Multi-omics analysis reveals Angelica dahurica radix extract alleviates migraine in rats via gut microbiota-metabolome-gut-brain axis regulation

Baizhi was sourced from Baoding, Hebei Province, authenticated as the dried root of Angelica dahurica (Fisch.ex Hoffm.) Benth. et Hook. f. by Professor Guihua Jiang (Chengdu University of Traditional Chinese Medicine). The species Angelica dahurica, the botanical drug investigated herein, is covered in the monograph of national pharmacopoeias (Chinese Pharmacopoeia Commission, 2020). After collection, the Baizhi was washed to remove soil and impurities, then dried in an oven at a set temperature of 50 °C until thoroughly dried. Voucher specimens deposited in the university's herbarium (voucher No. BZ-2024-01, consistent with Sun et al., 2024).

200 g Baizhi was measured and combined with 10 times its volume of double-distilled water. The mixture was soaked for 1 h, followed by decoction for 30 min. After filtration, 8 times the amount of water was added to the filtrate and boiled again for 30 min. The two resulting fractions were combined and concentrated to a 0.5 g crude drug/mL concentration. The preparation, stored in a 4 °C refrigerator, yielded the BZ extract. Through this protocol, the dry extract rate of the BZ extract was calculated to be 10.77%.

The study protocol received ethical approval (2019-08) from the Ethics Committee of Chengdu University of Traditional Chinese Medicine, with all procedures conducted in strict accordance with established animal welfare standards. Specific-pathogen-free (SPF)-grade adult male Wistar rats (190 ± 20 g) were obtained from Chengdu Dasuo Experimental Animal Co., Ltd (License No.: SCXK (Sichuan) 2020-0030), and were maintained at 22 °C ± 2 °C with a 12 h light/dark cycle with ad-libitum access to food and water in the animal facility. After 1 week acclimation period, the rats were randomly assigned to four groups (randomized using a computer-generated sequence): Control group (Normal), Model group (Model), BZ group (BZ) and positive drug group (Sumaputan succinate injection, SS). The Normal and Model groups received intragastric administration of sterile water (10 mL/kg) for 14 days, while the BZ group was administered BZ (7.2 g crude drug/kg, corresponds to 0.78 g extract/kg) orally for 14 consecutive days. Previous investigations into dose - response effects (Chen et al., 2023; Sun et al., 2024; Lei et al., 2021). offered crucial guidance for determining the dosage employed in this study. The Normal group received no treatment, while the Model group, SS group, and BZ group were subcutaneously injected with NTG to induce migraine - like symptoms (10 mg/kg, Beijing Yimin Pharmaceutical Co., Ltd., Beijing, China) every other day for 9 days (5 times in total) (Hao et al., 2022).

The concentrations of factors 5-HT, NO, CGRP1, ET-1, PGE2, TNF-α, and DA in both plasma and brain tissue of the rats were quantified using commercial enzyme-linked immunosorbent assay (ELISA) kits (Elabscience, Wuhan, China; Cat. No.: 5-HT, E-EL-0033; CGRP1, E-EL-R0135; TNF-α, E-EL-R2856; ET-1, E-EL-R1458; PGE2, E-EL-0034; DA, E-EL-0046; NO, E-BC-K035-M). All assays were conducted in strict adherence to the manufacturer's protocols. Some results of ELISA were expressed as the ratio of the target substance content (ng) to the total protein amount (mg) (ng/mg).

To evaluate intestinal barrier integrity, colonic tissues were fixed in 4% paraformaldehyde for 24 h, embedded in paraffin, and sectioned into 5 μm slices. Hematoxylin and eosin (H&E) staining was performed following standard protocols (Lanza et al., 2021). Histopathological changes (e.g., inflammatory cell infiltration, mucosal epithelial necrosis, crypt degeneration, and lamina propria edema) were examined under a light microscope (Olympus CX23, Japan) at ×200 magnification. Three randomly selected fields per sample were analyzed using ImageJ software (v1.53, NIH, USA) to quantify pathological features.

Tight junction proteins Occludin and ZO-1 were assessed to determine intestinal barrier function. Deparaffinized sections underwent antigen retrieval in citrate buffer (pH 6.0) at 95 °C for 20 min. After blocking with 5% bovine serum albumin (BSA), sections were incubated overnight at 4 °C with primary antibodies: anti-Occludin (1:200, Abcam, Cat# ab216327) and anti-ZO-1 (1:200, Abcam, Cat# ab276131). Subsequently, horseradish peroxidase (HRP)-conjugated secondary antibodies (1:5000, Servicebio, China) were applied for 1 h at room temperature. Diaminobenzidine (DAB) was used for chromogenic detection, and nuclei were counterstained with hematoxylin. Protein expression levels were quantified as integrated optical density (AOD) using ImageJ software.

Microbial DNA extraction from feces samples was performed using the E.Z. N.A.^® soil DNA kit (Omega Bio-tek, Norcross, GA, U.S.) following the manufacturer's protocol. The V3–V4 region of the bacterial 16S rRNA gene was amplified using a thermal cycler PCR system (GeneAmp 9,700, ABI, USA). The paired-end sequence data from MiSeq sequencing were merged based on their overlap. Following quality control and filtering, operational taxonomic units (OTUs) with 97% similarity were clustered using USEARCH (version 7.0 http://drive5.com/USEARCH/↗), and chimeric sequences were subsequently removed.

The UPARSE software was employed for operational taxonomic unit (OTU) clustering of the sequences based on a 97% similarity threshold (Chen et al., 2025). The Rhonin Biosciences cloud platform (http://www.biomediv.cn/login.html↗) was utilized for analysis. Principal coordinates analysis (PCoA) and Venn diagrams were used to evaluate the differences in bacterial community structure across the samples. Linear discriminant analysis coupled with effect size (LEfSe) was conducted to identify the dominant microbiota from phylum to genus, according to a standard LDA score of >3, p < 0.05 in different groups (Tian et al., 2023).

Serum samples from each group were collected in metabolic cages and stored in a −80 °C refrigerator until analysis. Four times the volume of cold methanol was added to the precisely weighed samples and homogenized thoroughly. The mixture was then subjected to ultrasonic treatment in an ice bath for 20 min. Subsequently, the supernatant was centrifuged at 13,000 rpm for 10 min at 4 °C. Metabolomic analysis was performed using a ThermoFisher UHPLC(ThermoFisher Scientific, MA, USA) system coupled with an Orbitrap mass spectrometer. For both Electrospray ion source (ESI) positive and negative modes, the mobile phases consisted of A (0.1% formic acid in water) and B (0.1% formic acid in acetonitrile), with a flow rate of 0.3 mL/min.

ESI was utilized for positive ion mode detection. The operating parameters were as follows: spray voltage at 3.2 kV, ion source temperature at 350 °C, sheath gas flow rate at 35 arb, auxiliary gas flow rate at 10 arb, and ion transport tube temperature at 320 °C. The instrument employed a full scan/data-dependent secondary scan (Full MS/dd-MS2) mode with a first-level resolution of 70,000 and a second-level resolution of 17,500. The scan range was set from m/z 100 to 1,500, with a collision energy gradient of 20, 40, and 60 eV.

Non-targeted serum metabolomics raw data were imported into the Compound Discoverer 3.3 software. Through its wizard setup and method template, a process for identifying unknown metabolites was established. Peak alignment and extraction of the raw data were performed, fitting the extracted molecular ion chromatography peaks and isotopic peaks to possible molecular formulas. The secondary fragmentation spectra were matched with mzCloud, mzvault network database, and the local TCM composition database OTCML. Filter parameters for the matching results were set as follows: peak area threshold of 80,000, primary and secondary mass deviations of 5 ppm, and a match degree score above 80. Statistical analysis was then conducted. Metaboanalyst 6.0 software (http://www.metaboanalyst.ca/↗) and Metware Cloud (https://cloud.metware.cn↗) were used to perform principal components analysis (PCA), partial least squares discriminant analysis (PLS-DA), heatmap analysis, and metabolic enrichment and pathway analysis. The selection of significantly different metabolites was determined based on the variable importance in the projection (VIP) obtained using the PLS-DA model and the p-value of the Student's t-test. Differential metabolites were screened via the Metware Cloud using criteria: VIP ≥1.0, p < 0.05, and False Discovery Rate(FDR) ≤ 0.2. These metabolites were imported into MetaboAnalyst 6.0 for pathway analysis. The analysis compared three groups (Normal vs. Model vs. BZ), prioritizing Model vs. BZ to dissect BZ - mediated metabolic rescue, while Normal vs. Model data defined baseline migraine - associated dysmetabolism.

The Spearman correlation coefficient was employed to illustrate the relationship between the parameters (linear correlation). This coefficient consistently ranges between −1 and +1, with values closer to either extreme indicating a stronger linear relationship. In this study, Spearman correlation coefficients r > 0.6 and r < −0.6 were designated to signify significant positive and negative correlations, respectively.

All data were presented as mean ± standard error of the mean (SEM). Statistical analyses were performed using GraphPad Prism 10.4 software. Parametric and non-parametric data were compared using the independent t-test and Mann-Whitney U test, respectively (2 groups). For comparisons involving three or more groups, one-way analysis of variance (ANOVA) was first conducted to evaluate overall differences among groups. Subsequent pairwise comparisons were performed using Tukey's multiple comparison test to identify specific group differences. A value of p < 0.05 was considered statistically significant.

As shown in Figure 1A; Table 1, 40 chemical constituents were detected by UPLC-Q-Exactive Orbitrap MS, whose structural formulas were displayed in Figures 1C–H. The contents of three target metabolites in the Baizhi extract were determined as follows: Oxypeucedanin hydrate was 0.2193 mg/g, Byakangelicol was 0.2101 mg/g, and Bergapten was 0.0152 mg/g (n = 3). For detailed information, please refer to Supplementary Material 1.

Throughout the experimental period (Figure 2A), the Model group exhibited a slight decrease in body weight, while the groups of BZ and SS treatment groups demonstrated an increasing trend similar to the Normal group. The frequency of head scratching increased after NTG injection, the model group showed a significantly increased total average head-scratching frequency (113 times within 120 min) compared with the normal group, while BZ group reduced the total average head-scratching frequency by 61.2% vs. Model group (p < 0.01) (Figure 2B). ELISA assay results revealed significantly elevated levels of 5-HT, NO, CGRP1, TNF-α, PGE2, and DA in the Model group compared to the Normal group. Notably, these levels were significantly reduced in the BZ group relative to the Model group. In brain tissue, BZ group reduced the levels of CGRP1 by 9.6% (p < 0.05), DA by 7.5% (p > 0.05), 5-HT by 11.5% (p < 0.05), TNF-α by 9.7% (p < 0.05), PGE₂ by 7.2% (p > 0.05), and NO by 20.5% (p < 0.01) compared with the Model group. In plasma, BZ group reduced the levels of CGRP1 by 20.5% (p < 0.01), DA by 20.3% (p < 0.05), 5-HT by 24.2% (p < 0.01), TNF-α by 31.2% (p < 0.01), NO by 20.2% (p < 0.01), and ET-1 by 17.6% (p < 0.05) compared with the Model group. Collectively, these findings suggest that BZ administration alleviated NTG-induced migraine-like symptoms.

Intestinal barrier dysfunction plays a significant role in the pathogenesis of migraine (Lanza et al., 2021). We hypothesize that the efficacy of BZ in NTG-induced migraine may be closely associated with enhanced intestinal barrier function. Because Occludin/ZO-1 are pivotal for intestinal tight - junctions, we measured their protein levels to assess intestinal barrier integrity in migraine mice and the impact of BZ treatment. Histological changes in the colonic tissue were evaluated using H&E-stained tissue sections (Figure 3A). As shown by the arrow - marked areas in Figure 3A, the Model group presented severe damage: black arrows marked disrupted mucosal layers, blue arrows showed glandular distortion, yellow arrows indicated massive inflammatory cell infiltration, and green arrows denoted epithelial cell shedding. In the BZ group, distinct improvements appeared. As indicated by arrow - marked areas, yellow arrows showed a significant reduction in inflammatory cells compared to the Model group, black arrows revealed restored mucosal integrity, blue arrows showed more regular glands, and green arrows indicated reduced epithelial shedding. Overall, BZ treatment significantly alleviated colonic tissue damage.

Subsequently, immunohistochemistry was employed to explore the expression of migraine in relation to Occludin and ZO-1. As depicted in Figures 3B–D, in nitroglycerin-induced migraine model rats, the protein expression levels of Occludin and ZO-1 were significantly elevated after BZ treatment. These results indicate that BZ restores the impaired intestinal barrier function in NTG - induced migraine rats.

In order to further confirm whether the anti-migraine effect of BZ is dependent on the presence of the gut microbiota, we treated migraine rats with BZ. On the 15th day, we measured the composition of gut microbiota. Venn analysis revealed substantial microbial community divergence, with only 203 core OTUs shared among Normal, Model, and BZ groups (Figure 4A). The BZ group exhibited 492 unique OTUs versus 804 in Normal and 620 in Model groups, indicating treatment-specific microbiota reorganization. PCoA analysis (PC1: 22.03%, PC2: 16.26%) confirmed BZ's capacity to shift microbial β-diversity toward normative clustering (Figure 4B). α-Diversity metrics demonstrated migraine-associated dysregulation: Model group showed elevated Shannon index and InvSimpson index, suggesting pathological hyperdiversity. BZ treatment normalized these indices, restoring microbial community equilibrium (Figure 4C). Microbiota analysis at phylum and genus levels (Figures 4D,E) identified distinct dysbiosis in migraine rats. At the phylum level, migraine rats had reduced Firmicutes and increased Bacteroidetes, lowering the Firmicutes/Bacteroidetes (F/B) ratio vs. Normal; BZ reversed this by elevating Firmicutes relative abundance and suppressing Bacteroidetes (Figure 4D). The F/B ratio is a key gut microbiota balance indicator. Previous studies (An et al., 2023; Spase et al., 2020; Koliada et al., 2017) note that an abnormal F/B increase links to intestinal dysbiosis and disease pathogenesis—e.g., it may drive ulcerative colitis progression or contribute to obesity (Spase et al., 2020). In our study, the Model group had a higher F/B ratio than the Normal group (indicating migraine-induced gut dysbiosis), while BZ reduced the F/B ratio, suggesting it restores microbiota balance and may aid its anti-migraine effect. At the genus level (Figure 4E), several genera showed significant differences among the groups. In the Model group, the relative abundances of genus Lactobacillus and genus Unassigned_Bacteroidales were significantly increased compared to the Normal group, while in the BZ group, the relative abundances of these genera were decreased, approaching the levels of the Normal group. These changes suggest that BZ treatment may regulate the composition of gut microbiota at the genus level.

LEfSe analysis (LDA >3.0, p < 0.05) identified 32 discriminatory taxa across groups (Figure 5A). The Normal group was dominated by Lactobacillus, reflecting gut homeostasis, while the Model group exhibited enrichment of pro-inflammatory taxa including [Ruminococcus] torques group and Bacteroides. Conversely, the BZ group demonstrated a unique microbiota profile characterized by Erysipelotrichaceae and Enterobacteriales, alongside significant suppression of migraine-associated [Ruminococcus] gnavus group. Notably, bar plots in Figure 5C display the relative abundance of key taxa confirmed BZ-mediated restoration of beneficial Lactobacillus abundance, correlating with 5-HT recovery, while pathogenic Porphyromonas decreased, aligning with Z O -1 upregulation.

Non-targeted metabolomics was employed to investigate the changes in non-volatile metabolites in serum. Figure 6A reveals significant alterations in numerous metabolites compared to the Normal group, indicating that the establishment of the migraine model resulted in substantial changes in metabolite profiles in rats. Differences and similarities among the Normal, Model, and BZ groups were evaluated using PLS-DA (Figure 6B). The analysis revealed a significant separation between the Normal and Model groups. Furthermore, abundance analysis and random forest analysis were utilized to identify the top 15 most relevant substances (Figures 6C,D), which verified the similarities and differences among the Normal, Model, and BZ group (Xing et al., 2024).

The volcano map in Figure 7A confirmed the similarities and differences between the Normal group and the Model group, as well as between the Model group and BZ group. Further analysis revealed that, compared to the Normal group, the relative peak areas of multiple metabolites in the Model group were significantly altered. Moreover, metabolomics pathway analysis of the inter-group differences demonstrated distinct enrichment patterns. Pathways such as arachidonic acid metabolism exhibited a high -log10(p) value and relatively high pathway impact, suggesting a significant association with the observed metabolic changes. Other pathways, including one carbon pool by folate and phenylalanine metabolism, also displayed varying degrees of significance and impact (Figure 7B). Following BZ treatment, the relative peak areas of numerous metabolites, including 4-Amino-3-hydroxybenzoic acid and ProstaglandinA1, showed a trend of reverting towards the levels observed in the Normal group. Among these, 2-AG associated with CGRP1 inhibition, and eight-epi-PGF2α associated with decreased NF-κB phosphorylation. This suggests that these metabolites may play crucial roles in the response to BZ treatment and the underlying biological processes among the different groups. Metabolomic pathway analysis identified significant enrichment in arachidonic acid metabolism, which is linked to neuroinflammatory responses in migraine.

Spearman correlation analysis employed to generate a correlation heatmap, elucidating the relationships among serum metabolites and genera, as well as the correlation between microbiota metabolites, as well as between microbiota metabolites and brain tissue migraine-related indices (5-HT, CGPR1, TNF-α, IL-1β, and NO, among others), and indicators closely associated with intestinal permeability (ZO-1, Occludin). Figure 8A illustrates distinct correlations among various bacterial genera and metabolites. N-Acetylvaline exhibited a positive correlation with Lactobacillus, while 2,3-Dinor-8-epi-prostaglandin F2α demonstrated a negative correlation with [Ruminococcus] gnavus group. Additionally, Ascorbic acid showed a positive association with Tyzzerella 4, and Monobutyl phthalate displayed a negative relationship with Bacteroides. Concerning the gut microbiota-migraine connection depicted in Figure 8B, significant effects on the expression of migraine-related factors were observed. Glutaric anhydride positively correlated with NO and TNF-α, while 6-Methyl [1,2,4]triazolo [4,3-b]pyridazin-8-ol negatively correlated with ZO-1 AOD. Furthermore, 4-Amino-3-hydroxybenzoic acid exhibited a positive relationship with DA, and Gramine negatively correlated with 5-HT. In conclusion, these findings reveal the complex interplay between fecal metabolite alterations, gut microbiota shifts, and migraine-related factors, highlighting the intricacy of the underlying mechanisms.