Bioinformatics Identification and Validation of Ferroptosis-Related Key Genes and Therapeutic Compounds in Septic Lung Injury

Introduction

Sepsis refers to lethal organ dysfunction caused by a dysregulated host response to infection, and is a major cause of morbidity and mortality in ICU patients.¹ Among the organs involved, the lungs are the most susceptible to infection,² and septic patients are prone to developing acute lung injury (ALI) or acute respiratory distress syndrome (ARDS), which leads to a high mortality rate of > 30%.³ Therefore, it is important to explore innovative mechanisms and pharmacological interventions for septic lung injury (SLI) to improve patient survival.

Regulated cell death (RCD) includes pyroptosis, apoptosis, ferroptosis, and regulated necrosis (necroptosis), all of which are involved in sepsis and organ dysfunction.⁴ Among these, ferroptosis, a unique form of programmed cell death characterized by abnormal iron metabolism and excessive lipid peroxidation,⁵ has emerged as an academic point of interest among researchers. Recent studies have confirmed its contribution to a range of diseases, including cancer, ischemic diseases, infections, and immune system disorders.^5–7 Recent studies have explored the role of ferroptosis in ALI development of acute lung injury. For instance, ferroptosis has been identified as a crucial form of cell death in a lipopolysaccharide (LPS)-induced lung injury model,⁸ and numerous ferroptosis-related key molecules and pathways have been identified in ALI caused by various factors.⁹ Notable work by Zhang et al explored the regulatory mechanisms of neutrophils and ferroptosis in SLI.¹⁰ Moreover, specific proteins such as MUC1 and AUF1 have been identified as crucial for mitigating ferroptosis-related SLI.^11,12 Relevant therapeutic targets are also being explored and studied. Esketamine was found to alleviate ferroptosis-mediated lipid peroxidation via upregulation of HIF-1α/HO-1 pathway in acute lung injury.¹³ Similarly, another study on acute lung injury revealed that sodium houttuyniae, an herbaceous plant, mitigated ferroptosis by through TRAF6-c-Myc signaling pathway.¹⁴ In addition, the activation of AMPK/Nrf-2 signaling pathway by upregulation of GPX4 (glutathione peroxidase 4) and SLC7A11 (solute carrier family 7 member 11) could inhibit ferroptosis in lung ischemia-reperfusion injury.¹⁵

Phloretin, a hydrolysis product of phlorizin derived from apple plants, exhibits numerous pharmacological properties including anticancer, antibacterial, and anti-inflammatory effects.¹⁶ Crucially, both in vivo and in vitro experiments have demonstrated that phloretin alleviates acute lung injury by reducing the production of inflammatory factors, blocking the NF-κB and MAPK pathways, and inhibiting glycolysis in macrophage.¹⁷ However, the specific mechanisms underlying the role of phloretin in attenuation of ferroptosis during SLI remain unclear.

Bioinformatics analysis (BA) uses clustering and statistics to reveal interdependent relationships between molecules, genes, and pathways.¹⁸ Researchers have utilized BA in fields of ischemic stroke, aortic valve disease and osteoporosis^19–21 to explore key genes related to diseases. Interestingly, HMOX1, SLC2A3, IL-6 and HIF-1α are frequently enriched as hub genes in similar studies. Such study approach not only enhances the reliability of the results but also provides new insights for disease diagnosis and treatment. However, studies of ferroptosis-related SLI are lacking. By analyzing the Gene Expression Omnibus (GEO) microarray database and intersecting it with the FerrDB database, we identified ferroptosis-related differentially expressed genes (DEGs) and conducted functional enrichment analyses. After validating the expression of hub genes in other datasets, the Connectivity Map (c-MAP) database was used to identify whether phloretin may affect key DEGs of ferroptosis during SLI. Finally, we established SLI animal models by intraperitoneal administration of LPS, and phloretin was used to verify its therapeutic effects in SLI. This study sheds light on the nascent molecular mechanisms and intervention strategies for ferroptosis-related SLI.

Materials and Methods

Microarray Data

We visited the GEO database (https://www.ncbi.nlm.nih.gov/geo/) and downloaded transcriptome data from SLI-related microarray datasets (GSE130936, GSE60088, and GSE137342). The search terms used were “sepsis-lung”. The GSE130936 dataset was analyzed with the Affymetrix Mouse Expression 430A Array, in which we used the expression profile data in lung samples harvested from mice 22 h after challenge with either saline as a control (n = 4) or LPS (n = 3) as a sepsis model.

Differentially Expressed Gene Analysis

We used GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r/) to compare the gene expression data between the LPS and Control group in the GSE130936 dataset with p_adjusted < 0.05 and |log₂FC|>1 as the threshold. Ferroptosis-related genes (including drivers, suppressors, and markers) were obtained from the FerrDb database (www.zhounan.org/ferrdb/index.html/). Finally, these genes were intersected with the DEGs of the GSE130936 dataset to identify ferroptosis-related DEGs.

Functional Enrichment Analysis

The ClusterProfiler package, org. The Hs.eg.db package and the Gene ontology GO plot package were loaded into R software to perform Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses on ferroptosis-related DEGs (p < 0.05).

PPI Network Analysis

Ferroptosis-related DEGs were uploaded to the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (www.string-db.org/) to predict the PPI networks (interaction score > 0.4). The results were then imported into Cytoscape software v.3.9.1, and the Molecular Complex Detection (MCODE) plugin was used to perform clustering analyses of the ferroptosis-related DEGs (Degree Cutoff = 2, Node Score Cutoff = 0.2, K-Core = 2, and Max. Depth = 100).

Molecular Docking

To analyze the binding affinities and modes of interaction between the drug candidate and their targets, AutodockVina 1.2.2, a silico protein–ligand docking software was employed.²² The molecular structures of phloretin was retrieved from PubChem Compound (https://pubchem.ncbi.nlm.nih.gov/).²³ The 3D coordinates of IL-6, TIMP1, HIF-1α and HMOX1 were downloaded from the PDB (http://www.rcsb.org/pdb/home/home.do). Molecular docking studies were performed by Autodock Vina 1.2.2 (http://autodock.scripps.edu/).

Animal Experiments and Grouping

Male Institute of Cancer Research (ICR) mice (weighing 25–30 g) were used in this study. Animals were kept in an environmentally controlled room (temperature, 24 ± 2°C; humidity, 55 ± 10%) with a 12 h light/dark cycle and free access to food and water. All experimental procedures were approved by the Animal Ethics Committee of Xinhua Hospital, Shanghai Jiao Tong University School of Medicine. Eighteen mice were randomly assigned to three groups: control (Con), LPS, and LPS + phloretin (LPS + PHL), each comprising six mice. Phloretin (CAS No. 60-82-2, MCE) was dissolved in dimethyl sulfoxide (DMSO) to a concentration of 2.5 mg/mL. Mice from the LPS + Ph group were injected intraperitoneally with the prepared phloretin (20 mg/kg) solution seven days in a row before LPS treatment. The phloretin solution was replaced with DMSO in the control and LPS groups. LPS (L2630, Sigma-Aldrich, St. Louis, MO, United States, 5 mg/kg) was intraperitoneally injected after 7 days of PHL pretreatment and the mice were sacrificed 24 h later. Animal studies were performed in accordance with the Guide for the Care and Use of Laboratory Animals published by the NIH (NIH publication No.85–23, revised 1996).

ELISA Analysis

24 hours after LPS treatment, blood samples were collected from the mice by cardiac puncture, centrifuged at 3000 rpm for 5 min, and the supernatant was stored at −80°C before use. Serum concentrations of TNF-α (Cusabio Biotech, Wuhan, China) and IL-6 (Cusabio Biotech) were measured using ELISA kits according to the manufacturer’s instructions. All samples were run in triplicate.

Histology

The left lower lobes of the lung specimens were fixed with 10% formalin, embedded in paraffin, and cut into 4 μm sections. The lung tissue was stained with hematoxylin and eosin, and histopathological changes were visualized using light microscopy (Olympus, Tokyo, Japan). The degree of lung injury was analyzed by two pathologists who were blinded to the study based on a scoring system described in a previous study.²⁴

Immunofluorescence of 8-OH-dG

Paraffin sections (4 μm) of lung tissues were rehydrated and microwaved to retrieve antigens. After incubation with 5% BSA for 1 h, the sections were incubated with primary antibodies against 8-hydroxy-2′-deoxyguanosine (8-OHdG) (Abcam, 1:200, Cambridge, MA, USA) overnight. Subsequently, the sections were incubated with secondary antibodies (Alexa Fluor 488: goat anti-mouse, 1:400, Servicebio, Wuhan, China) in the dark for 1 h and counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (Beyotime). The stained slices were observed under a fluorescence microscope (Nikon, Tokyo, Japan).

TdT-Mediated dUTP Nick-End Labeling (TUNEL) Assay

Lung samples used for immunofluorescence analysis were used for the TUNEL assay. Apoptotic staining was performed according to the manufacturer’s protocol (Beyotime). DAPI staining was performed to determine the number of nuclei present. TUNEL signals were observed under a fluorescence microscope (Nikon, Tokyo, Japan).

Malondialdehyde Detection

A lipid peroxidation malondialdehyde assay kit (Beyotime) was used to measure the MDA content in the lungs, following the manufacturer’s instructions. The absorbance was measured at 532 nm (n = 6 per group).

Detection of Glutathione Disulfide

Oxidized GSSG is a common marker for measuring levels of oxidative stress. The GSSG levels in the lungs were quantified using a detection kit (Beyotime). Absorbance was measured at 412 nm (n = 6 per group).

Western Blotting

Proteins were extracted from the lung tissues using a mixture of cold RIPA lysis buffer (Beyotime). Equal amounts of samples (40 μg) were resolved by electrophoresis on 10% or 15% SDS sulfate-polyacrylamide gels, and the proteins were subsequently transferred onto PVDF membranes (Millipore, Billerica, MA, USA). Primary antibodies against GPX4 (Abcam, 1:1000), VCAM-1 (Servicebio, 1:1000, Wuhan, China), VE-cadherin (Abcam, 1:1000), and β-actin (Sigma-Aldrich, 1:3000, Mo, USA) were used, followed by incubation with horseradish peroxidase-linked secondary antibodies (ProteinTech Group, Inc., HRP-goat anti-mouse, 1:5000; HRP-Goat Anti-Rabbit,1:5000, Wuhan, China). All samples were run in triplicate.

Extraction of RNA and Quantitative RT-PCR (qRT-PCR)

RNAiso Plus reagent (Takara Biotechnology, Dalian, China) was used to extract RNA from lung tissues. Complementary DNA was synthesized according to the manufacturer’s protocol. qRT-PCR was performed using ChamQ Universal SYBR qPCR Master Mix (Vazyme Biotech, Nanjing, China). Target gene mRNA levels were normalized to endogenous β-actin expression and quantified using the 2^−ΔΔct method. All samples were run in triplicate. The primer sequences used were as follows:

β-actin-F: 5′-CTGTATGCCTCTGGTCGTAC-3′

β-actin-R: 5′-TGATGTCACGCACGATTTCC-3′

HMOX1-F: 5′-GCCCCACCAAGTTCAAACAG-3′

HMOX1-R: 5′-GCTCCTCAAACAGCTCAATGT-3′

HIF-1α-F: 5′- CTGCCACTGCCACCACAACTG-3′

HIF-1α-R: 5′- TGCCACTGTATGCTGATGCCTTAG-3′

TIMP1-F: 5′- GCATCTCTGGCATCTGGCATCC-3′

TIMP1-R: 5′- CGCTGGTATAAGGTGGTCTCGTTG-3′

IL6-F: 5′- CGTGGAAATGAGAAAAGAGTTGTGC-3′

IL6-R: 5′- GGTACTCCAGAAGACCAGAGGA-3′

Statistical Analyses

Statistical analyses were performed using SPSS 26.0 (IBM, Armonk, NY, United States), Prism 7 (GraphPad, San Diego, CA, United States), and R. Data are displayed as the mean ± SEM. Statistical significance in experiments comparing only two groups was determined using two-tailed Student’s t-test. One-way ANOVA was used for comparisons between multiple groups, followed by post hoc analysis using the Student-Newman-Keuls test. Statistical significance was set at p < 0.05.

Result

Identification of Ferroptosis-Related Differentially Expressed Genes in Sepsis-Induced Lung Injury Models

Expression data between the LPS and Control groups in the GSE130936 were analyzed using GEO2R. The cut-off criteria for DEGs dataset were set as |log₂FC| >1 and p_adjusted < 0.05. Using these criteria, 460 genes were upregulated, and 525 genes were identified as downregulated, and these genes are displayed in a volcano plot (Figure 1A). Then, to explore the correlation between ferroptosis and SLI, 340 ferroptosis-related genes obtained from the FerrDb database were intersected with sepsis-associated DEGs, and 38 ferroptosis-related DEGs were obtained and are shown in a Venn diagram and heat map (Figure 1B and 1C), including 26 up-regulated and 12 down-regulated genes (Table 1).

Figure 1 Identification of ferroptosis-related DEGs. (A) Volcano plots of DEGs. (B) Venn diagram of DEGs from GSE130936 and ferroptosis-related genes. (C) Heatmap of DEGs in GSE130936.

Table 1 Ferroptosis-Related Differentially Expressed Genes of Sepsis

Functional Enrichment Analyses and PPI Networks for Ferroptosis-Related Differentially Expressed Genes

In order to gain a better understanding of the functions of these genes, GO and KEGG analyses of 38 ferroptosis-related DEGs (p < 0.05) were performed using R software. As shown in Figure 2A and 2B, the response to hypoxemia was significantly enriched in the biological process category. Ferroptosis, autophagy, and HIF-1 signaling pathways were ranked among the top 20 KEGG pathways. Subsequently, ferroptosis-related DEGs were further analyzed to construct a PPI network (Figure 3). The Cytoscape software and MCODE were used to identify the most significant cluster (cluster 1, containing eight genes). Another GO and KEGG analysis were conducted on these eight genes (Table 2), and we found that four out of the eight genes in cluster 1 (IL-6, HMOX1, HIF-1α, and TIMP1) were enriched in the HIF-1 signaling pathway. These genes were then selected as candidate genes for further analysis and were identified as the key ferroptosis-related genes involved in sepsis-induced lung injury.

Table 2 Functional Enrichment Analysis of Genes in Cluster 1

Figure 2 Enrichment analyses of ferroptosis-related DEGs. (A) Functions of ferroptosis-related DEGs with GO enrichment analysis. (B) Signaling pathways of ferroptosis-related DEGs with KEGG enrichment analysis. Notes: Full terms for (A) iconnective tissue replacement involved in inflammatory response wound healing; iiregulation of transcription from RNA polymerase II promoter in response to oxidative stress; iiivascular endothelial growth factor production; ivpositive regulation of vascular endothelial growth factor production; vsuperoxide-generating NAD(P)H oxidase activity. Full terms for (B) iAGE-RAGE signaling pathway in diabetic complications; iiPathways of neurodegeneration - multiple diseases; iiiAlanine, aspartate and glutamate metabolism.

Figure 3 PPI analysis of ferroptosis-related DEGs. (A) PPI network of ferroptosis-related DEGs. (B) Gene cluster 1 based on the MCODE algorithm. (C) Gene cluster 2 based on the MCODE algorithm.

Verification of the Key Ferroptosis-Related Genes in Sepsis-Induced Lung Injury

Next, in order to observe whether the above four candidate ferroptosis-associated genes exhibits the same trend of changes in previous related experiments, we validated them by selecting GSE60088 and GSE137342 as the validation datasets. The mRNA levels of IL-6, TIMP1, HIF-1α, and HMOX1 were increased significantly in the lung tissues of mice induced by sepsis (Figure 4A) and in the plasma of septic patients (Figure 4B) compared with the control group.

Figure 4 Verification of the four candidate genes (IL-6, TIMP1, HIF-1α, HMOX1) (A) in septic and control mice lung samples in GSE60088 dataset (B) and peripheral blood of healthy and septic patients in GSE137342. ***p < 0.001, **p < 0.01, *p < 0.05.

Phloretin Protects Against LPS-Induced Lung Injury

We then searched the c-MAP database to identify potential small molecules or drugs that might affect DEGs related to ferroptosis during SLI (Table 3). Phloretin, a natural product derived from apple plants, was selected as a potential compound to alleviate SLI because of its excellent anti-inflammatory effects introduced above. Simultaneously, its target SLC23A1 is highly expressed in lung,²⁵ while AQP9’s important pathophysiological role in sepsis has been identified,²⁶ which lead us to believe that PHL is a more reliable choice compared to other drugs filtered. To evaluate the affinity of the candidate substance for their target proteins, we performed molecular docking analysis. The binding poses and interactions of phloretin with IL-6, TIMP1, HIF-1α and HMOX1 were obtained with Autodock Vina v.1.2.2 and shown in Figure 5. As shown in Figure 6A–6D, qRT-PCR confirmed that IL-6, TIMP1, HIF-1α, and HMOX1 mRNA expression was upregulated in mouse lung tissues challenged with LPS compared to control mice, and PHL significantly inhibited the expression of four key genes. Histologically, LPS treatment resulted in edema, thickening of the alveolar wall, and neutrophil infiltration in the lung parenchyma; however, these effects were markedly inhibited by PHL treatment (Figure 6E and 6F). Existing literature has indicated that Vascular Endothelial cadherin (VE-cadherin) and Vascular Cellular Adhesion Molecule-1 (VCAM-1) are key in endothelial and epithelial pulmonary injury,^27,28 so we also included them in the observation indicators. PHL increased VE-cadherin and decreased VCAM-1 induced by LPS (Figure 6G and 6H). Serum levels of inflammatory factors IL-6 and TNF-α were also alleviated in PHL-treated mice challenged with LPS compared to those in the LPS group (Figure 6I and 6J).

Table 3 Potential Small Molecules or Drugs That May Have Effect on Ferroptosis-Related DEGs Enriched by c-MAP

Figure 5 Binding mode of phloretin to their targets by molecular docking. (A–D) Docking mode of phloretin to IL-6, TIMP-1, HIF-1α, HMOX1. The crystal structure of phloretin is marked within the red square in each figure.

Figure 6 Therapeutic effects of phloretin on SLI mice. (A–D) Relative mRNA levels of IL-6, TIMP-1, HIF-1α, and HMOX1 in lung tissues of mice by qRT-PCR. (E) H&E staining of mice lung tissues (scale bar is 50 μm). (F) Lung injury scores. (G and H) The expression levels of VE-cadherin and VCAM-1 in lung tissues were evaluated semi-quantitatively by Western blot analysis. Representative bands (G) and corresponding histogram (H) were shown. (I and J) IL-6 and TNF-α level in serum. Data are represented as mean ± SEM (n=6 per group). ***p < 0.001, **p < 0.01, *p < 0.05.

Phloretin Alleviated LPS-Stimulated Pulmonary Ferroptosis and Apoptosis

Finally, the effect of PHL on SLI and ferroptosis was evaluated. Western blot analysis of lung tissues revealed markedly higher levels of GPX4 in mice treated with PHL than those in the untreated group subjected to LPS (Figure 7A and 7B). Oxidative stress biomarkers, MDA and GSSG levels, were increased by LPS administration, whereas the results exhibited an obvious reduction in both MDA and GSSG levels following administration of PHL (Figure 7C and 7D). In addition, ferroptosis is closely related to apoptosis and oxidative stress.^29,30 8-OHdG immunoreactivity was measured as an index of membrane lipid peroxidation and oxidative DNA damage. As shown in Figure 7E and 7F, significant increases in 8-OHdG-positive cells were observed in the LPS group compared with those in the control group. PHL significantly alleviated LPS-induced pulmonary oxidative stress, as indicated by the decreased 8-OHdG immunoreactivity in the lung tissues of septic mice. TUNEL, a universal assay for multiple cell death, including ferroptosis, confirmed the increase in lung tissues of LPS-administered mice, whereas the proportion of TUNEL-positive cells decreased in the LPS plus PHL group (Figure 7G and 7H). These findings demonstrated that PHL is a candidate compound that alleviates SLI by inhibiting ferroptosis through the regulation of IL-6, TIMP1, HIF-1α, and HMOX1.

Figure 7 Effects of phloretin on LPS-stimulated pulmonary ferroptosis and apoptosis. (A and B) The expression levels of GPX4 in lung tissues was evaluated semi-quantitatively by Western blot analysis. Representative bands (A) and corresponding histogram (B) were shown. (C and D) MDA and GSSG levels were measured in serum of mice. (E and F) Lung tissues were stained with TUNEL (red). Nuclei were counterstained with 4’6-diamidino-2-phenylindole (DAPI) (blue). Percentages of TUNEL+ cells are shown as histograms. (G and H) Lung tissues were stained with 8-OHdG (green), Nuclei were counterstained with DAPI (blue). Percentages of 8-OHdG-positive cells are shown as histograms. Data are represented as mean ± SEM (n=6 per group). ***p < 0.001, **p < 0.01, *p < 0.05.

Discussion

The lung is the most vulnerable organ in the development of sepsis, and almost half of patients with sepsis progress to ALI or ARDS.³¹ Despite the constant developments in clinical diagnosis and treatment strategies, the prognosis of sepsis-associated ALI remains poor.

In recent years, ferroptosis has become a popular topic in the study of various diseases. According to latest studies, critical COVID-19 patients with higher levels of plasma MDA and catalytic iron have higher risk for 90-day mortality³² while research on critical acute pancreatitis and non-alcoholic fatty liver disease patients reveal that the biomarker of ferroptosis at all stages is significantly higher in non-survival patients than that in the survival group.³³ Moreover, a study on multiple organ dysfunction syndrome (MODS) patients showed that the maximum value of MDA and catalytic iron per patient had a significant positive correlation with the sequential organ failure assessment (SOFA) score. MDA values were consistently higher in the deceased group when analyzed everyday,³⁴ indicating that in long-term ferroptosis may increase the risk of death in septic patients. The inhibition of ferroptosis is considered an effective treatment for sepsis. As the concrete mechanisms between ferroptosis and sepsis-induced ALI/ARDS has not been completely elucidated,^35,36 we recognized the urgency of identifying ferroptosis-related biomarkers and efficacious medications to prevent and treat SLI. In the current study, we identified 38 ferroptosis-related genes associated with BA. The HIF-1 signaling pathway was one of the most significantly enriched pathways. HIF-1 is an oxygen-regulated transcriptional activator commonly the expression of HIF-1α surges in hypoxic cells.³⁷ HIF-1α is also involved in numerous diseases through diverse pathways. Dysregulation of Akt/mTOR/HIF-1 signaling has been observed in SARS-CoV-2 infection, suggesting an alternative therapy.³⁸ HIF-1α has been found to be an upstream regulator of BNIP3 in the hypoxia signaling pathway during myocardial ischemia-reperfusion injury.³⁹ Wu et al demonstrated that ferroptosis occurs in mouse models exposed to di-(2-ethylhexyl) phthalate (DEHP), a plasticizer, via the HIF-1α/HO-1 signaling pathway.⁴⁰ Therefore, we hypothesized that the HIF signaling pathway plays a significant role in SLI-induced ferroptosis. On account of more concrete clustering analysis results, we proposed that IL-6, TIMP1, HIF-1α and HMOX1 were key genes participated in ferroptosis regulation in SLI.

IL-6 is a 26-kD secreted protein relevant to inflammatory diseases.⁴¹ Elevated IL-6 is usually seen in sepsis,⁴² rheumatoid arthritis,⁴³ chronic obstructive pulmonary disease (COPD)⁴⁴ etc. Previous studies have shown that the expression of IL-6 is upregulated upon LPS-induced sepsis, which then amplifies STAT3 activation, leading to a vicious cycle of further IL-6 through TLR4/Mal signaling.⁴⁵ The therapeutic strategy mainly focuses on blocking monoclonal antibodies, and different inhibiting sites where IL-6 blockers act have been exhaustively summarized in the literature.⁴⁶ TIMP1 is a cancer-related chemokine associated with angiogenesis, cell survival, proliferation, and apoptosis.⁴⁷ TIMP1 is favorable for tumor progression in pancreatic cancer, as it promotes perineural invasion and triggers neutrophil extracellular traps (NETs) to promote tumorigenesis.^48,49 On the other hand, prior BA and experimental analyses have identified TIMP1 as an inflammatory biomarker in temporal lobe epilepsy.⁵⁰ Clinical research has also substantiated the connection between elevated serum TIMP1 levels, organ dysfunction, and mortality in septic patients.⁵¹ As an adaptation to hypoxia, HIF-1α increases the production of harmful inflammatory cytokines, including TNF-α, IL-1, and IL-4, and promotes angiogenesis, glycolysis, and erythropoiesis.⁵² HIF-1α is also known to mediate the inflammatory responses during LPS-induced sepsis.⁵³ Interestingly, enhanced HIF-1α expression differs under different conditions during ferroptosis. For instance, in cervical cancer, it leads to ferroptosis resistance,⁵⁴ while overactivation of HIF promotes ferroptosis in myocardial ischemia–reperfusion injury.³⁹ HMOX1 encodes heme oxygenase (HO), a stress-induced enzyme that catalyzes the process of heme degradation and exhibits anti-inflammatory effects in cardiovascular disease; metabolic disorders; and liver, kidney, and lung diseases.⁵⁵ In hind limb ischemia injury mouse models, HMOX1 influences both up-and downstream HIF-1α, and rescues ischemic injury by stabilizing HIF-1α. Similarly, the upregulation of HMOX1 due to excess production of heme is positively related to the degree of cardiac ferroptosis in sickle cell disease.⁵⁶ Such a correlation between HMOX1 and ferroptosis has also been observed in osteosarcoma⁵⁷ and diabetic atherosclerosis,⁵⁸ which indicates that flexible regulation of HMOX1 serves as a novel approach for the treatment of cancer and inflammatory diseases.

We established LPS-induced SLI mouse models to confirm the BA results and found that the expressions of IL-6, TIMP1, HIF-1α, and HMOX1 were significantly upregulated. We then used c-MAP, currently the largest database that connects related genes, drugs, and diseases,⁵⁹ to screen for possible natural medicines and verify its therapeutic effect on SLI mouse models. High-quality literature has summarized small molecules that target ferroptosis in vivo and clinical trials, and it differentiates targeted pathways and proposed mechanisms.⁶⁰ The side effects and off-target effects of conventional medicine or chemical agents are inevitable So more dietary and natural compounds may make up for this deficiency. Phloretin is one of the natural agents listed, and we chose it for its reported anti-inflammatory effects on lung injury.^17,61 Consistently, our study showed that phloretin decreased the expression of the four hub genes and ameliorated lung injury and inflammation in septic mice. VE-cadherin binds to adjacent lung endothelial cells and is important for basal lung microvascular stabilization.⁶² Downregulation of VE-cadherin induced by IL-1β has been shown to be the basis for septic lung barrier impairment.⁶³ In contrast, VCAM-1 is responsible for leukocyte adhesion and extravasation is upregulated during inflammation. In PM2.5-induced lung injury, increased polymorphonuclear leukocytes (PMN) via upregulated VCAM-1 expression leads to exacerbated inflammation and injury.⁴⁶ In our phloretin treatment group, VE-cadherin and VCAM-1 were upregulated and downregulated, respectively, compared to the LPS group, which further confirmed the value of phloretin in SLI treatment. In addition, we verified for the first time that phloretin could improve ferroptosis in SLI. This may lead to wider application of phloretin in other ferroptosis-related diseases. Since ferroptosis is related to oxidative stress and apoptosis,⁶⁴ immunofluorescence of TUNEL and 8-OHdG in lung tissue was performed, and the results fully illustrated the anti-ferroptosis effect of phloretin.

Four ferroptosis-related genes were identified to play important roles in SLI through BA and have been verified in other GEO datasets and animal models. In addition, we discovered a novel natural therapeutic compound by using the c-MAP tool. Identifying specific genes that are involved in ferroptosis and linking them to potential therapeutic compounds are novel aspects of our research. The use of a bioinformatics approach followed by experimental validation bridges computational predictions with practical, therapeutic insights. However, our study had several limitations. First, the experimental validation of human lung tissue is hindered by the difficulty in obtaining samples from patients with sepsis. Second, among the 38 ferroptosis-related genes, we focused only on four genes, leaving other genes to be explored comprehensively in the future. Third, we only adopted a single phloretin dosage in the study, and did not explore a dose-response relationship, which is crucial for therapeutic window and safety profile of phloretin. Thus, our results require further verification with more clinical data and tests and more in-depth and detailed exploration.

Conclusion

In conclusion, we identified key ferroptosis-related genes through a combination of bioinformatics analysis and experimental validation. Additionally, the natural compound phloretin significantly suppresses the expression of ferroptosis-related key genes, alleviates lung injury and inflammatory responses in LPS-induced SLI. These findings offer new strategies and theoretical support for the treatment of SLI.