (RFS) of gastric cancer patients. In the GSE26253 dataset, samples were divided into four subgroups according to the expression level of BGN, Q1, Q2, Q3, and Q4, respectively. BGN mRNA levels negatively correlated with RFS of gastric cancer patients (Figure 1B). Therefore, the BGN expression level was negatively correlated with the prognosis of gastric cancer patients in a dose-dependent manner. In a stratified survival analysis according to the pathological stage, samples were re-stratified as BGN-high (equal or greater than the median) and BGN-low (less than the median), according to the expression levels of BGN mRNA. The HR values were 1.44 (95%CI 1.02-2.06, p=0.038) and 2.16 (95% CI 1.22-3.87, p=0.007) for high BGN expression in stage I-III (n=365) and stage IV gastric cancer patients respectively (Figure 1C and 1D). These results suggested that high BGN mRNA levels were significantly related to poor prognosis in gastric cancer.
Bioinformatics analysis for the genes and proteins interaction network of BGN
To understand the biological functions of BGN, we conducted bioinformatics analysis for coexpressed genes of BGN on Oncomine. The analysis of genes coexpressed with BGN was conducted on Chen Gastric dataset(28). We screened out more than 10 genes with a strong correlation with BGN, such as THBS2, ARHGAP5, FN1, INHBA, and CDH11 (Figure 2A). Meanwhile, the bioinformatics analysis for the protein-protein interaction network was conducted on www.strig-db.org. Figure 2B showed the protein-protein interaction network of BGN, more than a dozen genes were reported interacting with BGN through text mining, such as VCAN, TLR4, HSPG2, TGFB1, and GPC1. Most of the above genes were involved in cell growth, cell communication, signal transduction, and cell adhesion (Figure 2C), which closely related to tumorigenesis.
GSEA of BGN in gastric cancer
To explore the cancer-related gene signatures of BGN, we performed a GSEA on the GSE26253 dataset, a downloaded microarray dataset of 432 gastric cancer cases. The expression of BGN was significantly associated with the following gene sets: Park hsc VS multipotent progenitors UP (Figure 3A), Nakamura metastasis model DN (Figure 3B), IVANOVA Hematopoiesis Stem Cell Long Term (Figure 3C) and RICKMAN Tumor Differentiated Moderately VS Poorly UP (Figure 3D) in GSE26253 dataset. GSEA results suggested that BGN significantly enriched metastasis and poor prognosis gene signatures, revealing that BGN might be associated with proliferation, poor differentiation, and high invasiveness of gastric cancer.
Prediction of putative transcription factors of BGN by bioinformatic analysis
In order to further understand the carcinogenic mechanism, it is essential to explore the upstream regulation of BGN in gastric cancer. The prediction of the BGN promoter region has used the website of http://gtrd.biouml.org/ and http://alggen.lsi.upc.es/cgi-bin/promo_v3/promo/promoinit.cgi?dirDB=TF_8.3. In Fig 4A, the promoter region of BGN would be around the signal of H3K27Ac, which located around the 1st exon and partially overlapped with CpG island. Meanwhile, the potential transcription factor binding sites (TFBSs) were predicted by Gene Transcription Regulation Database (GTRD) and ALGGEN - PROMO. By co-expression analysis, the eligible transcription factors were identified, including AR, CEBPA, CEBPB, E2F1, ELF1, GATA1, MAZ, PAX5, RXRA, SP1, STAT5A, TCF4, TP53, and YY1. Figure 4B showed the location of these eligible TFBSs on the promoter of BGN. A linear regression analysis indicated expression BGN significantly and positively associated with TCF4 level, while negatively associated with AR or E2F1(Figure 4C).
BGN expression in protein level of gastric cancer tissue
For gastric cancer, the protein expression level of BGN in tumor tissues was significantly higher than that in normal tissues (Figure 5). Unfortunately, BGN was mainly expressed in the extracellular matrix rather than in the intracellular matrix, which made quantitative analysis difficult.