Healthy adult female Sprague-Dawley rats (180–220 g, 8 weeks old) were obtained from the Animal Center of Nantong University. Animals were housed under standard conditions (12 h light/dark cycle, 23 ± 1°C) with ad libitum access to food and water, followed by a 1 week acclimation period. All procedures strictly adhered to the NIH Guide for the Care and Use of Laboratory Animals (8th ed., 2010) and were approved by the Animal Ethics Committee of Nantong University (Protocol Nos. S202112294-08 and S20230727-007). Animal welfare was prioritized, with measures implemented to minimize distress.

Prior to any surgical procedures that could potentially cause pain or discomfort, animals were administered appropriate anesthesia or analgesia. Following best practices recommended for the species and the nature of the surgery, the SCI model was established in rats through the intraperitoneal administration of avertin (Kuhn and Wrathall, 1998) (2,2,2-tribromoethanol, Sigma-Aldrich, 250 mg/kg) to induce anesthesia. Subsequently, a Walsh impactor weighing 35 g was used to apply compression at the T7–T9 level for 1 min. The successful induction of SCI was confirmed by the observation of hind limb flaccid paralysis. Serum and spinal cord samples were collected on the 3rd, 7th, and 14th day post-SCI. Each group, namely the sole SCI group and the Sham group, consisted of 6 rats.

In the second and third experiments, treatment compounds were administered to rats after SCI. The daily administration of compounds to rats at the lesion site was performed using a Hamilton 701N micropipette. The injection was made percutaneously into the injury site, with the injection point 1 mm from the edge of the lesion and at a depth of 1.5 mm. Starting from the day of the surgical procedure and for the following 7 days, treatment compounds were injected daily into the injured area. During the study, all rats were administered antibiotics for 7 days and underwent artificial bladder drainage twice daily until bladder function was restored. The detailed experimental grouping information was presented in the following table. All treatments in Table 1 utilized the same vehicle solution.

In accordance with the AVMA (American Veterinary Medical Association) guidelines for animal euthanasia, CO₂ asphyxiation was used for sedation and euthanasia when necessary. The experimental animals were placed in a CO₂ anesthesia chamber, and the CO₂ valve was opened (65% vol/min). Once the animals gradually lost consciousness, the CO₂ concentration was increased to 100%, resulting in an unconscious state in the rats. Ventilation was continued for an additional 2 min and cervical dislocation was performed to ensure death. A total of 84 rats were euthanized, and this was performed under the following conditions: upon obtaining experimental results or at the end of the animal experiment (80 rats); when, based on the veterinarian’s assessment considering factors such as body weight, appetite, infection, and dying symptoms, the degree of suffering reached or exceeded the predefined humane endpoint (four rats); and due to other reasons rendering them unsuitable for further housing (0 rats).

On the 3rd, 7th, 14th, 21st, and 28th day post-SCI, the rats were scored with the Basso-Beattie-Bresnahan (BBB) score (Basso et al., 1995). The score system assigns a score of 21 to normal rats and 0 to completely paralyzed rats. Two experienced experimenters conducted evaluations at a consistent time in the morning on each test day. They independently assessed the BBB scores within 5 min to determine the hind limbs’ motor function recovery following SCI. In addition, the limb muscle strength in rats was evaluated using the inclined plane test (Rivlin and Tator, 1978). The test involved gradually increasing the angle between the horizontal plane and the inclined plane until the rats were unable to maintain a constant posture for more than 5 s. The maximum angle achieved before losing posture was recorded as the result of the inclined angle for the rats.

TRIZOL reagent (No. A33250, Invitrogen) was used to extract RNA and reverse transcribed into cDNA according to the manufacturer’s instructions. The total RNA extracted is approximately 1 μg, with a reverse transcription system of 20 μl. The cDNA product after reverse transcription is diluted five times before PCR amplification. The QPCR was performed using SYBR green dye (No. #RR067A, Takara) in a thermal cycler with the following parameters: an initial denaturation step at 95°C for 30 min; followed by 40 cycles at 95°C for 5 s and 60°C for 30 s. Each sample underwent the entire experimental process independently. The QPCR results were quantified by 2^–ΔΔCT. The forward primer of GAPDH is 5′-ACA GCA ACA GGG TGG TGG AC-3′; the reverse primer of GAPDH is 5′-TTT GAG GGT GCA GCG AAC TT-3′. The forward primer of GAL3 is 5′-AGG CTC CTC CTA GTG CCT AT-3′; the reverse primer of GAL3 is 5′-CCT CCA GGC AAG GGC ATA TC-3′. The forward primer of CDC42 is 5′-AAA GAA AAG TGG GTG CCT GAG-3′; the reverse primer of CDC42 is 5′-AGC AGT CTC TGG AGT GAT AGG -3′.

The protein samples were obtained from rat spinal cord tissue and neurons, and the total soluble protein concentration was determined. The protein content of each experimental sample was approximately 100 μg. An immunoprecipitation experiment was conducted using the Co-IP kit (No.88804, Invitrogen). The whole cell lysate was pre-absorbed with protein A + G agarose beads (No.78610, Invitrogen) for 1 h, followed by high-speed centrifugation to remove any non-specifically bound proteins on protein A + G. Subsequently, the samples were incubated overnight at 4°C with the designated antibodies with gentle rotation. Protein samples were separated on 10% Tris- sodium dodecyl glycinate-polyacrylamide gel and transferred onto an immobile-PSQ transfer membrane (Merck Millipore, Darmstadt, Germany). Different membranes were mixed with GAPDH (1:1000, mouse IgG; No.10494, Proteintech), GAL3 (1:10000, rabbit IgG; No. 60207, Proteintech), CDC42 (1:1000, rabbit IgG; No.10155, Proteintech), ATG7 (1:1000, rabbit IgG; No.10088, Proteintech), P62 (1:1000, rabbit IgG; No.18420, Proteintech), and LC3 (1:1000, rabbit IgG; No. ab51520, Abcam). Then, the membrane was incubated with goat secondary antibody coupled with horseradish peroxidase against mouse IgG (1:10000; No. SA00001-1, Proteintech) or rabbit IgG (1:10000; No. SA00001-2, Proteintech). The signal was developed using a hypersensitive ECL chemiluminescence kit (No. G2020, Servicebio). Density measurement is performed by ImageJ software (version 1.53t, NIH, United States) to quantify the signal intensity.

Rats were euthanized by cardiac puncture under terminal CO₂ anesthesia (65% vol/min) with avertin (250 mg/kg), perfused with 0.9% saline, followed by 4% paraformaldehyde (pH 7.2). Then, the injured segment of the spinal cord was immediately dissected and separated, then fixed in 4% paraformaldehyde for 24 h before embedding into the paraffin section. After dewaxing the paraffin sections and performing heat antigen retrieval, immunofluorescence staining was performed. Next, the steps of immunofluorescence staining of paraffin sections or cell slides of tissues were the same. After incubation with blocking solution, samples were co-incubated overnight at 4°C with two primary antibodies simultaneously (a target protein antibody and a cellular marker antibody). All subsequent experimental procedures were performed under light-protected conditions: the tissues were washed with PBST three times, followed by incubation with corresponding secondary antibodies. Lastly, staining with DAPI (4’,6-diamidino-2-phenylindole) was performed for imaging. The primary antibody used targeted the following proteins: NeuN (1:500, mouse IgG; No.26975, Proteintech), GAL3 (1:200, mouse IgG; No. 60207, Proteintech), and CDC42 (1:200, rabbit IgG; No.10155, Proteintech). The secondary antibodies used were: Alexa fluor 488-implicated Donkey Anti-Rabbit IgG (1: 1000; No. SA00013-1, Proteintech) and Cy3–conjugated Affinipure Goat Anti-Rabbit IgG (1:1000; No. SA00009-2, Proteintech). Immunofluorescence images were captured using a fluorescence microscope (Carl Zeiss, Jena, Germany). Immunofluorescence intensity was quantified using ImageJ by calculating fluorescence intensity coefficients (FIC) for each group: (background-subtracted target protein intensity)/(background-subtracted cellular marker intensity). Statistical comparisons were made between experimental and control group FIC ratios, with cellular marker normalization eliminating positive cell number effects.

One-day-old Sprague-Dawley rats were purchased from the Experimental Animal Center of Nantong University, and spinal cord neurons were isolated following these procedures: (1) Neonatal rats were euthanized with CO₂ and immediately disinfected with 75% alcohol followed by decapitation. The spinal cord adventitia was carefully removed using a stereoscopic microscope in pre-cooled L-15 buffer (No. 11415064, Invitrogen). Subsequently, L-15 was aspirated, and the spinal cord was cut into pieces using 0.125% trypsin. The cord fragments were digested at 37°C for 30 min with gentle agitation every 10 min. (2) After tryptase treatment, the solution was removed, and the digestion was halted by adding Duchenne’s modified Eagle medium (DMEM, No. 11320033, Invitrogen) with 10% Fetal Bovine Serum (No. 30067334, Invitrogen) and 1% Penicillin-Streptomycin (No. C0222, Beyotime). After centrifugation at 1,000 rpm for 3 min, L-15 was added again to gently wash the resulting precipitate, followed by another centrifugation at 1,000 rpm for 3 min. (3) Gentle mixing of L-15 was followed by filtration of all liquids using a 70 μm sieve (No.FSTR070, Beyotime). The filtered liquids were carefully added in a 1:1 ratio on the upper layer of a motor neuron separation solution consisting of Nycoprep 1.077 (15%) (No.1114551, Axis-Shield), Hibernate-E medium (83%) (No. A1247601 Invitrogen), and Brewer’s B27 (2%) (No. 05711, Stemcell) (immiscible). Subsequently, the mixture was centrifuged at 2,000 rpm for 15 min. (4) The middle layer at the liquid surface was extracted and transferred into a new centrifuge tube. After centrifugation at 1,000 rpm for 3 min, the supernatant was discarded. Neurobasal-A culture medium (No. 10888022, Invitrogen), containing 2 mmol/L L-glutamine and 2% concentration of B27 neuron-specific additive, with final concentrations of 100 U/mL and 0.1 mg/mL for penicillin and streptomycin, respectively, were added. The solution was then inoculated into a culture plate for cell culture at 37°C in an incubator containing 5% carbon dioxide. (5) Replacement of the medium occurred 24 h post-plating and continued every 2 days. To enhance the neuron purity, Ara-C (No.HY-13605, MedChemExpress) at a concentration of 0.05 mg/mL was added 24 h after plating. (6) For the nerve injury model, after 3 days of cell culture, the experimental group and the cell transfection control group (the siR-Con group) were stimulated with a culture solution containing 10 μM Glutamate (No. 56-86-0, Sigma-Aldrich) for 1 h (Shahsavani et al., 2021; Yan et al., 2024), while the control group received an equal volume of phosphate buffer saline. Subsequently, both groups were cultured for an additional 16 h in a Neurobasal-A culture medium for glutamate-free neurons.

Total RNA was extracted from neonatal Sprague-Dawley rat spinal cord neurons using TRIZOL reagent (Invitrogen) in collaboration with Allwegene Co. Ltd (Nanjing, China). Sequencing was carried out using the Illumina high-throughput sequencing platform (HiSeqTM2500/4000). A total of six samples were sequenced, including three samples from neurons as the transfection control group and three samples from neurons with GAL3 knocked down. The data provided by the company underwent batch effect removal, standardization, and correction for intra-group differences, resulting in an expression matrix. The detailed sequencing steps conducted by the sequencing company were provided in the “Supplementary Figure 1.”

We downloaded the datasets of GSE2599 (Aimone et al., 2004), GSE20907 (Siebert et al., 2010), GSE45006 (Chamankhah et al., 2013), and GSE174549 (Li E. et al., 2022) from the Gene Expression Omnibus Datasets (GEO) available at https://www.ncbi.nlm.nih.gov/geo/. Table 2 displays the basic information for these four datasets. The 60 samples were re-divided into two groups based on the relative expression level of GAL3, and the role of GAL3 in SCI was investigated by single-gene analysis. Before dataset analysis, batch effects were removed, and standardization pretreatment was conducted. The R software (v4.3.0¹) was used to analyze and visualize data. The r package used to remove the batch effect is sva (Ge et al., 2011) (v3.46.0²). After obtaining the expression matrix, limma package (Ritchie et al., 2015) (v3.54.2³) was used for differential expression analysis. The screening criteria of differential expression genes (DEGs) are |logFC| > 1 and P-value < 0.05. We used Gene Set Enrichment Analysis (GSEA) (Han et al., 2019) as a calculation method to determine whether a set of a priori-defined genes show statistically significant consistent differences between two biological states by clusterProfiler (Wu et al., 2021) (v4.0.5⁴). We used ggplot2 (Ito and Murphy, 2013) (v3.4.1⁵) to screen out the candidate genes and used the STRING⁶ website and the Cytoscape software (v3.10.0⁷) to analyze the candidate genes by Protein-Protein Interaction Networks (PPI)⁴⁷.

Blood samples were collected from 13 patients diagnosed with SCI, and admitted between January 2022 and December 2023, on the first day of admission. Serum was extracted from these samples after high-speed centrifugation and stored at −80°C. The screening criteria for patients’ inclusion in the SCI group were as follows: diagnosis of spinal cord injury confirmed through CT, MRI, and other imaging examinations, in according with the International Standard for Neurological Classification of SCI (revised in 2019) (ASIA and ISCoS International Standards Committee, 2019); Absence of spinal malignant tumor, and tuberculosis-related pathological SCI, or visceral injury; No concurrent diseases such as craniocerebral injury. Additionally, 13 serum samples from individuals without SCI were collected as the control group, and the screening criteria included: Age difference from the SCI patients within 5 years and absence of underlying health conditions. All subjects volunteered for inclusion in this study and provided informed consent. Approval for this study was obtained from the Medical Ethics Committee of Nantong First People’s Hospital (Protocol number: 2020KY036).

After removing the cells from the cell culture incubator, collect the supernatant. Then gently wash the spinal cord neurons with PBS at 4°C. Add an appropriate amount of cell lysis solution mixed with 1/100 protease inhibitor and 1/50 phosphatase inhibitor, gently scrape off the cells, and place them on ice for 15 min for lysis. Centrifuge at 4°C and 13,000 rpm for 20 min. Collect the protein supernatant. The ELISA kit (GAL3, No. E0497h, ElAab; and CDC42, No. E0321r, ElAab) were utilized to measure the expression levels of GAL3 and CDC42 in human serum, rat serum, primary neuron cell lysate, and primary neuron cell supernatant. The samples were stored at −80°C before measurement. The optical density at 450 nm was calculated by subtracting the background value, and the standard curve was conducted.

All data were analyzed using GraphPad Prism 9.0 (GraphPad Software Inc., California, United States). All data are presented as mean ± standard error and were derived from at least three independent repeated experiments. The difference in BBB locomotor scores and inclined plane degrees between groups were assessed using two-way Repeated Measurement ANOVA. One-way ANOVA was used to evaluate the significance of the differences between groups for other data. The HSD (Honestly Significant Difference) test was used as a post hoc test following ANOVA. Unpaired Student’s t-test was used for comparisons between the two groups. A significant level of P < 0.05 was considered statistically significant.

After successfully establishing the rat SCI model, we examined GAL3 expression in spinal cord tissue and serum. The mRNA and protein were extracted at intervals of 3rd, 7th, and 14th days post-SCI. QPCR and Western Blot showed that the mRNA level and protein expression level of GAL3 were increased after SCI and peaked on day 3 after injury, gradually decreasing thereafter (Figures 1A–C). Notably, GAL3 level in serum was also increased post-SCI (Figure 1D). Immunofluorescence analysis revealed that GAL3 was co-localized with neurons in the ventral horn of the spinal cord (Figure 1E), notably significant at 3rd, 7th, and 14th days post-SCI (Figure 1F). However, GAL3 did not co-localize with astrocyte marker GFAP (glial fibrillary acidic protein) (Figure 1G) or microglia marker IBA1 (Figure 1H) post-SCI, suggesting a potential involvement of neuronal GAL3 in SCI pathogenesis.

Given that extracellular glutamate triggers excitotoxic neuronal cell death (Tehse and Taghibiglou, 2019), and is a significant factor in SCI-associated neurological dysfunction (Hausmann, 2003; Ribeiro and de Wit, 2017), we used glutamate to induce a neuron injury model (Rong et al., 2019; Shahsavani et al., 2021). Cell Counting Kit-8 (CCK8) experiments showed that 10 μM glutamate for 1 h yield the most suitable injury models (Figure 1I). QPCR showed that the mRNA of GAL3 was higher than that of the control group (Figure 1J). Western blot showed upregulated GAL3 protein levels (Figures 1K, L). ELISA further supported increased GAL3 secretion (Figure 1M). The immunofluorescence revealed heightened GAL3 fluorescence intensity in neurons (Figures 1N, O). These results collectively indicate GAL3 involvement in spinal cord neuron injury mechanisms after SCI.

To explore the specific role of GAL3 in the mechanism of spinal cord neuron injury, we used siRNA to knock down the GAL3 gene in spinal cord neurons under glutamate stimulation. The knock-down efficiency of siR-GAL3 resulted in a 90% reduction in GAL3 mRNA levels (Figure 2A), a 75% decrease in GAL3 protein expression (Figures 2B, C), and a 60% decline in secreted GAL3 (Figure 2D).

Subsequently, we administrated the siR-GAL3 or GAL3 inhibitor, TD139, into the injured spinal cord of rats for 7 consecutive days to observe motor function recovery post-SCI. Starting from the 7th-day post-SCI, both experimental groups displayed increased BBB scores compared to the simple SCI group (Figure 2E). Similarly, the degrees of the inclined plane angles also exhibited notable improvement (Figure 2F). These data suggest that knockdown or inhibition of GAL3 effectively promotes the recovery of motor function in rats following SCI. Interestingly, when administering GAL3 recombinant protein continuously for 7 days in the injured spinal cord of rats, we observed that starting from the 7th day following SCI, compared to the SCI + Vehicle group, the rats in the SCI + GAL3 group showed a decrease in BBB scores and incline angle, indicating that the injection of GAL3 hindered the recovery of motor function in SCI.

To unravel the role of GAL3 in the mechanism of SCI, we conducted a single-gene bioinformatics analysis of GAL3 involvement by rat SCI datasets obtained from GEO. Before formal dataset analysis, the batch effect was removed (Figures 3A, B). After batch effect removal and standardization, we obtained expression levels for 2,240 genes. Subsequent differential expression analysis revealed 158 differentially expressed genes (DEGs), of which 101 were up-regulated and 57 were down-regulated (Figure 3C). Biological Process (BP) analysis via GSEA, demonstrated significant enrichment of the highly expressed gene set functioned in pathways associated with programmed cell death mechanisms, such as GO:0008219 Cell Death, GO:0012501 Programmed Cell Death, GO:0006915 Apoptotic Process and Go: 0016239 Positive Regulation of Macroautophagy (Figure 3D).

To determine the relationship between these 158 DEGs and GAL3, we conducted the protein-protein interaction (PPI) analysis using the STRING website (Figure 3E). This analysis revealed evidence of interaction between GAL3 and 22 core nodes, with 21 of the core nodes showing up-regulated expression, except for the down-regulation of Eef1a2. These results strongly indicate that GAL3 is closely related to programmed cell death after SCI.

Considering the pivotal role of autophagy in programmed cell death, we validated whether GAL3 functions through autophagy after SCI. Initially, we used glutamate to stimulate neurons treated by siR-GAL3 (Figure 4A). In comparison with the siR-Con group, the siR-GAL3 group displayed significantly reduced GAL3 expression (Figure 4B), decreased ATG7 expression (Figure 4C), increased P62 expression (Figure 4D), and a decreased LC3 II/I ratio (Figure 4E).

We then examined autophagy in the spinal cord after injection with TD139. We extracted tissue protein and serum from rats in the Sham group, the SCI group, and the SCI + TD139 group on the 7th day after SCI (Figure 4F). Compared with the Sham group, expression levels of ATG7 in the SCI group increased (Figure 4G), whereas the P62 level decreased (Figure 4H), and the LC3 II/I ratio increased (Figure 4I), suggesting elevated autophagy level post-SCI. Furthermore, compared to the SCI group, expression levels of ATG7 decreased in the SCI + TD139 group (Figure 4G), while P62 expression and LC3 II/I ratio remained unchanged (Figures 4H, I). These results indicate that inhibition of GAL3 reduced autophagy levels at the injured site in rats.

To delve deeper into the specific role of GAL3 in the mechanism of spinal cord neuron injury, we extracted RNA from Glu + siR-Con and Glu + siR-GAL3 neurons for sequencing analysis. The volcano plot revealed 316 DEGs, with 151 up-regulated and 165 down-regulated genes (Figure 5A). Further, BP of GSEA showed significant enrichment in genes linked to autophagy mechanisms like Go: 0010506 Regulation of autophagy and Go: 0061684 Chaperone-mediated autophagy (Figure 5B). To establish connections between these 316 DEGs and GAL3, PPI analysis via the STRING website identified 29 core nodes interacting with GAL3, where 25 core nodes exhibited up-regulation, excluding Clu, Lamp1, Cdc42, and Ctnnb1 (Figure 5C). Subsequently, by overlapping 29 core nodes with 22 core nodes from the SCI dataset, eight core genes were screened out using a Venn diagram (Figure 6A). However, only CDC42 showed an upregulation with GAL3 in the SCI dataset, and down-regulated in the siR-GAL3-treated neuron dataset. Additionally, the expression level of GAL3 positively correlated with CDC42 expression in both the SCI dataset (R² = 0.712, P < 0.001) (Figure 6B) and the siR-GAL3-treated neuron dataset (R² = 0.809, P = 0.0146) (Figure 6C). Furthermore, Co-IP experiments confirmed the interaction between GAL3 and CDC42 (Figure 6D).

Then we stimulated neurons with GAL3 and detected the expression levels of CDC42 and autophagy markers. Western blot analysis showed, compared to the control group, expression levels of CDC42 and ATG7 increased, P62 expression decreased, and the LC3 II/I ratio increased in the GAL3 group (Figures 6E–I). However, the GAL3 + ML141 (the CDC42 molecular inhibitor) group displayed lower autophagy levels than the GAL3 group (Figures 6E–I). In addition, we detected the expression level of CDC42 in the supernatant of neurons, which followed a similar trend as the western blot (Figure 6J). These data indicate that GAL3 promotes neuron autophagy, and inhibition of CDC42 suppresses GAL3-induced autophagy.

We detected the expression level of CDC42 in rats post-SCI. We found an increase in both the mRNA level and protein level of CDC42 in the spinal cord tissue after SCI, peaking on the third day post-SCI, followed by a decrease over time (Figures 7A–C). Immunofluorescence analysis of spinal cord tissue showed co-localization of CDC42 within neurons in the ventral horn, statistically significant on the 3rd, 7th, and 14th days post-SCI (Figures 7D, E). Moreover, post 3rd-day of SCI immunofluorescence analysis showed a lack of co-localization between CDC42 and GFAP or IBA1 (Figures 7F, G). Similarly, in the glutamate-induced neuron injury model, both mRNA and protein expression levels of CDC42 were increased (Figures 7H-J), with an increased fluorescence intensity observed (Figures 7K, L), indicating a potential role of CDC42 in SCI.

To explore the mechanism of CDC42 in SCI, we injected siR-CDC42 and the CDC42 molecular inhibitor ML141, into the injured site of SCI rats, observing the recovery of hind limb motor function. The knock-down efficiency of siR- CDC42 resulted in an 85% reduction in CDC42 mRNA levels (Figure 8A), and a 50% decline in secreted CDC42 (Figure 8B). Analysis of BBB locomotor scores and inclined plane experiments revealed that siR-CDC42 and ML141 injection indeed promoted the recovery of motor function in SCI rats (Figures 8 C, D).

Finally, to understand the expression trend of GAL3 and CDC42 in human serum post-SCI, we collected samples from 8 individuals with SCI and 8 healthy controls. We found higher levels of GAL3 and CDC42 in the serum of SCI patients compared to those in the control group (Figures 8E, F).