3D bioprinted multifunctional GelMA/TMP scaffold integrated with neural stem cell-derived extracellular vesicles and neural progenitor cells for spinal cord injury repair

Abstract

Spinal cord injury (SCI) disrupts neural architecture through a cascade of inflammatory, vascular, and glial responses that collectively create a regenerative deadlock. Overcoming this complex, temporally evolving pathology requires the coordinated delivery of structural, cellular, and biochemical cues. Here, we present a 3D bioprinted multifunctional scaffold composed of gelatin methacryloyl (GelMA), tetramethylpyrazine (TMP), neural progenitor cells (NPCs), and neural stem cell-derived extracellular vesicles (NSC-EVs). This combinatorial construct mimics essential features of the neural niche and orchestrates reparative processes across multiple levels. Compared to adipose-derived EVs, NSC-EVs demonstrated a superior cytokine and neurotrophic profile that enhanced angiogenesis and neuronal differentiation. In vitro, the integrated scaffold promoted NPC survival, neurogenesis, angiogenesis and immunomodulation. In a complete transection rat SCI model, the scaffold supported locomotor recovery by reducing cystic cavitation, facilitating axonal regeneration and remyelination, preserving parenchymal integrity, and attenuating neuroinflammation. Our findings suggest that integrated, multimodal interventions can modulate the hostile post-injury microenvironment and stimulate endogenous repair mechanisms, offering a clinically translatable paradigm for SCI regeneration.

Introduction

Spinal cord injury (SCI) presents a major clinical and biological challenge because it involves multiple interrelated pathological processes that evolve dynamically over time and space. The primary mechanical insult disrupts both neural parenchyma and vasculature, while long-term disability primarily results from secondary injury cascades such as inflammation, oxidative stress, ischemia, excitotoxicity, and cell death, which collectively exacerbate damage to gray and white matter.^1,2 The ensuing microenvironment becomes highly unfavorable for regeneration, characterized by blood–spinal cord barrier (BSCB) breakdown, excessive cytokine release, reactive gliosis, and scar formation. ³

Recent single-cell and spatial transcriptomic studies have revealed that the injured spinal cord is a heterogeneous tissue comprising diverse cellular populations that undergo dynamic spatiotemporal changes.^4,5 These include pro-inflammatory microglial and macrophage subtypes, reactive astrocytes with divergent transcriptional states, dysregulated oligodendrocyte progenitor cells (OPCs), and perivascular fibroblasts that contribute to fibrotic scarring.^2,6 Ligand–receptor signaling among these populations shapes either supportive or inhibitory conditions for regeneration in a context-dependent manner. This intricate milieu not only impedes axonal regrowth and remyelination but also limits the survival and integration of both endogenous and transplanted regenerative cells. Efforts to modulate single pathological processes (e.g. inflammation, gliosis, or neurodegeneration) have typically produced only partial or transient recovery in preclinical models, underscoring the need for multifaceted therapeutic strategies.^7–10 Consequently, research has increasingly focused on tissue engineering approaches to address the broad challenges of SCI repair.^11,12

Biomaterial scaffolds provide both structural support and a tunable interface for regulating local cellular behavior.^13–15 Among them, gelatin methacryloyl (GelMA) hydrogels have attracted attention for their favorable physicochemical properties, including biocompatibility, mechanical adjustability, and resemblance to the native extracellular matrix.^16,17 These characteristics enable GelMA to support cell adhesion, survival, and differentiation in neural contexts. ¹⁸ However, given the multifactorial pathophysiology of SCI, hydrogel scaffolds alone often fail to induce substantial functional recovery, motivating their integration with regenerative cell populations and paracrine signaling components.^19,20

Neural progenitor cells (NPCs) are promising candidates for such combinatorial approaches. With their capacity for self-renewal, multipotency, and trophic factor secretion, NPCs can contribute both to cell replacement and to modulation of the injury environment. Preclinical studies have demonstrated that NPC transplantation enhances axonal growth, remyelination, and synaptic reconnection while mitigating inflammation.^21–23 Moreover, 3D bioprinting can improve NPC viability by embedding stem cells directly within bioinks during scaffold fabrication.^24,25 Nevertheless, the hostile post-injury microenvironment continues to limit NPC survival and integration, highlighting the need for additional modulators.

Extracellular vesicles (EVs) derived from neural stem cells (NSCs) have emerged as potent cell-free therapeutics.^26–28 Compared with whole-cell therapies, EVs offer greater stability and a reduced risk of immune rejection. By delivering bioactive proteins, regulatory RNAs, and lipid mediators, NSC-derived EVs (NSC-EVs) can modulate inflammation, oxidative stress, and apoptosis; promote neurogenesis; preserve mitochondrial function; and support neuronal survival. Their incorporation into biomaterial scaffolds allows localized, sustained release that complements NPC activity.^29–32

In parallel, modulation of intracellular stress responses using small molecules provides an additional axis of therapeutic control. Tetramethylpyrazine (TMP), a bioactive compound derived from Ligusticum wallichii, exhibits antioxidant, anti-inflammatory, and mitochondrial-protective effects.^33,34 TMP upregulates peroxisome proliferator-activated receptor-gamma coactivator 1-alpha (PGC-1α), a central regulator of cellular energy metabolism and neuronal survival, and suppresses apoptotic cascades in SCI models. These properties make TMP an attractive adjunct to cell- and EV-based therapies for mitigating secondary injury.

Here, we developed a multifaceted therapeutic platform based on a 3D bioprinted GelMA scaffold integrating structural, cellular, and biochemical components (NPCs, NSC-EVs, and TMP). In the complex spatiotemporal landscape of SCI, our approach seeks to synchronize neurogenesis, axonal elongation, remyelination, angiogenesis, and immunomodulation. We evaluate this strategy through in vitro assays and an in vivo rat transection model, providing new insights into combinatorial neuroregenerative design and its translational potential for SCI repair.

Materials and methods

Cell culture

NPCs, provided by Dong-Youn Hwang’s Laboratory at CHA University, were differentiated from H9 embryonic stem cells (WiCell, Madison, WI, USA; catalog no. WICE001; RRID: CVCL_9773). NPCs were cultured in DMEM/F12 medium supplemented with N-2 Supplement (1×; Thermo Fisher Scientific, MA, USA), B-27 Supplement (1×; Thermo Fisher Scientific), epidermal growth factor (EGF, 20 ng/ml; Peprotech, NJ, USA), and basic fibroblast growth factor (bFGF, 20 ng/ml; Peprotech). NSCs were cultured in DMEM/F12 medium containing 2% penicillin-streptomycin, B-27 Supplement (1×; Thermo Fisher Scientific), heparin (2 μg/ml; H3149, Sigma-Aldrich, MO, USA), bFGF (30 ng/ml), and EGF (30 ng/ml). Human adipose-derived mesenchymal stem cells (ASCs; Lonza, Basel, Switzerland) were maintained in α-MEM (HyClone Laboratories, UT, USA) supplemented with Antibiotic-Antimycotic (100×, 1%; AA, 15240062, Gibco, NY, USA) and 10% Premium Fetal Bovine Serum (FBS; 35015CV, Corning, NY, USA). Human umbilical vein endothelial cells (HUVECs; Lonza) were cultured using the Endothelial Growth Medium-2 BulletKit (EGM-2 BulletKit, CC-3162, Lonza, MD, USA). All cell types were incubated at 37°C in a humidified atmosphere with 5% CO₂.

Cell viability assay

Cell viability was assessed using the Cell Counting Kit-8 (CCK-8; Dojindo, Kumamoto, Japan). The assay was performed in accordance with the manufacturer’s protocol to determine relative cell viability. Absorbance was measured at 450 nm using a microplate reader (Molecular Devices, CA, USA).

Extracellular vesicle isolation

For ASC-EV isolation, ASCs were seeded and cultured until they reached more than 70% confluence. The culture medium was then replaced with serum-free medium, and conditioned media were collected every 12 h over a 48 h period (four total collections). For NSC-EV isolation, the entire culture medium was harvested at each media change throughout the culture period. NSCs, supplied by Seung-Woo Cho’s Laboratory at Yonsei University, were maintained until approximately 90% confluence, and all collected media were used for subsequent EV isolation.

Collected media were centrifuged at 1300 rpm for 3 min to remove cells, followed by filtration through a 0.22 μm vacuum filter/storage bottle system to eliminate large non-exosomal particles such as debris, microvesicles, and apoptotic bodies. EVs were then isolated using tangential flow filtration (TFF; Repligen, Waltham, MA, USA) with a 500 kDa molecular-weight-cutoff filter, and the diafiltration rate was set to 7. The isolated EVs were concentrated using an Amicon Ultra-15 centrifugal filter unit (Merck, Darmstadt, Germany).

Characterization of extracellular vesicles

The quantity and size distribution of EVs were analyzed using the MONO ZetaView^® system (PMX-120, Particle Metrix, Meerbusch, Germany) operated in 488 nm scatter mode. EV samples were diluted in filtered PBS (HyClone Laboratories, UT, USA) to a concentration of 10⁷–10⁸ particles/ml. Parameters were optimized for all samples with sensitivity = 75, shutter = 100, minimum trace length = 15, and cell temperature = 25°C. Transmission electron microscopy (TEM; Hitachi H-7600, 80 kV, Japan) was employed to visualize EV morphology. The EV suspension was placed on a Formvar/copper grid with a 150-mesh carbon coating (FCF150-CU, Electron Microscopy Sciences, USA) and dried. EVs were negatively stained with 7% uranyl acetate or gadolinium acetate solution and air-dried on the copper grid. The prepared grid was placed in a grid box for TEM examination after drying.

Cytokine array

Radio-immunoprecipitation assay (RIPA) buffer (Rockland Immunochemicals, PA, USA) was used to lyse equal numbers of EVs (1.0 × 10¹⁰ particles). The lysed EV solutions were applied to the nitrocellulose membrane of the Proteome Profiler Human XL Cytokine Array Kit (R&D Systems, MN, USA). After development with ChemiDoc XRS+, cytokine signal intensity was quantified using ImageLab software. The mean intensity values were used for analysis and expressed as cytokine array intensities in log₂ scale. Functional annotation and pathway enrichment analyses were performed using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases via DAVID (Database for Annotation, Visualization and Integrated Discovery; https://david.ncifcrf.gov/).

Enzyme-linked immunosorbent assay (ELISA)

Angiogenic factors (hepatocyte growth factor (HGF), angiopoietin-1 (Ang-1), and basic fibroblast growth factor (bFGF) in EVs were quantified using Quantikine™ ELISA kits (R&D Systems, MN, USA). Equal amounts of EVs (1 × 10⁹ particles) were loaded into each ELISA well. The procedure followed the manufacturer’s protocol. Absorbance was measured at 450 nm using a microplate reader, with background correction at 540 and 570 nm, respectively.

Tube formation assay

HUVECs (1 × 10⁵ cells/well) were seeded onto 24-well plates (Corning, NY, USA) coated with Matrigel and cultured for 18 h. ASC-EVs and NSC-EVs were administered at equivalent concentrations (1 × 10⁸ particles/ml) using a transwell system. After Calcein AM staining, tube formation was observed under a fluorescence microscope (CKX53, OLYMPUS, Japan). Image analysis was conducted using the Angiogenesis Analyzer plugin in ImageJ (Wayne Rasband, NIH, USA). For quantitative analysis, multiple imaging fields were acquired from each independent biological replicate. The values obtained from these fields were averaged to generate a single data point per replicate. Thus, the statistical unit (n) represents biological replicates, and multiple imaging fields were not treated as independent samples.

Scaffold 3D printing and characterization

A bioink consisting of 5 wt% GelMA, 3 wt% gelatin, 0.25 wt% LAP, TMP 400 μM, and 1 × 10⁷ cells/ml was 3D-printed using a Dr. INVIVO bioprinter (ROKIT Healthcare, Seoul, Korea) equipped with a 27 g nozzle. The printing bed temperature was maintained at 13°C. Following printing, the constructs were photo-crosslinked under UV light for 3 s and then trimmed using a 4 mm biopsy punch (Kai Industries Co., Ltd., Seki, Japan) to achieve uniform scaffold size.

Photo-crosslinking was performed using the built-in UV irradiation module of the DR. INVIVO 3D bioprinter. The printed constructs were exposed to UV light for 3 s directly on the printing bed immediately after fabrication. The UV wavelength, light intensity, exposure geometry, and source-to-sample distances are preset by the manufacturer and are not user-adjustable parameters. The GelMA group consisted of scaffolds containing only GelMA and gelatin. The NPC@GelMA/TMP/NSC-EV scaffold was fabricated by printing GelMA, gelatin, TMP, and NPCs, followed by post-loading with NSC-EVs.

Immunocytochemistry (ICC)

Cells were fixed in 4% paraformaldehyde at room temperature for 10 min and permeabilized for 10 min with 0.2% Triton X-100 in PBS. Samples were then blocked with 1% BSA in PBS and incubated overnight at 4°C with primary antibodies, followed by a 1 h incubation at room temperature in the dark with secondary antibodies. The antibodies used were anti-βIII-Tubulin (Tuj1; Abcam, ab18207, 1:300), MAP2 (Abcam, ab32454, 1:300), and Alexa Fluor 488 donkey anti-rabbit IgG (H + L; Invitrogen, A-21206, 1:200). Nuclei were counterstained with Hoechst (Thermo Scientific, 62,249, 1 µg/ml). Immunofluorescence images were captured using a fluorescence microscope (CKX53, OLYMPUS, Japan).

Confirming modification of inflammation factors

Tumor necrosis factor (TNF)-α (10 ng/ml) and the scaffolds were applied to NPCs 24 h after cell seeding, with the scaffolds positioned in the upper compartment using a transwell system. After 24 h, to determine the level of inflammation, RNA for qPCR (Quantitative real-time PCR) was extracted. The SYBR Green PCR reagent mix (Applied Biosystems, CA, USA) was applied to real-time PCR. QuantStudio 3 (Applied Biosystems, CA, USA) was employed to perform reactions with the following primers: NF-kB: forward, 5’- CGGGATGGCTTCTATGAGG-3’, and reverse, 5’-CTCCAGGTCCCGCTTCTT-3’; interleukin (IL)-6: forward, 5’-GATGAGTACAAAAGTCCTGATCCA-3” and reverse, 5’- CTGCAGCCACTGGTTCTGT-3”; IL-8: forward, 5’-AGACAGCAGAGCACACAAGC-3” and reverse, 5’-ATGGTTCCTTCCGGTGGT-3”; 18S rRNA: forward, 5’-GCAATTATTCCCCATGAACG-3” and reverse, 5’-GGGACTTAATCAACGCAAGC-3.” The data were quantified using the 2^−ΔΔCt method with 18S rRNA as a reference.

Animal care

All animal procedures were conducted in compliance with the Institutional Animal Care and Use Committee (IACUC) guidelines of CHA University School of Medicine (protocol no. 240146). To reduce stress and ensure consistent experimental outcomes, animals were acclimatized for 1 week before surgery. Environmental conditions were maintained at 24 ± 3°C and 55%−65% humidity with a 12 h light/dark cycle. Food and water were provided ad libitum. To prevent xenogeneic immune rejection, cyclosporine A (Cypol-N, Chongkundang, South Korea) was administered orally via drinking water starting 2 days before cell injection and continued throughout the experimental period.

Surgical procedures and postoperative care

A total of 56 Sprague-Dawley rats were randomly divided into the following groups: sham (n = 6), SCI (n = 14), GelMA (n = 14), GelMA/TMP/NSC-EV (n = 14), and NPC@GelMA/TMP/NSC-EV (n = 14). Anesthesia was induced by intraperitoneal injection of a zoletil (zolazepam/tiletamine) and xylazine mixture (1:1; Zoletil, 50 mg/kg, Virbac Laboratories, France; Rompun, 10 mg/kg, Bayer, Korea). Animals were placed on a heating pad maintained at 39°C until the end of surgery. After anesthesia induction, the dorsal hair was shaved and sterilized with povidone-iodine. A 2–2.5 cm longitudinal incision was made at the T10 spinal level. Paraspinal muscles surrounding the T9–T11 vertebrae were carefully detached, and the T10 lamina was removed using bone-cutting forceps to expose the spinal cord. Complete spinal cord transection was performed using a surgical blade in all groups except the sham group, which underwent only laminectomy. A 1.8 mm spinal cord segment was surgically excised in all groups. Following resection, the inherent elasticity of the spinal cord caused the stumps to retract, resulting in a final lesion gap of approximately 2.0 mm. This phenomenon allows for the implantation of a 2.0 mm scaffold that is retained securely between the transected segments, a method consistent with previous protocols. ³⁵ In the treatment groups, this post-retraction gap accommodated the scaffold implantation without additional compression. Hemorrhage was controlled by gentle gauze packing, and once hemostasis was achieved, the scaffold was implanted at the lesion site using a spatula and forceps.

After scaffold transplantation, the muscles and skin were sutured, and animals received 0.9% saline, cefazolin (Chongkundang, Korea), and ketoprofen (SCD, Korea) postoperatively. Manual bladder expression was performed three times daily until spontaneous bladder function recovered.

Behavior tests

Functional recovery of hindlimbs was evaluated using the Basso, Beattie, and Bresnahan (BBB) locomotor rating scale. Rats were recorded moving freely in an open field for at least 5 min, one trial per time point, on days 1, 4, and 7 DPI and subsequently once per week. To ensure objectivity and minimize observer bias, video recordings were analyzed independently by two investigators who were strictly blinded to the experimental groups throughout the 28-day period. The left and right hindlimbs were scored separately for each animal at each time point, and the scores were averaged to obtain a single BBB score per animal. The final value used for statistical analysis was calculated as the means of the scores assigned by the two blind investigators. Data are presented as mean ± standard error of the mean.

Molecular and histological analyses

RT-qPCR was conducted to assess gene expression. Three rats from each group were randomly selected and euthanized 28 days post-injury using CO₂ asphyxiation. A 2 cm spinal cord segment encompassing 1 cm rostral and caudal to the injury epicenter was excised and immediately frozen at −80°C. Before analysis, frozen tissue samples were thawed and homogenized in TRIzol reagent (#15596018, Thermo Fisher Scientific, Waltham, MA, USA) for RNA extraction. RNA was isolated via phase separation with chloroform and precipitation with isopropanol, and its concentration and purity were determined using a NanoDrop spectrophotometer. High-quality RNA was reverse-transcribed into complementary DNA (cDNA) using a reverse transcription premix (#25081, NtRON Biotechnology, Seongnam, Korea) at 45°C for 1 h, followed by enzyme inactivation at 95°C for 5 min in a thermal cycler (T100, Bio-Rad, Hercules, CA, USA). The synthesized cDNA was combined with gene-specific primers (Bioneer, Daejeon, Korea) and SYBR Green Master Mix (#EBT-1801, ELPIS-BIOTECH, Daejeon, Korea) in a reaction mixture that was dispensed into a 96-well plate and amplified using a CFX Connect Real-Time PCR Detection System (Bio-Rad). The qPCR amplification protocol included an initial denaturation step of 3 min, followed by 40 cycles of denaturation at 95°C for 10 s, annealing at 60°C for 10 s, and extension at 72°C for 30 s. Target gene expression levels were normalized to the housekeeping gene GAPDH, and relative expression was calculated using the 2^−ΔΔCt method.

Histological analysis

For hematoxylin and eosin (H&E) staining, 5 μm-thick paraffin-embedded spinal cord sections were deparaffinized, rehydrated, and sequentially incubated in hematoxylin, alcoholic HCl solution, ammonia solution, and eosin. Following dehydration and clearing through graded ethanol and xylene, sections were mounted using Canada balsam (Junsei, #23255-1210, Tokyo, Japan).

For Luxol Fast Blue (LFB) staining, deparaffinized sections were incubated overnight at room temperature in LFB dye (Abcam, #ab150675). The slides were differentiated in lithium carbonate solution, followed by alcohol reagent treatment. Nuclei were counterstained with cresyl echt violet for 5 min. The sections were dehydrated, cleared, and mounted in Canada balsam. Both H&E- and LFB-stained slides were scanned using a Zeiss Axio Scan Z1 digital slide scanner (Carl Zeiss, Oberkochen, Germany). Image acquisition and analysis were performed with ZEN software (ZEN 3.1 Blue Edition) for quantitative histological evaluation.

Immunofluorescence staining

Immunofluorescence staining was performed on longitudinal spinal cord sections (3 μm thick). Sections were deparaffinized, rehydrated, and subjected to heat-mediated antigen retrieval in Tris-EDTA buffer (Biosesang, #TR2220-050-90, Yongin, Korea) at 95°C for 15 min. To minimize nonspecific binding, sections were blocked with 3% bovine serum albumin (BSA; #A8806, Sigma-Aldrich) for 2 h at room temperature. Primary antibodies diluted in 1% BSA were applied and incubated overnight at 4°C. The following day, Alexa Fluor 488- and 568-conjugated secondary antibodies (1:200, Invitrogen) were applied for 1 h, followed by nuclear counterstaining with DAPI for 10 min. Slides were mounted using fluorescence mounting medium (Dako Omnis, Agilent, #S3023) and imaged using a Zeiss Axio Scan Z1 digital slide scanner (Carl Zeiss, Germany). Image analysis was performed with ZEN software (ZEN 3.1 Blue Edition) to evaluate fluorescence intensity and distribution.

Data analysis and statistical methods

All experiments were performed at least in triplicate to ensure reproducibility. Data are expressed as mean ± standard deviation (SD) unless otherwise indicated. Statistical significance was defined as follows: ns (not significant), *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. Quantitative analyses of H&E, LFB, and immunofluorescence images were performed using ZEN software (ZEN 3.1 Blue Edition) after digital imaging with a Zeiss Axio Scan Z1.

For behavioral assessments, including BBB locomotor scoring, a mixed-effects model with Geisser-Greenhouse correction was employed to account for potential violations of sphericity in repeated-measures data. Other quantitative datasets were analyzed using one-way analysis of variance (ANOVA) to detect differences among groups, followed by Tukey’s post hoc test for multiple comparisons. Statistical analyses were conducted using GraphPad Prism (GraphPad Software, San Diego, CA, USA), with p < 0.05 considered statistically significant.

Results

Extracellular vesicle isolation and characterization

To maximize the neuroregenerative potential of NPCs, the generation of suitable EVs was a critical prerequisite. NSCs are known to facilitate NPC differentiation and regeneration; therefore, NSC-EVs were employed in this study. To comparatively assess the regenerative and differentiation-supportive effects of NSC-EVs, ASCs were selected as a control, given their accessibility and well-documented regenerative capacity. ³⁶

EVs from both cell types were isolated using TFF, a method we have consistently employed in previous studies.^37,38 The particle size distribution of the isolated EVs is shown in Figure 1(a). The mean particle size of ASC-EVs was 143 nm, whereas that of NSC-EVs was 125 nm, both within the typical EV size range of 50–150 nm. ³⁹ To confirm the identity of the isolated vesicles, we performed Western blotting according to the MISEV 2023 guidelines, evaluating representative surface markers (CD63), intracellular proteins (tumor susceptibility gene 101; TSG101), and a negative marker (apolipoprotein A1; Apo-A1; Figure 1(b)). ⁴⁰ Both EV types expressed CD63 and TSG101, while Apo-A1 was absent, confirming successful EV isolation. Transmission electron microscopy (TEM) revealed the characteristic cup-shaped morphology typical of EVs in both ASC-EVs and NSC-EVs (Figure 1(c)). To explore the functional potential of the vesicles, cytokine array analysis was conducted to profile regeneration-related factors (Figure 1(d)). Quantification of array intensities (Figure 1(e)) demonstrated that NSC-EVs contained a broader spectrum of regenerative cytokines than ASC-EVs. The differential cytokine profile supports the interpretation that NSC-EVs have greater regenerative potential than ASC-EVs.

Figure 1.

Characterization of extracellular vesicles (EVs): (a) size distribution of ASC-EVs and NSC-EVs measured using ZetaView, (b) western blot analysis of representative EV markers, (c) TEM images showing EV morphology; scale bars = 100 nm, (d) representative cytokine array membranes from ASC-EVs and NSC-EVs, and (e) scatter plot comparing cytokine array expression profiles of ASC-EVs and NSC-EVs.

EV internal factors and functionality

To identify functionally relevant factors, we selected cytokines with an intensity value above 20 from the cytokine array membrane. These were considered as the top-expressed factors in each group. Consistent with the earlier observations, NSC-EVs exhibited a broader spectrum of highly expressed factors compared to ASC-EVs, as shown in Figure 2(a).

Figure 2.

Internal factors of extracellular vesicles (EVs): (a) comparison of representative cytokine expression levels. (“Protein level” represents the log2-transformed intensity values of cytokine array spots shown in Figure 1(d), as described in the Methods section, reflecting the relative abundance of individual cytokines contained in NSC-EVs, and (b) GO-BP, GO-CC, GO-MF, and KEGG pathway analyses from DAVID based on cytokines highly expressed in NSC-EVs.

To gain further insight into the biological functions of these EV-derived proteins, we performed functional enrichment analysis using the Database for Annotation, Visualization and Integrated Discovery (DAVID), including GO Biological Process (GO-BP), Cellular Component (GO-CC), Molecular Function (GO-MF), and KEGG pathway analysis (Figure 2(b)). The GO-BP terms such as “positive regulation of cell migration,” “positive regulation of angiogenesis,” “negative regulation of apoptotic process,” “cell-cell signaling,” and “angiogenesis” suggest that NSC-EVs may actively participate in various regenerative processes. Furthermore, the enrichment of “extracellular region” and “extracellular space” under GO-CC supports the extracellular vesicular origin of the identified factors. In the GO-MF terms, significant terms including “cytokine activity,” ⁴¹ “growth factor activity,” ⁴² “chemokine activity,” ⁴³ “receptor ligand activity,” ⁴⁴ and “chemoattractant activity” ⁴⁵ indicate that NSC-EVs contain a diverse array of signaling molecules that may modulate cellular communication, enhance neural cell survival, promote migration, and support differentiation thereby contributing to neural regeneration. KEGG pathway analysis further identified significant enrichment in pathways such as the JAK-STAT signaling pathway, ⁴⁶ PI3K-Akt signaling pathway, ⁴⁷ MAPK signaling pathway, Rap1 signaling pathway, and Ras signaling pathway. ⁴⁸ These pathways are broadly associated with tissue regeneration, immune and inflammatory regulation, vascular function, and wound healing, suggesting that NSC-EVs possess enhanced regenerative potential through coordinated activation of multiple signaling networks.

Taken together, these results indicate that NSC-EVs contain a wider range of bioactive factors compared to ASC-EVs and are likely to exert more potent regenerative effects through modulation of diverse signaling pathways.

Angiogenic effect of EVs

Gene ontology biological process (GO-BP) enrichment analysis from the earlier DAVID results identified significant terms including “positive regulation of angiogenesis” and “angiogenesis,” suggesting that NSC-EVs may enhance vascular formation. To validate this prediction, we hypothesized that NSC-EVs contain a higher abundance of pro-angiogenic factors. ELISA was performed to quantify HGF, Ang-1, and bFGF in NSC-EVs and ASC-EVs (Figure 3(a)). All three angiogenic factors were significantly more abundant in NSC-EVs compared with ASC-EVs.

Figure 3.

Angiogenic effect of extracellular vesicles (EVs): (a) quantification of angiogenesis-related factors (HGF, Angiopoietin-1, and bFGF) in ASC-EVs and NSC-EVs via ELISA, (b) representative fluorescence images of tube formation assays; scale bars = 200 μm, and (c) quantitative analysis of total tube length, total branching length, number of meshes, and number of nodes (mean ± SD). Statistical significance determined by one-way ANOVA followed by Tukey’s post hoc test. ns: not significant.

*p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001.

To further confirm these effects at the cellular level, a tube formation assay was performed to assess endothelial angiogenesis in vitro (Figure 3(b)). HUVECs treated with NSC-EVs exhibited markedly enhanced tube formation, characterized by increased network complexity and interconnectivity. Quantitative analysis revealed significant increases in total tube length, total branching length, mesh count, and node number in the NSC-EV-treated group relative to the ASC-EV group (Figure 3(c)). Collectively, these findings demonstrate that NSC-EVs possessed superior angiogenic capacity compared to ASC-EVs, providing a mechanistic basis for their enhanced neuroregenerative and pro-differentiation potential.

Enhancement of neural differentiation dependent on scaffold

To examine how scaffold composition affects NPC behavior, we investigated whether incorporating TMP and NSC-EVs could enhance the neurogenic potential of GelMA-based scaffolds. Based on these components, three distinct scaffold formulations were constructed, and their effects on NPC viability and lineage commitment were systematically compared.

We first evaluated the cytocompatibility of TMP in NPCs to determine its optimal concentration for scaffold integration (Figure 4(a)). Among the tested concentrations, 400 μM TMP yielded the highest cell viability, whereas 600 μM showed cytotoxic effects. Accordingly, 400 μM was selected for subsequent experiments.

Figure 4.

Scaffold characterization and neural differentiation: (a) NPC viability at various TMP concentrations, (b) SEM images of GelMA, GelMA/TMP, and GelMA/TMP/NSC-EVs scaffolds; scale bars = 500 μm, (c) immunocytochemical staining for Tuj1 expression in NPCs, (d) quantified Tuj1 fluorescence intensity, (e) immunocytochemical staining for MAP2 expression in NPCs, and (f) quantified MAP2 fluorescence intensity. Data presented as mean ± SD; statistical significance determined by one-way ANOVA followed by Tukey’s multiple comparison test; each data point represents an independent biological replicate, with multiple imaging fields averaged per replicate prior to statistical analysis. ns: not significant.

*p < 0.05. **p < 0.01. ***p < 0.001.

A GelMA-based bioink was then prepared by incorporating TMP at the optimal concentration and 3D-printed to fabricate scaffolds. In parallel, NSC-EVs, previously demonstrated to promote neuronal differentiation, were incorporated to produce three groups: GelMA, GelMA/TMP, and GelMA/TMP/NSC-EVs. Scanning electron microscopy (SEM) revealed that all scaffolds exhibited well-defined, interconnected porous architectures conducive to cell infiltration, viability, and differentiation (Figure 4(b)).

To address the second major limitation of NPC therapy, restricted differentiation, we examined the expression of neuronal markers. Tuj1, an early neuronal differentiation marker, was most strongly expressed in NPCs cultured on the GelMA/TMP/NSC-EVs scaffold (Figure 4(c) and (d)), indicating that the combined scaffold environment enhanced neurogenic commitment. This effect appeared to be primarily mediated by NPCs themselves rather than solely through EV signaling. Furthermore, MAP2, a neuronal maturation marker, showed the highest expression in the GelMA/TMP/NSC-EVs group, with a smaller yet notable increase in scaffolds containing NSC-EVs but no NPCs (Figure 4(e) and (f)), suggesting a direct contribution of NSC-EVs to neuronal differentiation.

Biocompatibility was further assessed by monitoring NPC viability over time (Supplemental Figure S1). While GelMA and GelMA/TMP scaffolds exhibited comparable viability on days 1, 3, and 5, the GelMA/TMP/NSC-EVs scaffold showed significantly improved cell survival on days 3 and 5, consistent with the pro-survival effects previously observed for NSC-EVs.

Collectively, these findings demonstrate that the NPC@GelMA/TMP/NSC-EV system enhanced both NPC survival and neuronal differentiation, providing a promising multifunctional platform for spinal cord regeneration.

Angiogenic and anti-inflammation effects of scaffolds

To evaluate the regenerative potential of the engineered scaffolds, angiogenic activity was assessed (Figure 5(a)). Among the tested groups, the NPC@GelMA/TMP/NSC-EV scaffold exhibited the most robust pro-angiogenic effect. Quantitative analyses confirmed this observation, as total tube length, branching length, mesh count, and node number were all significantly higher in the NPC@GelMA/TMP/NSC-EV group (Figure 5(b)), indicating that enhanced angiogenesis may contribute to the scaffold’s superior regenerative performance. Accordingly, the increased mesh number observed in this group is interpreted as the result of the combined actions of structural support provided by the GelMA scaffold, the cytoprotective and anti-inflammatory effects of TMP, and the paracrine signaling mediated by NSC-EVs, rather than the mechanistic effect of any single component. In addition to angiogenesis, we assessed the anti-inflammatory capacity of the implanted scaffolds to determine whether modulation of local inflammation could further promote tissue repair (Figure 5(c)). Expression levels of key pro-inflammatory cytokines, including IL-8, IL-6, and NF-κB, were markedly reduced in the NPC@GelMA/TMP/NSC-EV group compared to other groups, demonstrating potent anti-inflammatory effects. Together, these results indicate that the NPC@GelMA/TMP/NSC-EV scaffold promoted spinal cord regeneration through dual mechanisms: stimulation of angiogenesis and suppression of inflammation.

Figure 5.

Angiogenic and anti-inflammation effects of scaffolds: (a) representative images of tube formation assay evaluating angiogenesis using scaffolds, (b) quantitative analysis of total length, branching length, mesh count, and node number from the tube formation assay, and (c) expression levels of inflammatory markers IL-6, IL-8, and NF-κB. Data presented as mean ± SD; statistical significance determined by one-way ANOVA followed by Tukey’s multiple comparison test; each data point represents an independent biological replicate, with multiple imaging fields averaged per replicate prior to statistical analysis. ns: not significant.

**p < 0.01. ***p < 0.001. ****p < 0.0001.

In vivo transplantation of NPC@GelMA/TMP/NSC-EV promoted hindlimb functional recovery

To evaluate the therapeutic efficacy of the multifunctional scaffold in vivo, a complete spinal cord transection model was established at the T10 level, creating a 2 mm lesion gap (Figure 6(a)). Animals were randomly assigned to five groups: sham (laminectomy only), SCI (injury without treatment), GelMA scaffold, GelMA/TMP/NSC-EVs scaffold, and NPC@GelMA/TMP/NSC-EV scaffold. Locomotor recovery was assessed over 28 days post-injury (DPI) using the BBB locomotor rating scale. All SCI-induced rats initially exhibited complete hindlimb paralysis (BBB = 0 at 1 DPI). Gradual improvements were observed across treatment groups; however, significant intergroup differences emerged from 7 DPI onward. By 14 DPI, animals treated with the NPC@GelMA/TMP/NSC-EV scaffold demonstrated significantly higher BBB scores (3.02 ± 0.55, p < 0.05 vs SCI; 2.63 ± 0.59, p < 0.05 vs GelMA; Figure 6(b) and (c)), indicating accelerated functional recovery. At 28 DPI, the same group achieved a mean BBB score of 5.86 ± 0.66, indicative of coordinated hindlimb movement across all joints. This performance was significantly superior to the untreated SCI group (3.61 ± 0.64, p < 0.0001). The GelMA/TMP/NSC-EVs group also exhibited improved locomotor recovery compared with SCI animals (1.38 ± 0.48, p < 0.05), though the effect was less pronounced than that of the NPC-containing scaffold. These results suggest that the inclusion of NPCs was critical for functional restoration and that the combinatorial scaffold, integrating structural, cellular, and biochemical elements, provided a microenvironment conducive to neurological regeneration following SCI.

Figure 6.

Implanting NPC@GelMA/TMP/NSC-EV into the complete rat spinal cord transection model improved functional recovery and reduced cyst formation: (a) schematic diagram of in vivo experimental design, (b) representative images showing hindlimb functional recovery at 28 DPI, (c) BBB locomotor rating scale assessment, (d) gross morphology of perfused spinal cord tissue at 28 DPI, (e) longitudinal H&E-stained sections showing lesion-site morphology at 28 DPI, and (f) quantitative analysis of cystic volume at the injury epicenter. ImageJ analysis; data expressed as mean ± SD; statistical significance determined by one-way ANOVA followed by Tukey’s multiple comparison test. ns: not significant.

*p < 0.05. ****p < 0.0001.

To ensure that the long-term oral administration of Cyclosporine A did not induce systemic toxicity or gastrointestinal adverse effects that could confound behavioral assessments, body weight was monitored throughout the 28-day experimental period (Supplemental Figure S2). At baseline, no significant differences in body weight were observed among any of the groups (p > 0.05), ensuring a comparable starting condition. Following surgery, a significant difference was noted between the Sham group and the injury groups at 7, 14, and 21 DPI (p < 0.05); this weight reduction is attributable to the physiological and metabolic stress of the severe spinal cord transection procedure rather than drug toxicity, as all groups received identical CsA treatment. Crucially, no statistically significant differences were found between the untreated SCI group and any of the scaffold-treated groups (GelMA, GelMA/TMP/NSC-EV, or NPC@GelMA/TMP/NSC-EV) at any time point (p > 0.05). These findings confirm that CsA administration was well-tolerated across all experimental conditions and did not introduce confounding variables related to general health or growth retardation.

NPC@GelMA/TMP/NSC-EV bridged the injury gap and reduced cyst formation

At 28 DPI, spinal cord tissues were harvested to assess lesion morphology and tissue integrity. Gross anatomical inspection revealed distinct morphological differences among groups (Figure 6(d)). In the untreated SCI group, the lesion site displayed extensive cystic cavitation, tissue necrosis, and darkened areas characteristic of chronic neuroinflammation and glial scarring. In contrast, scaffold-treated groups (GelMA, GelMA/TMP/NSC-EV, and NPC@GelMA/TMP/NSC-EV) exhibited progressive lesion-site filling, reduced cavitation, and improved tissue continuity. Notably, the NPC@GelMA/TMP/NSC-EV group showed the most cohesive architecture, with the lesion area appearing morphologically continuous and better aligned with adjacent parenchyma, suggesting enhanced scaffold integration and architectural restoration.

Histological analysis using H&E staining further corroborated these findings (Figure 6(e)). The sham group maintained intact spinal cord architecture, whereas the SCI group showed profound disruption, including large cystic cavities and dense immune cell infiltration, indicative of persistent inflammation and neurodegeneration. GelMA scaffolds partially bridged the lesion but did not prevent cavity formation. The GelMA/TMP/NSC-EV group showed moderate preservation of parenchymal structure, with smaller cysts and reduced inflammatory infiltration, though residual disorganization and glial scarring remained. In contrast, the NPC@GelMA/TMP/NSC-EV scaffold yielded the most significant architectural restoration, with dense cellular alignment and continuous morphology at the host–scaffold interface, indicating effective neural tissue reconstruction.

Quantitative analysis of cystic cavity volume further supported these histological observations (Figure 6(f)). Compared with the sham group (4.14 ± 1.90%), the SCI group exhibited a dramatic increase in cystic area (29.15 ± 2.97%, p < 0.0001). While the GelMA scaffold (24.54 ± 2.83%) provided a modest reduction relative to SCI (p < 0.05), the multifunctional scaffolds elicited much more robust tissue preservation. Both the GelMA/TMP/NSC-EV (14.53 ± 2.14%, p < 0.0001) and NPC@GelMA/TMP/NSC-EV (12.15 ± 3.34%, p < 0.0001) groups significantly reduced lesion volume compared to the SCI and GelMA groups. Although the difference between the two combinatorial groups was not statistically significant, the NPC-loaded group exhibited the lowest absolute cystic volume. These data highlight the contribution of NPCs to scaffold-mediated preservation of spinal cord structure and integrity.

NPC@GelMA/TMP/NSC-EV enhanced axonal growth across the injury site by attenuating gliosis

To characterize the cellular composition within the lesion cavity, we first examined the presence of neurons and glial cells, critical mediators of functional recovery following SCI. Immunofluorescence staining for Tuj1, a marker of early neuronal differentiation, was performed at the lesion epicenter to assess neurogenesis (Figure 7(a)). In the SCI group, Tuj1 expression was markedly diminished, consistent with extensive neuronal and axonal loss. In contrast, both GelMA/TMP/NSC-EV and NPC@GelMA/TMP/NSC-EV scaffolds induced robust Tuj1⁺ signals, with the highest intensity observed in the NPC@GelMA/TMP/NSC-EV group. This finding suggests that the combinatorial scaffold markedly enhances neuronal regeneration. Quantitative analysis confirmed these observations: the SCI group exhibited significantly reduced Tuj1⁺ area compared with sham, while both GelMA/TMP/NSC-EV (p < 0.001) and NPC@GelMA/TMP/NSC-EV (p < 0.0001) treatments significantly increased Tuj1 expression (Figure 7(b)). These results highlight the scaffold’s potent neurogenic capacity and underscore the contribution of NSC-EVs and NPCs to neural repair.

Figure 7.

NPC@GelMA/TMP/NSC-EV enhance axonal regeneration by attenuating glial scar formation and promote remyelination: (a) Immunofluorescence staining of neuronal marker Tuj1 (green) and astrocytic marker GFAP (red), (b and c) quantitative analysis of Tuj1⁺ and GFAP⁺ areas at and around the lesion epicenter, (d) relative GFAP mRNA expression 28 days post-injury, (e) immunofluorescence staining of neurofilament (NF; green) and myelin basic protein (MBP; red), (f) quantitative analysis of NF⁺ axonal area, (g) relative GAP43 mRNA expression 28 days post-injury, (h) quantitative analysis of NF⁺MBP⁺ myelinated axons, (i) LFB staining of longitudinal spinal cord sections assessing myelination, and (j) quantification of LFB⁺ area at and around the epicenter. Nuclei counterstained with DAPI; analyzed using ZEN 3.1 Blue Edition; data presented as mean ± SD; statistical significance determined by one-way ANOVA followed by Tukey’s post hoc test.

ns: not significant.

*p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001.

Reactive gliosis and glial scar formation were evaluated by analyzing glial fibrillary acidic protein (GFAP) expression, a canonical marker of astrocytic activation (Figure 7(a)). GFAP⁺ signals were sparse at the lesion epicenter but concentrated at the lesion borders, consistent with the formation of a glial scar barrier. The SCI group displayed the highest GFAP immunoreactivity, indicating severe astrocytic activation and scar formation (Figure 7(c)). Both GelMA/TMP/NSC-EV and NPC@GelMA/TMP/NSC-EV scaffolds significantly reduced GFAP⁺ area (p < 0.0001), with comparable levels of attenuation between the two treatments (13.8 ± 1.15 and 14.3 ± 1.22, respectively). qRT-PCR analysis further confirmed these results, showing markedly elevated GFAP mRNA levels in the SCI group, while both scaffold-treated groups exhibited significant reductions (GelMA/TMP/NSC-EV: 1.18 ± 0.09, p < 0.0001; NPC@GelMA/TMP/NSC-EV: 1.29 ± 0.09, p < 0.0001; Figure 7(d)). These findings demonstrate that the multifunctional scaffold effectively suppresses reactive astrogliosis and glial scar formation, creating a microenvironment conducive to axonal regeneration.

NPC@GelMA/TMP/NSC-EVs preserved myelin in the intact spinal cord and promoted remyelination at the lesion epicenter

Restoration of neural connectivity following SCI requires both axonal regrowth and remyelination to re-establish functional signal conduction. To evaluate axonal regeneration, immunofluorescence staining for neurofilament (NF), a structural marker of mature axons, was performed (Figure 7(e)). The SCI group exhibited markedly diminished NF expression, reflecting severe axonal degeneration and disruption of longitudinal tracts. In contrast, the NPC@GelMA/TMP/NSC-EV group displayed substantial axonal regrowth, with a significantly larger NF⁺ area compared to the SCI group (6.60 ± 1.01; p < 0.0001; Figure 7(f)). qRT-PCR analysis of GAP43, a gene associated with axonal sprouting, further corroborated these results: GAP43 expression was severely downregulated in the SCI group but significantly upregulated in NPC@GelMA/TMP/NSC-EV-treated animals (1.68 ± 0.07; p < 0.0001; Figure 7(g)). These data collectively indicate that the engineered scaffold promoted reconnection of disrupted neural pathways and structural repair.

To assess remyelination, co-localization of NF and myelin basic protein (MBP) was evaluated as an indicator of functional myelin sheath regeneration (Figure 7(e)). In contrast to the SCI group, both GelMA/TMP/NSC-EV and NPC@GelMA/TMP/NSC-EV groups exhibited significantly increased NF–MBP co-localization (p < 0.0001; Figure 7(h)). The NPC@GelMA/TMP/NSC-EV group demonstrated the greatest extent of myelinated axons, confirming its efficacy in promoting remyelination. To further assess myelin integrity, LFB staining was performed (Figure 7(i)). The sham group displayed intense, uniform LFB staining, indicating preserved myelinated white matter, whereas the SCI group exhibited severe demyelination, with extensive loss of LFB-positive areas and large cystic cavities (Figure 7(j)). Scaffold-treated groups showed progressive improvement in myelin preservation. Both GelMA and GelMA/TMP/NSC-EV scaffolds exhibited partial restoration of LFB staining and reduced cavitation relative to the SCI group, while the NPC@GelMA/TMP/NSC-EV group showed the most extensive myelin preservation, with densely packed, continuous fibers indicative of enhanced remyelination and white matter repair. (Figure 7(i)) Together, these findings demonstrate that the NPC@GelMA/TMP/NSC-EV scaffold promotes both axonal regeneration and remyelination, restoring neuronal circuitry and improving structural and functional recovery following SCI.

NPC@GelMA/TMP/NSC-EV supported neuronal differentiation of grafted NPCs, restored vascular integrity, and promoted angiogenesis

To examine neuronal differentiation within the lesion site, immunofluorescence staining for MAP2 and HuNu was performed in the NPC-containing scaffold group (Figure 8(a)). The colocalization of HuNu⁺ and MAP2⁺ cells confirmed that implanted NPCs differentiated into neurons within the lesion area. Notably, MAP2⁺ neuronal signals were also observed in close proximity to HuNu⁺ cells without direct overlap, suggesting that transplanted NPCs may exert paracrine effects on host tissue, enhancing neuronal survival and neurite outgrowth. High-magnification imaging further revealed extensive neurite extension and integration into surrounding tissue, underscoring the scaffold’s ability to provide a neurogenic microenvironment that supports neuronal differentiation and maturation.

Figure 8.

NPC@GelMA/TMP/NSC-EV supports NPC survival and differentiation, while incorporating NSC-EV enhances the expression of angiogenic factors and tight-junction proteins at the lesion site: (a) immunofluorescence staining of MAP2 (green) and human nuclei (HuNu; red) indicating neuronal differentiation of grafted NPCs, (b) immunofluorescence staining of VEGF-A (green), an angiogenic marker, and occludin (red), a tight-junction protein associated with BSCB structure, and (c and d) quantitative analysis of VEGF-A⁺ and occludin⁺ areas at the epicenter. Nuclei counterstained with DAPI; analysis performed using ZEN 3.1 Blue Edition; data expressed as mean ± SD; statistical significance determined by one-way ANOVA followed by Tukey’s multiple comparison test. ns: not significant.

*p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001.

Effective spinal cord repair requires sufficient vascularization to sustain nutrient and oxygen delivery while supporting the re-establishment of BSCB-associated structures. Promoting vascular network formation is essential not only for metabolic support but also for potentially limiting inflammatory cell infiltration at the lesion site. To assess the scaffold’s angiogenic potential and the presence of barrier-associated proteins, immunofluorescence staining was conducted for vascular endothelial growth factor A (VEGF-A), a key angiogenic mediator, and occludin, a tight-junction protein indicative of BSCB structural composition (Figure 8(b)). Both scaffold-treated groups exhibited upregulated VEGF-A expression, with the GelMA/TMP/NSC-EV and NPC@GelMA/TMP/NSC-EV groups showing the most pronounced increases (6.19 ± 1.52 and 6.93 ± 1.56; p < 0.001; Figure 8(c)). Similarly, occludin expression was significantly elevated in these groups compared with SCI controls (5.77 ± 1.23 and 6.80 ± 1.32; p < 0.001; Figure 8(d)). These findings suggest that grafted NPCs, in conjunction with NSC-EV and TMP, not only support neuronal regeneration but also promote the expression of angiogenic factors and tight-junction proteins, together contributing to a favorable post-injury microenvironment.

Modulation of neuroinflammation and microglial activation

Neuroinflammation is a critical determinant of secondary injury and long-term outcomes following SCI. The phenotypic state of microglia and macrophages strongly influences the inflammatory milieu, dictating whether it favors neurodegeneration or repair. To assess this response, immunofluorescence staining for CD68 and iNOS was performed as indicators of pro-inflammatory (M1-type) microglial and macrophage activation (Figure 9(a)). In the SCI group, extensive CD68⁺iNOS⁺ cell infiltration was observed, indicating persistent neuroinflammation. In contrast, scaffold-treated groups exhibited reduced expression of these inflammatory markers, with the NPC@GelMA/TMP/NSC-EV group showing the greatest suppression. Quantitative analysis confirmed a significant reduction in CD68⁺iNOS⁺ area in the NPC@GelMA/TMP/NSC-EV group compared with the SCI group (39.82 ± 3.36; p < 0.0001). (Figure 9(b))

Figure 9.

NPC@GelMA/TMP/NSC-EV attenuates the immune response by modulating macrophage polarization and reducing microglial activation (a, c and f) representative immunofluorescence images of spinal cord tissue sections at the lesion epicenter 28 days post-injury. Nuclei are counterstained with DAPI (blue): (a) staining for pan-macrophage marker CD68 (green) and pro-inflammatory M1 macrophage marker iNOS (red); scale bars: 1000 µm (left), 100 µm and 200 µm (right columns), (b) quantitative analysis of the relative CD68⁺iNOS⁺ (M1-like) area, (c) staining for CD68 (green) and anti-inflammatory M2 macrophage marker CD163 (magenta/purple); scale bars: 100 µm, (d) quantitative analysis of the relative ratio of CD163⁺/CD68⁺ (M2-like) area, (e) relative mRNA expression levels of CD163 (M2 macrophage polarization marker), (f) staining for microglial marker Iba-1 (green) and neuroinflammatory receptor P2X7R (red); insets show high-magnification details of cell morphology; scale bars: 1000 µm (left), 10 µm (inset), 100 µm (right), (g) quantitative analysis of the relative ratio of Iba-1⁺ area, (h) quantitative analysis of the relative ratio of Iba-1⁺P2X7R⁺ area, (i) relative mRNA expression of Heme oxygenase 1 (HO-1; oxidative stress/anti-inflammatory marker), and (j) relative mRNA expression of Beclin 1. Data are presented as mean ± SD; statistical significance was determined using one-way ANOVA followed by Tukey’s multiple comparison test.

ns: not significant.

*p < 0.05. ***p < 0.001. ****p < 0.0001.

Beyond suppressing the pro-inflammatory phenotype, we investigated whether the scaffold actively promoted the repolarization of macrophages toward the reparative M2 state. To validate this, IF stining was performed for CD163, a marker for M2 macrophages, in conjunction with the pan-macrophage marker CD68 (Figure 9(c)). The SCI group was characterized by a massive infiltration of CD68⁺ cells with only a small proportion of CD163 expression, indicative of a pervasive, stalled pro-inflammatory M1 state. In contrast, while the GelMA/TMP/NSC-EV scaffold induced a moderate upregulation of CD163, the NPC@GelMA/TMP/NSC-EV group elicited the most robust phenotypic shift. Quantitative analysis of the CD163⁺ to CD68⁺ areal ratio confirmed a significant elevation in the NPC-containing group compared to both the SCI control (p < 0.0001) and the GelMA group (p < 0.0001; Figure 9(d)). qRT-PCR analysis further revealed that CD163 expression was markedly downregulated in the SCI group, while scaffold implantation progressively increased CD163 expression, with the NPC@GelMA/TMP/NSC-EV group showing the highest level of M2 polarization relative to the SCI group (0.65 ± 0.05; p < 0.0001; Figure 9(e)). This phenotypic shift from M1 to M2 macrophages suggests that the multifunctional scaffold fosters an immunomodulatory microenvironment conducive to tissue repair.

To further characterize the immune response, immunofluorescence staining for Iba-1 and P2X7R was performed to assess microglial activation (Figure 9(f)). The NPC@GelMA/TMP/NSC-EV group showed the most pronounced reduction in Iba-1⁺ area, indicating attenuation of microglial activation and a shift toward a less inflammatory state (Figure 9(g); 10.08 ± 0.65; p < 0.01). Moreover, P2X7R, a purinergic receptor implicated in neuroinflammatory signaling, was markedly elevated in the SCI group and remained high in the GelMA/TMP/NSC-EV group (Figure 9(h)). In contrast, the NPC@GelMA/TMP/NSC-EV group exhibited a significant decrease in Iba-1⁺P2X7R⁺ colocalization (6.67 ± 0.51; p < 0.0001), indicating that the scaffold effectively suppressed purinergic receptor–mediated neuroinflammation. The concurrent reduction in P2X7R-expressing microglia supports this conclusion, as P2X7R activation is closely associated with chronic neuroinflammation and microglial cytotoxicity.

Additionally, in the NPC@GelMA/TMP/NSC-EV group, expression of heme oxygenase-1 was reduced compared with the untreated SCI group, reflecting diminished oxidative stress and a reduced requirement for endogenous cytoprotective activation (Figure 9(i)). Conversely, beclin-1 expression was significantly elevated, indicating enhanced autophagy that may promote clearance of cellular debris and damaged organelles, thereby supporting neuroprotection and facilitating functional recovery (Figure 9(j)). Collectively, these findings demonstrate that the NPC@GelMA/TMP/NSC-EV scaffold modulates the post-injury immune milieu by suppressing pro-inflammatory microglial activation, mitigating purinergic receptor–mediated neurotoxicity, and promoting a reparative, anti-inflammatory phenotype that supports neural tissue recovery.

Discussion

Orchestrating regeneration via a phase-matched combinatorial strategy

This study establishes a proof of concept for a combinatorial therapeutic strategy that leverages 3D bioprinting to spatially and temporally coordinate regeneration in the injured spinal cord. We engineered a multifunctional scaffold composed of GelMA, TMP, NSC-EVs, and NPCs to address the complex, multifactorial pathology of SCI. Our findings demonstrate that diverse pathological barriers, including neuroinflammation, glial scarring, and neuronal loss, can be mitigated concurrently while promoting neuroregeneration. The interaction between scaffold components represents a multifunctional integration rather than a simple additive effect: each element enhances the bioactivity, signaling dynamics, and functional engagement of the others within the lesion microenvironment.

Previous studies have shown that 3D bioprinted scaffolds facilitate neural repair in preclinical models of SCI.^35,49–51 When combined with growth factors or bioactive molecules, such scaffolds have demonstrated improved regenerative outcomes.^52,53 However, most existing strategies fail to align with the phase-specific demands of spinal cord repair. SCI is not a static event but a dynamic cascade evolving from acute inflammation and vascular disruption to chronic scarring and atrophy. The failure of single-modality interventions often stems from a mismatch between therapeutic stimulus and the pathological phase. A key distinction of our platform lies in its phase-matched combinatorial design, rather than relying on complex programed release, the biological components naturally target distinct temporal phases of post-injury recovery, from early immune modulation mediated by TMP and NSC-EVs to mid-phase neural differentiation driven by NPCs, and late-stage structural support of neural and vascular networks. ⁵⁴

In the acute phase, the immediate availability of TMP and NSC-EVs modulates the hostile inflammatory milieu. The observed inhibition of oxidative stress markers dampens early excitotoxicity, protecting grafted cells. This immunomodulation is critical for facilitating a shift in the local immune profile from a pro-inflammatory M1 state to a reparative M2 phenotype, thereby creating a permissive window for cellular integration. Following this stabilization, metabolic demands are supported through revascularization driven by NSC-EV-derived angiogenic factors. The upregulation of VEGF-A and occludin indicates not only endothelial proliferation but also the expression of tight-junction proteins associated with BSCB structure, providing the necessary metabolic substrate for the survival and differentiation of implanted NPCs. Spatially, the scaffold addresses the barriers compartmentalizing the lesion. The GelMA hydrogel provides a structural bridge across the cystic cavity, preventing lesion core collapse. Moreover, the system modulates the lesion boundary by attenuating reactive astrogliosis and reducing GFAP expression at the margins, effectively lowering physical and biochemical barriers to axonal penetration. Thus, the strategy functions as an integrated system: the anti-inflammatory and angiogenic priming by TMP/NSC-EVs conditions the microenvironment to facilitate the effective survival and integration of the NPCs.

Engineering a permissive niche for structural support and NPC integration

Each constituent of the scaffold, GelMA, TMP, NSC-EVs, and NPCs, was deliberately chosen for its complementary role within the neuroinflammatory and neurodegenerative context of SCI. GelMA, a natural polymer widely used in biofabrication, served as the structural backbone of the scaffold. ⁵⁵ By emulating the extracellular matrix, GelMA created a mechanically stable yet biologically permissive environment that supported cellular adhesion, infiltration, and integration.^56–59 In our study, the GelMA scaffold effectively bridged the lesion gap, reduced cystic cavity formation, and improved transplanted cell retention at the injury site. GelMA hydrogels have been widely reported to exhibit tunable mechanical properties within the range relevant to soft tissues, including neural tissues, depending on formulation and crosslinking conditions. Although direct mechanical characterization was not conducted in this study, the biological performance observed herein suggests that the scaffold provides a permissive microenvironment for neural regeneration. ⁵⁹ The mechanical properties of GelMA, which approximate those of native spinal tissue, likely contributed to its ability to mitigate secondary collapse and establish a supportive niche for regeneration.^60,61 However, despite these structural advantages, GelMA alone failed to modulate inflammation or stimulate significant endogenous neural regeneration, underscoring the necessity of functionalization with bioactive components; however, while the GelMA hydrogel provides immediate physical filling of the lesion defect, our data indicate that the superior preservation of tissue architecture in the combinatorial groups is primarily attributable to biological intervention rather than passive structural support. Since cystic cavitation is largely driven by secondary necrosis resulting from unchecked inflammation and ischemia, the incorporation of TMP and NSC-EVs disrupted this cascade and supported vascular perfusion. Thus, our functionalized scaffold acts not merely as a bridge, but as a bioactive shield that preserves parenchymal integrity against the hostile post-injury microenvironment. These data highlight the cumulative contribution of NPCs and bioactive cues to the scaffold-mediated preservation of spinal cord structure.

Embedded within this hydrogel matrix, NPCs acted as a cellular source of neurogenic replacement. Their dual capacity for neuronal and glial differentiation, coupled with their secretion of trophic and immunomodulatory factors, positioned them as central to both structural reconstruction and microenvironmental homeostasis. ⁶² Nonetheless, NPC transplantation alone is frequently hindered by the hostile post-injury environment, which impairs cell viability and maturation.^21,63 Our data demonstrate that these limitations are markedly mitigated when NPCs are co-delivered with NSC-EVs within the GelMA/TMP scaffold. NPCs embedded in the composite scaffold survived for at least 4 weeks post-transplantation and differentiated into neurons NSC-EVs, enriched with cytokines, neurotrophic factors, and microRNAs, likely enhanced NPC viability and neuronal differentiation within the supportive GelMA microenvironment.^64,65 Moreover, the scaffold’s intrinsic anti-inflammatory properties further contributed to improved NPC survival and integration.

Immunomodulation and attenuation of the inhibitory microenvironment

Because post-traumatic inflammation is a major determinant of secondary SCI progression, TMP was incorporated into the hydrogel system to modulate oxidative and inflammatory stress. TMP, a bioactive alkaloid derived from Ligusticum wallichii, has long been recognized for its neurovascular protective, anti-inflammatory, antioxidant, and anti-apoptotic properties. ⁶⁶ In experimental SCI models, TMP suppresses NF-κB activation, reduces iNOS expression, and promotes PGC-1α expression, a master regulator of mitochondrial biogenesis and redox balance.^33,67 Consistent with these reports, our results show that TMP incorporation significantly downregulated TNF-α-induced expression of IL-6, IL-8, and NF-κB in NPCs in vitro. This suggests that TMP exerts anti-inflammatory effects not only through direct molecular signaling but also through diffusion-mediated modulation within the 3D scaffold architecture. In vivo, the NPC@GelMA/TMP/NSC-EV scaffold reduced inflammatory cytokine levels and significantly suppressed P2X7R expression, a purinergic receptor implicated in chronic microglial activation and neurotoxicity. ³⁴

A central finding of this work is the observed integration between NSC-EVs and NPCs in establishing a pro-regenerative microenvironment capable of overcoming the inhibitory landscape of SCI. NSC-EVs, derived from neural stem cells, carry CNS-specific miRNAs, neurotrophic factors, and synapse-associated proteins that more effectively promote neuronal survival and axonal regeneration than the more generalized molecular cargo of mesenchymal stem cell-derived EVs. ⁶⁸ The functional superiority of NSC-EVs over ASC-EVs observed in this study aligns with the parent cell hypothesis, which posits that EVs inherit a lineage-specific molecular signature from their tissue of origin. ⁶⁹ While ASC-EVs are typically rich in VEGF and generic immunomodulatory factors suited for stromal repair and adipose expansion, our proteomic analysis revealed a distinct enrichment of neuro-angiogenic factors, including HGF, Ang-1, and bFGF, in NSC-EVs.^41,70 This compositional divergence is mechanistically significant; for instance, whereas ASC-EV-derived VEGF promotes rapid angiogenesis that may exacerbate vascular permeability in the CNS, the high levels of Ang-1 in NSC-EVs are known to activate Tie2 signaling. This pathway promotes vessel maturation and tight junction upregulation, thereby stabilizing the BSCB and reducing secondary edema. ⁴² Similarly, HGF functions as a dual modulator, providing neuroprotection while specifically antagonizing TGFβ signaling to mitigate glial scarring that arrests regeneration. ⁴³ Beyond protein cargo, the differential efficacy is likely mediated by lineage-specific miRNAs. Existing literature suggests that NSC-EVs are reservoirs for neural-specific miRNAs, such as miR-124 and miR-9, which suppress the REST complex to promote neurogenesis and inhibit NF-κB-mediated inflammation.^69,71 In contrast, mesenchymal miRNAs predominant in ASC-EVs lack this specific neural code. Consequently, the NSC-EV cargo represents a synchronized repair program tailored for the neurovascular niche, whereas ASC-EVs provide primarily generalized trophic support.

Beyond their trophic influence, NSC-EVs suppressed pro-inflammatory and glial reactivity, as evidenced by reduced populations of CD68⁺iNOS⁺ and Iba-1⁺P2X7R⁺ microglia/macrophages, elevated CD163 expression, and downregulated NF-κB activity, hallmarks of a phenotypic shift from M1 to M2 macrophages. ⁷² Together with NPCs, NSC-EVs not only enhanced progenitor viability and neuronal differentiation, as demonstrated by HuNu⁺/MAP2⁺ colocalization, but also promoted host tissue remodeling through paracrine signaling involving factors such as BDNF and GDNF. Additionally, the combinatorial scaffold suppressed GFAP⁺ astrocyte accumulation and glial scar formation at the lesion boundary, thereby reducing physical and biochemical barriers to axonal regeneration.

Restoration of vascular integrity and the blood-spinal cord barrier

Alongside neurogenesis and axonal regeneration, vascular remodeling emerges as a pivotal determinant of successful spinal cord repair. Spinal cord injury is characterized by rapid and sustained disruption of the BSCB, leading to hemorrhage, leukocyte infiltration, and oxidative stress. ⁷³ Revascularization not only restores perfusion and oxygenation at the lesion site but also enables the delivery of systemic and local trophic factors essential for neural repair. In this study, the NPC@GelMA/TMP/NSC-EV scaffold markedly enhanced VEGF-A expression and tubular structure formation both in vitro and in vivo. The enrichment of angiogenesis-related proteins such as Ang-1 and bFGF within NSC-EVs likely contributed to this robust vascular response, consistent with previous findings that EVs promote endothelial proliferation, migration, and tube formation via activation of the PI3K–Akt and MAPK pathways. ⁷⁴ Of particular significance was the upregulation of occludin, a tight-junction protein critical for BSCB integrity, in scaffold-treated groups. This observation suggests that the newly formed vasculature was not merely reactive and permeable, as often occurs in pathological neovascularization, but rather structurally mature and functionally competent. Such vascular stabilization likely supports sustained metabolic exchange, tissue viability, and long-term regenerative outcomes.

Comparative analysis with emerging therapeutic strategies

In our study, the 3D bioprinting technology employed in our study offers a decisive advantage. By fabricating a lattice structure with pre-designed porosity and anisotropy, the NPC@GelMA/TMP/NSC-EV scaffold provides an immediate, robust physical highway for axonal guidance that does not depend on the chaotic post-injury host response. This pre-set topography acts as an architectural template, ensuring that regenerating axons and grafted NPCs align longitudinally rather than forming disarrayed neuroma-like structures. Additionally, the development of conductive scaffolds that mimic the electrophysiological microenvironment represents another major frontier. Recent studies, including Deng et al. and our own previous work, have utilized conductive materials (carbon nanotubes, MXenes) to restore electrical signal transmission. ⁷⁵ While conductive materials represent a significant leap forward in facilitating neuronal signaling, the design philosophy of our current study reflects a prioritization of acute pathological rescue. Following severe SCI, the immediate barriers to regeneration are metabolic collapse, oxidative toxicity, and the valley of death created by the inflammatory storm. Without effective antioxidant and anti-inflammatory protection, even a perfectly conductive scaffold cannot prevent the death of implanted cells. Therefore, we prioritized biochemical functionalization over intrinsic conductivity in this phase. By co-delivering TMP and NSC-EVs, our scaffold functions as a bioactive reservoir and provides a complex cocktail of neurotrophic factors.

The therapeutic potential of EVs has been elucidated by Wang et al., who identified specific molecular targets in athlete-derived EVs for mitigating ferroptosis. ⁷⁶ While targeting specific molecular pathways is powerful, the pathology of SCI is multifactorial. In this context, NSC-EVs offer a broad-spectrum therapeutic advantage. As parent signals of the neural lineage, NSC-EVs carry a comprehensive cargo of immunomodulatory and neurogenic factors that address multiple pathological axes simultaneously shifting macrophage polarization from M1 to M2 and inhibiting apoptosis via autophagy modulation. Furthermore, a critical challenge in EV therapy is the rapid clearance or washout effect observed with systemic injections. ⁷⁷ Our study overcomes this by anchoring NSC-EVs within the GelMA hydrogel network, achieving sustained in situ retention. This design ensures that the grafted NPCs and host tissues are continuously exposed to therapeutic concentrations of bioactive cues. ⁷⁸ Finally, regarding cellular strategy, while some approaches rely on neuronal relays formed by host interneurons, the complete transection model requires exogenous cell replacement to bridge the physical gap. Our NPC@GelMA/TMP/NSC-EV system functions as an exogenous neuronal relay, where grafted NPCs differentiate into MAP2+ neurons that physically and functionally reconnect the severed stumps, offering a robust solution for severe injuries where endogenous bridging capacity is lost.

Limitations and future directions

Despite these promising results, several limitations should be acknowledged. First, although the rat transection model provides reproducible lesioning and quantitative evaluation of scaffold efficacy, it does not fully capture the heterogeneity of human SCI, particularly in contusion or compression models where spared white matter and complex inflammatory dynamics may influence regeneration.

Second, regarding the experimental design, our study focused on evaluating the maximal therapeutic potential of the fully integrated platform (NPC@GelMA/TMP/NSC-EV). It was not designed to isolate the individual contributions of the scaffold, TMP, or NSC-EVs alone, but rather to evaluate the integrated, combinatorial effect of the multifunctional scaffold system; therefore, while the combined efficacy is evident, we cannot strictly differentiate between the individual versus integrated contributions of each bioactive component in the animal model.

Third, while we demonstrated the upregulation of tight-junction proteins and modulation of macrophage phenotypes via immunofluorescence and gene expression, we did not employ functional permeability assays or transgenic knockout models. Future studies are necessary to definitively establish the causal molecular pathways and functional integrity of the repaired BSCB. Furthermore, the mechanisms underlying the region-specific actions of NSC-EVs, specifically their roles in modulating OPC lineage commitment and microglial polarization, require deeper investigation using single-cell transcriptomics and proteomics.

Fourth, while NPC-derived neurogenesis and host tissue modulation were demonstrated, the long-term fate, electrophysiological integration, and functional synaptogenesis of regenerated neurons remain to be established through advanced electrophysiological or transsynaptic tracing studies.

Future research should explore multiple directions to address these challenges. Optimizing the spatial gradients and release kinetics of bioactive components could better replicate endogenous repair zones. Incorporating bioresponsive or degradable linkers within the GelMA framework may enable more precise, inflammation-triggered release of therapeutic cues. Additionally, engineering gene-edited or preconditioned NPCs could enhance subtype-specific differentiation and connectivity, particularly toward cholinergic and serotonergic lineages involved in locomotor control. Finally, long-term behavioral and safety studies will be essential to validate the durability and translational potential of this approach. Collectively, our findings underscore the promise of a dynamically responsive, phase-tuned scaffold platform for advancing combinatorial strategies in functional spinal cord regeneration.

Conclusion

In summary, this study presents a multifunctional, biomimetic scaffold integrating GelMA hydrogel, TMP, NSC-EVs, and NPCs into a unified platform for spinal cord injury repair. Through rational design and systematic evaluation, we demonstrated that this combinatorial construct effectively mitigates the multifactorial barriers of SCI by concurrently promoting neuroprotection, axonal regeneration, remyelination, angiogenesis, and immunomodulation. The collective interplay among scaffold components transforms the hostile post-injury microenvironment into one that fosters structural reconstruction and functional recovery. Importantly, the incorporation of NSC-EVs and TMP supports the therapeutic efficacy of NPCs, creating a permissive niche for survival, guiding lineage differentiation, and facilitating endogenous repair pathways. These findings highlight the value of integrated multimodal interventions that move beyond single-target paradigms, offering a clinically translatable framework for next-generation spinal cord regeneration therapies.