Targeting microglial PANoptosis through AMPK activation: Metformin as a promising therapy for spinal cord injury

Abstract

Spinal cord injury (SCI) triggers robust neuroinflammation, in which microglial activation and dysregulated programmed cell death exacerbate secondary damage. PANoptosis, a recently defined inflammatory cell death modality integrating pyroptosis, apoptosis, and necroptosis, has been implicated in central nervous system (CNS) injury, yet its cellular specificity and regulation remain unclear. In this study, integrative bioinformatic analyses identified microglia as the predominant PANoptotic cell population after SCI, with PANoptosis inversely correlated with adenosine monophosphate (AMP)-activated protein kinase (AMPK) pathway activity. Using metformin, a classical AMPK activator, we demonstrated that pharmacological activation of AMPK suppresses microglial PANoptosis, promotes a shift toward an anti-inflammatory microglial phenotype, and suppresses the associated pro-inflammatory cytokine cascade by inhibiting the nuclear factor κB (NF-κB) signaling pathway. This modulation of the immune microenvironment promotes axonal regeneration, remyelination, and functional recovery in a rat SCI model. Notably, the neuroprotective effects of metformin were abrogated by the AMPK inhibitor compound C (CC), confirming AMPK-dependence. Together, our findings demonstrate that metformin alleviates microglial PANoptosis in an AMPK-dependent manner, promoting tissue repair and functional recovery after SCI. This study uncovers AMPK as a key regulator of microglial PANoptosis and highlights the therapeutic potential of metformin for SCI repair.

Graphical abstract

Highlights

• •Microglial PANoptosis is identified as a key pathological process post-SCI.
• •AMPK activity is inversely correlated with microglial PANoptosis in SCI.
• •Metformin activates AMPK to suppress PANoptosis and neuroinflammation in microglia.
• •AMPK-dependent regulation by metformin promotes axonal regeneration and behavioral recovery.

1. Introduction

Following spinal cord injury (SCI), the formation of an inhibitory microenvironment severely compromises the already limited intrinsic regenerative capacity of the central nervous system (CNS) [1]. During the acute phase, the lesion site undergoes profound pathological changes, including ischemia, hypoxia, lipid peroxidation, excitotoxic neurotransmitter release, and ionic dysregulation. These events are rapidly followed by the infiltration of neutrophils and monocytes, activation of microglia and macrophages, and the release of axon growth-inhibitory molecules, collectively establishing a hostile environment for regeneration. After SCI, apoptosis initially starts in neurons with disrupted ionic homeostasis. These neurons release inflammatory factors to trigger apoptosis in astrocytes, which then induce oligodendrocyte apoptosis through reduced nutritional support and release of toxic substances. In turn, oligodendrocyte apoptosis further exacerbates neuronal damage and reciprocally accelerates astrocyte apoptosis, forming a vicious cycle that aggravates SCI [2]. Moreover, the death of infiltrating and resident immune cells further amplifies the secondary inflammatory cascade, leading to widespread neuronal apoptosis and necrosis, disruption of axonal conduction, and ultimately impaired tissue repair and functional recovery [3].

Microglia, the resident macrophage-like immune cells of the CNS, play a pivotal role in maintaining neural homeostasis, coordinating injury responses, and promoting tissue repair. Functionally, microglia exhibit remarkable phenotypic plasticity and dynamically adapt to microenvironmental cues. Under physiological conditions, they constantly survey the CNS milieu, preserving homeostasis by pruning redundant synapses, clearing metabolic waste, and regulating synaptic remodeling. Following SCI, microglia become rapidly activated by molecules such as high mobility group box-1 and release proinflammatory mediators, including tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β), and reactive oxygen species (ROS), thereby exacerbating neuronal injury. Conversely, microglia can also exert reparative effects by secreting cytokines such as IL-4 and IL-13, which promote macrophage polarization toward a pro-regenerative phenotype and facilitate tissue remodeling. However, indiscriminate depletion of microglia has been shown to reduce neuronal and oligodendrocyte populations, and hinder recovery [4]. Thus, maintaining microglial immune homeostasis is critical for supporting regenerative processes after SCI.

Inflammation-driven programmed cell death represents a central mechanism in immune regulation and has profound implications for neuroregeneration following CNS trauma [5]. PANoptosis, a recently characterized form of inflammatory cell death, integrates molecular features of pyroptosis, apoptosis, and necroptosis [6]. It is capable of inducing both cell lysis and a robust inflammatory cytokine storm. Recent studies have shown that bone marrow-derived macrophages can undergo PANoptosis through Z-DNA-binding protein 1-mediated PANoptosome formation, accompanied by mitochondrial dysfunction and the release of mitochondrial DNA [7]. In the context of neurological diseases, PANoptosis has been implicated in the pathogenesis of neurodegenerative disorders and acute CNS injuries [[8], [9], [10]]. However, the specific role of PANoptosis in regulating microglial function following SCI remains largely unexplored and warrants further exploration.

Adenosine monophosphate (AMP)-activated protein kinase (AMPK), a crucial intracellular energy sensor, functions as a heterotrimeric complex composed of a catalytic α-subunit and two regulatory β- and γ-subunits. It plays multifaceted roles in cellular homeostasis, including the regulation of neuroinflammation, oxidative stress, autophagy, mitochondrial integrity, and energy metabolism [[11], [12], [13]]. AMPK has been extensively studied in the context of aging, neural injury, and metabolic disorders [14,15]. The activation of AMPK signaling has been shown to suppress neuronal pyroptosis and inhibit M1 polarization of microglia, thereby promoting motor function recovery following SCI [16,17]. Emerging evidence also suggests a potential crosstalk between AMPK signaling and PANoptosis [18,19], although the precise mechanisms underlying this interaction in SCI remain largely undefined. Metformin, a classical AMPK agonist, activates the pathway primarily through inhibition of mitochondrial complex I, resulting in an increased AMP/adenosine triphosphate (ATP) ratio and subsequent AMPK phosphorylation [20]. Beyond its well-established role in the treatment of metabolic syndromes, metformin has demonstrated robust anti-inflammatory and neuroprotective effects in various neurological disorders, including neurodegenerative diseases and traumatic CNS injuries [21,22]. These findings lead us to hypothesize that metformin may facilitate recovery after SCI by suppressing microglial PANoptosis via AMPK activation.

In this study, we first applied integrative bioinformatics approaches to investigate the cellular specificity of PANoptosis following SCI, with a particular focus on its association with the AMPK signaling pathway. We then conducted a series of in vivo and in vitro experiments to determine whether metformin could suppress microglial PANoptosis by activating AMPK, specifically inhibiting the key components of pyroptosis, apoptosis, and necroptosis. Finally, we assessed whether AMPK activation by metformin facilitates axonal regeneration and promotes functional motor recovery after SCI.

2. Materials and methods

2.1. Dataset acquisition

Gene expression datasets GSE5296 and GSE47681 were retrieved from the Gene Expression Omnibus (GEO) database. Data were extracted for the Sham group and for mice at 1, 3, and 7 days post-thoracic SCI. GSE5296 included 17 samples (8 Sham and 9 SCI), while GSE47681 comprised 34 samples (9 Sham and 25 SCI). PANoptosis-related genes (PRGs) were obtained from GeneCards and relevant literature, totaling 41 genes.

2.2. Data processing and screening of PANoptosis related genes

All statistical analyses were performed using R software (v4.1.3). Batch effects between the merged RNA sequencing datasets GSE5296 and GSE47681 were corrected using the ComBat function from the sva package. To visualize the effectiveness of batch effect removal, principal component analysis (PCA) was conducted using the FactoMineR package and visualized with factoextra. Differentially expressed genes (DEGs) between the Sham and SCI groups were identified using the limma package with thresholds set at |log2(fold change)| > 0.33 and adjust P value < 0.05. The intersection between DEGs and the PRGs was determined using the UpSetR package. Gene Ontology (GO) analysis of the intersection genes was performed using the ClusterProfiler package.

2.3. Single-gene set variation analysis (GSVA) and immune infiltration analysis

GSVA was performed using the GSVA package to calculate PANoptosis enrichment scores for each sample. Immune cell infiltration was evaluated using the xCell algorithm, which provided estimated enrichment scores for various immune cell types across samples. The resulting immune cell infiltration profiles were visualized using the pheatmap package. Spearman's correlation analysis and result visualization were performed using the ggcor package to explore potential associations between PANoptosis-related activity, the AMPK signaling pathway, and the immune microenvironment within spinal cord tissues.

2.4. Weighted gene co-expression network analysis (WGCNA)

WGCNA was performed using the WGCNA package to identify gene modules associated with PANoptosis-related transcriptional changes. The appropriate soft threshold, set at 3, was chosen based on the criterion of approximate scale-free topology using the pickSoftThreshold function. The adjacency matrix was subsequently transformed into a topological overlap matrix (TOM) to measure the network connectivity among genes. Hierarchical clustering based on TOM dissimilarity was applied, and gene modules were defined using the dynamic tree cut algorithm. Each module was summarized by its eigengene, and Pearson's correlation analysis was performed between module eigengenes and PANoptosis scores to identify biologically relevant modules. The module exhibiting the strongest correlation was selected for further functional annotation. Genes within this module were subjected to pathway enrichment analysis using the clusterProfiler package to identify significantly enriched signaling pathways.

2.5. Cell culture

The murine BV2 microglial cell line was obtained from Wuhan Pricella Biotechnology Co., Ltd. (Wuhan, China). Cells were maintained in Dulbecco's modified Eagle medium (DMEM) (Gibco, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (FBS) (Gibco) and 1% penicillin-streptomycin (Thermo Fisher Scientific Inc., Waltham, MA, USA) at 37 °C in a humidified atmosphere containing 5% CO₂. For treatments, cells were pre-treated with metformin (1 μM; Selleckchem, Houston, TX, USA) and/or compound C (CC) (1 μM; Shanghai Yuanye Biotechnology Co., Ltd., Shanghai, China) for 8 h, followed by stimulation with lipopolysaccharide (LPS) (100 ng/mL; Sigma-Aldrich, St. Louis, MO, USA) for 16 h. Subsequently, ATP (5 mM; Sigma-Aldrich) was added and incubated for an additional 2 h to induce inflammasome activation. Subsequent experiments and data collection were conducted by investigators blinded to the experimental groups to ensure unbiased interpretation.

2.6. Western blot

Samples were lysed in radio immunoprecipitation assay (RIPA) buffer (Beyotime Biotechnology, Shanghai, China) containing protease and phosphatase inhibitors. Total protein lysates were separated by 7.5%–12.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto polyvinylidene fluoride (PVDF) membranes (Millipore, Billerica, MA, USA). After blocking with 5% skimmed milk (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) for 1 h, membranes were incubated overnight at 4 °C with the appropriate primary antibodies (Table S1). On the following day, membranes were incubated with horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature. Protein bands were visualized and quantified using ImageJ software (version 1.8.0, National Institutes of Health (NIH), Bethesda, MD, USA). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or β-tubulin was used as the internal loading control.

2.7. Real-time quantitative polymerase chain reaction (RT-qPCR) analysis

Total RNA was extracted from BV2 cells and spinal cord tissue samples using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and quantified with a NanoDrop spectrophotometer (Thermo Fisher Scientific Inc.). First-strand complementary DNA (cDNA) was synthesized from 1 μg of total RNA using the FastKing RT Kit (Tiangen Biotech (Beijing) Co., Ltd.‌‌, Beijing, China) according to the manufacturer's instructions. Quantitative PCR was performed using the UltraSYBR Mixture (Cowin Biotechnology Co., Ltd., Beijing, China) on a LightCycler real-time PCR system (Roche, Basel, Switzerland). Relative gene expression levels were calculated using the 2^−ΔΔC_T method. GAPDH was used as the internal reference gene. Primer sequences are listed in Table S2.

2.8. Enzyme-linked immunosorbent assay (ELISA)

The levels of pro-inflammatory cytokines, including IL-1β, TNF-α, IL-6, and IL-18, in spinal cord tissue homogenates and cell culture supernatants were quantified using ELISA kits following the manufacturers’ instructions. ELISA kits for IL-1β, TNF-α, and IL-6 were from Beyotime Biotechnology, and the IL-18 kit was from Jiangsu Meimian Industrial Co., Ltd. (Yancheng, China). Absorbance was measured using a BioTek multimode microplate reader, and cytokine concentrations were calculated based on standard curves generated for each assay.

2.9. Flow cytometry assays

For BV2 cell polarization analysis, M1 polarization was induced by treatment with LPS (100 ng/mL; Sigma-Aldrich) and interferon-γ (IFN-γ) (20 ng/mL; BioLegend, San Diego, CA, USA), whereas M2 polarization was induced by IL-4 (20 ng/mL, BioLegend, San Diego, CA, USA). Cells were incubated with Zombie near infrared (NIR) viability dye (BioLegend) for 15 min at 4 °C to exclude dead cells. After washing, surface staining was conducted for 40 min at 4 °C (Table S1). Cells were then fixed and permeabilized using Fix/Perm buffer (BioLegend) for 20 min at room temperature. Intracellular staining was performed for 20 min at room temperature.

For caspase-1 activity detection, BV2 single-cell suspensions were incubated with the fluorochrome-labeled inhibitor of caspases (FLICA) 660 caspase-1 assay kit (ImmunoChemistry, Bloomington, MN, USA) for 30 min at room temperature according to the manufacturer's instructions.

For apoptosis analysis, BV2 single-cell suspensions were incubated with annexin V-fluorescein isothiocyanate (FITC) (Solarbio, Beijing, China) for 5 min at room temperature, followed by propidium iodide (PI) staining. Samples were analyzed immediately by flow cytometry using a LSRFortessa cytometer (Becton, Dickinson and Company), and data were processed with FlowJo software (version X.0.7; Becton, Dickinson and Company).

2.10. Immunofluorescence staining

BV2 cells and 10 μm cryosections of spinal cord tissue were fixed with 4% paraformaldehyde for 15 min, then washed with tris-buffered saline with Tween 20. Samples were blocked and permeabilized using 5% bovine serum albumin (Solarbio) containing 0.25% Triton X-100 (Beyotime Biotechnology) for 1.5 h, followed by overnight incubation at 4 °C with primary antibodies (Table S1). After washing, appropriate secondary antibodies were applied for 1 h. Co-localization staining was performed using a Triple Label Four-Color Multiplex Fluorescence Staining Kit (Beyotime Biotechnology) according to the manufacturer's instructions. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). Images were acquired using a Zeiss LSM 800 confocal microscope (Zeiss, Oberkochen, Germany) or Leica DMi8 fluorescence microscope (Leica, Wetzlar, Germany) and analyzed with ImageJ. Specifically, positive regions were defined using a uniform threshold, and binary masks were generated as region of interests (ROIs). Mean fluorescence intensity within ROIs was measured after background subtraction. At least three randomly selected, non-overlapping fields per section were analyzed, and the average values were used for statistics.

2.11. Lactate dehydrogenase (LDH) release assay

LDH levels were used as an indicator of cell membrane damage and cytotoxicity. LDH release in BV2 cell supernatants was quantified using a commercial LDH assay kit (Solarbio), according to the manufacturer's protocol. Absorbance was measured at 450 nm using a BioTek multimode microplate reader.

2.12. TdT-mediated dUTP nick-end labeling (TUNEL) staining

Samples were fixed with 4% paraformaldehyde for 15 min and washed with phosphate-buffered saline (PBS) with Tween 20. Cells were then blocked and permeabilized with 5% bovine serum albumin (Solarbio) containing 0.25% Triton X-100 (Beyotime Biotechnology) for 1.5 h. TUNEL staining was performed using a commercial assay kit (Beyotime Biotechnology) according to the manufacturer's instructions, with incubation at 37 °C for 1 h. Nuclei were counterstained with DAPI. Fluorescence images were acquired using a Leica DMi8 microscope and analyzed with ImageJ software. For clarity of visualization, representative fluorescence images were uniformly adjusted for brightness and contrast across all groups, without altering the relative signal intensities. Quantitative analyses were performed using raw, unprocessed images to ensure accuracy.

2.13. Animals

Female Wistar rats (6–8 weeks old, 190–210 g) were obtained from Charles River Laboratories (Wilmington, MA, USA). Animals were housed under specific pathogen-free conditions at a controlled temperature of 22 ± 2 °C and relative humidity of 50%–55%, with a 12 h light/dark cycle. Food and water were available ad libitum. All animal procedures were approved by the Ethics Committee of the Institute of Radiation Medicine, Chinese Academy of Medical Sciences, Tianjin, China (Approval No.: IRM-DWLL-2021139) and conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals.

2.14. Spinal cord contusion model and treatment protocols

Animals were randomly assigned to five groups and anesthetized with isoflurane (induction at 2.5%, maintenance at 2.0%). A moderate spinal cord contusion at the T10 level was induced using the MASCIS Impactor Model III (Rutgers University, Piscataway, NJ, USA) based on a modified Allen's method. After T10 laminectomy, a 10 g impactor rod was dropped from a height of 25.0 mm. Following injury, muscles and skin were sutured in layers, and rats were placed on a heating pad for recovery. Successful SCI induction was confirmed by a characteristic tail flick response, bilateral hindlimb spasms, and subsequent complete paralysis of both hindlimbs, accompanied by visible venous congestion and hematoma formation at the injury site. On the first postoperative day, the absence of voluntary hindlimb movement and reduced muscle tone further verified the establishment of a consistent and reliable injury model. To prevent infection, cefotaxime (10 mg/kg) was administered intramuscularly once daily for three days. Manual bladder expression was performed twice daily until spontaneous urination was restored.

The Sham group received only laminectomy without SCI. The injury group received intraperitoneal saline daily after injury. The Met group received metformin (50 mg/kg/day, i.p.). The CC group received CC (2 mg/kg/day, i.p.). The Met + CC group received both metformin (50 mg/kg/day) and CC (2 mg/kg/day) via intraperitoneal injection. The number of animals used in each group, as well as their specific experimental applications, are detailed in Table S3. Subsequent sample processing, histological analyses, and data quantification were performed by independent investigators blinded to the experimental groups.

2.15. Basso, Beattie, and Bresnahan (BBB) locomotor rating scale

Hind limb motor function was evaluated using the BBB scale, which ranges from 0 to 21 points. Behavioral assessments were conducted on days 3, 7, 14, 21, 28, 42, and 56 post-injury. All evaluations were performed in a quiet, open-field environment, with each rat allowed to move freely for 4 min. Scoring was independently conducted by two experienced investigators blinded to group allocation.

2.16. Louisville swim scale (LSS)

At 56 days post-injury, motor function was evaluated using a swimming test, with scores ranging from 0 to 17 points. Key parameters included hind limb movement and alternation, forelimb dependency, trunk stability, and body angle. The test was conducted in a glass tank (60 cm × 33 cm × 38 cm) filled with water at a depth of approximately 30 cm and maintained at 24–25 °C. Each rat was tested individually for 4 min. Scoring was performed by two experienced observers blinded to the experimental groups. After testing, rats were dried thoroughly and placed on a heating pad to maintain body temperature.

2.17. Gait analysis

Gait performance at 56 days post-injury was assessed using the CatWalk XT automated gait analysis system (version 10.6; Noldus, Wageningen, The Netherlands) [23]. The system settings were as follows: camera gain, 17.5 dB; green walkway light intensity, 15.4 V; red ceiling light intensity, 17.1 V; green intensity threshold, 0.12; and time threshold range, 0.5–15 s. Under dim lighting conditions, each rat was placed at one end of the glass walkway, with a food reward positioned at the opposite end to encourage continuous movement. Each rat completed three successful runs, defined as uninterrupted crossings, and data were averaged for analysis. Gait performance and statistical analyses were conducted by investigators blinded to the experimental grouping.

2.18. Electrophysiological assessment

Nerve conduction function was evaluated using an electrophysiological recording system (Zhuhai Yiriki Co., Ltd., Zhuhai, China). Rats were anesthetized with isoflurane, and stimulating electrodes were inserted subcutaneously between the ears. Reference electrodes were placed subcutaneously on the dorsal surface, and recording electrodes were inserted into the bilateral gastrocnemius muscles. Motor evoked potentials (MEPs) were elicited by 10 mA current stimulus. Each rat underwent three repeated stimulations. Electrophysiological recordings and waveform analyses were performed by experimenters blinded to the treatment allocation. Waveforms were recorded and analyzed to determine latency (ms) and amplitude (mV) of the MEPs. Data were processed using ImageJ software.

2.19. Weight measurement and blood glucose test

After an overnight fast, body weight measurements were taken, and blood was collected from the rat's tail. Fasting blood glucose levels were then determined using the ACCU-CHEK Active meter (Roche).

2.20. Cytokine array analysis

Spinal cord tissue homogenates were applied to the cytokine array (RayBiotech, Norcross, GA, USA) and incubated at room temperature for 1 h. After washing, a biotinylated antibody cocktail was added and incubated for another 1 h. Following a second wash, cyanine 3 (Cy3)-equivalent dye-conjugated streptavidin was added and incubated for 1 h. Fluorescence signals were acquired using the InnoScan 300 Microarray Scanner (Innopsys, Toulouse, France) with a wavelength of 532 nm and a resolution of 10 μm. Data analysis was performed using ImageLab software (RayBiotech).

2.21. Magnetic resonance imaging (MRI)

MRI was performed on rats to assess the area of hyperintense lesions in each group, as described in previous studies [24]. Briefly, animals were anesthetized with isoflurane and placed in a supine position prior to scanning with a 3.0 T MRI machine (Discovery MR750; GE Healthcare, Waukesha, WI, USA). Sagittal T₂-weighted images of the spinal cord were acquired using the fast relaxation fast spin echo pulse sequence with the following MRI parameters: repetition time (TR)/echo time (TE), 3000 ms/110 ms; image matrix, 320 × 224; field of view (FOV), 6 mm; slice thickness, 2 mm; slice spacing, 0.5 mm; echo train length, 21; and number of excitations (NEX), 4. Data were processed using ImageJ software with the thresholding method.

2.22. Hematoxylin-eosin (HE) staining

At 56 days post-injury, spinal cord and bladder tissues were collected, fixed, dehydrated, paraffin-embedded, and sectioned at 10 μm. Sections were stained with HE (Solarbio) following the manufacturer's instructions. Images were captured using a light microscope and analyzed with ImageJ software.

2.23. Statistical analysis

Statistical analysis was performed with GraphPad Prism (version 8.0.1). Unless otherwise specified, multiple comparisons were evaluated by one-way analysis of variance followed by Tukey's post-hoc test. Data were presented as the mean ± standard error of mean (SEM), and P values < 0.05 were considered to be statistically significant, i.e., ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001.

3. Results

3.1. PRGs are dynamically activated after SCI and associated with immune cell infiltration

Bioinformatics analysis offers an unbiased approach to capture temporal transcriptional alterations and identify molecular pathways that may contribute to inflammation-driven cell death after SCI [25]. To identify PRGs involved in SCI, we integrated two publicly available transcriptomic datasets (GSE5296 and GSE47681). PCA revealed substantial batch effects that were effectively corrected, resulting in a harmonized dataset (Fig. 1A). Differential expression analysis between Sham and SCI groups at 1, 3, and 7 days post-injury showed significant upregulation of key PANoptosis mediators, including Casp1, Casp3, Ripk3, Gsdmd, and Pycard (Fig. 1B). The PRGs gene set was obtained from literature and GeneCards database searches [[26], [27], [28]]. By intersecting DEGs with known PRGs, we identified 28 key genes upregulated post-SCI (Fig. 1C). GO enrichment revealed the upregulated PRGs were enriched in pathways related to IL-1β production, inflammatory responses, cytokine signaling, and canonical inflammasome complexes (Fig. 1D), underscoring the pivotal role of inflammatory cell death in SCI pathology. To further explore the temporal pattern of PANoptosis activation and its relationship with the immune microenvironment, we calculated PANoptosis scores using GSVA and assessed immune cell infiltration via the xCell algorithm. PANoptosis scores showed time-dependent increases, particularly on days 3 and 7, and correlated with elevated infiltration of M1 macrophages, monocytes, dendritic cells (DCs), and T cells (Figs. 1E and F). These findings suggest a link between PANoptotic activity and macrophages/microglia during SCI.

Fig. 1.

Identification of PANoptosis related genes after spinal cord injury (SCI) and analysis of cell infiltration in the spinal cord microenvironment. PANoptosis-related genes (PRGs) are significantly upregulated after SCI and are closely associated with macrophages/microglia in the injured spinal cord microenvironment. (A) Principal component analysis (PCA) of GSE5296 and GSE47681 datasets pre- and post-batch effect correction: pre-correction (left) and post-correction (right). (B) Volcano plots of differentially expressed genes (DEGs) in the Sham-operated group versus the three groups at 1, 3, and 7 days after SCI. The labels indicate DEGs associated with PANoptosis (P value < 0.05 and |log2(fold change)| > 1). (C) Intersection of PANoptosis related genes and upregulated DEGs in Fig. 1B. (D) Gene Ontology (GO) enrichment analysis of 28 key PANoptosis related genes in Fig. 1C. (E) PANoptosis score (upper panel) and cellular infiltration abundance assessment (lower panel) of transcriptomic data at different time points after SCI using gene set variation analysis (GSVA) and xCell algorithms, respectively. (F) Spearman's correlation analysis of PANoptosis score and cell infiltration abundance. PC: principal component; aDC: activated dendritic cell (DC); NK: natural killer cell; NKT: natural killer T cell; Tregs: regulatory T cells.

3.2. Identification of PANoptosis-associated gene module via WGCNA

Exploration of PANoptosis-related transcriptional regulation following SCI was carried out using WGCNA on the integrated transcriptome. A total of 16 distinct gene co-expression modules were identified through hierarchical clustering and dynamic tree cut algorithms (Fig. 2A). The turquoise module exhibited the strongest correlation with PANoptosis, as well as significant associations with pyroptosis, necroptosis, and apoptosis, highlighting its potential involvement in regulated cell death (Fig. 2B). The turquoise module contained 529 genes, including Casp8, Bnip3, Ccr5, Alox5ap, and Camk2b, with high module membership and significance for PANoptosis (r = 0.9822, P < 0.001) (Figs. 2C and D). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis revealed that genes in this module were predominantly involved in the mitogen-activated protein kinases (MAPK), phosphatidylinositol 3-kinase/protein kinase B (PI3K-Akt), and AMPK signaling pathways. GO enrichment further highlighted their participation in the regulation of MAPK signaling, ROS metabolism, Wnt signaling, and calcium ion transmembrane transport (Figs. 2E and F). Notably, many of the signaling pathways enriched in this module, including MAPK, PI3K-Akt, Wnt, nuclear factor κB (NF-κB), and mammalian target of rapamycin pathways, have been reported in prior studies to be subject to AMPK-dependent regulation [29,30]. In addition, processes such as ROS metabolism, calcium transport, cytoplasmic metabolic regulation, glucose metabolism, and autophagy are known to be regulated in an AMPK-dependent manner in the literature [12,13]. Given its established role in modulating neuroinflammation and programmed cell death, as well as its translational potential, AMPK was prioritized for mechanistic investigation in this study. Correlation analysis demonstrated a significant inverse association between AMPK pathway activity and PANoptosis scores (r = −0.46, P < 0.001), suggesting a potential inhibitory role of AMPK signaling in PANoptotic cell death (Fig. 2G).

Fig. 2.

Weighted gene co-expression network analysis (WGCNA)-identified hub gene module correlates with PANoptosis and immune cell infiltration profiles. WGCNA revealed a turquoise module strongly associated with PANoptosis signatures, adenosine monophosphate (AMP)-activated protein kinase (AMPK) signaling, and macrophages/microglia in the injured spinal cord. (A) Hierarchically clustered dendrogram constructed using the dynamic tree cut algorithm identifies 16 gene co-expression modules. Color bars below denote module assignments. (B) Heatmap visualization of module-trait correlations. Values represent correlation coefficients and P-values between module genes and phenotypic traits. (C) Network construction of genes in the turquoise module. (D) Scatterplot showing the correlation between module members in the turquoise module and gene significance for PANoptois. (E) Bubble plots illustrating significantly enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (left) and Gene Ontology (GO) terms (right) for the turquoise module genes. (F) Network construction of genes and enriched KEGG pathways in the turquoise module. (G) Spearman's correlation analysis of PANoptosis score calculated by the gene set variation analysis (GSVA) algorithm with AMPK signaling pathway. (H) Bipartite correlation analysis of PANoptosis score and AMPK signaling pathway: correlation matrix among PANoptosis score (quantified by GSVA), AMPK signaling pathway, and expression levels of 28 PANoptosis-associated hub genes (left) and correlation matrix among PANoptosis score, AMPK signaling pathway, and immune cell infiltration abundance (xCell algorithm-derived) (right). The asterisks indicate the statistical significance of Spearman's correlations among PANoptosis-related genes (PRGs) in the left heatmap and among cell types in the right heatmap. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. MAPK: mitogen-activated protein kinases. PI3K-Akt: phosphatidylinositol 3-kinase-protein kinase B; mTOR: mammalian target of rapamycin; cGMP-PKG: cyclic guanosine monophosphate-protein kinase G; HIF-1: hypoxia-inducible factor-1; NF-κB: nuclear factor κB; ErbB: erythroblastic leukemia viral oncogene homolog; CI: confidence interval; NK: natural killer; aDC: activated dendritic cell (DC); NKT: NK T cell.

To further dissect the relationship between PANoptosis, AMPK signaling, and the immune microenvironment, we performed integrative correlation analyses. The results revealed that AMPK signaling was negatively correlated with several key PRGs, including Casp3, Ripk1, Casp1, Aim2, and Pycard, and inversely associated with the infiltration of monocytes/macrophages and astrocytes. Conversely, PANoptosis activity was positively correlated with the abundance of these pro-inflammatory cell types (Fig. 2H). Collectively, these findings suggest that the AMPK signaling pathway may serve as a critical negative regulator of macrophage/microglia-mediated PANoptosis following SCI, potentially modulating the neuroinflammatory milieu and contributing to neural repair.

3.3. Metformin attenuates inflammatory cytokine production and M1 polarization in BV2 microglia via AMPK/NF-κB signaling

The regulatory involvement of AMPK signaling in PANoptosis-associated inflammation was validated using an in vitro model, where BV2 cells were activated with LPS and ATP [31]. Co-localization immunofluorescence staining for NOD-like receptor protein 3 (NLRP3), cleaved caspase-3, and receptor-interacting protein kinase 3 (RIPK3) in BV2 cells showed these markers were co-expressed within the same cells following LPS + ATP treatment (Fig. S1). Cells were pretreated with metformin (an AMPK activator) or CC (an AMPK inhibitor) (Fig. S2A). LDH release assays revealed that metformin significantly reduced cytotoxicity in BV2 cells, whereas CC reversed this protective effect (Fig. S2B). Given the close interplay between AMPK and NF-κB signaling, and the central role of NF-κB in driving the expression of pro-inflammatory cytokines [32], we assessed the activation status of the NF-κB pathway in our study. Western blot analysis demonstrated that metformin markedly increased the phosphorylation of AMPK and concomitantly suppressed phosphorylation of NF-κB subunit p65 and IκB, indicating inhibition of NF-κB pathway activation (Figs. S2C–F). Consistent with protein expression findings, RT-qPCR and ELISA assays showed that metformin significantly downregulated the transcription and secretion of pro-inflammatory cytokines including IL-1β, TNF-α, IL-6, and IL-18, while CC abolished these effects (Figs. S2G–K). These findings confirm that metformin modulates inflammatory cytokine production in microglia via AMPK-dependent suppression of NF-κB signaling.

As AMPK signaling has been closely associated with microglial polarization [33,34], we hypothesized that metformin may exert its anti-inflammatory effects, at least in part, by regulating microglial phenotypic transition. To test this, M1 polarization was induced by LPS and IFN-γ, whereas M2 polarization was induced by IL-4, followed by flow cytometric analysis of M1 (CD86⁺) and M2 (CD206⁺) marker expression (Fig. S2L). Metformin significantly decreased the proportion of CD86⁺ F4/80⁺ M1 cells and increased CD206⁺ F4/80⁺ M2 cells (Figs. S2M–O). And CC abrogated the effects of metformin. Taken together, these findings suggest that metformin alleviates PANoptosis-associated inflammatory responses and M1 microglial polarization via modulation of the AMPK/NF-κB signaling axis.

3.4. Metformin attenuates microglial pyroptosis

To determine whether metformin attenuates microglial pyroptosis via AMPK signaling, we examined the expression of canonical pyroptotic markers in BV2 microglia. Immunofluorescence staining revealed that metformin significantly reduced the expression levels of NLRP3 and gasdermin D (GSDMD), while these reductions were reversed by CC (Figs. 3A–C). Furthermore, apoptosis-associated speck-like protein containing a CARD (ASC) speck formation, a hallmark of inflammasome assembly, was markedly increased in LPS + ATP-treated cells, and notably diminished by metformin treatment (Figs. 3D and E). The protective effect was abolished when CC was co-administered, underscoring the critical role of AMPK activation in inflammasome activation. Flow cytometry analysis using FLICA staining demonstrated that metformin substantially decreased the proportion of caspase-1-positive BV2 cells, whereas inhibition of AMPK with CC significantly increased caspase-1 activation (Figs. 3F and G). Consistently, Western blot analysis showed that metformin decreased the expression levels of NLRP3, full-length GSDMD, and caspase-1 in cell lysates, and reduced the abundance of GSDMD-N-terminal domain (NT), cleaved caspase-1, and IL-1β in the culture supernatants (Figs. 3H–N). These effects were reversed upon co-treatment with CC, further validating the involvement of AMPK in metformin-mediated suppression of pyroptosis. Collectively, these findings suggest that metformin effectively suppresses microglial pyroptosis through activation of the AMPK pathway, thereby inhibiting inflammasome assembly and downstream caspase-1-dependent inflammatory responses.

Fig. 3.

Metformin inhibits microglial pyroptosis through adenosine monophosphate (AMP)-activated protein kinase (AMPK) signaling. Metformin significantly suppresses pyroptotic activation in BV2 cells by reducing inflammasome assembly and caspase-1 activation in an AMPK-dependent manner. (A–C) Representative fluorescence images of BV2 cells immunostained for NOD-like receptor protein 3 (NLRP3) (red) and gasdermin D (GSDMD) (cyan) (A), and quantitative analysis of the relative fluorescence intensity of NLRP3 (B) and GSDMD (C) (n = 6–7). Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). The relative fluorescence intensity was expressed as fold changes compared to the phosphate-buffered saline (PBS) group. (D, E) Representative fluorescence images of BV2 cells immunostained for apoptosis-associated speck-like protein containing a CARD (ASC) (green) (D), and quantitative analysis (E) of the proportion of cells containing ASC specks (n = 6). The white squares in Fig. 3D indicate representative cells selected for high-magnification views shown in the enlarge panels, and the white arrows indicate ASC specks. Cell nuclei were stained with DAPI (blue). (F, G) Representative flow cytometry plots (F) and quantitative analysis (G) of the proportion of caspase-1-positive cells in BV2 cells (n = 3). (H–N) Representative Western blot images (H) and quantitative analysis of expression levels of NLRP3 (I), GSDMD (J) and caspase-1 (K) in BV2 cell lysates and cleaved caspase-1 (L), GSDMD-N-terminal domain (NT) (M), and interleukin-1β (IL-1β) (N) in culture supernatants (n = 3). The relative expression level of target proteins in cell lysates was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and then calculated as fold changes compared to the lipopolysaccharide (LPS) + adenosine triphosphate (ATP) group. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. Met: metformin group; CC: compound C; SSC-H: side scatter-height; FLICA: fluorochrome-labeled inhibitor of caspases.

3.5. Metformin suppresses microglial apoptosis

To evaluate whether metformin mitigates microglial apoptosis through AMPK activation, we evaluated key apoptotic markers in BV2 cells. TUNEL staining revealed a significant increase in apoptotic cells following LPS and ATP stimulation, while metformin treatment markedly reduced the proportion of TUNEL-positive cells (Figs. 4A and B). Consistently, immune-fluorescence analysis of cleaved caspase-3, a central effector of apoptosis, demonstrated that metformin significantly decreased the proportion of cleaved caspase-3-positive BV2 cells compared to the LPS + ATP group, an effect abolished by CC administration (Figs. 4C and D). Western blot analysis corroborated these findings, showing that metformin restored the expression of anti-apoptotic protein B-cell lymphoma 2 (Bcl-2) and caspase-3, and reduced the levels of pro-apoptotic proteins Bcl-2-associated X protein (Bax) in BV2 cells exposed to LPS + ATP (Figs. 4E–H). These molecular changes were reversed upon AMPK inhibition. Together, these results suggest that metformin effectively suppresses microglial apoptosis induced by inflammatory stimuli via activation of AMPK signaling, thus preserving microglial cell survival.

Fig. 4.

Metformin inhibits microglial apoptosis through adenosine monophosphate (AMP)-activated protein kinase (AMPK) signaling. Metformin reduces apoptosis in BV2 cells by decreasing caspase-3 activation and modulating B-cell lymphoma 2 (Bcl-2) family protein expression in an AMPK-dependent manner. (A, B) Representative TdT-mediated dUTP nick-end labeling (TUNEL) staining images (A) and quantitative analysis (B) of the proportions of TUNEL-positive cells in BV2 cells (n = 8). Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). (C, D) Representative fluorescence images (C) of BV2 cells immunostained for cleaved-caspase-3 (green), and quantitative analysis (D) of the proportion of cleaved-caspase-3-positive cells in BV2 cells (n = 10). Cell nuclei were stained with DAPI (blue). (E–H) Representative Western blot images (E) and quantitative analysis of expression levels of Bcl-2 (F), Bcl-2-associated X protein (Bax) (G), and caspase-3 (H) in BV2 cells (n = 3). The relative expression level of target proteins was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH)/β-tubulin and then calculated as fold changes compared to the lipopolysaccharide (LPS) + adenosine triphosphate (ATP) group. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. PBS: phosphate-buffered saline; Met: metformin group; CC: compound C.

3.6. Metformin inhibits microglial necroptosis

To explore whether metformin regulates microglial necroptosis through AMPK activation, we evaluated the expression of necroptosis-related markers in BV2 cells following LPS and ATP stimulation. Consistently, flow cytometry analysis revealed that LPS + ATP markedly increased both necrotic (PI⁺) and apoptotic (annexin V⁺/PI⁻) cell populations. Metformin treatment significantly reduced the proportion of necrotic and apoptotic BV2 cells, whereas co-treatment with CC largely reversed this protective effect (Figs. 5A–C). Immunofluorescence staining revealed that LPS + ATP stimulation markedly elevated RIPK3 expression in BV2 cells, while metformin significantly attenuated RIPK3 signal intensity (Figs. 5D and E). In addition, Western blot analysis demonstrated that metformin reduced the expression of necroptosis-associated proteins RIPK1, RIPK3, and mixed lineage kinase domain-like (MLKL) in BV2 cells stimulated with LPS + ATP (Figs. 5F–I). However, these effects were reversed by CC, indicating that AMPK activation is essential for metformin-mediated inhibition of necroptosis. These results suggest that metformin effectively inhibits microglial necroptosis via activation of AMPK signaling, thereby contributing to the maintenance of microglial viability under inflammatory conditions.

Fig. 5.

Metformin inhibits microglial necroptosis through adenosine monophosphate (AMP)-activated protein kinase (AMPK) signaling. Metformin reduces necroptosis in BV2 cells by suppressing receptor-interacting protein kinase 1 (RIPK1)/RIPK3/mixed lineage kinase domain-like (MLKL) expression in an AMPK-dependent manner. (A–C) Representative annexin V-fluorescein isothiocyanate (FITC)/propidium iodide (PI) flow cytometry plots (A) and quantitative analysis of the proportion of necrotic (propidium iodide (PI⁺)) (B) and apoptotic (annexin V⁺/PI⁻) (C) cells in BV2 cells (n = 3). (D, E) Representative fluorescence images (D) of BV2 cells immunostained for RIPK3 (cyan), and quantitative analysis (E) of the relative fluorescence intensity of RIPK3 (n = 5). Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). The relative fluorescence intensity was expressed as fold changes compared to the PBS group. (F–I) Representative Western blot images (F) and quantitative analysis of expression levels of RIPK1 (G), RIPK3 (H), and MLKL (I) in BV2 cells (n = 3). The relative expression level of target proteins was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and then calculated as fold changes compared to the lipopolysaccharide (LPS) + adenosine triphosphate (ATP) group. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. PBS: phosphate-buffered saline; Met: metformin group; CC: compound C.

3.7. Subacute low-dose metformin administration enhances motor recovery without inducing systemic side effects after SCI

To identify the optimal therapeutic window and dosage for metformin intervention following SCI, we administered low or high doses of metformin intraperitoneally at 0, 3, or 7 days post-injury, and assessed motor function recovery at 56 days post-injury using BBB score and CatWalk gait analysis (Fig. S3A). Rats treated with low-dose metformin at 3 days post-injury showed significantly improved BBB scores compared to other groups (Fig. S3B). CatWalk gait analysis further substantiated these findings, showing that the 3 days post-injury low-dose group demonstrated enhanced locomotor coordination and stability (Fig. S3C). Specifically, improvements were observed in stand time, maximal contact area, regularity index, and paw print positioning (Figs. S3D–G), reflecting better interlimb coordination and more consistent gait patterns.

In addition, long-term administration of low-dose metformin was well tolerated, as evidenced by the absence of significant changes in body weight or fasting blood glucose levels among all treatment groups (Fig. S4). These findings suggest that subacute administration of low-dose metformin, particularly when initiated at 3 days post-injury, maximizes functional recovery after SCI without causing adverse metabolic effects.

3.8. Metformin attenuates inflammatory cascades and microglia/macrophage M1 polarization after SCI via AMPK/NF-κB signaling

In order to investigate the anti-inflammatory mechanisms of metformin following SCI, we examined the AMPK/NF-κB signaling axis and associated inflammatory mediators. Based on the above results, rats received daily intraperitoneal injections of low-dose metformin, CC, or their combination, with therapeutic intervention initiated on day 3 post-injury to target secondary injury processes. Spinal cord samples were collected at 7 and 56 days post-injury, and motor function was evaluated 56 days post-injury (Fig. S5A). Western blot results demonstrated that SCI led to significant reductions in phosphorylated AMPK and concurrent increases in phosphorylation of p65 and IκB, indicating NF-κB activation. Metformin treatment markedly increased p-AMPK levels and reduced p-p65 and p-IκB, effects that were abolished by co-treatment with CC (Figs. S5B–E). Consistent with these findings, RT-qPCR analysis revealed that metformin significantly downregulated the messenger RNA (mRNA) expression of pro-inflammatory cytokines IL-1β, IL-6, TNF-α, and IL-18 compared to the injury group, and these effects were reversed by CC (Fig. S5F). Protein array heatmap analysis and corresponding quantitative evaluations corroborated these results, revealing reduced levels of IL-1β, IL-6, TNF-α, and IFN-γ following metformin treatment (Figs. S5G–K). ELISA further validated the decreased IL-18 protein levels in the metformin-treated group, an effect again negated by CC co-administration (Fig. S5L).

Immunofluorescence staining of spinal cord sections revealed that metformin enhanced arginase 1 (Arg1, a marker of M2 polarization) and suppressed inducible nitric oxide synthase (iNOS, a marker of M1 polarization) expression in CD68-positive microglia/macrophages, indicative of a shift from pro-inflammatory M1 to anti-inflammatory M2 polarization (Figs. S5M–O). Since Iba-1 and CD68 labeling do not allow clear distinction between resident microglia and infiltrating macrophages, the term microglia/macrophages is used operationally throughout this study to avoid overinterpretation. Collectively, these data suggest that metformin attenuates SCI-induced inflammation and microglia/macrophage M1 polarization via activation of AMPK and inhibition of the NF-κB pathway.

3.9. Metformin inhibits microglia/macrophage pyroptosis after SCI via AMPK-dependent regulation

To further investigate the role of metformin in regulating pyroptosis following SCI, we assessed the expression of key pyroptosis-related protein. Western blot analysis revealed that SCI markedly increased the protein levels of NLRP3, NIMA-related kinase 7 (NEK7), ASC, caspase-1, cleaved caspase-1, GSDMD, GSDMD-NT, pro-IL-1β, and IL-1β compared to the Sham group (Fig. 6A). Metformin treatment significantly reduced the expression of these proteins, indicating suppression of inflammasome activation and pyroptosis (Figs. 6B–I). These inhibitory effects were reversed by CC co-treatment, highlighting a critical role of AMPK signaling in metformin-mediated anti-pyroptotic activity. Furthermore, immunofluorescence staining showed that NLRP3 was predominantly localized in Iba-1⁺ microglia/macrophages (Fig. 6J). Metformin markedly decreased the fluorescence intensity of NLRP3 and GSDMD in Iba-1⁺ cells, while the presence of CC negated this effect (Figs. 6K–M). Together, these findings indicate that metformin suppresses inflammasome activation and subsequent pyroptosis in microglia/macrophages through AMPK-dependent mechanisms, thereby contributing to its neuroprotective effects in the context of SCI.

Fig. 6.

Metformin inhibits microglia/macrophage pyroptosis after spinal cord injury (SCI) via adenosine monophosphate (AMP)-activated protein kinase (AMPK)-dependent regulation. Metformin reduces pyroptosis in microglia/macrophages by suppressing NOD-like receptor protein 3 (NLRP3) inflammasome activation and downstream caspase-1/gasdermin D (GSDMD) signaling in an AMPK-dependent manner following SCI. (A–I) Representative Western blot images (A) and quantitative analysis of expression levels of NLRP3 (B), NIMA-related kinase 7 (NEK7) (C), caspase-1 (D), cleaved- caspase-1 (E), GSDMD (F), GSDMD-N-terminal domain (NT) (G), pro-interleukin-1β (IL-1β) (H), and IL-1β (I) of the spinal cord at 7 days post-injury (n = 3). The relative expression level of target proteins was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH)/β-tubulin and then calculated as fold changes compared to the injury group. (J, K) Representative fluorescence images of the spinal cord immunostained for NLRP3 (red) and Iba-1 (green) at 7 days post-injury (J) and quantitative analysis of the relative fluorescence intensity of NLRP3 in Iba-1⁺ cells (K) (n = 4). Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). The relative fluorescence intensity was expressed as fold changes compared to the injury group. (L, M) Representative fluorescence images of the spinal cord immunostained for GSDMD (red) and Iba-1 (green) at 7 days post-injury (L) and quantitative analysis of the relative fluorescence intensity of GSDMD in Iba-1⁺ cells (M) (n = 4). Cell nuclei were stained with DAPI (blue). The relative fluorescence intensity was expressed as fold changes compared to the injury group. The white squares in Figs. 6J and L indicate representative regions selected for high-magnification views shown in the enlarged panels. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. Met: metformin group; CC: compound C; ASC: apoptosis-associated speck-like protein containing a CARD.

3.10. Metformin inhibits microglia/macrophage apoptosis after SCI via AMPK-dependent regulation

To investigate the anti-apoptotic effects of metformin after SCI, we evaluated the expression of apoptosis-related proteins at 7 days post-injury. Western blot analysis showed a significant increase in Bcl-2 and caspase-3 expression in the Met group compared to the injury and Met + CC groups (Figs. 7A–C). Conversely, the Bax protein expression was significantly decreased in the Met group compared to the injury and Met + CC groups (Fig. 7D). To evaluate the effect of metformin on microglial/macrophage apoptosis, TUNEL staining was performed on spinal cord sections at 7 days post-injury. Quantification of TUNEL-positive cells revealed a significant reduction in apoptotic microglia/macrophages in the Met group compared to the injury group, while CC co-treatment negated this effect (Figs. 7E and F). Additionally, cleaved caspase-3, another key indicator of apoptosis, was assessed in Iba-1⁺ cells using fluorescence immunostaining. Metformin markedly decreased the fluorescence intensity of cleaved caspase-3 in Iba-1⁺ cells, while the presence of CC negated this reduction (Figs. 7G and H). Taken together, these results suggest that metformin treatment effectively inhibits microglial/macrophage apoptosis following SCI, an effect that appears to be mediated via AMPK-dependent pathways.

Fig. 7.

Metformin inhibits microglia/macrophage apoptosis after spinal cord injury (SCI) via adenosine monophosphate (AMP)-activated protein kinase (AMPK)-dependent regulation. Metformin reduces apoptosis in microglia/macrophages by modulating B-cell lymphoma 2 (Bcl-2) family proteins and caspase-3 activation in an AMPK-dependent manner following SCI. (A–D) Representative Western blot images (A) and quantitative analysis of expression levels of Bcl-2 (B), caspase-3 (C), and Bcl-2-associated X protein (Bax) (D) of the spinal cord at 7 days post-injury (n = 3). The relative expression level of target proteins was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH)/β-tubulin and then calculated as fold changes compared to the injury group. (E, F) Representative fluorescence images of the spinal cord immunostained for TdT-mediated dUTP nick-end labeling (TUNEL) (red) and Iba-1 (green) at 7 days post-injury (E) and quantitative analysis of the proportions of TUNEL-positive cells in Iba-1⁺ cells (F) (n = 4). Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). (G, H) Representative fluorescence images of the spinal cord immunostained for cleaved-caspase-3 (red) and Iba-1 (green) at 7 days post-injury (G) and quantitative analysis of the relative fluorescence intensity of cleaved-caspase-3 in Iba-1⁺ cells (H) (n = 4). Cell nuclei were stained with DAPI (blue). The relative fluorescence intensity was expressed as fold changes compared to the injury group. The white squares in Figs. 7E and G indicate representative regions selected for high-magnification views shown in the enlarged panels. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. Met: metformin group; CC: compound C.

3.11. Metformin inhibits microglia/macrophage necroptosis after SCI via AMPK-dependent regulation

To explore the effect of metformin on necroptosis after SCI, we examined the expression levels of key necroptosis-associated proteins RIPK1, RIPK3, and MLKL. Western blot analysis revealed a significant upregulation of RIPK1, RIPK3, and MLKL protein levels in the injury and CC groups compared to the Met group (Figs. 8A–D). However, co-treatment with CC partially reversed these effects, suggesting the involvement of AMPK signaling in mediating the anti-necroptotic action of metformin. Immunofluorescence staining further confirmed these findings. Co-localization of RIPK3 or MLKL with the microglia/macrophage marker Iba-1 demonstrated metformin treatment significantly decreased the relative fluorescence intensity of RIPK3 and MLKL in Iba-1⁺ cells, whereas this reduction was attenuated by CC co-treatment (Figs. 8E–H). These results further support the notion that metformin suppresses microglial/macrophage necroptosis in the injured spinal cord through AMPK-dependent signaling pathways.

Fig. 8.

Metformin inhibits microglia/macrophage necroptosis after spinal cord injury (SCI) via adenosine monophosphate (AMP)-activated protein kinase (AMPK)-dependent regulation. Metformin reduces necroptosis in microglia/macrophages by suppressing receptor-interacting protein kinase 1 (RIPK1)/RIPK3/mixed lineage kinase domain-like (MLKL) expression in an AMPK-dependent manner following SCI. (A–D) Representative Western blot images (A) and quantitative analysis of expression levels of RIPK1 (B), RIPK3 (C), and MLKL (D) of the spinal cord at 7 days post-injury (n = 3). The relative expression level of target proteins was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and then calculated as fold changes compared to the injury group. (E, F) Representative fluorescence images of the spinal cord immunostained for RIPK3 (red) and Iba-1 (green) at 7 days post-injury (E) and quantitative analysis of the relative fluorescence intensity of RIPK3 in Iba-1⁺ cells (F) (n = 4). Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). The relative fluorescence intensity was expressed as fold changes compared to the injury group. (G, H) Representative fluorescence images of the spinal cord immunostained for MLKL (red) and Iba-1 (green) at 7 days post-injury (G) and quantitative analysis of the relative fluorescence intensity of MLKL in Iba-1⁺ cells (H) (n = 4). Cell nuclei were stained with DAPI (blue). The relative fluorescence intensity was expressed as fold changes compared to the injury group. The white squares in Figs. 8E and G indicate representative regions selected for high-magnification views shown in the enlarged panels. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. Met: metformin group; CC: compound C.

3.12. Metformin treatment significantly promoted tissue repair and axon regrowth following SCI

We next examined the impact of metformin on long-term structural recovery following SCI via MRI and histological evaluation. T₂-weighted MRI images revealed a marked reduction in hyperintense lesion area in the Met group compared to CC and injury groups, whereas the combination therapy significantly reversed the protective effect of metformin alone (Figs. 9A and B). Similarly, histological analysis using HE staining demonstrated an increased ratio of spared tissue in the Met group relative to the injury and Met + CC groups (Figs. 9C, 9D, and S6A).

Fig. 9.

Metformin enhances tissue repair and axon regrowth after spinal cord injury (SCI) via adenosine monophosphate (AMP)-activated protein kinase (AMPK)-dependent regulation. Metformin promotes tissue preservation, axonal regeneration, remyelination, decreases glial scar formation, and ameliorates bladder wall thickening in an AMPK-dependent manner. (A, B) Representative T₂-weighted magnetic resonance imaging (A) of the spinal cord in the sagittal plane at 56 days post-injury, and quantitative analysis (B) of the area of hyperintense lesions (n = 3). (C, D) Representative Hematoxylin-eosin (HE) images of spinal cord sections at 56 days post-injury (C) and quantitative analysis of the ratio of spared tissue (D) (n = 3). (E–G) Representative fluorescence images of the spinal cord immunostained for neurofilament 200 (NF200) (red) and myelin basic protein (MBP) (green) at 56 days post-injury (E) and quantitative analysis of the relative fluorescence intensity of NF200 (F) and MBP (G) at lesion center (n = 3). Cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue). The relative fluorescence intensity was expressed as fold changes compared to the injury group. (H–J) Representative fluorescence images of the spinal cord immunostained for NF200 (red) and glial fibrillary acidic protein (GFAP) (green) at 56 days post-injury (H) and quantitative analysis of NF200⁺ (I) and GFAP⁺ (J) areas (n = 3). Cell nuclei were stained with DAPI (blue). (K–M) Representative fluorescence images B-cell lymphoma 2 (Bcl-2)-associated X protein (Bax) of the spinal cord immunostained for 5-hydroxytryptamine (5-HT) (cyan) and CS56 (grey) at 56 days post-injury (K) and quantitative analysis of serotonergic axons density (L) in the spinal cord from rostral to caudal relative to the injury site and CS56⁺ area (M) (n = 3). Cell nuclei were stained with DAPI (blue). The white squares in Figs. 9E, H, and K indicate representative regions selected for high-magnification views shown in the enlarged panels. The white arrows in Fig. 9E indicate NF200⁺ and MBP⁺ myelinated axons in the lesion center. (N, O) Representative HE images of bladder sections (N) and quantitative analysis of the thickness of bladder wall (O) (n = 6). Injury vs. Met: ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001; Met vs. Met + CC: ^#P < 0.05 and ^###P < 0.001. ns: no significance. Met: metformin group; CC: compound C.

Immunofluorescence analysis at the lesion center and adjacent rostral and caudal segments demonstrated increased neurofilament 200 (NF200) and myelin basic protein (MBP) expression in the Met group, indicative of enhanced axonal preservation and myelination. In contrast, the Met + CC group showed diminished levels of both markers, resembling the CC and injury groups (Figs. 9E–9G and S6B–S6G). Furthermore, immunostaining of longitudinal spinal cord sections revealed increased NF200-positive axons and decreased glial fibrillary acidic protein (GFAP)-positive reactive astrocytes in the Met group compared to others, whereas CC co-treatment abrogated these effects (Figs. 9H−J). To assess descending serotonergic axon regeneration, immunostaining for 5-hydroxytryptamine (5-HT) and CS56 (a marker of chondroitin sulfate proteoglycans (CSPGs)) demonstrated improved serotonergic axon density in the Met group at 56 days post-injury, particularly in the rostral regions and lesion core (Figs. 9K and L). Unfortunately, few 5-HT axons in the metformin group crossed the injury center to the caudal side. In parallel, the Met group exhibited a significant reduction in CS56-positive areas, suggesting a more permissive environment for axon regrowth, which was again attenuated by CC (Figs. 9K and M). While metformin markedly reduced lesion cavity size and attenuated glial scar formation, these structural and biochemical barriers were not completely eliminated, likely limiting axonal extension across the lesion. Additional strategies, such as targeting CSPGs or other inhibitory factors, may be required to achieve long-distance axonal regeneration.

Given that muscle denervation and urinary retention after SCI may lead to stretching and thinning of bladder muscle fibers, HE staining of bladder tissue were performed. Histological analysis of bladder tissue showed a significant increase in bladder wall thickness in the Met group compared to the injury group, suggesting a potential benefit in functional recovery (Figs. 9N and O). Collectively, these results indicate that metformin promotes tissue repair, axon regeneration, and glial remodeling after SCI, potentially through AMPK-dependent mechanisms.

3.13. Metformin enhances long-term locomotor and electrophysiological recovery after SCI via AMPK activation

Whether AMPK signaling contributes to metformin-induced functional recovery was addressed through behavioral and electrophysiological monitoring in a rat SCI model. Metformin administration significantly improved hindlimb function, as evidenced by higher BBB scores from day 14 onward compared with injury and CC groups (Fig. 10A). At 56 days post-injury, rats in the Met group achieved mean BBB scores of approximately 12, whereas scores in the injury and CC groups remained below 10; co-treatment with CC attenuated metformin's benefit (Fig. 10B). Consistent with BBB findings, LSS scores at 56 days post-injury were significantly higher in the Met group than in injury or CC group, indicating enhanced voluntary hindlimb kicking and coordination; this improvement was reversed by CC (Figs. 10C and D). CatWalk gait analysis revealed that metformin treatment restored interlimb coordination and paw placement, while rats in other groups showed more dragging (Fig. 10E). Stand time, maximum contact area, and regularity index in the Met group were all significantly increased compared to injury or CC rats. Additionally, print position analysis indicated more consistent and accurate paw placement in the Met group (Figs. 10F–I). Notably, CC co-administration partially abolished these metformin-induced gait improvements. Moreover, MEP recordings at 56 days post-injury demonstrated metformin shortened MEP latency and increased amplitude relative to injury and CC groups, reflecting improved neural conduction (Figs. 10J–L). And these electrophysiological gains were lost in the Met + CC group. Together, these data confirm that metformin promotes robust functional and electrophysiological recovery after SCI primarily via AMPK-dependent mechanisms.

Fig. 10.

Metformin promotes functional recovery after spinal cord injury (SCI) via adenosine monophosphate (AMP)-activated protein kinase (AMPK)-dependent regulation. Metformin improves locomotor performance, gait coordination, and electrophysiological function mediated by AMPK activation following SCI. (A, B) Basso, Beattie, and Bresnahan (BBB) scores at different time points (A) and on day 56 (B) after SCI (n = 9). (C, D) Representative swimming images (C) and quantitative analysis (D) of Louisville swim scale (LSS) scores at 56 days post-injury (DPI) (n = 9). (E) Representative images of CatWalk footprint views and timing views at 56 days post-injury. (F–I) Quantitative analysis of stand (F), max contact area (G), regularity index (H), and print position (I) in CatWalk footprints (n = 9). (J–L) Representative motor evoked potential (MEP) images at 56 days post-injury (J) and quantitative analyses of the latency (K) and amplitude (L) of MEP (n = 9). The comparison of BBB scores at different time points was conducted using two-way analysis of variance (ANOVA) with Tukey's post-hoc test. Other analyses were conducted using one-way ANOVA with Tukey's post hoc test. Injury vs. Met: ^∗∗P < 0.01 and ^∗∗∗P < 0.001; Met vs. Met + CC: ^#P < 0.05 and ^##P < 0.01. Met: metformin group; CC: compound C.

4. Discussion

Maintaining microglia homeostasis following SCI plays a critical role in supporting neuronal survival [4]. In this study, we performed integrative bioinformatics analyses on two publicly available transcriptomic datasets and identified the critical involvement of PANoptosis, characterized by pyroptosis, apoptosis, and necroptosis, in microglia/macrophage dysfunction following SCI, which was inversely correlated with AMPK signaling activity. Pharmacological activation of AMPK by metformin effectively suppressed microglia PANoptosis, thereby reducing the production of pro-inflammatory cytokines. This alleviation of the inhibitory microenvironment contributed to enhanced axonal regeneration, remyelination, and reduced glial scar formation, ultimately promoting functional recovery post-SCI. Notably, the beneficial effects of metformin were partially or completely abrogated by co-treatment with CC, a well-established AMPK inhibitor. Collectively, our findings reveal a novel regulatory role of AMPK signaling in microglial PANoptosis and uncover a previously unrecognized mechanism by which metformin facilitates neural repair after SCI.

Microglia and infiltrating macrophages are rapidly recruited and activated at the lesion site following SCI. Programmed cell death of immune cells releases damage-associated molecular patterns, which amplify local inflammation by recruiting more immune populations and contribute to irreversible neuronal damage. PANoptosis, a recently defined form of inflammatory cell death that integrates pyroptosis, apoptosis, and necroptosis, has been shown to play a crucial role in various inflammatory and neurotraumatic conditions [35]. Accumulating evidence has demonstrated that neuronal PANoptosis occurs after SCI and can be effectively mitigated by pharmacological agents, trace elements, stem cell therapies, and biomaterial-based interventions [8,10]. For instance, Wu and co-workers [9] reported that tripartite motif containing 56 interacts with and promotes ubiquitination of Y-box-binding protein 1, thereby attenuating neuronal PANoptosis and improving neurological outcomes following SCI. However, the specific contribution of PANoptosis to traumatic CNS injury and its cell-type specificity within the post-injury microenvironment have remained largely undefined. Our study provides novel insights by identifying microglia/macrophage as the principal cell population undergoing PANoptosis after SCI, as evidenced by the concurrent upregulation of canonical markers including Casp1, Casp3, Ripk3, Gsdmd, and Pycard. The enrichment of PANoptotic signaling in microglia/macrophage suggests that these cells may act as upstream amplifiers of inflammation-mediated secondary injury, underscoring the therapeutic value of targeting microglial PANoptosis in SCI.

Numerous molecular mechanisms have been identified that facilitate functional recovery after SCI. For example, neuronal phosphatase and tensin homolog deletion enhances intrinsic axonal regeneration capacity [36], while microglial Plexin-B2 has been shown to restrict inflammatory spread and promote wound healing following SCI [37]. As a central metabolic sensor, AMPK not only maintains cellular metabolism and homeostasis under physiological conditions as well as diverse pathological stimulus [11,34]. However, the precise role of AMPK in SCI remains controversial. Several studies have demonstrated that activation of AMPK promotes axonal regeneration and neuroprotection after SCI [38]. For instance, Chen and co-workers [39] reported that AMPK activation upregulates sirtuin 3, thereby maintaining mitochondrial homeostasis and enhancing neuronal survival. In contrast, Giovanni and co-workers [40] found that conditional deletion of AMPKα1 enhanced axonal growth by activating multiple regenerative signaling pathways at the injury site. In our study, pharmacological activation of AMPK by metformin significantly attenuated microglial PANoptosis and the expression of associated pro-inflammatory cytokines, potentially through suppression of NF-κB signaling. Additionally, metformin treatment enhanced axonal regeneration and locomotor recovery after SCI, while these beneficial effects were attenuated by co-administration of the AMPK inhibitor CC, further supporting the beneficial role of AMPK activation. These results are in line with prior findings that highlight AMPK's role in dampening inflammation and regulating cell death, and further extend its significance to the control of PANoptosis in neuroinflammatory conditions. Nevertheless, whether AMPK exerts divergent effects across distinct neural cell types warrants further investigation.

Metformin, a widely prescribed anti-diabetic agent, has recently drawn increasing attention for its immunomodulatory and neuroprotective properties beyond glucose regulation [41]. Previous studies indicate that metformin shifts microglia toward an anti-inflammatory phenotype by enhancing autophagy, promoting myelin debris clearance and neuroprotection after SCI [33]. However, its impact on microglial PANoptosis remains largely unexplored. In this study, we demonstrate that metformin can suppress microglial pyroptosis, apoptosis, and necroptosis, hallmarks of PANoptotic activation, and concurrently regulate microglial polarization. These immunomodulatory effects are accompanied by enhanced axonal regeneration, remyelination, and reduced glial scarring, ultimately leading to improved locomotor recovery. Furthermore, our results refine the therapeutic strategy by identifying the optimal dosing and timing of metformin administration. Through BBB scoring and CatWalk gait analysis, we demonstrate that subacute administration of low-dose metformin, particularly when initiated at 3 days post-injury, confers the greatest benefit in terms of functional recovery. Consistent with the result, our previous study demonstrated that intervention initiated at 3 days post-injury more effectively reduced cavity formation, attenuated microglia/macrophage activation, and limited neutrophil infiltration compared with immediate intervention [42]. We hypothesize that this temporal window aligns with a critical shift in microglial/macrophage behavior. In the acute phase, microglia/macrophages predominantly adopt a pro-inflammatory phenotype, and premature suppression of this response may paradoxically amplify inflammation and exacerbate neuronal and glial loss. By contrast, the subacute phase is characterized by a gradual shift of microglia/macrophages toward reparative phenotypes, which contributes to the resolution of inflammation and supports tissue repair [43,44]. The metformin dose used in this study (50 mg/kg/day) corresponds to an estimated human equivalent dose of approximately 500 mg/day, based on body surface area normalization [45]. This dosage falls within the clinically therapeutic range (500−2000 mg/day) commonly prescribed for humans, thereby supporting the translational relevance and feasibility of our experimental regimen [46]. Mechanistically, metformin has long been recognized to activate AMPK by inhibiting mitochondrial complex I, thereby decreasing intracellular ATP levels and increasing the AMP/ATP ratio [11]. More recently, presenilin enhancer 2-lysosome-dependent pathway involving the glucose-sensing protein ATP6AP1 has been proposed as an alternative mode of AMPK activation [22]. While the mechanism has been described in metabolic contexts, whether it also operates in the injured spinal cord microenvironment, particularly within microglia, remains speculative. Future studies will be required to clarify the contribution of these distinct mechanisms in SCI. Our previous studies have confirmed that metformin could promote the proliferation and differentiation of endogenous neural stem cells and inhibit ferroptosis through the AMPK pathway [21]. Current theories on PANoptosis regard it as an integrated concept that goes beyond the previous research on single cell death regulation. Previous studies mostly focused on a single form of cell death or only regulated mitochondria-dependent apoptosis, while this study explores PANoptosis as an integrated inflammatory cell death form. By detecting the coordinated expression changes of classic PANoptosis markers and combining with subsequent experiments on the simultaneous regulation of pyroptosis-apoptosis-necroptosis, we confirmed that metformin can simultaneously inhibit the coordinated activation of these three cell death forms through AMPK, rather than the isolated regulation of a single form. This systematic intervention is more in line with the actual pathological feature of cross-interaction of multiple pathological processes after SCI, and also provides a new idea to solve the problem of limited efficacy of previous single-target interventions. This study confirms that metformin, a clinically approved drug, can inhibit PANoptosis through AMPK. This not only provides a new mechanistic basis for the repurposing of this old drug in SCI treatment, but also proposes a systematic anti-inflammatory strategy centered on PANoptosis, achieving comprehensive regulation of secondary SCI injury and having higher feasibility for clinical translation.

Despite the promising findings, several limitations of this study should be acknowledged. First, our conclusions regarding the role of AMPK in microglial PANoptosis are primarily based on pharmacological interventions. Although metformin is a well-characterized AMPK activator, it exerts pleiotropic effects across multiple cell types within the injured spinal cord microenvironment [21,33,47]. Meanwhile, CC may also inhibit other kinases besides AMPK [48]. Future studies will employ conditional AMPK knockout or overexpression models to specifically manipulate AMPK signaling in microglia, thereby confirming its direct role in PANoptosis regulation following SCI. Second, infiltrating macrophages share many morphological and functional characteristics with resident microglia. Our study did not distinguish between these two populations, which may confound interpretation of cell-type–specific responses. Future investigations should adopt a multidimensional approach to better differentiate microglia from infiltrating macrophages. First, combinatorial molecular marker detection can be used to identify microglia based on their specific markers (transmembrane protein 119 (TMEM119) and purinergic receptor P2Y, G-protein coupled 12 (P2RY12)) and low CD45 expression, while labeling infiltrating macrophages by their high CD45 expression together with CCR2 or Ly6C [43]. Second, utilizing transgenic models such as Cx3cr1^CreER mice will be essential to selectively target and trace microglial populations [49], thereby refining our understanding of their specific contributions to PANoptosis and SCI pathology. In addition, it is also worth noting that sex differences may influence microglial activation and inflammatory responses following SCI [50]. Since only female rats were used in this study to minimize variability, future work should include both sexes to delineate potential sex-specific differences in PANoptosis regulation and metformin responsiveness. Meanwhile, we need to consider the potential influence of species differences on both the bioinformatics analysis and subsequent in vivo outcomes. Future studies utilizing transcriptomic or single-cell sequencing data derived from rat SCI models will further enhance the robustness and translational relevance of our findings. Lastly, further mechanistic exploration is warranted to elucidate how AMPK signaling interfaces with specific PANoptotic molecular complexes. Current evidence suggests that AMPK may suppress PANoptosis by preserving mitochondrial and redox homeostasis, enhancing autophagic clearance of damaged organelles, and maintaining calcium homeostasis (Fig. S7). These integrated regulatory effects position AMPK as a central checkpoint linking metabolic control to inflammatory cell death after SCI.

5. Conclusions

In conclusion, our study reveals a previously underrecognized role of AMPK signaling in modulating microglial PANoptosis following SCI. We demonstrate that pharmacological activation of AMPK by metformin attenuates PANoptosis-associated inflammation and promotes tissue repair and functional recovery. These findings provide new insights into the mechanisms underlying microglial-mediated secondary injury and highlight PANoptosis as a promising therapeutic target for SCI.

CRediT authorship contribution statement

Song Liu: Writing – original draft, Data curation, Conceptualization. Mi Zhou: Writing – original draft, Visualization. Cong Xing: Formal analysis, Conceptualization. Zhenxing Guo: Validation, Methodology. Qi Zhang: Validation, Methodology. Hongpeng Ma: Validation, Methodology. Hao Zhong: Validation, Formal analysis. Hongjiang Yang: Validation, Formal analysis. Guangzhi Ning: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.

Availability of data and materials

All data are available from the corresponding author on reasonable request.

Declaration of competing interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article.