All animal experiments were conducted in strict accordance with international guidelines for animal care and use and received approval from the Quanzhou Medical College Experimental Animal Ethics Committee (Approval Number 2024049). This investigation utilized a total of 18 female rats (age: 6 weeks) of the Sprague-Dawley (SD) strain. These animals were supplied by Sipeifu (Beijing) Biotechnology Co., Ltd., which operates under the animal license SCXK(Jing) 2024-0001. Following a 5-day acclimatization period, a randomization process was utilized to allocate the subjects into three distinct experimental arms, with each arm comprising six animals: a control group receiving vehicle, a positive control group treated with 0.2 mg/kg Ethinylestradiol/Cyproterone acetate and 0.23 g/kg Metformin, and the GSW group, which was administered Gui Shen Wan. The formula consists of Cuscutae Semen (Tu Si Zi), Eucommiae Cortex (Du Zhong), Lycii Fructus (Gou Qi Zi), Angelicae Sinensis Radix (Dang Gui), Corni Fructus (Shan Zhu Yu), Rehmanniae Radix Praeparata (Shu Di Huang), Dioscoreae Rhizoma (Shan Yao), and Poria (Fu Ling). The final aqueous extract was concentrated to 1 g crude drug/mL. It was administered at a dosage of 0.1 mL/100g, prepared by the Affiliated Hospital of Jiangxi University of Traditional Chinese Medicine. All treatments were administered once daily via oral gavage for seven days. One hour after the final administration, all rats were anesthetized, and blood samples were collected under sterile conditions from the abdominal aorta for subsequent analysis.

To isolate the serum, collected blood samples were initially allowed to clot at room temperature stand for 3 hours. Following this, the specimens were subjected to a 15-minute centrifugation step, applying a rotational speed of 3,000 rpm. The serum collected from rats within the same experimental group was then pooled and processed. The protocol required a 30-minute incubation at 56 °C within a water bath to achieve heat inactivation. Directly afterward, the serum underwent sterile filtration by passing it through a 0.22 µm micropore filter. The resulting medicated serum was then dispensed into aliquots and stored at -80 °C. Medicated serum from the ethinylestradiol/cyproterone acetate plus metformin group was used as a pharmacological reference reflecting commonly used clinical management for PCOS-associated endocrine and metabolic dysregulation.

To analyze the bioactive components in the GSW-medicated serum, sample preparation was first conducted. This involved precipitating serum proteins with an acetonitrile/methanol solution, after which the supernatant was collected by centrifugation. This analysis was performed on a UPLC-MS/MS system (ExionLC™ AD, SCIEX). Chromatographic separation was carried out on an Agilent SB-C18 column (1.8 µm, 2.1 mm × 100 mm), utilizing a gradient elution. The mobile phase was composed of 0.1% formic acid in water and acetonitrile. For detection, the mass spectrometry analysis was operated in both positive and negative electrospray ionization (ESI) modes. Metabolite identification was based on fragmentation patterns matched against a self-built database, and quantification was performed in Multiple Reaction Monitoring (MRM) mode by integrating chromatographic peak areas.

To elucidate GSW’s underlying mechanisms against PCOS, a strategy based on network pharmacology was employed. Specifically, the bioactive components identified in the GSW-medicated serum via UPLC-MS/MS served as the candidate library for this analysis. Initially, candidate active constituents were screened using pharmacokinetic parameters, specifically oral bioavailability (OB) ≥ 30% and a drug-likeness (DL) score ≥ 0.18. Putative targets for the selected compounds were subsequently predicted via the TCMSP, SwissTargetPrediction, and SymMap platforms. All retrieved target proteins underwent normalization against the UniProt knowledgebase. Concurrently, targets related to PCOS pathogenesis were aggregated from the GeneCards and OMIM resources. GSW’s primary therapeutic targets for PCOS were pinpointed by intersecting the set of drug-related targets with the disease-specific targets. An integrated “drug-component-target” map was thereafter generated and graphically rendered using Cytoscape software (v3.7.2).

For a deeper analysis of the interplay among these core targets, a protein-protein interaction (PPI) network was constructed utilizing information from the STRING database (requiring an interaction score > 0.4). Central (hub) genes within this network were pinpointed via the CytoHubba plugin, integrating the consensus results from five distinct algorithms (MCC, MNC, Degree, EPC, and BottleNeck). Finally, to investigate the pertinent biological functions and signaling cascades, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment evaluations were performed. These functional annotations were carried out using the DAVID 6.8 platform, with a P-value < 0.05 set as the criterion for statistical significance. Although these constituents were detected in medicated serum, OB/DL thresholds were applied as a standardized prioritization strategy in network pharmacology; this may bias toward compounds with higher predicted drug-like properties.

The human granulosa cell line KGN was cultured in DMEM/F-12 medium (Gibco, Cat. No. 11320033) supplemented with 10% fetal bovine serum (FBS; Gibco, 10099141) and 1% penicillin-streptomycin (Gibco, 15140122). The cells were maintained in a humidified incubator (Thermo Fisher Scientific, Heracell™ VIOS 160i) at 37°C with an atmosphere of 5% CO₂. To establish an in vitro PCOS model, KGN cells were seeded into culture plates and grown to 70-80% confluency. Subsequently, the culture medium was replaced with a medium containing 1 µmol/L dexamethasone (Sigma-Aldrich, D4902) and 100 nmol/L insulin (Sigma-Aldrich, I9278), and the cells were incubated for an additional 48 hours to induce a PCOS-like cellular phenotype before subsequent treatments. This dexamethasone- and insulin-induced system was used to model key PCOS-relevant cellular features (inflammation, impaired PI3K/Akt activation, apoptosis, and steroidogenic dysregulation) under metabolic stress, rather than to recapitulate the full systemic endocrine–immune phenotype of PCOS. Unlike other ovarian cell lines, KGN cells retain key physiological granulosa cell characteristics, including functional follicle-stimulating hormone receptors and aromatase activity, making them a valid model for investigating steroidogenic and inflammatory responses (22, 23). Model establishment was verified by increased apoptosis and inflammatory cytokine production, together with reduced PI3K/Akt phosphorylation in the Model group, as shown in the Results.

For the primary efficacy assessment, cells were divided into six groups: a control group treated with blank serum (Control); a model group treated with model serum (Model); a positive control group treated with Ethinylestradiol/Cyproterone acetate + Metformin-medicated serum (Ethinylestradiol/Cyproterone acetate + Metformin); and three experimental groups treated with low (5%), medium (10%), and high (20%) doses of GSW-medicated serum (Low-dose, Mid-dose, and High-dose), respectively. To investigate the causal relationship between inflammation and PI3K/Akt signaling, a mechanism-rescue experiment was performed with the following groups: Control, Model, Model + 20 ng/mL recombinant human TNF-α (TNF-α, PeproTech, 300-01A), Model cells + high-dose GSW serum (GSW), and Model cells + high-dose GSW serum + recombinant human TNF-α (GSW + TNF-α). For the validation of key active components, model cells treated with 15 μM embelin (Low-embelin), 30 μM embelin (High-embelin), 10 μM nobiletin (Low-nobiletin), 20 μM nobiletin (High-nobiletin) or a combination of 15 μM embelin and 10 μM nobiletin (embelin+nobiletin). The concentrations of these monomers were determined based on previous studies (24, 25). Embelin and nobiletin were procured from MedChemExpress under the product codes HY-N0115 and HY-N0145, respectively. For monomer treatments, compounds were dissolved in DMSO, and vehicle controls containing the same final DMSO concentration were included in all experiments.

The evaluation of cellular viability was conducted using the Cell Counting Kit-8 (CCK-8) assay (Dojindo, CK04). Initially, KGN cells were dispensed into 96-well plates, with each well containing 3.5 × 10³ cells suspended in 100 µL of medium. This was followed by a 24-hour incubation period, allowing cells to adhere and reach approximately 60% confluence. Subsequent to this, the cultures were exposed to differing concentrations of medicated serum for an additional 24-hour duration. Upon completion of the treatment incubation, every well received 10 µL of freshly diluted CCK-8 solution. A final 2-hour incubation of the plates was performed at 37°C, ensuring they were protected from light. Finally, to quantify cell viability, the optical density was recorded at 450 nm utilizing a BIOBASE-EL10A microplate reader.

For every sample, an aliquot of approximately 1 × 10⁶ cells was gathered. The cells underwent two washing cycles with ice-cold PBS (Gibco, 10010023) and were subsequently pelleted via centrifugation (1500 rpm, 3 min). The resulting cellular pellet was resuspended within 300 µL of cold 1× Binding Buffer. A staining solution, consisting of 5 µL of Annexin V-FITC and 10 µL of Propidium Iodide (PI), was then introduced to the cell suspension. After gentle mixing, the mixture was incubated for 10 minutes at ambient temperature while shielded from light. In the final preparation step, an additional 200 µL of cold 1× Binding Buffer was dispensed into each sample. Prompt analysis was then performed by flow cytometry using a FACSCanto II instrument (BD Biosciences).

The concentrations of secreted hormones and cytokines in the cell culture supernatant were quantified using commercial ELISA kits. Briefly, following the treatments, supernatants from each group were collected and centrifuged at 1000 × g for 20 minutes to remove cell debris. The levels of Estradiol (RuixinBio, RXJ106056H), LH (RuixinBio, RX106050H), FSH (RuixinBio, RX106046H), Testosterone (RuixinBio, JRX106316), TNF-α (RuixinBio, RX1600891C), and IL-6 (RuixinBio, RXQ96) were then determined. All assays were performed strictly in accordance with the manufacturer’s protocols.

To perform immunoblot analysis, total protein was harvested from KGN cells utilizing RIPA lysis buffer (Solarbio, R0020) supplemented with a protease inhibitor (PMSF, Sigma-Aldrich, P7626). Determination of the protein content within the lysates was accomplished via a BCA protein assay kit (Biosharp, BL521A). Identical quantities of protein from every sample were heat-denatured, resolved by 8%-12% SDS-PAGE, and thereafter electro-transferred onto PVDF membranes (Merck Millipore, IPVH00010). Non-specific binding sites on these membranes were blocked at ambient temperature for one hour using 5% non-fat milk dissolved in TBST. Following blocking, the membranes were probed overnight at 4 °C with the following primary antibodies: anti-PI3K (Bioss, BS-10657R, 1:800), anti-p-PI3K (CST, 17366, 1:800), anti-Akt (CST, 9279, 1:800), anti-p-Akt (CST, 4060, 1:1000), anti-PGR (Abcam, AB60954, 1:800), and anti-β-actin (Proteintech, 20536-1-AP, 1:5000). After washing, the membranes were incubated for 1 hour at room temperature with the appropriate HRP-conjugated secondary antibodies (1:10000, Proteintech). Finally, immunoreactive protein bands were visualized with an ECL detection system (Meilunbio, MA0186). Densitometric analysis was conducted using ImageJ, with β-actin serving as the internal loading control. Phosphorylation levels were expressed as the ratio of p-PI3K to total PI3K and p-Akt to total Akt, normalized to β-actin where appropriate.

The initial step involved extracting total RNA from KGN cells with the TransZol Up Plus RNA Kit (TransGen Biotech, ER501-01). Following quantification, reverse transcription was conducted using TUAFS SuperMix (TransGen Biotech, AU341); this kit notably includes a procedure for genomic DNA (gDNA) elimination. This newly synthesized cDNA served as the template for quantitative RT-PCR analysis. The reactions were executed on a CFX Connect System (Bio-Rad) utilizing the MagicSYBR Mixture (CWBIO, CW3008). The cycling protocol was programmed for an initial 10-minute denaturation at 95°C, preceding 40 amplification cycles (10 s at 95°C, 30 s at 60°C, and 30 s at 72°C). All oligonucleotide primers, which are detailed in Table 1, were commercially produced by Sangon Biotech (Shanghai, China). Calculation of relative mRNA abundance was achieved via the 2^−ΔΔCt method, wherein GAPDH served as the endogenous control for normalization.

All quantitative data are expressed as the mean ± standard deviation (SD), derived from a minimum of three independent experiments. The statistical analyses were carried out using GraphPad Prism software (version 9.0, GraphPad Software, USA). Prior to analysis, data normality and homogeneity of variance were verified using the Shapiro-Wilk test and Levene’s test, respectively. A one-way Analysis of Variance (ANOVA) followed by Tukey’s post-hoc test was applied for comparisons between multiple groups. A p-value less than 0.05 was set as the threshold for statistical significance.

To elucidate the key therapeutic targets of GSW in treating PCOS, we first analyzed its serum-absorbed components (Supplementary Material S1). 1477 potential drug targets for these components were predicted using the SwissTargetPrediction, TCMSP, and SymMap databases. This prediction was screened using pharmacokinetic parameters, requiring an OB of ≥ 30% and a DL score ≥ 0.18 (Figure 1A).

To illustrate these interactions, a “drug-component-target” map was generated utilizing the Cytoscape software (version 3.7.2). The network included 399 nodes and 3122 edges, with diamond, rectangular, and circular nodes representing the drug, its components, and the targets, respectively. The top five active components, ranked by degree, were embelin, sesamin, nobiletin, dopamine, and histamine. Notably, embelin, sesamin, and nobiletin were associated with the highest number of potential targets, suggesting that these multi-target compounds may play a pivotal role in the therapeutic mechanism of GSW (Figure 1B).

To analyze the interplay among these potential targets, a PPI network was constructed for the 272 common targets by querying the STRING database (interaction score > 0.4). This analysis produced a network of 271 nodes and 6863 edges (Figure 1C). Hub genes were then obtained from this network by the CytoHubba plugin, which employed five distinct algorithms (MCC, MNC, Degree, EPC, and BottleNeck). A consensus analysis of the leading genes identified by all algorithms yielded four central hub nodes: IL-6, TNF, INS, and GAPDH (Figures 1D–I). GAPDH emerged as a hub node likely due to its central role in cellular metabolism and its extensive interactions in protein–protein interaction networks, rather than being interpreted as a specific therapeutic target.

To elucidate the biological roles and molecular cascades that GSW modulates relative to PCOS, GO and KEGG functional enrichment evaluations were conducted on the 272 shared targets, utilizing the DAVID platform. Applying a statistical cutoff of P < 0.05, the Gene Ontology annotation identified 3066 BP items, 113 CC items, and 261 MF items. Within the BP classification, these shared targets showed notable enrichment in functions including the inflammatory response, drug response, response to lipopolysaccharide, steroid metabolism, and the positive control of cell proliferation, suggesting that GSW’s therapeutic effects are closely related to the modulation of inflammation, hormone metabolism, and cell growth status. In the CC category, target proteins were predominantly localized to the extracellular space, plasma membrane, and cytosol, indicating that GSW’s targets are widely distributed, encompassing cell surface receptors, extracellular signaling molecules, and intracellular proteins. In the MF category, the targets were primarily associated with cytokine activity, enzyme binding, and steroid binding, confirming their involvement in complex protein-protein interactions and signal transduction (Figure 2A).

The KEGG pathway enrichment evaluation identified 170 pathways showing significant enrichment (P < 0.05), with the leading ten being visualized in a bubble plot. These findings revealed that the common targets were significantly enriched in multiple cascades highly relevant to PCOS pathophysiology. Notably, the Lipid and atherosclerosis pathway was the most highly enriched, consistent with the dyslipidemia commonly associated with PCOS. Furthermore, the IL-17 and TNF signaling pathway were significantly enriched, directly indicating that inflammation is a key aspect of GSW’s mechanism of action. The PI3K-Akt signaling pathway, a critical intracellular hub whose dysfunction is directly linked to insulin resistance in PCOS, was also prominently identified. Other notable enriched pathways included the HIF-1 signaling pathway, Apoptosis, and MAPK signaling pathway (Figure 2B).

Collectively, these enrichment analyses suggest that the core mechanism of GSW in treating PCOS involves its broad anti-inflammatory effects and the modulation of key signaling pathways, particularly the PI3K-Akt and TNF pathways, which are closely associated with inflammation, cell metabolism, and survival.

To validate the therapeutic effects of GSW on PCOS, we treated the in vitro cell model with GSW-medicated serum. Flow cytometry analysis revealed that the induction of the PCOS model significantly increased the cellular apoptosis rate compared to the control group. Treatment with serum from GSW-administered rats, however, dose-dependently attenuated this pro-apoptotic effect (Figures 3A, B). Consistent with these findings, the CCK-8 assay demonstrated that the GSW-medicated serum significantly enhanced the viability of the PCOS model cells in a dose-dependent manner (Figure 3C). Furthermore, to elucidate the impact of GSW on hormone secretion, the levels of estradiol, LH, testosterone, and FSH were measured. These data indicated a significant hormonal imbalance in the model group compared to the control. Treatment with GSW-medicated serum led to a dose-dependent decrease in the secretion of estradiol, LH, and testosterone, while concurrently causing a dose-dependent increase in FSH secretion, suggesting that GSW-medicated serum partially restored the steroidogenic/hormone secretion profile in the granulosa cell PCOS-like model (Figures 3D–G). Notably, measurements of hormones/cytokines in the culture supernatant reflect in vitro outputs under the current experimental conditions and should not be extrapolated to systemic endocrine regulation in vivo.

Based on the network pharmacology analysis, we investigated the effects of GSW-medicated serum on inflammation and the PI3K/Akt signaling pathway in PCOS model. The results showed a potent anti-inflammatory effect that GSW-medicated serum treatment caused a dose-dependent reduction in both the protein secretion and mRNA expression levels of IL-6 and TNF-α when compared to the model group (Figures 4A–D). We then evaluated the impact on the PI3K/Akt signaling cascade. Western blot analysis indicated that GSW-medicated serum, in a dose-dependent manner, significantly increased the phosphorylation levels of both PI3K and Akt. We also assessed the expression of PGR, a critical protein for ovulation. Protein expression of PGR was significantly downregulated in the model group, but GSW-medicated serum treatment effectively restored its levels in a dose-dependent manner (Figures 4E–J).

To further validate the relationship between TNF-α and the PI3K/Akt pathway during GSW treatment, a rescue experiment was conducted using recombinant human TNF-α to interfere with the therapeutic effects of GSW. Flow cytometry analysis revealed that the administration of TNF-α alone exacerbated apoptosis in the PCOS model cells. Critically, when cells were co-treated with GSW and TNF-α, the protective anti-apoptotic effect of GSW was significantly reversed (Figures 5A, B).

Western blot analysis further delineated the molecular link between TNF-α and the PI3K/Akt pathway. The phosphorylation levels of PI3K and Akt were significantly downregulated in the model group compared to the control, and this suppression was further exacerbated by the addition of TNF-α. While GSW treatment markedly upregulated the phosphorylation of PI3K and Akt, this activated signaling effect was significantly attenuated by co-incubation with TNF-α (Figures 5C–G). The recombinant TNF-α rescue experiment was designed to specifically test the causal relationship between inflammatory suppression and PI3K/Akt activation, rather than to comprehensively map all upstream inflammatory mediators. Taken together, these results strongly suggest that the therapeutic activation of the PI3K/Akt pathway by GSW is dependent on its ability to suppress TNF-α, highlighting TNF-α as a critical therapeutic target of GSW in PCOS.

To further identify the key bioactive components of GSW, we treated the PCOS model cells with the top five active ingredients predicted by network pharmacology: embelin, sesamin, nobiletin, dopamine, and histamine. These candidates were selected because they were not only predicted as top-ranking compounds by network pharmacology but were also confirmed to be present in the GSW-medicated serum via UPLC-MS/MS analysis. ELISA results revealed that embelin and nobiletin significantly reduced the secretion of TNF-α, while sesamin showed a modest inhibitory effect. In contrast, dopamine had no significant effect, and histamine slightly increased TNF-α secretion (Figure 6A).

Based on these findings, the effects of embelin and nobiletin were investigated further. Western blot analysis demonstrated that both embelin, as well as nobiletin, stimulated the phosphorylation of both PI3K and Akt. This effect was observed in a dose-dependent fashion, as high concentrations yielded a more potent response than low concentrations. Notably, the combined application of low-dose embelin and low-dose nobiletin resulted in a significantly more potent activation of the PI3K/Akt pathway compared to the effect of either compound at low dose level. These results suggest that embelin and nobiletin are key components responsible for the anti-PCOS action of GSW and that they function cooperatively to exert an enhanced effect, consistent with the multi-component therapeutic theory of TCM (Figures 6B–F).