Integrated analysis reveals key circRNA-mediated regulatory axes related to prognosis and immune infiltration in lung adenocarcinoma

Abstract

Background: For lung adenocarcinoma (LUAD), the competitive endogenous RNA (ceRNA) networks mediated by circular RNAs (circRNAs) have been partially constructed. However, a mass of networks still need to be thoroughly investigated to uncover novel regulatory axes in LUAD.

Methods: Clinical information along with transcriptome data were obtained from open databases. The circRNA-mediated ceRNA subnetworks and a risk score were constructed through layer-by-layer screening and validating. Immune infiltration analysis was performed by CIBERSORT. Quantitative real-time PCR, immunohistochemical analysis, and dual-luciferase reporter assays confirmed the expression and relationships of hub genes.

Results: The expression of 9 circRNAs, 81 miRNAs, and 952 mRNAs significantly varied in LUAD tissues. Subsequently, 63 miRNA-mRNA interactions and 3 circRNA-miRNA interactions were employed to generate a ceRNA network. Two prognostic subnetworks mediated by hsa_circ_0072088 and hsa_circ_0049271 were obtained following hub genes screening and survival analysis. Then, a risk score consisting of MMP14 and DCN was successfully constructed and verified using a different dataset. Among the high-risk group, more deaths and poor prognosis occurred. The distribution of immune infiltrating cells varied between high- and low-risk groups, and they were correlated with both the expression of DCN and MMP14 and the risk score.

Conclusions: The two key regulatory axes, hsa_circ_0072088/hsa-miR-532-3p/MMP14 and hsa_circ_0049271/hsa-miR-224-5p/DCN, might be involved in carcinogenesis, prognosis and immune infiltration of LUAD.

Keywords: Immune infiltration; Lung adenocarcinoma; Prognosis; ceRNA network; circRNA.

Fig. 1

The workflow of this study

Fig. 2

Differentially expressed circRNAs, miRNAs, and mRNAs in LUAD by mining GEO and TCGA database. (A) The PCA plots of four datasets (GSE101586, GSE101684, TCGA_LUAD_mRNA, and TCGA_LUAD_miRNA). (B) Venn diagram of the common differentially expressed circRNAs between GSE101586 and GSE101684. (C) Heatmap of the 9 differentially expressed circRNAs in GSE101586 and GSE101684. (D, E) Volcano plot of differentially expressed miRNAs and mRNAs in TCGA LUAD project, respectively

Fig. 3

The circRNA-miRNA-mRNA regulatory network in LUAD. (A) The schematic structure of 3 circRNAs. (B) The ceRNA network consists of 3 circRNAs nodes, 3 miRNAs nodes and 63 mRNAs nodes. Red represents up-regulated expression, while blue represents down-regulated expression. (C, D) The Go and KEGG pathway enrichment analysis of differentially expressed mRNAs in ceRNA network, respectively

Fig. 4

Screening hub genes through the PPI network analysis and exploring prognostic value of hub genes in LUAD. (A) The PPI network of differentially expressed mRNAs in ceRNA network. Red represents up-regulated genes, while blue represents down-regulated expression. (B) Two subnetworks were retrieved from the whole PPI network. (C) The circRNA-miRNA-hub gene networks consist of 2 circRNAs, 2 miRNAs, and 8 hub genes were established. (D) Analyzing prognostic value of 8 hub genes (ATF3, MAFF, FOSB, NR4A1 (NGFIB), DCN, COL1A1, MMP14, and MET) in LUAD by the online Kaplan-Meier plotter

Fig. 5

Construction and validation of the risk score. (A) Venn diagram of the common prognostic genes screened by Kaplan-Meier plotter, TCGA database, and multivariate COX regression analysis. (B-D) The distribution of risk score, survival status and MMP14 and DCN expression between high-risk and low-risk groups in TCGA cohort. (E, F) Kaplan-Meier analysis of high-risk and low-risk groups in TCGA cohort and GSE68465, respectively. (G, H) Univariate and multivariate cox regression analyses of risk score and clinical features including age, gender, and TNM staging, respectively

Fig. 6

Immune cell infiltration analysis. (A) The landscape of 22 immune cell infiltrations in LUAD. (B) The boxplot of differences in 22 immune cell infiltrations between high risk and low risk groups. *P < 0.05, **P < 0.01, ***P < 0.001. (C) The correlations between risk score and immune cell infiltrations in LUAD. (D) The correlations between risk score and immune checkpoint genes (PD-L1, PD1, CTLA4, and LAG3) in LUAD

Fig. 7

Validating the differentially expressed circRNAs, miRNAs, and mRNAs by qPCR and immunohistochemical analysis. (A-F) The relative expression level of has_circ_0072088, has-miR-224-5p, MMP14, has_circ_0049271, has-miR-532-3p, and DCN in LUAD cells compared to those in normal BEAS-2B cell. ***P < 0.001. (G, H) Immunohistochemical analysis of DCN and MMP14 in normal tissues and LUAD tissues via the human protein atlas database

Fig. 8

Validation of the two key regulatory axes. (A) The binding sites between has_circ_0072088, has- has-miR-532-3p, and MMP14. (B) The binding sites between has_circ_0049271, miR-224-5p, and DCN. (C-D) Validation the target relationships of the two key regulatory axes via dual-luciferase reporter assays. (E) The relative expression of MMP14 was measured after co-transfection. (F) The relative expression of DCN was measured after co-transfection. **P < 0.01, ***P < 0.001