The blood-absorbed components used in this study were derived from our previous work (22), which systematically identified the prototype components absorbed into rat plasma after oral administration of XCHG. Briefly, in that study, a rat fever model was established by subcutaneous injection of 20% dry yeast suspension, and both normal and model rats were orally administered XCHG at a dose of 12.96 g·kg⁻¹. Blood samples were collected from the orbital vein at multiple time points (0.25, 0.5, 1, 2, 4, 6, and 8 h) after administration. Plasma samples were processed by protein precipitation with acetonitrile and analyzed using ultra-high performance liquid chromatography coupled with triple quadrupole mass spectrometry (UPLC-MS/MS) in multiple reaction monitoring (MRM) mode. Through comparison of reference standards, blank plasma, and plasma samples from both groups, 18 prototype blood-absorbed components were identified (Figure 3).

The SMILES structures of these 18 components were retrieved from the PubChem database¹ and then submitted to SwissTargetPrediction² for target prediction, with the species parameter set to Homo sapiens. After merging and removing duplicates, we obtained the potential targets of XCHG’s blood-absorbed components.

Fever-associated targets were systematically retrieved from four major databases: OMIM³, TTD⁴, DrugBank⁵, and GeneCards⁶. After merging non-redundant entries, we cross-referenced these with XCHG’s blood-absorbed component targets to identify potential anti-pyretic candidates. Target overlaps were visualized via a Venn diagram tool⁷.

Intersection targets were analyzed using the STRING database⁸ under Homo sapiens settings (interaction skey ≥ 0.4; free proteins hidden). The resulting interactions (TSV format) were imported into Cytoscape 3.7.0 for visualization and topological assessment. Network centrality parameters—degree (DC), betweenness (BC), and closeness centrality (CC)—were computed via the CentiScape 2.2 plugin. Targets with DC, BC, and CC values exceeding their respective means were defined as key targets, and a condensed PPI subnetwork was generated. Node properties (size/color intensity) in the visualization reflect interaction degrees, with larger, darker nodes indicating higher connectivity. Final topological validation was performed using Cytoscape’s ‘Network Analyzer’ plugin.

To elucidate the multi-scale therapeutic mechanisms of XCHG, we first constructed a blood-absorbed component-target network of XCHG using Cytoscape 3.7.0 to identify pharmacologically relevant interactions between bioavailable components and their molecular targets. Building upon this foundation, we further constructed an integrated “herb-component-target-fever” network by mapping herbal components to their shared targets with fever-related genes. This comprehensive network integrates multilayered information, including herbal sources, bioactive components, molecular targets, and fever-associated pathways. Topological analysis using the Network Analyzer plugin identified five key components exhibiting the highest degree and betweenness centrality values, indicating their pivotal roles in mediating XCHG’s antipyretic effects through multi-target modulation of fever-associated pathways.

Functional enrichment analysis of XCHG’s fever-related targets was performed using the DAVID platform⁹, encompassing Gene Ontology (GO) terms and KEGG pathways (23). Significant results were visualized as bubble plots and bar charts through the bioinformatics online platform (See Footnote text 7), highlighting key biological processes and signaling pathways modulated by XCHG.

To investigate the interaction mechanisms between key components of XCHG and their corresponding targets, molecular docking analysis was performed. The five key components were selected as ligands, and the six highest-degree targets from the network were used as receptors. Component structures were obtained from the PubChem database, while target protein structures were retrieved from the Protein Data Bank (PDB) using the following criteria: human origin (Homo sapiens), X-ray diffraction resolution ≤3 Å, and publications within the last decade. Molecular docking was performed using AutoDock Vina (version 1.2.5) and AutoDock Tools (version 1.5.7). Protein structures were prepared using PyMOL (version 2.4.0) to remove water molecules, original ligands, and heteroatoms, followed by addition of polar hydrogen atoms and assignment of Kollman charges in AutoDock Tools. Ligand structures were processed by adding hydrogen atoms, optimizing geometric conformations, assigning Gasteiger charges, and converting to PDBQT format. Docking calculations were conducted with exhaustiveness = 24 and number of binding modes = 9. Binding site definitions and docking box parameters are provided in Supplementary Table S1. Each docking calculation was performed in multiple replicates to ensure reproducibility, with statistical analysis presented in Supplementary Table S2. All docked complexes were visualized using PyMOL (Supplementary Figure S1), and the top three complexes were further analyzed using Discovery Studio 4.5 Client to generate 2D interaction maps.

Molecular dynamics (MD) simulations were performed using GROMACS 2021 to evaluate the structural dynamics and binding characteristics of the three most stable complexes identified from docking studies. Systems were parameterized using the AMBER14SB force field for proteins and GAFF2 for ligands, solvated with the SPC/E water model in periodic cubic boxes extending 1.2 nm from the complex surface, and neutralized with NaCl. Long-range electrostatic interactions were calculated using the particle mesh Ewald (PME) method. Prior to production runs, systems underwent energy minimization using the steepest descent algorithm (50,000 steps), followed by NVT and NPT equilibration (50,000 steps each at 310 K and 1 atm with a 2 fs timestep). Production simulations were conducted for 100 ns with LINCS constraints applied to hydrogen bonds, saving coordinates every 10 ps. Each simulation was performed in triplicate to ensure reproducibility, with statistical analysis presented in Supplementary Table S3. Binding free energies were calculated using the MM/PBSA method. Trajectory analysis included root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), radius of gyration (Rg), hydrogen bond formation, and solvent-accessible surface area (SASA). Principal component analysis (PCA) and dynamic cross-correlation matrices (DCCM) were employed to characterize collective motions. Free energy landscape (FEL) construction and residue-specific energy decomposition were performed, with conformational evolution monitored at 25-ns intervals throughout the simulation.

The commercial preparation utilized in this study was manufactured by Guangzhou Guanghua Pharmaceutical Co., Ltd. The composition of the granules consists of seven medicinal herbs in specific proportions: Bupleurum chinense DC. [Apiaceae] roots (Chaihu), Scutellaria baicalensis Georgi [Lamiaceae] roots (Huangqin), ginger-processed Pinellia ternata [Araceae] tubers (Jiangbanxia), Codonopsis pilosula [Campanulaceae] roots (Dangshen), fresh Zingiber officinale [Zingiberaceae] rhizomes (Shengjiang), Glycyrrhiza uralensis [Fabaceae] roots/rhizomes (Gancao), and Ziziphus jujuba [Rhamnaceae] fruits (Dazao). The crude drug-to-granule ratio is 0.486:1 (w/w), indicating that each gram of granules contains the equivalent of 0.486 g of raw herbal materials. RAW264.7 murine macrophages and DMEM complete medium were sourced from Wuhan Pricella Biotechnology Co., Ltd. (Wuhan, China), while LPS was acquired from Sigma-Aldrich (USA). Reagents including the NO assay kit and CCK-8 came from Beyotime Biotechnology (Shanghai, China), with ELISA kits for TNF-α, IL-6, COX-2, IL-1β, PGE2, and iNOS provided by Jiangsu Meimian Industrial Co., Ltd. (Jiangsu, China). Primary antibodies against p-EGFR, EGFR, p-SRC, SRC, and ESR1 were purchased from Abcam (Cambridge, UK). All primers were commercially synthesized by Genewiz Co., Ltd. (Suzhou, China).

RAW264.7 cells were seeded in DMEM supplemented with 10% FBS, 100 IU/mL penicillin, and 100 IU/mL streptomycin, cultured at 37 °C in 5% CO₂ for 1–2 days. When confluence reached about 80%, cells were passaged. Cells in the logarithmic growth phase were used for experiments.

RAW264.7 cells at 80–90% confluence were seeded in 96-well plates at a density of 1 × 10⁴ cells/well (100 μL/well) and allowed to adhere for 24 h at 37 °C under 5% CO₂. After aspiration of the supernatant, treatments were initiated. XCHG was directly dissolved in complete culture medium (RPMI 1640 supplemented with 10% FBS, 100 IU/mL penicillin, and 100 IU/mL streptomycin), followed by sterile filtration through 0.22 μm PVDF membranes. Experimental groups received 100 μL of XCHG-containing medium at concentrations of 0.5, 1, 1.5, 2, 2.5, and 3 mg/mL, while control groups were treated with equal volumes of freshly prepared medium without XCHG. Following 24 h incubation under identical conditions, the medium was aspirated and replaced with 10 μL CCK-8 solution. After 2 h incubation at 37 °C/5% CO₂, absorbance was measured at 450 nm using a microplate reader to determine cell viability.

RAW264.7 cells were plated in 96-well plates at a density of 1 × 10⁴ cells per well. The experiment comprised six groups: control, model, positive control (10 μM dexamethasone (Dex)), and XCHG-treated groups (low, medium, and high doses: 1, 1.5, and 2 mg/mL, respectively). After 24 h of culture, except for the control group, all groups were stimulated with LPS(1 μg/mL) and treated with corresponding drug concentrations for 24 h. Following 24 h of incubation at 37 °C with 5% CO₂, supernatants were collected. The supernatant was mixed with Griess reagent and incubated at room temperature for 5 min. Absorbance was measured at 540 nm using a microplate reader. A standard curve was generated using sodium nitrite standards, and NO concentrations in test samples were calculated based on their absorbance values.

Cells were seeded in 6-well plates at a density of 3 × 10⁵ cells/well and divided into the following groups: blank control, model, positive control (10 μM Dex), and XCHG-treated groups (low, medium, and high doses: 1, 1.5, and 2 mg/mL, respectively). After 24 h of culture, except for the control group, all groups were stimulated with LPS and treated with corresponding drug concentrations for 24 h. The cell supernatants were then collected, and the concentrations of TNF-α, IL-6, COX-2, IL-1β, PGE2, and iNOS were determined using ELISA kits according to the manufacturer’s instructions. The absorbance was measured at 450 nm, and the cytokine concentrations were calculated based on standard curves.

The culture grouping conditions of RAW264.7 cells were performed according to Section 2.12. Total RNA was isolated from the cells using Trizol reagent, followed by reverse transcription with a cDNA synthesis kit. For the quantification of TNF-α and IL-6 genes, qRT-PCR amplification was carried out under specific conditions. Initially, there was a denaturation step at 95 °C for 1 min and 30 s (1 cycle). This was followed by 40 cycles, each consisting of denaturation at 95 °C for 20 s, annealing at 64 °C for 20 s, and extension at 72 °C for 30 s. Finally, a melting curve analysis was performed, starting at 60 °C for 1 min and increasing by 0.4 °C every 15 s until 95 °C was reached, where it was maintained for 15 s (1 cycle). The primers employed in this study are presented in Table 1.

Cells were seeded in 6-well plates (4 × 10⁵ cells/well, 2 mL) and divided into the following groups: blank control, model, and XCHG-treated groups (low, medium, and high doses). After 24 h of culture, except for the control group, all groups were stimulated with LPS and treated with corresponding drug concentrations for 24 h Following treatment, cells were collected and total protein was extracted with RIPA buffer (30 min on ice), followed by centrifugation (12,000 rpm, 20 min, 4 °C) to obtain the supernatant. Samples were mixed with 5 × loading buffer and heat-denatured at 100 °C for 10 min. Equal protein quantities were separated by SDS-PAGE using 15-well precast gels and electrotransferred to PVDF membranes. Membranes were blocked for 30 min, then incubated overnight at 4 °C with primary antibodies against p-EGFR, EGFR, p-SRC, SRC, ESR1 (all at 1:1000 dilution), and GAPDH (1:5000). Following five TBST washes (6 min each), HRP-conjugated secondary antibodies (1:5000) were applied for 1 h at room temperature. After additional washing, bands were visualized by ECL and captured using a chemiluminescence imaging system. Band intensities were quantified with ImageJ.

The data are shown as mean ± standard deviation (SD). Statistical significance was assessed using one-way analysis of variance (ANOVA) via GraphPad Prism 10.1.2 software, with a p-value of less than 0.05 regarded as statistically significant.

Potential therapeutic targets of XCHG were identified through database mining and comparative analysis. The SwissTargetPrediction platform yielded 298 putative targets for the 18 blood-absorbed components. Concurrently, fever-associated genes (n = 2,079) were compiled from four disease databases (GeneCards, OMIM, TTD, DrugBank) following deduplication. Intersection analysis through Venn plotting revealed 120 shared targets between component and disease gene sets (Figure 4A), representing potential fever-modulating targets of XCHG.

The 120 shared targets were analyzed via STRING to generate a PPI network (120 nodes, 1,432 edges; Figure 4B). Network visualization in Cytoscape 3.7.0 incorporated node attributes scaled by topological importance, with size and color intensity reflecting connectivity (Degree). Network centrality parameters—degree (DC), betweenness (BC), and closeness centrality (CC)—were computed via the CentiScape 2.2 plugin. Targets with DC, BC, and CC values exceeding their respective means (Degree ≥23.87, Closeness≥0.00446, Betweenness≥109.13) were defined as key targets, and a condensed PPI subnetwork was generated. Through this analysis, we identified 17 key targets (Figure 4C) including IL6, TNF(TNF-α), EGFR, STAT3, SRC, ESR1, PTGS2, HSP90AA1, MMP9, PPARG, MAPK3, IL2, CYP3A4, APP, ABCB1, ABCG2, and NTRK2. These targets demonstrated substantial network influence, implicating their functional significance in XCHG’s antipyretic activity. The top six key targets by Degree value (Table 2) were subsequently selected for molecular docking with key components.

The “herb-component-target-fever” interaction network was constructed using Cytoscape 3.7.0 through topological analysis of overlapping targets. The blood-absorbed component-target network analysis revealed a network comprising 317 nodes, including 298 key targets, 18 components, and XCHG (Figure 4D), demonstrating the pharmacological relevance between bioavailable components and their molecular targets. The further constructed integrated “herb-component-target-fever” network comprised 140 nodes, including 120 key targets, 18 components, XCHG, and fever (Figure 4E). Based on degree values, key components were identified as primary therapeutic components: Oroxylin A, Wogonin, Baicalein, Liquiritigenin, and Enoxolone (Table 3). These components are predicted to be the major key components responsible for XCHG’s antipyretic effects.

Functional enrichment analysis revealed XCHG’s comprehensive regulatory profile across key biological domains. The analysis identified 528 significantly enriched biological processes (BP), prominently featuring protein phosphorylation and MAPK cascade regulation. Cellular component (CC) analysis (69 terms) demonstrated predominant localization to focal membrane receptor complexes and endoplasmic reticulum structures. Molecular function (MF) assessment (129 terms) highlighted crucial activities including tyrosine kinase function, ATP binding, and enzyme binding (Figure 5A). These findings collectively suggest that XCHG, while regulating cellular stress responses, also has the potential to regulate phosphorylation signaling (such as the MAPK cascade) and membrane receptor activity. KEGG pathway analysis from the DAVID database identified 160 pathways, with the top 20—including PI3K-Akt, Ras, Rap1, MAPK, C-type lectin receptor signaling, and EGFR tyrosine kinase inhibitor resistance (Figure 5B) —Although these pathways are broadly functional signaling cascades involved in diverse cellular processes, substantial evidence supports their specific roles in inflammatory fever pathophysiology. The PI3K-Akt pathway has been demonstrated to regulate macrophage inflammatory activation and cytokine production, with PI3K inhibition significantly attenuating LPS-induced TNF-α and IL-6 release (24). The Ras signaling pathway functions as an upstream activator of the MAPK cascade; upon Toll-like receptor (TLR) stimulation in macrophages, Ras activation triggers the Raf–MEK–ERK cascade, which promotes the transcription of pro-inflammatory cytokines including TNF-α and IL-6 through AP-1 activation (25, 26). The Rap1 signaling pathway, while primarily recognized for regulating integrin-mediated cell adhesion, has been shown to modulate inflammatory responses in immune cells by controlling leukocyte recruitment to inflammatory sites and regulating NF-κB-dependent cytokine production (27). The MAPK cascade, particularly p38 MAPK, directly controls COX-2 mRNA stability and protein expression in response to pyrogenic stimuli such as IL-1β (28). EGFR signaling has been shown to amplify inflammatory responses in macrophages through transactivation of NF-κB and subsequent upregulation of pro-inflammatory mediators (29). Furthermore, C-type lectin receptor pathways are pattern recognition receptors that detect pathogen-associated molecular patterns and initiate the inflammatory cascade leading to fever (30). Collectively, these pathways converge on the regulation of pro-inflammatory cytokine production (TNF-α, IL-1β, IL-6) and downstream mediators (COX-2, PGE2) that constitute the molecular basis of inflammatory fever. These findings suggest that XCHG may exert its antipyretic effects through modulation of these inflammation-related signaling networks, supporting its multi-target mechanism of action.

Molecular docking between XCHG’s key components and six highest-degree targets (EGFR, IL6, ESR1, STAT3, SRC, TNF-α) revealed binding energies ranging from −6.3 to −9.3 kcal/mol (Figure 6), all exceeding the favorable binding threshold (−5.0 kcal/mol). Three complexes demonstrated exceptionally strong interactions: EGFR-Enoxolone (−9.3 kcal/mol), ESR1-Liquiritigenin (−8.7 kcal/mol), and SRC- Baicalein (−8.4 kcal/mol), with binding energies significantly below the −7.0 kcal/mol high-affinity threshold (31). Detailed interaction analysis revealed multi-modal binding patterns (Figure 7). The EGFR-Enoxolone complex showed hydrogen bonds with CYS797 and ASN842, extensive hydrophobic interactions with multiple residues, and π-σ interactions involving PHE856. ESR1-Liquiritigenin formed hydrogen bonds through LEU387, comprehensive hydrophobic contacts, and π-π stacking with PHE404. SRC-Baicalein exhibited hydrogen bonding via GLU310 and ILE336, hydrophobic interactions with key residues, and π-σ/π-alkyl interactions stabilizing the flavonoid structure. These multi-modal interaction patterns validate the strong binding capabilities of these compounds to their respective targets, explaining their exceptionally high binding affinities. These three optimal complexes were selected for further MD simulations to investigate their dynamic binding characteristics.

The structural integrity of protein-ligand complexes was assessed through root-mean-square deviation (RMSD) analysis, with values under 1 nm signifying stable molecular interactions under physiological conditions (32). RMSD analysis demonstrated that all three complexes (EGFR-Enoxolone, ESR1-Liquiritigenin, and SRC-Baicalein) maintained high structural stability within 20-60 ns (Figure 8A), with RMSD values stabilizing at approximately 0.3 nm, 0.25 nm, and 0.4–0.6 nm, respectively. Compared to the other two complexes, SRC-Baicalein exhibited relatively larger RMSD fluctuations (0.4–0.6 nm), which may be attributed to inherent structural characteristics of different proteins under identical simulation conditions. RMSF was used to assess positional fluctuations of amino acid residues, reflecting regional flexibility (33). RMSF analysis revealed that all three complexes displayed overall fluctuations below1 nm, indicating limited residue mobility, with flexible regions exhibiting fluctuations between 0.4–0.6 nm (Figure 8B). Rg was utilized to evaluate overall structural compactness, where larger values indicate structural expansion while smaller values reflect tighter packing (34). Rg analysis further confirmed the structural compactness of the complexes, with Rg values maintained at approximately 2 nm, 1.75 nm, and 3 nm for the three complexes respectively, showing minimal fluctuations (Figure 8C). Hydrogen bonds (H-bonds), as critical non-covalent interactions governing complex stability (35), were quantitatively analyzed across the three complexes. The EGFR-Enoxolone complex formed 2 stable H-bonds, ESR1-Liquiritigenin exhibited 1–2 bonds, while SRC-Baicalein showed the most extensive network with 2–3 persistent H-bonds (Figure 8D). This H-bond patterning demonstrates stable ligand-receptor recognition, the SRC-Baicalein complex forming the most H-bonds, indicative of superior binding affinity. SASA reflects surface solvent accessibility, where stable SASA profiles indicate well-folded structures (36). All three complexes exhibited minimal SASA fluctuations (Figure 8E), further supporting their high structural stability. These results collectively demonstrate that all three complexes maintained excellent binding stability and structural compactness throughout the MD simulations.

The FEL was generated using RMSD and Rg data to characterize conformational changes of the complexes during simulations. In the FEL plots, dark-colored regions represent the lowest-energy states while light-colored areas correspond to higher-energy states (37). The results demonstrated that all three complexes (EGFR-Enoxolone, ESR1-Liquiritigenin, and SRC-Baicalein) formed single, well-defined low-energy clusters (Figure 9A), indicating high conformational stability and strong binding interactions during the simulations. The MM/PBSA approach quantified binding affinities, with more negative values corresponding to stronger receptor-ligand interactions (38). The MM/PBSA calculations revealed strongly favorable binding free energies for all three complexes (Figure 9B; Supplementary Table S4): EGFR-Enoxolone (−45.79 kcal/mol), ESR1-Liquiritigenin (−35.58 kcal/mol), and SRC-Baicalein (−32.9 kcal/mol), indicating robust molecular recognition at the binding interfaces. Per-residue energy decomposition analysis (Figure 9C) identified a key binding site residue that made a significant contribution to the interaction: PHE856 (−2.67 kcal/mol), LEU718 (−2.1 kcal/mol), and VAL726 (−1.99 kcal/mol) in EGFR-Enoxolone; LEU346 (−2.02 kcal/mol), PHE404 (−1.71 kcal/mol), and LEU387 (−1.7 kcal/mol) in ESR1-Liquiritigenin; and GLU310 (−4.0 kcal/mol) in SRC-Baicalein. The smaller the values corresponding to these residues, the greater their contribution to the binding energy. These energetically favorable interactions significantly stabilized the respective complexes. All three systems demonstrated excellent conformational and binding stability throughout molecular dynamics simulations, providing crucial insights for future studies of protein-ligand interactions.

To further investigate the conformational changes of proteins in the complexes, we employed PCA and DCCM to evaluate the dynamic behavior of three receptor-ligand systems. PCA identified dominant motion modes, with plot axes representing conformational space dimensions and color gradients (blue to red) indicating simulation progression time (39). DCCM revealed residue-residue cross-correlations, where color gradients (light blue to pink) represented motion coordination: values approaching 1 indicated synchronized movements, while values near −1 represented anti-correlated motions (40, 41) PCA results (Figure 10A) demonstrated that the three principal components (PC1, PC2, and PC3) accounted for 57.44, 43.82, and 80.93% of the total variance in the EGFR-Enoxolone, ESR1-Liquiritigenin, and SRC-Baicalein complexes, respectively. PC1 primarily captured large-scale conformational changes, while PC2 and PC3 were associated with medium-range motions and localized movements, respectively. Notably, the SRC-Baicalein complex exhibited the highest PC1 contribution (61.55%), indicating its most pronounced conformational flexibility and suggesting greater potential for dynamic responses and structural adaptability. DCCM analysis (Figure 10B) revealed distinct binding dynamics among the complexes. In the EGFR-Enoxolone complex, correlated motions were predominantly observed within residues 100–150 and 200–250, demonstrating significant cooperative movements in these regions with relatively independent motions elsewhere. The ESR1-Liquiritigenin complex showed strongly coupled dynamics confined to residues 50–100, while other regions displayed more independent motions. Notably, the SRC-Baicalein complex demonstrated the most pronounced correlation patterns, with strong positive couplings (residues 100–150) and anti-correlated motions (residues 300–450) that significantly exceeded those observed in other complexes. This unique dynamic profile, characterized by enhanced residue coordination and pronounced anti-correlations, likely underpins both its superior binding affinity and exceptional conformational adaptability during simulations. Furthermore, structural snapshots extracted at 0, 25, 50, 75, and 100 ns time points from the molecular dynamics trajectories (Figure 10C) demonstrated EGFR-Enoxolone, ESR1-Liquiritigenin, and SRC-Baicalein maintained stable binding conformations without significant positional drift, confirming their excellent binding stability. Collectively, these findings provide crucial insights into the dynamic behavior and binding characteristics of the protein-ligand complexes, while supporting the therapeutic potential of these ligands as drug candidates.

The selection of RAW264.7 macrophages as the in vitro model was based on several considerations. First, macrophages serve as sentinel cells in the innate immune system and are primary producers of pyrogenic cytokines (TNF-α, IL-1β, IL-6) upon pathogen recognition, directly initiating the fever cascade (42). Second, KEGG pathway enrichment analysis revealed significant involvement of PI3K-Akt, MAPK, and EGFR signaling pathways, all of which have been extensively documented to regulate macrophage inflammatory activation and cytokine secretion (26). Third, the enrichment of C-type lectin receptor signaling pathways further supports the relevance of macrophage models, as these pattern recognition receptors are predominantly expressed on macrophages and mediate pathogen-induced inflammatory responses (30). Fourth, RAW264.7 cells represent a well-established and widely validated model for studying anti-inflammatory mechanisms of natural products, enabling comparison with previous studies on herbal medicine compounds (43). Therefore, this cellular model provides a biologically relevant platform to investigate XCHG’s modulatory effects on the inflammatory-pyrogenic signaling network.

The CCK-8 assay results (Figure 11A) demonstrated that XCHG at concentrations below 2 mg/mL had minimal impact on RAW264.7 cell viability after 24-h treatment compared to the control group. However, cytotoxicity of the drug has been observed when the concentration exceeds 2 mg/mL. Therefore, we do not include this dose range to avoid interference caused by the cytotoxic mechanism. Based on these findings, subsequent experiments were conducted using XCHG concentrations of 1, 1.5, and 2 mg/mL.

NO is usually used to characterize the activation of inflammatory responses in RAW264.7 macrophages. As determined by Griess assay, XCHG significantly attenuated LPS-induced NO production in RAW264.7 cells (Figure 11B). The LPS-induced model group showed markedly increased NO secretion compared to the control group (p < 0.001), confirming successful model establishment. Notably, XCHG treatment at all tested concentrations (low dose: 1 mg/mL, medium dose: 1.5 mg/mL, and high dose: 2 mg/mL) and Dex dose-dependently suppressed NO release compared to the model group, demonstrating potent anti-inflammatory effects.

Figure 12 demonstrates that the LPS-treated model group exhibited significantly elevated levels of inflammatory biomarkers—TNF-α, IL-6, COX-2, IL-1β, PGE2, and iNOS—compared to untreated controls (p < 0.001). Both Dex and medium-to-high doses of XCHG significantly suppressed the expression of TNF-α and iNOS (p < 0.001), while the low-dose XCHG group showed moderate reduction in these parameters (p < 0.05). Moreover, treatment with Dex or any concentration of XCHG resulted in substantial suppression of IL-6, COX-2, IL-1β, and PGE2 relative to the LPS-induced group. These findings reveal that XCHG inhibits the release of pro-inflammatory factors induced by LPS in macrophages in a dose-dependent manner, thereby suppressing their inflammatory activation. Notably, the immunomodulatory efficacy of medium- and high-dose XCHG treatment approached that of the clinical reference compound Dex.

As shown in Figure 13, compared with the control group, the LPS-induced model group exhibited significantly upregulated mRNA expression levels of TNF-α and IL-6 (p < 0.001). Relative to the LPS model group, both the Dex group and the medium-and high-dose XCHG groups demonstrated marked downregulation of TNF-α mRNA expression (p < 0.001). Additionally, the Dex group and high-dose XCHG group significantly suppressed IL-6 mRNA levels (p < 0.05).

Western blot analysis revealed that (Figure 14), compared with the control group, the model group exhibited significantly elevated p-EGFR/EGFR and p-SRC/SRC ratios, accompanied by a marked reduction in ESR1 expression (all p < 0.001). XCHG treatment attenuated the p-EGFR/EGFR ratio in a dose-dependent manner, with the low-dose group demonstrating moderate inhibition (p < 0.01), while the medium- and high-dose groups exhibited more pronounced suppression (p < 0.001) (Figure 14A). Regarding SRC phosphorylation, all three doses of XCHG significantly suppressed the p-SRC/SRC ratio (p < 0.001) (Figure 14B), indicating a sustained inhibitory effect on SRC activation across the dosage range. Furthermore, both medium- and high-dose XCHG treatments significantly restored ESR1 expression (p < 0.001) (Figure 14C), suggesting a protective effect of XCHG on ESR1 protein levels. Collectively, these findings demonstrate that XCHG intervention effectively suppresses aberrant activation of the EGFR-SRC signaling pathway while restoring ESR1 expression in model cells.