Establishment of a novel brainstem ischemic dysphagia model: single-cell sequencing reveals the molecular mechanisms underlying mPES intervention Authors: Yueqin Tian, Qianqian Wang, Jiahui Hu, Jia Qiao, Chao Li, Xiangxiang Zhang, Tingting Jiang, Nenggui Xu, Hongmei Wen, Zulin Dou, Qiuping Ye Published: 2025 ABSTRACT AbstractObjectivePrevious animal models of post-stroke dysphagia (PSD) have limitations—these models are primarily induced by cortical strokes. Compared to cortical strokes, brainstem strokes are more likely to cause dysphagia, and the pathological mechanisms underlying dysphagia differ by focal infarction location. This study aimed to create a novel rat dysphagia model via brainstem ischemia (BSI) and explore modified pharyngeal electrical stimulation (mPES) therapy.MethodsRat brainstem ischemia was induced by photochemical embolization, confirmed by MRI and pathology. Swallowing function was assessed using Videofluoroscopic Swallowing Study (VFSS), while motor and neurobehavioral changes were evaluated through behavioral tests. Post-mPES treatment, VFSS was conducted to evaluate the swallowing function and single-cell transcriptomics was performed to explore therapeutic mechanisms.ResultsThe BSI model showed stable dysphagia characteristics, with prolonged pharyngeal transit time, increased inter-swallowing interval and smaller bolus size area. mPES significantly improved these parameters. Behavioral tests revealed BSI caused anxiety-like behavior and worse motor performance in rats. Single-cell transcriptomics indicated mPES treatment involved multiple biological mechanisms, possibly exerting therapeutic effects by influencing oligodendrocyte differentiation and myelin, or synapse regeneration and repair.ConclusionThe novel BSI-induced dysphagia model exhibited stable swallowing function deficits. mPES effectively improved these deficiencies, with therapeutic mechanisms potentially associated with oligodendrocytes.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s12984-025-01841-3.Keywords: Brainstem ischemia, Dysphagia model, Modified pharyngeal electrical stimulation, Single-cell transcriptomics, Oligodendrocytes Introduction Globally, stroke is the second leading cause of death and the third most common cause of death and disability combined [1]. Post-stroke dysphagia (PSD) may not only lead to serious complications such as malnutrition, respiratory infections, but also significantly affects patients' quality of life and rehabilitation outcomes, and increases the risk of hospitalization, death, and healthcare costs [2, 3]. However, pathogenesis of PSD remains incompletely understood and the efficacy of available treatments remains suboptimal due to various limitations. The construction of a suitable animal model of PSD is crucial for the in-depth study of its pathogenesis, the exploration of new therapeutic approaches, and the promotion of the development of this field. In recent years, with the increasing attention to the pathological mechanism and treatment of PSD, researchers have developed a variety of animal models, including transient middle cerebral artery occlusion (tMCAO), primary motor cortex (M1) photochemical embolization, bilateral common carotid artery ligation, and type VII collagenase-induced cerebral hemorrhage, etc. The tMCAO model can simulate typical brain injury and cause abnormal swallowing function, but the damage to the swallowing function often occurs 2–3 weeks after modeling, which differs somewhat from clinical patients [4]. Although the M1 photochemical approach is useful for investigating brain plasticity, the stability of the swallowing phenotype is relatively poor [5]. Bilateral common carotid artery ligation is simple to operate but lacks model specificity [6]. These models each have their advantages, but they fail to fully replicate the characteristics of dysphagia following clinical cerebral ischemia. Furthermore, these models exhibit instability or are subject to interference from other etiological factors. In addition, it has been previously recognized that swallowing control relies on dynamic interactions between the brainstem and cortex, although the functional emphasis differs between these two regions. The brainstem provides foundational rhythms and coordination, while the cortex finely modulates brainstem activity through sensory integration, motor initiation, and adaptive compensation, thereby ensuring the efficiency and safety of swallowing [7, 8]. Ischemic damage to either the cortex or brainstem is one of the key pathological bases for dysphagia [9, 10]. Epidemiological studies indicated that over 40% of stroke patients experience dysphagia [2, 11]. Although cortical strokes are more common, the probability of dysphagia in brainstem stroke patients could reach up to 81% [12], particularly in those with anterolateral or posterolateral medullary infarction (LMI), where the incidence of dysphagia could reach 92.5% [13]. However, previous studies revealed distinct patterns of swallowing-related cortical activation depending on lesion location: brainstem injury patients exhibited bilateral activation, whereas unilateral cerebral hemisphere stroke patients primarily showed activation in the unaffected hemisphere [14]. Therefore, there is an urgent need to establish a specific brainstem ischemia-induced (BSI) dysphagia model to directly investigate the impact of brainstem ischemia on swallowing function and to explore its pathophysiological and in-depth therapeutic mechanisms. Furthermore, a stable BSI model will significantly enhance the reproducibility and reliability of experimental results. Up to 80% of patients with brainstem injuries experience severe swallowing difficulties, making recovery particularly challenging [11]. Patients with dysphagia following brainstem infarction are often found to have reduced sensation in the pharynx, primarily manifesting as pharyngeal swallowing disorders [15–18]. Previous studies have shown that treatment methods targeting sensory input are more effective for swallowing disorders following BSI than those targeting motor muscles [19]. Pharyngeal electrical stimulation (PES) is a sensory stimulation technique that delivers electrical stimulation to the pharyngeal mucosa locally to increase sensory input and further induce swallowing movements. In recent years, PES has garnered significant attention in the field of neurogenic dysphagia treatment. Recent clinical study data confirm that PES demonstrates marked efficacy in treating severe PSD. It not only facilitates early removal of tracheostomy tubes and nasogastric tubes in patients with severe PSD but also significantly reduces the incidence of pneumonia [20]. Modified Pharyngeal Electrical Stimulation (mPES) is an improvement based on PES, which demonstrates comparable efficacy to conventional PES in the treatment of PSD [21]. Moreover, its neural mechanism is similar and is related to the brain network closely related to swallowing function [22]. Although the therapeutic effect of mPES on PSD has been confirmed clinically, its molecular mechanism of action is unknown and still lacks in-depth exploration in animal models. In this study, we intend to establish a model of dysphagia after BSI by photochemical embolization method. We will multidimensionally evaluate this model by combining multidimensional approaches, including MRI scan, pathology, behavior, and videofluoroscopy swallowing study, the latter being considered the gold standard of swallowing examination. In addition, we will validate the effect of mPES on the improvement of swallowing function in this model. Finally, we will explore the specific mechanisms by which mPES improves swallowing deficits after brainstem stroke through single-cell transcriptomics. Methods Experimental animals and groups 44 healthy SPF-grade 8-week-old SD male rats weighing 180–220 g were purchased from the Experimental Animal Center of Guangzhou University of Traditional Chinese Medicine. Housing conditions: temperature 20–24 °C, relative humidity 40–60%, 12 h/day lighting, free access to normal diet. All animal procedures complied with the regulations of the Guangzhou University of Traditional Chinese Medicine Ethics Committee, with ethics approval number 20220927005. This study was reported in accordance with the ARRIVE guidelines [23]. Experimental design As shown in the experimental flow chart (Fig. 1), rats underwent three days of behavioral training starting three days before modeling. On day 1 post-modeling, brain stem sections were taken for TTC staining. On day 3, magnetic resonance imaging (MRI), video fluoroscopic swallowing study (VFSS), and behavioral tests were done to validate the BSI dysphagia model. Then, mPES was applied for 3 days, followed by VFSS evaluation. On day 7, brain tissues were collected for pathology and single-cell transcriptomics. Experimental design flowchart Construction of a BSI model Similar to the previous description [24], the BSI model was made according to the following protocol. ① Anesthesia: 44 rats were anesthetized via intraperitoneal injection of pentobarbital sodium (1%, 30 mg/kg), and then intraperitoneally injected rose bengal solution (2%, 1 mL/100 g, Sigma, USA). After 15 min, the limbs were secured and the surgical area was disinfected. Anesthesia: 44 rats were anesthetized via intraperitoneal injection of pentobarbital sodium (1%, 30 mg/kg), and then intraperitoneally injected rose bengal solution (2%, 1 mL/100 g, Sigma, USA). After 15 min, the limbs were secured and the surgical area was disinfected. ② Dissection: Make a midline incision along the neck and perform blunt dissection of the fascia. Take the digastric muscle as a landmark and gently dissect the fascia alongside the trachea towards the cranial base. Mark the midline and the position where the internal carotid artery enters the cranium with the prominence as a transverse anatomical landmark. Utilize a cranial drill to create an approximately 0.8 mm diameter hole and employ a fine needle to carefully puncture the dura mater to avoid bleeding or damage to the underlying brain tissue; Dissection: Make a midline incision along the neck and perform blunt dissection of the fascia. Take the digastric muscle as a landmark and gently dissect the fascia alongside the trachea towards the cranial base. Mark the midline and the position where the internal carotid artery enters the cranium with the prominence as a transverse anatomical landmark. Utilize a cranial drill to create an approximately 0.8 mm diameter hole and employ a fine needle to carefully puncture the dura mater to avoid bleeding or damage to the underlying brain tissue; ③ Expose the drilled area, then gently place the laser probe (Laserware, China) vertically onto the surface, and irradiate with a laser of 0.66 A intensity (wavelength 560 nm) for 5 min; Expose the drilled area, then gently place the laser probe (Laserware, China) vertically onto the surface, and irradiate with a laser of 0.66 A intensity (wavelength 560 nm) for 5 min; ④ Seal the drill hole, suture the wound, disinfect, and return to the animal room for normal management after resuscitation. Seal the drill hole, suture the wound, disinfect, and return to the animal room for normal management after resuscitation. It should be noted that the superior laryngeal nerve and the pharyngeal branch of the glossopharyngeal nerve play a major role in triggering reflex swallowing from the pharynx [25], and damage for them should be avoided during surgery. 12 rats in the sham group were given intraperitoneal injections of equal amounts of rose bengal solution with equal concentrations and underwent the same surgical operation without laser irradiation. 10 rats died after surgery and were excluded from the experiment. 4 rats each in the model group and sham group were performed TTC staining, and the remaining rats were randomly divided into 2 groups: 14 in the BSI model group and 8 in the mPES group. Rats in the BSI model group were treated with rose bengal solution intraperitoneally and operated with laser irradiation, while mPES-treated rats were treated with mPES after being modeled. TTC staining Referring to previous studies [26], the brainstem was dissected and cut into 5 consecutive coronal Sects. 24 h after embolization. The slices were incubated with 2% TTC (Solarbio, China) at 37 °C for 15 min and were turned over once halfway through (ischemic areas stained to appear white, while non-ischemic areas remained red). Finally, figures were taken with a digital camera for documentation. Magnetic resonance imaging (MRI) scan The coronal and sagittal T2WI scans were performed using a 9.4-T small animal MRI scanner (Agilent Inc., USA) according to the previously described method [24]. Parameters were as follows: fast spin echo (FSE), TR 3000 ms, TE 38 ms, number of repetitions (NEX) 2, field of view (FOV) 30 × 30 mm, slice thickness 0.7 mm, slice spacing 0.2 mm. Subsequently, based on measurements by ImageJ software (National Institutes of Health, Baltimore, MD), the infarct volume was calculated using the formula “ischemic volume = ischemic area per slice × (slice thickness + slice spacing)” [27]. mPES treatment After anesthesia, platinum double-ring electrodes for small rodents were introduced into the oral cavity, and the electrodes were precisely positioned at the upper esophageal sphincter with the aid of a laryngoscope (Shanghai XinSoft, XR-XH201). Subsequently, the mPES therapeutic device manufactured by ZIMMER (Neu-Ulm, Germany) was used to stimulate the pharyngeal mucosa of the rats for 3 consecutive days according to the stimulation parameters (5 Hz, 10 min, 1 mA) set in the previous study of our team [21]. Behavioral test First, neurological deficits were evaluated in each group of rats 24 h after modeling according to the Longa scoring method [28]. Second, the elevated plus maze (EPM) test was designed to assess anxiety-related behaviors in rodents [29]. The open field experiment is a reliable behavioral test for estimating locomotion and anxiety-like behaviors in rats [30]. The rat's locomotor trajectory and spontaneous activity of these two tests were measured using automated video-tracking software (Shanghai Neosoft, China). Finally, the automated VisuGait gait analysis system (Shanghai NewSoft, China) quantified multiple gait parameters to accurately and objectively assess motor and coordination functions in rats post-BSI [31]. VFSS assessments VFSS is the gold standard for evaluating swallowing function [32]. As previously described [21], prior to the formal experiment, the rats were fasted for 12 h with water allowed, and their preferred food flavors were determined through palatability testing. Additionally, the rats were placed in the VFSS testing chamber for a 30-min pre-adaptation period, and the preferred flavor food was mixed with iopamidol at a 1:1 ratio. During the formal experiment, the prepared food was placed inside the testing chamber to allow the rats to freely eat. Swallowing was recorded by a digital fluoroscopic X-ray machine (Zhuhai Perlead, China), and swallowing function was assessed for each rat based on the average of 8–10 swallows. The bolus area was measured by ImageJ software (National Institutes of Health, Bethesda, Maryland, USA). Specific assessment indicators included [33, 34]: Bolus area (mm2); Pharyngeal transit time (PTT, seconds); Esophageal transit time (ETT, seconds); Inter-swallow interval (ISI, seconds). To minimize potential bias, blinding was applied to the analysts in this study. The VFSS examinations were performed by Yueqin Tian and Qianqian Wang, while subsequent image analysis and data processing were conducted by two independent analysts (Jiahui Hu and Qiuping Ye) who were blinded to the group assignments and true identity of the rats. In cases where significant heterogeneity was detected in the analyzed data, a re-analysis was performed by both analysts. HE staining As previously described [35], for HE staining, rats were sacrificed, and their brains were fixed and dehydrated for embedding. 4 μm thick sections were cut, dewaxed, and rehydrated. Then sections were stained with hematoxylin and eosin solutions, followed by dehydration, clearing, and mounting. For Nissl staining, sections were rehydrated, dehydrated, and cleared. Finally, all sections were observed and photographed using a CaseViewer scanner (3DHISTECH, HU). Single-cell sample preparation and sequencing On day 7 after modeling, 2 rats each from the BSI and mPES groups were randomly selected for single-cell RNA sequencing (scRNA-seq). After dissecting the ischemic foci of brainstem ischemia on the ischemic side, nuclei were isolated with Nuclei EZ lysis buffer (Sigma-Aldrich, USA) as well as protease inhibitors (Roche, Switzerland) and RNase inhibitors (Promega, USA, and Life Technologies, USA). Single-nuclei suspensions were loaded into 10 × Chromium to capture single cells according to the manufacturer’s instructions of Single-Cell 3’ kit (10 × Genomics Chromium, USA). Libraries were sequenced on an Illumina NovaSeq 6000 sequencing system (paired-end multiplexing run, 150 bp) at a minimum depth of 20,000 reads per cell. Quality control The 10X Cell Ranger [36] (version 3.1.0) was used to process the single-cell RNA sequencing raw data into FASTQ format. Subsequently, Cell Ranger output was loaded into Seurat [37] (version 3.1.1) for dimensionality reduction, clustering, and analysis of single-cell RNA sequencing data. Visualization, identification of marker genes and enrichment analysis of DEGs First, the t-SNE and UMAP algorithms in the Seurat R package [38] were First, the t-SNE and UMAP algorithms in the Seurat R package [38] were applied to project the high-dimensional sequencing data into a two-dimensional space in order to cluster and differentiate cells with similar expression patterns. Second, the clustered marker genes were identified by the FindAllMarkers function [39], and the screening criteria were |log2FC|> 0.26, p < 0.05, and expression in at least 10% of the target cell population. Furthermore, differentially expressed genes (DEGs) between the BSI group and the mPES group, as well as oligodendrocyte subpopulations were screened with |log2FC|> 0.26 and p < 0.05 as criteria. Finally, functional annotation and significant enrichment analysis were performed on the obtained genes, including GO and KEGG analysis [40]. Trajectory analysis and cell communication analysis Pseudotime analysis of single-cell transcriptomic data was performed via the Monocle2 R package [41] (version 2.14.0). The DDRTree method [42] was used for dimensionality reduction to construct cellular developmental trajectories, thereby simulating the dynamic changes in cells during this biological process. The BEAM algorithm was utilized to identify DEGs on specific branches of the pseudotime tree, thereby investigating potential key regulatory genes. The CellPhoneDB was used for cell communication analysis [43]. This tool comprehensively scanned the expression profiles of each subpopulation and identified the ligand-receptor molecular combinations involved in cellular communication, thus visualizing the potential communication links and their strengths among oligodendrocyte subpopulations. Statistical analysis Statistical analysis was performed using Prism 9.0 (GraphPad Software, USA). Measurement data were presented as mean ± standard deviation. The t-test was used for comparisons between two groups, and two-way ANOVA was applied to compare the levels of two or more factors. p < 0.05 was considered statistically significant. The number of rats used in each experiment is indicated in the legend and denoted by n. Results Establishment of a rat model of dysphagia induced by brainstem ischemia Establishment of BSI rat model In the present study, we established a rat model of BSI and validated the successful establishment of the model using several methods. First, TTC staining (Fig. 2A) indicated that significant ischemic injury had occurred in the brainstem of BSI rats. Meanwhile, MRI images (Fig. 2B, C) showed that abnormal signal areas consistent with the TTC staining results appeared in the brains of BSI rats, which contrasted with the surrounding normal brain tissues, reconfirming the formation of infarct lesions in the brainstem. Moreover, we observed that all BSI model rats had focal infarcts within the left brainstem (Fig. 2B). The location of the infarcts was highly reproducible, and the average infarct volume was approximately 5.636475 mm2 as quantified by MRI. The ischemic area mainly involved the facial nucleus, the lateral paragigantocellular nucleus and part of the raphe nuclei. The infarct geometry induced by this model appeared to match that of human brainstem infarcts caused by occlusion of basilar artery branches. Establishment of BSI model rats. A. TTC staining of coronal sections of the brainstem in rats with BSI. Red areas indicate normal tissue, and white areas indicate infarct regions (red arrows). B. Coronal and sagittal T2-weighted images show a hyperintense lesion in the brainstem of model rats (red arrows), suggesting acute ischemia. C. MRI-derived quantification of brainstem infarct volume (n = 4 biologically independent samples in each of two groups. Unpaired two-tailed t-test for two group comparisons). D–E. HE staining showed cellular structural abnormalities and tissue damage in the infarct region of BSI model. F–G. Nissl staining showed reduced, lighter Nissl bodies in neurons within the infarct area of BSI model, with abnormal cell shapes. Data are presented as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 In addition, we observed the pathological changes in brain tissue by HE staining and Niebuhr staining. The HE staining results (Fig. 2D, E) revealed that in the focal area of BSI, a large number of neuronal cells had solidified nuclei, deep staining, enhanced cytoplasmic eosinophilia, widened cellular gaps, and disruption of the organization, in contrast to the cellular morphology and arrangement of normal brain tissue. The results of Nissl staining (Fig. 2F, G) showed that the Nissl bodies in the cytoplasm of neurons in the infarct foci were significantly reduced and lightened in color. Some neurons lacked visible Nissl bodies, and the nuclear chromatin was condensed and marginalized, with abnormal cellular morphology. The above pathologic changes further confirmed the successful establishment of the rat model of BSI at the cellular level. Neuropsychological and function motor assessment in rats To explore the changes in neurological function and motor coordination in rats modeled with BSI, we assessed the neurological deficit score (NDS) in rats. As shown in Fig. 3A, we observed that BSI rats scored significantly higher in the NDS assessment than that in the sham group. This indicates that BSI resulted in a significant decrease in neurological function and motor coordination. In order to investigate whether the model rats had anxiety behaviors, we performed the EPM experiment and open field experiment. The results demonstrated that, compared with the sham group, rats in the BSI group spent a significantly lower percentage of time in the open arm (Fig. 3B–D), had significantly lower average spontaneous activity speed and total movement distance (Fig. 3E-G), and spent a significantly lower percentage of time in the central area (Fig. 3H), indicating that brainstem ischemia caused rats to exhibit anxiety-like behavior and poorer motor performance. Effects of the rats' BSI model on motor function. A Neurological severity scores. B–D EPM experiments were performed to assess anxiety-like behavior in rats. Representative motor trajectory diagrams of sham (B_left) and BSI (B_right) rats on the cross-elevated frame. C Percentage of time spent in open arms. D Percentage of entries into open arms. E-I. Anxiety-like behavior and locomotor mobility of rats were assessed by performing open-field experiments. Representative trajectory diagrams of activity within the open field in sham (E_left) and BSI (E_right) rats. F Average velocity. G Total distance traveled. H Percentage of time spent in the center area. I–P Automated CatWalk gait analysis was performed to assess motor function in rats. I Maximum intensity. J Max footprint area. K Average footprint area. L Walking trajectories and representative footprint maps of rats. M Footprint width. N Footprint length. O Stride length. P Swing time. n = 8 biologically independent samples in the sham group and n = 10 biologically independent samples in the BSI group. Unpaired two-tailed t-test for two group comparisons. Data are presented as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. LF, Left Forepaw; LH, Left Hindlimb; RF, Right Forepaw; RH, Right Hindlimb To further explore motor function changes in the model rats, we conducted gait analysis using the automatic CatWalk gait analysis system. Based on the static CatWalk gait parameters of the rats, we assessed the maximum contact intensity of the feet and the maximum and average contact areas. Compared to the sham group, the BSI group showed a significant decrease in the maximum contact intensity of both forepaws, a reduction in the maximum and average contact areas, and a marked decrease in the maximum contact area of both hindpaws (Fig. 3I-K). Observations of the footprint images revealed that the footprint width and stride length of the left forepaw in the BSI group were significantly lower than those in the sham group, while the footprint length remained unchanged (Fig. 3L-O). Notably, the swing time of all four limbs during walking was significantly longer in the BSI group than in the control group (Fig. 3P). These findings indicated that motor function was impaired in BSI rats. Evaluation of swallowing function in rats with BSI To assess swallowing function in rats with BSI, VFSS was conducted on day 5 post-modeling. 18 videofluoroscopic swallowing videos were recorded and manually analyzed, including 10 from the BSI group, with 10 consecutive swallows per rat analyzed and averaged. The Fig. 4A, B showed representative radiographic images of rats undergoing VFSS and illustrated bolus area during swallowing in the sham and BSI group (the bolus was outlined in white). Analysis (Fig. 4C–E) revealed that, compared to the sham group, rats in the BSI group exhibited significantly prolonged pharyngeal transit time and inter-swallow interval, as well as reduced bolus area. However, there was no significant difference in esophageal transit time between the two groups (Fig. 4F). In addition, after modeling, rats in the BSI group showed a significant decrease in body weight, while the sham group showed normal weight gain (Fig. 4G). These above data confirmed dysphagia induced by BSI. Verification of swallowing disorders in BSI rats. A, B Schematic representation of VFSS and illustration of single-swallow bolus size in sham and brainstem ischemic rats during VFSS examination (bolus is marked with white outlines). C–F VFSS examination showed that compared with the control group, the pharyngeal transit time, inter-swallowing interval time were significantly prolonged and the size of the food mass was remarkably reduced in brainstem ischemic rats, but there was no significant difference in the esophageal transit time. G Rats showed a significant decrease in body weight after modeling. n = 8 biologically independent samples in the sham group and n = 10 biologically independent samples in the BSI group. Unpaired two-tailed t-test for two group comparisons, and two-way RM ANOVA for comparisons including more than two groups. Data are presented as mean ± SD, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 Validation of the efficacy of mPES treatment for BSI rats Previous studies have found that mPES treatment improved dysphagia after stroke [21]. On this basis, this study aims to further explore whether mPES can also improve dysphagia in BSI model rats, with the expectation of providing a more solid experimental basis for clinical treatment. First, the outcomes of EPM test (Fig. 5A-C), open field experiment (Fig. 5D-G), and CatWalk gait analysis (Fig. 5H-O) collectively indicated that mPES intervention failed to produce statistically significant changes in motor function and anxiety-related behavior in BSI rats. Second, VFSS showed that prior to mPES treatment, there were no statistically significant differences between the mPES group and the BSI group in terms of pharyngeal transit time, esophageal transit time, inter-swallow interval, or bolus area during swallowing (seen in Additional file 1). After mPES treatment, the delivery time of the bolus in the pharynx and the inter-swallow interval were significantly shortened in mPES group (Fig. 5P-Q). And the bolus area was significantly increased (Fig. 5R), but the study did not find a significant change in the esophageal delivery time of the bolus (Fig. 5S). Third, after mPES intervention, there was a slight enhancement in the body weight of BSI rats (Fig. 5T). These demonstrated that mPES could effectively improve dysphagia in BSI rats. The effects of mPES treatment on the swallowing function and neurobehavioral outcomes in BSI model rats. A–C Elevated plus maze test. A Movement trajectories of the BSI and the mPES rats on the elevated-cross frame. B Percentage of time spent in open arms. C Percentage of entries into open arms. D–G. Open field test. D Movement trajectories of BSI and mPES rats in the open field. E Average speed. F Total distance traveled. G Percentage of time spent in central area. H–O Automated CatWalk gait analysis in the BSI and the mPES groups. H Max intensity. I Max print area. J Average print area. K Rat walking trajectory and representative footprint images. L Print width. M Print length. N Stride length. O Swing time. P-S VFSS examination showed that compared with the BSI group, the mPES treatment group had significantly shorter pharyngeal transit time (P), inter-swallowing interval time (Q), and larger food bolus area (R), but no significant difference in esophageal transit time (S). T. Compared with the BSI group, rats in the mPES group showed a slight increase in body weight. n = 10 biologically independent samples in the BSI group and n = 8 biologically independent samples in the mPES group. Unpaired two-tailed t-test for two group comparisons, and two-way RM ANOVA for comparisons including more than two groups. Data are presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. LF, Left Forepaw; LH, Left Hindlimb; RF, Right Forepaw; RH, Right Hindlimb Single-cell transcriptomics analysis of the mPES treatment for the dysphagia model induced by BSI Overall analysis To investigate the specific mechanism by which mPES improves dysphagia after BSI, we performed single-cell RNA transcriptomics. After rigorous quality control, we obtained a total of 37,192 cells. The UMAP plot illustrated the integration of data from different sources (Fig. 6A). Based on the expression of specific marker genes for different cell types, we identified eight distinct cell clusters, including neurons (Rtn1, Gad2, Map2, Snap25, Syp), oligodendrocytes (Mbp, Plp1, S100b, Vcan, C1ql1), microglia (Ctss, Cx3cr1, iba1, C1qa, C1qb), astrocytes (Slc1a3, Slc6a11, Slc7a10, Acsl6, Gja1), pericytes (Pecam1, Fn1, Abcc9, Kcnj8, Rgs5), fibroblasts (Dcn, Col1a1, Bgn, Osr1), neutrophils (S100a9, S100a8, Ltf, Ly6g), and B cells (Ebf1, Cd19, Cd79a, Cd74) (Fig. 6B). To determine the types of brain cells that may be affected by mPES treatment, we conducted a comprehensive comparison of the composition ratios of each cell type between the BSI group and the mPES group. As shown in Fig. 6C, D, the stacked plot and bar chart illustrated the changes in the relative expression of different cell types within each group. Compared with the BSI group, the number of oligodendrocytes in the brainstem ischemic lesion increased after mPES treatment (27.89% vs. 31.57%), which may be closely related to their key characteristics, including myelin repair, axon regeneration, and support for neurons [44, 45]. Additionally, the proportions of astrocytes, microglia, and fibroblasts decreased after mPES treatment. Transcriptome profiles of rat brainstem through single-cell RNA sequencing in BSI model rats. A UMAP plot demonstrating single-cell clusters colored according to sample origin. B UMAP plot showing single-cell clusters colored by cell type. C-D Bar graph displaying the proportion of each cell cluster identified in each group. E–F Scatter plots and volcano plots show the 10 significantly differentially expressed genes between groups (including both up- and down-regulated genes). G-H Bubble plots presenting the top 15 enriched GO terms and enriched signaling pathways involved in the DEGs between groups in order To investigate the DEGs and pathways or biological processes potentially involved in mPES intervention, we compared the significant changes of genes between the mPES and BSI groups, and performed GO and KEGG analyses on these DEGs. In comparison with the BSI group, mPES intervention up-regulated genes such as Mt-co1, Mt-co3, Mt-co2, Plp1, and S100b in oligodendrocytes, while down-regulating genes such as Prex2, Zbtb16, Sat1, Pdk4, and Fkbp5 (Fig. 6E, F). GO analysis implied that DEGs were significantly enriched in biological processes such as lipid biosynthetic process, regulation of dendrite morphogenesis, axon ensheathment and oligodendrocyte differentiation (Fig. 6G). KEGG analysis used p-value as screening criteria and found that DEGs were highly concentrated in signaling pathways including steroid biosynthesis, Wnt signaling pathway, and Phospholipase D signaling pathway (Fig. 6H). Subpopulation heterogeneity analysis of oligodendrocytes Myelin destruction is considered a key feature of ischemic stroke, and oligodendrocytes play an important role in maintaining myelin integrity. To gain a deeper understanding of the transcriptional changes in oligodendrocyte sub-populations following mPES intervention, we specifically isolated oligodendrocytes for subpopulation analysis. First, compared with the BSI group, mPES treatment primarily up-regulated genes such as Mt-co1, Mt-co3, Mt-co2, Cyp51, and S100b in oligodendrocytes, while downregulating genes such as Prex2, Zbtb16, Sat1, Pdk4, and Fkbp5 (Fig. 7A). Second, GO enrichment analysis revealed significant enrichment in processes such as lipid biosynthesis, regulation of dendrite morphogenesis, axonogenesis and synapse assembly (Fig. 7B). KEGG analysis primarily showed enrichment in pathways such as steroid biosynthesis, nicotinate and nicotinamide metabolism and Wnt signaling pathway (Fig. 7C). Third, by re-annotating and identifying genes up-regulated in oligodendrocyte subpopulations, we ultimately identified 3 distinct oligodendrocyte sub-populations: oligodendrocyte sub-cluster 0 (we named OL0, Grm3, Mast4, Mbp), oligodendrocyte sub-cluster 1 (we named OL1, Thsd7b, Mbp, S100b), and oligodendrocyte sub-cluster 2 (we named OL2, Nxph1, Pdgfra, Vcan) (Fig. 7D-H). We intuitively observed that, compared with the BSI group, the proportion of oligodendrocyte OL1 increased in the mPES group, while the proportion of OL2 decreased. However, these observed differences have not yet reached statistical significance (Fig. 7E-F). Finally, we conducted GO and KEGG analyses of the DEGs between these two sub-clusters of oligodendrocyte to explore their heterogeneity. Heterogeneity analysis results of oligodendrocyte subpopulations. A The scatter plot shows the 10 genes with significant differences in oligodendrocytes between different groups (including up- and down-regulated genes). B-C Bubble plots illustrating the top 15 enriched GO terms and enriched signaling pathways associated with DEGs in oligodendrocytes between groups. D UMAP plot showing single-cell clusters colored by cell type. E–F. Bar charts displaying the proportion of each identified cell cluster in each group. G. Volcano plot presenting the top 10 genes with the highest upregulation fold-change across cell subclusters. H. Ridgeplot depicting the relative expression of marker genes across subclusters. I-L Bubble plots showing the top 15 enriched GO terms and enriched signaling pathways associated with genes in OL1 cluster (I, K) and OL2 cluster (J, L) The GO enrichment analysis revealed that OL1 cluster was significantly enriched in processes involving lipid biosynthesis, signal transduction, phospholipid transport and myelin maintenance (Fig. 7I). In contrast, OL2 cluster was primarily enriched in processes such as neurotransmitter secretion, oligodendrocyte differentiation, nervous system development and axon guidance (Fig. 7J). KEGG enrichment analysis showed that OL1 cluster was concentrated in pathways such as steroid biosynthesis, Wnt signaling pathway, PI3K-Akt signaling pathway and sphingolipid signaling pathway (Fig. 7K). OL2 cluster exhibited enrichment in pathways including the calcium signaling pathway, PI3K-Akt signaling pathway, Wnt signaling pathway and sphingolipid signaling pathway (Fig. 7L). These suggested that OL1 cluster was preferentially enriched in lipid synthesis and myelin formation, while OL2 cluster was preferentially involved in oligodendrocyte differentiation and axon regeneration. Additionally, both cell clusters were co-enriched in signaling pathways such as the Wnt, PI3K-Akt, and sphingolipid signaling pathways. Based on marker genes and enrichment analysis, we tended to define OL1 cluster as oligodendrocytes and subcluster 2 as oligodendrocyte precursor cells. Developmental analysis of oligodendrocytes The Monocle2 software package was used to perform pseudo-time analysis of oligodendrocyte sub-clusters to investigate their developmental differentiation. Pseudo-time information displayed three distinct developmental stages and a branching point (Fig. 8A). State 1 represented the initial developmental stage, followed by states 2 and 3 (Fig. 8B). This analysis revealed sequential developmental correlations among oligodendrocyte sub-clusters, suggesting that OL2 cluster may differentiate into OL1 and OL0 cluster (Fig. 8C). This serves as a hint that oligodendrocyte precursor cells can mature into oligodendrocytes, a conclusion that resonates with prior studies [45, 46]. Additionally, we identified genes such as Nav2, Smad5, Sorcs5, Sox6, Spon1, and Xylt1 as fate-determining genes for differentiation from state 1 to states 2 and 3, thereby promoting the developmental differentiation of oligodendrocytes (Fig. 8D–F). Pseudotime analysis and cell communication analysis of oligodendrocytes. A Developmental timeline of oligodendrocyte subpopulations. Darker shades indicate lower pseudo-time values, representing earlier developmental stages. B Oligodendrocyte subpopulations can be divided into three distinct developmental stages. C Annotations for the three distinct developmental stages. D–F Differential expression of fate-determining genes Nav2, Smad5, Sorcs5, Sox6, Spon1, and Xylt1 during differentiation. G-H Heatmaps showing ligand-receptor relationships between cell subpopulations in the BSI group (G) and mPES group (H) I–J Network diagrams visualizing changes in ligand-receptor relationships between cell subpopulations in the BSI group (I) and mPES group (J) The size of the circles reflected the number of ligand-receptor pairs in each subpopulation. Thicker lines indicate stronger communication strength. K–L. Bubble plots depicted signaling pathway relationships between cell subpopulations in the BSI group (K) and mPES group (L) Analysis of the inter-cellular communication network of oligodendrocytes We employed the CellPhoneDB software to investigate ligand-receptor interactions between oligodendrocyte sub-clusters in different groups. Heatmaps and Network diagram plots indicated that the number and strength of ligand-receptor interactions were seemingly higher between the C0-C0, C0-C2, C1-C0, C1-C1, and C1-C2 subpopulations following mPES intervention (Fig. 8G–J). Bubble plots (Fig. 8K–L) further indicated that signaling pathways involving TENM4_ADGRL3, TENM3_ADGRL3, and TENM2-ADGRL3 were formed between subcluster 1 and 2 after mPES intervention. The signaling pathways formed between OL0 and OL2 cluster exhibited a similar pattern to those between OL1 and OL2 cluster. In addition, OL0 and OL1 cluster established signaling pathways with themselves via NRG3_ERBB4 and TENM2-ADGRL3, respectively. A signaling pathway involving NLGN3_PTPRD was observed between OL0 and OL1 cluster, which may reveal potential key communication pathways and cellular regulatory mechanisms of mPES. Discussion In this study, we employed a photochemical embolism method to establish a BSI model, and verified the phenotypic characteristics of swallowing disorders in rats using the gold-standard VFSS. This indicated that we have successfully established a model of dysphagia caused by BSI. Furthermore, our findings demonstrated that mPES can significantly improve swallowing function in the BSI rats. Additionally, we innovatively used single-cell RNA-sequencing to explore the therapeutic molecular mechanism of mPES in the BSI-induced dysphagia model and revealed that mPES might exert its therapeutic effect by regulating the differentiation of oligodendrocytes and the regeneration and repair of myelin and synapses. This study offered new perspectives for delving into the potential mechanisms underlying mPES treatment for post-stroke dysphagia. In previous studies, the most commonly used dysphagia model is the tMCAO-induced dysphagia model. It shows prolonged swallowing-initiation latency and reduced swallowing frequency [4]. However, the dysphagia phenotype in this model is often unstable, which may be related to the large fluctuations of the ischemic volume [47]. In addition, swallowing function impairment in this model typically emerges 2–3 weeks post-modeling, which is significantly different from the timing of dysphagia onset in stroke patients [4]. In addition, the M1 photochemical embolism model of dysphagia is characterized by a high survival rate and precise damage. However, it has relatively low stability of swallowing disorder phenotype [5, 48]. Further, bilateral common carotid artery ligation, as a model of chronic cerebral hypoperfusion, is simple to perform but suffers from a relative lack of specificity, probably due to the multiple brain regions affected by the injury [6]. It should be noted that the above PSD model is mainly triggered by ischemia in the cortex and subcortical tissues. Although there is an association between cortical ischemia and dysphagia, brainstem ischemia, as a key regulatory center of the swallowing response, renders patients more susceptible to dysphagia [9, 12], which may be attributed to the cranial nuclei in the brainstem that control and regulate the swallowing response [49]. Moreover, dysphagia caused by BSI is significantly different from that induced by cortical ischemia in terms of pathophysiological mechanisms and clinical manifestations [50, 51]. This study creatively focuses on the field of brainstem ischemia, addressing the limitations of previous related research. By simulating brainstem ischemia, we were able to observe more precisely the specific effects of brainstem injury on the swallowing reflex. In addition, the photochemical embolization approach was able to induce ischemic infarction in specific regions of the brainstem with greater precision, mimicking the pathological mechanisms of PSD in humans more closely and with high reproducibility. Although the female stroke patient population is also substantial, this study selected male rats exclusively for model exploration primarily based on scientific and experimental control considerations. On the one hand, female rats exhibit distinct estrous cycles, with significant fluctuations in estrogen and progesterone levels. However, these hormones exert neuroprotective effects in ischemic brain injury [52], potentially introducing confounding factors that could compromise the stability and reproducibility of experimental outcomes and interfere with the analysis of key signaling pathways [53]. On the other hand, previous ischemic stroke research has predominantly employed male animal models [54]. Utilizing male animals facilitates direct comparison and validation of our findings against existing literature. Additionally, although stroke patients are mostly elderly individuals, we chose 8-week-old rats as experimental subjects in this study. First, rats aged 6–8 weeks typically reach sexual maturity and are widely used in stroke research [55–57]. Second, previous studies on focal cerebral ischemia have shown that young rats are more sensitive to ischemic injury than aged rats [58]. This sensitivity facilitates the presentation of more pronounced pathological changes, thereby enhancing the model's operability and the sensitivity of intervention effect detection. Finally, aged rats may introduce greater experimental variability due to age-related physiological changes (such as vascular dysfunction, reduced swallowing-related muscle strength and coordination, or comorbidities), potentially compromising result consistency and reliability [59]. In this study, the infarcted areas of the model primarily involved the facial nucleus, the lateral paragigantocellular nucleus, and part of the raphe nucleus. Coincidentally, previous studies have found that these brain regions are closely associated with natural swallowing or reflexive swallowing pathways. First, interneurons in the facial nucleus project to the hypoglossal nucleus, while neurons in the paragigantocellular reticular formation project to both the facial nucleus and the hypoglossal nucleus [60, 61]. This may facilitate coordinated activity of facial and tongue muscles during swallowing [7], reflecting the integration of natural and reflexive swallowing. Second, studies have confirmed the existence of projection pathways from the laryngeal and pharyngeal motor cortex regions to the parabrachial nucleus (PBN) [7, 19, 62, 63]. Specifically, neurons in the lateral PBN directly project to pharyngeal motor neurons in the nucleus ambiguus, participating in the control of pharyngeal movement and thereby influencing the execution of reflexive swallowing [64, 65]. Additionally, research has shown that quipazine may stimulate raphe nucleus neurons, thereby enhancing the excitability of the swallowing network and increasing natural swallowing frequency [66, 67]. In summary, these brain regions participate in both voluntary and reflexive swallowing behaviors. Therefore, we can more comprehensively assess the functional integrity of the entire neural pathway from the cortex to the brainstem and periphery by evaluating voluntary swallowing through VFSS in this study. This also explains the occurrence of dysphagia symptoms in the model used in this study. Similar to previous studies, the present study also identified that mPES markedly improved swallowing indexes in rats with dysphagia after BSI. This further validates the therapeutic effect of mPES in dysphagia following cerebral ischemia. Our study also found that, compared with the control group, there was no significant change in the transit time of food bolus in the esophagus of rats in the BSI group, and mPES treatment had no obvious effect on esophageal transit time. This may be because dysphagia after brainstem ischemia mainly manifested as pharyngeal swallowing disorders, and mPES mainly acts on the pharynx [15]. Previous studies suggested that white matter damage following stroke may disrupt neural connections, impairing the function of swallowing-related brain networks such as the corticobulbar tract, thereby leading to dysphagia [68, 69]. Research has indicated that mPES may improve swallowing function by enhancing corticobulbar excitability and inducing cortical reorganization in the swallowing motor cortex [70, 71]; however, the specific cellular mechanisms involved remain poorly understood. In this study, we innovatively applied single-cell RNA sequencing technique to analyze the cellular landscape of brain tissue in BSI rats treated with or without mPES. A total of eight cell subpopulations were identified: neurons, oligodendrocytes, astrocytes, microglia, pericytes, fibroblasts, neutrophils, and B cells. Among these, only the proportion of oligodendrocytes showed an increase in mPES-treated rats compared to the BSI group. Previous studies suggest that oligodendrocytes and myelin integrity are crucial for axonal functional recovery following white matter injury [72, 73]. This indicates that the therapeutic effect of mPES for BSI may be related to oligodendrocytes. Recent studies reveal that myelin abnormalities can impair neural signal transmission, leading to the onset of dysphagia [74]. Furthermore, central inflammation following stroke triggers systemic immune suppression, potentially exacerbating dysphagia symptoms [75]. Similarly, this study found that mPES treatment upregulates genes closely associated with myelin formation maintenance—including apolipoprotein D, myelin lipoprotein, and myelin basic protein, and promotes expression of the S100b gene (Fig. 4E–F), which facilitates neurite growth, differentiation, and repair [76]. Concurrently, it downregulates genes associated with promoting apoptosis and inflammation—such as Prex2, Rn18s, and Fkbp5 [77]. This suggests that mPES may mitigate inflammatory damage, promote oligodendrocyte proliferation and differentiation, and repair myelin sheaths by regulating these gene expressions, thereby facilitating swallowing function recovery. However, more experiments are needed to validate this conclusion. Previous studies have demonstrated that myelin is a multilayered membrane structure enveloping neuronal axons, with its high lipid content being crucial for maintaining axonal insulation and efficient signal transmission [78]. Among these, sphingomyelin, as a key structural component of cell membranes, and its metabolites—such as ceramides (Cer) and sphingosine-1-phosphate (S1P)—participate in post-ischemic pathological processes by regulating apoptosis, inflammatory responses, and vascular function [79, 80]. Moreover, following cerebral reperfusion injury, the Wnt/β-catenin pathway inhibits cell apoptosis and promotes synaptic plasticity by regulating downstream factors including PI3K/Akt and GSK-3β [81]. Our study revealed that compared to the brainstem ischemia group, genes in the mPES treatment group were predominantly enriched in lipid synthesis, axogenesis, sphingomyelin metabolism, and Wnt signaling pathways. Coincidentally, this result closely resembles with the enrichment findings from oligodendrocyte subpopulation analysis. These suggest that mPES's improvement in swallowing function might relate to oligodendrocytes regulating lipid metabolism, apoptosis, and synaptic plasticity to mitigate pathological processes after cerebral ischemia. However, this hypothesis requires further experimental validation. Further subpopulation analysis revealed that oligodendrocytes are not a homogeneous population but comprise functionally specialized subpopulations. We identified the OL1 cluster as specifically overexpressing genes associated with lipid metabolism and myelin maintenance. In contrast, the OL2 cluster exhibited functional characteristics more closely associated with cell differentiation, axon guidance, and neuronal interactions. This finding aligns with recent studies highlighting the functional diversity of oligodendrocytes. For instance, beyond myelin formation, oligodendrocytes directly participate in maintaining axonal integrity and regulating neural circuits through mechanisms such as secreting neurotrophic factors [82, 83]. Pseudo-time analysis suggests that mPES treatment may influence the differentiation trajectory of oligodendrocytes. Furthermore, we identified genes such as Sox6, Nav2, Smad5, Xylt1, Sorcs5, and Spon1 as core fate-determining factors in this differentiation pathway. Of these, previous studies have shown that after oligodendrocyte precursor cells differentiate into immature oligodendrocytes, SOX6 promotes chromatin decondensation and transcriptional activation, thereby maintaining the immature cellular state. Inhibiting SOX6 accelerates the transition to the mature myelin-forming stage [84]. The widespread expression of Nav2 in central nervous system development suggests its potential involvement in oligodendrocyte development or differentiation processes [85]. Smad5 serves as a key mediator in the BMP/TGF-β signaling pathway, which critically regulates glial cell differentiation [86]. Furthermore, XYLT1 upregulation correlates with multiple pathological processes, where it modifies the microenvironment by regulating ECM components, thereby influencing cellular differentiation and behavior [87, 88]. Collectively, these genes form a complex regulatory network that precisely guides the fate decisions of oligodendrocytes. Analysis of cell-to-cell communication within oligodendrocyte subpopulations revealed that mPES intervention may remodel the communication networks within these subpopulations, potentially representing a key mechanism for its neuroprotective effects. First, following mPES intervention, interactions between specific subpopulation combinations (e.g., C0-C0, C0-C2, C1-C0, C1-C1, C1-C2) increased in both quantity and intensity. Second, we identified several signaling pathways. Among these, the TENM-ADGRL3 signaling pair was most notably established between OL1-OL2 and OL0-OL2 clusters. Tenm and ADGRL3 are recognized cell adhesion molecules that form trans-synaptic complexes. Their diverse expression across developmental stages and spatial axes mediates axon guidance, synapse formation, and neural circuit assembly [89–92]. Furthermore, the activation of the NRG3-ERBB4 pathway within OL0 clusters and the formation of the NLGN3-PTPRD pathway between OL0 and OL1 clusters add another layer of complexity to the interaction network. Beyond inducing receptor dimerization and autophosphorylation to activate downstream signaling pathways like PI3K/AKT and MAPK/ERK—thereby influencing cell survival, differentiation, and synaptic plasticity [93, 94]—the NRG3-ERBB4 interaction also regulates dendritic and axonal growth [95]. Recent studies reveal that NLGN3 and PTPRD not only function in synaptic formation and plasticity but also promote axon guidance and initiate axonal myelination [96–99]. The identification of these pathways among oligodendrocyte subpopulations suggests that mPES may activate a sophisticated mechanism, analogous to inter-neuronal communication, to regulate oligodendrocyte proliferation and differentiation while promoting synaptogenesis, axon guidance, and axonal myelination. However, additional studies are required to confirm this finding. Previous studies have primarily focused on the impact of cortical damage on swallowing function. This research, however, targets brainstem ischemia as a specific pathological condition, offering a novel perspective for a more comprehensive understanding of the pathogenesis of PSD. Furthermore, the cellular heterogeneity and cell-type-specific gene expression pattern changes revealed by single-cell sequencing following mPES treatment provide new clues for subsequent in-depth investigations into specific cell–cell interactions and signaling pathways. However, this study has several limitations. First, the long-term symptomatic manifestations of the model and the sustained efficacy of mPES in treating dysphagia following brainstem ischemia require further observation; Second, due to experimental limitations, the actual feeding efficiency and energy expenditure during feeding were unable to quantify. Third, the single-cell sequencing sample size in this study was insufficient. Future research should increase the sample size and conduct experimental validation to further explore the mechanisms underlying mPES treatment for brainstem ischemic dysphagia. For instance, specific modulation of oligodendrocyte subpopulations and molecular targets through gene editing or pharmacological interventions could elucidate how mPES exerts therapeutic effects via oligodendrocytes and target molecules, as well as interactions among oligodendrocytes subpopulations. Finally, integrating multimodal imaging techniques with single-cell sequencing data could enable deeper exploration of the spatiotemporal dynamics linking brainstem microstructural remodeling to cellular functional changes. Conclusion In this study, we succeeded in establishing a rat model of dysphagia induced by BSI, which exhibited symptoms of dysphagia, and mPES can effectively improve the swallowing function in this rat model. These results suggested that dysphagia model induced by BSI was expected to be a useful animal model for exploring the pathophysiological mechanisms and therapeutic methods of dysphagia after BSI. In addition, single-cell sequencing revealed that mPES might exert its therapeutic effect on dysphagia post-BSI by influencing the differentiation and the regeneration and repair of myelin and synapses. Supplementary Information Supplementary file 1. Supplementary file 2. Acknowledgements Thanks to Dr. Siqi Li for providing technical guidance on MRI scanning. Abbreviations Brainstem ischemia Modified pharyngeal electrical stimulation Post-stroke dysphagia Videofluoroscopic Swallowing Study Transient middle cerebral artery occlusion 2,3,5-Triphenyltetrazolium Chloride staining T2-weighted imaging Fast spin echo Number of Excitations Repetition Time Echo Time Field of view Magnetic resonance imaging Elevated plus maze Pharyngeal Transit Time Esophageal Transit Time Inter-swallow interval Ribonucleic acid T-distributed stochastic neighbor embedding Uniform manifold approximation and projection Differentially expressed genes Gene Ontology Kyoto encyclopedia of genes and genomes Discriminative dimensionality reduction with trees Branch expression analysis modeling Neurological deficit score Reticulon 1 Glutamic acid decarboxylase 2 Microtubule-associated protein 2 Synaptosomal-associated protein 25 Synaptophysin Myelin basic protein Proteolipid protein 1 S100 calcium-binding protein b Versed can Complement component 1 Q subcomponent-like 1 Cathepsin S C-X3-C motif chemokine receptor 1 Ionized calcium-binding adaptor molecule 1 Complement component 1 Q subcomponent A Complement component 1 Q subcomponent B Solute carrier family 1 member 3 Solute carrier family 6 member 11 Solute carrier family 7 member 10 Acyl-CoA synthetase long-chain family member 6 Gap junction protein alpha 1 Platelet endothelial cell adhesion molecule 1 Fibronectin 1 ATP-binding cassette subfamily c member 9 Potassium voltage-gated channel subfamily J member 8 Regulator of G-protein signaling 5 Decorin Collagen type I alpha 1 Biglycan Odd-skipped related transcription factor 1 S100 calcium-binding protein A9 S100 calcium-binding protein A8 Lactotransferrin Lymphocyte antigen 6 complex locus G Early B-cell factor 1 CD19 molecule CD79a molecule CD74 molecule Mitochondrially encoded cytochrome C oxidase I Mitochondrially encoded cytochrome C oxidase III Mitochondrially encoded cytochrome C oxidase II Hemoglobin subunit beta Phosphatidylinositol-3,4,5-trisphosphate-dependent Rac exchanger 2 Zinc finger and BTB domain containing 16 Spermidine/Spermine N1-Acetyltransferase 1 Pyruvate dehydrogenase kinase 4 FK509 binding protein 5 Glutamate metabotropic receptor 3 Macrophage stimulating 1 receptor Thrombospondin type I domain containing 7B Neurexophilin 1 Platelet-derived growth factor receptor alpha Wingless-type MMTV integration site family Phosphoinositide 3-Kinase-Akt signaling pathway Neuronal voltage-gated sodium channel 2 SMAD family member 5 Sorting and signaling coreceptor 5 SRY-box transcription factor 6 Spondin 1 Xylosyltransferase 1 Teneurin transmembrane protein 4 Adhesion G protein-coupled receptor L3 Teneurin transmembrane protein 3 Teneurin transmembrane protein 2 Neuroligin 3 Protein tyrosine phosphatase receptor type D Neuregulin 3 Erb-B2 receptor tyrosine kinase 4 Titin Cytochrome P450 Family 51 Apolipoprotein D Transthyretin Glutamate metabotropic receptor 8 Inter-alpha-trypsin inhibitor heavy chain 3 Heparan sulfate-glucosamine 3-sulfotransferase 4 Ring finger protein 220 Ubiquilin 1 Mitogen-activated protein kinase Extracellular regulated protein kinases Bone morphogenetic protein Transforming growth factor beta Extracellular matrix Author contributions Thank you for the contributions of all authors. YQT, ZLD and QPY conceptualized and designed the study. YQT, QQW, JHH, JQ and QPY established the model, performed treatment and evaluation, and data analysis. CL, XXZ, TTJ and NGX performed scRNA sequencing, and data analysis. YQT contributed to the initial manuscript writing. QPY, HMW and ZLD reviewed the manuscript. All the authors contributed to the manuscript revision and approved the submitted version. Funding This work was supported by the National Natural Science Foundation of China (82202807 to Qiuping Ye, 82572939, to Zulin Dou, 82372569 to Hongmei Wen). Clinical Collaboration Project of Traditional Chinese and Western Medicine for Major and Difficult Diseases (Post-Stroke Dysphagia) (ZDYN-2024-A-018 to Dou Zulin), Guangdong Provincial Clinical Research Center for Rehabilitation Medicine (2023B110003). Data availability The raw single-cell RNA-seq data reported in this paper have been deposited in the Gene Expression Omnibus (GEO) database under accession number GSE303298. The other data used and/or analysed during the current study is available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Yueqin Tian, Qianqian Wang and Jiahui Hu have contributed equally to this work. Contributor Information Hongmei Wen, Email: wenhm0625@126.com. Zulin Dou, Email: douzulin@mail.sysu.edu.cn. Qiuping Ye, Email: yeqp3@mail.sysu.edu.cn. References Associated Data Supplementary Materials Supplementary file 1. Supplementary file 2. Data Availability Statement The raw single-cell RNA-seq data reported in this paper have been deposited in the Gene Expression Omnibus (GEO) database under accession number GSE303298. The other data used and/or analysed during the current study is available from the corresponding author on reasonable request.