Neutrophil extracellular trap related risk score exhibits crucial prognostic value in skin cutaneous melanoma, associating with distinct immune characteristics

Abstract

Background

Neutrophil extracellular traps (NETs) are related to the prognosis of cancer patients. Nevertheless, the potential prognostic values of NETs in skin cutaneous melanoma (SKCM) remains largely unknown.

Materials and methods

The NET‐related gene signature was constructed by LASSO Cox regression analysis using the TCGA‐SKCM cohort. The overall survival (OS) and immune status in SKCM patients between the high‐ and low‐NET score (high‐score, low‐score) groups were explored. The scRNA‐seq dataset GSE115978 was used to understand the role of NET score in SKCM at single cell resolution.

Results

A five NET genes‐based signature (TLR2, CLEC6A, PDE4B, SLC22A4 and CYP4F3) was constructed as the NET‐related prognostic model for SKCM. The OS of SKCM patients with low‐score was better than that in patients with high‐score. Additionally, NET score was negatively associated with infiltration of some immune cells (e.g. type I Macrophages, CD8‐T cells, CD4‐T cells). Moreover, patients with high‐score had low stromal, immune and ESTIMATE scores. Furthermore, drug sensitivity analysis results showed that Lapatinib, Trametinib and Erlotinib may have better therapeutic advantages in patients with high‐score.

Conclusion

We established a NET‐related five gene signature in SKCM and found that the NET‐related signature may exhibit a good predictive ability for SKCM prognosis. The NET score may not only predict the survival outcome and drug sensitivity in SKCM, but also reflect the immune conditions of SKCM patients.

1. INTRODUCTION

Skin cutaneous melanoma (SKCM) is the deadliest form of skin cancer, and its incidence has been steadily rising in recent years worldwide. ¹ , ² Melanoma originates from the uncontrolled proliferation of melanocytes, ³ and it is characterized by a high degree of malignancy and poor prognosis. ⁴ It has been shown that surgery is a major option for the treatment of the primary tumor in melanoma. ⁵ , ⁶ Meanwhile, systemic therapies (including chemotherapy, immunotherapy and targeted therapy) are the mainstay of treatment in advanced (metastatic) melanoma. ⁷ , ⁸ , ⁹ However, patients in the advanced SKCM often have a worse prognosis. ¹⁰ Therefore, development of novel prognostic indicators for predicting the prognosis will help to improve the management of SKCM patients.

Evidence have shown that immune cell infiltrate populations in tumor tissues are often related to the prognosis in cancer patients. ¹¹ , ¹² In general, the growth and invasion of tumor cells greatly depends on the surrounding tumor microenvironment (TME). ¹³ , ¹⁴ Immune cells, a main cell type in the TME, play crucial roles in tumorigenesis and metastasis. ¹⁵ , ¹⁶ It has been shown that some immune cells (e.g. natural killer cells and cytotoxic CD8+ T cells) exert tumor‐antagonizing function, while some immune cells (e.g. M2 type macrophages and myeloid‐derived suppressor cells) possess tumor‐promoting activities. ¹⁵ , ¹⁷ Neutrophils are also an important immune cell type in the TME, ¹⁸ playing a dual role in antagonizing and promoting cancer. ¹⁹ , ²⁰

Previously, reports have shown that neutrophils can promote cancer metastasis through neutrophil extracellular traps (NETs). ²¹ NETs are net‐like structures composed of decondensed chromatin and granule proteins, which are formed and extruded by neutrophils to trap and kill pathogens. ²² It has been shown that NET are related to survival outcomes of cancer patients. ²³ Zuo et al. constructed a six NET genes‐based risk signature for predicting the prognosis of lung cancer patients, including G0S2, KCNJ15, S100A12, AKT2, CTSG and HMGB1, and demonstrated a prognostic value of this risk signature in lung cancer. ²⁴ Nevertheless, the role of NETs in SKCM remains largely unknown.

In this research, we established a NET‐related five gene signature in SKCM. Meanwhile, we found that the signature may exhibit a good predictive ability for SKCM prognosis. The NET score may not only predict the OS and drug sensitivity in SKCM, but also reflect the immune conditions of SKCM patients. It may lay a foundation for facilitating personalized and precise treatment of SKCM in the near future.

2. MATERIALS AND METHODS

2.1. Dataset preparation

The mRNA profile data (from 471 SKCM samples and 1 normal sample) and corresponding clinical information of patients in the TCGA‐SKCM cohort (as a training cohort) was retrieved and downloaded from the Cancer Genome Atlas (TCGA, https://tcga‐data.nci.nih.gov/tcga/) database. In the TCGA‐SKCM cohort, 454 patients had complete survival information.

The GSE65904 and GSE115978 datasets were acquired from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) database. The GSE65904 dataset contained 214 SKCM tissue samples from 214 SKCM patients, which was set as a validation cohort. Meanwhile, a total of 150 patients had complete survival information in the GSE65904 dataset. The GSE115978 dataset included 7186 cell samples from 31 SKCM patients.

Additionally, 69 NET‐related genes were collected from the previous study, which are listed in Table S1. ²⁵

2.2. Establishment of a NET‐related prognostic model

To screen candidate NET genes related to prognosis in SKCM, 69 NET‐related genes were first scored employing univariate Cox regression analysis based on 454 patients’ survival data from TCGA‐SKCM cohort. Next, LASSO Cox regression analysis was utilized to filter hub prognosis‐related NET genes using the “glmnet” R package (version 4.1.7). ²⁶ Based on the expression features of five hub prognosis‐related NET genes screened above, the NET‐related risk score for each patient was calculated [$\mathrm{NET}\mathrm{score}={\sum }_{i=1}^{\mathrm{n}}Coe{f}_i\times x_i$ (Coef_i , the regression coefficient of each gene, x_i , the gene expression level)]. The risk score was calculated for each patient in the TCGA‐SKCM cohort. According to the median NET score value, patients with the NET score higher than the median NET score were grouped into high‐score group, the others were assigned into low‐score group. The R packages “survival” and “survminer” and two‐sided log‐rank test were then employed to determine the value of NET score.

2.3. Survival analysis

The OS rates of SKCM patients in low‐ and high‐score groups were estimated using R packages “survival” (https://CRAN.R‐project.org/package = survival, version 3.5‐5) and “survminer” (https://CRAN.R‐project.org/package = survival, version 0.4.9) based on Kaplan‐Meier method. The significance of the OS rates of patients between two groups was tested by log‐rank test. The R package “ survival” (version 0.4) was applied for generating the receiver operating characteristics (ROC) curve, ²⁷ and then the area under curve (AUC) was calculated.

2.4. Screening of differential expressed genes (DEGs)

To identify DEGs between low‐ and high‐score groups, R package “ limma” (version 3.52.4) was used. ²⁸ |Log2FC| > 1 and FDR < 0.05 were considered as the threshold.

2.5. Functional enrichment analysis and gene set variation analysis (GSVA)

To explore the biological role of DEGs, Gene Ontology [GO, including biological process (BP), molecular function (MF), cellular component (CC)] and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted using R package “clusterProfiler” (version 4.7.1.2). ²⁹ The GSVA was performed using the c2.cp.kegg.v2023.1.Hs.symbols and c5.go.v2023.1.Hs.symbols datasets downloaded from the Molecular Signature Database (https://www.gsea‐msigdb.org/gsea/msigdb). |LogFC| > 0.5 and p‐value < 0.05 were considered as the threshold.

2.6. Analysis of immune cell characteristics

To calculate the abundance of 28 immune cells in two groups, ssGSEA algorithm from the R package “GSVA” (version 1.46.0) was utilized. ³⁰ Meanwhile, the stromal, ESTIMATE and immune scores between two groups were evaluated using the R package “estimate” (https://R‐Forge.R‐project.org/projects/estimate/, version 1.0.13).

2.7. Single‐cell RNA‐sequencing (scRNA‐seq) analysis

To obtain high‐quality single‐cell data, R package “Seurat” (version 4.3.0) was applied for data processing. ³¹ Next, data normalization was processed using the function “NormalizedData”, and then, principal component analysis (PCA) was conducted using the function “RunPCA”. After that, the unsupervised clustering of major cell subtypes was analyzed using the function “FindClusters” from R package “Seurat”, and then visualized by Umap. Later on, to annotate the single‐cell clusters manually, the CellMarker database (http://bio‐bigdata.hrbmu.edu.cn/CellMarker/) was used. Meanwhile, DEGs between two risk groups were screened by the function “FindMarkers”. The pseudotime trajectory analysis was then conducted using the R package “monocle”, ³² and cell communication pattern was generated by the R package “iTALK” (https://github.com/Coolgenome/iTALK).

2.8. Statistical analysis

Wilcoxon rank‐sum test (two‐side) was employed to compare the difference in gene levels between two groups. Pearson correlation analysis was performed using the function “cor” in R package. p‐value < 0.05 was considered statistically significant.

3. RESULTS

3.1. Construction of a NET‐related prognostic model in SKCM

To explore the prognostic value of NET‐related genes in SKCM patients, univariate Cox regression analysis was conducted. Twenty‐three NET‐related genes related to patients’ prognosis were identified in the TCGA‐SKCM cohort based on Cox regression analysis (Figure 1A). Next, these 23 genes were then subjected to LASSO regression analysis. Ultimately, five hub NET‐associated genes, including toll‐like receptor 2 (TLR2), phosphodiesterase 4B (PDE4B), solute carrier family 22, member 4 (SLC22A4), C‐type lectin domain family 6 member A (CLEC6A), and cytochrome P450 4F3 (CYP4F3), were selected for the establishment of the NET‐related prognostic model in SKCM (Figure 1B). Thereafter, according to the formula $({\sum }_{i=1}^{\mathrm{n}}Coe{f}_i\times x_i)$, the NET score of each patient in the TCGA‐SKCM cohort was calculated as follows: NET score = (−0.07484666 × TLR2 expression level) + (−0.12240557 × CLEC6A expression level) + (−0.06321539 × PDE4B expression level) + (−0.04372487 × SLC22A4 expression level) + (0.07549143 × CYP4F3 expression level). Differential expression of NET‐related genes among different stages (I, II, III, IV) in the TCGA‐SKCM cohort was shown by the heat map (Figure S1). Meanwhile, differential expression of signature genes between low‐ and high‐score groups in the TCGA‐SKCM cohort was shown by the heat map (Figure S2).

FIGURE 1.

Construction of a NET‐related prognostic model in SKCM. (A) Twenty‐three candidate genes were selected by univariate Cox regression analysis. (B) Lasso regression analysis was conducted to identify hub NET‐related genes. The optimal tuning parameter lambda value was determined using the Lasso regression analysis. The X axis represents Log (λ, lambda), the Y axis represents partial likelihood Deviance. The smallest Y value corresponds to the optimal lambda value (the value below the dotted line). The number of variables above the dotted line is the number of hub NET‐related prognostic genes. (C, D) Kaplan‐Meier curves showed the survival outcomes of SKCM patients in high‐ and low‐score groups in (C) TCGA‐SKCM cohort and (D) GSE65904 cohort. (E, F) Time‐dependent AUC curves for SKCM patients in (E) TCGA‐SKCM cohort and (F) GSE65904 cohort. The horizontal axis represents 1‐specificity, and the vertical axis represents sensitivity. AUC, area under the curve; CI, confidence interval; NET, neutrophil extracellular trap; ROC, receiver operating characteristics; SKCM, skin cutaneous melanoma; TCGA.

Based on the median value of NET score, patients in the TCGA‐SKCM cohort were distributed into low‐ and high‐score groups. The results in Figure 1C revealed that the OS in high‐score group was remarkably worse than that in low‐score group. Meanwhile, according to the optimum cutoff value of NET score, SKCM patients in the GSE65904 cohort were also grouped into low‐ and high‐score groups. Consistently, SKCM patients in high‐score group were related to a poor OS (Figure 1D). Additionally, time‐dependent ROC analysis further assessed the prognostic value of the NET score. As indicated in Figure 1E and F, the 1‐, 3‐ and 5‐year AUC values of OS were 0.64, 0.69 and 0.68 in the TCGA‐SKCM cohort and 0.58, 0.59 and 0.59 in the GSE65904 cohort. These data suggested that NET score was capable of predicting 1‐, 3‐ and 5‐year OS rates of SKCM patients in the TCGA‐SKCM cohort.

3.2. Relationship between the NET score and the clinicopathological characteristics of SKCM patients

In the TCGA‐SKCM cohort, significant differences in NET score were observed between age ≥58 and age <58 groups and between female and male groups (Figure 2A and B). The SKCM patients older than 58 years or male SKCM patients displayed higher NET scores compared to patients younger than 58 years or female SKCM patients, respectively (Figure 2A and B). Meanwhile, significant differences in NET score were also observed between stage I and II, between stage I and III, between stage II and III of SKCM patients (Figure 2C).

FIGURE 2.

Relationship between the NET score and the clinicopathological characteristics of SKCM patients. (A–C) The box plot showed the NET score in different groups in the TCGA‐SKCM cohort, including (A) age (> 58, < 58), (B) gender (female, male) and (C) stage (I, II, III, IV). (D, E) Kaplan‐Meier curves showed the survival outcomes of SKCM patients with (D) age > 58 or (E) age < 58 in high‐ and low‐score groups in TCGA‐SKCM cohort. (F, G) Kaplan‐Meier curves showed the survival outcomes of (F) female or (G) male SKCM patients in high‐ and low‐score groups in TCGA‐SKCM cohort. (H, I, J) Kaplan‐Meier curves showed the survival outcomes of SKCM patients in stage (H) II, (I) III and (J) IV in high‐ and low‐score groups in TCGA‐SKCM cohort. NET, neutrophil extracellular trap.

Thereafter, for validation, the prognostic value of the NET score in subgroups, SKCM patients in the TCGA‐SKCM cohort were grouped into different groups based on the clinicopathological characteristics: age (≥58 and <58 groups), gender (male and female groups) and stage (II, III, IV groups). As shown in Figure 2D‐J, in all these subgroups, SKCM patients with high‐score displayed worse OS relative to patients with low‐score, suggesting that the prognostic value of NET score was not influenced by the clinicopathological characteristics of SKCM patients.

3.3. DEGs and the functional pathways between two risk groups

Next, we screened DEGs between low‐ and high‐score groups. A total of 2383 DEGs, including 230 elevated and 2153 decreased genes, were identified in the high‐score group in the TCGA‐SKCM cohort compared to the low‐score group (Figure 3A and B). Next, GO and KEGG analysis were conducted on these DEGs. KEGG analysis showed that 11 pathways, such as “Ether lipid metabolism”, “Ras signaling pathway”, “Melanogenesis”, “Arachidonic acid metabolism”, were presented in high‐score group, while 70 pathways, such as “Cytokine−cytokine receptor interaction” and “Hematopoietic cell lineage”, appeared in low‐score group (Figure 3C and Table S2). GO analysis revealed that 92 pathways, such as “epidermis development”, “keratinocyte differentiation” and “skin development”, were presented in high‐score group, while 1213 pathways, such as “lymphocyte mediated immunity” and “immune response−regulating signaling pathway”, appeared in low‐score group (Figure 3D and Table S2). Moreover, GSVA results showed that 27 pathways, such as “Interleukin 15 mediated signaling pathway”, “Interleukin 2 mediated signaling pathway”, “Interleukin 12 receptor binding” and “Positive regulation of natural killer (NK) cell differentiation”, were more active in low‐score group (Figure 3E, Table S2). Meanwhile, GSEA results showed that 79 pathways, such as “Antigen processing and presentation” and “Natural killer cell mediated cytotoxicity”, were significantly enriched in low‐score group (Figure S3, Table S2).

FIGURE 3.

DEGs and the functional pathways between two NET score groups. (A) The volcano plot and (B) heat map showed the DEGs identified between high‐ and low‐score groups in the TCGA‐SKCM cohort. (C) KEGG, (D) GO and (E) GSVA analysis of DEGs identified between two groups in the TCGA‐SKCM cohort. GO, Gene Ontology; GOBP, Gene Ontology Biological Process; GOCC, Gene Ontology Cellular Component; GOMF, Gene Ontology Molecular Function; KEGG, Kyoto Encyclopedia of Genes and Genomes, NET, neutrophil extracellular trap; os, overall survival time (days).

3.4. Predictive value of NET score in immune microenvironment in SKCM

Next, we analyzed the correlation between NET score and each immune cell proportion in the TCGA‐SKCM cohort. The ssGSEA results showed that 14 immune cell types, such as effector memeory CD8 T cells, regulatory T cells and immature B cells, were negatively associated with the NET score, while 7 immune cell types, such as immature dendritic cells, neutrophils and monocytes, were positively related to the NET score (Figure 4A). Additionally, the data from the Xcell (https://xcell.ucsf.edu/) database showed that 12 immune cell types, such as osteoblasts, MEP and keratinocytes, were positively related to the NET score, while 35 immune cell types, such as activated dendritic cells (aDCs), M1 Macrophages, CD8 T cells, NK cells, CD4 memory T cells, were negatively associated with the NET score (Figure 4B and Table S3). Furthermore, the results of ESTIMATE algorithm showed that significant lower stromal, immune and ESTIMATE scores were observed in the high‐score group, compared to the low‐score group (Figure 4C). Meanwhile, the stromal, immune and ESTIMATE scores were all negatively correlated with the NET score (Figure 4D).

FIGURE 4.

Predictive value of NET score in immune microenvironment in SKCM. (A, B) The correlations between the NET score and the abundance of immune cell infiltration. (C) The violin plot showed the Stromal Score, Immune Score and ESTIMATE Score in two groups. (D) The scatter plot displayed the correlations between NET score and Stromal Score, Immune Score or ESTIMATE Score in the TCGA‐SKCM cohort. NET, neutrophil extracellular trap.

3.5. scRNA‐seq analysis for the NET score

The SKCM scRNA‐seq data from the GSE115978 dataset was used to understand the role of NET score in SKCM at single‐cell resolution. A total of 19 cell clusters were found in the tumor tissues from the GSE115978 dataset using the R package “Seurat” (Figure 5A). Seven major cell types, including B cells, endothelial cells, fibroblasts, melanocyte, myeloid cells, NK cells and T cells, were identified (Figure 5B), and NET score levels in these cell clusters are presented in Figure 5C.

FIGURE 5.

scRNA‐seq analysis for the NET score. (A) The UMAP algorithm identified 19 cellular clusters based on the scRNA‐seq data from the GSE115978 dataset. (B) UMAP plot showed clusters of 7 main cell types. (C) The levels of the NET score in different cell types. (D) KEGG and (E) GO analysis of DEGs identified between high‐ and low‐score groups in the GSE115978 dataset. (F) Pseudotime analysis of melanoma cells. GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; NET; neutrophil extracellular trap; UMAP, uniform manifold approximation and projection.

Next, DEGs between two low‐ and high‐score groups were screened by the function “FindMarkers”. Thereafter, GO and KEGG analyses were conducted on these DEGs. KEGG analysis showed that “Tyrosine metabolism”, “Melanogenesis”, “Cell adhesion molecules” and “Primary immunodeficiency” pathways appeared in high‐score group, while 37 pathways, such as “Leishmaniasis”, “Staphylococcus aureus infection” and “Tuberculosis”, were enriched in low‐score group (Figure 5D and Table S4). GO analysis showed that 49 pathways, such as “melanin biosynthetic process” and “melanin metabolic process”, were presented in the high‐score group, while 566 pathways, such as “MHC class II protein complex” and “MHC class II protein complex binding”, appeared in low‐score group (Figure 5E and Table S4). By using pseudotime analysis, SKCM cell lineage differentiation trajectory was constructed and five differentiation stages of SKCM cells identified (Figure 5F). As pseudotime increased, SKCM cells showed a tendency towards elevated NET score (Figure 5F).

3.6. Cell communication pattern related to the NET score

Based on the scRNA‐seq data, we constructed cell‐cell communications to reveal the molecular signatures underlying the complex cellular processes. The different cellular signaling pathways (CSPs) regarding checkpoints, cytokine, growth factor (GF) and other among different cells in the TME are shown in Figure 6A‐H. For diverse CSPs regarding checkpoints between two groups among different cells, TNFRSF14 was the most active signaling in the high‐score group (Figure 6A and B). For diverse CSPs regarding GF between two groups among different cells, CD44 and CTGF were the most active signaling in both high‐ and low‐score groups (Figure 6E and F). For diverse CSPs regarding other between two groups among different cells, CD44 and VIM were the most active signaling in the high‐score group (Figure 6G and H). No active cytokine‐related signaling pathway was observed between two groups among different cells in the TME (Figure 6C and D).

FIGURE 6.

Cell communication pattern related to the NET score. Different cellular signaling pathways regarding (A, B) checkpoint, (C, D) cytokine, (E, F) growth factor or (G, H) other among different cells in the tumor microenvironment between high‐ and low‐score groups.

3.7. Prediction of potential therapeutic agents against SKCM based on the NET score

To further link the NET score with clinical application, the R package “oncoPredict” was applied to analyze the correlation between the NET score and IC50 of the drugs in the TCGA‐SKCM cohort. NET score was significantly positively correlated with 168 drugs, such as AMG.319_2045, JQ1_2172, SB216763_1025 (Figure 7A and Table S5), and was negatively correlated with 11 drugs, including ERK_6604_1714, SB505124_1194, SCH772984_1564, OSI.027_1594, Lapatinib_1558, Trametinib_1372, VX.11e_2096, Sapitinib_1549, Erlotinib_1168, Ulixertinib_1908, PD0325901_1060 (Figure 7B and Table S5).

FIGURE 7.

Prediction of potential therapeutic agents against SKCM based on the NET score. (A, B) The scatter plot displayed the correlation of NET score and IC₅₀ values of different agents in the TCGA‐SKCM cohort. (C–G) The radar chart showed the inference score of NET‐related genes, including TLR2, CLEC6A, PDE4B, SLC22A4 and CYP4F3 in SKCM. NET, neutrophil extracellular trap.

Furthermore, Comparative Toxicogenomics Database (CTD) database (http://ctdbase.org/) was used to analyze the inference score of TLR2, CLEC6A, PDE4B, SLC22A4 and CYP4F3 in the NET score in SKCM. The results showed that all these NET‐associated genes might be the therapeutic targets for SKCM (Figure 7C‐G). The genes TLR2, PDE4B and CYP4F3, might have strong predictive value on the occurrence and development of SKCM based on the inference score (Figure 7C‐E).

4. DISCUSSION

NET has been found to be involved in the progression of melanoma. ³³ , ³⁴ , ³⁵ Modestino et al. found that the circulating levels of NETs were obviously increased in patients with metastatic melanoma. ³⁶ Meanwhile, Liu et al. found that NETs could impair the endothelial barrier and then lead to the vascular leakage, thereby facilitating the entrance of melanoma cells into the blood circulation. ³⁷ These findings showed important roles of NET in the development of melanoma. In the current research, for the first time, we constructed a NET‐related gene signature in the present study, and found that the NET score could predict the prognosis and the sensitivity of anti‐tumor drugs in SKCM.

Using the LASSO regression analysis, a total of five genes, including TLR2, CLEC6A, PDE4B, SLC22A4 and CYP4F3, were identified to construct a NET‐related gene signature for predicting prognosis in SKCM. The CTD database verified that these five hub genes could be the promising therapeutic targets for SKCM. TLR2 is a pathogen recognition receptor, and activation of TLR2 could mediate innate immune responses. ³⁸ , ³⁹ Recently, researchers have found that TLR2 level may be linked to the prognosis of cancers. ⁴⁰ , ⁴¹ Beilmann‐Lehtonen et al. found that high tissue TLR2, TLR5 and TLR7 levels were associated with a better prognosis. ⁴¹ PDE4B is also an immune regulatory molecule, and high level of PDE4B may be related to unfavorable outcomes in lymphoma. ⁴² Similar to TLR2, CLEC6A is also an innate immunity gene, and plays a crucial role in mediating innate immunity and adaptive immunity. ⁴³ Chen et al. indicated that breast cancer patients in the CLEC6A high expression group had a better prognosis compared to those in the low expression group. ⁴⁴ Additionally, SLC22A4, an organic cation transporter, plays a role in the development of immune‐related diseases, including cancers. ⁴⁵ , ⁴⁶ Quan et al. reported that low SLC22A4 level was related to the poor clinical outcome in patients with renal cell carcinoma. ⁴⁶ Furthermore, CYP4F3, a member of cytochrome P450 (CYP) superfamily, exerts a key role in lipid metabolism. ⁴⁷ Evidence have shown that CYP4F3 might be related to cancer diagnosis and prognosis. ⁴⁸ Zhao et al. identified eight hub TME‐related genes (including CYP4F3) using LASSO regression analysis, and found that the eight‐gene signature could predict the prognosis of lung cancer patients. ⁴⁸ These findings revealed that most of these hub genes are involved in cancer patient prognosis. Therefore, to verify the prognostic value of the NET score in SKCM in this study, the OS in SKCM patients in low‐ and high‐score groups were evaluated using the Kaplan‐Meier curves. Our results demonstrated that SKCM patients with high‐score had remarkably worse OS than those patients with low‐score. Excitingly, the NET score may exert a good ability for predicting the prognosis of SKCM patients.

Next, functional analyses, including GO, KEGG and GSVA, were performed to explore the mechanisms underlying the NET signature in SKCM. GO and GSVA results showed that immune‐related signaling pathways (e.g. NK cell‐ and lymphocyte‐related signaling and interleukin 2, 12 and 15 signaling pathways) were more active in low‐score group. Evidence have found that interleukin 2, 12 and 15 could induce abundant infiltration of immune effector cells (e.g. cytotoxic CD8+ T and NK cells) into tumors, thereby enhancing anti‐tumor immunity. ⁴⁹ , ⁵⁰ Additionally, we found that low NET score was correlated with high infiltration of some immune cells, such as type I Macrophages, CD8‐T cells, NK cells, CD4‐T cells. Meanwhile, low‐NET score was correlated to high stromal, immune and ESTIMATE scores. Zhuang et al. reported that hepatocellular carcinoma patients with higher immune and stromal scores displayed better survival rates. ⁵¹ To sum up, our findings showed that patients with low‐score had higher immune infiltration in tumors, suggesting that SKCM patients with low‐score may display favorable outcomes with high survival rates. Wang et al. constructed a ferroptosis‐related gene signature, and found that osteosarcoma patients with high score were related to worse prognosis and immunosuppression. ⁵² Our data showed that low stromal, immune and ESTIMATE scores were found in patients with high NET score, suggesting an immunosuppressive environment in high‐score group.

The SKCM scRNA‐seq data was then used to understand the role of NET score in SKCM at single‐cell resolution. Our results revealed that SKCM cells showed a progressive evolution to high NET score‐like tumors at the single‐cell resolution. GO results showed that MHC class II protein complex‐related pathways were more active in low‐score group. Evidence have shown that MHC class II protein is crucial for initiating antigen‐specific anti‐tumor immune response. ⁵³ Sun et al. showed that immune checkpoint blockade could enhance the antigen‐specific anti‐tumor immunity through inducing immunoactive status of the TME. ⁵⁴ These findings showed that at single‐cell resolution, low‐score group may display an immunoactive status, but high‐score group may display a reduced immune status, which was consistent with our results obtained from tumor tissues.

Furthermore, we then explored whether NET score displayed a role in the prediction of drug sensitivity. Uncovering drugs with higher sensitivity in high‐score group might improve worse prognosis of SKCM patients. Our results indicated that NET score was negatively related to the IC50 value of 11 drugs in SKCM, such as SB505124_1194 (TGFβR inhibitor), SCH772984_1564 (ERK1/2 inhibitor), OSI.027_1594 (mTORC1 and mTORC2 inhibitor), Lapatinib_1558 (ErbB2/EGFR inhibitor), Trametinib_1372 (MEK1/2 inhibitor), VX.11e_2096 (ERK inhibitor), Sapitinib_1549 (EGFR inhibitor), Erlotinib_1168 (EGFR inhibitor), Ulixertinib_1908 (ERK1/2 inhibitor), PD0325901_1060 (MEK inhibitor). These data revealed that Lapatinib, Trametinib, Sapitinib, Erlotinib and Ulixertinib may be promising therapeutic options for SKCM patients with high‐score. Among them, Lapatinib, Trametinib, Erlotinib and Ulixertinib have been investigated in vitro assay or in clinical trial. ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁸ However, no studies have shown the role of sapitinib in SKCM. For the first time, our data showed that sapitinib may be a potential therapeutic drug for SKCM patients, providing a new option for the treatment of SKCM. Nevertheless, further studies are needed to verify the responsiveness of SKCM patients to therapy.

5. CONCLUSION

In general, we established a NET‐related five gene signature in SKCM and found that the NET‐related signature may predict the prognosis of SKCM accurately. Patients with high NET score had worse prognosis. Meanwhile, the NET score may not only predict the OS and drug sensitivity in SKCM, but also reflect the immune conditions of SKCM patients. It may lay a foundation for facilitating personalized treatment of SKCM in the near future.

CONFLICT OF INTEREST STATEMENT

Authors declare no conflict of interests for this article.

ETHICS STATEMENT

No animal nor human was involved within this study.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Zhang H, Bi X, Yan P, Wang C. Neutrophil extracellular trap related risk score exhibits crucial prognostic value in skin cutaneous melanoma, associating with distinct immune characteristics. Skin Res Technol. 2024;30:e70008. 10.1111/srt.70008

DATA AVAILABILITY STATEMENT

The data generalized and analyzed during the current study are available in the public database The Cancer Genome Atlas (TCGA, https://tcga‐data.nci.nih.gov/tcga/) database, and the GSE65904 and GSE115978 datasets are available in the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) database.