Exploring the mechanisms of luteolin in treating polycystic ovary syndrome and endometriosis via network pharmacology, molecular docking, and molecular dynamics simulation

This study systematically elucidated the molecular mechanisms underpinning luteolin’s therapeutic effects on PCOS and EM through an integrated approach combining network pharmacology, molecular docking, and molecular dynamics simulations. Luteolin was found to exert synergistic therapeutic actions via 3 core pathways: restoration of endocrine balance in PCOS and attenuation of EM lesion proliferation through ESR1, modulation; suppression of ovarian granulosa cell hyperplasia and endometrial cell invasion via EGFR/PI3K/AKT axis inhibition; and metabolic syndrome amelioration in PCOS through lipid metabolism pathway regulation. Molecular validation confirmed high-affinity binding to key targets (ESR1, MMP9, and ERBB2) and demonstrated dynamic binding stability through molecular dynamics simulations, providing structural evidence for its pharmacological basis. Notably, the identification of shared risk gene across PCOS and EM offers novel targets for trans-disease therapeutic strategies, aligning with luteolin’s multi-target profile that matches the complex pathological networks of reproductive endocrine disorders. This study not only advances mechanistic understanding of natural products’ polypharmacology but also proposes luteolin as a candidate agent for gynecological metabolic diseases, grounded in the “treating different diseases with the same method” paradigm. Future investigations should prioritize translational validation to unlock its full clinical potential.

Polycystic ovary syndrome (PCOS), one of the most prevalent gynecological disorders, affects 8% to 13% of women of reproductive age. Characterized by ovarian polycystic morphology, hyperandrogenism, and chronic anovulation, it is frequently complicated by insulin resistance, obesity, and an elevated risk of long-term sequelae including endometritis, diabetes mellitus, hypertension, and atherosclerosis. [ 1 , 2 ] Endometriosis (EM), another common benign gynecological condition with a prevalence of 10% to 15% in reproductive-age women, involves the ectopic implantation of endometrial glands and stroma outside the uterine cavity. [ 1 , 2 ] Manifesting as pelvic pain, infertility, and dyspareunia, EM exhibits malignant-like behaviors such as invasion, metastasis, and recurrence, significantly impairing patients’ quality of life. [ 3 , 4 ] Current therapeutic strategies for PCOS primarily rely on insulin sensitizers, oral contraceptives, and antiandrogens, while EM management involves gonadotropin-releasing hormone agonists, progestins, or surgical resection. [ 5 , 6 ] Despite partial efficacy, these interventions are limited by adverse effects including gastrointestinal disturbances, thromboembolic risks, and bone loss. [ 6 ] Furthermore, while hyperandrogenism in PCOS may exacerbate EM progression, and insulin resistance in PCOS could modulate inflammatory responses in EM, existing studies predominantly focus on single diseases, with insufficient evidence for synergistic treatment approaches. [ 7 , 8 ] Luteolin, a naturally occurring flavonoid, has recently emerged as a promising multi-target therapeutic agent for metabolic gynecological disorders. [ 9 , 10 ] Its pharmacological activities include hormonal regulation, insulin resistance amelioration, and antioxidant effects, which collectively mitigate ovarian inflammation, preserve follicular development, and alleviate PCOS symptoms. [ 9 , 10 ] Additionally, luteolin demonstrates anti-endometriotic potential by inducing apoptosis, modulating immune responses, and suppressing inflammation, thereby disrupting the pro-EM microenvironment. [ 11 ] Genome-wide association studies have identified 12 shared risk loci between PCOS and EM, with genes such as EGFR implicated in both ovarian steroidogenesis and endometrial cell migration. Luteolin may thus exert synergistic effects by modulating these overlapping targets. [ 12 , 13 ] However, experimental validation of luteolin’s dual therapeutic actions in PCOS and EM remains scarce. Recent advances in integrative methodologies (combining network pharmacology, molecular docking, and molecular dynamics simulations) offer novel avenues for dissecting the complex mechanisms of natural products. [ 14 , 15 ] By constructing protein–protein interaction networks, enriching core pathways, validating ligand–target interactions via molecular docking, and analyzing ligand–receptor stability through molecular dynamics, this study systematically elucidates the molecular basis of luteolin’s regulatory effects on PCOS- and EM-associated signaling networks. Such insights may provide a theoretical foundation for innovative therapeutic strategies targeting these interconnected disorders. The detailed workflow is illustrated in Figure 1 . Flowchart of this study.

Conceptualization: Rong Chen, Li Tian. Methodology: Li Tian, Siying Pan, Guodong Zhao, Aravind Raveendran. Software: Rong Chen, Siying Pan. Supervision: Li Tian, Yingbo Xu, Juan Zhang, Guodong Zhao. Validation: Rong Chen, Siying Pan, Aravind Raveendran. Writing – original draft: Rong Chen, Li Tian, Yingbo Xu. Writing – review & editing: Rong Chen, Li Tian, Juan Zhang, Aravind Raveendran.

Ethical approval was waived or not necessary, all procedures performed in studies do not involve human participants or animals. To identify luteolin’s pharmacological targets, we integrated 3 databases: Traditional Chinese Medicine Systems Pharmacology ( https://tcmsp-e.com/ ), Encyclopedia of Traditional Chinese Medicine (ETCM) ( http://www.tcmip.cn/ETCM/index.php/Home/ ), and SwissTargetPrediction ( http://www.swisstargetprediction.ch/ ). Target genes were standardized using the Universal Protein Resource ( https://www.uniprot.org/ ) with the species restricted to Homo sapiens . Duplicate entries across databases were removed to compile the final target list. Disease-related genes for PCOS and EM were retrieved from the GeneCards database ( https://www.genecards.org/ ) using the search terms “Polycystic Ovary Syndrome” and “Endometriosis” (relevance score > 1). Overlapping targets between disease-associated genes and luteolin’s targets were identified to determine potential therapeutic targets. The intersection targets of luteolin, PCOS, and EM were uploaded to the STRING database ( https://cn.string-db.org/ ) with the species set to Homo sapiens and a confidence threshold >0.4. Disconnected nodes were hidden to generate the PPI network, which was subsequently imported into Cytoscape v3.8.0. Hub genes were identified using the CytoHubba plugin based on topological network degree values. Gene Expression Omnibus datasets ( https://www.ncbi.nlm.nih.gov/geo/ ) were queried using “Polycystic Ovary Syndrome” and “Endometriosis” to obtain PCOS- and EM-related expression profiles. Inclusion criteria required human samples, consistent tissue origins between cases and controls, and balanced group sizes. Hub gene diagnostic performance was evaluated using receiver operating characteristic (ROC) curves and area under the curve (AUC) values via the “pROC” package in R. Gene ontology (GO) and Kyoto encyclopedia of genes and genomes (KEGG) enrichment analyses were performed on intersection genes using the R (v4.2.2) clusterProfiler package. GO terms encompassed biological processes (BP), cellular components, and molecular functions (MF). KEGG pathways were analyzed to identify disease-relevant signaling cascades. Statistical significance was set at adj. P  < .05. To validate luteolin’s therapeutic interactions, molecular docking was performed between luteolin (ligand) and hub gene-encoded proteins (receptors). Protein structures (PDB format) were downloaded from RCSB ( http://www.rcsb.org/ ), selecting only reviewed human entries. Luteolin’s SDF structure was obtained from PubChem. Blind docking was conducted using CB-Dock2 ( https://cadd.labshare.cn/cb-dock2/index.php ), which automates binding site detection and box sizing. [ 16 , 17 ] Water molecules and heteroatoms were removed prior to docking with AutoDock Vina. Optimal poses were analyzed using PyMOL v.2.6.0 (Schrödinger, LLC, New York). Based on docking results, the top-ranked luteolin–protein complexes for PCOS and EM hub genes underwent molecular dynamics simulations. Proteins and ligands were parameterized with Amber14 sb and GAFF2 force fields, respectively, using the TIP4P water model and periodic boundary conditions (1.2 nm). Long-range electrostatics were calculated via the particle mesh Ewald method, and systems were neutralized with NaCl using Monte Carlo ion placement. Simulation protocols included energy minimization (50,000 steps), equilibration under NVT (310 K, 50,000 steps), and NPT (310 K, 1 atm, 50,000 steps) ensembles, followed by 100 ns production runs (2 fs time steps, coordinates saved every 10 ps). Trajectory analyses included root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), radius of gyration (Rg), hydrogen bonding, and solvent-accessible surface area (SASA). Binding free energies were estimated using the molecular mechanics/generalized born surface area method.

A total of 57 targets for luteolin were identified from the Traditional Chinese Medicine Systems Pharmacology database, 72 from ETCM, and 100 from SwissTargetPrediction. After matching with Universal Protein Resource and removing duplicates, 191 unique targets were retained (Table S1, Supplemental Digital Content, https://links.lww.com/MD/R735 , Fig 2 ). (A) Venn diagram of luteolin targets from 3 databases. (B) Luteolin-target interaction network. Red diamonds = luteolin, ochre triangles = databases, green circles = genes. STP = SwissTargetPrediction. From the GeneCards database, 5055 PCOS-related targets and 1333 EM-related targets were retrieved. Sixty-nine overlapping targets were identified between luteolin’s targets and disease-associated genes (Table S2, Supplemental Digital Content, https://links.lww.com/MD/R735 , Fig 3 ), deemed as luteolin’s primary therapeutic targets. Venn diagram of intersecting targets for PCOS, EM, and luteolin. EM = endometriosis, PCOS = polycystic ovary syndrome. A PPI network of 69 nodes and 977 edges was generated using STRING. Topological analysis via CytoHubba identified the top 10 hub genes based on degree values: ACTB, AKT1, EGFR, ESR1, IL6, TNF, TP53, MMP9, ERBB2, and JUN (Fig 4 and Table 1 ). Detailed information of 10 targets. Core PPI network of luteolin targets for PCOS and EM. EM = endometriosis, PCOS = polycystic ovary syndrome, PPI = protein–protein interaction. Gene Expression Omnibus datasets GSE34526 / GSE5850 (PCOS) and GSE51981 / GSE7305 (EM) were corrected for batch effects (Fig. 5 A and B). ROC analysis demonstrated AUC values >0.600 for all 10 hub genes in both diseases, indicating diagnostic utility, particularly for PCOS (Fig. 5 C and D). (A and B) Batch correction of GEO datasets for PCOS and EM. (C and D) Diagnostic performance of hub genes in PCOS and EM. EM = endometriosis, GEO = Gene Expression Omnibus, PCOS = polycystic ovary syndrome. GO enrichment identified 2001 BP (e.g., regulation of smooth muscle cell proliferation, smooth muscle cell proliferation, and cellular response to chemical stress), 35 cellular components (e.g., membrane raft, membrane microdomain, and chromosomal region), and 149 MF (e.g., RNA polymerase II-specific DNA-binding transcription factor binding, DNA-binding transcription factor binding, and nuclear receptor activity) (Table S3, Supplemental Digital Content, https://links.lww.com/MD/R735 , Fig. 6 A and B). KEGG analysis highlighted 239 pathways, including endocrine resistance, prostate cancer, proteoglycans in cancer, EGFR tyrosine kinase inhibitor resistance, fluid shear stress and atherosclerosis, bladder cancer, lipid and atherosclerosis, etc (Table S4, Supplemental Digital Content, https://links.lww.com/MD/R735 , Fig. 6 C and D). Enrichment analysis of intersecting targets. (A) GO bubble plot (BP/CC/MF). (B) GO circular plot. (C) KEGG bar plot. (D) KEGG chord diagram. BP = biological processes, CC = cellular components, GO = gene ontology, KEGG = Kyoto encyclopedia of genes and genomes, MF = molecular functions. The binding potential of luteolin (ligand) to hub targets (receptors) intersecting with PCOS and EM genes was validated using molecular docking. Docking scores ranged from ‐9.7 to ‐5.6 kcal/mol (Table 2 ), confirming luteolin’s capacity to bind to these hub proteins under physiological conditions. PyMOL visualization highlighted specific interactions with ERBB2, ESR1, and MMP9 (Fig. 7 ). ERBB2: luteolin formed 3 hydrogen bonds with residues THR-2, ARG-413, and ASN-467 (green dashed lines) and one hydrophobic interaction with LEU-292 (yellow dashed line). ESR1: 6 hydrogen bonds were observed with residues LEU-346, THR-347, GLU-353, ARG-394, and LEU-525, along with 2 hydrophobic interactions involving ALA-350 and LEU-525. MMP9: luteolin established 3 hydrogen bonds with TYR-420, ARG-424, and THR-426, complemented by 2 hydrophobic interactions with VAL-398 and THR-426. These multiple hydrogen bonds and hydrophobic interactions suggest robust binding between luteolin and ERBB2, ESR1, and MMP9, likely modulating the structural and functional properties of these proteins. Molecular docking results for luteolin and hub targets. Molecular docking visualization. Hydrogen bonds (green dashed lines) and hydrophobic interactions (yellow dashed lines) between luteolin. Molecular dynamics simulations were performed on ESR1, MMP9, and ERBB2 complexes with luteolin, prioritized based on docking results. For ERBB2 and ESR1 (Fig. 8 A and B), RMSD fluctuations remained below 1 nm throughout simulations, indicating stable ligand–receptor binding. Minor RMSF fluctuations (0.5 nm for ERBB2 residues 300–320, 0.6 nm for ESR1 residues 330–340) occurred at peripheral protein regions, while other residues maintained stability (<1 nm). Rg values for ERBB2 (2.50 nm) and ESR1 (1.90 nm) remained constant, confirming overall protein stability. Hydrogen bonding between luteolin and ERBB2, ESR1 persisted at 1 to 2 and 2 to 3 bonds, respectively, with minimal fluctuations. SASA values (230 nm 2 for ERBB2, 130 nm 2 for ESR1) remained stable, further supporting protein conformational stability. For MMP9 (Fig. 8 C), initial RMSD fluctuations (0–30 ns) stabilized at ~1.2 nm after 50 ns, coinciding with RMSF peaks (>0.6 nm) and transient Rg shifts (30–50 ns), suggesting localized structural rearrangements. Despite these motions, hydrogen bonding remained stable (2–4 bonds), and SASA fluctuations at 30 ns indicated transient solvent exposure without compromising luteolin binding. Molecular dynamics trajectories. (A) ERBB2-luteolin. (B) ESR1-luteolin. (C) MMP9-luteolin. Figure 9 illustrates the free energy landscapes, conformational snapshots at 5 simulation time points (0, 25, 50, 75, and 100 ns), average binding free energies, and residue contributions for the 3 complexes. For ERBB2 and ESR1 complexes (Fig. 9 A and B), single energy minima clusters were observed in their free energy landscapes, indicating stable ligand–receptor interactions. The MMP9 complex (Fig. 9 C) exhibited 2 energy minima, suggesting a transition from an initial stable state to a more stabilized conformation through localized structural adjustments (a process that did not compromise luteolin binding [as corroborated by Fig. 8 C]). Conformational alignment across all time points confirmed consistent luteolin positioning relative to ERBB2, ESR1, and MMP9, underscoring binding stability. (A) Free energy landscape, conformational snapshots (0, 25, 50, 75, and 100 ns), average binding free energy, and key residue contributions for the ERBB2-luteolin complex. (B) Equivalent analyses for the ESR1-luteolin complex. (C) Equivalent analyses for the MMP9–luteolin complex. Average binding free energies (‐20.82, ‐38.2, and ‐30.21 kcal/mol for ERBB2, ESR1, and MMP9, respectively) indicated strong binding affinity, particularly for ESR1. Residue contribution analysis highlighted key interactions, GLY-412 and PRO-279 in ERBB2 (‐2.33 and ‐1.46 kcal/mol), GLU-353 in ESR1 (‐8.52 kcal/mol), and TYR-420/TYR-423/LEU-418/HIS-401 in MMP9 (‐2.41 to ‐1.58 kcal/mol). Overlapping residues in docking and simulation results confirmed conserved binding modes and robust complex stability.

This study systematically elucidated the multi-target mechanisms of luteolin in treating PCOS and EM using an integrated approach encompassing network pharmacology, PPI analysis, bioinformatics validation, and molecular dynamics simulations. Initially, through multi-database cross-validation, 191 potential luteolin targets were identified, with 69 overlapping significantly with PCOS (5055 disease targets) and EM (1333 disease targets), suggesting dual therapeutic effects mediated by shared target clusters. The PPI network revealed modular hubs centered on ACTB, AKT1, EGFR, ESR1, IL6, TNF, TP53, MMP9, ERBB2, and JUN (degree > 46), which were validated as diagnostic biomarkers (AUC > 0.600). Notably, ESR1, the top-ranked hub, underscores luteolin’s potential to modulate hormonal signaling pathways, aligning with its role in mitigating PCOS-related endocrine dysfunction and EM proliferation. [ 18 ] These findings establish a systems-level rationale for luteolin’s “multi-component, multi-target, multi-pathway” synergistic action, providing a mechanistic foundation for clinical translation. At the functional and pathway levels, this study further delineated the molecular regulatory network underlying luteolin’s therapeutic effects on PCOS and EM through GO/KEGG enrichment analyses. GO analysis revealed significant enrichment of 69 core targets in BP and MF critical to both diseases, including regulation of smooth muscle cell proliferation, membrane microdomain organization, and DNA-binding transcription factor activity. Notably, smooth muscle cell hyperplasia constitutes a pathognomonic feature of EM lesion formation, while membrane microdomain dysfunction has been implicated in ovarian granulosa cell insulin resistance in PCOS. [ 19 ] Furthermore, enrichment of nuclear receptor activity aligns with previous findings that luteolin suppresses cAMP response element-binding protein-mediated aromatase expression, thereby reducing estrogen biosynthesis in follicle-stimulating hormone-stimulated human ovarian granulosa cells (a mechanism directly relevant to hormonal dysregulation in both pathologies). [ 20 ] KEGG pathway analysis reinforced these insights, identifying 239 significantly enriched pathways with particular relevance to endocrine resistance, EGFR tyrosine kinase inhibitor resistance, and lipid metabolism disorders. Endocrine resistance, a cardinal feature of PCOS, likely mediates luteolin’s benefits on insulin sensitivity and ovarian function. [ 21 ] In EM, EGFR signaling drives lesion progression through pro-proliferative and pro-angiogenic mechanisms, with luteolin’s inhibitory effects on EGFR-PI3K/Akt axis activity offering a plausible mechanism for its anti-invasive properties. [ 22 ] The PI3K/Akt pathway, a central effector downstream of EGFR, plays pivotal roles in tumor cell survival, proliferation, and drug resistance through Akt-mediated phosphorylation events that suppress proapoptotic proteins, activate mTOR-driven protein synthesis, and modulate FOXO-dependent cell cycle arrest. [ 22 ] Experimental validation confirms that luteolin improves insulin sensitivity in ovarian granulosa cells via PI3K/AKT pathway inhibition while attenuating EM lesion progression by downregulating PI3K/Akt signaling cascades. [ 23 , 24 ] Enrichment of oncogenic pathways (e.g., prostate/bladder cancer) suggests shared proliferative molecular signatures between PCOS and EM, consistent with luteolin’s documented antiproliferative effects through AKT1 phosphorylation inhibition and EGFR signal blockade. [ 25 ] Notably, lipid metabolism pathway enrichment (e.g., lipid and atherosclerosis) underscores luteolin’s multi-target metabolic regulatory capacity, potentially involving modulation of peroxisome proliferator-activated receptor gamma (a master regulator of lipid homeostasis). These findings collectively position luteolin as a multifunctional agent capable of addressing the hormonal, proliferative, and metabolic dysregulation inherent to both diseases. [ 9 ] These findings not only validate network pharmacology-predicted targets like AKT1 and EGFR but also elucidate luteolin’s tripartite synergistic mechanism: estrogen reduction via aromatase inhibition; antiproliferative effects through PI3K/AKT and EGFR pathway blockade; and metabolic reprogramming via insulin receptor/peroxisome proliferator-activated receptor gamma axis modulation. This multi-pathway coordination provides a novel theoretical framework for developing natural product-based therapies targeting reproductive endocrine disorders. Molecular docking revealed high-affinity binding between luteolin and key targets including ERBB2, ESR1, and MMP9, with ESR1 demonstrating exceptional binding energy (‐9.0 kcal/mol). This aligns with the hormonal dysregulation pathognomonic to both diseases and positions ESR1 as a critical therapeutic node. Molecular dynamics simulations further confirmed binding stability, with ERBB2/ESR1 complexes maintaining conformational integrity and persistent hydrogen bonding (1–2 and 2–3 bonds, respectively). The GLU-353 residue in ESR1 (‐8.52 kcal/mol) emerged as a pivotal regulatory site. For MMP9, transient RMSD fluctuations (30–50 ns) preceded stabilization into dual energy minima, indicating a dynamic adaptation mechanism where luteolin induces conformational reorganization to enhance binding. This may explain its dual regulatory effects on MMP9 (suppressing pathological activation while preserving physiological functions). [ 26 ] These structural insights demonstrate the power of computational approaches in elucidating luteolin’s multi-target therapeutic potential, providing critical mechanistic evidence at the interface of systems biology and structural pharmacology. These findings align with existing literature highlighting MMP9’s pivotal role in extracellular matrix remodeling and fibrosis. Luteolin’s potential to mitigate PCOS ovarian fibrosis and EM progression likely stems from MMP9 inhibition, as evidenced by its downregulation of MMP9 expression via PI3K/AKT pathway suppression in colorectal cancer models (a mechanism potentially applicable to PCOS given the pathway’s involvement in ovarian stromal remodeling). [ 23 , 27 ] In oral squamous cell carcinoma, MMP9 overexpression correlates with fibrotic-to-malignant transition, and luteolin’s high-affinity binding (as confirmed by molecular docking) suggests direct inhibitory effects on MMP9-mediated invasive growth in EM. [ 28 , 29 ] Consistent with our molecular docking results, luteolin exhibits high binding affinity with MMP9, suggesting its potential to directly inhibit MMP9 activity. Collectively, luteolin may mitigate ovarian fibrosis in PCOS and endometriotic progression by suppressing MMP9 expression/activity, thereby blocking excessive extracellular matrix degradation and fibrosis. Notably, ESR1 (estrogen receptor α) is frequently hyperactivated in PCOS due to estrogenic imbalance, contributing to hormonal dysregulation and ectopic endometrial proliferation. Luteolin likely restores estrogenic signaling homeostasis by modulating ESR1 activity. Further mechanistic insights reveal that luteolin abrogates the MKK3/6-p38 MAPK–cAMP response element-binding protein signaling axis through inhibition of tumor progression locus 2, leading to reduced CYP19A1 (aromatase) expression and diminished estrogen biosynthesis. This pathway effectively ameliorates hyperandrogenism and ovarian dysfunction in PCOS mouse models. [ 9 , 20 ] In parallel with breast cancer contexts where ESR1 overexpression correlates with therapeutic resistance, luteolin downregulates ESR1 expression and suppresses ESR1-mediated cellular proliferation (a mechanism applicable to hormone-sensitive tissues in PCOS patients). [ 30 , 31 ] Consequently, luteolin may restore endocrine balance in PCOS by attenuating ESR1-driven estrogenic hyperactivation while curbing EM-associated ectopic endometrial proliferation. Despite these insights, limitations persist: network pharmacology remains database-dependent, potentially omitting low-abundance or novel targets; static structural modeling neglects physiological dynamics (e.g., posttranslational modifications, protein interaction network fluctuations); and molecular mechanisms require validation through in vivo/in vitro pharmacodynamic and safety assessments. Future directions include CRISPR-engineered disease models, metabolomic pharmacokinetic profiling, and multicenter clinical trials to fully evaluate translational potential.

The authors would like to thank all authors of references.