Identification of Triple-Negative Breast Cancer Subtypes Based on Ferroptosis- Related Gene Signature for Targeted Treatment Prioritization

doi:10.21203/rs.3.rs-3236815/v1

Download PDF

Research Article

Identification of Triple-Negative Breast Cancer Subtypes Based on Ferroptosis- Related Gene Signature for Targeted Treatment Prioritization

https://doi.org/10.21203/rs.3.rs-3236815/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Ferroptosis, an iron-dependent form of programmed cell death, plays a crucial role in cancer therapies. However, its impact on chemotherapy, immune checkpoint inhibitor (ICI) treatments, and molecular subtypes of triple-negative breast cancer (TNBC) is poorly understood. In this study, we utilized the FUSCC TNBC cohort to classify TNBC patients into distinct subtypes based on the expression of eight ferroptosis-related genes (FRGs). We employed Gene Ontology (GO) and Gene Set Variation Analysis (GSVA) to characterize the immune phenotype and enriched pathways associated with these subtypes. Additionally, we developed the FerrScore model to identify potential compounds and predict the benefits of ICIs in TNBC patients. Our analysis revealed two ferroptosis-related subtypes with contrasting overall survival (OS) outcomes. Cluster 1 had superior OS and exhibited a "hot" tumor phenotype with increased immune cell infiltration and elevated expression of immune checkpoints compared to Cluster 2. We identified Everolimus as the most promising candidate drug for TNBC patients with a high FerrScore, considering CMap score, experimental evidence, and clinical trial status. Moreover, we validated FerrScore as a powerful metric for predicting the benefits of various ICIs. These findings highlight the influence of ferroptosis on the tumor microenvironment, enabling the classification of TNBC patients into subgroups with different OS outcomes. The FerrScore model has potential in screening compounds and predicting the benefits of ICIs in TNBC, offering valuable insights for treatment prioritization in clinical settings.

ferroptosis

triple-negative breast cancer

molecular subtype

immunotherapy

chemotherapy

Triple-negative breast cancer (TNBC), the most aggressive subtype of breast cancer characterized by high intra-tumoral heterogeneity, accounts for approximately 15–20% of all invasive breast tumors [1]. TNBC is characterized by the lack of expression of progesterone receptor (PR), estrogen receptor (ER), and human epidermal growth factor receptor 2 (HER-2) [2]. Unfortunately, patients diagnosed with advanced TNBC often face a grim prognosis, with a median overall survival (OS) of rarely exceeding 12 to 18 months [3]. Despite the implementation of aggressive therapeutic approaches, including surgery, radiotherapy, chemotherapy, and immunotherapy, breast cancer has now surpassed lung cancer as the leading cause of cancer-related death in women globally, as reported in the Global Cancer Statistics 2020 [4]. According to the 2022 version of the Chinese Society of Clinical Oncology (CSCO)-TNBC guidelines, chemotherapy remains the primary treatment for unresectable TNBC [5]. Numerous studies have demonstrated that immunotherapy enhances the sensitivity of tumor cells to chemotherapy [6]. Therefore, there is an urgent need for in-depth research and analysis to guide the clinical management of TNBC, particularly regarding the optimal utilization of chemotherapy and immunotherapy.

Molecular subtype identification serves as the foundation for a comprehensive understanding and precision therapy of TNBC [7]. Initially, Lehmann identified six distinct TNBC subtypes with unique gene expression profiles and ontologies through cluster analysis: basal-like 1 and 2 (BL1 and BL2), mesenchymal (M), mesenchymal stem-like (MSL), luminal androgen receptor (LAR), and immunomodulatory (IM) subtypes, which guided therapeutic decision-making [8]. Subsequently, Lehmann refined these subtypes into four tumor-specific subtypes (LAR, BL1, BL2, and M) [9]. Additionally, Professor Powel stratified TNBC into four subtypes based on RNA and DNA profiling analysis: basal-like immune-activated (BLIA), basal-like immunosuppressed (BLIS), LAR, and MES [10, 11]. However, these findings have yet to translate into revised clinical practice guidelines or improved patient outcomes. In a subsequent study, Professor Zhimin Shao classified TNBC into four subtypes (BLIS, MES, LAR, and IM) based on genomic and transcriptomic profiles and proposed precision therapeutic strategies specific to each FUSCC subtype [12]. More recently, Shao's group identified three distinct metabolomic subgroups, further advancing our understanding of TNBC and enabling precision therapy [13]. Despite the numerous proposed molecular subtyping approaches for guiding precision therapy, limitations and challenges persist due to the high intra-tumoral heterogeneity of TNBC. Thus, the exploration of novel classification systems remains an exciting area of focus for TNBC in the future.

Ferroptosis, an iron-dependent form of regulated cell death (RCD), has been shown to impact the response of tumors to radiotherapy, chemotherapy, and immunotherapy [14]. Previous studies have indicated that breast cancer cells are particularly sensitive to ferroptosis due to CD44-dependent iron endocytosis, which promotes the activity of iron-dependent demethylases and upregulates genes related to the epithelial-mesenchymal transition (EMT) pathway [15]. Notably, the human TNBC cell line MDA-MB-157 exhibits heightened sensitivity to RSL3, a ferroptosis activator, due to its high expression of ACSL4 [16]. Published research has revealed that TNBC is enriched in a ferroptosis-gene signature, and co-inhibition of BET and proteasome induces ferroptosis across major TNBC subtypes (BL1/2 and M/MSL), while targeting BET and CXCR2 induces tumor cell apoptosis specifically in the mesenchymal TNBC subtype [17]. Ferroptotic cancer cell-derived damage-associated molecular patterns (DAMPs) trigger inflammation-related immunosuppression and reshape the immune landscape in the tumor microenvironment (TME), thereby promoting tumor growth [18]. Moreover, recent studies have demonstrated that ferroptotic cancer cells recruit T cells, NK cells, and macrophages, thereby converting "cold" tumors into a "hot" phenotype by activating anti-tumor immune responses [19]. These findings highlight ferroptosis as a promising avenue for precision therapy in TNBC. Ferroptosis-related signatures have been utilized in the molecular subtyping of various cancer types, including pancreatic ductal adenocarcinoma, lung adenocarcinoma, and hepatocellular carcinoma [20–22]. Therefore, further investigations into the role of ferroptosis in molecular subtyping of TNBC and its impact on TME remodeling are of paramount importance.

In this study, our first objective was to identify eight ferroptosis-related genes (FRGs) and investigate their gene expression heterogeneity across different cancer types. We then focused on utilizing the mRNA expression levels of these eight FRGs for molecular subtype identification in TNBC and examined the immune landscape within each subtype. Additionally, we developed a scoring model, referred to as the FerrScore, which was based on the product of mRNA expression and the correlative coefficient of the FRGs. We used the FerrScore to identify six drugs from the Cancer Therapeutics Response Portal (CTRP) dataset and six drugs from the Profiling Relative Inhibition Simultaneously in Mixtures (PRISM) dataset as potential candidates for TNBC patients, based on their FerrScore values. Notably, Everolimus emerged as the most promising candidate drug for TNBC patients with high FerrScore. Furthermore, we evaluated the diagnostic value of FerrScore in predicting the benefits of immune checkpoint inhibitors (ICIs) therapy in immunotherapy cohorts. Our findings demonstrated that FerrScore is a valuable biomarker for predicting the efficacy of anti-PD-1, anti-PD-L1, and anti-PD-1&CTLA-4 therapies. In conclusion, the FRGs signature represents a potent metric for stratifying TNBC patients, while FerrScore serves as a reliable tool for screening candidate compounds and predicting the benefits of ICIs therapy, offering new perspectives for clinical treatment strategies.

Data acquisition and filtration

We retrieved the transcriptomic profiles and clinical data of patients with TNBC from the National Omics Data Encyclopedia (NODE) (https://www.biosino.org/node/) (FUSCC, Project ID: OEP000155) conducted by Zhimin Shao at Fudan University Shanghai Cancer Center. After excluding patients without matched information, a total of 358 tumor samples and 88 normal samples were included for further analysis [12]. For validation purposes, we utilized the GSE76124 and GSE21653 cohorts as independent datasets [10].

Genetic variations and expression profiles of FRGs

We initially identified 125 ferroptosis-related genes (FRGs) from the FerrDb database (http://www.zhounan.org/ferrdb/legacy). Next, we performed Kaplan-Meier survival analysis on the FUSCC dataset, resulting in the identification of 841 survival-related genes. Through the intersection of these gene sets, we selected nine hub genes. Ultimately, eight FRGs were used for unsupervised clustering analysis, with ALOX15B excluded due to its lack of expression difference between normal and tumor samples. To gain insights into the expression profiles of the eight FRGs at the single-cell level, we visualized them using the GSE118389 dataset. Furthermore, we validated these expression patterns in multiple single-cell RNA-seq datasets obtained from the Tumor Immune Single-cell Hub (TISCH) (http://tisch.comp-genomics.org/home/). Additionally, we examined the genetic variations of the eight FRGs in the TCGA TNBC cohort using the "maftools" R package. We also investigated the dysregulation and methylation profiles of the eight FRGs, along with the Pearson correlation between FRG expression and copy number variation (CNV) as well as FRG expression and promoter methylation in pan-cancer.

FRGs-based unsupervised clustering

We employed the "ConsensusClusterPlus" R package to perform unsupervised clustering analysis based on the mRNA expression of the eight FRGs. This analysis enabled the identification of molecular subtypes. To visualize the distribution of FRGs-related subtypes, we utilized principal component analysis (PCA). To assess the prognostic significance of the identified subtypes, we conducted Kaplan-Meier analysis to compare the survival probability between distinct groups. To ensure the reliability and stability of our FRGs-related subtype identification strategy, we employed the Non-negative Matrix Factorization (NMF) algorithm, implemented through the "CancerSubtypes" R package, in the FUSCC TNBC cohort. Additionally, we validated our findings in the GSE76124 and GSE21653 cohorts using the "ConsensusClusterPlus" R package.

Gene Ontology (GO) and Gene Set Variation Analysis (GSVA) between distinct subtypes of TNBC

We utilized the "limma" R package to determine the differentially expressed genes (DEGs) between the various ferroptosis subtypes. The DEGs were selected based on a fold-change threshold of 1.5 and an adjusted p-value of < 0.05. Subsequently, we employed the "ClusterProfiler" R package to conduct Gene Ontology (GO) analysis, aiming to elucidate the biological processes associated with the ferroptosis-related subtypes of TNBC. To gain further insights into the immune features of the distinct ferroptosis subtypes, we conducted pathway analysis using Gene Set Variation Analysis (GSVA) with the Molecular Signatures Database (MSigDB-v5.2) as a reference (https://www.gsea-msigdb.org/gsea/index.jsp).

Characterizing the Tumor Immune Landscape in Distinct Subtypes of Triple-Negative Breast Cancer

We utilized the "ESTIMATE" R package to compare various scores, including the immune score, stromal score, estimate score, tumor purity, TIDE score, exclusion score, and dysfunction score. Additionally, the Tumor Immune Dysfunction and Exclusion (TIDE) tool (http://tide.dfci.harvard.edu/) was employed. We also calculated the tumor inflammation signature (TIS) score based on the mRNA expression of the 18 genes [23]. To estimate immune cell infiltration across different groups, we employed multiple algorithms, including Cibersort, XCELL, ssGSEA, TIMER, and MCP counter.

Construction of the FerrScore scoring model

In order to further investigate the potential diagnostic value of ferroptosis signatures in prioritizing treatment strategies for TNBC, we employed the penalized Cox regression model, utilizing the least absolute shrinkage and selection operator (LASSO) penalties. This approach enabled us to calculate the correlative coefficients of the eight FRGs and define the ferroptosis-gene signature as the FerrScore. The FerrScore formula was established as follows: FerrScore = ∑ (correlative coefficient × gene's expression). Subsequently, TNBC patients were divided into high and low FerrScore groups based on an optimal cut-off value determined using the "Survminer" R package. We then compared the overall survival (OS) between these two groups. To evaluate the predictive accuracy of the FerrScore model, we utilized the "survivalROC" R package to plot the receiver operating characteristic (ROC) curve and estimate its performance.

FerrScore-based drug sensitivity screening

To identify potential therapeutic agents for patients with high FerrScore, we employed the FerrScore construction model in the PRISM (https://www.theprismlab.org/) and CTRP2.0 (https://portals.broadinstitute.org/ctrp/) databases. The expression profiles of cancer cell lines (CCLs) were obtained from the Broad Institute-Cancer Cell Line Encyclopedia (CCLE) portal (https://portals.broadinstitute.org/ccle/). We calculated the area under the dose-response curve (AUC) values to assess drug sensitivity, where lower AUC values indicated higher sensitivity to chemotherapy. By performing Spearman correlation analysis, we investigated the association between FerrScore and AUC values to identify candidate drugs for patients with high FerrScore.

Furthermore, we calculated the Connectivity Map (CMap) score of drugs by submitting the 300 differentially expressed genes (DEGs) between normal and tumor tissues, which comprised the most significantly upregulated and downregulated genes (150 each), to the CMap website (https://clue.io/query). To gather additional information, we obtained data on the clinical trial status, experimental evidence, and mRNA expression of drug targets from the National Center for Biotechnology Information (NCBI) (https://pubmed.ncbi.nlm.nih.gov/) and DrugBank (https://go.drugbank.com/).

FerrScore-based ICIs therapy benefits prediction

To further assess the diagnostic value of FerrScore in predicting the benefits of immune checkpoint inhibitors (ICIs) therapy, we utilized an urothelial cancer cohort with the "IMvigor210CoreBiologies" R package to investigate the correlation between FerrScore and the response to anti-PD-L1 immunotherapy. Additionally, we examined the potential role of FerrScore in anti-PD-1 and anti-PD-1 & CTLA4 ICIs therapy using the GSE78220 and PRJEB23709 cohorts, respectively. Furthermore, we explored the association between FerrScore and the response to adoptive T cell therapy in the GSE100797 cohort.

Histological analysis

Human TNBC samples were collected from the department of Breast Disease at the First Affiliated Hospital of Nanjing Medical University (Nanjing, China) between January 2015 and September 2022. The tissues were sectioned into 5 µm-thick paraffin sections and subjected to hematoxylin and eosin (H&E) staining as well as immunohistochemical (IHC) staining. For IHC staining, the following primary antibodies were used: anti-SOCS1 Rabbit pAb (Cat# GB11372, Servicebio), anti-NOX4 Rabbit pAb (Cat# GB11347, Servicebio), anti-PD-1 Rabbit pAb (Cat# GB11338, Servicebio), and anti-CD3 Rabbit pAb (Cat# GB11014, Servicebio).

Four-color immunofluorescence (IF)

Formalin-fixed paraffin-embedded sections were subjected to immunohistochemical (IHC) staining using primary antibodies including anti-SOCS1 Rabbit pAb (Cat# GB11372, Servicebio), anti-NOX4 Rabbit pAb (Cat# GB11347, Servicebio), anti-PD-1 Rabbit pAb (Cat# GB11338, Servicebio), and anti-CD3 Rabbit pAb (Cat# GB11014, Servicebio). The sections were incubated with the primary antibodies overnight at 4°C. Subsequently, secondary antibodies were applied and incubated for 1 hour at room temperature. The cell nuclei were counterstained with DAPI (Cat# GDP1024, Servicebio).

Statistical Analysis

RNA-seq data analysis and visualization were conducted using R version 4.0.4 software. To compare differences between two groups, Wilcoxon test and Student's t-test were performed, while Kruskal-Wallis test and one-way ANOVA were used for comparisons between three or more groups. Spearman and Pearson correlation analysis were utilized to examine positive or negative relationships. Kaplan-Meier survival analysis was applied to compare survival probabilities among groups. Statistical significance was defined as p < 0.05 for comparisons between groups, indicating significant differences.

Identification, genetic variations and expression profiles of the FRGs

Cytotoxic T cell-driven immunity promotes ferroptosis in cancer cells. For example, the anti-PD-L1 antibody induces lipid peroxidation-dependent ferroptosis in tumor cells (Fig. 1A), making ferroptosis-related anti-cancer therapy a promising target [24]. In this study, we aimed to explore the role of ferroptosis in molecular subtype identification of TNBC and construct a FerrScore model to screen promising drugs and predict response to immunotherapy (Supplementary Fig. 1). Initially, we identified 125 ferroptosis-related genes (FRGs) from the FerrDb, and intersected them with 841 survival-related genes screened by Kaplan-Meier survival analysis (Fig. 1B, Supplementary Fig. 2A-J, Supplementary table 1). In the FUSCC TNBC cohort, EMC2, FTH1, HMOX1, LPCAT3, NOX4, and SOCS1 were up-regulated in tumors, while BAP1 and ISCU were down-regulated (Fig. 1C). Eight hub genes were identified, while ALOX15B was excluded due to no difference between normal and tumor samples. The expression of the eight FRGs distinguished tumors from normal tissues as shown by principal component analysis (PCA) (Fig. 1D). Additionally, the expression of the eight FRGs varied across different grades, T and N stages of TNBC (Supplementary Fig. 3). The coding sequence (CDS) length and chromosomal location of the eight FRGs were obtained from NCBI (Supplementary table 2). In pan-cancer analysis, ISCU was significantly down-regulated in breast invasive carcinoma, and EMC2 showed the highest frequency of copy number variation (CNV) (Supplementary Fig. 4A-B). NOX4 exhibited the highest frequency of methylation, and the expression of SOCS1 positively correlated with promoter methylation across 20 types of cancer in TCGA (Supplementary Fig. 4C-D).

The protein-protein interaction (PPI) network revealed that HMOX1 interacted with FTH1 and NOX4, and the correlation of the eight FRGs was visualized in the network plot (Fig. 1E-G). Somatic mutation analysis of the eight FRGs showed that 6 (4.41%) of 136 samples had genetic variations, and EMC2 had a significantly high frequency (32.37%) of copy number gain in the TCGA cohort (Fig. 1H-J, Supplementary table 2). In the GSE118389 single-cell RNA-seq dataset, SOCS1 was primarily expressed in T cells, while HMOX1 was predominantly expressed in macrophages at the single-cell level (Fig. 1K-T), which was further validated in four additional datasets in TISCH (Supplementary Fig. 5A-H). At the protein level, FTH1, ISCU, and SOCS1 showed moderate staining in breast cancer tissues on the Human Protein Atlas portal (HPA) (Supplementary Fig. 5I). These results strongly indicated that dysregulation and imbalance of the FRGs contributed to the progression and malignancy of TNBC.

Ferroptosis-related subtypes identification and pathway enrichment analysis

We performed consensus clustering to identify ferroptosis-related molecular subtypes based on the mRNA expression of the eight FRGs in the FUSCC TNBC cohort. This analysis divided the patients into two distinct groups (K = 2) with high and low intergroup correlations (Fig. 2A-E). PCA analysis demonstrated that the distinct subgroups could be differentiated based on the expression of the eight FRGs (Fig. 2F). Kaplan-Meier analysis revealed that patients in cluster 1 had better overall survival (OS) compared to those in cluster 2 (Fig. 2G). Tumors in cluster 2 were primarily composed of immunomodulatory (IM) and BLIS mRNA subtypes, which were associated with a younger age distribution (Supplementary Fig. 6A-C, Supplementary table 3). Research from MD Anderson Cancer Center has suggested that the BLIS mRNA subtype has the worst outcome [11]. The expression of LPCAT3 and NOX4 was higher in cluster 2, while EMC2 showed no difference, and the remaining FRGs exhibited an opposite profile (Fig. 2H-I). Univariate Cox regression analysis indicated that cluster 2 was a risk factor (HR = 2.47) associated with worse OS (Fig. 2J). Consistent with this, validation using the NMF Clustering algorithm showed that patients in the FUSCC TNBC cohort could also be divided into two distinct subgroups with significantly different OS probabilities (Supplementary Fig. 7A-C). Similarly, unsupervised clustering analysis in the GSE76124 and GSE21653 datasets clearly divided TNBC patients into two groups with different OS, and PCA revealed the distribution of FRG-related subtypes (Supplementary Fig. 7D-K).

To further explore the involved Gene Ontology (GO) processes and enriched pathways contributing to the progression of the distinct subgroups, we identified differentially expressed genes (DEGs) using a fold change cutoff of 1.5, as illustrated in the volcano plot (Fig. 2K). GO analysis of the up-regulated DEGs showed a significant enrichment of iron ion binding molecular function (MF) in cluster 2, while the down-regulated DEGs were predominantly enriched in T cell activation and leukocyte proliferation biological processes (BP) (Fig. 3A-B). Collectively, tumors in cluster 2, which exhibited worse OS, showed an increased iron ion overload and decreased T cell activation, strongly suggesting their increased sensitivity to ferroptosis. Furthermore, using Gene Set Variation Analysis (GSVA), we investigated the impact of tumor cell ferroptosis on the reprogramming of the tumor microenvironment (TME). Tumors in cluster 2 exhibited higher enrichment scores for programmed cell death, metal ion SLC transporters, drug metabolism cytochrome p450, and epithelial-mesenchymal transition (EMT) pathways (Supplementary Fig. 8). Previous studies have confirmed that tumor cells in a high-mesenchymal state are more sensitive to ferroptosis due to enhanced iron endocytosis during EMT [15, 25]. Additionally, tumors in cluster 2 showed lower enrichment scores for immune-related pathways, such as natural killer cell-mediated cytotoxicity, leukocyte transendothelial migration, and T cell and B cell receptor signaling pathways (Fig. 3D-E, Supplementary table 4). Overall, tumors in cluster 2 represented a subgroup of cancers that were more sensitive to ferroptosis, exhibited an immunosuppressive phenotype, and had worse OS compared to those in cluster 1.

Immune landscape characterization within ferroptosis-related subtypes

Ferroptotic tumor cells release damage-associated molecular patterns (DAMPs) that can impact the infiltration, differentiation, and function of immune cells in the tumor microenvironment (TME), thereby influencing tumor growth [18]. To comprehensively illustrate the immune signature of the two ferroptosis-related subtypes and the role of ferroptosis in TME remodeling, we performed a comprehensive analysis. We observed marked differences in the expression of chemokines, interleukins, interferons, and receptors between the two distinct subgroups (Fig. 4A). Tumors in cluster 2 exhibited lower immune score, stromal score, ESTIMATE score, dysfunction score, and TIS score, as well as higher tumor purity, TIDE score, and exclusion score (Fig. 4B-I). Furthermore, by employing Cibersort, we inferred the immune infiltration in tumors and found lower abundance of resting NK cells and activated CD4 memory T cells in cluster 2 (Fig. 4J). This observation may be attributed to the low expression of SOCS1, which showed a positive correlation with the number of resting NK cells based on Spearman correlation analysis (Fig. 4K). Additionally, the network plot illustrated the complex communications among immune cells in the TME of TNBC (Fig. 4L). Finally, tumors in cluster 2 exhibited lower expression of the major histocompatibility complex (MHC) complex and higher expression of immune checkpoints (Fig. 4M-N). In summary, tumors in cluster 2 were more likely to display a cold tumor phenotype characterized by reduced immune infiltration and increased expression of immune checkpoints, thus establishing an immunosuppressive TME.

Construction of the FerrScore model in TNBC

We developed the FerrScore model based on the expression of eight FRGs to investigate how ferroptosis affects the tumor phenotype and TME remodeling, and to assess the predictive accuracy of the FRGs signature in chemotherapy and immunotherapy response in TNBC. By applying a cutoff value of 0.03, we identified high and low FerrScore groups with significantly different overall survival (OS), and the FerrScore model achieved an area under the curve (AUC) of 0.70 for 3-year survival prediction (Fig. 5A-C, Supplementary table 5). The two groups differed in terms of age distribution, mRNA type, intrinsic subtype, paclitaxel treatment status, tumor grade, and relapse-free survival (RFS) status of patients (Supplementary table 6). Among the two distinct TNBC subtypes, cluster 2 exhibited significantly higher FerrScore than cluster 1, and patients in cluster 2 were classified into the high FerrScore group with worse OS (Fig. 5D-E). Moreover, the high FerrScore group had a higher proportion of luminal androgen receptor (LAR) mRNA subtype and a lower proportion of basal intrinsic subtype (Fig. 5F). Among the eight FRGs, the expression of EMC2 positively correlated with tumor FerrScore, whereas HMOX1 showed a distinct negative correlation (Fig. 5G). Additionally, the expression of LPCAT3, EMC2, and NOX4 was higher in the high FerrScore group, while the other five FRGs were down-regulated in this group (Fig. 5H).

We further investigated the immune signature within the high and low FerrScore groups to examine the impact of FerrScore on TME reprogramming. Interestingly, FerrScore displayed a negative correlation with the expression of MHC complex molecules and immune checkpoints (Fig. 5I-J). Additionally, FerrScore exhibited a positive correlation with metal ion SLC transporters, tumor escape from immune attack, and PD-1 signaling pathways, while showing a negative correlation with T cell and B cell receptor signaling pathways (Fig. 5K). Furthermore, FerrScore was positively associated with stromal score, TIDE score, and exclusion score, while negatively correlated with TIS score and immune score (Fig. 5L). We inferred the immune signatures between the two groups using Cibersort, XCELL, ssGSEA, TIMER, and MCP counter algorithms, and found that tumors with low FerrScore had a higher abundance of immune cells in the TME (Supplementary Fig. 9). Overall, FerrScore served as an effective indicator to determine the immune phenotype of TNBC, which provides a solid foundation for its application in screening promising candidate drugs and predicting response to immune checkpoint inhibitor therapy.

A. Classification of patients into high and low FerrScore groups using the "ggrisk" R package with a cut-off value of 0.03. B. Kaplan-Meier curve demonstrating worse overall survival (OS) in patients with high FerrScore. C. ROC curve illustrating the specificity and sensitivity of FerrScore in predicting 1-, 3-, and 5-year survival of TNBC patients. D. Variations in FerrScore between the two distinct clusters; Wilcoxon test. E. Sankey diagram displaying the correlations among clusters, FerrScore, recurrence-free survival (RFS) status, intrinsic subtype, and mRNA subtype using the "ggalluvial" R package. F. Heatmap showing the correlation between FerrScore and CNA subtype, iCluster subtype, intrinsic subtype, mRNA subtype, mutation subtype, and SNF subtype in TNBC patients. G. Pearson correlation analysis between the expression of the eight FRGs and FerrScore. H. mRNA expression levels of the eight FRGs in the high and low FerrScore groups; Student’s t-test. I-J. Comparison of the expression levels of MHC complexes and immune checkpoints between the high and low FerrScore groups; Student’s t-test, *p < 0.05. K-L. Pearson correlation analysis between FerrScore and immune-related signaling pathways and scores.

Screening Candidate Drugs for Chemotherapy Guided by FerrScore

Large amounts of preclinical evidence have revealed the potential of inducing ferroptosis in tumors to prevent acquired chemotherapy resistance and enhance the efficacy of traditional cancer therapies. Synergistic effects have been observed between ferroptosis inducers and conventional drugs in suppressing tumor growth in mouse models of head and neck cancer [26]. We initially applied the FerrScore scoring model to human breast cancer cell lines in the CCLE database, which also revealed the presence of two distinct clusters (Supplementary Fig. 10A). Cluster 2 tumors exhibited higher sensitivity to NUTLIN-3A(-), AFATINIB, and AZD8055, albeit with lower AUC values, while cluster 1 tumors demonstrated greater sensitivity to KU-55933, MASITINIB, PAC-1, AZD7762, BI-2536, and PELITNIB (Supplementary Fig. 10B-J).

To further investigate the potential role of FerrScore in predicting chemotherapy response, we utilized the FerrScore model to screen candidate compounds that could guide chemotherapy for patients with TNBC. Firstly, we applied the FerrScore model to the CTRP and PRISM datasets, resulting in the identification of five CTRP-derived drugs (BRD5468, ZSTK474, AZD6482, Sildenafil, and Canertinib) and five PRISM-derived drugs (Temsirolimus, Idasanutlin, Everolimus, CGM097, and Nutlin-3) for tumors with high FerrScore, using Pearson correlation analysis (Fig. 6A-D, Supplementary table 7). Secondly, we calculated the fold changes of the target genes of the candidate agents, with higher fold changes indicating greater potential for candidate drugs. Thirdly, to determine the most promising agents for high FerrScore tumors, we calculated the CMap score by submitting the 300 differentially expressed genes (DEGs) with the most significant fold changes to the CMap website. Notably, Everolimus exhibited the highest CMap score of -90.58 (Fig. 6E-F). Finally, we conducted a search to explore published experimental evidence and the clinical trial status of the candidate drugs, which revealed that Everolimus has been recognized as a therapeutic agent for advanced breast cancer (Supplementary table 7), further affirming its status as the most promising compound for patients with higher FerrScore.

FerrScore-based prediction of ICIs therapy to optimize immunotherapy

ICIs therapy has revolutionized the clinical treatment of cancer patients. It has been demonstrated that anti-PD-L1 antibodies act synergistically with ferroptosis activators, such as erastin, to suppress tumor cell growth [24]. The release of DAMPs from ferroptotic tumor cells alters the TME and influences the response to ICIs therapy. In light of this, we utilized the FerrScore scoring model to predict the benefits of anti-PD-L1, anti-PD-1, and anti-PD-1&CTLA-4 ICIs immunotherapy, as well as adoptive T cell therapy.

Initially, we examined an immunotherapy cohort (IMvigor210) of urothelial carcinoma treated with anti-PD-L1. Patients were divided into high and low FerrScore groups, with those in the high FerrScore group exhibiting worse overall survival (OS) (Fig. 7A). Notably, patients with low FerrScore were more likely to benefit from anti-PD-L1 therapy (Fig. 7B-C). Furthermore, tumors with lower FerrScore demonstrated higher tumor mutational burden (TMB) and tumor neoantigen burden (TNB) (Fig. 7D-E). Increased TMB leads to a higher TNB, enhancing the chances of T cell recognition and clinically correlating with improved ICIs therapy outcomes [27]. The FerrScore model achieved an area under the curve (AUC) value of 0.56 for predicting anti-PD-L1 ICIs benefits in the IMvigor210 cohort (Fig. 7F).

Next, we applied the FerrScore model to an anti-PD-1 therapy cohort (GSE78220) comprising melanoma patients. Patients with high FerrScore had worse OS (Fig. 7G). Patients with low FerrScore displayed a more favorable response to anti-PD-L1 therapy, and the AUC value of this model for predicting therapy response was 0.74 (Fig. 7H-J). Subsequently, we divided patients with melanoma (PRJEB23709) undergoing anti-PD-1&CTLA-4 therapy into high and low FerrScore groups. Patients with low FerrScore exhibited better OS and derived greater benefits from anti-PD-1&CTLA-4 therapy (Fig. 7K-M). The AUC value of the FerrScore model to predict anti-PD-1&CTLA-4 therapy benefits was 0.61 (Fig. 7N). Lastly, we classified patients with melanoma (GSE100797) undergoing adoptive T cell therapy into high and low FerrScore groups but observed no difference in OS or therapy benefits between the two groups (Fig. 7O-Q). Overall, FerrScore emerges as a promising biomarker for predicting the benefits of ICIs therapy, facilitating the optimization of therapy options for patients.

SOCS1 and NOX4 are associated with immunity to TNBC

Our primary focus was to delve into the role of the FRGs signature in reprogramming the tumor microenvironment (TME). To accomplish this, we conducted a detailed examination of SOCS1 and NOX4, based on the above analysis. Pearson correlation analysis revealed a strong positive correlation between the expression of SOCS1 and the mRNA levels of PDCD1 (PD-1), CD274 (PD-L1), CTLA-4, LAG-3, and HAVCR2 (TIM3). Conversely, the expression of NOX4 showed a negative correlation with the mRNA levels of PDCD1, IDO1, and LAG3 (Fig. 8A-H). Given these findings, PD-1 was selected as the key marker to explore the relationship between the two FRGs and the shaping of immune phenotypes. We categorized TNBC samples into cold and hot immune phenotypes based on CD3 IHC staining and examined the expression of SOCS1 and NOX4 across the two tumor subtypes (Fig. 8I). Our analysis revealed that tumors with high expression of NOX4 and low expression of SOCS1 were characterized as cold tumors, while tumors with high expression of SOCS1 and low expression of NOX4 exhibited phenotypes associated with hot tumors (Fig. 8I-J). These results validate the active involvement of SOCS1 and NOX4 in TME infiltration and remodeling in TNBC.

Immunotherapy-induced IFNγ released by activated CD8 + T cells has been shown to trigger ferroptosis-specific lipid peroxidation in cancer cells, thereby promoting cancer immunotherapy [24]. The release of DAMPs from ferroptotic tumor cells has the ability to modulate the innate immune system and subsequently reshape the TME [28]. Ferroptosis in tumor cells has been demonstrated to enhance the response to radiotherapy, chemotherapy, and immunotherapy in various types of tumors [14]. In this study, we identified eight FRGs associated with survival in the FUSCC cohort. We explored the dysregulation and methylation patterns of these FRGs across multiple cancer types, as well as genetic variations and expression profiles specifically in TNBC. Our findings revealed imbalances in the expression profiles of FRGs among different cancer types, including TNBC, contributing to the phenotypic heterogeneity observed in TNBC within clinical practice.

In the FUSCC cohort, patients with TNBC were classified into two molecular subtypes of ferroptosis based on significantly different OS: cluster 1 (low FerrScore) and cluster 2 (high FerrScore). Tumors in cluster 2 exhibited poorer OS and were characterized by lower immune score, TIS score, and estimate score, as well as higher TIDE score, exclusion score, and tumor purity. These findings indicate a negative correlation between ferroptosis and tumor immune infiltration. We further observed a negative association between ferroptosis and immune-related pathways, such as natural killer cell-mediated cytotoxicity, T cell and B cell receptor signaling pathways. Additionally, MHC complexes and immune checkpoints were found to be rarely expressed in tumors of cluster 2, likely due to the limited infiltration of immune cells. Previous studies have suggested that tumors with abundant lymphocyte infiltration have better prognoses across a wide range of cancers [29]. Moreover, a negative correlation was identified between ferroptosis and the enrichment scores of programmed cell death pathway, metal ion SLC transporters pathway, and drug metabolism cytochrome p450 pathway. To further explore the role of ferroptosis in tumor molecular subtyping and TME reprogramming, we developed a scoring system based on FRGs, which we referred to as FerrScore. We observed a negative correlation between FerrScore and the expression of MHC complexes and immune checkpoints. Tumors with high FerrScore displayed higher stromal score, TIDE score, and exclusion score, while having lower TIS score and immune score. Furthermore, Pearson correlation analysis revealed that tumors with higher FerrScore were significantly enriched in metal ion SLC transporters pathway, tumor escape from immune attack pathway, and PD-1 signaling pathway. Our study suggests that ferroptosis in TNBC reshapes the TME into a "cold" phenotype characterized by reduced lymphocyte infiltration and immune checkpoint expression, thus limiting T cell recognition and responsiveness to anticancer therapy.

Chemotherapy remains the primary treatment option for advanced TNBC, as recommended by the 2022 version of the CSCO guidelines. However, chemotherapy resistance poses a significant challenge in cancer therapy. Previous studies have demonstrated that ferroptosis inducers synergize with traditional compounds, such as cisplatin, to inhibit tumor growth in mouse models of head and neck tumors [26]. By conducting Spearman correlation analysis, we identified six CTRP-derived and six PRISM-derived drugs that showed potential for patients with high FerrScore, offering valuable insights for clinical therapy options. Further assessment led us to confirm Everolimus as the most promising agent among the 12 candidate drugs for patients with high FerrScore, considering comprehensive factors such as CMap score, experimental evidence, and clinical trial status. Moreover, we applied FerrScore to four immunotherapy cohorts to predict the benefits of ICIs. Our study revealed that patients with lower FerrScore are more likely to benefit from anti-PD-L1, anti-PD1, and anti-PD-1&CTLA-4 ICIs therapy. In summary, FerrScore represents a potentially powerful metric for predicting the benefits of chemotherapy and ICIs therapy, providing new perspectives for the clinical treatment of TNBC.

Our focus centered on two ferroptosis genes, SOCS1 and NOX4, which exhibited a significant correlation with the expression of immune checkpoints in TNBC within the FUSCC cohort. Based on CD3 expression, TNBC displayed two immune phenotypes, and we observed significantly higher expression of SOCS1 in hot tumors, while NOX4 showed elevated expression in cold tumors of TNBC. Additionally, we investigated a positive correlation between the expression of SOCS1 and PD-1, whereas NOX4 demonstrated a negative association with PD-1 expression through IHC and IF staining. SOCS1, known as a negative feedback regulator of cytokine signaling, has been shown to suppress CD8⁺ T cell proliferation in a mouse model of acute inflammatory arthritis [30]. Additionally, SOCS1 was identified as a negative regulator of PD-L1 through a barcoding system [31]. Moreover, SOCS1 has been characterized as an intracellular negative checkpoint for CD4⁺ T cells, and its inactivation restored proliferation suppression and anticancer efficacy in both murine and human CD4⁺ T cells [32, 33]. Surprisingly, we observed a positive correlation between SOCS1 expression and the infiltration of activated CD4⁺ memory T cells and resting NK cells in TNBC, as well as PD-1 mRNA expression. Based on these findings, distinct roles and mechanisms of SOCS1 in PD-1 and PD-L1 expression emerged across different disease types. Previous studies have shown that siRNA knockdown or pharmacologic inhibition of NOX4 promoted intratumoral CD8⁺ T cell infiltration in various human cancer types, while also restoring immunotherapy response induced by cancer-associated fibroblasts (CAF) [34]. In summary, through a variety of experiments, we confirmed the significant influence of SOCS1 and NOX4 on immune landscape reprogramming and immune phenotype reshaping in TNBC, indicating their potential role in enhancing the effectiveness of immunotherapy for TNBC.

Although we utilized the largest worldwide mRNA sequencing cohort of TNBC to identify ferroptosis-related molecular subtypes, we acknowledge several limitations in our research. Firstly, our data primarily focused on patients from China, which may not fully represent the diverse ethnic populations affected by TNBC. Therefore, future analyses and verifications should include patients from multicenter clinical cohorts to ensure broader applicability and generalizability. Secondly, although we conducted some IHC and IF experiments, additional mechanistic studies are urgently required to elucidate the specific mechanisms through which SOCS1 and NOX4 influence T cell infiltration and PD-1 expression in the TME of TNBC. Our research group is currently engaged in additional studies to delve deeper into this subject.

In conclusion, our study revealed the existence of two distinct ferroptosis-related molecular subtypes in TNBC, which exhibited significant differences in overall survival and the immune landscape. These findings highlight the high degree of heterogeneity within TNBC. We developed a FerrScore based on eight ferroptosis-related genes, and discovered a negative correlation between FerrScore and immune infiltration, indicating a reshaping of the tumor microenvironment towards an immunosuppressive phenotype. Furthermore, our investigation confirmed Everolimus as the most promising candidate agent for patients with high FerrScore, and demonstrated the potential of FerrScore as a powerful metric for predicting the benefits of anti-PD-L1, anti-PD-1, and anti-PD-1&CTLA-4 ICIs therapies.

AUC, area under the dose-response curve; BLIS, basal-like immunosuppressed; CCLE, Cancer Cell Line Encyclopedia; CNV, copy number variation; CSCO, Chinese Society of Clinical Oncology; CTRP, Cancer Therapeutics Response Portal; DAMPs, damage-associated molecular patterns; DEGs, differentially expressed genes; EMT, epithelial-mesenchymal transition; FRGs, ferroptosis-related genes; GO, Gene Ontology; GSVA, Gene Set Variation Analysis; H&E, hematoxylin and eosin; ICIs, immune checkpoint inhibitors; IF, immunofluorescence; IHC, immunohistochemical; OS, overall survival; PCA, principal component analysis; PRISM, Profiling Relative Inhibition Simultaneously in Mixtures; RCD, regulated cell death; RFS, relapse-free survival; ROC, receiver operating characteristic; TIDE, Tumor Immune Dysfunction and Exclusion; TIS, tumor inflammation signature; TMB, tumor mutational burden; TME, tumor microenvironment; TME, tumor neoantigen burden; TNBC, triple-negative breast cancer

Author contributions

Conceptualization: TZ, QY, JJ; Data curation: YC, XH, TZ; Formal analysis: YC, XH, TZ; Funding acquisition: TZ, QY; Investigation: TZ, HW, YZ, LT; Methodology: ZS, TL, YX, JJ; Software: XH, TZ; Roles/Writing - original draft: TZ, QY; Writing - review & editing: TZ, QY.

Founding

This work was sponsored by Shanghai Sailing Program (23YF1406800), China Postdoctoral Science Foundation (2023M730675), Jiangsu medical scientific research project of Jiangsu Health Commission, the National Natural Science Foundation of China (81870409), Project of Educational Commission of Guangdong Province of China (2021KCXTD023), and the 789 Outstanding Talent Program of SAHNMU (789ZYRC202070102).

Availability of data and materials

The data are available within the article, supplementary information, or available from the authors upon request. Source data are provided with this paper.

Ethics approval and consent to participate

Not applicable.

Consent for publication

All authors have read and agreed to the published the manuscript.

Conflict of interest

The authors declare no competing interests.

Sharma P, Barlow WE, Godwin AK, Parkes EE, Knight LA, Walker SM, et al. Validation of the DNA Damage Immune Response Signature in Patients With Triple-Negative Breast Cancer From the SWOG 9313c Trial. J Clin Oncol. 2019; 37: 3484–92.
Kim C, Gao R, Sei E, Brandt R, Hartman J, Hatschek T, et al. Chemoresistance Evolution in Triple-Negative Breast Cancer Delineated by Single-Cell Sequencing. Cell. 2018; 173: 879–93 e13.
Traina TA, Miller K, Yardley DA, Eakle J, Schwartzberg LS, O'Shaughnessy J, et al. Enzalutamide for the Treatment of Androgen Receptor-Expressing Triple-Negative Breast Cancer. J Clin Oncol. 2018; 36: 884–90.
Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021; 71: 209–49.
Boughey JC, Ballman KV, McCall LM, Mittendorf EA, Symmans WF, Julian TB, et al. Tumor Biology and Response to Chemotherapy Impact Breast Cancer-specific Survival in Node-positive Breast Cancer Patients Treated With Neoadjuvant Chemotherapy: Long-term Follow-up From ACOSOG Z1071 (Alliance). Ann Surg. 2017; 266: 667–76.
Ferreira R, Napoli J, Enver T, Bernardino L, Ferreira L. Advances and challenges in retinoid delivery systems in regenerative and therapeutic medicine. Nat Commun. 2020; 11: 4265.
Mai WX, Gosa L, Daniels VW, Ta L, Tsang JE, Higgins B, et al. Cytoplasmic p53 couples oncogene-driven glucose metabolism to apoptosis and is a therapeutic target in glioblastoma. Nat Med. 2017; 23: 1342–51.
Lehmann BD, Bauer JA, Chen X, Sanders ME, Chakravarthy AB, Shyr Y, et al. Identification of human triple-negative breast cancer subtypes and preclinical models for selection of targeted therapies. J Clin Invest. 2011; 121: 2750–67.
Lehmann BD, Jovanovic B, Chen X, Estrada MV, Johnson KN, Shyr Y, et al. Refinement of Triple-Negative Breast Cancer Molecular Subtypes: Implications for Neoadjuvant Chemotherapy Selection. PLoS One. 2016; 11: e0157368.
den Hollander P, Rawls K, Tsimelzon A, Shepherd J, Mazumdar A, Hill J, et al. Phosphatase PTP4A3 Promotes Triple-Negative Breast Cancer Growth and Predicts Poor Patient Survival. Cancer Res. 2016; 76: 1942–53.
Burstein MD, Tsimelzon A, Poage GM, Covington KR, Contreras A, Fuqua SA, et al. Comprehensive genomic analysis identifies novel subtypes and targets of triple-negative breast cancer. Clin Cancer Res. 2015; 21: 1688–98.
Jiang YZ, Ma D, Suo C, Shi J, Xue M, Hu X, et al. Genomic and Transcriptomic Landscape of Triple-Negative Breast Cancers: Subtypes and Treatment Strategies. Cancer Cell. 2019; 35: 428 – 40 e5.
Xiao Y, Ma D, Yang YS, Yang F, Ding JH, Gong Y, et al. Comprehensive metabolomics expands precision medicine for triple-negative breast cancer. Cell Res. 2022; 32: 477–90.
Chen X, Kang R, Kroemer G, Tang D. Broadening horizons: the role of ferroptosis in cancer. Nat Rev Clin Oncol. 2021; 18: 280–96.
Muller S, Sindikubwabo F, Caneque T, Lafon A, Versini A, Lombard B, et al. CD44 regulates epigenetic plasticity by mediating iron endocytosis. Nat Chem. 2020; 12: 929–38.
Doll S, Proneth B, Tyurina YY, Panzilius E, Kobayashi S, Ingold I, et al. ACSL4 dictates ferroptosis sensitivity by shaping cellular lipid composition. Nat Chem Biol. 2017; 13: 91–8.
Verma N, Vinik Y, Saroha A, Nair NU, Ruppin E, Mills G, et al. Synthetic lethal combination targeting BET uncovered intrinsic susceptibility of TNBC to ferroptosis. Sci Adv. 2020; 6.
Tang D, Kang R, Coyne CB, Zeh HJ, Lotze MT. PAMPs and DAMPs: signal 0s that spur autophagy and immunity. Immunol Rev. 2012; 249: 158–75.
Yao L, Zhao MM, Luo QW, Zhang YC, Liu TT, Yang Z, et al. Carbon Quantum Dots-Based Nanozyme from Coffee Induces Cancer Cell Ferroptosis to Activate Antitumor Immunity. ACS Nano. 2022.
Tang R, Wu Z, Rong Z, Xu J, Wang W, Zhang B, et al. Ferroptosis-related lncRNA pairs to predict the clinical outcome and molecular characteristics of pancreatic ductal adenocarcinoma. Brief Bioinform. 2022; 23.
Zhang W, Yao S, Huang H, Zhou H, Zhou H, Wei Q, et al. Molecular subtypes based on ferroptosis-related genes and tumor microenvironment infiltration characterization in lung adenocarcinoma. Oncoimmunology. 2021; 10: 1959977.
Liang JY, Wang DS, Lin HC, Chen XX, Yang H, Zheng Y, et al. A Novel Ferroptosis-related Gene Signature for Overall Survival Prediction in Patients with Hepatocellular Carcinoma. Int J Biol Sci. 2020; 16: 2430–41.
Ayers M, Lunceford J, Nebozhyn M, Murphy E, Loboda A, Kaufman DR, et al. IFN-gamma-related mRNA profile predicts clinical response to PD-1 blockade. J Clin Invest. 2017; 127: 2930–40.
Wang W, Green M, Choi JE, Gijon M, Kennedy PD, Johnson JK, et al. CD8(+) T cells regulate tumour ferroptosis during cancer immunotherapy. Nature. 2019; 569: 270–4.
Viswanathan VS, Ryan MJ, Dhruv HD, Gill S, Eichhoff OM, Seashore-Ludlow B, et al. Dependency of a therapy-resistant state of cancer cells on a lipid peroxidase pathway. Nature. 2017; 547: 453–7.
Roh JL, Kim EH, Jang HJ, Park JY, Shin D. Induction of ferroptotic cell death for overcoming cisplatin resistance of head and neck cancer. Cancer Lett. 2016; 381: 96–103.
Jardim DL, Goodman A, de Melo Gagliato D, Kurzrock R. The Challenges of Tumor Mutational Burden as an Immunotherapy Biomarker. Cancer Cell. 2021; 39: 154–73.
Proneth B, Conrad M. Ferroptosis and necroinflammation, a yet poorly explored link. Cell Death Differ. 2019; 26: 14–24.
Ganesan AP, Clarke J, Wood O, Garrido-Martin EM, Chee SJ, Mellows T, et al. Tissue-resident memory features are linked to the magnitude of cytotoxic T cell responses in human lung cancer. Nat Immunol. 2017; 18: 940–50.
Egan PJ, Lawlor KE, Alexander WS, Wicks IP. Suppressor of cytokine signaling-1 regulates acute inflammatory arthritis and T cell activation. J Clin Invest. 2003; 111: 915–24.
Wroblewska A, Dhainaut M, Ben-Zvi B, Rose SA, Park ES, Amir ED, et al. Protein Barcodes Enable High-Dimensional Single-Cell CRISPR Screens. Cell. 2018; 175: 1141-55 e16.
Sutra Del Galy A, Menegatti S, Fuentealba J, Lucibello F, Perrin L, Helft J, et al. In vivo genome-wide CRISPR screens identify SOCS1 as intrinsic checkpoint of CD4(+) TH1 cell response. Sci Immunol. 2021; 6: eabe8219.
Guo H, Li R, Wang M, Hou Y, Liu S, Peng T, et al. Multiomics Analysis Identifies SOCS1 as Restraining T Cell Activation and Preventing Graft-Versus-Host Disease. Adv Sci (Weinh). 2022; 9: e2200978.
Ford K, Hanley CJ, Mellone M, Szyndralewiez C, Heitz F, Wiesel P, et al. NOX4 Inhibition Potentiates Immunotherapy by Overcoming Cancer-Associated Fibroblast-Mediated CD8 T-cell Exclusion from Tumors. Cancer Res. 2020; 80: 1846–60.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Identification of Triple-Negative Breast Cancer Subtypes Based on Ferroptosis- Related Gene Signature for Targeted Treatment Prioritization

Status:

Version 1

Abstract

Figures

Introduction

Materials and Methods

Data acquisition and filtration

Genetic variations and expression profiles of FRGs

FRGs-based unsupervised clustering

Gene Ontology (GO) and Gene Set Variation Analysis (GSVA) between distinct subtypes of TNBC

Characterizing the Tumor Immune Landscape in Distinct Subtypes of Triple-Negative Breast Cancer

Construction of the FerrScore scoring model

FerrScore-based drug sensitivity screening

FerrScore-based ICIs therapy benefits prediction

Histological analysis

Four-color immunofluorescence (IF)

Statistical Analysis

Results

Identification, genetic variations and expression profiles of the FRGs

Ferroptosis-related subtypes identification and pathway enrichment analysis

Immune landscape characterization within ferroptosis-related subtypes

Construction of the FerrScore model in TNBC

Screening Candidate Drugs for Chemotherapy Guided by FerrScore

FerrScore-based prediction of ICIs therapy to optimize immunotherapy

SOCS1 and NOX4 are associated with immunity to TNBC

Discussion

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1