ROS-responsive hydrogel delivering liposomal indole-3-propionic acid promotes spinal cord injury repair by enhancing CX3CL1-CX3CR1 mediated neuron-microglia communication

Abstract

Spinal cord injury (SCI) is a devastating neurological disorder characterized by excessive neuroinflammation and limited regenerative capacity. Here, we developed a reactive oxygen species (ROS)-responsive hydrogel for localized delivery of liposomal indole-3-propionic acid (IPA), a gut microbiota-derived metabolite with neuroprotective potential. The hydrogel exhibited favorable injectability, self-healing capability, and ROS-responsive degradation, enabling sustained release and prolonged retention of IPA at the lesion site. In vitro, IPA-loaded liposomes effectively attenuated lipopolysaccharide-induced inflammation and promoted axonal regeneration with favorable biosafety. In the rat SCI model, localized delivery of liposomal IPA significantly improved locomotor recovery, electrophysiological function, and tissue preservation while reducing cavity formation and secondary injury. Moreover, IPA promoted neuronal survival, enhanced axonal integrity, and shifted microglia toward a pro-repair phenotype. Single-nucleus RNA sequencing identified enhanced CX3CL1-CX3CR1 mediated neuron-microglia communication following IPA treatment, accompanied by enrichment of PI3K/AKT signaling pathway in both neurons and microglia. In addition, IPA facilitated microglial migration toward the lesion core, suggesting coordinated modulation of the neuroinflammatory microenvironment after SCI. Collectively, our study identifies a previously unrecognized therapeutic role of IPA in spinal cord repair and presents a ROS-responsive nanoplatform for localized metabolite delivery to regulate neuron-microglia crosstalk and promote functional recovery after SCI.

Highlights

• •A drug delivery system based on ROS-responsive hydrogel contains IPA-loaded liposomes.
• •LIP-IPA@PHA-4P-CHO can be used for SCI repair by regulating neuron and microglia.
• •SnRNA-seq reveals IPA enhances CX3CL1-CX3CR1 mediated neuron-microglia communication.
• •IPA activates PI3K/AKT signaling in neurons and microglia to promote SCI repair.

1. Introduction

Spinal cord injury (SCI) is a devastating condition that often leads to temporary or permanent change in spinal cord function, resulting in sensory and functional disorders and even paraplegia [1]. Globally, an estimated 250,000–500,000 new SCI cases occur each year, imposing substantial physical, psychological, and socioeconomic burdens on patients, their families, and healthcare systems [2,3]. Current clinical management, including early surgical decompression, maintenance of spinal cord perfusion, and pharmacological interventions to alleviate secondary injury, can only partially improve neurological outcomes [4]. However, these approaches fail to effectively restore the damaged neural circuitry or promote long-distance axonal regeneration. Therefore, developing therapeutic strategies capable of simultaneously mitigating secondary injury and enhancing neural regeneration remains a major challenge in SCI repair.

Emerging evidence has highlighted that diet and intestinal flora are of great significance to maintain homeostasis of the body [5]. The imbalance of gut microbiome is closely related to the occurrence and development of several kinds of diseases [[6], [7], [8], [9]]. Serger et al. confirmed that intermittent fasting was able to promote axon regeneration of sciatic nerve in mice, which could be contributed to the increase of the gut bacteria-derived metabolite indole-3-propionic acid (IPA) in the serum [10]. In addition, IPA has shown excellent anti-inflammatory effects in several central nervous system (CNS) disorders. In Alzheimer's disease, IPA attenuated neuroinflammation by suppressing NLRP3 inflammasome activation [11], while in multiple sclerosis, microbiota-derived IPA activated aryl hydrocarbon receptor (AHR) signaling in astrocytes to suppress CNS inflammation [12]. The anti-inflammatory effect of IPA shown in other diseases of CNS and its promoting effect on axon regeneration in the peripheral nervous system are very consistent with the needs of SCI repair, suggesting that IPA is expected to be applied for the treatment of SCI. Nevertheless, its poor aqueous solubility and limited stability restrict its therapeutic efficacy, highlighting the need for an efficient delivery strategy capable of improving its local bioavailability and sustained retention at the lesion site.

Biomaterial-based drug delivery systems have emerged as promising approaches for overcoming the hostile microenvironment after SCI and enhancing therapeutic efficacy [13]. Among them, hydrogels have attracted considerable attention owing to their excellent biocompatibility and ability to provide localized and sustained drug release [14,15]. Moreover, excessive accumulation of reactive oxygen species (ROS) is a hallmark of secondary SCI and contributes to persistent inflammation, neuronal apoptosis and impaired tissue regeneration. Exploiting this pathological feature to construct ROS-responsive biomaterials offers an attractive strategy for achieving lesion-specific drug release while minimizing off-target effects [16,17]. In addition, liposomal encapsulation provides an effective means to improve the stability, bioavailability, and cellular uptake of hydrophobic small molecules such as IPA, thereby further enhancing therapeutic performance [18,19].

Herein, we developed a ROS-responsive liposome-in-hydrogel delivery platform for localized and sustained administration of IPA after SCI (Scheme 1). IPA was first encapsulated into liposomes to improve its stability and delivery efficiency and subsequently incorporated into a hyaluronic acid (HA)-based hydrogel functionalized with phenylboronic acid (PBA), enabling ROS-triggered degradation and controlled drug release within the injured spinal cord [20]. We hypothesized that this integrated delivery system would enhance local IPA bioavailability, promote neuroprotection and axonal regeneration, and ultimately improve functional recovery after SCI. Furthermore, we investigated the molecular mechanisms underlying IPA-mediated neurorestoration using single-nucleus RNA sequencing (SnRNA-seq) and experimental validation, revealing that IPA enhances CX3CL1-CX3CR1 mediated neuron-microglia communication and activates PI3K/AKT signaling pathway to facilitate SCI repair.

Scheme 1.

Schematic illustration of LIP-IPA@PHA-4P-CHO for SCI repair.

2. Materials and methods

2.1. Preparation of IPA@LIP

40 mg distearoyl phosphorylcholine (DSPC), 5 mg poly(ethylene glycol) conjugated phospholipids (DSPE-PEG₂₀₀₀) and 5 mg cholesterol (mass ratio 8:1:1) were dissolved together in chloroform with 7 mg IPA. Next, the organic solvent was dried by a rotary evaporator to form a liposome film. Add ultra-pure water and remove the liposome film on the bottle wall through water bath ultrasound. Finally, use the probe ultrasound to fully self-assemble the above-mentioned components and obtain IPA@LIP.

2.2. Characterization of IPA@LIP

The surface morphology of IPA@LIP was observed under transmission electron microscopy (TEM, JEM-2100, JEOL, Japan). Size and zeta potential were measured using a dynamic light scattering (DLS) measurement (BI-200 SM, Brookhaven Instruments, USA) at a scattering angle of 90° at 25°C. Data at different time points was detected in order to evaluate its stability. The encapsulation efficiency and loading rate of IPA were tested by ultraviolet spectrophotometer.

2.3. Cell isolation and culture

PC12 cells (rat adrenal pheochromocytoma-derived cell line) were obtained from Zhongqiao Xinzhou Biotechnology Co., Ltd., and cultured in RPMI-1640 medium (Procell, China) supplemented with 10% fetal bovine serum (FBS, Gibco, USA) and 1% penicillin/streptomycin solution.

Primary cortical neurons were isolated from E17 pregnant Wistar rat embryos as described previously [21]. In brief, hippocampal tissues of fetal rats were cut into pieces under aseptic condition after removing the blood vessels and meninges, then tissues were digested with papain (Solarbio, China) and dsDNase (Solarbio, China) for 15 min at 37°C. Dulbecco modified Eagle medium (DMEM)/F-12 (Gibco, USA) with 10% FBS and 1% penicillin/streptomycin was used to terminate the digestion. After centrifuging at 1000 rpm for 5 min and filtering through a 70 μm cell sieve, neurons were seeded on plates precoated with 0.01% poly-D-lysine (PDL, Sigma-Aldrich, USA). DMEM/F12 was changed to neurobasal medium (Gibco, USA) with 2% B-27 (Gibco, USA), 1% L-glutamine (Gibco, USA) and 1% penicillin/streptomycin 4 h after spreading on plates. Half of the medium was changed every three days.

BV2 mouse microglia cell line, which was purchased from Procell Life Science & Technology Co. Ltd., was cultured in DMEM (Gibco, USA) containing 10% FBS and 1% penicillin/streptomycin.

2.4. Live/dead staining assay

Live/Dead cell staining kit (L6037, UElandy, China) was used to evaluate the safety of liposome and IPA@LIP. PC12 cells were co-cultured with IPA, liposome or IPA@LIP for 48 h. They were stained by 2 μm Calcein AM and 4.5 μm PI for 20 min at 37°C, and then a fluorescence microscope (IX73, Olympus, Japan) was used for photographs. Live cells stained by Calcein AM were green while dead cells stained by PI showed red color.

2.5. Cell viability assay

PC12 cells were seeded into 96-well plates, and the density of cells was 1 × 10⁴ cells/well. IPA and/or Lipopolysaccharide (LPS) were added to co-culture with PC12 cells. At the end of different treatments, 10% CCK-8 reagent (Dojindo, Japan) was added to test cell viability. Finally, a microplate reader was used to detect the absorbance at a wavelength of 450 nm after incubating 2 h at 37°C.

2.6. Cell proliferation and apoptosis

LPS was chosen to mimic the inflammatory microenvironment after SCI [[22], [23], [24]]. PC12 cells were divided into five groups (Control, IPA, LPS, LPS + LIP and LPS + IPA@LIP) and incubated for 24 h. The proliferation of PC12 cells was measured by EdU staining kit (C0071, Beyotime, China) [22]. TUNEL staining kit (C1088, Beyotime, China) was used to detect cell apoptosis according to the manufacturer's instructions after different treatments. PC12 cells were observed under a laser-scanning fluorescence confocal microscopy (LSM980, Zeiss, Germany). Finally, all images were analyzed by ImageJ software.

2.7. Flow cytometry

Flow cytometer (FACSCelesta, BD, USA) was used to analyze the apoptosis of PC12 cells and polarization of BV2 cells. Both of the two kinds of cells were divided into five groups as described above. PC12 cells were collected and washed twice with cold phosphate-buffered saline (PBS) after different treatments, and then 5 μl Annexin V-FITC and 5 μl PI were mixed with PC12 cells and incubated at room temperature for 20 min away from light according to Annexin V-FITC kit (C1383, Beyotime, China). BV2 cells were seeded into 6-well plates and incubated for 24 h under different conditions. FITC-CD86 (FITC-65068, Proteintech, China) and PE-CD206 (141,706, BioLegend, USA) were used to identify M1 and M2 types of BV2 cells. Flow cytometry data analysis software (FlowJo V10.9.0, FlowJo, LLC) was used for data analysis.

2.8. Immunofluorescence (IF) staining

The slides of PC12 cells and primary cortical neurons were prepared for IF. Cells were fixed in 4% paraformaldehyde (PFA) at 4°C for 30 min, and then they were permeabilized and blocked with 0.3% Triton X-100 (Sigma, Germany) and 5% bovine serum albumin (BSA, Solarbio, China) for 1 h at room temperature. Subsequently, samples were incubated overnight at 4°C with primary antibodies. After washing away the antibodies with PBS at the following day, cells were incubated with secondary antibodies at room temperature for 2 h, protected from light. Finally, DAPI (Abcam, USA) was added to stain the cell nucleus for 10 min at room temperature. Paraffin sections of spinal cord were also made for IF. They were dewaxed, rehydrated, permeabilized and blocked in turn. Next, sections were incubated with primary and secondary antibodies and DAPI with the same method mentioned above. The laser-scanning fluorescence confocal microscopy was used to observed slides.

The primary antibodies used were Tuj-1 (1:200, Abcam, USA), NeuN (1:200, Abways, China), NF200 (1:800, Proteintech, China), 5-HT (1:200, Abcam, USA), GFAP (1:200, Abways, China), iBa-1 (1:200, Abways, China), Arg-1 (1:500, Proteintech, China), CX3CL1 (1:1000, Abcam, USA), CX3CR1 (1:500, Bioss, China) and Fascin-1 (1:500, Santa Cruz, USA). The secondary antibodies used were Alexa Fluor 488/594/647-labeled goat anti-rabbit/mouse IgG (1:500, UElandy, China).

2.9. Lactate dehydrogenase (LDH) release assay

To assess the rescue effect of IPA in LPS-induced cytotoxicity of BV2 cells, LDH assay kit (C0018, Beyotime, China) was used to evaluate LDH release of BV2 cells in different groups. The absorbance at a wavelength of 450 nm was measured by a microplate reader after treatments according to the instructions. The percentage of LDH was calculated using the formula: LDH release (%) = [(treated cell LDH release - spontaneous LDH release)/(total LDH release - spontaneous LDH release)] × 100%.

2.10. Enzyme-linked immunosorbent assay (ELISA)

BV2 cells were divided into five groups and different treatments were used as mentioned above. Supernatant was collected after incubating for 24 h. Based on the manufacturer's instructions, ELISA kits (4 A Biotech, China) were used to measure the levels of TNF-α, IL-6 and IL-1β in the supernatant. At 28 days post injury (dpi), spinal cord tissues were collected from rats and homogenized. The concentration of soluble CX3CL1 in the tissue homogenates was measured using ELISA kit (Thermo Fisher Scientific Inc., USA). The absorbance was detected at 450 nm with a microplate reader. The concentrations of TNF-α, IL-6, IL-1β and soluble CX3CL1 were calculated in accordance with standard curves.

2.11. Western blot analysis

Proteins were extracted from BV2 cells and spinal cord tissue by RIPA lysis and extraction buffer (Thermo Fisher Scientific Inc., USA) and quantified by a BCA protein assay kit (Thermo Fisher Scientific Inc., USA). Next, equivalent amount of protein samples were separated by 10% SDS polyacrylamide gel and transferred to polyvinylidene fluoride membranes (PVDF, Merck Millipore, USA). The membranes were blocked with 5% skim milk at room temperature for 1 h and then incubated with primary antibodies overnight at 4°C. On the following day, the membranes were incubated with secondary antibodies at room temperature for 2 h. Finally, The relative expression of target proteins were visualized using a ChemiDicTM XRS + Imaging System (Bio-Rad, USA).

The primary antibodies used were TNF-α (1:1000, Proteintech, China), IL-6 (1:1000, Proteintech, China), IL-1β (1:1000, Abcam, USA), iNOS (1:1000, Proteintech, China), Arg-1 (1:1000, Proteintech, China), PI3K (1:1000, Proteintech, China), p-PI3K (1:1000, MCE, USA), Akt (1:2000, Proteintech, China), p-Akt (1:2000, Proteintech, China), mTOR (1:5000, Proteintech, China), p-mTOR (1:2000, Proteintech, China), TLR4 (1:1000, Proteintech, China), β-actin (1:1000, Proteintech, China). The secondary antibodies used were HRP-goat anti rabbit/mouse recombinant secondary antibody (1:5000, Proteintech, China).

2.12. Synthesis of LIP-IPA@PHA-4P-CHO

PHA was fabricated through the condensation reaction of HA, PBA and adipic dihydrazide (ADH) with condensing agents. 1.3 g HA was dissolved in 100 ml deionized water, and then 1-Ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDCI) and N-hydroxysuccinimide (NHS) were added to the HA solution and activated in the ice water bath for 30 min. Subsequently, 0.3 g of ADH and 0.15 g of PBA were added and stirred continuously at room temperature for 24 h. After dialysis, the solution was freeze-dried to get the solid PHA. To prepare 40% (w/v) LIP-IPA@4P-CHO solution, 0.4 g 4-arm-PEG-CHO was dissolved in 1 ml IPA@LIP and stirred continuously at room temperature until completely dissolved. Next, 150 mg, 225 mg and 300 mg PBA was dissolved in 3 ml deionized water to obtain the 5%, 7.5% and 10% PHA solution. Finally, LIP-IPA@PHA-4P-CHO was synthesized by mixing LIP-IPA@4P-CHO solution and PHA solution in equal volumes.

2.13. Characterization of LIP-IPA@PHA-4P-CHO

40% (w/v) LIP-IPA@4P-CHO was mixed 1:1 with 5%, 7.5% and 10% PHA respectively. The optimal ratio to form the hydrogel was determined according to the fluidity, hardness and viscoelasticity. The microstructure of LIP-IPA@PHA-4P-CHO was observed by scanning electron microscopy (SEM, EM-30C, COXEM, Korea). In brief, freeze-dried hydrogel was broken in liquid nitrogen and the cross-section was coated with a gold layer to enhance the conductivity for SEM. A rheometer (DSR502, Anton Paar, Austria) was used to test sample's shear thinning properties, mechanical properties and so on. The variations of energy storage modulus (G′) and loss modulus (G″) were recorded. The self-healing performance of hydrogel was detected by splicing two pieces of samples (one was stained with rhodamine B). The adhesion strength measurement was quantitatively analyzed by using the lap shear test. The porcine skin was cut into long strip at 10 × 50 mm and soaked in PBS solution for 24 h. The hydrogel precursor mixture was uniformly coated on the surface of the porcine skin, and then another piece of porcine skin was placed over it to ensure that the overlap area between them was 10 × 10 mm². Next, they were kept at room temperature for 2 h for fully bonding and curing. The lap shear test was carried out using a universal testing machine to measure the adhesion force between the hydrogel and porcine skin. The maximum load-bearing capacity of the force sensor was set at 50 N and the application speed was 20 mm/min. The steps to determine the swelling rate of LIP-IPA@PHA-4P-CHO were as follows. Dry hydrogel samples were weighed (W0) accurately, and then they were immersed in PBS solution at 37°C. They were taken out of PBS solution when they reached the swelling equilibrium state (i.e., the mass no longer changed over time). After sucking off the residual liquid and salt on the surface carefully, swelling mass (Wt) was immediately recorded. The swelling rate of LIP-IPA@PHA-4P-CHO was calculated by the following formula: Swelling rate = (Wt - W0)/W0 × 100%.

In order to measure the degradation rate of hydrogel, samples were freeze-dried and then cut into pieces of equal mass. Subsequently, they were immersed in PBS and H₂O₂ respectively and their remaining weight in different time points was recorded. The calculation method of degradation rate is similar to that of swelling rate. The release rates of IPA@LIP from hydrogel in PBS and H₂O₂ were evaluated by ultraviolet spectrophotometer. Briefly, the absorbance of the solution at 330 nm at different time points was recorded by the microplate reader and compared with the standard curve of IPA@LIP.

2.14. ROS response ability of PHA-4P-CHO

Fluorescent probe 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA, GC30006, Glpbio, USA) was used to detect the ROS response ability of PHA-4P-CHO. BV2 cells in experiment group were co-cultured with hydrogel for 24 h while cells in the other two groups were untreated. Next, cells were exposed to H₂O₂ for 3 h except for the control group. Subsequently, DCFH-DA was added and the cells were incubated at 37°C for 30 min and washed three times with PBS to eliminate excess probe. A fluorescence microscope was used for photographs.

2.15. Sustained release property of LIP-IPA@PHA-4P-CHO

Liposome was labeled with DiR (HY-D1048, MCE, USA) for tracking of IPA@LIP in vivo. To evaluate the sustained release ability of hydrogel, IPA@LIP and LIP-IPA@PHA-4P-CHO were placed at the lesion of rats’ spinal cords respectively. At several time points after surgery, rats were anesthetized and fluorescence intensity were detected by the IVIS Imaging System (Lumina XRMS Series III, PerkinElmer, USA).

2.16. Animals

8 weeks age female Wistar rats (180-200 g) were obtained from SPF (Beijing) Biotechnology Co., Ltd. They were housed in the specific pathogen-free condition under a standard light/dark cycle with food and water ad libitum. All animal experiments were conducted with approval from the Animal Welfare and Research Ethics Committee of the First Affiliated Hospital of Zhengzhou University (ZZU-LAC20250418 [11]). Animals were randomly assigned to the indicated experimental groups. Sample size was selected based on our previous experience and published studies using similar models, in which comparable group sizes were sufficient to detect biologically meaningful differences. In addition, the number of animals was determined in accordance with the principle of the 3Rs to avoid unnecessary animal use while maintaining adequate statistical power. To minimize potential bias, personnel performing surgical procedures, biomaterials administration, tissue collection and processing were blinded to group allocation whenever feasible. Behavioral assessments, histological evaluation, IF imaging, western blot quantification and data analyses were performed by investigators who were blinded to the groups.

2.17. SCI model establishment

All the rats were randomly divided into six groups: Sham, SCI, IPA@LIP, PHA-4P-CHO, LIP@PHA-4P-CHO and LIP-IPA@PHA-4P-CHO. SCI model establishment was performed according to the modified Allen method as previously described [25]. Anesthesia was induced using 1% pentobarbital sodium, and then the surgical region was prepared by shaving and disinfecting. Spinal cord was exposed at T10 level by laminectomy and a precise impactor (68,099 II, RWD, USA) was used to produce an injury on the dorsal surface of spinal cord. IPA@LIP or different hydrogel were used for lesion and the incision sites were sutured in layers. Rats in Sham group only underwent laminectomy. All rats received antibiotic and analgesia after operation, and the bladders of rats were manual emptied three times a day for a week.

2.18. Tissue preparation

Rats were perfused transcardially with 0.9% saline solution and 4% PFA successively under deep anesthesia, and then spinal cord and other tissues were dissected and fixed in 4% PFA. After fixation, they were dehydrated, paraffin-embedded and made into sections for following staining.

2.19. Hematoxylin and eosin (HE) staining

Paraffin sections of heart, liver, spleen, lung, kidney and spinal cord tissue were prepared for HE staining. They were dewaxed and rehydrated through xylene and gradient alcohol, washed three times with deionized water. Hematoxylin and eosin were used to stain for 6 min and 10 s respectively. Finally, sections were sealed by neutral resin and observed by a microscope.

2.20. Behavioral evaluation

The Basso, Beattie, and Bresnahan (BBB) score, inclined plane test and catwalk gait analysis were used to evaluate motor functional recovery of rats after SCI. For BBB score, rats were placed in an open field and two blinded observers recorded the score on a scale from 0 to 21 according to joint activities, coordination of front and hind limbs, trunk movement stability, tail elevation and so on. The assessment time points were −1, 0, 1, 3, 7, 14, 21, 28 dpi, which were also selected for inclined plane test. All rats were placed on an experimental board with a rubber surface. As the angle of the plate increased, the maximum allowable angle was recorded by two blinded observers, which was the threshold at which the rat could not hold a standing position for more than 5 s before falling. All the rats underwent the test for three times and they were provided with a 1 min break between every test. At 28 dpi, rats were placed on a horizontal glass track to walk from left to right in a dark room. All their footprints were recorded by the high-speed camera below the track and analyzed using the Catwalk XT system (Noldus, Netherlands).

2.21. Electrophysiology

Electrical activities of rats were assessed at 28 dpi. In brief, rats were anesthetized and then the latency and amplitude of motor evoked potentials (MEPs) were recorded by the electrophysiological apparatus (Blackrock Neurotech, USA) using a single 5 mA electrical pulse. The stimulation electrodes were embedded behind the head and neck, and the recording electrodes were placed in the gastrocnemius muscle of the hind limb.

2.22. SnRNA-seq

Rats were randomly divided into Sham, SCI and IPA group. LIP@PHA-4P-CHO was used in Sham and SCI group while LIP-IPA@PHA-4P-CHO was used in IPA group. Their spinal cords were isolated at 28 dpi and ground on ice to obtain the nuclear suspension, which was used for preparing droplets with beads by DNBelab C-TaiM. The processes of cell lysis and mRNA capture by magnetic beads were completed in the droplet. cDNA was generated by reverse transcription. After purification and amplification, DNA nano beads were sequenced by combined probe anchoring polymerization (cPAS) technology on the DNBseq T7 platform with a 150-bp paired-end reading strategy.

Cellranger (v9.0.1) was used to map raw sequencing data with reference genomes (rattus norvegicus, GCF 015227675.2 mRatBN7.2-2023-06-12) to obtain expression matrices and the Seurat package (v5.0.2) was used for further analysis. The background RNA contamination was filtered by SoupX (version: 1.6.2) and the double cells were labeled and filtered by DoubletFinder (version: 2.0.4). As for quality control procedure, Cells with less than 200 genes identified or more than 90% of the maximum number of genes, as well as the top 20% of cells sorted in descending order based on the proportion of mitochondrial DNA, were excluded. After normalizing and scaling the gene expression data, 4000 highly variable genes in the dataset were used for PCA analysis (n = 20), and then Uniform Manifold Approximation and Projection (UMAP) was conducted for cluster visualization. For each cluster, marker genes and differential genes were identified by FindAllMarkers function (logfc > 0.25, minPct > 0.1 and Padj ≤ 0.05). Easybio packages (version: 1.1.1) were conducted to cell type annotation and CellChat packages (version: 2.1.2) were employed for the analysis of cell interaction. Kyoto Encyclopedia of Genes and Genomes (KEGG, V93.0) pathway analysis and Gene Ontology (GO, v20190319) term analysis was performed to differential expressed genes (DEGs) using the clusterprofiler R package.

2.23. Statistical analysis

Statistical analysis and visualization for data of SnRNA-seq were executed by R software (version 4.4.2). Graphpad Prism 10.1.2 was utilized for statistic analysis to other data. Differences between groups were carried out by one-way analysis of variance (ANOVA), followed by Turkey's multiple comparison test. Behavioral data, including the results of the BBB score and inclined plane test, obtained from the same animals at multiple time points were analyzed using repeated measures two-way ANOVA followed by Tukey's multiple comparisons test. Each experiment was replicated more than three times independently, and the results were presented as mean ± standard deviation (SD). p < 0.05 was considered significant.

3. Results

3.1. Toxicity and effectiveness of IPA

IPA in a range of concentrations was added to the plates of PC12 cells for 24 h to prove its toxicity and effectiveness. The results of CCK8 test showed that there was no significant toxicity to PC12 cells when the concentration of IPA was lower than 2 mM (Fig. S1A). The concentration of LPS using in the following experiments was set as 5 μg/ml judged by similar method (Fig. S1B). Furthermore, after being treated by LPS (5 μg/ml), cell viability of PC12 cells was significantly increased when the concentration of IPA was 500 μM, which was selected for following studies (Fig. S1C).

3.2. Fabrication, characterization and biocompatibility of IPA@LIP

Owing to high lipid solubility of IPA based on its structure (Fig. 1A), The strategy of embedding it in the liposome membrane was proposed to overcome its defect of poor water solubility in SCI repair. In our study, IPA was mixed with DSPC, DSPE-PEG₂₀₀₀ and cholesterol to form the complex of IPA@LIP (Fig. 1B). The encapsulation efficiency and the loading rate of IPA were 62.14% and 8.70% respectively measured by ultraviolet spectrophotometer. Spherical morphology observed by TEM further showed that IPA@LIP was fabricated successfully (Fig. 1C). The diameter of IPA@LIP was 145.65 ± 4.76 nm and the polydispersity index (PDI) was 0.13, indicating that the distribution of particle size was uniform (Fig. 1D). Besides, zeta potential was −8.93 ± 3.44 mV according to the data of DLS. Stability tests over 14 days showed that there were minimal changes in both particle size and zeta potential, suggesting that IPA@LIP was stable for at least two weeks (Fig. 1E, F). Finally, in order to prove the biosecurity of IPA@LIP, PC12 cells were cultured in normal medium or medium containing IPA, liposome or IPA@LIP for 48 h, no significant difference was found in the percentage of dead cells among these groups (Fig. 1G).

Fig. 1.

Characterization and biocompatibility of IPA@LIP. (A) Chemical structure of IPA. (B) Schematic diagram of IPA loaded in liposome to form IPA@LIP. (C) Transmission electron microscopy (TEM) image of IPA@LIP. Scale bar = 100 nm. (D) Particle size distribution of IPA@LIP. (E) Variation of particle size of IPA@LIP with time (n = 3). (F) Change in zeta potential of IPA@LIP over time (n = 3). (G) Representative live/dead staining images in each group. Scale bar = 200 μm. Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA followed by Turkey's multiple comparison test.

3.3. IPA@LIP recovered cell proliferation, inhibited apoptosis and promoted axon regeneration in vitro

PC12 cells was selected to assess the effect of IPA@LIP on cell survival, proliferation and apoptosis in vitro. PC12 cells were divided into five groups (Control, IPA, LPS, LPS + LIP and LPS + IPA@LIP). Since inflammation is one of the major secondary injury mechanisms after SCI, LPS was applied to mimic the inflammatory microenvironment after SCI [[22], [23], [24]]. According to the results of EdU staining, IPA could slightly promote the proliferation of PC12 cells although there was no statistical difference between control group and IPA group. After treatment with LPS, the proliferation of PC12 cells was significantly inhibited and the effect of LPS could not be attenuated by the addition of liposome. However, IPA@LIP was able to reduce the inhibition on cell survival and proliferation caused by LPS to a certain extent (Fig. 2A, B). To evaluate the effects of IPA@LIP on alleviating cell apoptosis, TUNEL staining and flow cytometry were adopted. The result of TUNEL revealed that only few PC12 cells underwent apoptosis and apoptotic cells were hardly seen in many fields of view in the control and IPA group, wheres the number of apoptotic PC12 cells increased sharply after being stimulated by LPS. It was comforting that the percentage of apoptotic cells decreased effectively within the IPA@LIP group (Fig. 2C, D). Similar phenomenon observed by flow cytometry also indicated that IPA@LIP had the ability to inhibit apoptosis. In LPS group and LPS + LIP group, the apoptosis rates were close to 30%, which was much higher than that in control group. The usage of IPA@LIP could significantly decrease apoptosis rate of PC12 cells (Fig. 2E, F).

Fig. 2.

The function of IPA@LIP to PC12 cells and primary cortical neurons in vitro. (A) Representative EdU staining images of PC12 cells in each group. Scale bar = 100 μm. (B) Quantitative analysis of EdU staining of PC12 cells in each group (n = 3). (C) Representative TUNEL staining images of PC12 cells in each group. Scale bar = 100 μm. (D) Quantitative analysis of TUNEL staining of PC12 cells in each group (n = 3). (E) Flow cytometric analysis of apoptosis rate of PC12 cells in each group. (F) Quantitative analysis of apoptosis rate of PC12 cells in each group (n = 3). (G) Representative immunofluorescence staining images of PC12 cells in each group. Red represented Tuj-1. Scale bar = 50 μm. (H) Quantitative analysis of synapse lengths of PC12 cells in each group (n = 22). (I) Representative immunofluorescence staining images of primary cortical neurons in each group. Green represented NF200. Scale bar = 50 μm. (J) Quantitative analysis of neurite lengths of primary cortical neurons in each group (n = 14). Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA followed by Turkey's multiple comparison test. **p < 0.01, ****p < 0.0001.

To investigate the effect of IPA@LIP on promoting axon regeneration in vitro, PC12 cells and primary cortical neurons were used in our study. Synapse lengths of PC12 cells were 71.44 ± 9.12 μm and 75.39 ± 8.91 μm in control and IPA group. The pathological conditions produced by LPS dramatically inhibited the elongation of synapse, resulting in the lengths of synapse being only 33.05 ± 6.91 μm and 35.94 ± 6.71 μm in LPS and LPS + LIP group. Fortunately, synapse expanded to 50.33 ± 9.74 μm after being rescued by IPA@LIP (Fig. 2G, H). Considering that the synapses of PC12 cells were only morphologically similar to the axons of neurons, primary cortical neurons of rats were cultured for further study. Because of the ability of impeding axonal outgrowth of neurons, chondroitin sulfate proteoglycans (CSPGs) was applied to mimic inhibitory environment after SCI in vitro [26,27]. Neurite retraction of primary cortical neurons was observed clearly in CSPG and CSPG + LIP group and axons were significantly shortened to less than approximately 200 μm. Compared with control group, it was showed that the application of IPA alone was supposed to promote the elongation of neurites. In addition, IPA@LIP could antagonize the function of CSPGs in restricting axonal growth. In CSPG + IPA@LIP group, the length of neurites reached 330.21 ± 80.42 μm, proclaiming the function of IPA to foster axon regeneration (Fig. 2I, J).

3.4. IPA@LIP alleviated inflammation and regulated polarization of BV2 cells in vitro

To evaluate the influence of IPA@LIP to BV2 cells, the release of LDH was detected firstly and the result showed that the level of LDH release increased sharply after treatment with LPS. However, combined application of IPA@LIP was supposed to reduce this tendency, confirming the protection of IPA@LIP to BV2 cells (Fig. 3A). Then ELISA was used to measure the levels of inflammatory factors in the supernatant of medium. The levels of TNF-α, IL-6 and IL-1β in LPS group were similar with those in LPS + LIP group and much higher than those in LPS + IPA@LIP group. Besides, it was worth noting that the level of TNF-α in IPA group was substantially lower than that in control group (Fig. 3B–D). The protein expression levels of TNF-α, IL-6 and IL-1β were further examined by Western blot and the results of protein levels in BV2 cells showed a similar trend with the content of these inflammatory factors in supernatant. LPS increased their protein expression levels dramatically and IPA@LIP could rescue this phenomenon to a certain extent. In addition, the application of IPA alone decreased the level of IL-1β compared with that in control group (Fig. 3E–H). All of these results indicated that both IPA and IPA@LIP had the ability to alleviate inflammation and improve the immune microenvironment.

Fig. 3.

The function of IPA@LIP to BV2 cells in vitro. (A) The level of LDH release in each group (n = 6). (B) The level of TNF-α in each group (n = 3). (C) The level of IL-6 in each group (n = 3). (D) The level of IL-1β in each group (n = 3). (E) Representative Western blot results in each group. (F) Quantitative analysis of the protein expression level of TNF-α in BV2 cells (n = 3). (G) Quantitative analysis of the protein expression level of IL-6 in BV2 cells (n = 3). (H) Quantitative analysis of the protein expression level of IL-1β in BV2 cells (n = 3). (I) Quantitative analysis of the protein expression level of iNOS in BV2 cells (n = 3). (J) Quantitative analysis of the protein expression level of Arg-1 in BV2 cells (n = 3). (K) Flow cytometric analysis of CD86 positive cells in each group. (L) Quantitative analysis of CD86 positive cells rate in each group (n = 3). Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA followed by Turkey's multiple comparison test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Next, iNOS and Arg-1, which were representative markers of pro-inflammatory (M1) and anti-inflammatory (M2) phenotype respectively, were used to investigate the polarization of BV2 cells by Western blot. LPS or LPS + LIP stimulation increased the protein expression of iNOS and decreased that of Arg-1. Conversely, the expression level of iNOS was down-regulated and that of Arg-1 was up-regulated after the treatment of IPA@LIP. The increase in the protein expression of Arg-1 in IPA group was also detected compared with that in control group (Fig. 3E, I, J). To further confirm these findings, CD86 and CD206, which were also the markers of polarization, were selected to identify M1 and M2 types of BV2 cells by flow cytometry. It showed that the proportion of CD86 positive cells in IPA@LIP group was lower than that in LPS group or LPS + LIP group, indicating that the percentage of BV2 cells in M1 type decreased after the application of IPA@LIP (Fig. 3K, L). The results of both Western blot and flow cytometry demonstrated that IPA and IPA@LIP could regulate polarization of BV2 cells and transform BV2 cells from M1 type to M2 type.

3.5. Fabrication, characterization and biocompatibility of LIP-IPA@PHA-4P-CHO

Gelation was tested in tilted bottles after the mixture of 4-arm-PEG-CHO and PHA with different mass fractions. Obvious fluidity was still observed in hydrogel containing 5% PHA after being mixed for 5 min, indicating that it failed to form the gel in this condition. However, after the same period of time, samples with 7.5% and 10% PHA did not flow and maintained the stable gel state when the bottles were tilted at 45° for 10 s (Fig. 4A). Hydrogel with 7.5% PHA was selected as the material for the subsequent experiment. The cross-section of PHA-4P-CHO observed by SEM had a typical porous morphology, and its interior presented a highly interconnected and evenly distributed three-dimensional network structure (Fig. 4B). Fig. 4C showed the injectability and writing ability of PHA-4P-CHO, which were confirmed again by its shear thinning property. The apparent viscosity of the hydrogel decreased sharply with the increase of the shear rate, from 945.8125 to 3.9665 Pa s, reflecting a significant shear thinning behavior (Fig. 4D). G′ of hydrogel was maintained at 384.28 Pa under the small strain of 1%. However, as the strain increased, G′ and G″ also changed and eventually an intersection point (679.15%) was formed (Fig. 4E). Remarkably, when the strain was restored from 1000% to 1%, both G′ and G″ were able to return to their original levels (Fig. 4F). The cycling test showed that sample had a self-healing performance, which could also be confirmed by macroscopic experiment further. Under the condition of no external force, two pieces of samples could recover within 2 h (Fig. 4G). Moreover, the repaired hydrogel did not break after the tensile test. The maximum adhesion strength of PHA-4P-CHO was able to reach 38.1 kPa, demonstrating its excellent adhesion characteristic to the tissue (Fig. 4H). Fig. 4I showed the swelling rate of the hydrogel, which rose rapidly to 639.16 ± 15.40% within 12 h, and then the increasing rate slowed down gradually. At 18 h, it reached 668.04 ± 12.51% and tended to be stable.

Fig. 4.

Characterization of LIP-IPA@PHA-4P-CHO. (A) Gelation of hydrogel containing 5%, 7.5% and 10% PBA in tilted bottles. (B) Scanning electron microscopy (SEM) image of PHA-4P-CHO. Scale bar = 200 μm. (C) Injectability and writing ability of PHA-4P-CHO. (D) Shear thinning property of PHA-4P-CHO. (E) Mechanical property of PHA-4P-CHO. (F) Cycling test to detect the self-healing performance of PHA-4P-CHO. (G) Self-healing performance of PHA-4P-CHO. (H) Adhesion strength of PHA-4P-CHO. (I) Swelling rate of PHA-4P-CHO (n = 3). (J) Degradation rates of PHA-4P-CHO in PBS and H₂O₂ respectively (n = 3). (K) Intracellular ROS-responsive performance of PHA-4P-CHO. Scale bar = 200 μm. (L) Schematic diagram of IPA@LIP loaded in PHA-4P-CHO to fabricate LIP-IPA@PHA-4P-CHO. (M) Release rates of IPA@LIP from LIP-IPA@PHA-4P-CHO in PBS and H₂O₂ respectively (n = 3). (N) Representative in vivo images of rats using IPA@LIP or LIP-IPA@PHA-4P-CHO at several time points post injection.

To confirm the ROS-responsive property of PHA-4P-CHO, H₂O₂ solution was used to mimic pathological microenvironment with high level of ROS after SCI [28]. Samples of PHA-4P-CHO were immersed in PBS and H₂O₂ respectively to test their degradation rates. The result illustrated that the degradation rate of hydrogel in H₂O₂ solution was significantly faster than that in PBS. After immersing into solutions for 60 h, samples in H₂O₂ almost completely disappeared, at which point samples in PBS still had about 60% of their initial mass in contrast (Fig. 4J). Similar results were observed when determining intracellular ROS levels of BV2 cells using DCFH-DA as a probe, which in cells co-cultured with PHA-4P-CHO was dramatically lower than that in H₂O₂ group (Fig. 4K).

IPA@LIP was mixed with 4-arm-PEG-CHO and PHA to fabricate LIP-IPA@PHA-4P-CHO (Fig. 4L), and then it could be released continuously and slowly from hydrogel. The result demonstrated that the release of IPA@LIP accelerated in the environment with high level of ROS owing to the existence of PBA. The curves of release rates showed that the amount of IPA@LIP released in H₂O₂ reached the peak (appromixately 80%) at 60 h, which is consistent with the result of the degradation of PHA-4P-CHO in H₂O₂ mentioned above. However, at the same time point, only about 35% IPA@LIP was released in PBS. After that, the release of IPA@LIP in PBS continued to increase steadily until it reached the same peak of that in H₂O₂ at 336 h basically (Fig. 4M). Before confirming the characteristics of hydrogel in vivo, its biocompatibility was determined firstly. PHA-4P-CHO, LIP@PHA-4P-CHO and LIP-IPA@PHA-4P-CHO were placed on the dorsal surface of the spinal cords of rats respectively. After 28 days of operation, major organs of rats were taken for HE staining and the result declared that there was no significant difference among these three groups and Sham and SCI groups (Fig. S2). Finally, IPA@LIP was labeled with DiR and the IVIS Imaging System was used to evaluate the sustained release ability of this drug delivery system in vivo. Fluorescence signal was monitored continuously and the fluorescence intensity changes represented the remaining of IPA@LIP (Fig. 4N). The fluorescence signal weakened rapidly in mice injected with IPA@LIP solely and almost disappeared at 7 days post injection. In contrast, stable and significant fluorescence signal still could be seen in the spinal cords of rats using LIP-IPA@PHA-4P-CHO at 14 days post operation.

3.6. LIP-IPA@PHA-4P-CHO promoted functional recovery of SCI rats and improved the morphology of injured tissues

SCI model of rat at T10 level was established to explore the impacts of LIP-IPA@PHA-4P-CHO on functional recovery in vivo. After the formation of spinal cord hematoma, IPA@LIP, PHA-4P-CHO, LIP@PHA-4P-CHO and LIP-IPA@PHA-4P-CHO were placed on the dorsal surface of spinal cord to cover the lesion in different groups respectively. The recovery of motor function was evaluated by observing the movement of rats in the open field or on the inclined plane at different time points after SCI. BBB scores showed that rats in LIP-IPA@PHA-4P-CHO group had the best recovery of motor function and at each time points their BBB scores were higher than those of other groups. Compared with rats in SCI group, the motor function of rats in IPA@PHA group, PHA-4P-CHO group and LIP@PHA-4P-CHO group was improved slightly, and there was no significant difference in BBB scores among these groups (Fig. 5A). The results of inclined plane test demonstrated a similar trend with BBB scores (Fig. 5B). During the first 7 dpi, no significant difference was observed among SCI, IPA@LIP, PHA-4P-CHO and LIP@PHA-4P-CHO groups. From 7 dpi, rats in LIP-IPA@PHA-4P-CHO group exhibited significantly higher maximum maintained incline angles than all other groups. By 28 dpi, the maximum angle maintained by rats in LIP-IPA@PHA-4P-CHO group exceeded 30°. In contrast, rats in PHA-4P-CHO and LIP@PHA-4P-CHO groups performed slightly better than those in SCI and IPA@LIP groups, but their maximum maintained incline angle remained around 20°. In order to better analyze the motor function recovery of hind limbs in SCI rats in each group, Catwalk analysis was applied to record their footprints and related parameters at 28 dpi (Fig. 5C). Rats in SCI and IPA@LIP groups exhibited almost no voluntary hind limb movement, resulting in only sporadic hind limb paw prints being detected by the system. Although rats treated with PHA-4P-CHO and LIP@PHA-4P-CHO regained partial hind limb swing and weight support, their gait remained poorly coordinated. In contrast, rats in LIP-IPA@PHA-4P-CHO group displayed substantially improved hindlimb weight-bearing capacity during locomotion, accompanied by markedly enhanced gait coordination and joint mobility. Besides, MEPs were also recorded at 28 dpi to assess electrical activities of rats in each group (Fig. 5D–F). It was obvious that in SCI group, the latency prolonged significantly while the amplitude decreased sharply compared with Sham group. Administration of IPA@LIP alone produced only a slight increase in amplitude, with little effect on latency. Although there was certain recovery on both electrophysiological parameters after the application of PHA-4P-CHO or LIP@PHA-4P-CHO, the rescue effect was most significant in LIP-IPA@PHA-4P-CHO group.

Fig. 5.

Effects of LIP-IPA@PHA-4P-CHO on functional recovery in SCI rats and morphological changes in injured tissues in vivo. (A) The BBB scores in each group (n = 6). (B) The angle of inclined test in each group (n = 6). (C) Representative images of footprints and limbs' supporting timing view of Catwalk gait analysis at 28 dpi in each group. (D) Representative images of MEPs at 28 dpi in each group. (E) Quantitative analysis of the latency of MEPs at 28 dpi in each group (n = 6). (F) Quantitative analysis of the amplitude of MEPs at 28 dpi in each group (n = 6). (G) Representative images of gastrocnemius muscle at 28 dpi in each group. (H) Quantitative analysis of the weight of gastrocnemius muscle at 28 dpi in each group (n = 12). (I) Representative images of bladder tissue at 28 dpi in each group. (J) Quantitative analysis of the weight of bladder tissue at 28 dpi in each group (n = 6). Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA or repeated measures two-way ANOVA followed by Turkey's multiple comparison test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

The mass of gastrocnemius muscle was weighed at 28 dpi to evaluate the function of hind limb again from another perspective (Fig. 5G, H). The result indicated that muscle weight reduced significantly after SCI and treatment with IPA@LIP, PHA-4P-CHO or LIP@PHA-4P-CHO alone failed to produce a noticeable therapeutic effect. However, the atrophy of muscle could be alleviated owing to the treatment of LIP-IPA@PHA-4P-CHO. Likewise, the volume and weight of bladder increased sharply after SCI, which decreased only after the application of LIP-IPA@PHA-4P-CHO (Fig. 5I, J).

3.7. LIP-IPA@PHA-4P-CHO reduced spinal cord cavity formation, promoted neuronal survival and axonal regeneration, and regulated microglial polarization in vivo

HE staining was performed to observe the cavitation of spinal cord at 28 dpi, which was smallest in LIP-IPA@PHA-4P-CHO group. In IPA@LIP, PHA-4P-CHO and LIP@PHA-4P-CHO groups, there was no significant reduction in the damaged area of spinal cord compared with that in SCI group (Fig. 6A). Then IF staining was further conducted to identify the therapeutic effect of LIP-IPA@PHA-4P-CHO on neurons and microglia in vivo. Based on the IF staining of NeuN, only few neurons could be seen in the lesion of spinal cord in SCI, IPA@LIP, PHA-4P-CHO and LIP@PHA-4P-CHO groups at 28 dpi. However, the number of survival neurons increased sharply in LIP-IPA@PHA-4P-CHO group (Fig. 6B). In addition, the relative fluorescence intensity of NF200 was markedly reduced within the lesion cavities in SCI, PHA-4P-CHO and LIP@PHA-4P-CHO groups, whereas a modest increase was observed in IPA@LIP group. Notably, treatment with LIP-IPA@PHA-4P-CHO resulted in a significant increase of NF200 expression within the injured spinal cord regions (Fig. 6C). 5-HT staining was performed to evaluate axonal regeneration further. At 28 dpi, only limited axons were observed in SCI, PHA-4P-CHO and LIP@PHA-4P-CHO groups. In contrast, substantial axonal regeneration was found in LIP-IPA@PHA-4P-CHO group. Some regenerating axons were also detected distal to the lesion in IPA@LIP group, although the extent of regeneration was less pronounced than that observed following LIP-IPA@PHA-4P-CHO treatment (Fig. 6D). Next, iBa-1 and Arg-1 were applied together to detect the change in the polarization of microglia after the treatment of LIP-IPA@PHA-4P-CHO (Fig. 6E). Compared with Sham group, the relative intensity of Arg-1 in the damaged area were remarkably attenuated after SCI, which could not be rescued by IPA@LIP, PHA-4P-CHO or LIP@PHA-4P-CHO. But the rebound in fluorescence intensity was observed in LIP-IPA@PHA-4P-CHO group.

Fig. 6.

Effects of LIP-IPA@PHA-4P-CHO on cavitation, neurons and microglia of spinal cords in vivo. (A) Representative HE staining images of spinal cord in each group. Scale bar = 500 μm. (B) Representative immunofluorescence staining images of spinal cord in each group. Red represented GFAP, green represented NeuN and blue represented DAPI. Scale bar = 500 μm. (C) Representative immunofluorescence staining images of spinal cord in each group. Red represented GFAP, green represented NF200 and blue represented DAPI. Scale bar = 500 μm. (D) Representative immunofluorescence staining images of spinal cord in each group. Green represented 5-HT and blue represented DAPI. Scale bars = 500 μm (upper panels) and 100 μm (lower panels). (E) Representative immunofluorescence staining images of spinal cord in each group. Red represented iBa-1, green represented Arg-1 and blue represented DAPI. Scale bar = 500 μm.

3.8. The mechanism of IPA in SCI repair

SnRNA-seq was further utilized to explore the mechanism of IPA in SCI repair. After rigorous quality control and filtering, dimensionality reduction and cluster analysis were performed on the data from all samples and the results were showed in Fig. S3A, and then the 21 clusters were divided into 13 cell types according to marker genes (Fig. 7A, S3B). Fig. 7B displayed these marker genes used for annotating different cell types. The proportion of each cell type was calculated in each group (Fig. 7C). The three cell types with the highest proportion were oligodendrocyte, neuron and microglia. There was significantly higher proportion of microglia in SCI and IPA groups compared with Sham group. Next, the cell interactions between each cell types were analyzed and related ligand-receptor pairs were presented in Fig. S3C. As we focused on the cell chat between neuron and other cell types, it is worth noting that the ligand-receptor interaction of CX3CL1 and CX3CR1 only existed between neuron and microglia (Fig. 7D, E). Besides, the heat map demonstrated that the signal of intercellular communication mediated by CX3CL1-CX3CR1 was emitted by neurons and received by microglia (Fig. 7F). Then it was found that CX3CL1-CX3CR1, which existed in SCI and IPA groups, could not be observed in Sham group by comparing the ligand-receptor pairs in each group (Fig. 7G). Moreover, the further comparison between SCI and IPA group illustrated that the expression of CX3CL1-CX3CR1 was more significant in the latter group (Fig. 7H, S3D).

Fig. 7.

SnRNA-seq results of spinal cords. (A) UMAP plot of all samples with cell type annotation. (B) Dot plot of marker genes for each cell type. (C) Stacked bar chart of all cell types in each group. (D) Dot plot of the ligand-receptor pairs between neuron and other cell types. (E) Cell-cell interaction chord diagram of CX3CL1-CX3CR1. (F) Heat map of intercellular communication signals between neuron and microglia. (G) Dot plots of the ligand-receptor pairs between neuron and other cell types in each group. (H) Dot plot of the ligand-receptor pairs between neuron and other cell types in SCI and IPA group. (I) UMAP plot of cell subsets of neuron. (J) Stacked bar chart of cell subsets of neuron in each group. (K) Dot plot of KEGG pathways enriched among the DEGs of neuron between SCI and IPA group. (L) UMAP plot of cell subsets of microglia. (M) Stacked bar chart of cell subsets of microglia in each group. (N) Dot plot of KEGG pathways enriched among the DEGs of microglia between SCI and IPA group.

Furthermore, neuron was divided into 5 subsets, namely excitatory motor neuron, inhibitory neuron, CD9⁺ excitatory motor neuron, afferent neuron and interneuron (Fig. 7I, S4A). The marker genes utilized to distinguish cell subsets were shown in Fig. S4B. CD9⁺ excitatory motor neuron was differentiated from excitatory motor neuron because of the high expression of CD9. In addition, in SCI group, the proportion of this subset was significantly lower than that in Sham group, while in the IPA group, it rose to close to that in Sham group (Fig. 7J). Microglia was divided into 4 subsets, which were homeostatic microglia, disease-associated microglia, MHC-II microglia and proliferating microglia (Fig. 7L, S5A). The marker genes used to identify different cell subsets were displayed in Fig. S5B. The largest subset in Sham group was homeostatic microglia, whose proportion decreased sharply in SCI and IPA group. Correspondingly, the proportions of three other subsets increased in SCI and IPA groups, most notably disease-associated microglia (Fig. 7M). KEGG pathway analysis between SCI and IPA groups revealed that DEGs in neuron and microglia were both enriched in PI3K/AKT signaling pathway, axon guidance and neurotrophin signaling pathway. Additionally, DEGs in microglia were also enriched in chemokine signaling pathway, which might be related to CX3CR1 (Fig. 7K, N). As for GO term analysis, enriched pathways in neuron were related with PI3K/AKT pathway, synaptic vesicle, endocytosis and glial cell, such as phosphatidylinositol 3-kinase binding, vesicle-mediated transport in synapse, regulation of endocytosis and glial cell development (Fig. S4C). DEGs in microglia were enriched in pathways associated with neuron and transmembrane transport, such as myelination, distal axon and transmembrane transporter binding (Fig. S5C).

Next, the results of SnRNA-seq were confirmed further by IF staining, ELISA and Western blot. LY294002 was utilized as an inhibitor of PI3K. The distributions of neurons and microglia in spinal cord were observed and analyzed by IF staining (Fig. 8A). CX3CL1 was utilized to label neurons while microglia was tagged by CX3CR1 and Fascin-1 [29]. At 28 dpi, the expression of CX3CL1 in damaged area decreased sharply in SCI group. At the same time, although CX3CR1 could be observed in the injury site, Fascin-1 was only located at the border of lesion core and could not migrate towards the lesion center. In contrast, there was a significant increase in the concentration of CX3CL1 in lesion after the treatment of LIP-IPA@PHA-4P-CHO. Meanwhile, accompanied by the reduction of cavity area, CX3CR1 and Fascin-1 positive microglia exhibited pronounced centripetal migration toward the lesion center. Following LY294002 treatment, the expression of CX3CL1 in lesion decreased compared with IPA group. Moreover, the migration of CX3CR1 and Fascin-1 positive microglia toward the lesion center was markedly attenuated. To further quantify soluble CX3CL1, ELISA was performed using homogenates prepared from the injured spinal cord at 28 dpi. The level of soluble CX3CL1 increased progressively from Sham group to SCI group and reached its highest level in IPA group. Although LY294002 treatment partially reduced the concentration of soluble CX3CL1, its level remained higher than that in SCI group (Fig. 8B). The downstream signaling events associated with CX3CL1-CX3CR1 signaling were further evaluated by Western blot (Fig. 8C). The expression level of p-PI3K was significantly increased in IPA group and reduced in SCI group compared with Sham group (Fig. 8D). Similar expression patterns were observed for p-AKT and p-mTOR (Fig. 8E, F). In contrast, the protein expreesion of TLR4 was considerably increased in SCI group and decreased in IPA group (Fig. 8G). Notably, pharmacological inhibition of PI3K with LY294002 largely abolished the effects of IPA, resulting in reduced p-PI3K, p-AKT, and p-mTOR, accompanied by re-elevated TLR4 expression toward the levels observed in the SCI group.

Fig. 8.

The mechanism of IPA in SCI repair. (A) Representative immunofluorescence staining images of spinal cord in each group. Green represented CX3CL1, red represented CX3CR1, yellow represented Fascin-1 and blue represented DAPI. Scale bar = 500 μm. (B) The level of CX3CL1 in each group (n = 6). (C) Representative Western blot results in each group. (D) Quantitative analysis of the protein expression level of p-PI3K/PI3K (n = 6). (E) Quantitative analysis of the protein expression level of p-AKT/AKT (n = 6). (F) Quantitative analysis of the protein expression level of p-mTOR/TOR (n = 6). (G) Quantitative analysis of the protein expression level of TLR4 (n = 6). (H) Schematic illustration of the molecular mechanism of IPA in SCI repair. Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA followed by Turkey's multiple comparison test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

4. Discussion

Patients with SCI are often accompanied by intestinal dysfunction and imbalance of gut microbiota. The changes in intestinal flora after SCI are supposed to trigger persistent neurogenic inflammatory response to hinder nerve recovery and axon regeneration [30]. Some researches have indicated that fecal microbiota transplantation and dietary intervention are able to exert neuroprotective effects [31,32]. Thus it can be seen that the treatment targeting gut microbiota or its metabolites may be a potential approach to promoting SCI repair and functional recovery. As one of the three major metabolites of intestinal microorganisms, Tryptophan exhibits significant biological effects in neurological diseases [33]. IPA, which is produced by the decomposition of tryptophan by Clostridium sporogenes, Clostridium botulinum, and Clostridium caloritolerans in the intestinal tract, is an important metabolite of tryptophan and is expected to serve as a therapeutic target for SCI repair [34,35].

In order to overcome the deficit in practical applications and maximize the function of IPA in SCI repair, a kind of liposome was designed in this study and IPA was embedded in its membranes to form the complex of IPA@LIP. Our results of characterization show that not only are its encapsulation rate and drug loading rate both at relatively high levels compared with normal liposome, but also IPA@LIP is able to keep stable for a considerable period of time, which are consistent with our original design intention. In addition, the liposome we designed serves solely as a carrier for IPA without performing special biological functions, which is confirmed by the results of biosafety test and cell experiments. After cells were treated with LPS or CSPG, there was no significant difference between LPS/CSPG + LIP group and LPS/CSPG group. However, the rescue effects were always observed in LPS/CSPG + IPA@LIP group, indicating that it is IPA rather than liposome that regulates the cell phenotype. It is worth noting that the function of IPA does not be affected, even though it is encapsulated into the complex.

PC12 cells are usually selected for in vitro experiments in the research of SCI because they have the structure of synapses which are similar to neurites of neurons [36]. We firstly demonstrated that IPA can recover cell proliferation and inhibit apoptosis by PC12 cells, and then its function to promote axon regeneration was confirmed by both PC12 cells and primary cortical neurons. Subsequently, BV2 cells were used to mimic microglia in spinal cord. LDH release showed that IPA could protect BV2 cells against pyroptosis or excessive activation [37]. Further, the function of IPA on inhibiting inflammation and regulating polarization of microglia was verified by both Western blot and flow cytometry. All these results not merely align with previous research but point to the promising role of IPA in SCI repair.

In general, IPA is administered to animals by gavage or intraperitoneal injection [38,39]. However, both of these two methods involve the process of drug absorption and the dosage of IPA that can reach the lesion of spinal cord is uncertain, even little. Therefore, considering that in-situ release of IPA at the injury area of spinal cord may be a better approach, we incorporated the hydrogel into our biomaterial design. HA has been widely used in the basic research of SCI owing to its excellent biocompatibility and low immunogenicity [40]. However, the application of natural HA is limited because loaded substances can not be released directionally and slowly. Therefore, in this study HA was modified based on the pathological mechanism of SCI.

Taking into account that PBA can respond to high ROS level in the damaged area after SCI and complete the release of drugs through the breaking of chemical bonds [41], PBA-HA is designed for targeted release of IPA@LIP in this study by covalently attaching PBA group to HA through chemical modification. The reversible borate ester bond formed by PBA and the vicinal diol structure provides dynamic crosslinking points. Meanwhile, the stable acylhydrazone bond is generated by the dehydration and condensation of the aldehyde group (-CHO) of 4-arm-PEG-CHO and the acylhydrazine bond (-NH-NH₂) of ADH undergoing the Schiff Base reaction, achieving dual crosslinking of the hydrogel to enhance its stability. The hardness of the hydrogel increases and the viscoelasticity decreases with the increase of PHA. Additional pressure on the spinal cord caused by excessive hardness of hydrogel may lead to extra injury or inhibit its recovery. Therefore, hydrogel with 7.5% PHA was selected due to its appropriate hardness and viscoelasticity finally. Shear thinning property of PHA-4P-CHO indicates that hydrogel has excellent injectable performance and can reduce the viscosity due to the increase of shear stress during the injection process, which is convenient for minimally invasive implantation into the body through fine needles and accurately covering irregular surface of the spinal cord after injury. The results of mechanical property test directly proved that PHA-4P-CHO had enough strain resistance. The cycling test and the self-healing test shows further that PHA-4P-CHO had significant self-healing performance and it is supposed to still maintain enough mechanical strength after self-healing. The variation trend of the swelling rate indicates that water molecules diffuse rapidly into the hydrogel network at the initial stage. As time goes by, the swelling rate decreases owing to the increase of network extension resistance. In other words, this kind of hydrogel has the anti-swelling ability to some extent.

Porous morphology and three-dimensional network structure of PHA-4P-CHO observed by SEM endow the hydrogel with excellent long-lasting sustained-release performance. This kind of structure not only optimizes the loading environment of IPA@LIP, but also provide it with an ideal channel for the efficient diffusion. The ROS-responsive characteristic of PHA-4P-CHO was first confirmed by DCFH-DA in BV2 cells. Additionally, it was proved again by different degradation rates of hydrogel in PBS and H₂O₂. Due to the existence of PBA, IPA@LIP could be directed and guided to be released to locations with high ROS level. Both the release rate curve of IPA@LIP and the result of in vivo imaging demonstrates that this drug delivery system can achieve the sustained and slow release of IPA@LIP and effectively prolong the duration of IPA@LIP action in the lesion.

It should be noted that the pathological microenvironment after SCI is highly dynamic and contains elevated levels of ROS together with various enzymes, including hyaluronidase and matrix metalloproteinases. Since HA can undergo enzymatic degradation, the in vivo degradation of the HA-based hydrogel is unlikely to be exclusively dependent on oxidative stimuli. Therefore, the degradation and subsequent drug release observed in vivo are probably mediated by the combined effects of ROS and enzymatic processes. In this study, H₂O₂ was employed as a representative oxidative stimulus to verify the ROS-accelerated degradation behavior of the hydrogel. Consequently, the term “ROS-responsive” should be interpreted as preferentially accelerated degradation under oxidative conditions rather than absolute specificity toward ROS.

The results of in vivo experiments demonstrated that compared with rats in SCI group, rats treated with PHA-4P-CHO or LIP@PHA-4P-CHO exhibited modest improvements in locomotor function and electrophysiological performance. These limited improvements may be attributable to the interaction between ROS-responsive biomaterial PHA-4P-CHO and the microenvironment after SCI. However, these effects were relatively minor. In addition, there was no significant difference in histology results and IF staining among SCI, PHA-4P-CHO and LIP@PHA-4P-CHO group, indicating that ROS-responsive hydrogel we designed is an excellent drug delivery platform with little impact on SCI repair.

Only when hydrogel loaded with IPA@LIP was applied, could marked functional recovery be observed in rats, as evidenced by significantly improved locomotor performance, enhanced electrophysiological conduction, and reduced lesion size. Moreover, the expression levels of NeuN, NF200, 5-HT and Arg-1 were significantly increased, indicating enhanced neuronal survival, axonal preservation and regeneration, and polarization of microglia toward the reparative M2 phenotype. These findings demonstrate that IPA effectively promotes neurorestoration after SCI when delivered through the hydrogel-based delivery system. Notably, compared with SCI group, administration of an equivalent dose of IPA@LIP without hydrogel encapsulation failed to produce significant improvements in behavioral and electrophysiological outcomes even there was a slight effect on axonal regeneration. This limited efficacy is likely attributable to the inability of free IPA@LIP to achieve sustained retention and continuous therapeutic action within the hostile microenvironment after SCI. Therefore, the therapeutic benefits observed in LIP-IPA@PHA-4P-CHO group arise not only from IPA itself but also from the localized and sustained drug delivery enabled by the ROS-responsive hydrogel. These two components are complementary and both are indispensable for achieving optimal therapeutic efficacy.

Under the same condition of using hydrogel, the results of SnRNA-seq revealed that the application of IPA significantly enhanced the interaction of CX3CL1-CX3CR1, which only exists between neuron and microglia [42]. Therefore, one of the effects of IPA in promoting SCI repair may be achieved by regulating the interaction between neurons and microglia. CX3CL1, which is a kind of unique chemokine, is thought to be able to regulate the activation of microglia and subsequent inflammatory response by combining with its sole receptor CX3CR1 [43,44]. Microglia was simply believed to be detrimental to SCI repair by releasing pro-inflammatory factors in the past, but now increasing evidence declares that microglia plays a dual role in promoting SCI repair and neurotoxicity [45,46]. CX3CL1 is specifically expressed in neurons in CNS, while CX3CR1 may be expressed in macrophages and microglia [44]. So Fascin-1 was utilized in this study to distinguish microglia from macrophages with CX3CR1 together [29,47]. IF staining showed that after SCI there was notable decrease of CX3CL1 in lesion. Even CX3CR1 was able to be observed in the cavitation, the expression of Fascin-1 was extremely low, indicating that most CX3CR1 positive cells in the damaged area might be macrophages rather than microglia. Because previous studies found that due to the destruction of the blood-spinal cord barrier after SCI, macrophages can preferentially infiltrate the damaged area and hinder the centripetal migration of microglia, leading to the accumulation of microglia around the lesion and impeding reparative processes [48]. While in IPA group, apparent increase in the expression of both CX3CL1 and Fascin-1 could be seen in the injury site, proclaiming that IPA promotes the survival of neurons and the centripetal migration of microglia, which is the first mechanism of IPA in SCI repair.

Next, we explored the reason why IPA can enhance the intercellular communication between neuron and microglia mediated by CX3CL1-CX3CR1. It is worth noting that this kind of interaction also exists under physiological conditions rather than just emerging after SCI. CX3CL1 can be divided into membrane bound and soluble forms [42]. Under physiological conditions, membrane bound CX3CL1 predominates, whereas SCI promotes the generation of soluble CX3CL1 through proteolytic cleavage. The strengthened intercellular communication after the application of IPA may be attributed to the transformation of more CX3CL1 from membrane bound molecule to soluble molecule. Then the accumulation of soluble CX3CL1 in the damaged area of spinal cord promotes the centripetal migration of microglia and CX3CL1 binds to CX3CR1 on their surface. This can be further confirmed by ELISA results. Soluble CX3CL1 was significantly increased in the injured spinal cord after IPA treatment, suggesting that IPA enhanced the availability of CX3CL1 for neuron–microglia communication.

According to the results of KEGG pathway analysis, PI3K/AKT signaling was enriched in both neurons and microglia after IPA treatment. PI3K/AKT signaling pathway plays an important role in the development of SCI, which is associated with neuronal cell proliferation and differentiation, axon regeneration, inflammation, glial scar formation and so on [49,50]. Therefore, the enrichment of this pathway in neurons may reflect a direct neuroprotective and pro-regenerative effect of IPA and may contribute to the increased expression and availability of CX3CL1. Activation of this pathway can reduce the formation of glial scar, inhibit the cell apoptosis and regulate the polarization of microglia [51]. Combined with KEGG analysis and Western blot, it is deduced that IPA acts on neurons to increase the amount of soluble CX3CL1 and promote its binding with CX3CR1 to enhance the communication between neuron and microglia. Subsequently, in microglia that receive the signal from neuron, the PI3K/AKT pathway is also activated and TLR4 is inhibited, ultimately achieving the effects of axon regeneration and inhibition of neuroinflammation. The pharmacological inhibition experiment using LY294002 further supported this hypothesis. Notably, the direct neuroprotective effects of IPA mediated by activation of PI3K/AKT signaling in neurons and the downstream PI3K/AKT signaling activated in microglia through CX3CL1-CX3CR1 mediated neuron-microglia communication may represent two mechanistically distinct yet complementary pathways contributing to SCI repair. Consistent with this interpretation, LY294002 treatment partially attenuated the increase in soluble CX3CL1 induced by IPA, while the CX3CL1 level remained higher than that observed in SCI group. These findings suggest that PI3K/AKT signaling contributes to the regulation of CX3CL1 availability but is unlikely to be the sole mechanism responsible for the enhanced CX3CL1-CX3CR1 mediated intercellular communication induced by IPA.

Collectively, our findings support a model in which IPA exerts dual actions after SCI. On the one hand, IPA directly activates PI3K/AKT signaling in neurons, promoting neuronal survival and increasing the availability of CX3CL1. On the other hand, soluble CX3CL1 activates CX3CR1 on microglia, leading to activation of downstream PI3K/AKT signaling, suppression of TLR4-mediated neuroinflammation, enhanced microglial migration toward the lesion, and ultimately improved SCI repair (Fig. 8H).

The subsets of different cell types and potential mechanisms related to them are not be discussed thoroughly, which is a limitation in this study. The subset of neuron with the greatest difference among three groups is CD9⁺ excitatory motor neuron. CD9 is known as a marker of extracellular vesicle and the alteration in the proportion of CD9⁺ excitatory motor neuron subset may indicate that neuronal injury after SCI is associated with changes in extracellular vesicles. Previous research has demonstrated that microglia is able to participate in the process of injury and remodeling of synapse by internalizing the membranous-vesicular fragments which come from synaptic terminals [52]. The biological process of GO term analysis in our study also showed that the enriched pathways in neuron were connected with extracellular vesicle and glia, including vesicle-mediated transport in synapse, synaptic vesicle and glial cell development. Therefore, the function of extracellular vesicles which originate from neuron and its interaction with microglia may be a potential direction of SCI repair.

5. Conclusion

In summary, we developed a ROS-responsive hydrogel platform that enables localized and sustained delivery of IPA@LIP to the injured spinal cord, thereby maximizing the therapeutic efficacy of IPA. This treatment markedly promoted neurological recovery by preserving neuronal viability, facilitating axonal regeneration, modulating microglial polarization and alleviating neuroinflammation. Mechanistically, IPA exerted dual neurorestorative actions: it directly activated PI3K/AKT signaling pathway in neurons to support neuronal survival, while simultaneously enhancing CX3CL1-CX3CR1 mediated neuron-microglia communication, which was associated with activation of downstream PI3K/AKT signaling pathway and suppression of TLR4-mediated inflammatory responses in microglia. Collectively, these findings identify CX3CL1-CX3CR1 mediated neuron-microglia crosstalk as an important mechanism underlying the therapeutic effects of IPA and demonstrate the potential of ROS-responsive hydrogel-mediated IPA delivery as a promising strategy for spinal cord injury repair.

CRediT authorship contribution statement

Longyu Li: Writing – original draft, Visualization, Software, Methodology, Funding acquisition, Data curation, Conceptualization. Yuhao Zhang: Visualization, Software, Data curation, Conceptualization. Zhishuo Wang: Validation, Software, Data curation. Jiehan Li: Visualization, Software, Methodology, Formal analysis. Hao Han: Data curation. Jiaming Zhang: Data curation. Chang Liu: Investigation. Chunfeng Shang: Investigation. Guowei Shang: Investigation. Yanhui Ji: Investigation. Junjie Guo: Investigation. Yiran Xu: Resources, Methodology. Huan Zhang: Resources. Xiaohua Jiang: Resources. Keya Mao: Resources. Kin Chiu: Writing – review & editing, Validation. Kwow-Fai So: Writing – review & editing, Supervision, Project administration. Songfeng Chen: Writing – review & editing, Project administration, Funding acquisition, Formal analysis, Conceptualization. Zikuan Leng: Writing – review & editing, Project administration, Funding acquisition, Formal analysis, Conceptualization. Hongjian Liu: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.

Declaration of competing interest

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.

Data availability

Data will be made available on request.