Background
Lung adenocarcinoma (LUAD) is one of the most common malignant tumors in the world, demonstrating a rising trend in recent years [
1]. Due to the high recurrence and metastasis, traditional treatments, such as surgery, radiotherapy, and chemotherapy, could not meet all LUAD patients’ needs. Although immunotherapy has been shown to improve survival in LUAD patients, the 5-year overall survival rate is only 23% [
2]. The pathogenic mechanism of LUAD should be further elucidated to discover a new effective treatment strategy.
The tumor heterogeneity, including immune microenvironment and tumor mutation burden, could affect immunotherapy effectiveness. Ferroptosis is also involved in T cell immunity and cancer immunotherapy. The increased ferroptosis contributes to the anti-tumor efficacy of immunotherapy [
3].
Ferroptosis is an iron-dependent form of regulated cell death that is characterized by the excess reactive oxygen species (ROS) generation and lethal accumulation of lipid peroxidation [
4‐
6]. Ferroptosis has been implicated in multiple physiological and pathological processes, including cancer cell death and T-cell immunity [
7]. Autophagy-dependent ferroptosis is featured by excessive autophagy and lysosome activity [
8]. The influence of ferroptosis, especially autophagy-dependent ferroptosis, on the tumor microenvironment needs further study.
The iron metabolism and homeostasis could be influenced by immune cells and related molecules [
9]. Immune cells in the microenvironment play crucial roles in maintaining iron metabolism balance [
10]. The excessive activation of ferroptosis in tumor cells can lead to exposure to tumor antigens, which activate the immune system. Then, the immunogenicity of the microenvironment was improved, and the effectiveness of immunotherapy was enhanced [
11]. Immunotherapy can activate CD8 + T cells to enhance the lipid peroxidation in tumor cells, which further increases ferroptosis in turn [
3]. Therefore, targeting ferroptosis to improve the effectiveness of cancer immunotherapy might become a prospective strategy. In the clinical applications of immunotherapy, tumor mutation burden (TMB) is emphasized as an emerging feature and a biomarker of immunotherapy response [
12,
13]. TMB is defined as the total number of somatic, coding, base substitution, and indel mutations per megabase of genome examined [
14]. Each of these mutations results in the generation of one protein that is a new antigen and could be recognized by the immune system [
15]. Highly mutated tumors are more likely to carry neoantigens, making them become the targets for activated immune cells [
14].
In this study, we comprehensively analyze the genome of LUAD, identify autophagy-dependent ferroptosis-related genes closely associated with the prognosis and chemotherapy sensitivity, further construct and validate the predictive model of the key gene, and explore the relationship with immune infiltration and tumor mutation. Our findings may help generate personalized treatment and improve the clinical outcomes of LUAD patients.
Materials and methods
Workflow
A multi-step approach was used to identify and analyze the autophagy-dependent ferroptosis-related key gene in LUAD. The transcriptome and clinical information were downloaded from The Cancer Genome Atlas (TCGA) project and Gene Expression Omnibus (GEO) data. Autophagy-dependent ferroptosis-related genes were screened by the published articles. Differentially expressed genes (DEGs) related to autophagy-dependent ferroptosis were identified. Univariate and multivariate Cox analyses were applied to screen out the independent prognosis genes related to overall survival (OS). The key gene was identified by the intersection of the DEGs and the prognostic genes. The LUAD patients were classified into the high-risk and low-risk groups based on the key gene expression level. Kaplan-Meier (K-M) analysis and receiver operating characteristic (ROC) curve were conducted to analyze the survival prognosis of patients in TCGA and GEO cohorts. Chemotherapy sensitivity was predicted between different risk groups. Gene ontology (GO) and Kyoto encyclopedia of genes and genomes (KEGG) were conducted to investigate the potential bio-function of the key gene. ImmuneScore was calculated using the Tumor Immune Estimation Resource (TIMER) algorithm, and the TMB was counted as the total number of mutations per megabyte of tumor tissue.
LUAD patients dataset processing
All the RNA-Seq data were normalized as fragments per kilobase of transcript per million mapped reads. mRNAs ensemble gene identities were derived from the HUGO Gene Nomenclature Committee (HGNC) database. The corresponding clinical information includes age, gender, tumor grade, lymph node metastasis, AJCC TNM stages, and survival outcomes. Patients with insufficient clinical data were excluded. OS was estimated as the primary endpoint.
Construction and validation of an autophagy-dependent ferroptosis-related gene signature
Autophagy-dependent ferroptosis-related genes were retrieved from the literature published before January 2021. After combining the related mRNA expression and the clinical data, the gene expression files were obtained. The DEGs between LUAD and normal lung tissues were identified with a false discovery rate (FDR) < 0.05 in the TCGA cohort. Univariate and multivariate Cox analyses of OS were performed to screen the genes with prognostic values in TCGA-LUAD cohort. The key gene was identified by the intersection of the DEGs and the prognostic genes in the TCGA cohort. The cut-off score was defined as the median expression level of the key gene in the LUAD cohort. Patients were stratified into high-risk and low-risk groups based on the cut-off score. To choose appropriate matching cohorts to perform survival analysis before selecting prognostic-related genes, we performed propensity score matching to reduce the selection bias between the high- and low-risk groups. Propensity scores were estimated using age, gender (male versus female), TNM stage (I, II, III, IV), Tumor stage (T1, T2, T3, T4), Lymph Node stage (N0, N1, N2, N3) and Metastasis stage (M0, M1) in TCGA-LUAD cohort. In the same way, propensity scores were estimated using age, gender (male versus female), TNM stage (I, II, III, IV), and
TP53 (Wild versus Mutant) in the GEO cohort. The prognostic value and the clinical correlation of the key gene were both validated between the high- and low-risk groups in TCGA and GEO cohort (GSE116959). The time-dependent ROC curve analyses were conducted to evaluate the predictive power of the key gene. The mRNA expression level of the key gene in various types of cancers was identified in the Oncomine database [
16]. The mRNA and protein expression of the key gene in LUAD were determined using the Gene Expression Profiling Interactive Analysis (GEPIA) and The Human Protein Atlas (HPA) database [
17,
18]. To verify the correlation between ferroptosis and LUAD outcome, we also analyze the survival value of
GPX4 in the LUAD cohort, which is the master regulator of ferroptosis [
19].
Chemotherapeutic response prediction
We analyzed the commonly used chemotherapy drugs, including pemetrexed, cisplatin, gemcitabine, paclitaxel, vinorelbine, docetaxel, doxorubicin, etoposide, erlotinib, and gefitinib [
20]. The chemotherapeutic response prediction was made based on the TCGA-LUAD cohort using the “pRRophetic” R package [
21]. The half maximal inhibitory concentration (IC50) of patients in different risk groups were compared.
Functional enrichment analysis
The biological functions and pathways of the key gene were elucidated through the DEGs between the high-risk and low-risk groups. GO enrichment and KEGG pathway analyses [
22] were then assessed in DAVID database. The correlation analysis of the key gene with tumor proliferation and cell cycle markers was conducted in GEPIA database [
17].
Correlation between the key gene and tumor immune cell infiltration
The enrichment levels of immune cells were quantified by the Tumor Purity, Estimate Score, Immune Score, and Stromal Score in each sample. The tumor immune cell infiltration was calculated by Single Sample Gene Set Enrichment Analysis (ssGSEA). Then we analyzed the correlation between the key gene expression and the abundance of infiltrating immune cells (B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells) via The Tumor IMmune Estimation Resource (TIMER) database [
23].
Analyses of somatic mutations and TMB estimation
The somatic mutation profiles of LUAD patients were downloaded from TCGA database. The mutation frequency with the number of variants/the length of exons (38 million) were calculated for each sample. The OncoPlot of the top 10 mutated genes was plotted. The detailed mutational information, including the variant classification, the number of variant type, and the single-nucleotide variant (SNV) class, were displayed. Then we assessed the correlation between the key gene expression and the TMB levels.
Quantitative real-time PCR and immunohistochemistry
The mRNA and protein expression of the key gene in LUAD were determined using quantitative real-time PCR (qRT-PCR) and immunohistochemistry. The clinical tissue samples of LUAD were obtained from patients who received surgery in Thoracic Oncology Department of Sun Yat-sen University Cancer Center, which was approved by the Institutional Review Committee of Sun Yat-sen University Cancer Center. The detailed procedure was performed according to strict adherence to the manufacturers’ instructions.
For qRT-PCR, 24 paired LUAD and normal tissues were resected and stored in RNAlater immediately. Total RNA was extracted using TRIzol reagent. cDNA was synthesized from total RNA using cDNA reverse transcription kit (Thermo Fisher Scientific). qRT-PCR was performed using the SYBR Green PCR kit (Thermo Fisher Scientific). The housekeeping gene GAPDH was used as an endogenous control. Primer information: FANCD2: 5′- AAAACGGGAGAGAGTCAGAATCA-3′ (forward) and 5′- ACGCTCACAAGACAAAAGGCA-3′ (reverse); GAPDH: 5′- GGAGCGAGATCCCTCCAAAAT-3′ (forward) and 5′- GGCTGTTGTCATACTTCTCATGG-3′ (reverse). The cycle threshold (Ct) of each gene in samples was recorded. Relative quantification was calculated as 2-ΔCt (ΔCt values = target gene mean Ct value - control gene mean Ct value).
Twenty pairs of LUAD immunohistochemistry samples were fixed using 10% formalin and embedded in paraffin. Immunohistochemistry was carried out using the processed 5 μm continuous sections. Samples were dewaxed with decreasing concentrations of 100, 95, 75, and 50% ethanol and washed in deionized water. The sections were heated in a microwave with TE buffer pH 9.0 to retrieve antigens. Endogenous peroxidase was inhibited by incubation in goat serum. Then they were incubated in rabbit anti-FANCD2 (Proteintech, 204006–1-AP, 1:200) overnight at 4 °C. Next, the sections were incubated with horseradish peroxidase-coupled goat anti-rabbit secondary antibody and stained using DAB Detection Kit (Polymer). The following process is cell nucleus staining, dehydration, xylene infusion, and mounting [
24]. The immunohistochemical scores were scored by two independent pathologists. The intensity of FANCD2 expression was scored as zero, negative; one point, weak staining; two points, mild staining; three points, strong staining. The positive stained area percentage (PSAP) of FANCD2 expression was scored as 1, 0–25%; 2, 25–50%; 3, 50–75% and 4, 75–100%. FANCD2 IHC score = Intensity score × PSAP score.
Statistical analysis and R software packages
Significance analysis of microarrays was used to screen the differentially expressed genes between the LUAD and normal lung tissues. Univariate Cox proportional hazards model was used to analyze the association between gene expression level and prognosis. Kaplan-Meier method and Log-rank test were used to evaluate the difference between survival curves. The continuous and categorical variables between the two risk subtypes were compared using the two-sided Wilcoxon rank-sum test and chi-square test, respectively. Benjamini-Hochberg method was used to adjust for multiple hypothesis testing. All P values were 2-sided, and P < 0.05 was considered statistically significant.
Statistical analyses and result visualization were performed using R software v3.6.3, v4.0.5 and v4.1.2 (“pheatmap v4.0.5”, “limma v3.6.3 [
25]”, “survival v3.6.3”, “survminer v3.6.3”, “ggpubr v3.6.3”, “survivalROC v3.6.3”, “car v3.6.3”, “ggridges v3.6.3”, “genefilter v4.1.2”, “ggpubr v3.6.3”, “pRRophetic v4.1.2”, “ggplot2 v3.6.3”, “colorspace v4.0.5”, “stringi v4.0.5”, “clusterProfiler v4.1.2 [
26]”, “enrichplot v4.1.2” and “maftools v4.1.2” R package).
Discussion
LUAD is a common malignancy with high morbidity and mortality [
1]. The development of LUAD often involves genetic abnormalities and immune dysfunction [
32]. Iron metabolism could influence malignant biological behaviors and impact the tumor microenvironment [
33]. The increase of labile iron in cancer cells can facilitate DNA replication [
34] and induce the occurrence of ferroptosis to participate in and accelerate tumor progression [
4].
Ferroptosis is a programmed cell death in which multiple signaling molecules interact in the tumor microenvironment and synergistically regulate tumor progression [
35]. Ferroptosis has a dual role in tumor promotion and suppression [
36]. On the one hand, the induced tumor cell ferroptosis inhibit tumor metastases, is involved in drug resistance, and influences cancer immunotherapeutic efficacy [
37,
38] . On the other hand, ferroptotic damage could contribute to inflammation-related immunosuppression within the tumor microenvironment and promote tumors’ growth [
36,
39].
The role of ferroptosis in LUAD has not been elaborated. Our research provides a new perspective for the development of LUAD. Ferroptosis was once considered a novel cell death process distinct from apoptosis, necrosis, and autophagy [
4]. However, studies from autophagy-deficient cells suggested that ferroptosis was a type of autophagy-dependent cell death in some conditions [
8]. Autophagy, including ferritinophagy [
40,
41], lipophagy [
28], clockophagy [
42,
43], and chaperone-mediated autophagy [
42], could promote ferroptosis through lipid peroxidation. As we know,
GPX4 is an essential regulator of ferroptotic cancer cell death [
44].
GPX4 is closely related to the tumor stage and promotes acquired chemoresistance by suppressing ferroptosis [
45]. GPX4 inhibitor could augment the anticancer effect of platinum drugs in lung cancer brain metastasis. And GPX4 inhibition synergizes with radiation to induce ferroptosis in LUAD by enhancing cytoplasmic lipid peroxidation [
46]. Our study shows that the high expression of
GPX4 is related to a better prognosis in LUAD patients. And the functional analysis of
FANCD2 is mainly enriched in response to oxidative stress and ROS, which indicated the role of ferroptosis in LUAD.
FANCD2 (FA complementation group D2) contributes heterogeneously to Fanconi anemia (FA), a genetic disorder characterized by birth defects, progressive bone marrow failure, and cancer-prone phenotype [
47]. The patients with aberrant expression of FANCD2 possess abnormality in chromosomal breakage and hypersensitivity to DNA crosslinking agents [
48]. As a DNA damage response regulator, FANCD2 also regulates ferroptosis sensitivity by inhibiting iron accumulation and lipid peroxidation in an autophagy-independent manner [
49,
50]. FANCD2 has an intricate relationship with tumors. The heterozygous and somatic mutations of
FANCD2 were reported in various malignancies, including pancreatic cancers and squamous cell carcinomas [
51,
52]. The overexpression of
FANCD2 was involved in metastasis-prone melanomas [
53] and colorectal cancer [
54]. In our study,
FANCD2 was identified as an autophagy-dependent ferroptosis-related key gene in the LUAD occurrence after comprehensive analysis. However, the exact mechanism of how
FANCD2 influences LUAD outcome is complex, which probably includes more than its role in ferroptosis, DNA damage and cell cycle.
Immune cells could regulate tumor ferroptosis during cancer immunotherapy [
3]. Besides, ferroptosis also could regulate immunity activity within the tumor microenvironment [
39]. The potential connection between the behavior of immune cells in the tumor microenvironment and ferroptosis needs to be further studied. In our study, the high expression of
FANCD2 group achieved a high fraction of neutrophil, which revealed that the ferroptosis-related gene
FANCD2 might be closely associated with neutrophil-mediated tumor immunity.
Following the above findings, the antigen processing and presentation contents were enriched by KEGG analyses. In adaptive immunity, neutrophils play a significant role in internalizing antigen and regulating antigen-specific responses [
55]. When ferroptosis occurred, the dead cells released the immunogenic signals, such as lipid mediators. Subsequently, the antigen-presenting cells, including neutrophils, were attracted to the site of ferroptotical cells [
39]. A multitude of recruited neutrophils further activated the immune system to resist the invasion of pathogenic factors. Abnormal and uncontrolled ferroptosis may be implicated with invalid immunity [
39]. Our research indicated that high expression of
FANCD2 induced aberrant ferroptosis and further contributed to the abnormality of anti-tumor immunity in patients with LUAD.
TMB level has demonstrated utility in selecting patients for response to immunotherapy and has proven to be an essential biomarker for patient selection. Patients in high TMB benefit more from immunotherapy, which provided a new avenue to make LUAD treatment more precise [
56]. In our study, the LUAD patients with a high expression level of
FANCD2 achieved a high TMB, indicating that these patients may gain more benefit from immunotherapy than those with a low
FANCD2 expression level. Among the mutation genes, tumor suppressor gene inactivation, such as
P53, is very common in LUAD [
57]. P53 activation has been explored to be essential in some other activities to suppress tumor progression [
58,
59], whereas the anti-P53 activity traditionally drives cell senescence, cell cycle arrest, and apoptosis [
60].
Additionally,
P53 was correlated with ferroptosis, and it could inhibit cysteine uptake and sensitize cells to ferroptosis. Studies revealed that the sensitivity of ROS-induced ferroptosis was markedly increased in P53-activated cells [
61]. Our study found that
P53 is the most frequently mutated gene and positively correlated with a higher
FANCD2 expression level, which indicated that the
P53 mutation might activate the FANCD2-mediated ferroptosis and increase the response of immunotherapy in LUAD.
However, there are several limitations of our study. The role of ferroptosis in LUAD outcome and the role of FANCD2 in LUAD ferroptosis have not been fully clarified. The function of FANCD2 may include tumor ferroptosis and DNA damage, and cell cycle, which influence the outcome of LUAD together.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.