All male Sprague-Dawley rats (220 ± 10 g) were obtained from Si Pei Fu (Beijing) Biotechnology Co., Ltd. (Beijing, China) and they were housed under the conditions of controlled room temperature of around 25 °C and humidity at 55% ± 10%, with a 12-h light/dark cycle. The food and water were available to access unlimitedly. All animal experiments were ethically approved by the Animal Ethics Committee of the Yunnan Branch of the Institute of Medicinal Plant Development, Chinese Academy of Medical Sciences and Peking Union Medical College and followed the Compilation of Group Standards and Implementation Guidelines of the Chinese Society for Laboratory Animals.

Based on the method described by Longa et al. (1989), we used a tMCAO rat model to induce focal cerebral IR brain injury. After anes-thetizing the rats with isoflurane, the common and external carotid arteries were ligated on the right side of the rats, and the internal carotid artery was embolized from the common carotid artery using a threaded bolt (RWD Life Science Co., Ltd., China). After 1.5 h of embolization, the threaded bolt was withdrawn until the black marker point protruded through the skin and was cut off. The sham rats performed similarly to the model rats except for the embolization of the middle cerebral artery with a threaded bolt.

The experimental groups were the sham group (10 mL/kg, 0.5% CMC-Na), model group (10 mL/kg, 0.5% CMC-Na), B group (10 mL/kg, 90 mg/kg), DB group (540 mg/kg), different doses of DB + B group (10 mL/kg, DB, 270 mg/kg, 540 mg/kg; B, 30 mg/kg, 60 mg/kg, and 90 mg/kg), and nimodipine group (10 mL/kg, 10 mg/kg). DB and B were synthesized by Xishuangbanna Kanglong Pharmaceutical Co., Ltd., and nimodipine was synthesized by YABAO Pharmaceutical Co., Ltd. All groups were orally administered the therapy once daily for 7 days. On the seventh day, the tMCAO model was established after 1 h of administration.

After 22 h of IR, neurological deficits were assessed via mNSS by four people blinded to the experimental groups (Chen et al., 1996; Chen et al., 2001). The set of mNSS is listed in Supplementary Table S1. Briefly, neurologic deficits of rats can be graded according to score: 1–6 points indicated mild injury, 7–12 points indicated moderate injury, and 13–18 points indicated severe injury.

Rats were sacrificed after blood was collected from the abdominal aorta. Coronal sections, 2 mm thick, were cut through the cerebrum and subsequently incubated in a solution of 2,3,5-triphenyl tetrazolium chloride (TTC, J&K Scientific, China) for 15 min at 37 °C. Brain sections were fixed in 4% paraformaldehyde overnight. All sections were photographed and measured using ImageJ software to delineate the infarct.

The brain sections were dehydrated with a 15% sucrose solution for 2 h and then with a 30% sucrose solution until they sank to the bottom. These brain sections were then cut into frozen sections (CRYOSTAR NX50, Thermo Fisher Scientific) and subjected to HE staining for microscopic examination (DP43, Olympus).

Post-reperfusion cerebral blood flow was assessed using a laser speckle contrast imager (PeriCam PSI, PERIMED, Sweden), which provided high-resolution real-time blood flow images. The rats were anesthetized with isoflurane, and their fur was cut to expose the skull. Normal saline was administered to prevent skull dryness. Continually 1-min Speckle imaging (250 frames, 10s/frame) was obtained at 0.1 Hz by PIMSoft 1.11.2 software at baseline, after middle cerebral artery occlusion, after post-reperfusion, and after 24 h of post-reperfusion.

We selected the cerebral cortex of the ischemic hemisphere of the sham, tMCAO, and DB + B groups to sequence all samples using the DNBSEQ high-throughput sequencing service (Beijing Genomics Institution, China). We used the Beijing Genomics Institution’s Dr. Tom platform to perform differential gene, Gene Ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses.

Take 200 μL of the serum, add 800 μL of pre-cooled acetonitrile (at 4 °C), vortex for 10 s, leave to stand for 10 min at 4 °C, and then centrifuge at 4°C and 13,000 revolutions per minute for 15 min. Take 800 μL of the supernatant, dry with the Vacuum Centrifuge Concentrator, add 400 μL of 50% precooled acetonitrile to reconstitute, vortex for 10 s, centrifuge at 4 °C and 13,000 revolutions per minute for 15 min. Finally, extract the supernatant to obtain results. Quality control samples were prepared using equal amounts of the supernatant from each sample.

Online LC-MS analysis was conducted utilizing an Ultrahigh-performance Liquid Chromatograph (UPLC, Shimadzu Nexera Series) combined with a Quadrupole Time-of-Flight Mass Spectrometry (Q-TOF/MS, AB SCIEX X500B QTOF). Using the electric spray positive and negative ion modes with an ESI ion source. Nitrogen was utilized as the curtain gas (CUR), ion source gas 1 (GS1), ion source gas 2 (GS2), and collision gas (CAD). The MS parameters are listed in Table 1. The primary mass spectrum was scanned from 60 to 1,250 Da. The collection of primary and secondary MS data was based on information-dependent acquisition, and the 12 highest peaks with a response value of above 100 cps were selected to scan the secondary mass spectrum. The secondary mass spectrum was scanned from 50 to 1,250 Da.

The column was T3 with column temperature at 35 °C. The mobile phase A was water with 0.1% formic acid while the mobile phase B was acetonitrile with 0.1% formic acid. The gradient elution sequences are listed in Table 2. The injection volume is 5 μL at a flow rate of 0.3 mL per minute.

The intestinal contents of rats were collected aseptically. The samples were then sequenced using Metware Biotechnology Co., Ltd. for 16s rRNA amplicon sequencing. According to a previous report, the total genomic DNA of the microbiome was extracted using the cetyltrimethylammonium bromide method. 16S rRNA/18S rRNA/ITS genes of distinct regions were amplified with 15 L of Phusion^® High-Fidelity PCR Master Mix (New England Biolabs). A mixture of the PCR products was then purified. Sequencing libraries were generated usingTruSeq^® DNA PCR-Free Sample Preparation Kit (Illumina, USA). The quality of the library was evaluated using a Qubit@ 2.0 Fluorometer (Thermo Scientific) and an Agilent Bioanalyzer 2,100 system. Then, the library was sequenced on an Illumina NovaSeq platform, generating 250 bp paired-end reads. Finally, the analysis of the data was accomplished using the Metware Cloud platform (https://cloud.metware.cn).

All data were used GraphPad Prism 9 software for statistical analysis and presented as mean ± SEM (n ≥ 3). The comparison was analyzed using the t-test, one-way ANOVA, Mann-Whitney U test, or Wilcox test. p < 0.05, the indication of statistical significance.

The neuroprotective effects of different proportional doses of DB + B on IR brain injury were detected using the mNSS and cerebral infarct size. Compared to the model group, the administration of DB, different proportions of DB + B, and nimodipine significantly decreased the neurological deficit scores (Figure 1A). The results showed that DB + B improved motor nerve function in tMCAO rats, and the effect of 540 mg/kg DB + 90 mg/kg B was similar to that of nimodipine. Furthermore, TTC staining showed that the administration of DB, different proportions of DB and B, and Nimodipine significantly reduced the brain infarct volume (Figures 1B, C) and that the 540 mg/kg DB + 90 mg/kg B group had the most visible effect on alleviating neurologic deficits and infarction volumes in tMCAO rats.

To evaluate neuronal loss histopathologically, brain tissue slices from rats were processed using HE staining. Compared with the sham group, HE staining showed necrosis and pyknosis of nerve cells, deep staining of nuclei, inflammatory cell infiltration, and gliosis in the model group (Figure 1D). The brain tissue of rats in the different drug delivery groups showed a reduction in inflammatory cell infiltration or gliosis and some improvements in neuronal necrosis and loss in HE staining, notwithstanding that it still had some impairment, but it was not as serious as the model group. Combined with the results of mNSS and TTC staining, the 540 mg/kg DB + 90 mg/kg B group showed the most significant therapeutic effect against focal cerebral IR in rats. A dose of 540 mg/kg DB + 90 mg/kg B was administered for all subsequent experiments.

Additionally, to determine whether the DB + B improved stroke outcome in tMCAO rats was associated with improvement of local cerebral blood flow, we assessed the laser speckle imaging at different time points (Figure 1E, F). Compared to the model group, regional cerebral blood flow significantly improved after 24 h of reperfusion in the DB + B group. In a word, DB + B can alleviate IR brain injury in tMCAO rats.

RNA-sequencing was used to further explore the critical pathway for the neuroprotective effect of DB + B in IR brain injury. The significantly changed DEGs were clustering analyzed, and the thresholds of DEGs were set to log2FC ≥ 2, qvalue≤0.05. Approximately 3,677 genes differed significantly between the sham and model groups, of which 2,389 genes were upregulated and 1,288 genes were downregulated in the model group compared to the sham group (Supplementary Figure S1A). Meanwhile, 2,794 genes differed significantly between the DB + B and model groups, of these 1,024 genes were upregulated, and 1752 genes were downregulated in the DB + B group compared to the model group (Figure 2A). The VENN diagram shows the DEGs between the sham/model groups and DB + B/model groups (Figure 2B). We selected the top 100 significantly altered DEGs of the sham/model groups and DB + B/model groups, as shown in heatmaps (Figure 2C; Supplementary Figure S1B). Furthermore, we verified the results of RNA-Seq by the Western blot experiments of c-Jun and IL-6 (Supplementary Figures S1C–E).

GO is divided into biological processes, molecular functions, and cellular components. The DEGs were annotated using GO. According to GO functional enrichment analysis, it can be seen that the DEGs of the DB + B group compared to the model group are mainly concentrated in the biological processes of inflammatory response, immune response, immune system process, positive regulation of pro-inflammatory factor such as IL-6 and IL-8, chemokine-mediated signaling pathway, and so on (Figure 2D)). Similar results were found in the model group compared to the sham group (Supplementary Figure S1F).

According to KEGG analysis (Figure 2E; Supplementary Figure S1G), DB + B was found to play a role in cytokine-cytokine receptor interaction, hematopoietic cell lineage, neuroactive ligand-receptor interaction, cell adhesion molecules, chemokine signaling pathway, TNF signaling pathway, NF-κB signaling pathway, IL-17 signaling pathway, Th17 cell differentiation. Most were linked to the immune system, metabolism, signal transduction, nervous system, and infectious microbes (Figure 2F; Supplementary Figure S1H).

Serum metabolites were assessed by UPLC-Q/TOF-MS, and the total ion chromatograms of the three groups were separated well in both positive and negative ion modes within 21 min. First, PCA analysis of metabolomics demonstrated that the sham, model, and DB + B groups were distinct, and the DB + B group had a tendency to relocate to the sham group (Figure 3A, B). The OPLS-DA model was used to further analyze the metabolic differences between the sham, model and DB + B groups (Supplementary Figure S2A, C, E, G). In negative ion mode, the R²Y and Q² values were 0.965 and 0.823 when the model group was compared to the sham group, and the R²Y and Q² values were 0.925 and 0.809 when the DB + B group was compared to the sham group (Supplementary Figure S2B, D). In the positive ion mode, the R²Y and Q² values were 0.958 and 0.882 when the model group was compared to the sham group, and the R²Y and Q² values were 0.873 and 0.647 when the DB + B group was compared to the model group (Supplementary Figure S2F, H). Furthermore, the R²Y-intercepts and Q²-intercepts were lower than the original values, demonstrating the high predictivity and accuracy of fit of the established model.

Potential biomarkers in serum that helped to distinguish the model group from the sham group were selected with the following criterion: VIP ≥1 and p ≤ 0.05 (Figure 3C–F). A total of 30 endogenous metabolites with significant differences were identified using available biochemical databases (Table 3), the trend of these differential metabolites was shown in Figure 3G. MetaboAnalyst 5.0 was used to analyze the pertinent metabolic pathways of 40 potential biomarkers, including organic acids and derivatives, lysophospholipids, and organic heterocyclic compounds. The glycine, serine, and threonine metabolism; glycerophospholipid metabolism; arachidonic acid metabolism; linoleic acid metabolism; biosynthesis of unsaturated fatty acids; fatty acid biosynthesis; fatty acid elongation and degradation; pyrimidine metabolism; arginine and proline metabolism; and primary bile acid biosynthesis were found to be relevant to IR brain injury in rats (Figure 3H). In the presence of DB + B, 24 potential biomarkers were significantly regulated (p < 0.05), including creatine, 3-indolepropionic acid, lysoPC(15:0), lysoPC (18:2), lysoPC(18:0), tetradecanoic acid,γ-linolenic acid, palmitoleic acid, lysoPE (18:1), pentadecanoic acid, 1-heptadecanoylglycerophosphoethanolamine, lysoPC(18:3), N-(2-Furoyl)glycine, cytidine, lysoPC(22:4), arachidonic acid, cytarabine, lysoPE (18:0), 11.14.17-eicosatrienoic acid, oleic acid, palmitic acid, lysoPE (20:1), heptadecanoic acid, lysoPE (20:0).

In the KEGG analysis of DEGs, most of the pathways belonged to the immune system, so Spearman’s correlation analysis of DEGs in the immune system and metabolites was performed between the DB + B and model groups to further assess the molecule mechanism of DB + B in preventing IR brain injury. The clustered heatmap showed that the metabolites are mainly positively correlated with pro-inflammatory genes, including IL-17f, Il-17ra, Jun, IL-6, Myd88, and Tlr4 (Figure 4A). We showed the correlation network graph between these DEGs and metabolites (Figure 4B), in which the lysoPC(18:3) and lysoPE (22:4) had the most significant correlations with most of the DEGs. Most of these DEGs belong to the IL-17 signing pathway (Figure 4C).

16S rRNA amplicon sequencing analysis was carried out to investigate the effect of DB + B on the dysbiosis of intestinal flora in tMCAO rats. We observed differences in the abundance of these groups at each classification level (i.e., phylum, class, order, family, genus, and species) when analyzing the species composition of the intestinal flora in the groups (shown in Figure 5A; the species composition of intestinal microbes in each sample is shown in Supplementary Figure S3A). The ɑ-diversity and β-diversity analysis was performed to detect significant differences in the intestinal flora within and among samples. The treatment of DB + B of IR brain injury had no apparent effect on ɑ-diversity (Figure 5B). PCoA indicated that the DB + B and model groups were partially separated (Figure 5C). Based on the metastasis analysis, f_Selenomonadaceae, g_Anaerovibrio, s_Bifidobacterium_adolescentis, and s_Brevundimonas_vesicularis were significantly altered in the intestinal flora (Figure 5D). Based on the linear discriminant analysis effect size, we found that the intestinal flora that differed significantly between the groups could be divided into 16 categories (Figure 5E). Statistically significant biomarkers between groups were species with an LDA value of >2 (Figure 5F). The t-test results showed that f_Prevotellaceae, P_Actinobacteriota, g_Muribaculum, and s_Muribaculum_ intestinale significantly changed between the model and sham groups (Supplementary Figure S3B), and g_Eisenbergiella and o_unidentified_Bacilli significantly changed between the DB + B and model groups (Supplementary Figure S3C). Additionally, the heat map shows the differential intestinal flora found by metastasis analysis, linear discriminant analysis effect size, and T-test (Figure 5G).

We found significant associations between serum metabolites and intestinal flora using Spearman’s correlation analysis between the DB + B and model groups (Figure 6A). lysoPC(18:3) N-(2-Furoyl)glycine and cytidine are positively correlated with o_unidentified_Bacilli and s_Bacteroides_caecimuris. LysoPC(18:1), lysoPE (18:0), palmitic acid, cytarabine, and 3-indole propionic acid are positively correlated with o_Pseudomonadales, f-Pseudomonadaceae, and g-Pseudomonas. The network graph showed the correlation between these flora and metabolites (Figure 6B). o_Pseudomonadales and s_Bacteroides_caecimuris had the most significant correlations with the most serum metabolites. The results indicated that the DB + B improvement in IR brain injury was associated with changes in the intestinal flora and its metabolites, and o_Pseudomonadales and s_Bacteroides_caecimuris are the most important intestinal flora.