Single-cell transcriptome analysis reveals mechanisms by which hippocampal deep brain stimulation promotes neurorepair and microglial subpopulation remodeling in ischemic stroke

Abstract

Background

In patients with ischemic stroke, the hippocampus is one of the most severely damaged regions of the brain; however, the optimal parameters for hippocampal deep brain stimulation in these patients remain unknown. The underlying mechanism of hippocampal deep brain stimulation in the treatment of ischemic stroke is not fully understood.

Objective

To investigate the effect and underlying mechanisms by which hippocampal deep brain stimulation enhances the repair of brain damage from ischemic stroke.

Methods

We utilized multi-channel electrode arrays to record local field potentials and to monitor motor symptoms in mice with ischemic stroke during deep brain stimulation in hippocampus. Using 10x single-cell transcriptome sequencing and RNA-seq in stroke mice with and without electrostimulation treatment, we explored the underlying mechanisms.

Results

After being treated with hippocampal deep brain stimulation, the stroke mice showed significant improvements in motor ability, neuronal function, and behaviors related to anxiety/depression. Bioinformatics analysis and subsequent investigations revealed a significant increase in the quantity of excitatory neurons and astrocytes in the hippocampus following deep brain stimulation while the number of oligodendrocytes, microglia and inhibitory neurons was markedly diminished. The microglia subset 5 showed a significant increase following deep brain stimulation treatment, while subsets 3 and 4 experienced a significant reduction. Within subset 5, markers such as Cd93, Ccr2, Emilin2, Pirb, Lgals3, Thbs1, Gpnmb, and Ccr1 were associated with immune inflammation and angiogenesis.

Conclusion

Collectively, our findings imply that hippocampal deep brain stimulation is highly effective for treating ischemic stroke and offer deeper insights into the potential mechanisms by which hippocampal deep brain stimulation improves stroke symptoms. These findings support hippocampal deep brain stimulation as a promising novel therapeutic approach for stroke and offer valuable insights for the treatment of other neurological disorders.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-07388-0.

Introduction

Stroke, a prevalent and severe cardiovascular disease, stands as the second most common cause of morbidity and mortality worldwide [1]. It accounts for a substantial 5% of disability-adjusted life years [2]and for 10% of global mortality [3]. Defined as a condition stemming from cerebral hemorrhage or vascular occlusion, stroke encompasses two broad categories based on pathogenesis: ischemic stroke (IS) and hemorrhagic stroke (HS) [4]. In roughly 80% of cases, stroke arises from a deficiency in blood supply caused by cerebral artery obstruction and/or occlusion [1]. Current data indicate a poor prognosis for stroke, with the majority of survivors experiencing long-term and significant neurological impairments [5]. These encompass a plethora of sensory and motor sequelae, as well as cognitive and psychological deficits. Post-stroke depression (PSD) is the most common psychological issue, affecting approximately 33% of stroke survivors [6]. PSD patients frequently experience symptoms of anxiety and cognitive impairment [6, 7]. Currently, acute stroke treatment focuses on rapid reperfusion through intravenous thrombolysis and endovascular thrombectomy [8], followed by a variety of rehabilitative interventions. Nonetheless, investigations reveal that the risk of post-stroke mortality continues to escalate within the first year, and numerous stroke survivors experience incomplete recovery due to residual functional deficits and disruptions, which lead to compromised quality of life [9]. This underscores the importance of exploring novel therapeutic approaches to alleviate the burden imposed by stroke.

Deep brain stimulation (DBS) is an invasive form of neuromodulation that involves the placement of electrodes within dysfunctional neural circuits to provide electrical stimulation (ES) [10]. This process suppresses aberrant activity and/or drives underactive networks. The ES in DBS can be administered either continuously or intermittently, and the stimulation parameters (frequency, pulse width, voltage) as well as contacts are programmable. Currently, DBS has emerged as the standard treatment for movement disorders, particularly Parkinson’s disease (PD) [10, 11], essential tremor, and dystonia [12, 13], and is also under investigation for its potential application in treating other psychiatric disorders including stroke [14–16]. Clinical studies have suggested that DBS may hold therapeutic potential for the treatment of PSD, including neuropathic pain, tremors, dystonia, and movement disorders [16, 17]. The primary targets of DBS include the posterior limb of the internal capsule (PLIC), the sensory thalamus and the periventricular/periaqueductal grey (PVG/PAG) [18]. However, these studies have primarily focused on using DBS at various targets to address specific post-stroke sequelae and on assessing treatment efficacy based on symptom improvement. This approach is highly targeted towards specific symptoms and may not be suitable for most stroke patients. Additionally, there is limited research on the mechanisms underlying the effects of DBS on stroke symptoms, and the optimal DBS targets are still under investigation. The hippocampus, a crucial neurogenic region in the brain, not only plays a key role in memory, spatial navigation, and emotional regulation but also exhibits extensive neural connections with other brain regions [19]. In IS, the hippocampus is one of the most prominently affected areas [20]; nonetheless, the exploration of post-stroke DBS targeting the hippocampal region has been largely neglected. The hippocampus exhibits high neural plasticity, making DBS targeting this region a potentially powerful intervention for stroke recovery.

This study aims to employ single-cell RNA sequencing technology to investigate the association between hippocampal DBS (Hip-DBS) and IS, with the goal of further elucidating the underlying mechanisms. Specifically, the study will explore the potential therapeutic effects of Hip-DBS on microglia in IS by examining how Hip-DBS modulates microglial state. Through this research, our aim is to clarify the functional role and potential effects of Hip-DBS in IS, thereby contributing to a deeper understanding of the therapeutic mechanisms underlying stroke treatment.

Methods

Animals and establishment of the IS model

Male C57BL/6 (WT) mice weighing 22–25 g (8–10 weeks old) were used. Mice were housed in a specific pathogen-free facility at Shanghai University and maintained at a constant temperature (22 ± 1 °C), humidity (60–80%), and 12 h/12 h light/dark cycle, with free access to water and food. For IS, a transient middle cerebral artery occlusion (tMCAO) model was established using classic methods [21]. Briefly, mice were anesthetized with 1% sodium pentobarbital. After skin preparation and disinfection, an incision was made in the cervical skin and the external carotid artery was separated. A silk suture was advanced from the external carotid artery into the internal carotid artery until the origin of the middle cerebral artery. During the entire experiment, murine body temperature was maintained at 37℃. All experiments were approved by and conducted in accordance with the ethical guidelines of the Shanghai University Animal Experimentation Committee and were in complete compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Multi-channel neuroelectrophysiological recording in vivo

To record neural activity, bilateral multi-channel macroelectrodes (3 channels in each hemisphere) were implanted into the hippocampal region (anterior/posterior: 2 mm; mediolateral: ±1.5 mm; dorsal/ventral: 2 mm), and local field potential signals (LFPs) were recorded using the AlphaLab SnR system (AlphaOmega), which captures the electrical activity of neuronal populations. The ES parameters were 0.15 mA, 0.06 ms, 10 Hz, 40 Hz, 130 Hz for 60 min.

Preparation of the oxygen-glucose deprivation/reperfusion (OGD/R) model

The complete medium was discarded, the cells were washed twice with PBS, and then replaced with glucose-free DMEM. The cells were placed in a hypoxic chamber, and a mixed gas of 95% N₂-5% CO₂ was introduced to maintain the oxygen concentration at 0.3%. The inlet and outlet tubes of the hypoxic chamber were clamped. BV2 cells were subjected to OGD treatment for 4 h, followed by a 2 h OGD/R treatment. ES with different parameters (10 Hz, 40 Hz, 130 Hz, stimulation intensity 0.05 mA, pulse width 0.1 ms, continuous stimulation lasting 20 min) was applied to BV2 cells after OGD/R treatment.

Pretreatment of local field potential signals

All signals were manually checked before further processing, and data containing significant electromyography, linear motion, and other artifacts were discarded. A 90 Hz low-pass filter was used to remove the high-frequency noise from the signal. Next, the signal was downsampled from 44 kHz to 1000 Hz by linear interpolation. The power frequency of the signal was filtered by the way of notch; the center frequency of the notch was 50 Hz and the stopband width was 2 Hz. Finally, the signal was high-pass filtered at 2 Hz to remove baseline drift and other low-frequency noise.

Power spectral analysis

The Welch periodogram method was used to estimate the power spectral density of LFP signals. Sliding window sampling, a window length of 2 s, an overlap of 1 s, and a Hamming window were used. The number of fast Fourier transform points was 2048. Considering the physiological characteristics of LFP signals, a filter was used to extract the 2–90 Hz frequency band for analysis. In addition, to reduce the influence of inter-sample variability, the power spectrum was normalized by PSD integration of 2–90 Hz.

Statistical analysis of different frequency bands

After completing the power spectrum analysis, different frequency bands were analyzed separately according to the needs. The frequency bands used in this study were divided into: Theta (4–8 Hz), Alpha (8–12 Hz), low Beta (12–20 Hz), high Beta (20–30 Hz), Beta (12-30 Hz), low Gamma (30–55 Hz), high Gamma (55–90 Hz), and Gamma (30–90 Hz). When comparing the power of different frequency bands, the required frequency band was selected and the power within the frequency band was integrated for further statistical comparison. The PSD of LFPs was compared using non-parametric paired Wilcoxon test.

Real-time quantitative PCR (RT-qPCR)

Total RNA was extracted from hippocampal tissue using RNA Isolater Total RNA Extraction Reagent (Vazyme, Nanjing, China) according to the manufacturer’s instructions and subsequently reverse-transcribed to cDNA. The fluorescence RT-qPCR method was used to amplify the target genes iNos, Arg1, Thbs-1, and Ccn1 with a SYBR^® Green qPCR kit (Yeasen, Shanghai, China). β-actin was used as an internal reference. The total volume of the PCR reaction was 20 µL, which included 2 µL cDNA, 1 µL forward and reverse primers, 10 µL SYBR^® Green qPCR reagent, and 6 µL ddH₂O. The reaction conditions were as follows: an initial preheating at 95 °C for 1 min, followed by 40 cycles of denaturation at 95 °C for 20 s, and annealing and elongation at 60 °C for 30 s. The relative expression of each target gene was calculated using the 2^-ΔΔCT method.

Immunofluorescence

Brain tissue sections were dried at 37℃ for 10 min and washed three times with PBS for 5 min each. Subsequently, sections were immersed in 0.6% immunostaining permeable solution containing Triton™ X-100 for 30 min and then blocked with 5% FBS in PBS at room temperature for 2 h. Sections were incubated overnight at 4℃ with primary antibodies against Iba1 (1:1,000; Servicebio; GB111205), JMJD3 (1:100; Abclonal; A19776), H3K27me (1:100; Abclonal; A17424). The following day, sections were rinsed three times in PBS for 10 min each and then incubated with goat anti-rabbit IgG H&L (ab150077, Abcam, UK) at 37℃ for 2 h. Sections were washed again three times with PBS for 10 min each and then incubated with DAPI solution at room temperature for a further 20 min. Sections were rinsed again three times with PBS for 10 min and mounted on a coverslip with an anti-fluorescence quencher. Sections were analyzed by fluorescence microscopy (Zeiss, Germany).

Nissl staining

The sections were removed from − 20 °C and allowed to thaw at room temperature. The sections were fixed in 4% paraformaldehyde for at least 10 min at room temperature, washed with water for 2 min, stained with Nissl solution for 5 min, and then washed again with water until the water ran clear. The sections were differentiated with 0.1% glacial acetic acid; appropriate differentiation was determined according to the color of the tissue. Brain tissue differentiated to dark blue Nissl bodies with a light blue or colorless background. Excess water was removed and the sections were placed in a 65 °C oven to dry. After thorough drying, the sections were sealed with neutral gum and observed under a microscope.

The open-field test (OFT)

The OFT was carried out using an open-field box (40 cm × 40 cm), which was divided into a central area and a peripheral area. To facilitate accurate identification of the animals by the software, attention was paid to maintaining uniform and stable light in all areas of the open-field box throughout the experiment. Smart software was used to assess the time spent in the central zone as well as the total distance traveled for 10 min in the open field. The distance that each mouse moved was referred to as the exercise ability, and the time spent in the central zone was used to assess anxiety-like behavior. Between tests, the apparatus was cleaned with 75% ethanol to mask the scent of previously tested animals.

Identification of differentially expressed genes (DEGs)

Our sequencing dataset includes samples divided into a non-stimulated control group and experimental groups stimulated at 10 Hz, 40 Hz, and 130 Hz. Using the DEseq2 package (version 1.30.1) in R (version 4.0.2), we defined DEGs between the control and experimental groups based on a P-value < 0.05 and |Fold change| >1.2. Additionally, we performed hierarchical clustering analysis on the DEGs using the ComplexHeatmap package (version 2.6.2) in R (version 4.0.2) based on the centered Pearson correlation algorithm.

Functional enrichment analysis

Using the ClusterProfiler package (version 3.18.1) in R (version 4.0.2), we performed gene ontology (GO) analysis for biological processes (BP)/cellular components (CC)/molecular functions (MF) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis on the DEGs identified between the control group and the three experimental groups. P-adjust < 0.05 was considered statistically significant as the threshold for enriched pathways. Additionally, we used the GO plot package (version 1.0.2) to visualize the selected GO terms and corresponding DEGs using GO chord diagrams.

Protein–protein interaction (PPI) network analysis

According to the DEGs identified in the key GO chord comparisons of the three experimental groups, we used the STRING database (https://string-db.org/cgi/input.pl) to analyze the interaction relationships among the encoded proteins, with a confidence score of 0.4. The results were imported into Cytoscape (https://www.cytoscape.org/) to generate a protein–protein interaction network.

scRNA-seq data processing

Mouse hippocampal tissue samples were sent to Singleron Biotechnologies (Nanjing, China) for single-cell RNA sequencing. Within 30 min after surgery, freshly harvested tissues were placed in ice-chilled sCelLive™ Tissue Preservation Solution (Singleron Bio Com, Nanjing, China) for storage. Subsequently, the tissue specimens were rinsed three times using HBSS. For tissue dissociation, 2 mL of sCelLive™ Tissue Dissociation Solution (Singleron) was added to the specimens, and the dissociation process was carried out at 37 °C for 15 min using the Singleron PythoN™ Automated Tissue Dissociation System (Singleron). Following dissociation, 2 mL of GEXSCOPE^® red blood cell (RBC) lysis buffer (Singleron) was introduced into the mixture, and the cells were incubated at 25 °C for an additional 10 min to eliminate RBCs. The resulting cell suspension was then centrifuged at 500 × g for 5 min, and the cell pellet was gently resuspended in PBS. Finally, the cell samples were stained with trypan blue (Sigma, United States), and cell viability was assessed through microscopic observation. Single-cell suspensions were added to the microfluidic device, and using the Singleron GEXSCOPE^® protocol, single-cell RNA sequencing libraries were constructed with the GEXSCOPE^® Single-cell RNA Library Kit (Singleron Biotechnologies). The single-cell RNA library was diluted to 4 nM, and Illumina HiSeq X was used for sequencing [22]. For quality control and preprocessing of single-cell RNA sequencing data, we employed the Seurat R package (version 4.0) [23]. Initially, we used Seurat to conduct a comprehensive quality assessment and preprocessing to identify and remove low-quality cells, which were determined based on factors like low gene detection numbers, high mitochondrial gene expression proportions, or low library sizes, aiming to minimize noise and bias in subsequent analyses. After eliminating these low-quality cells, we normalized and merged the data from all samples into a unified dataset. Specifically, we utilized the NormalizeData function in Seurat to scale gene expression by total cellular expression, enabling meaningful cross-cell gene expression comparisons. The normalized counts were then multiplied by a scale factor of 10,000 and log-transformed to stabilize variance and enhance the interpretability of results for data visualization and further analysis. Following this, we annotated cell types using the SingleR package (version 1.4.1) with reference transcriptomic datasets of pure cell types for accurate classification [24]. Cluster annotations were validated against manually curated gene markers from the CellMarker database to ensure reliability [25], and we identified DEGs in each cell type with thresholds of expression in at least 20% of cells in either sample, |log2FoldChange| >0.585, adjusted p value < 0.01 to pinpoint significantly differentially expressed genes.

Results

Impaired neuronal activity in mice with ischemic stroke

Mouse model with left hemisphere IS was established using the transient middle cerebral artery occlusion method. Subsequently, multi-channel electrodes were implanted into the hippocampus of both hemispheres to deliver DBS. Behavioral tests and bulk RNA-seq were used to evaluate the effects of Hip-DBS on IS (Fig. 1A). To evaluate the impairment of neural activity caused by IS, LFPs were recorded by multi-channel electrodes in hippocampus (Fig. 1B). Power spectrum analysis represented that stroke-affected (left) exhibited a significantly decreases in LFPs power across the entire frequency range (2–90 Hz) compared to the non-stroke (right), which served as the control (Fig. 1C, D). This reduction in power indicated that neural activity had indeed been compromised by IS. This finding suggests a significant reduction in the bioelectrical signals involved in information transmission between neurons on the injured side of the brain in mice with stroke, indicating pronounced neural impairment.

Fig. 1.

Impairment of neural activity in mouse model with IS.(A) Schematic diagram of experiment process. (B) Raw LFPs in hippocampus of both hemispheres, L: left side, R: right side. (C) Power spectrum of LFPs in hippocampus of both hemispheres, data are represented as mean ± SEM. (D) Quantifications of power of each frequency band, compared stroke side with control side, data are represented as boxplots. *P < 0.05, **P < 0.01, ***P < 0.001, two tailed independent Wilcoxon test. n = 15

Hip-DBS ameliorates motor impairments and attenuates anxiety and depressive behaviors in mice with IS

To assess the impact of DBS at different frequencies (two weeks post-stroke) on stroke-affected mice, we conducted OFT to evaluate their spontaneous activities. Our data revealed that in comparison with the IS group, mice subjected to DBS exhibited significantly increased total travel distances in the open field (Fig. 2A-C), indicating enhanced autonomous motor capability. Moreover, DBS treatment notably increased the percentage of time spent by mice in the center zone (Fig. 2D-F) and the duration of center zone residence (Fig. 2G-I), implying a reduction in anxiety and depressive-like symptoms. Furthermore, it is also noteworthy that mice receiving DBS at 40 Hz performed better in terms of both total travel distance and center zone residence distance/time (%) in comparison with those receiving DBS at 10 or 130 Hz.

Fig. 2.

Hip-DBS ameliorates motor impairments and attenuates anxiety and depressive behaviors in mice with IS. (A-C) The total travel distances in the open field for mice with stroke following treatment with 10 Hz, 40 Hz, or 130 Hz DBS, respectively, in comparison with IS and WT. Notably, the 40 Hz DBS treatment group exhibited a significantly increased distance traveled in the open field. (D-I) The center zone residence distance (%) (D-F) and center zone residence time (%) (G-I) in the open field for stroke mice after DBS treatment at 10 Hz, 40 Hz, and 130 Hz, respectively, in comparison with IS. Mice tended to engage in movement within the center zone and spent more time in the center zone following DBS treatment in comparison with the IS. Data were analyzed using one-way ANOVA. Results are expressed as the mean ± SEM; n = 6; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns, no significant difference

Significant transcriptome alterations are evident in mice with stroke following Hip-DBS treatment

To further investigate the effects of Hip-DBS on IS, we conducted transcriptome sequencing of the hippocampus in the IS and the 10 Hz, 40 Hz, and 130 Hz stimulated groups. Volcano plots and heatmaps presenting the DEGs among the different groups (Fig. 3A-F) highlight the top 10 genes based on their respective P values. Consistent with our experimental results, the expression levels of c-fos (Fos), Nestin (Nes), and Arg1, which are linked to neurogenesis, were upregulated in the 10 Hz, 40 Hz, and 130 Hz DBS groups compared to the IS group, with all three markers showing significant differences. Additionally, at 40 Hz DBS group, upregulation of Vegfb and Prkn, which are involved in neurodevelopment, was observed. However, 130 Hz DBS led to a more marked differential expression of genes associated with neurogenesis. It also caused the upregulation of inflammatory-related genes such as Il1b and Nlrp3, along with the downregulation of the anti-inflammatory gene Il-4. Through GO analysis of different groups, it was found that 130 Hz was more enriched in pro-inflammatory related pathways, while 40 Hz was enriched in anti-inflammatory related pathways, further indicating that 40 Hz had a better effect (Fig. 3G-I).

Fig. 3.

Transcriptome changes in mice with stroke at different frequencies of DBS. (A-C) Volcano plots showing the differential gene expression at 10 Hz, 40 Hz, and 130 Hz DBS in comparison with the IS group. Downregulated genes are shown in blue, upregulated genes are shown in red. The top 10 DEGs and genes of interest are labeled according to their respective P values. DEGs were defined as those with P < 0.05 and |Fold change| >1.2. (D-F) Heatmaps displaying the differential gene expression in each stimulated group. (G-I) GO enrichment analysis of differential genes at 10 Hz, 40 Hz and 130 Hz

Identification of DEGs and functional enrichment analysis reveals potential pathways for Hip-DBS treatment of stroke

The differentially expressed genes among the three stimulation groups were visualized using a Venn diagram, revealing a significantly greater number of such genes in the 130 Hz DBS treatment group (Fig. 4A). A total of 348 differential genes were selected for protein interaction network analysis, GO analysis and KEGG analysis (Fig. 4B-D), which included Fos, Nes and Arg1. The analysis suggested that these genes were highly interconnected with other proteins, and their interacting proteins formed clusters, indicating that numerous proteins shared similar functions and might be key factors influencing the overall lineage metabolism or signal-transduction pathway. To further explore the functional differences under different frequency ES at the molecular level, KEGG analysis (Fig. S1A-C) suggested that metabolic pathway changes in the stroke-affected brain were less obvious with 10 Hz DBS, but more pronounced with 130 Hz DBS. This indicates that 130 Hz may trigger more complex metabolic mechanisms with the brain system. GO enrichment analysis (Fig. S1D-F) results suggested that the primary processes under 10 Hz DBS were injury response, axon generation and glial development. Under 40 Hz DBS, the processes were predominantly associated with gliogenesis, autophagy, negative regulation of neuronal death and positive regulation of endothelial cell proliferation. At 130 Hz DBS, neuronal apoptosis, lymphocyte proliferation and inflammatory response regulation were observed. In the GO string diagram of the 10 Hz DBS group, Arg1 primarily performed functions related to collagen biosynthesis and organic acid transport functions, and the PPI analysis an interaction between Arg1 and Fn1, a key protein involved in wound healing (Fig. S1G-I). In the 40 Hz group, Arg1 mainly regulated the active proliferation of endothelial cells and was involved in functions related to collagen metabolism. It interacted with Fn1, Akt1, and Stat3 protein, which are known to regulate the differentiation of neural progenitors. In the 130 Hz group, although we found that Arg1 regulated epithelial cell proliferation, unlike in the 10 Hz and 40 Hz groups, Arg1 also negatively regulated leukocyte proliferation and response to external stimulation. The PPI analysis supported that the proteins interacting with Arg1 were encoded genes associated with the inflammatory response, including the pro-inflammatory factors such as Il1b, Rela, and Nlrp3. Additionally, there was a downregulation of the inflammatory inhibitor Il-4. These genes were only differentially expressed under 130 Hz stimulation. The results implied that under 10 Hz and 40 Hz DBS, the differentially expressed genes primarily function in neural development, glial cell differentiation, negative regulation of neuronal death, apoptosis and autophagy for neural repair in mice with, for stroke-induced damaged. However, at 130 Hz DBS, while there were also signs of damage response, neuronal proliferation development and axonal regeneration of nerve repair, there was concurrently an upregulation of genes promoting inflammation and an enrichment of inflammatory response.

Fig. 4.

Identification of DEGs and functional enrichment analysis reveals potential pathways for Hip-DBS treatment of stroke. (A) Venn diagram showing the intersection of DEGs between the control and the 10 Hz, 40 Hz, and 130 Hz stimulated groups. (B, D) Functional and KEGG pathway enrichment analyses of the shared set of 348 DEGs. (C) PPI network of the 348 shared DEGs. Key differentially encoded proteins are highlighted in yellow, while proteins with direct interactions are shown in purple. The strength of the interaction is represented by the thickness and darkness of the connecting lines, with thicker, darker lines indicating stronger interactions. Medium confidence level = 0.4

Therefore, we believe that the high frequency of the 130 Hz stimulation may bring secondary damage to mice, inducing an inflammatory reaction, which greatly reduces the efficiency of ES in repairing damage in stroke mice. In contrast, low frequency electrical stimulation, particularly 40 Hz DBS may yield better treatment effect, inspiring us to further explore other related mechanisms during DBS-mediated brain damage repair.

Changes in single-cell transcriptome levels in stroke mice following DBS treatment

To further explore the effect of DBS on cell type-specific gene expression in the brain of mice with IS, we performed single-cell sequencing on the hippocampal tissue of both WT and 40 Hz DBS group mice, and subsequently analyzed the data. After filtering out low quality cells, we obtained a total of 37,531 cells and identified 22 clusters (Fig. 5A). Based on a review of the literature, the cell types of different clusters were annotated (Fig. 5B), identifying excitatory neurons, inhibitory neurons, microglia, astrocytes, oligodendrocytes, oligodendrocyte precursor cells, and stromal cells. Furthermore, we calculated the proportion of each cell type in each sample (Fig. 5C). The results showed that the DBS treatment group altered the proportion of cell subsets in the brains of IS mice, with varying trends across different cell types. The proportion of microglia and astrocytes increased significantly after stimulation. Most current studies focus on the relationship between astrocytes and stroke, with microglia receiving comparatively less attention. Consequently, we directed our focus towards them. Subsequently, analysis of microglia (Fig. 5D-F) showed that subsets 2 and 3 emerged post-stroke and returned to WT level after DBS treatment, suggesting that these subsets were pathological microglia (Fig. S2C-D). Subsets 0,4 and 5 were only present in the stroke and DBS groups (Fig. S2A-B, E-F), and subset 5 increased after DBS, suggesting that this subset played a role in post-stroke treatment. Subset 1 significantly decreased in both the stroke and DBS groups, suggesting that this subgroup represented steady-state microglia. Next, we combined the top five genes from each subset with preliminary experimental data and found an interaction between the THBS-1 gene and CCN1 (Fig. 5G), and THBS-1 was significantly increased in the DBS group (Fig. 5H-I).

Fig. 5.

Landscape analysis of 37,531 single cells from WT, IS, and IS with electric stimulation in mice. (A) UMAP plot depicting 37,531 cells clustered into 22 cell clusters. (B) UMAP plots illustrating expression profiles across 7 cell types. (C) Bar graph showing the distribution of identified cell types across WT, IS, and IS + Stim (STIM) in mice. (D) Bar graph representing the composition of cell across different disease states. (E) UMAP plot illustrating the distribution of microglial subpopulations. (F) Heatmap displaying the relative expression of top marker genes for each microglial subtype. (G) The PPI network of DEGs shown among the top 5 marker genes. (H-I) The violin plots and PCR results respectively show the expression of THBS-1 in mice (WT, IS, IS + Stim) and cells (OGD/R, OGD/R + Stim)

Hip-DBS facilitates the transition of microglia toward an anti-inflammatory phenotype

The DEG analysis suggested that Hip-DBS has a positive effect on nerve cell repair in the hippocampus of stroke mice. Nissl staining analysis implied that, in comparison to stroke mice not receiving DBS treatment, the hippocampal cells of stroke mice treated with DBS exhibited a more orderly arrangement, along with a significant increase in the number of neurons (Fig. 6A). In order to study the effect of DBS on cell viability after stroke, CCK-8 results showed that microglia viability after DBS treatment was significantly higher than that in the OGD/R group. This suggests that DBS treatment promotes cell proliferation without causing cellular damage (Fig. 6B). We hypothesized that microglia might play a role in the repair mechanism of DBS-induced brain damage in stroke mice. Therefore, we proceeded to examine microglial responses after DBS treatment by staining brain slices with IBA1 and TUNEL (Fig. 6C). The results showed that, compared with the stroke group, mice treated with 40 Hz DBS exhibited reduced apoptosis. This implies that DBS treatment significantly diminishes the stroke-induced hippocampal cell apoptosis and that different frequencies of DBS treatment notably improve neurodegeneration in stroke mice.

Fig. 6.

Hip-DBS facilitates the transition of microglia toward to an anti-inflammatory phenotype. (A) Nissl staining of representative sections of the hippocampus from IS and 10 Hz, 40 Hz, 130 Hz DBS group, and DBS treatment increased the number of Nissl-positive neurons in stroke mice. (B) BV2 cells were subjected to OGD treatment for 4 h, followed by a 2 h OGD/R treatment. Cell viability curves of BV2 after OGD/R and OGD/R + Stim. (C) Representative micrographs of IBA1 (red), TUNEL (green), DAPI (blue) immunofluorescence in the hippocampus of IS and 10 Hz, 40 Hz Stim group, with the insets showing that apoptosis-positive cells were also significantly reduced in the Stim group. (D) mRNA expression changes of proinflammatory factors in BV2 OGD/R and OGD/R + Stim. (E) mRNA expression changes of inflammatory suppressors in BV2 OGD/R and OGD/R + Stim groups. (F-G) Western blot of iNOS and ARG1 in BV2 OGD/R and OGD/R + Stim. (H) mRNA expression levels of iNos and Arg1 in mouse brain tissues of the IS and 40 Hz Stim groups. Data were analyzed using a Student’s t-test and are expressed as the mean ± SEM; n ≥ 3; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001

Studies have shown that IS can trigger a strong inflammatory response at the site of cerebral infarction, and potentially throughout the entire brain [26]. Accordingly, the result showed that the levels of Il6, Tnf-α, Il12 and Nlrp3 were decreased significantly after DBS treatment (Fig. 6D), while those of Il-10 and Il-4 were observed to increase (Fig. 6E). M1 and M2 subtypes play different functions, so the levels of iNos and Arg1 in M1 and M2 microglia were examined at protein and RNA levels, respectively, to assess the involvement of M1 and M2 microglia in inflammation of the nervous system [27]. The results showed that iNOS did not show a significant difference between the two data sets, however, the level of ARG1 increased notably (Fig. 6F-H) (Fig. 7A), which was consistent with the RNA-seq data, and marked the increase of M2 type microglia, suggesting the enhancement of the immune modulatory capacity and tissue repair capacity of microglia. Consequently, we concluded that DBS may mediated the modulation of stroke induced brain damage by activating the brain’s autoimmune response and improving the neural microenvironment through microglial phenotypic transition.

Fig. 7.

DBS treatment regulates the expression of JMJD3 through the CCN1/PI3K/AKT pathway, causing a transition of microglia to the M2 isoform. (A) mRNA expression of Arg1, inos in the OGD/R, and OGD/R + Stim. (B) Protein expression of CCN1 in BV2. (C) mRNA expression of CCN1 in BV2 of OGD/R, and OGD/R + Stim groups. (D) Quantification of protein expression of the PI3K/AKT pathway in mouse brain lysates from IS and IS + Stim, and the mRNA expression levels of JMJD3. (E) Immunofluorescence staining of JMJD3 (red), H3K27Me (red), and DAPI (blue) in BV2 of OGD/R, and OGD/R + Stim groups. (F) Immunoprecipitation was performed with anti-JMJD3 antibody and ChIP DNA levels were analyzed using semiquantitative PCR and normalized to starting DNA levels. Data were analyzed using a Student’s t-test and are expressed as the mean ± SEM; n ≥ 3; *P < 0.05, **P < 0.01, ****P < 0.0001; ns indicates no significant difference

DBS treatment regulates the expression of JMJD3 through the CCN1/PI3K/AKT pathway, causing a transition of microglia to the M2 isoform

To explore the mechanism underlying the subtype transition of microglia, we further analyzed the top10 differential genes based on the RNA-seq data. Cysteine-rich protein 61 (CYR 61) is a tissue growth factor belonging to the CCN family (CYR61/CTGF/NOV) that regulates cell proliferation, inflammation, migration, embryogenesis, and wound healing [28]. In the context of IS, the results showed that DBS treatment could significantly increase the mRNA and protein expression level of CCN1 (Fig. 7B-C), which may be related to the anti-inflammatory response of microglia. It was founded that CCN1 induces IL-6 production via the α6β1/PI3K/AKT/NF-κB signaling pathway [29], and the carboplatin-stimulated dephosphorylation of PI3K and AKT was significantly rescued by CYR61 [28]. To further explore the effect of CCN1 on PI3K/AKT signaling, the increased phosphorylation of AKT, observed at the protein level (Fig. 7D) showed that DBS treatment could activate this pathway, potentially enhancing the expression of JMJD3 [30]. We next studied whether epigenetic mechanisms were involved in M2 microglia polarization. Given the demethylation of histones was discovered in recent years and its emergences as a critical factor in various human diseases, we focused on the histone demethylases that may regulate this process. We examined the levels of H3K27me3 methylase and demethylase JMJD3, and the results of western blot, qPCR and immunofluorescence analysis showed decreased levels of H3K27me3 methylase and increased levels of JMJD3 (Fig. 7D-E). It was found that M2 markers such as ARG1 and CD206 also significantly decreased when they were knocked down [31]. To further investigate the role of JMJD3 in the M2 polarization in microglia, we performed the chromatin immunoprecipitation (ChIP) experiments, to assess the recruitment of H3K27me3 and JMJD3 to the ARG1 promoter. After immunoprecipitation by using the anti-H3K27me3 and anti-JMJD3 antibodies, the results showed that JMJD3 was significantly more enriched under stim conditions than under OGD/R conditions, this may imply an increased binding of JMJD3 to the ARG1 promoter under stimulation conditions (Fig. 7F). In conclusion, DBS treatment may regulate JMJD3 expression by potentially activating the CCN1/PI3K/AKT pathway, could promote ARG1 demethylation, and may subsequently facilitate the transition of microglia to the M2 anti-inflammatory subtype. This process may offer new strategies for the treatment of stroke.

Discussion

Over the past three decades, DBS has emerged as a mainstream surgical intervention for movement disorders such as PD and essential tremor [32, 33]. Its application has also been increasingly extended to a diverse range of motor, mood, and cognitive disorders [33]. Despite its demonstrated efficacy in certain neurological conditions, the potential of DBS for stroke treatment remains under exploration [34]. Employing an IS mouse model, we conducted an investigation into the effects of Hip-DBS. Neurophysiological measurements revealed a reduction in neural activity within the ischemic hemisphere. By applying Hip-DBS at frequencies of 10 Hz, 40 Hz, and 130 Hz, we determined that 40 Hz DBS was more effective than 130 Hz DBS in enhancing mobility and reducing anxiety-like and depressive-like symptoms in stroke-injured mice. Our results demonstrate that Hip-DBS can ameliorate motor impairments in mice with stroke, particularly at 40 Hz stimulation, and may improve brain injury symptoms by stimulating neurorepair and regeneration and activating the immunomodulatory capacity of microglial cells. These findings demonstrate the feasibility of neural ES in improving IS outcomes and provide a preclinical mechanistic basis for the translational application of 40 Hz stimulation in stroke rehabilitation.

Following a stroke, patients experience functional impairments as a consequence of neuronal death and disruption of neural circuits [35]. Neural repair and regeneration are of critical importance, and a variety of preclinical strategies, such as stem cell transplantation and targeted drug therapies, have been explored [36–39]. Our bioinformatics analysis demonstrated that Hip-DBS upregulated genes associated with neurorepair and neuroregeneration. Microglia, which play a crucial role in neurogenesis and immune modulation following stroke, were impacted by DBS. Research has shown that depletion of microglia in the mouse brain using the colony-stimulating factor 1 receptor (CSF1R) antagonist PLX3397 resulted in an approximately 60% increase in infarct size after MCAO [40], indicating that microglia play a significant role in neuroprotection. During ischemic injury, microglia assume an M1 phenotype and secrete various pro-inflammatory molecules. Conversely, microglia possessing the M2 phenotype are primarily responsible for clearing cellular debris, facilitating the resolution of post-ischemic inflammation, and releasing abundant growth factors related to brain repair [27, 41]. Our results are consistent with these findings. We propose that the mechanisms underlying DBS-induced functional recovery in stroke-affected brains may be multifaceted, although further validation with larger sample sizes is needed to substantiate these observations.

Utilizing transcriptome sequencing data, we conducted further analysis of the differential molecular mechanisms between different frequencies of DBS. The results imply that 10 Hz and 40 Hz DBS tend to promote brain neurodevelopment, negative regulation of neuronal death, apoptosis and other neurorepair-related functions. In contrast, 130 Hz DBS leads to the enrichment of pro-inflammatory genes and activation of inflammatory response pathways. This suggests that different parameters of DBS may result in highly divergent effects, highlighting the need to explore appropriate parameters and consider the potential for side-effects when applying DBS in clinical treatment. To further explore the changes occurring in the brain tissue after DBS in microglia, single-cell sequencing analysis was conducted. The results showed that the DBS treatment group changed the composition of cell subsets in the brain of stroke mice. The trend of these changes varied among different cell types, with a notable increase in the proportion of microglia and astrocytes post-stimulation. This may be related to the immune response of microglia and the energy metabolism of astrocytes.

Despite numerous studies on the mechanisms of post-stroke neuronal repair, effective therapeutic strategies remain elusive [42]. According to Figs. 3, 4 and 5, in addition to the CCN1-AKT-JMJD3-ARG1 pathway, the apoptotic signaling pathway and mTOR signaling pathway may also be involved in the IS process. Among these, the high expression of the Clca3a1 in the 40 Hz group might enhance immune function by inducing the activation of TAZ, a co-activator of the Hippo pathway. Further studies are required to fully elucidate the underlying mechanisms.

Certainly, this study has certain limitations. As an immortalized microglial cell line, BV2 cells exhibit distinct expression profiles for certain genes compared to primary microglia. Furthermore, the in vitro single-cell culture environment cannot recapitulate the complex cellular interactions present in the hippocampal microenvironment in vivo. Nevertheless, leveraging its advantages of ease of culture, stable cellular state, and ability to mimic responses associated with central nervous system injury, our results still provided a reliable in vitro basis for validating Hip-DBS regulated microglial transcriptomic changes. These findings lay a critical foundation for subsequent in-depth investigations. Future studies will incorporate primary microglial culture systems, thereby more comprehensively and accurately deciphering the molecular mechanisms underlying DBS in stroke treatment and providing more compelling experimental evidence for clinical translation.

Conclusions

Our study suggests that Hip-DBS improves stroke outcomes and promotes neuronal repair by regulating microglial phenotype transition via the CCN1/PI3K/AKT signaling pathway. Mechanistically, this intervention drives a shift from pro-inflammatory to anti-inflammatory microglial states, evidenced by decreased pro-inflammatory cytokines and increased anti-inflammatory cytokines, ultimately restoring neurofunctional deficits associated with IS. These findings highlight the substantial therapeutic potential of Hip-DBS for stroke injury and provide mechanistic insights into its efficacy, establishing a critical foundation for translating Hip-DBS into clinical post-stroke rehabilitation. As the reparative and immunomodulatory effects of DBS on ischemic brain injury are further elucidated, future research may focus on optimizing stimulation parameters or developing combinatorial approaches to more efficiently activate the brain’s endogenous repair mechanisms. Notably, current literature reveals high heterogeneity in DBS stimulation parameters across studies, with distinct frequencies often yielding non-overlapping effects. This suggests that multi-frequency alternating periodic therapies could be a promising direction for maximizing therapeutic consistency. In clinical settings, integrating Hip-DBS into comprehensive rehabilitation plans-including traditional physical therapy and pharmacological interventions-may synergize to achieve optimal functional recovery.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Abbreviations

DBS

Deep brain stimulation

IS

Ischemic stroke

Hip-DBS

Hippocampal DBS

LFP

Local field potentials

HS

Hemorrhagic stroke

ES

Electrical stimulation

PSD

Post-stroke depression

PD

Parkinson’s disease

PLIC

Posterior limb of the internal capsule

PVG/PAG

Periventricular/periaqueductal grey

OFT

Open field tests

GO

Gene ontology

DEGs

Differentially expressed genes

PPI

Protein-protein interaction network

CYR 61

Cysteine-rich 61

ChIP

Chromatin immunoprecipitation

Author contributions

J.W. and X.L. conceptualized and supervised experiments. X.Z. and Y.C. designed and performed the experiments, analyzed data, and wrote manuscript. P.W. and J.Z. analyzed data and edited manuscript. H.H. and Q.Z. performed the experiments. J.W., J.H. and R.M. polished manuscript. J.W. was responsible for funding acquisition, supervision, validation, and manuscript revision. All authors read and agreed to the published version of the manuscript.

Funding

This work was sponsored by The National Key Research and Development Program of China (2020YFA0113000), Basic Research Program of Shanghai (25J22800700, 24JS2810400, 20JC1412200), the Shandong Provincial Natural Science Foundation (Grant No. ZR2021MH304), Guangxi Natural Science Foundation (Grant No. 2025GXNSFHA069271).

Data availability

All data needed to evaluate the conclusions in the paper are present in the paper. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All experiments were approved by and conducted in accordance with the ethical guidelines of the Shanghai University Animal Experimentation Committee and were in complete compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Consent to participate

Not applicable.

Competing interests

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.