Xianyu Capsule (XYC, Lot #20220401) was obtained from Xi’an Chiho Pharmaceutical Co., Ltd. (Xi’an, China). Pilocarpine hydrochloride (#HY-B0726) was purchased from MCE (Shanghai, China). Malondialdehyde (MDA, #A003-1) and Superoxide dismutase (SOD, #A001-3) assay kits were purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China). TUNEL BrightRed Apoptosis Detection Kit (#A113-02) were obtained from Vazyme Biotech Co., Ltd. (Nanjing, China). Akt1-Specific Recombinant antibody (#80816-1-RR) and Alpha Tubulin polyclonal antibody (#11224-1-AP) were acquired from Proteintech (Wuhan, China). ELISA assay kit for IL-6 (#E-EL-R0015), TNF-α (#E-EL-R2856), IL-1β (#E-EL-R0012) were obtained from Elabscience Biotechnology Co., Ltd. (Wuhan, China).

XYC constituents were analyzed with UPLC-Orbitrap-MS as previously described (33). Chromatography on Waters HSS T3 column (100 × 2.1 mm, 1.8 μm). High-resolution mass spectrometric data were acquired using Q Exactive HFX Hybrid Quadrupole-Orbitrap instrument (Thermo Fisher Scientific) equipped with a heated ESI source and operating in Full-ms/ddMS² mode.

SD rats (180–220 g) were purchased from SLAC Laboratory Animal Co., Ltd. (Shanghai, China), and housed under specific pathogen-free (SPF) conditions at the laboratory animal center of the First Affiliated Hospital of Zhengzhou University, with 22 ± 2°C temperature, 50% humidity, 12 h light/dark cycle, and ad libitum access to food and water. According to clinical application, the dosage of XYC and CBZ is 6.6 g and 1.2 g per person per day, respectively. Based on the body surface area normalization from rodent-to-human equivalent dose conversion, the equivalent dose for rat is 0.70 g/kg and 0.125 g/kg per day. Thirty rats were randomly assigned to five groups (Control, Model, CBZ, XYC 0.35 g/kg and 0.70 g/kg group) using a computer-generated block randomization scheme (block size = 6) to ensure balanced distribution of body weights and baseline metabolic parameters. Randomization was performed by an independent researcher not involved in data collection, and group allocation was concealed until interventions began. After 7 days of adaptive feeding, the rats were administered with CBZ or XYC by gavage for 14 days. The dosage of CBZ was 0.125 g/kg, the high dosage of XYC was 0.70 g/kg, and the low dosage was 0.35 g/kg. On the 15th day, epilepsy was induced as previously described (34, 35). Following an initial intraperitoneal (i.p.) administration of lithium chloride (3 mmol/kg), rats received pilocarpine (35.3 mg/kg, i.p.) 20 h post-injection. Seizures manifested 15–35 min post-induction and were quantified over 30 min using the established modified Racine scale (36). The seizures were terminated with 10% chloral hydrate saline solution 1 h after the seizures. The epileptic rats were housed with food and water available ad libitum. After 7 days of continuous intragastric administration, the rats were sacrificed, blood was collected for metabonomic detection. The cecal contents were collected for 16S rDNA sequencing. The hippocampal tissues were collected for the determination of MDA, SOD and pathological observation. Throughout the experiment, investigator blinding and outcome assessor blinding was implemented. Ethical approval (2023-KY-1341) was granted by the Research Ethics Committee at Zhengzhou University First Hospital, with informed consent obtained following NIH guidelines for animal welfare.

The catalase activity analysis was added here: “Catalase (CAT) activity was quantified by measuring the absorbance at 405 nm of a yellow-colored complex formed between residual H₂O₂ and ammonium molybdate, which reflects the remaining H₂O₂ after catalase-mediated decomposition.”

The serum concentrations of proinflammatory cytokines (IL-6, TNF-α, and IL-1β) were determined via ELISA according to manufacturer protocols. Absorbance measurements at 450 nm were performed on a Varioskan LUX multimode microplate reader.

Following intervention, hippocampal specimens were immediately immersion-fixed in 4% paraformaldehyde. Gradient ethanol dehydration preceded paraffin embedding. Serial coronal sections (5 μm thickness) were obtained for subsequent histological processing, including hematoxylin & eosin (H&E) and cresyl violet-based Nissl staining (37). Histopathological scoring was performed by two independent pathologists blinded to treatment conditions.

The isolation of total DNA from intestinal microbiota, followed PCR reaction and sequencing were performed as described before (38). RDP classifier¹ was used to annotate the representative sequences after clustering, and a threshold of 70% was set against the Silva database (SSU128/16 s bacteria). Using the R language tool to draw the community histogram, we determined the structural composition of different groups of the bacterial community at the phylum and generic levels.

For metabolomics analysis, serum were collected after treatment. The chromatographic conditions in UHPLC system and the QExactive high-resolution system for mass spectrometry systems were conducted as previously described (39, 40).

Briefly, Serum were vortexed with methanol/acetonitrile (1:1, v/v), centrifuged (4°C 14,000 × g for 20 min), and the resultant supernatant lyophilized before reconstitution in acetonitrile/water (1:1, v/v).

UHPLC system (Vanquish, Thermo) with HILIC column were used. Mobile phase: A = 25 mM ammonium acetate/hydroxide, B = acetonitrile. Gradient: 98% B (1.5 min) to 2% B (10.5 min), held 2 min, rapidly restored to 98% B (0.1 min) with 3 min re-equilibration in both ESI modes.

Processed raw MS data (ProteoWizard) using XCMS with centWave (peakwidth = 10–60, m/z = 10 ppm) and grouping parameters (mzwid = 0.025, bw = 5). CAMERA annotated isotopes/adducts. Metabolites were identified via m/z (<10 ppm) and MS/MS matching against an authentic standards database. Normalized data underwent multivariate analysis (Pareto-scaled PCA/OPLS-DA) via ropls. Model robustness was verified by 7-fold cross-validation/permutation tests.

Bioactive constituents of XYC’s constituent herbs were systematically curated from the Traditional Chinese Medicine Systems Pharmacology (TCMSP, https://tcmspw.com/tcmsp.php) or BATMN-TCM² using the keywords “Hedysarum Multijugum Maxim.,” “Codonopsis Radix,” “Radix Salviae,” “Radix Bupleuri,” “Ziziphi Spinosae Semen,” “Polygala tenuifolia Willd.,” “Rhizom Gastrodiae,” “Uncariae Ramulus Cumuncis,” “Acoritataninowii Rhizoma,” “Arisaema Cum Bile,” “Angelicae Sinensis Radix,” “Bombyx Batryticatus,” “Massa Medicata Fermentata,” “Curcumae Radix,” “licorice” and “Typhonii Rhizoma.” Then screen the retrieved compounds by ADME parameters (oral bioavailability (OB) ≥ 30%, drug-likeness (DL) ≥ 0.18) (41). Epilepsy-related disease targets were obtained from OMIM,³ DisGeNET,⁴ and GeneCards databases.⁵ Then XYC-epilepsy-related target screening and PPI network construction were obtained as previously described (42).

Total RNA from hippocampal tissues were isolated using TRIzol and reversed into cDNA. RT-qPCR was conducted using SYBR qPCR Master Mix (#A57155, Applied Biosystems, MA, United States). Primers sequences were listed in Table 1.

Hippocampal protein was extracted using ice-cold RIPA lysis buffer (#R0010, Solarbio) supplemented with 1% protease/phosphatase inhibitor cocktail (#P1260, Solarbio, Beijing, China), and then separated with SDS-PAGE, and PVDF membranes were used for the incubation of primary antibodies followed by secondary antibodies, and ECL substrate for visualizing the results.

Statistical analyses were conducted with SPSS Statistics (v19.0, IBM Corp.) using the mean ± standard error of mean (SEM). The significance between three or more groups was determined by one-way ANOVA, with p < 0.05 considered statistically significant.

The chemical constituents of XYC were identified by UPLC-MS analysis. Our study revealed that XYC contains abundant constituents, including organic oxygen compounds, phenylpropanoids and polyketides, benzenoids, organoheterocyclic compounds and lipids and lipid-like molecules. A total of 904 compounds were identified, the positive and negative ion flow chromatography were shown in Figure 1 and the top ten anionic and cationic constituents, ranked by peak area, were presented in Table 2.

To delineate XYC’s anti-epileptic pharmacodynamics, we established a lithium-pilocarpine rat model of acute epileptogenesis. A schematic overview of the treatment regimen and experimental workflow is presented in Figure 2A. Treatment with CBZ and XYC significantly prolonged the seizure latency and reduced the seizure frequency induced by pilocarpine in rat models of epilepsy (Table 3). H&E staining results found that, the hippocampal CA1 region tissues in the control group exhibited normal cellular morphology with neatly arranged cells, abundant cytoplasm, and large, round nuclei. In contrast, the model group showed a noticeable reduction in cell count, accompanied by a loose and disorganized arrangement. The cytoplasm appeared deeply stained, with evident nuclear dissolution and fragmentation, along with vacuolated manifestations. Both the CBZ and XYC groups demonstrated improved hippocampal cellular characteristics compared to the model group. Specifically, these groups exhibited increased cell density, gradual restoration of normal cellular morphology, and better-organized cellular alignment. Furthermore, the pathological manifestations of nuclear fragmentation and pyknosis were significantly alleviated (Figure 2B).

As shown in Figure 2C, Nissl staining revealed morphological changes of hippocampal CA1 neurons. Neurons in control group showing neatly arranged with intact morphology and clearly discernible Nissl bodies. While, model group exhibited a significant neuronal loss accompanied by disorganized cellular alignment and enlarged intercellular spaces. However, the CBZ and XYC treatment groups exhibited partial preservation of neuronal architecture in the hippocampal CA1 region, though with distinct pathological features. While neurons in these groups maintained relatively intact cellular morphology with preserved structural integrity, mild disorganization in cellular alignment was observed. Notably, Nissl body staining intensity demonstrated moderate attenuation, coupled with discernible enlargement of intercellular spaces. Despite these morphological alterations, the majority of neurons retained detectable viability and functional capacity. Quantitative analysis of TUNEL-positive cells combined with nuclear counterstaining revealed that both CBZ and XYC treatments significantly reduced apoptosis in CA1 neurons compared to the Model group (Figure 2D). Furthermore, the H&E staining of the liver and kidneys of each group were observed, and found that CBZ and XYC treatment did not changed the tissue morphology and showing no toxicity on liver and kidney (Figure 2E). These consistent findings across multiple staining techniques demonstrate that XYC pretreatment effectively mitigates seizure-induced neuronal injury in the hippocampal CA1 region.

Neuroinflammation and oxidative stress represent pivotal mechanisms underlying epileptogenesis, with their pathophysiological contributions being linked to elevated disease susceptibility (43). Consequently, interventions that targeting these dual pathological processes may provide effective therapeutic strategies for mitigating epileptogenesis. As shown in Figure 3, biochemical analyses revealed that epilepsy induction significantly elevated MDA levels while reducing SOD activity and CAT content in hippocampal tissues, indicating pronounced oxidative stress. Both CBZ and XYC treatments effectively attenuated these pathological changes, demonstrating significant reductions in MDA content and restoration of SOD activity and CAT contents compared to untreated epileptic controls (Figure 3A). Furthermore, ELISA assay found that IL-6, TNF-α and IL-1β pro-inflammation cytokine levels were significantly increased in epileptic rats, while both CBZ and XYC treatments effectively reversed this upward trend (Figure 3B). These findings demonstrate that XYC treatment not only induced anti-oxidant but also anti-inflammatory properties in epilepsy rats.

Emerging research has revealed that sophisticated interactions between gut microbiota and cerebral functions are mediated through an integrated network of cellular signaling mechanisms and neural communication circuits, collectively termed the “bidirectional microbiota-gut-brain axis” (44). Elucidating the microbiota’s dual role in both epileptogenesis processes and anticonvulsant therapeutic interventions could fundamentally advance our understanding of the neurobiological foundations underlying seizure disorders.

To elucidate the impact of XYC on intestinal homeostasis, 16S rDNA were amplified followed by bioinformatic analysis. Alpha diversity analysis (Chao1 index, Shannon index, and Observed OTUs) demonstrated significant microbial community changes following epilepsy induction and XYC treatment. While rank abundance curves and diversity indices showed comparable bacterial composition patterns among Control, Model, and XYC 0.7 groups (Figures 4A–C), quantitative analysis revealed distinct differences. The Model group exhibited significantly reduced species richness (Chao1 index, p < 0.05) compared to controls, consistent with established microbiota depletion in epilepsy (45). XYC treatment effectively restored microbial diversity, significantly increasing both the Chao1 index and Shannon diversity (p < 0.05 versus Model group, Figure 4D). Beta diversity analysis using PLS-DA demonstrated clear separation of gut microbiota communities among the three groups (Figure 4E). Venn diagram analysis at the OTU level identified 2067, 1,591, and 1972 operational taxonomic units in Control, Model, and XYC groups respectively, revealing both shared and group-specific microbial signatures (Figure 4F).

Microbiome profiling demonstrated marked taxonomic restructuring of gut microbial architecture across phylum and genus hierarchies post-epileptogenesis and XYC intervention. At the phylum level, epileptic rats exhibited marked reductions in Tenericutes, Patescibacteria, Epsilonbacteraeota, and Proteobacteria abundance compared to controls (p < 0.05). XYC treatment substantially restored these depleted phyla to near-normal levels (Figures 5A,B). Genus-level analysis demonstrated that XYC effectively counteracted epilepsy-induced dysbiosis, significantly preserving populations of beneficial taxa including Lactobacillus, Ramboutsia, Staphylococcus, and Lachnospiraceae NK4A136 group (Figures 5C,D).

Microbial community differences among the three experimental groups were visualized through LEfSe analysis, generating a cladogram that highlights taxonomically distinct features at multiple phylogenetic levels (Figure 6A). Using stringent criteria (LDA score > 2, p < 0.05), we identified group-specific microbial signatures: Staphylococcus, Ruminococcaceae_UCG_008 and Blautia abundance characterized the Control, Model and XYC group, respectively (Figure 6B).

Serum metabolite profiling was conducted using UHPLC-QTOF-MS/MS, revealing distinct metabolic patterns among experimental groups. OPLS-DA demonstrated clear separation among groups (R²X = 0.35, R²Y = 0.952, Q² = 0.512 for negative ion mode, R²X = 0.566, R²Y = 0.993, Q² = 0.532 for positive ion mode, Figure 7A). Using stringent criteria (VIP > 1, p < 0.05), we identified 970 significantly altered metabolites. Metabolomic profiling showed 157 differentially expressed metabolites between groups (FC ≥ 2 or ≤ 0.5, p < 0.05), with Model specimens exhibiting 73 upregulated versus 84 downregulated species relative to Controls. Notably, XYC treatment modulated 149 metabolites compared to the Model group, including 72 elevated and 77 reduced species (Figure 7B). Hierarchical clustering analysis and fold-change distributions further illustrated these metabolic shifts (Supplementary Figures S1, S2).

KEGG pathway enrichment analysis demonstrated that XYC treatment significantly modulated multiple metabolic pathways in epileptic rats (Figure 8A). The most affected pathways included glycerophospholipid metabolism, pyrimidine metabolism, ABC transporters, proximal tubule bicarbonate reclamation, choline metabolism, and mineral absorption. The relative content of metabolites in the glycerophospholipids pathway were shown in Supplementary Table S1. Targeted analysis of glycerophospholipid metabolism revealed three significantly elevated metabolites in the Model group compared to controls: glycerophosphate (log₂FC = 1.05, p = 0.004), glycerophosphocholine (log₂FC = 0.89, p = 0.003), and sn-glycerol-3-phosphoethanolamine (log₂FC = 1.01, p = 0.021). XYC treatment effectively normalized these aberrant metabolite levels (p < 0.05 vs. Model group; Figure 8B).

To investigate potential microbiota-metabolite interactions, spearman correlation analysis was performed to examine associations between the 19 most abundant bacterial genera and 9 differentially expressed metabolites. Figure 9 delineates statistically robust associations between specific microbial taxa and glycerophospholipid metabolites: Candidatus_Saccharimonas exhibited negative correlations with glycerophosphate, glycerophosphocholine and sn-Glycerol 3-phosphoethanolamine; Bacteriovorax showed positive correlations with glycerophosphate and glycerophosphocholine; and Ruminococcaceae_UCG_008 demonstrated a distinct pattern, correlating positively with sn-glycerol 3-phosphoethanolamine.

Using the TCMSP database, we systematically screened herbal components for the XYC formulation based on pharmacokinetic properties, while excluding compounds lacking biological targets. The analysis identified bioactive constituents from 16 medicinal herbs: Glycyrrhiza uralensis (88 compounds), Salvia miltiorrhiza (46), Astragalus membranaceus (17), Uncaria rhynchophylla (32), Bupleurum chinense (13), Gastrodia elata (14), Polygala tenuifolia (18), Codonopsis pilosula (16), Ziziphus jujuba var. spinosa (7), Acorus tatarinowii (4), along with fewer components from Angelica sinensis (2), Arisaema erubescens (3), Aconitum coreanum (3), Curcuma aromatica (3), Bombyx batryticatus (1), and Massa Medicata Fermentata (2). Target integration and deduplication yielded 773 unique protein targets, representing the potential therapeutic target space of the XYC formulation.

Through comprehensive integration of disease target databases (GeneCards, OMIM, and DisGeNET), we identified 9,615 epilepsy-associated targets, which were subsequently normalized using UniProt identifiers. Comparative analysis revealed 558 overlapping targets between these epilepsy-related genes and the XYC’s potential therapeutic targets (Figure 10A), representing the putative molecular targets through which the formulation may exert its anti-epileptic effects.

Ingredient-target network (283 active ingredients from XYC and 588 targets in epilepsy) was constructed with Cytoscape (version 3.10.3) (Supplementary Figure S3), and the PPI network was shown in Supplementary Figure S4. Topological analysis the top 10 targets as shown in Figure 10B, and the protein–protein interaction (PPI) network was constructed using Cytoscape, with the top 25 hub genes identified by node degree analysis (Figure 10C). KEGG pathway enrichment analysis of these hub genes revealed the top 30 significantly enriched pathways (Figure 10D), among which neuroactive ligand-receptor interactions, pathways in cancer, and lipid metabolism/atherosclerosis were most prominent. Notably, AKT1, INS, and IL-6 emerged as central nodes, which were functionally associated with lipid metabolic and inflammation processes, particularly glycerophospholipid metabolism, suggesting their potential as key therapeutic targets of XYC for epilepsy treatment.

To investigate the targets underlying the anti-epileptic effects of XYC, we performed RT-qPCR and Western blot analyses. RT-qPCR revealed elevated mRNA levels of Il-6, Ins, and Akt1 in the hippocampal tissues of model groups compared to controls. Notably, XYC intervention significantly attenuated these increases (Figure 10E). Consistent with the mRNA findings, immunoblotting showed upregulated AKT1 protein expression in the epileptic models. Importantly, this elevated expression was normalized by XYC treatment (Figure 10F).

These findings suggest that XYC effectively alleviated seizure in lithium-pilocarpine-induced acute epilepsy by preserved neuro-inflammatory and glycerophospholipid metabolism.