Acupuncture may reduce brain inflammation in chronic migraine by changing gut bacteria and their chemical activity

Nitroglycerin injection (Approval No. H11020289) was purchased from Beijing Yimin Pharmaceutical Co., Ltd. (Beijing, China). The mixed Lactobacillus preparation consisted of Lactobacillus plantarum, Lactobacillus paracasei, Lactobacillus acidophilus, and Lactobacillus delbrueckii subsp. bulgaricus and was obtained from Beijing Beina Chuanglian Biotechnology Research Institute (Beijing, China). Enzyme‐linked immunosorbent assay (ELISA) kits were supplied by Shanghai Zhuocai Biotechnology Co., Ltd. (Shanghai, China).

Healthy male Sprague–Dawley (SD) rats (200 ± 20 g, SPF grade) were obtained from Chengdu Enswire Experimental Animal Co., Ltd. (Chengdu, China). All animals were housed under standard laboratory conditions (temperature: 24°C ± 2°C; humidity: 40%–70%; 12 h light/dark cycle) with ad libitum access to food and water. Rats were acclimatized to the housing environment for 7 days prior to experimentation. To avoid cross‐interference of gut microbiota, animals from different experimental groups were housed separately.

A total of 42 rats were randomly allocated using a random number table into seven groups (n = 6 per group): control (Con), CM model (Mod), acupuncture (Acu), model + probiotics (Mod + Pro), model + antibiotics (Mod + Anti), acupuncture + probiotics (Acu + Pro), and acupuncture + antibiotics (Acu + Anti). All experimental procedures were approved by the Experimental Animal Ethics Committee of Chengdu University of Traditional Chinese Medicine (Approval No. 2019‐23) and were conducted in accordance with the Guidelines for the Care and Use of Laboratory Animals of the Chinese Pharmacological Society.

The CM model was induced by subcutaneous injection of nitroglycerin (NTG, 10 mg/kg) once every other day for nine consecutive days. During the intervention phase, rats in the acupuncture group received daily acupuncture treatment at bilateral Shuaigu (GB8; located at the midpoint between the auricular apex and the parietal tubercle) and Yanglingquan (GB34; located inferior to the fibular head on the lateral aspect of the knee) acupoints. A single‐use sterile acupuncture needle (0.25 ∗ 25 mm, Suzhou Huatuo Medical Equipment Co., Ltd., Suzhou) was used for each session. Acupuncture was performed bilaterally at both acupoints and nonacupoint sites, with a needle insertion depth of approximately 5‐6 mm. Acupuncture treatment began on the first day of modeling and was performed once per day for 20 min per session, lasting for a total of 9 days. The blank and model groups were similarly restrained for 20 min daily, but no acupuncture treatment was applied.

For depletion of gut microbiota, rats in the Anti group were given ad libitum access to an antibiotic cocktail dissolved in sterile drinking water, consisting of ampicillin (1.0 g/L), neomycin (1.0 g/L), metronidazole (1.0 g/L), and vancomycin (0.5 g/L). The antibiotic solution was replaced every 1‐2 days, and administration commenced 2 weeks prior to model induction to maintain drug efficacy and stability. To ensure adequate antibiotic intake, in addition to measuring and recording the animals’ water consumption daily, the concentration of antibiotics in the drinking water is adjusted every 5 days based on the animals’ water consumption and body weight, ensuring consistent antibiotic intake. Rats in the Pro group received mixed Lactobacillus supplementation via oral gavage (2.7 × 10⁹ CFU/rat/day) during the modeling period for a total of 9 days. A schematic of the experimental design is shown in Figure 1A.

All pain‐related behavioral tests were conducted in a blinded manner and included the paw withdrawal mechanical threshold (PWMT), facial withdrawal mechanical threshold (FWMT), and tail‐flick latency (TFL). PWMT and FMWT were assessed using Von Frey filaments to stimulate the central plantar surface and periorbital region of the rats, respectively, to evaluate changes in mechanical sensitivity. TFL was determined by applying a thermal stimulus to the tail and recording the latency from stimulus onset to tail withdrawal, reflecting thermal nociceptive sensitivity. Behavioral assessments were performed on Days 1, 3, 5, 7, and 9 after model induction to dynamically monitor the progression of pain hypersensitivity and the therapeutic effects in each group.

For microbiota and metabolomics analyses, at least five fresh fecal pellets were collected directly from the rectum of each rat, placed into prechilled sterile centrifuge tubes, and immediately stored at −80°C. Under anesthesia with 3% sodium pentobarbital, rats underwent cardiac perfusion, after which brain tissues were rapidly dissected. The spinal trigeminal nucleus caudalis (Sp5C) region was quickly isolated, snap‐frozen in liquid nitrogen, and stored at −80°C until further analysis.

The concentrations of calcitonin gene‐related peptide (CGRP), interferon‐gamma (IFN‐γ), interleukin‐1 beta (IL‐1β), interleukin‐3 (IL‐3), and tumor necrosis factor‐alpha (TNF‐α) in brain tissue were measured using commercially available ELISA kits according to the manufacturer’s protocols.

Total DNA was extracted from fecal samples, and the target regions of the 16S rRNA gene were amplified using specific primers. Amplicons were pooled at equimolar concentrations and subjected to high‐throughput sequencing. Raw sequencing data underwent quality control (QC), including sequence filtering, chimera removal, and operational taxonomic unit clustering. Amplicon sequence variant analysis was then performed to identify unique variants, followed by taxonomic annotation using reference databases. Alpha (α) and beta (β) diversity metrics were calculated using the QIIME2 platform to evaluate intra‐ and intersample microbial diversity. Linear discriminant analysis effect size (LEfSe) was employed to identify taxa with significantly different abundances (LDA ≥ 4.0, p ≤ 0.05), while PICRUSt2 was used for functional prediction and annotation of Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathways, as well as for differential pathway analysis.

Fecal metabolomic profiling was performed using a Waters Acquity I‐Class PLUS ultra‐performance liquid chromatography (UHPLC) system coupled with a Xevo G2‐XS QToF high‐resolution mass spectrometer. Data acquisition was conducted in both positive and negative ionization modes, using MSe data‐independent acquisition (DIA). Raw data were processed with Progenesis QI software for peak extraction and metabolite identification. An orthogonal partial least squares‐discriminant analysis (OPLS‐DA) model and its permutation test were employed to evaluate the robustness and discriminatory power of the model. Differential metabolites were selected based on the following criteria: variable importance in projection (VIP) > 1.0, p < 0.05, and fold change (FC) ≥ 1.2 or ≤ 0.83. The identified metabolites were further subjected to pathway enrichment analysis using the KEGG database to evaluate their biological functions and potential mechanisms. In addition, log₂FC values were calculated for the differential metabolites, and the top 10 metabolites ranked by the absolute value of log₂FC were selected to generate radar plots illustrating their variation trends.

Microbe–metabolite pairs with an absolute correlation coefficient (|CC|) > 0.8 and P (CCP) < 0.05 were first identified to generate a correlation heatmap between differential metabolites and bacterial genera. Pairs meeting p < 0.05 were then further filtered, and the occurrence frequency of each metabolite and bacterial genus was calculated. The top 30 most frequently occurring metabolites and genera were selected for further analysis. To ensure that critical associations were not overlooked, all microbe–metabolite pairs containing at least one CC ranked within the top 30 in absolute value were retained. Finally, a chord diagram was constructed to visualize the correlations between metabolite modules and bacterial genera.

All statistical analyses were performed using IBM SPSS Statistics Version 25.0, and figures were generated with GraphPad Prism version 8.0. Data are presented as mean ± standard deviation (mean ± SD). Comparisons between two groups were conducted using a two‐tailed unpaired Student’s t‐test, while comparisons among three or more groups were performed using one‐way or two‐way analysis of variance (ANOVA), depending on the experimental design. Unless otherwise specified, statistical significance was defined as: ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001 and ^∗∗∗∗p < 0.0001.

To validate the establishment of the CM model and evaluate the therapeutic effects of acupuncture on NTG‐induced pain hypersensitivity and central neuroinflammation, pain‐related behavioral parameters—including PWMT, FWMT, and TFL—were assessed on Days 1, 3, 5, 7, and 9 after model induction.

Compared with the Con, the Mod exhibited a sustained and significant decrease in pain thresholds over the 9‐day modeling period. Marked differences in mechanical and thermal responses were observed from Day 3 onward (p < 0.0001), confirming that the CM model successfully induced a stable pain hypersensitivity state (Figure 1B–D; Tables S1–S3). The Acu group produced a significant analgesic effect in PWMT and FWMT from Day 3 (p < 0.0001), with thresholds continuing to improve as the intervention progressed. Similarly, TFL was significantly prolonged from day 3 (p < 0.05), with more pronounced differences by Day 9 (p < 0.0001), indicating that acupuncture exerted a stable and sustained effect in alleviating central sensitization.

To further explore the anti‐inflammatory mechanism of acupuncture, ELISAs were performed to quantify inflammation‐related cytokines in the Sp5C region. Compared with the Con group, the Mod group exhibited markedly elevated expression of CGRP, IFN‐γ, IL‐1β, IL‐3, and TNF‐α (p < 0.0001), indicating that NTG induction triggered a robust central inflammatory response. Acupuncture significantly reduced the expression levels of these inflammatory mediators, with CGRP, IFN‐γ, IL‐3, and TNF‐α showing significant decreases (p < 0.05) and IL‐1β exhibiting a more pronounced reduction (p < 0.01) compared with the Mod group (Figure 1E). Collectively, these findings demonstrate that acupuncture effectively attenuates NTG‐induced central neuroinflammation and has the potential to alleviate pain hypersensitivity associated with CM.

To investigate whether the therapeutic effects of acupuncture on CM are associated with modulation of the gut microbiota, 16S rRNA gene sequencing was performed on fecal samples from the Con, Mod, and Acu groups. Multiple α‐diversity indices, including Chao1, Shannon, Simpson, and ACE, were calculated to assess species richness and diversity among groups (Figures–). The results showed no significant differences in these indices across groups, suggesting that neither CM induction nor acupuncture intervention markedly altered the overall species count or evenness of the gut microbiota. S1A S1D

To further characterize differences in microbial community composition, principal coordinates analysis (PCoA) and nonmetric multidimensional scaling (NMDS) were employed for β‐diversity visualization. This combined approach allows simultaneous evaluation of absolute differences in microbial communities and their ordination relationships, providing a more comprehensive and robust assessment of intergroup microecological variation. As shown in (Figure 2A,B), the Mod group exhibited clear separation from the Con group at the species level, indicating that CM induced significant gut microbiota dysbiosis. Notably, the Acu group, following acupuncture intervention, displayed a microbial community structure that shifted toward that of the Con group, suggesting a trend toward restoration of normal microecological status. These findings imply that acupuncture may exert its antimigraine effects, at least in part, through the modulation of gut microbial community composition.

In the overall compositional analysis of the gut microbiota, taxonomic bar plots were generated to illustrate the dominant taxa at different classification levels in each group. At the class level, Bacilli, Bacteroidia, Clostridia, Verrucomicrobiae, and Gammaproteobacteria represented the five most abundant taxa (Figure 2C). At the species level, Lactobacillus murinus, unclassified_Muribaculaceae, Limosilactobacillus reuteri, Lactobacillus acidophilus, Akkermansia muciniphila, and Lactobacillus johnsonii were predominant (Figure 2D). These bacterial taxa are widely recognized for their critical roles in gut–brain axis signaling and the maintenance of metabolic homeostasis.

Based on the overall microbial distribution patterns, ANOVA was performed at the class level (Figure 2E, Table S4) to identify core taxa exhibiting significant dysregulation in the CM model that could be modulated by acupuncture. The results revealed that Clostridia was significantly decreased in the Mod group compared with the Con group (Con vs. Mod, p < 0.01), a change markedly reversed by acupuncture (Acu vs. Mod, p < 0.001). Bacilli were significantly enriched in the Acu group (Acu vs. Mod, p < 0.01), whereas Coriobacteriia showed significant alterations in the Mod group (Con vs. Mod, p < 0.01) and was partially restored following acupuncture treatment (Acu vs. Mod, p < 0.05). These differential taxa are potentially linked to key functional pathways such as SCFA synthesis, lipid metabolism, and the transport of aromatic compounds.

To pinpoint specific microbial biomarkers modulated by acupuncture at higher taxonomic resolution, LEfSe analysis was conducted (Figure 2F). In the Con group, dominant taxa were mainly clustered within the lineage s_Lactobacillus acidophilus ⟶ g_Lactobacillus. In contrast, the Mod group displayed a scattered distribution without well‐defined dominant lineages, reflecting disrupted gut microecological architecture and reduced functional stability. Notably, the Acu group exhibited relatively complete and systematic phylogenetic lineages, including s_Akkermansia muciniphila ⟶ g_Akkermansia ⟶ f_Akkermansiaceae ⟶ o_Verrucomicrobiales and s_Limosilactobacillus reuteri ⟶ g_Limosilactobacillus ⟶ f_Lactobacillaceae ⟶ o_Lactobacillales, suggesting that acupuncture preferentially enriches mucin‐degrading bacteria and lactic acid bacteria—key taxa with recognized roles in metabolism and barrier protection—during the restoration of gut homeostasis.

To explore potential functional shifts underlying the observed community changes, functional predictions were performed using PICRUSt2. The analysis revealed significant differences between the Con and Mod groups in core KEGG pathways such as pyrimidine metabolism, biosynthesis of amino acids, ribosome, carbon metabolism, and purine metabolism (Figure 2G), indicating that CM modeling not only reshaped microbial composition but also potentially restructured fundamental metabolic networks associated with nucleic acid and protein biosynthesis, as well as energy and carbon flow. Correspondingly, differences between the Mod and Acu groups were also concentrated in purine/pyrimidine metabolism, ribosome, amino acid biosynthesis, and carbon metabolism, with additional enrichment in microbial metabolism in diverse environments (Figure 2H), suggesting that acupuncture may facilitate functional recovery by reprogramming the material and energy metabolic potential of the gut microbiota.

Having confirmed that acupuncture modulates the composition and abundance of gut microbiota, we next investigated whether this intervention also alters the metabolic profile in CM rats. Using an LC–QTOF platform, we performed qualitative and quantitative metabolomic analyses on 18 fecal samples, detecting a total of 10,509 peaks, of which 3720 metabolites were successfully annotated. To ensure data reliability, overall quality was rigorously assessed by evaluating the reproducibility of QC samples, the stability of intra‐group samples, and the discrimination between groups (Figures–). S2A S2C

Principal component analysis (PCA) revealed clear separation among the Con, Mod, and Acu groups at the metabolomic level (Figure 3A,B), indicating that both disease modeling and acupuncture intervention induced measurable and distinct metabolic alterations. OPLS–DA further demonstrated robust group separation (Figures S2D–S2G). Between the Con and Mod groups, the model exhibited excellent goodness‐of‐fit (R²Y = 0.967) and strong predictive power (Q²Y = 0.565), confirming its ability to capture the metabolic distinctions with high accuracy and predictive validity. Similarly, the Mod vs. Acu comparison yielded a high R²Y value (0.961), with a slightly lower but still substantial predictive capacity, reflecting significant and predictable metabolic changes following acupuncture treatment.

Volcano plot screening, using the criteria of FC > 1.2 or < 0.83, p < 0.05, and VIP > 1, identified 338 differentially abundant metabolites (203 upregulated, 135 downregulated) between Con and Mod groups (Figure 3C; Table S5) and 297 metabolites (115 upregulated, 182 downregulated) between Mod and Acu groups (Figure 3D; Table S6). These results indicate that acupuncture may influence the metabolic profile in CM, particularly by affecting specific metabolic pathways.

Radar plots illustrated pronounced alterations among the top 10 differential metabolites across the groups. Specifically, comparison between the Con and Mod groups revealed significant changes in dihydrolipoamide, isobutyric acid, 28‐glucopyranosyl‐3‐methyloleanolic acid, C16 sphingosine, and saponin H (Figure 3E). In the comparison between the Mod and Acu groups, representative differential metabolites included dihydrolipoamide, pimelate, 1‐phenylcyclohexylamine, 5‐amino‐5‐(4‐hydroxybenzyl)‐6‐(D‐ribitylimino)‐5,6‐dihydrouracil, and nitrosoguanidine (Figure 3F). The observed shifts in their abundance imply that acupuncture may indirectly modulate central nervous system function by influencing these metabolic processes, thereby alleviating CM‐associated symptoms.

To further explore the biological significance of the differential metabolites, KEGG pathway enrichment analysis was performed. In the comparison between the Con and Mod groups, significantly enriched pathways included drug metabolism–other enzymes, ubiquinone and other terpenoid‐quinone biosynthesis, Th17 cell differentiation, and the intestinal immune network for IgA production (Figure 3G; Table S7). These results suggest that the CM model may induce disruptions in antioxidant capacity, as well as impair intestinal immune homeostasis and Th17 cell differentiation, potentially leading to altered neuroimmune regulation and neuroinflammation associated with chronic pain. In the comparison between the Mod and Acu groups, significantly enriched pathways included the degradation of valine, leucine, and isoleucine and the mTOR signaling pathway (Figure 3H; Table S8). Specifically, the modulation of branched‐chain amino acid (BCAA) degradation and aminoacyl‐tRNA biosynthesis suggests that acupuncture may restore neurometabolic homeostasis by regulating amino acid availability and protein synthesis. Enrichment of the mTOR signaling pathway and central carbon metabolism reflects recovery of cellular energy metabolism and neuronal function. Additionally. Taken together, these results indicate that acupuncture may exert its therapeutic effects in CM by selectively modulating specific metabolic pathways, thereby improving metabolic and neurofunctional disturbances at the molecular level.

LEfSe analysis revealed that the Acu group exhibited altered taxa predominantly clustered within specific lineages, suggesting that acupuncture induces directional and selective modulation of the gut microbiota in the CM model. Therefore, in the subsequent correlation analysis, we focused specifically on the relationships between metabolites and microbiota in the Acu and Mod groups. By integrating multidimensional data from differential metabolites and microbial communities, we aimed to comprehensively elucidate the complex interactions between gut microbiota and metabolites, thereby providing mechanistic insights into the potential therapeutic effects of acupuncture in CM.

In the correlation heatmap (Figure 4A), Limosilactobacillus and unclassified_Lactobacillaceae displayed markedly distinct distribution patterns from other taxa, showing trends opposite to those of most microbial groups. This suggests that these two taxa may play unique roles within the gut microbial community. Moreover, both Limosilactobacillus and unclassified_Lactobacillaceae exhibited strong correlations (p < 0.01) with key differential metabolites between the Mod and Acu groups, such as cis‐5‐dodecenoic acid, dihydrolipoamide, pimelate, and hexadecasphingosine, highlighting their potential importance in metabolic regulation.

Further analysis using frequency bar plots of metabolite–microbe associations (Figure 4B,C) revealed that unclassified_Lactobacillaceae and Limosilactobacillus occurred at significantly higher frequencies than other taxa, underscoring their prominent roles in modulating CM‐related metabolites. Among the top 30 most frequently observed metabolites, dihydrolipoamide, hexadecasphingosine, and cis‐5‐dodecenoic acid were all enriched, with dihydrolipoamide ranking fifth, indicating its central position in the metabolic network.

The metabolite–microbe chord diagram (Figure 4D) clearly illustrated a significant association between unclassified_Lactobacillaceae and cis‐5‐dodecenoic acid, as well as a positive correlation between Limosilactobacillus and cis‐5‐dodecenoic acid and a negative correlation with dihydrolipoamide. This finding was further validated by the metabolite–microbe network analysis (Figure 4E), in which Limosilactobacillus emerged as a central network node, showing negative correlation with dihydrolipoamide and positive correlation with cis‐5‐dodecenoic acid.

Based on the preceding correlation analysis, we identified significant interactions between Limosilactobacillus and unclassified_Lactobacillaceae with key differential metabolites, with abundance changes strongly associated with the metabolomic profile of CM. This finding highlights the potential value of Lactobacillus taxa in CM intervention, particularly through the regulation of gut microbiota and metabolic pathways to synergistically alleviate chronic pain and neuroinflammation. Accordingly, we implemented exogenous supplementation with a mixed Lactobacillus preparation on top of the established CM model and acupuncture intervention to further validate its effects. Notably, Limosilactobacillus and Lactobacillus murinus belong to distinct clades within the Lactobacillaceae family, and our taxonomic analysis also revealed that CM‐associated dysbiosis involves significant changes across multiple species within the Lactobacillus genus, rather than being confined to a single species. Based on these characteristics, we selected a multistrain probiotic formulation composed of Lactobacillus plantarum, Lactobacillus paracasei, Lactobacillus acidophilus, and Lactobacillus delbrueckii subsp. bulgaricus. These strains are among the most widely used Lactobacillus representatives worldwide, with well‐established clinical safety and functional profiles. Previous studies have demonstrated that different Lactobacillus species share substantial functional similarities and exhibit complementary effects in maintaining gut barrier integrity, promoting SCFA production, modulating immune responses, and influencing gut–brain axis signaling [23–26]. Therefore, supplementation with multiple commonly used Lactobacillus species could not only mimic and amplify acupuncture’s modulatory effects on key Lactobacillus taxa at the functional level but also offer high reproducibility and strong potential for clinical translation.

We found that Lactobacillus supplementation significantly enhanced the analgesic effects of acupuncture. Compared with the Acu group, the Acu + Pro group exhibited greater improvements in PWMT and FWMT (p < 0.0001; Figure 5A,B, Tables S9 and S10), as well as significantly prolonged TFL (p < 0.001; Figure 5C, Table S11), suggesting that Lactobacillus may further potentiate the therapeutic effects of acupuncture in CM via modulation of the gut microbiota. Compared with the Mod group, Lactobacillus supplementation in the Mod + Pro group also significantly improved pain sensitivity (p < 0.0001), indicating that probiotic supplementation alone can effectively ameliorate pain‐related behaviors in the CM model. In the antibiotic‐treated groups, Mod + Anti rats, in which gut microbiota were depleted, showed only slight improvements in PWMT and FWMT (p < 0.05), with no significant improvement in TFL. Similarly, in the Acu + Anti group, antibiotic‐mediated depletion of the gut microbiota failed to produce significant pain relief and even resulted in intermittently lower pain thresholds than the Acu group during days 3–7 (p < 0.05), further underscoring the critical importance of gut microbial homeostasis.

ELISA results (Figure 5D) showed no significant differences in inflammatory cytokine expression between the Acu + Pro and Acu + Anti groups compared with the Acu group. This finding suggests that, under conditions where acupuncture has already markedly suppressed inflammatory responses, additional probiotic or antibiotic interventions do not further enhance its anti‐inflammatory effects. Interestingly, the Mod + Pro group exhibited significant differences from the Mod group in the expression levels of CGRP, IFN‐γ, IL‐1β, and IL‐3 (p < 0.05), indicating that probiotic supplementation alone can partially reverse model‐induced neuroinflammation and central sensitization. Although this effect may not fully match the magnitude of acupuncture or acupuncture combined with probiotics, the results further establish gut microbial homeostasis as a potential therapeutic target. In the Mod + Anti group, despite gut microbiota depletion by antibiotic treatment, IL‐1β, IL‐3, and TNF‐α levels remained significantly lower than those in the Mod group (p < 0.05), suggesting that gut microbial depletion per se can alter central inflammatory cytokine expression patterns. The underlying mechanism may involve the removal of pathogenic bacteria, the remodeling of gut microbial metabolite profiles, or the downregulation of immune response thresholds.

Data are available on request due to privacy/ethical restrictions.