Identification and validation of ferroptosis-related lncRNA signature as a prognostic model for skin cutaneous melanoma

Abstract

Background: Melanoma is a type of skin cancer, which originates from the malignant transformation of epidermal melanocytes, with extremely high lethality. Ferroptosis has been documented to be highly related to cancer pathogenesis and the effect of immunotherapy. In addition, the dysregulation of lncRNAs is greatly implicated in melanoma progression and ferroptosis regulation. However, the significance of ferroptosis-related lncRNA in melanoma treatment and the prognosis of melanoma patients remains elusive.

Methods: Via Least Absolute Shrinkage Selection Operator (LASSO) regression analysis in the TCGA SKCM database, a cutaneous melanoma risk model was established based on differentially-expressed ferroptosis-related lncRNAs (DEfrlncRNAs). The nomogram, receiver operating characteristic (ROC) curves, and calibration plots were conducted to examine the predictive performance of this model. Sequentially, we continued to analyze the differences between the high- and low-risk groups, in terms of clinical characteristics, immune cell infiltration, immune-related functions, and chemotherapy drug sensitivity. Moreover, the expressions of DEfrlncRNAs, PD-L1, and CD8 were also examined by qRT-PCR and immunohistochemical staining in melanoma tissues to further confirm the potential clinical implication of DEfrlncRNAs in melanoma immunotherapy.

Results: 16 DEfrlncRNAs were identified, and a representative risk score for patient survival was constructed based on these 16 genes. The risk score was found to be an independent prognostic factor for the survival of melanoma patients. In addition, the low-risk group of patients had higher immune cell infiltration in the melanoma lesions, higher sensitivity to chemotherapeutic agents, and a better survival prognosis. Besides, the high expression of the identified 5 DEfrlncRNA in the low-risk group might suggest a higher possibility to benefit from immune checkpoint blockade therapy in the treatment of melanoma.

Conclusion: The DEfrlncRNA risk prediction model related to ferroptosis genes can independently predict the prognosis of patients with melanoma and provide a basis for evaluating the response of clinical treatment in melanoma.

Keywords: cutaneous melanoma; ferroptosis; immune microenvironment; immunotherapy; lncRNA.

Figure 1

Presentation of differentially expressed lncRNAs between SKCM samples and samples from normal tissues (log2 fold change>2, adjusted p-value<0.05). (A) Flow chart of the analytical process in this study. (B) A volcano map of lncRNAs that are differentially expressed. Red abd blue dots represent the genes that are significantly up-regulated and downregulated, respectively. (C) A Venn map for the up-regulated ferroptosis-related lncRNAs.

Figure 2

A prognostic risk model was established using LASSO regression and Cox regression analyses. (A) Cross-validation was used to tune the parameter screening in the LASSO regression model. (B) The LASSO coefficient profiles for the 32 DEfrLncRNAs. (C) Forest plots of HRs and P-values of selected DEfrLncRNAs using univariate Cox regression analysis. Sixteen of the DEfrLncRNAs were found to be prognostic factors, and all are protective factors in SKCM with HR< 1. (D) Distribution heat map of the 16 DEfrLncRNA expressions, RS, and clinical status (alive or dead) relative to SKCM.

Figure 3

Risk assessment model for prognosis prediction. (A) The RS distribution for the SKCM patients. (B) The number of patients with different RS who were alive or had died. A larger number of deaths were observed in the group with a higher RS. (C) The OS rates in patients with SKCM who were included in the LR and HR groups were evaluated using the Kaplan-Meier method. The ROC curves as well as the AUCs at 2-, 4-, and 5-year survival as the predictive signature. (D) The OS rate for SKCM patients in the first internal cohort, as assessed using the Kaplan-Meier method; ROC curves and AUCs at 2-, 4-, and 5-year survival as the predictive signature. (E) The SKCM patient OS rate in the second internal cohort was determined using the Kaplan-Meier method; the ROC curves and AUCs at 2-, 4-, and 5-year survival as the predictive signature.

Figure 4

Clinical assessment using the risk assessment model. (A) A heatmap reveals the distribution of tumor purity, T stage, sex, pathologic stage, N and M stages, and age, along with RS. (B) The univariate Cox regression analysis was used to examine the associations among the RS, clinical features, and OS of SKCM patients. (C) Multivariate Cox regression analysis was employed to reveal the associations present among the RS, clinical features, and OS of SKCM patients. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 5

Nomogram construction and verification. (A) A nomogram that combined the clinical parameters and RS was applied to estimate the 2-, 4-, and 5-year OS for patients with SKCM. (B) The consistency of the calibration curve tests between the observed rates of OS and the survival rates that were predicted. (C) The OS rates for the nomogram samples for the low and high Nom-risk groups were analyzed using the Kaplan-Meier method. (D) The ROC curves and the AUCs at 2-, 4-, and 5-year survival as the predictive signature. (E) ROC curve of the RS and clinicopathological features.

Figure 6

The difference observed between the HR and LR groups in the immune microenvironment. (A) A correlation heatmap for the 22 different types of immune cells. The degree of correlation is represented by the size of the colored squares. Blue indicates the existence of a negative correlation, while red indicates a positive correlation. The intensity of the color represents the strength of the correlation; a darker color represents a stronger correlation. (B) The CIBERSORT algorithm was employed to determine the level of infiltration that the 22 immune cells exhibited in the LR and HR groups. (C) This figure reveals the correlation between the RS and the immune checkpoint genes. The dot size indicates how strong the correlation is between the RS and the immune checkpoint genes. Larger dots represent stronger correlations, and smaller dots indicate weaker correlations. Furthermore, the dot color and its intensity is indicative of the P-value. A more intense purple color represents a lower P-value, and a more intense green color represents a larger P-value. A P-value<0.05 indicates statistical significance. (D) A distribution heat map that shows the expression of immune checkpoint genes in the LR and HR groups. *p < 0.05, **p < 0.01.

Figure 7

(A) The expression of five DEfrlncRNAs in melanoma tissues and nevi. (B, C) Correlation analysis between the five DEfrlncRNAs and PD-L1/CD8α in melanoma tissues. The Spearman correlation was used to calculate the r value. *p < 0.05, **p < 0.01, ***p < 0.001.