In recent years, various prognostic signatures have been proposed to predict the prognosis and immune infiltration of malignant tumors[26]. Including predicting the immune infiltration state and prognosis of malignant tumors by screening immune-related genes or FRGs, and few studies have used a method of combining the FRGs with the immune-related genes to predict the immune infiltration state and prognosis of tumors. Therefore, in this study, an ferroptosis-related prognosis signature was constructed by combining ferroptosis with immunity to better predict the prognosis and immune response of GC patients.
Firstly, we obtained 16 GC prognostic genes related to ferroptosis, and divided GC patients into two subgroups by consistent clustering. Cluster C2 showed a significant median survival advantage, while Cluster C1 showed a poor prognosis. Compared with Cluster C1, GC patients in Cluster C1 had significantly higher ESTIMATE score, higher immunocyte infiltration and higher matrix score. MCPCounter, TIMER and CIBERSORT algorithms also indicated that Cluster C1 had more immunocyte infiltration, which indicated that high-risk groups have higher levels of immunocyte infiltration and interstitial components in the tumor microenvironment. TME is generally divided into three categories: immune inflammation, immune rejection and immune desert [27]. In this study, based on the high-risk group's manifestations of high immune cell infiltration abundance and large proportion of interstitial components but with poor prognosis, it was speculated that TME in Cluster C1 of the high-risk group met the immune exclusion subtype. Although the TME of the patients in the Cluster C1 group had a large number of immunocyte infiltrates, they were not effective at penetrating the tumor parenchyma to eliminate tumor cells. Therefore, the prognosis of high-risk group was often poor. Subsequently, a risk signature was constructed based on LASSO analysis, and 13 genes were finally obtained, which showed that TUBE1, NFE2L2, ACSL4 genes were protective factors, while ZFP36, NOX5, and MIR9-3 genes were risk factors. Kaplan-Meier analysis showed that all the constructed prognostic genes were independent prognostic markers for GC patients.The ROC curve showed the reliability and stability of the risk signature constructed. The ESTIMATE algorithm was used to evaluate the TIME between the high-risk group and the low-risk group. The results showed that the high-risk group had higher ESTIMATE score, higher immune cell infiltration and higher matrix score, which were the same as the results of the above study, suggesting that TME in the high-risk group met the immunologic exclusion subtype. The TIMER database showed that the key prognostic genes for ferroptosis were closely related to the infiltration of macrophages, B cells, T cells, dendritic cells, and neutrophils. The HPA database examined the immunohistochemical staining of 13 critical prognostic genes for ferroptosis and found that the protein expressions of ZFP36, TUBE1, NFE2L2, GCH1, GABARAPL2, CHAC1, CAPG, ACSL4, ACO1, and SLC1A4 in GC and normal tissues were significantly different, and there was no expression of NOX5, MIR9-3, and NOX4 proteins in the HPA. A study had shown that autophagy promotes ferroptosis by degrading anti-iron death factors [28], and ZFP36 was a key protein for autophagy and considered to be related to ferroptosis [29]. NFE2L2, a known transcription factor involved in the encoding of GC development, is overexpressed as a prognostic marker of GC [30]. OS rate in GC patients with NRF2 positive expression was significantly reduced[31]. The experiment conducted by Wei [32] proved that GCH1 induced immunosuppression through a 5-HTP-AHR-ID01-dependent mechanism, and that the combination of metabolic intervention and immunotherapy of this pathway might be a promising strategy for the treatment of triple-negative breast cancer (TNBC), and the GCH1 inhibitor could be used as an analgesic [33]. Members of the GABARAP family (GABARAP, GABARAPL1/GEC1 and GABARAPL2/GATE-16) are one of the subfamilies of the ATG8 protein family, which are related to the receptor and autophagy pathway[34]. The high-expression of GABARAP is related to the good prognosis of tumors[35].CHAC1 is an enzyme related to the activity of γ-glutamyl cyclotransferase that can degrade intracellular GSH and promote ferroptosis of tumor cells[36], which has been proved to be related to glioma [37]and breast cancer [38]. CAPG is particularly abundant in macrophage expression [39], and CAPG had been proved to be related to tumor cell invasion and tumorigenic [40]. SLC1A4 is one of the members of solute carrier family 1(SLC1), and SLC1A4 is one of the important roles of amino acid transporter [41]. SLC1A4 is highly expressed in pancreatic ductal adenocarcinoma and liver cancer cells, and some studies have suggested that SLC1A4 may promote the process of ferroptosis[42]. ACSL4, a long-chain fatty acyl coenzyme, is closely related to the proliferation and migration of tumor cells[43]. ACSL4 had been shown to be overexpressed in breast cancer [44], GC[45], and liver cancer[46]. ACO1(Cytoplasmic aconitic acid hydratase) is a protein that participates in cytoplasmic and mitochondrial metabolism and, when down-regulated, leads to cell death [47]. NOX is a family of encoded oxidases, NOX4 is a catalytic subunit of nicotinamide adenine dinucleotide phosphate (NADPH) oxidase complex, and NOX5 mainly encodes calcium-dependent NADPH oxidase, produces superoxide, and acts as a calcium-dependent proton channel. The ROS produced by NOX4 is involved in a variety of biological functions, including signal transduction, cell differentiation and tumor cell growth [48, 49], and NOX4 plays an important role in the process of ferroptosis [50]. Inhibition of NOX4 can significantly block ferroptosis[51]. MiRNA plays an important role in tumors. MiR-9 is overexpressed in lung cancer tissues[52], and MiR-9 acts as a biomarker for poor prognosis in lung cancer and thyroid papillary carcinoma[53]. There is no report about the relationship between TUBE1 and tumor in the literature. These results indicated that the risk signature constructed had a strong potential for prognosis prediction of GC patients, and there was a strong correlation between this prognosis signature and GC immunity.
We verified the accuracy of the constructed prognostic signature by verifying the cohort, and the results showed that this prognostic signature had a strong potential for prognosis prediction of GC patients, and there was a strong correlation between this prognostic signature and GC immunity. Nomograms that combine risk signatures with clinical features can accurately predict the prognosis of patients with GC.
Finally, KEGG enrichment analysis showed that DEGs was mainly enriched in the Cell cycle, p53 signaling pathway, IL-17 signaling pathway, MAPK signaling pathway, and PI3K-Akt signaling pathway. p53 is a tumor suppressor gene, and p53 mutations have been reported in many cancers[54]. When p53 mutations occur, cells proliferate abnormally and transform into cancer cells. GC patients with p53 mutation have worse prognosis than those without mutation [55]. More and more evidences support the pathogenic role of IL-17 in cancer formation, including colon cancer [56] and lung cancer [57]. Wu [58]found that IL-17 could promote tumor angiogenesis by mediating the up-regulation of VEGF in GC through STAT3 pathway. It has been confirmed that MAPK and PI3K-Akt pathways are involved in many processes of the occurrence and development of GC[59–61]. The results of GSEA analysis showed that compared with Cluster C2, Cluster C1 had low expression in lipid metabolism and glutathione metabolism, which were important metabolic pathways in the occurrence of ferroptosis, which might be related to the poor prognosis of GC.
The above results indicated that the expression of prognosis genes related to ferroptosis was related to the immune microenvironment of GC patients. It can be used to accurately predict the prognosis of GC patients and provide a new strategy for immunotherapy of GC patients.
However, this study has certain limitations. Our signature was constructed and validated based on retrospective data, without relevant experimental verification.