Amantadine modulates novel macrophage phenotypes to enhance neural repair following spinal cord injury

Abstract

Background

Spinal cord injury (SCI) triggers a complex inflammatory response that impedes neural repair and functional recovery. The modulation of macrophage phenotypes is thus considered a promising therapeutic strategy to mitigate inflammation and promote regeneration.

Methods

We employed microarray and single-cell RNA sequencing (scRNA-seq) to investigate gene expression changes and immune cell dynamics in mice following crush injury at 3 and 7 days post-injury (dpi). High-dimensional gene co-expression network analysis (hdWGCNA) and slingshot trajectory analysis were employed to identify key gene modules and macrophage differentiation pathways. Subsequently, immunofluorescence staining, flow cytometry, and western blotting were performed to validate the identified effects of amantadine on macrophage differentiation and inflammation.

Results

To elucidate the molecular mechanisms underlying the injury response at the transcriptional level, we performed a microarray analysis followed by gene set enrichment analysis (GSEA). The results revealed that pathways related to phagocytosis and macrophage activation are significantly involved post-injury, shedding light on the regulatory role of macrophages in SCI repair. To further investigate macrophage dynamics within the injured spinal cord, we conducted scRNA-Seq, identifying three distinct macrophage subtypes: border-associated macrophages (BAMs), inflammatory macrophages (IMs), and chemotaxis-inducing macrophages (CIMs). Trajectory analysis suggested a differentiation pathway from Il-1b⁺ IMs to Mrc1⁺ BAMs, and subsequently to Arg1⁺ CIMs, indicating a potential maturation process. Given the importance of these pathways in the injury response, we utilized molecular docking to hypothesize that amantadine might modulate this process. Subsequent in vitro and in vivo experiments demonstrated that amantadine reduces Il-1b⁺ IMs and facilitates the transition to Mrc1⁺ BAMs and Arg1⁺ CIMs, likely through modulation of the HIF-1α and NF-κB pathways. This modulation promotes neural regeneration and enhances functional recovery following SCI.

Conclusions

Amantadine modulates macrophage phenotypes following SCI, reduces early inflammatory responses, and enhances neural function recovery. These findings highlight the therapeutic potential of amantadine as a treatment for SCI, and provide a foundation for future translational research into its clinical applications.

Background

Spinal cord injury (SCI) is a severe condition with an estimated global incidence of 250,000–500,000 new cases annually, primarily caused by traumatic events such as motor vehicle accidents, falls, and sports injuries [1, 2]. SCI commonly results in permanent neurological deficits and significant complications, including infections and deep vein thrombosis [3]. Current treatments such as surgical decompression, corticosteroids, and rehabilitation have limited recovery promotion effects [4, 5]. The underlying pathophysiology involves primary mechanical trauma followed by secondary injury processes such as ischemia, inflammation, and apoptosis, which worsens the initial damage [6]. Multiple studies have identified neuroinflammation as a key contributor to secondary injury following spinal cord injury (SCI), suggesting that mitigating neuroinflammation could enhance neural repair and functional recovery, thereby offering a promising therapeutic approach for SCI [7, 8].

SCI unfolds in three distinct phases—acute, subacute, and chronic—each marked by specific pathological changes: the acute phase involves immediate tissue damage, inflammation, and the onset of secondary injury; the subacute phase is characterized by extended inflammation, apoptosis, and glial scar formation; and the chronic phase features inhibited nerve regeneration, ongoing inflammation, and persistent functional deficits [9, 10]. In the subacute phase of SCI(48 h to 2 weeks), pronounced neuroinflammation is mainly regulated by macrophages [11]. Macrophages are particularly critical and traditionally classified into M1 (pro-inflammatory) and M2 (anti-inflammatory) phenotypes [12, 13]. However, this classification system does not fully capture the complexity of macrophage changes after post-SCI. Prior studies utilizing single-cell RNA sequencing (scRNA-seq) have identified novel macrophage subtypes, including chemotaxis-inducing, inflammatory, and border-associated macrophages, each displaying unique gene expression profiles and functions [14]. In addition, recent studies have identified Fabp5⁺ macrophage subsets and thrombospondin-sensing macrophage subsets [15, 16]. Despite these findings, the roles of these subtypes in SCI repair and their therapeutic potential remain unclear.

Several promising strategies have been proposed to regulate macrophage phenotypes to mitigate neuroinflammation and promote SCI repair involves [17]. Pharmacological agents, such as phosphodiesterase inhibitors (e.g., rolipram) as well as other drugs, such as minocycline and statins, have shown potential for modulating macrophage activity to improve SCI outcomes [18,19,20]. In this study, we applied bulk transcriptomics and scRNA-seq in mice following crush injury at 3 and 7 days post-injury (dpi) with pseudotime analysis, high-dimensional gene co-expression network analysis (hdWGCNA), and molecular docking to identify therapeutic candidates. Amantadine has been shown to promote neural repair by modulating neurotransmitter release and reducing neuroinflammation, with a well-established safety profile in both animal and clinical studies [21, 22]. In our research, we found that amantadine is a promising drug capable of shifting macrophages from the inflammatory (IMs) subtype to the border-associated (BAMs) and chemotaxis-inducing (CIMs) subtypes. As confirmed by in vivo and in vitro experiments, this transition occurs via the HIF-1α and NF-κB pathways and helps reduce neuroinflammation and enhance neural recovery in SCI. The workflow of this study is illustrated in Scheme 1.

Scheme 1

Schematic illustration of Amantadine and its therapeutic role in spinal cord injury. Step One: Hub gene identification through molecular docking. Genes such as CFH, LY86, CSF1R, HEXB, CX3CR1, and FCRLS are found to be associated with spinal cord injury. Amantadine is subsequently identified as a candidate compound for treatment. Step Two: Compression injury is induced in C57 mice at 3dpi and 7dpi. Step Three: Macrophage subtypes are identified, including inflammatory macrophages (IMs), border-associated macrophages (BAMs), and chemotaxis-inducing macrophages (CIMs). Detection markers like Mrc1, Ccl12, Il-1b, Cd81, Arg1, and Spp1 are used for macrophage characterization. Step Four: Amantadine modulates macrophage activity through the Notch signaling pathway, affecting Hif-1α and p-NF-κB. This promotes the recovery process by regulating anti-inflammatory factors and preventing neuron apoptosis, ultimately aiding spinal cord repair

Methods

Microarray and scRNA-seq datasets

In this study, we downloaded microarray data from spinal cord tissues of young (2–3 months old) female C57BL/6 mice. The microarray analysis included data from 8 sham operation group samples and 14 subacute phase thoracic moderate contusion SCI group samples (comprising 7 samples at 3dpi and 7 samples at 7 dpi) from two microarray datasets (GSE47861, GSE5296). All datasets were sequenced using the GPL1261 platform. Additionally, single-cell RNA sequencing data were downloaded from young (2–3 months old) female C57BL/6 sham operation group and subacute phase thoracic moderate contusion SCI group (3dpi and 7dpi) samples (GSE205038), focusing on CD45⁺ immune cells following the protocol for CD45⁺ microbeads (Miltenyi Biotec) and LS columns (Miltenyi Biotec) [23]. All data were hosted on the Gene Expression Omnibus database (https://www.ncbi.nlm.nih.gov/geo/), and were not generated specifically for this publication, and we gratefully acknowledge the original authors for providing the raw data.

Microarray data Processing and differential analysis

All statistical analyses and visualizations were performed using R software version 4.2.3. To mitigate batch effects following matrix normalization, we utilized the combat function from the sva package (v3.44.0). Differentially expressed genes (DEGs) were identified using the limma package (v3.52.1), applying a selection criterion of |log₂(fold-change)| > 1.0 and a p-value < 0.05. Heatmaps were generated using the pheatmap package (v1.0.12). Gene function enrichment for Mus musculus was conducted using the clusterProfiler package, complemented by the GeneMANIA website (https://genemania.org/). Gene Set Enrichment Analysis (GSEA) was performed using the clusterProfiler package (v4.9.3), with the gseGO function applied using the org.Mm.eg.db database for Gene Ontology (GO) terms (https://www.gsea-msigdb.org/gsea/index.jsp). The analysis considered gene sets ranging from 5 to 500 genes, with a p-value cutoff of 1. The GOSemSim package was employed to conduct Friend analysis for hub gene identification (v2.27.2).

Single-cell RNA-seq data processing

We processed single-cell RNA sequencing (scRNA-seq) raw count matrix using the Seurat package (v4.3.0) [24]. Quality control criteria was retaining cells with gene counts between 300 and 10,000, mitochondrial gene content under 15%, and excluding cells expressing genes in fewer than three cells or with a red blood cell gene expression proportion over 0.1. Highly variable genes were identified with the FindVariableGenes function, followed by principal component analysis (PCA) for dimensionality reduction. Data visualization was conducted using Uniform Manifold Approximation and Projection (UMAP). To integrate data from different tissue samples and mitigate batch effects, we employed the Harmony package (v0.1.1) [25]. Differential gene expression analysis was performed using the FindAllMarkers function, with thresholds of |log (fold-change)| > 0.3 and p-value < 0.05. We visualized single-cell distributions with the DimPlot function, and gene expression patterns with the FeaturePlot function. Cell subgroups were annotated using the SingleR package (Ref: MouseRNAseqData, v1.10.0). For dot plot visualization of grouped data, the SCP package (v0.5.1) was used.

Monocle2 analysis

We utilized Monocle version 2.22.0 for trajectory analysis of single-cell RNA sequencing data. Expression matrices and cell metadata were extracted from the Seurat objects to create CellDataSet objects. The data were normalized, and size factors and dispersion parameters were estimated. Highly variable genes were selected based on a mean expression threshold of ≥ 0.1 and empirical dispersion equal to or exceeding the fitted dispersion. Dimensionality reduction was performed using the DDRTree algorithm with the maximum number of components set to two. Cells were ordered along the trajectory using the orderCells function, and the root state was specified based on prior biological knowledge of macrophage development. This approach allowed us to infer pseudotemporal ordering and identify key transitional states among macrophage populations.

Slingshot analysis

Slingshot version 1.8.0 was applied for unsupervised pseudotime trajectory inference [26]. Seurat objects were converted into SingleCellExperiment objects while retaining UMAP embeddings for consistency. The slingshot function was executed using cell type annotations (cell_type) as cluster labels and UMAP coordinates as the reduced dimensional space. Primary pseudotime lineages (slingPseudotime_1) were extracted and integrated back into the Seurat objects for downstream analyses. Visualization was performed using ggplot2, overlaying the inferred trajectories onto UMAP plots to illustrate the pseudotemporal progression of macrophage differentiation.

PAGA analysis

We conducted Partition-based Graph Abstraction (PAGA) analysis using Scanpy version 1.9.1 to construct a graph-based representation of cellular relationships. The normalized expression matrix underwent selection of highly variable genes and principal component analysis for dimensionality reduction. A k-nearest neighbor graph was computed with k set to 15 to capture local cell-cell relationships. PAGA was then utilized to assess the connectivity between clusters, generating an abstracted graph that highlights major lineage pathways. UMAP embeddings were calculated for visualization, and PAGA connectivity was overlaid onto UMAP plots to depict potential developmental trajectories among macrophage subpopulations.

High dimensional weighted gene co-expression network analysis (hdWGCNA)

We employed hdWGCNA (v0.2.1) to identify and characterize gene modules associated with macrophage subtypes in SCI (https://github.com/smorabit/hdWGCNA). Annotated macrophage data were loaded, and genes expressed in at least 5% of cells were filtered using the SetupForWGCNA function from the hdWGCNA package. To address data sparsity, we constructed metacells by MetacellsByGroups function, normalized the metacell expression matrix using the NormalizeMetacells function, and transposed the expression matrix using the SetDatExpr function. The appropriate soft-thresholding power was determined using the TestSoftPowers function, and a soft power of 8 was selected to construct the co-expression network with the ConstructNetwork function. Module detection was carried out by generating a dendrogram with the PlotDendrogram function and calculating module eigengenes (MEs) using the ModuleEigengenes and ModuleConnectivity functions. Hub genes within each module were identified using the GetHubGenes function and visualized with the PlotKMEs function. Module-trait relationships were explored through module correlations and gene scoring methods using the ModuleFeaturePlot and HubGeneNetworkPlot functions.

Molecular modeling analysis

We used the Connectivity Map (CMAP) database to predict interactions between FDA-approved compounds and hub genes of inflammatory macrophages, using a significance threshold of p < 0.05 [27]. The two-dimensional structures of the compounds were sourced from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/), while protein structures were obtained from the RCSB Protein Data Bank. AutoDock 4 software was employed to model molecular interactions between small molecules and target proteins, with visualization performed using PyMol software.

Cell culture

RAW 264.7 cells (Procell, Wuhan, China, Cat# CL-0190) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Sigma-Aldrich, St. Louis, MO, USA, Cat# D0819) supplemented with 10% fetal bovine serum (FBS) (Invitrogen, Waltham, MA, USA, Cat# 10100147) and 100 U/mL penicillin-streptomycin (Invitrogen, Cat# 15070063) at 37 °C in a 5% CO₂ atmosphere. For the control group, the growth medium was changed without any additional treatment. For the LPS + IFN-γ group, RAW 264.7 macrophages were incubated with LPS (100 ng/mL, #L3129, Sigma-Aldrich) and IFN-γ (20 ng/mL, #L17001, Sigma-Aldrich) for 24 h [28]. For the LPS + IFN-γ + amantadine group, cells were incubated with LPS, IFN-γ, and amantadine (50 µM, #768-94-5, Sigma-Aldrich) for 24 h.

Cell counting Kit-8 (CCK-8) assay

According to the manufacturer’s instructions, the cell viability and cytotoxicity of RAW 264.7 cells were assessed using the Cell Counting Kit-8 (CCK-8) assay (Cat. No. C0038, Beyotime Biotech). RAW 264.7 cells (1 × 10⁵ cells/mL) were seeded into 96-well plates. Cells were exposed to 5 µM, 10 µM, and 50 µM amantadine [29]. After 24 h of incubation, 10 µL of CCK-8 reagent was added to each well and incubated at 37 °C for another 4 h. Cell viability was measured using a microplate reader (Bio-Rad) at a wavelength of 450 nm.

Animals

Female C57BL/6 mice, aged six to eight weeks and weighing approximately 18–20 g, were obtained from the Animal Experiment Centre of The Fourth Military Medical University, China. The mice were housed in clean, warm cages at a temperature of 22–25 °C, under a 12-hour light/dark cycle, with ad libitum access to food and water. All animal protocols were reviewed and approved by the Ethics Committee of The Fourth Military Medical University. We adhered to the principles of the 3Rs (Replacement, Reduction, and Refinement) by minimizing the number of animals used, refining procedures to reduce pain and distress, and ensuring ethical compliance through institutional approval. The mice were randomly assigned to three groups, with 20 mice in each group: (1) Sham group: Mice underwent surgery without SCI; (2) SCI group: The untreated SCI mice; (3) SCI + amantadine group: Mice received SCI followed by daily intraperitoneal injections of 2 mg/kg amantadine for 7 days. These groups were further divided into different experimental conditions, with 5 mice sacrificed at each time point: 3 dpi, 7 dpi, 14 dpi, and 28 dpi. Different histological and behavioral analyses were performed at each time point, such as inflammation analysis at 3 dpi, Nissl staining and NeuN/Tunel analysis at 7 dpi, histological staining at 14 dpi, and electrophysiological analysis at 28 dpi.

Spinal cord injury model

A standardized bilateral spinal cord compression injury model was established under aseptic conditions based on previous studies [30]. Mice were anesthetized with intraperitoneal injections of sodium pentobarbital and positioned prone. The back hair was shaved, and the area was disinfected with iodophor. After disinfection, the skin, subcutaneous tissue, and paravertebral muscles were sequentially separated to expose the T9 vertebra. The T9 spinous process and vertebral plate were removed using Fiber bite forceps to expose the spinal cord without causing injury. Dumont #5 forceps (Fine Science Tools, Foster City, CA, USA) were used to compress the spinal cord for 5 s vertically [31]. Successful compression was confirmed by stretching of the lower limbs and tail twitching. Post-compression, muscles, and skin were sutured with sterile band sutures, and cefuroxime was administered intraperitoneally. Mice were allowed to recover naturally at room temperature. Post-surgery, mice in the amantadine group received daily intraperitoneal injections of 2 mg/kg amantadine for one week. Additionally, manual bladder massage was performed twice daily until reflex bladder emptying was restored.

Mice were humanely sacrificed by administering an overdose of isoflurane at 3, 7, 14, and 28 dpi). A median thoracoabdominal incision was performed, followed by perfusion with 10 mL of cold saline through the left ventricle and then 100 mL of cold 4% paraformaldehyde (Beyotime, Shanghai, China, Cat# P0099-100 ml). The spinal cord segment spanning 0.5 cm above and below the lesion site was collected and fixed in 4% paraformaldehyde overnight.

BBB score

The BBB score is an internationally recognized index for evaluating the functional recovery process following SCI in mice, with a total score ranging from 0 to 21. The BBB scores of the mice were observed and recorded weekly. In the eighth week, the movement trace of the treated group was also captured. The hindlimb movement and weight-bearing ability were recorded when mice walked on horizontal plane. The observers were blinded to the SCI mice condition, and the score evaluations were repeated five times and recorded [32]. Three or more researchers independently completed the scoring as reported in previous literature on a weekly basis for statistical analysis [33].

Electrophysiological analysis

Electrophysiological tests were performed at 28 dpi to evaluate the functional status of sensorimotor signal conduction. The sciatic nerve of the animals was exposed under anesthesia. For motor evoked potential (MEP) measurements, a stimulating electrode was inserted into the spinal motor cortex (SMC), and a recording electrode was placed into the sciatic nerve (20 mA, 0.1 ms, 1 Hz). The waveforms, amplitude, and latency of MEP were recorded and analyzed.

H&E staining and nissl staining

Spinal cord tissue samples were first fixed in 4% paraformaldehyde for 24 h at room temperature. After fixation, the tissues were dehydrated through a graded series of ethanol solutions (70%, 80%, 95%, and 100%) to remove water content. The tissues were then cleared twice in xylene for 30 min. Following clearing, the samples were infiltrated with molten paraffin at 60 °C for 1 h, repeated three times to ensure complete paraffin infiltration.

The infiltrated tissues were then embedded in fresh paraffin and allowed to solidify at room temperature. Once solidified, the paraffin-embedded tissue blocks were stored at room temperature until sectioning. Sections were cut at a thickness of 5 μm using a microtome and subsequently mounted on glass slides for further histological analysis. Sections were then treated with xylene I and II for 10 min each, followed by sequential immersion in various concentrations of alcohol solutions (100%, 95%, 90%, 80%, and 70%) for 5 min each, and rinsed in distilled water for 1 min. The sections were stained with hematoxylin (Solarbio, #G1120, Beijing, China) for 2 min, differentiated with 1% hydrochloric acid, washed in distilled water for 2 min, and dyed with 0.5% eosin (Solarbio, #G1120, Beijing, China) for 1 min. The dehydration process was completed by soaking the sections in a gradient ethanol solution (70%, 80%, 90%, and 100%). Slides were treated with xylene I and II for 5 min each and sealed with neutral resin (Solarbio, #G8590, Beijing, China). Finally, the samples were observed and photographed using a fully automated tissue in situ multi-label landscape quantification analyzer (Vectra Polaris, PerkinElmer). Nissl staining was conducted according to the manufacturer’s guidelines (Beyotime, Cat# C0117). Images were then taken using an inverted microscope to determine the number of neurons in the anterior horn of each spinal cord section.

Indirect flow cytometry (FACS)

RAW264.7 cells were filtered through a 70-µm cell strainer to obtain a single-cell suspension. The cells were incubated with anti-rabbit IL-1β antibody (Abcam, ab283818) and anti-rabbit Liver Arginase1 antibody (Abcam, ab96183) for 30 min. Cells were then permeabilized using Cyto-Fast™ Fix/Perm Buffer Set (Biolegend, #426803, USA) for 15 min. After permeabilization, cells were incubated with anti-mouse MRC1 antibody (Biolegend, #141705, USA) and FITC anti-rat F4/80 antibody (Biolegend, #201805, USA) for 30 min at room temperature in the dark. Subsequently, cells were incubated with goat anti-mouse IgG H&L (Alexa Fluor^®647/APC) (Abcam, #150115, MA, USA) and goat anti-rabbit IgG H&L (Alexa Fluor^® 594) (Abcam, #150080, MA, USA). Samples were collected using a BD LSRFortessa flow cytometer (BD Biosciences, US), and data were analyzed using FlowJo software.

Quantitative PCR

Cells were resuspended in 300 µL of RLT buffer before RNA purification using the Qiagen RNeasy kit. The purified RNA was converted into cDNA using a high-capacity cDNA reverse transcription kit (Life Technologies). Transcript levels were quantified via SYBR Green qPCR performed on an iQ Thermocycler (Bio-Rad) or ABI using Quant Studio Design v1.3 software. Primer sequences used for qPCR are listed in Supplementary Table 1.

Immunofluorescence staining

For cell immunofluorescence staining, cells were fixed, permeabilized, and blocked before being exposed to primary antibodies at 4 °C overnight, followed by secondary antibodies. For spinal cord tissue immunofluorescence staining, frozen spinal cord sections were permeabilized and blocked, followed by treatment with primary antibodies and secondary antibodies. The sections were examined using an ultra-high resolution laser confocal microscope (ZEISS LSM 900, Germany). Primary and secondary antibodies used were: Rabbit anti-IL-1β (1:100; Abcam, ab283818, MA, USA); Rabbit anti-Mannose Receptor (1:100; Abcam, #64693, MA, USA); Rat anti-F4/80 (1:400; Abcam, ab6640, MA, USA); Rabbit anti-Liver Arginase1 (1:400; Abcam, ab96183, MA, USA); Chicken anti-GFAP (1:400; Abcam, ab4674, MA, USA); Rabbit anti-MAP2 (1:400; Abcam, ab183830, MA, USA); Goat anti-rabbit IgG H&L Cy3 (1:400; Abcam, #6939, MA, USA); Goat anti-mouse IgG H&L Alexa Fluor 488 (1:400; Abcam, #150113, MA, USA); Goat anti-chicken IgY H&L Alexa Fluor 488 (1:400; Abcam, ab150173, MA, USA).

Western blot analysis

Proteins from cultured cells were extracted using RIPA lysis buffer supplemented with a protease and phosphatase inhibitor cocktail (Thermo Fisher). Protein concentration was determined using the BCA method. Equal amounts of protein (30–50 µg) from each sample were separated by SDS-PAGE (10%) and transferred to NC membranes (EMD Millipore Corp). Membranes were blocked with 5% skimmed milk powder for 1 h at room temperature and incubated overnight at 4 °C with primary antibodies: Arg-1 (ab96183, Abcam, 1:1000), β-Actin (3700 S, Cell Signaling Technology, 1:1000), HIF-1α (bs-0737R, Bioss, 1:1000), NF-κB (8242 S, Cell Signaling Technology, 1:1000), and p-NF-κB (3033 S, Cell Signaling Technology, 1:1000). After washing with TBST, membranes were incubated with corresponding secondary antibodies for 1 h. Bands were developed using western blotting-enhanced chemiluminescent solution (Millipore), and images were captured with an Amersham Imager 600 Station (GE Healthcare, Stockholm, Sweden). The greyscale value of each band was measured using ImageJ software.

Statistical analysis

Statistical analysis was performed using GraphPad Prism 9.2.0 software (GraphPad Software, San Diego, CA, USA). Data analysis between different groups was conducted using Student’s t-test and one-way analysis of variance (ANOVA), followed by Tukey’s multiple comparison post hoc test. Significance levels were defined as follows: *p < 0.05, **p < 0.01, ***p < 0.001. Each experiment was repeated at least three times, and the results were reported as the mean ± SD.

Microarray and single-cell RNA sequencing data were obtained from the GEO database. Microarray data (GSE47861, GSE5296) were sourced from young (2–3 months old) female C57BL/6 mice with SCI. All samples were collected from regions 1 cm surrounding the lesion site and included both sham operation and SCI groups. The microarray datasets did not involve specific cell type enrichment and represented whole spinal cords and adjacent areas. Single-cell RNA sequencing data (GSE205038) were also obtained from young (2–3 months old) female C57BL/6 mice with SCI. This dataset focused on CD45 + immune cells, which were specifically enriched. The data were processed using the Seurat package (v4.3.0) for quality control and analysis. We acknowledge the original authors for providing the raw data.

Results

Differential gene expression and pathway analysis post-SCI

In this study, we integrated Bulk RNA-seq datasets (GSE47681 and GSE5296) to identify key differentially expressed genes in the subacute phase following SCI at 3 days post-injury (dpi) and 7 dpi. At 3 dpi, 936 genes were significantly upregulated, while 424 genes were downregulated (Fig. 1A). At 7 dpi, 978 upregulated genes and 498 downregulated genes were identified (Fig. 1B). The heat maps of the differentially expressed genes at both time points are presented in Fig. 1C-D. Gene set enrichment analysis (GSEA) for both 3 dpi and 7 dpi revealed a significant downregulation of pathways related to neuronal regeneration, including the regulation of axonogenesis and neurotransmitter secretion (Fig. 1E-F). Interestingly, pathways associated with macrophage activation and migration were also consistently downregulated at both time points (Fig. 1G-H). These findings align with previous studies on macrophage behavior during SCI. Specifically, Yuan et al. highlighted the potential of modulating macrophage activity to promote SCI repair [34].

Fig. 1

Differential Gene Expression and Pathway Analysis Post SCI. (A-B) Volcano plot showing differentially expressed genes (DEGs) at subacute phase (3dpi and 7dpi) versus sham group (|log2(FC)|>1 and P < 0.05). (C-D) Heatmap illustrating the expression levels of DEGs across different samples. (E-H) Gene Set Enrichment Analyses (GSEA) of DEGs post-3dpi and 7dpi

Single-cell RNA sequencing reveals immune cell dynamics post-SCI

To further elucidate the complex immune response following SCI, we performed an in-depth analysis of the immune cells isolated from the scRNA-seq dataset (GSE205038) using CD45⁺ flow cytometry. After quality control, 55,159 immune cells were identified in both the control and subacute phase (3 dpi and 7 dpi) samples, including microglia, macrophages, neutrophils, NK cells, and B cells (Fig. 2A). Following injury, flow cytometry masked specific changes in the immune microenvironment. However, we observed an initial increase, followed by a decrease in microglia, a significant increase in neutrophils at 3dpi, and a sustained increase in macrophages throughout the subacute phase (Fig. 2B). Figure 2C shows the results for marker genes for each cell subset: Ms4a7⁺ macrophages, Siglech⁺ microglia, Ltf⁺ neutrophils, Nkg1⁺ NK cells, and Ly6d⁺ B cells.

Fig. 2

Single-Cell RNA Sequencing Reveals Immune Cell Dynamics Post SCI. (A) UMAP (uniform manifold approximation and projection) plots of immune cells in sham, 3dpi, and 7dpi samples. (B) The percentage of each immune cell subtype in sham, 3dpi, and 7dpi samples among all immune cells. (C) Feature plots displaying the expression of marker genes for each cell subset. The color intensity indicates the level of expression. (D) UMAP shows three macrophage subtypes: Border-associated macrophages (BAMs), Inflammatory macrophages (IMs), and Chemotaxis-inducing macrophages (CIMs). (E-F) Heatmap (E) and volcano plots (F) revealing the z-score normalized expression of top marker genes for each macrophage subtype (Mrc1 for BAMs, Il-1b for IMs, Arg1 for CIMs) across different samples (|log2(FC)|>0.3 and p < 0.05). (G) PAGA trajectory analysis of the three macrophage subsets. (H-K) Monocle2 trajectory analysis for each macrophage subtype. (L) Slingshot trajectory analysis of macrophages indicating a differentiation pathway from Il-1b⁺ IMs to intermediate Mrc1⁺ BAMs and finally to Arg1⁺ CIMs. (M) Hierarchical clustering analysis (HCA) and GO analysis identified key distinct features along the differentiation trajectory from Il-1b⁺ IMs to Mrc1⁺ BAMs and Arg1⁺ CIMs

Further sub-clusters of the macrophage population identified three distinct subtypes: BAMs, IMs, and CIMs (Fig. 2D). Mrc1⁺ BAMs were predominantly involved in myeloid cell differentiation, antigen presentation, and postsynaptic translation; Il-1b⁺ IMs were associated with immune cell activation, inflammatory responses, gliogenesis, and immune cell migration; Arg1⁺ CIMs were linked to wound healing, oxidative phosphorylation, and synaptic structure maturation (Fig. S1A-C). Heatmaps and volcano plots were used to display highly expressed genes in Mrc1⁺ BAMs, Il-1b⁺ IMs, and Arg1⁺CIMs (Fig. 2E, F).

We subsequently tracked the differentiation trajectory of various macrophage subpopulations. The PAGA analysis revealed distinct macrophage clusters, each associated with different functional states during SCI. The thickness of the edges in the network reflects the strength of the connections between clusters, with the strongest interaction observed between BAMs and IMs, followed by BAMs and CIMs (Fig. 2G). Using Monocle2 for pseudotime analysis of macrophage subpopulations, we observed a transition from the sham group to the SCI group, with macrophages differentiating from IMs and BAMs into CIMs following SCI (Fig. 2H-K). Slingshot trajectory analysis further confirmed a differentiation pathway from Il-1b⁺ IMs to intermediate Mrc1⁺ BAMs, and finally to Arg1⁺ CIMs (Fig. 2L). A trajectory heatmap (Fig. 2M) demonstrated a shift in gene expression from negative regulation of tissue remodeling and promotion of macrophage migration to negative regulation of inflammatory cytokines in cells transitioning from Il-1b⁺ IMs to Arg1⁺ CIMs. These single-cell analysis results underscore the potential therapeutic value of modulating the transition from Il-1b⁺ IMs to Arg1⁺ CIMs, which may improve the inflammatory response and facilitate spinal cord injury repair.

HdWGCNA identifies key gene modules linked to macrophage subtypes post-SCI

HdWGCNA is a powerful tool for uncovering relationships between complex gene modules and cell phenotypes, thereby aiding in the discovery of biomarkers and therapeutic targets through the analysis of cellular heterogeneity [35]. In our study, we set the soft-threshold power to eight and classified the genes into nine distinct modules, the turquoise module (S100a6, Pgam1, Capg); blue module (Rpl23, Rpl37a, Rpl27a); brown module (Marcks, Mef2c, Fcrls); red module (Btf3, Rhog, Lsp1); magenta module (Klf4, Ms4a6b, Fos); green module(Gas6, Glb1, Gnas); yellow module (Syngr1, Cd63, Creg1); black module (Cdk2ap2, Osm, Nktr); pink module (Hcls1, Tm6sf1, Clec4a3) (Fig. 3A-C).

Fig. 3

High-Dimensional Gene Co-expression Network Analysis (hdWGCNA) Identifies Key Gene Modules Linked to Macrophage Subtypes Post SCI. (A) Plots showing the determination of the soft-threshold power used in hdWGCNA. A power of 8 was selected as the optimal threshold. (B) Dendrogram showing hierarchical clustering of genes into distinct modules based on their co-expression patterns using WGCNA. (C) The bar plots display the module eigengene (kME) values for genes within different color-coded modules (turquoise, blue, brown, red, magenta, green, yellow, black, pink). (D) Correlation analysis illustrating the relationships between different gene modules. Positive and negative correlations are indicated by blue and red colors, respectively. (E) Dot plot showing the correlation of key modules across different macrophage subtypes (BAMs, IMs, CIMs). (F-I) Network plots for key gene modules: Brown module (F), associated with Il-1b⁺ IMs. Blue module (G) and Magenta module (H) are associated with Mrc1⁺ BAMs. Yellow module (I), associated with Arg1⁺ CIMs. (J) Functional enrichment analysis of the brown module and yellow module. (K) Module association analysis of key genes between key modules. (L) Functional enrichment analysis of the yellow module

Among the nine modules analyzed, the yellow and green modules exhibited the strongest positive correlation (Cor = 0.73), whereas the yellow and magenta modules displayed the strongest negative correlation (Cor = -0.58) (Fig. 3D). Further, the blue module was most strongly associated with Mrc1⁺ BAMs, while the brown and magenta modules were primarily linked to Il-1b⁺ IMs. The yellow module showed a predominant association with Arg1⁺ CIMs (Fig. 3E). Figure 3F-I illustrate the key genes and interaction networks within these four modules. We further examined the functional roles of these modules and assigned them the following designations: the brown module as the “Inflammatory module,” the blue module as the “Ribosome module,” the magenta module as the “RNA Modulation module,” and the yellow module as the “Remodeling module” (Fig. 3J-L, Fig. S2A-B).

Subsequent functional enrichment analysis revealed that the inflammatory module, most associated with Il-1b⁺ IMs, is involved in the regulation of inflammatory responses, cytokine production, gliogenesis, and processes detrimental to wound healing (Fig. 3J-K). The remodeling module, which is predominantly related to Arg1⁺CIMs, was implicated in tissue homeostasis, extracellular vesicle formation, autophagy, angiogenesis, and lipid metabolism (Fig. 3K-L). Analysis of hdWGCNA data further indicated the potential to enhance SCI repair by modulating the transition from Il-1b⁺ IMs to Arg1⁺CIMs.

Modulating macrophage differentiation to promote SCI repair

Based on macrophage trajectory analysis and single-cell WGCNA results, we hypothesized that the suppression of Il-1b⁺IMs post-injury could promote their differentiation into Arg1⁺CIMs. To investigate this, we performed a combined analysis of DEGs from bulk RNA-seq data, genes from the brown module most associated with IMs, and characteristic marker genes of IMs. This analysis identified 11 core genes which showed the strongest correlations with IMs (Fig. 4A). Co-expression analysis revealed that these 11 genes were primarily involved in glial cell activation, neuroinflammatory responses, and complement activation (Fig. 4B). Subsequently, Friends analysis screened eight of these genes (Olfml3, Csf1r, Fcrls, Ly86, Cfh, Cx3cr1, Selplg, and Hexb) as hub genes of IMs (Fig. 4C). Heatmaps and expression trend plots of transcriptome further demonstrated that these hub genes were highly expressed post-injury (Fig. 4D-E). Further validation using RT-qPCR was consistent with the trend of the transcriptome (Fig. 4F). Notably, these genes were highly expressed in Il-1b + IMs but showed low expression in the Mrc1⁺ BAMs and Arg1⁺CIMs populations along the differentiation trajectory (Fig. 4G).

Fig. 4

Identification and Validation of Key Genes and Potential Therapeutic Targets Post SCI. (A) Venn diagram showing the overlap of DEGs, the brown module and Inflammatory macrophages (IMs) marker genes. (B) Co-expression network of the eleven hub genes of IMs identified in the Venn diagram. (C) Friends’ analysis reveals the core genes of IMs, cut off > 0.3. (D) Heatmap showing the expression levels of the hub genes in SCI samples compared to sham controls. (E): Box plots displaying the transcriptome expression levels of the hub genes in SCI and sham samples. The red and blue boxes represent SCI and sham samples, respectively. (F) RT-qPCR validation of the expression of hub genes in sham and SCI mice, normalized to Gapdh (n = 3). Statistical analyses were done using Student’s t test. *p < 0.05, **p < 0.01, ***p < 0.001. (G) Expression trend plots of hub genes along the macrophage differentiation trajectory. (H) Molecular docking analysis predicting the binding sites of Amantadine with six hub proteins (CFH, CSF1R, CX3CR1, FCRLS, HEXB, and LY86)

Using the CMAP database, we identified amantadine (p = 0.034) as a potential small-molecule candidate capable of targeting Il-1b⁺ IMs hub genes. AutoDock software was successfully applied to predict the potential binding sites of amantadine in six of the eight proteins (CFH, CSF1R, CX3CR1, FCRLS, HEXB, and LY86), as detailed in Fig. 4H. Therefore, amantadine might be a promising drug in regulating the transition of macrophages phenotype from the IMs to the BAMs population.

Amantadine modulates macrophage differentiation via the HIF-1α and NF-κB pathways

To better understand the effect of amantadine on macrophage phenotype transition, we treated RAW264.7 cells with LPS and IFN-γ to simulate the inflammatory microenvironment following SCI in vitro. We conducted the CCK-8 assay to evaluate the impact of amantadine on the viability of RAW264.7 cells. The results revealed that following LPS + IFN-γ stimulation, the cell viability of RAW264.7 cells decreased to approximately 60%. This decrease was reversed by amantadine treatment in a dose-dependent manner. Due to its much more enhanced protective effect, a dose of 50 µM was selected for subsequent experiments (Fig. 5A).

Fig. 5

Amantadine Modulates Macrophage Differentiation and Inhibits Inflammatory Responses in vitro. (A) Cell viability of RAW264.7 cells was estimated by CCK-8 assay (n = 3, ***P < 0.001 for comparison with the LPS + IFN-γ group. (B-D) Integrated fluorescence signal of Il-1b⁺IMs(B), Mrc1⁺BAMs (C), and Arg1⁺CIMs (D). (E-G) Representative immunofluorescent staining images showing the effects of different treatments on macrophage differentiation. Il-1b (red), Mrc1 (red), Arg1 (red), and F4/80 (green) are shown, with DAPI (blue) staining nuclei. Scale bar, 50 μm. (H-I) The flow cytometry analysis and quantification results of the subtypes of RAW264.7 cells (gated on F4/80⁺) treated with LPS and IFN-γ, with or without Amantadine. (J-K) Western blot analysis showing the expression levels of Arg1, HIF-1α, NF-κB p65, and phosphorylated-NF-κB p65 (p-p65). Statistical differences were determined by using the Analysis of Variance (ANOVA) with Tukey’s multiple comparison post hoc test (*p < 0.05, **p < 0.01, ***p < 0.001, and ns: no significance)

Immunofluorescence staining was performed to assess the effects of the different treatments on macrophage differentiation. We utilized Il-1b (interleukin 1 beta), Mrc1 (mannose receptor C-type 1), and Arg1 (arginase 1) as markers for IMs, BAMs, and CIMs, respectively. As demonstrated in Fig. 5B-G, following treatment with LPS and IFN-γ, we observed a marked increase in the pro-inflammatory factors Il-1b from macrophage. This finding suggested that a combination of LPS and IFN-γ was sufficient to convert macrophages into inflammatory macrophages (IMs). The amantadine group further exhibited a noteworthy decrease in the integrated fluorescence signal of Il-1b to approximately 50.1% compared to the LPS + IFN-γ group. In contrast, the integrated fluorescence signals of Mrc1 and Arg1 were found to be significantly higher in the amantadine group, with Mrc1 being approximately 3.3- fold higher and Arg1 being approximately 18-fold higher compared to the LPS + IFN-γ group. Flow cytometry confirmed these findings, revealing that amantadine decreased the proportion of Il-1b⁺/F4/80⁺ macrophages (IMs) to approximately 52.8% of the proportion in the LPS + IFN-γ group, while simultaneously increasing the proportions of Mrc1⁺/F4/80⁺ macrophages (BAMs) by 1.1-fold and Arg1⁺/F4/80⁺ macrophages (CIMs) by 1.3-fold (Fig. 5H-I). Collectively, these results indicate that amantadine inhibits inflammatory responses by suppressing IMs and enhancing CIMs modulation.

Previous research has indicated that the HIF-1α signaling pathway is a crucial regulator of macrophage polarization, particularly in promoting the pro-inflammatory phenotype [36, 37]. Additionally, the NF-κB signaling pathway plays essential roles in immune responses, inflammation, and cell survival by promoting the transcription of pro-inflammatory genes [38, 39]. Next, we investigated the possible mechanism by which amantadine repopulates macrophages using western blotting. The western blotting results showed a significant increase in Arg1 levels following amantadine treatment (Fig. 5J). Compared to the control group, LPS + IFN-γ significantly upregulated the expression of Hypoxia-inducible factor 1-alpha (HIF-1α) and phosphorylated-NF-κB p65 (p-p65). Notably, amantadine partially blocked these effects (Fig. 5K). These findings indicate that amantadine could modulate the transition from Il-1b⁺ IMs to Arg1⁺ CIMs through the HIF-1α and NF-κB pathways.

Effects of amantadine on neuroprotection after SCI

We examined the effects of amantadine on early-stage neural inflammation in vivo. The experimental timeline is shown in Fig. 6A. At 3 dpi, the SCI + amantadine group exhibited a significant decrease in the local recruitment of Il-1b⁺ IMs to approximately 73.1% of the levels observed in the SCI group (Fig. 6B-C). Additionally, treatment with amantadine led to approximately 5- and 3.2-fold increases in the fluorescence intensity of Mrc1⁺ BAMs (Fig. 6D-E) and Arg1⁺ CIMs (Fig. 6F-G), respectively, at various lesion sites compared to the SCI group. These findings suggest that amantadine may reshape the inflammatory environment caused by SCI by promoting the transition of Il-1b⁺ IMs to Mrc1⁺ BAMs and Arg1⁺ CIMs.

Fig. 6

Amantadine Reshapes the Inflammatory Environment and Promotes Neuronal Survival and Regeneration Post SCI in vivo. (A) Schedule of SCI mice treated with Amantadine. Amantadine was administered intraperitoneally (i.p) at a dose of 2 mg/Kg daily. The timeline includes key time points for BBB score assessment, inflammation analysis, Nissl staining, Tunel assay, MAP2/GFAP staining, and MEP analysis. (B-G) Representative immunofluorescent staining images and quantification of the relative fluorescence intensity of Il-1b⁺ IMs (B-C), Mrc1⁺ BAMs (D-E), and Arg1⁺ CIMs (F-G) at 3dpi at lesion sites. Il-1b (red), Mrc1 (red), Arg1 (red), and F4/80 (green) are shown, with DAPI (blue) staining nuclei. Scale bar, 50 μm. (H-I) Representative Nissl staining images and quantification of the average optical density of Nissl-stained neurons at 7 dpi. Enlarged images of boxed areas are shown below. Scale bar, 50 μm. (J-K) Representative Tunel staining images and the quantification of the proportion of Tunel⁺ neurons at 7 dpi. Enlarged images of boxed areas are shown below. Scale bar, 50 μm. (L) Representative immunofluorescent staining images of sagittal sections of the injured spinal cord at 14 dpi. MAP2 (red) and GFAP (green) are shown, with DAPI (blue) staining nuclei. (M) Quantification of the relative fluorescence intensity of MAP2 and GFAP is shown. Statistical differences were determined by using the Analysis of Variance (ANOVA) with Tukey’s multiple comparison post hoc test(*p < 0.05, **p < 0.01, ***p < 0.001, and ns: no significance)

In addition to macrophage differentiation, spinal cord regeneration involves a variety of other processes, including neuronal survival and maturation [40]. As such, Nissl staining and Tunel assays were applied to assess the neuroprotective effects of amantadine at 7dpi in vivo. Nissl staining revealed that the group treated with amantadine had an approximately 1.5-fold increase in the number of neurons compared to the SCI group (Fig. 6H-I). The Tunel assay confirmed a significantly lower percentage of apoptotic cells in the SCI + amantadine group (49.7 ± 6.2%) compared with the SCI group (78.2 ± 5.7%) (Fig. 6J-K). These findings indicate that amantadine exerts neuroprotective effects on injured spinal cords.

To further investigate the neuronal distribution at the lesion site, we stained sagittal sections of the injured spinal cord with glial fibrillary acidic protein (GFAP) to visualize the activated astrocytes and microtubule-associated protein 2 (MAP2) to reveal mature neurons at 14dpi. A significantly higher intensity of MAP2⁺ cells was observed in the lesion area of mice in the SCI + amantadine group compared with the SCI group (Fig. 6L-M). In summary, our results demonstrated that amantadine can improve macrophage phenotypes following SCI, thereby mitigating early inflammatory responses and protecting neural survival.

Amantadine enhances neurological function recovery post-SCI

We assessed neurological function recovery at 28 dpi through a comprehensive evaluation of behavior, motor function, and neurophysiology. The SCI + amantadine group demonstrated significantly better motor function recovery than the SCI group during 4 weeks post-SCI (Fig. 7A). In the climbing test, mice in the SCI group showed post-surgery paralysis and continued to drag both hind limbs, whereas mice in the SCI + amantadine group displayed improved palmar weight-bearing movement (Fig. 7B). Enhanced waveforms were further observed in the SCI + amantadine group, with increased amplitude and reduced latency in MEP testing compared to mice in the SCI group, indicating improved neural conduction (Fig. 7C-E). Finally, to evaluate the biomedical safety of amantadine, we performed a histological analysis of the major organs. Histopathological staining revealed no visible signs of cellular or tissue disruption, necrosis, or inflammation to the heart, liver, spleen, lung, or kidney, confirming the biocompatibility and safety of amantadine (Fig. 7F).

Fig. 7

Amantadine Enhances Neurological Function Recovery and Ensures Biomedical Safety Post SCI in vivo. (A) The Basso, Beattie & Bresnahan (BBB) locomotor rating scale of hindlimbs of mice from different groups during the 28-day post-SCI period (n = 5). The Amantadine-treated group demonstrated significantly better motor function recovery compared to the SCI group (***P < 0.001). (B) Representative images from the climbing test showing the motor function of mice. (C-D) Motor evoked potential (MEP) testing to assess neural conduction function. Quantification of MEP amplitude (C) and latency (D) compared to the SCI group. (E) Representative MEP waveforms for each group. (F) Histological analysis of major organs (heart, liver, spleen, lung, kidney) to evaluate the biomedical safety of Amantadine. Statistical differences were determined using Analysis of Variance (ANOVA) with Tukey’s multiple comparison post hoc test.(*p < 0.05, **p < 0.01, ***p < 0.001, and ns: no significance)

Discussion

Microarray analysis revealed significant changes in gene expression during the acute phase of SCI. GSEA at both 3 dpi and 7 dpi confirmed the upregulation of pathways involved in macrophage activation and recruitment at both time points. However, pathways related to axonal regeneration and neurotransmitter function were progressively downregulated from 3 dpi to 7 dpi, indicating a shift in the cellular response during the injury progression. These findings indicate that the targeted modulation of these pathways could potentially reduce local inflammation and promote neural repair [41, 42].

Further, scRNA sequencing identified three distinct macrophage subtypes previously reported in the literature: BAMs, associated with myeloid cell differentiation and antigen presentation; IMs, involved in immune activation and inflammatory responses; and CIMs, linked to wound healing and synaptic maturation [14]. These findings transcend the traditional M1/M2 classification and highlight the complex role of macrophages in SCI [43]. Using Slingshot trajectory analysis and gene expression heat maps, we proposed a differentiation pathway from Il-1b ⁺ IMs to Mrc1⁺ BAMs, and finally to Arg1⁺ CIMs, suggesting that the modulation of this pathway may reduce inflammation and promote neural repair. Additionally, analysis of the CMAP database identified amantadine as a potential small-molecule candidate for the targeting Il-1b + IMs, thus providing new therapeutic insights.

Amantadine, commonly used as a central nervous system stimulant as well as an anti-Parkinsonian medication to improve motor symptoms in patients with Parkinson’s disease, also serves as an adjunct therapy for conditions such as chronic fatigue syndrome and certain neuropathic pain disorders [21, 44]. While previous studies have demonstrated the therapeutic potential of amantadine in neurological conditions, our study further elucidates its specific role in modulating macrophage phenotypes and promoting neural repair in the context of SCI. The effect of amantadine on early-stage neural inflammation was evident from the significant decrease in Il-1b⁺ IMs and increase in Arg1⁺ CIMs and Mrc1⁺ BAMs at the lesion sites identified by immunofluorescence imaging. This reshaping of the inflammatory environment is crucial in promoting a shift from a pro-inflammatory to a reparative macrophage phenotype and enhancing the potential for neural repair.

Behavioral and electrophysiological assessments further revealed significant improvements in motor function and neural conduction in the amantadine-treated group compared to the SCI group. These functional outcomes were supported by immunohistochemical analyses showing increased numbers of MAP2⁺ neurons and decreased numbers of apoptotic cells in the spinal cord tissue, indicating enhanced neuronal survival and reduced apoptosis. Furthermore, higher numbers of GFAP⁺ astrocytes were observed, indicating a supportive environment for neural repair. Histological analysis confirmed the biomedical safety of amantadine, with no apparent tissue damage observed in the major organs. This is consistent with the safety profile of the drug in clinical use for other conditions, supporting its potential repurposing for SCI therapy [45].

Although this study provides promising results, several limitations must be addressed before advancing toward clinical trials. First, one of the limitations of this study is the lack of focus on microglia, whose dynamic and distinct roles alongside macrophages during various stages of SCI recovery remain underexplored, highlighting the need for further research to differentiate the contributions of microglia and infiltrating macrophages [46]. Second, while we explored the cytotoxic effects of amantadine on RAW264.7 macrophages in vitro, comprehensive toxicity testing is still necessary to evaluate both short-term and long-term safety in mice. Lastly, a limitation of this study is the use of intraperitoneal injections in our animal model, underscoring the need for future research to develop more precise drug delivery systems, such as nanoparticles or biocompatible materials to enhance drug retention and efficacy at the injury site [47].

Conclusions

In conclusion, our study highlights the therapeutic potential of amantadine in improving macrophage phenotypes after SCI, reducing early inflammatory responses, and enhancing neural functional recovery. Overall, these findings provide a foundation for future translational research and clinical applications, offering hope for improved treatment of patients with SCI.