DNA damage repair-related gene signature predicts prognosis and indicates immune cell infiltration landscape in skin cutaneous melanoma

Abstract

Background: DNA damage repair plays an important role in the onset and progression of cancers and its resistance to treatment therapy. This study aims to assess the prognostic potential of DNA damage repair markers in skin cutaneous melanoma (SKCM).

Method: In this study, we have analyzed the gene expression profiles being downloaded from TCGA, GTEx, and GEO databases. We sequentially used univariate and LASSO Cox regression analyses to screen DNA repair genes associated with prognosis. Then, we have conducted a multivariate regression analysis to construct the prognostic profile of DNA repair-related genes (DRRGs). The risk coefficient is used to calculate the risk scores and divide the patients into two cohorts. Additionally, we validated our prognosis model on an external cohort as well as evaluated the link between immune response and the DRRGs prognostic profiles. The risk signature is compared to immune cell infiltration, chemotherapy, and immune checkpoint inhibitors (ICIs) treatment.

Results: An analysis using LASSO-Cox stepwise regression established a prognostic signature consisting of twelve DRRGs with strong predictive ability. Disease-specific survival (DSS) is found to be lower among high-risk patients group as compared to low-risk patients. The signature may be employed as an independent prognostic predictor after controlling for clinicopathological factors, as demonstrated by validation on one external GSE65904 cohort. A strong correlation is also found between the risk score and the immune microenvironment, along with the infiltrating immune cells, and ICIs key molecules. The gene enrichment analysis results indicate a wide range of biological activities and pathways to be exhibited by high-risk groups. Furthermore, Cisplatin exhibited a considerable response sensitivity in low-risk groups as opposed to the high-risk incidents, while docetaxel exhibited a considerable response sensitivity in high-risk groups.

Conclusions: Our findings provide a thorough investigation of DRRGs to develop an DSS-related prognostic indicator which may be useful in forecasting SKCM progression and enabling more enhanced clinical benefits from immunotherapy.

Keywords: DNA damage repair; immunotherapy; prognostic factor; skin cutaneous melanoma; tumor microenvironment.

Figure 1

The flow diagram of this study.

Figure 2

Construction of DRRGs signature. (A) Significant enrichment of DNA repair-related pathways in SKCM patients. NES, normalized enrichment score; (B) The forest plot of the univariate Cox regression depicted 30 DRRGs associated with SKCM survival; (C) Lasso regression for DRRGs in univariate Cox regression; (D) Coefficients of selected features denotes the risk coefficient is shown by lambda parameter.

Figure 3

Construction of the DRRGs signature and prognostic analysis. (A) Risk score distribution among patients with SKCM; (B) Spearman correlation analysis of risk score and disease-specific survival (DSS); (C) Survival status of each patient with SKCM. Blue signifies low risk or alive while red signifies high risk or dead; (D) Heatmap of gene expression between the high and low-risk cohort; (E) Relative gene expression between the high and low-risk cohort; (F) AUC values for risk score, age, gender, stage, TMB, ESTIMATE score, tumor purity, and CYT classification; (G) Immunohistochemical staining images from The Human Protein Atlas of 2 key genes in SKCM.

Figure 4

DRRGs signature based on training (TCGA-SKCM) and testing (GSE65904) cohorts. (A) KM plot of DSS premised on the risk scores; (B) ROC for DSS; (C)PCA plot; (D) t-SNE analysis in the training cohort (TCGA-SKCM); (E) KM plot of DSS premised on the risk scores; (F) ROC for DSS; (G) PCA plot; (H) t-SNE analysis in the test cohort (GSE65904).

Figure 5

Construction of a nomogram based on the DRRGs signature. (A) The univariate Cox analysis illustrated that the age, stage, tumorpurity, CYT and riskScore were statistically distinct; (B) The multivariate cox analysis illustrated the age, stage and riskScore were 3 independent predictors of SKCM prognosis; (C) Survival curve for patients with distinct TMB and risks; (D) The nomogram integrated with the parameters (riskScore, stage, and age) amongst patients from the TCGA cohort; (E) Calibration curve of the nomogram at 1, 3, and 5 years; (F) AUC values for 1-, 3-, and 5- year survival rates premised on the nomogram.

Figure 6

GSEA of SKCM patients premised on the DRRGs prognostic signature. (A–D) The expression differences of immune score, stromal score, ESTIMATE score, and tumor purity in high- vs. low-risk cohort; (E–H) Spearman correlation analysis of risk score with immune score, stromal score, RNAss, and DNAss; (I–L) GSEA outcomes illustrated substantial enrichment of immune-associated biological mechanism in the low-risk patients; (M) A Barplot of the 22 immune fractions designated by dissimilar colors in each SKCM sample; (N) Correlation heatmap of the ratio of TIICs; (O) Wilcoxon test of 22 immune fractions in high- vs. low-risk cohort; (P) Link between risk score and immune cell infiltration and related roles through the ssGSEA analysis. The score denotes the immune score, with elevated scores signifying a deeper extent of immune cell infiltration; (Q) ICB molecules expressed differently between high- and low-risk groups; (ns, not significant, *P < 0.05, **P < 0.01, and ***P < 0.001).

Figure 7

The mutation profile and TMB among low-risk and high-risk groups. (A, B) Mutation profile of low-risk and high-risk groups. (C, D) The relationship between the immune-related risk signature and TMB. (E–G) The proportion of mutation of TTN, MUC16 and DNAH5 in both low-/high-risk groups form the TCGA-SKCM dataset. (H–J) Comparison of the risk score between mutation and wild groups.(ns, not significant, *P < 0.05, **P < 0.01, and ***P < 0.001).

Figure 8

Relationship Between the Prognosis-Associated Immune Signature and Drug Response in SKCM. The differential expression of (A) PDCD1, (B) CD274, (C) CTLA4, (D) TGFB1 in the two subgroups, correspondingly; Spearman correlation analysis of risk score and (E) PDCD1, (F) CD274, (G) CTLA4, (H) TGFB1; (I) IPS score distribution plot; (J) IPS-PD1 blocker score distribution plot; (K) IPS-CTLA4/PD1 blocker score distribution plot; (L) IPS-CTLA4 blocker score distribution plot; (M) Boxplot comparing patient response to docetaxel chemotherapy; (N) Boxplot comparing patient response to cisplatin chemotherapy. (ns, not significant, *P < 0.05, **P < 0.01, and ***P < 0.001).