Identification of metabolism-related long non-coding RNA (lncRNA) signature predicts prognosis and immune infiltrates in hepatocellular carcinoma

Abstract

Background: Cancer-associated metabolic reprogramming promotes cancer cell differentiation, growth, and influences the tumor immune microenvironment (TIME) to promote hepatocellular carcinoma (HCC) progression. However, the clinical significance of metabolism-related lncRNA remains largely unexplored.

Methods: Based on The Cancer Genome Atlas (TCGA) Liver hepatocellular carcinoma (LIHC) dataset, we identified characteristic prognostic long non-coding RNAs (lncRNAs) and construct metabolism-related lncRNA prognostic signature for HCC. Gender, age, grade, stage and TP53 status were used as covariates were used to assess the prognostic capacity of the characteristic lncRNA signature. Subsequently, the molecular and immune characteristics and drug sensitivity in metabolism-related lncRNA signature defined subgroups were analyzed.

Results: We identified 34 metabolism-related lncRNAs significantly associated with the prognosis of HCC (P<0.05). Subsequently, we constructed a multigene signature based on 9 characteristics prognostic lncRNAs and classified HCC patients into high- and low-risk groups based on cutoff values. We found the lncRNA signature [hazard ratio (HR) =3.55 (2.44-5.15), P<0.001] to be significantly associated with survival. The receiver operating characteristic curve (ROC) curves area under the curve (AUC) values for 1-, 3-, and 5-year survival were 0.811, 0.773, and 0.753, respectively. In univariate and multivariate Cox regression analyses, prognostic characteristic lncRNAs were the most crucial prognostic factor besides the stage. The prognostic signature was subsequently validated in the test set. In addition, Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and gene set enrichment analysis (GSEA) analyses revealed potential biological features and signaling pathways associated with the prognostic signature. We constructed a nomogram including risk groups and clinical parameters (age, gender, grade, and stage). Calibration plots and decision curve analysis (DCA) showed that our nomogram had a good predictive performance. Finally, we found reduced expression of immune-activated cells in the high-risk group.

Conclusions: The metabolism-related lncRNA signature is a promising biomarker to distinguish the prognosis and an immune characteristic in HCC.

Keywords: Hepatocellular carcinoma (HCC); The Cancer Genome Atlas (TCGA); cancer-related metabolic reprogramming; long non-coding RNA (lncRNA); tumor immune microenvironment (TIME).

Figure 1

Acquisition of hub metabolism-related lncRNAs and biological functions based on TCGA database in LIHC patients. (A) Flowchart for bioinformatics analysis of publicly available data from TCGA databases. (B,C) The heatmap showed differential expression in metabolism-related genes (B) and metabolism-related lncRNAs (C). Red means up-regulated, green means down-regulated, and grey means no significance. (D,E) GO, and KEGG analysis demonstrated the biological functions of hub metabolism-related lncRNAs. lncRNA, long non-coding RNA; TCGA, The Cancer Genome Atlas; LIHC, liver hepatocellular carcinoma; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Figure 2

Construction of the risk score model in the training cohort. (A) Univariate Cox analysis of 34 hub metabolism-related lncRNAs. (B) LASSO coefficient profiles. The two dotted vertical lines indicate the optimal values using the minimum and 1-SE criteria. (C) Candidate metabolism-related lncRNAs from the univariate Cox regression analysis were filtered by the LASSO algorithm. Each colored line represents a lncRNA, and the axis above indicates the number of nonzero coefficients at the current lambda(λ). lncRNA, long non-coding RNA; LASSO, least absolute shrinkage and selection operator.

Figure 3

Evaluation of the prognostic signature model in different cohorts. (A) Kaplan-Meier survival curve analysis showed the relationship of survival time of LIHC patients between high- and low-risk scores (upper). The time-dependent ROC curves for 12-, 36-, and 60-month OS predictions by the prognostic signature model in the training cohort (middle). Cases were ranked according to the risk score, and the correlation between survival time and risk scores was demonstrated using scatter plots. Heatmap showed the correlation between characteristic lncRNAs and the prognostic signature model (bottom). The same approach was applied to the test cohort (B) and the entire cohort (C). lncRNA, long non-coding RNA; LIHC, liver hepatocellular carcinoma; ROC, Receiver Operating Characteristic; OS, overall survival; TPR, true positive rate; FPR, false positive rate.

Figure 4

Assess the accuracy and clinical applicability of the prognostic signature model in LIHC patients. (A) Univariate and multivariate Cox regression analysis showed the correlation between OS and various clinicopathological parameters such as age, gender, grade, stage, TP53, and characteristic lncRNAs prognostic signature risk group. The stage and prognostic signature risk group significantly correlated with prognosis (P<0.001). (B) The prognostic nomogram with characteristic lncRNAs prognostic signature risk group and clinicopathological features were constructed to predict the prognosis of LIHC patients. (C) Calibration curve for nomogram-predicting 1-, 3-, and 5-year OS. The X-axis is nomogram-predicted survival probability, and the Y-axis is observed survival probability. (D) Decision curve analysis for nomogram and stage. (E) A net reduction in intervention per 100 patients results from the two risk prediction models. lncRNA, long non-coding RNA; LIHC, liver hepatocellular carcinoma; OS, overall survival; DCA, decision curve analysis.

Figure 5

Stratification analysis based on clinical parameters. Kaplan-Meier analysis demonstrated the survival of LIHC patients according to different ages, gender, stage, and TP53 status based on different prognostic signature risk groups. The respective P values are displayed under the survival curves. LIHC, liver hepatocellular carcinoma.

Figure 6

Association between prognostic signature risk score and TIICs. (A) Heatmap for metabolism-related lncRNAs prognostic signature risk score, MUC4, TMB, and clinicopathological features based on CIBERSORT. TP53, status, time, stage, grade, gender, and age are patient annotations. (B) ssGSEA for the association between TIICs and related functions in different risk groups. Adjusted P values were showed as: ns, not significant; *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. TIICs, tumor-infiltrating immune cells; TMB, tumor mutational burden; ssGSEA, single-sample gene set enrichment analysis; CIBERSORT, Cell-type Identification by Estimating Relative Subsets of RNA Transcripts.

Figure 7

Molecular characteristics of patients in the high- and low-risk group. (A,B) GSEA analysis showed that the high-risk group was negatively associated with metabolic-related pathways and positively associated with tumor proliferation-related pathways. (C) IC50 values of cytotoxic chemotherapeutic agents between different risk groups based on the pRRophetic algorithm. The high-risk group had significantly lower IC50 values for cisplatin and sorafenib. (D) Significantly mutated genes in LIHC patients in the high- and low-risk group. The top 10 mutated genes in 2 groups ranked by mutation rate are shown. The mutation rate is shown on the right, and the mutation counts are shown on the top. Adjusted P values are shown as: *, P<0.05; **, P<0.01; ****, P<0.0001. GSEA, gene set enrichment analysis; LIHC, liver hepatocellular carcinoma.