Development and validation of a disulfidptosis and M2 TAM-related classifier for bladder cancer to explore tumor subtypes, immune landscape and drug treatment

Abstract

Background

Disulfidptosis, as a new mode of programmed cell death, is closely associated with tumorigenesis. Meanwhile, M2 tumor-associated macrophage (TAM) plays an important role in tumor progression. Here, we propose to combine these two perspectives to detect novel disulfidptosis and M2 TAM-related biomarkers in bladder cancer (BCa) to identify various tumor subtypes, construct prognostic features, reveal immune and somatic mutational landscapes, and screen for drugs in BCa.

Methods

We used weighted gene co-expression network analysis (WGCNA) to mine M2 TAM-related genes. Consensus unsupervised clustering was performed to identify potential tumor subtypes. The least absolute shrinkage and selection operator (LASSO) regression and multivariate Cox regression analyses were utilized to build the risk model. We then explored the immune cell, immune function, immune checkpoint expression patterns and somatic mutational landscape in clusters and risk groups. In addition, we performed sensitivity analysis for anti-cancer drugs.

Results

We identified 3057 M2 TAM-related genes and intersected them with disulfidptosis-related genes to obtain 95 disulfidptosis and M2 TAM-related genes (DMRGs). In terms of tumor subtypes, two molecular clusters were identified. Cluster 1 showed stronger immunogenicity and higher tumor mutational burden (TMB). We also predicted 50 drugs with high sensitivity in cluster 1. On the basis of risk grouping, the high-risk group had poor overall survival in the training, test, and validation groups. Ten screened anti-cancer drugs were more sensitive in the high-risk group. A nomogram predicting survival of BCa patients was also established.

Conclusion

By combining two hotspot perspectives, disulfidptosis and M2 TAM, we provide a valuable risk score signature for establishing individualized treatment regimens and drug choices. The risk score may serve as an independent risk factor for BCa patients.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00432-023-05352-3.

Introduction

Bladder cancer (BCa) is one of the ten most common types of cancer worldwide and the second most common urological malignancy, which has a high prevalence rate, with more than 1.6 million people worldwide suffering from this disease. BCa poses a serious risk to human health, with 550,000 new cases and 200,000 deaths annually (Richters et al. 2020). Although recent explorations of biomarkers or subtypes of BCa have proliferated, there is still a lack of accepted molecular clusters and personalized scoring criteria to guide treatment and predict prognosis (Compérat et al. 2022).

In recent years, many emerging forms of programmed cell death have entered the picture, such as ferroptosis, cuproptosis, and pyroptosis (Galluzzi et al. 2018). It is known that cancer progression is significantly associated with a decrease in cell death (Messmer et al. 2019), and disorders of ferroptosis and pyroptosis have also been observed in BCa (Chen et al. 2021; Liu et al. 2022). A recent study by Liu et al. has identified a new mode of programmed cell death associated with disulfide proteins, termed disulfidptosis, which cannot be suppressed by any known inhibitors of cell death (Liu et al. 2023). They found that in the presence of glucose starvation, rapid depletion of intracellular nicotinamide adenine dinucleotide phosphate (NADPH) prevents SLC7A11-mediated cystine reduction from working normally, leading to cystine accumulation. Intracellular disulfide accumulation disrupts the normal binding of disulfide bonds between cytoskeletal proteins in SLC7A11^high cells, thereby inducing cell death. CRISPR screening and functional studies have unearthed a series of disulfidptosis-associated proteins whose role in BCa is unclear.

Tumor-associated macrophages (TAMs) are a primary component of the tumor microenvironment (TME) (Bruni et al. 2020). Macrophages are distinguished into M1 and M2 types based on different functions, surface receptors and secretory characteristics (Wynn et al. 2013). The M1/M2 macrophage pattern is tightly correlated with tumor progression. M1 macrophages are considered anti-tumor, whereas M2 macrophages are commonly regarded as TAMs with tumorigenic effects (Boutilier and Elsawa 2021). TAM has anti-inflammatory properties and immunosuppressive effects. In addition, TAM can promote tumor progression and metastasis by affecting epithelial-mesenchymal transition (EMT), extracellular matrix remodeling and neoangiogenesis (Chen et al. 2018). When tumor metastasis occurs, TAM facilitates the extravasation and growth of tumor cells. Hence, it is significant to explore the role of M2 TAM in the progression of BCa.

In this study, we combined disulfidptosis-related genes and M2 TAM-related genes to obtain 95 intersecting genes. Based on these 95 genes, we classified 289 BCa patients into two clusters. Cluster 1 was an immunologically hot tumor compared to cluster 2. Subsequently, we used the Lasso-Cox model to construct a risk model associated with disulfidptosis and M2 TAM for the first time to provide a foundation for establishing individualized treatment protocols and drug selection.

Materials and methods

Data source

Transcriptomic and clinical data of BCa patients were obtained from The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/) (Tomczak et al. 2015). The data for the validation group were the dataset GSE13507 from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). Referring to the recent study by Junjie Chen et al. 808 genes with the absolute value of NormZ greater than 2.0 were regarded as disulfidptosis-related genes (DRGs) for downstream analysis (Supplementary Data sheet 1) (Liu et al. 2023).

Acquisition of 95 DMRGs

Gene Ontology (GO) analyses were performed using the “clusterProfiler” and “circlize” R packages. Based on the transcriptome matrix of the TCGA cohort, we calculated the relative content of macrophages for each patient on CIBERSORTx (https://cibersortx.stanford.edu/) (Newman et al. 2019).

Weighted Gene Co-expression Network Analysis (WGCNA) was used to identify highly correlated gene modules, summarizing associations between modules and external sample traits. Subsequently, the “WGCNA” package was used to mine the gene modules associated with macrophages (Langfelder and Horvath 2008). Finally, the M2 TAM-related genes (MRGs) were intersected with DRGs to obtain 95 disulfidptosis and M2 TAM-related genes (DMRGs).

Consensus clustering

Relying on 95 DMRGs, we divided 289 BCa patients into two clusters using the “ConsensusClusterPlus” R package (Wilkerson and Hayes 2010). Principal component analysis (PCA) and T-distributed Stochastic Neighbor Embedding (t-SNE) were undertaken using the “Rtsne” R package.

Development and validation of a disulfidptosis and M2 TAM-related risk model for overall survival

To reduce the number of genes included in the risk model, we first performed univariate Cox regression analysis on 95 DMRGs in the training group. Subsequently, least absolute shrinkage and selection operator (LASSO) and multivariate Cox regression analysis were carried out. The following formula was used to calculate the risk scores for the training and test groups to group patients into high- and low-risk groups:

$$\begin{array}{cc}\hfill \text{risk score}& =\text{gene}(A)\text{expression}\ast \text{coef}(A)+\text{gene}(B)\text{expression}\ast \text{coef}(B)\hfill \\ \hfill & +\cdots +\text{gene}(i)\text{expression}\ast \text{coef}(i).\hfill \\ \hfill \end{array}$$

Patients were divided into high- and low-risk groups based on the median risk score of the training group. Kaplan–Meier survival analysis and the log-rank test were performed using the “survminer” R package. Time-dependent receiver operator characteristic (ROC) curves for 1, 3, and 5 year were generated using the timeROC package. The nomogram incorporating age, stage, T stage, Nstage and risk was constructed using the “rms” R package.

Landscape of immune cell infiltration

Referring to the file named “infiltration estimation for tcga” from the TIMER2.0 (http://timer.cistrome.org/) (Li et al. 2020), the “limma,” “scales” and “pheatmap” R packages were used to calculate the infiltration abundance of immune cells in the two risk groups or two clusters. The “GSVA” and “GSEABase” R packages were utilized to perform single-sample gene set enrichment analysis (ssGSEA) and gene set variation analysis (GSVA). “c2.cp.kegg.v7.4.symbols.gmt” was the reference file for GSVA analysis (Hänzelmann et al. 2013; Subramanian et al. 2005). The tumor microenvironment (TME) scores (including stromal score, immune score, and ESTIMATE score) for each patient were calculated by the “estimate” R package.

Analyses of somatic mutation and drug sensitivity

To explore somatic mutations in two risk groups or two clusters, we pooled mutational information of representative genes and calculated the tumor mutational burden (TMB) for each patient using the “maftools” R package. Lastly, we conducted targeted drug sensitivity analysis for patients in risk groups or clusters by the half-maximal inhibitory concentrations (IC50) on Genomics of Drug Sensitivity in Cancer (GDSC) (https://www.cancerrxgene.org/) (Yang et al. 2013).

Results

GO analysis of 808 DRGs and mining for M2 TAM-related genes by WGCNA in BCa

Figure 1 illustrates the main process of this study. The clinical characteristics of the training and testing groups are presented in Table 1. We extracted 808 DRGs and performed GO functional enrichment analysis on them (Fig. 2A). The top 10 of terms which the 808 genes were enriched in are presented in Fig. 2B. The biological processes (BP) included “generation of precursor metabolites and energy” and “mitochondrial respiratory chain complex assembly,” and the cellular components (CC) included “transmembrane transporter complex” and “respiratory chain complex.” The molecular functions (MF) included “electron transfer activity” and “NADH dehydrogenase (ubiquinone) activity.” Subsequently, we performed WGCNA to explore M2 TAM-related module genes in BCa. In this research, the soft-threshold power was calibrated to 6 (Fig. 2C) and 22 modules were screened by WGCNA (Fig. 2D, E). According to Fig. 2E, the grey60 module (cor = 0.5, P < 0.001) and the turquoise module (cor = 0.43, P < 0.001) had strong positive correlations with M2 macrophage subset. Therefore, 3075 genes in the turquoise (Supplementary Data sheet 2) and grey60 (Supplementary Data sheet 3) module were considered as MRGs in this study. Finally, 808 DRGs and 3075 MRGs were combined to obtain 95 DMRGs for downstream analysis (Fig. 2F). Figure 2G shows the co-expression network of the 95 DMRGs in BCa patients.

Fig. 1.

Flowchart of this research. TCGA the Cancer Genome Atlas, GO gene ontology, DRGs disulfidptosis-related genes, TAM tumor-associated macrophage, DMRGs disulfidptosis and M2 TAM-related genes, GSVA gene set variation analysis, LASSO least absolute shrinkage and selection operator, ROC receiver operating characteristics

Table 1.

The clinical characteristics of BCa patients

Characteristics	Training cohort	Testing cohort
Sample	145	144
Age	–	–
<  = 65	62	65
> 65	83	79
Gender	–	–
Male	112	106
Female	33	38
T stage	–	–
T0	1	0
T1	1	2
T2	49	46
T3	62	67
T4	16	19
Unknown	16	10
N stage	–	–
N0	103	84
N1	11	15
N2	17	25
N3	3	3
Unknown	11	17
M stage	–	–
M0	75	79
M1	2	5
Unknown	68	60

Fig. 2.

GO analyses and identification of DMRGs. A, B GO analysis of DRGs. GO gene ontology, BP biological process, CC cellular component, MF molecular function. C 6 was selected as the soft threshold power. D, E 22 modules screened by WGCNA. WGCNA weighted gene co-expression network analysis. F Venn diagram of DRGs and M2-like TAM-related genes. DRGs disulfidptosis-related genes. G Co-expression network of the 95 DMRGs. DMRGs disulfidptosis and M2-like TAM-related genes

Identification of two disulfidptosis and M2 TAM-related molecular clusters

Dependent on these 95 DMRGs, 289 BCa patients were classified into two clusters (Fig. 3A). The t-SNE analysis demonstrated that the two clusters can be clearly distinguished (Fig. 3B). There was no significant difference in overall survival between the two clusters (Fig. 3C). However, cluster 1 had a higher abundance of immune infiltration compared to cluster 2 based on analyses of different platforms (Fig. 3D). Subsequently, we visualized the signaling pathways enriched in clusters by GSVA (Fig. 3E). Signaling pathways, such as regulation of actin cytoskeleton, bladder cancer, pantothenate and CoA biosynthesis, natural killer cell mediated cytotoxicity and B cell receptor signaling pathway, were enriched in cluster 1. SsGSEA showed that cluster 1 had a higher percentage of a series of immune cells, such as CD8 + T cells, T helper cells, NK cells and Tregs than cluster 2 (Fig. 3F). In addition, cluster 1 also appeared to be more active in several immune functions, including APC co-stimulation, inflammation promoting and T-cell co-stimulation (Fig. 3G). A range of immune checkpoints expressed more activity in cluster 1, such as CD274, CD276, CTLA4, LAG3, PDCD1 and TIGIT (Fig. 3H). Consistently, cluster 1 had a higher stromal score, a higher immune score and a higher ESTIMATE score (Fig. 3I).

Fig. 3.

Identification of two molecular clusters based on 95 DMRGs. A Patients divided into two clusters by consensus unsupervised clustering. B T-SNE was undertaken to distinguish the two clusters. T-SNE t-distributed stochastic neighbor embedding. C Survival analysis. D Immune cell infiltration analysis in clusters based on different platforms. E GSVA gene set variation analysis. F, G SsGSEA of immune cells and immune function. SsGSEA single-sample gene set enrichment analysis. H Expression of immune checkpoints. I The comparison of stromal score, immune score and ESTIMATE score in clusters. ESTIMATE, Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data

Furthermore, we explored the somatic mutation landscape and precision treatment in clusters. Figure 4A–D showed the mutation landscape in the cluster 1 and cluster 2, respectively. The top five genes with the highest mutation rates in cluster 1 were TP53 (64%), TTN (48%), KMT2D (34%), EP300 (28%), PIK3CA (28%), whereas in cluster 2 they were TTN (45%), TP53 (43%), MUC16 (29%), KMT2D (27%), ARID1A (26%). Missense mutations were the most common mutation form in clusters. Cluster 1 had higher TMB compared to cluster 2 (P < 0.05) (Fig. 4E). Figure 4F demonstrated that the low mutation group was associated with worse survival (P < 0.01). Finally, we screened 50 targeted agents that had lower IC50 in cluster 1, 12 of which were displayed (Fig. 4G), such as AICAR (Acadesine) and Mitomycin C.

Fig. 4.

Tumor mutations diagram and drug sensitivity prediction in clusters. Somatic mutations in A, B cluster 1 and C, D cluster 2. E Cluster 1 has a higher TMB than cluster 2. TMB tumor mutational burden. F Low mutation group has poor survival. G Immunotherapy prediction showing significant IC50 difference in clusters

Establishment and validation of a disulfidptosis and M2 TAM-related prognostic model

Lasso-Cox regression was utilized to reduce the number of DMRGs and ultimately three genes (including LRRC49, RWDD1 and NPC1) were included in the risk equation (Fig. 5A).

Fig. 5.

Construction of the risk model. A Variable selection by LASSO regression algorithm. LASSO least absolute shrinkage and selection operator. B The PCA and t-SNE of risk groups. PCA principal component analysis. The division of high- and low-risk groups, survival status, survival time, expression levels of three risk genes and ROC curves in the C training and D test groups. ROC receiver operating characteristic. E Survival analysis and ROC curves in the validation group. F Distribution of clinical characteristics in clusters and risk groups. Somatic mutations in the G high- and H low-risk groups. I TMB in risk groups. J Survival analysis based on risk level and TMB

We calculated risk score with the formula:

$$\begin{array}{cc}\hfill \text{risk score}& =\left(-0.921042275988898\right)\ast LRRC49+\left(-0.937784070441489\right)\ast RWDD1\hfill \\ \hfill & +\left(0.552308328841134\right)\ast NPC1.\hfill \\ \hfill \end{array}$$

As illustrated in Fig. 5B, PCA and t-SNE were able to distinguish patients in high- and low-risk groups. Figure 5C, D shows the division of high- and low-risk groups, survival status, survival time, expression levels of three related genes and time-dependent receiver operating characteristic (ROC) curves in the training and test groups, respectively. Results for the training, test and validation (Fig. 5E) groups indicated that the high-risk group had worse overall survival. The 1-, 3-, and 5-year area under the ROC curve (AUC) of the training group were 0.715, 0.662, and 0.736, of the test group were 0.685, 0.637, and 0.639, and of the validation group were 0.713, 0.680, and 0.631 (Fig. 5E), respectively. Moreover, we found that the clinicopathological characteristics of patients were asymmetrically distributed in the high- and low-risk groups, including stage, T stage and N stage (Fig. 5F).

To explore the differences in TMB between high- and low-risk groups, we further analyzed the mutational information of representative genes in BCa patients (Fig. 5G, H). The top five genes with the highest mutation rates in high-risk group were TP53 (51%), TTN (49%), ARID1A (30%), KMT2D (29%), MUC16 (28%). Genes such as TP53 (45%), TTN (42%), KMT2D (29%), MUC16 (28%) and KDM6A (25%) had the top five mutation frequencies in the low-risk group. The high-risk group had significantly higher TMB compared to the low-risk group (P < 0.01) (Fig. 5I). Finally, we combined TMB and risk to group patients and performed survival analysis (Fig. 5J).

Construction of a nomogram predicting the 1-, 3- and 5-year overall survival

We included clinicopathological factors and risk score in univariate (Fig. 6A) and multivariate (Fig. 6B) Cox regression. As illustrated in Fig. 6A, B, risk score was an independent predictor of overall survival in BCa patients (univariate: HR 1.684, P < 0.001; multivariate: HR 1.587, P < 0.001). Additionally, we found that age (univariate: HR 1.044, P 0.001; multivariate: HR 1.038, P 0.006) and N stage (univariate: HR 1.713, P < 0.001; multivariate: HR 1.607, P = 0.043) were also independent prognostic elements (Fig. 6B). Then, according to a series of clinical factors (including age, stage, T stage and N stage) and risk, we constructed a nomogram for predicting the 1-, 3-, and 5-year overall survival of BCa patients (Fig. 6C). The calibration plot demonstrated the outstanding predictive power of the nomogram (Fig. 6D).

Fig. 6.

Construction of the nomogram relying on a range of clinical parameters and risk level. A Univariate and B multivariate Cox regression analysis of risk level and other clinical characteristics in the entire TCGA cohort. Risk scores are independently associated with survival of BCa patients. C The nomogram based on risk level, age, T stage, N stage and stage. D Calibration plot

Immune-related analysis and precision treatment in risk groups

As presented in Fig. 7A, according to different software platforms, the correlation coefficients between risk scores and the abundance of immune cell infiltration were visualized in the bubble plot. In addition, we depicted the landscape of immune cell infiltration in risk groups according to different methods (Fig. 7B). SsGSEA showed that high-risk group had a higher percentage of macrophages, Th1 cells and Tregs (Fig. 7C). Besides, high-risk group also expressed more activity in some immune functions, such as APC co-inhibition, parainflammation and type I IFN response (Fig. 7D). Immune checkpoints such as CD276, CD47, LAIR1, TIGIT and TNFSF9, appear to be more active in the high-risk group (Fig. 7E). The high-risk group had a higher stromal score and a higher ESTIMATE score compared to the low-risk group. However, there was no difference in the immune score between the two groups (Fig. 7F). By drug sensitivity analysis, we selected 10 targeted drugs that had significantly lower IC50 in the high-risk group (Fig. 7G).

Fig. 7.

Immune infiltration landscape and immunotherapy prediction in risk groups. A The bubble plot and B heat map of immune cell abundance dependent on different software platforms. SsGSEA of C immune cells and D immune function in risk groups. E Immune checkpoints activation. F Stromal score, immune score, and ESTIMATE score for risk groups. G Ten immunotherapeutic drugs showing lower IC50 in high-risk group

Discussion

While non-muscle-invasive bladder cancer (NMIBC) is usually managed by endoscopic resection and intravesical therapy (Lenis et al. 2020), systemic treatment for muscle-invasive bladder cancer (MIBC) consists mainly of gemcitabine combined with cisplatin chemotherapy (Patel et al. 2020). Over the past few decades, rapid advances in genomics have allowed researchers to better understand the pathogenesis of BCa and to develop immunotherapy and targeted therapeutic regimens by mining predictive biomarkers. On the one hand, cell death disorder is a significant condition for the high proliferative state of tumor cells (Faubert et al. 2020). As a new form of cell death, the role of disulfidptosis in bladder carcinogenesis is still unclear with high potential to become a hot spot for future research. Due to glucose starvation, NADPH deficiency leads to abnormal cystine reduction. Disulfide stress subsequently causes rapid cell death (Liu et al. 2023). It has been revealed that SLC7A11-mediated redox state plays a significant role in tumor progression, multidrug resistance and tumor-related ferroptosis (Liu et al. 2020). On the other hand, several studies have indicated that survival and therapeutic response of cancer patients are closely related to immune cell components (Fridman et al. 2017; Kurebayashi et al. 2018; Thorsson et al. 2018). TAM (M2 macrophage) is a major constituent of the TME and is capable of promoting cancer cell proliferation. It has been shown that bone morphogenetic protein-4 (BMP-4) secreted by BCa cells can trigger M2 polarization and drive BCa progression (Martínez et al. 2017). It has also been reported that accumulation of TAM is associated with lymphangiogenesis and lymph node metastasis of tumor (Chen et al. 2018; Schoppmann et al. 2002). In this paper, for the first time, we combined two perspectives of disulfidptosis and M2 macrophage to identify individualized risk scoring criteria that can help clinical treatment decisions for patients with BCa. We included three genes in our research and provided a simpler prognostic risk formula from these two perspectives.

In this study, we first screened two gene modules significantly associated with TAM using WGCNA and obtained 95 DMRGs by taking intersections with 808 DRGs. We applied consensus unsupervised clustering to these 95 genes and identified two molecular clusters, namely cluster 1 and cluster 2. Although there was no significant difference in overall survival between the two clusters, it surprised us that cluster 1 showed a more brightly colored immune landscape and had a higher TMB compared to cluster 2. GSVA showed that cluster 1 was enriched in tumor-related pathways, such as DNA replication and bladder cancer, while cluster 2 was enriched in metabolism-related pathways. In addition, we screened 50 targeted drugs with lower IC50 in cluster 1. Subsequently, by Lasso-Cox regression we screened three genetic traits which were used to construct a predictive risk model. In the training group, test group and validation group, the high-risk group had poor overall survival compared to the low-risk group. The predictive risk model consists of three genes, LRRC49, RWDD1 and NPC1. The role of leucine rich repeat containing 49 (LRRC49) in cancer has not been studied in depth to date. LRRC49 is located on chromosome 15q23 and shares the same promoter as the THAP domain containing 10 (THAP10) gene. A study found that promoter methylation leading to LRRC49/THAP10 silencing is a common event in breast cancer, while the mechanism remains to be elucidated (Souza et al. 2008). Sulfide-quinone oxidoreductase (SQR) is responsible for the oxidation of sulfide to thiosulfate and the transfer of the generated electrons to ubiquinone. It has been indicated that RWD domain-containing 1 (RWDD1) may act as a transcription factor that binds to the proximal region of the sqr promoter and subsequently affects sulfide metabolism (Li et al. 2018). Also, Vuong et al. observed that melanoma patients with high RWDD1 expression had poor survival rates (Vuong et al. 2014). Niemann-Pick C1 (NPC1) is highly expressed in triple-negative breast cancer (TNBC) and promotes aggressive features of TNBC (O'Neill et al. 2022). According to our study, NPC1 was highly expressed in the high-risk group, but its relationship with survival in BCa was unknown.

We also explored the somatic mutation profile of the entire TCGA cohort. TP53 and TTN were the two genes with the highest mutation frequency. P53 alterations are among the most common genetic alterations in invasive urothelial carcinoma and are associated with tumor progression and chemotherapy sensitivity (Nishiyama et al. 2008). It has been shown that p53 mutations in combination with KDM6A deletion activate cytokine pathways which promote M2 macrophage polarization and cause BCa (Kobatake et al. 2020). He et al. demonstrated defects in p53 and PTEN promote basal/squamous muscle invasive bladder cancer (MIBC) (He et al. 2022). Additionally, it has been suggested by machine learning that spontaneous mutations in Titin (TTN) are associated with late replication time and represent high TMB condition (Oh et al. 2020). It is reported that high somatic TMB is associated with better survival and a high response rate to immune checkpoint inhibitors (ICIs) in BCa (Samstein et al. 2019), which is consistent with our findings.

Moreover, our risk score reveals the immune landscape of risk groups. With regards to immune cells, macrophages, Th1 cells, and Treg cells were more abundant in the high-risk group. In terms of immune function, cytolytic activity, inflammation promoting, MHC class I, T-cell co-inhibition, and type I IFN response were more active in the high-risk group. Th1 cells can produce pro-inflammatory cytokines and thereby assist CD8 + T cells, thus Th1 cells are often associated with a favorable survival for tumors, such as ovarian, renal cell, and hepatocellular cancers (Fridman et al. 2012; Katsuta et al. 2020). In contrast, Treg cells promote tumor growth by secreting cytokines that inhibit the function of anti-cancer immune cells. High abundance of Treg cells is commonly associated with poor outcome for tumors (Nishikawa and Sakaguchi 2014). Despite such notoriety of Treg cells, there still exists a subpopulation of Treg cells, termed non-Tregs, which do not have suppressive activity. As some reports suggest, high levels of Treg cells are correlated with improved prognosis in colon and head and neck cancers (Nishikawa and Sakaguchi 2014).

For the first time, our paper proposed to combine two perspectives of disulfidptosis and M2 TAM to construct a risk model to predict the overall survival for BCa patients. The risk formula includes fewer genes than similar research, making its application simple and easy. However, it is undeniable that there are still some limitations in this study. First, the present study relied on retrospective BCa cohorts, a large prospective cohort was necessary to validate the precision of the prognostic model despite the presence of internal and external validation. Second, more in vivo and in vitro experiments are necessary to validate the role of disulfidptosis and TAM in tumors, which will be the direction of our future work.

Conclusion

In this research, two molecular clusters were identified based on 95 DMRGs. Combining both perspectives of disulfidptosis and M2 TAM, we constructed a risk score signature which could be a promising biomarker for predicting the survival outcome of BCa patients. This risk score reveals the immune landscape of BCa to some degree and provides a basis for personalized treatment regimens and drug choices.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

CZR and XQL designed the study. QHW, ZNX and YP collected the data. CZR, YZL were responsible for bioinformatics analysis. All authors wrote the manuscript. All authors have read and approved the final manuscript.

Funding

None.

Data availability

The datasets analyzed in this study are available in the Cancer Genome Atlas (https://portal.gdc.cancer.gov/), Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) and GDSC (https://www.cancerrxgene.org/).

Declarations

Conflict of interest

The authors declare no competing interests.