A novel cuproptosis-related prognostic gene signature and validation of differential expression in hepatocellular carcinoma

Abstract

Background: Cuproptosis is a newly discovered form of programmed cell death, which is characterized by accumulation of intra-cellular copper ion leading to the aggregation of lipoproteins and destabilization of Fe-S cluster proteins in mitochondrial metabolism, thereby affecting the prognosis of patients with cancer. However, the role of cuproptosis-related genes (CRGs) in hepatocellular carcinoma (HCC) remains elusive. Methods: Mutation signature, copy number variation and the expression of 10 CRGs were assessed in HCC from TCGA-LIHC dataset. ICGC-LIRI-JP dataset was used as further validation cohort. The least absolute shrinkage and selection operator (LASSO) was used to construct the prognostic model. Kaplan Meier curves, time-ROC curves, nomogram, univariate and multivariate Cox regression were utilized to evaluate the predictive efficacy of CRGs-score. Immune infiltration was analyzed by CIBERSOFT, ssGSEA algorithm, and TIMER database. The expression of prognostic CRGs was validated by qPCR both in-vitro and in-vivo. Drug sensitivity analysis was performed by pRRophetic. Results: All of the CRGs were differentially expressed in HCC and 5 out of them (CDKN2A, DLAT, GLS, LIPT1, MTF1) correlated with patient survival. These signature genes were selected by LASSO analysis to establish a prognosis model to stratify HCC patients into high and low CRGs-score subgroups. High CRGs-score was associated with a worse prognosis. Subsequently, univariate and multivariate Cox regression verified that CRGs-score was an independent cancer risk factor that correlated with clinical factors including stage and grade. Nomogram integrating the CRGs-score and clinical risk factors performed well to predict patient survival. Immune infiltration analysis further revealed that the expression of immune checkpoint genes was significantly enhanced in high CRGs-score group, especially PD-1 and PD-L1. An independent validation cohort (ICGC) confirmed that CRGs-score as a stable and universally applicable indicator in predicting HCC patient survival. Concordantly, the expression of five confirmed signature genes were also differentially expressed in human HCC cell lines and mouse HCC model. In addition, we also analyzed the sensitivity of 10 clinical targeted therapies between high and low CRGs-score groups. Conclusion: This study elucidated the role of dysregulated CRGs in HCC cohort, with validation with in-vitro and in-vivo models. The CRGs-score might be applied as a novel prognostic factor in HCC.

Keywords: cuproptosis; drug sensitivity; hepatocellular carcinoma; prognostic model; tumor immunity.

FIGURE 1

Landscape of genetic alterations of CRGs in HCC. (A) The landscape of mutation profiles of 364 HCC patients from TCGA-LIHC cohort. The upper barplot represented the mutation burden. The right barplot showed mutation frequency individually. (B) The mutation summary plot of CRGs. The barplot displayed the variant classification, variant types, SNV class, and top mutated CRGs. (C) The location of CNV alteration of 10 CRGs on 23 chromosomes. (D) The CNV frequency of CRGs in TCGA cohort. The height of the column showed the proportions of gain or loss variations.

FIGURE 2

Functional enrichment analysis of CRGs in HCC. (A) The top 20 enriched items of CRGs in gene ontology analysis in biological process. The size of circles represented the number of genes enriched. (B) The top 10 enriched pathways of CRGs in KEGG database.

FIGURE 3

Determination the differentially expressed CRGs and prognostic signatures in HCC. (A) The expression of 10 CRGs in HCC (n = 368) and normal liver tissues (n = 50) in TCGA-LIHC cohort. Tumor was shown in orange, while normal liver was shown in blue (t-test, *p < .05; **p < .01; ***p < .001; ****p < .0001; ns, not statistically significant). (B) Univariate Cox regression analysis of CRGs. The right boxplots represented the Hazard ratio with 95% confidence interval, the p-value was calculated by univariate cox regression.

FIGURE 4

Construction of a prognostic CRGs model. (A) The LASSO-COX model screened out prognostic CRGs and carried out 10-fold cross-validation. The λ value was confirmed as .0014 where the optimal lambda resulted in five non-zero coefficients. (B) LASSO coefficient profiles of the five CRGs. (C) The survival curves for the different CRGs-score subgroups with the cut-off value 2.309 among 368 HCC patients (Log-rank test, p = .00018, HR = 1.94). The mean OS for the high and low CRGs-score group were 41.8 and 81.7 months, respectively. (D) The distribution of CRGs-score, survival status, and the expression of five prognostic CRGs in HCC. (E) The time-dependent receiver operating characteristic (ROC) analysis of CRGs-score. The AUC values were .72, .66, .62 at 1 year, 3 years, and 5 years, respectively.

FIGURE 5

The clinical features of the CRGs-score model. The forest plot for univariate Cox (A) and multivariate Cox regression (B) considering clinical indicators and CRGs-score in HCC cohort. (C) Nomogram incorporating age, gender, stage, grade and CRGs-score was a predictor of 1-, 3-, and 5-year overall survival probabilities in HCC patients. (D) Calibration curves of 1-, 3-, and 5-year of survival outcomes.

FIGURE 6

The clinical application of CRGs-score. The KM curves of (A) DFI (B) DSS (C) PFI time between low and high CRGs-score subgroups in the TCGA cohort. (D) CRGs-score between the early stage and the advanced stage subgroups. (E) CRGs-score between the low-grade and the high-grade subgroups.

FIGURE 7

The association between the CRGs and immune microenvironment in HCC. (A) The proportion of immune infiltrating cells between the low and the high CRGs-score subgroups by CIBERSOFT analysis. (B) The enrichment score of thirteen immune pathways between the low and the high CRGs-score groups. (C) The boxplots were utilized for visualizing the expression of seven immune checkpoint genes in the low and the high CRGs-score subgroups.

FIGURE 8

Immune infiltration analysis of CRGs signatures. Correlation between (A) CDKN2A (B) DLAT (C) GLS (D) LIPT1, and (E) MTF1 expression and immune infiltration in HCC from the TIMER database.

FIGURE 9

An independent validation of the CRGs-score with the ICGC cohort. (A) The KM curves for overall survival of 240 HCC patients between low and high CRGs-score subgroups (Log-rank test, p = .015, HR = 2.2). (B) The time-dependent ROC analysis of CRGs-score. The AUC values were .65, .68, .68 at 1 year, 3 years, and 5 years, respectively. (C) The increasing of CRGs-score, survival status, and heatmap of five prognostic CRGs in the ICGC cohort.

FIGURE 10

Validation of the expression of the five prognostic CRGs in-vitro HCC cell lines model. (A–E) RT-qPCR was performed to detect the expression of CDKN2A, DLAT, GLS, LIPT1 and MTF1 in normal liver cell (LO2) and HCC cells (Huh7 and MHCC97H) (Data are presented as mean ± SEM, t-test, *p < .05; **p < .01; ***p < .001; ****p < .0001; ns: not statistically significant).

FIGURE 11

Validation of the expression of the five prognostic CRGs in vivo HCC mouse model. (A) HCC mouse model diagram: C57BL/6 J male mice were hydrodynamic injected with normal saline or AKT/NRas/SB plasmids, respectively. (B) Representative mouse liver tissues of normal (left) and HCC (middle) and liver/body ratio between two groups (right). (C–G) The mRNA expression of CDKN2A, DLAT, GLS, LIPT1 and MTF1 in normal and HCC liver. (H–I) Immunohistochemistry (IHC) staining of CD8 between normal (H) and HCC (I) liver tissues. (J) The statistical analysis of CD8⁺ T-cell in filed between normal and HCC group.

FIGURE 12

The IC50 of 10 commonly used chemotherapeutic drugs and targeted drugs in the low and high CRGs-score subtypes of HCC. (A–J) The IC50 of Elesclomol, Sorafenib, Temsirolimus, 5-Fluorouracil, Sunitinib, Pazopanib, Gemcitabine, Erlotinib, Bleomycin, Axitinib between low and high CRGs-score subgroups.