Construction of a novel disulfidptosis-associated lncRNAs signature for risk features and immunotherapy in breast cancer

Introduction

According to 2022 data, breast cancer (BRCA) accounts for 2.31 million new cases globally among women, representing a quarter of female cancer cases and 10% of all cancer cases worldwide (1,2). BRCA is not only the most common cancer type among women globally but also a leading cause of cancer-related mortality in women. In China, the incidence of BRCA has been steadily increasing, with approximately 357,000 new cases and 75,000 deaths in 2022, showing a trend toward affecting younger individuals (3). Due to the typically subtle early symptoms of BRCA, patients often miss the optimal treatment window, leading to late-stage metastasis to critical organs such as the liver and lungs (4). Despite modern medical interventions covering surgery, radiotherapy, chemotherapy, targeted therapy, and immunotherapy, challenges such as rapid progression of BRCA, treatment-related complications, drug resistance, and poor prognosis continue to make recovery challenging for patients.

Metabolic reprogramming is a significant hallmark of cancer, often resulting in increased uptake of essential nutrients for biosynthesis and energy metabolism in cancer cells, such as glucose and glutamine, primarily achieved through upregulation of the expression levels of transporters for glucose and amino acid uptake (5-7). Consequently, when glucose or amino acid sources are limited, certain cancer cells undergo cell death while normal cells can survive. This nutritional dependence offers potential metabolic vulnerabilities for targeted cancer therapy. Under glucose deprivation, insufficient nicotinamide adenine dinucleotide phosphate (NADPH) supply leads to the abnormal intracellular accumulation of disulfides, such as cysteine, inducing cell death. This type of cell death, characterized by disulfide stress, is termed “disulfidptosis” (8-10). According to Gan’s research team, elevated SLC7A11 expression encourages metabolic vulnerability leading to disulfidptosis in cancer cells, indicating a potentially effective approach for treating tumors (8,11). However, further exploration of disulfidptosis in BRCA is warranted.

Long non-coding RNAs (lncRNAs), which do not encode proteins, are RNA transcripts longer than 200 nts (12). Great attention has been attached to them due to their powerful biological activities. Compared with traditional genes, action modes of lncRNAs vary and can regulate the expression of target genes by regulating chromosomal reorganization, gene modification, messenger RNA (mRNA) stability, gene transcription, and post-transcriptional regulation (13). They also regulate the formation of various diseases, which have become a hot issue in tumor research. Currently, studies have reported that lncRNAs act a vital role in the biological characteristics of BRCA, such as proliferation, invasion and metastasis, and drug resistance (14,15).

However, research on disulfidptosis-related lncRNA in BRCA is currently limited. Recently, one study focused on the development of a prognostic model based on disulfidptosis-related lncRNAs, which enhances the accuracy of predicting the prognosis of BRCA patients (16); while the shortage of data processing, analysis methods, result validation limits the clinical application potential. Hence, exploring key differential lncRNAs with prognostic functions is crucial for BRCA patients. This study utilized The Cancer Genome Atlas (TCGA) database, performed consensus clustering analysis for identifying differential RNA of long length (DRLs), and constructed a DRLs prognostic model using the least absolute shrinkage and selection operator (LASSO)-Cox method. Additionally, we validated this feature for predicting the prognosis of BRCA patients and analyzed differences in the survival outcomes of BRCA patients, tumor immune microenvironment, and responses to immunotherapy. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2024-2377/rc).

Methods

Data acquisition and preprocessing

RNA-seq data, clinical information, and tumor mutational burden (TMB) data for BRCA patients were obtained from TCGA database (https://portal.gdc.cancer.gov/). This included 1,118 tumor samples and 113 normal samples. Thirty-six differential relevance genes (DRGs) from previous studies were included (ABI2, ACTB, ACTN4, AJAP1, CAPZB, CD2AP, DSTN, FLNA, FLNB, GYS1, INF2, IQGAP1, LRPPRC, MYH3, MYH10, MYH9, MYL6, NADPH, NCKAP1, NDUFA11, NDUFS1, NUBPL, OXSM, PDLIM1, RPN1, SLC3A2, SLC7A11, TLN1D, TLN1, TM9SF2, WASF2, BRK1, RAC, CYFIP1, TRIP6, PSMD2) (17-19). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

lncRNA and gene expression analysis

Pearson correlation analysis (R>0.30, P<0.001) was used to screen 1,430 DRGs-lncRNAs. Differential analysis was conducted using the “limma” package [log fold change (FC) =1, false discovery rate (FDR) =0.05], resulting in 277 significantly differentially expressed DRGs-lncRNAs in BRCA. Utilizing the “ggplot2” package, we visualized the differential expression of DRGs, lncRNAs, and genes.

Construction and validation of prognostic model

Data from clinical records and lncRNA expression were merged, after which the createDataPartition function was employed to randomly allocate the combined dataset into training and testing sets. Single-factor Cox regression analysis was performed on each variable in the training set to select lncRNAs associated with BRCA prognosis. LASSO regression and cross-validation (cv. glmnet) were applied to select the optimal regularization parameter λ and perform multivariable Cox regression analysis to optimize the model. The “glmnet” package was used to calculate the risk score (RS) for BRCA patients based on the model, and patients were stratified into high-risk and low-risk groups according to the median RS. The formula for the RS was:

Riskscore=∑iln(expi×βi)[1]

where expi represents the expression level of each lncRNA, and βi represents the respective coefficient within the feature. The survival rates of two groups were compared by generating Kaplan-Meier survival curves using the “survival” package. The “ggrisk” package was used to generate a risk plot.

Clinical feature and prognostic ability assessment

Single-factor and multi-factor Cox regression analyses were performed to evaluate the independence of the RS and other clinical features. Time-dependent receiver operating characteristic (tROC) curves were developed using the “timeROC” R package to assess the accuracy of the model in predicting patient survival at different time points. A stratified survival analysis of clinical and pathological features was conducted.

Development of forest plot and principal component analysis (PCA)

Through the utilization of the “rms” and “regplot” packages in R, forest plots were generated to anticipate the 1-, 3-, and 5-year overall survival (OS) of patients considering various clinical features such as age, sex, and stage. Calibration curves were plotted to evaluate the predictive accuracy of the forest plots. PCA analysis was used to explore the major sources of variation in the data and assess differentiation between different risk groups.

Functional enrichment analysis

Gene Ontology (GO) enrichment analyses were performed for lncRNAs associated with high-/low-RS. Gene set enrichment analysis (GSEA) was used to analyze biological processes (BPs) and metabolic pathways in high-RS groups.

Consensus clustering analysis

Employing the “ConsensusClusterPlus” package, clustering analysis was performed on the samples using the K-means algorithm with a maximum cluster number set to 9 (maxK =9). The optimal number of clusters, in this case 2, was determined based on the consistency matrix of clustering results. To reveal the intrinsic structure and clustering characteristics of the data, PCA and t-distributed stochastic neighbor embedding (tSNE) techniques were utilized for visualization purposes.

Tumor immune microenvironment analysis

Utilizing the “estimate” package, scores for the tumor microenvironment were computed for each sample alongside the use of the cellular composition of tumor samples by estimating relative subsets of cells (CIBERSORT) algorithm to determine the proportion of immune infiltrating cells in each BRCA sample. Following this, differential analyses of tumor microenvironments were conducted to assess disparities in immune cells, immune-related functions, and immune checkpoint-related gene expressions between different risk groups and subtypes. Further analysis included the evaluation of the linear relationship between immune cells and BRCA patient RS, employing Spearman rank correlation analysis. Visualization of the correlation between various immune cell types and RS was achieved using bubble plots, providing correlation coefficients (R values) and P values. Additionally, the correlation between the 8 DRG-lncRNAs used in constructing the prognostic model and immune cells was also examined.

TMB assessment

Mutation data for BRCA patients were downloaded from the TCGA database, and waterfall plots were created to display the mutation status of each gene in the samples. Assessment of TMB for high-/low-risk groups involved comparing mutation frequencies and genes between different risk groups. Samples were stratified into different subgroups based on the median TMB score, and the impact of different tumor mutation loads and risk level combinations on patient survival probability was analyzed. Kaplan-Meier survival analysis was conducted to evaluate the correlation between TMB and BRCA patient OS.

Statistical analysis

All data analyses were performed using the R software (version 4.3.3). The Chi-squared test was used to assess the statistical significance of differences in survival curves. Pearson correlation analysis was used to determine gene expression correlations. Univariate, LASSO, and multivariable Cox regression analyses were conducted to establish predictive features. Spearman rank correlation analysis was used to evaluate the correlation between immune cells and RS. A P value of less than 0.05 was considered statistically significant, and all tests were two-tailed (*, P<0.05; **, P<0.01; ***, P<0.001).

Results

Detection of DRGs-lncRNAs and differentially expressed gene (DEGs)

We obtained transcriptome data and clinical information for 1,118 BRCA samples and 113 adjacent normal samples from the TCGA database. Initially, 16,261 lncRNAs were extracted. Through Pearson correlation analysis (R>0.30, P<0.001), 1,430 DRGs-lncRNAs were identified. A Sankey diagram was used to show the co-expression network of 32 DRGs and lncRNAs (Figure 1A). By comparing gene expression levels between normal and tumor tissues and conducting differential analysis using the “limma” package (logFC =1, FDR =0.05), 277 significantly differentially expressed DRGs-lncRNAs in BRCA were identified. A heatmap was used to visualize the expression levels of the top 50 DRGs-lncRNAs in BRCA and normal samples (Figure 1B). Additionally, a volcano plot was employed to mark the top 10 significantly differentially expressed lncRNAs (AUXG01000058.1, MAGI2-AS3, AC107959.1, ADAMTS9-AS2, AC009806.1, MIR99AHG, CARMN, CADM3-AS1, MIR100HG, GAS1RR) (Figure 1C).

Figure 1 Differential analysis of DRGs-lncRNAs in normal and BRCA samples. (A) Sankey diagram illustrating the co-expression of 32 DRGs and lncRNAs. (B) Heatmap displaying the differential expression of DRGs-lncRNAs in BRCA and normal samples. (C) Volcano plot showcasing the top 10 differentially expressed DRGs-lncRNAs. BRCA, breast cancer; DRGs, differential relevance genes; lncRNAs, long non-coding RNAs.

Construction and validation of prognostic model

A dataset combining clinical data with lncRNA expression data obtained from TCGA was created. Using the createDataPartition function, this dataset was randomly divided into training (n=547) and testing (n=546) sets. Single-factor Cox regression analysis of variables in the training set led to the identification of 15 lncRNAs associated with BRCA prognosis (Figure 2A). Identified as high risk were two lncRNAs (AC120498.10 and AC121247.1) with a hazard ratio (HR) >1, while 13 lncRNAs with HR <1 were categorized as low risk. LASSO regression was used on the significant variables and the optimal regularization parameter (λ) was chosen through cross-validation (cv. glmnet) (Figure 2B,2C). Subsequently, a multivariable Cox regression analysis optimized the model, resulting in eight lncRNAs (AC121247.1, AC120498.10, AL358472.3, AC090181.2, AP005131.3, AC004816.2, AL451123.1, and AL137847.1) being identified as independent prognostic biomarkers (Figure 2D). In addition, we conducted an in-depth analysis of these eight DRGs-lncRNAs and explored their association with 33 disulfide death genes (Figure 2E). An RS was calculated using the “glmnet” package. The formula for the model was: BRCA patient’s RS = −0.510780742030587 * AL358472.3 expression + 0.417112425039459 * AC120498.10 expression − 1.58840837521972 * AP005131.3 expression − 1.048029024742 * AC090181.2 expression + 1.92488431351569 * AC121247.1 expression − 1.60643907303238 * AC004816.2 expression − 3.95217556313913 * AL451123.1 expression − 4.80212579449422 * AL137847.1 expression. Patients were categorized into high- and low-risk groups based on the median RS. The model’s predictive ability for BRCA patient OS was assessed using training, testing, and overall cohorts. Subsequently, we plotted risk heat maps (Figure S1A-S1C), survival status maps (Figure S1D-S1F), and risk curves (Figure S1G-S1I) for all patients. As the risk increased, the number of patients who died gradually increased. Kaplan-Meier analysis indicated that patients in the high-risk group exhibited poorer OS in all three cohorts (Figure S1J-S1L). These results collectively suggest that the eight DRGs-lncRNAs used to construct the model are accurate predictors of BRCA patient prognosis.

Figure 2 Risk model based on disulfidptosis-related lncRNAs for BRCA patients. (A) Results of single-factor Cox regression analysis demonstrating significant association of selected lncRNAs with clinical prognosis. (B) LASSO regression used to construct model based on optimal parameter (lambda). (C) LASSO regression coefficient curve. (D) Histogram illustrating feature coefficients of the prognostic model. (E) Bubble plot displaying correlation analysis of model lncRNAs with DRGs. BRCA, breast cancer; CI, confidence interval; DRGs, differential relevance genes; FDR, false discovery rate; LASSO, least absolute shrinkage and selection operator; lncRNAs, long non-coding RNAs.

Clinical feature and prognostic ability assessment

The forest plot resulting from single-factor and multi-factor Cox regression analyses indicated that the BRCA patient’s RS and age were independent prognostic factors (P<0.001) (Figure 3A,3B). ROC curve analysis showed AUCs of 0.726, 0.705, and 0.693 for 1-, 3-, and 5-year RS, demonstrating the model’s high accuracy in predicting patient survival at different time points (Figure 3C). ROC curves were also jointly plotted with other clinical features to compare the ability of the eight DRGs-lncRNAs to assess BRCA patient survival. The results revealed an AUC of 0.726 for the RS, indicating that the RS outperformed other clinical features besides age (Figure 3D). To comprehensively evaluate the clinical effectiveness of the prognostic model, Kaplan-Meier survival analysis of clinical features for high- and low-risk groups was performed. Even within suBRCAategories defined by age and tumor stage, the low-risk group exhibited better OS than the high-risk group (Figure 3E-3H).

Figure 3 Evaluation of clinical features and prognostic ability. (A) Single-factor Cox regression analysis of risk score with clinical features. (B) Multi-factor Cox regression analysis of risk score with clinical features. (C) 1-, 3-, and 5-year AUCs based on OS risk score. (D) ROC curves for risk score and other clinical features based on OS. (E,F) Kaplan-Meier curves for differences in OS stratified by gender between high-risk and low-risk groups. (G,H) Kaplan-Meier curves for differences in OS stratified by stage between high-risk and low-risk groups. AUC, area under the curve; CI, confidence interval; OS, overall survival; ROC, receiver operating characteristic; TNM, tumor-node-metastasis.

Construction of forest plot and PCA

A forest plot was constructed by adding up scores for each clinical feature (age, sex, stage) to predict patient survival (Figure S2A). Calibration curves were plotted to evaluate the forest plot’s predictive capacity, showcasing its accuracy in predicting patient survival at 1, 3, and 5 years (Figure S2B). PCA analysis was used to investigate the overall distribution of BRCA patients, revealing the most distinct separation between high- and low-risk groups, indicating that the DRGs-lncRNAs used in the model were more effective in identifying low- and high-risk BRCA patients (Figure S2C-S2F).

Functional enrichment analysis of DRG-lncRNAs

To study the distribution patterns of gene sets in BPs, molecular functions (MFs), or cellular components (CCs), GO enrichment analyses of lncRNAs based on high-/low-RS were conducted. Endopeptidase inhibitor activity, peptidase inhibitor activity, and endopeptidase regulator activity exhibited significant enrichment according to the GO analysis. Additionally, carboxylic acid binding, organic acid binding, and peptidase regulator activity showed notable enrichment as well (Figure 4A,4B). Subsequent GSEA analysis explored BPs and metabolic pathways in high-RS groups, revealing that processes such as cardiac muscle contraction, metabolism of xenobiotics by cytochrome p45, oxidative phosphorylation, and steroid hormone biosynthesis were more prevalent in the high-risk subgroup (Figure 4C).

Figure 4 Pathway enrichment analysis of DRGs-lncRNAs in BRCA patients. (A) Bar graph demonstrating top GO signaling pathways involved in BP, MF, and CC biological processes. (B) Bubble plot showcasing top GO signaling pathways involved in BP, MF, and CC biological processes. (C) GSEA pathways significantly enriched in the high-risk subgroup. BP, biological process; BRCA, breast cancer; CC, cellular component; DRGs, differential relevance genes; GO, Gene Ontology; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; lncRNAs, long non-coding RNAs; MF, molecular function; snRNP, small nuclear ribonucleoprotein particles.

Subtyping based on model lncRNA expression levels

Using the “ConsensusClusterPlus” package’s consensus clustering method, BRCA samples were categorized into different subtypes based on the expression levels of DRG-lncRNAs used to construct the prognostic model. Clustering stability was best at k =2, with the highest intra-group correlation and lowest inter-group correlation. Therefore, all BRCA samples were divided into two groups (Figure S3A). Cumulative distribution function (CDF) curve analysis and CDF delta area curve analysis were employed to assess the clustering results (Figure S3B,S3C). A Sankey diagram revealed the correspondence between subtypes and different risk groups of BRCA patients (Figure S3D). PCA analysis and tSNE subtyping techniques demonstrated that the subtypes effectively separated the samples into two distinct parts, confirming the effectiveness of the subtyping (Figure S3E,S3F).

Analysis of tumor immune microenvironment

Differential analysis of the tumor microenvironment was performed for different risk groups and subtypes. Tumor microenvironment (TME) scoring results indicated statistically significant differences in stromal cell scores and immune scores between the high- and low-risk groups (Figure 5A). Significant differences in stromal cell scores, immune cell scores, and composite scores were also observed between C1 and C2 subtypes (Figure 5B). Furthermore, differential expression analysis of immune cells, immune-related functions, and immune checkpoint-related genes was conducted between different risk groups and subtypes (Figure 5C-5H).

Figure 5 Association of tumor immune microenvironment patterns in different risk groups and subtyping groups. (A,B) Differential evaluation of tumor microenvironment scores. (C,D) Differential analysis of immune cells. (E,F) Differential analysis of immune-related functions. (G,H) Differential analysis of immune checkpoint-related genes. *, vs. control group (low-risk or C1 group) P<0.05; **, vs. control group (low-risk or C1 group) P<0.01; ***, vs. control group (low-risk or C1 group) P<0.001. aDCs, activated dendritic cells; APC, antigen-presenting cell; CCR, chemokine receptor; DCs, dendritic cells; HLA, human leukocyte antigen; iDCs, immature dendritic-like cells; IFN, interferon; MHC, major histocompatibility complex; NK, natural killer; pDCs, plasmacytoid dendritic cells; Tfh, follicular helper T cells; TIL, tumor infiltrating lymphocytes; TME, tumor microenvironment; Treg, regulatory T cells.

Correlation analysis of immune cells with RS

To investigate the correlation between immune cells and BRCA patient RS, bubble plots were employed to demonstrate the performance of different tools (Figure S4A). An immune cell heatmap based on subtyping showed downregulation of T cell CD4, T cell CD8, neutrophil, macrophage, and myeloid dendritic cells in C2 (Figure S4B). B cells (R=0.18, P<0.001), neutrophils (R=0.14, P<0.001), and uncharacterized cells (R=0.13, P<0.001) showed a positive correlation with RS, while macrophage M1 (R=0.16, P<0.001), cancer-associated fibroblasts (R=0.14, P<0.001), and natural killer (NK) cells (R=0.14, P<0.001) exhibited a negative correlation with RS (Figure S4C-S4H).

Correlation of prognostic model genes with immune cells

Through the correlation lollipop chart, we could intuitively identify the immune cell types significantly associated with the eight DRG-lncRNAs (Figure S5A-S5H). For example, we found that naïve B cells, and activation of mast cells, CD8⁺ T cells, plasma cells, NK cells, monocytes, CD4⁺ T cells, macrophages M1 and macrophages M0, were significantly associated with lncRNA AC004816.2. In particular, naïve B cells showed the strongest correlation with AC004816.2. The final correlation analysis diagram combines the correlation analysis results of eight DRG-lncRNAs (Figure S5I). This provides important clues for us to understand the role of lncRNA in immune regulation further.

Assessment of TMB

Mutation data for BRCA patients were obtained from the TCGA dataset, and waterfall plots were used to display the mutation status of each gene in the samples. The overall tumor mutation rate in all BRCA samples was 87.5%. Missense mutations and nonsense mutations were more common (Figure 6A,6B). The tumor mutation burden for high- and low-risk groups was evaluated to understand the changes in tumor mutation status between different risk categories. The mutation frequency was higher in the high-risk group (60.85%) than in the low-risk group (56.17%) (Figure 6C,6D). A comparison of the top 15 mutated genes revealed TTN (20% vs. 14%), TP53 (36% vs. 27%), and MUC16 (9% vs. 12%) as the most common mutated genes. TMB analysis showed a significant difference in TMB status between the two groups (P<0.001) (Figure 6E). BRCA samples were stratified into different subgroups based on the median TMB score. The survival probability differed significantly among the four combinations of different tumor mutation burdens and risk levels (P<0.001) (Figure 6F). Kaplan-Meier survival analysis indicated a significant increase in OS for BRCA patients in the high TMB group compared to the low TMB group (P=0.02) (Figure 6G).

Figure 6 Mutation patterns in different risk populations. (A) Mutation information of all genes in all samples. (B) Somatic mutation spectrum of all BRCA patients. (C,D) Waterfall plots of somatic mutations between different risk score populations. (E) Violin plot of differences in TMB between high-risk and low-risk individuals. (F) Kaplan-Meier curve for OS based on TMB+ risk. (G) Kaplan-Meier curve for OS among subgroups of H-TMB and L-TMB patients. BRCA, breast cancer; H, high; L, low; OS, overall survival; TMB, tumor mutational burden; SNV, single nucleotide variation.

Discussion

Disulfidptosis is a novel form of cell death mediated by SLC7A11-induced intracellular over-accumulation of cysteine, resulting in disulfide stress (20,21). This mode of cell death is independent of existing programmed cell death types such as apoptosis, ferroptosis, necroptosis, and cuproptosis (22,23). Its characteristic features include the overexpression of SLC7A11 in cells under conditions of glucose deprivation, inducing the excessive accumulation of disulfides in cells, leading to the formation of excessive disulfide bonds between actin cytoskeletal proteins, causing actin filament contraction and detachment from the cell membrane, disrupting the cell cytoskeleton structure, ultimately leading to cell death (24). Evading cell death is considered a hallmark of cancer. The importance of other forms of cell death in BRCA has been demonstrated, but the role of disulfidptosis in BRCA remains unclear.

We investigated the predictive significance of disulfidptosis concerning the prognosis, survival duration, and immunotherapeutic response in BRCA patients as part of our study. The results of this study provide a foundation for precision treatment of BRCA.

In this study, we first performed lncRNA-related expression analysis of 36 disulfidptosis-related genes, further identifying 277 differentially expressed disulfidptosis-related lncRNAs between BRCA and normal tissues. Using LASSO regression and cross-validation, we developed a model comprising eight DRLs (AC121247.1, AC120498.10, AL358472.3, AC090181.2, AP005131.3, AC004816.2, AL451123.1, and AL137847.1) to predict the prognosis of BRCA patients. Using TCGA databases, we evaluated the model’s prognostic value through survival curves, ROC curves, survival status plots, and heatmaps. Previously reported for diagnosing and prognosing BRCA, AL358472.3 has been identified as an immune-related biomarker. Additionally, it has been recognized as a potential therapeutic target for tumor immunotherapy. Dai et al. suggested that AP005131.3 could serve as an oncogene in BRCA, and inhibiting AP005131.3 might be a treatment strategy for BRCA (25). Jiang et al. proposed that AC004816.2 could predict the prognosis and immune microenvironment of BRCA, indicating that PIGR may be a risk factor for BRCA (26). AL137847.1 has also been used as a biomarker for diagnosing BRCA (27). The area under the curve (AUC) was used to assess the predictive ability of the RS for patient prognosis. A larger AUC area indicates that the model has good classification capability. For 1, 3, and 5 years, the model exhibited AUC values of 0.693, 0.705, and 0.726, surpassing the majority of predictive models. A distinct ferroptosis model, consisting of 16 genes, demonstrated AUC values of 0.756, 0.752, and 0.723 for the corresponding time frames. In contrast, another model yielded AUC values of 0.685, 0.678, and 0.678 for 1, 3, and 5 years, respectively (28,29). There are too many model genes with high AUC value, which is not conducive to clinical application. In conclusion, our model could be considered an appropriate prognostic signal, and its mechanism of action in BRCA warrants further exploration and validation.

Following this, we carried out a clustering analysis on a cohort of BRCA patients. Consensus clustering, a widely used unsupervised method for cancer subtype classification research, was employed. This method enabled the categorization of samples into different subtypes based on varied omics datasets, facilitating the identification of new disease subtypes or the comparison of existing ones. Within the scope of this study, we identified two distinct DRG clusters. Furthermore, our analysis revealed that Cluster 1 exhibited a closer resemblance to the high-risk group, while Cluster 2 corresponded with the low-risk group. As immunotherapy holds a pivotal role in treating BRCA patients, we conducted an immune analysis of the two patient groups to identify immune factors influencing the prognosis of the high-risk and low-risk groups. Through single-sample gene set enrichment analysis (ssGSEA) analysis, we observed extensive enrichment and infiltration of immunocytes in Cluster 1, potentially indicating an immune response inhibition of tumor cells. In contrast, the lack of immune cells and immune suppression in Cluster 2 could be associated with poor patient prognosis. PCA analysis demonstrated a considerable difference between Cluster 1 and Cluster 2. The high-risk group exhibited generally higher expression levels of immune checkpoint genes. In previous studies, immune checkpoints could inhibit the effective recognition of tumor cells by immune cells and suppress immune response and immunotherapy (30,31). This difference may be one of the reasons for the differential survival rates between the high-risk and low-risk groups.

BRCA is a highly heterogeneous disease, and its TME plays a crucial role in tumor progression, metastasis, and response to therapy, which is a complex ecosystem comprising various cellular, including cancer-associated fibroblasts, tumor-associated macrophages, immune cells, and endothelial cells, and non-CCs, for instance, extracellular matrix, cytokines, and growth factors (32). In this study, the TMB study showed a total mutation rate of 60.85% in the high-risk group and 56.17% in the low-risk group. The mutation rate of TP53 in patients of the high-risk group was 36%, whereas the mutation rate in the low-risk group was only 27%. Previous researches have reported that TP53 gene mutations are associated with poor treatment outcomes and prognosis in BRCA, confirming that patients with a high mutation rate of TP53 in the high-risk group have a poor prognosis (33,34). Personalized medicine approaches, including genomic profiling and liquid biopsies, are increasingly being used to tailor treatment strategies based on individual patient characteristics. Future directions include a deeper understanding of the TME, the development of new therapeutic targets and combination strategies, the discovery of robust biomarkers to predict treatment response, and the conduct of well-designed clinical trials to evaluate the efficacy and safety of emerging therapies.

Our study showed certain advantages. The model focuses on disulfidptosis, a relatively new and emerging area of research in cancer cell death mechanisms. By linking disulfidptosis-related lncRNAs to BRCA prognosis, it explores a previously under-investigated field, potentially uncovering new biological insights and therapeutic targets. Moreover, the model demonstrates a strong ability to predict patient survival outcomes, with AUC values of 0.726, 0.705, and 0.693 for 1-, 3-, and 5-year survival, respectively. These values indicate a relatively high accuracy in prognosis compared to other clinical features. In addition, the study employs a robust methodology, including co-expression network analysis, LASSO regression, and multivariate Cox regression, to identify and validate the lncRNAs. The use of both training and testing datasets ensures that the model’s predictive ability is not overfitted to a specific cohort.

Moreover, there were several limitations in this study. Firstly, the bioinformatics analysis was performed on the public databases, which remain the lack of in vitro validation of disulfidptosis-related genes in BRCA prognosis. Then, an independent cohort should be carried out to further confirm this prognostic prediction model. Therefore, further experimental verification of the expression levels of prognostic model genes is warranted.

Conclusions

The study systematically examined the involvement of disulfidptosis-related genes in BRCA prognosis, tumor microenvironment correlation, and clinical attributes, culminating in the development of a robust prognostic prediction model. This model effectively anticipated survival outcomes, tumor immune microenvironment, and responses to immunotherapy in BRCA. The study contributes to personalized and precise treatment for BRCA.

Acknowledgments

We are very grateful for the data from databases such as TCGA, and GEO. Thanks to reviewers and editors for their sincere comments.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2024-2377/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2024-2377/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-2024-2377/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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

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. Arnold M, Morgan E, Rumgay H, et al. Current and future burden of breast cancer: Global statistics for 2020 and 2040. Breast 2022;66:15-23. [Crossref] [PubMed]
2. Siegel RL, Miller KD, Wagle NS, et al. Cancer statistics, 2023. CA Cancer J Clin 2023;73:17-48. [Crossref] [PubMed]
3. Han B, Zheng R, Zeng H, et al. Cancer incidence and mortality in China, 2022. J Natl Cancer Cent 2024;4:47-53. [Crossref] [PubMed]
4. Zhi S, Chen C, Huang H, et al. Hypoxia-inducible factor in breast cancer: role and target for breast cancer treatment. Front Immunol 2024;15:1370800. [Crossref] [PubMed]
5. Parida PK, Marquez-Palencia M, Nair V, et al. Metabolic diversity within breast cancer brain-tropic cells determines metastatic fitness. Cell Metab 2022;34:90-105.e7. [Crossref] [PubMed]
6. Pei S, Zhang P, Yang L, et al. Exploring the role of sphingolipid-related genes in clinical outcomes of breast cancer. Front Immunol 2023;14:1116839. [Crossref] [PubMed]
7. Li W, Jaga S, Shah S, et al. Cancer's Metabolic Hijack: The Under-Recognized Hematologic Emergency Related to the Warburg Effect. Cureus 2025;17:e77869. [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. Zheng P, Zhou C, Ding Y, et al. Disulfidptosis: a new target for metabolic cancer therapy. J Exp Clin Cancer Res 2023;42:103. [Crossref] [PubMed]
10. 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]
11. Koppula P, Zhang Y, Zhuang L, et al. Amino acid transporter SLC7A11/xCT at the crossroads of regulating redox homeostasis and nutrient dependency of cancer. Cancer Commun (Lond) 2018;38:12. [Crossref] [PubMed]
12. Nojima T, Proudfoot NJ. Mechanisms of lncRNA biogenesis as revealed by nascent transcriptomics. Nat Rev Mol Cell Biol 2022;23:389-406. [Crossref] [PubMed]
13. Wei L, Sun J, Zhang N, et al. Noncoding RNAs in gastric cancer: implications for drug resistance. Mol Cancer 2020;19:62. [Crossref] [PubMed]
14. Huarte M. The emerging role of lncRNAs in cancer. Nat Med 2015;21:1253-61. [Crossref] [PubMed]
15. Huo X, Han S, Wu G, et al. Dysregulated long noncoding RNAs (lncRNAs) in hepatocellular carcinoma: implications for tumorigenesis, disease progression, and liver cancer stem cells. Mol Cancer 2017;16:165. [Crossref] [PubMed]
16. Zheng Y, Lin Y, Zhang Y, et al. Determining new disulfidptosis-associated lncRNA signatures pertinent to breast cancer prognosis and immunological microenvironment. Transl Cancer Res 2024;13:5815-29. [Crossref] [PubMed]
17. Pan Y, Jin X, Xu H, et al. Developing a prognostic model using machine learning for disulfidptosis related lncRNA in lung adenocarcinoma. Sci Rep 2024;14:13113. [Crossref] [PubMed]
18. Wei J, Wang M, Wu Y. A disulfidptosis-related lncRNAs cluster to forecast the prognosis and immune landscapes of ovarian cancer. Front Genet 2024;15:1397011. [Crossref] [PubMed]
19. Yang X, Cai Z, Wang C, et al. Integrated multiomic analysis reveals disulfidptosis subtypes in glioblastoma: implications for immunotherapy, targeted therapy, and chemotherapy. Front Immunol 2024;15:1362543. [Crossref] [PubMed]
20. Koppula P, Zhuang L, Gan B. Cystine transporter SLC7A11/xCT in cancer: ferroptosis, nutrient dependency, and cancer therapy. Protein Cell 2021;12:599-620. [Crossref] [PubMed]
21. Yan Y, Teng H, Hang Q, et al. SLC7A11 expression level dictates differential responses to oxidative stress in cancer cells. Nat Commun 2023;14:3673. [Crossref] [PubMed]
22. Peng F, Liao M, Qin R, et al. Regulated cell death (RCD) in cancer: key pathways and targeted therapies. Signal Transduct Target Ther 2022;7:286. [Crossref] [PubMed]
23. Zhang P, Pei S, Liu J, et al. Cuproptosis-related lncRNA signatures: Predicting prognosis and evaluating the tumor immune microenvironment in lung adenocarcinoma. Front Oncol 2022;12:1088931. [Crossref] [PubMed]
24. 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]
25. Dai YW, Wen ZK, Wu ZX, et al. Amino Acid Metabolism-Related lncRNA Signature Predicts the Prognosis of Breast Cancer. Front Genet 2022;13:880387. [Crossref] [PubMed]
26. Jiang T, Wang Y, Chen X, et al. Neutrophil extracellular traps (NETs)-related lncRNAs signature for predicting prognosis and the immune microenvironment in breast cancer. Front Cell Dev Biol 2023;11:1117637. [Crossref] [PubMed]
27. Han X, Chen Y, Xie J, et al. Characteristics of m(6)A-related LncRNAs in breast cancer as prognostic biomarkers and immunotherapy. J Cancer 2023;14:2919-30. [Crossref] [PubMed]
28. Guo LW, Lyu ZY, Meng QC, et al. Construction and Validation of a Lung Cancer Risk Prediction Model for Non-Smokers in China. Front Oncol 2021;11:766939. [Crossref] [PubMed]
29. Jiao Y, Liu X, Shi J, et al. Unraveling the interplay of ferroptosis and immune dysregulation in diabetic kidney disease: a comprehensive molecular analysis. Diabetol Metab Syndr 2024;16:86. [Crossref] [PubMed]
30. Li B, Chan HL, Chen P. Immune Checkpoint Inhibitors: Basics and Challenges. Curr Med Chem 2019;26:3009-25. [Crossref] [PubMed]
31. Topalian SL, Taube JM, Anders RA, et al. Mechanism-driven biomarkers to guide immune checkpoint blockade in cancer therapy. Nat Rev Cancer 2016;16:275-87. [Crossref] [PubMed]
32. Zhang XG, Li WT, Jin X, et al. Comprehensive Analysis Reveals Midnolin as a Potential Prognostic, Therapeutic, and Immunological Cancer Biomarker. Biomedicines 2025;13:276. [Crossref] [PubMed]
33. Mahdavi M, Nassiri M, Kooshyar MM, et al. Hereditary breast cancer; Genetic penetrance and current status with BRCA. J Cell Physiol 2019;234:5741-50. [Crossref] [PubMed]
34. Shiovitz S, Korde LA. Genetics of breast cancer: a topic in evolution. Ann Oncol 2015;26:1291-9. [Crossref] [PubMed]