Primary antibodies including myeloperoxidase (MPO, 22225-1-AP), tumor necrosis factor-alpha (TNF-α, 60291-1-Ig), interleukin-1 beta (IL-1β, 16806-1-AP), brain-derived neurotrophic factor (BDNF, 66292-1-Ig), PSD95 (20665-1-AP), synaptophysin (SYN, 67864-1-Ig), choline acetyltransferase (CHAT, 20747-1-AP), choline transporter 1(CHT1, 21848-1-AP), zonula occludens-1 (ZO-1, 66452-1-Ig), Occludin (66378-1-Ig) and β-tubulin (80713-1-RR) were purchased from Proteintech Biotechnology Co., Ltd. (Wuhan, China). Commercial kits including total superoxide dismutase (SOD, A001-3-2), malonaldehyde (MDA, A003-1-2), reduced glutathione (GSH, A006-2-1), acetyl choline (ACh, A105-1-1), acetylcholinesterase (AChE, BC2025-100T/48S) were purchased from Nanjing Jiancheng Biotechnology Co., LTD. (Nanjing, China). AB-PAS staining assay (G1285) was purchased form Beijing Solarbio Science & Technology Co.,Ltd. Scopolamine (Scop, HY-N0296) was purchased from MedChemExpress LLC (China) (Shanghai, China). Xinshubao tablets (2011024) were obtained from Zhangzhou Pien Tze Huang Pharmaceutical Co., Ltd. (Zhangzhou, China). Donepezil hydrochloride tablets (2203026) were purchased from Eisai (China) Pharmaceutical Co., LTD.

Male C57BL/6J mice (6–8 weeks old) were procured from Beijing SPF Biotechnology Co., Ltd. (Certificate No.: SCXK (Beijing) 2019-0010). The animals were housed in a controlled environment with a temperature maintained at 23°C–25°C, relative humidity of 40%–60%, and a 12-h light/dark cycle. Standard laboratory chow and water were provided ad libitum. Following a week acclimatization period, the experimental procedures were initiated.

All the experiments were approved by the Ethical Committee of Experimental Animal Welfare of Experimental Research Center China Academy of Chinese Medicine Science (ERCCACMS21-2201-01), and conducted in accordance with the ARRIVE guidelines, and were carried out in accordance with Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85-23, revised 1996).

The experiment was divided into 6 groups: Control group, Model group, XSB low-dose, medium-dose, and high-dose groups, and a Positive Control group, with 12 mice in each group. Mice in the Control group received intraperitoneal injections of normal saline, while all other groups were administered scopolamine hydrobromide solution (3 mg/kg) via intraperitoneal injection once daily for 4 consecutive weeks to induce cognitive impairment (Xu et al., 2019). Simultaneously, the XSB-treated groups were orally administered XSB at doses of 0.13 g/kg, 0.26 g/kg and 0.52 g/kg once daily (Xiao et al., 2024). The Control and Model groups were given an equivalent volume of CMC-Na via oral gavage, whereas the Positive control group received 0.65 mg/kg donepezil hydrochloride suspension in CMC-Na by oral gavage (Xiao et al., 2020). n = 12 mice in each group.

Commencing on the 21st day of drug administration, the cognitive functions of the mice were systematically assessed through a series of three sequential behavioral paradigms: the novel object recognition (NOR) test, the spontaneous alternation Y-maze test, and the Morris water maze (MWM) test.

Spontaneous alternation Y-maze test: The spontaneous alternation Y-maze test was conducted to evaluate working memory in mice. Each mouse was placed at the distal end of one arm and allowed to explore freely for 5 min. Behavioral trajectories were recorded using a video tracking system, and the spontaneous alternation rate was calculated with the ANY-maze software.

NOR test: Following the Y-maze test, each mouse was placed individually in a chamber for 5 min to adapt to the environment. After 24 h, two identical objects were introduced, and mice were allowed to explore freely for 5 min (training phase). One hour later, in the testing phase, one of the identical objects was replaced with a novel object of similar size but different shape. The exploration time for the novel object (T1) and familiar object (T2) was recorded using ANY-maze software, and the recognition index (RI) was calculated as RI = T1/(T1 + T2).

MWM test: MWM test was conducted to evaluate spatial learning and memory. During the spatial acquisition phase, a hidden platform was placed 1 cm below the water surface (20°C–21°C). Mice were trained once daily for 5 days, with trials initiated from random quadrants. If the platform was not located within 60 s, mice were guided to it and allowed to remain for 20 s. On the 6th day, the platform was removed, and a probe trial was performed. The escape latency, platform crossings, and time spent in the target quadrant were recorded to assess learning and memory.

After the MWM test, 3 mice per group were randomly selected for histological analysis. Paraffin-embedded brain sections (4 μm) were prepared using a rotary microtome and stored at 4°C for histopathological analysis.

Following the completion of the MWM test, 8 mice from each group were sacrificed under deep anesthesia, and their brains were rapidly harvested. The superoxide dismutase (SOD), malondialdehyde (MDA), glutathione (GSH), acetylcholine (ACh) and acetylcholinesterase (AChE) levels in brain tissue were quantified following the manufacturer’s instructions.

Hippocampal tissues were collected from 3 mice per group. The hippocampal samples were rapidly dissected and fixed in 2% glutaraldehyde (pH 7.2). Tissue blocks from the CA1 region, measuring approximately 1.0 × 1.0 × 1.0 mm³, were further fixed in 1% osmium tetroxide at room temperature for 2 h. After sequential dehydration, infiltration, embedding, and polymerization, ultrathin sections (70 nm) were prepared. These sections were stained with 2% uranyl acetate in ethanol and 2.6% lead citrate. Finally, the sections were visualized and imaged using a TEM (HT7700, Hitachi, Tokyo, Japan). Quantitative analysis of mitochondrial density and synaptic numbers was performed using ImageJ software.

Hippocampal total proteins were separated by SDS-PAGE and subsequently transferred onto PVDF membranes via wet transfer at 250 mA for 80 min. The membranes were blocked with 5% BSA at room temperature for 1 h, followed by incubation with primary antibodies against CHAT, CHT1, BDNF, SYN, PSD95, TNF-α, IL-1β, and MPO at 4°C overnight. After thorough washing, membranes were incubated with the corresponding HRP-conjugated secondary antibodies. The protein bands were visualized using a gel imaging system, and grayscale intensity was quantified using ImageJ software. Protein expression levels were normalized to internal controls (n = 3 in each group).

The detailed information was included in Supplementary Material. In short, fecal samples from 6 mice per group were aseptically collected. Amplicon sequence variants (ASVs) were identified through exact sequence variants, and taxonomic classification was performed against the SILVA 16S rDNA gene reference database. Alpha diversity indices, such as Chao1, Shannon, and Simpson, were calculated to assess species richness and diversity. Beta diversity was evaluated using weighted UniFrac distances. Differentially abundant taxa were identified using linear discriminant analysis effect size (LEfSe), with an LDA score threshold of 3.0. Spearman correlation analysis was performed to investigate the associations between differentially abundant taxa and cognitive function-related indicators in the behavior tests, including escape latency, platform crossing number, time spent in the target quadrant, and spontaneous alternation. A correlation coefficient threshold of |R| > 0.3 and a significance level of P < 0.05 were used to identify potential associations.

The detailed information was included in Supplementary Material. In short, serum samples from 6 mice per group were collected. Metabolomics profiling was conducted using liquid chromatography coupled with mass spectrometry (LC-MS). Chromatographic separation was performed on an ACQUITY UPLC system, with mass spectrometric detection carried out on a Q Exactive Plus Orbitrap mass spectrometer. Data acquisition was conducted in both positive and negative ion modes, and full-scan mass spectra were recorded across a specified m/z range. Differential metabolites (DMs) were determined based on the variable importance in projection (VIP) scores from the PLS-DA model, with a threshold of VIP >1.0. Pathway enrichment analysis was performed using KEGG database. Pearson correlation analysis was conducted to explore the relationships between DMs and cognitive function-related indicators. A correlation coefficient threshold of |R| > 0.5 and statistical significance of P < 0.05 were used to identify significant associations.

mRNA-seq was conducted to analyze transcriptomic variations in the hippocampal tissues. Total RNA was extracted using a commercial RNA extraction kit (RC101-01, Vazyme Biotech Co., Ltd., Nanjing, China). Gene expression levels were quantified by calculating the TPM value. Differentially expressed genes (DEGs) were identified using DESeq2 with an absolute log₂ fold change (log₂ FC) ≥ 1 and P value <0.05. Functional annotation of DEGs was conducted through Gene Ontology (GO) and KEGG pathway enrichment using DAVID database (https://davidbioinformatics.nih.gov/tools.jsp). The Volcano, Heatmap, GO and KEGG pathway enrichment plots were generated using the online tool available at https://www.bioinformatics.com.cn.

All data were presented as mean ± standard error of the mean (mean ± SEM). Statistical analyses were performed using GraphPad Prism 9.4 software. Escape latency data in the MWM test, collected across multiple trials, were analyzed using repeated-measures two-way ANOVA followed by Bonferroni post-hoc test. For comparisons among multiple groups, one-way ANOVA with Tukey’s post-hoc test was applied. A p-value of less than 0.05 was considered statistically significant. Fluorescence intensity quantification was conducted using ImageJ software.

The NOR test results showed treatment with XSB-M (Figures 1A–C, P <0.05 vs. Model group), and XSB-H (Figures 1A–C, P < 0.01 vs. Model group) significantly improved the recognition index, with effects similar to the positive control drug donepezil (Figures 1A–C P < 0.01 vs. Model group). Similarly, in the Y-maze test, XSB-H significantly improved spontaneous alternation behavior (Figures 1D–F, P < 0.05 vs. Model group). In the MWM test, the typical swimming track diagrams of mice were as shown in Figure 1G. XSB treatment decreased escape latency over time, test, XSB treatment decreased escape latency over time, with significant differences observed on days 4 and 5 compared to the Model group (Figure 1H, P < 0.001 on day 5). Additionally, in the spatial probe trial, XSB treatment increased platform crossings (Figure 1I, P < 0.01 vs. Model group) and time spent in the target quadrant compared (Figure 1J, P < 0.01 vs. Model group) to the Model group. These results suggest that XSB can enhance short-term memory, working memory, and spatial learning and memory in model mice.

H&E staining results revealed that the XSB-L, XSB-M, and XSB-H treatment groups exhibited well-organized and compact neuronal arrangements, with round nuclei and evenly distributed staining in both the cytoplasm and nuclei (Figure 2A). Nissl staining further demonstrated that in the XSB-treated groups, neurons were more densely packed and more intense staining than that in Model group. Notably, the XSB-M and XSB-H groups showed a significant increase in the number of Nissl bodies in the cortex (Figures 2B,C, P < 0.05) and hippocampal CA1 (Figures 2B,C, P < 0.05) and DG (Figures 2B,C, P < 0.05) regions compared to the Model group, with effects comparable to the positive control drug donepezil.

TEM analysis revealed that following XSB treatment, mitochondrial morphology was restored to typical elongated, fusiform shapes, with well-defined mitochondrial cristae and an intact nuclear membrane structure. Furthermore, the number of injured mitochondria was markedly decreased (Figures 2D,E, P < 0.001 vs. Model group), with effects similar to the positive control drug donepezil (Figures D,E, P < 0.001 vs. Model group). These results demonstrate that XSB effectively alleviates both pathological and ultrastructural damage in the cortex and hippocampus of model mice.

As shown in Figure 3, treatment with XSB-H significantly increased ACh levels (Figure 3A, P < 0.05 vs. Model group) and decreased AChE activity (Figure 3B, P < 0.05 vs. Model group) in the Model mice. Moreover, both XSB-M and XSB-H treatment significantly elevated the expression of ChAT and CHT1 in the cortex and hippocampus evidenced by western bolt and immunofluorescence data (Figures 3C–H), suggesting that XSB can effectively improve the cholinergic system in the brain of model mice.

As shown in Figure 4, treatment with XSB, particularly XSB-M and XSB-H, significantly improved microbial diversity, as reflected by lower Chao 1 (Figure 4A), Shannon (Figure 4B) and Simpson indices (Figure 4C). At the phylum level, XSB significantly decreased the relative abundance of Firmicutes while increasing Bacteroidota (Figures 4D,E), suggesting a shift towards a more balanced microbiota in these groups. The UniFrac distance analysis (Figure 4F) showed that the gut microbiota composition in XSB-treated groups was significantly closer to that of the Control group compared to the Model group.

Additionally, LEfSe analysis identified that s-Lactobacillus-johnsonii, g-Lactobacillus, o-unidentified Clostridia, f-Lachnospiraceae, g-unidentified-Lachnospiraceae, s-Clostridiales-bacterium-CIEAF-020, and dentified-Ruminococcaceae were enriched in Model group, while g-Enterococcus, o-Coriobacteriales, s-Enterococcaceae, g-unidentified-Muribaculaceae, p-Actinobacteriota, f-Eggerthellaceae which were identified beneficial for cognitive function were enriched in Control group (Figure 4G). Compared to Model group, Bacteroidia, Helicobacter-sp-MIT-02-6899, g-unidentified-Muribaculaceae, s-Faecalibacterium-prausnitzii and s-Lachnospiraceae-bacterium-610, associated with gut health and cognitive function, were enriched in XSB-H group (Figure 4H).

Moreover, correlation analysis between cognitive function-related indicators and beneficial bacteria revealed several taxa significantly associated with cognitive performance. For instance, g_Enterococcus, p_Actinobacteriota, o_Coriobacteriales, and f_Eggerthellaceae exhibited a positive correlation with cognitive function (Figure 4I, P < 0.001), whereas s_Helicobacter_rodentium and g_unidentified_Lachnospiraceae showed a negative correlation (Figure 4I, P < 0.01). Furthermore, in the XSB-treated groups, the abundance of the cognitively harmful taxa, including s_Helicobacter_rodentium and g_unidentified_Lachnospiraceae, was slightly decreased, while the cognitively beneficial taxa, including g_Enterococcus, p_Actinobacteriota, o_Coriobacteriales, and f_Eggerthellaceae, was slightly elevated (Figure 4J). These findings suggest that scopolamine-induced cognitive impairment leads to gut microbiota dysbiosis, while XSB administration partially restores the abundance of beneficial bacteria, thereby re-establishing microbial balance and subsequently improving cognitive function in the model mice.

As shown in Figure 5A, we identified 439 DMs between the Model and Control groups (218 upregulated and 221 downregulated), 127 DMs between the XSB-L and Model groups (37 upregulated and 90 downregulated), 158 DMs between the XSB-M and Model groups (88 upregulated and 70 downregulated), and 256 DMs between the XSB-H and Model groups (141 upregulated and 115 downregulated). The clustering heatmap in Figure 5B and the PLS-DA score plot in Figure 5C reveal that the XSB-H and XSB-M groups cluster more closely with the Control group. This suggests that intraperitoneal injection of scopolamine induces alterations in serum metabolites, and that XSB treatment can partially restore these metabolic disturbances. Figure 5D shows the top 30 metabolites with VIP >1 from the PLS-DA analysis. Pearson correlation analysis of these metabolites with cognitive function-related indicators indicated several significant correlations. Specifically, HMDB0000748 (L-3-Phenyllactic acid), HMDB0040891 (3′,4′,5′-Trimethoxycinnamyl alcohol acetate), HMDB0114073 (PE-NMe2 (18:2 (9Z, 12Z)/18:2 (9Z, 12Z))), and HMDB0052711 (TG (18:2 (9Z, 12Z)/22:4 (7Z,10Z,13Z, 16Z)/18:2 (9Z, 12Z))) were negatively correlated with cognitive function, whereas HMDB0015673 (Carglumic acid), HMDB0005772 (Postin), HMDB0000755 (Hydroxyphenyllactic acid), and HMDB0005781 (Glycitein) exhibited positive correlations (Figure 5E). Additionally, the abnormal levels of these metabolites were partially corrected by XSB treatment (Figure 5F).

KEGG enrichment analysis of the DMs with VIP >1 further revealed that XSB primarily restores metabolic balance in model mice by modulating key metabolic pathways, including glycerophospholipid metabolism, linoleic acid metabolism, alpha-Linolenic acid metabolism, ether lipid metabolism, glutathione metabolism, pentose phosphate pathway and folate biosynthesis (Figure 5G).

As illustrated in Figures 6A–D, the Model group exhibited 307 DEGs compared to the Control group, with 58 genes upregulated and 249 downregulated. In contrast, the XSB-L group demonstrated 139 DEGs relative to the Model group, comprising 43 upregulated and 96 downregulated genes. The XSB-M group exhibited 217 DEGs, with 115 upregulated and 102 downregulated, while the XSB-H group displayed 286 DEGs, including 179 upregulated and 107 downregulated, when compared to the Model group. Subsequently, these DEGs underwent hierarchical clustering analysis. As shown in Figure 6E, the XSB-H and XSB-M groups clustered closely with the Control group, indicating a potential reversal of gene expression patterns by XSB treatment.

The GO functional annotation revealed that the DEGs in the XSB-H vs. Model comparison were predominantly enriched in cellular components such as the dopaminergic synapse, DNA packaging complex, membrane raft, integral component of the presynaptic membrane, and acetylcholine-gated channel complex. Additionally, these genes were associated with diverse molecular functions, including receptor-ligand activity, cytokine activity, neuropeptide receptor binding, passive transmembrane transporter activity, oxidoreductase activity, and acetylcholine-gated cation-selective channel activity. Functionally, these DEGs were implicated in key biological processes such as the regulation of nerve impulse transmission, response to extracellular stimuli, megakaryocyte differentiation, hippocampal development, learning and memory, neuropeptide signaling pathways, superoxide metabolic processes, and interleukin-1 receptor binding (Figure 6G). Furthermore, KEGG pathway enrichment analysis indicated that the DEGs in the XSB vs. Model group were significantly associated with neuroactive ligand-receptor interactions, cytokine-cytokine receptor interactions, cell adhesion molecules, and the TNF signaling pathway (Figure 6I). The GO and KEGG enrichment analysis results of Model vs. Control group were as shown in Figures 6F,H. These findings suggest that intraperitoneal scopolamine injection disrupts hippocampal gene expression, while XSB treatment partially restores homeostasis through the regulation of neuroactive ligand-receptor interactions and modulation of neuroinflammatory pathways, both of which are critical for maintaining cognitive function.

Using MetaboAnalyst, joint pathway analysis was performed to integrate DEGs and DMs Figure 7A. The results indicated that XSB primarily regulates metabolic pathways, including linoleic acid metabolism, retinol metabolism, biosynthesis of unsaturated fatty acids, tyrosine metabolism, and glycerophospholipid metabolism (Figure 7B). To explore potential associations between DMs and gut microbiota, Spearman correlation analysis was performed. The results revealed significant associations between specific bacterial species and DMs (Figure 7C). For instance, bacteria beneficial to cognitive function, such as g_Enterococcus, p_Actinobacteriota, o_Coriobacteriales, and f_Eggerthellaceae, exhibited negative correlations with metabolites detrimental to cognitive function, including HMDB0114073, HMDB0040891, and HMDB0000748, while showing positive correlations with metabolites beneficial to cognition, such as HMDB0005781 and HMDB0015673. In contrast, bacteria detrimental to cognitive function, such as g_unidentified_Lachnospiraceae and s_Helicobacter_rodentium, demonstrated positive correlations with HMDB0114073, HMDB0040891, and HMDB0000748, but negative correlations with HMDB0005781 and HMDB0015673. These results highlight the critical role of XSB in modulating key metabolic pathways and the interplay between gut microbiota and metabolites, suggesting a potential mechanism through which XSB improves cognitive function by restoring microbial and metabolic homeostasis.

Based on the enrichment of inflammatory-related signaling pathways, such as the TNFα signaling pathway and arachidonic acid metabolism, identified in the KEGG analysis, we further assessed the levels of key inflammatory mediators-TNF-α, IL-1β, and MPO-in the hippocampus using Western blot analysis. As shown in Figures 8A,B, XSB treatment significantly reduced the protein expression levels of TNF-α (P < 0.01 vs. Model group), IL-1β (P < 0.05 vs. Model group), and MPO (P < 0.05 vs. Model group) in the hippocampus. Similar results were observed in immunofluorescence experiments (Figures 8C–E), further confirming the anti-neuroinflammatory effects of XSB. Moreover, XSB treatment increased the activity of SOD (P < 0.01 vs. Model group), elevated the levels of GSH (P < 0.01 vs. Model group), and reduced MDA levels (P < 0.01 vs. Model group), thereby alleviating oxidative stress damage in the brains of model mice (Figures 8F–H).

Given that KEGG enrichment analysis revealed the involvement of neuroactive ligand-receptor interaction signaling pathways, which are crucial for synaptic signaling, we then evaluated the expression of critical synaptic proteins-BDNF, SYN, and PSD95-using Western blot and immunofluorescence methods. As shown in Figures 9A,B, XSB treatment notably increased the expression of BDNF (P < 0.05), SYN (P < 0.001), and PSD95 (P < 0.001) compared to Model group. Immunofluorescence results (Figures 9C,D) supported these findings, further confirming the beneficial effects of XSB on synaptic damage. Additionally, XSB treatment significantly increased synaptic density (Figures 9E,F, p < 0.05 vs. Model group). These results suggest that XSB promotes the expression of BDNF, SYN, and PSD95 and enhances synaptic density, potentially contributing to the improvement of cognitive function in model mice.

The results demonstrated that transplantation of XSB-FM significantly decreased the escape latency (Figures 10B,C, p < 0.05 vs. Scop-FM group on 5th day), increased the platform crossing number (Figure 10D, P < 0.05 vs. Scop-FM group) and the time spent in target square (Figure 10E, P < 0.05 vs. Scop-FM group). Moreover, compared with Scop-FM group, transplantation of XSB-FM markedly boosted the AB-PAS positive cells (Figure 10F) in the intestinal mucosa, and upregulated ZO-1 (Figures 10G,I, p < 0.05) and Occludin (Figures 10H,J, p < 0.05) protein expressions in dementia mice induced by scopolamine. These findings indicated that transplantation of XSB-FM rescues cognitive dysfunction and intestinal barrier injures in AD model mice, which further confirmed that gut microbiota remodeling is essential for XSB’s therapeutic efficacy.