Introduction
The ferroptosis is an iron-dependent form of regulated cell death (RCD) that has been recently discovered. It differs from the apoptosis, necrosis, and autophagy [
1]. The implementation of ferroptosis requires the activation of the following three ferroptosis features: The oxidation of phospholipids containing polyunsaturated fatty acids (PUFA); The presence of redox active iron; and the loss of lipid peroxide repair abilities [
2]. With the in-depth analysis of ferroptosis, the induction of ferroptosis has been identified as a vital event involved in pathological progressions, including human tumors. Preliminary evidences have suggested the regulatory effect of ferroptosis on the growth of various types of cancers like renal cancer, pancreatic cancer, non-small-cell lung cancer (NSCLC) and diffuse large B-cell lymphoma [
3]. The ferroptosis has been identified to suppress tumor growth and the progression, and as a result, the induction of ferroptosis has emerged as a promising anti-cancer treatment [
4].
Lung cancer is the most prevalent one among malignancies and it is the chief leading cause of tumor-related deaths worldwide. Pathologically categorized, about 85% of lung cancer cases belong to NSCLC, of which lung adenocarcinoma (LUAD) is one of the most frequent subtypes [
5]. With the wide-spreading application of targeted drugs and immune checkpoint inhibitors, therapeutic options of patients with LUAD have revolutionarily changed. However, the prognosis of metastatic or recurrent LUAD is still far away from satisfying [
6]. Besides, the overall survival of lung cancer patients significantly varies across the world, with a 5-year survival of 21.2% in the United States, which can be higher than that in China [
7]. Recent studies have reported that up-regulation of the GSH synthesis pathway in NSCLC cells can suppress ferroptosis [
8],
9. NFS1, as a ferroptosis-related gene, is detected highly expressed in LUAD cells [
10]. In addition, ferroptosis is also correlated to the prognosis of renal carcinoma and hepatocellular carcinoma [
11],
12. We therefore speculated whether ferroptosis is correlated to the prognosis of LUAD, and the possible involvement of ferroptosis-related genes.
RNA sequencing data of ferroptosis-related genes and clinical information of LUAD patients were downloaded from the public databases. It is shown that expression levels of ferroptosis-related genes were correlated to survival outcomes of LUAD. In the present study, LUAD patients were divided into a training set and a validation set based on the random stratified sampling of tumor stages. Then, we established the multi-gene LUAD prognosis prediction model and calculated risk scores through the LASSO regression with tenfold cross-validation, and univariate and multivariate Cox regression analyses. Finally, the established model was verified in the validation set and the overall sample set, aiming to test the fitting degree of the model. To explore the underlying molecular mechanism of the difference in the prognosis of LUAD, we further performed immune and biological functional enrichment analyses.
Discussion
The current study systematically identified the correlation between 60 ferroptosis-related genes and the prognosis (overall survival) of LUAD. In this study, the LUAD dataset was divided into the training set and validation set by the random stratified sampling method. The prognostic prediction model involving 5 genes was established for the training set samples through the LASSO regression with tenfold cross-validation and univariate and multivariate Cox regression analysis, which was validated in the actual clinical practice.
Previous evidences have confirmed the vital functions of ferroptosis-related genes in the entire ferroptosis process. Nevertheless, the specific influence of a single ferroptosis-related gene on the prognosis of a certain type of cancer remains unclear. We combined the mRNA expression levels of each gene with actual clinical characteristics to analyze the differential expression between LUAD and paracancerous specimens. Moreover, quantitative data of gene expressions were analyzed for their predictive potential in the survival of LUAD. Subgrouped by the risk score, identified genes between the high-risk group and low-risk group were analyzed for their differential variations, biological functions and pathways they were mainly enriched in. In conclusion, our screened DEGs in LUAD patients may be potential targets involved in the occurrence and development of LUAD.
In the constructed predictive model, a total of 5 genes (
ACSL3,
ACSL4,
GSS,
PEBP1,
PGD) were involved in, which were closely associated with the process of ferroptosis. The
ACSL3 gene is responsible for exogenous monounsaturated fatty acids to protect cells against ferroptosis, and it is negatively correlated with ferroptosis sensitivity [
23]. The
ACSL4 gene is essential for proferroptosis. Knockdown of ACSL4 inhibits erastin-induced ferroptosis, and its overexpression can restore ferroptosis sensitization. Re-expression of flag-tagged human wild-type (WT) ACSL4 (ACSL4-Flag) in Acsl4 KO (Acsl4
−/−) Pfa1 cells restores full sensitivity to ferroptosis induction, and knockdown of it significantly prolongs survival compared to vehicle-treated mice. Knockout of ACSL4 in ferroptosis-sensitive cells protects erastin- and RSL3-induced cell death [
24],
25. The
GSS gene provides instructions for making glutathione synthetase. The glutathione-dependent lipid hydroperoxidase GPX4 contributes to prevent ferroptosis by converting lipid hydroperoxides into non-toxic lipid alcohols. Overexpression of PEBP1 increases the sensitivity of HK2 cells to RSL3, and knockdown of PEBP1 in HAEC and HT22 cells yields an opposite result [
26]. The
PGD gene is involved in erastin-induced ferroptosis [
1].We modeled and calculated the corresponding risk score based on the expression data of 5 candidate genes, and then divided LUAD patients into high-risk group and low-risk group. Our established model effectively predicted the survival of LUAD patients in the training set, the validation set and the total cases. Meanwhile, the gender, age, and tumor stage of LUAD patients were taken into consideration, and thus a composite nomogram was established, which was much closer to the actual clinical practice. Among them, the gender and age of LUAD patients had relatively a small effect on the prognosis, whereas the tumor staging and risk score posed a greater one. The above results were consistent with our investigation expectations.
To explore the factors for the prognosis difference between the high-risk group and low-risk group, we compared the SNV levels and analyzed the differences in biological functions of the two groups. After subgrouping the acquired SNV statistics according to high-risk and low-risk scores, the mutation frequency of each gene and the number of mutations in each sample were calculated, which were displayed as the mutation landscape. The results revealed that LUAD patients in the high-risk group presented a higher frequency of each gene mutation. Moreover, the top 5 genes with the highest mutation frequency in the high-risk group, involving the TP53 (53%), TTN (50%), MUC16 (42%), CSMD3 (40%) and RYR2 (39%) genes presented 5–13% higher mutation frequency than those in the low-risk group. Calculating the mutation frequency easily identified the source of the variations between groups. Genomic analyses about the prognosis difference between high-risk and low-risk LUAD patients need to be performed in the future.
Functional enrichment analysis and immune infiltration score on DEGs between the two groups were further performed. DEGs were mainly enriched in the cytochrome P450, steroid hormones, IL-17 signaling pathways, staphylococcus aureus infection pathways, etc. The cytochrome P450 oxidoreductase (POR) is necessary for triggering the ferroptosis in cancer cells [
27]. Moreover, POR boosts the execution of ferroptosis by engaging in the peroxidative modification of phospholipids in the cell membrane. PPARα, as a member of the steroid hormone receptor superfamily, can inhibit iron overload and lead to ferroptosis by combining with the GPX4 and facilitating its expression. The preliminary evaluations indicated that the steroid hormones in adrenal sebaceous cells are significantly affected by ferroptosis induction [
28]. Iron plays a crucial role in the continuous evolution process for
staphylococcus aureus by establishing efficient iron transportation systems. On the one hand,
staphylococcus aureus consumes hemoglobin in the red blood cells of the host and serum through the transport systems encoded by the iron-regulated surface determinants located in cell wall. On the other hand, it acquires the iron by the siderophores with a high affinity for iron. A previous study has shown that the reactive oxygen species can result in the drug resistance in
staphylococcus aureus infection [
29]. Previous GO enrichment revealed that biological functions like microtubule bundle formation and keratinization are mainly enriched in the iron [
30,
31]. Therefore, we concluded that the variations in these pathways might be the inherent factors for the differences between the high-risk group and low-risk group. Our findings provide theoretical references for underlying the mechanism of ferroptosis in lung cancer patients.
Although constructing a prognostic model is of great significance in the TCGA-LUAD cohort, only internal verification and overall verification were performed in the verifying process of the model. Specifically, the external validation set was not included in the examination of the final results. In functional analysis, enrichment analysis was performed in DEGs between the high-risk group and low-risk group, while functions of these pathways were not be experimentally verified. As a result, we only identified which pathways and functions were responsible for triggering the prognosis difference in high-risk and low-risk LUAD patients, but how they induced it needs to be further explored.
In general, we established a prognostic prediction model based on 5 ferroptosis-related genes by analyzing the LUAD dataset. In the training and validation set, this model was found effectively predict the overall survival of LUAD, and more important, clinical features of LUAD patients were taken into consideration, which remarkably simulated the actual clinical practice. Our findings provided a promising tool in predicting the prognosis of LUAD patients, and theoretical references for explaining the prognosis difference between high-risk and low-risk LUAD.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.