Thirty-six male Sprague-Dawley (SD) rats (aged 6-8 weeks) were purchased from the Guangdong Medical Laboratory Animal Center (Guangzhou, China). The rats were housed under a 12 h light-dark cycle (lights on 07:00) with controlled temperature (20-26°C), relative humidity (40-70%), and had free access to food and water. The rats were randomly assigned to experimental groups. Rats were allowed to acclimate for five days before the experiments were started. All animal experiments were approved by the Animal Care Committee of the South China Agricultural University (Guangzhou, China. Approval number NO.2020d030) and were performed in accordance with the ARRIVE Guidelines.

The rat PSD model was established using transient middle cerebral artery occlusion (tMCAO). Thirty SD rats were randomly selected for left transient middle cerebral artery occlusion. After administering intraperitoneal anesthesia with 0.4% pentobarbital (40 mg/kg), the rats were fixed on a flat plate, the hair on the neck region was removed and disinfected with iodine. An incision (10 mm long) was made on the left side of the neck to expose the common carotid artery (CCA) and external carotid artery (ECA). The proximal ends of the left CCA and the ECA were ligated. A monofilament nylon line (50 mm in length; 0.26 mm in diameter) coated with 5-mm silica gel in the distal segment was used. Monofilament lines were inserted into the internal carotid artery (ICA) through a left CCA incision. The line was then inserted at a depth of 18-20 mm (beginning at the carotid bifurcation) to block the left middle cerebral artery. Finally, the skin was sutured and the rats were placed on a thermal insulation blanket. After 90 min of ischemia, the monofilament lines were carefully extracted from the ICA to induce reperfusion. Seventeen rats survived beyond 7 days postoperatively without mortality. Based on Videofluoroscopy swallowing study (VFSS) assessments, 12 of the 17 surviving rats that displayed dysphagia were randomly divided into two groups: the PSD group (n = 6) and the PSD+HF-rTMS group (n = 6). During the VFSS enrollment assessment of the 12 model rats, stroke severity was evaluated using the Longa scale. The scores obtained ranged from 1 to 2 points, indicating uniformity in neurological function severity among the rats (Zhong et al., 2022). Additionally, the sham operation group (n = 6) of SD rats was subjected to the same protocol without monofilament insertion. The detailed experimental schedule is shown in Figure 1A .

HF-rTMS was administered using a customized stimulator (YRDCCI, Wuhan, China). On the 7^th day post-operation, the magnetic stimulation group received rTMS treatment, which was continued until the 14^th day post-surgery. A round prototype coil (23mm diameter with 3.5T peak magnetic welds) was positioned perpendicular to the cortical surface projection area of the right primary motor cortex (M1). TMS-induced electric field modeled in the animal’s brain is 140 V/m. The treatment protocol was as follows: stimulation for 5 s and then, rest for 60 s (repeated 10 times); stimulation intensity, 33% of the maximum stimulator output; stimulation frequency, 10Hz. The animal special coil mechanism diagram and pictures of rTMS administered in animal models in Figures 2A-C .

The rats underwent VFSS on the day before surgery (baseline) and 2 weeks after surgery (day 14). The rats were fasted for 12 h before the VFSS assessments. A mixture of experimental rat breeding feed and iohexol (1:1) was placed on a platform in a transparent plastic box. During ad libitum feeding of the feed mixture, videos were obtained at 30 fps using swallowing radiography equipment (digital fluoroscopic X-ray machine PLD8100C, Zhuhai, China). Only swallows with clear sagittal views were analyzed. The mean fluoroscopic results for each rat were calculated using 8-10 swallows.

ImageJ software (National Institutes of Health, Bethesda, MD) was used to assess swallowing function, including 1) bolus area (mm²): bolus size measured after swallow initiation and before the head of the bolus passed through the second cervical vertebra (C2); 2) Pharyngeal Transit Time (PTT, s): initiated on bolus passing through the posterior oropharynx and ended on it completely passing through C2, while the position was near the upper esophageal sphincter; 3) pharyngeal bolus speed (mm/s): the maximum speed at which the head of the bolus traveled from the initiation point to C2; and 4) inter-swallow interval (s): time interval between two successive swallows during uninterrupted eating (Russell et al., 2013; Welby et al., 2021).

All rats were sacrificed on the 14^th days after surgery. The Intestinal tissues (3 samples of each group) were placed in 4% paraformaldehyde for 24 h, mounted with gelatin, and then, embedded in paraffin. The tissues were cut into 10 μm slices; after which the sections were deparaffinized and stained with histopathological and eosin (Servicebio, G1076).

We collected feces early in the morning after lights were turned on at the housing facility. The rats were temporarily placed individually in empty autoclaved cages and allowed to defecate. Approximately two fecal pellets were collected aseptically from each rat and placed in a sterile 2-mL tube. Fecal samples from four rats were collected from each group and stored at -80.0°C refrigerator. The time point for fecal sample collection coincided with that of the VFSS assessments.

The stool samples of each group were removed from the -80°C refrigerator. Total genomic DNA from stool samples was extracted using the CTAB/SDS method, and its concentration and purity were monitored using 1% agarose gels according to the concentration. The DNA was diluted to 1 ng/µL using sterile water. Primers were designed to amplify the variable V3-V4 regions of the 16S rRNA gene. PCRs was performed with 15 mL of Phusion^® High-Fidelity PCR Master Mix (New England Biolabs, USA), 0.2 mM forward and reverse primers, and approximately 10 ng of template DNA. The initial denaturation temperature was set at 98°C for 1 min, followed by denaturation at 98°C for 10 s (30 cycles), annealing at 50°C for 30 s, extension at 72°C for 1 min, and final elongation at 72°C for 8 min and storage at 10°C. PCR products were electrophoresed on a 2% agarose gel and purified using the Qiagen Gel Extraction Kit (Qiagen, Germany). The TruSeq DNA PCR-Free Sample Preparation Kit (Illumina, USA) was used to construct the library, which was assessed using an Agilent Bioanalyzer 2100 system. After the library was qualified, the Illumina NovaSeq was used for 16S rRNA V3-V4 sequencing. We spliced and filtered the original data to remove contaminated data and to obtain accurate and reliable data.

Uparse software was used to cluster clean reads from all samples, and sequence clustering was converted into OTUs) by default with 97% identity. Species annotation analysis was performed using the Silva Database based on the Mothur algorithm to evaluate the differences and community composition of the dominant species in different samples. Raw data were analyzed using BioTree Biotechnology Co., Ltd. (Shanghai, China). Alpha and beta diversities were calculated using the breakaway method implemented in Quantitative Insights into Microbial Ecology, version 2.0. Finally, linear discriminant analysis effect size and linear discriminant analysis were used to identify the dominant bacterial taxa in the different groups of rats.

A total of 25 mg of fecal sample was weighed and placed in an EP tube. Then, 200 μL of water and 480 μL of extract solution (MTBE: MeOH = 5:1) were added sequentially. After vortexing for 30 s, the samples were homogenized for 4 min at 35 Hz and sonicated for 5 min in an ice-water bath. Homogenization and sonication cycles were repeated three times. Next, the samples were incubated at -40°C for 1 h and centrifuged at 3000 rpm for 15 min at 4°C. The supernatant (300 μL) was transferred to a fresh tube and dried under vacuum at 37°C. Then, The dried samples were reconstituted in 200 μL of 50% methanol in dichloromethane by sonication for 10 min in ice water bath. The mixture was then centrifuged at 13000 rpm for 15 min at 4°C, and 150 μL of the supernatant was collected and stored in a refrigerator at -20°C until LC-MS/MS analysis was performed.

LC-MS/MS analysis was performed using a UHPLC system (Vanquish, Thermo Fisher Scientific) with a UPLC HSS T3 column (2.1 mm × 100 mm; 1.8 μm) coupled with a Q Exactive HFX mass spectrometer (Orbitrap MS, Thermo). The Mobile phase A consisted of 40% water, 60% acetonitrile, and 10 mM ammonium formate. The Mobile phase B consisted of 10% acetonitrile and 90% isopropanol, which was added with 50 mL of 10 mmol/L ammonium formate for every 1000 mL mixed solvent. Analysis was performed using an elution gradient. The column temperature was 55°C. The autosampler temperature was maintained at 4°C. A Triple TOF 5600 + mass spectrometer (AB SCIEX, USA) was used for the electrospray ionization (ESI) positive and negative ion mass spectrometry analyses. ESI conditions: ion source gas 1:60, ion source gas 2:60, temperature 600°C, voltage ± 5500 V in positive and negative modes, mass scanning range m/z:60-1000 Da, first-level scanning range 25-1000 Da, scanning time 0.2 s, and high-sensitivity mode was used. For the second mass spectrometry, the delustering voltages of the positive and negative modes were ± 60 V and the high collision energy was (35 ± 15) eV.

The raw data files were converted to files in the mzXML format using the ‘msconvert’ program from ProteoWizard. The CentWave algorithm in XCMS was used for peak detection, extraction, alignment, and integration; the minfrac for annotation was set at 0.5, and the cutoff for annotation was set at 0.3. Supervised orthogonal projections to latent structure discriminant analysis (OPLS-DA) was used to screen for the differential metabolites. Variable importance in projection (VIP) ≥ 1.0 and absolute fold change ≥ 3.0, were used as criteria for differential metabolite selection.

Data are presented as mean ± standard error of the mean (SEM). One-way ANOVA variance was performed to compare the continuous variables across multiple groups. All statistical analyses and correlation graphs were generated using GraphPad Prism 8 (GraphPad Software, San Diego, CA, United States).

Three different parameters (observed operational taxonomic units (OTUs) and Shannon and Simpson indices) were used to assess the alpha diversity. Beta diversity between samples was assessed using principal coordinate analysis (PCoA) and permutational multivariate analysis of variance (PERMANOVA). Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) was utilized to explore differences in the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways in bacterial taxa between groups. MetaboAnalyst 5.0 (Xia Lab, McGill University, Canada) was used to analyze metabolomics data, including a pathway analysis overview. The association between the main differential metabolites and gut butyrate-producing bacteria was assessed using Spearman’s rank correlation and correlation heat maps were drawn using R Studio (version 3.6.1, Boston, Massachusetts, USA). p-value < 0.05.

A total of 24 recorded swallowing videos (six rats at baseline and six rats each from the PSD, PSD+HF-rTMS, and Sham groups on the 14^th postoperative day) were manually analyzed. As shown in Table 1 , compared with the baseline group, the bolus area and pharyngeal bolus speed were significantly decreased in the PSD group, and there were significant increases in the bolus area and decreases in PTT following HF-rTMS intervention (p < 0.05). Representative radiographic images of the rats undergoing videofluorography are shown in Figure 1B . According to these results, middle cerebral artery occlusion can result in dysphagia, which can be mitigated by HF-rTMS intervention.

H&E staining revealed that the colon epithelial cells of rats in the sham group had intact structures and were tightly arranged. In contrast, the rats in the PSD group exhibited damaged surface epithelial cells and disrupted cryptal glands. However, in the PSD+HF-rTMS group, there was a reduction in damage, thickening of the intestinal mucous layer, and recovery of the appearance of the surface epithelium ( Figure 1C ). These findings suggest that HF-rTMS intervention may have a positive restorative effect on intestinal tissues in rats with PSD, thereby further reinforcing intestinal epithelial cell integrity.

The rarefaction, Shannon-Wiener, and rank-abundance curves in the three groups tended to be flat or plateaued, thus demonstrating satisfactory sequencing depth ( Supplementary Figure 1 ).

Alpha diversity analysis revealed no significant differences in the observed OTUs and Chao1 and Shannon indices among the baseline, PSD, and HF-rTMS groups ( Figures 3A-C ). The PCoA and PERMANOVA analysis of variance for beta diversity revealed a significant difference in the composition and abundance of the gut microbiota between the baseline and PSD groups. (Bray-Curtis p < 0.05; R² = 0.287) ( Figure 3D ). The PCoA and PERMANOVA analyses did not show significant differences in the composition and abundance of the gut microbiota between the PSD+HF-rTMS group and PSD group (Bray-Curtis p > 0.05, R² = 0.173) ( Figure 3E ).

Taxon-dependent analysis revealed the top 10 phyla in the Baseline, PSD, and PSD+HF-rTMS groups, with Firmicutes, Bacteroidetes, Verrucomicrobia, Tenericutes, and Proteobacteria being the most dominant phyla ( Figures 4A, E ). Although no significant differences in the gut microbiota composition at the phylum level were observed among the three groups, the abundance of the phylum Firmicutes was significantly lower in the PSD group than in the baseline group (p < 0.05) ( Figure 4C ), and the abundance of Verrucomicrobia showed an increasing trend without statistical significance (p > 0.05). Notably, the abundance of Firmicutes increased after HF-rTMS treatment in the PSD+HF-rTMS group (p < 0.05) ( Figure 4G ).

The most relative abundant genera (> 1.0%) in the three groups are shown in Figures 4B, F . At the genus level, the proportion of Lactobacillus, Anaerotruncus, Romboutsia, Turicibacter, Ruminococcaceae_NK4A214_group, Allobaculum, Dubosiella, and Faecalibaculum (p < 0.05) decreased, whereas those of Ruminococcaceae_UCG-014 and Odoribacter increased in the PSD group (p < 0.05) ( Figure 4D ). In the PSD+HF-rTMS group, the relative abundance of Lactobacillus and uncultured_bacterium_f_Prevotellaceae showed an increasing trend; however, the difference was not statistically significant. Although the relative abundance was < 1.0%, the proportion of uncultured_bacterium_Peptococcaceae and Marvinbryantia increased, whereas the abundance of [Eubacterium]_xylanophilum_group decreased (p < 0.05) ( Figure 4H ).

Linear discriminant analysis effect size (LEfSe) was used to identify potential microbiota biomarkers among different groups. Firmicutes exhibited a relatively higher abundance at the phylum level in the baseline group than in the PSD group. At the genus level, Quinella and Lactobacillus were the most abundant at baseline, whereas Ruminococcaceae_UCG_014, Enterococcus, and Phascolarctobacterium were enriched in the PSD group (linear discriminant analysis (LDA) > 3, p < 0.05) ( Figures 5A, C ). These microbes might be considered as biomarkers for the Baseline and PSD groups.

At the phylum level, Firmicutes was remarkably enriched in the PSD+HF-rTMS group compared to that in the PSD group. At the genus level, the abundance of uncultured_bacterium_Peptococcaceae was relatively high in the PSD+HF-rTMS group. The abundance of Streptococcus, Acutalibacter, and Ruminiclostridium_9 were relatively high in the PSD group (LDA > 3, p < 0.05) ( Figures 5B, D ).

Several KEGG pathways, such as cell growth, cancer-specific types, and immune diseases, were significantly different between the Baseline and PSD group (p < 0.05) ( Supplementary Figure 2A ). The pathways in the PSD group and PSD+HF-rTMS group differed in cardiovascular and infectious diseases (viral) (p < 0.05) ( Supplementary Figure 2B ).

LC-MS/MS was used to determine the metabolic characteristics of rats in the three groups. OPLS-DA models showed that the metabolic profile of the PSD group was significantly different from that of the baseline group in both the negative and positive ion modes ( Figures 6A, D ). Similarly, the PSD+HF-rTMS group was significantly different from the PSD group ( Figures 7A, D ).

Differential metabolites were identified using VIP > 1 and p < 0.05. A total of 699 differential metabolites [357 upregulated and 342 downregulated]) were screened and identified in the PSD group compared with those in the baseline group in the negative mode ( Figure 6B ). In Figure 6C , the heatmap displays a subset of 17 named metabolites (15 upregulated and two downregulated) out of the 699 differential metabolites analyzed. Additionally, in positive-mode analysis ( Figure 6E ), 403 metabolites (234 upregulated and 169 downregulated) were identified and compared. Furthermore, among the 403 differentially expressed metabolites, three upregulated metabolites were visualized with their names in the form of a heat map, as presented in Figure 6F . Briefly, the expression levels of differential metabolites in lipid classes, such as ceramides (Cer), free fatty acids (FA), phosphatidylethanolamine (PE), triacylglycerol (TAG), and sulfoquinovosyl diacylglycerol, were increased in PSD.

In the negative mode, a comprehensive analysis of the PSD+HF-rTMS group compared to the PSD group identified 1470 differential metabolites (963 upregulated and 507 downregulated) ( Figure 7B ). Among these 1470 metabolites, 26 specifically named metabolites (11 upregulated and 14 downregulated) were selected for visualization in the form of a heat map ( Figure 7C ). Additionally, in the positive-mode analysis, 919 metabolites (260 upregulated and 659 downregulated) were identified ( Figure 7E ). Figure 7F presents a heatmap that displays 20 named metabolites (11 upregulated and 15 downregulated) from the total pool of 919 differential metabolites analyzed. Briefly, the levels of the differential lipids Cer, FA, and DG decreased after HF-rTMS treatment, whereas TAG levels increased.

The MetaboAnalyst 5.0 was used for MetaboAnalyst pathway analysis of group differential metabolites, with p < 0.05. As shown in Figure 8A , metabolic pathways, such as glycerolipid, sphingolipid, glycerophospholipid, and urine metabolism, were altered in rats with PSD. HF-rTMS induced changes in sphingolipid, cysteine, methionine, glycerophospholipid, and pyrimidine metabolism in the PSD+HF-rTMS group ( Figure 8B ).

Spearman correlation analysis was performed to explore the association between gut microbiota biomarkers and the aforementioned group of differential metabolites (including negative and positive modes). At the genus level, a relationship between the four genera and 21 metabolites was observed between the PSD and Baseline groups ( Figure 6G ). Specifically, Enterococcus was positively correlated with Cer, FA, and PE in the PSD rats. Similarly, a relationship between the four genera and 45 metabolites was observed between the PSD+HF-rTMS and PSD groups. Peptococcaceae was negatively correlated with Cer, Streptococcus was negatively correlated with DG, and Acutalibacter was positively correlated with FA and Cer ( Figure 7G ).