Breast cancer has been the most common cancer type and the first leading cause of cancer-associated deaths in females worldwide [47, 48]. Due to the heterogeneity of breast cancer, particularly in terms of carcinogenesis, molecular subtypes, treatment responses, and clinical outcomes, it is difficult to predict prognosis of breast cancer [49–52]. Currently, even after treatments with surgical resection, endocrinotherapy, chemotherapy, radiotherapy, and immunotherapy, recurrence or malignant progression is still unavoidable. The application of prognostic models is beneficial for guiding clinical decisions and precision medicine.
Iron metabolism has recently been found to play a vital role in the development, progression, and tumor microenvironment of breast cancer [8–11, 53]. Iron metabolism has also emerged as the possible therapeutic target and prognostic marker for breast cancer. Nowadays, the model based on ferroptosis-related genes in breast cancer has been found to be established in many studies [18–20], and the iron metabolism signature has been employed in lung adenocarcinoma, glioma, clear cell renal cell carcinoma, and hepatocellular carcinoma [21–24]. However, prognostic value and therapeutic importance of iron metabolism-related genes in breast cancer remain unclear.
As the method to assess whole-genome expression profile microarray data, GSEA can integrate data scientifically from different levels and sources without significant differences in gene thresholds [54]. In this study, GSEA was performed using mRNA expression profile data of 970 breast cancer patients. Analysis of the 514 genes found to be related with iron metabolism and their potential functions revealed that GSEA were mainly enriched in the three gene sets. 7 prognostic genes for breast cancer patients were identified using univariate, multivariate Cox, and LASSO regression models. Thus, based on these seven most valuable prognostic genes, we established and verified a model to predict clinical information in breast cancer patients. Advantageously, survival analysis confirmed significant differences between high and low risk groups, which proves the value of the signature. The model was then validated in the GEO datasets, which confirmed a favorable clinical prediction ability. In addition, this prognostic model has been proved to be an independent prognostic risk factor through univariate and multivariate Cox analyses. We also found that the patients with higher risk score were inclined to be older, advanced disease progress and poorer prognosis. Moreover, we established a novel nomogram by integrating our prognostic model and clinical information, which provided admirable assessment of prognosis. Furthermore, the results displayed that C-index, ROC and calibration plot revealed favorable outcomes in our study. These results confirmed that the calculation of risk score is of great significance to the prognosis of breast cancer patients. The underlying causes of this result should be further investigated in the future. These findings revealed that calculating risk scores has a significant prognostic and clinical importance for breast cancer patients. This not only improves the ability to forecast prognosis, but it may also assist physicians in selecting more appropriate treatment options for patients.
The treatments of breast cancer involves a multi-mode strategy, combining surgery, chemotherapy, targeted molecular therapy, radiotherapy, and hormone therapy [55, 56]. Thus, identifying predictive indicators for breast cancer treatments is critical. In our study, we found that high risk group was positively associated with worse survival in breast cancer patients with surgery, chemotherapy, targeted molecular therapy, radiotherapy, and hormone therapy, suggesting that our risk signature might serve as a biomarker to predict the responses to multi-mode strategies of breast cancer patients. Since iron is required for cellular proliferation, cellular iron metabolism has been linked to cancer development and resistance to therapies [57]. This also demonstrated the extensive clinical applications of our iron metabolism-related signature.
Considering risk signature derived from iron metabolism, which was substantially connected with adaptive and innate immunity, the potential function of risk model in complexity of immune and stromal cells, which are the main components of the TME, was further investigated [58–60]. Herein, when compared to the low risk group, the high risk group had significantly higher immune, stromal, and estimate scores, but lower tumor purity. We found that elevated infiltration levels of macrophages M0, mast cell resting and NK cells activated were related to high-risk group of patients, suggesting that patients with a high risk score had an immunosuppressive phenotype. Moreover, we found that high risk group was positively connected with the phenotype of CAFs, which are considered as a tumor-promoting component of the stroma in most malignancies [61, 62]. These findings suggested that iron metabolism-related signature might be linked to the regulation of TME, hence influencing tumor development and progression.
Clinical outcomes from the trials studying anti-PD-1, anti-PD-L1 or anti-CTLA-4 treatment for patients with solid tumors have arracted the interest of researchers in the current era of immunotherapy. Only a small number of patients, nevertheless, benefitted from immunotherapy. It is necessary to identify the effective biomarkers for patients who were likely to benefit from immunotherapy. In this study, we found there were significant differences between immune checkpoints, TMB, TIDE and our risk model in the TCGA breast cancer cohort. Importantly, iron metabolism-related signature was found to be predictive of outcome following anti-PD-L1 in urothelial carcinoma, indicating that it might serve as a biomarker to predict the patient's response to immunotherapy.
The 7 iron metabolism-related genes identified in this study included CA1, CYP2A6, CYP2C9, P4HA3, PQLC1/SLC66A2, RIOK3, and UBC. Of these genes, CA1 (carbonic anhydrase I) has been identified as a potential oncogene that promotes abnormal cell calcification, apoptosis, and migration in breast cancer via regulating the expression of the androgen receptor (AR) and down-regulation of X-box binding protein 1 (XBP1). MCF-7 cells treated with anti-CA1 siRNA showed increased migration and apoptosis [63]. P4HA3 (prolyl 4-hydroxylase subunit alpha 3) is a component of prolyl 4-hydroxylases (P4HAs) involved in collagen synthesis, and the knockdown of P4HA3 significantly inhibited breast cancer progression and metastasis in vitro and vivo [64]. High P4HA3 expression is associated with poor prognosis in breast cancer [65]. RIOK3 (RIO kinase 3) has been reported to be an oncogene which leads to the invasion and metastasis of breast cancer in an HIF1α-dependent manner during hypoxia exposure [66]. Bioinformatics analysis found that high RIOK3 expression predicted poor prognosis in breast cancer [67]. UBC (ubiquitin C) is high expressed in breast cancer cell lines, including MCF7, SKBR3 and MDA-MB231 [68], and the hub gene of HER2 positive breast cancer with brain metastasis [69]. CYP2A6 and CYP2C9 are clinically important isoforms of cytochrome P450s (CYPs), which are associated with the development and progression of breast cancer and the effectiveness of cancer treatments [70]. Due to the function in the metabolism of tamoxifen, paclitaxel, and other treatments, CYP2C9 has been found to affect breast cancer-free survival [71]. Bieche et al. found that CYP2A6 is one of seven genes that code for important xenobiotic-metabolising enzymes implicated in breast tumorigenesis and that have a greater expression in ER-positive breast tumors compared to ER-negative tumors and normal mammary tissues [72]. However, PQLC1/SLC66A2 gene was discovered to be predictive in breast cancer patients for the first time. Further studies on the biological activities of these genes in breast cancer are required.
To the best of our knowledge, this study is the first to find and thoroughly assess prognostic iron metabolism-related genes for the prediction of survival in breast cancer patients using data from the public TCGA database. A novel risk signature based on 7 iron metabolism-related genes was found and confirmed. In addition, we successfully created an iron metabolism-related nomogram incorporating clinical variables to predict the OS of breast cancer patients using a quantitative approach. We also explored the associations between our risk model and TME, and used an existing credible immunotherapy cohort of urothelial cancer to evaluate our gene signature. Our findings were important for directing future research as well as for improving our understanding of the genetics of breast cancer.
However, our research has some limitations. First, it was a retrospective study, and all breast cancer patients were found using public databases. Second, to verify the accuracy of our model in predicting outcomes and assess how well it can be used in practice to manage breast cancer, large-scale multicenter cohorts are required. Furthermore, to confirm our results and clarify the functional roles of 7-iron metabolism-related genes involved in the onset and progression of breast cancer, additional basic research will need to be conducted in future.