Integrating bioinformatic analyses and experimental validation of disulfidptosis-related genes and TMX4 in melanoma progression

Abstract

Background

Melanoma is a highly metastatic and lethal malignancy originating from melanocytes. Disulfidptosis is a recently discovered type of programmed cell death with promising potential in cancer therapy. This study aims to establish a prognostic risk model based on disulfidptosis-related gene signatures and investigate their roles in melanoma progression.

Materials and methods

Transcriptomic data from the TCGA-SKCM and GEO databases were analyzed to identify disulfidptosis-related gene expression profiles in melanoma. Consensus clustering categorized samples into two molecular subtypes, followed by comprehensive analyses of genomic and immune infiltration characteristics. A prognostic risk model was developed using Lasso regression based on DEGs. The expression and function of TMX4 were further validated using IHC and in vitro experiments, including siRNA-mediated knockdown in A375 and A875 melanoma cell lines. Cell proliferation, migration, and indicators of redox homeostasis (GSH, MDA, and ROS) were evaluated, and SLC7A11 expression was measured to explore underlying mechanisms.

Results

Melanoma samples were divided into two subtypes with distinct tumor microenvironments. The eight-gene prognostic model accurately predicted patient outcomes in both training and validation cohorts. TMX4 was highly expressed in melanoma tissues, and its knockdown significantly inhibited the proliferation and migration of melanoma cells. Silencing of TMX4 enhanced oxidative stress by reducing GSH and increasing MDA and ROS, while upregulating SLC7A11 protein expression.

Conclusion

The prognosis model based on DRGs has demonstrated potential predictive capability in melanoma. Functional experiments confirmed that TMX4 plays a role in regulating melanoma cell proliferation, migration, and redox homeostasis. These findings suggest that targeting TMX4 and its related pathways (potentially involving disulfidptosis) could be a promising therapeutic strategy for melanoma. However, further studies are required to validate its clinical applicability.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12935-026-04182-w.

Introduction

Melanoma is a malignant tumor originating from melanocytes, most frequently occurring in the skin [1]. However, due to the widespread distribution of melanocytes throughout the body, melanoma can theoretically develop in any anatomical location [2]. Over the past few decades, the incidence of melanoma has risen sharply. From 2012 to 2020, the global number of melanoma cases increased from 230,000 to 325,000, representing a 41% increase [3]. Although more than 85% of melanomas do not metastasize, melanoma remains one of the most metastatic types of cancer in humans and is also the leading cause of death from skin tumors [4, 5]. Currently, surgical resection, chemotherapy, and radiotherapy are the primary treatment modalities for melanoma. In recent years, the treatment landscape for melanoma has undergone a significant transformation with the introduction of immunotherapy and targeted therapy, leading to improved survival rates for patients [6]. However, due to challenges such as metastasis and drug resistance, current treatment options do not fully meet clinical needs [7]. Further elucidation of the mechanisms underlying melanoma progression and identification of novel therapeutic targets remains essential.

Disulfidptosis is a newly identified form of programmed cell death (PCD), first reported in 2023 [8]. Disulfidptosis is triggered by disulfide stress caused by excessive cystine accumulation [9]. The typical characteristics of disulfidptosis include high expression of SLC7A11, depletion of NADPH, and disruption of the actin cytoskeleton [10]. SLC7A11 functions as an antiporter that imports cystine into the cell while exporting glutamate [11]. Upon entering the cell, cystine is reduced to cysteine through NADPH-dependent reactions, and cysteine then participates in the synthesis of glutathione (GSH). GSH acts as a crucial antioxidant that maintains intracellular redox homeostasis and protects cells from oxidative damage [12]. Under glucose-deprived conditions, the intracellular supply of NADPH becomes limited. In cells with high expression of SLC7A11, the large amount of cystine imported into the cell cannot be efficiently reduced to cysteine due to the insufficient NADPH supply, leading to abnormal accumulation of disulfides and disulfide stress [13]. Disulfide stress induces the formation of abnormal disulfide bonds between actin filaments, causing actin filament contraction and disruption of the cytoskeletal structure [14]. As a result, cells fail to maintain normal morphology and function, ultimately leading to cell death [12]. Recent studies have shown a close relationship between disulfidptosis and the occurrence and progression of tumors. The discovery of the disulfidptosis mechanism has opened new pathways for cancer treatment [15]. Therefore, further investigation into the association between melanoma and disulfidptosis may provide important insights for the development of melanoma biomarkers and therapeutic targets.

However, there are little experimental reports of disulfidptosis in melanoma progression at present. To address this gap, we aim to identify potential disulfidptosis-related genes (DRGs) and their functions in melanoma. This study systematically analyzes DRGs in melanoma using bioinformatics approaches, identifying two distinct melanoma subtypes based on gene expression profiles. A prognostic risk model was constructed using differentially expressed genes (DEGs) from these subtypes, and the expression of TMX4 was further investigated in melanoma cell lines through siRNA-mediated knockdown, revealing its role in regulating redox homeostasis. Our findings suggest that targeting DRGs may provide novel therapeutic opportunities for improving melanoma treatment outcomes.

Materials and methods

Data collection

The transcriptomic data of melanoma were collected from public databases. Transcriptomic data of tumor tissue samples obtained from the TCGA-SKCM project were downloaded as the bulk RNA-seq training set, along with the corresponding clinical information of the samples. Additionally, GSE65904 and GSE72056 datasets were downloaded from the GEO database and used as the bulk RNA-seq validation set and single-cell RNA-seq (scRNA-seq) validation set. A total of 15 DRGs were identified from the study by Liu et al., including ACTA2, ACTB, ACTN4, CAPZB, CD2AP, DSTN, FLNA, FLNB, INF2, IQGAP1, MYH10, MYH9, PDLIM1, SLC7A11, and TLN1 [16].

Identification of disulfidptosis-related tumor subtypes

Consistent clustering analysis of TCGA-SKCM samples was performed based on the expression profiles of DRGs using the R package CancerSubtypes. To ensure the stability and reliability of clustering, the maximum number of clusters was set to K = 10, and the proportion of ambiguous clustering (PAC) method was employed to evaluate and determine the optimal number of clusters.

Genomic and transcriptomic characterization of tumor subtypes associated with disulfidptosis

To explore the genomic differences between different subtypes, genomic mutation (short variant) and copy number variation (CNV) data from TCGA-SKCM were downloaded. Waterfall plots and lollipop charts were created to visually present the mutation and CNV profiles. Additionally, PCA, t-SNE, and UMAP dimensionality reduction algorithms were applied to the transcriptomic expression profiles to generate scatter plots, providing an intuitive representation of sample distribution in two-dimensional space.

Immune infiltration analysis of disulfidptosis-related tumor subtypes

The ESTIMATE algorithm was utilized to analyze gene expression profiles of tumor samples, thereby assessing the infiltration levels of stromal and immune cells. The xCell algorithm was applied to evaluate the infiltration levels of 64 distinct cell types using single-sample gene set enrichment analysis (ssGSEA). The Wilcoxon rank-sum test was used to determine the statistical significance of differences between the two subtypes, and the results were visualized using violin plots, boxplots, and clustered heatmaps.

DEGs and GSEA analysis of disulfidptosis-related tumor subtypes

DEGs in different tumor subtypes were identified using the limma package for transcriptome analysis. A filtering criterion was applied with a P < 0.05 and a fold change greater than 1 to determine the DEGs. The p-values were adjusted using the Benjamini-Hochberg (BH) method. Subsequently, significant KEGG pathways with alterations across different tumor subtypes were calculated using GSEA, with a threshold of P < 0.05.

Establishment and validation of a melanoma prognosis model associated with disulfidptosis

First, a univariate Cox regression analysis was performed on the DEGs associated with disulfidptosis-related tumor subtypes to identify genes correlated with prognosis. Subsequently, a prognostic risk model consisting of eight DEGs was constructed using Lasso regression, and the risk score calculation formula was ΣExp*Coef, where Exp and Coef represent the expression level and coefficient of each gene, respectively. Univariate survival analysis was conducted for the eight genes associated with the risk model, along with Time-dependent receiver operating characteristic (ROC) curve analysis of the risk score. The model validity was further validated using the GSE65904 dataset. Finally, a nomogram model was developed by integrating clinical variables gender, age, and stage, with the risk score. Additionally, genomic heterogeneity data from TCGA-SKCM, including microsatellite instability (MSI), tumor mutational burden (TMB), homologous recombination deficiency (HRD), and loss of heterozygosity (LOH), were used to perform differential tests between high- and low-risk groups. A heatmap was employed to display the correlation coefficients between the eight risk model genes and fifteen DRGs.

scRNA-seq validation

The scRNA-seq data from the GSE72056 dataset were used as the research object, and preprocessing was performed using the Seurat package. First, the raw data were normalized using the NormalizeData method. Next, dimensionality reduction was performed on the data matrix using RunPCA and RunTSNE. Subsequently, cells were clustered into subgroups based on different resolutions. The optimal resolution was determined to be 0.3, and the cells were annotated and classified. A Sankey diagram was used to display the associations between cell subgroups, malignant tumor cells, and cell types. Finally, the specific expressions of risk model-related genes across different cell types were visualized using heatmaps and violin plots.

Collection of clinical samples and IHC staining

Following approval by the Medical Ethics Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology (Ethics Approval No: TJ-IRB202308125), this study collected paraffin-embedded tissue sections of melanoma from three different patients and normal melanocytic nevus tissues from another three different patients from the Department of Pathology at Tongji Hospital. Immunohistochemistry (IHC) was performed to evaluate the expression of TMX4 in the collected tissues. The paraffin-embedded sections underwent sequential processing, including antigen retrieval, blocking of endogenous peroxidase activity, and blocking of non-specific binding sites. Subsequently, the sections were incubated with a primary antibody against TMX4 (21348-1-AP, Proteintech, USA; 1:100) at 4 °C overnight. On the following day, the sections were incubated with an HRP-conjugated secondary antibody (Goat Anti-Rabbit IgG(H + L), SA00001-2, Proteintech, USA; 1:5000) at room temperature for 1 h. Finally, DAB staining was performed, followed by dehydration, clearing, and mounting of the sections. The stained sections were observed under an inverted optical microscope (CX23, Olympus, Japan), and images were captured.

Cell culture

The human malignant melanoma cell lines A375 and A875 used in this study were obtained from the cell bank of Suzhou Haixin Biotechnology. (Resource numbers TCH-C114 and TCH-C203, respectively). Cells were cultured in DMEM medium (KGL1211-500, KeyGEN Bio TECH, China) supplemented with 10% fetal bovine serum (A5256701, Gibco, USA) and 1% penicillin-streptomycin solution (15140122, Gibco, USA). Cultures were maintained in a humidified incubator at 37 °C with 5% CO₂.

Transfection of cells

Specific siRNA targeting the TMX4 gene was synthesized (siRNA sequences are shown in Table S1). According to the manufacturer’s instructions, cells were incubated with the TMX4-targeting siRNA using Lipofectamine 3000 reagent (L3000015, Invitrogen, USA).

qRT-PCR

Total RNA was extracted in an RNase-free environment using the FastPure Cell/Tissue Total RNA Isolation Kit V2 (RC112-01, Vazyme, China) following the manufacturer’s protocol, which includes cell lysis, filtration, centrifugation, and washing steps to obtain high-purity RNA. The RNA concentration was measured using a Thermo Scientific NanoDrop spectrophotometer (Kaiao, China). Subsequently, cDNA was synthesized using the Hifair^® III 1 st Strand cDNA Synthesis SuperMix for qPCR (gDNA digester plus) kit (11141ES10, Yeasen, China) according to the manufacturer’s instructions. Quantitative PCR was performed using the Hieff^® qPCR SYBR Green Master Mix (Low Rox Plus) kit (11202ES03, Yeasen, China), with primers (sequences listed in Table S2) added to the reaction mixture. Amplification was conducted using a PCR instrument (Applied Biosystems, USA), and the relative expression levels of the TMX4 gene in melanoma cells under different treatments were analyzed using the 2^−ΔΔCt method.

CCK-8 assays and EdU assays

CCK-8 assays were performed using the CCK-8 kit (CA1210, Solarbio, China) according to the manufacturer’s instructions. Pre-treated cells were seeded in a 96-well plate, and the absorbance at 450 nm was measured at 0, 24, 48, and 72 h using a microplate reader to determine the optical density (OD) values of each well.

Cell proliferation was assessed using the EdU Cell Proliferation Imaging Kit (KTA2030, Abbkine, China). After seeding the cells in a 96-well plate, the cells were incubated with EdU working solution for 2 h. Following fixation, washing, permeabilization, and a second washing step, the Click-iT reaction mixture was added. After washing the samples, Hoechst 33,342 nuclear staining was performed. Finally, images were captured using a laser scanning confocal microscope (FV1000, Olympus, Japan), and the cell proliferation capacity was evaluated by calculating the proliferation index.

Cell scratch assay and transwell cell migration assay

For the cell scratch assay, cells were seeded in 6-well plates and allowed to reach confluence. A vertical scratch was then created across the cell monolayer. The wells were gently washed with PBS to remove detached cells, and serum-free high-glucose DMEM was added. Images of the scratch area were captured at 0, 24, 48, and 72 h using an inverted microscope to monitor scratch closure. Cell migratory capacity was evaluated by calculating the migration rate.

For the Transwell migration assay, 500 µL of high-glucose DMEM was first added to the lower chamber of each well in a 24-well plate. The prepared cell suspension was then added to the upper chamber of the Transwell insert, and the plate was incubated for 48 h. After incubation, the inserts were fixed with 4% paraformaldehyde, and the remaining cells on the upper surface of the membrane were gently removed using a cotton swab. The inserts were subsequently stained with crystal violet. Finally, migrated cells on the lower surface of the membrane were observed and counted under an inverted microscope.

Measurement of intracellular GSH, MDA, and ROS levels

Intracellular GSH levels were measured using a reduced GSH assay kit (A006-2-1, Nanjing Jiancheng, China). The absorbance of each well was measured at 405 nm using a microplate reader. Intracellular malondialdehyde (MDA) levels were measured using a MDA assay kit (ADS-F-YH002-96, AIDISHENG, China). After following the manufacturer’s instructions, the absorbance of each well was measured at wavelengths of 532 nm and 600 nm using a microplate reader.

Intracellular reactive oxygen species (ROS) levels were measured using a ROS detection assay kit (CA1410, Solarbio, China). The DCFH-DA probe was diluted in DMEM high-glucose medium to a final concentration of 10 µmol/L at a 1:1000 ratio. When the cell confluence reached 80–90%, the culture medium was removed, and 1 mL of the diluted DCFH-DA was added. After incubating in a cell culture incubator for 20 min, cells were washed three times with DMEM high-glucose medium. Cellular fluorescence was observed using a confocal laser microscope, and the average fluorescence intensity was quantified.

Western blot (WB) analysis

After lysing the collected melanoma cells, the supernatant was centrifuged and collected. A protein loading buffer (G2075-1ML, Servicebio, China) was added, and the mixture was heated to extract total protein. A 10% PAGE gel was prepared using a PAGE gel quick preparation kit (PG112, Epizyme, China), and electrophoresis was performed after preparing the running buffer. The proteins were subsequently transferred to a PVDF membrane. After transfer, the membrane was blocked using a rapid blocking solution (PS108P, Epizyme, China). The membrane was incubated with primary antibodies: TMX4 Polyclonal Antibody (21348-1-AP, Proteintech, USA; 1:1000), SLC7A11 Polyclonal Antibody (26864-1-AP, Proteintech, USA; 1:2000), and Beta Actin Monoclonal Antibody (66009-1-Ig, Proteintech, USA; 1:50000), followed by incubation with corresponding secondary antibodies: HRP-conjugated Goat Anti-Rabbit IgG (H + L) (SA00001-2, Proteintech, USA; 1:5000) and HRP-conjugated Goat Anti-Mouse IgG (SA00001-1-A, Proteintech, USA; 1:5000). Finally, the membrane was developed using enhanced chemiluminescent substrate (BL523B, Biosharp, China) and photographed.

Statistical analysis of the data

For statistical methods in bioinformatics analysis, differential analysis was conducted using the Wilcoxon test, survival significance was evaluated using the Log-rank test, and correlation coefficients of gene expression levels were calculated using the Pearson method. Hierarchical clustering was employed for clustering analysis. All statistical methods related to bioinformatics analysis were implemented using R software.

In addition, image processing and measurements were performed using Image J software. Experimental data calculations and graphical representations were generated using GraphPad Prism 8 software. For normally distributed data, the results are expressed as mean ± standard error (x ± s). Between-group comparisons were conducted using the t-test, while comparisons among multiple groups were analyzed using one-way analysis of variance (ANOVA). In cases of homogeneity of variance, ANOVA was applied, while Welch’s test was used when variance was heterogeneous. A p-value less than 0.05 was considered statistically significant, and a p-value less than 0.01 was considered highly statistically significant.

Results

Identification and characterization of disulfidptosis-related tumor subtypes

Consensus clustering analysis was performed on TCGA-SKCM samples with the help of the R package, identifying 2 as the optimal number of clusters. Based on the expression patterns of DRGs, the TCGA-SKCM samples were clearly divided into 2 groups: Subtype 1 (n = 274) and Subtype 2 (n = 194) (Fig. 1A–C). Using PCA, t-SNE, and UMAP dimensionality reduction algorithms to analyze the expression profiles of DRGs, the two tumor subtypes were clearly separated (Fig. 1D-F), confirming the ability of DRGs to effectively distinguish melanoma subtypes.

Fig. 1.

Identification of melanoma subtypes based on DRGs. (A-C) Clustering process and heatmap of the optimal consensus clustering results. TCGA-SKCM samples are clearly classified into two groups based on the expression profiles of disulfidoptosis-related genes. (D-F) Two-dimensional scatter plots of dimensionality reduction using PCA, t-SNE, and UMAP, with colors indicating different subtypes

In tumor microenvironment (TME) research, the stromal score is typically used to estimate the relative abundance of stromal cells, while the immune score reflects the relative abundance of immune cells. The ESTIMATE score integrates both to evaluate the overall state of the TME [17]. Using the ESTIMATE algorithm, studies revealed that Subtype 2 exhibited a higher ESTIMATE score compared to Subtype 1, indicating a more abundant presence of stromal cells and immune cell infiltration (Fig. 2A). Further validation using the xCell algorithm showed a significant increase in the infiltration abundance of various immune cells in Subtype 2 (Fig. 2B-C). The higher ESTIMATE score and greater cell infiltration may suggest that Subtype 2 is associated with a more favorable prognosis.

Fig. 2.

Immune infiltration analysis of disulfidoptosis-related tumor subtypes. (A) Differences in stromal score, immune score, and ESTIMATE score between the two melanoma subtypes. (B) Clustered heatmap of 18 differentially infiltrated cell types. (C) Identification of 18 cell types with significant intergroup differences using xCell. (^* P < 0.05; ^** P < 0.01; ^*** P < 0.001; ^**** P < 0.0001)

At the genomic level, the two subtypes also showed differences in the mutation profiles of DRGs and copy number variation (CNV) patterns (Figure S1). Meanwhile, the expression patterns of DRGs themselves also exhibited distinct characteristics between the subtypes (Figure S2). Notably, the differences in immune features between the subtypes were reflected not only in the levels of cell infiltration but also extended to the expression of immune regulatory molecules (Figure S3). The dual heterogeneity of the transcriptome and genome together outlines the unique biological profiles of two disulfidoptosis-related melanoma subtypes, providing a molecular foundation for further functional analysis and clinical translation.

DEGs and GSEA of disulfidptosis-related tumor subtypes

Transcriptomic data were analyzed using the Limma package to identify DEGs among the distinct subtypes (Fig. 3A). Based on GSEA, KEGG pathways significantly upregulated in Subtype 1 were found to be associated with DNA double-strand break repair, ubiquitin-mediated proteolysis, cell growth and differentiation, learning and memory formation, and regulation of gene expression (Fig. 3B). In contrast, pathways significantly upregulated in Subtype 2 included graft-versus-host disease, autoimmune thyroid disease, immune rejection, autoimmune destruction of insulin-secreting cells, and chronic airway inflammation (Fig. 3C). These markedly upregulated pathways might have played critical roles in the initiation and progression of different melanoma subtypes.

Fig. 3.

DEGs and GSEA of tumor subtypes associated with disulfidptosis. (A) Volcano plot of DEGs between the two melanoma subtypes. (B) KEGG pathways significantly enriched in Subtype 1. (C) KEGG pathways significantly enriched in Subtype 2

Construction and validation of the prognostic risk model for melanoma

Univariate Cox regression analysis was conducted on DEGs associated with disulfidoptosis-related tumor subtypes, and 19 DEGs significantly correlated with the overall survival of melanoma patients were initially identified (Fig. 4A). Among the identified DEGs, TYRP1 was the only factor recognized as a risk contributor. Using Lasso regression analysis, 8 key DEGs were identified when the optimal λ value was set to 8 (Fig. 4B), including CD19, CD79A, IFI27, ITGA10, TMX4, TYRP1, UCP2, and WWTR1. Survival analysis demonstrated that the high expression of genes in the model except TYRP1, was associated with longer survival (Fig. 4C).

Fig. 4.

Identification of prognostic genes in disulfidptosis-associated tumor subtypes. (A) Univariate Cox regression forest plot showing that 19 DEGs are significantly associated with melanoma prognosis. (B) Lasso variable selection process. The minimum partial likelihood deviance is achieved when the number of variables is 8. (C) Survival curves of genes included in the Lasso-based risk model. Patients were stratified into high- and low-expression groups based on the median expression level

A prognostic risk model was constructed based on Lasso regression, with the risk score calculated using the formula ΣExp*Coef. Using the median value of the risk score as the cutoff, cancer samples from two cohorts (TCGA-SKCM and GSE65904) were divided into high-risk and low-risk groups. Survival analysis results confirmed that the high-risk group in the training set (TCGA-SKCM) exhibited a more rapid decline in survival probability, indicating a shorter overall survival time for the high-risk group (Fig. 5A). Similar results were observed in the validation cohort (GSE65904), where the high-risk group also demonstrated worse prognosis (Fig. 5B). To evaluate the model’s predictive accuracy, we conducted a time-dependent ROC analysis and achieved a highest AUC of 0.73, demonstrating the model’s strong predictive performance (Fig. 5C-D). Furthermore, we combined the risk score with clinical parameters of the samples, such as gender, age, and cancer stage, to construct a nomogram (Figure S4), enabling a more intuitive and rapid evaluation of patients’ treatment outcomes and survival expectations.

Fig. 5.

Validation of the melanoma prognosis-related risk model. (A) Survival curves of high- and low-risk groups in the TCGA-SKCM cohort. (B) Survival curves of high- and low-risk groups in the GSE65904 cohort. (C) 1- to 3-year ROC curves for survival prediction in the TCGA-SKCM cohort. (D) 1- to 3-year ROC curves for survival prediction in the GSE65904 cohort

Expression of risk model-associated genes in different cell types

To validate the expression characteristics of key genes in the prognostic risk model at the single-cell level, we analyzed the GSE72056 dataset. The data were processed and clustered using the standard Seurat workflow, successfully identifying eight major cell types in the melanoma microenvironment, including malignant melanoma cells, T cells, B cells, macrophages, and others (Figure S5).

At the end of the study, we analyzed the differential expression of the eight risk model-related genes across different cell types and generated an overall heatmap (Fig. 6A) as well as individual violin plots for each gene (Fig. 6B). The TMX4 gene is primarily involved in regulating the intracellular redox state and protein folding [18]. Cadherin 19, encoded by the CDH19 gene, functions as a key mediator of intercellular adhesion [19]. Comparative analysis demonstrated that TMX4 and CDH19 showed significantly higher expression in malignant melanoma cells compared to seven other distinct cell types. The elevated expression of TMX4 and CDH19 suggested a potential role in the initiation and progression of melanoma. The UCP2 gene encodes a mitochondrial membrane transporter that is involved in the regulation of energy homeostasis [20]. The expression of the UCP2 gene was relatively high in immune cells, suggesting a potential involvement of UCP2 in regulating the metabolic state of immune cells within the melanoma TME. Using genomic heterogeneity data from TCGA-SKCM to conduct differential tests between high-risk and low-risk groups, we found no significant differences in the distribution of TMB, MSI, HRD, and LOH between the two groups. This indicated that these genomic heterogeneity markers did not have a significant impact on the prognostic risk model of melanoma (Figure S6A). In addition, the correlation analysis between the expression of risk model-associated genes and DRGs revealed significant co-expression patterns among certain genes (Figure S6B). These associations provided novel insights into understanding the molecular characteristics of melanoma.

Fig. 6.

Expression of risk model–associated gene. (A) Heatmap showing the expression profiles of the eight genes associated with the risk model. (B) Violin plots illustrating the expression levels of these eight genes across different cell types

High expression of TMX4 in human melanoma tissues

In bioinformatics analysis, we found that the TMX4 gene is highly expressed in malignant melanoma cells. We used IHC to detect the expression of TMX4 in the tissue samples of melanoma patients. Compared to the control group of normal melanocytic nevus sections, the melanoma tissue sections from patients exhibited higher staining intensity (Fig. 7A-C). It indicated that TMX4 expression was significantly elevated in human melanoma tissues compared to normal melanocytic nevus tissues.

Fig. 7.

The expression level of the TMX4 gene. (A) IHC images of pathological sections from the patient group (melanoma) under light microscopy at 100× and 200× magnifications. (B) IHC images of pathological sections from the control group (melanocytic nevus) under light microscopy at 100× and 200× magnifications. (C) Bar graph showing the IHC staining intensity in the patient and control groups. (D) Relative expression levels of the TMX4 gene in A375, A375 transfected with negative control siRNA, and A375 transfected with siRNA TMX4-157. (E) Relative expression levels of the TMX4 gene in A875, A875 transfected with negative control siRNA, and A875 transfected with siRNA TMX4-611. (ns indicates no significant difference; ^* P < 0.05; ^** P < 0.01; ^*** P < 0.001)

Inhibitory effect of siRNA on TMX4 gene expression in melanoma

Bioinformatics analysis and IHC results demonstrated that TMX4 was highly expressed in melanoma. To investigate the role of TMX4 in melanoma, we utilized siRNA to knock down TMX4 expression and examined the effects of both high and low expression levels of TMX4. After transfecting A375 and A875 cells with specific siRNA, qRT-PCR analysis revealed that siRNA TMX4-157 and siRNA TMX4-611 were the most effective in inhibiting TMX4 expression in A375 and A875 cells, respectively (Fig. 7D-E). In subsequent experiments, we used these two siRNAs to model melanoma cell lines.

Suppression of TMX4 gene expression reduces melanoma cell proliferation and migration

In the CCK-8 assay, A375 cells transfected with siRNA TMX4-157 and A875 cells transfected with siRNA TMX4-611 showed significantly lower absorbance on days 2 and 3 compared to the control group (Fig. 8A-B), indicating that the inhibition of TMX4 expression reduced the proliferative capacity of melanoma cells. The EdU assay demonstrated a higher proliferation index in cells transfected with the negative control siRNA (Fig. 8C-F), which was consistent with the results from the CCK-8 assay.

Fig. 8.

The effect of the TMX4 gene on the proliferation of melanoma cells. (A) Absorbance at 450 nm of A375 cells transfected with siRNA TMX4-157 or negative control siRNA at different time points. (B) Absorbance at 450 nm of A875 cells transfected with siRNA TMX4-611 or negative control siRNA at different time points. (C) Fluorescence microscopy images of A375 cells transfected with siRNA TMX4-157 or negative control siRNA. Cells labeled with AbFluor 488 were observed using Ex/Em = 501/525 nm, and nuclei stained with Hoechst 33342 were observed using Ex/Em = 360/460 nm (400× magnification). (D) Proliferation index of A375 cells transfected with different siRNAs. (E) Fluorescence microscopy images of A875 cells transfected with siRNA TMX4-611 or negative control siRNA. Cells labeled with AbFluor 488 were observed using Ex/Em = 501/525 nm, and nuclei stained with Hoechst 33342 were observed using Ex/Em = 360/460 nm (400× magnification). (F) Proliferation index of A875 cells transfected with different siRNAs. (^* P < 0.05; ^** P < 0.01; ^**** P < 0.0001)

The cell scratch assay showed that the scratch closure rate was significantly delayed in A375 cells transfected with siRNA TMX4-157 and A875 cells transfected with siRNA TMX4-611 (Fig. 9A-D). The Transwell migration assay further validated that siRNA-mediated knockdown of TMX4 resulted in a significant reduction in the number of transmembrane A375 and A875 cells compared to the control group (Fig. 9E-G). These two independent functional assays collectively demonstrate that suppressing TMX4 expression markedly weakens the migratory potential of melanoma cells.

Fig. 9.

Effect of TMX4 gene on the migration of melanoma cells. (A) Representative images of scratch wound closure in A375 cells for the experimental and control groups (40× magnification). (B) Representative images of scratch wound closure in A875 cells for the experimental and control groups (40× magnification). (C) Quantitative analysis of cell migration rates in A375 cells for the experimental and control groups. (D) Quantitative analysis of cell migration rates in A875 cells for the experimental and control groups. (E) Crystal violet staining images of A375 and A875 cells in the experimental and control groups (200× magnification). (F) Quantification of migrated cells in the experimental and control groups of A375 cells. (G) Quantification of migrated cells in the experimental and control groups of A875 cells. (^* P < 0.05; ^*** P < 0.001; ^**** P < 0.0001)

The effect of TMX4 gene on redox homeostasis in melanoma cells

Disulfidptosis is closely associated with intracellular redox homeostasis. Disulfidptosis induces cell death by depleting reducing molecules such as GSH and NADPH, leading to the formation of aberrant disulfide bonds and the accumulation of ROS [21]. TMX4 protein is a reductase encoded by the TMX4 gene [22], and previous bioinformatics analyses have suggested a potential link between TMX4 and disulfidptosis. Therefore, we aimed to assess the effect of TMX4 on redox homeostasis in melanoma cells by measuring key molecules associated with intracellular redox status and oxidative stress, including GSH, MDA, and ROS.

Compared to the control group, A375 cells transfected with siRNA TMX4-157 and A875 cells transfected with siRNA TMX4-611 showed significant decreases in GSH levels (Fig. 10A), increased MDA accumulation (Fig. 10B), and elevated intracellular ROS concentrations (Fig. 10C-D). These results indicated that the expression of the TMX4 gene in melanoma cells was closely associated with cellular redox homeostasis. The inhibition of TMX4 significantly enhanced oxidative stress, thereby increasing the risk of cellular oxidative damage.

Fig. 10.

The effect of TMX4 gene on redox homeostasis in melanoma cells. (A) GSH levels in the experimental and control groups of A375 and A875 cells. (B) MDA levels in the experimental and control groups of A375 and A875 cells. (C) Green fluorescence of melanoma cells in the experimental and control groups of A375 and A875 observed under a fluorescence microscope using Ex/Em = 488/525 nm (400×). (D) Average green fluorescence intensity in the experimental and control groups of A375 and A875 cells. (^* P < 0.05; ^*** P < 0.001; ^**** P < 0.0001)

Inhibition of TMX4 expression upregulates SLC7A11 expression

Previous studies have identified a potential association between the TMX4 gene and disulfidptosis, and inhibiting the expression of TMX4 in melanoma cells significantly reduces the intracellular antioxidant GSH levels. SLC7A11 is the critical subunit of system Xc⁻ that mediates the antiporter activity transporting extracellular cystine into cells in exchange for intracellular glutamate. Once extracellular cystine enters the cell through SLC7A11, cystine is reduced to cysteine via an NADPH-dependent reduction reaction, and cysteine serves as a critical precursor for GSH synthesis [23]. Furthermore, the high expression of SLC7A11 is one of the hallmark features of disulfidptosis. Therefore, we employed WB analysis to investigate the impact of TMX4 downregulation on the expression of SLC7A11 in melanoma cells by examining protein band patterns.

Finally, WB results showed that TMX4 expression levels were significantly lower in A375 cells transfected with siRNA TMX4-157 and in A875 cells transfected with siRNA TMX4-611 compared to the control group (Fig. 11A-D), which was consistent with the qRT-PCR results. Meanwhile, the expression levels of SLC7A11 protein were markedly increased in these cells relative to the control group (Fig. 11A-D), suggesting that TMX4 inhibition in melanoma cells might upregulate SLC7A11 expression.

Fig. 11.

The effect of the TMX4 gene on the expression of SLC7A11. (A) Protein electrophoresis bands of β-Actin (internal control) and target protein TMX4 and SLC7A11 in the experimental and control groups of A375 cells. (B) Bar graph showing the relative expression levels of TMX4 and SLC7A11 protein in the experimental and control groups of A375 cells. (C) Protein electrophoresis bands of β-Actin (internal control) and target protein TMX4 and SLC7A11 in the experimental and control groups of A875 cells. (D) Bar graph showing the relative expression levels of TMX4 and SLC7A11 protein in the experimental and control groups of A875 cells. (^* P < 0.05; ^** P < 0.01)

Discussion

Currently, basic research on disulfidptosis in melanoma is still in the early stages of exploration. Similar studies focused on establishing prognostic prediction models for melanoma based on disulfidptosis typically involve constructing and validating models directly after screening disulfidptosis-related features [24]. For instance, models have been constructed using DRGs and disulfidptosis-related lncRNAs to predict melanoma prognosis and the efficacy of immune therapy [25–27]. In contrast, the present study did not directly construct a prognostic model based on DRGs after screening. Instead, DRGs were used to classify melanoma samples into subtypes, and the accuracy of the classification was subsequently validated. DEGs between the tumor subtypes were then analyzed, and a prognostic prediction model was constructed based on the identified DEGs. This approach not only ensured the connection with disulfidptosis but also expanded the analytical features used for model construction. The composition and functional state of the tumor immune microenvironment are key factors influencing melanoma progression and therapeutic response [28]. Previous studies have shown that metabolic reprogramming of tumor cells, particularly alterations in redox states, can significantly impact the activity and function of immune cells, thereby shaping an immunosuppressive or immune-activating microenvironment [29, 30]. This study observed that distinct tumor subtypes associated with disulfidptosis exhibit unique immune infiltration patterns, providing insights into how disulfidptosis may contribute to tumor-immune interactions. The prognostic model developed in this study not only enables personalized evaluation of patient outcomes but also identifies key genes and pathways that hold potential as therapeutic targets, offering new perspectives for overcoming heterogeneity and expanding treatment strategies.

The TMX4 gene is located on human chromosome 20 and encodes thioredoxin-related transmembrane protein 4, a member of the protein disulfide isomerase (PDI) family [31]. PDI assists in protein folding by catalyzing the formation of intramolecular and intermolecular disulfide bonds (oxidation), and also corrects structural errors by breaking non-native disulfides and facilitating their rearrangement into native disulfides (isomerization) [32]. By reducing intramolecular and intermolecular disulfide bonds, PDI facilitates the translocation of terminally misfolded polypeptide chains across the endoplasmic reticulum membrane, thereby promoting their degradation by proteases. In addition to these functions, PDI also acts as a regulator of calcium homeostasis within the endoplasmic reticulum lumen and participates in the formation of oligomeric structures [33]. As a reductase, TMX4 can participate in the dissociation of the LINC complex under endoplasmic reticulum stress through the reductive activity of TMX4, thereby regulating the dynamic remodeling of the nuclear envelope [34]. Besides, recent studies have demonstrated that TMX4 not only enhances platelet-mediated activation of the coagulation system and promotes the formation of thrombin and fibrin, but also reduces the formation of disulfide bonds in platelet integrin αIIbβ3 through the reductase activity of TMX4, thereby enhancing platelet aggregation [35].

TMX4 possesses reductase activity and can regulate the formation and reduction of disulfide bonds in proteins. One of the hallmark features of disulfidptosis is the aberrant formation of disulfide bonds within the cytoskeleton. In addition, bioinformatics analyses have demonstrated that TMX4 significantly affects the survival of melanoma patients and contributes to the construction of prognostic risk models. TMX4 is also highly expressed in melanoma. Therefore, based on these bioinformatics findings, this study further investigates the role of TMX4 in melanoma. Through cellular experiments, we observed that knockdown of the TMX4 gene could inhibit the proliferation and migration of melanoma cells. Moreover, TMX4 knockdown was accompanied by a decrease in intracellular antioxidant GSH levels, an increase in the levels of the lipid peroxidation product MDA, elevated ROS levels, and upregulation of SLC7A11 expression. Combined with bioinformatics analysis suggesting a specific association between the TMX4 gene and disulfidptosis, we propose that these experimental findings collectively indicate a possibility: the SLC7A11 upregulation and redox imbalance caused by TMX4 knockdown may render cells more susceptible to disulfidptosis. We speculate that this disruption of redox homeostasis is a potential factor leading to the inhibition of cell proliferation and migration. Of course, the upregulation of SLC7A11 may also be involved in other cellular pathways, and the specific molecular mechanisms by which TMX4 influences disulfidptosis require further elucidation.

Although the prognostic risk model constructed in this study demonstrates a certain predictive capability, allowing for differentiation of patient outcomes and identification of high-risk individuals to support personalized treatment strategies, it still faces several challenges in clinical application. The modeling data are derived from multiple databases, introducing heterogeneity that may affect the model accuracy and stability [36]. Furthermore, the limited sample size in the independent validation cohort raises concerns about the model generalizability. In practical applications, the integration of genetic characteristics with clinical variables is required to address the challenge of reduced predictive accuracy caused by individual variability [37]. Future research should focus on large-scale, multi-center prospective studies to optimize the prognostic risk model and validate the clinical efficacy of the improved model, thereby maximizing the potential contribution to melanoma diagnosis and treatment.

Finally, we must acknowledge some specific issues in terms of the methods and results within this study. The detailed molecular mechanisms underlying the disruption of intracellular redox homeostasis and the reduction in cell proliferation and migration capacity caused by TMX4 gene knockdown, are not thoroughly investigated. Future studies are needed to further explore the regulatory role of TMX4 in the expression of SLC7A11. In addition, the impact of TMX4 gene knockdown on the melanoma immune microenvironment is not evaluated. In bioinformatic analysis, high expression of the TMX4 gene is associated with favorable patient prognosis, whereas in cellular validation experiments, high TMX4 expression promotes functions of melanoma cells. This seemingly contradictory observation may indicate tissue-specific TMX4 expression patterns or tumor heterogeneity. In different cell types or within distinct TME, TMX4 may function through diverse molecular mechanisms. Future studies are needed to explore the role of the TMX4 gene in melanoma more comprehensively through in-depth investigations using cellular and animal models. Such research will provide a scientific basis for the development of novel therapeutic strategies.

Conclusions

In this study, DRGs in melanoma were systematically analyzed using bioinformatic approaches, leading to the identification of tumor subtype characteristics based on disulfidptosis. Furthermore, DEGs among these melanoma subtypes were used to construct an effective prognostic risk model. By integrating bioinformatics findings with cellular experiments, we observed that TMX4 knockdown in melanoma cells induces upregulation of SLC7A11 expression, accompanied by disruption of intracellular redox homeostasis and inhibition of tumor cell proliferation and migration.

Abbreviations

AUC

Area under the curve

BH

Benjamini-Hochberg

CNV

Copy number variation

DEGs

Differentially expressed genes

DRGs

Disulfidptosis-related genes

GSH

Glutathione

HRD

Homologous recombination deficiency

IHC

Immunohistochemistry

LOH

Loss of heterozygosity

MDA

Malondialdehyde

MSI

Microsatellite instability

ANOVA

One-way analysis of variance

OS

Overall survival

scRNA-seq

Single-cell RNA sequencing

PCD

Programmed cell death

PAC

Proportion of ambiguous clustering

PDI

Protein disulfide isomerase

ROS

Reactive oxygen species

ROC

Receiver operating characteristic

ssGSEA

Single-sample gene set enrichment analysis

t-SNE

t-distributed stochastic neighbor embedding

TME

Tumor microenvironment

TMB

Tumor mutational burden

WB

Western blot

Author contributions

Y.Y. and R.T. performed the experimental data analysis and wrote the main manuscript text. Z.S. and L.C. collected clinical samples and conducted bioinformatics analysis. Z.Z. and M.W. coordinated the project and revised the manuscript. All authors reviewed the manuscript.

Funding

This work was supported by the Online Education and Teaching Research Project for Graduate Students in Chinese Medicine (Grant No. B_YXC2024-02-03_10).

Data availability

The bioinformatics analysis data are obtained from public databases. Tumor tissue sample data are sourced from the TCGA-SKCM project. Datasets GSE65904 and GSE72056 are retrieved from the GEO database. The DRGs are obtained from the literature by Liu et al. Other relevant experimental data in this study are available upon request from the corresponding author.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with medical ethics standards and approved by the Ethics Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology (Approval No: TJ-IRB202308125). All participants provided informed consent and authorized the use of their samples for research purposes.

Consent for publication

All authors have approved of the consents of this manuscript and provided consent for publication.

Competing interests

The authors declare no competing interests.