Breast cancer (BRCA) has entered the era of precision treatment at the molecular level. The molecular hallmarks of BRCA, including ER, PR, HER2, Ki-67, and PD-1, and PD-L1, have been employed for personalized and individualized treatment (Harbeck et al., 2019; Waks and Winer, 2019). For instance, the endocrine therapy, HER2 targeted therapy, and immune checkpoint therapy are employed for ER/PR-positive, Her2-positive, and PD-1/PD-L1-positive BRCA tumors, respectively².

Since the advancement of microarray and high-throughput sequencing, multi-gene prediction, such as the PAM50 signature (Cheang et al., 2015), 70-gene assay, and 21-gene recurrence score², has been widely used to guide decision making in the therapeutic approach for various BRCA subtypes. Multi-gene prediction has been commonly used to predict the benefits of chemotherapy or to estimate patient’s prognosis. Thus, new biomarker combinations have been developed to improve the precision of current diagnostic and therapeutic modalities. Currently, researchers are looking for new therapeutic targets and biomarkers for BRCA treatment. Moreover, there is still an insurmountable therapeutic challenge for triple negative breast cancer (TNBC) because therapeutic targets and biomarkers have not yet been identified. A few previous studies found that several ferroptosis-related genes (FRG) could be promising therapeutic targets in BRCA (Hangauer et al., 2017; Zhang et al., 2019), especially for the TNBC subtype (Chen et al., 2017; Zhu et al., 2020; Ding et al., 2021; Zhang et al., 2021). Therefore, it is necessary to develop a panel involving FRG biomarkers for risk stratification and identification of new targets.

Ferroptosis refers to cell death resulting from iron-mediated lipid peroxidation and is characterized by intracellular accumulation of reactive oxygen species (ROS) (Supplementary Figure S3) (Dixon et al., 2012). According to preliminary data, ferroptosis inhibits tumor development and proliferation; hence, ferroptosis can be targeted for cancer therapy (Stockwell et al., 2017). Correspondingly, researchers have increasingly focused on the role of ferroptosis in BRCA, especially in TNBC and Her2-positive BRCA. Due to the high recurrence and metastatic rate, TNBC and Her2-positive BRCA have been regarded as refractory and aggressive BRCA subtypes (Foulkes et al., 2010; Cesca et al., 2020). Ma et al. (Dixon et al., 2012) found that siramesine and lapatinib induced ferroptosis more than other canonical ferroptotic reagents did, implying that lapatinib participated in modulating ferroptosis without targeting EGFR and HER2 (Ma et al., 2016). Moreover, a recent study discovered that neoadjuvant neratinib induced ferroptosis and prevented brain metastasis in Her2-positive BRCA (Nagpal et al., 2019). Furthermore, several studies reported that ferroptosis could be a useful therapeutic target in the treatment of TNBC (Chen et al., 2017; Zhu et al., 2020; Ding et al., 2021; Zhang et al., 2021). It has been found that through the stimulated GCN2-eIF2-ATF4 pathway, CHAC1 degradation of GSH increases cystine-starvation-induced ferroptosis in TNBC cells (Chen et al., 2017). The anti-TNBC impact of DMOCPTL was demonstrated in a cell death method test by triggering ferroptosis via GPX4 ubiquitination (Ding et al., 2021). Chen et al. found that treatment of TNBC cells with holo-Lf increased total iron concentration, boosted ROS production, increased the lipid peroxidation end product malondialdehyde, and improved ferroptosis (Zhang et al., 2021). In addition, chemotherapy, radiotherapy, and immunotherapy were all influenced by ferroptosis; thus, targeting both ferroptosis and the identified biomarkers could be an effective treatment strategy for BRCA (Chen et al., 2021).

The aim of this research was to develop a panel consisting of FRG that could be used to predict the disease-specific survival (DSS) of patients with BRCA and to develop a risk stratification system that could aid diagnosis and provide novel therapeutic strategies.

An overview of our methodology is summarized in Figure 1.

Cell death plays an essential role in the homeostasis of the body. An advantage of cell death is the prevention of the uncontrolled growth and proliferation of cancer cells, which have excessive energy demands to maintain their infinite self-renewal potential. Cancer is associated with alterations in energy metabolism, antioxidants, and intake of iron (Stockwell et al., 2017). Therefore, tumor cells are more vulnerable to iron-induced necrosis, which is also known as ferroptosis, due to their iron-dependent growth mechanism (Dixon et al., 2012).

A few study developed a similar model to predict BRCA prognosis and validated the FRG expression level using cell lines (Zhu et al., 2021a; Wu et al., 2021). However, it did not consider the molecular subtype and was not based on DSS. Further, only three genes in that model (G6PD, FLT3, and SLC1A4) overlapped with those in our DSS model. After the multivariate Cox regression analysis, the ferroptosis-related nine-gene panel that we discovered demonstrated an excellent prognostic prediction capability in the TCGA, METABRIC, and GSE3494 cohorts. The high-risk subgroup was significantly associated with a high tumor grade, an advanced TNM stage, and TNBC and Her2-positive subtypes, which were all related to poor survival and refractory treatment response. Conversely, the low-risk subgroup was positively associated with positive statuses of ER and PR and a negative HER2 status (Table2), which all corresponded to favorable survival outcomes in the traditional classification. Due to the high recurrence and metastatic rate, Her2-positive BRCA and TNBC have been regarded as refractory and aggressive subtypes (Foulkes et al., 2010; Cesca et al., 2020). Hence, researchers are focusing on exploring more therapeutic methods to fill this gap. Several studies suggested that targeting ferroptosis may be a useful therapeutic approach in the treatment of TNBC and Her2-positive BRCA (Chen et al., 2017; Nagpal et al., 2019; Zhu et al., 2020; Ding et al., 2021; Zhang et al., 2021).

The predictive capability of the nine-gene panel was more significant than those of traditional indicators (Figure 4A) in terms of tumor size, lymph node metastasis, and ER/PR and HER2 statuses, especially for TNBC and Her2-positive BRCA subtypes. This finding suggested that treatment could be escalated or de-escalated depending on patients’ risk stratification combined with canonical methods. However, the relationship between the nine genes and ferroptosis needs further exploration. The median follow-up of the TCGA was 2.5 years when the events was only 8% (Table 1), while METABRIC and GSE3494 cohorts were about 10 years, resulting in the AUC value for 3 years and 5 years DSS prediction is lower than the 2 years in TCGA. The nine-gene panel was a more reliable predictor of 2 years DSS in the TCGA (AUC = 0.806), METABRIC (AUC = 0.706), and GSE3494 cohorts (AUC = 0.624) (Figures 3D–F) because it had a high predictive accuracy for DSS, which was even better than those of the statuses of ER, PR, HER2, and lymph nodes as well as tumor size (Figures 4A,B).

A total of nine FRG constituted our panel. These FRG were ALDH3A2, SIAH2, G6PD, SLC1A4, FLT3, SQLE, EGLN2, SFXN5, and CHAC1. G6PD, SQLE, and CHAC1 were unfavorable genes for BRCA prognosis in this gene panel, and they were significantly overexpressed in the tumor compared with levels in the adjacent tissues (Figures 2B, 7). Contrastingly, the other genes were protective. A few studies have discovered that these nine genes are all associated with ferroptosis. The SQLE gene encodes squalene epoxidase to catalyze the oxidation of squalene that could change the lipid profile of tumor cells and protect them from ferroptosis (Garcia-Bermudez et al., 2019). In addition, it is one of the most significantly upregulated genes in many tumors, especially in BRCA (Xu et al., 2020). Moreover, Qin et al. (Shen et al., 2017) found that SQLE mRNA is stabilized by lnc030 in collaboration with poly (rC) binding protein 2(PCBP2), resulting in an increase in cholesterol synthesis. They suggested that targeting SQLE might be a potential mechanism of terbinafine in the treatment of BRCA (Qin et al., 2021). The G6PD gene encodes glucose-6-phosphate dehydrogenase, which produces NADPH to keep the glutathione (GSH) in balance to inhibit ferroptosis. In addition, G6PD eliminates ROS (Hao et al., 2018). Hence, the upregulation of G6PD facilitates cancer development and is thus associated with a poor prognosis in many forms of carcinomas (Ju et al., 2017; Chen et al., 2018; Zhu et al., 2021b). Likewise, the upregulated expression of CHAC1 could prognosticate unfavorable outcomes in BRCA (Goebel et al., 2012; Li et al., 2021). Additionally, CHAC1 has been discovered to decease intracellular GSH levels, enhancing tumor cell ferroptosis (Chen et al., 2017). Moreover, FLT3, ALDH3A2, and SIAH2 gene depletion or deficiency can result in ferroptosis (Kang et al., 2014; Hassannia et al., 2019; Chillappagari et al., 2020). A loss in ALDH3A2 triggers ferroptosis and cooperates with GPX4 inhibition (Yusuf et al., 2020). SIAH2-deficient cells exhibit increased vulnerability to ferroptosis, and re-expression of GPX4 can rescue these cells from ferroptosis (Chillappagari et al., 2020). Equally, EGLN2 knockdown inhibited ferroptosis in mice, researchers validated the mRNA level of prostaglandin-endoperoxide synthase two b y qPCR, it is as a marker for assessing ferroptosis in vivo; and EGLN2 gene could mediate HIF1A downregulation to promote ferroptosis (Yang et al., 2019). Only few researches have focused on SLC1A4 and SFXN5. However, they did not find an association between these genes and ferroptosis. Nonetheless, SLC1A4 may promote ferroptosis and may function as a marker of ferroptosis, according to the FerrDb website data. Moreover, SFXN5 may be involved in cellular iron ion homeostasis (Tifoun et al., 2021). These two genes need further study to confirm their relationship with ferroptosis. Further, these genes are all linked to the promotion or prevention of ferroptosis in various cancers via multiple mechanisms (Supplementary Figure S3); however, it is unclear whether these genes influence the prognosis of BRCA patients through ferroptosis.

Recently, researchers have been exploring the role of ferroptosis in tumor therapy. However, the possible regulation of tumor immunity and ferroptosis remains a mystery. The idea that immunity stimulates or inhibits cancer cells is widely known, and targeting the immune checkpoint has become a promising and potential therapeutic approach in recent years. Studies that involved the TCGA and METABRIC cohorts revealed that the majority of the types of immune cells in the high-risk subgroup had higher immune scores than did those in the low-risk subgroup, except for mast cells (Figure 6A). Furthermore, the high-risk subgroup also had excessively higher scores for most immune functions (Figure 6B). There may be a crosstalk between the ferroptosis of cancer cells and those of infiltrating immune cells. One theory is that ferroptotic cancer cells emit distinct signals, which cause phagocytosis and induce antigen presentation by dendritic cells (Friedmann Angeli et al., 2019). Meanwhile, the suppression of ferroptotic activity impairs the capacity of CD8⁺ T and natural killer cells to destroy cells in vivo (Wang et al., 2019).

The results of the GO analysis suggested that many immune-related pathways and biological processes, such as humoral immune response, circulating immunoglobulin-mediated human immune response, complement activation, and classical pathways were enriched. The enriched KEGG pathways were the IL−17 signaling pathway and the PPAR signaling pathway (Figures 5A,B). The IL−17 signaling transduction regulated the infiltration of CD8⁺ T cells and PD-1/PD-L1 in BRCA patients (Shuai et al., 2020), while the PPAR signaling pathway was activated in TNBC patients (Lin et al., 2021). Our results demonstrated that the TNBC subtypes (91% in TCGA, 98.7% in METABRIC, Figure 4C) were mostly found in the high-risk subgroup (Table2). The clinical trial of combination anti-PD-1/PD-L1 showed that the progression-free survival was significantly improved in TNBC patients with metastasis (Cortes et al., 2020). The mechanism of anti-PD-L1 antibodies was to trigger ferroptosis, which subsequently enhanced the efficacy of immunotherapy (Wang et al., 2019). The resistance to anti-PD-1/PD-L1 in TNBC cells inhibited ferroptosis and changed the proportion of macrophage cells (Jiang et al., 2021). Therefore, ferroptosis may be induced through changes in the immune system. According to our panel, BRCA patients are stratified into the high- and low-risk subgroups. The patients in the high risk subgroup, whose CTLA4 and PD-1statuses were both positive or negative, demonstrated a substantial clinical benefit from combination therapy with anti-CTLA4 and anti-PD-1 (Figures 6D,F). Therefore, the high-risk subgroup should be administered intensive treatment or immunotherapies, whereas the low-risk subgroup should be administered de-escalated treatment. The relationship between immunity and ferroptosis has not been thoroughly clarified. However, there might be a strong relationship between the immune microenvironment of the tumor and ferroptosis in BRCA patients. Thus, further research is needed to validate the above findings.

Our study has several limitations. First, since ferroptosis research is a new and rapidly expanding field, more FRG are likely to be discovered in the future. Second, because of the observed heterogeneity between different populations, the findings of this retrospective and cross-cohort research need validation by further prospective reviews involving multicenter cohorts. Third, since the data were downloaded from public databases, several important clinical details were not accessible. These inaccessible data included chemotherapy regimens, drug information, and tumor burden, and the lack of these data limited a more in-depth comparison among the TCGA, METABRIC, and GSE3494 data. Finally, since the results are based on RNA sequence, verification of protein expression in terms of immunohistochemistry is needed to conveniently apply our findings in clinical practice.

We constructed a DSS prognostic prediction panel involving nine FRG genes in BRCA. This panel was based on FRG and DSS events. In the TCGA, METABRIC, and GSE3494 cohorts, our panel was independently correlated with DSS prognosis. We also found that the tumor immune microenvironment and ferroptosis may have a strong inherent connection with BRCA. This panel could be used to evaluate prognosis and to select patients for escalation/de-escalation treatment. Our research provides a preliminary theory for clinically individualized therapy by targeting ferroptosis genes.