Comprehensive analysis of the competing endogenous circRNA-lncRNA-miRNA-mRNA network and identification of a novel potential biomarker for hepatocellular carcinoma

Abstract

Background: The competing endogenous RNAs (ceRNAs) hypothesis has received increasing attention as a novel explanation for tumorigenesis and cancer progression. However, there is still a lack of comprehensive analysis of the circular RNA (circRNA)-long non-coding RNA (lncRNA)-miRNA-mRNA ceRNA network in hepatocellular carcinoma (HCC).

Methods: RNA sequencing data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) database were employed to identify Differentially Expressed mRNAs (DEmRNAs), DElncRNAs, and DEcircRNAs between HCC and normal tissues. Candidates were identified to construct networks through a comprehensive bioinformatics strategy. A prognostic mRNA signature was established based on data from TCGA database and validated using data from the GEO database. Then, the HCC prognostic circRNA-lncRNA-miRNA-mRNA ceRNA network was established. Finally, the expression and function of an unexplored hub gene, deoxythymidylate kinase (DTYMK), was explored through data mining. The results were examined using clinical samples and in vitro experiments.

Results: We constructed a prognostic signature with seven target mRNAs by univariate, lasso and multivariate Cox regression analyses, which yielded 1, 3 and 5-year AUC values of 0.797, 0.733 and 0.721, respectively, indicating its sensitivity and specificity in the prognosis of HCC. Moreover, the prognostic signature could be validated in GSE14520. The prognostic ceRNA network of 21 circRNAs, 15 lncRNAs, 5 miRNAs, and 7 mRNAs was established according to the targeting relationship between 7 hub mRNAs and other RNAs. Our experiment results indicated that the depletion of DTYMK inhibited liver cancer cell proliferation and invasion.

Conclusions: The network revealed in this study may help comprehensively elucidate the ceRNA mechanisms driving HCC, and provide novel candidate biomarkers for evaluating the prognosis of HCC.

Keywords: DTYMK; HCC; ceRNA network; prognostic signature.

Figure 1

Identification of differential genes. (A) Heatmap of DEcircRNAs from GEO databases. (B) Venn diagram of the intersection of DEcircRNAs. (C) Volcano maps of DEmRNAs from TCGA. (D) Volcano maps of DElncRNAs from GSE138178. (E) Volcano maps of DElncRNAs from TCGA. (F) Venn diagram of the intersection of DElncRNAs.

Figure 2

Prediction of common targeted miRNAs and their targeted DEmRNAs. (A) Flow chart of common pre-miRNAs prediction. (B) The relationship between DEcircRNAs and targeted miRNAs. (C) The relationship between DElncRNAs and targeted miRNAs. (D) Venn diagram of the intersection of circ-pre-miRNAs and lnc-pre-miRNAs. (E) The relationship between the common miRNAs and their targeted DEmRNAs.

Figure 3

Development the prognostic model. (A, B) Lasso regression analysis results. The trajectory of each independent variable, the horizontal axis represents the log value of the independent variable lambda, and the vertical axis represents the coefficient of the independent variable. The tuning parameter (λ) was calculated based on the partial likelihood deviance with ten-fold cross validation. The dotted vertical lines are drawn at the optimal values by minimum criteria and 1-SE criteria. (C) The risk scores distribution, survival status, and gene expression patterns of patients in high and low-risk groups. The dot presents patient status ranked by the increasing risk score. The X axis is patient number and Y axis is survival time. (D) Kaplan–Meier survival curve of two groups. (E) The time-dependent ROC curves analyses of two groups.

Figure 4

Validation of the prognostic model. (A) Forrest plot of the univariate Cox regression analysis in TCGA. (B) Forrest plot of the multivariate Cox regression analysis in TCGA. (C) The risk scores distribution, survival status, and gene expression patterns of HCC patients in GSE14520. The dot presents patient status ranked by the increasing risk score. The X axis is patient number and Y axis is survival time. (D) Kaplan–Meier survival curve of two groups in GSE14520. (E) The time-dependent ROC curves analyses of two groups in GSE14520.

Figure 5

Gene set enrichment analyses between high and low risk group in TCGA. (A) The top ten significantly enriched cancer hallmarks in high-risk group. (B) The significantly enriched GO terms in high-risk group. (C) The significantly enriched oncological signatures in high-risk group.

Figure 6

Bioinformatics analysis of DTYMK in HCC. (A) Gene expression profiles of DTYMK in the HCCDB database. (B) Representative immunohistochemistry (IHC) images from the HPA with the DTYMK antibody. (C) The expression level of DTYMK was positively correlated with tumor stage in HCC patients. (D) The expression level of DTYMK was positively correlated with tumor grade in HCC patients. (E) Overall survival analysis of DTYMK in GEPIA. (F) Disease free survival analysis of DTYMK in GEPIA. *** represents p < 0.001, ** represents P < 0.01.

Figure 7

Validation of DTYMK’s expression and function. (A) The expression levels of DTYMK in HCC and adjacent noncancer tissues were evaluated by Western blot (n=47). (B) Statistical analysis of relative DTYMK levels in HCC tissues compared to normal tissue controls (n= 47). (C, D) Transfection efficiency was verified after transfection of siDTYMK or negative control siRNA. (E) HCC cell viability was evaluated with CCK-8 assays. (F) EdU assay showed change of proliferative rate after transfection with siDTYMK. (G) The number of HCC cell colonies was reduced after transfection with siDTYMK. *** represents p < 0.001, ** represents P < 0.01, * represents P < 0.05.

Figure 8

(A, B) Transwell assays were used to detect HCC cells invasion and migration. (C) Effects of DTYMK knockdown on HCC cell migration, as evaluated by wound healing assay. (D) Cell cycle was arrested in G0/G1 phase after transfection with siDTYMK in HCC cells. *** represents p < 0.001, ** represents P < 0.01, * represents P < 0.05.

Figure 9

Prognostic circRNA-lncRNA-miRNA-mRNA ceRNA Network in HCC.