The flavonoid astragalin induces apoptosis in lung adenocarcinoma and correlates with a prognostic cell cycle gene signature associated with PANoptosis

Introduction

Lung cancer remains the leading cause of cancer-related mortality worldwide (1), with lung adenocarcinoma (LUAD) representing the most common histological subtype of non-small cell lung cancer (NSCLC), accounting for over 38.5% of all cases (2). Despite significant advances in targeted therapies and immune checkpoint inhibitors (3), clinical outcomes for LUAD patients remain suboptimal due to high rates of recurrence, intrinsic or acquired drug resistance, and pronounced tumor heterogeneity (4). These challenges underscore the urgent need for novel therapeutic agents and robust prognostic biomarkers that can guide personalized treatment strategies (5).

Recent conceptual advances have introduced the term “PANoptosis”—an integrative framework encompassing multiple forms of regulated cell death, including apoptosis, necroptosis, and pyroptosis (6). Unlike traditional views that treat these pathways in isolation, PANoptosis emphasizes their molecular crosstalk, functional redundancy, and synergistic roles in tumor suppression and immune modulation (7). Dysregulation of PANoptosis-related genes (PARGs) has been increasingly implicated in LUAD pathogenesis, while their restoration or activation represents a promising avenue for anticancer therapy (8). Notably, targeting PANoptosis not only directly eliminates malignant cells but may also reshape the tumor microenvironment (TME), thereby enhancing antitumor immunity (9).

Astragalin (kaempferol-3-O-glucoside), a naturally occurring flavonoid abundant in traditional medicinal plants such as Astragalus membranaceus and Cuscuta chinensis (10), has demonstrated anti-inflammatory, antioxidant, and antitumor properties in preclinical models (11,12). Emerging evidence suggests that astragalin can inhibit proliferation and induce mitochondrial apoptosis in various cancer cell lines (13,14). However, its potential role in modulating PANoptosis in LUAD and whether its effects are mediated through specific molecular targets within this network remain unexplored. Moreover, the clinical relevance of astragalin-associated genes in LUAD prognosis has not been systematically investigated. The integration of large-scale cancer genomics with advanced computational approaches, such as co-expression network analysis, protein-protein interaction (PPI) mapping, and machine learning-based survival modeling (15), provides unprecedented opportunities to uncover disease-relevant molecular drivers (16). When coupled with experimental validation, this integrative strategy bridges the gap between traditional herbal medicine and modern oncology by transforming bioactive compounds into mechanism-informed therapeutics.

In this study, we established an integrative computational and experimental framework to systematically identify key genes at the intersection of LUAD pathogenesis, PANoptosis regulation, and astragalin pharmacology. By combining bioinformatics prioritization with in vitro validation, we aimed to elucidate the molecular basis of astragalin’s antitumor activity and develop a clinically relevant prognostic model. Our approach exemplifies a rational, data-driven paradigm for repurposing natural products in cancer therapy while advancing the mechanistic understanding of regulated cell death in LUAD. We present this article in accordance with the TRIPOD and MDAR reporting checklists (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0150/rc).

Methods

Data collection

The RNA-seq data of LUAD samples were obtained from The Cancer Genome Atlas (TCGA) database, comprising 542 tumor and 48 adjacent normal samples. The list of PARGs was compiled using data from the Molecular Signatures Database and GeneCards database. Astragalin target genes (ATGs) were acquired from the Traditional Chinese Medicine Systems Pharmacology (TCMSP) database (http://tcmspw.com/tcmsp.php). Additionally, the SMILES notation of astragalin was retrieved from PubChem (https://pubchem.ncbi.nlm.nih.gov/) and subsequently submitted to SwissTargetPrediction (http://www.swisstargetprediction.ch/) to predict its potential protein targets. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Differentially expressed genes (DEGs)

We identified the DEGs between LUAD and control samples through the “DESeq2” R package, with the criteria of |log₂fold change (FC)| >1.5 and adjusted P<0.05.

WGCNA for key gene modules

Based on the DEGs identified above, the “WGCNA” R package was employed to identify gene co-expression modules. Genes with the lowest 50% median absolute deviation (MAD) were removed to retain those with the highest expression variability. Pairwise correlations among the remaining genes were then converted into an adjacency matrix for co-expression network construction. Using topological overlap measures (TOM), genes were clustered into distinct modules based on their expression similarity, with each module representing a group of genes sharing high co-expression and functional coherence. Modules significantly associated with LUAD were then identified. Finally, the genes within these LUAD-related modules were intersected with both PARGs and ATGs to pinpoint candidate hub genes.

Functional enrichment analysis

The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis on the target gene set was conducted using the “clusterProfiler” R package. Significantly enriched KEGG pathways were identified based on the criterion of P<0.05 and visualized by the “ggplot2” R package.

PPI network

A PPI network was built using the STRING database (https://string-db.org) based on the DEGs. The network was visualized in Cytoscape, and hub genes were determined by analyzing its topological features.

Identification of key candidate genes

To identify key candidate genes, we employed an integrative approach combining multiple computational methods. Least absolute shrinkage and selection operator (LASSO) regression was first implemented using the “glmnet” R package, with hub gene expression profiles from the PPI network as input, 10-fold cross-validation was used to select the optimal penalty parameter (λ), and genes retaining non-zero coefficients at one standard error above the minimum (λ₁SE) were retained as candidates. Second, the “xgboost” R package implemented an extreme gradient boosting (XGBoost) model to evaluate each gene’s importance in predicting patient prognosis risk. Third, the “Boruta” R package conducted robust feature selection by generating randomized shadow features and comparing their importance against original features in a random forest framework. Finally, univariate Cox proportional hazards regression (“survival” R package) was applied to assess prognosis in LUAD patients, with genes achieving P<0.05 considered prognostically relevant. The final set of key target genes was defined by the intersection of results from all four methods and served as input features for subsequent modeling.

Construction of a PANoptosis-related prognostic model using machine learning

A PANoptosis-related prognostic model was developed using the “Mime” R package, which integrates multiple machine learning algorithms to generate a comprehensive ensemble of 101 predictive models. The algorithms included: elastic net (Enet), Lasso regression, Ridge regression, stepwise Cox (StepCox), random survival forest (RSF), super principal components (SuperPC), survival support vector machine (survivalSVM), gradient boosting machine (GBM), partial least squares Cox regression (plsRcox), and CoxBoost. To evaluate predictive performance, the concordance index (C-index) was calculated for each model in both the training and validation cohorts. Model selection was guided by 10-fold cross-validation to ensure robustness and minimize overfitting. Based on the optimal model, a risk score was computed for each patient, and samples were stratified into high- and low-risk groups accordingly. To elucidate the biological and immunological basis of our four-gene risk stratification, we performed functional enrichment and immune microenvironment analyses between high- and low-risk groups using GSVA with the MSigDB “c2.cp.kegg.v7.4.symbols.gmt” gene set.

Immune cell infiltration analysis

Immune cell composition across samples—stratified into high- and low-risk groups—was deconvoluted using the CIBERSORT algorithm. Using a predefined signature matrix of 547 barcode genes, CIBERSORT applies a deconvolution approach to infer immune cell composition from bulk gene expression profiles. The estimated fractions of all 22 immune cell types sum to 1 for each sample. In addition, pairwise correlations among immune cell types were calculated to assess their interrelationships. Furthermore, the associations between the expression levels of the identified key target genes and the abundance of each immune cell type were evaluated to explore potential links between these targets and the tumor immune microenvironment.

Molecular docking

To predict the potential binding modes and interactions between astragalin and its key predicted target proteins, molecular docking was performed. Astragalin’s three-dimensional structure was acquired from PubChem (https://pubchem.ncbi.nlm.nih.gov/), and atomic coordinates of the target proteins were downloaded from the RCSB Protein Data Bank (https://www.rcsb.org/), with protein–ligand preparation and parameter setup performed via AutoDock Tools. This allowed for the prediction of binding poses and affinity scores between astragalin and each target protein.

Cell culture and treatment

Human NSCLC A549 cells were obtained from Yimo Biotechnology (Cat. No. IM-H113, Shanghai, China). Cells were cultured in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin at 37 ℃ in a humidified atmosphere containing 5% CO₂. For experimental treatments, cells were seeded into six-well plates and allowed to adhere overnight. The control group received vehicle only [e.g., dimethyl sulfoxide (DMSO) <0.1%], while the astragalin-treated group was exposed to a defined concentration of astragalin (dissolved in DMSO) for a specified duration (40 µg/mL, 48 h) (17). Following treatment, cells were harvested for total RNA and total protein extraction to assess gene and protein expression, respectively.

Quantitative real-time polymerase chain reaction (qRT-PCR) for gene expression

A549 cells were cultured in the presence or absence of astragalin, and total RNA was subsequently extracted with TRIzol. After chloroform-induced phase separation and isopropanol-mediated RNA precipitation, the purified RNA was dissolved in diethyl pyrocarbonate (DEPC)-treated water and its concentration measured using a NanoDrop spectrophotometer. Reverse transcription was conducted with 2 µg of total RNA using a commercial 2× room temperature (RT) master mix [including oligo(dT) primers and reverse transcriptase], incubated at 50 ℃ for 15 min and terminated by brief heating at 85 ℃ (5 s).

For qRT-PCR, each 20 µL reaction—set up in triplicate—included 1 µL of diluted complementary DNA (cDNA) template, 0.4 µL each of 10 µM gene-specific primers, 10 µL Hieff^® SYBR Green Master Mix (No Rox; Yeasen, Cat. No. 11201ES08), and nuclease-free water. Amplification was run on a qRT-PCR system with an initial denaturation at 95 ℃ for 30 s, followed by 40 cycles of 95 ℃ (10 s) and 60 ℃ (30 s). Transcript levels of TOP2A, PLK1, CCNB1, and CDKN3 were evaluated relative to the reference gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) using the comparative 2^−ΔΔCt algorithm. All results reflect the mean ± standard deviation derived from three biologically independent experiments. Primer sequences are provided in Table 1.

Western blot

Total protein was extracted from A549 cells treated with or without astragalin using ice-cold RIPA lysis buffer (R0010, Solarbio, Beijing, China) supplemented with phosphatase inhibitors. Lysates were sonicated on ice and centrifuged at 12,000 g for 10 min at 4 ℃. Protein concentrations were determined using a bicinchoninic acid (BCA) protein assay kit (PC0020, Solarbio). Equal amounts of protein (20 µg per lane) were separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto polyvinylidene difluoride (PVDF) membranes via wet electroblotting (110 mA, 8 h, 4 ℃). Membranes were blocked with 5% non-fat milk in TBST for 1 h at RT and then incubated overnight at 4 ℃ with primary antibodies: anti-Bax (A19684, ABclonal), anti-caspase-3 (A19654, ABclonal), and anti-Bcl-2 (A0208, ABclonal). After three washes with Tris-buffered saline with Tween 20 (TBST), membranes were incubated for 1 h at 37 ℃ with horseradish peroxidase (HRP)-conjugated goat anti-rabbit immunoglobulin G (IgG) secondary antibody (1:5,000). Immunoreactive bands were detected using enhanced chemiluminescence (ECL) reagents and imaged with a chemiluminescence detection system.

Statistical analysis

Data processing, statistical testing, and figure generation were carried out in R (version 4.4.1). Depending on data distribution, group comparisons for continuous variables employed either the Wilcoxon rank-sum test or Student’s t-test. Correlations were quantified using the Spearman method. All statistical tests were two-tailed, with significance set at P<0.05.

Results

Identification of LUAD-associated candidate genes

We analyzed the DEGs between LUAD and control samples and identified 5,211 upregulated and 1,396 downregulated genes (Figure 1A, table available at https://cdn.amegroups.cn/static/public/tcr-2026-1-0150-1.xlsx). The top 100 DEGs expression across samples were showed in a heatmap (Figure 1B). To further pinpoint genes involved in LUAD pathophysiology, we performed WGCNA in these DEGs expression matrix with a soft-thresholding power β=4 for scale-free topology (Figure 1C), and identified 10 gene modules. Modules with inter-modular correlation coefficients >0.75 were merged, yielding a final set of nine distinct modules (Figure 1D). We then evaluated module-trait relationships by correlating each module’s eigengene with the binary phenotype and identified four modules (yellow, pink, green, and turquoise) (Figure 1E), as significantly associated with LUAD phenotype. Finally, the strong positive association between each module and the LUAD phenotype was evaluated by correlating module membership (MM) with gene significance (GS) (Figure 1F), suggesting that highly connected hub genes are closely linked to LUAD pathogenesis, underscoring their functional importance.

Figure 1 Identification of key gene modules associated with LUAD. (A) Volcano plot showing DEGs between tumor and adjacent normal tissues. (B) Heatmap of the top 100 most significantly dysregulated DEGs across samples. (C) Selection of the soft-thresholding power (β). (D) Hierarchical clustering dendrogram of genes and corresponding module assignment. (E) Heatmap showing correlations between WGCNA modules and clinical traits. (F) Scatter plots of MM versus GS for the yellow, pink, green, and turquoise modules. DEGs, differentially expressed genes; GS, gene significance; LUAD, lung adenocarcinoma; MM, module membership; NC, normal control; WGCNA, weighted gene co-expression network analysis.

Identification of key DEGs associated with PANoptosis and drug action

To pinpoint key DEGs in LUAD that are involved in PANoptosis and potentially modulated by astragalin, we intersected three gene sets [WGCNA modules, PARGs (table available at https://cdn.amegroups.cn/static/public/tcr-2026-1-0150-2.xlsx), and ATGs (table available at https://cdn.amegroups.cn/static/public/tcr-2026-1-0150-3.xlsx)], yielding 59 core overlapping genes (Figure 2A). Functional enrichment analysis revealed that these upregulated overlapping genes (DEGs in WGCNA module and PARGs) were significantly enriched in cancer-related pathways, including the PI3K-Akt signaling pathway, cell cycle, and viral carcinogenesis (Figure 2B), while these downregulated overlapping genes (DEGs and PARGs) were primarily associated with complement and coagulation cascades, malaria, and lipid metabolism, highlighting involvement in immune and metabolic processes (Figure 2C). In addition, KEGG pathway analysis of all 59 core genes showed significant enrichment in p53 signaling, cell cycle, cellular senescence, HIF-1 signaling, and focal adhesion (Figure 2D), indicating that astragalin may exert anti-LUAD effects by modulating central networks governing cell cycle progression and cell fate decisions. Moreover, pathway analysis focusing specifically on the intersection of PARGs and ATGs revealed strong enrichment in apoptosis, NSCLC pathways, MAPK signaling, as well as immune-related pathways such as T cell receptor signaling and receptor interaction (Figure 2E).

Figure 2 Identification of hub genes associated with PANoptosis and astragalin. (A) Venn diagram showing the overlap between WGCNA module genes, PARGs, and ATGs. (B) Pathway enrichment analysis of upregulated DEGs that intersect with PARGs. (C) Pathway enrichment analysis of downregulated DEGs that intersect with PARGs. (D) Pathway enrichment analysis of the 59 genes identified as the intersection of DEGs, PARGs, and ATGs. (E) Pathway enrichment analysis of the intersection between PARGs and ATGs. ATGs, astragalin target genes; DEGs, differentially expressed genes; PARGs, PANoptosis-related genes; WGCNA, weighted gene co-expression network analysis.

Identification of potential candidate genes

To prioritize critical targets among the 59 core genes, we first constructed a PPI network with a confidence score threshold of 0.7 and removed isolated nodes, resulting in a refined network of 53 nodes and 243 edges (Figure 3A). The network was imported into Cytoscape for topological analysis using the CytoHubba plugin, and the top 30 hub genes were selected for downstream analysis (Figure 3B). We then applied LASSO regression to these 30 genes, which narrowed the list to 12 genes with non-zero coefficients (Figure 3C). To enhance robustness, we further employed the Boruta algorithm, a shadow-feature-based feature selection method—which identified six genes as significantly relevant: ACHE, IL17A, CDKN3, TOP2A, CCNB1, and PLK1 (Figure 3D). In parallel, the importance of the 12 genes was ranked based on the XGBoost algorithm (Figure 3E). Finally, univariate Cox regression showed that eight were significantly associated with the prognosis of patients (P<0.05) (Figure 3F).

Figure 3 Identification and prioritization of candidate genes associated with prognosis in LUAD. (A) PPI network of 59 hub genes. (B) Subnetwork of the top 30 hub genes ranked by degree centrality. (C) LASSO Cox regression model tuning. (D) Feature selection using the Boruta algorithm. (E) Feature importance ranking of candidate genes derived from the XGBoost model. (F) Univariate Cox regression analysis of candidate genes. CI, confidence interval; LASSO, least absolute shrinkage and selection operator; LUAD, lung adenocarcinoma; PPI, protein-protein interaction; XGBoost, extreme gradient boosting.

Development of a prognostic risk model for LUAD

Based on integrative multi-algorithm screening and survival analysis, we identified TOP2A, PLK1, CDKN3, and CCNB1 as key candidate genes, all significantly upregulated in tumor tissues compared to normal controls (Figure 4A). To translate these findings into a clinically applicable tool, we developed a machine learning-based prognostic risk model using these four genes. Using the Mime framework in R, we constructed 101 models across various algorithms (e.g., StepCox, RSF, Enet, GBM). Each model generated a composite risk score, stratifying patients into high- and low-risk groups. The cohort was randomly split into training (60%) and validation (40%) sets for robustness. The StepCox(forward) + RSF ensemble achieved the highest predictive performance with the best C-index in both sets (Figure 4B). Kaplan-Meier analysis showed significant differences in overall survival between the two groups (P<0.001), with worse prognosis in the high-risk group (Figure 4C). Time-dependent ROC analysis demonstrated excellent accuracy, with high area under the curve (AUC) values >0.90 at 1-, 3-, and 5-year timepoints in both sets (Figure 4D,4E).

Figure 4 Validation of the prognostic model in LUAD. (A) Expression profiles of four key targets across LUAD and normal tissues. (B) Integration of multiple algorithms to evaluate the predictive performance of each model through C-index scores in both training and validation cohorts. (C) Kaplan-Meier survival curves for overall survival in different risk groups. (D,E) ROC curve analysis illustrating the prediction performance of the developed model in the training and validation cohorts. ****, P<0.0001. AUC, area under the curve; C-index, concordance index; Enet, elastic net; GBM, gradient boosting machine; Lasso, least absolute shrinkage and selection operator; LUAD, lung adenocarcinoma; plsRcox, partial least squares Cox regression; ROC, receiver operating characteristic; RSF, random survival forest; StepCox, stepwise Cox; SuperPC, super principal components; survivalSVM, survival support vector machine.

Functional enrichment and immune microenvironment analysis across risk groups

GSVA revealed distinct pathway activation patterns between risk groups. The p53 signaling pathway and autophagy-related pathways were significantly less active in the high-risk group (Figure 5A,5B). Immune cell analysis showed that the low-risk group exhibited higher infiltration of plasma cells, resting memory CD4⁺ T cells, and monocytes, whereas the high-risk group was enriched for CD8⁺ T cells, M1-polarized macrophages, and regulatory T cells (Tregs) (Figure 5C). A correlation network further illustrated unique interrelationships among immune subsets in each group (Figure 5D). All four signature genes showed strong positive pairwise correlations (Figure 5E). Strikingly, each gene was significantly positively correlated with M1 macrophage infiltration (Figure 5F), suggesting a potential association that warrants further investigation into its functional role in the TME of advanced LUAD.

Figure 5 Functional enrichment analysis and immune microenvironment characterization. (A) Heatmap showing enrichment scores of differentially activated pathways between high- and low-risk groups. (B) Violin plots illustrating the difference of p53 and autophagy in the two risk groups. (C) Immune infiltrating cell differences in the two risk groups. (D) Heatmap depicting the correlations among immune cell populations. (E) Heatmap depicting the correlation between the expression levels of the four key genes. (F) Correlations between each of the four key genes and specific immune cell subsets. *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001.

Molecular docking analysis

We evaluated the binding potential of astragalin with the four key LUAD targets (TOP2A, PLK1, CDKN3, and CCNB1), the in silico docking simulations predicted strong binding affinities between astragalin and all four targets, with binding energies of −8.5 kcal/mol (TOP2A), −8.4 kcal/mol (PLK1), −7.4 kcal/mol (CDKN3), and −7.5 kcal/mol (CCNB1), all well below the −5 kcal/mol threshold, indicating stable and favorable interactions (Figure 6). These computational results suggest potential direct binding, providing a theoretical foundation for the observed gene regulation and guiding future validation of target engagement and functional inhibition.

Figure 6 Molecular docking. (A) TOP2A. (B) PLK1. (C) CDKN3. (D) CCNB1.

Astragalin suppressed cell cycle progression and induced apoptosis

Treatment with astragalin significantly downregulated the messenger RNA (mRNA) expression of cell cycle-related genes TOP2A, PLK1, CCNB1, and CDKN3 in A549 cells (Figure 7A). Western blot revealed that astragalin treatment significantly altered the expression levels of key apoptosis-related proteins in A549 cells. Specifically, astragalin downregulated the anti-apoptotic protein Bcl-2 and upregulated the pro-apoptotic proteins Bax and Caspase-3 compared to the control group (Figure 7B).

Figure 7 Validation of gene and protein expression changes in A-549 cells treated with astragalin. (A) qRT-PCR analysis showing the relative mRNA expression levels. (B) Western blot analysis of apoptosis-related proteins caspase-3, Bax, and Bcl-2 expression. *, P<0.05; **, P<0.01. ASG, astragalin; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; mRNA, messenger RNA; qRT-PCR, quantitative real-time polymerase chain reaction.

Discussion

Emerging evidence suggests that coordinated induction of multiple regulated cell death modalities termed “PANoptosis” may overcome the limitations of single-pathway targeting in cancer therapy (18). Natural compounds like astragalin, with their polypharmacological profiles, represent promising candidates to engage this integrated cell death network (19). Meanwhile, LUAD remains a formidable clinical challenge due to its aggressive biology, therapeutic resistance, and lack of reliable prognostic biomarkers (20). In this study, we integrated TCGA transcriptomic data, PARG sets, astragalin target predictions from systems pharmacology databases, and multiple machine learning algorithms to identify a four-gene signature, including TOP2A, PLK1, CDKN3, and CCNB1 that robustly predicts LUAD prognosis Collectively, our bioinformatic analysis revealed that these genes are enriched in PANoptosis-related pathways, supporting the hypothesis that the mechanism of astragalin may extend beyond canonical apoptosis to involve a broader network of programmed cell death.

The four hub genes identified in our model are all well-established regulators of mitosis and genomic stability. TOP2A resolves topological stress during DNA replication and is frequently overexpressed in cancers, correlating with poor survival (21). PLK1 orchestrates multiple stages of mitosis and is a validated oncogene in NSCLC (22). CCNB1 forms a complex with CDK1 to drive the G2/M transition (23), while CDKN3 dephosphorylates CDKs to promote cell cycle progression—paradoxically acting as an oncogene despite its name (24). Their consistent upregulation in LUAD tumors and strong association with adverse outcomes align with their roles in sustaining proliferative signaling, a hallmark of cancer. Notably, all four genes were significantly downregulated upon astragalin treatment in A549 cells, suggesting an association between astragalin treatment and the suppression of these key cell cycle regulators (25). Definitive proof of cell cycle arrest, however, would require direct functional assays such as flow cytometric analysis of cell cycle distribution or measurement of cyclin-dependent kinase (CDK) activity.

Overall, beyond the individual functional roles of the genes, our integrative bioinformatics approach yielded several key insights that warrant in-depth discussion (Table 2). WGCNA identified the four-gene signature as a component of a tightly co-expressed module, implying their coordinated involvement in the pathogenesis of LUAD (26). Their robust and consistent upregulation in tumor tissues, combined with the high predictive accuracy of the machine learning-derived prognostic model, underscores their clinical relevance as reliable prognostic biomarkers (27). Furthermore, GSVA revealed suppressed p53 signaling and autophagy pathways in the high-risk group, which provides a biological rationale for the unfavorable prognosis observed in this subgroup and links the gene signature to the inactivation of critical tumor suppressor pathways (28). Finally, while the positive correlation between the expression of signature genes and M1 macrophage infiltration remains correlative, it generates a novel hypothesis regarding the potential crosstalk between tumor cell-intrinsic cell cycle dysregulation and the composition of the tumor immune microenvironment (29). This systems-level perspective, derived from comprehensive computational analyses, represents a critical outcome of our study and establishes a multi-faceted framework for deciphering the molecular mechanisms underlying LUAD progression.

Additionally, our functional enrichment analyses revealed that these genes intersect with PANoptosis-related pathways, particularly p53 signaling and apoptosis. While astragalin has been reported to induce mitochondrial apoptosis (13,17,30), our study expands this understanding by positioning it within the broader context of regulated cell death. The significant enrichment of PANoptosis pathways, including pyroptosis and ferroptosis in the overlapping gene set, implies that astragalin may trigger multiple death modalities simultaneously (19,31), potentially overcoming the redundancy and compensatory mechanisms that often limit single-pathway targeting. This multi-death induction could enhance therapeutic efficacy and reduce the likelihood of resistance (32). Importantly, molecular docking simulations predicted high-affinity binding between astragalin and all four target proteins, suggesting a structural plausibility for potential direct interaction. While experimental confirmation is needed, the docking and network-based predictions indicate that astragalin has the potential to bind multiple proteins associated with cell cycle progression and oncogenic signaling, warranting further mechanistic investigation.

The prognostic model built on these four genes exhibited exceptional performance. Moreover, risk stratification was biologically meaningful: the high-risk group showed suppressed p53 and autophagy pathways, processes critical for tumor suppression and stress adaptation (33), while exhibiting elevated infiltration of M1 macrophages and Tregs. Although M1 macrophages are classically considered anti-tumor (34), their persistent presence in advanced tumors may reflect a state of chronic inflammation that paradoxically supports immunosuppression and tissue remodeling (35). The positive correlation between all four signature genes and M1 macrophage infiltration observed in our bioinformatic analysis is intriguing. It suggests a feed-forward loop in which oncogene-driven proliferation shapes an inflammatory yet permissive TME (36). Although these results imply a potential interplay, they should be interpreted with caution due to their purely correlative nature. Further experimental validation, including cytokine profiling or macrophage co-culture assays following astragalin treatment, will be required to clarify the functional significance of this association.

Despite these advances, several limitations merit consideration. First, our in vitro validation was restricted to a single cell line (A549), which substantially limits the generalizability and translational relevance of our findings. Future studies should incorporate additional LUAD cell lines with diverse genetic backgrounds and non-malignant lung epithelial cells to evaluate the selectivity and reproducibility of astragalin’s effects. Most importantly, while our bioinformatic analysis indicated the potential involvement of PANoptosis, this study did not experimentally validate the activation of pyroptosis or necroptosis. Future investigations should employ functional assays for key PANoptosis markers, such as caspase-1 activation, gasdermin cleavage, and RIPK1/RIPK3/MLKL phosphorylation, to determine whether astragalin genuinely triggers this integrated cell death network. Second, our in silico target predictions for astragalin are based on computational docking and database mining, which do not account for compound stability, cellular uptake, or metabolic transformation, factors that could significantly influence its actual biological activity. Third, the observed downregulation of TOP2A, PLK1, CDKN3, and CCNB1 could be a consequence of reduced proliferation following apoptosis induction. Future studies should include cell cycle analysis and functional rescue experiments to establish causality.

Conclusions

In conclusion, our integrative approach bridges traditional herbal medicine and modern oncology by transforming astragalin from a phenotypic hit into a mechanism-informed candidate. The identified four-gene signature not only offers a clinically actionable prognostic tool but also reveals a potential link between a prognostic cell cycle signature, PANoptosis-related pathways, and immune modulation, providing a rationale for future studies to explore these interconnected networks as therapeutic targets.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD and MDAR reporting checklists. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0150/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0150/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0150/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0150/coif). The authors have no conflicts of interest to declare.

Ethical Statement: 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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Marshall HM, Fong KM. Lung cancer screening - Time for an update? Lung Cancer 2024;196:107956. [Crossref] [PubMed]
2. Skřičková J, Kadlec B, Venclíček O, et al. Lung cancer. Cas Lek Cesk 2018;157:226-36.
3. Chen Y, Zhou Y, Ren R, et al. Harnessing lipid metabolism modulation for improved immunotherapy outcomes in lung adenocarcinoma. J Immunother Cancer 2024;12:e008811. [Crossref] [PubMed]
4. Zhang L, Cui Y, Mei J, et al. Exploring cellular diversity in lung adenocarcinoma epithelium: Advancing prognostic methods and immunotherapeutic strategies. Cell Prolif 2024;57:e13703. [Crossref] [PubMed]
5. Zhang L, Cui Y, Zhou G, et al. Leveraging mitochondrial-programmed cell death dynamics to enhance prognostic accuracy and immunotherapy efficacy in lung adenocarcinoma. J Immunother Cancer 2024;12:e010008. [Crossref] [PubMed]
6. Shi C, Cao P, Wang Y, et al. PANoptosis: A Cell Death Characterized by Pyroptosis, Apoptosis, and Necroptosis. J Inflamm Res 2023;16:1523-32. [Crossref] [PubMed]
7. Zhu P, Ke ZR, Chen JX, et al. Advances in mechanism and regulation of PANoptosis: Prospects in disease treatment. Front Immunol 2023;14:1120034. [Crossref] [PubMed]
8. Miao Z, Yu W. Significance of novel PANoptosis genes to predict prognosis and therapy effect in the lung adenocarcinoma. Sci Rep 2024;14:20934. [Crossref] [PubMed]
9. Gao J, Xiong A, Liu J, et al. PANoptosis: bridging apoptosis, pyroptosis, and necroptosis in cancer progression and treatment. Cancer Gene Ther 2024;31:970-83. [Crossref] [PubMed]
10. Chen J, Zhong K, Qin S, et al. Astragalin: a food-origin flavonoid with therapeutic effect for multiple diseases. Front Pharmacol 2023;14:1265960. [Crossref] [PubMed]
11. Fu YQ, Li C, Ding DN, et al. Astragalin: A promising herbal compound with broad anticancer potential Oncol Lett 2026;31:39. (Review). [Crossref] [PubMed]
12. Yang CZ, Wang SH, Zhang RH, et al. Neuroprotective effect of astragalin via activating PI3K/Akt-mTOR-mediated autophagy on APP/PS1 mice. Cell Death Discov 2023;9:15. [Crossref] [PubMed]
13. Zheng L, Zhao Z, Chen L, et al. Astragalin induces immunogenic cell death in liver cancer by targeting NQO2 to promote ROS-mediated endoplasmic reticulum stress pathway. Free Radic Biol Med 2025;239:317-35. [Crossref] [PubMed]
14. Yang S, Zhang Y, Qiu H, et al. Astragalin Attenuates Bone Destruction and the Progression of Bone Metastasis in Breast Cancer. Cancers (Basel) 2025;17:3442. [Crossref] [PubMed]
15. Zhang C, Liu Y, Lu Y, et al. Identification of potential biomarkers for lung adenocarcinoma: a study based on bioinformatics analysis combined with validation experiments. Front Oncol 2024;14:1425895. [Crossref] [PubMed]
16. Shen Z, Liu S, Liu J, et al. Weighted Gene Co-Expression Network Analysis and Treatment Strategies of Tumor Recurrence-Associated Hub Genes in Lung Adenocarcinoma. Front Genet 2021;12:756235. [Crossref] [PubMed]
17. Bai J, Chen Y, Zhao G, et al. In Vitro and Vivo Experiments Revealing Astragalin Inhibited Lung Adenocarcinoma Development via LINC00582/miR-140-3P/PDPK1. J Biochem Mol Toxicol 2024;38:e70042. [Crossref] [PubMed]
18. Cai H, Lv M, Wang T. PANoptosis in cancer, the triangle of cell death. Cancer Med 2023;12:22206-23. [Crossref] [PubMed]
19. Xu G, Yu B, Wang R, et al. Astragalin flavonoid inhibits proliferation in human lung carcinoma cells mediated via induction of caspase-dependent intrinsic pathway, ROS production, cell migration and invasion inhibition and targeting JAK/STAT signalling pathway. Cell Mol Biol (Noisy-le-grand) 2021;67:44-9. [Crossref] [PubMed]
20. Wang H, Han G, Chen J. Heterogeneity of tumor immune microenvironment in malignant and metastatic change in LUAD is revealed by single-cell RNA sequencing. Aging (Albany NY) 2023;15:5339-54. [Crossref] [PubMed]
21. Zhou T, Niu Y, Li Y. Advances in research on malignant tumors and targeted agents for TOP2A Mol Med Rep 2025;31:50. (Review). [Crossref] [PubMed]
22. Kim DE, Shin SB, Kim CH, et al. PLK1-mediated phosphorylation of β-catenin enhances its stability and transcriptional activity for extracellular matrix remodeling in metastatic NSCLC. Theranostics 2023;13:1198-216. [Crossref] [PubMed]
23. Yang W, Wu X, Cai F, et al. NFATc4 Promotes Lung Adenocarcinoma Progression via the CCNB1/CDK1 Pathway and Is a Potential Prognostic Biomarker. Cancer Sci 2025;116:2400-12. [Crossref] [PubMed]
24. Chen Q, Chen K, Guo G, et al. A critical role of CDKN3 in Bcr-Abl-mediated tumorigenesis. PLoS One 2014;9:e111611. [Crossref] [PubMed]
25. Xie Q, He Z, Tan L, et al. Hesperetin induces apoptosis in lung squamous carcinoma cells via G(2)/M cycle arrest, inhibition of the Notch1 pathway and activation of endoplasmic reticulum stress. Int J Mol Med 2025;55:77. [Crossref] [PubMed]
26. Dai D, Shi R, Han S, et al. Weighted gene coexpression network analysis identifies hub genes related to KRAS mutant lung adenocarcinoma. Medicine (Baltimore) 2020;99:e21478. [Crossref] [PubMed]
27. Liang M, Zhang Z, Wu L, et al. Evolving prognostic paradigms in lung adenocarcinoma with brain metastases: a web-based predictive model enhanced by machine learning. Discov Oncol 2025;16:117. [Crossref] [PubMed]
28. Ji NN, Li SN, Shao L, et al. MDMX enhances radiosensitivity in lung adenocarcinoma and squamous cell carcinoma by inhibiting P53-mediated autophagy. Cell Oncol (Dordr) 2025;48:1067-88. [Crossref] [PubMed]
29. Wang F, Gao X, Wang P, et al. Immune Subtypes in LUAD Identify Novel Tumor Microenvironment Profiles With Prognostic and Therapeutic Implications. Front Immunol 2022;13:877896. [Crossref] [PubMed]
30. Chen M, Cai F, Zha D, et al. Astragalin-induced cell death is caspase-dependent and enhances the susceptibility of lung cancer cells to tumor necrosis factor by inhibiting the NF-кB pathway. Oncotarget 2017;8:26941-58. [Crossref] [PubMed]
31. Huang Q, Shi Z, Zheng D, et al. Astragalin Inhibits Oxidative Stress-Induced Pyroptosis and Apoptosis in Mouse Models of Renal Ischemia/Reperfusion Injury by Activating the SIRT1/Nrf2 Pathway. Phytother Res 2025;39:5025-42. [Crossref] [PubMed]
32. Tufail M, Jiang CH, Li N. Immune evasion in cancer: mechanisms and cutting-edge therapeutic approaches. Signal Transduct Target Ther 2025;10:227. [Crossref] [PubMed]
33. Liu Y, Stockwell BR, Jiang X, et al. p53-regulated non-apoptotic cell death pathways and their relevance in cancer and other diseases. Nat Rev Mol Cell Biol 2025;26:600-14. [Crossref] [PubMed]
34. Wang X, Huang R, Lu Z, et al. Exosomes from M1-polarized macrophages promote apoptosis in lung adenocarcinoma via the miR-181a-5p/ETS1/STK16 axis. Cancer Sci 2022;113:986-1001. [Crossref] [PubMed]
35. Xu R, Lee YJ, Kim CH, et al. Invasive FoxM1 phosphorylated by PLK1 induces the polarization of tumor-associated macrophages to promote immune escape and metastasis, amplified by IFITM1. J Exp Clin Cancer Res 2023;42:302. [Crossref] [PubMed]
36. Locati M, Curtale G, Mantovani A. Diversity, Mechanisms, and Significance of Macrophage Plasticity. Annu Rev Pathol 2020;15:123-47. [Crossref] [PubMed]