Novel Key Genes In Gastric Cancer Identified By Constructing A Weighted Gene Co-expression Network And The Relationship Between The Expression Of COL8A1 And Gastric Cancer

BackgroundGastric cancer(GC)is one of the most common malignant gastrointestinal tumors worldwide.The morbidity and mortality of GC ranked 5th and 3rd among all malignant tumors,respectively.China is an epidemic area of GC with higher incidence and death rate than the global average level.Due to atypical early symptoms,most patients are diagnosed with GC at an advanced stage and others may experience recurrence and metastasis after radical surgery.For patients with advanced GC,chemotherapy is the primary treatment,but the efficacy is limited and the prognosis of GC is poor.With the development of molecular biology,some molecular mechanisms of GC have been gradually clarnfied.Signaling pathways associated with GC include PI3K/AKT signaling pathway,MAPK signaling pathway,erbB signaling pathway,vascular endothelial growth factor signaling pathway,c-Met signaling pathway,cyclooxygenase 2/NF-?B signaling pathway,etc.Some drugs targeting these pathways,such as trastuzumab,ramucirumab and apatinib,have been approved for the treatments of advanced GC patients.However,only a small subset of patients can benefit from these target therapies.Other target drugs,such as everolimus targeting the mTOR pathway,cetuximab targeting the EGFR pathway and rilotumumab targeting the c-MET pathway,have failed to show significant efficacy in large-sample clinical trials.Therefore,it is important to explore the underlying mechanisms of GC and screen for novel treatment targets.In recent years,the developments of next-generation sequencing and gene microarray technologies as well as the systems biology analytical methods based on high-throughput data have provided new strategies for exploring molecular mechanisms and identifying drug targets of malignant tumors.These methods can screen for tumor-related driving genes and therapeutic targets more efficiently.Weighted gene co-expression network analysis(WGCNA)is a kind of systems biology analytical methods by constructing scale-free networks based on power law.Compared with random networks,weighted gene co-expression networks are more closed to biological regulatory networks.Moreover,WGCNA can also identify modules and genes related to external characteristics.Therefore,it has been widely and successfully applied in various biological contexts.In this study,a weighted gene co-expression network was constructed based on transcriptome big data of GC obtained from public databases,and novel key genes related to GC were identified.Furthermore,the pro-tumorigenic role of the key gene in GC was also validated through molecular biology experiments.Part 1.Novel key genes in gastric cancer identified by constructing a weighted gene co-expression networkObjectiveTo construct a weighted gene co-expression network in GC based on big data and identify novel key genes related to GC for further studies.Materials and methodsData acquisition Transcriptome data and clinicopathological information of GC patients were obtained from The Cancer Genome Atlas(TCGA)database and GEO(Gene Expression Omnibus)database,which served as the training dataset and validation datasets,respectively.Data preprocessingThe gene expression data was normalized and the samples with incomplete clinicopathological information were excluded.The 5000 most variant genes were filtered for network construction according to median absolute deviation of genes.Constructing a weighted gene co-expression networkThe WGCNA package in R software was used to construct the co-expression network:(1)A sample network was constructed,and the outlier samples were removed according to Euclidean distances.(2)Scale-free networks were fitted and the optimal soft threshold was selected.(3)The expression correlations of all genes were calculated,and the gene expression matrix was converted into the gene dissimilarity of topological overlap matrix.(4)A hierarchical clustering tree based on the gene dissimilarity of topological overlap matrix was constructed,and the dynamic tree cut method was performed for branch cutting to generate modules.Meanwhile,the module eigengenes of modules were calculated,and highly correlated modules were merged to obtain the co-expression network.Identifying preserved modules with clinical significanceThe correlations between module eigengenes and clinicopathological characteristics were analyzed.Then,the preservations of modules in the validation datasets were evaluated.Finally,preserved modules with clinical significance were identified.Enrichment analysesGene ontology(GO)and Kyoto Encyclopedia of Genes and Genomes(KEGG)signal pathway enrichment analyses were performed on the genes in the preserved modules with clinical significance to explore the potential functions and pathways.Identifying candidate genes of GCThe hub genes of the module were identified according to module memberships.Then,the correlations between the hub genes and clinicopathological characteristics were analyzed in the training dataset and validation dataset to identify candidate genes of GC.Identifying key genes of GCThe differences of candidate gene expression between in normal tissues and in GC tissues were analyzed.Then,the prognostic values of the candidate genes were evaluated through overall survival(OS)and relapse-free survival(RFS)analyses using the TCGA dataset and Kaplan Meier plotter.Finally,the candidate genes which were differentially expressed between in normal tissues and in GC tissues and correlated to OS and RFS in GC patients were identified as key genes.Results1.Three GC datasets were obtained from the public databases,including the TCGA dataset,the GSE15459 dataset and the GSE26942 dataset.2.After excluding outlier samples and filtering of genes with less variation,a total of 20 modules were generated based on the weighted gene co-expression network.3.Among these modules,ten were related to histological grade,six were related to pathological T stage,and five were highly preserved both in the GSE15459 dataset and in the GSE26942 dataset.Finally,two modules which were highly preserved and related to clinicopathological characteristics were identified.The module consisting of 506 genes was further analyzed while the other one was discarded because of the low module membership values of genes in it.4.Enrichment analyses showed that genes in the module were involved in endothelial cell proliferation and growth factor binding.Moreover,they also participated in some cancer-related signaling pathways such as PI3K/AKT signaling pathway,TGF-beta signaling pathway,TNF signaling pathway,etc.5.Twenty-six genes with module membership values greater than 0.8 in the module were identified as hub genes.Among these hub genes,PDGFRB,COL8A1,EFEMP2,FBN1,EMILIN1,FSTL1,KIRREL and MRC2 were identified as candidate genes for they were correlated with histological grade and pathological T stage both in the TCGA dataset and in the GSE15459 dataset(all P<0.05).6.Among the candidate genes,PDGFRB,COL8A1,FBN1,FSTL1 and KIRREL were overexpressed in GC tissues compared with normal tissues(all P<0.05).Among these differentially expressed genes,PDGFRB,COL8A1 and FSTL12 were associated with OS of GC patients(all P<0.05),and PDGFRB and COL8A1 were associated with RFS of GC patients(both P<0.05).Therefore,PDGFRB and COL8A1 were identified as key genes of GC.ConclusionsPDGFRB,COL8A1,EFEMP2,FBN1,EMILIN1,FSTL1,KIRREL and MRC2 were the candidate genes of GC and may relate to the proliferation and differentiation of GC.Most of them were first reported in GC.Among these candidate genes,PDGFRB and COL8A1 were also upregulated in GC and associated with RFS and OS of GC patients.They may be novel key genes of GC and deserve further studies.Part 2.The relationship between the expression of COL8A1 and gastric cancerObjectiveIn our previous study,we found that the expression of COL8A1 was related to clinicopathological characteristics of GC patients.The objective of the present study was to explore the expression,biological functions and potential molecular mechanisms of COL8A1 in GC.Materials and methods1.The difference of COL8A1 mRNA expression levels between in GC and in normal tissues was analyzed through the Oncomine database.Real-time fluorescence quantitative PCR was used to detect the expression levels of COL8A1 mRNA and the difference of COL8A1 mRNA expression levels between in GC cells and in normal gastric mucosa cells was analyzed.2.The expression of COL8A1 protein in the tissue microarray of GC was detected by immunohistochemical staining.Then,the differences of COL8A1 protein expression levels between in GC cells and in normal gastric mucosa cells and between in GC tissues and in normal gastric mucosa tissues were analyzed.Moreover,correlations between COL8A1 protein expression and clinicopathological characteristics as well as prognosis of patients were also analyzed.3.Lentiviral shRNA vectors were constructed to knock down the expression of COL8A1 gene in GC cells.Then,the proliferation of GC cells was detected by Cell Counting Kit-8;the cloning formation of GC cell was detected by plate cloning formation test;the apoptosis of GC cells was detected by Annexin V-APC staining and flow cytometry;the cell cycle of GC cells was detected by PI staining and flow cytometry;the invasion of GC cells was determined by transwell invasion assay;the migration of GC cells was detected by transwell migration assay and wound-healing assay.4.The expression correlations between COL8A1 and 29 genes related to PI3K/AKT signal pathway in the co-expression module identified in the previous study were analyzed.The phosphorylation of Akt protein in GC cells was detected by Western-blot.Results1.The mRNA of COL8A1 was upregulated in GC tissues and cells compared with normal gastric mucosa tissues and cells(all P<0.05).2.The protein of COL8A1 was overexpressed in GC cells compared with normal gastric mucosa cells(P<0.001).However,no significant difference of COL8A1 protein expression levels between in GC tissues and in normal gastric mucosa tissues was observed.COL8A1 protein was significantly upregulated in the pT3/4 patients compared with the pT1/2 patients(P =0.049).It was also significantly upregulated in the G3 patients compared with the G1/2 patients(P =0.017).There were no significant correlations between gender,age,tumor size,regional lymph node metastasis and pathological stage and COL8A1 protein expression.Log-rank test showed that the patients with high expression of COL8A1 protein had shorter overall survival time(P =0.01).Univariate Cox analysis indicated that patients with high expression of COL8A1 protein had increased risks of death(HR:1.024,95%CI:1.004-1.044).Multivariate Cox analysis suggested that the independent prognostic value of COL8A1 protein was marginally significant(P =0.067).3.After COL8A1 knockdown,compared with the control group,the proliferation of GC cells was significantly suppressed on day 2,day 3,day 4 and day 5(all p<0.001);the cloning formation capabilities of GC cells significantly decreased(P=0.03);the apoptosis of GC cells significantly increased(P<0.001);the proportion of cells in G0/G1-phase significantly increased(P =0.008)while the proportion of cells in G2/M-phase significantly decreased(P =0.01),the proportion of cells in S-phase also decreased although there was no statistical difference(P = 0.109);the invasion and migration of GC cells were significantly depressed(all P<0.05).4.Through expression correlation analyses,it was found that COL8A1 significantly co-expressed with the 29 genes which related to PI3K/AKT signaling pathway(all P<0.05).Furthermore,the phosphorylation level of Akt protein in GC cells was significantly reduced after COL8A1 knockdown(P<0.001).ConclusionsCOL8A1 mRNA was upregulated in GC cells and tissues;COL8A1 protein was overexpressed in GC cells and correlated with pathological T stage,histological grade and OS of GC patients.COL8A1 may promote the proliferation,invasion and migration and inhibit the apoptosis of GC cells by activating PI3K/AKT signaling pathway.