Predictive Prognostic Model for Hepatocellular Carcinoma Based on Seven Genes Participating in Arachidonic Acid Metabolism

Abstract

Background: The occult onset and rapid progression of hepatocellular carcinoma (HCC) lead to an unsatisfactory overall survival (OS) rate. Established prognostic predictive models based on tumor-node-metastasis staging and predictive factors do not report satisfactory predictive efficacy. Arachidonic acid plays pivotal roles in biological processes including inflammation, regeneration, immune modulation, and tumorigenesis. We, therefore, constructed a prognostic predictive model based on seven genes linked to arachidonic acid metabolism, using samples of HCC patients from databases to analyze the genomic profiles. We also assessed the predictive stability of the constructed model.

Methods: Sample data of 365 patients diagnosed with HCC were extracted from The Cancer Genome Atlas (TCGA, training set) and HCCDB18, GSE14520, and GSE76427 databases (validation sets). Patient samples were clustered using ConsensusClusterPlus analysis based on the expression levels of 12 genes involved in arachidonic acid metabolism that were significantly associated with HCC prognosis. Differentially expressed genes (DEGs) within different clusters were distinguished and compared using WebGestaltR. Immunohistochemistry (IHC) analysis was performed using a human HCC tissue microarray (TMA). Tumor immune microenvironment assessment was performed using ESTIMATE, ssGSEA, and TIDE.

Results: Samples of patients with HCC were classified into three clusters, with significant differences in OS. Cluster 2 showed the best prognosis, whereas cluster 1 presented the worst. The three clusters showed significant differences in immune infiltration. We then performed Cox and LASSO regression analyses, which revealed CYP2C9, G6PD, CDC20, SPP1, PON1, TRNP1, and ADH4 as prognosis-related hub genes, making it a simplified prognostic model. TMA analysis for the seven target genes showed similar results of regression analyses. The high-risk group showed a significantly worse prognosis and reduced immunotherapy efficacy. Our model showed stable prognostic predictive efficacy.

Conclusions: This seven-gene-based model showed stable outcomes in predicting HCC prognosis as well as responses to immunotherapy.

Keywords: arachidonic acid metabolism; hepatocellular carcinoma; immunotherapy responses; prognosis; tumor microenvironment.

FIGURE 1

Analysis of HCC subtypes based on arachidonic acid metabolism. (A–C) Consensus clustering results of TCGA‐LIHC samples. (D) Principal component analysis distribution demonstrating clear boundaries for the three HCC subtypes in both TCGA‐LIHC and HCCDB18‐LIHC samples. (E) Kaplan–Meier survival analysis showing significant differences in OS for each subtype. HCC: hepatocellular carcinoma; LIHC: liver hepatocellular carcinoma; OS: overall survival; TCGA: The Cancer Genome Atlas.

FIGURE 2

Analysis of the expression of prognosis‐related genes and immune conditions of HCC subtypes. (A) Heatmap showing the expression levels of 12 prognosis‐related genes in each TCGA‐LIHC subtype. (B, C) Immune cell infiltration condition in each subtype. (D) Analyses of acquired and innate immune responses in each subtype. HCC: hepatocellular carcinoma; LIHC: liver hepatocellular carcinoma; TCGA: The Cancer Genome Atlas. *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 3

Construction of the seven‐gene–based HCC prognostic model. (A–D) DEGs among subtypes pairwise. A total of 63 common DEGs were identified. (E) A total of 32 common DEGs significantly associated with HCC prognosis, 11 of which were risk factors while 21 were protective factors. (F–H) LASSO regression model used for compressing the genes in the HCC prognostic model. Seven genes (CYP2C9, G6PD, CDC20, SPP1, PON1, TRNP1, and ADH4) were selected as hub genes. DEG: differentially expressed gene; HCC: hepatocellular carcinoma; LASSO: least absolute shrinkage and selection operator.

FIGURE 4

Evaluation and validation of the predictive efficacy of the seven‐gene–based model, and correlation between risk score and gene mutation. (A–D) Kaplan–Meier survival analysis results confirmed significant differences in the OS of patients with HCC in the TCGA‐LIHC training dataset and HCCDB18, GSE14520, and GSE76427 validation datasets. Time‐dependent ROC curve (timeROC) showing satisfactory prediction accuracy of the model on the training and validation datasets. (E) Top 20 mutated genes in both the high‐ and low‐risk groups. (F) Risk group and TMB levels having significant effects on patient OS. HCC: hepatocellular carcinoma; LIHC: liver hepatocellular carcinoma; OS: overall survival; ROC: receiver operating characteristic; TCGA: The Cancer Genome Atlas; TMB: tumor mutational burden.

FIGURE 5

Clinical pathological characteristics of patients with different risk scores, and the results of TMA analysis, and Kaplan–Meier survival analysis based on clinical HCC samples. (A) Patients with different risk scores having significantly different T stage and tumor stage and grade. (B) Patients in different Kyoto Encyclopedia of Genes and Genomes arachidonic acid metabolism pathway score groups having significantly different T stage and tumor stage and grade. (C) IHC images showing the expression of seven proteins in HCC tumor tissues and adjacent normal tissues in TMA analysis. (D) Kaplan–Meier survival analysis indicating that patients with lower risk scores had better outcomes than those with higher risk scores. HCC: hepatocellular carcinoma; IHC: immunohistochemistry; TMA: tissue microarray. *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 6

Results of Kaplan–Meier survival analysis based on samples of patients with HCC who received immunotherapy. (A–D) High‐ and low‐risk groups having significant difference in OS in the IMvigor210, GSE135222, and GSE78220 samples but not in the GSE91061 samples. The PD/SD ratio in the high‐risk group was higher than that in the low‐risk group, indicating a lower likelihood of patients within the high‐risk group to benefit from immunotherapy. HCC: hepatocellular carcinoma; OS: overall survival. *p < 0.05.

FIGURE 7

Risk score combined with clinical pathological characteristics further improves the HCC prognostic model. (A, B) Univariate and multivariate Cox analyses of RiskScore and clinicopathological features. (C) A decision tree constructed using risk group and T stage as pivotal parameters categorizing the samples into four different subtypes. (D–F) Samples from each subtype displaying significant differences in survival analysis. (G) A nomogram constructed based on risk scores and other clinicopathological features revealing risk score as the most important prognostic factor for HCC prognosis. (H) Calibration curve for the analysis of the model's accuracy. (I) DCA to assess the reliability of the prognostic model. DCA: decision curve analysis; HCC: hepatocellular carcinoma. *p < 0.05.