An angiogenesis-related long noncoding RNA signature correlates with prognosis in patients with hepatocellular carcinoma

Abstract

Hepatocellular carcinoma (HCC) is one of the most prevalent and lethal cancers worldwide. Neovascularization is closely related to the malignancy of tumors. We constructed a signature of angiogenesis-related long noncoding RNA (lncRNA) to predict the prognosis of patients with HCC. The lncRNA expression matrix of 424 HCC patients was downloaded from The Cancer Genome Atlas (TCGA). First, gene set enrichment analysis (GSEA) was used to distinguish the differentially expressed genes of the angiogenesis genes in liver cancer and adjacent tissues. Next, a signature of angiogenesis-related lncRNAs was constructed using univariate and multivariate analyses, and receiver operating characteristic (ROC) curves were used to assess the accuracy. The signature and relevant clinical information were used to construct the nomogram. A 5-lncRNA signature was highly correlated with overall survival (OS) in HCC patients and performed well in evaluations using the C-index, areas under the curve, and calibration curves. In summary, the 5-lncRNA model can serve as an accurate signature to predict the prognosis of patients with liver cancer, but its mechanism of action must be further elucidated by experiments.

Keywords: GSEA; angiogenesis; mRNA signature.

Figure 1. GSEA of starvation-related gene sets

(A–C) Enrichment map of one angiogenesis-related gene set between liver cancer and paired adjacent tissues identified by GSEA. (D) Heat map of 123 genes in liver cancer and normal tissue.

Figure 2. Construction and validation of the angiogenesis-related lncRNA prognostic signature in patients with HCC

(A and B) The distribution and scatter plots of risk scores based on angiogenesis-related lncRNA prognostic characteristics of high- and low-risk HCC patients show the correlation between survival time and risk scores based on the angiogenesis-related lncRNA prognostic characteristics of HCC patients. (C) The KM curves show that the survival time of patients with low risk scores based on the angiogenesis-related lncRNA prognostic signature is significantly longer than that of patients with high-risk scores. (D) Time-dependent ROC curves at 1, 3 and 5 years show the accuracy of the signature in predicting the survival times (prognosis) of HCC patients from the TCGA database.

Figure 3. Correlation of our signature with the clinicopathological characteristics of HCC

(A) Age (≥ 65 vs. <65 years; P=0.4375); (B) Sex (male vs. female; P=0.920); (C) Tumor grade (grade 1-2 vs. 3-4; P=0.032); (D) Tumor stage (stage 3-4 vs. 1-2; P=0.005); (E) Alpha-fetoprotein(AFP) (≤20 vs. >20 ng/ml; P=0.291). (F) Vascular invasion (none vs. micro vs. macro; P=0.030).

Figure 4. Estimation of the prognostic accuracy of the signature and other clinicopathological variables

(A and B) In the training and validation sets, univariate and multivariate analyses were performed for risk scores and each clinical feature. (C) The time-dependent ROC curves of risk scores and clinical features were predicted in the training and validation sets at 5 years.

Figure 5. Survival rates of high- and low-risk HCC patients stratified by different clinicopathological characteristics

(A–H) Kaplan–Meier survival curve analysis shows the overall survival (OS) rates of high- and low-risk HCC patients from the TCGA database stratified by age (≤65 vs. >65 years), sex (male vs. female), tumor grade (high grade vs. low grade), stage (stages I and II vs. stages III and IV) and T stage (T1/T2 vs. T3/T4).

Figure 6. An established nomogram for predicting OS

(A) Construction and validation of the prognostic nomogram with the starvation-related mRNA prognostic signature risk score as one of the parameters in the training set. (B and C) Calibration curves of the nomogram for the prediction of 3- and 5-year OS.

Figure 7. Construction of the angiogenesis-related lncRNA–mRNA coexpression network and functional enrichment analyses

(A) Diagrammatic representation of the angiogenesis-related lncRNA–mRNA network shows 27 lncRNA–mRNA coexpression. The red circles correspond to angiogenesis-related lncRNAs, and the blue diamonds correspond to the mRNAs. (B) The Sankey diagram shows the connection degree between the 27 mRNAs and 5 angiogenesis-related lncRNAs (risk/protective). (C) Risk factor score, clinical features and expression of nine mRNAs in each patient. (D–F) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway, hallmark and Reactome analyses show the enriched signaling pathways associated with the high-risk group.

Figure 8. Correlation between signature and immune cells

(A and B) GO analysis results showing the functions and enriched signaling pathways associated with the signature. (C–H) Correlation between immune cells and risk score.