Clinical and functional significance of a novel ferroptosis‐related prognosis signature in lung adenocarcinoma

Dear Editor, Lung cancer remains the most common and fatal type of cancer around the world, and lung adenocarcinoma (LUAD) is the most common subtype.1 Although genetargeted therapies and immunotherapies have made great advances, most patients still suffer from drug-resistance or insensitivity to current regimens.2 Ferroptosis is a new regulated cell death form defined in 2012,3 recently, its relevance in multiple pathological process including malignant tumors4 has been reported, which is of great prognostic and therapeutic interests. Here, we aim to identify novel prognostic genes and construct a ferroptosis-related prognostic signature for LUAD. As shown in the flow chart (Figure 1A), RNA-seq data of LUAD were retrospectively downloaded from The Cancer Genome Atlas (TCGA) database as discovery cohort, which contained 505 tumor tissues and 59 normal lung tissues after exclusion of samples without clinical or survival information. GSE72094 datasets were obtained from gene expression omnibus, and a total of 398 cases were used as an independent validation cohort. Ferroptosis-related genes were selected based on the ferroptosis database (http://www.zhounan.org/ferrdb/) and literature review

Patients were then divided into high-or low-risk group according to mean value of risk score. TP53 and KRAS, which were reported to be closely related with ferroptosis, 5 were more frequently mutated in high-risk group (Table  S3). Kaplan Meier (KM)-plot showed that patients in highrisk group correlated with a significantly worse overall survival (OS), no matter in overall population (Figure 3A) or subgroup analysis ( Figure S1). UCA showed that stage, EGFR mutation status, and risk score were prognosis-related ( Figure 3B); multivariate cox analysis Only data with significance that is p < 0.05 were showed here due to space limitation. (G) Transcription factors of ferroptosis-related genes predicted by RcisTarget. From left to right: Recovery curve of the gene-set on the motif ranking; regulating network of TF; motif sequence diagram. More motifs could be seen in Figure S1. (H) Volcano plot of correlation between hub gene expression and drug response AUC value in LUAD cell lines. p < 0.05 and Pearson correlation coefficient > 0.4 were considered significantly correlated, and a negatively correlation means higher expression of this gene was correlated with better response (smaller response AUC value). *p < 0.05, **p < 0.01, ***p < 0.001. Abbreviations: AUC, area under curve; GO, gene ontology; GSVA, gene set variation analysis; KEGG: Kyoto Encyclopedia of Genes and Genomes; LUAD, lung adenocarcinoma; ssGSEA, single sample gene set enrichment analysis; TF, transcription factor (MCA) revealed that stage and risk score were independent predictive factors for OS ( Figure 3C). Time dependent receiver operating curves (tROCs) showed that risk score has a maximum area under curve (AUC) of 0.7439144 at 1-year ( Figure 3D).
These results were then validated in the external cohort, KM-plot showed a consisted result as discovery cohort ( Figure 3E), UCA indicated stage, KRAS status, and risk score were prognosis-related ( Figure 3F), stage and risk score were then proved to be independent risk factors for F I G U R E 4 Continued OS by MCA, and tROCs showed a max AUC of 0.7176205 for risk score at 1-year ( Figure 3G).
To further investigate underlying mechanism of this prognosis relevance, functional analysis was implemented.
Differential-expressed genes (DEGs) bewteen high-and low-risk patients were identified ( Figure 4A), and further KEGG analysis with DEGs mainly enriched in metabolic pathways, for instance steroid hormone biosynthesis, fat digestion, and absorption ( Figure 4B). GO enrichment mainly enriched in ion transportation and leukotriene metabolism pathways ( Figure 4C). Immuno-infiltration status deconvolution by single sample gene set enrichment analysis (ssGSEA) (Table S4) showed high-risk patients were associated with a decreased score of type I and II T helper cell, immature and activated dendritic cell, and effector memory CD8 T cell (Figures 4D and 4E); gene set variation analysis showed decreased score of antigen processing and presentation pathway in high-risk patients, which is consistent with ssGSEA results (Figure 4F), and pathways related to glutathione metabolism, steroid biosynthesis, sphingolipid metabolism, and chemokine signaling were also varied in two groups.
RcisTarget, 6 an R package usually used in single-cell transcriptome data, was then applied on 274 ferroptosisrelated genes to identify potential transcription factors (TFs). A total of 74 potential TFs were identified, and five TFs with maximum AUCs were annotated as Erythroid Derived 2 Like Protein 2, Nuclear Factor-Erythroid 2, MAF BZIP Transcription Factor K, Activating Transcription Factor 4, and CCAAT Enhancer Binding Protein Gamma, respectively ( Figures 4G and S3). Then we carried out integrated analysis of transcriptome data from cancer cell line encyclopedia and drug response data (response AUC value) from cancer therapeutics response portal respectively to identify potential candidate drugs for risky genes in our signature ( Figure 4H). Interestingly, Tozasertib (an Aurora Kinase inhibitor) and PHA-665752 (a c-MET inhibitor) were found strongly correlated with both GCLC and SLC3A2 expression, as well as VDAC2, although not statistically significant (p > 0.05).
In conclusion, novel ferroptosis-related prognosisrelated molecules for LUAD have been identified, a novel eight-gene signature was constructed and validated, prove to have good capacity in predicting OS in LUAD. Functional analysis showed that immune-related pathways may be involved in the regulation of ferroptosis. Potential TFs regulating ferroptosis and several candidate drugs targeted the risk gene were identified. Our findings could provide hints for further study and warrant further research on ferroptosis as a functional and therapeutic target in LUAD.

ETHICS APPROVAL AND CONSENT TO PARTICIPATE
All the data involved were publicly available, so the ethics approval was waived. The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.