A new ferroptosis-related genetic mutation risk model predicts the prognosis of skin cutaneous melanoma

Abstract

Background: Ferroptosis is an iron-dependent cell death mode and closely linked to various cancers, including skin cutaneous melanoma (SKCM). Although attempts have been made to construct ferroptosis-related gene (FRG) signatures for predicting the prognosis of SKCM, the prognostic impact of ferroptosis-related genetic mutations in SKCM remains lacking. This study aims to develop a prediction model to explain the relationship between ferroptosis-related genetic mutations and clinical outcomes of SKCM patients and to explore the potential value of ferroptosis in SKCM treatment. Methods: FRGs which significantly correlated with the prognosis of SKCM were firstly screened based on their single-nucleotide variant (SNV) status by univariate Cox regression analysis. Subsequently, the least absolute shrinkage and selection operator (LASSO) and Cox regressions were performed to construct a new ferroptosis-related genetic mutation risk (FerrGR) model for predicting the prognosis of SKCM. We then illustrate the survival and receiver operating characteristic (ROC) curves to evaluate the predictive power of the FerrGR model. Moreover, independent prognostic factors, genomic and clinical characteristics, immunotherapy, immune infiltration, and sensitive drugs were compared between high-and low-FerrGR groups. Results: The FerrGR model was developed with a good performance on survival and ROC analysis. It was a robust independent prognostic indicator and followed a nomogram constructed to predict prognostic outcomes for SKCM patients. Besides, FerrGR combined with tumor mutational burden (TMB) or MSI (microsatellite instability) was considered as a combined biomarker for immunotherapy response. The high FerrGR group patients were associated with an inhibitory immune microenvironment. Furthermore, potential drugs target to high FerrGR samples were predicted. Conclusion: The FerrGR model is valuable to predict prognosis and immunotherapy in SKCM patients. It offers a novel therapeutic option for SKCM.

Keywords: ferroptosis; genetic mutation; prognosis; single nucleotide variant; skin cutaneous melanoma; tumor immunity.

FIGURE 1

Identification of prognosis-related key FRGs in SKCM. (A) Heatmap to show the SNV landscape of the top 30 FGRs with the most frequent SNV mutations in the TCGA-SKCM cohort. (B) Forest plots showing the results of the univariate Cox regression analysis between FGRs and prognosis. (C) Pie charts depicting the proportions of wild-type and DDIT3-mutant patients. (D) Kaplan-Meier survival analysis of the wild-type and DDIT3-mutant patients.

FIGURE 2

Construction and prognostic analysis of the ferroptosis-related genetic mutation risk (FerrGR)model in the training cohort and validation cohort. (A) Lasso coefficient spectrum of 19 key FRGs in the FerrGR model (B,D) SNV heatmap and clinicopathologic features of 19 key FRGs in the (B) training cohort and (D) validation cohort. (C,E) Kaplan-Meier curves of the FerrGR model for SKCM patients with different risk groups in the (C) training cohort and (D) validation cohort. (F,G) ROC analysis of the FerrGR model compared with TMB and MSI.

FIGURE 3

Relationships between the FerrGR score and clinicopathological features. The boxplots showed whether the FerrGR score was correlated with pathological features in SKCM patients, including (A) sample type, (B) tumor stage, (C) gender, (D) Clark level, (E) T stage, (F) N stage, (G) M stage, and (H) TCGA subtype.

FIGURE 4

The connection between FerrGR score and conventional clinical characteristics. (A,B) Univariate and multivariate regression analysis of FerrGR score and clinical characteristics in prognostic value showed FerrGR score had excellent prognostic independence, (C) Prognostic nomogram for SKCM patients with factors, including Breslow depth value, age, FerrGR score and N stage, (D) Test for the proportional hazards hypothesis, (E) Calibration maps for predicting patient survival at 1, 2, and 3 years. The x-axis and y-axis represent the expected and actual survival rates of the nomogram.

FIGURE 5

Analysis of ferroptosis-related genomic variation in the high FerrRG group and low FerrRG group (A,B) waterfall plots represent mutation information of FRGs in each sample of the high FerrGR group and low FerrGR group SKCM patients, (C) Heatmap of top 30 differentially expression FRGs between the high- and the low-FerrGR groups, (D,E) The CNV mutation proportion of (D) RELA, and (E) NOX4 between groups, Amp: Gene amplification, Del: gene deletion (F) Heatmap showed the methylation sites of FRGs with top 30 significantly different methylation levels between groups.

FIGURE 6

Evaluation of immunotherapy in SKCM with FerrGR model. (A) Heatmap displaying the expression levels of 61 immune checkpoint genes in the high FerrGR group and low FerrGR group. (B) Kaplan-Meier curves of the high-FerrGR and low-FerrGR group patients in the GSE91061 dataset. (C) The proportion of immune response to immunotherapy of high- and low-FerrGR groups in the TCGA cohort. CR, complete response; PR, partial response; SD, stable disease; PD, progressive disease. (D) Kaplan - Meier survival analysis of OS among patients within each of the three indicated subgroups (Both high: FerrGR-high/TMB -high; Both low: FerrGR-low/TMB-low; Single high: FerrGR-high/TMB-low or FerrGR-low/TMB-high). (E) Kaplan-Meier survival analysis of OS among patients within each of the three indicated subgroups (Both high: FerrGR-high/MSI -high; Both low: FerrGR-low/MSI-low; Single high: FerrGR-high or MSI-high).

FIGURE 7

Tumour immune microenvironment analysis of FerrGR model. (A) The heatmap showed the correlations between FerrGR score and clinicopathological features. (B) Boxplot exhibited the distribution patterns of immune cell infiltration between two FerrGR groups. (C–G) violin plots visualizing the expression of immune checkpoint genes in the high—and Low—FerrGR groups, (C) CD80, (D) CD86, (E) CD274, (F) PDCD1LG2, (G) CTLA4 (H) Gene set enrichment analysis (GSEA) results depicting enrichment of immune-related pathways based on FerrGR score.

FIGURE 8

Potential targeted drugs prediction on basis of the FerrGR model. (A) The Venn chart showed the number of drugs in the PRISM dataset and CTRP v 2.1 databases. (B) The correlation between the AUC value of potential drugs and the FerrGR score. (C,D) The distribution of the AUC value of each potential drug from (C) RISM dataset and (D) CTRP v 2.1 database according to the FerrGR model. (E) The correlation between the IC50 value of potential drugs and the FerrGR score. (F) The distribution of the IC50 value of each potential drug on basis of the FerrGR model.