Hua-Feng-Dan is supplied by Zunyi Liao Yuan He Tang Pharmaceutical (batch 20230601, Zunyi, China). The prescription of Hua-Feng-Dan is as follows: Yaomu, cinnabar, Moschus, Perilae Folium, Bombyx Batryticatus, Scorpio, Arisaematis Rhizoma Preparatum, Atractylodis Rhizoma, realgar, borax, Crotonis Semen Pulveratum, Borneolum Syntheticum, Gastrodile Rhizoma, Schizonepetae Herba, and Santali Albi Lignum (Table 1). In addition to the initial three drugs, the remaining twelve drugs are finely crushed into powder (100 mesh, >95%), mixed with Yaomu and glutinous rice powder (100 mesh, >95%), and formed into pills by adding water; cinnabar (200 mesh, >95%) and Moschus are rapidly ground well and uniformly spread on the surface of the pills to obtain vermilion-red water pills, 0.12 g/pill.

The Yaomu is made by crushing Typhonii Rhizoma, Pinelliae Rhizoma, Arisaematis Rhizoma, Aconiti Radix and Curcumae Radix into a fine powder (100 mesh, >95%), thoroughly mixing them, and then adding into Cow Bile and Medicated Leaven (Shen-Qu) for fermentation. To characterize the compositions of Hua-Feng-Dan, we utilized UPLC-Q-Exactive (Thermo, MA, United States) to detect the water-soluble sample of Hua-Feng-Dan (Supplementary Figure S1; Supplementary Table S1).

Male Sprague-Dawley rats free of specific pathogens (180–220 g) were purchased from the Institute of Laboratory Animals at the Guizhou University of Traditional Chinese Medicine (Guizhou, China), operating under license SCXK (Qian) 2021-0003. Animal procedures were approved by the Ethics Committee for Animal Experiments at Guizhou University of Traditional Chinese Medicine (approval 20230067). Rats were housed in autoclaved, ventilated cages with autoclaved bedding and sterilized feed. They were maintained on a 12-h light/dark cycle at a temperature of 23°C ± 2°C and relative air humidity of 60%–70%.

The 120 rats were randomly assigned to 20 rats each in sham-operated (Sham), model (MCAO), Hua-Feng-Dan at low dose (HFD-L, 0.162 g/kg/day), Hua-Feng-Dan at middle dose (HFD-M, 0.324 g/kg/day), Hua-Feng-Dan at high dose (HFD-H, 0.648 g/kg/day), and nimodipine (NMDP, 0.19 g/kg/day) groups. Rats were randomized to receive, by oral gavage, saline instead of drug treatment or one of the following drugs dissolved in saline: the calcium channel blocker nimodipine (NMDP, batch 211274, Yabao Pharmaceutical, Shanghai, China), which is widely used to ischemic cerebrovascular disease; and Hua-Feng-Dan. 10 days was the period of the treatment, with modeling conducted 1 h after the drug administration on the 7th day. Three rats were used for 2,3,5-triphenyltetrazolium chloride (TTC) staining, three rats were used for histopathological test, four rats were utilized for brain water content, four rats were utilized for brain metabolomics and 16S rRNA sequencing analysis, six rats were chosen to undergo enzyme-linked immunosorbent assay (ELISA) and gut metabolomics analysis in each group.

Rats were fasted for 12 h, anesthetized with isoflurane inhalant (4% for induction; 1.5% for maintenance), and placed in a supine position. Throughout the following procedures, animals were placed on a thermostatic heating pad to maintain their body temperature at about 37°C. A longitudinal incision was made along the middle of the neck, then the fascia and muscles were separated to expose the left common carotid artery, internal carotid artery, and external carotid artery. The left common and external carotid arteries were ligated, while the internal carotid was clamped with a microvascular clamp. A nylon wire of diameter 0.25 mm was inserted from the left common carotid into the internal carotid for approximately 18.5 ± 0.5 mm, then the incision was rapidly closed. After 2 h of occlusion, the nylon wire was pulled out a certain length to allow reperfusion. Rats were kept under the oven lamp until they recovered from anesthesia. As controls, sham-operated rats underwent the same procedures except that no nylon wire was inserted.

Rats were assessed for nerve damage using a previously described scoring system (Longa et al., 1989). Points were assigned as follows: 0, normal walking without apparent nerve damage; 1, weak contralateral forelimb and inability to fully extend it; 2, circling toward the contralateral side; 3, tilting toward the contralateral side; and 4, failure to walk spontaneously and diminished or no consciousness. Only animals scoring 1–3 immediately after awakening from reperfusion were randomized to receive drug treatments. The assessment was repeated at 72 h after reperfusion.

At 72 h following reperfusion, the brains were frozen at −20°C for 20 min and then sliced into six sections along the coronal plane, immersed in 2% (v/v) TTC (batch 23206277, Sigma, St Louis, MO, United States), stained at 37°C without light for 15 min to visualize the infarct volume, and fixed in 4% paraformaldehyde solution (batch 22182401, Biosharp, Anhui, China). Infarct volume and total brain volume across all six sections were determined using ImageJ 2.9.0 (US National Institutes of Health, Bethesda, MD, United States), and extent of cerebral infarction was calculated as. Cerebral infarction extent  % = infarct volume  /  whole brain volume × 100 % .

The brains of rats were immediately removed 72 h after modeling. Carefully removing the leptomeninges, cerebellum, and brain stem, any liquid that remained on the surface was absorbed using filter paper. The cerebral hemispheres of the infarcted side were promptly weighed to determine their wet weight using an analytical balance. The brain was placed in an electric thermostat where it was baked at 110°C for 24 h; this dry weight was recorded. The water content of ischemic brain was calculated using the following formula: Brain water content  % = wet weight − dry weight   /  wet weight × 100 % .

Brain and colon tissues were fixed in 4% paraformaldehyde solution, embedded in paraffin, cut into sections 5 μm and 4 μm thick, respectively, and stained with hematoxylin-eosin (H&E, batch G1003, Servicebio, Hubei, China). Additionally, we stained brain sections utilizing the TUNEL assay kit (batch G1504, Servicebio). Histopathology in the cerebral cortex, hippocampus and colon was assessed using an orthoptic light microscope (Eclipse E100, Nikon, Tokyo, Japan) equipped with a DS-U3 imaging system (Nikon). Observations and image acquisition of TUNEL were performed using an ECLIPSE C1 orthogonal fluorescence microscope (Nikon). The number of TUNEL positive cells at three distinct locations within each section was quantified using Image-Pro Plus 6.0 (Media Cybernetics, MD, United States).

At 72 h of reperfusion, rat serum was assayed for tumor necrosis factor (TNF)-α, interleukin (IL)-1β and IL-6 as indices of inflammation. It was also assayed for lipopolysaccharide (LPS), diamine oxidase (DAO), and d-lactate (d-LA) as indices of permeability of the intestinal epithelium; for total cholesterol (T-CHO) as an index of dyslipidemia; and for total superoxide dismutase (T-SOD) and malondialdehyde (MDA) as indices of oxidative stress. The first six assays were ELISA with kits purchased from Shenzhen Ziker Biological Technology (Shenzhen, China), while the last three assays were utilized kits provided by the Nanjing Jiancheng Bioengineering Institute (Nanjing, China).

After 72 h of modeling, cecal contents were collected into sterile 1.5-mL microcentrifuge tubes, labeled, snap-frozen in dry ice, and stored at −80°C. DNA was extracted from thawed samples using the Fast Pure Stool DNA Isolation Kit (MJYH, Shanghai, China), the variable region V3-V4 within 16S rRNA was amplified using primer 338F-806R through PCR, and the PCR products were used to create a paired-end (PE) library, which was sequenced on a NextSeq 2000 system (Illumina, California, United States). The detailed parameters of the 16S rRNA sequencing can be found in the Supplementary Material. Raw sequences were deposited into the Sequence Read Archive at the National Center for Biotechnology Information under accession code PRJNA1129823 (www.ncbi.nlm.nih.gov).

Sequencing PE reads were split, and double-ended raw sequences were checked for quality using FASTP 0.19.6 software (Chen et al., 2018), then spliced using FLASH 1.2.11 software (Magoč and Salzberg, 2011) and subjected to noise reduction using the plugin DADA2 (Callahan et al., 2016) within the Qiime2 pipeline (Bolyen et al., 2019), resulting in amplicon sequence variants (ASV). The number of sequences from each sample was rarefied to 28,770, and the ASVs were analyzed for species taxonomy based on the Sliva 16S rRNA database (version 138; www.arb-silva.de), utilized naïve Bayes classifier in the Qiime2.

Based on relative abundances of microbial taxa, we calculated α-diversity using MOTHUR 1.30.2 software (Schloss et al., 2009), and inter-group differences were assessed for significance using the Wilcoxon rank sum test. We also calculated β-diversity using principal coordinate analysis (PCoA) based on Bray-Curtis dissimilarities. Bar charts were used to compare gut microbiomes among groups at the levels of phylum, family and genus. “Linear discriminant analysis (LDA) effect size (LEfSe)” (Segata et al., 2011) was used to determine gut microbial taxa with significant differences in abundance from phylum to genus level; differential bacteria were defined as those whose relative abundance differed between groups with a LDA score >4 and P < 0.05. Based on the reconstructed gut microbiomes, we predicted functional information of the microbial communities in the samples using PICRUSt2 2.2.0-b software (Douglas et al., 2020). Differences between groups were assessed for significance using the Wilcoxon rank-sum test.

Aliquots of cecal contents (see Section 2.9) and brain tissues weighing 50 ± 5 mg were respectively placed into 2-mL microcentrifuge tubes, to which were added a grinding bead with diameter of 6 mm and 400 µL of methanol/water (4/1, v/v) containing the following internal standard l-2-chlorophenylalanine (0.02 mg/mL, Adamas-beta), respectively. The sample was ground for 6 min using a freezing tissue grinder (-10°C, 50 Hz), cryo-sonicated for 30 min at 5°C (40 kHz), left at -20°C for 30 min, then centrifuged for 15 min at 13,000 g at 4°C. Reproducibility of analysis results was assessed by mixing 20 µL equal aliquots of different samples to obtain quality control (QC) samples.

The resulting supernatant was injected onto a Vanquish Horizon ultra-high performance liquid chromatography (Thermo, MA, United States) equipped with a ACQUITY UPLC HSS T3 column (100 mm long, 2.1 mm inner diameter, 1.8 µm pore size; Waters, Milford, MA, United States). The following chromatographic parameters were used: the mobile phase A was 95% water and 5% acetonitrile (containing 0.1% formic acid), while mobile phase B was 47.5% acetonitrile, 47.5% isopropanol, and 5% water (containing 0.1% formic acid), injection volume, 3 μL; column temperature, 40°C. The detailed parameters of the LC-MS program can be found in the Supplementary Material.

Eluted compounds were identified using Q-Exactive HF-X mass spectrometry (Thermo) at a sheath gas flow rate, 50 Arb; auxiliary gas flow rate, 13 Arb; capillary temperature, 325°C; a mass scan range, 70-1,050; full resolution, 60,000; MS/MS resolution, 7,500; collision energy, 20/40/60 Ev; and spray voltage, ±3.5 kV. Raw chromatographic data were corrected for retention time and peaks were aligned using Progenesis QI 3.0 software (Waters). Peaks were identified through comparison with the Human Metabolome Database (HMDB, www.hmdb.ca), Metlin database (https://metlin.scripps.edu), and a self-built database (Majorbio Bio-Pharm Technology CO., LTD., Shanghai, China), based on a mass error threshold of 10 ppm and secondary mass spectrometry match scores. Peak annotations were verified manually. Data were analyzed in both positive and negative ion modes.

The stability of the internal standards was assessed by checking whether the z-score values fell within two times the standard deviation. Principal component analysis (PCA) and orthogonal partial least-squares discriminant analysis (OPLS-DA) were utilized to assess the quality of the LC-MS data. Statistical significance between groups of metabolites was assessed using a two-tailed unpaired Student’s t-test. Differential metabolites screening criteria included OPLS-DA Variable Importance in the Projection (VIP) > 1 and P < 0.05. Metabolic pathway annotation was conducted using the KEGG database (www.kegg.jp/kegg/pathway.html). Pathway enrichment analysis was performed using the Python package, and relevant metabolic pathways were identified through Fisher’s exact test.

Data were analyzed statistically using SPSS 26.0 software (IBM, Chicago, IL, United States). Data were reported as mean ± standard deviation (SD). Differences between two groups were assessed for significance using Student’s two-tailed t-test, while differences among at least three groups were assessed using a one-way ANOVA and Tukey’s post hoc tests. Differences were considered statistically significant at P < 0.05. Potential correlations between relative abundance of microbiota and levels of gut metabolites were explored using the Spearman algorithm. Data were plotted using GraphPad Prism 9.0 (GraphPad Software, San Diego, CA, United States).

As expected, MCAO created a cerebral infarct area on the infarct side of the brain (Figure 1A), which Hua-Feng-Dan significantly reduced in a dose-dependent manner (Figure 1B). In fact, the infarct volume after treatment with the highest dose was comparable to that after treatment with nimodipine. Similarly, the highest doses of Hua-Feng-Dan and nimodipine significantly reduced brain water content following ischemia, while HFD-L group did not differ considerably from MCAO groups (Figure 1C). In a dose-dependent manner, Hua-Feng-Dan also reduced neurological score (Figure 1D).

Hua-Feng-Dan ameliorated the ischemia-induced sparse structuring, neuronal necrosis, tissue disorganization, and invasion of macrophages and glia in the cerebral cortex; and it ameliorated the crumpling, intracellular vacuolization and excessive intercellular spacing of pyramidal cells in the hippocampus (Figure 2A). The mechanism behind nerve cell death was examined by TUNEL staining (Figure 2B). The MCAO group had a significant number of apoptotic nerve cells, and that the damage was decreased in a dose-dependent manner by Hua-Feng-Dan treatment (Figure 2C). These findings presented Hua-Feng-Dan can lessen both the neurological and histopathological damage caused by ischemia.

Neither MCAO nor treatment with the highest dose of Hua-Feng-Dan significantly altered the richness or diversity of the gut microbiome (Figures 3A–D), based on analysis of 908 ASVs (616 unique), 737 ASVs (512 unique), and 1,013 ASVs (716 unique) were identified in the Sham, MCAO, and HFD groups, respectively (Figure 3E). Nevertheless, principal coordinate analysis (PCoA) indicated that the MCAO significantly altered the β-diversity of gut microflora, with the highest dose of Hua-Feng-Dan shifting the structure of gut microbiota closer to that of sham-operated animals (Figure 3F). These results suggest that IS can cause dysbiosis of the gut flora, which Hua-Feng-Dan can renormalize.

Specifically, MCAO was associated with a significant increase in relative abundance of Proteobacteria, Enterobacteriaceae, Enterococcaceae, Erysipelotrichaceae, Escherichia-Shigella, Enterococcus and Clostridium_innocuum_group (Figures 3G–I). It was also associated with a significant decrease in relative abundance of Firmicutes, Actinobacteriota, Lachnospiraceae, Atopobiaceae, Blautia, Bacillus, Candidatus_Stoquefichus and Lactobacillus. Hua-Feng-Dan significant reversed these changes in relative abundance, while also increasing the abundance of unclassified__f__Lachnospiraceae, Ruminococcus_torques_group, norank__f__norank__o___Clostridia_UCG-014, Monoglobus and Akkermansia.

Comparison across treated, untreated and sham-operated animals identified several genera whose relative abundance differed the most, among them Escherichia-Shigella, Enterorhabdus, Prevotella, norank_f__Butyricicoccaceae, Coprococcus (Figure 4A). We identified several species that predominated in untreated animals but not in the other two groups: p_Proteobacteria, f_Enterobacteriaceae, g_Escherichia-Shigella, g__Clostridium_innocuum_group, g__Allobaculum, g__Romboutsia and g__Enterococcus. We also identified several species that predominated in treated animals but not in untreated ones, among them c_Clostridia, o_Lachnospirales and f_Lachnospiraceae (Figures 4B–E).

The abovementioned changes in the composition of gut microbiota were associated with predicted changes in microbial function. MCAO was associated with biosynthesis of aminoacyl-tRNAs and amino acids, as well as metabolism involving purines, pyrimidines, alanine, aspartate and glutamate (Figure 4F). Hua-Feng-Dan treatment was also associated with metabolism of alanine, aspartate and glutamate, as well as metabolism of butanoate and histidine, biosynthesis of arginine, and degradation of lysine (Figure 4G).

The z-score values of the QC samples were within ±2 SD, and the total ion chromatogram peak shapes were well-defined and consistent distributions, suggesting that the equipment is stable and the data quality is reliable (Supplementary Figures S2, S3). To explore these metabolic changes in detail, we performed metabolomics on cecal contents from the different animal groups. Metabolomic profiles differed significantly across untreated, treate d and sham-operated animals (Figures 5A, B), with the data suggesting that Hua-Feng-Dan partially reversed ischemia-induced changes to renormalize gut microbial metabolism. These results are unlikely to reflect overfitting based on permutation testing (Figures 5C, D) and are therefore likely to be reliable. Indeed, volcano plots highlighted obvious differences in metabolite levels across the three groups (Figures 5E, F). Metabolic data in the positive ion mode are presented in the Supplementary Figure S4.

Extrapolating from levels of individual metabolites to entire metabolic pathways, we found that ischemic brain injury significantly affected pathways involving metabolism of lipids and amino acids, biosynthesis of other secondary metabolites, which primarily affect the body’s endocrine, nervous, and immune systems (Figures 5G, H). The MCAO also affected the biosynthesis of steroid hormones, isoflavone, and primary bile acids; metabolism of glycerophospholipids, sphingolipids, arachidonic acid, tryptophan, histidine, nucleotides, α-linolenic acid, and butanoate; and lysine degradation (Figures 5I, J). Hua-Feng-Dan significantly reversed these ischemia-induced changes, while also altering pathways involving degradation of steroids, biosynthesis of plant-derived secondary metabolites, and metabolism of arginine and proline.

Using carefully defined cut-offs (see Methods), we identified 39 metabolites whose levels in the gut were significantly altered by brain ischemia and were partially renormalized by Hua- Feng -Dan (Table 2). These potential biomarkers are 3-methylindole and other metabolites known to be involved in metabolic pathways involving amino acids (e.g., prolyl-glutamate, N2-succinyl-l-ornithine, N-arachidonoyl phenylalanine), SCFAs [e.g., 3-pyridylacetic acid, l-2-amino-3-(1-pyrazolyl)propanoic acid, 2-hydroxy-3-methylbutyric acid, 4-(glutamylamino) butanoate, mevalonic acid], essential fatty acids (dihomo-α-linolenic acid, dihomolinoleic acid, 10-nitrolinoleic acid), and nucleosides (1-methylpseudouridine, 1-methyladenosine, 5-acetylamino-6-amino-3-methyluracil).

We searched for Spearman correlations between the relative abundance of the 30 most abundant microbial taxa and the levels of the 39 potential metabolite biomarkers (Figure 6). Among taxa predominating after ischemic injury, the relative abundance of g_Escherichia-Shigella, g__Clostridium_innocuum_group, g__Romboutsia, g__Allobaculum and g__Enterococcus correlated positively with levels of 5-acetylamino-6-amino-3-methyluracil, chenodeoxycholylthreonine, 2′-deoxyadenosine 5′-phosphate, deoxycytidylic acid, phenylalanyl-prolyl-arginine nitrile, 1-methyladenosine, and 3-methylindole. Conversely, the abundance of these taxa correlated negatively with 4-(glutamylamino) butanoate, 2-hydroxy-3-methylbutyric acid, dihomo-α-linolenic acid, dihomolinoleic acid, and 10-nitrolinoleic acid. Among taxa predominating after treatment with Hua-Feng-Dan, the relative abundance of g__unclassified_f__Lachnospiraceae, g__norank_f__noranko__Clostridia_UCG-014 and g__Ruminococcus_torques_group showed almost the opposite trend compared with the predominant bacteria in MCAO group. The abundance of these taxa correlated positively with 4-(glutamylamino) butanoate, 2-hydroxy-3-methylbutyric acid, and 10-nitrolinoleic acid.

In the Sham group, the intestinal glands were tightly arranged, intact structure of the connective tissue, abundant goblet cells, and there was no obvious inflammatory cell infiltration. Hua-Feng-Dan treatment significantly improved ischemia/reperfusion-induced mucosal epithelial cell edema and necrosis, goblet cell reduction, intestinal gland dilatation and disorganization, lamina propria edema, plasma layer vasodilatation, and lymphocyte infiltration, while nimodipine did not alleviate the mucosal epithelial lesions (Figure 7A). The findings indicated that Hua-Feng-Dan had a protective effect on the integrity of the intestinal barrier.

MCAO significantly increased levels of lipopolysaccharide (LPS), diamine oxidase (DAO) and d-lactate (d-LA) in serum, consistent with translocation of gut microbiota into the circulation as a result of partial permeabilization of the gut epithelium (Figures 7B–D). Consistently, it significantly increased levels of the pro-inflammatory factors tumor necrosis factor (TNF)-α, interleukin (IL)-1β and IL-6 (Figures 7E–G); it increased total cholesterol (T-CHO) level, indicating dyslipidemia; and it increased the level of malondialdehyde (MDA) but decreased total level of superoxide dismutase (T-SOD), indicating greater oxidative stress (Figures 7H–J). All these ischemia-induced changes were significantly reversed by Hua-Feng-Dan or nimodipine, suggesting that the traditional Chinese medicine can repair ischemia-induced permeabilization of the intestinal barrier, which in turn mitigates systemic inflammatory responses and oxidative stress.

We searched for Spearman correlations between the relative abundance of the 30 most abundant microbial taxa and the levels of the 39 potential metabolite biomarkers with the levels of the three markers of gut barrier permeability and three pro-inflammatory cytokines. Among taxa predominating after ischemic injury, the relative abundance of Escherichia-Shigella, Clostridium_innocuum_group, Allobaculum, Romboutsia and Enterococcus correlated positively with levels of the six marker molecules (Figures 8A, C). Conversely, among taxa predominating after treatment with Hua-Feng-Dan, the relative abundance of unclassified_f__Lachnospiraceae and norank_f__norank_o__Clostridia_UCG-014 correlated negatively with the six marker levels.

Levels of the six marker molecules correlated negatively with levels of the metabolites 1-methylpseudouridine, 3-methylindole as well as metabolites from pathways involving metabolism of amino acids (e.g., prolyl-glutamate, N2-succinyl-l-ornithine, N-arachidonoyl phenylalanine), SCFAs [e.g., 3-pyridylacetic acid, l-2-amino-3-(1-pyrazolyl)propanoic acid, 2-hydroxy-3-methylbutyric acid, 4-(glutamylamino) butanoate, mevalonic acid], and essential fatty acids (e.g., dihomo-α-linolenic acid, dihomolinoleic acid, 10-nitrolinoleic acid) (Figures 8B, D). Conversely, marker levels correlated positively with levels of 5-acetylamino-6-amino-3-methyluracil, chenodeoxycholylthreonine, phenylalanyl-prolyl-arginine nitrile, 2′-deoxyadenosine 5′-phosphate, deoxycytidylic acid, propylenediamine tetra-acetic acid, and 1-methyladenosine. The analyses in Section 3.3 (Table 2) have shown that Hua-Feng-Dan downregulates these seven metabolites, while upregulating the related amino acids and SCFAs metabolites.

Metabolomics analysis of brain tissue was also carried out in several groups to investigate the alterations in rat brain metabolism. The MCAO group and the other two groups were clearly separated, but the HFD group and Sham were hardly separated at all (Figure 9A). Significant variations were seen in the metabolomic profiles of the sham-operated vs. untreated and treated vs. untreated animals (Figures 9B, C). Hua-Feng-Dan largely undid changes in brain metabolism brought on by ischemia. Permutation testing (Figures 9D, E) supports reliability of the finding by indicating that they are unlikely to be the result of overfitting. In fact, metabolite levels in the three groups varied significantly, as shown by volcano plots (Figures 9F, G). Metabolic data obtained in positive ion mode are presented in Supplementary Figure S5. There were 89 (44 unique) differential metabolites in HFD vs. Sham, 192 differential metabolites (8 unique) in MCAO vs. Sham, and 188 differential metabolites (4 unique) in HFD vs. MCAO (Figure 9H).

We identified a total of 38 potential biomarkers within brain of rats subjected to MCAO and treated with Hua-Feng-Dan (Table 3). These potential biomarkers encompass a variety of amino acids (e.g., citrulline, cys-arg-glu-lys-ala, glu glu, N-acetylaspartylglutamic acid, N-acetyl- l-alanine, l-tryptophan, l-phenylalanine, l-histidine, l-valine), SCFAs (2-hydroxy-3-methylbutyric acid, 2-aminohexanedioic acid), nucleosides (guanosine, isoguanosine, didanosine, adenine, 1-methyladenosine, dihydrothymine), glu-gln, and indole-3-carboxaldehyde. To explore the expression of brain metabolites within different groups, we performed metabolite clustering analysis. The expression of the two clusters was nearly opposite; the MCAO group gathered for the first cluster, whereas the HFD and Sham groups gathered for the second cluster (Figure 9I). In particular, MCAO was positively connected with l-phenylalanine, l-isoleucine, l-proline, l-valine, and l-tryptophan, whereas HFD was positively connected with l-aspartic acid, gamma-glutamylglutamic acid, N-acetyl-alanine, and inosine. These analyses demonstrated that the metabolism of brain underwent notable changes following IS, which were subsequently ameliorated through Hua-Feng-Dan intervention.