1 Introduction
Lung cancer (LC) is one of the fastest growing malignant tumors in term of morbidity and mortality, and one of the most threatening to human health and life. According to statistics, LC patients accounted for 13% of all new diagnoses and 24% of all cancer deaths in 2019 [
1]. The vast majority of patients diagnosed with non-small cell lung cancer (NSCLC), the most common form of LC and accounting for 85%, are already at advanced or distant stages [
2,
3]. Despite the current advances in surgery, radiotherapy and immunotherapy, the 5-year survival rate is only 4–17% [
1,
4]. In recent years, diagnosis and prognosis models have become increasingly abundant in medical research, and they provide a useful reference for cancer diagnosis and prediction of cancer recurrence or death [
5,
6]. Therefore, this study aims to construct predictive models that provide potential key biomarkers for LC.
Ferroptosis was first proposed in 2012, and is a novel form of programmed cell death [
7]. Unlike autophagy and apoptosis, ferroptosis is a type of iron- and reactive oxygen species (ROS)-dependent cell death, and its mainly characterized by cytological changes [
8]. New evidences suggested that ferroptosis maybe an adaptive process that was essential for eliminating cancer-causing cells [
9‐
11]. In LC, ferroptosis was first triggered using Erastin (an ferroptosis activator) in A549 cells with K-ras mutant [
12]. Subsequently, Erastin was found to sensitize LC cells to the apoptosis-inducing agent cisplatin by inhibiting glutathione peroxidase to reduce glutathione in an ferroptosis manner [
13]. Therefore, ferroptosis-related genes (FRGs) are highly likely to serve as biomarkers with great potential in diagnostic and prognostic models of LC.
Cancer biology and immunosurveillance are inextricably linked. In the process of tumor development, the complex tumor immune microenvironment (TIME) closely interacts with tumor cells and tumor stroma, which have an invaluable role in monitoring and preventing tumor growth [
14]. A central link between cancer biology and TIME is the iron competition between tumor cells and the immune system [
15]. Iron is closely associated with the regulation of innate and adaptive responses in TIME, particularly in T cells and macrophages [
16]. Macrophages resident in tissues are the “gatekeepers” of iron homeostasis, which absorb, metabolize, store, and export iron to meet the needs of the surrounding cells [
17]. In tumor immunity, iron is necessary for t cell proliferation and effector function [
18]. Hence, it is important to explore the correlation between FRGs and TIME for the diagnosis and prognosis of LC.
To explore the role of FRGs in the clinical diagnosis and prognosis of LC, in current study, a comprehensive analysis of LC cohorts in the cancer genome atlas (TCGA) and gene expression omnibus (GEO) databases was performed by bioinformatics methods to identify FRGs that are closely associated with LC prognosis. In addition, a FRGs-based LC diagnosis and prognosis prediction model was constructed, and the relationship between FRGs and immune infiltration of LC was explored. Our diagnosis and prognosis model may enhance early diagnosis of LC and ameliorate personalized prognostic assessment.
4 Discussion
LC incidence and mortality rates are increasing year by year, posing a significant health burden to society. Ferroptosis is a novel programmed cell death closely associated with excess iron loading and may have novel molecular mechanisms in tumor immunity and tumor suppression [
7]. As previously described that unfavorable prognosis in LC patients was significantly correlated with iron dysregulation [
36]. Epidemiological and clinical studies also revealed that iron acts as an essential function in the evolution of LC [
37]. Immune system is a dynamic and complex network, and tumor progression and its reaction to treatment are closely monitored by the immune system [
38]. Previous studies have shown that cancer cells undergoing ferroptosis could release oxidized lipid that modulate antitumor immunity [
10]. Therefore, it is a necessary to identify the key FRGs biomarkers affecting the prognosis of LC, which is of great importance for early diagnosis and improvement of clinical outcome for LC.
In this study, we identified 8-FRGs, and the prognostic model constructed based on 8-FRGs could independently predict the prognosis of LC patients and had good predictive performance. The corresponding nomograms also help clinicians improve clinical decision-making and formulation of treatment plans. In addition, the diagnostic model based on 18-FRGs has high specificity and sensitivity for early diagnosis of LC. In the immune infiltration analysis, we found that the prognostic model had higher proportions of Macrophages_M0, Mast_cells_activated, Neutrophils and NK_cells_resting in the high-risk group than the low-risk group, which indicated the correlation between ferroptosis and TIME, and suggested that the poor prognosis of high-risk groups may be related to strong immunosuppressive effects. Furthermore, we found that immune checkpoints B7H3, CD112, CD155, B7H5, and ICOSL in the high-risk group were increased. These differences promoted the progression of LC and led to poor prognosis of LC.
The 8 FRGs (ACSL3, FADS2, GLS2, HSF1, PANX1, PHKG2, VDAC2, and CDKN1A) that we selected to be associated with the diagnosis and prognosis of lung cancer have been shown to play important roles in cancer and tumor immunity. Acyl‐CoA synthetase long‐chain family member 3 (ACSL3) is a member of the long-chain acyl-COA synthetase family and a lipid-metabolizing enzyme that converts free fatty acids to fatty acid-CoA. The expression of ACSL3 increased in prostate cancer cells, which promoted the growth of CRPC by promoting the synthesis of dehydroepiandrosterone and preventing the catabolism of active androgen [
39]. ACSL3 facilitated growth of LC cell, and was exceptionally high expression in LC tissues [
40]. ACSL3-mediated fatty acid oxidation was essential for lung carcinogenesis with KRAS mutant [
41], and numerous articles have shown that ACSL3 is a key gene in the prognostic model of LUAD [
42]. Knockdown of ACSL3 impeded pancreatic ductal carcinoma progression, which regulated fibrotic and ratio of immune cells in TIME [
43]. Fatty acid desaturase 2 (FADS2) acted as a desaturating agent mainly by introducing a double bond at the δ6 position of the fatty acid chain, which was the first rate-limiting enzyme for the conversion of upstream fatty acids to PUFA. As previously described, FADS2 was higher expression in LC tissues than in paraneoplastic tissues [
44], and knockdown of FADS2 led to a remarkable increase in iron and lipid ROS in LC cells, and eventually LC cells underwent ferroptosis [
45]. Glutaminase 2 (GLS2) was a highly mobile and multiple positioning protein that transfer to both the mitochondria and the nucleus, and nuclear translocation of GLS2 was associated with proliferation inhibition and cell differentiation in LC [
46]. Dias et al. [
47] reported that GLS2 promoted breast cancer development by promoting the proliferation and metastasis of breast cancer cells. What’s more, GLS2 could be used as a therapeutic target of ferroptosis in cardiomyocytes [
48]. Heat shock factor 1 (HSF1) as a major regulator in protein homeostasis, and it has been demonstrated to be up-regulated in LC cells, and was necessary for brain metastasis in vivo [
49]. A clinical study showed that overexpression of HSF1 was a biomarker of unfavorable prognosis in LC [
50]. In tumor immunity, HSF1 inhibition triggers loss of NK cell activation ligand MICA/B [
51].
Purines from pannexin 1 (PANX1) was a channel-forming glycoprotein found in tumor cells and other cells in TIME, including immune cells, which played an important role in the exchange of information between cells, due to its main function of forming large-pore single-membrane channels that related release of ATP and metabolites [
52]. PANX1 has been revealed to promote metastasis in a variety of tumors including hepatocellular carcinoma [
53], testicular cell carcinoma [
54], and breast cancer [
55]. Phosphorylase kinase G2 (PHKG2) can be used as a biomarker for thyroid cancer [
56], endometrial cancer [
57], renal clear cell carcinoma [
58] and colorectal cancer [
59]. However, the function of PANX1 in LC is still to be further investigated. Voltage dependent anion channel 2 (VDAC2) acts as a mediator of oxidative stress response and regulates production of ROS, translocation of Bax and release of cytochrome c during ME-344 (a therapeutic isoflavone)-induced mitochondria-mediated apoptosis in LC cells [
60]. Mcl-1 was upregulated in NSCLC, and Mcl-1 promoted migration by increasing mitochondrial Ca2
+ uptake and ROS production through direct interaction with VDAC2 [
61]. Cyclin-dependent kinase inhibitor 1A (CDKN1A) encodes a potent cyclin-dependent kinase inhibitor. At the same time, the expression of CDKN1A is tightly regulated by tumor suppressor protein P53, which mediates p53-dependent cell cycle G1 arrest in response to a variety of stress stimuli [
62]. The expression of CDKN1A was increased in NSCLC, and knock-down of CDKN1A can significantly promote apoptosis and G1 phase arrest [
63].
Compared with previous prognostic models [
64,
65], our prognostic model had a larger sample size and was more comprehensive, and we constructed a diagnostic model for the diagnosis of LC. In this study, the diagnostic and prognostic models have excellent predictive performance, and can help clinicians improve clinical decision-making and formulation of treatment plans. Unfortunately, there are still more limitations in this study. For example, we combined the samples of lung squamous cell carcinoma and LUAD in non-small cell lung cancer for analysis, but the fact was that there were some differences in the prognosis of lung adenocarcinoma and lung adenocarcinoma. In subsequent studies, we will conduct further individual analyses of the subtypes of lung squamous cell carcinoma and lung adenocarcinoma in order to obtain a more rigorous diagnostic and prognostic model. In addition, this study is only retrospective and requires prospective studies to corroborate each other's results; functional experiment of 8 FRGs in LC are lacking for validation.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.