GC is a relatively common gastrointestinal tumor, and its incidence is gradually increasing. It is characterized by a high recurrence rate and a large difference in prognosis[26, 27]. Currently, treatment options for GC is mainly radical excision, supplemented simultaneously by postoperative radiotherapy, chemotherapy, targeted therapy and immunotherapy. Accurate prediction of patient prognosis can benefit patients in subsequent treatment. Therefore, it is possible to develop an appropriate treatment plan for the individual differences in order to improve the prognosis. Traditional clinicopathological factors have been widely used to predict the prognosis and progression of tumors at present[30, 31]. While, individual difference in reaction is more genetic difference. With the standardized quantification of gene testing procedures, gene expression can be used to predict the prognosis of tumors and even served as a target for new treatments[32, 33]. As a newly discovered form of regulatory cell death different from apoptosis, necrosis and autophagy in terms of morphologic, biochemical and genetic aspects, ferroptosis is induced by abnormal accumulation of iron-dependent lipid reactive oxygen species. Studies have found that ferroptosis was closely related to tumor development, and it could promote apoptosis in tumor cell by inducing ferroptosis. With the study of the mechanism of ferroptosis, increasing ferroptosis-related genes have been found and the genes associated with prognosis may well predict prognosis of clinical patients[35, 36]. Compared with traditional prediction models, this study combines the ferroptosis-related prognostic gene risk score and clinical tumor pathological information to predict the prognosis of GC, which will be more accurate and be beneficial for clinical decision-making and individualized treatment.
In this study, analysing in TCGA dataset and ferroptosis-related genes, we screened 59 ferroptosis-related DEGs in gastric cancer. Functional enrichment analysis showed that the DEGs were strongly linked biological processes of tumor cell oxidation, and AGE-RAGE signaling pathway and HIF-1 signaling pathway were the most abundant signaling pathways. The univariate survival prognostic analysis showed that 14 genes were associated with prognosis and the further lasso-COX regression analysis revealed 4 hub genes associated with ferroptosis that influenced the prognosis of gastric cancer. The down-regulation of MYB and CHAC1 expression was identified as tumor protection genes, and the up-regulation of NOX4 and ATF3 expression was related to the poor prognosis of tumors. In the results, the risk scores of the four genes were independent prognostic factors for GC, and the prognosis of patients in the low-risk group was significantly higher than that of the high-risk group. The training cohort in TCGA and GSE26901 in GEO were respectively used to be internal and external validation, which confirmed the predictive property of four hub genes. At last, the nomogram was constructed by a combination of the risk score with clinicopathological information, and performed to accurately predict overall survival (OS).
Four hub genes that predict the prognosis of GC in this study have been previously reported to be related to tumor development. MYB has been proved to be associated with tumor-associated macrophages in breast cancer, and high expression of MYB can be regarded as evidence of macrophage aggregation. MYB is also related with ferroptosis, and some study have shown that the ferroptosis produced by a classic inducer erastin could reduce through inhibition of MYB. The specific mechanism of the action may be related to cysteine dioxygenase 1 (CDO1). Studies have shown that artesunate can enhance the ferroptosis in DAUDI and CA-46 cells by activating ATF4-CHOP-CHAC1 pathway, and reducing expression of CHAC1 can enhance resistance of tumor cell to ferroptosis. NOX4 knock-down could significantly improve remodeling of left ventricular and reduce cardiomyocyte death, meanwhile, knocking down NOX4 could inhibit ferroptosis due to NOX4 mainly involved in hydrogen peroxide and lipid peroxidation in biological processes[40–42]. Wang et al. found that ATF3 could accelerate ferroptosis induced by Erastin by inhibiting the XC-system, which was essential for glutamate synthesis. In previous studies, the four hub genes mentioned above (MYB, CHAC1, NOX4, AFT3) were all promoters for ferroptosis, while MYB and CHAC1 with down-regulation were considered to be tumor protective genes and NOX4 and ATF3 with up-regulation were associated with poor prognosis. Whether these genes play a role in the prognosis of GC patients by influencing the process of iron death remains to be further investigated.
Immunotherapy is a conventional method for tumor treatment. Ferroptosis has been shown to play a significant role in tumor development, but further studies are remained to be explored to uncover the potential mechanisms between ferroptosis and tumor immunity. Therefore, we also analyzed the correlation of tumor immunity between high and low risk groups. In this study, there are significant differences in the content of the antigen presentation process between the high- and low-risk groups. The mechanism may be that ferroptosis cells attract antigen-presenting cells (APCs) to accumulate inside by releasing different signals, such as lipid signals. Furthermore, we found both in TCGA dataset and GSE26901, the content of Macrophages reached a high level in the high-risk groups. Previous studies have also shown that the aggregation of tumor macrophages is closely related to the occurrence and development of tumors, as it may play a role in immune invasion[44, 45]. In addition, the low expression of MHC_class_I in high-risk patients was found in both databases, which may be related to tumor immune impairment.
Combining the characteristics of the four hub genes and clinicopathological parameters, our nomogram can be used by clinicians to determine the prognosis of each patient. Its scoring system is easy to understand and facilitates customized treatments and medical decisions. Our research can provide new insights into the molecular mechanism and prognosis prediction of GC. In addition, the DEGs obtained in this study comes from a comprehensive analysis of multiple data sets and has high reliability. Prior to this study, four genes signature has not yet seen the report related to GC. These genes may be potential molecular targets against GC.
However, we must acknowledge that the study has some limitations. First of all, our main source of clinical information is from the data sets of TCGA and GEO databases, in which most of patients are white, African or Latino. So, we must be cautious when extending our findings to patients of other races. Secondly, the establishment and validation of the nomogram were based on existing database. Therefore, it is necessary to use prospective data with complete clinical information and gene expression information to verify its clinical effectiveness in the future. In addition, it should be emphasized that the link between ferroptosis-related DEGs risk score and immune activity has not been resolved in experiments, which is also one of our regrets.