Construction Of CeRNA Network In The Carcinogenesis Of Cirrhosis And Identification Of Key Genes

Background Long non-coding RNA(lncRNA)could competitively sponge miRNAs,and thus up-regulate the expression of downstream target genes.These RNAs were also called competitive endogenous RNA(ceRNA).Studies have indicated an important role of ceRNA in the development of liver cirrhosis.Despite of an increased risk of hepatocellular carcinoma(HCC)in cirrhosis,no studies systematically constructed the ceRNA network in the carcinogenesis of cirrhosis.This study aimed to use the microarray data to construct a ceRNA network related with the carcinogenesis of cirrhosis,and identify key genes.Method The microarray data including several stages from cirrhosis to HCC were downloaded from the database of Gene Expression Omnibus(GEO).LncRNA-mRNA expression profile was obtained by probe re-annotation.Then,the genes were divided into different modules by weighted gene co-expression network analysis(WGCNA)according to the expression pattern.In the modules related with the carcinogenesis of cirrhosis,the ceRNA network was constructed respectively.The common hub genes between the WGCNA and ceRNA network analysis were regarded as key genes in the carcinogenesis of cirrhosis.In validation dataset,the expression difference of key genes was evaluated between cirrhosis and HCC,as well as their roles in the prognosis of HCC.Finally,molecular experiment was conducted to validate one of the key genes.Results Sixty-five liver samples included seven stepwise stages from cirrhosis to HCC.In WGCNA analysis,four modules were identified in significant association with the carcinogenesis of cirrhosis(R~2>0.7,P<0.05).In each significant module,the ceRNA network was constructed.The 11 common hub genes between the WGCNA and ceRNA network were regarded as key genes,among which 8 genes were positively correlated with the carcinogenesis(ASPM,CCNB1,CDC6,CDC23,EIF3H,EXO1,RAD21,TOP2A),and 3 genes were negatively correlated(CXCL12,DCN,HGF).In validation dataset(N=37),7 genes had a significantly higher expression level in HCC than cirrhosis(ASPM,CCNB1,CDC6,CDC23,EIF3H,EXO1,TOP2A)(P<0.05),while 3 genes had a lower expression(CXCL12,DCN,HGF)(P<0.05)and no significant change was found in RAD21(P>0.05).The signature of two key genes showed a good performance in the diagnosis between cirrhosis and HCC(area under ROC curve(AUC)=0.821).In HCC,the high expression in 6 genes indicated a poorer prognosis(ASPM,CCNB1,CDC6,CDC23,EIF3H,EXO1,TOP2A)(P<0.05),and high expression of CXCL12 and DCN indicated a better prognosis(P<0.05).The cases with high RAD21 expression had a lower 3-year survival rate(P<0.05),while the cases with high HGF expression had a higher 3-year survival rate(P<0.05).The 11-gene signature showed a good(concordance index(C-index):0.683,95%CI:0.628~0.738)and stable(P between groups=0.640)performance in predicting the prognosis of HCC.In the experimental validation of ASPM,there existed a positive association between ASPM expression and hepatic fibrosis severity in the CCL4-induced fibrosis rat model(P=0.042).After knock-down of ASPM on hepatoma cell lines(7721 and M9),there found a significant decrease in the ability of proliferation,clone formation,migration and invasion(P<0.05),and an increase in apoptosis(P<0.05).After knock-down of ASPM on the hepatic stellate cell line(LX2),there found a significant disease in the expression ofα-SMA and COL1A1(P<0.05),and an increase in apoptosis(P<0.05).Conclusions This study combined two methods of biological network analyses to construct a ceRNA network related with the carcinogenesis of cirrhosis,and identified11 key genes.The subsequent validation by the dataset and molecular experiment also proved the robustness of the method and results in this study.