Intratumoral disulfidptosis heterogeneity in triple-negative breast cancer, a multiomics integration analysis

Introduction

Breast cancer is a common malignant tumor that significantly threatens women’s health (1). Triple-negative breast cancer (TNBC), which accounts for approximately 20% of all breast cancer cases, is distinguished by its high degree of malignancy (2). For TNBC, chemotherapy remains the cornerstone of treatment. Despite the emergence of a spectrum of novel therapeutic approaches in recent years, such as PARP inhibitors, antibody-drug conjugates, and immunotherapy, a subset of patients still exhibit suboptimal responses. Consequently, there is an urgent need to identify new therapeutic targets and develop creative treatment solutions.

Previous studies have revealed that breast cancer exhibits significant metabolic reprogramming and metabolic heterogeneity. From the perspective of energy metabolism, breast cancer can be classified into two subtypes, and further insights into its metabolic heterogeneity can be gleaned from single-cell data (3). The TNBC subgroup can be subdivided on the basis of metabolic genes into the lipid synthesis subtype, glycolytic subtype, and mixed subtype. Targeting fatty acid synthesis or glycolysis pathways in combination with immunotherapy represents a potential strategy for treating lipid synthesis and glycolytic subtypes (3). Moreover, from a metabolomic standpoint, TNBC can be categorized into fatty acid and sphingolipid types, active group transport types, and low metabolism types, which aligns well with metabolic gene classification. Key metabolites affecting TNBC tumor progression, such as sphingosine-1-phosphate and N-acetyl-aspartyl-glutamate, have been identified on this basis (4). Additionally, previous research has focused on microbial-related metabolites in TNBC, revealing that trimethylamine N-oxide plays a crucial role in antitumour immunity (5). Concurrently, studies have highlighted the heterogeneity of ferroptosis in TNBC, delineated its ferroptotic characteristics, and suggested that GPX4 inhibitors may be potential therapeutic strategies for TNBC (6). Our previous research revealed that FLAD1 regulated lipid metabolism reprogramming, indicating that FLAD1 is expected to become a biomarker for the clinical application of LSD1 and SREBP1 inhibitors (7). These studies underscore the pivotal role of metabolic reprogramming and heterogeneity in the progression of breast cancer. However, the relationship between metabolic reprogramming and tumor heterogeneity remains incompletely understood, and there is a need to further dissect the heterogeneity of breast cancer from other metabolic perspectives.

Disulfidptosis is an emerging concept in the field of metabolic reprogramming and represents a novel form of regulated cell death (RCD) that is distinct from traditional pathways such as apoptosis and necroptosis. Characterized by the formation of disulfide bonds in cellular proteins, particularly in the actin cytoskeleton, this process leads to cellular dysfunction and eventual death. In this process, the cystine transporter known as SLC7A11 is key, as it imports cystine, which is later turned into cysteine, a precursor for making glutathione. Under conditions of oxidative stress or glucose starvation, the depletion of nicotinamide adenine dinucleotide phosphate (NADPH) can lead to an imbalance in redox homeostasis, resulting in disulfide bond formation and subsequent disulfidptosis (8,9). Recent studies have highlighted the importance of disulfidptosis in tumor biology, particularly in how cancer cells exploit redox mechanisms to survive under stress conditions. In tumor cells, there is often an increase in the abundance of reactive oxygen species (ROS), which can induce oxidative damage but also serve as signalling molecules that promote tumorigenesis (10,11). The imbalance of redox processes in cancer cells contributes not only to their survival but also to their ability to evade conventional therapies, making the understanding of disulfidptosis particularly relevant for the development of new therapeutic strategies (12,13). However, the precise mechanism through which disulfidptosis regulates TNBC progression, especially TNBC heterogeneity, requires further exploration.

Our research included a thorough examination of a TNBC multiomic cohort, with the goal of identifying disulfidptosis variability and offering potential therapeutic insights for TNBC models. We present this article in accordance with the MDAR and TRIPOD reporting checklists (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-844/rc).

Methods

Study cohorts

The TNBC cohort utilized in this research was previously described (4,14). All the samples were untreated, and complete clinical information was provided for each patient in the dataset.

Cell lines

TNBC cell lines were obtained from the American Type Culture Collection (ATCC). High-glucose DMEM containing 10% fetal bovine serum (FBS) was used to grow TNBC cells. With respect to mycoplasma detection, regular tests confirmed that none of the cell lines were contaminated with mycoplasma. The cells were incubated at 37 ℃ with 5% carbon dioxide. Moreover, short tandem repeat (STR) profiling was used to confirm the authenticity of the cell lines.

Glucose deprivation

To test the effect of glucose deprivation on TNBC cell viability, TNBC cells were cultured in six-well plates. After two weeks, the medium was replaced with glucose-free DMEM. After 24 hours, the TNBC cells were fixed with methanol and then stained with 1% crystal violet. ImageJ software was used to count the colonies.

Chemical products

BAY-876, a GLUT1 inhibitor, was obtained from MedChemExpress (Cat No. HY-100017).

Susceptibility experiment

To verify the sensitivity of the cells to BAY-876, we conducted drug susceptibility experiments in vitro. Specifically, 10,000 cells were seeded in 96-well plates. The cell culture medium was replaced, and drugs were added after 24 h. After 96 h, the cells were incubated for 1 h after CCK-8 was added to the wells at a concentration of 10 µL/well. The percentage of viable cells was determined via the following formula: Cell viability = (OD_treated / OD_untreated) × 100%, where OD represents optical density.

Consensus clustering analysis

To identify disulfidptosis clusters in TNBC patients, we conducted consensus clustering analysis via the R package ‘ConsensusCluster Plus’. The cluster number (k) was adjusted between 2 and 4, and the best k value was chosen by finding the inflection point of the sum of squared errors (SSE). The stability of disulfidptosis clusters was validated via the t-distributed stochastic neighbor embedding (t-SNE) algorithm. Furthermore, Kaplan-Meier survival analysis was employed to evaluate the recurrence-free survival (RFS) of patients within the different disulfidptosis clusters.

Differential expression enrichment analysis

Gene expression differences between the two disulfidptosis-related clusters were analysed via the limma package in R software (P<0.05 and |logFC| >1). Metabolite expression differences between the two disulfidptosis-related clusters were analysed via ANOVA (P<0.05 and |logFC| >0.58). The execution of the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses was carried out with the clusterProfiler package.

Gene Set Enrichment Analysis (GSEA) and gene mutation analysis

Pathway activation or suppression between the two clusters was determined via GSEA. The gene set ‘c2.cp.kegg.v7.5.1.symbols.gmt’ was retrieved from the MSigDB database (https://www.gsea-msigdb.org/gsea/msigdb/). The maftools R package facilitated the analysis of genetic mutation differences between the two clusters.

Immune infiltration analysis

Immunostimulants and inhibitors were obtained from previous research (14,15). Differences in immune cell infiltration between the two clusters were analysed via the CIBERSORT algorithm (16), and the results were visualized via the ggplot2 package.

DA score

The DA score indicates how likely a pathway is to exhibit higher metabolite levels than those in a control group (17). DA score analysis was carried out with reference to our previous study (4).

Construction and validation of the nomogram

To establish the most accurate and practical prediction model, the LASSO technique can be employed. Weights are assigned to the model parameters via this technique to pinpoint significant variables. We utilized the R package ‘glmnet’ to conduct LASSO analysis (18). Risk scores were calculated via the coefficients linked to each gene and metabolite associated with disulfidptosis.

In this model, patient data were analysed via the R survival package, with the log-rank test and Kaplan-Meier curves applied to assess group differences. The receiver operating characteristic curve is a standard tool for assessing how well a model can predict outcome events by measuring its sensitivity and specificity. We employed the R package ‘survival’ for ROC analysis (19). Using the median risk score as a threshold, we correlated clinical outcomes with risk scores in breast cancer patients. The research involved creating ROC curves and measuring the area under the curve (AUC) for survival at 3 and 5 years. The calibration ability of the nomograms was evaluated via 1,000 bootstrap resamplings. A line at a 45-degree angle represented perfect calibration, with its closeness indicating the calibration quality.

Statistical analysis

Statistical analysis was conducted via GraphPad Prism 9.3. The Kaplan-Meier method was used to create survival curves, which were then compared via log-rank tests. The interval between the surgery and recurrence dates was used to define the RFS. Patients who did not experience events were censored at their most recent follow-up date. A P value of less than 0.05 was considered statistically significant. The figure legends provide details on the specific statistical methods used. Three biological replicates of the experiments were carried out, and the results displayed are typical.

Results

Transcriptomic study uncovered the prognostic importance of disulfidptosis-associated genes (DRGs) in TNBC

We included 10 DRGs defined in previous studies (20,21) and analysed the expression of the indicated genes in TNBC and paracancerous tissues. The results indicated that SLC7A11, SLC3A2, PRN1, NDUFA11, LRPPRC, OXSM and GYS1 mRNA levels were upregulated in TNBC. Conversely, NCKAP1 and NUBPL were downregulated in TNBC (Figure 1A). We then investigated the prognostic value of the indicated genes in TNBC (Figure 1B-1K). The optimal cut-off for continuous gene expression analysis was determined to generate the greatest survival difference. The results of the prognostic analysis revealed that the upregulation of NCKAP1, NUBPL and LRPPRC was associated with a poor prognosis (Figure 1E,1F,1H). Conversely, high RPN1 and OXSM expression was associated with a better prognosis (Figure 1D,1I). The remaining DRGs had no significant prognostic value (Figure 1B,1C,1G,1J,1K). The above results suggest that DRGs with different expression characteristics have different prognostic values.

Figure 1 mRNA expression and prognosis of 10 DRGs in TNBC. (A) mRNA expression of 10 DRGs in TNBC tumor tissues and paratumor tissues. (B-K) The prognostic value of SLC7A11 (B), SLC3A2 (C), RPN1 (D), NCKAP1 (E), NUBPL (F), NDUFA11 (G), LRPPRC (H), OXSM (I), NDUFS1 (J) and GYS1 (K) in the TNBC cohort. The data are presented as the mean ± SEM. Statistical significance was determined by Student’s t-test (A) and log-rank test (B-K). ***, P<0.001. DRGs, disulfidptosis-associated genes; RFS, recurrence-free survival; SEM, standard error of the mean; TNBC, triple-negative breast cancer.

Molecular subtypes of TNBC in relation to DRGs expression

We next wondered whether TNBC has disulfidptosis heterogeneity. On the basis of the relative variation in the area beneath the consensus distribution function (CDF) curve, we determined the optimal K of 2 by the first “elbow” rule (Figure S1A). On the basis of the results above, we classified 360 TNBCs into two transcriptional subtypes on the basis of the mRNA expression of DRGs (Figure 2A). PCA revealed significant transcriptome differences between the two subtypes (Figure S1B). These two subtypes consisted of the following: (I) cluster 1 (n=130) and (II) cluster 2 (n=230). According to the survival analysis, patients in cluster 1 had a better prognosis than those in cluster 2 (Figure 2B). Taken together, our molecular typing of DRGs revealed disulfidptosis heterogeneity in TNBC, and we next aimed to characterize the molecular features of each subtype.

Figure 2 Identification of the molecular subtypes of TNBC on the basis of the mRNA expression of DRGs. (A) Heatmap of the expression characteristics of the two disulfidptosis-related subtypes. The samples were also annotated on top by RFS status, T stage, N stage and FUSCC TNBC subtype. (B) Kaplan-Meier analysis of the two subtypes. Statistical significance was determined by log-rank test. BLIS, basal-like immune-suppressed; DRGs, disulfidptosis-associated genes; FUSCC, Fudan University Shanghai Cancer Center; IM, immunomodulatory; LAR, luminal androgen receptor; MES, mesenchymal; RFS, recurrence-free survival; TNBC, triple-negative breast cancer.

Cluster 1 was characterized by mutation TP53 status and activation of the cell cycle

First, we evaluated the mutational signatures between the two clusters. The results indicated that cluster 1 had more TP53 mutations (TP53 mutation frequency: 82%) than did cluster 2 (TP53 mutation frequency: 70%) (Figure S2A). Next, we performed differential gene analysis for the two clusters (absolute value logFC >1 and P value <0.05). GO enrichment analysis revealed that genes that were differentially expressed between the two clusters were involved in the positive regulation of lymphocyte activation as well as the immunoglobulin complex and antigen binding (Figure S2B-S2D). KEGG pathway analysis revealed that the most upregulated pathway in cluster 1 was the cell cycle pathway (Figure 3A). GSEA revealed that the MYC target and E2F target pathways were activated in cluster 1 (Figure 3B). Hence, the above results suggest that cluster 1 is characterized by increased mutation of TP53, activation of the cell cycle and other cancer-promoting pathways, which may lead to a worse prognosis.

Figure 3 Differences in the molecular characteristics of the two clusters based on multiomics. (A) Examining KEGG enrichment in genes that show differential expression across the two clusters. (B) Analysis of differentially expressed genes between the two clusters for KEGG enrichment. (C) The expression of immunostimulatory molecules in the two clusters. (D) Volcanic graphs depicting metabolites with different abundance levels between the clusters. (E) Pathway-focused analysis of metabolomic variations between two clusters. The DA score indicates the average alterations in all metabolites within a pathway. If the score is 1, it means every metabolite in the pathway has risen in cluster 1 relative to cluster 2, whereas a score of −1 implies a decrease in all measured metabolites. For the DA score calculation, pathways containing at least three measured metabolites were utilized. (F) The expression of DTT in two clusters. The data are presented as the mean ± SEM. Statistical significance was determined by Student’s t-test. *, P<0.05; **, P<0.01; ***, P<0.001. DA, differential activity; DTT, dithiothreitol; FC, fold change; FDR, false discovery rate; FPKM, fragments per kilobase of transcript per million mapped reads; KEGG, Kyoto Encyclopedia of Genes and Genomes; SEM, standard error of the mean.

Cluster 1 represented an immunosuppressive microenvironment and activated the pentose phosphate pathway

Given the significant impact of the tumor microenvironment (TME) on the advancement of tumors, we compared the differences in immune cell infiltration between the two clusters. Combined analysis with CIBERSORT and differential expression profiling revealed that cluster 2 contained a high concentration of both immune-activated cells and immunostimulants (Figure 3C, Figure S3A). Moreover, expression profiling demonstrated that immunostimulants, particularly TNFRSF14, TGFB1 and CD96, were significantly overexpressed in cluster 2 (Figure S3B), providing further justification for the use of immune checkpoint blockade as a treatment strategy.

High SLC7A11 expression results in the transport of a large amount of cystine into the cytoplasm, cells reduce cystine to cysteine by consuming NADPH, and the generation of NADPH depends on the pentose phosphate pathway, which oxidatively decomposes glucose. Glucose deprivation causes NADPH shortage and abnormal accumulation of cystine, resulting in the formation of disulfide bonds in actin (F-actin) to collapse the network and ultimately cause disulfidptosis (21,22). Therefore, the cysteine and pentose phosphate metabolism pathways are key regulators of disulfidptosis. We next analysed the differences in the cysteine and pentose phosphate metabolism pathways between the two clusters. Analysis of the differentially abundant metabolites revealed that cystine was downregulated in cluster 1 (Figure 3D). Analysis of the differential abundance score (DA score) revealed that the pentose phosphate pathway was significantly activated in cluster 1 (Figure 3E). Moreover, we observed that dithiothreitol (DTT), a key disulfidptosis inhibitor (22), was significantly upregulated in cluster 1 (Figure 3F). Taken together, our results indicate that cluster 1 TNBCs are expected to be sensitive to disulfidptosis.

Cluster 1 was more sensitive to disulfidptosis induced by glucose deprivation and a GLUT1 inhibitor

We next aimed to assess disulfidptosis sensitivity among the two clusters. We investigated whether there was a strong correlation between our disulfidptosis-based subtypes and metabolic pathway-based subtypes (MPSs) on the basis of previous studies by Gong et al. (3). The Sankey diagram suggests that cluster 1 mainly consisted of MPS2, and cluster 2 was predominantly composed of MPS1 and MPS3 (Figure 4A). We next evaluated the sensitivity of TNBC cell lines to disulfidptosis. Considering the lack of direct disulfidptosis-targeting drugs, glucose deprivation was used to mimic disulfidptosis (22). The results indicated that HCC1806 and HCC1937 cells (cluster 1/MPS2) were the most susceptible to glucose deprivation, followed by MDA-MB-468 and MDA-MB-231 cells (cluster 2/MPS3) and finally MDA-MB-453 and BT-20 cells (cluster 2/MPS1) (Figure 4B-4E). Furthermore, we tested the sensitivity of two clusters of cell lines to BAY-876 (a GLUT1 inhibitor)-induced disulfidptosis. Susceptibility tests revealed that the cell lines represented by cluster 1 exhibited significantly greater sensitivity to BAY-876 compared to those in cluster 2 (Figure 4F). These results highlighted that cluster 1 was more sensitive to disulfidptosis.

Figure 4 Sensitivity of disulfidptosis induced by glucose deprivation and BAY-876 in the two clusters. (A) Sankey diagram of our disulfidptosis-related subtypes and MPS subtypes. (B-D) Colony formation experiments to detect the effect of glucose deprivation on HCC1806 and HCC1937 (B), MDA-MB-468 and MDA-MB-231 (C), MDA-MB-453 and BT-20 (D) cells. Cells were fixed with methanol and stained with 1% crystal violet. Bar graph (E) demonstrating the relative viability of the indicated cells under glucose deprivation conditions for 24 h; n=3 biological replicates. (F) Bar graph demonstrating the relative viability of TNBC cells treated with various concentrations of BAY-876; n=3 biological replicates. The data are presented as the mean ± SEM. Statistical significance was determined by Student’s t-test. *, P<0.05; **, P<0.01; ***, P<0.001; ns, not significant. MPS, metabolic pathway-based subtype; SEM, standard error of the mean; TNBC, triple-negative breast cancer.

Genes related to the disulfidptosis prediction model can accurately predict the prognosis of TNBC patients

Through univariate Cox regression analysis, 26 genes were found to have prognostic value among the genes differentially expressed between the two clusters (Figure 5A). LASSO Cox regression analysis identified 10 genes linked to disulfidptosis: CHIT1, ATP8B2, TNC, C10orf116, NPR3, RDH10, PKP1, F12, TNS1, and C8orf55 (Figure 5B,5C). We then applied these genes to construct a prognostic risk model (Figure S4A). To determine the risk score for each person, risk score = (The mRNA level of CHIT1 × −0.12) + (The mRNA level of ATP8B2 × −0.04) + (The mRNA level of TNC × −0.12) + (The mRNA level of C10orf116 × 0.09) + (The mRNA level of RDH10 × 0.03) + (The mRNA level of NPR3 × 0.02) + (The mRNA level of PKP1 × 0.12) + (The mRNA level of F12 × 0.03) + (The mRNA level of TNS1 × 0.31) + (The mRNA level of C8orf55 × 0.17). The predicted AUC values for the 1-, 3-, and 5-year operating curves were 0.887, 0.812 and 0.778, respectively (Figure S4B). A total of 360 patients were divided into high-risk and low-risk groups on the basis of the median risk value. Moreover, the calibration graph revealed that the forecasted probability from the nomogram was consistent with the actual RFS probability (Figure S4C). We subsequently split the patients into two categories on the basis of the median risk score. Next, we illustrated the distribution of risk scores among the two categories (Figure 5D). We then presented the survival status and duration for patients across different risk groups (Figure 5E). Moreover, as illustrated in Figure 5F, the mRNA levels of 10 genes linked to disulfidptosis were assessed for each patient. These findings revealed that patients who experienced recurrence had increased risk scores (Figure 5G). According to the survival analysis, individuals in the high-risk category experienced lower RFS rates than those in the low-risk category did (Figure 5H). Overall, genes associated with the disulfidptosis prognostic model can precisely predict the survival outcomes of TNBC patients.

Figure 5 Construction of prediction models based on disulfidptosis-related genes. (A) Analysis of RFS-associated genes linked to disulfidptosis using univariate Cox regression. (B,C) The LASSO Cox regression analysis reveals the best coefficients (B) and the smallest lambda (C) for RFS-related genes associated with disulfidptosis. (D) A depiction of how risk scores are distributed. (E) The survival status and duration of survival for each patient. (F) The expression of 10 genes associated with disulfidptosis is depicted in a heatmap for low- and high-risk categories. (G) Risk scores in groups with and without recurrence. (H) Analysis of survival rates in both low- and high-risk categories. The data are presented as the mean ± SEM. Statistical significance was determined by Student’s t-test (G). Log-rank test (H). CI, confidence interval; HR, hazard ratio; LASSO, least absolute shrinkage and selection operator; RFS, recurrence-free survival; SEM, standard error of the mean.

Metabolites related to the disulfidptosis prediction model can accurately predict the prognosis of TNBC patients

We next constructed another prediction model based on the metabolites related to disulfidptosis. We conducted univariate Cox regression analysis on the differentially expressed metabolites (absolute value LogFC >0.58 and P value <0.05) between the two clusters, leading to the discovery of 16 metabolites linked to survival (Figure 6A). Variables were subjected to additional screening via LASSO regression to discard those without clinical relevance, and 12 metabolites (Phe-Val, 2(1H)-pyridinone, Ser-Gly, D-gluconate, flavone, Lys-Cys, tridecanoic acid, melatonin, oleoyl-CoA, mevalonic acid, metanephrine and His-Met) were incorporated into the model (Figure 6B,6C). We next used the indicated metabolites to construct the model and construct a nomogram (Figure S5A). The risk score for each patient was determined according to the following formula:

Riskscore=abundanceofPhe-Val×(−0.15)+abundanceof2(1H)-Pyridinone×(−0.15)+abundanceofSer-Gly×(−0.02)+abundanceofD-gluconate×(−0.25)+abundanceofflavone×(−0.08)+abundanceofLys-Cys×(−0.02)+abundanceoftridecanoic acid×(−0.14)+abundanceofmelatonin×(−0.01)+abundanceofoleoyl-CoA×(−0.02)+abundanceofmevalonicacid×0.18+abundanceofmetanephrine×0.09+abundanceofHis-Met×0.26[1]

Figure 6 Construction of prediction models based on disulfidptosis-related metabolites. (A) Metabolites related to RFS and disulfidptosis were examined through univariate Cox regression analysis. (B,C) Through LASSO Cox regression analysis, the optimal coefficients (B) and the minimum lambda (C) for metabolites associated with RFS and disulfidptosis are determined. (D) An illustration depicting the spread of risk scores. (E) Each patient’s survival duration and their survival status. (F) Expression patterns of 12 disulfidptosis-related metabolites in low- and high-risk groups are shown in a heatmap. (G) Groups with recurrence and without recurrence risk scores. (H) Assessment of survival in low-risk and high-risk groups. The data are presented as the mean ± SEM. Statistical significance was determined by Student’s t-test (G) and Log-rank test (H). CI, confidence interval; HR, hazard ratio; LASSO, least absolute shrinkage and selection operator; RFS, recurrence-free survival; SEM, standard error of the mean.

The areas under the receiver operating characteristic (ROC) curves were 0.801, 0.743 and 0.736 for 1, 3 and 5 years, respectively (Figure S5B). A total of 360 patients were divided into a high-risk group and a low-risk group based on the median risk value. Calibration plots demonstrated that the nomogram worked effectively in relation to an ideal model (Figure S5C), indicating that the compound nomogram has excellent reliability and veracity. We subsequently split the patients into two groups according to the median risk score. After that, we showed the distribution of risk scores between the two groups (Figure 6D). Afterward, we demonstrated the survival status and timeline for patients at various risk levels (Figure 6E). Moreover, for each patient, the expression levels of 12 metabolites were assessed, as depicted in Figure 6F. These findings indicate that patients who experienced recurrence had elevated risk scores (Figure 6G). Furthermore, patients classified as high risk had a poorer prognosis (Figure 6H). Collectively, these findings confirmed that metabolites related to the disulfidptosis prediction model accurately predict the prognosis of patients with TNBC.

Discussion

Recognizing tumor heterogeneity is vital for obtaining mechanistic insights into tumor growth and for designing appropriate treatments. In our study, we analysed the heterogeneity of TNBC disulfidptosis through molecular typing. We determined that cluster 1, a group of patients with a relatively poor prognosis, may benefit from disulfidptosis-targeted treatment. The research we conducted offers more comprehensive insight into precision therapy for TNBC.

A new form of cell death, disulfidptosis, is defined by the quick loss of NADPH inside cells and the heightened expression of SLC7A11 during glucose deprivation, causing an unusual buildup of cysteine and the creation of incorrect disulfide bonds in actin cytoskeletal proteins, which eventually leads to the breakdown of the actin network (9). Recent studies have investigated the role of disulfidptosis in cancer, which may provide a new perspective for tumor treatment. For example, the disulfidptosis process is accompanied by a significant endoplasmic reticulum stress response and depends on cystine uptake mediated by SLC7A11. The combination of endoplasmic reticulum stress inhibitors and glucose transporter inhibitors has been confirmed in an in vivo tumor model to effectively promote the disulfide bonding of cytoskeletal proteins and inhibit tumor growth (23). Furthermore, the interplay between disulfidptosis and other forms of RCD, such as pyroptosis, has been explored. For example, ROS-induced pyroptosis can inhibit the NADPH supply by regulating glucose metabolism, induce cystine accumulation, and lead to disulfidptosis (24). This crosstalk between different cell death pathways underscores the complexity of the cellular response to disulfidptosis and highlights the potential for targeting these pathways in therapeutic strategies. In our study, we observed that the pentose phosphate pathway was activated in cluster 1, which provided NADPH to protect cells from endogenous disulfidptosis (21). Moreover, the abundance of DTT, which is a suppressor that prevents disulfidptosis, was increased in cluster 1 (22). We inferred that cluster 1 is characterized by the activation of endogenous disulfidptosis due to compensatory activation of the indicated pathway and suppressor. Our study also revealed that the cell lines belonging to cluster 1 are more sensitive to disulfidptosis induced by glucose deprivation, which indicates that cluster 1 may benefit from disulfidptosis-targeted treatment.

The TME is a key factor in the effectiveness of immunotherapy (25). Recent studies have highlighted the importance of understanding the complex interactions within the TME to predict and enhance the response to immunotherapy. The presence and activity of tumor-infiltrating immune cells, such as cytotoxic T lymphocytes and macrophages, are critical components that can influence treatment outcomes. For example, the density of Tbet+ tumor-infiltrating lymphocytes has been shown to reflect an effective preexisting adaptive antitumour immune response, which can be reinvigorated by treatments such as anti-PD1 therapy, irrespective of the microsatellite status of colorectal cancer (26). In our study, we observed that cluster 2 had abundant infiltration of immune-activated cells and upregulation of immune checkpoint molecules. This finding suggested that there could be a link between disulfidptosis and the TIME, which needs to be revealed by more in-depth investigations of the underlying mechanisms in the future.

Clinical prediction models have become essential tools in modern medicine, providing valuable insights into patient prognosis across various diseases. These models utilize a range of data inputs, including clinical, genetic, and imaging data, to predict outcomes such as survival rates, disease progression, and treatment responses. For example, in hepatocellular carcinoma, a model in which immune system-associated genes were used to predict clinical outcomes was developed. This model demonstrated that patients with a high immune risk score had poorer overall survival rates, highlighting the potential of immune signatures in prognostic predictions (27). For gastric cancer, a model incorporating m⁶A-related genes was constructed to predict overall survival. This model not only provides prognostic insights but also suggests potential therapeutic targets, highlighting the dual utility of such predictive frameworks (28). Another study focused on focused on lung adenocarcinomas, where a model based on methylation-related lncRNAs was developed. This model links the expression of specific lncRNAs to patient outcomes, offering a novel approach to understanding and predicting cancer progression (29). In our study, we developed two models for predicting the prognosis of TNBC patients, which could simplify treatment decision-making for patients with TNBC.

Our study has the following limitations: (I) Currently, there are no targeted drugs for disulfidptosis directly, so we did not assess whether the heterogeneity of TNBC leads to differential sensitivity to drugs targeting disulfidptosis. (II) We did not elucidate the mechanisms mediating the heterogeneity of disulfidptosis in TNBC. Understanding these mechanisms is crucial, as the heterogeneity of TNBC is a major challenge in the development of effective treatments (30). (III) Our clinical prediction model requires large-scale clinical cohort studies for further validation and generalization.

Conclusions

In summary, our study delves into the heterogeneity of disulfidptosis in TNBC and provides insights and references for precision treatment targeting disulfidptosis in TNBC. Furthermore, our predictive models, which are based on disulfidptosis-related genes and metabolites, offer valuable insights for clinical decision-making in TNBC patients.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the MDAR and TRIPOD reporting checklists. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-844/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-844/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-844/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-844/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Miller KD, Fidler-Benaoudia M, Keegan TH, et al. Cancer statistics for adolescents and young adults, 2020. CA Cancer J Clin 2020;70:443-59. [Crossref] [PubMed]
2. Boyle P. Triple-negative breast cancer: epidemiological considerations and recommendations. Ann Oncol 2012;23:vi7-12. [Crossref] [PubMed]
3. Gong Y, Ji P, Yang YS, et al. Metabolic-Pathway-Based Subtyping of Triple-Negative Breast Cancer Reveals Potential Therapeutic Targets. Cell Metab 2021;33:51-64.e9. [Crossref] [PubMed]
4. Xiao Y, Ma D, Yang YS, et al. Comprehensive metabolomics expands precision medicine for triple-negative breast cancer. Cell Res 2022;32:477-90. [Crossref] [PubMed]
5. Wang H, Rong X, Zhao G, et al. The microbial metabolite trimethylamine N-oxide promotes antitumor immunity in triple-negative breast cancer. Cell Metab 2022;34:581-594.e8. [Crossref] [PubMed]
6. Yang F, Xiao Y, Ding JH, et al. Ferroptosis heterogeneity in triple-negative breast cancer reveals an innovative immunotherapy combination strategy. Cell Metab 2023;35:84-100.e8. [Crossref] [PubMed]
7. Song XQ, Yu TJ, Ou-Yang Y, et al. Copy number amplification of FLAD1 promotes the progression of triple-negative breast cancer through lipid metabolism. Nat Commun 2025;16:1241. [Crossref] [PubMed]
8. Liu X, Zhuang L, Gan B. Disulfidptosis: disulfide stress-induced cell death. Trends Cell Biol 2024;34:327-37. [Crossref] [PubMed]
9. Xiao F, Li HL, Yang B, et al. Disulfidptosis: A new type of cell death. Apoptosis 2024;29:1309-29. [Crossref] [PubMed]
10. Helfinger V, Schröder K. Redox control in cancer development and progression. Mol Aspects Med 2018;63:88-98. [Crossref] [PubMed]
11. Kirtonia A, Sethi G, Garg M. The multifaceted role of reactive oxygen species in tumorigenesis. Cell Mol Life Sci 2020;77:4459-83. [Crossref] [PubMed]
12. Montero AJ, Jassem J. Cellular redox pathways as a therapeutic target in the treatment of cancer. Drugs 2011;71:1385-96. [Crossref] [PubMed]
13. Castaldo SA, Freitas JR, Conchinha NV, et al. The Tumorigenic Roles of the Cellular REDOX Regulatory Systems. Oxid Med Cell Longev 2016;2016:8413032. [Crossref] [PubMed]
14. Jiang YZ, Ma D, Suo C, et al. Genomic and Transcriptomic Landscape of Triple-Negative Breast Cancers: Subtypes and Treatment Strategies. Cancer Cell 2019;35:428-440.e5. [Crossref] [PubMed]
15. Angelova M, Charoentong P, Hackl H, et al. Characterization of the immunophenotypes and antigenomes of colorectal cancers reveals distinct tumor escape mechanisms and novel targets for immunotherapy. Genome Biol 2015;16:64. [Crossref] [PubMed]
16. Newman AM, Liu CL, Green MR, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 2015;12:453-7. [Crossref] [PubMed]
17. Hakimi AA, Reznik E, Lee CH, et al. An Integrated Metabolic Atlas of Clear Cell Renal Cell Carcinoma. Cancer Cell 2016;29:104-16. [Crossref] [PubMed]
18. Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw 2010;33:1-22.
19. Kamarudin AN, Cox T, Kolamunnage-Dona R. Time-dependent ROC curve analysis in medical research: current methods and applications. BMC Med Res Methodol 2017;17:53. [Crossref] [PubMed]
20. Hu K, Ou Y, Xiao L, et al. Identification and Construction of a Disulfidptosis-Mediated Diagnostic Model and Associated Immune Microenvironment of Osteoarthritis from the Perspective of PPPM. J Inflamm Res 2024;17:3753-70. [Crossref] [PubMed]
21. Liu X, Nie L, Zhang Y, et al. Actin cytoskeleton vulnerability to disulfide stress mediates disulfidptosis. Nat Cell Biol 2023;25:404-14. [Crossref] [PubMed]
22. Zheng T, Liu Q, Xing F, et al. Disulfidptosis: a new form of programmed cell death. J Exp Clin Cancer Res 2023;42:137. [Crossref] [PubMed]
23. Wang J, Chen J, Fan K, et al. Inhibition of Endoplasmic Reticulum Stress Cooperates with SLC7A11 to Promote Disulfidptosis and Suppress Tumor Growth upon Glucose Limitation. Adv Sci (Weinh) 2025;12:e2408789. [Crossref] [PubMed]
24. Zhu Y, Wang X, Feng L, et al. Intermetallics triggering pyroptosis and disulfidptosis in cancer cells promote anti-tumor immunity. Nat Commun 2024;15:8696. [Crossref] [PubMed]
25. Xu H, Zheng X, Zhang S, et al. Tumor antigens and immune subtypes guided mRNA vaccine development for kidney renal clear cell carcinoma. Mol Cancer 2021;20:159. [Crossref] [PubMed]
26. Ott E, Bilonda L, Dansette D, et al. The density of Tbet+ tumor-infiltrating T lymphocytes reflects an effective and druggable preexisting adaptive antitumor immune response in colorectal cancer, irrespective of the microsatellite status. Oncoimmunology 2019;8:e1562834. [Crossref] [PubMed]
27. Huang R, Chen Z, Li W, et al. Immune system-associated genes increase malignant progression and can be used to predict clinical outcome in patients with hepatocellular carcinoma. Int J Oncol 2020;56:1199-211. [Crossref] [PubMed]
28. Yang J, Wu Z, Wu X, et al. Constructing and validating of m6a-related genes prognostic signature for stomach adenocarcinoma and immune infiltration: Potential biomarkers for predicting the overall survival. Front Oncol 2022;12:1050288. [Crossref] [PubMed]
29. Yang K, Liu H, Li JH. A methylation-related lncRNA-based prediction model in lung adenocarcinomas. Clin Respir J 2024;18:e13753. [Crossref] [PubMed]
30. Guo M, Sinha S, Wang SM. Coupled Genome-Wide DNA Methylation and Transcription Analysis Identified Rich Biomarkers and Drug Targets in Triple-Negative Breast Cancer. Cancers (Basel) 2019;11:1724. [Crossref] [PubMed]