Identification of circRNA-lncRNA-miRNA-mRNA Competitive Endogenous RNA Network as Novel Prognostic Markers for Acute Myeloid Leukemia

Abstract

Background: Acute myeloid leukemia (AML) is one of the most common malignant and aggressive hematologic tumors, and its pathogenesis is associated with abnormal post-transcriptional regulation. Unbalanced competitive endogenous RNA (ceRNA) promotes tumorigenesis and progression, and greatly contributes to tumor risk classification and prognosis. However, the comprehensive analysis of the circular RNA (circRNA)-long non-coding RNA (lncRNA)-miRNA-mRNA ceRNA network in the prognosis of AML is still rarely reported.

Method: We obtained transcriptome data of AML and normal samples from The Cancer Genome Atlas (TCGA), Genotype-tissue Expression (GTEx), and Gene Expression Omnibus (GEO) databases, and identified differentially expressed (DE) mRNAs, lncRNAs, and circRNAs. Then, the targeting relationships among lncRNA-miRNA, circRNA-miRNA, and miRNA-mRNA were predicted, and the survival related hub mRNAs were further screened by univariate and multivariate Cox proportional hazard regression. Finally, the AML prognostic circRNA-lncRNA-miRNA-mRNA ceRNA regulatory network was established.

Results: We identified prognostic 6 hub mRNAs (TM6SF1, ZMAT1, MANSC1, PYCARD, SLC38A1, and LRRC4) through Cox regression model, and divided the AML samples into high and low risk groups according to the risk score obtained by multivariate Cox regression. Survival analysis verified that the survival rate of the high-risk group was significantly reduced (p < 0.0001). The prognostic ceRNA network of 6 circRNAs, 32 lncRNAs, 8 miRNAs, and 6 mRNAs was established according to the targeting relationship between 6 hub mRNAs and other RNAs.

Conclusion: In this study, ceRNA network jointly participated by circRNAs and lncRNAs was established for the first time. It comprehensively elucidated the post-transcriptional regulatory mechanism of AML, and identified novel AML prognostic biomarkers, which has important guiding significance for the clinical diagnosis, treatment, and further scientific research of AML.

Keywords: acute myeloid leukemia; ceRNA network; prognosis.

Figure 1

Identification and functional analysis of differentially expressed mRNAs (DEmRNAs). (a,b) Heatmap of acute myeloid leukemia (AML) DEmRNAs analyzed from The Cancer Genome Atlas (TCGA), Genotype-tissue Expression (GTEx), and Gene Expression Omnibus (GEO) databases. Red color represents increased expression, blue color represents decreased expression. The darker the color is, the greater the difference of mRNAs expression. (c) Bar diagram the GO function annotation result of common DEmRNAs. (d) Bar diagram of the Kyoto Encyclopedia of Genes and Genomes (KEGG) function enrichment of common DEmRNAs.

Figure 2

Identification of DElncRNAs and prediction of targeted miRNAs. (a) Heatmap of AML DElncRNAs from TCGA and GTEx databases. (b) Volcano plot of AML DElncRNAs from TCGA and GTEx databases. (c) Prediction of DElncRNAs targeting miRNAs. The interaction network was constructed consisting of 58 lncRNAs and 85 miRNAs.

Figure 3

Identification of DEcircRNAs and prediction of targeted miRNAs. (a) Heatmap of AML DEcircRNAs from GEO databases. (b) Structure annotation of circRNAs. (c) Venn diagram of the intersection of pre-miRNAs. A total of 49 miRNAs were commonly targeted by both lncRNAs and circRNAs. (d) Flow chart of common pre-miRNAs prediction. (e) Prediction of circRNA-targeted miRNAs. The interaction network was constructed with 6 circRNAs and 323 miRNAs.

Figure 4

Identification of hub prognostic mRNAs through Cox regression and survival analysis. (a) The pre-mRNAs and common DEmRNAs were intersected, and 69 hub mRNAs were preliminary obtained. (b) Survival curve of patients in different risk groups. The patients were divided into high and low risk groups with median risk value as the boundary. Kaplan–Meier survival analysis showed that the survival rate of patients in the high-risk group was lower (p < 0.0001). (c) Receiver operating characteristic (ROC) curve was drawn to verify the accuracy of Cox survival analysis model, and the area under the curve for predicting 3-year survival rate was 0.784, indicating the high predictive power. (d) Heatmap of 6 hub mRNAs’ expression between high and low risk samples. (e) Correlation diagram of 6 hub mRNAs, red color represents positive correlation, blue color represents negative correlation, and the size of the circle reflects the p-value. (f) Expression of six hub mRNAs between AML and normal control samples.

Figure 5

Identification of hub prognostic mRNAs’ expression difference between different AML subtypes and cytogenetic risk groups and construction of nomogram. (a) The expressions of the 6 mRNAs in different subtypes were significantly different, and the expression distribution trend are obvious differences among these mRNAs. (b) The distribution of 6 mRNAs in different cytogenetic risk groups was significantly different. (c) Nomogram was drawn to establish a more refined scoring system to evaluate the impact of hub mRNAs on prognosis.

Figure 6

Establishment of AML prognostic competitive endogenous RNA (ceRNA) regulatory network. (a) ceRNA network construction process. Determined the miRNAs, long non-coding RNAs (lncRNAs), and circular RNAs (circRNAs) in ceRNA network based on 6 prognostic hub mRNAs. (b) The number of each RNAs in ceRNA network. There were 6 circRNAs, 32 lncRNAs, 8 miRNAs, and 6 mRNAs. (c) Interaction diagram of multiple endogenous RNAs in ceRNA network.

Figure 7

Molecular mechanism of the AML hub prognostic ceRNA network.

Figure 8

Analysis flow chart of this study.