Ferroptosis-related lncRNA signature predicts prognosis and immunotherapy efficacy in cutaneous melanoma

Abstract

Purpose: Ferroptosis-related lncRNAs are promising biomarkers for predicting the prognosis of many cancers. However, a ferroptosis-related signature to predict the prognosis of cutaneous melanoma (CM) has not been identified. The purpose of this study was to construct a ferroptosis-related lncRNA signature to predict prognosis and immunotherapy efficacy in CM.

Methods: Ferroptosis-related differentially expressed genes (FDEGs) and lncRNAs (FDELs) were identified using TCGA, GTEx, and FerrDb datasets. We performed Cox and LASSO regressions to identify key FDELs, and constructed a risk score to stratify patients into high- and low-risk groups. The lncRNA signature was evaluated using the areas under the receiver operating characteristic curves (AUCs) and Kaplan-Meier analyses in the training, testing, and entire cohorts. Multivariate Cox regression analyses including the lncRNA signature and common clinicopathological characteristics were performed to identify independent predictors of overall survival (OS). A nomogram was developed for clinical use. We performed gene set enrichment analyses (GSEA) to identify significantly enriched pathways. Differences in the tumor microenvironment (TME) between the 2 groups were assessed using 7 algorithms. To predict the efficacy of immune checkpoint inhibitors (ICI), we analyzed the association between PD1 and CTLA4 expression and the risk score. Finally, differences in Tumor Mutational Burden (TMB) and molecular drugs Sensitivity between the 2 groups were performed.

Results: We identified 5 lncRNAs (AATBC, AC145423.2, LINC01871, AC125807.2, and AC245041.1) to construct the risk score. The AUC of the lncRNA signature was 0.743 in the training cohort and was validated in the testing and entire cohorts. Kaplan-Meier analyses revealed that the high-risk group had poorer prognosis. Multivariate Cox regression showed that the lncRNA signature was an independent predictor of OS with higher accuracy than traditional clinicopathological features. The 1-, 3-, and 5-year survival probabilities for CM patients were 92.7%, 57.2%, and 40.2% with an AUC of 0.804, indicating a good accuracy and reliability of the nomogram. GSEA showed that the high-risk group had lower ferroptosis and immune response. TME analyses confirmed that the high-risk group had lower immune cell infiltration (e.g., CD8+ T cells, CD4+ memory-activated T cells, and M1 macrophages) and lower immune functions (e.g., immune checkpoint activation). Low-risk patients whose disease expressed PD1 or CTLA4 were likely to respond better to ICIs. The analysis demonstrated that the TMB had significantly difference between low- and high- risk groups. Chemotherapy drugs, such as sorafenib, Imatinib, ABT.888 (Veliparib), Docetaxel, and Paclitaxel showed Significant differences in the estimated IC50 between the two risk groups.

Conclusion: Our novel ferroptosis-related lncRNA signature was able to accurately predict the prognosis and ICI outcomes of CM patients. These ferroptosis-related lncRNAs might be potential biomarkers and therapeutic targets for CM.

Keywords: cutaneous melanoma; ferroptosis; immune checkpoint; immune infiltration; long non-coding RNA.

Scheme 1

Flowchart of the analysis process.

Figure 1

FDEG and FDEL identification and functional enrichment analyses. (A) The heatmap illustrates 53 FDEGs. (B) GO analysis of the FDEGs. (C) KEGG analysis of the FDEGs. (D) Co-expression network of FDEGs and FDELs. Yellow nodes indicate FDEGs; blue arrows FDELs. (E,F) A heatmap and volcano plot show 161 FDELs and 5 studied LncRNAs.

Figure 2

Construction of the prognostic ferroptosis-related lncRNA signature. (A) Univariate Cox regression revealed that 50 of the 161 FDELs were associated with OS. (B) LASSO regression identified 8 FDELs from the 50 lncRNAs, Dynamic LASSO coefficient profiling. (C) Selection of the tuning parameter (lambda); and the dashed lines on the left and right indicate the “lambda. min” and “lambda.1se” criteria, respectively. (D) Stepwise Cox regression finally identified 5 FDELs (AATBC, AC145423.2, LINC01871, AC125807.2, and AC245041.1) to construct the risk score (P < 0.001). (E) Sankey diagram shows the associations among FDELs, FDEGs, and risk types.

Figure 3

Evaluation and validation of the lncRNA signature in the training, testing, and entire cohorts. (A) AUC of the lncRNA signature for predicting 5-year survival in the training cohort. (B) The rank of the calculated risk score in the training cohort. (C) The survival status and survival time in the training cohort. (D) Kaplan-Meier survival curves of the high-risk patients and low-risk patients in the training cohort. (E–H) The lncRNA signature was validated in the testing cohort. (I–L) The lncRNA signature was validated in the entire cohort. (M,N) PCA and t-SNE analysis of the training cohort. (O,P) PCA and t-SNE analysis of the test cohort. (Q,R) PCA and t-SNE analysis of the entire cohort.

Figure 4

Independent prognostic value of the lncRNA signature and the comprehensive nomogram. (A) Univariate Cox regression identified individual factors related to OS. (B) Multivariate Cox regression identified independent predictors of OS. Our lncRNA signature was an independent predictor of OS with a hazard ratio of 1.338 (95% CI, 1.154-1.552, P < .001). (C) The nomogram including the lncRNA signature, age, T stage, and N stage for the practical prediction of the prognosis of CM patients. (D) AUCs of nomogram, signature and other clinicopathological characteristics for predicting 3-year OS. (E) A calibration curve analysis showed that the nomogram predictions were consistent with actual observations. (F) The ROC comparison of the studied signature with existing signatures. (G) The C-index of Nomogram.

Figure 5

Gene set enrichment analysis. Immune-related and ferroptosis-related pathways were significantly enriched in the low-risk group. (A) Immune system process. (B) Immune response. (C) Ferroptosis. (D) Apoptosis. (E) Antigen processing and presentation. (F) T-cell receptor signaling pathway. (G) Natural killer cell-mediated cytotoxicity. (H) B-cell receptor signaling pathway.

Figure 6

Tumor microenvironment analyses. (A) Immunity heatmap including the Tumor purity score, ESTIMATE score, immune score, stromal score and 7 other algorithms (ssGSEA, TIMER, CIBERSORT, MCPcounter, QUANTISEQ, XCELL, and EPIC). (B) The high-risk group had significantly higher tumor purity. (C–E) The high-risk group had significantly lower ESTIMATE, stromal, and immune scores. (F) A radar plot shows that the high-risk group had significantly fewer CD8+ T cells, CD4+ memory activated T cells and M1 macrophages, but more M0 and M2 macrophages. *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 7

ssGSEA analyses (A) comparison of ssGSEA scores for 16 immune cells between the high- and low-risk groups. (B) A comparison of the 2 groups’ ssGSEA scores for 13 immune functions indicates that the high-risk group had a significantly lower the checkpoint score. *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 8

Immune checkpoint differences. (A) The expression of 44 immune checkpoint genes differed significantly between the high- and low-risk groups. The low-risk group had significantly higher expression of CD274 (PDL-1) and CLAT-4 (P < .001). (B) The difference in ICI efficacy was not significant between the 2 groups if PD-1 and CLAT4 were both negatively expressed. (C–E) Low-risk patients with PD-1 or CTLA4 expression were likely to respond better to ICIs than high-risk patients were. (F,G) The TIDE and Dysfunction scores were higher in low-risk group than high-risk group. (H) The prognostic performance in immunotherapy cohorts.

Figure 9

TMB and drugs sensitivity analysis. (A) The TMB in different risk groups. (P < .001). (B,C) The prognostic performance of TMB scores in survival cures. (D–E) The mutation rates of the top 20 genes in different risk groups. (F) The IC50 score with mutation. (G) The target drugs sensitivity in different risk groups. (H) The Chemotherapy drugs sensitivity in different risk groups.